doi
stringlengths 3
59
| license
stringclasses 95
values | pmc
stringclasses 0
values | pubmed_id
stringclasses 0
values | subset
stringclasses 1
value | text
stringlengths 301
1.8M
|
|---|---|---|---|---|---|
10.1128/mSystems.00754-21
|
CCBY
| null |
34463566
|
s2orc_pubmed_articles
|
Bacteria-Stimulated Metamorphosis: an Ocean of Insights from Investigating a Transient Host-Microbe Interaction
Recent research on host-microbe interactions has focused on intimate symbioses. Yet transient interactions, such as the stimulation of animal metamorphosis by bacteria, can have significant impacts on each partner. During these short-lived interactions, swimming animal larvae identify a desirable location on the seafloor and undergo metamorphosis into a juvenile based on the presence of specific bottom-dwelling bacteria. While this phenomenon is critical for seeding new animals to establish or maintain benthic ecosystems, there is an ocean of fundamental questions that remain unanswered. Here, I propose an updated model of how bacteria stimulate animal metamorphosis based on evidence that bacteria inject a stimulatory protein that prompts tubeworm metamorphosis. I consider what we hope to learn about stimulatory bacterial products, how animals recognize these products, and the consequences for both partners. Finally, I provide examples of how studying an enigmatic host-microbe interaction can serve as an engine for scientific discovery.
R ecent research on host-microbe interactions has focused on intimate symbioses, where partners in close contact can promote both pathogenic and beneficial outcomes. However, it has become clear that environmental bacteria can also provide cues during transient interactions that regulate essential processes in diverse animals. These fleeting host-microbe interactions have shaped animal development in an array of animal and microbial lineages; however, the mechanisms that underpin these interactions remain mysterious. A currently understudied example of one such transient microbe-animal interaction is the stimulation of animal metamorphosis by bacteria. During these interactions in marine environments, swimming animal larvae identify a suitable location on the seafloor and undergo metamorphosis into a juvenile based in part on the presence of specific bottom-dwelling bacteria forming biofilms attached to submerged surfaces. Representative animals from each major branch of the animal tree of life have been shown to undergo metamorphosis in response to bacteria (e.g., corals, tubeworms, and urchins). It is therefore plausible that the phenomenon of bacteria-stimulated metamorphosis evolved long ago and continues to shape where and when marine animal larvae undergo metamorphosis.
In marine habitats, bacteria-stimulated metamorphosis is a critical process for seeding new animals to maintain or establish populations. This process might, in part, dictate the distribution of animals and ecosystems in the environment. Bacteria-stimulated metamorphosis likely contributes to the economically costly process of biofouling of ship hullsand is important for the life cycle of aquaculture species such as oysters. This phenomenon is not restricted to marine animals, as symbiotic bacteria have also been implicated in insect metamorphosis, and an analogous phenomenon occurs when seaweed zoospores settle in response to bacteria. Despite the importance of this process, three major questions about the phenomenon remain unanswered: (i) What are the bacterial products that promote metamorphosis? (ii) How do animals recognize bacterial products? (iii) Which partners benefit from this interaction and how (i.e., commensalism, mutualism, parasitism)? To address these questions, my research program uses emerging host-microbe model interactions to determine the mechanistic basis of bacteria-stimulated metamorphosis.
## A surprisingly different way that bacteria stimulate metamorphosis
Historically, bacterial products that stimulate metamorphosis were described as small soluble molecules or products associated with the bacterial cell surface or biofilm matrices. However, we discovered a surprisingly different way that some bacteria stimulate animal metamorphosis; the bacterium Pseudoalteromonas luteoviolacea produces syringe-like structures called metamorphosis-associated contractile structures (MACs) that inject stimulatory proteins into target animals. MACs belong to a class of syringe-like structures termed contractile injection systems (CIS) that bear homology to the contractile tails of bacteriophage (the viruses of bacteria) and are produced by diverse bacteria. CIS often translocate protein payloads into target cells, which may target a specific molecule or process to exert their effect. Until our recent work, CIS were known to mediate antagonism among competing microbes or between bacteria and their eukaryotic hosts. However, we discovered that a single protein loaded within the MACs complex is translocated into the larvae of a tubeworm,
## Commentary
Hydroides elegans, and stimulates normal metamorphosis. This single protein effector, termed metamorphosis-inducing factor 1 (Mif1), represents the first protein from a bacterium identified to stimulate the metamorphosis of an animal.
## An ocean of possibilities
Bacteria from diverse lineages have been shown to stimulate the metamorphosis of marine animal larvae. These bacteria include strains from the Gammaproteobacteria and Alphaproteobacteria classes, as well as bacteria from the Bacteroidetes group and Gram-positive Firmicutes phylum. Intriguingly, some bacteria are potent stimulants of metamorphosis, while closely related strains can be unable to stimulate metamorphosis under equivalent conditions. The diversity of bacterial species that stimulate metamorphosis suggests that there are numerous stimulatory bacterial products remaining to be discovered.
So far, genes for the biosynthesis of two products from bacteria that stimulate metamorphosis have been identified. These genes promote the biosynthesis of a brominated natural product tetrabromopyrrole (TBP) or encode Mif1, which stimulate coral and tubeworm metamorphosis, respectively. Some bacteria possess the genes and ability to produce both TBP and Mif1. These products are chemically different; TBP is a brominated aromatic hydrocarbon, while Mif1 is a proteinaceous effector. Yet both have been shown to stimulate metamorphosis, suggesting that very different bacterial products can promote a dramatic developmental transition in diverse animals.
A small number of purified products from bacteria have been shown to stimulate the metamorphosis of the tubeworm, Hydroides elegans, and the cnidarian, Hydractinia echinata. These products include outer membrane vesicles, lipopolysaccharides (LPS), extracellular polysaccharides, and lysophospholipids. These products induce metamorphosis when provided to larvae in a purified form, but it is currently unknown if the products stimulate metamorphosis when produced by, and in the context of, whole bacteria. Whether animal larvae respond to specific bacterial products from living bacteria within biofilms will be an interesting avenue of future research.
## An updated model of bacteria-stimulated metamorphosis
The phenomenon that bacteria stimulate animal metamorphosis was discovered over 80 years ago. Since this initial discovery, one model explaining how bacteria stimulate metamorphosis has gained traction. This model predicts that animals are stimulated to undergo metamorphosis in response to bacterial products that result from normal growth or metabolism. This model implies an "animal-driven" process where bacteria serve as passive features of the environment that animals use as an indicator of a preferable habitat. It is unknown whether this animal-driven model explains most of the interactions mediating bacteria-stimulated metamorphosis in the environment. Our discovery that P. luteoviolacea stimulates tubeworm metamorphosis by producing a CIS that translocates a bioactive protein into the animal larvae builds on the previous model of how bacteria stimulate metamorphosis. This bioactive protein mechanism implies that a "bacteria-driven" process also exists where bacteria drive the interaction by injecting stimulatory proteins into animal larvae. Importantly, both mechanisms of bacteria-stimulated metamorphosis might exist simultaneously in the environment and drive the process of animal recruitment to new habitats. This updated model of how bacteria stimulate animal metamorphosis leads to broader questions of whether and how the partners of this transient interaction are harmed or benefit from the encounter. Whether biofilm bacteria that stimulate metamorphosis colonize the juvenile and adult organism was questioned only recentlyand remains a critical gap in knowledge about the relationship.
## Anticipated advances in the next 5 years and beyond
Within the next 5 years, I envision that there will be three main advances that will push the field of studying bacteria-stimulated metamorphosis forward as follows: (i) Diverse bacterial products that stimulate animal metamorphosis and the genetic basis of their biosynthesis will be identified. These products will differ in their chemical composition, structure, and function (e.g. a specialized metabolite such as tetrabromopyrrole or a protein effector, like Mif1). The mode of action of the bacterial product and the mode of perception by the larvae will vary widely. (ii) Model systems focused on microbe-animal symbioses will be used more frequently, and methods to study them will catch up to established model organisms. Technical advances to study the animal partners will help to determine how animal larvae perceive bacterial products that stimulate metamorphosis. (iii) Discoveries about the diverse bacterial products that stimulate metamorphosis and how animals recognize these products will help to flesh out the broader question of whether the relationship between the bacteria and animal partners is a parasitic, commensal, or mutualistic interaction. It remains unknown whether bacteria that stimulate metamorphosis continue their interaction with the animal once it has undergone metamorphosis.
## Unexpected insights from studying bacteria-stimulated metamorphosis
Studying a mysterious host-microbe phenomenon has led to two unexpected insights. First, in addition to targeting tubeworm larvae, we found that MACs are capable of targeting very different types of eukaryotic cells, including insect cells and mouse macrophages, ex vivo. CIS such as MACs might therefore be amenable to engineering for biotechnology purposes as protein delivery devices to target eukaryotic cells. Second, while studying CIS that promote tubeworm metamorphosis, my lab made a fortuitous discovery-we found that a poorly described class of CIS genes is present within Bacteroidales bacteria from the gut microbiomes of nearly all healthy human adults from the United States and Europe. Further, we show that individuals suffering from irritable bowel disease have fewer CIS genes than healthy individuals, hinting at their role in human health. Our discoveries provide important instances of the power of fundamental research as an engine for scientific discovery.
# Conclusion
Numerous marine bacteria forming multispecies biofilm communities on submerged surfaces likely serve as an indicator of a preferable habitat for, and trigger the metamorphosis of, marine larvae. Studying such interactions will provide a wealth of foundational knowledge with profound health, economic, and biotechnology applications.
|
10.1074/JBC.272.14.8937
|
CCBY
| null |
9083015
|
s2orc_pubmed_articles
|
Oligodendrocytes Direct Glycosyl Phosphatidylinositol-anchored Proteins to the Myelin Sheath in Glycosphingolipid-rich Complexes*
The myelin sheath synthesized by oligodendrocytes insulates central nervous system axons and is a specialized subdomain of the plasma membrane, containing a restricted pattern of proteins and lipids. Myelin is enriched in glycosphingolipids and cholesterol, a lipid environment favored by glycosylphosphatidylinositol (GPI)-anchored proteins, which associate with these lipids in detergent-insoluble complexes in many cell types. Since proteins regulating oligodendroglia-neuron interaction are largely unknown and GPI-anchored proteins are often involved in cell-cell interactions, we examined oligodendrocytes and myelin for their expression of these proteins. Oligodendrocyte precursors and maturing oligodendrocytes express a similar pattern of GPIanchored proteins, which unlike the majority of oligodendrocyte plasma membrane proteins, accumulate in myelin. To elucidate mechanisms underlying the expression of GPI-anchored proteins in myelin, we analyzed detergent-insoluble complexes from cells and myelin using TX-100 extraction and sucrose density gradients. In precursor cells, the GPI-anchored proteins are not incorporated in detergent-insoluble complexes. In contrast, GPI-anchored proteins from maturing oligodendrocytes and from myelin were isolated as complexes associated with glycosphingolipids and cholesterol. These results show a specific association of GPI-anchored proteins with glycosphingolipids and cholesterol during oligodendrocyte maturation and suggest sorting of these macromolecular complexes to myelin.
Myelin is the multilamellar sheath insulating selective axonal processes in higher animals, facilitating fast propagation of the action potential. In the central nervous system (CNS)it is synthesized by oligodendrocytes, and in the peripheral nervous system it is synthesized by Schwann cells as a specialized domain of the plasma membrane. In contrast to most plasma membranes, myelin is enriched in lipids, which comprise 70% of the dry weight, and contains a restricted pattern of proteins when compared with the oligodendrocyte. During myelination, large diameter axons are encircled by the myelinating glial cell process. This recognition event is assumed to involve specific adhesion molecules, which may be contained in the restricted fraction of myelin proteins. These molecules are still largely unknown, although the main protein components of the myelin sheath are defined and cloned. The relatively normal development and myelination of mice deficient in the myelin-associated glycoprotein (MAG), which was previously thought to be essential for glia-axon recognition during the early phases of myelination, has underscored this point. Hence, as yet unknown and possibly minor oligodendrocyte/myelin proteins and their direction to specialized membrane domains may play important roles in the onset and maintenance of myelination. The cell biological processes underlying the segregation of specific membrane components to myelin are thus of functional importance.
GPI-anchored proteins represent a diverse and growing group of membrane proteins, which includes parasite and lymphocyte surface antigens, cell adhesion molecules, membraneanchored ectoenzymes, and receptors . They are inserted into the outer leaflet of the lipid bilayer via covalent linkage to phosphatidylinositol. The insolubility of GPI-anchored proteins at 4°C in nonionic detergents such as TX-100 has been ascribed to their incorporation into large macromolecular complexes at the level of the trans-Golgi network by agglomeration with glycosphingolipids and cholesterol. As shown for many cell types, such detergent-insoluble glycosphingolipidenriched complexes (DIGs; Ref.can be isolated by flotation in low density fractions on sucrose gradients. The complexes also include cytoplasmic signaling molecules such as src family tyrosine kinases, heterotrimeric GTP-binding proteins and small GTPases. It is thought that glycosphingolipidrich complexes represent functional microdomains within the plasma membrane of intact cells, which are sorted to the apical surface in polarized epithelial cellsor participate in transmembrane signaling in lymphocytes.
The myelin sheath synthesized by oligodendrocytes and Schwann cells has a similar lipid composition to the detergentinsoluble microdomains described above. It has a high glycosphingolipid content (approximately 30% of total lipid weight) and is rich in cholesterol (26% of total lipid weight). This prompted us to analyze the expression of GPI-anchored proteins by oligodendrocyte lineage cells and myelin; such molecules could be candidates for the adhesion molecules involved in glial cell-neuron interaction and in the synthesis and maintenance of the myelin sheath. We report here that oligodendrocytes, both precursors and more mature cells, synthesize a broad spectrum of GPI-anchored proteins. Furthermore, we show that oligodendrocytes and myelin express the same pattern of GPI-anchored proteins. We investigated the flotation of detergent-insoluble complexes from precursor cells, maturing oligodendrocytes, and myelin on sucrose density gradients and observed that with maturation of the cells, GPI-anchored proteins become associated with glycosphingolipids and cholesterol in DIGs. This association may be crucial for the sorting of GPI-anchored proteins to the myelin sheath.
## Experimental procedures
Materials-Biotinylation reagents were from Pierce; growth factors were from Collaborative Research (Bedford, MA); [ 3 H]ethanolamine,H]glucosamine, 125 I-labeled protein A, and 125 I-labeled streptavidin were from Amersham-Buchler (Braunschweig, Germany); recombinant phosphatidylinositol-specific phospholipase C (PI-PLC) from Bacillus thuringensis was from Oxford Glycosystems; Triton X-100 and lipid standards were from Sigma (Deisenhofen, Germany); Triton X-114 was from Serva (Heidelberg, Germany); polyvinylidene difluoride membrane was from Millipore Corp. (Bedford, MA); Silica Gel 60 F 254 TLC plates were from Merck (Darmstadt, Germany).
Antibodies-The following rabbit polyclonal antibodies were used that recognized NCAM, MAG (a kind gift of F. Kirchoff), the C-terminal peptide of PLP (a kind gift of C. Linington), and F3 (a kind gift of G. Gennarini). Monoclonal antibodies used were as follows: murine monoclonal antibody against the myelin protein MOG (clone 818C5) and mouse monoclonal antibody against myelin basic protein (MBP) (clone 9 -8) kindly provided by C. Linington; murine monoclonal antibodies O4 and O1recognizing sulfatide and galactocerebroside, respectively, and in addition recognition of a sulfated protein epitope by O4 (1); murine monoclonal antibody O10 against a cell surface epitope expressed by more mature oligodendrocytes; and rat monoclonal antibody 324 against the neuronal cell surface glycoprotein L1. Secondary antibodies were from Dianova (Hamburg, Germany).
Primary Cultures and Cell Lines-Primary cultures of oligodendrocytes were prepared from embryonic day 14 -16 mice as described. Oligodendrocytes growing on top of astrocyte monolayers were shaken off and plated in modified Sato mediumcontaining 1% horse serum (HS) on poly-L-lysine-coated dishes. To increase the proportion of precursor cells as well as to promote survival, 10 ng/ml platelet-derived growth factor and 5 ng/ml basic fibroblast growth factor were added immediately after the shake and after 24 h in vitro. Precursor cells were harvested 30 h after the shake. Mature cells were cultured for 8 days in vitro without further growth factor additions before they were harvested. Cortical astrocytes described were derived from the monolayers remaining after shaking off the oligodendrocytes. The cell line Oli-neuwas cultured in Sato medium containing 1% HS. PC12 cells were cultured in Dulbecco's modified Eagle's medium with 5% HS and 5% fetal calf serum. 3T3 fibroblasts were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum.
Immunofluorescent Labeling of Cells-Staining of primary oligodendrocytes for indirect immunofluorescence was carried out as described previously. To determine the differentiation state of the cultures, cells growing on poly-L-lysine-coated glass coverslips were stained with O4, O1, O10, and MBP antibodies.
Preparation of Oligodendrocyte Membranes-For preparation of oligodendroglial membranes, cultures of primary oligodendrocytes (2 days after shaking off) were rinsed twice with PBS and harvested with a cell scraper. The cell pellet was resuspended in 10 mM Tris, pH 7.4 and left for 10 min on ice, and cells were disrupted using a 22-gauge syringe. Nuclei were removed by centrifugation for 10 min at 2000 ϫ g and 4°C. Membranes were pelleted by ultracentrifugation for 1 h at 100,000 ϫ g and 4°C. The pellet was resuspended in SDS-PAGE sample buffer.
Myelin Preparation-Myelin was isolated from the brains of adult NMRI mice of both sexes according to standard procedures. Initially, brains were homogenized in ice-cold 10.5% sucrose with the help of an Ultra-Turrax T25 (IKA, Staufen, Germany). Myelin was removed from the interface between 10.5 and 30% sucrose gradients and subjected to two rounds of hypoosmotic shock by resuspension in a large volume of ice-cold water and reisolation on the step gradient; this separates myelin membranes from axolemma. Purified myelin was collected from the final sucrose interface, washed twice with cold water, resuspended in a small volume of water, and immediately used or frozen in small aliquots at Ϫ80°C. The myelin preparations were analyzed for the absence of contaminating axolemmal and plasma membrane fragments by Western blot analysis using antibodies against the cell adhesion molecule L1 and the 180-kDa isoform of NCAM. In addition, the absence of contaminating astrocyte cytoskeletal material was monitored with antibodies against glial fibrillary acidic protein.
Detection of GPI-anchored Proteins by Biosynthetic Radiolabeling-Oli-neu cells or primary cultures of oligodendrocytes 2 days after plat-ing were incubated for 18 h with either 100 Ci/ml [1-3 H]ethan-1-ol-2amine hydrochloride (28 Ci/mmol) or 100 Ci/ml D-[6-3 H]glucosamine hydrochloride (25 Ci/mmol). The cell monolayer was then incubated for a further hour at 37°C in RPMI medium containing 10 mM HEPES either with or without the addition of 1 unit/ml PI-PLC. The culture supernatants were then removed and cleared of debris by centrifugation, and the proteins were precipitated with cold acetone. The total cell pellet as well as the proteins precipitated from the supernatants were subject to SDS-PAGE and autoradiography.
Protein Biotinylation-Biotinylation of cell surface proteins was carried out as described. After repeated washings with ice-cold PBS containing 1 mM CaCl 2 and 1 mM MgCl 2 (PBS-CM), cells were incubated twice for 30 min with membrane-impermeable Sulfo-NHS-LC-Biotin (1 mg/ml; approximately 2.5 mg of biotin/10 6 cells) at 4°C. They were then washed three times with cold PBS-CM, harvested with a cell scraper, and pelleted.
Myelin diluted to a final protein concentration of 200 g/ml was lightly sonicated to disrupt compacted lamellae and to facilitate access of biotin. Proteins were biotinylated with two additions of 1 mg/ml membrane-permeable NHS-LC-Biotin II for 45 min with shaking. Free biotin was removed by pelleting myelin membranes by ultracentrifugation for 30 min at 100,000 ϫ g and 4°C.
Isolation of GPI-anchored Proteins Using Triton X-114 Phase Separation-TX-114 phase separation was performed according to a protocol established by Bordier, which was adapted for the purification of GPI-anchored proteins. Cells (5 ϫ 10 6 to 4 ϫ 10 7 , after radiolabeling or cell surface biotinylation) or myelin (200 g of total protein, biotinylated) were solubilized at 4°C in 2 ml of Tris-buffered saline containing 1 mM EDTA, 1% TX-114, and a mixture of protease inhibitors. This mixture was shaken for 30 min at 37°C and centrifuged (5 min, 13,000 ϫ g) to separate detergent and aqueous phases. Detergent phases were reextracted twice with 1 mM Tris-buffered saline/EDTA. After dilution with Tris-buffered saline/EDTA to a final TX-114 concentration of 2%, the detergent phase was incubated without (control), or with 2 units/ml PI-PLC for 1 h at 37°C with shaking to specifically release GPI-anchored proteins into the aqueous phase. After additional separation of the phases (centrifugation for 5 min at 13000 ϫ g) the second aqueous phase was reextracted twice with 2% TX-114. The proteins of the second aqueous and detergent phases were precipitated with cold acetone and analyzed by SDS-PAGE followed by autoradiography in the case of radiolabeling experiments or followed by Western blotting and detection of biotinylated proteins with 125 I-labeled streptavidin (0.12 Ci/ml, 1 h at room temperature) and autoradiography. PI-PLC-treated and untreated second aqueous phases were compared to specifically identify GPI-anchored proteins.
Preparation of Detergent Extracts-Detergent extracts were prepared as described. In brief, cells (3-4 ϫ 10 7 ) or sonicated myelin (0.5 mg/ml protein) were solubilized at 4°C in 2 ml of buffer containing 10 mM Tris/HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 2% TX-100 (TNE/TX-100). The extracts were shaken for 30 min either at 4 or 37°C. Soluble and insoluble material was separated by centrifugation for 10 min at 13,000 ϫ g and 4°C or room temperature, respectively, in a microcentrifuge. Pellets and supernatants were analyzed by SDS-PAGE and Western blotting.
Sucrose Density Gradients-0.5-1 ml of total detergent extracts were adjusted to 40% sucrose by adding equal volumes of 80% sucrose in TNE without TX-100 and placed into an ultracentrifuge tube. A linear gradient from 5 to 30% sucrose (in TNE without TX-100) was layered over the lysate. Gradients were centrifuged for 16 -20 h at 33,000 rpm at 4°C in a Beckmann SW 40 TI rotor. Fractions of 1 ml were harvested from the top, and the density was determined by measurement of the refractive index. Proteins and lipids in each fraction were analyzed by SDS-PAGE followed by Western blot and TLC.
Immunoblotting-Proteins blotted onto polyvinylidene difluoride membrane were detected by incubation with primary antibodies overnight at 4°C. In the case of primary monoclonal antibodies, the blots were incubated with a second rabbit anti-species antibody for 1 h at room temperature. Bound antibodies were detected with 125 I-labeled protein A (0.12 Ci/ml, 1 h of incubation at room temperature).
Lipid Analysis-Lipid was extracted by the method of Bligh and Dyer. Dried samples were dissolved in 10 -20 l of chloroform/ methanol (1:1, v/v) and spotted on activated Silica Gel 60 F 254 plates. After resolution of the lipids in methyl acetate, propane-2-ol, chloroform, methanol, 0.25% (w/v) KCl (25:25:25:10:9, by volume; Ref., the plates were air-dried, and lipids were visualized after exposure of the plates to 10% sulfuric acid, 5% methanol and charring.
Electron Microscopy-Electron microscopic analysis was performed as described. In brief, detergent-insoluble complexes pelleted from low density fractions by ultracentrifugation for 1 h at 100,000 ϫ g and 4°C were fixed with 2.5% glutaraldehyde plus 2% formaldehyde in 0.1 M sodium cacodylate, pH 7.2, overnight at 4°C. Samples were postfixed in 1% OsO 4 plus 1.5% potassium ferricyanide for 1 h and incubated for 30 min in 1.5% magnesium uranyl acetate. After dehydration in ethanol and Epon embedding, ultrathin sections were cut, contrasted with uranyl acetate and lead citrate, and examined using a Zeiss electron microscope (EM 10CR).
# Results
## Oligodendroglial cells synthesize a variety of gpi-anchored
Proteins-To investigate the expression of GPI-anchored proteins by oligodendrocyte lineage cells, we took advantage of the selective incorporation of [ 3 H]ethanolamine into GPI anchors. The cell line Oli-neu was generated by immortalization of mitotic oligodendrocyte precursor cells with retroviral vectors containing the t-neu oncogene. It adheres to and ensheathes axons in vitro and in vivo in an oligodendroglial specific manner. In vitro metabolic labeling of Oli-neu cells and subjection of the total cell lysate to SDS-PAGE yielded a variety of distinct bands on the autoradiograph with prominent components of 135 and 120 kDa and a broad band at 50 -70 kDa probably consisting of several components [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref] , lane 2). Several minor components were also visible. PI-PLC, which selectively cleaves GPI-anchored proteins from the lipid anchor, was used to examine the susceptibility of the [ 3 H]ethanolamine-labeled proteins expressed by Oli-neu to enzyme-dependent release prior to cell lysis. Although cleavage was not complete (compare with [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref] , lane 1), the majority of the different [ 3 H]ethanolamine-labeled proteins was released into the culture supernatant after treatment of the cells with PI-PLC [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref] , confirming the presence of the GPI anchor. The heavily labeled band at 45 kDa not cleavable by PI-PLC is probably the cytoplasmic protein synthesis elonga- tion factor 1␣, which is an exception in that it incorporates ethanolamine but is not GPI-anchored.
When [ 3 H]glucosamine was used to metabolically label cells of the line Oli-neu, the total cell homogenate contained many more 3 H-labeled proteins [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref] than in the ethanolamine-labeling experiments [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref]. This result was expected, since this metabolite is not specific for the GPI anchor and after acetylation will also be incorporated into the carbohydrate side chains of many glycoproteins. However, when the GPI-anchored proteins were released from intact cells by PI-PLC treatment prior to lysis, analysis of such supernatants yielded a pattern of labeled proteins similar to that seen in the [ 3 H]ethanolamine experiments [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref]. In particular, the dominant signals at 135 and 120 kDa are clearly visible.
Primary cultures of murine oligodendrocytes (2 days after plating) consisting of a spectrum of differentiation stages were then analyzed. Metabolic labeling with [ 3 H]glucosamine followed by incubation of the cells with PI-PLC yielded a pattern of radiolabeled proteins in the supernatant [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref] similar to that seen with Oli-neu [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref].
Purification of Oligodendrocyte GPI-anchored Proteins-GPI-anchored proteins that are susceptible to PI-PLC cleavage can be specifically enriched in the detergent phase of a TX-114 phase separation protocol . They are subsequently cleaved from their hydrophobic lipid anchors and partition into the aqueous phase after a second extraction. Ethanolaminelabeled Oli-neu cells were subjected to this purification protocol. Labeled proteins were separated into the TX-114 detergent phase [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref] and could be specifically although not completely released into the second aqueous phase with PI-PLC [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref]. Aqueous phases derived from untreated detergent phases did not yield any ethanolamine-labeled protein bands [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref]. It was important to determine the pattern of proteins seen when [ 3 H]ethanolamine-labeled cells were subjected to this technique, since non-GPI-anchored proteins that are associated strongly with a GPI-anchored protein will also be co-purified. Such proteins would be visible in the biotinylation experiments described below but would not be visible when ethanolamine labeling is used.
The Expression Pattern of GPI-anchored Proteins Is Specific for Oligodendroglial Cells-We used biotinylation to detect protein components independent of their metabolic turnover. Oli-neu cells and primary oligodendrocytes were subjected to cell surface biotinylation followed by isolation of the GPI-anchored proteins according to the TX-114 phase separation protocol and detection of individual proteins with iodinated streptavidin. The results reproduced the data of the radiolabeling experiments shown previously . The 135-kDa protein was also found in the PI-PLC untreated second aqueous phase, but the signal was of much lower intensity. Precursor cells and more mature cultures showed the same pattern of bands (data not shown). Similar experiments conducted with other cells such as primary astrocytes from mouse cortex, NIH-3T3 fibroblasts, and PC12 cells yielded a different pattern of GPI-anchored proteins cleavable by PI-PLC for each of the cell types tested, although a few proteins may be common to several cell types. Thus, oligodendrocyte lineage cells express a specific and distinct set of GPI-anchored proteins that are susceptible to PI-PLC cleavage.
The GPI-anchored Proteins Expressed by Oligodendrocytes Are Also Found in Myelin-To test whether the GPI-anchored proteins expressed by oligodendrocytes are included in the restricted fraction of proteins that are present in the myelin sheath, CNS myelin was prepared from adult mice and subjected to biotinylation, TX-114 phase separation, and PI-PLC treatment. The pattern of GPI-anchored proteins observed for myelin was strikingly similar to that seen for primary oligodendrocytes [fig_ref] FIG. 3: Oligodendrocyte GPI-anchored proteins are localized in myelin [/fig_ref] and for the Oli-neu cell line . Again, clear protein bands were discernible at 135, 120, and 100 kDa, a broad band probably consisting of several proteins between 50 and 70 kDa, and several minor components of lower molecular mass. This suggests that all GPI-anchored proteins expressed by oligodendrocytes are also located in myelin. Comparison of total biotinylated oligodendrocyte plasma membrane proteins and biotinylated myelin proteins [fig_ref] FIG. 3: Oligodendrocyte GPI-anchored proteins are localized in myelin [/fig_ref] , reveals a totally different overall protein pattern. The majority of the oligodendrocyte plasma membrane proteins are not visualized in the myelin fraction as previously reported (3). Silver staining of total oligodendrocyte membrane proteins and myelin proteins [fig_ref] FIG. 3: Oligodendrocyte GPI-anchored proteins are localized in myelin [/fig_ref]. The band pattern matches the metabolic labeling data [fig_ref] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins [/fig_ref]. B, different cell types express different patterns of GPI-anchored proteins. Cultures of Oli-neu (lanes 1 and 5), cortical astrocytes (lanes 2 and 6), PC 12 cells [fig_ref] FIG. 3: Oligodendrocyte GPI-anchored proteins are localized in myelin [/fig_ref] , and NIH 3T3 fibroblasts (lanes 4 and 8) were assayed as described above. Second aqueous phases after PI-PLC treatment (ϩ) from the different cell lines exhibit different patterns of GPI-anchored proteins. primary oli., primary oligodendrocytes; Astroc., primary cortical astrocytes.
GPI-anchored proteins are located in myelin suggests that the GPI anchor may be an address signal for the myelin sheath.
The Identity of GPI-anchored Proteins Expressed by Oligodendrocytes and Myelin-Western blotting and peptide sequencing was used to identify some of the different GPI-anchored proteins synthesized by oligodendrocytes. NCAM 120 and 5Ј-nucleotidase were confirmed as the 120-and 65-kDa bands, respectively (data not shown). It is well established that both the GPI-anchored isoform of NCAM (NCAM 120; Ref.and 5Ј-nucleotidase are synthesized by oligodendrocytes and found in myelin.
Tryptic digests of the 135-kDa protein from Oli-neu after separation by two-dimensional gel electrophoresis yielded sequence data corresponding to the neuronal cell adhesion molecule F3 (data not shown; Ref.. In Western blots with antibodies against F3, the expression of this protein was confirmed in myelin and oligodendrocytes. Tryptic digests of the 67-kDa band yielded sequences identical to the GPI-anchored complement restriction factor CD55 (decay-accelerating factor), whose expression by human Schwann cells has been recently reported.
GPI-anchored NCAM 120 and F3 Are Resistant to Solubilization in TX-100 at 4°C-For a variety of cell types (e.g. epithelial cells, neuroblastoma cells, cerebellar cells, and lymphocytes) it has been shown that GPI-anchored proteins associate with glycosphingolipids and cholesterol in detergent-insoluble complexes. Such complexes are insoluble in the nonionic detergent TX-100 at 4°C but soluble in TX-100 at 37°C and in n-octyl glucoside. We extracted oligodendrocytes and myelin with a buffer containing TX-100 and analyzed soluble (supernatants) and insoluble (pellet) fractions by Western blotting. Extraction of oligodendrocytes at 4°C yielded a substantial quantity of GPI-anchored NCAM 120 and F3 in the insoluble fraction . In contrast, the transmembrane isoform of NCAM, NCAM 140, like MAG and MOG showed the opposite behavior and was exclusively located in the superna-tant. A small fraction of PLP remained in the pellet at 4°C. Extraction at 37°C or extraction with n-octyl glucoside (data not shown) completely solubilized all of the proteins. The precursor cell line Oli-neu again demonstrates the different behavior of the two NCAM isoforms, with a small fraction of the NCAM 120 but no NCAM 140 remaining insoluble in TX-100 at 4°C.
Extraction of myelin yielded slightly different results. As for oligodendrocytes, NCAM 120 and F3 were partially insoluble and substantially enriched in the pellet after detergent extraction at 4°C. However, differing proportions of the other myelin proteins MAG, MOG, and PLP also remained associated with the pellet, although the bulk of these proteins was found in the supernatant. Repeated extractions of the pellet did not solubilize further material. Extraction of myelin at 37°C revealed that a minor fraction of NCAM 120 and PLP remained insoluble, whereas all other proteins examined were completely soluble . Lipid analysis of the TX-100 insoluble pellet from myelin confirmed the presence of glycosphingolipids and cholesterol in the pellet (data not shown).
The insolubility of the GPI-anchored proteins described above could be due to the association of the proteins with glycosphingolipids and cholesterol. It is well established that the expression of galactocerebroside and sulfatide is up-regulated during the differentiation of oligodendrocytes from pre-. If maintained in culture for 8 days, the proportion of mature cells increases, as evidenced by the staining for O1, O10, and MBP antigens and a more complex morphology (Refs. 22 and 27; [fig_ref] FIG. 5: Morphology and antigen profile of oligodendrocyte primary cultures and the cell line... [/fig_ref]. In the following experiments, we will define the early cultures as "precursors" and the latter cultures as "oligodendrocytes." The purity of the cultures is always at least 95% for oligodendrocyte lineage cells, as determined by the absence of cells expressing neuronal markers or glial fibrillary acidic protein.
The antigen profile of the Oli-neu line confirms that the majority of the cells in the line are of the precursor cell phenotype [fig_ref] FIG. 5: Morphology and antigen profile of oligodendrocyte primary cultures and the cell line... [/fig_ref] , but approximately 7% of the cells express the O1 antigen; these probably arise by spontaneous differentiation.
Oligodendrocytes Acquire Glycosphingolipid-rich Low Density Detergent-insoluble Complexes with Maturation-To analyze the association of GPI-anchored proteins with lipids, 4°C detergent extracts of oligodendrocyte precursor cells, oligodendrocytes, or myelin were subjected to 5-30% linear sucrose density gradient centrifugation. Gradient fractions were analyzed by Western blotting for the presence of NCAM and F3 as examples of GPI-anchored proteins. Analysis of precursor cells revealed NCAM and F3 at high density in fractions 1 and 2 of the gradient (bottom fractions). Occasionally, after very long exposure times, a barely detectable signal for NCAM 120 but no signal for NCAM 140 was seen in fractions 4 -7. On gradients from oligodendrocytes, a large amount of NCAM 120 but no NCAM 140 was found in fractions 5-10, with a clear peak in fraction 6 corresponding to 21% sucrose . F3 co-localized with NCAM 120 ; analysis of biotinylated material showed that the complete spectrum of GPI-anchored proteins synthesized by oligodendrocytes was found in the gradient fractions 5-10 (data not shown). Other transmembrane proteins analyzed (MAG and MOG) behaved similarly to NCAM 140 and were exclusively located at high densities (data not shown). Gradients prepared from myelin revealed a substantial amount of floating material, clearly visible as an opaque band. Nearly all of the NCAM 120 and F3 protein was found in fraction 8 corresponding to 17% sucrose . Interestingly, the DIGs prepared from myelin were less dense than those from the cells. The distribution of the myelin proteins, MAG and MOG (not shown), showed that the majority of each of these proteins was localized at the bottom of the gradient in fraction 1 and 2; however, a significant fraction of each protein co-localized with NCAM 120 and F3. In the case of PLP a substantial fraction of the total protein was distributed along the entire gradient, reflecting the known difficulty in completely dissociating this protein from lipid . This is consistent with the experiments in which a fraction of these non-GPI-linked proteins in myelin was also revealed to be resistant to solubilization with TX-100 at 4°C . Flotation of the GPI-anchored proteins in the gradient suggests their specific association with lipids.
The lipid composition of each fraction was analyzed by TLC. The fractions enriched in the GPI-anchored proteins from mature oligodendrocytes and myelin were also enriched in glycosphingolipids and cholesterol. Both sulfatides and galactocerebrosides (hydroxylated and nonhydroxylated forms) were almost exclusively located in these fractions together with cholesterol (oligodendrocytes, , middle part, fractions 5-9; myelin, , bottom part, fractions 6 -9), whereas the phospholipids (e.g. phosphatidylethanolamine) were predominantly found in fractions 1 and 2. Precursor cells showed no detectable lipid signal in low density fractions , all of the lipids including cholesterol were located in high density fractions, and no signal for galactocerebroside or sulfatide was detected.
Examination of the DIG-containing sucrose fractions by electron microscopy showed the presence of vesicular membranous structures 100 -500 nm in diameter. Interestingly, the DIGs from oligodendrocytes were largely unilamellar structures, whereas the DIGs from myelin in addition contained a high proportion of multilamellar structures [fig_ref] FIG. 7: Morphological analysis of DIGs [/fig_ref].
These results demonstrate that low density DIGs enriched in GPI-anchored proteins, glycosphingolipids, and cholesterol are up-regulated during oligodendrocyte maturation and are also found in the mature myelin sheath. The precursor cells do not form these complexes, and this may be due to the different lipid content of precursor cells and maturing oligodendrocytes.
# Discussion
## Identification of oligodendrocyte and myelin gpi-anchored
Proteins-Biochemical studies demonstrated that oligodendrocytes and myelin contain many different GPI-anchored proteins. Some are known proteins, including cell adhesion molecules of the Ig superfamily and proteins involved in cell-cell interactions. These include NCAM 120, and 5Ј-nucleotidase, which have previously been reported to be expressed by oligodendrocytes and to be present in myelin. In addition, we identified two proteins as F3 and CD55. The oligodendrocyte myelin glycoprotein has a molecular mass of 120 kDa, which is reduced to 105 kDa after PI-PLC cleavageand may also be included in our preparations. Other candidate molecules for some of the observed components are the ciliary neurotrophic factor receptor ␣ subunit and prion protein, which have been reported to be expressed by oligodendrocyte lineage cells.
## Gpi-anchored proteins are included in detergent-insoluble complexes in mature oligodendrocytes and myelin but not in
Oligodendroglial Precursor Cells-We observed a clear difference between the behavior of GPI-anchored proteins of the myelin sheath and other myelin proteins, reminiscent of the behavior of GPI-anchored proteins included in glycosphingolipid-rich microdomains. Simple detergent extraction studies of oligodendroglial cells and myelin revealed that a fraction of the GPI-anchored NCAM 120 and F3 was insoluble in the detergent TX-100 at 4°C, whereas transmembrane proteins such as NCAM 140 and other oligodendroglial or myelin proteins were almost completely solubilized. In contrast, GPIanchored proteins could be completely solubilized in extracts prepared at 37°C or with the detergent n-octyl glucoside. It is known that myelin is poorly soluble in detergents compared with classical plasma membranes and that the cholesterol and glycosphingolipid component remain largely insoluble.
Analysis of detergent extracts on sucrose density gradients showed that low density complexes, containing GPI-anchored proteins, could be isolated from mature cells and myelin but not from precursor cells. The very small amount of DIGs that could occasionally be isolated from precursor cell cultures may result from the low percentage of more mature cells in these cultures (7% O1-positive cells). The appearance of DIGs thus correlates with maturation of the cells. The DIGs from myelin were of lower density, indicating a higher lipid:protein ratio. The complete pattern of GPI-anchored proteins found in myelin and oligodendrocytes is, however, already present in precursor cells. GPI-anchored proteins are not organized in DIGs in precursor cells, which are migrating cells establishing initial axon contacts, but are recruited to DIGs by myelinating cells and are present in DIGs in myelin. This may reflect a change in function of GPI-anchored proteins during oligodendrocyte development where the association in DIGs may regulate the signal transduction capacities of the GPI-anchored molecules by association with signaling molecules. In support of this concept, we have observed that the DIGs contain significant amounts of fyn kinase.It is well established in lymphocytes that crosslinking GPI-anchored proteins with antibodies can transmit signals inside the cell, possibly mediated by src family kinases.
Interestingly were prepared at 4°C, adjusted to 40% sucrose, overlaid with a 5-30% linear sucrose gradient, and centrifuged to equilibrium. Equal amounts of total protein were loaded on gradients of oligodendrocyte precursor cells and mature oligodendrocytes (1.5 mg), whereas the myelin gradients were loaded with less protein (650 g). Gradient fractions (1 ml) were collected and immunoblotted to detect NCAM (A) and F3 (B). Fraction 1 is of high density from the bottom of the gradient. In mature oligodendrocytes and myelin, but not in oligodendrocyte precursor cells, GPI-anchored NCAM 120 and F3 are present in detergent-insoluble low density complexes. Note that the transmembrane NCAM 140 is exclusively found at the bottom of the gradient. C, sucrose density gradients of TX-100 extracts prepared from murine CNS myelin were analyzed by Western blot for the distribution of the myelin proteins MAG and PLP/DM20. the neuronal cell adhesion molecule L1, and fyn tyrosine kinase was shown. However, the authors reported the "contamination" of the lighter fraction with myelin and assumed that the F3 signal was exclusively of neuronal origin. Our data show that myelin and oligodendrocyte membranes include such DIGs containing NCAM 120, F3 and fyn kinase.
Myelin May Contain Several Different Populations of Microdomains-Different subpopulations of detergent-insoluble microdomains may exist in myelin, consistent with the view that myelin consists of several distinct compartments, e.g. paranodal loops and compact myelin, with specific biological functions. In contrast to the GPI-anchored proteins, which were exclusively localized in the DIGs, the proteins MAG and PLP were found predominantly at high densities at the bottom of the sucrose density gradients from myelin. However, significant fractions of each of these proteins were also found in floating fractions containing the DIGs. The DIGs from myelin appear to be more heterogeneous than those from oligodendrocytes and include a population with a multilamellar structure.
Specific Lipid Association of GPI-anchored Proteins in Oligodendrocytes and Myelin Suggests Sorting Mechanism-The biological relevance of the GPI anchor as a mode of attachment of cell surface proteins is still a matter of debate. It has been suggested that it is involved in intracellular targeting. Thus, in differentiated hippocampal neurons in culture, it has been proposed that the GPI anchor sorts Thy-1 to the axonal domain and excludes it from the somatodendritic compartment. The original postulate that GPI-anchored proteins are sorted in rafts in association with glycosphingolipids and cholesterolhas been demonstrated in the polarized epithelial line Madin-Darby canine kidney cells. It was shown that the rafts are generated in the Golgi apparatus.
Recent studies on sorting in Madin-Darby canine kidney cells show however, that not all glycosphingolipids are directed apically. The lipids of oligodendrocyte and myelin DIGs consist largely of the glycosphingolipids galactocerebroside, sulfatide, and sphingomyelin, as well as cholesterol, lipids that are enriched in myelin. Oligodendrocyte precursor cells as well as mature oligodendrocytes also express low levels of gangliosides, but these are not all sorted to myelin. Oligodendrocyte precursor cells do not associate with the GPI-anchored proteins in DIGs; concomitant with the appearance of the myelin lipids galactocerebroside and sulfatide during maturation of the precursor cell to an oligodendrocyte, DIGs are up-regulated. We suggest that the association of GPI-anchored proteins with these glycosphingolipids and cholesterol in oligodendrocytes is responsible for the direction of these proteins to myelin. The complexes isolated from oligodendrocytes were shown to contain specifically the 120-kDa GPI-anchored but not the 140-kDa transmembrane isoform of NCAM. Since these isoforms have the same extracellular region but are differently anchored in the membrane, this observation supports the targeting hypothesis in which the GPI anchor is responsible for directing the proteins to DIGs and thus to myelin. Accordingly, myelin contains only NCAM 120. In addition, the GPI-anchored protein F3, galactocerebroside and sulfatide appear to be co-localized in double immunofluorescence stainings of cultured oligodendrocytes.Two groups have recently generated mice deficient in the enzyme ceramide galactosyl transferase. These mice are unable to synthesize galactocerebroside and sulfatide. Although they synthesize myelin around appropriate axons, they show severe clinical symptoms and have a shortened life span. It will be interesting to analyze the DIGs from such animals and, in particular, to see if the transport and location of GPI-anchored proteins is abnormal.
The prerequisites for the formation of DIGs are still unresolved. VIP 21/caveolin, which was initially thought to play a role in the formation of such complexes, has been reported to be absent from brainand expressed neither in oligodendrocytes nor myelin.For several cell types, including neuroblastoma cells, lymphocytes, and epithelial cells, glycosphingolipid-rich, detergent-insoluble microdomains have been shown to exist independently of caveolae and the expression of VIP 21/caveolin. For such caveolin-independent microdomains, it is feasible that other unknown proteins might be involved in their formation. A 17-kDa protein has been isolated from detergent-insoluble complexes of oligodendrocytes and myelin (MVP 17 or rMAL; Refs. 63 and 64) that belongs to the family of proteolipid proteins. It was proposed by that it may be involved in regulating protein sorting in myelinating oligodendrocytes. However, the association of GPI-anchored proteins with specific glycosphingolipids and cholesterol may be sufficient to generate detergent-insol-T. Koch and J. Trotter, unpublished results. uble complexes; the saturated character of the fatty acyl chains appears to be very important.
During myelinogenesis the oligodendrocyte synthesizes copious amounts of myelin membrane and segregates specific proteins and lipids to the myelin compartment. For the cytoplasmic myelin protein MBP, it has been shown that the messenger RNA is targeted to the oligodendroglial processes forming the myelin sheath. Additionally, the mechanism for the segregation of myelin proteins and lipids may include a combination of selective signals on proteins destined for myelin as well as retention of specific components in the oligodendroglial cell body.
We show here that oligodendrocyte precursor cells, oligodendrocytes, and central nervous system myelin display a strikingly similar pattern of GPI-anchored proteins, whereas oligodendrocyte plasma membrane and myelin exhibit a markedly different pattern of total proteins. As proposed for epithelial cells, we suggest that the GPI anchor is responsible for the selective association of GPI-anchored proteins with glycosphingolipid-rich microdomains during maturation of oligodendrocytes and targets these molecules to the myelin sheath, thus acting as a myelin sorting signal. GPI-anchored proteins, often involved in cell-cell interactions, are expressed at the external face of the plasma membrane in oligodendrocytes, which in myelin also apposes the axonal membrane. They are thus candidates for molecules involved in recognition and signal transduction between axons and the ensheathing glial cells, a process still largely unresolved at the molecular level. Their insertion in the external leaflet of the lipid bilayer may permit fluidity within the plane of the membrane, thus retaining adhesive contacts during the spiralling of the glial process around the axon and the laying down of the multilamellar sheath.
[fig] FIG. 1: In vitro biosynthetic radiolabeling of oligodendrocyte lineage cells to detect GPI-anchored proteins. A, autoradiogram of [ 3 H]ethanolamine-incorporating proteins synthesized by the Oli-neu cell line (C, lane 2) and ethanolamine-containing proteins released into the culture supernatant (S) without (Ϫ, lane 4) or after (ϩ, lane 3) incubation of the radiolabeled cells with PI-PLC. Lane 1 shows the signal remaining in the cells (C) after treatment with PI-PLC. The heavily labeled band at 45 kDa is probably the protein synthesis elongation factor 1␣ (40). Not all of the ethanolamine-labeled protein is released under these conditions. B, same as A but after biosynthetic radiolabeling of the Oli-neu cell line with [ 3 H]glucosamine hydrochloride. C, biosynthetic radiolabeling of primary cultures of oligodendrocytes, cultured for 2 days, labeled with [ 3 H]glucosamine hydrochloride and subsequent incubation of the cells with PI-PLC. The culture supernatant (S) with (ϩ, lane 9) and without (Ϫ, lane 10) PI-PLC incubation of the cells is shown. D, [ 3 H]ethanolamine-labeled Oli-neu cells were subjected to TX-114 phase separation. The TX-114 detergent phase was treated with PI-PLC, resulting in the specific release of GPI-anchored proteins into the aqueous phase after a second phase separation. The figure shows the [ 3 H]ethanolamine-labeled proteins as visualized by SDS-PAGE and autoradiography in the second detergent phase (lanes 11 and 12) and the second aqueous phase (lanes 13 and 14) with (ϩ) or without (Ϫ) PI-PLC treatment. The band pattern in lane 13 is similar to that in Fig. 1A, lane 3, in which the [ 3 H]ethanolamine-labeled proteins released from intact cells by PI-PLC treatment are shown. Not all of the ethanolamine-labeled protein is released from the detergent phase into the second aqueous phase by PI-PLC treatment under these conditions. C, cells; S, supernatant. [/fig]
[fig] FIG. 3: Oligodendrocyte GPI-anchored proteins are localized in myelin. Biotinylation and TX-114 phase separation of murine CNS myelin proteins (lanes 3 and 4) with (ϩ) or without (Ϫ) PI-PLC treatment shows the same pattern of GPI-anchored proteins as primary oligodendrocyte cultures (lanes 1 and 2). In contrast, if total biotinylated oligodendroglial cell surface proteins (o, lane 5) and total biotinylated myelin proteins (m, lane 6) are compared, or silver-stained proteins of oligodendrocyte membranes (o, lane 7) and myelin (m, lane 8) are compared, myelin displays a restricted and different pattern of proteins. primary oli., primary oligodendrocytes; o, oligodendrocytes; m, myelin. FIG. 4. Incomplete solubilization of NCAM 120 and F3 with TX-100. A, primary oligodendrocytes (cultured for 3 days; left panel) representing a mixture of differentiation stages and murine CNS myelin (right panel) were extracted either at 4 or 37°C with a TX-100containing buffer. After centrifugation, insoluble pellets and supernatants were analyzed by Western blot for the distribution of GPIanchored proteins compared with other transmembrane proteins. A fraction of the GPI-anchored proteins NCAM 120 and F3 remains in the insoluble pellet after extraction at 4°C, while the transmembrane NCAM isoform (NCAM 140) and other oligodendroglial transmembrane proteins MAG, MOG, and PLP are released into the supernatant. Solubilization of myelin is incomplete at 4°C for both GPI-anchored and the main myelin proteins. Extraction at 37°C solubilizes the oligodendrocyte and myelin proteins almost completely. B, the cell line Oli-neu was analyzed as described in A. GPI-anchored NCAM 120 is resistant to complete solubilization with TX-100 at 4°C, while transmembrane NCAM 140 is completely solubilized. P, pellet; SN, supernatant. cursor cells. It is thus possible that association of the GPIanchored proteins with lipids could change during maturation of the cells. Characterization of Cultures of Oligodendrocyte Lineage Cells of Different Maturation Stages-To study more closely changes in association of the GPI-anchored proteins with other components concomitant with maturation of the cells and to specifically investigate their inclusion in detergent-insoluble complexes containing glycosphingolipids and cholesterol, cultures consisting largely of precursor cells or containing more mature cells were studied. The antigens O4, O1, O10, and MBP are sequentially expressed during oligodendrocyte differentiation. Oligodendrocyte cultures 30 h after plating are enriched in precursor cells. These are O4-positive, O1-negative cells with three or more processes, or O4-negative bipolar precursor cells (Refs. 22 and 27; Fig. 5). Many bipolar oligodendrocyte precursor cells do not stain with any of the cell surface markers available [/fig]
[fig] FIG. 5: Morphology and antigen profile of oligodendrocyte primary cultures and the cell line Oli-neu. Primary cultures of oligodendrocyte precursors, cultured for 30 h (left micrograph, phase contrast picture) and oligodendrocytes, cultured for 8 days (middle micrograph, phase contrast picture) were stained by indirect immunofluorescence in parallel with sucrose density gradient experiments (Fig. 7) using O4, O1, O10, and MBP antibodies. Immunopositive cells were counted (10 fields of vision containing at least 400 cells total from duplicate coverslips) and displayed as percentage of positive cells of the total cell number. Standard deviations were less than 5%. Precursor cultures largely represent a pre-O4 differentiation stage. A small percentage of cells in the precursor cultures are positive for O1, whereas the majority of the cells in the mature cultures express the markers for late differentiation stages (O1, MBP, O10). The data shown for the Oli-neu cells (right micrograph, phase contrast picture) were adapted from Jung et al. (30). Bar, 30 M. [/fig]
[fig] FIG. 7: Morphological analysis of DIGs. DIGs isolated from oligodendrocytes (A) and myelin (B) by centrifugation on continuous sucrose density gradients were analyzed by thin section electron microscopy. Myelin DIGs are enriched in multilamellar structures. Bar, 500 nm. [/fig]
|
10.1074/jbc.RA118.007201
|
CCBY
|
6556574
|
31023823
|
s2orc_pubmed_articles
|
Hepatocyte-specific deletion of lysosomal acid lipase leads to cholesteryl ester but not triglyceride or retinyl ester accumulation
Experimental proceduresQuantification of plasma neutral lipids, glycerol, and β-hydroxybutyrate by colorimetric testsTotal CHOL (Cholesterol CHOD-PAP kit, Roche), TG (Triglyceride TM infinity kit, Thermo Scientific), NEFA (NEFA-HR(2) R1 and R2 Set, Wako Chemicals), glycerol (Free glycerol reagent, Sigma), and β-hydroxybutyrate (β-Hydroxybutyrate Colorimetric Assay Kit, Cayman Chemical) levels were analyzed by commercial enzymatic colorimetric kits according to manufacturer's instructions.Isolation of total DNA and analysis of mitochondrial DNA contentTotal DNA of liver tissues (20 mg) from ad libitum fed (chow or VitA/HFD) hep-LAL-ko mice and WT littermates was isolated using DNeasy Blood and Tissue Kit (Qiagen). Evaluation of relative copy number of mitochondrial DNA (mt-DNA) and nuclear DNA (n-DNA) was performed by qPCR according to (1). As marker gene of mt-DNA the mitochondrial cytochrom c oxidase (Mt-Co1) and of n-DNA the NADH dehydrogenase flavoprotein 1 (Ndufv1) were amplified. Primer sequences are listed inTable S2. Expression levels were calculated using the ΔΔCt-method and relative mt-DNA copy numbers were analyzed by determining the mtDNA/n-DNA ratios.
## Preparation of mitochondrial enriched fractions and fatty acid oxidation assay
Preparation of mitochondrial enriched fractions and determination of FA oxidation rates were performed as described [bib_ref] regulates mitochondrial fatty-acid oxidation by reversible enzyme deacetylation, Hirschey [/bib_ref]. Briefly, liver (big lope) from hep-LAL-ko mice and WT littermates fed a standard chow diet or VitA/HFD were excised and washed in ice cold STE buffer (0.25 M sucrose, 10 mM Tris, 1 mM EDTA, 20 µg/ml leupeptin, 2 µg/ml antipain, and 1 µg/ml pepstatin). Tissues were homogenized by 4 strokes using a glass douncer and centrifuged for 10 min at 450 x g and 4°C. Infranatants, excluding floating fatty layer, were collected and aliquoted (=liver homogenate). For preparation of a mitochondrial enriched fraction, liver homogenates were centrifuged for 15 min at 3,000 x g and 4°C. Resulting pellets were washed twice in STE buffer and re-suspended in 100 µl STE buffer. Protein concentrations of liver homogenates and mitochondrial enriched fractions were determined by Bio-Rad protein assay (Bio-Rad) according to manufacturer's instructions using BSA as standard. For FA oxidation assay, mitochondria were incubated with the reaction mix (100 mM sucrose, 10 mM Tris, 5 mM KH 2 PO 4 , 0.2 mM EDTA, 0.3% BSA, 80 mM KCl, 1 mM MgCl 2 , 2 mM L-carnitine, 0.1 mM malate, 0.05 mM CoA, 2 mM ATP, 1 mM DTT, 100 μM palmitic acid, and 0.1 μCi/reaction [1-14 C]-palmitic acid) for 30 min and 350 rpm at 37°. Then, the reaction mix was transferred into a new tube containing 70% perchloric acid and a Whatman® paper soaked with 50 μl of 5 N NaOH placed into the lid. CO 2 was trapped for 60 min at RT. After removing the filter papers, tubes were centrifuged for 10 min at max. speed. The amount of 14 C in the CO 2 (filter papers) and acid soluble metabolites (ASM) (supernatants) fractions were determined by liquid scintillation counting. . Unchanged plasma lipid and glycerol levels of mice lacking LAL specifically in hepatocytes. Blood was collected from ad libitum fed and overnight fasted, age-(5 months) and gender-matched (male) hepatocyte-specific LAL-deficient mice (hep-LAL-ko) and littermates (WT). (A) Total cholesterol (total CHOL), (B) triglyceride (TG), (C) non-esterified fatty acid (NEFA), and (D) glycerol plasma levels were determined by commercial kits. For the measurement of (E) retinol (ROH) and (F) retinyl palmitate (RP) plasma samples were n-hexane extracted and analyzed by HPLC-FD. Data are presented as mean ± S.D. for duplicate determinations (n=4 for all groups). Statistically significant differences were determined between genotypes and feeding status by Student´s unpaired t-test (two tailed; *.. p < 0.05, **.. p < 0.01). . Statistically significant differences were determined between genotypes and treatment by Student´s unpaired t-test (two tailed; *.. p < 0.05, **.. p < 0.01, ***.. p < 0.001).
## Figure s4: mitochondrial protein content and fatty acid oxidation rate is unaltered in hepatocyte-specific lal-deficient mice upon vitamin a excess/hfd feeding. (a-c) female
hepatocyte-specific LAL deficient (hep-LAL-ko) mice and littermates (WT) 5 months of age were ad libitum fed a standard chow or vitamin A excess (100,000 IU VitA)/high fat diet (HFD) for 3 weeks. (A) Liver homogenates (450 × g supernatant) were prepared and subjected to Western blotting analyses using antibodies specific against mitochondrial marker protein NADH:ubiquinone oxidoreductase core subunit S1 (NDUFS1) and mitochondrially encoded cytochrome c oxidase I (MT-CO1). GAPDH and Coomassie blue stain were used as loading controls. (B) Liver was homogenized and mitochondrial enriched fractions (3,000 × g pellet) were prepared. Western blot analysis was performed using antibodies specific against carnitine O-palmitoyltransferase 1A (CPT1A) and loading control Cytochrome c oxidase 4 (COX4). (C) Total DNA of liver was isolated and qPCR was performed. Relative mt-DNA content was calculated from copy numbers of the mitochondrial encoded Mt-Co1/nuclear DNA-encoded Ndufv1 gene ratios. (D) Male hep-LAL-ko and WT mice, 5 months of age, were fed a VitA/HFD for 3 weeks. Mice were fasted overnight, liver was freshly excised and mitochondrial enriched fractions were prepared. Fatty acid oxidation assay using 14 C-palmitic acid as substrate was performed. Trapped 14 C-CO 2 and 14 C-acid soluble metabolites (ASM) were determined by scintillation counting. Data are presented as mean ± S.D. for duplicate determinations (n=3-8 for hep-LAL-ko and n=3-8 for WT on VitA/HFD). Statistically significant differences were determined between genotypes by Student´s unpaired t-test (two tailed; **.. p < 0.01). : Primer sequences used for determination of relative gene expression levels and mt-DNA and n-DNA contents.
## Mt-co1
Mitochondrially encoded cytochrome c oxidase I NC_005089.
[fig] Figure S2: Vitamin A excess/HFD feeding decreases phospholipid levels in hepatocyte-specific LAL-deficient mice. Female hepatocyte-specific LAL deficient mice (hep-LAL-ko) and littermates (WT) 5 months of age were ad libitum fed a standard chow or vitamin A excess (100,000 IU VitA)/high fat diet (HFD) for 3 weeks. Lipids of liver homogenates were Folch-extracted and (A) phosphatidylcholine (PC), (B) phosphatidylinositol (PI), (C) sphingomyelin (SM), (D) lysophosphatidylcholine (LPC), (E) lysophosphatidylinositol, and (F) lysophosphatidylethanolamine (LPE) were analyzed by QQQ/MS. Date were normalized to g tissue and are presented as mean ± S.D.of duplicate determinations (n=3 of chow fed control mice; n=5 for hep-LAL-ko and n=7 for WT on VitA/HFD). Statistically significant differences were determined between genotypes and treatment by Student´s unpaired t-test (two tailed; *.. p < 0.05, **.. p < 0.01, ***.. p < 0.001). [/fig]
[fig] Figure S3: Vitamin A excess/HFD feeding leads to comparable changes in plasma lipid and glycerol levels of hepatocyte-specific LAL-deficient mice and WT littermates. Female hepatocytespecific LAL deficient (hep-LAL-ko) mice and littermates (WT) 5 months of age were fed a vitamin A excess (100,000 IU VitA)/high fat diet (HFD) for 3 weeks. Blood was drawn from ad libitum fed mice before (=basal) and after 3 weeks of VitA/HFD feeding. (A) Total cholesterol (CHOL), (B) triglyceride (TG), (C) non-esterified fatty acid (NEFA), (D) glycerol, and (E) β-hydroxybutyrate levels were determined by commercial kits. For (F) retinol (ROH) measurement, plasma lipids were nhexane extracted and analyzed by HPLC-FD. Data are presented as mean ± S.D. for duplicate determinations (n=5-11 for hep-LAL-ko and n=7-13 for WT mice) [/fig]
[fig] Figure S5: Breeding strategy for the generation of the LAL-deleter mouse strain (= mice globally lacking functional LAL) and validation of the LAL knockout. (A) Lipa tm1a mice were crossed with mice expressing Cre recombinase under the control of the CMV promoter. Offspring positive for the Lipa deleter allele globally lack functional LAL (LAL-del). (B-C) Liver was excised. (B) mRNA was transcribed and expression of Lipa was determined by qPCR. (C) Macroscopic liver depiction. Scale bars represent 1 cm. Expression levels were calculated by the Ct-method using 36b4 as ribosomal housekeeping gene. Lipa expression was normalized to WT expression levels. Data are presented as mean + S.D. for duplicate determinations (n=4 for both LAL-del and WT mice). Abbreviation: n.d.= not-detectable [/fig]
[table] Table S2: Primer sequences for the genotyping of hep-LAL-ko mice, LAL-del, and respective WT littermates. [/table]
[table] Table S3: Antibodies used for immunoblotting. [/table]
|
10.1186/1477-7827-10-59
|
CCBY
|
3495228
|
22905699
|
s2orc_pubmed_articles
|
A comparison of human chorionic gonadotropin and luteinizing hormone releasing hormone on the induction of spermiation and amplexus in the American toad (Anaxyrus americanus)
Background: Captive breeding programs for endangered amphibian species often utilize exogenous hormones for species that are difficult to breed. The purpose of our study was to compare the efficacy of two different hormones at various concentrations on sperm production, quantity and quality over time in order to optimize assisted breeding. Methods: Male American toads (Anaxyrus americanus) were divided into three separate treatment groups, with animals in each group rotated through different concentrations of luteinizing hormone releasing hormone analog (LHRH; 0.1, 1.0, 4.0 and 32 micrograms/toad), human chorionic gonadotropin (hCG; 50, 100, 200, and 300 IU), or the control over 24 hours. We evaluated the number of males that respond by producing spermic urine, the sperm concentration, percent motility, and quality of forward progression. We also evaluated the effects of hCG and LHRH on reproductive behavior as assessed by amplexus. Data were analyzed using the Generalized Estimating Equations incorporating repeated measures over time and including the main effects of treatment and time, and the treatment by time interaction.Results: The hormone hCG was significantly more effective at stimulating spermiation in male Anaxyrus americanus than LHRH and showed a dose-dependent response in the number of animals producing sperm. At the most effective hCG dose (300 IU), 100% of the male toads produced sperm, compared to only 35% for the best LHRH dose tested (4.0 micrograms). In addition to having a greater number of responders (P < 0.05), the 300 IU hCG treatment group had a much higher average sperm concentration (P < 0.05) than the treatment group receiving 4.0 micrograms LHRH. In contrast, these two treatments did not result in significant differences in sperm motility or quality of forward progressive motility. However, more males went into amplexus when treated with LHRH vs. hCG (90% vs. 75%) by nine hours post-administration.Conclusion: There is a clear dichotomy between the two hormones' physiological responses on gamete production and stimulation of amplexus. Understanding how these two hormones influence physiology and reproductive behaviors in amphibians will have direct bearing on establishing similar breeding protocols for endangered species.
# Background
The worldwide decline in amphibians has resulted in the establishment of numerous captive breeding assurance colonies by zoological, aquaria and governmental organizations. These assurance colonies are meant to serve as Arks that retain some portion of the populations' biological diversity should a species go extinct in the wild. The goal of these arks is to repatriate the animals once suitable habitat is restored, a causative agent of the decline (e.g. disease) is no longer a threat, or simply to supplement a fragmented population with additional brood stock. Thus, the near-term objective for these amphibian assurance colonies is to keep the animals alive long enough so that founder populations can reproduce and be sustainable. While there are certainly numerous breeding success stories, there are also several programs that require exogenous hormone stimulation of individuals for reproduction (e.g. Anaxyrus baxteri, Anaxyrus boreas boreas, and Peltophryne lemur) and may require assisted reproductive technologies (ART) such as invitro fertilization (IVF) for their continued propagation (e.g. Rana sevosa) [bib_ref] Applied reproductive technologies and genetic resource banking for amphibian conservation, Kouba [/bib_ref]. For example, the endangered Wyoming toad (Anaxyrus baxteri) has only reproduced naturally twice in captivity and at least one gender has always received hormone therapy; yet, nearly 20,000 tadpoles were produced for release in 2009 and more than 100,000 for the program over a ten year period (personal communication, Bruce Foster SSP coordinator). Unfortunately, for many captive assurance species there is a growing need for ART because of the lack of knowledge about their ecology, especially the environmental and social cues that stimulate reproduction and how nutrition correlates to reproductive fitness.
Exogenous hormones have been used to stimulate amphibian breeding for more than 50 years [bib_ref] Applied reproductive technologies and genetic resource banking for amphibian conservation, Kouba [/bib_ref] [bib_ref] Artificial fertilization for amphibian conservation: Current knowledge and future considerations, Kouba [/bib_ref] [bib_ref] The male frog, Rana pipiens, as a new test animal for early..., Wiltberger [/bib_ref]. The two hormones commonly employed for stimulating spermiation and ovulation in live animals are luteinizing hormone releasing hormone (LHRH) analog and human chorionic gonadotropin (hCG) with their effects extensively reviewed by several investigators [bib_ref] Applied reproductive technologies and genetic resource banking for amphibian conservation, Kouba [/bib_ref] [bib_ref] Artificial fertilization for amphibian conservation: Current knowledge and future considerations, Kouba [/bib_ref] [bib_ref] The USSR programme for breeding amphibians, including rare and endangered species, Goncharov [/bib_ref]. Whitakerprovides a detailed review of trial and error hormone concentrations selected for assisted breeding in 16 different anuran species. His summary clearly revealed that the range of hormone concentrations used, most of them empirically chosen, appeared to be species-specific but were often passed down by habit or tradition (about 33% were not published studies). Similarly, Goncharov et al. [bib_ref] The USSR programme for breeding amphibians, including rare and endangered species, Goncharov [/bib_ref] stimulated spawning in 37 different anuran species, using primarily LHRH but also hCG, and showed that successful gamete production for each species varied by the dosage (2 μg/kg up to 8 mg/ kg), the number of treatments and the injection interval. In addition to the concentration of hormone appearing to vary widely between species, the response time and duration of gamete production following hormone administration has rarely been studied in detail for any species.
Although several good reviews exist describing how hormone administration for assisted breeding is speciesspecific [bib_ref] Applied reproductive technologies and genetic resource banking for amphibian conservation, Kouba [/bib_ref] [bib_ref] Artificial fertilization for amphibian conservation: Current knowledge and future considerations, Kouba [/bib_ref] [bib_ref] The USSR programme for breeding amphibians, including rare and endangered species, Goncharov [/bib_ref] , very few publications are available showing how sperm production and quality or the animals' reproductive behavior are affected with variation in hormone concentration. This information is critical to captive breeding programs within zoos or aquariums that are regularly employing exogenous hormones for breeding as many pairings often fail to elicit amplexus behavior, exhibit poor egg fertilization rates, or animals release gametes in the absence of the other gender. For the most part, zoos and aquariums have typically used LHRH for assisted natural breeding, while academic studies employ hCG for collecting gametes to study fertilization and early embryonic development. The disparity between these two different approaches by institutional type may be related to their specific objectives. Whereas most university studies are using hormones as a tool to collect gametes for assisted fertilization or IVF, zoos are attempting to use the same compounds for assisted semi-natural breeding. The captive husbandry guidelines for Anaxyrus baxteri, Peltophryne lemur, and Anaxyrus boreas boreas all recommend LHRH concentrations in the range of 0.1-0.3 μg/g body weight. Having a better understanding of whether these programs are indeed using the optimal hormone concentration or correct hormone combination for breeding may improve the poor fertilization rates these programs frequently experience. The present study was conducted on the common American toad (Anaxyrus americanus), as a model species, for determining baseline hormone concentrations and potential toxicity issues before testing on endangered species. Obringer et al. [bib_ref] Characterization of the spermiation response, luteinizing hormone release and sperm quality in..., Obringer [/bib_ref] tested three different concentrations of LHRH on sperm production in Anaxyrus americanus and found a dose-dependent effect; however, a direct comparison of LHRH with hCG was not evaluated and sperm concentration was low compared to what we have found using hCG on another common species, Anaxyrus fowleri [bib_ref] Applied reproductive technologies and genetic resource banking for amphibian conservation, Kouba [/bib_ref].
The objectives of this study were to compare the efficacy of four different concentrations of LHRH and hCG on sperm production over time. We measured the number of male Anaxyrus americanus that respond by producing spermic urine, and evaluated the quantity and quality of the semen by measuring the concentration, motility and forward progression of the spermatozoa. Once the best hormone concentrations from our treatment groups were determined, we then evaluated their effect on inducing reproductive behavior over time as determined by the number of animals in amplexus with females. The results from this study may have direct bearing on establishing more successful breeding protocols for numerous endangered species that a global community of zoos and aquariums are working with including Rana sevosa, Anaxyrus boreas boreas, Anaxyrus baxteri, Atelopus zeteki, and Peltophryne lemur.
# Methods
## Animals
Adult male American toads (Anaxyrus americanus) were collected from Northern Kentucky during the breeding season between May and June. The presence of calling, nuptial pads and pigmented vocal sacs indicated adult reproductive status. The toads were housed in groups of four in covered plastic enclosures (30.48 H x 38.1 W x 55.88 L cm) maintained at 26-28°C with water bowls and substrate. Animals were fed crickets three times a week that had been dusted with Reptical W and vitamin D supplementation. Males were selected of similar weight for the study with an average of 44.3 ± 2.4 g. Experiments were conducted in the lab from September-March, outside of normal breeding season to avoid any additional influence of seasonality on the effect of the hormones on spermiation. The project was approved by the Memphis Zoo's IACUC (#001-11-01).
## Sperm collection and analysis
Spermic urine was collected from individual male toads following hormone administration treatments by holding the animal over a 150 mm Petri dish and gently spreading the hind legs apart with thumb and index finger which typically resulted in spermiation within 1-3 min as previously described [bib_ref] Applied reproductive technologies and genetic resource banking for amphibian conservation, Kouba [/bib_ref]. Spermic urine was immediately placed into a 1.5 ml eppendorf tube, the volume measured and the sperm evaluated by placing 10 μl on a slide and viewed at 400x magnification using an Olympus CX41 phase-contrast microscope. Variables measured included number of males producing sperm, percent motility, concentration and quality of forward progression (FP). Forward progression was based on a qualitative scale of 0 to 5, where 0 = no movement, and 5 = rapid forward movement. Spermatozoa exhibiting beating flagella were considered motile, even if no FP was observed. This qualitative scale is commonly used to assess the quality of sperm progressive motility in mammals [bib_ref] Sperm cryopreservation in three species of endangered gazelles (Gazella cuvieri, G. dama..., Garde [/bib_ref] as well as amphibians [bib_ref] Hormonal priming, induction of ovulation and in-vitro fertilization of the endangered Wyoming..., Browne [/bib_ref] [bib_ref] Structural and functional aspects of Bufo americanus spermatozoa: effects of inactivation and..., Kouba [/bib_ref] and is used to develop a sperm motility index (SMI) as previously described [bib_ref] Sperm cryopreservation in three species of endangered gazelles (Gazella cuvieri, G. dama..., Garde [/bib_ref]. The SMI was calculated as follows [% individual motility + (quality of motility x 20)] x 0.5. Sperm concentration was evaluated using a Neubaeur hemacytometer. In brief, 10 μl of sperm was mixed with 90 μl of 0.9% saline to inhibit sperm motility for counting; 10 μl of diluted sperm was placed in each chamber and the four corner squares of the grid counted for sperm. The average count for the two sides of the hemacytometer were then multiplied by the dilution factor and the conversion factor of 2500 and concentration expressed as number/ml.
## Study 1: evaluation of hcg administration on sperm collection
Study 1 examined the dose-dependent effects of hCG administration on sperm production and motility characteristics over time. Lyophilized hCG (Sigma-Aldrich Co.; 2500 IU; C1063) was rehydrated using various amounts of sterile saline such that 100 μl contained 50, 100, 200, or 300 IU of hormone (expressed on an average body weight (BW) basis the treatments would be 1.13, 2.28, 4.51 and 6.77 IU/g BW). Following intraperitoneal injection, animals were housed in plastic containers with aged tap water, approximately 2-3 cm deep, such that their abdomen was immersed to promote urine formation. Urine was collected from live American toads (n = 16/treatment) immediately after injection to confirm the absence of sperm and was subsequently collected at 3, 5, 7, 9, 12 and 24 h after injection and evaluated for the presence of spermatozoa, as previously described [bib_ref] Artificial fertilization for amphibian conservation: Current knowledge and future considerations, Kouba [/bib_ref] [bib_ref] Characterization of the spermiation response, luteinizing hormone release and sperm quality in..., Obringer [/bib_ref] [bib_ref] Structural and functional aspects of Bufo americanus spermatozoa: effects of inactivation and..., Kouba [/bib_ref]. In some cases animals did not give a urine sample, thus the presence of sperm was not determinable and the percentage of responders was adjusted based on the number of urine collections obtained.
## Study 2: evaluation of lhrh administration on sperm collection
For study 2 we evaluated the dose-dependent effects of LHRH administration on sperm production and motility characteristics over time. Lyophilized LHRH (Sigma-Aldrich Co.; 1 mg; L4513) was rehydrated using various amounts of sterile saline such that 20 μl contained 0.1, 1.0, 4.0, or 32 μg of hormone (expressed on an average BW basis the treatments would be 0.002, 0.022, 0.090, and 0.722 μg/g BW). The LHRH chosen for this study is a synthetic analog ([des-Gly10, D-Ala6]-LHRH ethylamide acetate) and is the standard LHRH used in many studies to date for inducing spermiation in amphibians as well as captive breeding programs worldwide. Following intra-peritoneal injection, animals were housed in plastic containers with aged tap water, approximately 2-3 cm deep, such that their abdomen was immersed to promote urine formation. Similar to Study 1, urine was collected from live American toads (n = 20/treatment) at 0, 3, 5, 7, 9, 12 and 24 h after injection and evaluated for the presence of spermatozoa, as previously described. There are four additional animals per treatment group in study 2 compared to study 1, as study 2 was based on an earlier pilot study conducted for power analysis, which included these four additional animals.
## Study 3: effects of hcg and lhrh on amplexus behavior
Study 3 was designed to evaluate how the two hormones, hCG and LHRH, affect reproductive behavior by counting the number of males in amplexus (grasping of the female by the male) at the same time points that sperm production and quality had been evaluated in Studies 1 and 2. We evaluated amplexus behavior in toads receiving either 300 IU hCG (n = 12/trt) or 4 μg LHRH (n = 11/trt), compared to control (n = 12/trt) in which animals received 100 μl of sterile saline only. Hormone concentrations were determined in Studies 1 and 2, from which we chose the best concentration based on number of responders and sperm quantity. Following hormone administration, each male was paired with a single female Anaxyrus americanus in a small plastic shoebox enclosure (34.3 cm x 20.3 cm x 12.7 cm) assuring visual and physical contact throughout the duration of the experiment, and the number of animals in amplexus recorded for each time point by visual inspection at 0, 3, 5, 7, 9, 12 and 24 hrs for the three treatments. Females were not treated with hormones.
# Statistical analysis
Data were analyzed using the Generalized Estimating Equations (GEE) incorporating repeated measures over time and including the main effects of treatment and time, and the treatment by time interaction. We used GEE to accommodate for correlated longitudinal data over time using the linear model, which could also account for missing samples at specific time points due to the lack of urine collection at those time points. For the GEE regression parameters we report the Wald Chisquare as W. For binary data (sperm production and amplexus in response to hormones) we used the binary logistic model within GEE. Percent motility data were arcsine transformed prior to analysis. Differences between treatment means were evaluated by Fisher's Protected Least Significant Difference test. All tests were two-tailed. Only those animals producing sperm were used for further treatment comparisons within concentration, motility, FP and sperm motility index (SMI). Values are expressed as Mean ± SEM. We considered differences significant when P < 0.05 and trending towards significance when P was between 0.05 and 0.10. All statistic analyses were performed using SPSS PASW 18 for Windows.
# Results
## Study 1: effects of hcg on sperm characteristics
Male Anaxyrus americanus administered with four different concentrations of hCG produced sperm in a dosedependent manner, with 100% of the animals producing sperm when given 300 IU hCG [fig_ref] Figure 1: Percentage of male Anaxyrus americanus producing sperm over 24 hrs after hormone... [/fig_ref]. All of the toads receiving 300 IU hCG were producing sperm at the first collection time point, 3 hrs post-administration of hormone. In comparison, the percentage of toads producing sperm at 3 hrs with 200 IU (36%), 100 IU (0%) or 50 IU hCG (8%) was much less [fig_ref] Figure 1: Percentage of male Anaxyrus americanus producing sperm over 24 hrs after hormone... [/fig_ref]. No animals in the control group, receiving sterile saline only, produced sperm. We found a significant difference in the number of male toads producing sperm depending on the hCG concentration (W 3 = 161.38, P < 0.0005) with the number of males producing sperm being highest after administrating 300 IU hCG (P < 0.0005 for all pairwise comparisons between 300 IU hCG and any other concentration). Only one toad given 50 IU hCG produced sperm; thus, we did not include the 50 IU hCG concentrations in the analyses for sperm characteristics (motility, forward progression or concentration). The number of male toads producing sperm did not significantly vary from 3 to 12 hrs (P > 0.05), but decreased at 24 hours (P < 0.0005). At 24 hours post-administration of hormone, none of the toads treated with 50 or 100 IU hCG were releasing sperm and the number of toads having received 200 IU hCG declined in sperm production from 69% to 14% between 12 and 24 hrs, respectively [fig_ref] Figure 1: Percentage of male Anaxyrus americanus producing sperm over 24 hrs after hormone... [/fig_ref]. Yet, 53% of the toads administered 300 IU hCG were still producing sperm 24 hrs post-administration.
Sperm concentration of male Anaxyrus americanus was highest after administering 300 IU hCG [fig_ref] Figure 2: Sperm concentration [/fig_ref] ; W 2 = 8, P = 0.018) compared to other hormone regimens. Furthermore, the concentration varied at different times after hormone administration (W 6 = 101.48, P < 0.0005). For example, after administering 300 IU hCG, sperm concentration was significantly higher (P < 0.05) at 7 and 9 hrs than at any other time points tested. There was a treatment by time interaction (W 10 = 59.73, P < 0.0005), indicating that the magnitude and direction of the response to the hormone levels varied over the collection periods. [fig_ref] W 3 = 36: [/fig_ref] shows the sperm motility after administration of 100-300 IU hCG. Sperm motility differed depending on the hCG concentration (W 2 = 12.98, P = 0.002) and the time after hormone administration (W 5 = 76.65, P < 0.0005). The treatment by time interaction for motility trended towards significance (W 8 = 13.76, P = 0.09). Pair-wise comparisons of motility data differed significantly between the 300 IU hCG and lower concentrations of hCG (100 and 200 IU; P < 0.01 for both comparisons). Furthermore, the quality of FP sperm motility for male American toads differed among the 100, 200, and 300 IU concentrations (W 2 = 8.41, P = 0.015). Sperm FP motility was highest when administering 300 IU hCG [fig_ref] W 3 = 36: [/fig_ref] , with the 200 IU and 300 IU concentrations almost differing statistically (P = 0.053), yet both of these concentrations differed from the 100 IU concentration (P < 0.05). Forward progressive sperm motility was highest between 5 hrs and 12 hrs after hormone administration [fig_ref] W 3 = 36: [/fig_ref] ; W 5 = 102.68, P < 0.0005). Similar to motility, the treatment by time interaction for FP trended towards significance (W 8 = 14.46, P = 0.07).
## Study 2: effects of lhrh on sperm characteristics
We found significant differences in the number of male toads producing sperm depending on the LHRH concentrations tested, 0.1-32 μg/per animal [fig_ref] Figure 1: Percentage of male Anaxyrus americanus producing sperm over 24 hrs after hormone... [/fig_ref] ; [fig_ref] Figure 2: Sperm concentration [/fig_ref]. Sperm concentration was also similar between 3 hrs and 24 hrs after hormone administration (P > 0.05 for all pairwise comparisons). [fig_ref] W 3 = 36: [/fig_ref] shows that sperm motility was not dependent on the dose of LHRH tested (W 3 = 1.3, P = 0.73;). However, there were significant differences in sperm motility between different times after hormone administration (W 5 = 31.4, P < 0.0005). Sperm motility differed between 3 hrs and 9 hrs (P = 0.012), was similar between 5 hrs and 12 hrs (P > 0.05 for all pair-wise comparisons), and decreased between 12 hrs and 24 hrs (P = 0.038). In addition, a treatment by time interaction was detected for sperm motility (W 14 = 21733, [fig_ref] W 3 = 36: [/fig_ref] shows the results of LHRH hormone administration on sperm Forward Progression. Forward progressive sperm motility of male Anaxyrus americanus was significantly different depending on the concentration of LHRH (W 3 = 11.31, P = 0.01). Specifically, forward progression differed between dosages of 4 μg LHRH and either 0.1 μg LHRH or 32 μg LHRH (P < 0.05 for the two pair-wise comparisons) but not between 4 μg LHRH and 1 μg LHRH (P = 0.15). There were significant differences between different times after hormone administration (W 5 = 28.99, P < 0.0005), as well as a treatment by time interaction (W 14 = 139.38, P < 0.0005). Specifically, forward progression increased from 3 hrs to 5 hrs (P = 0.016), remained similar between 5 hrs and 12 hrs (P > 0.05) and then decreased between 12 hrs and 24 hrs (P = 0.001).
## Comparison between 300 iu hcg and 4 μg lhrh
Next, we chose the best hormone concentrations from each treatment group (hCG or LHRH) to compare sperm characteristics. While 300 IU hCG was clearly the best concentration in our experimental design for the first treatment group, the LHRH results were less clear. A concentration of 4 μg LHRH was chosen for direct comparison as this hormone concentration is closest to the level commonly used in breeding programs (for 25-40 g/body weight toads) and there was no difference in the number of animals responding, concentration, or motility between 4 μg and the 1 μg or 32 μg LHRH treatments. We found that male toads administered hCG had a greater number of responders (W 1 = 63.15, P < 0.0005), and much higher sperm concentration (W 1 = 18.6, P < 0.0005) than those males receiving LHRH. Average sperm concentration was 12 times higher in spermic urine collected from males receiving 300 IU hCG (12 x 10 6 /ml) compared to those receiving 4 μg LHRH (1 x 10 6 /ml). In contrast, these two treatments did not result in significant differences in sperm motility (W 1 = 0.35, P = 0.55) or the quality of forward progressive motility (W 1 = 3.63, P = 0.057). Similarly sperm motility index, which is an overall measure to evaluate the quality of sperm, did not differ between the hCG and LHRH concentrations evaluated [fig_ref] Figure 4: A comparison of the sperm motility index [/fig_ref] ; W 1 = 2.54, P = 0.11). [fig_ref] Figure 5: The percentage of male Anaxyrus americanus amplexing females over time following hormone... [/fig_ref] shows the reproductive behavioral response of male toads injected with either hCG (300 IU) or LHRH (4 μg). No male toads were observed in amplexus following administration of the control (saline only). The number of pairs amplexing was not significantly different between the control and either treatment at 3 hrs or 5 hrs post-hormone administration (Fisher's probability tests: P > 0.05 for all tests). In contrast, at 7 hrs, 9 hrs, 12 hrs, and 24 hrs, the number of pairs amplexing was higher when males were administered either 300 IU hCG or 4 μg LHRH compared to the control group (Fisher's probability tests: P ≤ 0.01 for all tests). When considering the overall interaction effect of treatment over time between the two treatments, hCG or LHRH, there was not a significant difference in the number of pairs' amplexing for 300 IU hCG and 4 μg LHRH (GEE: W 1 = 0.35, P = 0.35). However, the number of males that were observed in amplexus was significantly different at specific individual time points after The percentage of male Anaxyrus americanus amplexing females over time following hormone administration. Males were treated with the optimal hormone concentrations from Study 1 and Study 2, specifically, 300 IU hCG (n = 12/trt) or 4.0 μg LHRH (n = 11/trt). Females were not treated with any hormones. treatment (W 4 = 27.34, P < 0.0005). The number of toads observed in amplexus was lower at 3 hrs and 5 hrs post-hormone administration than at any later time (P < 0.0005 for all pair-wise comparisons). As shown in [fig_ref] Figure 5: The percentage of male Anaxyrus americanus amplexing females over time following hormone... [/fig_ref] , the number of toads observed in amplexus was highest at 9 hrs post-hormone administration (91% when treated with LHRH and 75% when treated with hCG). Although the number of toads observed in amplexus was still relatively high at 24 hrs (73% when treated with LHRH and 50% when treated with hCG), there was a significant drop between 9 hrs and 24 hrs (P = 0.015; [fig_ref] Figure 5: The percentage of male Anaxyrus americanus amplexing females over time following hormone... [/fig_ref]. No treatment by time interaction was detected (W 4 = 2.83, P = 0.59).
## Study 3: effects of hcg and lhrh on amplexus behavior
# Discussion
Our research demonstrates that amphibian captive breeding programs that seek to utilize exogenous hormones for assisted reproduction need to carefully evaluate the optimal hormone concentration for their particular species. If the goal is assisted natural breeding, rather than in vitro fertilization, meticulous consideration must also be given to which hormones stimulate reproductive behaviors as well. In our study, the best hCG concentration tested, where 100% of the animals responded by producing sperm, was 300 IU. This same concentration of hCG has also been found to be extremely effective in stimulating spermiation in Anaxyrus baxteri [bib_ref] Applied reproductive technologies and genetic resource banking for amphibian conservation, Kouba [/bib_ref] [bib_ref] Artificial fertilization for amphibian conservation: Current knowledge and future considerations, Kouba [/bib_ref] [bib_ref] Hormonal priming, induction of ovulation and in-vitro fertilization of the endangered Wyoming..., Browne [/bib_ref] , Anaxyrus boreas boreas [bib_ref] Applied reproductive technologies and genetic resource banking for amphibian conservation, Kouba [/bib_ref] , Anaxyrus fowleri [bib_ref] Applied reproductive technologies and genetic resource banking for amphibian conservation, Kouba [/bib_ref] , Anaxyrus houstonensis (personal communication, Paul Crump) and Peltophryne lemur (personal communication, Andrew Lentini). Although this concentration of hCG is extremely effective in stimulating male Bufonidae to produce spermic urine, species-specific differences can be found in the density of sperm produced. In our lab, we have found that seven Bufonidae species (in the 20-75 g range) produce spermic urine after hCG administration of 300 IU or higher; however, one Ranid species of similar weight (Rana pipiens) produced low quantities of sperm with often variable responses when treated with similar hCG concentrations [bib_ref] Artificial fertilization for amphibian conservation: Current knowledge and future considerations, Kouba [/bib_ref]. This further supports our concept that developing a protocol in one taxonomic species may not extrapolate exactly to another, but that fine-tuning is necessary on hormone efficacy to optimize a breeding strategy. However, the basic principles learned about one taxa may provide a good starting point for more developed studies.
By testing various concentrations of hCG from 50-300 IU we were able to show a clear dose-dependent and dramatic increase in the number of male Anaxyrus americanus producing sperm, along with an increase in concentration of sperm. In contrast, LHRH did not show a clear dose-dependent effect on the sperm parameters measured. None of the LHRH treatments tested produced as high a concentration of sperm as shown with the 100-300 IU hCG treatments. Moreover, the number of animals producing sperm in response to LHRH (~35%) was much less than the 100% achieved with 300 IU hCG. Interestingly, the individual sperm characteristics, motility, forward progression, and SMI from the optimal LHRH treatment group were not different from the 300 IU hCG group. This suggests that hormone type and concentration can be altered to induce more animals to produce greater quantities of sperm, but that these hormones did not directly affect the movement patterns or quality of mature ejaculated spermatozoa. While it is possible that increasing the concentration of hCG above 300 IU/animal might increase sperm concentration or possibly the number of animals in amplexus, we chose to stop at a point where 100% of the animals were producing sperm of good quality that would lead to high rates of natural or artificial fertilization without compromising the health of the animals by over-dosing them. Furthermore, we recognize that our behavioral responses were not optimized (75% of the pairs in amplexus with hCG compared to 91% with LHRH) and that higher doses may have increased numbers of males exhibiting amplexus. Future studies will examine the effect that combinations of the two hormones might have on reproductive behaviors at these concentrations without increasing hormone levels to possibly dangerous concentrations. This method of cautious experimentation and our stopping point within our treatments is necessary considering we are outlining a hormonal strategy for developing such protocols in critically endangered toad species where losses to the assurance colonies could be catastrophic.
Very few studies have been conducted in male amphibians directly testing these two different hormones at four or more concentrations. Michael et al. [bib_ref] Induced ovulation and egg deposition in the direct developing anuran Eleutherodactylus coqui, Michael [/bib_ref] evaluated different concentrations of hCG and LHRH in female Eluetherodactylus coqui and, in contrast to our results with male toads, found that LHRH stimulated more female coqui to ovulate than the hCG concentrations they tested. The difference between these results and ours are likely due to species or gender differences in the response to these hormones. Our lab has found that toads within the family Bufonidae tend to respond in a similar fashion to hCG across species; yet, Rana pipiens seem to respond better to LHRH [bib_ref] Artificial fertilization for amphibian conservation: Current knowledge and future considerations, Kouba [/bib_ref] and these results are often concentration dependent. Understanding the a priori hormone levels that stimulate optimal spermiation or ovulation is critical for long-term breeding programs where captive facilities may utilize these protocols for endangered or threatened species. Without this prior knowledge on how a species will respond to varying hormone concentrations (possibly spanning several orders of magnitude), investigators may be tempted to conclude that their species' lack of ovulation or spermiation may limit hormone induction as a conservation tool. For example, Mann et al. [bib_ref] Hormonal induction of spermiation, courting behavior and spawning in the southern bell..., Mann [/bib_ref] concluded that a suite of hormones tested, at a single concentration, did not work well in Litoria raniformis. However, a more robust level of hormone testing may have yielded a larger number of animals ovulating, and could be explored.
Recent studies on Anaxyrus boreas boreas suggest that higher LHRH or hCG concentrations, or even priming regimens, may be more effective than what was historically employed in the captive breeding and reintroduction program for this threatened species (Natalie Calatayud and Kevin . Knowing that the optimal hormone concentration could be orders of magnitude higher than historically employed for inducing spermiation or ovulation in this species, raises the question of how to interpret other studies that are dependent upon hormone therapy to answer other physiological processes, such as the importance of hibernation in captive breeding [bib_ref] Effects of age, weight, hormones, and hibernation on breeding success in boreal..., Roth [/bib_ref]. Often times, how investigators arrived at optimal hormone levels is not clear and reports indicate that concentrations were: 1) chosen from previously published reports on other species; 2) determined in a pilot study; 3) established in a limited number of animals; 4) measured but the data was not shown; or 5) inconclusive leading to the need to go back and understand hormone efficacy due to less than hoped for results [bib_ref] Hormonal induction of spawning in 4 species of frogs by coinjection with..., Trudeau [/bib_ref] [bib_ref] Motility and cryopreservation of spermatozoa of European common frog, Rana temporaria, Mansour [/bib_ref] [bib_ref] Hormonal induction of gamete release, and in-vitro fertilisation, in the critically endangered..., Byrne [/bib_ref]. To be used as a conservation tool for the growing number of endangered amphibian species that will need reproductive intervention to prevent extinction, it is valuable that hormone efficacy trials be continued so that future studies on other species can be modeled likewise. Studies beginning with very low concentrations of hormone and working up to higher levels is imperative to strike a balance between using enough hormone to obtain gametes for reproduction while not administering too high a dose that will cause health problems or even possibly death of the animal.
Another important finding from our study was that the hormones' effects on sperm concentration, motility and forward progression display a time-dependent trend within a treatment (more visibly obvious with hCG than LHRH). American toad sperm concentration peaks between 7-9 hrs post hCG administration at 12 x 10 6 sperm/ml and declines steadily to about 4.5 x 10 6 and 2 x 10 6 sperm/ml at 12 and 24 hrs, respectively. However, sperm concentrations for our LHRH treatment were low compared to hCG and typically were below 1 x 10 6 sperm/ml. This low sperm concentration following administration of 4.0 μg LHRH in our study is similar to levels reported by Obringer et al. [bib_ref] Characterization of the spermiation response, luteinizing hormone release and sperm quality in..., Obringer [/bib_ref] for Anaxyrus americanus administered 4.0 μg LHRH via subcutaneous injections. However, they report an average concentration of 4.9 x 10 6 sperm/ml when providing the same hormone dose via intra-peritoneal injections. The range in sperm concentration for their study was 0.1 to 24 x 10 6 sperm/ml, although 33% of their values were below 1 x 10 6 sperm/ml indicating that one third of their data readings were comparable to our values when using the same hormone concentration and route of administration. The average sperm concentration value reported by Obringer et al. [bib_ref] Characterization of the spermiation response, luteinizing hormone release and sperm quality in..., Obringer [/bib_ref] are nearly five times higher than what we observed, due to a couple animals with very high sperm concentration. The discrepancy in our results from theirs with LHRH may be due to seasonal timing issues. Obringer et al. [bib_ref] Characterization of the spermiation response, luteinizing hormone release and sperm quality in..., Obringer [/bib_ref] initiated their study on newly captured animals during the breeding season with all experiments conducted shortly thereafter (April-June). In contrast, our animals had been held in captivity for more than a year prior to the start of the study, were not hibernated, and all experiments were conducted outside of the breeding season. For endangered toads held longterm in captivity as part of a breeding program, our study scenario is likely more realistic of the challenges confronted by using non-hibernated individuals without the appropriate environmental cues to stimulate natural reproduction (hence the need for hormones in the first place).
The time an amphibian takes to respond to a hormone treatment has direct bearing on how amphibian captive breeding programs time their series of hormone injections. Information from our studies highlights not only the importance of testing various hormone concentrations, but also the importance of carrying out a timed experimental sampling protocol such that the peak behavioral or physiological responses being measured are known. For example, studies in our lab with Rana pipiens discovered that sperm production following hormone administration occurred within 30 minutes and had stopped by 2 hours (unpublished data). If the same sampling period had been followed in the Rana pipiens experiments as with this study on Anaxyrus americanus, one might incorrectly assume that the hormone was ineffective, when the reality would be that the optimal time for sperm collection had been missed. Studies like these allow for protocol development on gamete synchronicity or the "timed release" of both sperm and eggs to maximize artificial fertilization or assisted natural breeding. After a review of several Bufonid captive breeding programs it became apparent that both male and female amphibians were receiving hormone stimulation simultaneously during the morning. In the case of the Anaxyrus americanus, this would result in an asynchronous release of gametes. Typically, female Anaxyrus americanus release eggs 12-24 hours post-hormone administration, but optimal sperm production in the male occurs 7-9 hours post-hormone administration and would be declining rapidly after 12 hours. These results indicate that the effect of hormone on spermiation has a limited duration. Several U.S. zoos working with the authors report that often females lay eggs following hormone stimulation but that poor fertilization was noted. This poor fertilization may be a result of the males already being in the refractory period and no longer producing sperm. Thus, an understanding of optimal or peak gamete release in relation to hormone administration assists in the development of a timed AI program for amphibians and how best to stagger the hormone injections for optimal fertilization and inception of reproductive behaviors.
While we were fairly confident that hCG was the optimal hormone for collecting sperm from Anaxyrus americanus to use for artificial fertilization experiments we needed to test whether replacing LHRH with hCG in several captive breeding programs for other Bufonids would affect their induced mating strategies. Several pilot studies were conducted at Central Park Zoo, Fort Worth Zoo and Sybille Wildlife Recovery Center in Wyoming to see whether amplexus or the percent of fertilized eggs increased in Anaxyrus baxteri or Peltophryne lemur. Every institution reported seeing fewer animals in amplexus; thus, fewer fertilized eggs, when administering hCG instead of LHRH. Hence, we went back and created a series of experiments to understand the impact these two hormones had on reproductive behavior. We found that over 91% of the pairs given LHRH were in amplexus 9 hrs post-administration, while 75% of the animals given hCG were in amplexus at the same time period. These two protein hormones have very different binding receptors and modes of action, yet both have nongonadal targeted tissues with potential overlapping results on reproductive behaviors with varying degrees of responsiveness.
In the rat brain hippocampus, LHRH acts as a neurotransmitter linking actions inside the central nervous system where it facilitates reproductive behaviors to peripheral endocrine effects [bib_ref] Brain gonadotropin releasing hormone receptors: localization and regulation, Jennes [/bib_ref]. In addition, LH/ hCG receptors have also been found in the rat hippocampus, impacting specific reproductive behaviors [bib_ref] Neural actions of luteinizing hormone and human chorionic gonadotropin, Lei [/bib_ref]. One of the hippocampus' roles is chemical and hormonal sensing, whereby it exerts cognitive control over specific aspects of the hypothalamic-pituitary-adrenal axis impacting reproduction as well as memory cognition and arousal. In some amphibian species, it is likely that similar actions are occurring and that both LHRH and LH/hCG homologues modify neuronal excitability within the hippocampus, thereby modulating hippocampal function and downstream behaviors, although at varying levels of stimulation. Androgen receptors are also located within the hippocampusand steroid biosynthesis following exogenous hormone administration may be another indirect route for stimulating amphibian spermiation or onset of reproductive behaviors. Binding of hCG to receptors in the testis has been shown to initiate steroid production and spermiation in several amphibians including Rana nigromaculata [bib_ref] 17 alpha,20 alpha-Dihydroxy-4-pregnen-3-one is the naturally occurring spermiation-inducing hormone in the testis..., Kobayashi [/bib_ref] , Xenopus laevis [bib_ref] Hormone effects on male sex behavior in adult South African clawed frogs,..., Kelley [/bib_ref] and Bufo marinus. Similarly, administration of LHRH has been shown in several amphibian species to increase androgen production, resulting in increased reproductive behaviors and spermiation [bib_ref] Seasonal changes in luteinizing hormone-releasing hormone concentrations in microdissected brain regions of..., Zoeller [/bib_ref] [bib_ref] Historical perspective: Hormonal regulation of behaviors in amphibians, Moore [/bib_ref]. Hence, both LHRH and hCG stimulate similar steroidogenic pathways although the effectiveness of each exogenous hormone on reproductive behaviors or spermiation may be species-, dosage-, or hormonespecific.
Our results suggest that although hCG may be optimal for collecting gametes for artificial fertilization, the use of this hormone alone is not sufficient for inducing breeding and overcoming reproductive behavioral challenges. Previous studies by our lab have used both hormones in a cocktail mixture to induce Anaxyrus baxteri to ovulate large numbers of eggs when given as a priming hormone [bib_ref] Hormonal priming, induction of ovulation and in-vitro fertilization of the endangered Wyoming..., Browne [/bib_ref]. Future studies will test whether combining the two hormones in a cocktail formula may provide higher fertilization rates by increasing the sperm concentration while not sacrificing the number of animals that can be induced to amplex. Although we were able to achieve a 91% amplexus rate with LHRH up to 9 hrs in this series of experiments with Anaxyrus americanus, many programs using these same hormones, sometimes at higher concentrations than we tested, still report low numbers of animals in amplexus.
# Conclusions
In summary, research on appropriate hormone types and concentrations can be beneficial to captive breeding programs for stimulating optimal sperm and egg production in amphibians. In addition, the hormone of choice may depend upon whether the purpose is to collect gametes for artificial fertilization or to assist natural breeding and amplexus. Regardless of the hormone chosen, understanding when the peak and refractory period occurs for gamete production will likely dictate the timing of 'when' and 'if' successful reproduction takes place. These protocols should be developed early in a research program such that future studies on physiological or behavioral control mechanisms are not impacted by inappropriate hormone regimens. This is especially important for reintroduction programs of endangered species where producing brood stock is a key component for recovery. Another priority for amphibian conservation programs should be the development of techniques for shipping sperm between institutions for artificial fertilization, hence allowing for greater maintenance of genetic diversity and reducing stress from transporting animals. Appropriate hormone protocols for targeted species will be the key to the success of any particular conservation effort. Lastly, more research is needed to understand how often animals can be administered hormones to stimulate sperm production or ovulation and whether the exogenous hormones affect fertilization rates, tadpole development, metamorphosis and/or production of subsequent generations.
[fig] Figure 1: Percentage of male Anaxyrus americanus producing sperm over 24 hrs after hormone administration with either hCG (Panel A; n = 16/trt) or LHRH (Panel B; n = 20/trt). Male toads were treated with four different concentrations of hCG at 50 IU (circles); 100 IU (triangles); 200 IU (squares) or 300 IU (diamonds). Alternatively, males were treated with four different concentrations of LHRH at 0.1 μg (circles); 1.0 μg (triangles); 4 μg (squares) or 32 μg (diamonds). [/fig]
[fig] Figure 2: Sperm concentration (x 10 6 /ml) produced by male Anaxyrus americanus over 24 hrs after hormone administration with either hCG (Panel A) or LHRH (Panel B). Male toads were treated with 3 different concentrations of hCG at 100 IU (triangles); 200 IU (squares) or 300 IU (diamonds). Alternatively, males were treated with 4 different concentrations of LHRH at 0.1 μg (circles); 1.0 μg (triangles); 4 μg (squares) or 32 μg (diamonds). Only one male administered 50 IU hCG produced sperm and was not included in Panel A. Data are expressed as Mean ± SEM; lower SEM bars not shown. For number of animals comprising each treatment Mean see Figure 1. [/fig]
[fig] W 3 =: 36.6, P < 0.0005). Specifically, the number of male toads producing sperm differed between the lowest LHRH concentration (0.1 μg) and all the higher concentrations (P < 0.05) for the majority of time points where samples were collected. When compared to 300 IU hCG, all four concentrations of LHRH tested had low numbers of animals producing sperm at the first 3-hour time point; 0.1 μg (10%), 1.0 μg (30%), 4.0 μg (20%) and 32 μg (30%). While 300 IU of hCG induced 100% of the animals to produce sperm, the best response to LHRH had a maximum of only 35% of the animals producing sperm (Figure 1B). There was an effect of time on sperm production (W 5 = 470.42, P < 0.0005), and a treatment by time interaction (W 14 = 5982.61, P > 0.0005) indicating that the number of responders producing sperm changed over time. Sperm concentration of male Anaxyrus americanus was not significantly different between any of the four LHRH concentrations ( 4.72, P = 0.19; [/fig]
[fig] Figure 3: Percent sperm motility from male Anaxyrus americanus treated with either hCG (Panel A) or LHRH (Panel B) over time after administration with either: 100 (triangles), 200 (squares), or 300 (diamonds) IU hCG compared to 0.1 (circles), 1.0 (triangles), 4 (squares) or 32 (diamonds) μg LHRH. Forward Progression for these same sperm samples are shown in Panel C for hCG or Panel D for LHRH. Only one male American toad administered 50 IU hCG produced sperm for quality assessment and was not included in Panel A and C. Data are expressed as Mean ± SEM; lower SEM bars not shown. For number of animals comprising each treatment Mean see Figure 1. P < 0.0005). [/fig]
[fig] Figure 4: A comparison of the sperm motility index (SMI) for Anaxyrus americanus treated with either 300 IU hCG (diamonds) or 4 μg LHRH (squares). Data are expressed as Mean ± SEM. For number of animals comprising each treatment Mean seeFigure 1. [/fig]
[fig] Figure 5: The percentage of male Anaxyrus americanus amplexing females over time following hormone administration. Males were treated with the optimal hormone concentrations from Study 1 and Study 2, specifically, 300 IU hCG (n = 12/trt) or 4.0 μg LHRH (n = 11/trt). Females were not treated with any hormones. [/fig]
|
10.3389/fimmu.2021.765965
|
CCBY
|
8551858
|
34721437
|
s2orc_pubmed_articles
|
Role of Gut Microbiome in COVID-19: An Insight Into Pathogenesis and Therapeutic Potential
# Introduction
Since it was first recognized, Coronavirus Disease 2019 , caused by the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), remains a global affliction. Although vaccines offer hope in curbing the pandemic (2), an improved understanding of its pathogenesis and concurrent efforts to explore preventive and therapeutic strategies remain a priority to consolidate the success of mass vaccination and herd immunity.
The clinical spectrum of COVID-19 ranges from asymptomatic to severe, life-threatening disease. Current understanding of pathogenesis postulates a rapid and intense hyperactivation of the immune system, resulting in critical illness and mortality. Older age, burden of comorbidities, obesity, immunocompromised states, malignancy or ongoing cancer treatment, and being a transplant recipient, have been strongly linked with severe, and sometimes fatal, outcomes. Evolving data suggest that a state of chronic inflammation or baseline activation of the immune system might influence the course of COVID-19 more than direct cytopathic effects of the SARS-CoV-2. Furthermore, a subgroup of patients have been noted to develop auto-inflammatory symptoms (such as Kawasaki-like disease in children and multi-system inflammatory syndrome) long after clearance of the SARS-CoV-2 virus from body, suggesting an immune dysregulation.
In the human body, the gastrointestinal tract (GIT) is the largest immune organ. The pool of resident microorganisms (bacteria, viruses, and fungi) in the GIT, collectively known as the gut microbiota, not only supports mucosal immunity but also modulates the systemic immune response of the host. Current evidence from other respiratory illnesses indicates that the gut microbiota affects the immunity and inflammation in the lungs. Lately, some studies have examined the association between gut microbiota and SARS-CoV-2. In this review, we present the existing data related to the intersection of gut microbiome and the host's immune response to SARS-CoV-2. We further explore the role of gut microbiome diversity and its compositional differences as diagnostic biomarkers, and the potential of the gut microbiome as an interventional target in modifying COVID-19 outcomes.
## Significance of gut microbiota
The human GIT is home to about 10 4 -10 5 bacteria per millimeter of content in the small intestine, and 10 11 bacteria per gram of colonic content. In a healthy person, the gut microbiota comprises more than 100 bacterial phyla and the majority of bacteria belong to Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria, with Firmicutes and Bacteroidetes phyla constituting over 90% of the entire gut microbiota. The microorganisms and their combined genetic material make up the gut microbiome, which outnumber the human genome by about 150 times. The proportion of the various phyla remains quasi-stable and unique for an individual, although a shift can be observed during a change in health status. For example, the gut microbiome in the elderly has been observed to drift away from Firmicutes and towards Proteobacteria and Alistipes.
The gut microbiota exists in a symbiotic relationship with its host. It facilitates the synthesis of vitamins and fermentation of carbohydrates and other undigested nutrients and aids in the delivery of essential nutrients like short-chain fatty acids (SCFAs) to colonic epithelial cells. In addition, it also regulates mucosal permeability and provides deterrence against pathogenic microbes. More importantly, the microbiota plays an indispensable role in the preservation of intestinal homeostasis by modulating local and systemic immune responses of the host. The microbiota protects the GIT by (a) acting as a competitor against binding of pathogenic microbes, (b) neutralizing pathogens with their anti-microbial metabolites, (c) keeping the local immune system in a perpetual vigilant state, and (d) regulating the innate and adaptive immunity.
In a healthy person, the proportion of the various phyla mostly remains quasi-stable and unique for an individual. An imbalanced state is described as "gut dysbiosis", a condition characterized by an alteration in the abundance or composition of the microbiota. Gut dysbiosis may occur with aging, dietary effects, drugs, gastrointestinal infections, and anatomical alterations of the GIT. A significantly dysbiotic state may predispose to the diseases of GIT, such as Clostridioides difficile enterocolitis, which is associated with prolonged and recurrent broad-spectrum antibiotic usage. Since gut microbiota modulates the fine balance between pro-and antiinflammatory systemic responses, a dysbiotic state has also been associated with non-gastrointestinal systemic illnesses such as malignancy, type 2 diabetes mellitus, nonalcoholic steatohepatitis, coronary artery disease (23), neurodegenerative diseases, and depression.
## The gut-lung axis
The GI and respiratory tracts share a common mucosal immune system, known as the gut-lung axis. From birth, both tracts receive their quota of microbiota via the oral route, and subsequently establish a differing but internally quasi-stable genre of microorganisms or microbiota. Although the microbiota of both tracts consists of similar phyla, they differ at the level of species in composition and density. Understandably, studies on respiratory microbiota have been complicated by tedious and invasive methods for collection of uncontaminated lower respiratory samples, and most data have been derived from mice models where lung tissue can be aseptically obtained. Consequently, there is growing excitement in understanding this complex immunological intersection.
Throughout the lifespan of an individual, established microbiota of both tracts contribute to the gut-lung axis, modulating both local and systemic immune responses when faced with a pathogenic threat. The axis is believed to be bidirectional, affecting the immune response of either tract when one site is activated. Using a germ-free murine model, Ichinohe et al. demonstrated potentially deleterious effects on respiratory immune responses after alteration of the gut microbiota with antibiotics. Other studies have also found that gut microbiota alterations result in abnormal activation of the immune system, predisposing to respiratory illnesses such as asthma, lung allergic responses, and chronic respiratory diseases. Conversely, animal studies have also revealed an alteration in the gut microbiota after respiratory viral and bacterial infections. This distant effect is believed to be communicated by activation of the systemic immune system, with dysbiosis of either tract feeding into the other.
understanding of the impact of the gut microbiota in patients with COVID-19 may be extrapolated by examining the existing evidence of its role in non-SARS-CoV-2 respiratory virus infections.
## Evidence from sars-cov-1 infection
Many respiratory viral illnesses are commonly accompanied by GI symptoms. Previous studies during the severe acute respiratory syndrome (SARS) outbreak in 2002 showed that diarrhea was a common symptom and occurred in 16%-73% of patients. The Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV-1) was not only known to infect the lung epithelial cells but also the immune cells, triggering an intense immune response with elevation in Th2 cytokines. It was postulated that high levels of circulating pro-inflammatory cytokines altered the gut microbiota and compromised intestinal integrity. The resultant "leaky" gut allowed translocation of bacterial products and antigens into the circulation, further exacerbating the illness. Due to the bidirectional nature of the gut-lung axis, an alteration in gut microbiota further augments the respiratory immune responses, conceivably resulting in a vicious perpetuation of systemic inflammatory response.
## Evidence from other community respiratory viruses
Using a mouse-model, Deriu et al. demonstrated that respiratory viral infection due to influenza resulted in gut dysbiosis predisposing to secondary Salmonella infection via circulatory type I interferons. Similarly, Wang et al. demonstrated indirect intestinal inflammation with influenza infection in a mouse-model occurring via microbiota-mediated Th17 celldependent inflammation. Several studies have reported gut dysbiosis after respiratory viral infection. showed that gut dysbiosis, in the form of an increase in Bacteroidetes and a decrease in Firmicutes phyla abundance, occurred in mice models with respiratory syncytial and influenza virus infections, but not in those vaccinated with live attenuated influenza viruses. Furthermore, elevated levels of colonic Muc5ac and fecal lipocalin-2 in the pathogenic infection group suggest the presence of low-grade gut inflammation during respiratory virus infection.
Respiratory virus infection may also cause dysbiosis in lung microbiota, modulating both local immune responses within the lung parenchyma and systemically. Due to the difficulties in sampling of lung microbiota, only a few studies have examined the role of respiratory pathogens in altering lung microbiota. Molyneaux et al. reported an increased proportion of Proteobacteria and potentially pathogenic Haemophilus influenzae in the lower respiratory tract microbiota in rhinovirus-infected patients with chronic obstructive pulmonary disease. Using a mouse-model inoculated with intranasal H1N1 influenza virus, Gu et al. demonstrated a bacterial class shift in the lung microbiota, which persisted even during the recovery period. Notwithstanding the limitations in available studies, a common theme has emerged showing a link between respiratory virus infection and an alteration in the gut and respiratory tract microbiota, with the presence of inflammation of the GIT.
## Role of gut microbiota in the pathogenesis of covid-19
Although no specific interaction between any gut microbial species and SARS-CoV-2 has been identified to date, there is indirect evidencethat gut microbiota may have a role in the overall pathogenesis of COVID-19, as summarized in . Taking the corollary further from non-SARS-Cov-2 virus-mediated gut dysbiosis, it is conceivable that infection with SARS-CoV-2 may also be impacted by immunological interactions with the gut microbiota.
## Pathogenesis of gi symptoms in covid-19
About half of patients with COVID-19 develop GI symptoms, which often precede the respiratory symptoms. In the respiratory tract, the SARS-CoV-2 infects the alveolar cells by binding to angiotensin-converting enzyme 2 (ACE-2) receptors. Interestingly, these receptors are also abundantly expressed on the surface of enterocytes, where they play an important role in maintaining the homeostasis of microbiota and mucosal inflammation. In one patient with COVID-19, Xiao et al. detected SARS-CoV-2 RNA, viral nucleocapsid protein, and ACE-2 in the epithelial cells of the esophagus, stomach, duodenum, and rectum. Several authors have also reported detection of SARS-CoV-2 RNA fragments, but not the whole virus, in stool samples. A study by Lamers et al. also demonstrated that SARS-CoV-2 was able to infect enterocyte lineage cells in a human intestinal organoid model. Unfortunately, due to the scarcity of autopsy studies and pragmatic restrictions on endoscopic examinations of the GIT, there are only limited data to support this hypothesis in vivo. It also remains unknown how SARS-CoV-2 may survive the acidic gastric environment to directly infect enterocytes. Overall, it is unclear whether the GI symptoms of COVID-19 are a result of primary infection of the GIT or derived from other indirect mechanisms mentioned below.
Fecal Shedding of SARS-CoV-2 SARS-CoV-2 is principally transmitted by respiratory droplets, but evidence is accumulating for fecal-oral transmission. The hypothesis is supported by the presence of GI symptoms, detection of SARS-CoV-2 nuclear fingerprints in the GI mucosa, and detection of viral fragments in fecal samples. Interestingly, the first COVID-19 patient in the United States tested positive for SARS-CoV-2 via stool samples, and subsequent studies have consistently documented shedding of viral RNA in the stool samples in COVID-19 patients. Furthermore, viral shedding in stool samples has been observed to persist longer than that in respiratory samples.
## Hypercytokinemia and the leaky gut
Elevated serum levels of pro-inflammatory markers, such as interleukin-6 (IL-6) and interleukin-10 (IL-10), are hallmarks of severe COVID-19 infection. These cytokines predispose to dysbiosis, which consequently alters intestinal permeability, a state known as the "leaky gut". This enables further entry of a multitude of bacterial products and toxins, activating a proinflammatory cascade. In a study in 204 patients with COVID-19, Pan et al. reported that GI symptoms worsened with increasing severity of COVID-19. In another study, fecal calprotectin levels (a marker of GI mucosal inflammation) were elevated in patients who had diarrhea during COVID-19 illness. A recent study by Prasad et al. measured several markers of gut permeability in the plasma. Levels of FABP2, PGN, and LPS were significantly higher among COVID-19 patients compared to healthy subjects, suggesting translocation of proinflammatory antigens from a leaky gut.
## Association between gut dysbiosis and systemic inflammation
Gu et al. first presented evidence of an altered gut microbiota in COVID-19 patients by using high-throughput sequencing of 16S ribosomal RNA to compare the gut taxa of patients with COVID-19, H1N1 influenza, and healthy controls. Compared to healthy controls, COVID-19 patients had significantly reduced bacterial diversity, increased abundance of opportunistic pathogens (such as Streptococcus, Rothia, Veillonella, and Actinomyces), and significantly less diverse symbiotic species. Interestingly, the altered microbial signature in COVID-19 was different from patients with the H1N1 strain.
In another study, Zuo et al. analyzed fecal samples of 15 patients with COVID-19 using shotgun metagenomic sequencing. The study revealed a marked increase in opportunistic pathogens and depletion of beneficial microbes as compared to healthy controls, which persisted even after clearance of SARS-CoV-2. These findings suggest an inverse correlation between gut dysbiosis and COVID-19 severity and are congruent with the conclusions of a larger multi-center study in which Yeoh et al. examined the gut microbiota in 100 patients using shotgun sequencing. Fecal samples from 27 patients were analyzed longitudinally over 30 days, showing significant dysbiosis that persisted despite clearance of SARS-CoV-2. Correlative blood samples demonstrated an association between gut dysbiosis, elevation in the inflammatory mediators, and severity of systemic inflammation.
Another study by Newsome et al. compared microbiota composition from stool samples of 50 COVID-19 patients with uninfected patients. Significant perturbations in the microbiota composition in the COVID-19 patients were observed, independent of antibiotic exposure. The gut FIGURE 1 | SARS-CoV-2 and the lung-gut axis: SARS-CoV-2 virus enters the alveolar cells by binding with ACE2 receptors, which are also abundant on the surface of enterocytes. The implication of direct infection of enterocytes by SARS-CoV-2 is still being explored. The circulatory cytokines from alveolitis (and/or direct viral infection of the enterocytes) cause the GI dysbiosis with resultant alterations in GI mucosal barrier. The entry of bacterial products and toxins from the GIT floods the circulatory system with more pro-inflammatory cytokines. Image was created with Biorender.com.
"metabolome", a biochemical signature derived from bacterial metabolic activity in the gut, is another method for detecting an alteration in the gut microbiota composition. In a recent study by Lv et al., fecal samples of COVID-19 patients had altered metabolomes, suggesting malnutrition and intestinal inflammation. These results provide new insights into the pathogenesis of COVID-19.
## Gut dysbiosis as a biomarker of viral replication
Although it was previously believed that gut dysbiosis in COVID-19 was mainly driven by inflammatory mediators from respiratory tract infection, a recent study suggests that active replication of SARS-CoV-2 in the gut may be driving the dysbiosis. Using in vitro transcriptional analysis in a SARS-CoV-2-infected cell model (with samples obtained from stools), the 3' end of the SARS-CoV-2 genome was detected more than the 5' end, suggesting active viral replication. Interestingly, majority of the patients had no GI symptoms, suggesting a quiescent GI infection despite active replication of SARS-CoV-2 in the GIT with dysbiosis. Moreover, on functional analysis of the gut microbiota, fecal samples with signatures of high SARS-CoV-2 burden demonstrated high de novo nucleotide and amino acid biosynthesis, correlating with increased bacterial proliferation. Although this was a pilot study comprising only 15 patients, further studies on alterations in the functionality of the gut microbiota may unearth the pathophysiology of COVID-19 illness.
Overall, currently expanding evidence suggests that patients with COVID-19 suffer from an alteration in the gut microbiota during and after the illness. Both systemic inflammation and replicative potential of SARS-CoV-2 in the gut may contribute towards dysbiosis.
## Clinical implication of gut microbiota in covid-19
Though limited in number, studies to date have consistently demonstrated gut dysbiosis in patients with COVID-19. A more important clinical implication lies in understanding whether, and how, gut microbiota predisposes to varying degrees of COVID-19 severity.
## Potential role of gut microbiota in asymptomatic/subclinical and mild covid-19
As previously mentioned, clinical spectrum of COVID-19 ranges from asymptomatic to severe, life-threatening disease. A recent systematic review demonstrated that about one-third of patients remain clinically asymptomatic after infection with SARS-CoV-2 (68). However, a possibility of subclinical inflammatory process remains. In a systematic review involving 231 asymptomatic COVID-19 patients, almost two-thirds (63%) had inflammatory changes in the lungs on computed tomography (CT) scan. Irrespective of subtle inflammatory changes, a subset of patients may not mount the intense inflammatory response that portends severe illness. As this heterogeneity in clinical severity is less likely due to the existence of less virulent strains of SARS-CoV-2, or the protection from adaptive immunity given the novel nature of the virus, the immune response of the host remains the most probable factor in determining disease severity. It is unclear if any specific pattern of gut microbiota protects individuals from mounting a severe inflammatory state when infected with SARS-CoV-2. Kumar et al. highlighted a potential link between the environmental microbiota of a population and the burden of COVID-19. With data from 122 diverse countries, lower COVID-19-associated mortality was observed in countries with a higher percentage of rural population (alluding to higher gut microbial diversity), higher proportion of population residing in slums, and a lower water quality and sanitation score. While such observational data can be prone to confounders, these results offer some insight into the potential role of gut microbiota on the disease burden of COVID-19.
## Gut microbiota in severe covid-19
There is mounting evidence that being elderly and having a chronic inflammatory state (from chronic medical conditions) predisposes to a pro-dysbiotic state. It is unlikely a coincidence that the highest rates of morbidity and mortality from COVID-19 have also been observed in the elderly, those with underlying chronic medical conditions, and among immunosuppressed patients with cancers (5-9). COVID-19 disease severity is likely host dependent and driven by the inflammatory response. In autopsy samples from a patient with severe COVID-19, inflammatory cells were observed in the lungs, suggesting an intense inflammatory response. Furthermore, studies have also reported elevated plasma levels of pro-inflammatory cytokines such as interleukin-1 (IL-1), IL-6, and tumor necrosis factor alpha, in severe COVID-19. In patients prone to gut dysbiosis, further inflammatory triggers may tip the balance over to a leaky gut, resulting in a self-perpetuating inflammatory feedback circle. Notably, two small studies showed a direct correlation between severe COVID-19 and gut dysbiosisAnother two studies have shown that patients with severe COVID-19 experienced more pronounced GI symptoms, along with higher levels of stool calprotectin (an indicator of GI inflammation and disrupted mucosal integrity). This supports the concept of an immunological crosstalk between the lungs and gut, presumably moderated by the gut microbiota. There is still a lack of data assessing the role of gut microbiota in a high-risk cohort (such as elderly or cancer patients) with COVID-19, although a study on the impact of probiotics on health and immunity in elderly and diabetic patients, and response to COVID-19 vaccination, is underway.
## Therapeutic potential of gut microbiome for covid-19
Given the association between gut dysbiosis and COVID-19 severity, the therapeutic potential for modulation of the gut microbiome to modify disease outcomes holds promise. However, there is no microbiota-directed therapy that has demonstrated efficacy in preventing the development or progression of COVID-19 currently. A statistically significant negative correlation was observed between COVID-19 mortality and populations that had a high percentage of rural residents, and a high proportion of diarrhea secondary to inadequate sanitation. As a result, a high microbial exposure to gramnegative bacteria was proposed to confer protective effects against COVID 19, possibly due to increased interferon type I levels.
Cross-sectional data based on national population health indicators. Inferential hypothesis based on effect of environmental microbiological prevalence rather than direct sequencing of human GI microbiota samples. No analysis of interferon type-I levels in study populations. Microbial DNA extraction and 16S rRNA sequencing in the plasma samples. Levels of gut permeability markers were also measured.
In the plasma samples of about 65% patients with COVID-19, abnormal signatures of gut microbes were seen. As compared with the healthy controls, patients with COVID-19 had significantly elevated plasma levels of gut permeability markers (such as FABP2, PGN, and LPS).
## Potential role of prebiotics
Plant-based fibers exert a prebiotic effect by promoting the growth of beneficial microorganisms in the gut microbiota (e.g., Bifidobacterium and Lactobacillus spp.) while decreasing the proportion of harmful species (e.g., Clostridia). Moreover, the fermentation of soluble dietary fibers by certain bacterial species yields several beneficial metabolites, such as SCFAs, which serve to maintain colonic mucosal integrity and modulate the immune system. By-products of SCFAs are also absorbed into the circulatory system and have anti-inflammatory effects. In mice models, a high-fiber diet with elevated SCFA levels was protective against allergic inflammation in the lungs, while a low-fiber diet with decreased SCFA levels resulted in increased allergic airway disease. Interestingly, studies from other respiratory diseases have demonstrated a reduction in mortality with intake of whole grains. Although the beneficial effects of dietary fibers are intuitive, there is currently no direct evidence that any specific amount or type of dietary fiber is beneficial in COVID-19 illness.
## Potential role of probiotics
Oral probiotics are live bacteria of specific species that alter the composition of gut microbiota after reaching the intestines. A shift to beneficial bacterial species modulates the local and systemic inflammatory balance, with several studies demonstrating a positive impact on respiratory infections and other extra-intestinal illnesses. Using a probiotic bacterium, Lactobacillus gasseri SBT2055 in mouse models, prevention of infection with respiratory syncytial virus was demonstrated. In another study on 30 elderly volunteers, Bifidobacterium lactis HN019 ingestion was shown to enhance the cellular immunity. Placebo controlled clinical trials with probiotics (using Lactobacillus rhamnosus GG, Bacillus subtilis, and Enterococcus faecalis) have also demonstrated significant improvement in patients with ventilator-associated pneumonia.
Naturally, if dysbiosis is indeed involved in the pathogenesis of severe COVID-19, probiotics appear to be among the more convenient, efficient, and potentially safe strategies. After initial reports of gut dysbiosis in patients with severe COVID-19, the National Health Commission (of China) recommended the use of probiotics to maintain gut microbial homeostasis and prevent secondary bacterial infections. Given the dearth of data pertaining to SARS-CoV-2 and the relative safety of probiotics, the rapid promulgation in favor of probiotics seems reasonable while awaiting further evidence. Although there is no direct evidence yet showing the efficacy of any specific strain of probiotic against COVID-19, several registered trials are currently examining the therapeutic potential of various probiotics formulations in COVID-19.
## Potential role of fecal microbiota transplantation
FMT is a process of actively transferring colonies of fecal bacteria from a healthy person into the GIT of another individual. The process aims to restore the composition of gut microbiota back to a healthy state. As mentioned earlier, FMT is an effective therapy for recurrent or refractory C. difficile enterocolitis. Given the proposed role of gut microbiota in the abnormal activation of immune responses in the COVID-19, FMT can potentially be explored as a therapeutic strategy. A recent case reportdescribed two patients with rapid resolution of COVID-19 after FMT was undertaken to treat concomitant C. difficile infection. However strong may be the hypothesis and surrounding speculations, FMT should not presently be recommended as a therapy against COVID-19 due to the scarcity of large-scale studies. To explore further, a clinical trial (FeMToCOVID) is currently registered at the clinicaltrials.org (NCT04824222), though it has not yet started recruiting patients.
## Unanswered questions and future directions
Despite the efficient pace of clinical trials evaluating new and repurposed agents for COVID-19, success has been modest at best. Although still in nascent stages, evolving evidence signals a probable link between gut microbiota and the host's immune response to COVID-19. However, the exact mechanism and extent of the role of gut dysbiosis in disease severity remain to be elucidated. This is further compounded by the inherent challenges associated with designing microbiome studies. Another possible angle would be to define the state of a leaky Stool samples were analyzed for various microbial biochemical products (or metabolome) using gas chromatographymass spectrometry.
Differences in the metabolomes of COVID-19 patients were observed compared with healthy controls.
Small sample size. Absence of demonstration of gut dysbiosis in the stools. No control for diet and anti-microbial agents. Does gut microbiota play a role in the onset of infection with SARS-CoV-2?
To characterize the diversity of gut microbiome across separate cohorts of individuals with varying risk of exposure to SARS-CoV-2.
## Onset of symptoms after infection with sars-cov-2
Does gut microbiota influence development of symptoms in COVID-19?
To study differences of gut microbiome between asymptomatic and symptomatic individuals.
## Development of severe covid-19 illness
Does pre-existing gut microbiota predispose to different levels of severity?
To examine the baseline gut microbiota and correlate with severity of COVID-19 illness. Alteration in gut microbiota by the SARS-CoV-2
Does gut microbiota get altered by the SARS-CoV-2?
To evaluate the temporal trend of gut microbiome in the COVID-19 illness.
## Persistence of gut dysbiosis after sars-cov-2
Is SARS-CoV-2-induced gut dysbiosis temporary?
To examine the long-term trend in the gut dysbiosis and its associated implications.
## Alteration in intestinal permeability with sars-cov-2
Does SARS-CoV-2-led inflammation lead to alteration in intestinal permeability.
Detection and measurement of gut-derived bacteria and/or their bacterial products in the circulation or extra-intestinal tissues (such as mesenteric lymph nodes). Therapeutic and preventive roles of prebiotics and probiotics Do prebiotics and/or probiotics have potential to alter course of COVID-19 illness?
Assessing variety of probiotics in terms of prevention and optimization of COVID-19 illness.
## Therapeutic role of fmt in covid-19
Can a reset of gut dysbiosis to normal homeostasis with the FMT mitigate severe COVID-19 illness?
To assess the potential therapeutic role of the FMT in severe COVID-19 illness. Observational study looking at the effect of oropharyngeal and gut microbiota, host genotype, and immune characteristics and SARS-CoV-2 viral genome sequences on outcomes of COVID-19 infection.
## Nct04598334 cytokine storm among bangladeshi patients with covid-19
Prospective study evaluating the relationship of inflammatory markers and cytokine levels in addition to gut microbiota on COVID-19 infection severity in Bangladeshi patients, at various points of illness progression.
## Nct04597736 relationship between biological profiles and clinical evolutions
Within the Same Cluster COVID-19 (COVIDCOLLECT)
Cohort study examining the relationship between the biological profiles observed from analysis of nasopharyngeal, saliva, blood, stool, and urine samples and the clinical evolutions within the same cluster of COVID-19 cases and their contact subjects. NCT04581135 Study to Investigate Long-term Pulmonary and Extrapulmonary Effects of COVID-19
Prospective study investigating long-term pulmonary and extrapulmonary effects of COVID-19, including changes to gut microbiota. Observational study looking at differences in the biomarkers of different sexes during SARS-CoV-2 infection, including the gut microbiome. Randomized controlled trial evaluating the effects of Lactobacillus coryniformis K8 consumption on the incidence and severity of COVID-19 in health workers exposed to the SARS-CoV-2 virus.
(Continued) gut more clearly with biomarkers and cutoff criteria, to also enable clarification on whether patients who have developed an inflammatory cascade are still amenable to therapeutic gut microbiota modulation. Besides exploration of any potential role in the management of COVID-19, long-term consequences of gut dysbiosis should also be explored with longitudinal follow-up. A search of ongoing clinical trials at the US National Library of Medicine reveals 24 registered studies assessing microbiota-targeted therapeutic options.
# Conclusion
The role of resident gut microbiota in other respiratory illnesses has been well recognized. Furthermore, the brunt of unfavorable COVID-19 outcomes has been on elderly patients and those with chronic medical diseases, both scenarios known to have senescence-driven gut dysbiosis. Increasingly, evidence is mounting for gut dysbiosis as a predisposing factor for severe COVID-19, through a leaky gut phenomenon and resultant spillage of bacterial products and toxins. Evidence is emerging that the degree of dysbiosis correlates with the severity of COVID-19 illness. This behooves us to explore potentially preventive and therapeutic targets, such as dietary intervention and probiotics. Several ongoing trials are evaluating various pathogenic routes and therapeutic approaches. While efforts for direct anti-viral agents and vaccines are of prime significance, the gut-lung axis could still hold therapeutic potential. Randomized controlled trial studying a probiotic formula in enhancing immunity and reducing hospitalization in elderly and diabetic patients. B. Ongoing studies utilizing microbiota as therapeutic target
|
10.1007/s10689-020-00201-5
|
CCBY
|
8064993
|
32770331
|
s2orc_pubmed_articles
|
Is a colorectal neoplasm diagnosis a trigger to change dietary and other lifestyle habits for persons with Lynch syndrome? A prospective cohort study
## Supplemental
[formula] CRN ¥ 0.3 ± 1.8 -0.2 (-0.8, 0.4) -0.1 (-0.6, 0.4) Physical activity level ¶ , mean ± SD No CRN ¥ 0.3 ± 1.2 Reference Reference CRN ¥ 0.4 ± 1.3 0.1 (-0.3, 0.5) -0.1 (-0.4, [/formula]
[fig] Supplemental table S3: Sensitivity analysis excluding those with a colorectal neoplasm diagnosis before baseline. Body mass index (BMI) at baseline and at followthose without missing values in BMI. [/fig]
[table] table S2: Sensitivity analysis excluding those with a colorectal neoplasm diagnosis before baseline. Changes in lifestyle characteristics and multivariable linear regression models for differences in change in lifestyle and dietary factors among before baseline in those with and without a CRN diagnosis during follow-up. α [/table]
|
10.1002/ieam.1955
|
CCBY
| null |
28600872
|
s2orc_pubmed_articles
|
Anatomy of a decision II: Potential effects of changes to Tier I chemical approaches in Canadian Disposal at Sea program sediment assessment protocols
The effects of possible changes to the Canadian 2-tiered assessment framework for dredged material based on outcomes of the 2006 Contaminated Dredged Material Management Decisions Workshop (CDMMD) are evaluated. Expanding on the "data mining" approach described in a previous paper, which focused solely on chemical lines of evidence, the efficacy of Tier 1 approaches (increases to the number of chemical analytes, use of mean hazard quotients, and the use of a screening bioassay) in predicting toxicity are evaluated. Results suggest value in additional work to evaluate the following areas: 1) further expanding minimum chemical requirements, 2) using more advanced approaches for chemical interpretation, and 3) using a screening-level bioassay (e.g., Canadian solid-phase photoluminescent bacteria test) to determine whether it would complement Tier 1 chemistry as well as or better than the solvent-based Microtox TM test method evaluated in the present study. Integr Environ
# Introduction
Canada's Disposal at Sea (DaS) Program hosted a Contaminated Dredged Material Management Decisions Workshop (CDMMD) in 2006 [bib_ref] Towards the assessment and management of contaminated dredged materials, Agius [/bib_ref]. The resulting recommendations from the 50 sediment assessment and management experts in attendance addressed the development of sediment assessment tools, the interpretation of these tools, and the essential attributes of a comparative risk assessment process for dredged material (DM) management [bib_ref] Towards the assessment and management of contaminated dredged materials, Agius [/bib_ref]. Canada has since been working to evaluate the feasibility and effect on decision making of each workshop recommendation, with a particular focus on exploring and validating a novel DM assessment framework that maintains consistency, efficacy, and transparency in decision making, and that protects the environment without posing an undue burden on project proponents.pointed out that many options for refining a DM assessment framework (e.g., changes to chemical action lists or specific action levels, or rules for interpreting Tier 2 biology) are interdependent and that the optimal approaches for a regulator would depend on a range of policy choices, informed by available science. [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] report on the initial findings of the "data mining" approach being used by Canada to evaluate CDMMD workshop recommendations relating to its current 2-tiered assessment framework for dredged material (comprised of Tier 1 chemical testing for "no-effects," and Tier 2 biological testing to evaluate bioavailability and toxicity), and provide details on the development of a database built for this purpose and its initial use. After compiling co-occurring sediment chemistry data sets from external sources (primarily the US National Oceanic and Atmospheric Association , [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] evaluated how various chemical assessment approaches performed relative to one another in terms of potential chemical regulatory outcomes (i.e., chemical pass/fail). [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] identified potential changes to Tier 1 chemical protocols and reported findings that include the benefits associated with expanding the list of sentinel metals beyond Hg and Cd, the significance of the list of analytes (vs the specific sediment quality guidelines used) in chemistry-based outcomes, and the potential for chemical upper action levels to avoid unnecessary toxicity testing of sediments with very high contaminant levels. The analysis demonstrated that the inclusion of other metals in a chemical action list would improve the overall detection of metal-contaminated sediments.
This article includes online-only Supplemental Data.
Although [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] considered potential outcomes using chemical data, they did not consider outcomes in the context of a tiered decision framework that considers both sediment chemistry and sediment toxicity. To assist with these endeavors, the work described here further explores CDMMD workshop recommendations by combining the database on sediment chemistry with data on colocated toxicity measures to evaluate how effectively various Lower Action Level Tier 1 approaches predict sediment toxicity, and to evaluate potential regulatory outcomes for a range of tiered DM assessment protocols that include both chemical and toxicological evaluation. Specifically, the present paper evaluates the performance of bioassay results similar to Canada's current Tier 2 battery of bioassays against current chemical screening results, and the outcomes and implications of various potential changes to the Canadian chemical and toxicological assessment protocols for DaS, including the assessment of a broader suite of metal and organic contaminants. Outcomes following the application of a range of Tier 1 Lower Action Level (AL1) decision rules, including the use of mean hazard quotients (mHQs) and the addition of a screening bioassay, are also evaluated. The potential efficacy of chemical Upper Action Levels (AL2s) in predicting toxicity failures will be addressed in a subsequent paper.
Glossary of terms (adapted from Action level (AL). A decision rule or set of decision rules that integrates findings related to multiple lines of evidence (characteristics), after comparison to their respective benchmarks, to yield a single, "overall" decision.
Chemical action level. Uses only chemical benchmarks or sediment quality guidelines (SQGs).
Lower action level (AL1). Action levels that identify levels below which there is "negligible environmental concern" in relation to disposal decisions.
Upper action level (AL2). Action levels used to avoid acute and chronic effects in relation to decisions about disposal at sea.
Action list. Comprised of a number of characteristics to be considered for measurement in the dredged material.
Benchmark. A point on the range of the metric (e.g., 4 mg/kg Cu, 20% amphipod mortality) that is used to identify where environmental concern may be low or high for that characteristic. These can be referred to as the "lower benchmark" and "upper benchmark."
Characteristic. Ann attribute of dredged material (e.g., Cu, Hg, silt, petroleum compounds, pathogens) or a biological response to dredged material (e.g., mortality, growth, bioaccumulation).
## Metric.
A measurement that can be made on a characteristic (e.g., concentration, percent survival). [fig_ref] Figure 1: Figure 1 [/fig_ref] depicts the steps followed to collect and analyze data in the present study.
# Methods
## Database development
The database was developed as described in [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref]. As reported there, biotest results were collected for subsequent analysis but had not at that point been interpreted or validated. A compact data set was generated with all the sediment records that had the minimum chemical data set designated in [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] and also contained a full "matched" bioassay battery with results for amphipod (Ampelisca) survival, sea urchin fertilization, and Microtox TM . These particular bioassays were selected because they were the closest available matches to the current Canadian protocols. Data for bioassays in the database were assigned pass or fail values on the basis of the criteria described in [fig_ref] Table 1: Canadian DaS bioassays and the "matched" bioassays used from the database a... [/fig_ref].
The resultant compact data set contained a broad range of sediment physical, chemical, and biological data. Data sets were reviewed to ensure that all results for a given parameter were in the same units, and anomalous data (such as nonnumerical results or impossible values such as negative concentrations) were eliminated unless they could be corrected in correspondence with relevant database coordinators. Because the objective was to generate a plausible and realistic proxy for data that might potentially be encountered, but not to draw specific risk conclusions from specific data or sites, no further data validation or quality control were carried out. The number of records with both biological and chemical data was about half of the number that were available for the chemical analysis in [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] but was still substantial; there were 1081 records with the complete data, with a reasonable distribution around the US coastline.
## Chemical parameters of interest
Apitz and Agius (2013) described the methods and assumptions used to select chemical parameters of interest and the range of chemical action levels tested for each parameter. Briefly, the focus was on comparing sediment data to a set of chemical ALs that might be used as AL1 values in a decision framework (although these were called "Lower Action Levels" in [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref]. The set of chemical parameters used included metals for which other regulators have established benchmarks, total PAH (defined as the sum of the US Environmental Protection Agency priority PAHs for which data were available), total PCB (defined as the sum of congeners , and other organics for which both data and benchmarks were available. A broader list of other chemical analytes was included in the database for potential future use.
## Lower action level (al1) values used
To test how various changes to the DaS chemical assessment protocols would affect potential regulatory outcomes for the sediments in the database, contaminant levels were compared to hypothetical chemical AL1 for each constituent under consideration. A list of "consensus" ALs was developed by [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] to provide a consistent basis of comparison throughout the assessment scenarios; details of their development can be found there. The consensus values were based upon the geometric mean of chemical ALs used internationally in dredging programs, or, if not available, in other sediment assessment frameworks. The ALs tested in the present study, as well as those currently used in the Canadian DaS program (with the total PCB [tPCB] value adjusted for congener-based values as described in [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] are described in [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] ; they are also briefly discussed in Supplemental Data S1.
Different countries develop ALs on the basis of different sediment size fractions and analytical methods. Because most sediment contaminants tend to associate with the finegrained sediment fraction, these differences could result in different pass/fail interpretations in various countries. However, overall sediment pass/fail outcomes using different AL sets with the same narrative intent (e.g., AL1, AL2) do not Steps followed to test assessment protocols comprised of various parameters and decision rules, and to determine resulting regulatory outcomes. AL1 ¼ lower action level; DaS ¼ Disposal at Sea program; mHQ ¼ mean hazard quotient. differ nearly as much as outcomes using different chemical action lists and decision rules [bib_ref] The assessment of sediment screening risk in Venice Lagoon and other coastal..., Apitz [/bib_ref] [bib_ref] Integrated risk assessments for the management of contaminated sediments in estuaries and..., Apitz [/bib_ref] [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref]. The consensus AL1 values used here provide a consistent set of hypothetical ALs for the full suite of contaminants in the present study; they should not be regarded as regulatory proposals.
## Use of mean hazard quotients (mhqs)
Part of the present study examines the current "one out/all out" chemical decision rule used in the Canadian DaS program, wherein the exceedance of any single benchmark, by any margin, triggers a "failure" of that sample for the entire tier. By failing a sediment for even a marginal exceedance of the AL by a single constituent, it is possible that one out/all out rules are overconservative; at the same time, by treating a sediment with a very large degree of exceedance of a contaminant in the same manner as one with only a minor exceedance, they may also be underconservative. Also, a chemical-by-chemical assessment may fail to account for the cumulative or synergistic effects of low to moderate levels of multiple contaminants, compared to slightly higher levels of a single contaminant.
The use of mHQs could potentially address these issues with one out/all out rules. Mean hazard quotients have been developed in an attempt to account for the potential effects of multiple contaminants (assuming simple additivity), as well as the magnitude of exceedance above relevant ALs. In this method, the sum of HQs for all chemicals considered is divided by the number of contaminants measured in a sample. In general, sediments with low mHQs for thresholdeffects ALs (i.e., Effects-Range Low or Threshold Effects Concentration [TEC]) are probably, but not definitely, nontoxic, and those with relatively high mHQs for probable-effects ALs (i.e., Effects-Range Medium or Probable Effects Concentration [PEC]) are often, but not always, toxic. Samples with contaminant levels between these may or may not be acutely toxic, depending on their specific chemical characteristics.
In the present paper, in which mHQ values are used to evaluate the likelihood of nontoxicity (AL1), mHQ values were calculated on the basis of chemical AL1 values. Although a range of values and approaches were considered (see Supplemental Data), in the present paper, the mHQ pass benchmark was mHQ AL1 < 0.2.
## Selecting canadian-equivalent bioassays
Canadian DaS proponents whose sediments exceed or fail Tier 1 chemical testing and who still wish to be considered for open water disposal are required to conduct 3 bioassays to proceed through Tier 2. The 3 bioassays must include an acute lethality assay , and 2 of the other sublethal or bioaccumulation tests in the Canadian DaS battery, which is comprised of polychaete survival and growth in whole sediment , luminescent bacterial inhibition in solid phase , echinoid fertilization in porewater , and bivalve bioaccumulation in whole sediment . To "pass" Tier 2 and be considered for open water disposal, proponents must pass the lethality test and at least one of the other tests. If 1 sublethal test fails, sediments are considered "sublethally toxic," and open water disposal can only be considered with "special handling" to mitigate risk. Failure of the acute lethality test and/or both sublethal tests is considered "acutely toxic" and not considered suitable for unconfined disposal at sea.
Within the database, a range of bioassay results in a variety of units was available. These data were reviewed to determine how they compared to those used in the Canadian DaS program, and to develop scenarios that were as similar as possible to those likely to be encountered within the Canadian DaS program. Although exact matches could not be found in the original database, the first column of [fig_ref] Table 1: Canadian DaS bioassays and the "matched" bioassays used from the database a... [/fig_ref] lists the bioassay tests currently being used in the Canadian DaS program for which proxies were available for inclusion in the present study database. [fig_ref] Table 1: Canadian DaS bioassays and the "matched" bioassays used from the database a... [/fig_ref] summarizes the resulting "matched battery" of similar bioassays that was identified, developed, and used for an initial assessment of Tier 1 and 2 performance. Although the studies that fed into the database for the present work used different species, methods, and endpoints than did the Canadian DaS ones, these species, methods, and endpoints are similar enough to those used in the Canadian DaS program to be able to represent a set of plausibly similar acute and sublethal bioassays.
The most significant difference between the bioassays used in the Canadian DaS program and those in the matched database battery is associated with the Microtox TM test. For this test, Canada uses a solid-phase sediment as a test medium, whereas the data available and included in the database were generated using a solvent extract. It is recognized that these results cannot be directly compared. However, the substitution of another test, using a different organism and/or test endpoint would have rendered the matched battery even less comparable to the one used by the Canadian DaS program, and would have introduced additional complications and limitations when determining the implications of results on potential changes to the Canadian DaS assessment framework. Therefore, as a first step in evaluating the existing Canadian DaS bioassay battery within an overall assessment framework, it was decided that the solvent-based Microtox TM test using a bioluminescence endpoint would make a suitable proxy for the Canadian DaS solid-phase sediment test with the same endpoint.
## Developing canadian das-equivalent pass/fail criteria
For Ampelisca survival data, if survival rates were above 64.7% of control values, sediments were deemed to have passed; otherwise they failed. These pass rates were based upon those specified in the Environment Canada method for situations when grain size corrections cannot be made (data for grain size and potentially confounding factors were not consistently available in the database), and with species-specific corrections based upon the pass rate for Eohaustorius washingtonianus. Sea urchin fertilization rates were deemed to have passed if they exceeded 70% in 100% porewater or were 70% of control values, depending upon the data available, based upon EC (2011) decision rules.
Determining Microtox TM pass criteria was not as straightforward for a number of reasons. Firstly, the data downloaded from the NOAA website had no metadata on Microtox TM analyses, methods, or units. Also, detailed examination of the data revealed disparities of several orders of magnitude in the ranges of data reported from different studies and regions that fed into the database; these disparities could not be explained by contaminant levels, nor did they correspond with results for other bioassays.
Because these data are up to decades old, and many of the people involved in its generation are retired or in other jobs, extensive correspondence with the database administrator was needed to resolve Microtox TM -related issues. As a result of these efforts, it was confirmed that the data in the database were based on assays conducted with solvent extracts. Also, it was determined that the data from 4 regions had been reported in different units than those in other regions. Corrected data were received and the database was updated (as were, we believe, the data in the online data source itself). The resulting Microtox TM data set was more consistent; regional data ranges were more comparable and corresponded better with contaminant level ranges.
Because the EC (2002) photoluminescent bacteria method is a solid-phase test, whereas the database results were from a Microtox TM solvent-extract test, Canada's pass/fail criterion could not be readily applied. As a result, the present paper used pass/fail criteria for the database Microtox TM measures that were based upon values developed by [bib_ref] Silver Spring (MD): US Dept of Commerce National Oceanic and Atmospheric Adm,..., Long [/bib_ref]. [bib_ref] Silver Spring (MD): US Dept of Commerce National Oceanic and Atmospheric Adm,..., Long [/bib_ref] generated 2 new Microtox TM critical values, both based upon statistical analyses of data from a subset of the original data from which the NOAA surveys (n ¼ 1013) that the present study draws upon were compiled. The first value (0.06 mg/mL) represents the 90% lower prediction limit (LPL) of the entire data set. The probability that a future observation from this data distribution that is less than the LPL would be more toxic (i.e., an EC50 < 0.06 mg/ mL) would be 90%. Therefore, a sample with an EC50 less than 0.06 mg/mL would be extremely toxic in this test. This value is used as a Tier 2 pass/fail criterion for the Microtox TM sublethal test (Sublethal 2 in [fig_ref] Table 1: Canadian DaS bioassays and the "matched" bioassays used from the database a... [/fig_ref].
The second [bib_ref] Silver Spring (MD): US Dept of Commerce National Oceanic and Atmospheric Adm,..., Long [/bib_ref] value (0.51 mg/mL) represents the 80% LPL with the lowest (most toxic) 10% of the data values removed from the database to eliminate their influence on the distribution of the data. Samples with EC50 values <0.51 mg/mL or >0.06 mg/mL would be considered as moderately toxic in this test. This value is used in the present study as a conservative Tier 1 screen, as noted in the It should be pointed out that both the extremely and the moderately toxic narrative intents are less conservative than those typically used for AL2 and screening, respectively, in DaS programs. However, in the absence of a comparable DaS benchmark for these endpoints, the decision was made to use published benchmarks used for sediment risk assessment to enable an initial evaluation of scenarios, including a Microtox TM screening bioassay in Tier 1.
# Data analysis
A set of regulatory scenarios was developed to evaluate the efficacy (in terms of the ability to correctly predict acute and sublethal toxicity) and implications of various potential changes to the Canadian chemical assessment protocols for DaS. The scenarios tested and evaluated in the present paper include the assessment of a broader metal and organic chemical action list, the application of a range of Tier 1 decision rules, and the addition of a screening bioassay; these scenarios were then defined as test protocols. As a baseline, the records in the data set were evaluated using the current Canadian DaS protocol (i.e., the status quo), the expanded DaS chemical action list, alternative chemical rules, and the matched bioassay battery. Subsequent scenarios then changed Tier 1 protocols to examine changes in outcomes. Detailed descriptions of scenarios evaluated in the present study can be found in Supplemental Data Tables S1-2a to S1-2e. Brief descriptions of scenarios selected for discussion in the present paper can be found in [fig_ref] Figure 2: Scenarios evaluating the effects of expanded chemical action lists with one out/all... [/fig_ref].
Scenarios evaluating regulatory outcomes. [fig_ref] Table 2: Dredged material regulatory outcomes following Tier 1 and 2 assessment with the... [/fig_ref] summarizes the Canadian DaS tiered decision rules, considering both chemical and bioassay outcomes, and the resulting [fig_ref] Table 1: Canadian DaS bioassays and the "matched" bioassays used from the database a... [/fig_ref]. Unless otherwise noted, ALs are "AL1 consensus values" as in [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] and Supplemental Data Table S1-1. "Successful" scenarios maximize Classes I, V, and VI (correct decisions), minimize Classes III (acutely toxic false negatives) and II (sublethally toxic false negatives), and do not result in disproportionate numbers of Class IV (correct decisions), which represent additional assessment for proponents. Color bars represent the proportion of samples falling in each class. Scenario descriptions describe the Lower Action Level action lists (chemicals considered and presence or absence of screening bioassessment) and the chemical decision rule (one out/all out or mHQ). Formulae are for sediment j with contaminant i at concentration ij . AL i is contaminant i specific. For multiple chemicals, the number is n. A database of 1081 samples (both chemistry and toxicity) was used. AL ¼ action level; AL1 ¼ lower action level; DaS ¼ Disposal at Sea program; HCB ¼ hexachlorobenzene; mHQ ¼ mean hazard quotient; SQG ¼ sediment quality guideline; TBT ¼ tributyltin. 1 Cd, Hg, tPAH, tPCB. 2 As, Cd, Cr, Ni, Pb, Cu, Zn, Hg, tPAH, tPCB. 3 Cd, Hg, tPAH, tPCB, TCDD, tDDT, tTBT, lindane, dieldrin, chlordane, aldrin, HCB. 4 As, Cd, Cr, Ni, Pb, Cu, Zn, Hg, tPAH, tPCB, TCDD, tDDT, tTBT, lindane, dieldrin, chlordane, aldrin, HCB. 5 Chemical rule PASS for each chemical, if
[formula] ([C] ij / AL i ) 1. 6 Chemical rule PASS for each chemical, if ((S i ¼ 1 -n([C] ij / AL1 i )) / n) < 0.2. [/formula]
regulatory outcomes: unconfined open water ocean disposal, open water ocean disposal with special handling restrictions, and no ocean disposal without confinement or treatment. Briefly, sediments that pass the chemical tier (i.e., <AL1) are considered acceptable for ocean disposal without further bioassessment. If they fail the chemical tier (i.e., >AL1), to pass Tier 2 and be considered nontoxic and acceptable for unrestricted open water disposal, proponents must pass the lethality test and both of the other (sublethal) tests. If a single sublethal test is failed, sediment is considered sublethally toxic, and special handling is required for unconfined disposal at sea. Sediments that fail either the acute toxicity assay or both sublethal assays are considered acutely toxic, and open water disposal is not permitted.
Tier 1 scenario options tested Expanded chemistry. Various Tier 1 scenarios were evaluated by comparing assessments made using the following: 1) only the regulated Canadian DaS chemical action list (Cd, Hg, total PAH, and total PCB); Mean hazard quotients chemical assessments. One out/all out chemical decision rules and mHQ chemical assessments were compared. For each scenario, hazard quotients (HQs) were calculated for all samples and chemical constituents to be considered. For contaminant i and sample j,
[formula] HQ ij ¼ ½C ij =AL i ;ð1Þ [/formula]
where [C] ij is the concentration of contaminant i in sample j, and AL i is the AL1 of interest. For one out/all out Tier 1 scenarios, the sample fails if this value is !1 for any analyte included in the chemical action list. In other scenarios, chemical rules based upon mHQs were tested. The mHQs are calculated with the following formula:
[formula] mHQ ¼ X i¼1Àn C ½ ij =AL i =n;ð2Þ [/formula]
where n is the number of analytes in the chemical action list being considered for the sample and scenario. Depending on the scenario, either AL1 or AL2 is used for this calculation (only AL1 results are discussed in the present paper; results for other mHQ approaches can be seen in Supplemental Data Tables S1-3a to S1-3d). To calculate mHQ values and conduct the work presented in here, a set of consensus AL values was used as described in [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref]. Given the origins and objectives of the ALs used in the present study, the selection of an appropriate value for an mHQ decision rule is somewhat arbitrary; although a number of mHQ values were evaluated, only the mHQ AL1 <0.2 results are shown in the present paper. Results based on mHQs calculated using AL2 values generally performed more poorly than the one out/all out rule in terms of Class II and III assignments (see Supplement Data for additional information).
Addition of Microtox TM screening bioassay. A Microtox TM screening bioassay with a very conservative pass threshold (described in [fig_ref] Table 1: Canadian DaS bioassays and the "matched" bioassays used from the database a... [/fig_ref] was added to some Tier 1 scenarios. When used, a sample was required to pass 1) the chemical rule (one out/all out or mHQ < X) and the Microtox TM bioassay screening criterion, or 2) only the conservative Tier 1 Microtox TM screening criterion (Microtox TM -only scenarios). Various combinations of these parameters were evaluated.
# Evaluating scenario results
To determine how each Tier I scenario performed in terms of predicting toxicity, scenario decision rules were applied to each sediment sample in the data set (1081 samples), and scenario outcomes were classified on the basis of their Tier 1 and Tier 2 outcomes, and whether these tiers agreed or disagreed. Although screening bioassessment Tier 1 scenarios are examined, for simplicity, Tier 1 outcomes will frequently be called "chemistry," whereas Tier 2 outcomes are referred to as "biology" outcomes in this discussion. Scenario outcomes were classified as I to VI, as in [fig_ref] Figure 1: Figure 1 [/fig_ref]. Clearly, the most undesirable classifications are II and III, false negatives in which a chemical pass (and thus no subsequent bioassessment) results in toxic sediment being approved for unconfined open water ocean disposal (DaS). Class IV, a false positive in which nontoxic samples fail the chemical tier but subsequently pass bioassessment and are thus correctly approved for DaS, result in "extra" bioassessment but are ultimately successfully classified, as are Class V and VI sediments, which fail both chemical and bioassessment and are thus subject to regulatory controls. Although a tiered program should seek to minimize false positives for cost effectiveness, this must be balanced against the overarching need to minimize false negatives to ensure that the tiered approach is sufficiently protective of the marine environment. Thus, in discussions of various assessment scenarios in the next section, effective scenarios are considered to be those that maximize the rates of Classes I, V, and VI (correct decisions) while reducing relative rates of Class II (sublethally toxic false negative) and III (acutely toxic false negative) sediments, with a secondary focus on minimizing Class IV (false positive) rates, which represent correct framework decisions made at additional Tier 2 assessment cost to proponents.
The approach taken in the present paper is based on the assumption that the Tier 2 bioassessment is the baseline measure of "truth"-that toxicity measured in field sediments evaluates whether the chemical benchmarks are appropriate. This assumption is necessary to provide some basis of comparison, but the toxicity measures themselves contain uncertainty. Bioassessment can be subject to a range of confounding factors and may be sensitive to unmeasured contaminants in the sediments; errors or misclassifications can occur in any measurement [bib_ref] Integrated risk assessments for the management of contaminated sediments in estuaries and..., Apitz [/bib_ref]. Scenarios using mHQ filters to evaluate the possibility that sediments failing a single sublethal assessment (not described here but results in Supplemental Data Table S1-3) suggest that some proportion of sediments classified as sublethally toxic (i.e., failing 1 sublethal test) could be the result of confounding factors (the authors were able to confirm that hydrogen sulfide toxicity, which has a strong confounding effect in Microtox TM bioassays done using solvent extracts, was unlikely to be the culprit because standard procedures used in the National Status and Trends Program require the removal of hydrogen sulfide prior to solvent extraction [bib_ref] Silver Spring (MD): US Dept of Commerce National Oceanic and Atmospheric Adm,..., Long [/bib_ref]. The Canadian DaS tiered framework, and many others, allow for an evaluation of potential confounding factors when interpreting Tier 1 results, but data were not available for such an assessment in the present study. Thus, although substantial changes in efficacy between scenarios may be indicative of better approaches, minor differences that could be due to unique conditions in a handful of samples should be interpreted with caution. [fig_ref] Figure 2: Scenarios evaluating the effects of expanded chemical action lists with one out/all... [/fig_ref] illustrate the number and percentage of samples that fall in each sediment class for selected scenarios. Combined outcome classifications are based upon the agreement or disagreement of Tier 1 and Tier 2 classifications, as in [fig_ref] Figure 1: Figure 1 [/fig_ref]. Regulatory outcomes at the far right are shown as the percent of samples that would result in each regulatory decision after a full Tier 1 and Tier 2 assessment as per [fig_ref] Table 2: Dredged material regulatory outcomes following Tier 1 and 2 assessment with the... [/fig_ref]. "% Ocean Disposal" outcomes are the sum of Class I to IV outcomes. Percent "Special Handling" outcomes are comprised only of Class V samples. "% No Ocean Disposal" outcomes are comprised of Class VI samples. The results using the existing Canadian DaS decision-making rules and ALs ("Status Quo") are shown in the first row of [fig_ref] Figure 2: Scenarios evaluating the effects of expanded chemical action lists with one out/all... [/fig_ref]. All other assessment protocols use consensus ALs, or mHQs calculated using the consensus ALs.
# Results
## Expanded chemistry
The classification rates using "Expanded Chemistry" assessment protocols are shown just below the Status Quo (Scenario 1) in [fig_ref] Figure 2: Scenarios evaluating the effects of expanded chemical action lists with one out/all... [/fig_ref]. These scenarios evaluate different action lists, using a one out/all out rule (if a single analyte exceeds its AL, this is a Tier 1 failure) and the bioassessment protocol in Tier 2. Using the 4 Canadian DaS analytes and the consensus AL1s would result in about 1% of samples being acutely toxic false negatives (acutely toxic sediments receiving a permit for unconfined open water disposal; Class III) and almost 15% sublethally toxic false negatives (Class II). About 18% of the time, Class IV false positives would have resulted, such that applicants would need to carry out Tier 2 assessment for sediments that would ultimately receive unconfined ocean disposal permits. Special handling requirements and no ocean disposal decisions (sublethally and acutely toxic true positives) would result about 5.4% and 4% of the time, respectively (Class V and VI).
The addition of metals only to the assessment protocol (Scenario 3: DaS þ metal) results in a reduction of both Class II and III false negatives, but an increase in Class IV false positives. The addition of organics only to the assessment protocol (Scenario 4: DaS þ organics) results in a greater reduction in Class II and III false negatives than does the addition of metals, but also a much higher rate of Class IV false positives. When both metals and organics are added to the assessment protocol (Scenario 5: Full), the Class II and III rates are lower and the Class IV rates are higher than when a single class of chemicals is added, suggesting that shorter lists of chemical constituents are less effective than the longer lists in predicting sediment toxicity, and that the addition of constituents to the Tier 1 chemical assessment substantially improves the protectiveness of the assessment protocol. The "Full" protocol does this by more than doubling the Class IV false positive rate (therefore requiring Tier 2 assessment more often). However, among these Tier 2 assessments, a large proportion of the samples will ultimately pass bioassessment and be deemed suitable for ocean disposal. Specifically, these Tier 2 assessments will pass (as Class IV) more than twice as often as they will fail and fall in Classes V or VI.
## Mean hazard quotient chemical assessments
The bottom grouping in [fig_ref] Figure 2: Scenarios evaluating the effects of expanded chemical action lists with one out/all... [/fig_ref] shows the results of assessment protocols using a mean hazard quotient (mHQ AL1 ) instead of the one out/all out rule to define chemical pass or fail.
For all sets of analytes, the performance of the mHQ AL1 is much better than the one out/all out rule, substantially reducing the Class II and III false negative rates, but at a cost of much higher Class IV false positive rates.
The fact that the sublethal mHQ performance for DaS þ organics (Scenario 8) is worse than for DaS þ metals (Scenario 7), although DaS þ organics (Scenario 13) is better than for DaS þ metals (Scenario 12) is better in the one out/all out rule protocols, suggests that there is still scope for work in deriving the appropriate organic ALs for a tiered assessment. The focus of the present work, however, is not to suggest the specific ALs to apply, but to demonstrate how different general approaches affect framework performance. [fig_ref] Figure 3: Scenarios evaluating expanded chemical action lists with Tier 1 decision rules using... [/fig_ref] illustrates the classifications that would result from adding a Microtox TM screening bioassay to Tier 1. In these assessment protocols, a conservative pass/fail criterion was applied to the database Microtox TM data. This assessment was added to the 2 sets of Tier 1 chemical decision rules as follows: To pass Tier 1, the sediment would have to pass the chemical decision rule (either one out/all out or the mHQ AL1based rule), and pass the screening bioassay. The results allow an evaluation of whether this screening bioassessment could complement Tier 1 chemistry by potentially flagging sediments that are contaminated with constituents absent from the chemical list.
## Addition of a screening bioassay
The row just below the status quo on [fig_ref] Figure 3: Scenarios evaluating expanded chemical action lists with Tier 1 decision rules using... [/fig_ref] illustrates results when a screening Microtox TM bioassay is used alone in Tier 1 without any chemistry (Scenario 10). The rest of the grouping below the status quo in [fig_ref] Figure 3: Scenarios evaluating expanded chemical action lists with Tier 1 decision rules using... [/fig_ref] shows classifications following the simplest application of a screening bioassay, by adding it to a simple one out/all out approach. For all chemical action lists, the addition of the screening Microtox TM to the Tier 1 assessment increases agreement between chemical and biological screening performance in terms of predicting toxicity, but not in terms of predicting nontoxicity. Acutely toxic false negative (Class III) rates were substantially reduced; in the cases of the DaS þ organics (Scenario 13) and Full (Scenario 14) list of analytes, Class III rates were reduced to 0.1%; this represents a single sample out of the 1081 examined. At the same time, the addition of the screening bioassay to the full chemical action list (Scenario 14) reduces sublethally toxic false negatives (Class II) from 6.9% to 4.4%.
The bottom [fig_ref] Figure 3: Scenarios evaluating expanded chemical action lists with Tier 1 decision rules using... [/fig_ref] grouping shows classifications following the combination of a screening bioassay with an mHQ in Tier 1. This assessment protocol yielded the lowest Class II (sublethally toxic false negative) rates of any scenario set, and consistently low Class III (acutely toxic false negative) rates, but the highest Class IV (false positive) rates of up to 50%.
The question of whether a screening bioassay alone would be enough, and whether chemical measures are unnecessary, is addressed by the increasingly better performance when the chemical list is expanded, even with the addition the screening bioassay. [fig_ref] Figure 3: Scenarios evaluating expanded chemical action lists with Tier 1 decision rules using... [/fig_ref] illustrates an assessment protocol (Scenario 10; mTox only Tier 1) in which only the screening bioassay (without chemical assessment) is carried out in Tier 1. As can be seen, the screening bioassay alone performs as well for Class III as the Status Quo (Scenario 1), which uses only the current Canadian DaS ALs, and somewhat better than the Status Quo for Class II. The screening bioassay alone (Scenario 10) performs substantially better than the DaS assessment protocol (Scenario 2 in [fig_ref] Figure 2: Scenarios evaluating the effects of expanded chemical action lists with one out/all... [/fig_ref] , which uses only consensus AL values) at minimizing Class II sediments, but performs slightly less well at minimizing Class III sediments (by 2 samples).
# Discussion and conclusions
## Expanding minimum chemistry
The present study has found that for most scenarios, increasing the number of chemical analytes not only increased Tier 1 failure but also better predicted Tier 2 toxic (but not nontoxic) outcomes, suggesting that the rate of false negatives could be reduced by adding chemicals to the current Canadian DaS action list.
It is worth noting that the increased protectiveness achieved through the addition of chemical analytes comes at the cost of a (sometimes) greatly increased false positive rate compared with the current Canadian DaS framework. Among the increased number of samples subjected to bioassessment, generally well over half of these pass Tier 2 and are ultimately designated as suitable for unconfined ocean disposal. In other words, additional Tier 2 assessment does not drastically reduce the number of projects that are ultimately allowed for open water disposal. The results of this work suggest that requiring additional analyses of the chemical constituents examined in the present study is warranted because the additional information maximizes the rate at which correct "toxic" assignments are made (i.e., maximal reduction in Class II and III sediments is achieved). Also, the addition of only a few chemical constituents is needed to achieve improvements in the rate of correct decisions, and so the increased analytical burden to proponents is expected to be minimal.
The appropriate balance between maximizing the rate of correct environmental decisions and minimizing the assessment burden placed on proponents (i.e., how many false positives and negatives are acceptable) is ultimately a policy decision that environmental managers have to make when designing and applying DM assessment frameworks. When considering potential increases to the assessment burden on proponents, it may be possible to streamline assessment in other areas of the framework. For example, in the Canadian DaS context, one of the CDMMD recommendations not considered in the present paper or evaluated to date, was to streamline and reduce assessment requirements for routine, low-risk permit applications by creating a separate prescreening assessment that is distinct from Tier 1. This separate assessment would help to focus any increased assessment efforts on permit applications associated with greater uncertainty and/or greater risk of contamination, while ensuring that low-risk projects are assessed using a streamlined process.
Of course, even with the extended chemical action list examined in the present paper, only a small fraction of the millions of chemicals that might be found in the environment are represented [bib_ref] Pharmaceuticals and personal care products (PPCPs) as environmental pollutants -Pollution from personal..., Daughton [/bib_ref] ; the inclusion of further organic contaminants in an action list was addressed to some extent in [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref]. It has been noted that regulatory programs are skewed toward those chemicals that have the most data on human health risk, though these may not be the highest risk compounds [bib_ref] The science charade in toxic risk regulation, Wagner [/bib_ref] and that, for example, in European rivers, contaminants on priority lists could explain only a small fraction of the observed toxicity [bib_ref] Responses in sediment bioassays used in the Netherlands: Can observed toxicity be..., Lahr [/bib_ref] [bib_ref] Effect-directed analysis of key toxicants in European river basins: A review, Brack [/bib_ref]. Generally speaking, when considering which contaminants to include in an action list, it would be worth evaluating the inclusion of so-called "emerging contaminants" in addition to the legacy contaminants typically considered by DM regulatory frameworks around the world. However, such an approach is not possible with this data-mining exercise, which relies on both extensive data on colocated sediment chemistry and toxicology and risk-relevant ALs to carry out an assessment.
## Mean hazard quotient chemical assessments
Although broad differences in the mHQ approach versus the one out/all out protocol can be illustrative, they should be viewed and generalized with caution. The performance of an mHQ approach is sensitive to the chemical action list and the ALs applied. Thus, any detailed work to examine or optimize an mHQ approach should be carried out only after chemical action lists and both AL1 and AL2 values have been selected for a program. The metal consensus ALs used in the present paper are based upon 7 to 8 disposalat-sea ALs from around the world. With the exception of tPAH and tPCB, the organic consensus ALs, on the other hand, are based on far fewer values, and often, on values not generated for DaS programs. Thus, although they can be used here to illustrate differences in approaches, the values that should be used in a DaS program have yet to be determined.
Differences in the performance of mHQ decision rules for additional metals versus additional organics suggested that the ALs applied in the present study might not be conservative enough for some organics; this may explain why [bib_ref] Anatomy of a decision: Potential regulatory outcomes from changes to chemistry protocols..., Apitz [/bib_ref] observed that the addition of many organics to the chemical action lists had a much less dramatic effect on Tier 1 chemical outcomes than did the addition of metals. For many organic compounds, including dioxins and furans, there are very few ALs related to dredging. Given this situation, the organic consensus values used in the present work were based on ecological risk assessment [ERA] rather than on dredging values, and were derived differently and intended for a different purpose.
The effects of applying an mHQ, rather than a one out/all out rule, in the Tier 1 assessment were promising, although not completely straightforward. Further work and consideration to identify the best approaches to the use of mHQs in a tiered framework is recommended. Incorporating the latest dredging-specific organic ALs and approaches to regulating organic levels in sediment may identify potential DM framework improvements beyond those identified as a result of the present work.
## Addition of a tier 1 screening bioassay
A screening bioassessment alone did not improve Tier 1 decisions for acutely toxic sediments compared with status quo chemistry alone (although it did so for sublethally toxic sediments); the best decision outcomes were observed using the screening bioassessment and an expanded chemical action list together. Specifically, the addition of a screening bioassay in the form of a solvent-based Microtox TM assay with a conservative pass criterion substantially reduced the rates of both acutely and sublethally toxic false negatives; its use, along with an expanded chemical action list reduced the number of acutely toxic sediments that were missed in Tier 1 (i.e., Class III sediments) to very low levels.
The screening bioassay, with the addition of constituents to the chemical list, consistently improved the performance, decreased Class II and III sediments, but with a cost of higher false positive (Class IV) rates; these increased 5% to 7% compared to the same chemical lists without the screening bioassessment. Thus, it seems that regulatory performance would be relatively unchanged if the current chemical action list were replaced with this screening bioassay in Tier 1, but it would be substantially enhanced by adding a screening bioassay to the current chemical parameters tested in the first tier, and progressively improved with each extended chemical action list. The true positive rates for DaS alone (Scenario 2 in [fig_ref] Figure 2: Scenarios evaluating the effects of expanded chemical action lists with one out/all... [/fig_ref] , Microtox TM alone in Tier 1 (Scenario 10 in [fig_ref] Figure 3: Scenarios evaluating expanded chemical action lists with Tier 1 decision rules using... [/fig_ref] , and their combination (Scenario 11 in [fig_ref] Figure 3: Scenarios evaluating expanded chemical action lists with Tier 1 decision rules using... [/fig_ref] suggest that chemistry and the screening bioassay capture different groupings of sediments; this is much clearer when one compares Class V (sublethally toxic true positives) to the smaller population of Class VI sediments.
It should be noted that acutely toxic sediments (those that failed the acute bioassay and/or both sublethal assays) make up only 5% of the samples in the database (55 out of 1081). This is a relatively small population of samples from which to draw definitive conclusions based on small shifts in Class III and Class VI assignments (because acutely toxic samples will be classed as one or the other, depending upon the scenario). Thus, although small changes in these assignments could indicate more effective protocols, they may also be affected by very sample-specific differences that may not be generalizable.
It is possible that the screening bioassay is identifying a population of sediments with contaminants not on the chemical action lists, but it is also possible that the screening bioassay is sensitive to potential confounding factors that also are causing higher levels of bioassay failure not driven by chemical contaminants. On the other hand, it is also possible that the very conservative bioassay threshold is failing a largeenough proportion of samples (without a real link to toxicity), which coincidentally then also fail Tier 2 bioassessment, along with the large proportion of Class IV sediments, which then ultimately pass bioassessment. It is difficult to distinguish between these potential explanations with the available data, but results do suggest that the use of a screening bioassessment in conjunction with chemistry warrants further investigation. Because a key assumption behind the present study is that Tier 2 bioassessment represents the true state, a deeper evaluation of this question is outside the scope of this work, but investigations into indicators of confounding effects is ongoing in related work.
To further consider the addition of a Microtox TM screening assessment to Tier 1, it will be necessary to examine the performance of the solid-phase photoluminescent bacteria method, which will not necessarily comprise an assessment tool that is equivalent to the solvent-based Microtox TM method evaluated in the present work. Because the solventbased method extracts most chemical constituents from the chemical matrix (even those that may not otherwise be bioavailable), solvent-based assays are arguably more conservative, and therefore better suited to a screening level assay, than are their solid-phase counterparts. However, solid-phase tests are likely more ecologically relevant than solvent-based extractions.
In evaluating the performance of a Microtox TM toxicity test in Tier 1, the present work revealed that Microtox TM alone (Scenario 10 in [fig_ref] Figure 3: Scenarios evaluating expanded chemical action lists with Tier 1 decision rules using... [/fig_ref] achieved similar pass/fail rates as the current DaS chemical action list (Scenario 2 in [fig_ref] Figure 2: Scenarios evaluating the effects of expanded chemical action lists with one out/all... [/fig_ref] at predicting Tier 2 acute toxicity. Although this is not the optimal performance for Tier 1, it may be good enough to replace chemistry in a new, CDMMD-recommended "pre-Tier 1, screening assessment" that is distinct from Tier 1, and intended to confirm only that sediments previously subjected to an expanded chemical assessment and identified as lowrisk do not warrant extensive further assessment. Given differences in the Microtox TM and photoluminescent bacteria methods used by the USEPA and EC, a careful evaluation of the performance of the solid-phase test method as a low-risk sediment screen would be needed before deciding to proceed with a Microtox-only screening assessment, but doing so could potentially represent assessment cost savings for low-risk dredging projects.
The present paper examined the implications of adding a Microtox TM test to Tier 1. It is recognized, however, that there is an opportunity for future work exploring how other biological tests would perform as screening or prescreening tools. Such work could focus on other sublethal tests in the existing Canadian DaS bioassessment battery (e.g., , screening bioassays used to steer chemical testing in other regulatory programs, cell-based bioassays capable of detecting genotoxicity or endocrine disruption, or genomics methods being developed for toxicity identification and evaluation (TIE) purposes.
## The need for tiers
The analyses reported here suggest that an increase in the number of chemical parameters results in a moderate increase in the number of samples requiring Tier 2 bioassessment, resulting in an increased rate of correct environmental decisions. Following from this, one could argue that it may be more time and cost effective to carry out a full suite of bioassessments in parallel with chemical analyses at the screening stage. The data in the present study do not fully support such a conclusion for 2 reasons. First, most scenarios still had 20% to 40% of samples that passed in Tier 1 and would not need further assessment. If a full suite of bioassessments was required in Tier 1, these Class I sediments would be subject to bioassessment with no change to their rate of correct environmental decisions, while the rate of misassignment of the Class II and III sediments in some scenarios was quite low. Secondly, due to the perceived complexity and uncertainty of Tier 2 biological assessments, potential DaS permit applicants in Canada sometimes withdraw their applications when Tier 1 fails, choosing either not to dredge (potentially inhibiting development) or to go directly to land-based disposal, which falls into a different regulatory framework but which may or may not have fewer overall ecological and economic impacts.
Ultimately, the balance between the objective of minimizing false negatives and minimizing undue analytical burdens on applicants is a policy decision. It is not clear to what extent the higher numbers of Tier 2 outcomes will affect the decisions and behavior of applicants, but given past trends, it is not unreasonable to assume that any increase in the number of applicants required to conduct bioassays could result in a decrease in applications for disposal-at-sea permits. Such a decrease would likely result in proponents pursuing alternative and potentially less ecologically desirable disposal options (e.g., land based). Requiring bioassessment only when indicated based upon chemical results may maintain current rates of DaS permit applications. Still, some of the better performing scenarios evaluated here result in significant increases of samples requiring Tier 2 assessment (although in those scenarios, a large proportion of samples would eventually be permitted for DaS). Thus, the relatively high rate of DM found suitable for DaS following bioassessment in the present study may encourage proponents who fail Tier 1 to pursue bioassessment when they may currently do otherwise.
## The need for canadian data
The present work assumes that the sediments analyzed in this US-based database are representative of what might be encountered in the Canadian DaS program. To assist with any updates to Canada's sediment characterization process, future efforts will be made to integrate into the data set as much Canadian data as possible. Efforts are also underway to centralize a repository of DM characterization data, including data from a range of "clean" and "contaminated" sites, so that it will be available for future data-mining studies such as this one.
associates for their support in resolving questions on the data sets, and Lorraine Brown Read for her thoughtful review and suggestions for future analyses.
Disclaimer-This paper does not necessarily represent the views of Environment Canada or any affiliations represented by the authors. References to brand names and trademarks in this document are for information only and do not constitute endorsements by Environment Canada or the authors. The authors do not intend to suggest conclusions on the potential ecological risk or regulatory status of the sediments from which the database was drawn; these samples were not collected for the assessment of ocean disposal and this review represents an analysis of only a small fraction of the data available. These data are used only to provide a data set that might realistically represent the range of sediment types that might be encountered by the Canadian DaS program, and to evaluate the potential performance of a range of DM DaS decision rules.
Data Accessibility-Data are from NOAA Status and Trends (NS&T) and Mussel Watch datasets recently placed online (National Status and Trends Program, http://ccma. nos.noaa.gov/about/coast/nsandt/download.aspx).
## Supplemental data
Supplemental file provides scenario details and discusses the issue of potential confounding factors. [fig_ref] Figure 1: Figure 1 [/fig_ref] -2. Percentage outcomes for scenarios with various Tier 2 approaches. Scenario labels, numbers, conditions and outcomes as described in . Classifications as in [fig_ref] Figure 1: Figure 1 [/fig_ref] (paper).
[fig] Figure 1: Figure 1. Steps followed to test assessment protocols comprised of various parameters and decision rules, and to determine resulting regulatory outcomes. AL1 ¼ lower action level; DaS ¼ Disposal at Sea program; mHQ ¼ mean hazard quotient. [/fig]
[fig] Figure 2: Scenarios evaluating the effects of expanded chemical action lists with one out/all out and mHQ Tier 1 decision rules. Bioassay pass or fail criteria as in [/fig]
[fig] Figure 3: Scenarios evaluating expanded chemical action lists with Tier 1 decision rules using a Microtox TM screening bioassay and their results. Bioassay pass/fail criteria as inFigure 1. Unless otherwise noted, ALs as in [/fig]
[fig] 2: the DaS list as in 1) with the addition of 6 other metals (Cr, Ni, Pb, Cu, Zn, and As); 3) the DaS list with the addition of 7 other organics (total DDT [tDDT, the sums of dichlorodiphenyldichlorethane (DDD), dichlorodiphenyldichloroethylene (DDE), and DDT values when reported], total tributyltin [tTBT, the sum of tributyltin and dibutyltin], lindane, dieldrin, chlordane [the sum of alpha and gamma chlordane when reported], aldrin and hexachlorobenzene (HCB]); or 4) the DaS list with the addition of all available metals and organics (full). [/fig]
[table] Table 1: Canadian DaS bioassays and the "matched" bioassays used from the database a ASTM ¼ American Society for Testing and Materials; DaS ¼ Disposal at Sea program; EC ¼ Environment Canada; IC50 ¼ ; PW ¼ pore water; SOP ¼ standard operating procedure. a The 4 left columns describe the bioassays and "fails if" criteria used in this study to match those used in the Canadian DaS framework. Endpoints, organisms, and roles in the study are also described. The last 2 columns describe the equivalent tests used in the Canadian DaS program. last row of Table 1. In the present paper, Microtox TM values are compared to this more conservative value to provide a hypothetical Tier 1 "screening" bioassay in a number of scenarios tested. [/table]
[table] Table S1 - 1: ALs used in the present study Table S1-2a. Scenario conditions, Scenarios 1 to 14 Table S1-2b. Scenario conditions, Scenarios 15 to 27 Table S1-2c. Scenario conditions, Scenarios 28 to 40 Table S1-2d. Scenario conditions, Scenarios 41 to 47 Table S1-2e. Scenario conditions, Scenarios 48 to 54 Table S1-3a. Scenario outcomes, Scenarios 1 to 14 Table S1-3b. Scenario outcomes, Scenarios 15 to 27 Table S1-3c. Scenario outcomes, Scenarios 28 to 40 Table S1-3d. Scenario outcomes, Scenarios 41 to 54 [/table]
|
10.1038/s43705-021-00066-4
|
CCBY
|
9723692
|
37938239
|
s2orc_pubmed_articles
|
Xylan utilisation promotes adaptation of Bifidobacterium pseudocatenulatum to the human gastrointestinal tract
Dietary carbohydrates impact the composition of the human gut microbiota. However, the relationship between carbohydrate availability for individual bacteria and their growth in the intestinal environment remains unclear. Here, we show that the availability of long-chain xylans (LCX), one of the most abundant dietary fibres in the human diet, promotes the growth of Bifidobacterium pseudocatenulatum in the adult human gut. Genomic and phenotypic analyses revealed that the availability of LCXderived oligosaccharides is a fundamental feature of B. pseudocatenulatum, and that some but not all strains possessing the endo-1,4-β-xylanase (BpXyn10A) gene grow on LCX by cleaving the xylose backbone. The BpXyn10A gene, likely acquired by horizontal transfer, was incorporated into the gene cluster for LCX-derived oligosaccharide utilisation. Co-culturing with xylanolytic Bacteroides spp. demonstrated that LCX-utilising strains are more competitive than LCX non-utilising strains even when LCX-derived oligosaccharides were supplied. In LCX-rich dietary interventions in adult humans, levels of endogenous B. pseudocatenulatum increased only when BpXyn10A was detected, indicating that LCX availability is a fitness determinant in the human gut. Our findings highlight the enhanced intestinal adaptability of bifidobacteria via polysaccharide utilisation, and provide a cornerstone for systematic manipulation of the intestinal microbiota through dietary intervention using key enzymes that degrade polysaccharide as biomarkers. ISME Communications; https://doi.
# Introduction
The human gut microbiota (HGM) and host health are interconnected in various manners [bib_ref] The impact of the gut microbiota on human health: an integrative view, Clemente [/bib_ref]. Dietary glycan is a major nutrient for the HGM and significantly affects its composition [bib_ref] How glycan metabolism shapes the human gut microbiota, Koropatkin [/bib_ref]. One feasible strategy for manipulating the HGM is dietary intervention using indigestible carbohydrates that selectively increase endogenous health-promoting microorganisms [bib_ref] Probiotics, prebiotics, and synbiotics-approaching a definition, Schrezenmeir [/bib_ref]. Therefore, understanding glycan metabolic ability of specific HGM species could contribute to establishing a systematic means of manipulating gut microbiota. The HGM possesses a wide variety of carbohydrate-active enzymes (CAZymes) used to break down carbohydrates [bib_ref] The abundance and variety of carbohydrate-active enzymes in the human gut microbiota, Kaoutari [/bib_ref] [bib_ref] Microbial degradation of complex carbohydrates in the gut, Flint [/bib_ref] , indicating that carbohydrate availability plays an important role in microbiota adaptation to the intestinal environment. Although specific carbohydrate availability influences niche acquisition in the gut of animal models and human infants [bib_ref] An exclusive metabolic niche enables strain engraftment in the gut microbiota, Shepherd [/bib_ref] [bib_ref] Genetic determinants of in vivo fitness and diet responsiveness in multiple human..., Wu [/bib_ref] [bib_ref] A key genetic factor for fucosyllactose utilization affects infant gut microbiota development, Matsuki [/bib_ref] , the relationship between carbohydrate availability for individual types of bacteria and their growth remains obscure in the adult human gut, the complex bacterial composition of which widely varies among individuals.
Bifidobacteria, which belong to the phylum Actinobacteria, represent one of the core constituents of the HGM [bib_ref] Distribution of bifidobacterial species in human intestinal microflora examined with 16S rRNA-genetargeted..., Matsuki [/bib_ref]. The genus Bifidobacterium is characterised by an abundance of genes related to carbohydrate metabolism [bib_ref] Genomic encyclopedia of type strains of the genus Bifidobacterium, Milani [/bib_ref] ; thus, glycan metabolism is thought to be important for their adaptation to the intestinal environment [bib_ref] Genomics of the genus Bifidobacterium reveals species-specific adaptation to the glycan-rich gut..., Milani [/bib_ref]. Beyond some uncharacterised species such as B. pseudocatenulatum, glycan availability of the major HGM bifidobacterial species has been actively investigated. Infant-dominant taxa (e.g., B. breve, B. bifidum and B. longum subsp. infantis) can utilise human milk oligosaccharides (HMO) [bib_ref] The genome sequence of Bifidobacterium longum subsp. infantis reveals adaptations for milk..., Sela [/bib_ref] [bib_ref] Insights from genomes of representatives of the human gut commensal Bifidobacterium bifidum, Duranti [/bib_ref] [bib_ref] Comparative genomics and genotype-phenotype associations in Bifidobacterium breve, Bottacini [/bib_ref] , whereas adult-dominant taxa (e.g., B. adolescentis and B. longum subsp. longum) can utilise plantderived polysaccharides, including starch and the degradation products of long-chain xylans (LCX) [bib_ref] Gene-trait matching across the Bifidobacterium longum pan-genome reveals considerable diversity in carbohydrate..., Arboleya [/bib_ref] [bib_ref] Evaluation of genetic diversity among strains of the human gut commensal Bifidobacterium..., Duranti [/bib_ref].
LCX are primary components of various plant cell walls, and are the typical indigestible polysaccharides found in the human diet [bib_ref] REVIEW: variability in fine structures of noncellulosic cell wall polysaccharides from cereal..., Collins [/bib_ref]. LCX consists of a β-1,4-bonded xylose backbone with decorations such as arabinose. In the HGM, a limited number of species of Bacteroidetes and Firmicutes have been known to hydrolyse LCX [bib_ref] Characterization of the xylan-degrading microbial community from human faeces, Chassard [/bib_ref] , and the respective degradation mechanisms have been elucidated in detail [bib_ref] Glycan complexity dictates microbial resource allocation in the large intestine, Rogowski [/bib_ref] [bib_ref] Differential bacterial capture and transport preferences facilitate co-growth on dietary xylan in..., Leth [/bib_ref]. Endo-1,4-β-xylanase, belonging to the glycoside hydrolase (GH) 10 family, is a key enzyme that extracellularly cleaves the xylose backbone of LCX, which allows oligosaccharide transporters to take up degradation products such as xylooligosaccharides (XOS) [bib_ref] Glycan complexity dictates microbial resource allocation in the large intestine, Rogowski [/bib_ref] [bib_ref] Differential bacterial capture and transport preferences facilitate co-growth on dietary xylan in..., Leth [/bib_ref]. Parts of short-chain oligosaccharides are released during this process into the extracellular milieu, where they are metabolised by other intestinal bacteria [bib_ref] Glycan complexity dictates microbial resource allocation in the large intestine, Rogowski [/bib_ref] [bib_ref] Prebiotic and other health-related effects of cereal-derived arabinoxylans, arabinoxylan-oligosaccharides, and xylooligosaccharides, Broekaert [/bib_ref]. Therefore, a food web of LCX degradation functions in the human gut, and HGM species with endo-1,4-β-xylanase are generally classified as primary degraders that metabolise LCX and secondary consumers that can only metabolise LCX degradation products.
Bifidobacteria are considered secondary LCX consumers because a strain that can cleave the xylose backbone has not been identified, but several strains can utilise LCX-derived oligosaccharides [bib_ref] Prebiotic and other health-related effects of cereal-derived arabinoxylans, arabinoxylan-oligosaccharides, and xylooligosaccharides, Broekaert [/bib_ref] [bib_ref] Biochemistry of complex glycan depolymerisation by the human gut microbiota, Ndeh [/bib_ref]. However, in this study, we found that some strains of B. pseudocatenulatum, an adult-dominant species [bib_ref] Distribution of bifidobacterial species in human intestinal microflora examined with 16S rRNA-genetargeted..., Matsuki [/bib_ref] [bib_ref] The gut microbiome of healthy Japanese and its microbial and functional uniqueness, Nishijima [/bib_ref] [bib_ref] Quantitative PCR with 16S rRNA-gene-targeted species-specific primers for analysis of human intestinal..., Matsuki [/bib_ref] whose carbohydrate utilisation has not been thoroughly investigated, possess endo-1,4-β-xylanase and could be primary degraders of LCX. Our data revealed the genetic background and ecological significance of a previously overlooked phenotype of bifidobacteria. Furthermore, we confirmed via dietary intervention that LCX availability affects the growth of B. pseudocatenulatum in the adult human intestine.
## Materials and methods genome sequencing
We sequenced the genomes of 35 strains of B. pseudocatenulatum (Supplementary . These strains were isolated at the Yakult Central Institute and the species were identified based on the 16S rRNA gene sequence analysis. These strains have been isolated in the course of various studies over the past few decades, including many studies on infants and adults. B. pseudocatenulatum cultures were anaerobically incubated in modified Gifu anaerobic medium (Nissui Pharmaceutical, Tokyo, Japan) supplemented with lactose and glucose (both 0.5% wt/vol) at 37°C for 16 h. These culture conditions were applied throughout the study unless stated otherwise. The detailed procedures for genomic DNA extraction, library preparation for MiSeq (Illumina, San Diego, CA, USA), MinION (Oxford Nanopore Technologies, Oxford, UK) and PacBio RS2 (Pacific Biosciences, Menlo Park, CA, USA), and sequencing are described in the Supplementary Methods.
## Genome assembly, gene prediction and pangenome analysis
We used Unicycler [bib_ref] Unicycler: resolving bacterial genome assemblies from short and long sequencing reads, Wick [/bib_ref] with default parameters for both short-read and hybrid assembly, and Prokka [bib_ref] Prokka: rapid prokaryotic genome annotation, Seemann [/bib_ref] with default parameters for annotating the reconstructed genomes and those downloaded from the RefSeq database. The annotated genomes were then processed with Roary [bib_ref] Roary: rapid large-scale prokaryote pan genome analysis, Page [/bib_ref] with a default gene identity cut-off parameter of 95% for species level pangenome analysis. A representative sequence from each gene cluster was translated into a protein sequence, and CAZymes were identified using the dbCAN2 server [bib_ref] dbCAN: a web resource for automated carbohydrate-active enzyme annotation, Yin [/bib_ref]. Proteins were considered CAZymes if they were identified using HMMER, DIAMOND and Hotpep with default parameters. We then built a CAZyme gene distribution matrix based on the gene presence-absence table determined using Roary.
## Carbohydrate utilisation assays
Strains of B. pseudocatenulatum were cultured until they reached the exponential phase, centrifuged, and then, the resulting pellets were suspended to an OD 600 of 0.2 in modified peptone yeast extract (PY) medium (100 mM PIPES, pH 6.7, 2 g/L peptone, 2 g/L BBL trypticase peptone, 2 g/L bacto-yeast extract, 8 mg/L CaCl 2 , 19.2 mg/L MgSO 4 - 7H 2 O, 80 mg/L NaCl, 4.9 mg/L hemin, 0.5 g/L L-cysteine hydrochloride and 100 ng/L vitamin K1). These suspension cultures were inoculated (1% vol/vol) into modified PY medium supplemented with 0.5% (wt/vol) XOS (Xylo-Oligo95P, B Food Science, Aichi, Japan) (PY-XOS), wheat arabinoxylan (Megazyme, Bray, Ireland) (PY-AX) or beechwood xylan (Sigma-Aldrich, Darmstadt, Germany) (PY-XY) and covered with sterile mineral oil (50 μL) to prevent evaporation. Growth was monitored anaerobically by measuring the OD 600 using a PowerWave 340 plate reader (BioTek, Winooski, VT, USA) every 30 min in an anaerobic chamber for 48 h. The organic acids produced in PY-XY were analysed using high-pressure liquid chromatography as described [bib_ref] A key genetic factor for fucosyllactose utilization affects infant gut microbiota development, Matsuki [/bib_ref].
## Cloning, expression and purification of recombinant bpxyn10a
The GH10 domain of the BpXyn10A gene was amplified by PCR using the primers xynA-GH-F (5'-CATCATCATCATCATGCGGAAGGCGACGCCGTA-3') and xynA-GH-R (5'-AGCAGAGATTACCTAATCCTTGAATGCGTTCATGC-3'), with the genomic DNA of YIT 11027 as a template. A linearised vector was synthesised by PCR using primers pColdII-F (5'-GTAATCTCTGCTTA AAAGCACAGAATCTA-3') and pColdII-R (5'-ATGATGATGATGATGATGCAC TTTGT-3'), and the pColdII vector (Takara Bio, Otsu, Japan) as a template. These fragments were ligated using In-Fusion HD Cloning Kits (Takara Bio, Otsu, Japan), resulting in pColdII-xynA. Escherichia coli BL21 was transformed with pColdII-xynA and cultured to express recombinant BpXyn10A as described by the manufacturer. Bacterial cells were harvested by centrifugation and lysed with B-PER Bacterial Cell Lysis Reagent (Thermo Fisher Scientific, Waltham, MA, USA) containing lysozyme at 100 µg/mL and 10 U/mL of DNase I. Recombinant BpXyn10A was further purified using Ni-NTA Spin Column (Qiagen, Hilden, Germany) and analysed by SDS-PAGE.
Endo-xylanase activity assay B. pseudocatenulatum YIT 11027, YIT 11952 and YIT 4072 T cells were grown anaerobically in PY-AX or PY-XOS medium for 16 h. Cultures (1.5 mL) were centrifuged (8000× g for 2 min at room temperature); then, supernatants were sterilised by passage through a 0.22-μm filter. Pelleted cells were washed with modified PY medium and resuspended in 1.5 mL of the same medium. The endo-xylanase activity of the supernatant and the cell fractions were assayed using Xylanase Assay kits (XylX6 method) (Megazyme, Bray, Ireland) as described by the manufacturer. According to the manufacturer, this kit is designed to specifically detect only endoxylanase activity, and not xylosidase or exo-xylanase enzyme activity.
Purified BpXyn10A-added culture B. pseudocatenulatum YIT 4072 T and Ba. ovatus YIT 6161 T cells were cultured anaerobically until they reached the exponential phase. Thereafter, cultures (200 μL) were centrifuged (8000× g for 2 min at room temperature), then pelleted cells were resuspended in modified PY medium (500 μL), and inoculated (1% vol/vol) into PY-AX medium supplemented with 0, 10, 100 and 1000 ng/mL purified recombinant BpXyn10A. Growth was monitored anaerobically by measuring the OD 600 using the PowerWave 340 plate reader.
# Rna-seq analysis
B. pseudocatenulatum YIT 11952 was cultured in modified PY medium supplemented with 0.5% (wt/vol) lactose, xylose, XOS, beechwood xylan or arabinoxylan and harvested at mid-to late-log phase. The detailed procedures for total RNA extraction, rRNA removal and sequencing using MiSeq are described in the Supplementary Methods. We obtained a total of 23 million paired-end reads. Low-quality bases (average quality <30) were trimmed off at the 3', and the resulting reads with N bases, or <70 bp long, were filtered out using cutadapt [bib_ref] Cutadapt removes adapter sequences from high-throughput sequencing reads, Martin [/bib_ref]. Ribosomal RNA reads were removed using SortMeRNA [bib_ref] SortMeRNA: fast and accurate filtering of ribosomal RNAs in metatranscriptomic data, Kopylova [/bib_ref]. Filtered reads were then mapped to the complete genome sequence of B. pseudocatenulatum YIT 11952 using Bowtie2 [bib_ref] Fast gapped-read alignment with Bowtie 2, Langmead [/bib_ref] , and read counts of each gene were determined with featureCounts [bib_ref] featureCounts: an efficient general purpose program for assigning sequence reads to genomic..., Liao [/bib_ref]. The expression levels of each gene were quantified as transcripts per million calculated using Microsoft Excel 2013.
Batch co-cultures of B. pseudocatenulatum and Ba. ovatus
The 36 strains of B. pseudocatenulatum and Ba. ovatus YIT 6161 T were separately grown before co-culture. Cells harvested at the mid-to late-log phase were resuspended in modified PY medium adjusted to an OD 600 of 0.2 (B. pseudocatenulatum strains; equivalent to 1 × 10 8.4-9.0 cells/mL) or 0.02 (Ba. ovatus, equivalent to 1 × 10 8.7-9.0 cells/mL) so that the numbers of cells were similar. Equal amounts of each suspension were inoculated at 2% (vol/vol) into the PY-AX medium. After 48 h of anaerobic co-culture, DNA was extracted using the beads-phenol method as described above. Each sample was analysed by quantitative PCR using an AB7500 real-time qPCR system (Thermo Fisher Scientific, Waltham, MA, USA) to determine cell numbers with specific primers for B. pseudocatenulatum (BiCATg-1: 5'-CGGATGCTCCGACTCCT-3' and BiCATg-2: 5'-CGAAGGCTTGCTCCCGAT-3') [bib_ref] Distribution of bifidobacterial species in human intestinal microflora examined with 16S rRNA-genetargeted..., Matsuki [/bib_ref] and Ba. ovatus (g-Bfra-f: 5'-ATAGCCTTTCGAAAGRAAGAT-3' and g-Bfra-r: 5'-CCAGTATCAACTGCAATTTTA-3') [bib_ref] Development of 16S rRNA-gene-targeted group-specific primers for the detection and identification of..., Matsuki [/bib_ref] , respectively.
Co-culture time-series experiment between B. pseudocatenulatum and B. longum subsp. longum B. pseudocatenulatum YIT 11952 and B. longum subsp. longum H11-1 (isolated from an infant) were separately grown before co-culture. Cells harvested at the mid-to late-log phase were resuspended in modified PY medium adjusted to an OD 600 of 0.2. Equal amounts (2% vol/vol) of each suspension were inoculated into 500 μL of PY-AX medium. After 0, 8, 24 and 48 h of anaerobic co-culture, DNA was extracted from subsamples (50 μL) for quantitative real-time PCR using specific primers for B. pseudocatenulatum (BiCATg-1 and BiCATg-2) and B. longum subsp. longum (BiLON-1: 5'-TTCCAGTTGATCGCATGGTC-3' and BiLON-2: 5'-GGGAAGCCGTATCTCTACGA-3') [bib_ref] Distribution of bifidobacterial species in human intestinal microflora examined with 16S rRNA-genetargeted..., Matsuki [/bib_ref] using the ABI PRISM 7500 PCR system to determine cell numbers.
Y. Watanabe et al.
## Cereal intervention
Experimental design. This study was approved by the ethical committee of the Yakult Central Institute, in accordance with the committee guidelines and the Declaration of Helsinki (2013). This study is registered at UMIN Clinical Trials Registry (number UMIN 000043680). Written informed consent was obtained from 30 Japanese adult participants to participate in this study. Three participants who were prescribed with medications during the experimental period were excluded from data analysis. The experiment was completed by 27 (19 males and 8 females) participants aged 28-65 years who consumed 30 g of All Bran Original wheat bran-rich cereal (Kellogg Japan, Tokyo, Japan) with 180 mL of Accadi lactose-digested milk (Megmilk Snow Brand, Tokyo, Japan) once in the morning and once in the afternoon daily. According to the manufacturer, 60 g of cereal provided approximately 6.6 g of wheat bran arabinoxylan per day. Because almost the entire Japanese population has the genotype for low lactase activity [bib_ref] Association between functional lactase variants and a high abundance of Bifidobacterium in..., Kato [/bib_ref] , we used 80% lactose-depleted milk according to the manufacturer to minimise the effect of lactose on the growth of Bifidobacterium independently of cereal consumption. The experiment comprised preintervention (days 1-7), intervention (days [bib_ref] A key genetic factor for fucosyllactose utilization affects infant gut microbiota development, Matsuki [/bib_ref] [bib_ref] Distribution of bifidobacterial species in human intestinal microflora examined with 16S rRNA-genetargeted..., Matsuki [/bib_ref] [bib_ref] Genomic encyclopedia of type strains of the genus Bifidobacterium, Milani [/bib_ref] [bib_ref] Genomics of the genus Bifidobacterium reveals species-specific adaptation to the glycan-rich gut..., Milani [/bib_ref] [bib_ref] The genome sequence of Bifidobacterium longum subsp. infantis reveals adaptations for milk..., Sela [/bib_ref] [bib_ref] Insights from genomes of representatives of the human gut commensal Bifidobacterium bifidum, Duranti [/bib_ref] and post-intervention (days 15-21) periods. Throughout the 21 days of the experiment, the participants were instructed to refrain from consuming fermented milk containing bifidobacteria, prebiotic products and foods rich in LCX, such as other bran-rich cereal, whole grain or brown rice.
Faecal samples were collected on days 4, 7, 11, 14, 18 and 21. To reduce daily individual variations, we averaged the results of individual samples from each period. On days without defecation, faecal samples were collected the next day. If no defecation occurred on days 7, 14 or 21, the previous period was extended until a faecal sample was collected. None of the participants skipped the regime for more than 2 days. Faecal samples collected immediately after defecation were placed in sterile tubes, kept on ice and then stored at −80°C. Thereafter, DNA was extracted from ten-fold diluted faecal samples using the beads-phenol method as described above.
Quantitation of B. pseudocatenulatum in faecal samples. B. pseudocatenulatum cells were quantified using real-time PCR with a BiCATg primer that targets the 16S rRNA gene. This primer set has been confirmed to be specific for B. pseudocatenulatum and B. catenulatum and it does not react non-specifically with other bifidobacteria [bib_ref] Distribution of bifidobacterial species in human intestinal microflora examined with 16S rRNA-genetargeted..., Matsuki [/bib_ref]. Although this primer set detects both B. pseudocatenulatum and B. catenulatum [bib_ref] Distribution of bifidobacterial species in human intestinal microflora examined with 16S rRNA-genetargeted..., Matsuki [/bib_ref] , we regarded the measured value as that of only B. pseudocatenulatum because none of the participants had amplicon sequence variants (ASVs) close to B. catenulatum in the 16S rRNA gene amplicon analysis described below. The PCR mixture (20 μL total volume) contained 1 × TB Green Premix Ex Taq II (Takara Bio, Otsu, Japan), 0.2 μM of each primer and 2 μL of template DNA. The thermocycling conditions used were 94°C for 10 s, followed by 40 cycles of 94°C for 20 s, 55°C for 20 s and 72°C for 50 s, then a meltingcurve programme. Amplification was performed using the ABI PRISM 7500 Real-Time PCR System. For absolute quantification, a standard curve was calculated using a ten-fold dilution series of DNA extracted from the B. pseudocatenulatum strain with BpXyn10A, for which cell numbers were determined. The detection limit for this system was 5.0 × 10 4 cells/g faeces.
Quantification of cells with BpXyn10A in faecal samples. The numbers of cells with BpXyn10A were quantified by real-time PCR with BpXyn10A genetargeted oligonucleotide primers. To design specific primers, we used the sequences of all BpXyn10A genes and their homologues obtained from an NCBI blastn search, aligned them, and detected the specific region of the BpXyn10A gene. Then, we designed pBpXyn10A-F (5'-CGAGAATGC GAACACGTACTTC-3') and pBpXyn10A-R (5'-CTGCTCGGTGTTGTAATCGTT G-3'), which provided a 94 bp amplicon. Using Primer-BLAST [bib_ref] Primer-BLAST: a tool to design target-specific primers for polymerase chain reaction, Ye [/bib_ref] , we confirmed that there were no non-specific sequences that could be amplified in the sequences registered in the NCBI nr database. Furthermore, we performed PCR using DNA from the B. pseudocatenulatum 35 strains used in our study, and confirmed that specific amplification products were obtained only from strains with the BpXyn10A gene. The PCR mixture (20 μL total volume) contained 1 × TB Green Premix Ex Taq II (Takara Bio, Otsu, Japan), 0.4 μM of each primer and 2 μL of template DNA. The thermocycling conditions used were 95°C for 30 s, followed by 40 cycles of 95°C for 5 s and 60°C for 34 s, and a melting-curve programme. Amplification was performed using the ABI PRISM 7500 Real-Time PCR System.
16S rRNA gene amplicon analysis. The V1-2 regions of the 16S rRNA gene were amplified. Primer sequence, PCR conditions and details on library preparation are described in the Supplementary Methods. Pooled amplicons were sequenced using the MiSeq platform with the MiSeq reagent kit v2 500 cycle. Obtained reads were analysed using the QIIME2-2020.8 platform [bib_ref] Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2, Bolyen [/bib_ref]. Briefly, the sequence data were denoised using DADA2, and the resulting ASVs were assigned taxonomy in QIIME2 using the SILVA138 database. Alpha-diversity (observed_features, faith_pd and Shannon_index) and beta-diversity (unweighted UniFrac) were analysed using QIIME2. Homology of the ASVs was also analysed using vsearch with the 16S RefSeq nucleotide sequence records to determine the closest species. To analyse species level trends, we grouped ASVs with >97% homology to closely related species as a single species. No ASVs were more closely related to B. catenulatum than to B. pseudocatenulatum.
## Statistics
Data were statistically analysed using GraphPad Prism 7.0 (GraphPad Software, San Diego, CA, USA). Average CAZyme gene copy numbers of each species in the genome, amounts of organic acids produced in PY-XY medium and numbers of cells with and without BpXyn10A in the coculture experiments were compared using Mann-Whitney U tests. The effects of the presence or absence of Bacteroides species on the growth of B. pseudocatenulatum in co-culture experiments, and the increase or decrease of each bacterial species in the cereal intervention study were verified using Wilcoxon signed-rank tests.
# Results
## Composition of b. pseudocatenulatum cazymes
To understand the characteristics of carbohydrate utilisation of B. pseudocatenulatum within the genus Bifidobacterium, we determined draft or complete genomes of 35 isolates of this species . We then integrated these genomes with 451 publicly available genomes of the major HGM bifidobacteria and compared repertoires of CAZyme genes [fig_ref] Figure 1: Comparative analysis of CAZyme composition among human-resident bifidobacteria [/fig_ref] and . We calculated the average number of gene copies for each species and identified ten CAZyme genes that were characteristically abundant in B. pseudocatenulatum [fig_ref] Figure 1: Comparative analysis of CAZyme composition among human-resident bifidobacteria [/fig_ref]. Five of them contained enzymes that were involved in bacterial xylan degradation [bib_ref] Diversity in xylan-degrading prokaryotes and xylanolytic enzymes and their bioprospects, Verma [/bib_ref]. Among these, the following three GH families contained enzymes that degrade the xylose backbone of XOS in bifidobacteria: GH43 (xylosidase) [bib_ref] Multiple transporters and glycoside hydrolases are involved in arabinoxylanderived oligosaccharide utilization in..., Saito [/bib_ref] , GH8 (exooligoxylanase) [bib_ref] Recombinant expression and characterization of a reducing-end xylosereleasing exo-oligoxylanase from Bifidobacterium adolescentis, Lagaert [/bib_ref] and GH120 (xylosidase) [bib_ref] Characterization of two β-xylosidases from Bifidobacterium adolescentis and their contribution to the..., Lagaert [/bib_ref]. The two remaining GH families (GH10 and GH146) have not been verified in bifidobacteria; however, GH10 reportedly contains endo-1,4-βxylanase, which is a key enzyme for LCX degradation in HGM xylanolytic species [bib_ref] Molecular and biotechnological aspects of xylanases, Kulkarni [/bib_ref]. Therefore, the availability of oligosaccharides derived from LCX and LCX assimilation might be characteristic phenotypes of B. pseudocatenulatum. Conversely, genes encoding CAZymes for HMO utilisation [bib_ref] Bifidobacterial enzymes involved in the metabolism of human milk oligosaccharides, Kitaoka [/bib_ref] such as GH29 and GH95 (fucosidase), GH33 (sialidase) and GH112 (galacto-N-biose/ lacto-N-biose I phosphorylase) were detected in only a few B. pseudocatenulatum strains. These data suggest that B. pseudocatenulatum has adapted to the environment of the adult human gut, especially those with a diet rich in LCX.
## Lcx-related carbohydrate utilisation by b. pseudocatenulatum
We examined the availability of LCX-related carbohydrates in 35 isolates and the type strain of B. pseudocatenulatum. All 36 strains grew on a mixture of short-chain XOS (degree of polymerisation, 2-4), indicating that XOS availability is a universal feature of this species . Of the 36 strains, 12 could utilise arabinoxylan, a representative LCX . Furthermore, the strains that utilised arabinoxylan grew on xylan . Although the turbidity of several other strains gradually increased, only these 12 strains produced substantial amounts of organic acids , indicating that xylan availability was clearly distinguished between these 12 and other strains. The availability of LCX was completely consistent with the presence of the GH10 gene . The GH10 proteins detected in our isolates had 99-100% amino acid sequence similarity among each other. To confirm the enzymatic activity, we prepared a recombinant GH10 domain protein and mixed it with three types of LCX. Because of the accumulation of short-chain oligosaccharides , and because all known enzymatic activities in the GH10 family are endo-type (http://www. cazy.org/GH10.html), the enzyme was identified as endo-1,4-βxylanase, which we named BpXyn10A. This enzyme contains a GH10 catalytic domain, a carbohydrate-binding module family 9 (CBM9) domain and transmembrane regions . Since transmembrane regions were detected, BpXyn10A was expected to be associated with the membrane, and localise to the cell surface, similar to other HGM xylanolytic species [bib_ref] Glycan complexity dictates microbial resource allocation in the large intestine, Rogowski [/bib_ref] [bib_ref] Differential bacterial capture and transport preferences facilitate co-growth on dietary xylan in..., Leth [/bib_ref]. However, enzyme activity was detected not only on the cell surface but also in the filtered culture supernatant . Adding purified recombinant BpXyn10A allowed a non-LCX-utilising strain to grow on LCX . Therefore, LCX availability is determined by the BpXyn10A gene in the B. pseudocatenulatum genome.
## Genetic background of lcx utilisation
We analysed the genomic locus of the BpXyn10A gene to clarify the genetic background of the LCX-utilising strain. The BpXyn10A gene in all LCX-utilising B. pseudocatenulatum strains was located in the arabinoxylan-hydrolysate (AXH) utilisation gene cluster II, which is one of the gene clusters the expression of which is induced by AXH [bib_ref] Multiple transporters and glycoside hydrolases are involved in arabinoxylanderived oligosaccharide utilization in..., Saito [/bib_ref]. The BpXyn10A gene was adjacent to the transposase gene in all LCX-utilising strains. These genes were not detected in any other genomic regions of B. pseudocatenulatum. Considering that the horizontal transfer of a single transposase gene together with an adjacent gene between bifidobacteria strains in vitro has been confirmed [bib_ref] Comparative analysis of sequences flanking tet(W) resistance genes in multiple species of..., Kazimierczak [/bib_ref] , the BpXyn10A gene might have been horizontally transferred along with a transposase from another taxon. A homology search of public databases showed that the BpXyn10A was more homologous to proteins in other species of bifidobacteria (maximum, 86.41%) and non-HGM Actinobacteria (maximum, 47.55%) than those detected in Bacteroidetes (maximum, 32.95%) and Firmicutes (maximum, 35.29%) (Supplementary . These findings indicated that the origin of BpXyn10A is within the phylum Actinobacteria, but not Bacteroidetes or Firmicutes. However, no genomes with both BpXyn10A and transposase genes were found in the database; thus, the acquisition source of these genes remains unclear. In phylogenetic tree analysis, strains with BpXyn10A did not form a single cluster but rather were spread out , suggesting that the BpXyn10A gene was not acquired in a single horizontal transfer event in the past, but may have been acquired at multiple occasions from different genetic sources.
## Expression control of bpxyn10a
In order to assess possible transcriptional regulation of the BpXyn10A gene, we performed RNA-seq on the LCX-utilising strain cultured with LCX-associated carbohydrates. The expression of BpXyn10A and its surrounding genes was upregulated in the presence of xylose, XOS, xylan and arabinoxylan in B. pseudocatenulatum [fig_ref] Figure 4: The LCX utilisation system of B [/fig_ref] and . A common feature among known HGM xylanolytic bacteria is that intact LCX induces the expression of endo-1,4-β-xylanase [bib_ref] Glycan complexity dictates microbial resource allocation in the large intestine, Rogowski [/bib_ref] [bib_ref] Differential bacterial capture and transport preferences facilitate co-growth on dietary xylan in..., Leth [/bib_ref]. The unique feature of B. pseudocatenulatum was that xylose, a monosaccharide, also functions as an induction substrate. The expression induced by LCX-associated carbohydrates was similar to that of the AXH utilisation gene cluster II [fig_ref] Figure 4: The LCX utilisation system of B [/fig_ref] , indicating that this cluster acquired the BpXyn10A gene without disrupting its original gene regulation. Furthermore, two other AXH utilisation gene clusters were upregulated by LCX [fig_ref] Figure 1: Comparative analysis of CAZyme composition among human-resident bifidobacteria [/fig_ref]. Based on the present and previous findings [bib_ref] Multiple transporters and glycoside hydrolases are involved in arabinoxylanderived oligosaccharide utilization in..., Saito [/bib_ref] , we propose a model for LCX utilisation by B. pseudocatenulatum, in which BpXyn10A cleaves the xylan backbone and then, resulting oligosaccharides are taken up and further degraded by the AXH utilisation system [fig_ref] Figure 4: The LCX utilisation system of B [/fig_ref].
Fig. 2 LCX utilisation by the strains that possess a GH10 (BpXyn10A) gene. Growth curves of our isolates and the type strain in the medium supplemented with XOS (a), arabinoxylan (b) or xylan (c) as the only carbohydrate source. In panel c, total organic acid production in the supernatant is also shown. ***P value of Mann-Whitney U test <0.001. d The association between the presence/absence of predicted GH10 gene and growth phenotype. + indicates an exponential increase in turbidity of >0.1. e Endo-xylanase activity of purified recombinant GH10 analysed using thin-layer chromatography. A sample was collected over time after mixing each substrate, and the enzyme was applied. X1: xylose. X2: xylobiose. X3: xylotriose. X4: xylotetraose. X5: xylopentaose. X6: xylohexaose. f Domain organisation of BpXyn10A identified using dbCAN2 [bib_ref] dbCAN: a web resource for automated carbohydrate-active enzyme annotation, Yin [/bib_ref] and the PSORT web application (http://psort.hgc.jp/form.html); GH10, catalytic module of glycoside hydrolase 10; CBM9, carbohydrate-binding module 9; TM, transmembrane region. g Localisation of endo-1,4-β-xylanase activity of the LCX-utilising and non-utilising strains. h Growth of the non-LCX-utilising strain (YIT 4072 T ) in the arabinoxylan medium supplemented with multiple concentrations of purified recombinant BpXyn10A.
## Fig. 3
Genetic locus and strain distribution of BpXyn10A. a The gene arrangement around the BpXyn10A gene. Nucleotide identity of each gene was visualised using GenomeMatcher software [bib_ref] GenomeMatcher: a graphical user interface for DNA sequence comparison, Ohtsubo [/bib_ref]. Gene annotation details are also shown in [fig_ref] Figure 4: The LCX utilisation system of B [/fig_ref] using the YIT 11952 as an example. b A phylogenetic tree based on the alignment of core genes in 35 isolates and the type strain. Strains with the BpXyn10A gene are shown in blue. Local support values at the branch nodes were computed using the Shimodaira-Hasegawa test with the default parameter settings of fasttree [bib_ref] FastTree: computing large minimum evolution trees with profiles instead of a distance..., Price [/bib_ref].
## Impact of bpxyn10a on interspecies interaction
To assess the effect of BpXyn10A on interspecies interactions, we performed co-culture experiments on medium supplemented with arabinoxylan as the only carbohydrate source. Similar to previous findings of non-LCX-utilising Bifidobacterium species [bib_ref] Glycan complexity dictates microbial resource allocation in the large intestine, Rogowski [/bib_ref] , the growth of strains without BpXyn10A was stimulated by coculture with Bacteroides ovatus, an established HGM xylanolytic species [bib_ref] Glycan complexity dictates microbial resource allocation in the large intestine, Rogowski [/bib_ref] , whereas that of strains with BpXyn10A was not [fig_ref] Figure 5: Co-culture of B [/fig_ref]. The cell numbers of strains with BpXyn10A in co-cultures were higher than those of the strains without BpXyn10A [fig_ref] Figure 5: Co-culture of B [/fig_ref] , indicating that BpXyn10A provided an ecological advantage even in environments where coexisting bacteria supplied breakdown products. In contrast to B. pseudocatenulatum, the numbers of Ba. ovatus cells decreased when co-cultured with B. pseudocatenulatum strains with BpXyn10A rather than without this enzyme [fig_ref] Figure 5: Co-culture of B [/fig_ref]. Furthermore, adding purified recombinant BpXyn10A reduced the numbers of Ba. ovatus cells . Because the expression of xylanase in Bacteroidetes spp. is induced by long-chain XOS but not by short-chain XOS [bib_ref] Medium-to large-sized xylo-oligosaccharides are responsible for xylanase induction in Prevotella bryantii B14, Miyazaki [/bib_ref] , the decreased number of Ba. ovatus can be explained by a reduced degree of xylose backbone polymerisation by BpXyn10A, resulting in a decreased expression of the xylanase. We then examined whether a strain with BpXyn10A produces a growth substrate for other non-LCX-utilising bifidobacteria. The results of co-culture on arabinoxylan showed that the strain with BpXyn10A promoted the growth of B. longum subsp. longum strain H11-1, which utilised XOS but not arabinoxylan [fig_ref] Figure 5: Co-culture of B [/fig_ref]. Furthermore, we confirmed that the LCX-utilising strain released short-chain oligosaccharides from LCX into the culture supernatant . These results indicate that strains with BpXyn10A promote the growth of other secondary LCX consumers by releasing breakdown products. Therefore, the presence of the BpXyn10A gene in the genome defined whether the B. pseudocatenulatum strain would act as a keystone species or a recipient of LCX-derived oligosaccharides, and potentially affected the growth of other direct and indirect LCX-degrading microorganisms in the HGM.
Ecological role of BpXyn10A in the human intestine Based on the findings of our co-culture experiments, we speculated that supplementation with LCX would increase the population of B. pseudocatenulatum cells with BpXyn10A in the human intestine. B. pseudocatenulatum has been detected in most Heatmap depicts log 2 fold changes of the expression on xylose (X), XOS, xylan (XY) and arabinoxylan (AX) relative to lactose. Annotation information is from Prokka [bib_ref] Prokka: rapid prokaryotic genome annotation, Seemann [/bib_ref]. Asterisks indicate functional proteins that were not annotated in Prokka, but were found to have >95% amino acid homology by Blastp searches against the NCBI nr database. Locus ID YIT11952_xxxxx are abbreviated with the last numbers after the underscore. b LCX utilisation model using arabinoxylan as an example. Initial degradation of LCX occurs by both cell-anchored and extracellularly released BpXyn10A. Degradation products were imported to the cytoplasm and further degraded by the AXH utilisation system as described previously [bib_ref] Multiple transporters and glycoside hydrolases are involved in arabinoxylanderived oligosaccharide utilization in..., Saito [/bib_ref]. Imported degradation products induce the expression of the BpXyn10A and AXH utilisation system.
Japanese adults [bib_ref] Distribution of bifidobacterial species in human intestinal microflora examined with 16S rRNA-genetargeted..., Matsuki [/bib_ref] [bib_ref] The gut microbiome of healthy Japanese and its microbial and functional uniqueness, Nishijima [/bib_ref] , and about one-third of our isolates from Japanese adults possesses BpXyn10A . Therefore, we assumed that some Japanese individuals harbour B. pseudocatenulatum cells with BpXyn10A and others harbour those without BpXyn10A. Under this assumption, we provided Japanese adults with commercially available LCX-rich cereal food made mainly from wheat bran, then compared the behaviour of B. pseudocatenulatum cells using qPCR [fig_ref] Figure 6: Effect of the presence or absence of the BpXyn10A gene in the... [/fig_ref]. As expected, 10 out of 27 participants (37%) possess B. pseudocatenulatum cells with BpXyn10A (BpXyn10A+ group), 10 (37%) possess B. pseudocatenulatum cells without it (BpXyn10A-group) and 7 (26%) possess no detectable B. pseudocatenulatum (Bp-group) [fig_ref] Figure 6: Effect of the presence or absence of the BpXyn10A gene in the... [/fig_ref]. The number of cells with BpXyn10A in the BpXyn10A+ group ranged from 1 × 10 5.9 to 1 × 10 9.7 /g faeces [fig_ref] Figure 6: Effect of the presence or absence of the BpXyn10A gene in the... [/fig_ref]. The number of cells with BpXyn10A increased during the intervention in all participants [fig_ref] Figure 6: Effect of the presence or absence of the BpXyn10A gene in the... [/fig_ref]. Furthermore, the total number of B. pseudocatenulatum cells (regardless of the presence and absence of BpXyn10A) significantly increased during the intervention in the BpXyn10A+, but not in the BpXyn10Agroup [fig_ref] Figure 6: Effect of the presence or absence of the BpXyn10A gene in the... [/fig_ref]. These findings indicate that BpXyn10A defines the growth potential of B. pseudocatenulatum in LCX-rich intestinal environments. We then applied 16S rRNA gene amplicon analysis to determine the impact of BpXyn10A on the composition of HGM. Alpha-and beta-diversity significantly differed between the three groups [fig_ref] Figure 4: The LCX utilisation system of B [/fig_ref]. In particular, the BpXyn10A+ and BpXyn10Agroups differed not only during, but also before the intervention, suggesting that BpXyn10A affected the HGM composition even at the level of LCX supplied in a regular daily diet. The responses of the Bifidobacterium genus to the intervention obviously differed among the three groups, with an increase of relative abundance only in the BpXyn10A+ group [fig_ref] Figure 4: The LCX utilisation system of B [/fig_ref]. This mainly resulted from the increase of B. pseudocatenulatum, and no changes were evident in other bifidobacterial species such as B. longum or B. adolescentis [fig_ref] Figure 5: Co-culture of B [/fig_ref]. The relative abundance of the representative LCX-utilising genera Bacteroides and Roseburia in the HGM did not significantly increase in either the BpXyn10A+ or BpXyn10Agroups [fig_ref] Figure 5: Co-culture of B [/fig_ref]. In summary, the ability of BpXyn10A to promote or inhibit the growth of other bacterial species, such as those found in co-cultures in vitro, was limited in the complex human gut environment. Nevertheless, the results of the 16S rRNA gene amplicon analysis showed that BpXyn10A affected the growth potential of B. pseudocatenulatum.
# Discussion
This study showed that LCX availability is essential to promote adaptation of Bifidobacterium pseudocatenulatum in the adult human gut. In infants, HMO availability is known to be a key colonisation factor for bifidobacteria [bib_ref] A key genetic factor for fucosyllactose utilization affects infant gut microbiota development, Matsuki [/bib_ref] , and in adults, LCX availability contributes to niche expansion of the representative adult-dominant species B. pseudocatenulatum. This finding extends our understanding of the adaptive strategy of bifidobacteria to the human gut environment. In addition, our findings provide new insights into the food web of LCX degradation in HGM.
We showed that bifidobacteria, which are considered secondary LCX consumers, behave as primary degraders if they possess the BpXyn10A gene. To the best of our knowledge, the present study is the first to show that HGM organisms other than Bacteroidetes and Firmicutes possess key enzymes for LCX degradation. This leads to the question as to whether LCX metabolism is a characteristic only of B. pseudocatenulatum among bifidobacteria. At present, B. pseudocatenulatum is the only species expressing biochemically confirmed endo-1,4-β-xylanase in the Bifidobacterium genus. Conversely, a BpXyn10A homologue was also identified in a few strains of other species of HGM Bifidobacterium (Supplementary . These findings suggested that the availability of LCX by HGM Bifidobacterium may have been overlooked in previous studies.
The BpXyn10A gene was integrated into the AXH utilisation gene cluster II (Figs. 3a and 4a), which contains genes related to the uptake and further degradation of oligosaccharides produced by BpXyn10A, indicating that AXH utilisation gene cluster II has evolved into an LCXutilisation gene cluster. Because the LCX-utilisation gene cluster was induced by xylose or XOS [fig_ref] Figure 4: The LCX utilisation system of B [/fig_ref] , horizontal transfer of the BpXyn10A gene generated an LCX-utilisation gene cluster with the regulatory mechanisms for secondary consumers. Because xylose and XOS are products of BpXyn10A enzymatic activity and , the LCX-utilisation gene cluster of B. pseudocatenulatum can be regarded as having a positive feedback regulation mechanism. This is a feature that is not found in other known LCX-utilisation gene clusters of xylanolytic HGM species, in which intact LCX is the inducible substrate but not low-molecule carbohydrates such as xylose [bib_ref] Differential bacterial capture and transport preferences facilitate co-growth on dietary xylan in..., Leth [/bib_ref] [bib_ref] Xylan degradation by the human gut Bacteroides xylanisolvens XB1AT involves two distinct..., Despres [/bib_ref] [bib_ref] Xylan utilization in human gut commensal bacteria is orchestrated by unique modular..., Zhang [/bib_ref]. Having a positive feedback regulation mechanism might explain the obvious increase in B. pseudocatenulatum cells with BpXyn10A during the LCX-rich dietary intervention, even though other representative xylanolytic genera in HGM did not significantly increase.
We showed that BpXyn10A contributes to increasing numbers of B. pseudocatenulatum cells in vitro [fig_ref] Figure 5: Co-culture of B [/fig_ref] and in the human intestine [fig_ref] Figure 6: Effect of the presence or absence of the BpXyn10A gene in the... [/fig_ref]. A previous study found that LCX availability is a core fitness determinant in the Bacteroides species in mice fed with a high-plant polysaccharide or purified arabinoxylan diet [bib_ref] Genetic determinants of in vivo fitness and diet responsiveness in multiple human..., Wu [/bib_ref]. These findings indicate that LCX availability is a pivotal feature common to different bacterial species among various host gut environments. Therefore, it is reasonable to assume that there is positive selection pressure for BpXyn10A and that strains with this enzyme will become predominant among the isolates. Nevertheless, strains with BpXyn10A accounts for only~33% of our isolates. This discrepancy might be related to the release of BpXyn10A into the extracellular milieu, where LCX degradation products can be shared by cells producing the enzyme and secondary consumers. Therefore, when an intestinal environment includes cells with BpXyn10A, negative selection pressure on cells without BpXyn10A might be reduced. Compared with the membrane-bound endo-1,4-β-xylanase of other HGM xylanolytic species, free BpXyn10A is more altruistic, especially for secondary LCX consumers.
Although intact LCX has not been considered a substrate for bifidobacteria in the human intestine [bib_ref] Prebiotic and other health-related effects of cereal-derived arabinoxylans, arabinoxylan-oligosaccharides, and xylooligosaccharides, Broekaert [/bib_ref] , we found that LCX could be a substrate for the growth of bifidobacteria in the gut of people with detectable BpXyn10A. Furthermore, during cereal intervention, the number of B. pseudocatenulatum cells with BpXyn10A increased in the entire BpXyn10A+ group [fig_ref] Figure 6: Effect of the presence or absence of the BpXyn10A gene in the... [/fig_ref]. These findings suggest that the BpXyn10A in faecal samples serve as an accurate biomarker to predict responders, in whom the cell numbers of B. pseudocatenulatum would increase under LCX-rich food intervention. Non-responders might also become responders by consuming live B. pseudocatenulatum with BpXyn10A. These possibilities provide a framework that could guide intestinal bifidobacteria recalibration to gain health benefits. In particular, the selective growth of B. pseudocatenulatum may contribute to the alleviation of type 2 diabetes [bib_ref] Gut bacteria selectively promoted by dietary fibers alleviate type 2 diabetes, Zhao [/bib_ref]. Our study paves the way for systematic manipulation of the intestinal microbiota through dietary intervention using key polysaccharide degradative enzymes as biomarkers.
## Data availability
Sequence data that support the findings of this study have been deposited at GenBank and Sequence Read Archive under BioProject accession no. PRJNA745059. The authors declare that all other data supporting the findings of this study are available within the article and its supplementary information files, or from the corresponding author upon request.
[fig] Figure 1: Comparative analysis of CAZyme composition among human-resident bifidobacteria. The heatmap shows copy numbers of the CAZyme gene per genome of 486 strains belonging to B. pseudocatenulatum (102 strains), B. adolescentis (49 strains), B. longum subspecies longum (151 strains), B. breve (99 strains) and B. bifidum (85 strains). Each column represents a strain, and each row represents a CAZyme gene detected from at least one strain in this data set. CAZymes marked in bold and underlined represent xylan degradation-related CAZymes. Black circles indicate the gene with the highest average copy number in B. pseudocatenulatum, among other species (Mann-Whitney U test; P < 0.01). Five of these CAZymes were also shown in separated plots with mean ± standard error. SeeSupplementary Table S2for the actual copy number of each CAZyme gene.Y. Watanabe et al. [/fig]
[fig] Figure 4: The LCX utilisation system of B. pseudocatenulatum. a Expression profiles of genes around the BpXyn10A gene from strain YIT 11952. [/fig]
[fig] Figure 5: Co-culture of B. pseudocatenulatum and other HGM species in medium with arabinoxylan. a Cell numbers of mono-and co-cultures of B. pseudocatenulatum strains with or without the BpXyn10A gene. P a : P value of Wilcoxon signed-rank test, P b : P value of Mann-Whitney U test. b Cell numbers of Ba. ovatus co-cultured with B. pseudocatenulatum strains with or without the BpXyn10A gene. P b : P value of Mann-Whitney U test. c Cell number of B. longum subsp. longum co-cultured with B. pseudocatenulatum strain with (YIT 11952) or without (YIT 4072 T ) the BpXyn10A gene. Data are expressed as mean of quadruplicate experiments ± standard deviation. [/fig]
[fig] Figure 6: Effect of the presence or absence of the BpXyn10A gene in the human intestine on the population of B. pseudocatenulatum and genus Bifidobacterium during LCX-rich cereal intervention. a Study design. Participants consumed a 60 g bowl of LCX-rich food per day (see Methods). b B. pseudocatenulatum cells with BpXyn10A and total B. pseudocatenulatum, and relative abundance of genus Bifidobacterium on the participants grouped based on the presence or absence of B. pseudocatenulatum and BpXyn10A gene. Cell numbers were determined using the quantitative value of qPCR targeting the BpXyn10A and 16S rRNA genes of B. pseudocatenulatum. Relative abundance was obtained from 16S rRNA gene amplicon analysis. **P value of Wilcoxon signed-rank test <0.01; *<0.05. n.s.: not statistically significant. c PERMANOVA exploring the differences in microbial composition between groups based on unweighted Unifrac distance. d Comparison of Shannon index between groups. P values of Kruskal-Wallis test are shown. [/fig]
|
10.3390/ijerph17176228
|
CCBY
|
7504218
|
32867209
|
s2orc_pubmed_articles
|
Role of Meteorological Parameters in the Diurnal and Seasonal Variation of NO2 in a Romanian Urban Environment
The main purpose of this study was to investigate whether meteorological parameters (temperature, relative humidity, direct radiation) play an important role in modifying the NO 2 concentration in an urban environment. The diurnal and seasonal variation recorded at a NO 2 traffic station was analyzed, based on data collected in situ in a Romanian city, Braila (45.26 - N,27.95 - E), during 2009-2014. The NO 2 atmospheric content close to the ground had, in general, a summer minimum and a late autumn/winter maximum for most years. Two diurnal peaks were observed, regardless of the season, which were more evident during cold months. Traffic is an important contributor to the NO 2 atmospheric pollution during daytime hours. The variability of in situ measurements of NO 2 concentration compared relatively well with space-based observations of the NO 2 vertical column by the Ozone Monitoring Instrument (OMI) satellite for most of the period under scrutiny. Data for daytime and nighttime (when the traffic is reduced) were analyzed separately, in the attempt to isolate meteorological effects. Meteorological parameters are not fully independent and we used partial correlation analysis to check whether the relationships with one parameter may be induced by another. The correlation between NO 2 and temperature was not coherent. Relative humidity and solar radiation seemed to play a role in shaping the NO 2 concentration, regardless of the time of day, and these relationships were only partially interconnected.Int. J. Environ. Res. Public Health 2020, 17, 6228 2 of 15 thunderstorms, through lightning, in the upper troposphere[3]. Other natural NO x sources include volcanic activity emissions, bacteria decay, soil processes and stratospheric sources [1], while the rest is the result of fossil fuel combustion processes: fires, transportation, industrial-such as power generation, nitric acid manufacture, welding processes, the use of explosives and iron and steel industry and refineries[1,2]. According to[4], about 65% of NO x emissions are of anthropogenic origin. The NO 2 reaction with the atmospheric hydroxyl radical (-OH), and the NO reaction with hydroperoxyl (HO 2 ), followed by the formation of nitric acid and its wet deposition are important NO sinks (almost 60%), according to[4]. The contribution of road traffic to nitrogen oxides (NO x ) emissions has been described by various studies[5][6][7][8][9][10][11]. Some authors assume that about half of the anthropogenic NO x comes from traffic[12]. Although several measures of emission control were implemented by various European Commission (EC) regulations, emissions of nitrogen oxides from road traffic have not significantly decreased across Europe; sometimes, NO 2 concentrations have even increased[10,11]. Some recent papers showed that the real reduction of nitrogen oxides and other pollutants following Euro 5 and 6a/b introduction was not as important as expected[10].The concentration of NO 2 (and, in general, of any atmospheric pollutants) in an urban area and its variation are influenced by the industrialization level, the number of inhabitants and traffic density, but also by the topography of the area and meteorological parameters[5,9,11]. It is expected that meteorological factors would affect both natural and anthropogenic emissions of NO 2 (e.g., the reduced solar radiation during wintertime will impact the rate of NO 2 photochemical reactions and the reduced temperatures during the cold season will increase NO 2 emissions associated with heating). Moreover, meteorological parameters are interlinked in various ways: the relative humidity decreases with increasing temperature, the effective and saturation air vapor pressures increase with atmospheric temperature, etc. Thus, it is difficult to assess the effect of each separate meteorological factor on NO 2 pollution. None of these influences were yet quantified in a reliable way and a clear relationship is far from being established. Studies of the link between meteorological factors and atmospheric pollutants, in general, or NO 2 , in particular, are rather scarce, inconsistent and strongly depend on local factors[5,[13][14][15][16][17][18][19]. According to[13], the NO 2 concentration is slightly higher at a lower relative humidity, whereas other authors found that the NO 2 concentration correlates positively with the relative humidity in all seasons, especially during winter[20]. The dependence between NO 2 and temperature was found to be weak, in general, except for two considerable positive correlations, which were obtained for July and December, and for which no particular explanation was found in[13]. Other studies found strong positive correlations between the NO 2 concentration and temperature for all seasons[20]. A negative correlation between wind speed and NO 2 concentration in all seasons except for summer was found by[19]. This paper presents a study of the diurnal, monthly, seasonal and interannual evolution of nitrogen dioxide concentration at an urban site in the southeast of Romania (Braila, 45.3 - N, 27.9 - E), which is influenced by road traffic mainly during the daytime. The role played by anthropogenic sources and the most relevant meteorological parameters that may affect the level of NO 2 pollution at the local scale is investigated. A comparison between in situ measurements of the NO 2 concentration near the ground and satellite measurements of NO 2 vertical columns is also performed. The timeframe for both in situ and satellite data is 2009-2014.
# Introduction
Monitoring atmospheric pollution and the possibility to predict its evolution are of high interest. One major atmospheric pollutant is nitrogen dioxide (NO 2 ), which may cause many health problems [bib_ref] SCIAMACHY tropospheric NO 2 over Switzerland: Estimates of NO x lifetimes and..., Schaube [/bib_ref]. The NO 2 gas has a reddish-brown color, is nonflammable, and has a detectable, pungent odor, perceptible from concentrations of approximately 190 µg/m 3. Nitrogen dioxide reacts with the hydroxyl radical (-OH) in the atmosphere, forming the highly-corrosive nitric acid, but it can also form toxic organic nitrates. Nitrogen oxides known as NO x (i.e., NO + NO 2 ) are involved in the formation of tropospheric ozone and smog, mediated by light through photolysis. Due to the significant environmental impact, the NO x compounds are strictly monitored and legislative limits were set at EU and national levels. Part of the NO 2 molecules in the atmosphere are of primary nature (i.e., directly emitted), while most NO 2 results from nitrogen monoxide (NO) [bib_ref] SCIAMACHY tropospheric NO 2 over Switzerland: Estimates of NO x lifetimes and..., Schaube [/bib_ref] [bib_ref] The global lightning-induced nitrogen oxides source, Schumann [/bib_ref]. The latter is produced via natural and anthropogenic processes. About one-fifth of NO x is released in the atmosphere during The number of vehicles in Braila has increased during the last few years, which led to a rather increased concentration of atmospheric pollutants. BR1 is a traffic station; it is located on Calea Galaţi No. 53 (The Agency of Environmental Protection Braila: http://apmbr.anpm.ro/) and monitors on a continuous basis the pollution levels generated mainly by traffic emissions, with medium and high flows, from the neighboring streets. Calea Galati Street is one of the busiest traffic streets in Braila. The width of the street is 16 m and it is covered with an asphalt carpet. The area around the air quality monitoring station has apartment blocks with four floors and some green spaces. The air quality monitoring station is located approximately 20 m from Calea Galati Street and 15 m from the apartment buildings. The surface is flat, characteristic of a plain area. The sources of pollution in the area consist mostly of road traffic and domestic heating.
The hourly atmospheric concentrations of NO 2 are determined in situ, by using the chemiluminescence technique [bib_ref] Modeling results of atmospheric dispersion of NO 2 in an urban area..., Dragomir [/bib_ref]. This method implies the reduction of NO 2 to NO in the presence of a Molybdenum surface, at a temperature of roughly 310 - C. The resulted NO reacts with ozone (O 3 ), leading to the formation of fluorescent NO 2 , whose emission is detected by a sensor [bib_ref] Key chemical NO x sink uncertainties and how they influence top-down emissions..., Stavrakou [/bib_ref]. Data between 2009 and 2014 were used in this study. Besides, concentrations of NO 2, meteorological parameters were also recorded.
Satellite measurements were provided by the Ozone Monitoring Instrument (OMI) onboard the Earth Observing System Aura satellite [bib_ref] The ozone monitoring instrument, Levelt [/bib_ref] [bib_ref] Algorithm for NO/sub 2/ vertical column retrieval from the ozone monitoring instrument, Bucsela [/bib_ref]. The OMI measures several important pollutants, such as O 3 , NO 2 , SO 2 , and aerosols, with a spatial resolution of 13 km × 24 km, i.e., at a near urban scale resolution, https://aura.gsfc.nasa.gov/omi.html [bib_ref] A Brief Tutorial on Using the Ozone Monitoring Instrument (OMI) Nitrogen Dioxide..., Duncan [/bib_ref]. OMI is a remote sensing space instrument onboard the Aura satellite that provides a raw NO 2 product called DSCD (Differential Slant Column Density), which is retrieved using Differential Optical Absorption Spectroscopy (DOAS) in the 405-465 nm range. Satellite measurements provide valuable data about atmospheric pollutants, including NO 2 [bib_ref] Ground-level nitrogen dioxide concentrations inferred from the satellite-borne Ozone Monitoring Instrument, Lamsal [/bib_ref] [bib_ref] Validation of urban NO 2 concentrations and their diurnal and seasonal variations..., Boersma [/bib_ref] [bib_ref] Validation of Ozone Monitoring Instrument NO2 measurements using ground based NO2 measurements..., Gruzdev [/bib_ref] [bib_ref] Satellite Observations of NO 2 Trend over, Constantin [/bib_ref] [bib_ref] Daily Ambient NO 2 Concentration Predictions Using Satellite Ozone Monitoring Instrument NO..., Lee [/bib_ref] [bib_ref] Aura OMI observations of regional SO 2 and NO 2 pollution changes..., Krotkov [/bib_ref] [bib_ref] Estimating daily surface NO 2 concentrations from satellite data -a case study..., Anand [/bib_ref] at a large scale. The most refined product of OMI is the tropospheric Vertical Column Density (VCD), which is the result of a near real-time retrieval algorithm that gives a 0.7 × 10 15 molec./cm 2 uncertainty for each individual pixel [bib_ref] Validation of urban NO 2 concentrations and their diurnal and seasonal variations..., Boersma [/bib_ref]. In our study, we have used level 3 data, which are a conversion of OMI 13 × 24 km 2 data to a 0.25 - × 0.25 - resolution [bib_ref] Aura NO2 Cloud-Screened Total and Tropospheric Column L3 Global Gridded 0.25 •..., Krotkov [/bib_ref]. Such data collected between 2005 and 2015 by OMI confirmed a reduction in NO 2 concentration over Northern China, Eastern Europe, and USA, while an increase in NO 2 concentration was detected over the Persian Gulf and India [bib_ref] Aura OMI observations of regional SO 2 and NO 2 pollution changes..., Krotkov [/bib_ref].
# Methods
Meteorological parameters recorded at an hourly rate were used, in order to verify to what extent they are relevant for modulating the NO x variability by means of correlation analysis. Correlations/anticorrelations (i.e., positive/negative correlation coefficients) between two variables are not the decisive proof for direct cause-effect relationships; however, they suggest that a link may exist if a physical mechanism supports it. When two variables are not correlated, the first thought is that there is no link between the two variables, which, however, may not always be true. Moreover, meteorological parameters (temperature, humidity, solar radiation) are not independent (e.g., solar radiation at the ground is a measure of the cloud cover, which is linked to relative humidity but also to temperature). Consequently, the correlation analysis was refined and a partial correlation analysis was used [bib_ref] Correlation between clouds at different altitudes and solar activity: Fact or Artifact?, Usoskin [/bib_ref] , complementary to the bivariate correlation. Such an analysis is useful when the relationship between two variables, e.g., X and Z, measured by the bivariate correlation coefficient, C XZ , may be induced or suppressed by a third (intervening) variable, e.g., Y, which affects both variables X and Z. The partial correlation P X(Y)Z corresponds to the link between X and Z when Y is constant. To be more specific, in this particular case the main variables, X and Z, were the NO 2 concentration and one meteorological factor (e.g., temperature) and the intervening variable, Y, was another meteorological parameter (e.g., radiation). The radiation (Y) may affect both the NO 2 concentration (X) and the temperature (Z). The difference D X(Y)Z = P X(Y)Z − C XZ between the partial correlation coefficient, P X(Y)Z , and the direct correlation coefficient, C XZ , measures the degree of intervention for Y. If P X(Y)Z is smaller than C XZ , i.e., the difference and the direct coefficient have opposite signs and C XZ ×D X(Y)Z < 0, then Y is responsible for part of the correlation, or it is said that Y "intervenes". If the two coefficients, C and D, have the same sign, i.e., C XZ × D X(Y)Z > 0, then the correlation is real and Y even suppresses, partially, the correlation. If the partial and direct coefficients are equal, i.e., D = 0, then the Y variable does not intervene. [fig_ref] Table 1: Type of correlation for various results of the partial correlation analysis [/fig_ref] summarizes possible results and interpretations for various combinations of direct and partial correlation coefficients. The correlation analysis has been applied to standardized data (by their standard deviation) in order to identify possible links between NO 2 concentration in the atmosphere and the variations of several meteorological parameters. . These values are smaller than those measured by traffic stations in Bucharest [bib_ref] Satellite Observations of NO 2 Trend over, Constantin [/bib_ref].
# Results and discussion
## Diurnal and seasonal variation
The NO 2 content varied with year; it was low in 2009 and in the first part of 2010, reached a maximum in the winter between 2010 and 2011, and then slightly decreased towards the end of the interval. The period is too short for assessing whether a trend does exist.
The seasonal variation of NO 2 was evident; NO 2 concentrations were clearly higher during winter than during summer. The seasonal variation may be explained by: (1) the strong variation of the anthropogenic sources contributing to NO 2 emissions, i.e., traffic and combustion, but also by (2) a longer lifetime of the NO 2 during winter, taking into account that the NO 2 lifetime and its concentration in the atmosphere are affected by the seasonal variation of the photochemical activity [bib_ref] Diurnal and seasonal variations of NO, NO 2 and PM 2.5 mass..., Kendrick [/bib_ref] [bib_ref] Modeling results of atmospheric dispersion of NO 2 in an urban area..., Dragomir [/bib_ref] [bib_ref] Validation of urban NO 2 concentrations and their diurnal and seasonal variations..., Boersma [/bib_ref] [bib_ref] Estimating daily surface NO 2 concentrations from satellite data -a case study..., Anand [/bib_ref] , which is reduced during winter. Additionally, during winter, the soil is cold, and thus the nearby air is heavy so that the emitted pollutants may stay close to the ground longer. Another explanation for the seasonal NO x variation relates to the variation of the boundary layer height with the season, which influences the wind pattern that, in turn, has a very important role in pollutant dispersion [bib_ref] Diurnal and seasonal variations of NO, NO 2 and PM 2.5 mass..., Kendrick [/bib_ref].
Two diurnal peaks of the NO 2 concentration could be observed, centered around 09:00 Local Time (LT) and 20:00 LT, which were more evident during cold months. The diurnal peaks are in agreement with previous findings [bib_ref] Modeling results of atmospheric dispersion of NO 2 in an urban area..., Dragomir [/bib_ref] and are associated with peaks in traffic [bib_ref] Diurnal and seasonal variations of NO, NO 2 and PM 2.5 mass..., Kendrick [/bib_ref]. Morning peaks, observed around 9:00, were smaller than the evening peaks. The NO 2 concentration increased during the afternoon-evening period, between 17:00 and 24:00, with maxima progressing towards earlier time during winter months. This was seen in all seasons except summer. In December, these diurnal maxima were smaller, for all years, which can be linked to the reduced traffic and industrial activity due to holidays. earlier time during winter months. This was seen in all seasons except summer. In December, these diurnal maxima were smaller, for all years, which can be linked to the reduced traffic and industrial activity due to holidays. [fig_ref] Figure 2: Int [/fig_ref] shows the variation of the tropospheric NO2 VCD (Vertical Column Density) over Braila, derived from OMI, together with in situ measurements at 11:00 LT, which is the approximate hour when the satellite overpasses the city. The tropospheric NO2 VCD varied between 0.8 × 10 15 molec./cm 2 for February 2011 and 4.7 × 10 15 molec./cm 2 for December 2012. Note that the comparison in [fig_ref] Figure 2: Int [/fig_ref] is strictly qualitative, since OMI measures the number of NO2 molecules in a column over a grid of 0.25° [bib_ref] Aura NO2 Cloud-Screened Total and Tropospheric Column L3 Global Gridded 0.25 •..., Krotkov [/bib_ref] , while in situ instruments measure the volumetric concentration near the ground; thus a direct quantitative comparison makes no sense. The annual NO2 VCD ranges between 1.73 and 2.23 × 10 15 molec./cm 2 . Previous studies have shown that OMI measurements give reliable results about NO2 emissions of anthropogenic sources at a large scale in cities [bib_ref] Validation of urban NO 2 concentrations and their diurnal and seasonal variations..., Boersma [/bib_ref] [bib_ref] Evolution of NO 2 in five major cities in Europe using remote..., Rosu [/bib_ref] , above large industrial platforms (located away from cities) [bib_ref] Ground-level nitrogen dioxide concentrations inferred from the satellite-borne Ozone Monitoring Instrument, Lamsal [/bib_ref] [bib_ref] NO 2 and SO 2 observations in SouthEast Europe using mobile DOAS..., Constantin [/bib_ref] and given by vehicle emissions and also by residential emissions [bib_ref] Verification of anthropogenic emissions of China by satellite and ground observations, Wang [/bib_ref]. However, other studies have shown that satellite observations do not correctly evaluate the NO2 pollution caused by traffic or by other point sources [bib_ref] Validation of Ozone Monitoring Instrument NO2 measurements using ground based NO2 measurements..., Gruzdev [/bib_ref] [bib_ref] Satellite Observations of NO 2 Trend over, Constantin [/bib_ref]. [fig_ref] Figure 2: Int [/fig_ref] shows that the two time series agreed relatively well except for a period between the winter of 2009-2010 and the summer of 2011. The correlation coefficient between the two series was close to 0.40 (after values in December 2010 were removed) and did not change when NO2 concentrations measured earlier, between 08:00 and 11:00 LT, were considered for the mean.
The seasonal variation of the tropospheric VCD was more regular: it was small during warm months and high during cold months. In 2010, the two time series did not coincide, which is mainly due to a peculiar behavior of the in situ measurements that showed an unexpected wide maximum during summer and autumn, and a minimum in December. The OMI instrument "sees" a large surface area (312 km 2 ), and thus integrates the emissions of many sources. Consequently, it cannot accurately measure the local variability, captured by the in situ measurements. This opposing variation suggests that a punctual, temporary, source of NO2 was added in the second part of 2010 close to the measuring station. [fig_ref] Figure 2: Int [/fig_ref] shows the variation of the tropospheric NO 2 VCD (Vertical Column Density) over Braila, derived from OMI, together with in situ measurements at 11:00 LT, which is the approximate hour when the satellite overpasses the city. The tropospheric NO 2 VCD varied between 0.8 × 10 15 molec./cm 2 for February 2011 and 4.7 × 10 15 molec./cm 2 for December 2012. Note that the comparison in [fig_ref] Figure 2: Int [/fig_ref] is strictly qualitative, since OMI measures the number of NO 2 molecules in a column over a grid of 0.25 - [bib_ref] Aura NO2 Cloud-Screened Total and Tropospheric Column L3 Global Gridded 0.25 •..., Krotkov [/bib_ref] , while in situ instruments measure the volumetric concentration near the ground; thus a direct quantitative comparison makes no sense. The annual NO 2 VCD ranges between 1.73 and 2.23 × 10 15 molec./cm 2 . Previous studies have shown that OMI measurements give reliable results about NO 2 emissions of anthropogenic sources at a large scale in cities [bib_ref] Validation of urban NO 2 concentrations and their diurnal and seasonal variations..., Boersma [/bib_ref] [bib_ref] Evolution of NO 2 in five major cities in Europe using remote..., Rosu [/bib_ref] , above large industrial platforms (located away from cities) [bib_ref] Ground-level nitrogen dioxide concentrations inferred from the satellite-borne Ozone Monitoring Instrument, Lamsal [/bib_ref] [bib_ref] NO 2 and SO 2 observations in SouthEast Europe using mobile DOAS..., Constantin [/bib_ref] and given by vehicle emissions and also by residential emissions [bib_ref] Verification of anthropogenic emissions of China by satellite and ground observations, Wang [/bib_ref]. However, other studies have shown that satellite observations do not correctly evaluate the NO 2 pollution caused by traffic or by other point sources [bib_ref] Validation of Ozone Monitoring Instrument NO2 measurements using ground based NO2 measurements..., Gruzdev [/bib_ref] [bib_ref] Satellite Observations of NO 2 Trend over, Constantin [/bib_ref]. [fig_ref] Figure 2: Int [/fig_ref] shows that the two time series agreed relatively well except for a period between the winter of 2009-2010 and the summer of 2011. The correlation coefficient between the two series was close to 0.40 (after values in December 2010 were removed) and did not change when NO 2 concentrations measured earlier, between 08:00 and 11:00 LT, were considered for the mean.
The seasonal variation of the tropospheric VCD was more regular: it was small during warm months and high during cold months. In 2010, the two time series did not coincide, which is mainly due to a peculiar behavior of the in situ measurements that showed an unexpected wide maximum during summer and autumn, and a minimum in December. The OMI instrument "sees" a large surface area (312 km 2 ), and thus integrates the emissions of many sources. Consequently, it cannot accurately measure the local variability, captured by the in situ measurements. This opposing variation suggests that a punctual, temporary, source of NO 2 was added in the second part of 2010 close to the measuring station. In general, the diurnal variation was relatively regular for fall, when a relatively small data spread was seen. A higher variability was observed in winter, spring and summer, when the data spread was larger. There were no outliers in the middle part of the day, when the traffic is slightly reduced compared to the morning and afternoon. The outliers during morning and afternoon may be associated with peaks or drop-offs of the traffic. The least regular season was summer, when the number of outliers was high regardless of the hour and their large majority lied in the upper part of the plot. This suggests that the NO2 loading was significantly higher than the average for a short period of time. This is confirmed by the next analysis [fig_ref] Figure 4: Departures of hourly NO2 concentration from the hourly seasonal means for each... [/fig_ref]. The explanation might relate to road construction works that resulted in deviation of the traffic. Starting with 2010, landslides occurred on streets near the measuring station. In 2011, the asphalt was removed and some excavations were done, in order to check the underground water and sewerage pipes. Subsequently, utility pipes were installed, the foundation was compacted and the traffic flow returned to normal values.
Apparently, the NO2 concentration and the temperature were anticorrelated, especially during daytime: high NO2 concentrations during morning corresponded to the lowest temperatures, while the afternoon reduction in NO2 coincided with the highest temperatures. However, this is just coincidental, since the daytime variation of NO2 is governed by traffic [bib_ref] Modeling results of atmospheric dispersion of NO 2 in an urban area..., Dragomir [/bib_ref] , whose peaks happen to occur at times when the temperature is lowest. Moreover, during nighttime, the NO2 concentration and the temperature seemed to be correlated.
## Monthly deviation of no2 concentration from the seasonal mean
Obviously, the diurnal variation of NO2 was not the same for each year. [fig_ref] Figure 4: Departures of hourly NO2 concentration from the hourly seasonal means for each... [/fig_ref] shows the difference between the hourly NO2 concentrations at ground level and the hourly seasonal mean during each month of each year. Hourly seasonal means are averages of hourly NO2 values measured during 2009 and 2014 over those months that define a season, i.e., November-February for winter, March-April for spring, May-August for summer and September-October for fall. In general, the diurnal variation was relatively regular for fall, when a relatively small data spread was seen. A higher variability was observed in winter, spring and summer, when the data spread was larger. There were no outliers in the middle part of the day, when the traffic is slightly reduced compared to the morning and afternoon. The outliers during morning and afternoon may be associated with peaks or drop-offs of the traffic. The least regular season was summer, when the number of outliers was high regardless of the hour and their large majority lied in the upper part of the plot. This suggests that the NO 2 loading was significantly higher than the average for a short period of time. This is confirmed by the next analysis [fig_ref] Figure 4: Departures of hourly NO2 concentration from the hourly seasonal means for each... [/fig_ref]. The explanation might relate to road construction works that resulted in deviation of the traffic. Starting with 2010, landslides occurred on streets near the measuring station. In 2011, the asphalt was removed and some excavations were done, in order to check the underground water and sewerage pipes. Subsequently, utility pipes were installed, the foundation was compacted and the traffic flow returned to normal values.
Apparently, the NO 2 concentration and the temperature were anticorrelated, especially during daytime: high NO 2 concentrations during morning corresponded to the lowest temperatures, while the afternoon reduction in NO 2 coincided with the highest temperatures. However, this is just coincidental, since the daytime variation of NO 2 is governed by traffic [bib_ref] Modeling results of atmospheric dispersion of NO 2 in an urban area..., Dragomir [/bib_ref] , whose peaks happen to occur at times when the temperature is lowest. Moreover, during nighttime, the NO 2 concentration and the temperature seemed to be correlated. Int. J. Environ. Res. Public Health 2020, 17, x 7 of 15 The analysis of the temporal variation of the aforementioned differences may be useful for identifying months or periods of the day when the NO2 loading departed from the expected variability, which in turn may help in finding the cause for the outliers seen in [fig_ref] Figure 3: shows the average diurnal variation of the NO2 concentration during each season... [/fig_ref]. Values lying in the positive/negative part mean that the NO2 concentration were higher/lower than the average at that particular time.
A clear pattern could be observed for equinox seasons: the afternoon peak was higher during colder months (March and October) than during warmer months (April and September) for each year. This was not true for the morning peak or for solstice seasons. The month of May, which is part of the summer season, was the least regular; the NO2 content in 2010 was much lower, while in 2011 it was much higher than the average. The unusual increase during the second part of 2010, shown in [fig_ref] Figure 4: Departures of hourly NO2 concentration from the hourly seasonal means for each... [/fig_ref] for July 2010 and, partially, for June 2010. The explanation might relate to the construction works previously described, which resulted in significant alterations of the traffic flow during 2010-2011. In general, the major contributor to the summer mean for all years came from the NO2 content in August, since the departures from the seasonal mean were positive for all years except 2011.
## Monthly deviation of no 2 concentration from the seasonal mean
Obviously, the diurnal variation of NO 2 was not the same for each year. [fig_ref] Figure 4: Departures of hourly NO2 concentration from the hourly seasonal means for each... [/fig_ref] shows the difference between the hourly NO 2 concentrations at ground level and the hourly seasonal mean during each month of each year. Hourly seasonal means are averages of hourly NO 2 values measured during 2009 and 2014 over those months that define a season, i.e., November-February for winter, March-April for spring, May-August for summer and September-October for fall.
The analysis of the temporal variation of the aforementioned differences may be useful for identifying months or periods of the day when the NO 2 loading departed from the expected variability, which in turn may help in finding the cause for the outliers seen in [fig_ref] Figure 3: shows the average diurnal variation of the NO2 concentration during each season... [/fig_ref]. Values lying in the positive/negative part mean that the NO 2 concentration were higher/lower than the average at that particular time.
A clear pattern could be observed for equinox seasons: the afternoon peak was higher during colder months (March and October) than during warmer months (April and September) for each year. This was not true for the morning peak or for solstice seasons. The month of May, which is part of the summer season, was the least regular; the NO 2 content in 2010 was much lower, while in 2011 it was much higher than the average. The unusual increase during the second part of 2010, shown in [fig_ref] Figure 2: Int [/fig_ref] , is confirmed by the higher values seen in [fig_ref] Figure 4: Departures of hourly NO2 concentration from the hourly seasonal means for each... [/fig_ref] for July 2010 and, partially, for June 2010. The explanation might relate to the construction works previously described, which resulted in significant alterations of the traffic flow during 2010-2011. In general, the major contributor to the summer mean for all years came from the NO 2 content in August, since the departures from the seasonal mean were positive for all years except 2011.
The NO 2 content in winter also depended on month and year. November seemed to be the month with the highest contribution to the winter seasonal average of NO 2 concentration for the first part of the interval (2009-2011). In December, the NO 2 diurnal variability changed with the year: the NO 2 content was lower during the first three years and higher afterwards. We assume that the explanation lies in a combination of meteorological conditions and important variations of traffic and industrial emissions during these particular years. During the analyzed period, the distribution of thermal energy and hot water in Braila municipality was mainly controlled by the heating operator SC "CET" SA. Starting with 2012, the lack of investments in modernizing the heating system affected the control of emissions: severe losses and the evolution of the methane gas tariff led to an increase of classical housing heating systems, whose impact is important especially during the cold season.
## Correlation between no 2 concentration and meteorological parameters
The effect of meteorological parameters on the variability of NO 2 concentration for urban sites is, still, a conundrum. [fig_ref] Table 2: Correlations between NO 2 and meteorological parameters for various locations [/fig_ref] shows some examples of correlations between NO 2 and meteorological factors for different locations [bib_ref] Dependence of urban air pollutants on meteorology, Elminir [/bib_ref] [bib_ref] Effect of Meteorological Conditions on Air Pollution of Surat City, Verma [/bib_ref] [bib_ref] An assessment of influence of meteorological factors on PM 10 and NO..., Dominick [/bib_ref] [bib_ref] Influence of Meteorological Parameters on Air Pollution in Isfahan, Hosseinibalam [/bib_ref] [bib_ref] Correlation of Various Gaseous Pollutants with Meteorological Parameter (Temperature, Relative Humidity and..., Srivastava [/bib_ref] [bib_ref] The Interaction between Air Quality and Meteorological Factors in an Arid Environment..., Habeebullah [/bib_ref] [bib_ref] Relationships between meteorological parameters and criteria air pollutants in three megacities in..., Zhang [/bib_ref]. The correlation between temperature and humidity, on the one side, and NO x , on the other side, was positive, negative or insignificant. The correlation with the wind speed was more consistent, i.e., is negative for all sites, which is rather normal, since a stronger wind will disperse the NO 2 at a local urban site. This is the reason for investigating whether meteorological parameters may be linked to the variability of the NO 2 in a relatively small city, with an average level of pollution [bib_ref] Estimating daily surface NO 2 concentrations from satellite data -a case study..., Anand [/bib_ref] and whether this relationship depends on seasons or on time of day. [fig_ref] Figure 2: Int [/fig_ref] show that during about 07:00 and 21:00 LT, the NO 2 variability was strongly influenced by traffic, which is also confirmed by [bib_ref] Modeling results of atmospheric dispersion of NO 2 in an urban area..., Dragomir [/bib_ref] [bib_ref] NO 2 and SO 2 observations in SouthEast Europe using mobile DOAS..., Constantin [/bib_ref] [bib_ref] Tropospheric nitrogen dioxide measurements in South-East of Romania using zenith-sky mobile DOAS..., Rosu [/bib_ref] [bib_ref] Correlations between NO 2 distribution maps using GIS and mobile DOAS measurements..., Rosu [/bib_ref] [bib_ref] Mobile measurements of nitrogen dioxide using two different UV-Vis spectrometers, Rosu [/bib_ref]. The correlation coefficients were computed separately for the full 24 h time (black bars), for daytime (8-21 LT, red bars) and for the nighttime (22-7 LT, blue bars). The separation between daytime and nighttime was done because the influence of the road traffic should be lower during nighttime, and thus meteorological parameters may play a different role in modulating the NO 2 concentration during the night. Obviously, this is not valid for radiation, which is absent during nighttime. However, one should keep in mind that the NO 2 content is largely controlled by traffic, especially during the day, and this will definitely affect the correlation with meteorological parameters (or lack thereof). However, we assumed that the traffic pressure does not change for the analysis period. [fig_ref] Figure 5: Full bivariate correlation of NO 2 with the temperature [/fig_ref] shows the variation of the direct bivariate correlation between NO 2 and temperature (left), and humidity (right) with time. Correlation coefficients calculated for the full day are shown by black bars, while for day (night) these are shown by red (blue) bars. Only coefficients that were significant at 90% are shown. There was no clear NO2 dependence on temperature, since correlation coefficients were positive for March, May, August and November, while for February and September these were negative. During daytime most correlations were negative; however, this was already discussed as being artificially induced by rush-hour traffic. Correlations did not change from day to night, except for May, when the correlation changed from negative for the full-day to positive for the full day and night. The full 24 h correlation was rather small or insignificant for most months and changed from positive to negative during months that had similar meteorological characteristics, e.g., March (positive), April (negative) and May (positive). Unsurprisingly, NO2 and temperature were anticorrelated during the day, but this was already discussed as being mostly artificial. Significant correlation during the night was positive in March, May, August and November, and negative in January, February and September. All in all, temperature seemed to play no clear role in the variability of NO2.
Negative coefficients were found for the NO2-temperature dependence by [bib_ref] Effect of Meteorological Conditions on Air Pollution of Surat City, Verma [/bib_ref] [bib_ref] Influence of Meteorological Parameters on Air Pollution in Isfahan, Hosseinibalam [/bib_ref] [bib_ref] Correlation of Various Gaseous Pollutants with Meteorological Parameter (Temperature, Relative Humidity and..., Srivastava [/bib_ref] [bib_ref] The Interaction between Air Quality and Meteorological Factors in an Arid Environment..., Habeebullah [/bib_ref] , without a clear association with seasons. The NO2 lifetime is higher during winter; thus the "night" analysis may be less relevant during winter because of the lower concentrations of the N2O and N2O5 species. These species are key factors in the removal of NO2 during nighttime and in its transformation into nitric acid [bib_ref] Differential Absorption Spectroscopy, Platt [/bib_ref]. The amount of NOy species is higher during warmer seasons when bacteria and agriculture activities are intensified [bib_ref] A global inventory of nitric oxide emissions from soils, Davidson [/bib_ref]. However, [bib_ref] An assessment of influence of meteorological factors on PM 10 and NO..., Dominick [/bib_ref] and [bib_ref] Relationships between meteorological parameters and criteria air pollutants in three megacities in..., Zhang [/bib_ref] found that an increase in temperature would be followed by an increase in NOx.
The correlation of the NO2 concentration with the humidity, when significant, was positive for most months. The only exception was May, during which the NO2 variability departed significantly from the expected behavior [fig_ref] Figure 4: Departures of hourly NO2 concentration from the hourly seasonal means for each... [/fig_ref]. The correlation did not change from day to night except in September.
In the following the intervening effect on direct correlations is analyzed, to see whether the existing correlations were spuriously induced, especially for the link to humidity. The partial There was no clear NO 2 dependence on temperature, since correlation coefficients were positive for March, May, August and November, while for February and September these were negative. During daytime most correlations were negative; however, this was already discussed as being artificially induced by rush-hour traffic. Correlations did not change from day to night, except for May, when the correlation changed from negative for the full-day to positive for the full day and night. The full 24 h correlation was rather small or insignificant for most months and changed from positive to negative during months that had similar meteorological characteristics, e.g., March (positive), April (negative) and May (positive). Unsurprisingly, NO 2 and temperature were anticorrelated during the day, but this was already discussed as being mostly artificial. Significant correlation during the night was positive in March, May, August and November, and negative in January, February and September. All in all, temperature seemed to play no clear role in the variability of NO 2 .
Negative coefficients were found for the NO 2 -temperature dependence by [bib_ref] Effect of Meteorological Conditions on Air Pollution of Surat City, Verma [/bib_ref] [bib_ref] Influence of Meteorological Parameters on Air Pollution in Isfahan, Hosseinibalam [/bib_ref] [bib_ref] Correlation of Various Gaseous Pollutants with Meteorological Parameter (Temperature, Relative Humidity and..., Srivastava [/bib_ref] [bib_ref] The Interaction between Air Quality and Meteorological Factors in an Arid Environment..., Habeebullah [/bib_ref] , without a clear association with seasons. The NO 2 lifetime is higher during winter; thus the "night" analysis may be less relevant during winter because of the lower concentrations of the N 2 O and N 2 O 5 species. These species are key factors in the removal of NO 2 during nighttime and in its transformation into nitric acid [bib_ref] Differential Absorption Spectroscopy, Platt [/bib_ref]. The amount of NO y species is higher during warmer seasons when bacteria and agriculture activities are intensified [bib_ref] A global inventory of nitric oxide emissions from soils, Davidson [/bib_ref]. However, [bib_ref] An assessment of influence of meteorological factors on PM 10 and NO..., Dominick [/bib_ref] [bib_ref] Relationships between meteorological parameters and criteria air pollutants in three megacities in..., Zhang [/bib_ref] found that an increase in temperature would be followed by an increase in NO x .
The correlation of the NO 2 concentration with the humidity, when significant, was positive for most months. The only exception was May, during which the NO 2 variability departed significantly from the expected behavior [fig_ref] Figure 4: Departures of hourly NO2 concentration from the hourly seasonal means for each... [/fig_ref]. The correlation did not change from day to night except in September.
In the following the intervening effect on direct correlations is analyzed, to see whether the existing correlations were spuriously induced, especially for the link to humidity. The partial correlation was used to identify whether the effects of meteorological parameters (temperature, relative humidity and solar radiation) on NO 2 concentration were interconnected. The main and intervening variables (described in Section 2) are shown in [fig_ref] Table 3: Meteorological parameters as main/intervening variables, for the partial correlation analysis [/fig_ref]. The results are shown only for four out of the six possible combinations, because these are the most relevant for identifying possible intervening effects. The correlation between NO 2 and temperature was inconsistent and when assessing the intervening effect of radiation on correlation with humidity one can infer that the intervening effect of humidity was similar. . Partial correlation between NO2 concentration and meteorological parameters. Full bars describe the bivariate correlation and empty bars describe the difference (D) between the partial and bivariate correlation. Correlation coefficients for the whole day are shown in black, while red/blue correspond to day/night, respectively-see text for details. Coefficients are shown for the >90% confidence level. Partial correlation between NO2 and humidity is shown in (a) (temperature constant) and (b) (radiation constant). Partial correlation between NO2 and radiation is shown in (c) (temperature constant) and the effect of radiation on the correlation with temperature is in (d). [fig_ref] Table 3: Meteorological parameters as main/intervening variables, for the partial correlation analysis [/fig_ref]. Meteorological parameters as main/intervening variables, for the partial correlation analysis.
## Variable 1 (x) variable 2 (z)
Intervening Variable (Y) Plot in [fig_ref] Figure 2: Int [/fig_ref] relative humidity temperature a NO2
relative humidity radiation b . Partial correlation between NO 2 concentration and meteorological parameters. Full bars describe the bivariate correlation and empty bars describe the difference (D) between the partial and bivariate correlation. Correlation coefficients for the whole day are shown in black, while red/blue correspond to day/night, respectively-see text for details. Coefficients are shown for the >90% confidence level. Partial correlation between NO 2 and humidity is shown in (a) (temperature constant) and (b) (radiation constant). Partial correlation between NO 2 and radiation is shown in (c) (temperature constant) and the effect of radiation on the correlation with temperature is in (d). shows the results of the partial correlation analysis for the combinations in [fig_ref] Table 3: Meteorological parameters as main/intervening variables, for the partial correlation analysis [/fig_ref]. Full bars describe the direct correlation, while empty bars stand for the differences between partial and direct coefficients. According to [fig_ref] Table 1: Type of correlation for various results of the partial correlation analysis [/fig_ref] , when these two are opposite, the correlation is partially mediated by the intervening variable. When these both have the same sign, the correlation is real.
No consistent direct correlation between NO 2 and humidity was found for the whole 24 h period , full black bars are completely absent). Notably, D NO(t)H was rather high and positive for the whole 24 h day during May, June and December. This means that the temperature suppressed a potentially positive NO x -humidity correlation. During daytime, NO 2 and humidity were positively correlated (red full bars) for a large part of the year, except for February, August and November. The differences between partial and direct correlation, D NO(t)H , albeit small, were of opposing signs for most cases. This suggests that the temperature was partially responsible for the NO 2 -humidity correlation.
The important role played by the humidity in the variation of NO 2 is supported by , where the effect of radiation on the same correlation (NO 2 -humidity) is shown only for daytime. The solar radiation partially induced a positive correlation during some months, but the effect was not important, since all D NO(R)H were pretty small. Our results agree with [bib_ref] Relationships between meteorological parameters and criteria air pollutants in three megacities in..., Zhang [/bib_ref] or, but contradict [bib_ref] Dependence of urban air pollutants on meteorology, Elminir [/bib_ref] , who showed that NO 2 was negatively correlated with relative humidity. They argued that NO 2 concentrations are slightly higher at a lower relative humidity because the reactions between NO 2 and OH are less frequent, and thus the NO 2 persists more in the atmosphere. However, this was not confirmed by our results.
The NO 2 concentration was negatively correlated with solar radiation during the entire year. A higher direct radiation implies, usually, a higher air temperature; thus the intervening effect of temperature and radiation was analyzed. shows that the temperature did not artificially induce the anticorrelation with solar radiation (small or absent empty bars), except for May and September, when the effect was, however, small. This holds for the intervening effect of solar radiation on the anticorrelation with temperature, , which was also small.
Based on observations in [bib_ref] An observation of seasonal and diurnal behavior of O 3 -NO x..., Pancholi [/bib_ref] , which showed that, for a site in India, O 3 correlated negatively with both NO x and the humidity during all seasons, we suggest that the positive correlation between NO x and humidity may be an indirect result of the photochemical effect of solar radiation and humidity on ozone. Unfortunately, ozone measurements were not available to confirm this hypothesis. This also may partly explain the observed anticorrelation between NO 2 and solar radiation. Increased solar radiation favors the production of O 3 , which, in turn, reduces the NO 2 loading in the atmosphere [bib_ref] An observation of seasonal and diurnal behavior of O 3 -NO x..., Pancholi [/bib_ref].
In general, comparisons with other studies are not straightforward, since we are not aware of similar investigations of the intervening effect of meteorological parameters. Additionally, most studies did not consider monthly changes. Moreover, correlations between the NO 2 concentration and the meteorological parameters should be different for different cities, since the anthropogenic landscape and microclimate change significantly from one urban location to another [bib_ref] Estimating daily surface NO 2 concentrations from satellite data -a case study..., Anand [/bib_ref].
# Conclusions
This paper describes the diurnal, monthly, seasonal and annual evolution of the NO 2 concentration for 2009-2014, measured in situ by an urban traffic station in southeast Romania. The role that meteorological parameters might play in modulating the NO 2 variability was investigated in an attempt to separate the anthropogenic effects (which are well-known) from the effect of the local microclimate.
As expected, the NO 2 was higher during the cold season, except for one year, 2010, when summer NO 2 levels were highest. This suggests that natural factors, such as the effect of temperature on the NO 2 lifetime, are less important than the anthropogenic ones at urban sites. Some annual variation also existed, with low values at the beginning of the interval (2009-2010), most likely caused by a severe reduction of industrial activity.The summer minima and winter maxima have both anthropogenic and natural causes and the departure from these may relate to temporary changes of the local traffic and/or construction activities. The NO 2 diurnal variability was clearly shaped by the local transport: two diurnal peaks were observed, one around 8-9 LT and another one around 20-21 LT, and both were associated with increased road traffic, confirming previous observations at other urban sites. The afternoon peak was higher during the colder months (March and October) than during the warmer months (April and September) for each year. An irregular diurnal variation of the NO 2 concentration was seen in May and December. The most consistent season was autumn, with a relatively similar diurnal variation in all years.
Additionally, we found that over Braila, space observations of OMI followed the in situ observations during most of the selected interval (R = 0.4).
The analysis of the correlations between the NO 2 concentration and temperature, relative humidity and radiation has shown that the association with temperature is the least relevant. The correlation changed from positive to negative throughout the year without a clear pattern.
Obviously, the contribution of traffic cannot be disregarded and may mask or suppress the impact of temperature variations on the NO 2 concentration. Our assumption that during the night, the situation may change due to traffic disappearance, was not confirmed. The correlations with the humidity and radiation, on the other hand, were notably consistent: the NO 2 concentration correlated with the relative humidity and was anticorrelated with radiation for almost the entire year. Moreover, most of these relationships were real and the intervening effect of the other meteorological parameters was small.
Our results showed that finding a link between meteorological parameters and NO 2 variability for an urban site is a difficult task. Attempts to predict the NO 2 behavior based on meteorological data, even combined with traffic flux data, cannot be very successful at the local or regional scale or on a short-term basis, since landscape, infrastructure, traffic, local activities, and population clearly affect the NO 2 concentration. However, this may be useful for assessing trends or long-tern variability at a large scale, since these are averaging over a large area, thus reducing local and short-term contributions.
[fig] Figure 1: shows the diurnal and seasonal variation of NO 2 concentration (top) and temperature (bottom) during 2009-2014. The hourly averaged NO 2 concentration varied between 14 µg/m 3 (July 2009) and 41 µg/m 3 (January 2011) [/fig]
[fig] Figure 2: Int. J. Environ. Res. Public Health 2020, 17, Variation of standardized monthly averages of tropospheric NO2 (VCD measured by OMI, blue, dash) and of NO2 ground concentration measured in situ at 11:00 LT (black, solid) starting with January 2009 until December 2014. [/fig]
[fig] Figure 3: shows the average diurnal variation of the NO2 concentration during each season for the entire period. Seasons are defined here in a slightly different manner than usual: spring and fall/autumn consist of two months each (March-April and September-October), while solstice seasons have four months: the summer covers the period between May and August, and the winter starts in November and ends in February. This definition of seasons is justified by the weather profile during the last 20 years in Braila County and in South-Eastern Romania, which shows a fast transition from winter to summer and then back to winter. [/fig]
[fig] Figure 4: Departures of hourly NO2 concentration from the hourly seasonal means for each month, January to December. Each year is shown with different colors/line styles (see legend). [/fig]
[fig] 15, Figure 5: Int. J. Environ. Res. Public Health 2020, 17, x 10 of Full bivariate correlation of NO2 with the temperature (left) and humidity (right). Coefficients (>90% confidence level) are shown for 24 h by black bars, while red/blue bars correspond to day/night, respectively-see text for details. [/fig]
[fig] Figure 5: Full bivariate correlation of NO 2 with the temperature (left) and humidity (right). Coefficients (>90% confidence level) are shown for 24 h by black bars, while red/blue bars correspond to day/night, respectively-see text for details. [/fig]
[fig] Author: Contributions: Conceptualization, M.V.; methodology, M.V. and D.-E.C.; formal analysis, M.V.; investigation, M.V. and V.C.; resources, D.-E.C., C.M.D.B., S.C.-B., V.C. and A.R.; data curation, C.M.D.B.; writing-original draft preparation, M.V.; writing-review and editing, M.V., D.-E.C., S.C.-B.; funding acquisition, M.V. All authors have read and agreed to the published version of the manuscript. [/fig]
[fig] Funding: This work was supported by the EXPERT project, contract no. 14PFE/17.10.2018 with the Romanian Ministry of Research and Innovation. [/fig]
[table] Table 1: Type of correlation for various results of the partial correlation analysis. [/table]
[table] Table 2: Correlations between NO 2 and meteorological parameters for various locations. [/table]
[table] Table 3: Meteorological parameters as main/intervening variables, for the partial correlation analysis. [/table]
|
10.1371/journal.pone.0286373
|
CCBY
|
10228762
|
37253027
|
s2orc_pubmed_articles
|
SARS-CoV-2 spike protein diversity at an intra-host level, among SARS-CoV-2 infected individuals in South Africa, 2020 to 2022
Intra-host diversity studies are used to characterise the mutational heterogeneity of SARS-CoV-2 infections in order to understand the impact of virus-host adaptations. This study investigated the frequency and diversity of the spike (S) protein mutations within SARS-CoV-2 infected South African individuals. The study included SARS-CoV-2 respiratory samples, from individuals of all ages, received at the National Health Laboratory Service at Charlotte Maxeke Johannesburg Academic hospital, Gauteng, South Africa, from June 2020 to May 2022. Single nucleotide polymorphism (SNP) assays and whole genome sequencing were performed on a random selection of SARS-CoV-2 positive samples. The allele frequency (AF) was determined using TaqMan Genotyper software for SNP PCR analysis and galaxy.eu for analysis of FASTQ reads from sequencing. The SNP assays identified 5.3% (50/948) of Delta cases with heterogeneity at delY144 (4%; 2/50), E484Q (6%; 3/50), N501Y (2%; 1/50) and P681H (88%; 44/50), however only heterogeneity for E484Q and delY144 were confirmed by sequencing. From sequencing we identified 9% (210/2381) of cases with Beta, Delta, Omicron BA.1, BA.2.15, and BA.4 lineages that had heterogeneity in the S protein. Heterogeneity was primarily identified at positions 19 (1.4%) with T19IR (AF 0.2-0.7), 371 (92.3%) with S371FP (AF 0.1-1.0), and 484 (1.9%) with E484AK (0.2-0.7), E484AQ (AF 0.4-0.5) and E484KQ (AF 0.1-0.4). Mutations at heterozygous amino acid positions 19, 371 and 484 are known antibody escape mutations, however the impact of the combination of multiple substitutions identified at the same position is unknown. Therefore, we hypothesise that intra-host SARS-CoV-2 quasispecies with heterogeneity in the S protein facilitate competitive advantage of variants that can completely/partially evade host's natural and vaccine-induced immune responses.
[formula] a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 [/formula]
# Introduction
The primary function of the coronavirus spike glycoprotein (S protein) is to facilitate virus attachment and infection of host cells [bib_ref] Human Coronavirus : Host-Pathogen Interaction, Fung [/bib_ref]. Variations occurring within its amino acid (aa) sequence could influence the interaction with the host receptor, viral pathogenesis, transmissibility and infectivity [bib_ref] Human Coronavirus : Host-Pathogen Interaction, Fung [/bib_ref]. Similarly, since the emergence of SARS-CoV-2 in 2019, the S protein has evolved rapidly, resulting in diverse lineages of which some manifested as variants of concern (VOC). Globally waves of infections were mainly driven by VOCs of the Alpha, Beta, Delta, Gamma, and Omicron lineages. In South Africa, SARS-CoV-2 Wuhan-Hu1 wild-type strain dominated the first coronavirus disease of 2019 wave, the second and third waves were dominated by Beta and Delta VOCs respectively, and the fourth and slowly emerging fifth waves were dominated by Omicron BA.1, BA.2, BA.4 and BA.5 lineages [bib_ref] Detection of a SARS-CoV-2 variant of concern in South Africa, Tegally [/bib_ref] [bib_ref] Sixteen novel lineages of SARS-CoV-2 in South Africa, Tegally [/bib_ref] [bib_ref] Rapid epidemic expansion of the SARS-CoV-2 Omicron variant in southern Africa, Viana [/bib_ref] [bib_ref] Identification of SARS-CoV-2 Omicron variant using spike gene target failure and genotyping..., Subramoney [/bib_ref] [bib_ref] Early assessment of the clinical severity of the SARS-CoV-2 omicron variant in..., Wolter [/bib_ref]. The continuous emergence of new SARS-CoV-2 lineages overtime and the reversion of mutations together with recombination within the SARS-CoV-2 genome, predominantly in the S protein, could result in strains increasing resistant to antibody-mediated neutralisation. This highlights the need for characterisation of intra-host SARS-CoV-2 S protein diversity.
Several studies have shown that studying intra-host SARS-CoV-2 S protein diversity leads to further insight and understanding of SARS-CoV-2 transmission dynamics and could be important for the improvement of SARS-CoV-2 vaccines [bib_ref] Intra-host variation and evolutionary dynamics of SARS-CoV-2 populations in COVID-19 patients, Wang [/bib_ref] [bib_ref] Intrahost evolution during SARS-CoV-2 prolonged infection, Voloch [/bib_ref] [bib_ref] SARS-CoV-2 withinhost diversity and transmission, Lythgoe [/bib_ref] [bib_ref] Transmission dynamics of SARS-CoV-2 within-host diversity in two major hospital outbreaks in..., San [/bib_ref]. The Alpha, Beta, and Delta variants showed escape from neutralising antibodies induced by the dominant vaccine immune epitopes in the S protein. Circulation of these variants also resulted in varying efficacy (50-90%) of licenced SARS-CoV-2 vaccines, which comprised the Wuhan-Hu1 derived S protein. In addition, spatiotemporal analysis of intra-host single nucleotide variants (iSNVs) showed increased genetic diversity following symptomatic infections. In South Africa, iSNVs analysis coupled with bottleneck transmission, was successfully used to determine the transmission dynamics of SARS-CoV-2 in nosocomial outbreaks [bib_ref] Transmission dynamics of SARS-CoV-2 within-host diversity in two major hospital outbreaks in..., San [/bib_ref]. The study identified patients with similar iSNV profiles together with the evolution of SARS-CoV-2 infections. A limitation of the study was that intra-host variants were analysed at a nucleotide level with limited focus being placed on the impact of variants on possible immune escape based on aa variations.
Although SARS-CoV-2 variants that dominated the waves from 2020 to 2022 in South Africa have been described, the frequency of heterogeneous infections and heterogeneity within each defined lineage and VOC lineage is not well defined. Therefore, the factors that contribute to the emergence of dominant variants, including factors that could affect variant persistence or eventual replacement is still unclear. Intra-host diversity studies will assist with identifying and characterising the frequency of mutations and possible heterogeneity within the SARS-CoV-2 S protein. This will broaden our understanding of the virus-host adaptations, virulence factors, and pathogenicity and potentially provide information for more effective vaccine design. In this study we aimed to determine the diversity of SARS-CoV-2 S protein among SARS-CoV-2 infected individuals and if heterogeneity has an impact on intra-host adaptations.
# Methods
## Study population
The study included persons of all ages from whom respiratory samples were submitted for SARS-CoV-2 diagnosis to the National Health Laboratory Service (NHLS) at Charlotte Maxeke Johannesburg Academic hospital (CMJAH), Gauteng, South Africa, from June 2020 to May 2022. CMJAH received SARS-CoV-2 samples within the Gauteng province, as well as referrals from the Eastern Cape, Free-State KwaZulu-Natal, and Western Cape provinces of South Africa. This study contributed towards the national surveillance for SARS-CoV-2 in South Africa for which formal patient consent was not required. The study was approved by the University of Witwatersrand Human Research Ethics Committee (M210119) and all participants' data was anonymised and presented in an aggregated form.
## Plos one
## Study samples and sample size
Respiratory specimens submitted for SARS-CoV-2 diagnosis included nasal and nasopharyngeal swabs, throat swabs and oropharyngeal swabs, sputum, tracheal and lung aspirates, bronchoalveolar aspirates or lavages. Residual samples were stored for further characterisation of the infecting strains. Approximately 50-80 SARS-CoV-2 positive samples were randomly selected per week for genomics surveillance from 2020-2022. We strived to include at least 10 samples from each district from which samples were collected per day, for detection of single nucleotide polymorphisms (SNP) in the S protein.
## Sars-cov-2 diagnosis
The respiratory samples were extracted using three extraction platforms and compatible
## Genotyping to detect single nucleotide polymorphisms (snps) associated with specific amino acid mutations in the s protein
Total nucleic acids were extracted from randomly selected samples using the automated King-Fisher Flex purification system with the MagMAX™ Viral/Pathogen II nucleic acid extraction kit (Thermo Fisher Scientific), as per manufacturer's instructions. The TaqMan SARS-CoV-2 S gene singleplex mutation panels (Thermo Fisher Scientific) were used to identify 11 mutations (T20N, delH69/70, delY144, K417N, L452R, E484K, E484Q, N501Y, D614G, P681H and P681R). Results from this assay allowed us to determine the presence of homogeneous and/ or heterogenous (i.e., mixed mutations associated with more than 1 genotype) virus populations within an individual. Analysis was performed using the QuantStudio 5 design and analysis and TaqMan1 Genotyper software's. Alleles that clustered along the x-axis (allele 1 VIC) were represented as homozygous wild type with AF� 0.5, alleles that clustered along the y-axis (allele 2 FAM) represented homozygous mutants with AF� 0.5, and alleles that clustered between the x-axis and y-axis represented heterogeneous wild type/mutant.
## Next-generation sequencing of sars-cov-2 strains
All samples selected for sequencing where submitted to the KwaZulu-Natal Research Innovation and Sequencing Platform (KRISP) for sequencing or were sequenced in-house at the NHLS, Virology Laboratory. Samples were amplified using the ARTIC V3, V4 and V4.1 protocols [bib_ref] Detection of a SARS-CoV-2 variant of concern in South Africa, Tegally [/bib_ref]. At KRISP, libraries were prepared using the Nextera DNA library preparation kits (Illumina, United States) for sequencing performed on the MiSeq platform (Illumina) and ligation sequencing kits (Oxford Nanopore technologies, Oxford, United Kingdom) for libraries prepared for sequencing with the MinION (Oxford Nanopore technologies, United Kingdom) [bib_ref] Rapid epidemic expansion of the SARS-CoV-2 Omicron variant in southern Africa, Viana [/bib_ref] [bib_ref] Emergence of SARS-CoV-2 Omicron lineages BA.4 and BA.5 in South Africa, Tegally [/bib_ref]. For in-house sequencing, libraries were prepared using the Nextera DNA kits (Illumina) and 8pM of pooled libraries were loaded together with 1% PhiX control which were sequenced on the MiSeq platform (Illumina). Genome assembly was done using genome detective (https://www.genomedetective.com/app/typingtool/virus/) by KRISP and Exatype (https://sars-cov-2.exatype.com/) online tools. Consensus sequences were analysed using Nextstrain (https://clades.nextstrain.org) and Pangolin (https://pangolin.cog-uk.io/) online tools for variant assignment.
## Variant mutation analysis and identification of heterogeneous infections
FASTQ reads generated from sequencing were analysed using the COVID-19 workflows with galaxy.eu [bib_ref] The Galaxy platform for accessible, reproducible and collaborative biomedical analyses: 2018 update, Enis [/bib_ref]. To generate allele frequencies for major and minor variants at the aa level across the S protein: for Illumina pair-end reads we ran the "COVID-19: variation analysis on ARTIC PE data (V0.5)" together with the "COVID-19: variation analysis reporting" workflow; and the "COVID-19 variation analysis of ARTIC ONT (V0.3.1)" workflow was used for Min-ION single reads. The final output from galaxy.eu indicated if the sequence data passed the quality checks (indicated as "PASS"). Only the sequences that passed were included in this analysis. Heterogenous mutations were identified if the allele frequency ranged from 0.1 to 1.0 for more than 1 mutation.
# Data analysis
Descriptive analysis was done across different age groups (<5, 5-14, 15-24, 25-44, 45-60, and >60 years), gender, province, and specimen collection site (individuals that sought community screen and test services, healthcare workers staff clinic, in-patient wards, out-patient wards, mortuary). Intra-host sequences and genotyping data were analysed to determine the proportion of individuals infected with homogeneous and/or heterogeneous variant populations. A box-whisker analysis of heterogeneous aa mutations was used to display the intra-host frequency of heterogeneous SNPs across the S protein. A time-tree was generated by running our sequence dataset through a NextStrain custom build (https://github.com/nextstrain/ncov) and the most prevalent heterogeneous mutation positions were selected for analysis. The positions with heterogeneous mutations were compared to the global prevalence of mutations at the same positions from 2020 through to 2022, using GISAID CoVServer (https://www.gisaid.org/ epiflu-applications/covsurver-mutations-app/; [bib_ref] GISAID's Role in Pandemic Response, Khare [/bib_ref].
# Results
## Demographic characteristics of study population
Over the study period a total of 667 269 respiratory samples were tested, of which 107 495 (16.1%) tested positive for SARS-CoV-2. A total of 2982/107 495 (2.8%) SARS-CoV-2 positive samples were analysed to assess the proportion with mixed infections. Of these, 20.3% (601/ 2982) were analysed by SNP PCR only, 68.2% (2034/2982) were sequenced and 11.6% (347/ 2982) were both were analysed by SNP PCR and sequenced. Just under 50% of SARS-CoV-2 positive samples belonged to adults in age group 25-44 years (1351/2982; 45.3%), followed by older adults aged 45-60 years (594/2982; 20.0%) . The proportion of females (57.9%; 1726/2982) was greater than males (39.6%; 1181/ 2982). The majority of samples were from individuals residing in Gauteng province (2830/ 2982; 95.0%). Individuals that were out-patients and in-patients represented 11.5% (344/2982) and 8.1% (242/2982) respectively, while 77.9% (2324/2982) of individuals were from community screen and testing (CST) . ARV clinic and cardiac clinic attendees represented 25.0% (86/344) and 0.6% (2/344) of out-patients respectively.
Five
## Table 1. demographics of individuals with sars-cov-2 included in this study (n = 2982).
## Demographics genotyped (n/n, (%))
## Age group (years)
## Sequence-based analysis of heterogeneity in s protein
Fifty homogeneous SNPs associated with the primary VOCs detected in South Africa were identified (S3 . SNPs identified in Beta included D80A, D215G, K417N, E484K, N501Y, D614G, and A701V with an AF ranging from 0.6-1.0. Individuals infected with Delta had T19R, G142D, delFR157-158, L452R, and T478K (AF ranging from 0.5-1.0) mutations. Fortyone mutations (S3 with an AF of 0.6-1.0, were detected among Omicron-infected individuals. Important mutations at positions R346 and K444 were not detected.
Close to 9% (210/2381; 8.8%) of South African SARS-CoV-2 cases sequenced displayed heterogeneity at one or more SNP positions. A total 7 aa positions across the S protein, with AF scores ranging from 0.1 to 1.0, were identified with multiple heterogeneous mutations [fig_ref] Fig 3: Box and whisker plot showing the mutant allele frequencies [/fig_ref].
## Plos one
Diversity of SARS-CoV-2 spike protein cases with heterogeneous mutations had a high AF >0.8-0.9 for 371F (C>T; 22674) and 371P (T>C, 22763) [fig_ref] Fig 3: Box and whisker plot showing the mutant allele frequencies [/fig_ref]. [fig_ref] Fig 4: Fig 4 [/fig_ref]. The sample with T19KR was identified in October 2021 but could not be classified (unassigned) due to lower sequence quality at other regions of the genome. S371FP heterogeneous mutations were identified from June 2021 through to May 2022 of which 1/195 (0.5%) was classified as Delta. The S371FP heterogeneous mutation rapidly increased from the Delta (June to August 2021) to the Omicron waves and dominated during the Omicron BA.1/BA.2 wave from November 2021 .0%) to December 2021 (122/195; 62.6%) [fig_ref] Fig 5: Prevalence of samples with S371FP over time [/fig_ref]. From January 2022 to May 2022 the S371FP mutation was observed but at lower frequencies from 2.1% (4/195) to 5.1% of the cases due to low sequence data quality in some regions of the S gene [fig_ref] Fig 4: Fig 4 [/fig_ref]. The majority (146/195; 74.9%) of cases with the S371FP were from individuals that sought CST, while the remaining cases (25.2%) were from deceased individuals (1/ 195; 0.5%), HCW (4/195; 2.1%), in-patients .3%) or out-patients .3%). [fig_ref] Fig 4: Fig 4 [/fig_ref]. E484KQ (1/4; 25%) was observed in an out-patient from an ARV clinic in August 2022 and classified as Delta [fig_ref] Fig 4: Fig 4 [/fig_ref]. E484AK (2/4; 50%) and E484AQ (1/4; 25%) were identified among individuals that sought CST. E484AK was observed among samples classified as Beta and Delta in April 2021 and October 2021, respectively. While E484AQ (1/4; 25%) was identified in October 2021, and classified as Delta.
## Mutations at dominant heterogeneous snps positions compared to global data
The consensus sequence data was used to compare our data to global data. Globally, from January 2020 through to May 2022, multiple aa changes were seen at positions 19, 371, and 484 among 11 140 058 SARS-CoV-2 genome sequences uploaded on GISAID (accessed 22 June 2022). At aa position 19, the T19I mutation was detected in 15.8% of global sequences (January 2022 through May 2022) in Omicron BA.2 (21L clade) and T19R was identified among 38.3% of Delta lineages from March 2021 through March 2022, similarly identified in our study sequences [fig_ref] Fig 6: Frequency of mutations occurring at aa positions 19, 371, and 484 in... [/fig_ref]. At position 371, the S371F mutation was globally dominant (15.21%) and identified from December 2021 to May 2022 among Omicron BA.2 lineages, while our study identified S371F primarily among Omicron BA.4 (22A clade) which clustered out from Omicron BA.2 from June 2021 to May 2022. S371P is not yet reflected since it was only observed in June 2022 in 396 global genomes and was not observed from our data due to S371F being the dominant mutation from the consensus sequences. S371L was identified among our consensus sequences but was not observed among the heterogeneous mutations. At position 484, E484A
# Discussion
Intra-host SARS-CoV-2 heterogeneity results in greater viral diversity and viral evolutionary advantage to enhance virus-host adaptation, immune evasion, replication, transmission, and pathogenicity [bib_ref] SARS-CoV-2 withinhost diversity and transmission, Lythgoe [/bib_ref] [bib_ref] Quasispecies of SARS-CoV-2 revealed by single nucleotide polymorphisms (SNPs) analysis, Gao [/bib_ref] [bib_ref] Patient-derived SARS-CoV-2 mutations impact viral replication dynamics and infectivity in vitro and..., Yao [/bib_ref]. Diversity studies identified high variability of iSNVs in several regions of the genome such as the ORF and N genes of SARS-CoV-2, with a high frequency of Frequency of the heterozygous mutations (4A) and VOC/lineages (4B) identified from sequence data over time. 4A) X-axis represents the date during which the heterozygous SNP was identified, and the y-axis represents the percentage of samples that were observed with a particular heterozygous SNP. 4B) X-axis represents the period during which the VOC/lineages were identified, and the y-axis represents the percentage of samples with a specific VOC/lineage.
## Plos one
Diversity of SARS-CoV-2 spike protein
## Plos one
Diversity of SARS-CoV-2 spike protein allelic variations, especially within ORF8 and negative selective pressure towards the S gene. The mutation profiles of intra-host quasispecies were also shown to change from the onset of symptoms which could contribute to viral evolution occurring during the course of infection [bib_ref] SARS-CoV-2 withinhost diversity and transmission, Lythgoe [/bib_ref] [bib_ref] Quasispecies of SARS-CoV-2 revealed by single nucleotide polymorphisms (SNPs) analysis, Gao [/bib_ref] [bib_ref] Patient-derived SARS-CoV-2 mutations impact viral replication dynamics and infectivity in vitro and..., Yao [/bib_ref]. Our study assessed the intra-host diversity based on the heterogeneity of SNPs across the S gene, not previously described in South Africa, to postulate the impact on host immune responses.
In contrast to earlier studies, our study period from June 2020 to May 2022 encompassed five SARS-COV-2/COVID-19 infection waves in South Africa. The 4 th wave, driven by the Omicron BA.1 and BA.2 lineages, had the highest detection rate peaking at 58%, while the lowest peak detection rate (24%) was observed during the Omicron BA.4/BA.5 driven wave of infections. Although the majority of samples included in this study were from the Gauteng province, the overall distribution of SARS-CoV-2 lineages and VOCs was similar in other Provinces of South Africa [bib_ref] Network for Genomic Surveillance in South Africa (NGS-SA), Ngs-Sa [/bib_ref]. Individuals that seek CST are usually asymptomatic or present with mild symptoms as a result of a causative agent other than SARS-CoV-2, therefore a drop in the detection rate was observed. It has been reported that individuals infected with Omicron have reduced severity of infection compared to Delta as well as immunity resulting from natural infection and vaccination within the population [bib_ref] Population Immunity and Covid-19 Severity with Omicron Variant in South Africa, Madhi [/bib_ref] [bib_ref] Delta Coronavirus Variant : Scientists brace for impact, Callaway [/bib_ref]. Of the cases that tested positive for SARS-CoV-2 in this study, both males and females were predominantly from the 25-44 year age group and CST, with the majority belonging to females (59%).
SARS-CoV-2 S protein SNP analysis identified mixed alleles indicative of heterogeneous populations or quasispecies among 5.3% of cases (which define heterogeneous populations or quasispecies). Heterogeneous genotypes were detected mainly at one or more of the S protein aa positions: 144, 484, 501 and 681, where P681H was the most frequently observed. Overall, sequencing-based variant calling analysis identified 8.8% heterogeneous infections among our study population are heterogeneous infections, which included 3 samples analysed by SNP PCR. These results contrast with a study conducted in 2020 in Australia which reported quasispecies diversity in S protein at a frequency of 12% (13/107), although their sample size was much smaller [bib_ref] Intra-host diversity of sars-cov-2 should not be neglected: Case of the state..., Armero [/bib_ref]. We were unable to ascertain if study participants were infected with mixed virus populations, had multiple exposures/infections closely timed, or if the mixed populations observed were due to de novo evolution in the infected host during the course of illness.
The lower frequency of heterogeneity reported here for SNP-based approaches could be ascribed to the lower sensitivity of SNP assays. In some cases, more than 50% of samples could not be analysed for SNPs despite acceptable cycle threshold values obtained with the diagnostic tests. Of the heterogeneous SNPs identified, only aa position 144 was truly heterogeneous from sequencing. P681H was the dominant heterogeneous SNP observed during the tail-end of the 3 rd COVID-19 wave, but sequencing confirmed that this position had the wild type or mutant at position 681, but not both. Amino acid mutations that usually occur adjacent to each other in the SARS-CoV-2 genome, including P681H and P681R, make genotyping analysis challenging since the probes in the assay for one mutation failed to bind to viral sequence with the adjacent mutations [bib_ref] SARS-CoV-2 testing in the community: Testing positive samples with the TaqMan SARS-CoV-2..., Ashford [/bib_ref]. Mutations at this position appear as mixed (wild type/ mutant) and moved away from wild type/wild type or mutant/mutant assignment as a result of weak amplification due to non-specific probe activity was evident. The latter study tested K417N, L452R, E484K, N501Y, and P681R; and only identified heterozygosity for E484K (1/ 42; 0.6%) and P681R .0%) from the SNP assays, not observed in our findings. Another study that tested 13 S protein TaqMan SNP assays, excluding P681H but including additional T20N, del242, and T478K; showed a 100% concordance when comparing the SNP assay to sequencing data [bib_ref] SARS-CoV-2 variants detection using TaqMan SARS-CoV-2 mutation panel molecular genotyping assays, Neopane [/bib_ref]. However, only 9/150 samples were confirmed using sequencing, and comparisons were based on lineage/VOC calling rather than the frequency of wild type and/or mutation.
Of the samples sequenced, heterogeneity was primarily identified at aa positions 19 (T19IR and T19KR), 371 (S371FP), and 484 (E484AK, E484AQ, E484KQ), similar to the SNP PCR which identified heterogeneity of the S protein at the aforementioned aa positions, before the 3 rd wave through to the 5 th wave of infections in South Africa. We identified T19IR among individuals that tested during CST and had Omicron BA.2.15 infection, while the lineage information for those with T19KR mutations was unknown due to missing sequence data. Amino acid position 19 is present in the N-terminal domain (NTD) of the S gene, and mutations at this position previously identified in Delta and Omicron BA.2 [bib_ref] Emergence of Omicron third lineage BA.3 and its importance, Desingu [/bib_ref]. The site I ("supersite") on the NTD is usually the target for NTD-specific neutralising antibodies. However, the T19R mutation disrupts S protein conformation, which reduces virus neutralisation by antibodies [bib_ref] A comprehensive overview of identified mutations in SARS CoV-2 spike glycoprotein among..., Eslami [/bib_ref]. When comparing the mutations observed in our study to the global context, high frequencies of T19I were observed among Omicron BA.2 lineages and T19R among Delta. T19K was observed at 0.01% and in Alpha VOC that circulated at low frequencies in South Africa [bib_ref] Identification of SARS-CoV-2 Omicron variant using spike gene target failure and genotyping..., Subramoney [/bib_ref].
Our data showed that S371FP was identified among Omicron BA.1, BA.2.15, and BA.4 lineages. S371F was observed at much higher frequencies compared to S371P, both of which were identified in the Omicron BA.1 and BA.2. S371L was not associated with heterogeneous infections in our study but was identified as the dominant mutation globally among Omicron BA.1 lineage strains. Mutations at position S371 led to a reduced recognition of the receptor-binding domain (RBD) antibodies [bib_ref] Emergence of Omicron third lineage BA.3 and its importance, Desingu [/bib_ref]. Similar to our study, S371P, and S371F were reported among Omicron BA.2 lineages in Europe and Italy, and in BA.3 lineage which was sporadic in South Africa [bib_ref] Emergence of Omicron third lineage BA.3 and its importance, Desingu [/bib_ref]. Despite the fact that the region around S371 to F541 does not have any N-linked glycosylation sites, aa changes at position 371 still decrease neutralising antibody recognition [bib_ref] Recombinant antigens based on non-glycosylated regions from rbd sars-cov-2 as potential vaccine..., Núñez-Muñoz [/bib_ref]. S371P is recognised as a mutation that results in virus escape from neutralisation by antibodies [bib_ref] A potent SARS-CoV-2 neutralising nanobody shows therapeutic efficacy in the Syrian golden..., Huo [/bib_ref] , while S371F alters the conformation of the RBD which results in neutralising antibody escape [bib_ref] A structural dynamic explanation for observed escape of SARS-CoV-2 BA.2 variant mutation..., Miller [/bib_ref]. The most frequently observed heterogeneity was among individuals in age groups 5-14 (22.0%), 15-24 (23.3%), and 25-44 (32.9%). Although limited information on HIV-1 status among study participants including out-patients that visited ARV clinics and had heterogeneous mutations, recent data showed that the HIV-1 prevalence among South Africans aged 15 to 49 years remains high at 19.5%. A shedding study done in South Africa showed that persons living with HIV/AIDS with low CD4 counts (<200cells/μl) had significantly longer periods of SARS-CoV-2 shedding, ranging between 7 and 43 days, during the 2020 pandemic period in South Africa [bib_ref] Prolonged Shedding of Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) at High..., Meiring [/bib_ref]. A longer period of virus shedding provides an opportunity for the virus to evolve in its efforts to evade the SARS-CoV-2 specific immune responses [bib_ref] COVID-19 in an immunocompromised host: persistent shedding of viable SARS-CoV-2 and emergence..., Leung [/bib_ref] [bib_ref] Enhanced in fl ammation and suppressed adaptive immunity in COVID-19 with prolonged..., Tang [/bib_ref] [bib_ref] Prolonged viral shedding of SARS-CoV-2 and related factors in symptomatic COVID-19 patients:..., Long [/bib_ref]. Omicron BA.1 was initially detected in HIV-infected patients in Botswana, South Africa, suggesting that individuals with underlying medical conditions such as HIV infection are at a higher risk of SARS-CoV-2 infection and severe disease as well as a possible predisposition for evolving SARS-CoV-2 mutants [bib_ref] SARS-CoV-2 infection in HIV-infected patients: potential role in the high mutational load..., Tarcsai [/bib_ref].
We identified E484AK, E484AQ, and E484KQ heterogeneity at very low frequencies among Delta and Beta variants from CST persons and an ARV clinic patient. Unlike our study, E484 mutations were not previously detected in Delta but we could speculate that the virus underwent adaptive changes since E484AK/AQ/KQ were identified among cases just after the Delta (3 rd ) wave, leading towards the Omicron wave. Globally, E484A was observed at the highest frequency and was primarily observed in Omicron BA. [bib_ref] Human Coronavirus : Host-Pathogen Interaction, Fung [/bib_ref] [bib_ref] Competent immune responses to SARS-CoV-2 variants in older adults following two doses..., Jergović [/bib_ref] [bib_ref] Waves and variants of SARS-CoV-2: understanding the causes and effect of the..., Thakur [/bib_ref]. The E484A, E484K, and E484Q mutations reduce sensitivity to vaccine-induced antibodies through immune escape from class 2 neutralising antibodies, with increased transmissibility due to higher affinity for the host receptor [bib_ref] Competent immune responses to SARS-CoV-2 variants in older adults following two doses..., Jergović [/bib_ref] [bib_ref] Two-step fitness selection for intra-host variations in SARS-CoV-2, Li [/bib_ref] [bib_ref] SARS-CoV-2 Omicron-B.1.1.529 leads to widespread escape from neutralizing antibody responses, Dejnirattisai [/bib_ref]. Of note is that E484 was the only site for which vaccine data was available compared to T19 and S371. The E484Q mutation is associated with reduced resistance to neutralisation by post-vaccination sera, and two doses of mRNA vaccines provide protection against E484Q variants in adults [bib_ref] Competent immune responses to SARS-CoV-2 variants in older adults following two doses..., Jergović [/bib_ref] [bib_ref] Waves and variants of SARS-CoV-2: understanding the causes and effect of the..., Thakur [/bib_ref]. E484K has shown evidence of neutralisation by the monoclonal antibodies bamlanivumab and etesevimab as well as convalescent sera, but reduced neutralising antibody activity was observed among Pfizer-BioNTech and Moderna vaccine recipients [bib_ref] Waves and variants of SARS-CoV-2: understanding the causes and effect of the..., Thakur [/bib_ref].
Other studies reported significant non-synonymous SNVs in the S gene reported early in the pandemic during 2020, which include mutations at nucleotide position G22899T, which translates to amino acid mutation G446V in the RBD, and G24557T which translates to a G999C mutation in the heptad repeat region associated with membrane fusion during virus entry [bib_ref] Intra-host diversity of sars-cov-2 should not be neglected: Case of the state..., Armero [/bib_ref]. Similarly, another study identified nucleotide substitutions at positions 22899 (translates to G446V), 21575 (L5F associated with increased infectivity in-vitro), and 24198 (A879V with reported reduced sensitivity to neutralisation by convalescent sera in-vitro) which are present in the RBD of the S protein [bib_ref] SARS-CoV-2 variants detection using TaqMan SARS-CoV-2 mutation panel molecular genotyping assays, Neopane [/bib_ref]. However, these mutations were not observed in our study. Several mutations in the spike protein are responsible for neutralising antibody escape. Group A-D antibodies comprise the strongest antibodies which disrupt binding to the ACE-2 receptor and escape neutralisation from Omicron lineages [bib_ref] Omicron escapes the majority of existing SARS-CoV-2 neutralizing antibodies, Cao [/bib_ref]. Omicron lineages are associated with escaping neutralisation from various antibody groups. Group D antibodies were affected by mutations at positions 440, 444, 446, and 448 found in the Omicron lineages [bib_ref] Omicron escapes the majority of existing SARS-CoV-2 neutralizing antibodies, Cao [/bib_ref]. While group E neutralising antibodies are sensitive to mutations at positions including 339, 345, and 346, with the frequency of R346K mutations rapidly increasing with emerging Omicron lineages. The primary mutations known to evade neutralisation by antibodies in Beta and Delta include the mutations at positions 484 and 478 [bib_ref] Reduced sensitivity of SARS-CoV-2 variant Delta to antibody neutralization, Planas [/bib_ref] [bib_ref] Evidence of escape of SARS-CoV-2 variant B.1.351 from natural and vaccine-induced sera, Zhou [/bib_ref]. Of the described mutations, we identified changes at aa homogeneous positions 339, 478, and 484 where 484 was also heterogeneous.
Heterogeneity among our study samples was predominantly observed from June 2021 through to May 2022 covering the Delta to Omicron dominant waves. It was also evident among individuals that sought CST and were expected to be asymptomatic or present with mild influenza-like illness symptoms. However, in this study the majority of cases were from CST for which clinical data were not collected; therefore, we could not conclude that heterogeneous SARS-CoV-2 infections were related to the severity of the disease. Another contributing factor to the observed heterogeneity may be immune selective pressures and adaptation when virus transmission occurs between persons with different HLA backgrounds [bib_ref] Intra-host diversity of sars-cov-2 should not be neglected: Case of the state..., Armero [/bib_ref].
The first limitation of the study was the performance of the T20N, P681H, E484K, and E484Q genotyping assays. The cycle-threshold (Ct) values, of samples from diagnostic SARS-CoV-2 testing, were compared to the genotyping assay results to determine the success of each assay. We genotyped most samples that tested positive for SARS-CoV-2 with Ct-values ranging from 13 to 37 from diagnostic PCR assays. However, it is important to note that only two of these assays relate to heterogeneous SNP positions identified by variant calling analysis. We, therefore, concluded that the reduced sensitivity of some Thermo Fisher SNP assays to distinguish wild-type from mutant might be due to the negative impact of other mutations surrounding the SNP of interest [bib_ref] SARS-CoV-2 testing in the community: Testing positive samples with the TaqMan SARS-CoV-2..., Ashford [/bib_ref]. Secondly, incomplete or low-quality S gene sequence coverage was generated from whole genome based NGS due to primer mismatches as a result of novel variants emerging and becoming dominant in each wave of infection. We acknowledge that this problem could have been overcome by updating amplification primers or spiking the primers, but we did not re-sequence samples due to cost. Thirdly, our study is not a longitudinal study but rather describes the heterogeneity across the entire population, looking at the diversity and dominance of heterogeneous mutations within lineages and VOCs over time. Fourth, samples included in this study were primarily from the Gauteng Province, although the lineages and VOCs circulated similarly to that observed in the other Provinces of South Africa, we could not assume that similar heterogeneity would occur throughout the country. A fifth limitation was that we did not have the clinical data at the time of the study for individuals that were tested as part of community testing and from out-patients and encompassed most of the cohort. Lastly, we did not perform antibody neutralisation assays to determine the significance of immune adaptations during the different waves of infection among individuals infected with the heterogenous mutations detected.
# Conclusion
Unlike other intra-host diversity studies that focused on mutations of the SARS-CoV-2 genome, our study analysed the S protein aa changes as it is the primary target for neutralising antibodies and this focus for viral immune evasion. Substitutions at aa positions 19, 371, and 484 have previously been shown to reduce virus neutralisation by antibodies. However, we are uncertain if the impact is less or greater among persons infected with SARS-CoV-2 that display heterogeneity in the S protein. Low-frequency variants at heterogeneous aa positions across the S protein provide opportunities for the emergence of SARS-CoV-2 variants that can overcome or outcompete when conditions favour survival and transmission. Our findings show that heterogeneous mutations may be a result of intra-host adaptation and may drive the evolution of the S protein which could contribute to immune evasion. This may contribute to the ongoing emergence of new variants associated with continued outbreaks globally. Accounting for diversity in vaccine design and universal vaccine design strategies may be necessary to ensure effective vaccines to protect against all variants of SARS-CoV-2.
## Supporting information
[fig] Fig 1: SARS-CoV-2 detection rate among respiratory samples tested from June 2020 to May 2022. A) Blue bars represent the number of samples received for SARS-CoV-2 diagnostics and the red line graph represents the detection rate over time. B) Green bars represent the number of SARS-CoV-2 samples that were tested by SNP PCR and/or sequenced and the red line graph represents the detection rate over time.https://doi.org/10.1371/journal.pone. [/fig]
[fig] Fig 2: Frequency of the heterozygous SNPs identified from June to October 2021 (N = 50). X-axis represents the date during which the heterozygous SNP was identified. Y-axis: the line graph represents the number of samples that were observed with a particular heterozygous SNP, and the bar graphs represent the total number of samples (N) in which the SNP was observed on each day.https://doi.org/10.1371/journal.pone.195; 22.1%), 15-24 (48/195; 24.6%), and 25-44 (58/195; 29.7%) were the predominant age groups with S371FP (S1 and S2 Tables). Three percent (5/195) of out-patients visited ARV clinics. Two cases with multiple heterogeneous mutations which included S371FP/P681HR were classified as B.1.1 and Omicron BA.1 lineages; detected in November 2021 and December 2021 from CST, respectively. Whereas T19IR/S371FP/P681HR and S371FP/D389EN mutation profiles were identified in March 2022 among sequences from 2 cases classified respectively as Omicron BA.2.15 and Omicron BA.4 from CST [/fig]
[fig] Fig 3: Box and whisker plot showing the mutant allele frequencies (AF) at amino acid positions within the S protein that have more than 1 amino acid change. Circles represent the outliers, line within the box represents the mean AF. Mutations (T19K, P681S D950, D950K, and D950R) that only have a single AF value were excluded from the figure. https://doi.org/10.1371/journal.pone.in the majority (34.1%) of global sequences from November 2021 to May 2022 and classified as Omicron BA.1 and BA.2 lineages, similar to our findings but also identified among Omicron BA. 4 and BA.5 (22B clade). Whereas E484K was identified in 2.3% of global sequences belonging to all lineages except Omicron lineages. E484Q was observed among 0.2% of sequences from Beta and Delta lineages detected from May 2021 to December 2022. [/fig]
[fig] Fig 4: Fig 4. Frequency of the heterozygous mutations (4A) and VOC/lineages (4B) identified from sequence data over time. 4A) X-axis represents the date during which the heterozygous SNP was identified, and the y-axis represents the percentage of samples that were observed with a particular heterozygous SNP. 4B) X-axis represents the period during which the VOC/lineages were identified, and the y-axis represents the percentage of samples with a specific VOC/lineage. Omicron BA.1, BA.1.1, BA.1.10, BA.1.13, BA.1.14, BA.1.17, BA.1.17.2, BA.1.18, BA.1.19 and BA.1.21 were grouped as Omicron BA.1. [/fig]
[fig] Fig 5: Prevalence of samples with S371FP over time. The black line graph represents the total number of samples with S371FP heterogeneity each month. The bar graphs represent the SARS-CoV-2 VOCs and lineages observed with the S371FP heterogeneous mutation from 2021 to 2022. https://doi.org/10.1371/journal.pone.0286373.g005 [/fig]
[fig] Fig 6: Frequency of mutations occurring at aa positions 19, 371, and 484 in the S protein. Custom NextStrain time-tree (left) displays the sequence data from this study from June 2020 to May 2022, with specific aa changes at the heterogeneous positions. The phylogenetic tree (middle) represents the global SARS-CoV-2 sequence data from January 2020 to May 2022, with specific aa changes at the heterogeneous positions (https://nextstrain.org/ncov/gisaid/global/ all-time, accessed 22 June 2022). Clades displayed in phylogenetic trees represent the following SARS-CoV-2 pangolin lineages or VOCs (https://covariants. org/variants): 20A and 20C include B.1 lineages, 20B includes B.1.1 and its sub-lineages; 20D includes B.1.1.1 lineage, Delta 21A, and Delta 21J includes B.1.617.2 and AY lineages; Alpha 20I includes B.1.1.7; Omicron 21M includes B.1.1.529 lineage; Omicron 21K includes BA.1 lineage; Omicron 21L includes BA.2 lineages and branches out to Omicron 22A (BA.4) and Omicron 22B (BA.5). Top right graph shows the frequency of the aa changes over time (https:// nextstrain.org/ncov/gisaid/global, accessed 22 June 2022). Tables show the frequency of wild-type aa and mutations present globally from GISAID (https:// www.gisaid.org/epiflu-applications/covsurver-mutations-app/, accessed 22 June 2022). https://doi.org/10.1371/journal.pone.0286373.g006 [/fig]
[fig] S1: Fig. Distribution of Ct-values for each SARS-CoV-2 mutations tested. (TIF) S2 Fig. Allelic discrimination plots of mutants, wild type and heterogeneous SNPs (mutant/wild type). (TIF) [/fig]
[table] Table 2 ;: Fig 2). These included the E484Q (2/948; 0.2%), N501Y (1/948; 0.1%), P681H (45/511; 8.8%) and delY144 (2/331; 0.6%). P681R/L452R/P681H* profile of mutations was primarily identified from 17 July to 20 August 2021 and 13 September 2021, whereas the P681R/L452R/P681H*/D614G mutation profile was observed from 9 to 17 September 2021(Fig 2). Samples with heterogeneous SNPs that were sequenced confirmed heterogeneity at positions 144 (YY144Y, AF = 0.2) and 484 (E484Q, AF = 0.4).We were unable to determine the SNP for 2.0% (21/948) of samples (S1 and S2 Figs) on any of the SNP assays. In addition, undetermined results (negative or no amplification for a SNP) were obtained for 67.5%(345/511) of samples analysed with the P681H SNP assay, 66.3% (479/722) for E484K, 59.0% (196/332) for T20N, 53.2% (505/948) for E484Q, 20.7% (196/948) for L452R, 20.5% (179/873) for del69/70, 17.2% (163/948) for N501Y, 16.0% (152/948) for K417N, 11.2% (27/332) for D614G, 8.1% (77/948) for P681R, and 7.3% (24/331) for delY144. This is despite original PCR diagnostic results with ct values between 13 and 37 (S1 Fig). [/table]
[table] Table 2: Summary of SNP combinations with heterogeneous amino acids.*heterogeneous amino acids. Poor quality: a VOC or lineage could not be determined due to poor sequence.1 Sequence data absent at the amino acid position of interest https://doi.org/10.1371/journal.pone.0286373.t002 [/table]
[table] Table: Patient status for cases where heterogeneity in SARS-CoV-2 S protein was observed. (TIF) S2 Table. Age groups for cases for which heterogeneity in S was identified. (TIF) S3 Table. Allele frequency and prevalence of homogeneous mutations identified among all individuals in the population. (TIF) S4 Table. Accession numbers for fastq data for heterogeneous samples. (TIF) [/table]
|
10.1155/2016/7054276
|
CCBY
|
4745977
|
26904137
|
s2orc_pubmed_articles
|
Noise Exposure and Hearing Capabilities of Quarry Workers in Ghana: A Cross-Sectional Study
Introduction. Although quarry operations have high economic significance, the effects they cause to the workers in terms of excessive noise production cannot be overlooked. This cross-sectional study assessed the extent of noise exposure and its influence on hearing capabilities among quarry workers in Ashanti region. Methods. The study involved 400 workers randomly selected from five quarries in Ashanti region from April to June 2012. Data was collected using structured questionnaires, physical examination, and audiological assessments. A logistic regression model was fitted to assess independent predictors of hearing loss. Results. All the machines used at the various quarries produced noise that exceeded the minimum threshold with levels ranging from 85.5 dBA to 102.7 dBA. 176 (44%) of study respondents had hearing threshold higher than 25 dBA. 18% and 2% of these were moderately (41-55 dBA) and severely (71-90 dBA) impaired, respectively. Age, duration of work, and use of earplugs independently predicted the development of hearing loss. Use of earplugs showed a protective effect on the development of hearing loss (OR = 0.45; 95% CI = 0.25, 0.84). Conclusion. This study provides empirical evidence on the extent of damage caused to quarry workers as a result of excessive noise exposure. This will support the institution of appropriate protective measures to minimize this threat.
# Background
Occupational health is an important public health concern for the working population. Various aspects of the working environment could expose an individual to potential hazardous elements. Noise is considered as one of these potential hazards and it is currently seen as a global health concern. Noise, which is defined loosely as annoying sounds, is part of the everyday human activity. Excessive noise beyond tolerated levels from all these sources is hazardous and could cause hearing impairment. This is a widespread occupational hazard, which could result in noise-induced hearing loss (NIHL). Other associated health effects include elevated blood pressure, sleeping difficulty, annoyance and stress, and temporary threshold shift (TTS) [bib_ref] Air bag deployment and hearing loss, Morris [/bib_ref] [bib_ref] Selected occupational risk factors, " in Comparative Quantification of Health Risks: Global..., Concha-Barrientos [/bib_ref].
The growing attention for NIHL is due to the fact that, unlike many injuries or illnesses, hearing loss may be permanent and irreversible. Exposure to excess noise is first observed as an increase in the threshold of hearing (threshold shift), as assessed by audiometry [bib_ref] Selected occupational risk factors, " in Comparative Quantification of Health Risks: Global..., Concha-Barrientos [/bib_ref]. This is defined as a change in hearing thresholds of an average 10 dB or more at 2000, 3000, and 4000 Hz in either ear (poorer hearing). There are two types of NIHL, a temporary threshold shift, which is a temporary loss of hearing, and a permanent threshold shift, which involves a shift in the person's ability to hear soft sounds. This is as a result of long-term exposure to loud sounds of slightly lower intensity, such as factory noise or rock music.
Although noise is associated with almost every work activity, some activities are associated with particularly high 2 Journal of Environmental and Public Health levels of noise. In general, sounds above 85 dB are considered harmful, depending on how long and how often one is exposed to them and whether you wear hearing. Previous literature shows that workers in mines, quarries, sawmills, textile factors, printing presses, and many others work with machines that produces noise much higher than the tolerated levels and therefore expose workers to potential hearing loss [bib_ref] Noiseinduced hearing loss among quarry workers in a north-eastern state of Malaysia:..., Ismail [/bib_ref] [bib_ref] Industrial noise pollution and its effects on the hearing capabilities of workers:..., Boateng [/bib_ref] [bib_ref] Noise levels in the Ethiopian woodworking industries, Mulugate [/bib_ref]. In the large coal industry, about 76% are exposed to hazardous noise. This causes about 25% of severe hearing problems and about 80% of hearing impairment in the workers' retirement age.
Although the situation could be improving in developed countries as a result of more widespread appreciation of the hazard and the institution of protective measures, evidence from developing countries suggests that average noise levels are well above the occupational level recommended in many developed nations [bib_ref] Standards and regulations, Suter ; E [/bib_ref]. Increasing industrialization might exacerbate this situation in developing countries and therefore the need to assess the industrial noise pollution and its impact on the hearing capabilities of workers in such areas.
In Ghana, there is very little data on the effect of noise exposure on quarry industrial workers. More so, the increasing number of patients seeking medical attention for their ears at Hearing Assessment Centre at the Komfo Anokye Teaching Hospital (KATH) in Kumasi between 1995 to 1998 [bib_ref] Hearing-impairment among workers in a surface gold mining company in Ghana, Amedofu [/bib_ref] has called for the need to focus attention on noise pollution on workers in the industrial settings such as quarry industries. Furthermore, no study has focused specifically on occupational hearing loss among workers of quarry industries even though the impact of noise pollution on hearing in mining company has been studied [bib_ref] Hearing-impairment among workers in a surface gold mining company in Ghana, Amedofu [/bib_ref]. This study seeks to provide empirical evidence on the extent of noise exposure and its influence on hearing capabilities of workers in the quarry industry. This would help in formulating policies and institute preventive measures that will help minimize the risk of occupational hearing loss among the exposed population.
# Methodology
## Study design and settings.
This was a descriptive crosssectional study conducted in Kwabre East District of Ashanti region of Ghana. Kwabre District, carved out of the former Kwabre Sekyere District in 1988, is located almost in the central portion of Ashanti region. It is within latitude 6 ∘ 44 north and longitude 1 ∘ 33 to 1 ∘ 44 west. The district shares common boundaries with Afigya-Sekyere District to the north, Kumasi metropolitan area to the south, Ejisu-Juaben District to the southeast, Atwima District to the west, and Offinso District to the northwest. The District has a total land area of 246.8 square kilometers constituting about 1.01% of the total land area of Ashanti region. Kwabre District is part of the Greater Kumasi City region, which is made up of Kumasi Metropolitan Area and the surrounding Districts. According to 2000's population and housing census, the district has a total population of about 205,372 people. The physiography of the area consists of scarps formed by the extensive quarry activities going on in the area. The main land usage in the area can be categorized into stone quarrying and agriculture. The quarries are mainly located near the mountains, where granite deposits have been detected. The quarry developers lease the land from the owners at a fee until the time the quarry work is finished. The main commercial purposes of quarrying are building stones and quarry dust. Agricultural activities in the area are mostly subsistence and cash crop farming.
## Study
Population and Sampling. The study population was drawn from workers of the five (5) quarry companies, namely, Northern mine quarry, A. Kannin mine, Taysec quarry company, KAS quarry company, and Siemens quarry company. Workers who have been working at the respective quarry for more than six months were included in the study. Workers who had previous history of hearing loss and those not directly involved in the quarrying process were excluded. The study comprised 400 workers from the various quarry companies. The composition of respondents from the quarry companies was as follows: A. Kannin (120 respondents; 20.1%), KAS quarry (80 respondents; 13.4%), Siemens quarry (60 respondents; 10%), Taysec (40 respondents; 6.7%), and Northern mines and quarry (98 respondents; 16.4%).
The surveys were conducted from April to June 2012. Based on the average number of clients per day for a specific facility and the desired sample size, a recruitment interval, , was developed for each quarry. Based on this interval, the th worker was interviewed on each interview day until the required sample size was reached for each quarry. The administration of the study was focused mainly on the industrial noise pollution and its impact on the hearing capabilities of workers. The desired sample size was calculated based on Kirkwood and Sterne, assuming a 40% proportion for the event of interest and 10% nonresponse rate.
## Data collection techniques and instruments.
Data were collected mainly through interviewing. Participants were given consent form to sign and had all their concerns and questions answered before data collection began. All tools employed in the research were developed using standard procedures, pretested, and revised to ensure their validity and reliability and research assistants were trained to ensure uniformity in the administration of questionnaires. The various instruments used for measuring sound levels were the Casella sound level meter (in the recording of the noise levels in the various companies) and pure-tone audiometer (to evaluate the hearing threshold of subjects).
## Casella sound level
Meter. The weighted networks in a meter are electronic circuits whose response to low frequencies and to very high frequencies is reduced in a specified manner. In general, four different weightings have been standardized internationally as A, B, C, and D. Network A is used in industrial setting and this was used throughout the study.
To check for vibration, shock, and excessive heat, a calibrator was attached to the microphone of the meter and the reading was then compared with the calibrator's value. The meter was adjusted when required to bring it into calibration.
## Pure-tone audiometry.
A biologic calibration of the pure-tone audiometer was performed daily using supposedly normal people with normal hearing prior to any audiological evaluation of the subjects. An otoscopic examination was also performed on each subject to exclude wax or any discharge in the ear canal or perforation of the eardrum to rule out possibility of conductive hearing loss. To overcome bias on the hearing acuity by ambient noise influence, five (5) of the subjects each from the various companies were randomly selected to undergo audiometric retesting at Hearing Assessment Centre at Komfo Anokye Teaching Hospital (KATH). The hearing thresholds obtained for these groups were found not to be different from those obtained within the premises of the factories. Audiometric tests were performed on employees working in the area with noise level exceeding the threshold limit value of 85 dBA using screening Audiometer, (AS608) obtained from the Hearing Assessment Centre of the Department of Ear, Nose and Throat (ENT) of KATH. Hearing acuity was measured at 5 dBA interval over a range of octave band frequencies from 500 to 8000 Hz. Hearing was considered normal if the threshold level was less than or equal to 25 dBA at the frequencies 250, 500, 1000, 2000, 4000, and 8000 Hz. However, the intensity of the stimuli was increased beyond 25 dBA at any frequency until a response was obtained. Individuals with a characteristic notch of four (4) KHz depicting the classical sign of NIHL were analysed. The degree and type of hearing loss were also determined using Goodman's [bib_ref] Reference levels for pure-tone audiometer, Goodman [/bib_ref] and Carhart's [bib_ref] An improved method for classifying audiograms, Carhart [/bib_ref] approaches, respectively. Both ears of each subject were tested to establish pure-tone hearing sensitivity. All subjects were tested at the beginning of each work shift to ensure that those whose hearing had been "fatigued" might have gained some recovery after being away from the noise exposure.
## Description of variables and data
Analysis. The outcome variable was hearing threshold (dBA). Hearing threshold beyond 25 dBA is classified as having hearing loss and hearing loss was further classified based on its severity (26-40 dBA, mild; 41-55 dBA, moderate; 56-70 dBA, moderate-to-severe; 71-90 dBA, severe; and above 90, profound). The explanatory variables were age, duration of exposure, gender, and use of earplugs. The response of each subject and the data obtained from the administration of the physical instruments were scrutinized carefully and categorized in tables and graphs. Data was analyzed with R 3.1.1. The analysis involved a description of the baseline characteristics of respondents and the noise levels measured on various machines at the quarries. A chi-square test was conducted to see the association between workplace and hearing loss. Logistic regression analysis was done to look at the influence of age, duration of exposure, and use of earplugs on the odds of hearing loss among the quarry workers. Significance of associations was at value of <0.05.
## Model estimation and assumptions.
A correlation between age and duration of exposure was assumed and tested using Pearson's correlation. This was because the variable appeared to be approximately normally distributed. The explanatory variables included in the model were also tested for collinearity using the variance inflation factor (VIF). The choice of best-fitted model was based on Akaike's Information Criterion (AIC). This was proposed by Akaike [bib_ref] A new look at the statistical model identification, Akaike [/bib_ref] as a measure of the goodness of fit of an estimated statistical model. It takes into account both the statistical goodness of fit and the number of parameters that have to be estimated to achieve this particular degree of fit by imposing a penalty for increasing the number of parameters. The AIC is given as −2 * (log likelihood) + 2 * (number of parameters in the model), that is, −2 + 2 . Lower AIC is an indication of a better likelihood.
# Ethical
# Results
## Background characteristics of respondents.
Majority of the respondents were males (81.4%). The mean age (SD) of the respondents was 41.7 . Most of the respondents were Christians (59%) and about 29% were Muslims. With respect to their level of education, 47.2% had junior secondary or middle school education. About 17% had tertiary education and 14 (2.3%) had no formal education. Majority had worked in the quarry for up to 10 years and 24.7% had worked in the quarry for less than 5 years. 132 (33%) of the respondents used earplugs and 61% of respondents who wore no earplugs had hearing thresholds above 25 dB as against 36% among those who wore earplugs.
## Noise measurement at various facilities under study.
All the companies had different production units with more or less the same type of machinery. [fig_ref] Table 1: Measured noise levels [/fig_ref] displays the noise levels obtained from the machines in the various companies. The measurement values range from 85.5 dBA to 102.7 dBA, demonstrating that the noise levels produced exceed the limiting threshold level of 85 dBA. It was realized that all the five companies visited produced an excessive amount of noise capable of damaging the hearing status of the workers.
## Hearing thresholds at various quarries.
The mean hearing threshold among all workers was 27.32 dBA and 176 respondents (44%) had hearing thresholds higher than 25 dBA. Comparatively, a higher mean threshold was observed among workers at Taysec company (29.98 dBA) followed by Northern mines and quarry (26.43 dBA) [fig_ref] Table 2: Hearing threshold levels among various groups of respondents [/fig_ref]. More than 50% of respondents from Taysec company and KAS company had hearing threshold of more than 25 dBA (58% and 69%, resp.). The proportion of respondents with hearing threshold > 25 dBA was comparatively low among A. Kannin and Siemens (29% and 33%, resp.). This difference in hearing thresholds among respondents from various working environments was statistically significant ( < 0.001), indicating the influence of the working environment on the hearing threshold of a respondent. As shown in , majority of the respondents had mild hearing loss (132 out of 176 respondents). Thirty-two respondents (18%) had moderate hearing loss, whereas 4 (2%) respondents had severe cases of hearing loss. [fig_ref] Table 3: Logistic regression analysis of predictors of hearing loss [/fig_ref] presents results of the univariable and multivariable logistic regression analysis of factors influencing hearing loss among quarry workers. Two models were fitted in the multivariable analysis. In model 1, the place of work was adjusted for, while model 2 involved adjusting for place of work as well as adjusting for family history of hearing loss. The scatterplot of age and duration exposure showed a positive correlation, which was slightly moderate ( = 0.465) . The VIF for all variables (shown under [fig_ref] Table 3: Logistic regression analysis of predictors of hearing loss [/fig_ref] were very low and implied little or no existence of multicollinearity. To this end, all variables and an interaction between age and duration of exposure were included in the multivariable analysis. Based on the AIC, the best-fitted model (i.e., the model with the lowest AIC) in the multivariable analysis was a model without gender. Age, duration of exposure, and use of earplugs significantly predicted hearing impairment among quarry workers [fig_ref] Table 3: Logistic regression analysis of predictors of hearing loss [/fig_ref]. In the univariable analysis, an increase in the age of quarry workers by one year resulted in 8% increase in the odds of hearing impairment (OR = 1.08; 95% CI = 1.05, 1.11). The odds of hearing impairment with respect to age increased to 1.15 in the multivariable analysis, where the place of work was adjusted for, and it remained unchanged in model 2. In model 1, an increase in duration of working at the quarry by one year results in about two times increase in the odds of hearing impairment (OR = 1.96; 95% CI = 1.2, 3.23). The strength of association, however, increased in model 2, indicating a possible confounding effect of the variables adjusted for in the association between duration of exposure and hearing impairment. The use of earplugs showed a protective effect on the odds of having hearing loss.
## Influence of age, duration of exposure, and use of earplugs on hearing capabilities of quarry workers.
In model 1, use of earplugs resulted in 54% decrease in odds of having NIHL (OR = 0.46; 95% CI = 0.26, 0.80) and this did not change much in model 2 (OR = 0.40). The interaction between age and duration of exposure was just significant with value of 0.05.
# Discussion
Occupational health is an important concern of the working person. Various elements concerning a person's working environment can predispose one to developing a disease process. Quarries are such organizations with high noise production levels as a result of their activities. The aim of this study was to look at NIHL among quarry workers in Ashanti region of Ghana. This study found a high prevalence of hearing impairment among the quarry workers and all machines used at the various sites produced noise levels that exceed the limiting threshold.
## Noise measurement at various facilities under study.
Results from this study indicate that all companies studied have different production units with similar type of machinery and the noise levels ranged from 85.5 dBA to 102.7 dBA. This reveals that all the machines used at the various companies produced sound that exceeds the minimum threshold of 85 dBA as recommended by World Health Organization, thereby having the potential of damaging the hearing status of workers. These noise levels also exceed the limit set by the Ghana Environmental Protection Agency (EPA) under the EPA Act of 1994 (Act 490) [18] which permits light industrial areas 70 dB during the day and 60 dB at night and heavy industrial areas 70 dB noise during the day and 70 dB noise at night. Consistent with this study, other studies from different countries reported exposure to high levels of noise pollution at the workplace. This includes the study by Ismail et al. [bib_ref] Noiseinduced hearing loss among quarry workers in a north-eastern state of Malaysia:..., Ismail [/bib_ref] among quarry workers in Malaysia, where sound levels exceeded the level that may cause NIHL to the workers.
These, however, might not be limited to quarries alone. Studies on noise levels from other work settings including the study in Ghana by Boateng and Amedofu [bib_ref] Industrial noise pollution and its effects on the hearing capabilities of workers:..., Boateng [/bib_ref] on the impact of noise levels on hearing capabilities of workers in saw mills, corn mills, and printing houses revealed that noise level in corn mills exceeds the limiting value. Similarly, a study of working industries in Ethiopia by also reported noise levels higher than permissible levels of 90 dBA. Other studies from industrialized countries also indicate an overexposure to high noise levels among coalminers. These noise levels are, however, potentially hazardous and might result in hearing impairment among workers in that environment.
## Noise exposure and hearing
Loss. This study revealed that 44% of the quarry workers had hearing thresholds of more than 25 dB. This was, however, higher than the other prevalence reported from Sudan [bib_ref] Occupational deafness of workers in a heavy engineering industry of West Bengal,..., Manna [/bib_ref] , 30.6%, and from studies by Amedofu [bib_ref] Hearing-impairment among workers in a surface gold mining company in Ghana, Amedofu [/bib_ref] , 23.0%, and Boateng and Amedofu [bib_ref] Industrial noise pollution and its effects on the hearing capabilities of workers:..., Boateng [/bib_ref] , 23.0%. On the other hand, higher prevalence was observed by Chaddha and Singh (50.0%) [bib_ref] Survey of noise assessment and its effect on hearing of workers in..., Chaddha [/bib_ref] and Hong (60.0%) [bib_ref] Hearing loss among operating engineers in American construction industry, Hong [/bib_ref]. Three out of the five companies involved in the study also recorded a mean hearing threshold level of more than 25 dBA. All the companies had more than 25% of their respondents having NIHL with the percentage being highest among workers at KAS quarry. There was a significant association between the various working environments and HTL and this could be due to the type and quantities of noise generating equipment used in the various quarries and how these influence hearing capabilities. Analysis of the extent of hearing loss among respondents with HTL > 25 dB indicated that 75%, 18%, 5%, and 2% had mild, moderate, moderately severe, and severe hearing loss, respectively. These indicate the extent of ear damage caused to workers in severely noise-exposed environments like the quarry and the need to institute appropriate interventions to curb this.
## Relationship between age, duration of exposure, and use of earplugs and hearing loss among respondents.
Generally, there is an established association between the age of workers and hearing loss [bib_ref] The causes and prevalence of preschool deafness in Ghana, Amedofu [/bib_ref] [bib_ref] Joint effects of smoking, noise exposure and age on hearing loss, Ferrite [/bib_ref]. In this study, increasing age of quarry workers was associated with increasing odds of hearing loss among the quarry workers. This is consistent with the findings of an industry-specific study in USA [bib_ref] Joint effects of smoking, noise exposure and age on hearing loss, Ferrite [/bib_ref] , which showed that 90% of coal miners have hearing impairment by the age of 52 years. Also, it is estimated that 70% of male metal/nonmetal miners will have hearing impairment by the age of 60 years. In the study by Ahmed et al. [bib_ref] Occupational noise exposure and hearing loss of workers in two plants in..., Ahmed [/bib_ref] , age was the secondary predictor of hearing loss among workers in two plants in eastern Saudi Arabia.
The study also revealed a significantly positive association between duration of work at the quarry and hearing impairment. According to the United States National Institute of Deafness and Other Communicable Disorders, long or repeated exposure to sounds at or above 85 decibels can cause hearing loss. The louder the sound is, the shorter the time period before NIHL can occur becomes. Duration of exposure to noise at the working environment showed strong positive relationship with the hearing threshold among respondents in the multivariable analysis.
Wearing earplugs or other protective devices has been recommended for individuals involved in a loud activity. The proportion of respondents with hearing impairment was higher among those who did not wear earplugs. The multivariable analysis showed that use of ear protection decreased the odds of having hearing impairment among quarry workers. This is consistent with the outcome of the study by Hong [bib_ref] Hearing loss among operating engineers in American construction industry, Hong [/bib_ref] , which found an inverse relationship between hearing protector use and hearing status of the employee. Ahmed et al. [bib_ref] Occupational noise exposure and hearing loss of workers in two plants in..., Ahmed [/bib_ref] also showed that wearing hearing protection devices is among the important factors that influence the measured hearing threshold values at low frequencies. Subjects who did wear hearing protection devices had lower measured hearing threshold values than subjects who did not wear hearing protection. According to a WHO bulletin on environmental burden of disease [bib_ref] Selected occupational risk factors, " in Comparative Quantification of Health Risks: Global..., Concha-Barrientos [/bib_ref] , the first priority in minimizing hearing loss is to reduce noise through technical measures such as introducing hearing protection for workers when engineering controls are not applicable or are insufficient. It was, however, advocated that the protective equipment must be properly selected, worn, and maintained.
# Limitations
Although the study accessed the exposure to noise, there might still be other underlying factors, which might have contributed to hearing loss. These sources of noise produce what is called sociocusis and their effect on hearing loss might not differ from occupational hearing loss. The authors believed that the inclusion of variables on other sources of loud noise measured major external sources of noise that could contribute to noise induced hearing loss. There was also a possibility of recall bias with respect to previous history of hearing problems. Research assistants were however trained to ensure comparability in the administration of questionnaires across the study centers.
# Conclusion
This study found that most of the machines used in the quarry exceed the tolerable threshold of sound thereby having the potential of damaging the hearing status of workers. Noise levels measured among the quarries workers studied also exceed the limit set by the Ghana Environmental Protection Agency (EPA) under the EPA Act of 1994 (Act 490) which permits light industrial areas. It is recommended that EPA embark on regular monitoring to access noise levels and ensure that companies do not emit noise greater than the tolerable limits. Use of earplugs showed a protective effect on development of hearing loss. Efforts to ensure access and use of earplugs by quarry workers and quarry companies could be beneficial in reducing the absolute prevalence of hearing impairment especially among the elderly and long serving workers who have been shown to be at an increased risk of developing hearing loss.
[fig] Figure 1, Figure 2: Level of hearing impairment among quarry workers. Scatterplot of age versus duration of exposure. [/fig]
[table] Table 1: Measured noise levels (in dBA) on selected machines in five quarry companies. [/table]
[table] Table 2: Hearing threshold levels among various groups of respondents. [/table]
[table] Table 3: Logistic regression analysis of predictors of hearing loss (HTL > 25 dB) among quarry workers. Adjusted for type of quarry (place of work). $ Adjusted for type of quarry (place of work) and family history. VIF = age (1.244); duration of exposure (1.250); ear plugs (1.047). [/table]
|
10.2174/1573403X11309020001
|
CCBY
|
3682401
|
23701038
|
s2orc_pubmed_articles
|
Heart Failure: The Need for Global Health Perspective
Heart Failure: The Need for Global Health Perspective
2013Heart Failure: The Need for Global Health Perspective
Editorial Current Cardiology Reviews
92972013Heart failureepidemiologyvalve diseasetreatmentprognosisprevention
The global burden of cardiovascular disease (CVD), whether in terms of mortality or morbidity, is not in question and the scale of the threat to global health and security has been well documented and highlighted by both the scientific literature and last year's United Nations High-Level Meeting[1][2][3][4][5].Heart failure (HF) is an important cardiovascular disease in its own right, a risk factor for disease (atrial fibrillation, stroke, coronary artery disease), and a consequence of other diseases (e.g. rheumatic, hypertensive or coronary artery disease), therefore posing a triple threat to public health[6][7][8][9][10][11][12]. Heart failure is an important cause of hospitalisation and incurs huge costs to the individual and to society[4]. HF represents the common final pathway for all forms of cardiac disease, regardless of the aetiology. The role of HF in all stages in the development of CVD makes it an important focus for prevention and control of CVD, and for broader health policy approaches.This timely issue examines the burden of heart failure in terms of epidemiology, treatment and trends in policies in different regions of the world. Together, these contemporary reviews present an opportunity to compare and contrast the challenges faced by all areas of the world[13][14][15][16][17][18][19][20]. Although there are geographic differences in aetiology, with Chagas disease in South America, rheumatic heart disease in the Indian subcontinent and nutritional cardiomyopathies in Africa, common problems of inadequate access to proven primary and secondary prevention therapies are ubiquitous [21] particularly in low and middle income countries.Western countries are over-represented in population-based studies of epidemiology, aetiology, prognosis and management of HF, and in clinical trials, whereas the regions of the world carrying the greatest burden are under-represented and prospective studies are rare. While advances in device therapy and diagnostics are included in consensus practice guidelines in developed countries[22,23], many countries of the world struggle to offer the inexpensive, essential medicines to prevent and treat HF patients[24]. This lack of access to generic diuretics and beta-blocker therapy is even more striking given the inclusion of HF patients from low-resource settings in trials of novel therapies[25]. However, this series of reviews highlights that in all parts of the world, there is considerable room for improving equity of diagnosis and treatment.As well as lack of population-based, prospective data for HF, the existing data are heterogeneous in terms of population (hospital, community and specific patient subpopulations by age, sex or aetiology), definition and outcome (hospitalisation, mortality, prevalence, incidence). These factors make direct comparison of data from different settings challenging. Clear consensus guidelines exist for definition and diagnosis of HF [22-23] but should be adhered to in future epidemiologic studies, particularly from low-and middle-income settings in order to give an accurate picture of the global burden of HF and in order to improve comparability. Estimates of incidence rates in South Asia, East Asia, Africa and the Middle East are likely to under-estimate true burden of disease due to these methodological issues. Where burden of HF is estimated from data regarding risk factors for HF, such as coronary artery disease and hypertension[26], consistency in the definitions of these risk factors is also required in order to improve accuracy and comparability of data[27]. Encouragingly, there are increasing numbers of registries which aim to better quantify burdens of rheumatic heart disease, coronary artery disease and atrial fibrillation, particularly in low-and middle-income settings[8,28,29]. However maintaining the data quality of these registries in resource constrained settings remains a major challenge.HF is likely to grow as a major clinical and public health challenge due to demographic changes as well as rising prevalence of causative risk factors in aging patients, including hypertension, coronary artery disease, degenerative valve disease and obesity. Early identification of patients at risk for HF and asymptomatic patients with structural heart disease through a primary health care approach is critical if the human morbidity and mortality toll and the economic burden that HF causes are to be decreased. Improved management of individuals with HF is a key component of primary prevention of stroke, atrial fibrillation and other important causes of CVD[30], and is recognized as an essential intervention even for resource constrained primary care settings[31]. The need for intersectoral and integrated approaches has been repeatedlystressed by successive global health consensus documents[2,3,12,32]. Heart failure has earned its place as part of these intersectoral approaches (across cardiac diseases, other disease areas and other disciplines), which should be global in their scope and the need is urgent.In 2011 the World Health Assembly adopted a global target of 25% reduction of premature mortality from major noncommunicable diseases including cardiovascular disease by 2025. Strengthening primary health care worldwide, for primary and secondary prevention of HF will be critical for achieving this ambitious global target.
|
10.3389/fmicb.2023.1286626
|
CCBY
|
10663253
|
38029103
|
s2orc_pubmed_articles
|
Evolutionary flexibility and rigidity in the bacterial methylerythritol phosphate (MEP) pathway
Terpenoids are a diverse class of compounds with wide-ranging uses including as industrial solvents, pharmaceuticals, and fragrances.Efforts to produce terpenoids sustainably by engineering microbes for fermentation are ongoing, but industrial production still largely relies on nonrenewable sources.The methylerythritol phosphate (MEP) pathway generates terpenoid precursor molecules and includes the enzyme Dxs and two iron-sulfur cluster enzymes: IspG and IspH.IspG and IspH are rate limiting-enzymes of the MEP pathway but are challenging for metabolic engineering because they require iron-sulfur cluster biogenesis and an ongoing supply of reducing equivalents to function.Therefore, identifying novel alternatives to IspG and IspH has been an on-going effort to aid in metabolic engineering of terpenoid biosynthesis.We report here an analysis of the evolutionary diversity of terpenoid biosynthesis strategies as a resource for exploration of alternative terpenoid biosynthesis pathways.Using comparative genomics, we surveyed a database of 4,400 diverse bacterial species and found that some may have evolved alternatives to the first enzyme in the pathway, Dxs making it evolutionarily flexible.In contrast, we found that IspG and IspH are evolutionarily rigid because we could not identify any species that appear to have enzymatic routes that circumvent these enzymes.The ever-growing repository of sequenced bacterial genomes has great potential to provide metabolic engineers with alternative metabolic pathway solutions.With the current state of knowledge, we found that enzymes IspG and IspH are evolutionarily indispensable which informs both metabolic engineering efforts and our understanding of the evolution of terpenoid biosynthesis pathways.
# Introduction
Terpenoids are a diverse class of biomolecules with varied cellular functions, ranging from basic roles such as maintaining membrane fluidity and supporting electron transport to highly specialized roles such as hormone signaling and cellular defense [bib_ref] Kofam KOALA: KEGG Ortholog assignment based on profile HMM and adaptive score..., Aramaki [/bib_ref].All terpenoids are built from five carbon-building block precursors [dimethylallyl diphosphate (DMADP) and isopentenyl diphosphate (IDP)] and exhibit high structural diversity with over 95,000 different terpenoid compounds identified to date [bib_ref] Assembly-line catalysis in bifunctional terpene synthases, Faylo [/bib_ref].With applications as diverse as their molecular structures, terpenoids are used extensively in support of modern-day life such as commodity solvents, pharmaceuticals, and fuels [bib_ref] Terpenoids: opportunities for biosynthesis of natural product drugs using engineered microorganisms, Ajikumar [/bib_ref] [bib_ref] Hydrogenated monoterpenes as diesel fuel additives, Tracy [/bib_ref] [bib_ref] High-density renewable fuels based on the selective dimerization of Pinenes, Harvey [/bib_ref] [bib_ref] Isoprenoid drugs, biofuels, and chemicals--artemisinin, farnesene, and beyond, George [/bib_ref] and for exploring life's earliest record on Earth [bib_ref] Archean molecular fossils and the early rise of eukaryotes, Brocks [/bib_ref] [bib_ref] Steroids, triterpenoids and molecular oxygen, Summons [/bib_ref] [bib_ref] Terpenes, hormones and life: isoprene rule revisited, Hillier [/bib_ref].
Terpenoids are currently harvested from plants, generated by microbial fermentation, or produced chemically from nonrenewable petroleum [bib_ref] Developing fermentative terpenoid production for commercial usage, Leavell [/bib_ref].To date, variable yields, inefficiencies, and cost have made natural terpenoid production from plants or microbes unattractive leading to a global terpenoid supply chain that favors production from petroleum at the expense of sustainability and environmental impact [bib_ref] Developing fermentative terpenoid production for commercial usage, Leavell [/bib_ref].Engineered microbes have the potential for improved yields and increased diversity in the types of terpenoids that are produced, with some important successes to date [bib_ref] Isoprenoid drugs, biofuels, and chemicals--artemisinin, farnesene, and beyond, George [/bib_ref] [bib_ref] Metabolic engineering and synthetic biology for isoprenoid production in Escherichia coli and..., Navale [/bib_ref].Current efforts in engineering microbes for terpenoid production use one or both of two characterized "canonical" biosynthetic pathways (the mevalonate and methylerythritol phosphate pathways) and aim to produce higher yields and an ever-increasing range of terpenoid products as sustainably and economically as possible.However, knowledge about the evolutionary flexibility of terpenoid biosynthesis or possible alternative metabolic routes for terpenoid biosynthesis is lacking.
Terpenoids are used in membranes in all known domains of life [bib_ref] Isoprenoid biosynthesis: the evolution of two ancient and distinct pathways across genomes, Lange [/bib_ref] [bib_ref] Isoprenoids: remarkable diversity of form and function, Holstein [/bib_ref] and are present in molecular fossils billions of years old [bib_ref] Hopanoids. 1. Geohopanoids: the most abundant natural products on earth?, Ourisson [/bib_ref] [bib_ref] Archean molecular fossils and the early rise of eukaryotes, Brocks [/bib_ref] [bib_ref] Biomarker evidence for green and purple sulphur bacteria in a stratified palaeoproterozoic..., Brocks [/bib_ref].Given their universality and presence in the fossil records, terpenoids are identified as biomarkers, providing insight into life's earliest evolution on Earth [bib_ref] Isoprenoid biosynthesis: the evolution of two ancient and distinct pathways across genomes, Lange [/bib_ref] [bib_ref] Origins and early evolution of the mevalonate pathway of isoprenoid biosynthesis in..., Liu [/bib_ref] [bib_ref] Lipid biomarkers: molecular tools for illuminating the history of microbial life, Summons [/bib_ref].Yet, important questions remain about the evolution of terpenoid biosynthesis, including the identity of the primordial pathway for terpenoid biosynthesis and how horizontal gene transfer may have contributed to the evolution of terpenoid biosynthetic pathways [bib_ref] The eukaryotic MEP-pathway genes are evolutionarily conserved and originated from Chlaymidia and..., Zeng [/bib_ref] [bib_ref] Four billion years of microbial terpenome evolution, Hoshino [/bib_ref].Greater understanding of the variations within terpenoid biosynthetic pathways may further elucidate the evolution of this important ancient pathway.
Of the two known modern metabolic routes that generate the five carbon building-block precursor molecules from which all terpenoids are derived, the MVA pathway is generally used by eukaryotes and archaea [bib_ref] The eukaryotic MEP-pathway genes are evolutionarily conserved and originated from Chlaymidia and..., Zeng [/bib_ref] and the MEP pathway is generally used by bacteria and plastids [bib_ref] The eukaryotic MEP-pathway genes are evolutionarily conserved and originated from Chlaymidia and..., Zeng [/bib_ref].The MEP pathway can theoretically generate more terpenoid product from the same starting material as the MVA pathway (Figures [fig_ref] FIGURE 1: Reactions of MEP and MVA metabolic pathways [/fig_ref] ; [bib_ref] TECHNOLOGY UPDATE: development of a gas-phase bioprocess for isoprene-monomer production using metabolic..., Whited [/bib_ref].As such, the MEP pathway may have greater potential for producing terpenoids efficiently, but engineering efforts have yet to surmount several important barriers to increasing flux through the pathway.
The first enzyme of the MEP pathway, Dxs, is an initial bottleneck to carbon flux (Figure [fig_ref] FIGURE 1: Reactions of MEP and MVA metabolic pathways [/fig_ref] ; [bib_ref] Expression of prokaryotic 1-deoxy-dxylulose-5-phosphatases in Escherichia coli increases carotenoid and ubiquinone biosynthesis, Harker [/bib_ref] [bib_ref] Strategies of isoprenoids production in engineered bacteria, Li [/bib_ref].This bottleneck can be relieved by overexpression of Dxs, at which point IspG which converts methylerythritol 2, 4-cyclodiphosphate (MEcPP) to hydroxymethylbutenyl 4-diphosphate (HMBPP) becomes rate limiting [bib_ref] Metabolite profiling identified Methylerythritol Cyclodiphosphate efflux as a limiting step in microbial..., Zhou [/bib_ref] [bib_ref] Origins and early evolution of the mevalonate pathway of isoprenoid biosynthesis in..., Liu [/bib_ref] [bib_ref] Balanced activation of IspG and IspH to eliminate MEP intermediate accumulation and..., Li [/bib_ref] [bib_ref] Investigation of the methylerythritol 4-phosphate pathway for microbial terpenoid production through metabolic..., Volke [/bib_ref].IspG and the next enzyme downstream, IspH, are both iron-sulfur cluster containing [4Fe-4S] proteins [bib_ref] 1-methylthio-D-xylulose 5-phosphate methylsulfurylase: a novel route to 1-deoxy-D-xylulose 5-phosphate in Rhodospirillum rubrum, Wang [/bib_ref].Iron-sulfur cluster biogenesis and insertion into proteins is a complex process [bib_ref] Iron-Sulfur Clusters: Biogenesis, Molecular Mechanisms, and Their Functional Significance, Xu [/bib_ref].After IspG and IspH have been loaded with iron-sulfur clusters, they require an external supply of reducing equivalents to function.Altogether, simple overexpression of IspG and IspH may not relieve the bottleneck because these proteins require extensive cellular processes in the generation and loading of iron-sulfur clusters and the maintenance of substrates for these proteins to function.Due to the barriers to increasing flux through IspG and IspH, identifying alternative enzymes to IspG and IspH would be advantageous for engineering of the MEP pathway.
Multiple routes that allow native or acquired circumvention of the first MEP pathway enzyme Dxs have been identified but to date no alternative routes circumventing IspG or IspH have been found [bib_ref] A mutant pyruvate dehydrogenase E1 subunit allows survival of Escherichia coli strains..., Sauret-Güeto [/bib_ref] [bib_ref] A new family of enzymes catalyzing the first committed step of the..., Sangari [/bib_ref] [bib_ref] A rubis CO-like protein links SAM metabolism with isoprenoid biosynthesis, Erb [/bib_ref] [bib_ref] Mutations in Escherichia coli aceE and ribB genes allow survival of strains..., Perez-Gil [/bib_ref] [bib_ref] Evolutionary diversification and characterization of the eubacterial gene family encoding DXR type..., Carretero-Paulet [/bib_ref] [bib_ref] Enhancing terpene yield from sugars via novel routes to 1-deoxy-d-Xylulose 5-phosphate, Kirby [/bib_ref].In contrast, multiple alternative routes for the MVA pathway for terpenoid precursor biosynthesis have been discovered, making it a pertinent demonstration of the potential for variations on a metabolic pathway to evolve using alternative enzymes and metabolites (Figure [fig_ref] FIGURE 1: Reactions of MEP and MVA metabolic pathways [/fig_ref] ; [bib_ref] Modified mevalonate pathway of the archaeon Aeropyrum pernix proceeds via trans-anhydromevalonate 5-phosphate, Hayakawa [/bib_ref] [bib_ref] Four billion years of microbial terpenome evolution, Hoshino [/bib_ref].
Here we define evolutionary flexibility as steps in metabolism for which multiple enzymatic solutions have evolved, as is seen in the MVA pathway (Figure [fig_ref] FIGURE 1: Reactions of MEP and MVA metabolic pathways [/fig_ref].In contrast, evolutionarily rigid metabolic steps are those for which alternative enzymes that would allow circumvention of that step have not evolved.To inform efforts to engineer the MEP pathway, we investigated the evolutionary diversity of terpenoid biosynthesis strategies in bacteria.We explored bacterial diversity by assessing an available database of bacterial genomes for terpenoid biosynthesis strategies and alternative enzymatic routes to steps in the MEP pathway.In particular, we asked if enzymes have evolved to circumvent IspG and IspH which pose known engineering constraints, or to Dxs, for which some alternatives exist.We identified bacteria with apparently incomplete MEP pathways as candidates for non-canonical terpenoid biosynthesis routes.Our study provides a comprehensive survey of the known bacterial species for non-canonical MEP pathways and underscores the evolutionary significance of IspG and IspH.
# Materials and methods
## Database of species and hmmer profiles
Fasta files of the NCBI database of reference and representative bacterial genomes available at the assembly level "complete chromosome" were downloaded (access date 6/19/2023).Profile Hidden Markov Models (pHMM) from the KOfam database [bib_ref] Kofam KOALA: KEGG Ortholog assignment based on profile HMM and adaptive score..., Aramaki [/bib_ref] were downloaded from KEGG using genome.jp/ftp/db/kofam/ (access date 6/19/2023, KEGG release 106.0) as well as the "ko_list" file which contains adaptive score thresholds that define an ortholog match for each pHMM as determined by KOfam and described in more detail in section 2.3.The "ko_list" file was trimmed to only include the MEP and MVA genes of interest which increased computational analysis speed by limiting ortholog detection to only these genes.DxrII did not have a pHMM from KOfam, so we used hmmbuild to construct a pHMM model [bib_ref] Accelerated profile HMM searches, Eddy [/bib_ref] using the set of 7 genes predicted to be DxrII as annotated in NCBI's conserved protein domain family in COG4091.Full bioinformatic pipeline is available at https://github.com/bamarshall2/evolutionary_flexibility.
## Open reading frame (orf) and kofamscan for ortholog identification
For each genome, orfipy was used to identify ORFs of at least 150 nucleotides in length and generate an output of the peptide translation using the specifications "--min 150 --pep $file_name.fa" [bib_ref] Orfipy: a fast and flexible tool for extracting ORFs, Singh [/bib_ref].All genomes were translated using standard codon table 1 except genera Spiroplasma, Mycoplasma, Mycoplasmoides, Malacoplasma, and Ureaplasma which were translated with codon table 4 according to their taxonomy using the additional orfipy argument "--table 4. " Kofam_scan tool was run on each orfipy output using the curated KO list of MEP and MVA genes of interest and pHMM profiles from the KOfam database.The kofam_scan command used was "./exec_annotation --format = detail-tsv -o ~ path_to_ output/$output_file_name ~path_to_input/$input_file_name."
## Identification of orthologous genes
For Dxs, DxrI, IspE, IspDF, IspG, IspH, HMGS, HMGRI, HMGRII, MVK, DPMD, Rubisco-like protein MTRu-1P isomerase, and cupin type I MTXu-5P methylsulferase a gene ortholog was determined to be present if the similarity score from alignment exceeded the KOfam score threshold for identifying similarity [bib_ref] Kofam KOALA: KEGG Ortholog assignment based on profile HMM and adaptive score..., Aramaki [/bib_ref].This task was completed using custom scripts, all code is available at GitHub. 1 The computation of the KOfam score thresholds is described in detail in [bib_ref] Kofam KOALA: KEGG Ortholog assignment based on profile HMM and adaptive score..., Aramaki [/bib_ref].In brief, to compute the score threshold for a protein family in the KEGG Orthology database, the manually curated set of proteins is randomly divided into three groups, one for use as a positive training set and the other two for generation of the pHMM.Proteins that are not part of the protein family are used as a negative training dataset.Bit scores of alignments between the pHMM and the training data sets allows determination of a score threshold that maximizes the F measure which is the harmonic mean of precision and recall.The process of determining a threshold score is repeated three times by swapping the positive training set with one used to train the pHMM until each group of data has been used to both train and test the pHMMs.A final threshold score is then determined as the average of each iteration's threshold score.For the proteins for which we used the score thresholds, the average F measure was 0.979 with MVK having the smallest F measure of 0.897 and DxrI having the highest F measure of 1. IspD and IspF occasionally exist as a bifunctional fusion protein called IspDF.We observed that presence/absence calls for these two proteins were sometimes incorrect when using the KOfam adaptive thresholds which is reflected by lower F measures for IspD (F measure = 0.720) and IspF (F measure = 0.704).These inaccurate calls may have been due to attempts to distinguish whether a monofunctional or bifunctional protein is present, which requires high thresholds for alignment scores and causes reduced accuracy.For our purposes, identification of an IspD or IspF domain is of key relevance, but whether it is in a monofunctional or bifunctional version of the protein is not.To best identify IspD and IspF, we use the KOfam adaptive threshold for identification of bifunctional IspDF (F measure = 0.997), and for IspD or IspF we set a cutoff threshold of ORFs with alignment e-values of less than 1 −10 .The KOfam database did not have an adaptive threshold cutoff for DxrII for determination of orthology.For DxrII, we set the score cutoff at alignments with e-values of less than 1 −10 as well.
To determine a species' MEP and MVA gene status, we sorted for presence and absence using Excel.Full table output of gene presence or absence is available in Supplementary Table .
## Determining mva pathway status
Diversity exists in the lower part of the MVA pathway (Figure [fig_ref] FIGURE 1: Reactions of MEP and MVA metabolic pathways [/fig_ref] but complete characterization of the MVA pathway was not the goal of this research so determination of MVA pathway presence was more relaxed than analyses for the MEP pathway.For the MVA pathway, if 2 out of 4 genes (encoding HMGS, HMGRI/II, MVK, and DPMD) were present, the organism was considered to possess the MVA pathway.
## Phylogenetic tree building
Species trees were constructed using Genome Taxonomy Database (GTDB) master tree and only the species included in this study were retained.The GTDB master tree included 2,210 of the 4,400 species (~50%) in the presented analyses, so the phylogenetic trees show half of the species analyzed.
For protein trees, the protein sequences corresponding to the entries from the species tree were extracted and aligned by MAFFT [bib_ref] MAFFT: a novel method for rapid multiple sequence alignment based on fast..., Katoh [/bib_ref] using default parameters with iteration refinement over 1,000 cycles.Columns with >95% gaps were removed using TrimAL [bib_ref] Trim Al: a tool for automated alignment trimming in large-scale phylogenetic analyses, Capella-Gutiérrez [/bib_ref].These sequences were used to build the phylogenetic tree for the MEP pathway proteins.A maximum-likelihood phylogenetic tree was constructed for each of the protein using IQ-Tree 2 [bib_ref] Metabolic engineering and synthetic biology for isoprenoid production in Escherichia coli and..., Navale [/bib_ref] and ModelFinder plus was implemented for selecting the best-fit evolutionary model.Branch support values were calculated using the Shimodaira-Hasegawa approximate likelihood-ratio test (SH-aLRT) with 1,000 bootstrap replicates and 1,000 ultrafast bootstrap (UFBoot) replicates optimized by the nearest neighbor interchange (NNI).The phylogenetic trees were visualized using FigTree v1.4.4 [bib_ref] From molecular fossils of bacterial Hopanoids to the formation of isoprene units:..., Rambaut [/bib_ref].The reconstructed trees were rooted using midpoint rooting and annotated in R studio [bib_ref] Estimating phylogenetic trees from distance matrices, Farris [/bib_ref].
# Results
## Some bacteria use a non-canonical mep pathway lacking dxs
To find candidate species for a non-canonical MEP pathway, we started with an available database of representative bacterial genomes [bib_ref] Reference sequence (RefSeq) database at NCBI: current status, taxonomic expansion, and functional..., O'leary [/bib_ref] and extracted all open reading frames (ORFs) from these genomes [bib_ref] Orfipy: a fast and flexible tool for extracting ORFs, Singh [/bib_ref].Next, we used a HMMER based search tool called KofamScan for identification of orthologs to MEP and MVA genes (Figure; [bib_ref] Kofam KOALA: KEGG Ortholog assignment based on profile HMM and adaptive score..., Aramaki [/bib_ref].
Of 4,400 species in the database, 3,411 (78%) had all 7 MEP genes detected (Dxs, DxrI/II, IspD, IspE, IspF, IspG, and IspH) but lack the MVA pathway according to our criteria (see Methods) (Figure.Some species (541; 12%) had the MVA pathway and no MEP pathway genes or an incomplete MEP pathway detected whereas others (83; 1.8%) had all genes for the MEP pathway and at least two MVA pathway genes.Of 136 species (3.1%) that lacked a complete set of identifiable orthologs for both the MEP and MVA pathways, many had small genomes with a median size of 843,495 base pairs and included known endosymbiotic species Mycoplasma, Rickettsia, and Spiroplasma.The remaining 229 species (5.2%) had at most one MVA gene and an incomplete MEP pathway where at least one gene of Dxs, DxrI/II, IspD, IspE, IspF, IspG, or IspH appeared to be missing making them candidates for species using a non-canonical MEP pathway (Figure.
Of these 229 candidates, most were missing Dxs only (181, 79%) (Figureand thus might use an alternative to Dxs.One such potential alternative is the 5-methylthioadenosine (MTA) isoprenoid shunt first identified in Rhodospirillum rubrum (Figure [fig_ref] FIGURE 1: Reactions of MEP and MVA metabolic pathways [/fig_ref] ; [bib_ref] A rubis CO-like protein links SAM metabolism with isoprenoid biosynthesis, Erb [/bib_ref].The MTA-isoprenoid shunt uses the enzymes Rubisco-Like protein MTRu-1P isomerase and Cupin type I protein MTXu-5P methylsulferase to convert polyamine biosynthesis intermediate MTA to isoprenoid biosynthesis intermediate 1-deoxy-D-xylulose 5-phosphate (DXP) .For the candidate non-canonical MEP genomes lacking Dxs, 4 (2%) had only the MTXu-5P methylsulfurase and 5 (3%) had both the MTXu-5P methylsulfurase and the MTRu-1P isomerase (Supplementary Table .In these species, it is possible the MTA-isoprenoid shunt may substitute for Dxs, but this hypothesis will require further testing.
For the remaining 172 candidate species (95%) missing Dxs, the specific mechanism that allows the absence of Dxs requires further investigation to determine.Regardless of the mechanism, the identification of candidate species that contain all other MEP enzymes except Dxs suggest that the MEP pathway has evolutionary flexibility for the enzymatic activity that Dxs provides.
## Ispg and isph are evolutionarily rigid
Dxs is the only MEP pathway enzyme for which we identified a subset of candidate species that lack Dxs suggesting that an alternative strategy that circumvents Dxs has evolved.For IspG and IspH, there were no promising candidates for identification of alternative MEP pathways that circumvent one or both enzymes (Figure.Thus, we define IspG and IspH as evolutionary rigid.The single candidate genome lacking IspG only was Lelliotia steviae (Figure.When investigated, we found that the fasta file for this species was 1 million base pairs shorter than the sequence originally published and that other closely related Lelliottia species all had IspG present (Supplementary Table .Thus, we suggest the full Lelliotia steviae genome likely encodes IspG but that it may be in the 1 million missing base pairs.We interpret this single genome lacking IspG to be indicative of the level of inaccuracy within a large, collaborative database of species.
The two species lacking IspH only were Mycoplasma tullyi and Mycoplasmoides gallisepticum (Figures.Both species have genomes smaller than 1 million base pairs and all other Mycoplasma and Mycoplasmodies included in this database had In silico identification of terpenoid biosynthesis metabolic pathways.(A) Bioinformatic analysis pipeline.The HMMER search for orthologs used profile Hidden Markov Models (pHMM) from the KOfam database [bib_ref] Kofam KOALA: KEGG Ortholog assignment based on profile HMM and adaptive score..., Aramaki [/bib_ref].For ortholog identification cutoff values of genes Dxs, DxrI, IspE, IspDF, IspG, and IspH, we relied on the KOfam database of adaptive score thresholds determined for each gene [bib_ref] Kofam KOALA: KEGG Ortholog assignment based on profile HMM and adaptive score..., Aramaki [/bib_ref].For DxrII, IspD, and IspF we defined orthologs as alignments to the pHMM with e-values lower than 1 −10 .See Methods for more details.(B) Terpenoid biosynthesis strategy genotypes sorted.Circle areas are scaled to the total number of species for each genotype."+" indicates all enzymes of the MEP pathway or at least two enzymes of the MVA pathway were detected."-"indicates none of the enzymes of the MEP pathway or fewer than two enzymes of the MVA pathway were detected.Incomplete indicates that only some of the enzymes of the MEP pathway were detected.For the 541 MVA+ species, some have no detectable MEP orthologs while others do.This is indicated by the "−/incomplete" description of the MEP pathway.neither the MVA nor the MEP pathway consistent with all being endosymbionts and lacking terpenoid biosynthesis metabolisms (Figure;.It appears that though closely related Mycoplasma and Mycoplasmoides species have evolved to rely on host isoprenoid biosynthesis, Mycoplasma tullyi and Mycoplasmoides gallisepticum have retained a near complete MEP pathway despite having comparably small genomes.Mycoplasmoides gallisepticum has been demonstrated to generate MEP pathway intermediate HMBPPthough the essentiality of the MEP pathway in either organism remains to be confirmed.With only two species that are outliers in their genera, they do not constitute a compelling lead for identifying an alternative enzyme to IspH.It is possible that these species are making use of a near complete MEP pathway that ends with the generation of HMBPP and it is also possible that they do possess an alternative enzyme that circumvents IspH.These two species may have their own interesting evolutionary rationale for maintaining a nearly complete MEP pathway, but it will require further study to understand.The remaining 45 candidate species have no compelling patterns that suggest the presence of a non-canonical MEP pathway that would aid in the engineering of the MEP pathway (Figure.
We conclude that an alternative enzyme to Dxs probably exists and is in use in some of the 181 candidate species lacking Dxs.We interpret this to mean that an alternative to Dxs has probably evolved making it evolutionarily flexible.Meanwhile, given our approach, we cannot identify any species that may be using alternative enzymes to IspG or IspH.
## Phylogenetic distribution of species lacking dxs
Candidate species that might encode a non-canonical MEP pathway that lack Dxs are distributed across different phylogenetic groups (Figure .Species lacking Dxs can be found in the phyla Firmicutes, Actinobacteria, Alpha-proteobacteria and others.A complete list of these species is presented in Supplementary Figure .These species may use the same or different strategies to circumvent Dxs, but additional analysis will be required to identify these pathways.
## Within a genus terpenoid biosynthesis strategies vary at the species level
Generally, all species in a genus use either the MEP or MVA pathway (e.g., Bacillus and Weissella; Figure [fig_ref] FIGURE 4: Genera with species possessing the MEP and MVA pathways [/fig_ref].Another common pattern is for species to possess one complete pathway, and some of the enzymes of the other pathway.In other genera, most species use the same isoprenoid biosynthesis pathway although a subset encode enzymes for both pathways (e.g., Microbacterium; Figure [fig_ref] FIGURE 4: Genera with species possessing the MEP and MVA pathways [/fig_ref].For 10 genera with multiple represented species, terpenoid biosynthesis strategies are bifurcated (Figure [fig_ref] FIGURE 4: Genera with species possessing the MEP and MVA pathways [/fig_ref].In these genera, some species use one terpenoid biosynthesis strategy whereas other species use the other terpenoid biosynthesis strategy.
## Figure 3
Species tree highlighting candidate species lacking Dxs.Species tree of species included in the Genome Taxonomy Database (GTDB) and in the NCBI reference and representative database of complete genomes analyzed in this report (see Methods for details).Species with red circles at tip lack Dxs.For example, Staphylococcus was represented by 35 species in the database, 27 of which (77%) used the MVA pathway and possess one or two MEP pathway enzymes whereas the remaining 8 species (23%) encode complete MEP pathways but no MVA pathway enzymes (Figure [fig_ref] FIGURE 4: Genera with species possessing the MEP and MVA pathways [/fig_ref].In Cellulomonas, 9 of the species encode a complete MEP pathway with no MVA enzymes but a single species, Cellulomonas taurus, encodes an incomplete MEP pathway and a complete MVA pathway (Figure [fig_ref] FIGURE 4: Genera with species possessing the MEP and MVA pathways [/fig_ref].The MVA pathway genes in Cellulomonas taurus are in two operons in the genome, roughly 1 million base pairs apart.
These data suggest that selective pressure exists to retain a terpenoid biosynthesis strategy, but that a species may use MEP, MVA, or both based on evolutionary trajectory.
## Ispg protein tree differs from species tree and the three-domain ispg proteins cluster together
To understand the impact of horizontal gene transfer on the MEP pathway in bacteria, we generated evolutionary trees from protein sequences for Dxs, DxrI, IspE, IspG, and IspH (Supplementary Figure [fig_ref] FIGURE 5: Protein tree of IspG [/fig_ref].The Dxs, DxrI, IspE, IspG, and IspH trees differ from the species tree in that Actinobacteria have clustered with Proteobacteria in the protein trees instead of Bacteroidota as was the case of the species tree (Figures [fig_ref] FIGURE 5: Protein tree of IspG [/fig_ref] and Supplementary Figure [fig_ref] FIGURE 5: Protein tree of IspG [/fig_ref].Also, a notable feature for these protein trees is the paraphyletic nature of the Proteobacteria phylum, which is most clearly observed in the IspG tree.In the IspG protein tree, the Proteobacterial phylum is split into two groups; one Proteobacterial group clusters with Bacteroidota and the other with Firmicutes.Further, the Proteobacterial phylum is more widely distributed in the IspG protein tree with some γ-Proteobacteria clustering within β-Proteobacteria and the α-Proteobacteria being split.
IspG is generally a two-domain protein in Bacteria, but a third domain called A* may be present (Figure [fig_ref] FIGURE 5: Protein tree of IspG [/fig_ref] ; [bib_ref] Structure, function and inhibition of the two-and three-domain 4Fe-4S IspG proteins, Liu [/bib_ref].The A* domain has a TIM barrel fold, but its exact role has yet to be elucidated [bib_ref] Structure, function and inhibition of the two-and three-domain 4Fe-4S IspG proteins, Liu [/bib_ref].The phyla Bacteroidota, Chlorobi, Verrucomicrobia, and Chlamydiae all possess the A* domain and cluster together (Figures [fig_ref] FIGURE 5: Protein tree of IspG [/fig_ref].Plancomycetota, Verrucomicrobia, and Chlamydiae are the major phyla that constitute superphylum PVC [bib_ref] The Planctomycetes, Verrucomicrobia, Chlamydiae and sister phyla comprise a superphylum with biotechnological..., Wagner [/bib_ref] , however only Verrucomicrobia and Chlamydiae possess the three-domain form of IspG whereas Planctomycetota possess the 2-domain form.This splitting of the PVC superphylum with regards to IspG has been illustrated previously [bib_ref] The eukaryotic MEP-pathway genes are evolutionarily conserved and originated from Chlaymidia and..., Zeng [/bib_ref].Absence of the three-domain form of IspG only in Planctomycetota of the PVC superphylum could be due to horizontal gene transfer as the phylogenetic placement of the Plancomycetota IspG sequence in the protein tree is not congruent with the PVC clade in the species tree.An alternative possibility is that the A* domain evolved in a common ancestor of Bacteroidota, Plancomycetota, Verrucomicrobia, and Chlamydiae that was then lost in Planctomycetota.With the data presented here, these two possibilities cannot be distinguished.
Based on the differences between the MEP protein trees and the species tree, MEP pathway inheritance is not strictly vertical.Therefore, we suggest that horizontal gene transfer may have played a role in the evolution of this metabolic pathway.
# Discussion
In this study, we assessed the MEP and MVA pathways and protein diversity across bacteria.Our goal was to inform options for engineering the precursor pathways for microbial terpenoid biomanufacturing by determining if alternatives exist for some steps in precursor pathways.In particular, we were interested in alternatives to Dxs and whether non-canonical MEP pathways can circumvent the IspG and IspH enzymes, which are known to pose major engineering challenges.To achieve this goal, we identified genomes lacking the MVA pathway, with an incomplete MEP pathway referred to as candidate species for a non-canonical MEP pathway.We found that Dxs is evolutionarily flexible and that alternatives may exist whereas for IspG and IspH we could not identify any candidate species that appear to circumvent these enzymes.These results may aid the ongoing quest to understand early evolution of terpenoid biosynthetic pathways.
Identified alternatives to the canonical MEP enzymes include DxrII, the MTA-isoprenoid shunt, and mutations that arise when Dxs is knocked out (Figure [fig_ref] FIGURE 1: Reactions of MEP and MVA metabolic pathways [/fig_ref] ; [bib_ref] A mutant pyruvate dehydrogenase E1 subunit allows survival of Escherichia coli strains..., Sauret-Güeto [/bib_ref] [bib_ref] A new family of enzymes catalyzing the first committed step of the..., Sangari [/bib_ref] [bib_ref] A rubis CO-like protein links SAM metabolism with isoprenoid biosynthesis, Erb [/bib_ref] [bib_ref] Mutations in Escherichia coli aceE and ribB genes allow survival of strains..., Perez-Gil [/bib_ref] [bib_ref] Evolutionary diversification and characterization of the eubacterial gene family encoding DXR type..., Carretero-Paulet [/bib_ref] [bib_ref] Enhancing terpene yield from sugars via novel routes to 1-deoxy-d-Xylulose 5-phosphate, Kirby [/bib_ref].The suppressor mutations that have been documented to arise when We identified 181 candidate species for a non-canonical MEP pathway missing Dxs.In contrast, we identified fewer than 40 species missing IspG, IspH, or both in combination with other MEP enzymes and no compelling patterns among these species (Figure.We conclude that within this database of 4,400 species, all MEP utilizing species possess IspG and IspH as part of their MEP pathways.On the other hand, some of the 181 candidate species missing Dxs likely have an alternative MEP pathway strategy.This result would align with previous research where alternative routes for circumventing Dxs have been identified (see above).
Taken together, our results suggest that within the MEP pathway, evolutionary flexibility exists for Dxs.Alternative strategies to circumvent Dxs have likely evolved whereas IspG and IspH are evolutionarily rigid and no alternatives appear to have evolved or can be identified at this time.DxrI/II, IspD, IspE and IspF are also evolutionarily rigid as they appear to be retained by essentially all species using the MEP pathway.Whether the evolutionary rigidity of IspG, IspH, DxrI/II, IspD, IspE and IspF is due to a functional constraint hindering biochemical ability or due to the lack of sequence diversity available for the enzyme to explore under an evolving landscape remains an open question.
The evolutionary rigidity of enzymes in the MEP pathway may be valuable for unraveling the evolution of terpenoid biosynthesis.Analysis of the protein phylogenetic trees for the MEP enzymes was consistent with the possibility that horizontal gene transfer has played a role in the evolution of these enzymes (Figure [fig_ref] FIGURE 5: Protein tree of IspG [/fig_ref] ; Supplementary Figure [fig_ref] FIGURE 5: Protein tree of IspG [/fig_ref].The role of horizontal gene transfer in extant species is highlighted in Figure [fig_ref] FIGURE 4: Genera with species possessing the MEP and MVA pathways [/fig_ref] where within some genera bifurcation of the terpenoid biosynthesis strategies has occurred.
Species retaining all MEP enzymes except Dxs, are distributed across different phylogenetic groups (Figure .It is possible that these species acquired the same alternative to Dxs via horizontal gene transfer.Another possibility is that multiple strategies have evolved.Further work will be required to determine how these species circumvent Dxs as well as the relationship of these metabolic routes to each other.Possible explanations for the lack of Dxs include incomplete sequencing, promiscuous activity of related enzymes at levels adequate to support loss of Dxs, or an alternative enzyme evolved that is functionally equivalent to Dxs which should be sought.
A limitation of our bioinformatic approach is that it will only identify alternative MEP pathways in species lacking redundancy.Species with a complete MEP or MVA pathway in addition to an alternative non-canonical MEP pathway would not be identified.Additionally, we only surveyed the current database of bacterial species, which is inherently biased toward culturable bacteria and does not capture the full diversity of life [bib_ref] Identifying microbial diversity in the natural environment: a molecular phylogenetic approach, Hugenholtz [/bib_ref] [bib_ref] A new view of the tree of life, Hug [/bib_ref].It is possible that in organisms not yet sequenced, alternative MEP pathways may exist.Finally, we note that sequencebased identification of orthologs could miss functional enzymes that lack sequence similarity.Regardless, our results provide valuable insights into MEP pathway constraints and identify a set of species that are promising candidates for an alternative MEP pathway that circumvents Dxs.
Our findings also inform metabolic engineering of the MEP pathway in two ways.First regarding Dxs, we find that alternative enzymes to Dxs may exist.As Dxs governs MEP pathway flux via feedback regulation [bib_ref] Feedback inhibition of deoxy-d-xylulose-5-phosphate synthase regulates the methylerythritol 4-phosphate pathway, Banerjee [/bib_ref] and requires pyruvate and glyceraldehyde-3-phosphate as the MEP pathway substrates, it would be of value to metabolic engineers to identify Dxs alternatives.An alternative enzyme to Dxs may consume different substrates or avoid regulation.For IspG and IspH, ongoing efforts to engineer these enzymes to increase their flux are justified, as no alternatives to their role in the MEP pathway were apparent.
The development of sequencing technologies has enabled the determination of unprecedented numbers of bacterial genomes.This database of bacterial genomes is a repository of knowledge with astounding potential.Here, we leveraged the current repository of bacterial genomes as a resource for identifying alternative enzymes to enable engineering of a key metabolic pathway.Future efforts in microbial engineering should also utilize the database of bacterial genomes to search for naturally evolved alternatives for cellular metabolism that go beyond currently characterized literature.
[fig] FIGURE 1: Reactions of MEP and MVA metabolic pathways.(A) MEP pathway.Asterisks indicate a mutant version of an enzyme is necessary for completing a reaction.Enzymes circled in blue, Dxs, IspG, and IspH, are the focus of this report.GAP, glyceraldehyde 3-phosphate; Dxs, DXP synthase; DXP, 1-deoxyxylulose 5-phosphate; Dxr, DXS reductoisomerase; MEP, 2-C-methyl-erythritol 4-phosphate; CDP-ME, 4-(cytidine 5′-diphospho)-2-C-methylerythritol; CDP-MEP, 2-phospho-4-(cytidine 5′-diphospho)-2-C-methyl-erythritol; MEcPP, 2-C-methyl-erythritol-2,4-cyclodiphosphate; HMBPP, 4-hydroxy-3-methyl 2-butenyl diphosphate; IDP, isopentenyl diphosphate; DMADP, dimethylallyl diphosphate.(B) MVA pathway.AACT, Acetoacetyl-CoA thiolase; HMGS, HMG synthase; HMG-CoA, hydroxymethylglutaryl-CoA; HMGR, HMG reductase; MVK, mevalonate kinase; M5P, mevalonate 5-phosphate; M3K, mevalonate 3-kinase; M3P, mevalonate 3-phosphate; PMVK, phosphomevalonate kinase; MDP, mevalonate 5-diphosphate; M5PDH, M5P dehydratase; AM5P, trans-anhydromevalonate 5-phosphate; AMPD, AM5P decarboxylase; MPD, M5P decarboxylase; IP, isopentenyl phosphate; M3PK, M3P 5-kinase; MBP, mevalonate 3,5-bisphosphate; MBD, MBP decarboxylase; DPMD, diphosphomevalonate decarboxylase. [/fig]
[fig] FIGURE 4: Genera with species possessing the MEP and MVA pathways.(A) All genera with bifurcated terpenoid biosynthesis strategies.Representative nonbifurcating genera Bacillus, Weissella, and Microbacterium are shown for comparison on the left.Some of the species summarized in this bar chart have enzymes from the alternative terpenoid biosynthesis pathway.For complete description of each species' terpenoid biosynthesis gene status, see Supplementary TableS1.(B) MEP and MVA genotypes of Staphylococcus and Cellulomonas species as examples showing that within a single genus, species may bifurcate with regards to their terpenoid biosynthesis strategy. [/fig]
[fig] FIGURE 5: Protein tree of IspG.(A) Phylogenetic tree based on alignment of IspG protein identified in each organism.Species that have A* domain present in their IspG are highlighted in blue.(B) IspG domain configurations.IspG may either have two domains A and B or three domains A, A* and B. [/fig]
|
10.1074/jbc.RA120.015833
|
CCBY
|
7948570
|
33187980
|
s2orc_pubmed_articles
|
Lacritin proteoforms prevent tear film collapse and maintain epithelial homeostasis
[bib_ref] EMC3 coordinates surfactant protein and lipid homeostasis required for respiration, Tang [/bib_ref] [bib_ref] Novel FOXC2 mutation in hereditary distichiasis impairs DNA-binding activity and transcriptional activation, Zhang [/bib_ref] [bib_ref] Genotypic heterogeneity and clinical phenotype in triple A syndrome: a review of..., Brooks [/bib_ref] [bib_ref] LADD syndrome is caused by FGF10 mutations, Milunsky [/bib_ref] [bib_ref] Mutations in different components of FGF signaling in LADD syndrome, Rohmann [/bib_ref] [bib_ref] Mutations in NGLY1 cause an inherited disorder of the endoplasmic reticulumassociated degradation..., Enns [/bib_ref] [bib_ref] A new mutation in TP63 is associated with age-related pathology, Holder-Espinasse [/bib_ref] [bib_ref] A novel TRAPPC11 mutation in two Turkish families associated with cerebral atrophy,..., Koehler [/bib_ref] [bib_ref] Meibum lipid composition in Asians with dry eye disease, Lam [/bib_ref] [bib_ref] Hyaline membrane disease, Farrell [/bib_ref] [bib_ref] The real reason for having a meibomian lipid layer covering the outer..., Millar [/bib_ref] [bib_ref] Differential proteomic analysis of bronchoalveolar lavage fluid in asthmatics following segmental antigen..., Wu [/bib_ref] [bib_ref] In-depth proteomic analysis of human bronchoalveolar lavage fluid toward the biomarker discovery..., Sim [/bib_ref] [bib_ref] ) cDNA and genomic cloning of lacritin, a novel secretion enhancing factor..., Sanghi [/bib_ref] [bib_ref] Lacritin, a novel human tear glycoprotein, promotes sustained basal tearing and is..., Samudre [/bib_ref] [bib_ref] Topical administration of lacritin is a novel therapy for aqueous-deficient dry eye..., Vijmasi [/bib_ref] [bib_ref] A thermo-responsive protein treatment for dry eyes, Wang [/bib_ref] [bib_ref] Lacritin rescues stressed epithelia via rapid forkhead box o3 (FOXO3)-associated autophagy that..., Wang [/bib_ref] [bib_ref] Lacritin rescues stressed epithelia via rapid forkhead box o3 (FOXO3)-associated autophagy that..., Wang [/bib_ref] [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref] [bib_ref] Matrix metalloproteinase 9 and transglutaminase 2 expression at the ocular surface in..., Aragona [/bib_ref] [bib_ref] Transglutaminase 2: a new player in bronchopulmonary dysplasia?, Witsch [/bib_ref] [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref] [bib_ref] Human basal tear peptidome characterization by CID, HCD, and ETD followed by..., Azkargorta [/bib_ref] [bib_ref] Human basal tear peptidome characterization by CID, HCD, and ETD followed by..., Azkargorta [/bib_ref] [bib_ref] Human basal tear peptidome characterization by CID, HCD, and ETD followed by..., Azkargorta [/bib_ref]
# Results
## Lacritin immunodepleted normal tears prematurely collapse
Tears comprise a loose aqueous polymer of lipids and glycoproteins, including the basal tearing [bib_ref] ) cDNA and genomic cloning of lacritin, a novel secretion enhancing factor..., Sanghi [/bib_ref] [bib_ref] Lacritin, a novel human tear glycoprotein, promotes sustained basal tearing and is..., Samudre [/bib_ref] [bib_ref] Topical administration of lacritin is a novel therapy for aqueous-deficient dry eye..., Vijmasi [/bib_ref] [bib_ref] A thermo-responsive protein treatment for dry eyes, Wang [/bib_ref] and prohealth [bib_ref] Lacritin rescues stressed epithelia via rapid forkhead box o3 (FOXO3)-associated autophagy that..., Wang [/bib_ref] [bib_ref] Lacritinmediated regeneration of the corneal epithelia by protein polymer nanoparticles, Wang [/bib_ref] agonist lacritin whose bioactive C-terminus is dominated by two amphipathic α-helices [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] similar to the N-and C-termini of processed lipid stabilizing pulmonary [bib_ref] Restricted epithelial proliferation by lacritin via PKCα-dependent NFAT and mTOR pathways, Wang [/bib_ref] or PSIPRED (v4.0) predicted (N-terminal half) α-helices or beta strand (arrow head); dotted lines representing respective antigens of monoclonal "anti-N-term" antibody "1F5" and polyclonal "anti-C-term" antibody [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref] ; and alignment of lacritin synthetic peptides "C-95," "N-64/C-31," "N-94," and "N-94/C-6." B, lacritin N-and C-terminal proteoforms detected in normal human tears by topdown mass spectrometry ((30); used with permission). C, schematic diagram of lipid (orange) and aqueous (aqua) portions of normal tears on the eye, and pooled collections of normal human basal tears from 50 different individuals for Langmuir trough compression/expansion isocyling (schematic diagrams at right) without or after immunodepletion over immobilized "anti-N" term, "anti-C" term lacritin antibodies or preimmune (pre-imm) Ig. Dashed box on the eye indicates the region highlighted in [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref]. D, tear films (3 μl) on a 80 ml PBS subphase were equilibrated for 15 min under cover and then subjected to compression and expansion isocycling by respectively advancing or retreating dual opposing barriers at 1.37 cm 2 per second with surface pressure monitored by a Wilhelmy wire probe with sensitivity exceeding 0.01 mN/m [bib_ref] Surface relaxations as a tool to distinguish the dynamic interfacial properties of..., Georgiev [/bib_ref]. In this and subsequent studies, each isotherm represents the mean of triplicate experiments with individual experiments representing over 1200 data points. (i) indicates the "lift off area" at which compression elevates surface pressure from baseline to at least 1 mN/m. (ii) indicates the highest surface pressure attained. Inset, comparative highest surface pressure attained (left) and lift-off area (right) (mean with S.D [n = 3], ****p < 0.0001 [one-way ANOVA with Tukey multiple comparisons test]). E, expansion isotherm of the same samples, with bracket highlighting touch-down areas (n = 3). F, schematic diagram of compression and expansion isocycling of anti-C-term lacritin-depleted tear films or human aqueous deficient dry eye basal tears (pooled collections each from 50 different individuals) without or supplemented with 6 μM lacritin synthetic peptides C-95, N-64/C-31, N-94, or N-94/C-6. G, compression isotherms of anti-C-term lacritin-depleted tear films without or with each peptide per color code in A and H. H, comparative lift-off area (mean with S.D [n = 3], ****p < 0.0001 [one-way ANOVA with Tukey multiple comparisons test]). (1) and [bib_ref] A mutation in the surfactant protein c gene associated with familial interstitial..., Nogee [/bib_ref] are two different anti-C-terminal lacritin-immunodepletions. I, expansion isotherm of dry eye tear films without or with N-94 or N-94/C-6 (n = 3). J, LI-COR Western blot analysis of relative levels of lacritin monomer and C-terminal proteoforms in individual basal tears from (i) 21 normal (shown are monomer and proteoform data respectively from 14 and 21 individuals), or 21 aqueous deficient dry eye individuals, and tears collected without anesthesia from (ii) 21 Secondary Sjögren's syndrome dry eye patients. Values were normalized to equal protein ([mean with S.D], ****p < 0.0001; ***p = 0.0001; **p = 0.0013 [oneway ANOVA with Tukey multiple comparisons test]). Data for D-E, G-J in Data files [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref]. surfactant protein B (32-34) necessary for lung function and life [bib_ref] The molecular era of surfactant biology, Whitsett [/bib_ref]. Together, both lacritin α-helices partially comprise the antigen of polyclonal antibody "anti-C-term" [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] ; [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref] , the latter effective for lacritin immunodepletion [bib_ref] Lacritin rescues stressed epithelia via rapid forkhead box o3 (FOXO3)-associated autophagy that..., Wang [/bib_ref] [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref]. To gain insight into premature tear film collapse and whether lacritin deficiency in dry eye may be contributory, Langmuir surface balance studies were performed. Such studies are performed in a "Langmuir-Blodgett trough" consisting of a shallow (225 cm 2 ) reservoir of PBS (40 ml) with surface tear film that is compressed or expanded by symmetric movement of opposing barriers. As area changes, surface pressure is monitored by a Wilhelmy wire probe with sensitivity exceeding 0.01 mN/m. Surface pressure represents the surface tension of the buffer minus the surface tension of the film. We floated onto PBS normal human tears (3 μl) pooled from over 50 different individuals, or the same tear pools passed over immobilized anti-C-term lacritin antibodies (C-term depleted; [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] , or over a preimmune Ig column [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] ; mock depleted). Each was subjected to compression and expansion isocycling [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] , right) that mimics blinking. Isocycling is the synchronized repeated inward or outward movement of barriers at constant speed. Isocycling was performed at 35 C, the surface temperature of the human eye [bib_ref] Temperatures of the ocular surface, lid, and periorbital regions of Sjögren's, evaporative,..., Abreau [/bib_ref]. Normal and mock depleted films [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] , D-E) were more stable as per a higher "lift off" area ($50%; [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] , D inset, right) and maximum surface pressure ($25 mN/m; [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] (ii), D inset, left) achieved with compression. "Lift off" is when surface pressure first reaches 1 mN/m as compression progressively reduces film area (37)-an indicator of the capability of film constituents to restructure at the interface. Maximum surface pressure is when constituents are the most densely compressed in a stable film. Tears lacking lacritin were less resistant to compression (respectively $25% and $15-20 mN/m; [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] (i and ii), D inset, right and left) and expansion [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref]. Film stability is considered in terms of molecular packing density at the interface [bib_ref] Lipid monolayers: why use half a membrane to characterize protein-membrane interactions?, Brockman [/bib_ref].
Anti-C-term lacritin depleted tears behave like dry eye tears and can be rescued with C-Terminal lacritin peptides that are deficient in dry eye Normal basal tears contain different forms of lacritin as monomer, polymer, and proteoforms (26) -of which five N-and 42 different C-terminal proteoforms have been detected to date by top-down mass spectrometry ((30); [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] : "lacrt tear proteoforms"). Lacritin dimers, trimers, and larger polymers are attributable to cross-linking by tear tissue transglutaminase, a calcium-dependent glutamine γ-glutamyltransferase that is active in normal tears [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref] and whose ocular surface expression is elevated in patients with Sjögren's syndrome dry eye [bib_ref] Matrix metalloproteinase 9 and transglutaminase 2 expression at the ocular surface in..., Aragona [/bib_ref]. Which form of lacritin contributes to tear film stability? Anti-Cterm lacritin antibody detects lacritin monomer, polymer, and C-terminal proteoforms-but not N-terminal proteoforms. Anti-N-term lacritin antibody 1F5 detects N-terminal proteoforms, monomer, and polymer, but not C-terminal proteoforms [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] ; [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref]. We subjected normal human tears to 1F5 immunodepletion [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref]. Film stability under compression was unaffected per "lift off" area ($48%; [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] , D inset, right) and maximum surface pressure ($26 mN/m; [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] (ii), D inset, left) suggesting that C-terminal proteoforms maybe contributory. To test this possibility, we synthesized C-terminal synthetic peptides "N-94," "N-94/C-6," and "N-64/ C-31" that together span all detected C-terminal proteoforms in tears [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] and therefore serve as surrogates. N-terminal 24 amino acid peptide "C-95" lacking 95 C-terminal amino acids [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] was included as a negative control. "N-94" represents lacritin's C-terminal 25 amino acids with two amphipathic α-helices and a six amino acid C-terminal random coil domain, whereas "N-94/C-6" lacks the latter. "N-64/C-31" is a more internal amphipathic α-helix ((29); [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref]. In total, 6 μM of each was added to anti-C-term lacritin-depleted tears [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref]. In total, 6 μM is the suspected concentration of C-terminal lacritin proteoforms in human basal tears, versus $18 to 27 μM for all anti-Pep Lac N-Term detectable lacritin [bib_ref] Lacritin and the tear proteome as natural replacement therapy for dry eye, Karnati [/bib_ref]. N-94 and N-94/ C-6 ("lift off" area: [bib_ref] Reduced levels of tear lacritin are associated with corneal neuropathy in patients..., Mcnamara [/bib_ref] Lacritin monomer is selectively deficient in tears of almost all forms of dry eye (reviewed by . Are C-terminal proteoforms also lacking, and if so is their absence contributory to tear film instability? Dry eye is generally most severe in Sjögren's syndrome, an autoimmune form of dry eye disease [bib_ref] Reduced levels of tear lacritin are associated with corneal neuropathy in patients..., Mcnamara [/bib_ref]. We tested basal tears from 21 normal or 21 non-Sjögren's dry eye individuals [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] (i)) and tears collected without anesthesia from 21 Secondary Sjögren's syndrome dry eye patients [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref]. C-terminal proteoforms and monomer were deficient or absent in most non-Sjögren's and Sjögren's syndrome dry eye tears [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref]. Sjögren's syndrome samples were kindly provided by the Sjögren's International Collaborative Alliance for which volume was limiting. Non-Sjögren's dry eye tears (from the Warfighter Refractive Eye Surgery Program and Research Center, Fort Belvoir VA) were sufficient for compression and expansion isocycling. Compression values ("lift-off" area 26% [ [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] , H-I]; highest surface pressure 15%; [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] were essentially identical to anti-C-term lacritindepleted normal tears, as was the benefit of supplementation with 6 μM N-94 or N-94/C-6 (42-43% [ [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] , H-I] and 21-22 mN/m; [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref]. Similar benefit was apparent in expansion profiles [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref]. Thus, lacritin C-terminal proteoforms, but not apparently lacritin monomeric or polymeric forms, lower tear surface tension and when selectively absent or deficient in dry eye tears contribute to decreased film stability upon compression that can be largely rescued with N-94 or N-94/C-6.
## C-terminal lacritin peptide rescue restores elasticity
Viscoelasticity underlies tear compression and expansion properties that in turn modulate the quality of light refraction necessary for visual acuity-a feature compromised in dry eye [bib_ref] Optical and visual impact of tear break-up in human eyes, Tutt [/bib_ref]. We monitored the time-dependent relaxation of surface tension after pre-equilibrated films at a surface pressure of $15 mN/m were subjected to a sudden step compression of less than 5% of the prior surface area [fig_ref] Figure 2: Figure 2 [/fig_ref] , inset). The initial maximal surface pressure as the instantaneous stress response differed substantially between normal ($1.7 mN/m) and both anti-C-term lacritin-depleted normal ($0.6 mN/m) and dry eye tears ($0.6 mN/m; [fig_ref] Figure 2: Figure 2 [/fig_ref] , A-B]). As per compression/expansion isocycling, spiking N-94 or N-94/C-6 into anti-C-term lacritin-depleted normal tears was fully corrective [fig_ref] Figure 2: Figure 2 [/fig_ref] , but only partially so in dry eye tears in which the initial reading was elevated (respectively $1.2 and 1.6 mN/m) but then fell off precipitously with relaxation [fig_ref] Figure 2: Figure 2 [/fig_ref]. Viscoelasticity was assessed by Fourier transform of the relaxation data [bib_ref] Fourier transform surface viscoelastic modulus of dilute aqueous solutions of surfactants. Improved..., Loglio [/bib_ref] [bib_ref] Surface relaxations as a tool to distinguish the dynamic interfacial properties of..., Georgiev [/bib_ref] thereby yielding the stored elastic modulus ('E R ') and loss modulus ('E IM ') [fig_ref] Figure 2: Figure 2 [/fig_ref] , C-E)-the latter a measure of viscous behavior. N-94 and N-94/C-6 elevated the stored elastic modulus and diminished the loss modulus of both anti-C-term lacritin-depleted normal tears and dry eye tears [fig_ref] Figure 2: Figure 2 [/fig_ref] , D-E). The "dilatational elastic Tear film stabilizing activity of lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore normal viscoelasticity. A, time-dependent relaxation of surface tension after pre-equilibrated anti-C-term lacritin-depleted (without or with 6 μM N-94 or N-94/C-6) or normal human basal tear films formed as per [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] were subjected to a sudden step compression of less than 5% of the prior surface area. ****p < 0.0001; ns, not significant [two-way ANOVA with Sidak multiple comparisons test]. B, same procedure applied to human dry eye basal tear films (without or with 6 μM N-94 or N-94/C-6). For both A and B, shown are individual replicates for each experiment performed in triplicate (>2300 data points per variable with the exception of anti-C-term lacritin-depleted tears plus 6 μM N-94 [one replicate] and anti-C-term lacritin-depleted tears alone [1260 data points]). ****p < 0.0001 with top bracket a comparison to normal [two-way ANOVA with Sidak multiple comparisons test]. Inset, schematic diagram of time-dependent relaxation. C-G, Fourier transform (44) of the relaxation data in A-B as the stored elastic (E R ) or loss (E IM ) moduli, or loss factor (tan Φ)-both as a function of the log frequency. Color code of data is per A-B. C, stored elastic and loss moduli of normal human basal tears. D, stored elastic and loss moduli of anti-C-term lacritin-depleted human basal tears without or with N-94 or N-94/C-6 supplementation. E, stored elastic and loss moduli of human dry eye basal tears without or with N-94 or N-94/C-6 supplementation. F, loss factor of normal human basal tears versus anti-C-term lacritin-depleted tears without or with added 6 μM N-94 or N-94/ C-6; horizontal line: values less than or greater than 1 are respectively elastic or viscous. G, loss factor of human dry eye basal tears without or with added 6 μM N-94 or N-94/C-6. H, light incident to the Brewster angle reflects in a manner corelative to regional lipid thickness (top schematic). Shown is the reflection off normal, anti-C-term lacritin-depleted (without or with 6 μM N-94 or N-94/C-6), or dry eye (without or with 6 μM N-94 or N-94/C-6) basal tear films. Bar = 1 μm. Representative of triplicate experiments. Data for A-G in Data files [fig_ref] Figure 2: Figure 2 [/fig_ref]. modulus" ("E*") is the sum of E R and E IM , and the quotient of E IM /E R , ("tan Φ") is the "loss factor" such that values less than or greater than 1 are respectively elastic or viscous. At all frequencies, normal tears (maximal loss factor 0.5) as well as N-94 or N-94/C-6 supplemented anti-C-term lacritin-depleted normal (respective maximal loss factors 0.2, 0.1) and supplemented dry eye tears (respective maximal loss factors 0.5, 0.6) are elastic [fig_ref] Figure 2: Figure 2 [/fig_ref] , F-G). This contrasts with anti-C-term lacritin-depleted normal and dry eye (respective maximal loss factor 2.7, 2.8) tears that are respectively viscous between 10 −3.5 and 10 −2.2 Hz, and 10 −3.7 and 10 −2.3 Hz [fig_ref] Figure 2: Figure 2 [/fig_ref]
[formula] , F-G). [/formula]
Curious about the role of N-94/C-6 amino acids with nonpolar side chains, we synthesized "N-94/C-6-ser" in which seven of eight (not alanine) were replaced with serine. Serine is identical to alanine, but with polar -OH group. Also, serine is uncharged and neither hydrophobic nor hydrophilic. Lacritindepleted tears supplemented with N-94/C-6-ser (maximal loss factor 0.3) were surprisingly elastic [fig_ref] Figure 2: Figure 2 [/fig_ref] , suggesting that prevention of rupture is not a property of hydrophobicity, but instead likely due to charge. Indeed, N-94/C-6 contains six amino acids with charged side chains at neutral pH (four lysines and two glutamic acids). To ask whether normalized viscoelasticity was manifested in tear film structure, we applied Brewster angle microscopy. Light incident to the Brewster angle reflects in a manner proportional to the square of the regional lipid thickness [bib_ref] Direct visualization of monolayers at the air-water interface by Brewster angle microscopy, Hoenig [/bib_ref] such that white areas are thick and gray areas less so. Black areas lacking lipid islands are nonreflective [fig_ref] Figure 2: Figure 2 [/fig_ref] , top schematic; Figs. S2 and S3), as per regions in anti-C-terminal lacritin-depleted and dry eye tears. Supplementation of anti-C-terminal lacritin-depleted tears with N-94 or N-94/C-6 yields images similar to normal tears, whereas the pattern is more complex after supplementation of dry eye tears with islands packed with globular structures. By quantitation, normal tears present as a single population, in contrast to three populations for anti-C-terminal lacritindepleted tears and dry eye tears [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref]. With N-94 or N-94/ C-6 supplementation, these coalesce into two (anti-C-terminal lacritin-depleted) or one (dry eye) peaks. Two peaks were also observed in N-94/C-6-ser supplemented anti-C-terminal lacritin-depleted tears [fig_ref] Figure 2: Figure 2 [/fig_ref]. Thus N-94 and N-94/C-6 largely correct for loss of elasticity of dry eye tears by possibly acting as a surfactant, although the supplemented dry eye films are morphologically distinctive.
## C-terminal lacritin peptides interact with and stabilize meibomian gland secretions
If N-94 and N-94/C-6 are capable of acting as a tear lipid surfactant, one likely destination is the aqueous/lipid interface thereby accommodating respective hydrophilic and hydrophobic faces of their two amphipathic α-helices-as per pulmonary surfactant protein B (32-34), although per N-94/ C-6-ser the hydrophobic face appears to contribute little to stability. Tear lipids largely derive from eyelid meibomian glands, a form of sebaceous gland with characteristic holocrine secretion. To address this possibility directly, we collected and then pooled meibomian gland lipid secretions into chloroform (1 mg/ml; [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] from four normal individuals for adsorption [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] , relaxation [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] , compression [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] , D-E)/ expansion isocycling, Brewster microscopy [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] , and viscoelasticity [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] , G-H) studies-again all at 35 C. We also performed Raman microscopy [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] , I-K) using in part meibum collected from 27 other normal individuals. N-94 and N-94/C-6 introduced into the PBS subphase rapidly penetrated into the overlying meibum film with kinetics [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] ; line fit R 2 ≥ 0.96) in keeping with a two-step reaction model:
[formula] A ! k1 B ! k2 C [/formula]
in which A to B could be due to docking and B to C to incorporation. The equation describing such a mechanism [fig_ref] Figure 5: Slow tear release of topically applied 125 I-N-94 and 125 I-N-94/C-6 from... [/fig_ref] derives respective N-94 and N-94/C-6 k1 values of 2.678e −02 and 2.139e −02 s −1 , as well as 3.434e −04 and 2.411e −03 s −1 for k2. The implication is that putative docking is more rapid than incorporation and that N-94/C-6 does the latter much more quickly than N-94 [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] -in keeping with N-94/C-6's superior performance in tear viscoelasticity studies [fig_ref] Figure 2: Figure 2 [/fig_ref] , D-E). Yet both peptides are film stabilizing, as per elevated compression values [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] , D-E), superior film thickness [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] , and a slight reduction in the maximal loss factor to 0.4 (lg frequency 1.5) from 0.5 (lg frequency 3.3; [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref]. To confirm this interaction, we flowed 250 μM N-94/C-6 in PBS onto C 18 -functionalized fused-silica coverslips dosed with sufficient meibum to form a 10 to 20 μm film for analysis by Raman microscopy at 35 C. Higher concentration N-94C-6 was necessary for detection. Raman microscopy monitors inelastic light scattering at frequencies less than or greater than incident light with the difference from incident known as a Raman shift [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref]. A Raman shift not prominent in meibum but characteristic of aromatic groups contributed by N-94/C-6's three phenylalanines provided evidence for incorporation [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] that in turn transformed meibum into a continuous, thicker film [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] -the latter in agreement with an increase in the meibum CH2-twisting mode at 1300 cm −1 [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref]. Thus, N-94 and N-94C-6, as surrogates for natural C-terminal lacritin proteoforms in normal tears but substantially lacking in dry eye tears, rapidly interact with and stabilize the tear lipid layer.
Surfactant protein B is thought to superficially associate with underlying anionic phospholipids of pulmonary surfactant (46), whereas surfactant protein C tilts into the membrane (47)-both to ease spreading during respiration. How might N-94/C-6-and N-94-like proteoforms associate and with what affinity in films not subject to prior restructuring by isocycling? We synthesized N-94/C-6, or negative control C-95, with polyethylene glycol-linked Cy3 thiol-coupled to an added Nor C-terminal cysteine [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref]. We then reconstituted peptides to 6 μM in PBS and overlaid each with meibum or (O-acyl)-ω-hydroxy fatty acid (OAHFA) film. OAHFAs are thought to reside at the meibum/aqueous interface [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref] , inset), in keeping with their amphiphilic (both lipophilic and hydrophilic) nature. Here they are presumed to contribute to an elastic monolayer (48) essential for tear stability. OAHFAs appear to be the only lipid class downregulated in dry eye [bib_ref] Meibum lipid composition in Asians with dry eye disease, Lam [/bib_ref] , and transgenic mice lacking fatty acid ω-hydroxylase in the cornea and meibomian gland (and thus ω-O-C16:1 OAHFA's and type 2ω wax diesters) develop a form of dry eye characterized by increased blinking, corneal damage, meibomian orifice plugging, and decreased tear breakup time although tearing is normal. We synthesized the OAHFA 16-(O-oleoyloxy)hexadecanoic acid (also known as 16-(O-oleoyloxy) palmitic acid; [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref] as per Balas et al. [bib_ref] Regiocontrolled syntheses of FAHFAs and LC-MS/MS differentiation of regioisomers, Balas [/bib_ref] and estimated that $14.3 Â 10 [bib_ref] A new mutation in TP63 is associated with age-related pathology, Holder-Espinasse [/bib_ref] 16-(O-oleoyloxy)hexadecanoic acid molecules should be required to cover a PBS subphase with a surface area of $9.5 Â 10 13 nm 2 assuming $1.5 molecules per nm 2 -as determined for dipalmitoylphosphatidylcholine or phosphatidylcholine [bib_ref] Molecular dynamics simulations of phosphatidylcholine membranes: a comparative force field study, Piggot [/bib_ref]. Prior to overlaying OAHFA or meibum onto the PBS subphase, we measured the fluorescence of [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] with "meibum" lipid layer (orange) and underlying aqueous layer (aqua), and covering eyelid containing multiple meibomian glands from which meibum derives. Small dashed box indicates the region highlighted in [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref] (inset). Human meibum was collected by lid compression and dissolved in chloroform (1 mg/ml). B, 47 μg of human meibum spotted on a 80 ml PBS subphase was equilibrated for 15 min under cover while chloroform evaporated and then subjected to compression isocycling reaching 15 mN/m surface pressure at which time surface pressure was monitored as N-94 or N-94/C-6 was introduced into the subphase at a final concentration of 6 μM. Shown are individual replicates for each experiment performed in triplicate (>29,000 data points per variable). Inset, comparative surface pressure attained (mean with S.D [n = 3], ****p < 0.0001; [two-way ANOVA with Sidak multiple comparisons test]). C, time-dependent relaxation of surface tension after preequilibrated human meibum films (without or with 6 μM N-94 or N-94/C-6) were subjected to a sudden step compression of less than 5% of the prior surface area. Shown are individual replicates for each experiment performed in triplicate (>8000 data points per variable). Cy3-N-or -C-terminal labeled N-94/C-6 or C-95 in PBS as the "T0" values [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref]. Topping this with OAHFA or meibum under gentle rotation for 30 min at 35 C tested the affinity of peptides for each film, as inversely reflected by the level of Cy3 fluorescence in the PBS subphase ("T1") according the equation below. K a is the association constant:
[formula] K a ¼ T0−T1 T1 [/formula]
K a 's of Cy3-N-94/C-6 (respectively 0.35 ± 0.03 and 0.21 ± 0.04 for OAHFA and meibum) exceeded those of N-94/C-6-Cy3 (0.14 ± 0.03 and 0.03 ± 0.03) by 2.5-to 7-fold [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref] , D-E), implying preferential association of the C-terminal half. This is in keeping its C-terminal net positive charge (pI of 10.3 versus 4.6 for C-terminal ten amino acids versus nine N-terminal amino acids) and superior hydrophobicity (five versus three amino acids with nonpolar side chains). In contrast, C-95 Ka's were at background levels similar to Cy3 alone [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref].
While not (or minimally) disturbing the OAHFA or meibum film, we next replaced 2/3 of the subphase with fresh PBS [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref]. Our purpose was to ask whether peptide dissociation from OAHFA or meibum was detectable although not apparent by Raman microscopy [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] ("T2"). The dissociation constant (K d ) of N-94/C-6 was estimated as:
[formula] K d ¼ 1 ðT1 À T2Þ=T2 [/formula]
This suggested a K d of 1.1 from meibum and OAHFA [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref].
For validation, advantage was taken of N-94's penultimate Cterminal tryptophan [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref] that is accordingly absent from N-94/C-6. In PBS, N-94 displays a fluorescence optimum of 361 nm after excitation at 280 nm implying exposure (51) appropriate for quenching. Tryptophan quenching in meibum was impractical with background expected from its numerous apparent protein constituents-including lacritin [bib_ref] Proteomic analysis of human meibomian gland secretions, Tsai [/bib_ref]. Chloroform is a polar solvent often employed in partition studies with aqueous solutes. When 6 μM N-94 was added for 30 min to PBS overlying a chloroform subphase at 34 C, $72 ± 10% became quenched [fig_ref] Figure 6: Tear serine proteases, aminopeptidases, and metalloproteinases may contribute to the generation of... [/fig_ref] and a thin film formed at the interface suggesting affinity but inability to partition. Studies were accordingly performed with OAHFA as described above in which 52% of N-94 and only 15% of equimolar "Ctrl pep" were quenched [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref]. This corresponded to respective (T0 − T1)/ T1 K a 's of 1.21 ± 0.45 and 0.2 ± 0.09 and an N-94 1/(T1 − T2)/ T2 K d of 1.20 ± 0.27 [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref]. "Ctrl pep" was syndecan 1 "Pep30-50" peptide (53), also with a single tryptophan [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref] but with PSIPRED-predicted random coil structure and fewer amino acids with nonpolar (30 versus 44%) or basic (5 versus 25%) side chains. The fluorescent contribution of 1/(T1 − T2)/ T2 disassociated N-94 was calculated via the equation below that takes into consideration peptide in residual PBS. Dissociated N-94 was intact by matrix-associated laser desorption/ ionization (MALDI) mass spectrometry [fig_ref] Figure 6: Tear serine proteases, aminopeptidases, and metalloproteinases may contribute to the generation of... [/fig_ref] and restored homeostasis to interferon-γ and TNF-stressed human corneal epithelial cell cultures [fig_ref] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and... [/fig_ref] , as performed after calculating the quantity released per the equation:
[formula]  K ðT0−T1Þ a  K ðT1−T2Þ d  fluorescence T1 − ð0:33Þfluorescence T1 à =  K ðT1−T2Þ d þ 1 à [/formula]
Molar amounts were then obtained by coupling this value to the extinction coefficient, as confirmed by immunodot blot analysis [fig_ref] Figure 6: Tear serine proteases, aminopeptidases, and metalloproteinases may contribute to the generation of... [/fig_ref]. Thus N-94 and N-94/C-6, as surrogate C-terminal lacritin proteoforms, appear to preferentially associate with meibum and OAHFA of the tear lipid layer through their C-termini and (at least in nonstructured films) are subject to release at a level not detected by Raman microscopy [fig_ref] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or... [/fig_ref] nor by Langmuir surface balance (not shown) but sufficient to restore epithelial homeostasis.
## Release time slows with repeated topical application
Topical application of 5-Dodecanoylaminofluorescein (94 mM) and sodium fluorescein (0.13 mM) onto human eyes suggests respective turnover rates of 0.93 ± 0.36% and 10.3 ± 3.7% per minute, respectively (54) over a total of 108 ± 39 and 9.7 ± 4 min from respective lipid and aqueous portions of tears. The former value can also be considered the lipid release time. We synthesized N-94 and N-94/C-6 each with an added Cterminal tyrosine for iodination and performed ocular and systemic pharmacokinetic studies in rabbits. Human meibum and rabbit meibum display significant compositional differences, likely in keeping a much slower blink rate and greater tear stability in rabbits. Nonetheless, both contain OAHFA [bib_ref] Toward an animal model of the human tear film: biochemical comparison of..., Butovich [/bib_ref]. By assessing residence time in tears, information can be gained on tear lipid affinity in vivo. 4 μM 125 I-N-94 or 44 μM 125 I-N-94/C-6 were applied twice daily to rabbit eyes for 3 and 4 days, respectively [fig_ref] Figure 5: Slow tear release of topically applied 125 I-N-94 and 125 I-N-94/C-6 from... [/fig_ref] , A-C), or respectively as three 1.3 μM doses over 10 min or as a single 44 μM dose [fig_ref] Figure 5: Slow tear release of topically applied 125 I-N-94 and 125 I-N-94/C-6 from... [/fig_ref]. Concentrations chosen were in support of a phase 2 human trial. Turnover was exponential with data best fitting the equation:
[formula] R ¼ R 0 þ A:EXP½ À t=τ [/formula]
where R = radioactivity, R 0 = baseline radioactivity, A is a coefficient, t = time and τ = release (turnover) time. Accordingly, N-94 and N-94/C-6 release times were respectively 400 ± 95 and 2000 ± 500 min for multiday dosing, and 13 ± 2 and 22 ± 4 min following 10 min of dosing. The same was true for the meibomian gland although at lower levels that were not statistically different from baseline [fig_ref] Figure 5: Slow tear release of topically applied 125 I-N-94 and 125 I-N-94/C-6 from... [/fig_ref]. Radioactivity in other tissues and fluids remained near or at baseline [fig_ref] Figure 5: Slow tear release of topically applied 125 I-N-94 and 125 I-N-94/C-6 from... [/fig_ref]. (54)). If N-94 and N-94/C-6, as proxy for lacritin proteoforms, have a long residence time in tears, are they resistant to tear proteases? We incubated 50 μM N-94, N-94/C-6, or as positive control similar-sized laminin "SN-peptide" ((56); predicted β-sheet) with normal basal tears for 4 h at 37 C. By MALDI mass spectrometry, only N-94 and N-94/C-6 remained intact [fig_ref] Figure 5: Slow tear release of topically applied 125 I-N-94 and 125 I-N-94/C-6 from... [/fig_ref] -perhaps aided by their helical structure that makes backbone amide bonds less accessible to proteases and/or by relevant inhibitors resident in tears. Thus N-94 and N-94C-6 are relatively protease-resistant, a property possibly enhanced by their association with the tear lipid layer.
Tear serine proteases, aminopeptidases, and metalloproteinases may contribute to the generation of C-terminal lacritin proteoforms C-terminal processing of lacritin's putative homolog dermcidin is the responsibility of extracellular carboxypeptidases, an endopeptidase and the aspartyl protease cathepsin D-all in human sweat [bib_ref] Cathepsin D is present in human eccrine sweat and involved in the..., Baechle [/bib_ref]. The outcome is the bactericidal peptide "SSL-25." Tear proteases include: alanyl aminopeptidase, arginyl aminopeptidase, complement factor B, cathepsins (B, D, G, S), dipeptidyl-peptidase 4, HtrA serine peptidase 1, matrix metallopeptidase-9 and -10, plasma kallikrein, plasminogen (plasmin), serine protease 8, serine carboxypeptidase, coagulation factor II (thrombin) and trypsin 1 [bib_ref] Lacritin and the tear proteome as natural replacement therapy for dry eye, Karnati [/bib_ref]. Tears are also rich in protease inhibitors [bib_ref] Lacritin and the tear proteome as natural replacement therapy for dry eye, Karnati [/bib_ref]. Of lacritin's 42 known C-terminal proteoforms (30), 22 (including the longest) share N-termini residing within the lacritin sequence "LKSIVEKSILLTEQALAKAGKGMH" represented by synthetic peptide N-64/C-31 (Figs. 1A and 6A; amino acids 65-88 of lacritin's 119 amino acids). In silico analysis by PROSPER (58) predicts LK|SIVE and AGKG|MH cleavage by matrix metallopeptidase-9 (M10.004) and SILL|TEQA by chymotrypsin-like serine protease cathepsin G (S01.133). N-64/C-31 is also predicted (ExPASY PeptideCutter (59)) to be sensitive to the serine protease glutamyl endopeptidase I (S01.269; SIVE|KSIL and LLTE|QALA) from eye commensal S. epidermidis and pathogen S. aureus, and to trypsin-like serine proteases (IVEK|SILL; ALAK|AGKG; KAGK|GMH). To assess which ones may contribute to processing, we first optimized tear volume such that all N-64/C-31 (50 μM) was fully hydrolyzed by 48 h at 35 C and neutral pH [fig_ref] Figure 6: Tear serine proteases, aminopeptidases, and metalloproteinases may contribute to the generation of... [/fig_ref]. The assay was then repeated in the absence or presence of eight different proteolytic inhibitors at standard (1Â) or fivefold higher concentrations (except EDTA at 0.5 and 1x) for endpoint analysis by semiquantitative MALDI mass spectrometry in which 4 μM N-94/C-6 was spiked into the digest immediately before mixing with MALDI matrix sinapinic acid [bib_ref] Protein quantitation using mass spectrometry, Zhang [/bib_ref]. Processing was inhibited in a dose-dependent manner by AEBSF, bestatin, EDTA, leupeptin, or fully by boiling. Not effective were acivicin, antipain, chymostatin, and pepstatin. This is in keeping with involvement of tear cysteine proteases of the C1 and C2 families, metalloproteinases of the M1 and M10 B families, and serine proteases of the S1 family. Candidate N-64/C-31 tear proteases therefore include: cathepsin B (C1); calpain (C2); alanyl aminopeptidase, arginyl aminopeptidase (M1); MMP9, MMP10 (M10 B); cathepsin G, plasma kallikrein, plasmin, thrombin, and trypsin (S1).
# Discussion
How and which proteins help prevent premature collapse of the complex lipid film at the air/liquid interface of the eye necessary for vision? Using proxy synthetic peptides N-94 and N-94/C-6, we report that proteoforms from the C-terminus of the tear glycoprotein lacritin are essential through their rapid and stable insertion into tear lipids, including C16:1 OAHFA presumed to reside at the lipid/aqueous interface. This minimizes the loss factor between 10 −3.7 and 10 −2.3 Hz as a measure of viscosity by restoring elasticity to dry eye tears that otherwise are subject to premature collapse. We further report that C-terminal lacritin proteoforms are selectively deficient in dry eye. Interestingly, as proxy proteoforms gradually cycle off OAHFA in nonpressured films, bioactivity sufficient to restore homeostasis of corneal epithelial cells is retained-a slow release role for extracellular lipid films never previously appreciated.
C-terminal lacritin proteoforms were first noted as proteolytic fragments by western blotting of human tears [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref] and in tear bactericidal assays leading to the discovery of cleavagepotentiated C-terminal "N-104". This was later validated by Azkargorta et al. [bib_ref] Human basal tear peptidome characterization by CID, HCD, and ETD followed by..., Azkargorta [/bib_ref] through top-down sequencing of tears and identification of smaller "N-106" and "N-107" proteoforms that were also bactericidal. Their discovery of at least 40 additional C-terminal lacritin proteoforms of increasing size [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] paralleled a smaller library of lacritin synthetic peptides and recombinant fragments previously generated to dissect lacritin's ocular mitogenic [bib_ref] Restricted epithelial proliferation by lacritin via PKCα-dependent NFAT and mTOR pathways, Wang [/bib_ref] , prosecretory [bib_ref] ) cDNA and genomic cloning of lacritin, a novel secretion enhancing factor..., Sanghi [/bib_ref] [bib_ref] Lacritin, a novel human tear glycoprotein, promotes sustained basal tearing and is..., Samudre [/bib_ref] [bib_ref] Topical administration of lacritin is a novel therapy for aqueous-deficient dry eye..., Vijmasi [/bib_ref] [bib_ref] A thermo-responsive protein treatment for dry eyes, Wang [/bib_ref] [bib_ref] Lacritin-induced secretion of tear proteins from cultured monkey lacrimal acinar cells, Fujii [/bib_ref] , prohomeostatic [bib_ref] Lacritin rescues stressed epithelia via rapid forkhead box o3 (FOXO3)-associated autophagy that..., Wang [/bib_ref] , and cleavage potentiated bactericidal [bib_ref] Human basal tear peptidome characterization by CID, HCD, and ETD followed by..., Azkargorta [/bib_ref] activities. All focused attention on lacritin's two C-terminal amphipathic α-helices required for ligation of heparanasemodified cell surface syndecan-1 (53, 63) necessary for epithelial targeting. Their similarity to N-and C-termini of processed lipid stabilizing pulmonary surfactant protein B [bib_ref] Processing of pulmonary surfactant protein B by napsin and cathepsin H, Ueno [/bib_ref] [bib_ref] The role of charged amphipathic helices in the structure and function of..., Waring [/bib_ref] [bib_ref] Peptide-based synthetic pulmonary surfactant for the treatment of respiratory distress disorders, Braide-Moncoeur [/bib_ref] , together with lacritin's detection in broncheoalveolar lavage [bib_ref] Differential proteomic analysis of bronchoalveolar lavage fluid in asthmatics following segmental antigen..., Wu [/bib_ref] [bib_ref] In-depth proteomic analysis of human bronchoalveolar lavage fluid toward the biomarker discovery..., Sim [/bib_ref] , and selective lacritin and C-terminal proteoform deficiency in dry eye were rationale for testing with tear lipids.
Early studies of floating lipid films with an air interface debated whether nonmucinous glycoproteins (64), or any protein [bib_ref] The physical properties of an effective lung surfactant, Bangham [/bib_ref] , contributed substantially to integrity whereby rupture is minimized at low and high shear rates in a characteristic non-Newtonian viscoelastic manner [bib_ref] The viscosity of human tears, Tiffany [/bib_ref]. Later attribution of lung alveolar collapse in newborns to a variety of genetic mutations [bib_ref] The molecular era of surfactant biology, Whitsett [/bib_ref] and biophysical analyses with both lung surfactant proteins and candidate tear proteins (albumin [bib_ref] Insertion of tear proteins into a meibomian lipids film, Miano [/bib_ref] [bib_ref] Surface pressure measurements of human tears and individual tear film components indicate..., Tragoulias [/bib_ref] , keratin (69), lipocalin-1 [bib_ref] Structural changes in human tear lipocalins associated with lipid binding, Gasymov [/bib_ref] [bib_ref] Insertion of tear proteins into a meibomian lipids film, Miano [/bib_ref] [bib_ref] Surface pressure measurements of human tears and individual tear film components indicate..., Tragoulias [/bib_ref] [bib_ref] Human tear viscosity: an interactive role for proteins and lipids, Gouveia [/bib_ref] [bib_ref] Adsorption of human tear lipocalin to human meibomian lipid films, Millar [/bib_ref] , lactoferrin [bib_ref] Insertion of tear proteins into a meibomian lipids film, Miano [/bib_ref] [bib_ref] Surface pressure measurements of human tears and individual tear film components indicate..., Tragoulias [/bib_ref] [bib_ref] Human tear viscosity: an interactive role for proteins and lipids, Gouveia [/bib_ref] , lysozyme [bib_ref] Insertion of tear proteins into a meibomian lipids film, Miano [/bib_ref] [bib_ref] Surface pressure measurements of human tears and individual tear film components indicate..., Tragoulias [/bib_ref] [bib_ref] Human tear viscosity: an interactive role for proteins and lipids, Gouveia [/bib_ref] [bib_ref] Adsorption of lysozyme to phospholipid and meibomian lipid monolayer films, Mudgil [/bib_ref] validated how non-Newtonian behavior necessary to resist rupture is a consequence of complex protein-lipid interactions. Much remains to be learned about such interactions. Lipocalin-1 (10 μM) alone with captured lipid or mixtures of lactoferrin (21 μM) and lysozyme (136 μM) are non-Newtonian (70), yet tears without lipids are not [bib_ref] Human tear viscosity: an interactive role for proteins and lipids, Gouveia [/bib_ref]. Soluble bovine ocular mucins can interact with and stabilize tear lipids [bib_ref] The surface activity of purified ocular mucin at the air-liquid interface and..., Millar [/bib_ref] [bib_ref] Contribution of mucins towards the physical properties of the tear film: a..., Georgiev [/bib_ref] , but alone lack non-Newtonian behavior at physiological concentrations [bib_ref] The surface activity of purified ocular mucin at the air-liquid interface and..., Millar [/bib_ref]. That tear lipid binding proteins may functionally interact in modules, as suggested via Cytoscape for lacritin with lipocalin-1, apolipoprotein, lactoferrin, and others (75) is intriguing, although direct binding of these is not apparent in BioGrid. A coupled immunodepletion/rescue approach offered the novel opportunity to explore complex films and fluids in an otherwise undisturbed condition, making possible discovery of and dissection of lacritin's contribution through C-terminal proteoforms to the stability of whole tears and to tear lipids. That C-terminal lacritin proteoforms are tear stabilizing aligns with changes of the tear lipid spreading rate and stability in health and disease [bib_ref] Differentiation of lipid tear deficiency dry eye by kinetic analysis of tear..., Goto [/bib_ref] as it is well known that tightly packed, water insoluble lipid films (and hence those with higher lift-off areas and maximum surface pressure) are in general more elastic and able to recover after blink-like deformation [bib_ref] The monolayer technique: a potent tool for studying the interfacialproperties of antimicrobial..., Maget-Dana [/bib_ref]. Further, the capacity of C-terminal proteoforms to restore elasticity is in keeping with dilatational properties of surfactant layers that define the resistance of the air/water surface of wetting tear, alveolar [bib_ref] Lung surfactants and different contributions to thin film stability, Hermans [/bib_ref] , or other films to extensional deformations caused by capillary waves or hydrodynamic phenomena and play a key role in the overall stability [bib_ref] Mechanism of tear film rupture and formation of dry spots on cornea, Sharma [/bib_ref]. Indeed dilatational rheology differed substantially [bib_ref] Surface relaxations as a tool to distinguish the dynamic interfacial properties of..., Georgiev [/bib_ref] [bib_ref] Tear lipids interfacial rheology: effect of lysozyme and lens care solutions, Svitova [/bib_ref] between meibum from healthy individuals and from patients with meibomian gland disease. This was also true for contact lens lipid extracts collected from Caucasians versus those from Asians-the latter with a higher risk for dry eye disease. Thus, C-terminal proteoforms act as surfactants to promote tear viscoelasticity. As surfactants, they also reduce surface tension toward maintenance of a stable, "healthy" tear film. Stability suffers from lacritin downregulation in dry eye and would be further exacerbated by loss of OAHFA at the lipid/aqueous interface.
## Experimental procedures
Synthetic peptides "N-94" (KQFIENGSEFAQKLLKKFSLLKPWA) and "N-94/ C-6" (KQFIENGSEFAQKLLKKFS), respectively representing the C-terminal active 25 or 19 amino acids of human lacritin [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] were manufactured by PolyPeptide Group (San Diego, CA) with amino terminal acetylation and carboxy terminal amidation (N-94/C-6; i.e., absent of 94 N-terminal and 6 C-terminal amino acids) or only amino terminal acetylation (N-94), both under GMP conditions, and with trifluoracetic acid removed in place of acetate. Purity was respectively 98.9 and 97.2%. Numbering of synthetic peptides [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] and recombinant proteins excludes the signal peptide. N-94/C-6-Y and N-94-Y with added C-terminal tyrosine, control lacritin peptides N-64/C-31 (LKSIVEKSILLTEQALAKAGKGMH) with amino terminal acetylation and carboxy terminal amidation and C-95 (EDASSDSTGADPAQEAGTSKPNEE) with carboxy terminal amidation, as well as additional N-94 and N-94/C-6-both with terminal modifications as noted abovewere synthesized by Genscript (Piscataway, NJ) and completed as acetate salts with respective purity of 95.9, 95.4, 98.1, 97.9, 95.7, and 96.4%. Also, Genscript synthesized Cy3-labeled N-94/C-6 (Cy3-{PEG2}CKQFIENGSEFAQKLLKKFS ["Cy3-N-94/C-6"] with amidated C-terminus and KQFIENGSE FAQKLLKKFSC{PEG2}-Cy3 ["N-94/C-6-Cy3"] with acetylated N-terminus, as well as as Cy3-labeled C-95 (Cy3-{PEG2}C EDASSDSTGADPAQEAGTSKPNEE ["Cy3-C-95"] with amidated C-terminus and EDASSDSTGADPAQEAGTSKP NEEC{PEG2}-Cy3 ["C-95-Cy3C"] as trifluoracetic acid salts with respective purity of 95, 97.2, 95.1 and 97%. Syndecan-1 peptide "Pep30-50" (QDITLSQQTPSTWKDTQLLT; [bib_ref] Targeting of heparanasemodified syndecan-1 by prosecretory mitogen lacritin requires conserved core GAGAL..., Zhang [/bib_ref] and laminin peptide "SN-peptide" (SINNNRWHSIYITR FGNMGS; (56)) -all with amino terminal acetylation and carboxy terminal amidation were synthesized as trifluoracetic acid salts with respective purity of 97.6, 96.4, and 96.1%. All synthetic peptides were validated by electrospray ionization mass spectrometry. Aliquotted synthetic peptides were stored lyophilized at −70 C in a dry environment.
Tears, antibodies, meibum, OAHFA synthesis, Langmuir surface balance experiments, and Raman microscopy Collection of all human samples was approved by Institutional Review Boards as specified below and abides by the Declaration of Helsinki ethical principles. Eighty-five basal tear samples from over 50 individuals (median age 30.2 years; 54% female) were collected after 0.5% proparacaine anesthesia from both eyes by wicking onto a paper "Schirmer" strip that had been carefully inserted under the center lower lid of the eye for 5 min. Wicking of >15 mm or ≤6 mm of tears was respectively considered normal or evidence of dry eye. Tears on strips were immediately stored at −80 C. Approval was from the Walter Reed National Military Medical Center Institutional Review Board with informed consent. Tears collected on Schirmer strips without anesthesia from 21 patients with Secondary Sjögren's Syndrome were kindly provided by the Sjögren's International Collaborative Clinical Alliance (University of California, San Francisco) with Institutional Review Board approval at each of the Alliance collection sites. Only samples that had wicked ≤6 mm were tested. Tears were eluted from Schirmer strips immediately prior to experiments, by soaking in 25 μl of ice-cold PBS with protease inhibitors (Roche Complete Mini Inhibitor Cocktail; Sigma Chemical Co, St Louis MO) for 30 min on ice, and subsequent centrifugation at 20,000g for 30 min. A pool of 50 collected tears was subdivided into four pools. One was passed over a preimmune rabbit Ig column ("mock depletion"). Others were subjected to lacritin immunodepletion with rabbit polyclonal antibody "ab C-term" with specificity for lacritin's C-terminal 54 amino acids [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref] , as previously performed [bib_ref] Lacritin rescues stressed epithelia via rapid forkhead box o3 (FOXO3)-associated autophagy that..., Wang [/bib_ref] [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref]. Another pool of five collected tears was subject to immunodepletion with mouse monoclonal antibody 1F5 directed against lacritin N-terminal synthetic peptide DPAQEAGTSKPNEEIS (amino acids [bib_ref] Mutations in different components of FGF signaling in LADD syndrome, Rohmann [/bib_ref] [bib_ref] Mutations in NGLY1 cause an inherited disorder of the endoplasmic reticulumassociated degradation..., Enns [/bib_ref] [bib_ref] A new mutation in TP63 is associated with age-related pathology, Holder-Espinasse [/bib_ref] [bib_ref] A novel TRAPPC11 mutation in two Turkish families associated with cerebral atrophy,..., Koehler [/bib_ref] [bib_ref] Meibum lipid composition in Asians with dry eye disease, Lam [/bib_ref] [bib_ref] Hyaline membrane disease, Farrell [/bib_ref] [bib_ref] The real reason for having a meibomian lipid layer covering the outer..., Millar [/bib_ref] [bib_ref] Differential proteomic analysis of bronchoalveolar lavage fluid in asthmatics following segmental antigen..., Wu [/bib_ref] [bib_ref] In-depth proteomic analysis of human bronchoalveolar lavage fluid toward the biomarker discovery..., Sim [/bib_ref] [bib_ref] ) cDNA and genomic cloning of lacritin, a novel secretion enhancing factor..., Sanghi [/bib_ref] [bib_ref] Lacritin, a novel human tear glycoprotein, promotes sustained basal tearing and is..., Samudre [/bib_ref] [bib_ref] Topical administration of lacritin is a novel therapy for aqueous-deficient dry eye..., Vijmasi [/bib_ref] [bib_ref] A thermo-responsive protein treatment for dry eyes, Wang [/bib_ref] [bib_ref] Lacritin rescues stressed epithelia via rapid forkhead box o3 (FOXO3)-associated autophagy that..., Wang [/bib_ref] [bib_ref] Tissue transglutaminase is a negative regulator of monomeric lacritin bioactivity, Velez [/bib_ref]. 1F5 (144 μg IgG1) or 432 μg ab C-term or preimmune rabbit Ig in 120 μl was immobilized on 0.2 ml protein A/G spin columns (Thermo Nab #89950, Thermo Scientific, Rockford, IL) using end-over-end mixing for 10 min at room temperature. Columns were washed with 20 bead volumes of 200 μl each of binding buffer and then similarly incubated end-over-end with pooled normal tears (315 μl/column) for 18 h at 4 C. The "lacritin N-term-" and "lacritin C-term-depleted" or mockdepleted tear flow through were collected by centrifugation (5,000 x g, 1 min at 4 C) with validation by ab C-term western blotting using LI-COR. No column leaching of antibody was detected by secondary alone western blotting. Dry eye, normal, or normal depleted tears were lyophilized for shipment on dry ice (−79 C) for Langmuir surface balance experiments. Also shipped in this manner were lyophilized N-94, N-94/C-6, N-64/C-31, and C-95. Quantitation of lacritin monomer and Cterminal proteoforms was done by densitometery on tear Western blots immunostained with "anti-C-term" lacritin antibodies. All data were normalized to the maximum monomer per blot.
Meibum was collected from four normal individuals under the auspices of the Kyoto Prefectural University of Medicine Institutional Review Board (per a collaboration with GAG) and 27 other normals per the University of Virginia Insitutional Review Board, both with informed consent. For this purpose, the lower lid was squeezed using opposing cotton applicators or meibomian gland expressor forceps. Collection was into glass vials with 500 μl chloroform (1 mg/ml final concentration) for flash freezing on dry ice and storage at −80 C.
16-(O-oleoyloxy)hexadecanoic acid (also known as 16-(Ooleoyloxy)palmitic acid) was synthesized as described by using a two-step process involving esterification of oleic acid with 1,16-hexadecanediol followed by oxidation of the primary hydroxy group to a carboxylic acid. Purity by proton NMR was ≥95% with a yield of 1.41 gm. 1 H NMR spectra were recorded on a Varian Inova 600 (600 MHz) spectrometer in CDCl 3 with chemical shifts referenced to internal standards (CDCl 3 : 7.26 ppm 1 H).
Langmuir surface balance experiments were performed as previously described (81) using a computer controlled Microtrough XL (Kibron, Helsinki Finland) with a 40 ml volume, an area of 225 cm, and a Wilhelmy wire probe with sensitivity exceeding 0.01 mN/m. Briefly, lyophilized intact normal or dry eye or N-term-or C-term-lacritin-depleted or mock-depleted pooled tears were reconstituted to their initial eluted volume in PBS with water. C-term-lacritin-depleted tears were also reconstituted with added N-94, N-94/C-6, N-64/C-31, or C-95 at a final peptide concentration of 6 μM. Three microliters of tears or 47 μg of pooled meibum, each without or with N-94 or N-94/C-6, were gently deposited as small submicro droplets at the air/PBS surface prewarmed to 35 C, the corneal surface temperature [bib_ref] Temperatures of the ocular surface, lid, and periorbital regions of Sjögren's, evaporative,..., Abreau [/bib_ref]. After equilibration for 15 min under cover to minimize evaporation and to keep dust-free, and with a continuous supply of ozone-depleted air, the film was subjected to experimental manipulation. After advancing or retreating dual opposing barriers at 1.37 cm 2 per second over ten consecutive cycles, surface pressures were compared during compression and expansion isocycling. In other experiments, time-dependent relaxation of surface tension was monitored after pre-equilibrated films at a surface pressure of $15 mN/m were subjected to a sudden step compression of less than 5% of the prior surface areas. Fourier transformation made possible comparative appreciation of dilatational elastic moduli as per Loglio et al. [bib_ref] Fourier transform surface viscoelastic modulus of dilute aqueous solutions of surfactants. Improved..., Loglio [/bib_ref]. Further analyses were performed by Brewster angle microscopy (UltraBAM; Accurion GmbH, Göttingen Germany) to distinguish film regions by depth and continuity and via the pendant drop technique in which 2 μl of normal or lacritin-depleted tears were allowed to equilibrate for $3 min in a saturated vapor measurement cell to prevent evaporation, after which surface pressure was monitored for 60 s. Brewster angle microscopy images were analyzed by ImageJ.
For Raman confocal microscopy, 10 to 20 μm meibum films were created by application of 10 μl of 10 mg/ml meibum in chloroform to a coverslip on a 1 mm  7 mm flow channel (82) followed by drying under nitrogen and then briefly under vacuum with thickness between coverslip-meibum and meibum-buffer interfaces estimated by incident laser beam reflection. The Raman microscopic probe [bib_ref] Measuring diffusion of molecules into individual polymer particles by confocal Raman microscopy, Bridges [/bib_ref] [bib_ref] Single layer graphene for estimation of axial spatial resolution in confocal Raman..., Korzeniewski [/bib_ref] has a depth resolution of ±1.2 μm (FWHM), a diameter of $500 nm, and a probe volume of $500 fl, which is well-matched to the meibum film deposited on the coverslip interface. For analyses, the probe was brought to 1.5 μm below the meibum-buffer interface with spectra collected at a laser power of 100 mW with an integration time of 2 min. After wetting for 10 min with PBS, a meibum spectrum was collected, followed by flow of 250 μM N-94/C-6 in PBS at 0.2 ml/min.
## In vitro and in vivo release kinetics
Cy3 fluorescence signals (excitation = 550 nm; detection = 570 nm) of 100 μl of Cy3 or Cy3-labeled N-94/C-6 (N-or Cterminal labeled) or Cy3-labeled C-95 (N-or C-terminal labeled) in PBS were collected in a black 96-well plate at 35 C in a SpectraMax M3 microplate reader. Low photomultiplier tube power and 1 flash per read (reading from bottom) were applied to minimize bleaching. Wavelength cutoff was 570 nm. After a T0 fluorescence value of 6 μM peptide in PBS was obtained, 6 μM peptide in PBS was added to a final volume of 750 μl in a glass tube. The fluid surface area was 95 mm 2 , similar to the corneal surface area of $132 mm 2 . Onto this, 10 μl of 1 mg/ml OAHFA or meibum in acetone was allowed to spread, followed by rotation at 100 rpm for 30 min (35 C). After removal of 500 μl of the PBS subphase for determination (in a 100 μl aliquot) of the T1 fluorescence, fresh PBS was gently injected into the subphase for an additional 30 min rotation at 35 C. A 100-μl aliquot of 500 μl of the subphase provided the T2 fluorescence value. Fluorescence of 100 μl of PBS was subtracted as background signal. T0, T1, and T2 values facilitated calculation of the association (Ka) and dissociation (Kd) constants of Cy3 or Cy3-labeled peptides into and from the OAHFA or meibum layer, as described in the Results section.
Similarly, tryptophan fluorescence signals (excitation = 280 nm; detection = 325-500 nm) of N-94 or SDC1 "Pep30-50" peptide ("Ctrl pep") in PBS were collected in a quartz cuvette (282 QS 1.000) at 35 C in the cuvette chamber of the same SpectraMax M3 microplate reader at low photomultiplier tube power, 1 flash per read, and wavelength cutoff of 325 nm. Area under the emission spectra ≥325 nm was used as signals. T0, T1, and T2 fluorescence values were collected as described above for Cy3-labeled peptides but with the whole 500 μl sample. Fluorescence of 500 μl of PBS was subtracted as background signal. Only OAHFA films were used.
For cell culture studies, T0 and T2 subphase samples containing N-94 were filter-sterilized for inclusion in human corneal epithelial (HCE-T) cell viability experiments and MALDI mass spectrometry. HCE-T cells were validated by short-tandem repeat profiling. Cells were seeded overnight in 96-well plates at a density of 1.5 x 10 4 cells/ml and then treated with interferon-γ (1000 U/ml; Sigma-Aldrich, St Louis MO) and tumor necrosis factor (100 ng/ml; Peprotech, Cranbury NJ) with or without equal molar amounts of T0 or T2-released N-94, or with C-95, every 2 days for 7 days. Viability was assessed by the alamarBlue assay (Thermo Fisher Scientific, Waltham MA). Quantity of T2 released N-94 was quantitated from fluorescence values fitted into the equation noted in the Results section with mass values obtained via the extinction coefficient. This value was confirmed by immunodot blot analysis versus an N-94 standard curve [fig_ref] Figure 6: Tear serine proteases, aminopeptidases, and metalloproteinases may contribute to the generation of... [/fig_ref].
Rabbit pharmacokinetic studies were performed by Covance Laboratories Inc using Covance standard operating procedures in accordance with the Wisconsin Department of Health Services, Radiation Protection Section, as licensed to Covance, and in compliance with Animal Welfare Act Regulations (9 CFT 3). 44 μM stocks of N-94-Y and N-94/C-6-Y were radioiodinated by Perkin-Elmer (Shelton, CT) to a final specific activity of respectively 24.2 and 2.72 μCi/μg with initial radiopurity respectively 69.62 and 97.65%. At Covance Laboratories Inc (Madison, WI), 35 μl (3.3 μCi) of 1.3 μM 125 I-N-94 in PBS repeated three times over 10 min or a single 35 μl (10 μCi) dose of 44 μM 125 I-N-94/C-6 in 10 mM sodium citrate, 137 mM sodium chloride (pH 6.5) were topically added to each eye of respectively 12 and 14 female pigmented New Zealand White/New Zealand Red F1 cross rabbits (>3 months old, >2500 g each; Covance Research Products; Denver PA). 0.25, 0.5, 1, 3, 6, or 12 h after the first dose, tears were collected (30 s from inferior cul de sac) using Tear Flo Test strips (Accutome, Malvern PA) without prior anesthesia. Blood was also collected (0.5, 1, 3, 6, and 12 h postdose) via an auricular artery canula into tubes containing K 2 EDTA and centrifuged to obtain plasma. Later in the day, 125 I-N-94/C-6 treated eyes were further treated with 35 μl (10 μCi) of 44 μM of 125 I-N-94/ C-6 twice daily for three ( 125 I-N-94/C-6) additional days. The final treatment was on the morning of the fifth day after which tears were collected at 0.5, 1, 3, 6, 12, and 24 h postdose from two animals per time point who were then euthanized for collection of blood and ocular tissues. Eyes of a second group of 14 rabbits were treated twice daily with 35 μl of 4 μM 125 I-N-94 (8 μCi) for 3 days with the final treatment on the morning of the fourth day after which tears were collected at 0.5, 1, 3, 6, 12, and 24 h postdose from two animals per time point who were then euthanized for collection of blood and ocular tissues. Radiopurity determined upon dosing by sizeexclusion chromatography was 67.2% on Day 1 and 45.2% on Day 4 for 125 I-N-94 and 98.8% on Day 1 and 97.9% on Day 5 for 125 I-N-94/C-6. TCA precipitable radioactivity of formulated 125 I-N-94 and 125 I-N-94/C-6 respectively averaged 78% and 94% over the 4 and 5 days of dosing. TCA precipitable radioactivity of samples was assessed by scintillation counting and expressed per tissue wet weight.
## Mass spectrometry
Stability of 50 μM N-94 and N-94/C-6 in 10 μl of human basal tears for 4 h at 37 C was assessed by MALDI TOF mass spectrometry (Bruker MicroFlex) by diluting the digest 1:20 with 0.1% trifluoracetic acid for 1:1 application in a sinapinic acid matrix in 50% acetonitrile and 0.1% trifluoracetic acid for ionization. Mass measurements were taken in linear, positive mode. To appreciate how C-terminal proteoforms may be processed, 50 μM N-64/C-31 was incubated at 35 C for 48 h with 38 μl of basal normal tears without or with individual inhibitors (G Biosciences, St Louis MO) diluted from stock as AEBSF (0.2 or 1 mM), antipain (74 or 370 μM), bestatin (130 or 650 μM), chymostatin (100 or 500 M), EDTA (2.5 or 5 mM), leupeptin (10 or 50 μM), and pepstatin (1 or 5 μM) in a reaction volume of 50 μl buffered by 20 mM HEPES, 150 mM NaCl, 10 mM CaCl 2 (pH 7.4), or instead with 5% DMSO in the same buffer as vehicle control. Samples were then processed as above after spiking in fresh 4 μM N-94/C-6 during the 0.1% TFA dilution for combination with sinapinic acid matrix.
## Statistical analyses
All experiments were performed at least three times, with the exception of some Brewster angle microscopy and 125 I-N-94 and -N-94/C-6 pharmacokinetic studies. Data are reported as the mean ± SD with statistical approaches detailed in figure legends, as performed in Prism 8.4.3.
## Data availability
Source data for [fig_ref] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film... [/fig_ref] and Figures S1-S4 and S6 are provided in Data files, as indicated in each figure legend. Other supportive data of this study are available by request to the corresponding author.
[fig] Figure 1: Lacritin C-terminal proteoforms in normal human tears are essential for tear film stability. A, linear diagram of secreted lacritin (lacrt) monomer with rectangles indicative of demonstrated (C-terminal half; [/fig]
[fig] Figure 2: Figure 2. Tear film stabilizing activity of lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore normal viscoelasticity. A, time-dependent relaxation of surface tension after pre-equilibrated anti-C-term lacritin-depleted (without or with 6 μM N-94 or N-94/C-6) or normal human basal tear films formed as per Figure 1 were subjected to a sudden step compression of less than 5% of the prior surface area. ****p < 0.0001; ns, not significant [two-way ANOVA with Sidak multiple comparisons test]. B, same procedure applied to human dry eye basal tear films (without or with 6 μM N-94 or N-94/C-6). For both A and B, shown are individual replicates for each experiment performed in triplicate (>2300 data points per variable with the exception of anti-C-term lacritin-depleted tears plus 6 μM N-94 [one replicate] and anti-C-term lacritin-depleted tears alone [1260 data points]). ****p < 0.0001 with top bracket a comparison to normal [two-way ANOVA with Sidak multiple comparisons test]. Inset, schematic diagram of time-dependent relaxation. C-G, Fourier transform (44) of the relaxation data in A-B as the stored elastic (E R ) or loss (E IM ) moduli, or loss factor (tan Φ)-both as a function of the log frequency. Color code of data is per A-B. C, stored elastic and loss moduli of normal human basal tears. D, stored elastic and loss moduli of anti-C-term lacritin-depleted human basal tears without or with N-94 or N-94/C-6 supplementation. E, stored elastic and loss moduli of human dry eye basal tears without or with N-94 or N-94/C-6 supplementation. F, loss factor of normal human basal tears versus anti-C-term lacritin-depleted tears without or with added 6 μM N-94 or N-94/ C-6; horizontal line: values less than or greater than 1 are respectively elastic or viscous. G, loss factor of human dry eye basal tears without or with added 6 μM N-94 or N-94/C-6. H, light incident to the Brewster angle reflects in a manner corelative to regional lipid thickness (top schematic). Shown is the reflection off normal, anti-C-term lacritin-depleted (without or with 6 μM N-94 or N-94/C-6), or dry eye (without or with 6 μM N-94 or N-94/C-6) basal tear films. Bar = 1 μm. Representative of triplicate experiments. Data for A-G in Data files Figure S2. [/fig]
[fig] Figure 3: Lacritin C-terminal proxy proteoforms N-94 and N-94/C-6 restore viscoelasticity in whole or part through rapid insertion into the tear lipid layer. A, schematic diagram of human basal tears from dashed boxed in [/fig]
[fig] Figure 4: Slow release of N-94 and Cy3-N-94/C-6 from floating, gently rotating OAHFA and meibum films. A, synthetic peptides with fluorescent Cy3 tag (i), or tryptophan (ii). B, 1 H NMR spectrum (600 MHz; deuterated chloroform) of 16-(oleoyloxy)hexadecanoic acid (OAHFA) synthesized using the method of Balas et al.(49). Inset left, suspected location of OAHFA's at the aqueous (aqua)/lipid (orange) interface; for orientation, refer to box inFigure 3A. Inset right, linear structure of synthesized 16-(oleoyloxy)hexadecanoic acid. C, procedure with PBS and (i) or (ii) individual peptide-containing quartz cuvettes (actually rectangular) that were overlayed with OAHFA or meibum films before and after fluorescent excitation with intermittent rotation for 30 min. Two-thirds of the PBS were then replaced with fresh PBS, followed by rotation and fluorescent excitation.D, calculated Ka or Kd of Cy3 tagged peptides in cuvettes with OAHFA (left graph: mean with S.D [n = 3], ****p < 0.0001; **p = 0.0021 [one-way ANOVA with Dunnett's multiple comparisons test]; right graph: mean with S.D [n = 3], *p = 0.0117 [unpaired t test]), or E, or with meibomian gland secretion (mean with S.D [n = 3], ****p < 0.0001 [unpaired t test]). F, tryptophan emission spectra of N-94 or "Ctrl pep" in the absence or presence of OAHFA. G, calculated K a or K d of N-94 or Ctrl pep with OAHFA (mean with S.D [n = 3], **p = 0.0064 and 0.0016 [respectively left and right graphs; unpaired t test]). H, human corneal epithelial (HCE-T) cells without or stressed with interferongamma and tumor necrosis factor in the presence of 1.38 μM C-95, T0 N-94, or dissociated T2 N-94. 1.38 μM is the concentration of T2 N-94 released from OAHFA over 30 min (mean with S.D [n = 3], ****p < 0.0001; ns, not significant [one-way ANOVA with Tukey's multiple comparisons test]). Data for D−H in Data filesFigure S4. [/fig]
[fig] Figure 5: Slow tear release of topically applied 125 I-N-94 and 125 I-N-94/C-6 from rabbit eyes. A, 35 μl (10 μCi) of 44 μM 125 I-N-94/C-6 or 4 μM 125 I-N-94were topically added to each eye of respectively 14 rabbits twice daily for 4 ( 125 I-N-94/C-6) or 3 ( 125 I-N-94) days with the final treament on the morning of the fifth ( 125 I-N-94/C-6) or fourth ( 125 I-N-94) day (respective dark green or red downward arrows) after which tears were collected at 0.5, 1, 3, 6, 12, and 24 h postdose from two animals per time point who were then euthanized for collection of blood and ocular tissues (respective green or red downward arrowheads). B-C, TCA precipitable radioactivity of samples was assessed by scintillation counting and expressed per tissue wet weight. ****p < 0.0001; ***, respectively 0.0007 and 0.0003, *, respectively 0.0365 and 0.0198 (two-way ANOVA with Tukey's multiple comparisons test comparing tears versus retina); study was performed once each for rabbits treated with 125 I-N-94 or with 125 I-N-94/C-6). D, a single 35 μl (10 μCi) dose of 44 μM 125 I-N-94/C-6 or 4 μM 125 I-N-94 (as three 1.3 μM doses over 10 min) was topically added to each eye of respectively 14 or 12 rabbits. Tears were collected 0.25, 0.5, 1, 3, 6, or 12 h after. E, TCA precipitable radioactivity of samples from D were assessed as in B-C (****p < 0.0001; *p = 0.0431 [one-way ANOVA with Tukey's multiple comparisons test]; study was performed once each for rabbits treated with 125 I-N-94 or with 125 I-N-94/C-6). F, 50 μM of N-94, N-94/C-6, or positive control "SN-pep"(56) was incubated for 4 h at 37 C with 10 μl of human basal tears. Samples before (left column) or after tear treatment were assessed by MALDI mass spectrometry ([n = 1). Data for B-E in Data filesFigure S5. [/fig]
[fig] Figure 6: Tear serine proteases, aminopeptidases, and metalloproteinases may contribute to the generation of C-terminal lacritin proteoforms. A, N-64/C-31 with predicted protease cleavage sites as N-termini for proteoforms. B, left, N-64/C-31 digestion and MALDI mass spectrometry analysis scheme using 38 μl of pooled basal tears. Right, N-64/C-31 MALDI profiles normalized to N-94/C-6 spiked in postdigestion. Incubations included two different concentrations of AEBSF (0.2 or 1 mM), antipain (74 or 370 μM), bestatin (130 or 650 μM), chymostatin (100 or 500 μM), EDTA (2.5 or 5 mM), leupeptin (10 or 50 μM), and pepstatin (1 or 5 μM). Representative of three optimization experiments. [/fig]
|
10.3390/nano12111893
|
CCBY
|
9182164
|
35683747
|
s2orc_pubmed_articles
|
Microstructural Study of MgB2 in the LiBH4-MgH2 Composite by Using TEM
Citation: Jin, O.; Shang, Y.; Huang, X.; Mu, X.; Szabó, D.V.; Le, T.T.; Wagner, S.; Kübel, C.; Pistidda, C.; Pundt, A. Microstructural Study of MgB 2 in the LiBH 4 -MgH 2 Composite by Using TEM. Nanomaterials 2022, 12, 1893.
# Introduction
Hydrogen is a clean and reproducible energy carrier with the highest gravimetric energy density of~120 kJ g −1 . For extensive applications of hydrogen, advanced hydrogen storage materials are demanded to store hydrogen safely and efficiently. Reactive hydride composites (RHCs) have been studied intensively due to their exceptionally reversible hydrogen storage capacity [bib_ref] Complex hydrides for energy storage, Milanese [/bib_ref]. These materials were initially derived from light metal complex hydrides (e.g., LiBH 4 , LiNH 2 , NaAlH 4 , etc.) in combination with a second hydride (e.g., LiH, MgH 2 , etc.) [bib_ref] Interaction of hydrogen with metal nitrides and imides, Chen [/bib_ref] [bib_ref] Unexpected kinetic effect of MgB 2 in reactive hydride composites containing complex..., Barkhordarian [/bib_ref] [bib_ref] Thermodynamic destabilization and reaction kinetics in light metal hydride systems, Vajo [/bib_ref]. Among various RHCs, the LiBH 4 -MgH 2 composite is one of the most competitive candidates for both on-and off-board applications, based on the International Energy Agency Task 22 [bib_ref] Task 22 of IEA HIA-Fundamental and Applied Hydrogen Storage Materials Development, Hauback [/bib_ref]. According to prior studies, the related decomposition reaction occurs in two steps [bib_ref] Hydrogen sorption properties of MgH 2 -LiBH 4 composites, Bösenberg [/bib_ref] : 2LiBH 4 + MgH 2 → 2LiBH 4 + Mg + H 2 → 2LiH + MgB 2 +4H 2 [bib_ref] Complex hydrides for energy storage, Milanese [/bib_ref] Compared with the hydrogen capacity of~18.5 wt% in pristine LiBH 4 , about 11.4 wt% of hydrogen can still be yielded with the LiBH 4 -MgH 2 composite, while the thermodynamic properties are significantly improved by the addition of MgH 2 , resulting in overall superior performance in hydrogen storage. For LiBH 4 alone, the standard enthalpy of decomposition is about 70 kJ mol −1 H 2 [bib_ref] Stability and reversibility of LiBH 4, Mauron [/bib_ref]. This value corresponds to a decomposition temperature of about 400 - C under atmospheric pressure [bib_ref] Stability and reversibility of LiBH 4, Mauron [/bib_ref] [bib_ref] Dehydriding and rehydriding reactions of LiBH 4, Orimo [/bib_ref] [bib_ref] Hydrogen storage in destabilized chemical systems, Vajo [/bib_ref]. For the LiBH 4 -MgH 2 composite, the standard decomposition enthalpy is reduced to about 45 kJ mol −1 H 2 [bib_ref] Hydrogen storage in destabilized chemical systems, Vajo [/bib_ref] [bib_ref] Reversible storage of hydrogen in destabilized LiBH 4, Vajo [/bib_ref] , which is ascribed to the exothermic formation of MgB 2 during the endothermic two-step decomposition process of the LiBH 4 -MgH 2 composite (see Equation (1)) [bib_ref] Unexpected kinetic effect of MgB 2 in reactive hydride composites containing complex..., Barkhordarian [/bib_ref] [bib_ref] Reversible storage of hydrogen in destabilized LiBH 4, Vajo [/bib_ref]. This results in a notable reduction in the decomposition temperature down to about 170 - C under atmospheric pressure [bib_ref] Thermodynamic destabilization and reaction kinetics in light metal hydride systems, Vajo [/bib_ref] [bib_ref] Hydrogen storage in destabilized chemical systems, Vajo [/bib_ref].
However, in contrast to the thermodynamic predictions, the decomposition of the LiBH 4 -MgH 2 composite for hydrogen release is kinetically limited and barely triggered at moderate temperatures [bib_ref] Thermodynamic destabilization and reaction kinetics in light metal hydride systems, Vajo [/bib_ref] [bib_ref] Hydrogen sorption properties of MgH 2 -LiBH 4 composites, Bösenberg [/bib_ref]. Bösenberg et al. ascribed the sluggish kinetic behavior to the nucleation of MgB 2 as the rate-limiting step during dehydrogenation [bib_ref] Role of additives in LiBH 4 -MgH 2 reactive hydride composites for..., Bösenberg [/bib_ref] [bib_ref] Pressure and temperature influence on the desorption pathway of the LiBH 4..., Bösenberg [/bib_ref]. This is prominently reflected in the long incubation period, after the complete decomposition of MgH 2 into Mg (the first dehydrogenation step in Equation (1)) and before the massive decomposition of LiBH 4 (the second dehydrogenation step in Equation (1)).
Current research on enhancing the kinetics of hydride compounds mainly focuses on nanoconfinement using various carbon scaffolds and on utilizing transition metalbased additives [bib_ref] Enhanced hydrogen storage kinetics of LiBH 4 in nanoporous carbon scaffolds, Gross [/bib_ref] [bib_ref] Effective nanoconfinement of 2LiBH 4 -MgH 2 via simply MgH 2 premilling..., Gosalawit-Utke [/bib_ref] [bib_ref] Building robust architectures of carbon-wrapped transition metal nanoparticles for high catalytic enhancement..., Huang [/bib_ref] [bib_ref] Microstructural study of the LiBH 4 -MgH 2 reactive hydride composite with..., Deprez [/bib_ref]. The additives, in particular, provide an impressive boost to the dehydrogenation kinetics of RHCs, with their hydrogen storage capacities being wellpreserved. As for the LiBH 4 -MgH 2 composite, it has been reported that the enhanced kinetics can be attributed to a significant promotion of the heterogeneous nucleation of MgB 2 by additives [bib_ref] The catalytic effect of additive Nb2O5 on the reversible hydrogen storage performances..., Fan [/bib_ref] [bib_ref] Design of a nanometric AlTi additive for MgB 2 -based reactive hydride..., Le [/bib_ref]. However, their role in the decomposition path has not yet been explicitly understood due to a lack of microscopic investigations; thus, the mechanism of MgB 2 nucleation and growth in terms of crystallography has remained vague [bib_ref] Role of additives in LiBH 4 -MgH 2 reactive hydride composites for..., Bösenberg [/bib_ref]. Complementing the missing knowledge is essential for understanding the reaction process behind the kinetic improvement of the material.
In this study, the additive effect on MgB 2 nucleation and growth in the LiBH 4 -MgH 2 composite is clarified by determining the MgB 2 morphology and its crystallographic orientations toward nucleation centers. Transmission electron microscopy (TEM) studies were carried out on the microstructural evolution of MgB 2 in samples with and without Ti-and Al-based additives using a very high content. This allowed us to reveal the details of how the microstructural boundary conditions determine the decomposition kinetics of the system.
## Experimental section
# Material preparations
The reactants were provided in powder form by the following commercial suppliers: MgH 2 (95% purity) from Rockwood Lithium GmbH, LiBH 4 (95% purity) from Sigma-Aldrich, and 3TiCl 3 ·AlCl 3 (about 76-78% purity) from Fischer Scientific. The LiBH 4 -MgH 2 composite was prepared with a molar content of x% 3TiCl 3 ·AlCl 3 (x = 0, 0.625, and 20). The large additive content of 20 mol% is chosen to maximize the additive effect on the MgB 2 morphology. To achieve a fine mixing of the reactants and an even dispersion of the additives, the prepared material mixtures (3 g)-namely, 2LiBH 4 -MgH 2 or 2LiBH 4 -MgH 2 -3TiCl 3 ·AlCl 3 -were charged into stainless-steel vials with stainless steel balls in a ball to powder ratio of 20:1. The milling proceeded for 400 min using a Spex 8000 M Mixer Mill. Both the powder handling and milling were always performed under an argon atmosphere in a glovebox (O 2 , H 2 O < 0.5 ppm).
## Kinetic measurements
The volumetric measurements were performed using a custom-built Sievert's-type apparatus. The milled sample (~170 mg) was charged into the stainless-steel sample holder of the measuring apparatus. The samples were annealed from room temperature to 400 - C at a heating rate of 10 - C min −1 under a hydrogen atmosphere of 4 bar. After reaching the target temperature of 400 - C, the materials were kept under isothermal conditions for several hours.
## Xrd experiments
The ex situ XRD experiments were based on a Bruker D8 Discover diffractometer equipped with a Cu X-ray source (λ = 1.54184 Å) and a 2D VANTEC detector. The operating voltage and current were 50 kV and 1000 mA. The diffraction patterns were acquired in the 2θ range from 10 - to 90 - with a step size of 0.005 - , ∆2θ = 10 - , and the exposure time for each step of 400 s. To prevent oxidation phenomena during the acquirement of the XRD pattern, the specimens were sealed in an argon-filled sample holder made of poly (methyl methacrylate) (PMMA).
## Tem experiments
TEM experiments were performed using a Themis-Z 60-300 (Thermo Fisher Scientific Inc., Waltham, MA, USA) equipped with a monochromator and double aberration correctors (probe and image Cs correctors), operated at 300 kV. TEM sample preparation was carried out under an argon atmosphere in a glovebox (O 2 , H 2 O < 0.5 ppm). Sample powders were dispersed in toluene and ultra-sonicated for 1 min before being distributed on lacey-carbon coated gold TEM grids S166-A3-V (Ted Pella Inc., Redding, CA, USA). Subsequently, they were transferred under argon from the glovebox into the microscope with a vacuum transfer holder 648 (Gatan Inc., Pleasanton, CA, USA).
The beam current varied from 50 to 100 pA throughout all TEM experiments. Selected area electron diffraction (SAED) patterns, TEM, and high-resolution TEM (HRTEM) images were collected using a OneView camera (Gatan Inc., Pleasanton, CA, USA). Scanning TEM (STEM) images were recorded via a high-angle annular dark-field (HAADF) detector with a convergence angle of 21.5 mrad and a camera length of 93 mm. Energy-dispersive X-ray spectroscopy (EDX) spectrum-imaging (SI) was executed with a Super-X windowless EDX detector (Thermo Fisher Scientific Inc., Waltham, MA, USA) using the same parameters as in STEM mode.
Electron tomography was carried out using a Fischione 2020 tomography holder in STEM mode with the same STEM parameters as mentioned above. HAADF-STEM Tilt series with image dimensions of 2048 × 2048 pixels were collected using Xplore3D (Thermo Fisher Scientific) over a tilt range in increments of 2 - from −72 - to 78 - for the desorbed 2LiBH 4 -MgH 2 without additives and from −72 - to 68 - for the desorbed 2LiBH 4 -MgH 2 with 20 mol% 3TiCl 3 ·AlCl 3 . The alignment of the tilt series was performed by IMOD using 10 nm gold colloidal particles as fiducial markers, which leads to the respective mean residual error of 0.736 and 0.129 voxels [bib_ref] Computer visualization of three-dimensional image data using IMOD, Kremer [/bib_ref]. The aligned tilt series were then reconstructed using the algorithm simultaneous iterative reconstruction technique (SIRT) with 100 iterations by Inspect3D (Thermo Fisher Scientific) [bib_ref] Iterative methods for the three-dimensional reconstruction of an object from projections, Gilbert [/bib_ref]. The 3D visualizations were realized by Avizo 2020.2 (Thermo Fisher Scientific).
In addition, 4D-STEM measurements were carried out in µ probe mode using the OneView camera, with a convergence angle of 0.47 mrad, a camera length of 580 mm, and an acquisition time of 60 ms for each diffraction pattern, and a dose of~1.5 × 10 5 e nm −2 . STEM electron energy-loss spectroscopy (EELS) SI was acquired using a Gatan Continuum 970 HighRes imaging filter (GIF) (Gatan Inc., Pleasanton, CA, USA) in dual-EELS mode with 5 ms acquisition time for each low-loss spectrum, 20 ms for each high-loss spectrum, 21.5 mrad convergence angle, 40 mrad collection angle, and 0.3 eV per channel energy dispersion, leading to a measured energy spread of 2.0 eV on the camera. Both the EDX SI and EELS SI data were denoised via principal component analysis (PCA), which effectively reduces the random noise generated during the signal recording [bib_ref] Principal component analysis, Wold [/bib_ref] [bib_ref] Principal component analysis, Abdi [/bib_ref].shows the X-ray diffraction (XRD) results of the system 2LiBH 4 -MgH 2 with x mol% 3TiCl 3 ·AlCl 3 (x = 0, 0.625, and 20) in two different states: as-milled and after desorption. In general, no significant variation is visible for the sample without additives and the one containing 0.625 mol% 3TiCl 3 ·AlCl 3 . However, after increasing the additive content to a high value of 20 mol%, the LiBH 4 peaks for the as-milled state and the MgB 2 peaks for the after-desorption state noticeably weakened, whereas strong LiCl peaks appeared for both states. This difference can be attributed to the reaction between LiBH 4 and 3TiCl 3 ·AlCl 3, which consumed a significant amount of LiBH 4 that is responsible for the generation of MgB 2 [bib_ref] Design of a nanometric AlTi additive for MgB 2 -based reactive hydride..., Le [/bib_ref].
# Results and discussion
## Material characterization via xrd and kinetic performance
EDX SI and EELS SI data were denoised via principal component analysis (PCA), which effectively reduces the random noise generated during the signal recording [bib_ref] Principal component analysis, Wold [/bib_ref] [bib_ref] Principal component analysis, Abdi [/bib_ref].shows the X-ray diffraction (XRD) results of the system 2LiBH4-MgH2 with x mol% 3TiCl3- AlCl3 (x = 0, 0.625, and 20) in two different states: as-milled and after desorption. In general, no significant variation is visible for the sample without additives and the one containing 0.625 mol% 3TiCl3- AlCl3. However, after increasing the additive content to a high value of 20 mol%, the LiBH4 peaks for the as-milled state and the MgB2 peaks for the after-desorption state noticeably weakened, whereas strong LiCl peaks appeared for both states. This difference can be attributed to the reaction between LiBH4 and 3TiCl3- AlCl3, which consumed a significant amount of LiBH4 that is responsible for the generation of MgB2 [bib_ref] Design of a nanometric AlTi additive for MgB 2 -based reactive hydride..., Le [/bib_ref]. The kinetic performance of the corresponding materials is visualized in. In comparison with the pristine material, the incubation period was reduced from ~25 h to ~8 h with an additive content of 0.625 mol%. By adding 20 mol% 3TiCl3- AlCl3, the incubation stage disappeared, and the kinetics of the second desorption step for hydrogen release changed. The reduced hydrogen storage capacity of ~5 wt% for the 20 mol% 3TiCl3- AlCl3 sample is due to the significant consumption of LiBH4 during the reaction with additives. . The absence of crystallographic information on LiH is likely due to its instability under electron illumination and the low scattering power of Li and H. One impressive feature in [fig_ref] Figure 2a: presents the morphology of the decomposed LiBH4-MgH2 compositenamely, LiH and MgB2 [/fig_ref] is the bar-like morphology of crystals with a parallel arrangement. They were identified as MgB2 by HRTEM [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref] , its fast Fourier transform (FFT) pattern, and the EDXS elemental map of Mg acquired from the corresponding area in [fig_ref] Figure 2a: presents the morphology of the decomposed LiBH4-MgH2 compositenamely, LiH and MgB2 [/fig_ref] (see. The primary growth direction of MgB2 bars is thus determined to be , which is along the long axis of a MgB2 bar (the inset in [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref]. The kinetic performance of the corresponding materials is visualized in. In comparison with the pristine material, the incubation period was reduced from~25 h to~8 h with an additive content of 0.625 mol%. By adding 20 mol% 3TiCl 3 ·AlCl 3 , the incubation stage disappeared, and the kinetics of the second desorption step for hydrogen release changed. The reduced hydrogen storage capacity of~5 wt% for the 20 mol% 3TiCl 3 ·AlCl 3 sample is due to the significant consumption of LiBH 4 during the reaction with additives. . The absence of crystallographic information on LiH is likely due to its instability under electron illumination and the low scattering power of Li and H. One impressive feature in [fig_ref] Figure 2a: presents the morphology of the decomposed LiBH4-MgH2 compositenamely, LiH and MgB2 [/fig_ref] is the bar-like morphology of crystals with a parallel arrangement. They were identified as MgB 2 by HRTEM [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref] , its fast Fourier transform (FFT) pattern, and the EDXS elemental map of Mg acquired from the corresponding area in [fig_ref] Figure 2a: presents the morphology of the decomposed LiBH4-MgH2 compositenamely, LiH and MgB2 [/fig_ref] (see. The primary growth direction of MgB 2 bars is thus determined to be 1210 , which is along the long axis of a MgB 2 bar (the inset in [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref]. Nanomaterials 2022, 12, 1893 5 of 15 For a more comprehensive understanding of the MgB2 morphology, electron tomography was conducted. [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref] shows a volume rendering of MgB2 bars, reconstructed from the STEM-HAADF tomographic data, and [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref] exhibits a single MgB2 bar extracted from the selected region in [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref]. It should be noted that MgB2 bars possess a rectangular shape. This can be understood in combination with the study by Lee et al., claiming that a growth constraint exists for MgB2 along the c-axis [0002], which is consistent with our investigation, as the constrained direction [0002]
# Results and discussion
## Material characterization via xrd and kinetic performance
## Mgb2 formation without the influence of additives
## Mgb 2 formation without the influence of additives
(z-axis in [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref] lays perpendicular to the observed growth direction ] (x-axis). Given the rectangular MgB2 bars, another growing direction along the y-axis corresponds to
[formula] [101 0] [/formula]
, along which the growth is also limited to some extent. This is discussed in Section 3.5.
The observation of parallel MgB2 bars in [fig_ref] Figure 2a: presents the morphology of the decomposed LiBH4-MgH2 compositenamely, LiH and MgB2 [/fig_ref] implies that their nucleation and growth initially proceeded from the same class of atomic layers of one crystalline nucleation center. Given the two-step decomposition of the LiBH4-MgH2 composite (Equation (1)), Mg and LiBH4 are candidates for this nucleation center. Since LiBH4 has a melting point of approximately 275 °C and, therefore, remains in the liquid state during the desorption at 400 °C, Mg has to be the nucleation center for the heterogeneous nucleation of these MgB2 bars. Another requirement for the nucleation center is that it has to provide a sufficiently large superficial plane to nucleate several MgB2 bars with a lateral size of about 200 nm [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref]. The observed Mg grains are large enough, with a size up to about 1 µm to meet this requirement (see [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref].
## Mgb2 formation under the influence of additives
To reveal the additive's effect on MgB2 nucleation and growth, an overdose of 20 mol% 3TiCl3- AlCl3 was taken for the LiBH4-MgH2 composite. [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] shows the material morphology after desorption. According to the XRD resultand the local electron diffraction pattern [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] , the element Mg exists only as MgB2. Therefore, the Mg EDX map [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] can be directly regarded as representing the distribution of MgB2, For a more comprehensive understanding of the MgB 2 morphology, electron tomography was conducted. [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref] shows a volume rendering of MgB 2 bars, reconstructed from the STEM-HAADF tomographic data, and Figure 2d exhibits a single MgB 2 bar extracted from the selected region in [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref]. It should be noted that MgB 2 bars possess a rectangular shape. This can be understood in combination with the study by Lee et al., claiming that a growth constraint exists for MgB 2 along the c-axis [0002] MgB 2, which is consistent with our investigation, as the constrained direction [0002] MgB 2 (z-axis in [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref] lays perpendicular to the observed growth direction 1210 MgB 2 (x-axis). Given the rectangular MgB 2 bars, another growing direction along the y-axis corresponds to 1010 MgB 2 , along which the growth is also limited to some extent. This is discussed in Section 3.5.
The observation of parallel MgB 2 bars in [fig_ref] Figure 2a: presents the morphology of the decomposed LiBH4-MgH2 compositenamely, LiH and MgB2 [/fig_ref] implies that their nucleation and growth initially proceeded from the same class of atomic layers of one crystalline nucleation center. Given the two-step decomposition of the LiBH 4 -MgH 2 composite (Equation (1)), Mg and LiBH 4 are candidates for this nucleation center. Since LiBH 4 has a melting point of approximately 275 - C and, therefore, remains in the liquid state during the desorption at 400 - C, Mg has to be the nucleation center for the heterogeneous nucleation of these MgB 2 bars. Another requirement for the nucleation center is that it has to provide a sufficiently large superficial plane to nucleate several MgB 2 bars with a lateral size of about 200 nm [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref]. The observed Mg grains are large enough, with a size up to about 1 µm to meet this requirement (see [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref].
## Mgb 2 formation under the influence of additives
To reveal the additive's effect on MgB 2 nucleation and growth, an overdose of 20 mol% 3TiCl 3 ·AlCl 3 was taken for the LiBH 4 -MgH 2 composite. [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] shows the material morphology after desorption. According to the XRD resultand the local electron diffraction pattern [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] , the element Mg exists only as MgB 2 . Therefore, the Mg EDX map [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] can be directly regarded as representing the distribution of MgB 2 , which is comparable with the bright contrast in the HAADF image [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref]. In general, the MgB 2 morphology changed significantly, compared with what was previously reported in Section 3.2.
Nanomaterials 2022, 12, 1893 6 of 15 which is comparable with the bright contrast in the HAADF image [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref]. In general, the MgB2 morphology changed significantly, compared with what was previously reported in Section 3.2. In this case, two different MgB 2 morphologies can be distinguished. Orienting along the x-axis, the parallel-lying MgB 2 bars grow into a hollow tube. Inside the tube, the second morphology of platelet-like MgB 2 can be identified. It can be seen in [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] that these MgB 2 platelets grow from the tube walls of MgB 2 bars in a more or less random orientation. For a closer look at these MgB 2 platelets, one piece was cut out from [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] and displayed as a segmented surface rendering in [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref]. In contrast to the rectangular shape of the MgB 2 bars, these MgB 2 platelets with a lateral size of about 500 nm possess a more irregular or semi-circular shape. In other words, no primary growth direction is observed for the MgB 2 platelets. In addition, the MgB 2 morphology of the sample with 0.625 mol% 3TiCl 3 ·AlCl 3 can be found in [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] for comparison.
The platelet-like MgB 2 morphology hints at another nucleation center that facilitates the formation of the MgB 2 platelets. We suggest that this nucleation center stems from the additives, so the identification and location of these additives is another point of interest. As reported by T.-T. Le et al., the additive 3TiCl 3 ·AlCl 3 can react with LiBH 4 and generate either AlTi 3 or TiB 2 and AlB 2 [bib_ref] Design of a nanometric AlTi additive for MgB 2 -based reactive hydride..., Le [/bib_ref]. Accordingly, rather than the initial additive 3TiCl 3 ·AlCl 3 , these reaction products are considered to be the heterogeneous nucleation centers for the MgB 2 platelets. Based on the STEM-EDX map in [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] , both Ti and Al are dispersed inside the MgB 2 tube, which is also indicative of their role as the nucleation center (also see [fig_ref] Figure 4: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 [/fig_ref].
For further studies, the EELS SI was acquired. [fig_ref] Figure 4: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 [/fig_ref] displays the summed elemental map based on the EEL spectrum in [fig_ref] Figure 4: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 [/fig_ref] , comprising B K-edge in red, Ti L 2,3 -edge in blue, and Mg K-edge in yellow [fig_ref] Figure 4: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 [/fig_ref]. Therefore, the orange platelets that result from an overlap between Mg (yellow) and B (red) indicate the elemental correlation between Mg and B and suggest the location of MgB 2 . Furthermore, the background-subtracted B K-edge recorded in the orange area exhibits a pre-peak at about 190 eV [fig_ref] Figure 4: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 [/fig_ref]. This fine structure is evidence of a high and unfilled p-like density of states of Boron, which indicates the bonding between Mg and B and verifies the existence of MgB 2 [bib_ref] Electron energy-loss spectroscopy characterization of the boron p-like density of states in..., Kong [/bib_ref].
Similarly, the purple agglomerates in [fig_ref] Figure 4: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 [/fig_ref] represent the spatial elemental correlation between Ti (blue) and B (red), which disperse around the orange MgB 2 platelets. According to the HRTEM image and its FFT pattern [fig_ref] Figure 4: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 [/fig_ref] , the presence of TiB 2 rather than AlTi 3 can be confirmed already after the ball milling process. The HRTEM image corresponding to the purple agglomerates in [fig_ref] Figure 4: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 [/fig_ref] can be found in [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] , which shows a further growth of TiB 2 particles up to~20 nm after desorption likely due to Ostwald ripening. The beneficial effect of TiB 2 on accelerating the decomposition of the LiBH 4 -MgH 2 composite was already reported by some studies, which also supports our characterizations [bib_ref] Oxidation state and local structure of Ti-based additives in the reactive hydride..., Deprez [/bib_ref] [bib_ref] High catalytic efficiency of amorphous TiB 2 and NbB 2 nanoparticles for..., Fan [/bib_ref]. Since AlB 2 and TiB 2 have the same space group (P6m/mm, No. 191), with almost the same lattice constants, AlB 2 might have the same effect as TiB 2 , although they cannot be distinguished here.
In [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] , the relationship between the crystallography and the morphology of the MgB 2 bars and MgB 2 platelets is revealed via 4D-STEM. [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] -e provide local electron-diffraction patterns acquired from areas B-E in [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref]. A fragment of the MgB 2 bar with a rectangular shape can be found in area B of [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] , with the short axis along 1010 MgB 2 [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref]. [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] ,e indicate the long axis direction of the MgB 2 bars being along 1210 MgB 2 . These investigated orientations are consistent with [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref]. Thus, it can be summarized that the morphology of MgB 2 bars with a rectangular shape has the long axis direction oriented along 1210 , the short axis direction along 1010 MgB 2 , and the thin direction along [0002] MgB 2 . For MgB 2 platelets, the thin direction of a platelet, which is also regarded as the growth-restricted direction, is along [0002] MgB 2 [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref]. This growth restriction observed for more other MgB 2 platelets can be seen in. In [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] , the relationship between the crystallography and the morphology of the MgB2 bars and MgB2 platelets is revealed via 4D-STEM. [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] -e provide local electrondiffraction patterns acquired from areas B-E in [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref]. A fragment of the MgB2 bar with a rectangular shape can be found in area B of [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] , with the short axis along [101 0] [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref]. [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] ,e indicate the long axis direction of the MgB2 bars being along . These investigated orientations are consistent with [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref]. Thus, it 5a,c). This growth restriction observed for more other MgB2 platelets can be seen in.
## Analysis of orientation relationships
Given the small size of TiB2 or/and AlB2 nanoparticles only up to about 20 nm [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] and the complexity of the overlapping phases, the interface between the nucleation centers and MgB2 cannot be investigated in the experiment. For this reason, their orientation relationships (ORs) cannot be experimentally determined in a conventional way by focusing on the interface. To determine the orientation relationship between the different nucleation centers and MgB2, we took advantage of the edge-to-edge matching model [bib_ref] Edge-to-edge matching model for predicting orientation relationships and habit planes-the improvements, Zhang [/bib_ref] [bib_ref] Edge-to-edge matching and its applications: Part II. Application to Mg-Al, Mg-Y and..., Zhang [/bib_ref] [bib_ref] Crystallographic study of grain refinement in aluminum alloys using the edge-to-edge matching..., Zhang [/bib_ref] [bib_ref] Edge-to-edge matching-The fundamentals, Kelly [/bib_ref] [bib_ref] Extended application of edge-to-edge matching model to HCP/HCP (α-Mg/MgZn 2 ) system..., Yang [/bib_ref].
The edge-to-edge matching model builds on minimizing the energy of coherent interfaces between two adjacent materials, in this case, the nucleation center and the nucleating particle [bib_ref] Edge-to-edge matching model for predicting orientation relationships and habit planes-the improvements, Zhang [/bib_ref] [bib_ref] Edge-to-edge matching-The fundamentals, Kelly [/bib_ref]. According to this model, the heterogeneous nucleation between the two phases is controlled by their interatomic misfit and their interplanar mismatch. The interatomic misfit is determined along the matching directions of the two phases. These matching directions are selected among their close or nearly close-packed directions identified by the linear atomic density. The interplanar mismatch (d-value mismatch) is determined between the matching planes, which are chosen from the close or nearly closepacked planes. As a rule of thumb, the interatomic misfit and the interplanar mismatch should generally not exceed 10% and 6%, which are regarded as the critical values [bib_ref] Edge-to-edge matching model for predicting orientation relationships and habit planes-the improvements, Zhang [/bib_ref]. On the other hand, minimizing the interatomic misfit has a higher priority than the dvalue mismatch from an energetic perspective. A slight excess of the d-value mismatch above 6% is still acceptable, which can be subsequently refined by an additional rotation between the two phases about their matching directions [bib_ref] Edge-to-edge matching-The fundamentals, Kelly [/bib_ref].
Since MgB2, Mg, and TiB2 (and AlB2) have a similar hexagonal close-packed (hcp) crystal structure, their close or nearly close-packed directions and planes are similarly indexed. Based on their crystallographic characteristics, there are four possible close-packed directions 〈101 0〉, 〈0002〉, 〈12 10〉, and 〈11 23〉 and four possible close-packed planes {101 0}, {101 1}, {0002}, and {12 10} [bib_ref] Crystallographic study of grain refinement in aluminum alloys using the edge-to-edge matching..., Zhang [/bib_ref]. According to our observations, two distinct MgB2 morphologies were introduced:
## Analysis of orientation relationships
Given the small size of TiB 2 or/and AlB 2 nanoparticles only up to about 20 nm [fig_ref] Figure 5: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] and the complexity of the overlapping phases, the interface between the nucleation centers and MgB 2 cannot be investigated in the experiment. For this reason, their orientation relationships (ORs) cannot be experimentally determined in a conventional way by focusing on the interface. To determine the orientation relationship between the different nucleation centers and MgB 2 , we took advantage of the edge-to-edge matching model [bib_ref] Edge-to-edge matching model for predicting orientation relationships and habit planes-the improvements, Zhang [/bib_ref] [bib_ref] Edge-to-edge matching and its applications: Part II. Application to Mg-Al, Mg-Y and..., Zhang [/bib_ref] [bib_ref] Crystallographic study of grain refinement in aluminum alloys using the edge-to-edge matching..., Zhang [/bib_ref] [bib_ref] Edge-to-edge matching-The fundamentals, Kelly [/bib_ref] [bib_ref] Extended application of edge-to-edge matching model to HCP/HCP (α-Mg/MgZn 2 ) system..., Yang [/bib_ref].
The edge-to-edge matching model builds on minimizing the energy of coherent interfaces between two adjacent materials, in this case, the nucleation center and the nucleating particle [bib_ref] Edge-to-edge matching model for predicting orientation relationships and habit planes-the improvements, Zhang [/bib_ref] [bib_ref] Edge-to-edge matching-The fundamentals, Kelly [/bib_ref]. According to this model, the heterogeneous nucleation between the two phases is controlled by their interatomic misfit and their interplanar mismatch. The interatomic misfit is determined along the matching directions of the two phases. These matching directions are selected among their close or nearly close-packed directions identified by the linear atomic density. The interplanar mismatch (d-value mismatch) is determined between the matching planes, which are chosen from the close or nearly closepacked planes. As a rule of thumb, the interatomic misfit and the interplanar mismatch should generally not exceed 10% and 6%, which are regarded as the critical values [bib_ref] Edge-to-edge matching model for predicting orientation relationships and habit planes-the improvements, Zhang [/bib_ref]. On the other hand, minimizing the interatomic misfit has a higher priority than the d-value mismatch from an energetic perspective. A slight excess of the d-value mismatch above 6% is still acceptable, which can be subsequently refined by an additional rotation between the two phases about their matching directions [bib_ref] Edge-to-edge matching-The fundamentals, Kelly [/bib_ref].
Since MgB 2 , Mg, and TiB 2 (and AlB 2 ) have a similar hexagonal close-packed (hcp) crystal structure, their close or nearly close-packed directions and planes are similarly indexed. Based on their crystallographic characteristics, there are four possible closepacked directions 1010 , 0002 , 1210 , and 1123 and four possible close-packed planes 1010 , 1011 , {0002}, and 1210 [bib_ref] Crystallographic study of grain refinement in aluminum alloys using the edge-to-edge matching..., Zhang [/bib_ref]. According to our observations, two distinct MgB 2 morphologies were introduced: [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref].
Given the large interatomic misfit of 0002 MgB 2 , compared with the close-packed directions of Mg and TiB 2 (and AlB 2 ) (see Tables S1 and S2), {0002} MgB 2 , as one of the close-packed planes, is supposed to be the matching plane for the heterogeneous nucleation of both MgB 2 bars and MgB 2 platelets on Mg and TiB 2 (and AlB 2 ).
For the MgB 2 bars, the interplanar mismatch of {0002} MgB 2 concerning the possible matching plane of Mg is listed in . For a refined matching, a tilting angle between {0002} MgB 2 and 1210 Mg should also be considered (see . In summary, the possible orientation relationships between MgB 2 and Mg are predicted to be as follows: The orientation relationships between MgB 2 and TiB 2 (and/or AlB 2 ) can be derived in the same way. The related interplanar mismatch is given in [fig_ref] Table 3: The [/fig_ref]. In this case, it appears that {0002} MgB 2 pairing with {0002} TiB 2 (and/or {0002} AlB 2 ) results in the least mismatch. According to [fig_ref] Table 4: The [/fig_ref] , the possible matching direction are 1010 and 1210 for both MgB 2 and TiB 2 (and/or AlB 2 ). The least interatomic misfit is thus obtained for the matching directions 1010 MgB 2 || 1010 TiB 2 (AlB 2 ) and 12 10 MgB 2 || 1210 TiB 2 (AlB 2 ) . Considering the rotation between the two matching planes (see , the possible orientation relationships between MgB 2 and TiB 2 (and/or AlB 2 ) are as follows: , both of which are in-plane directions at the MgB 2 /Mg interface, resulting in a rectangular-shaped MgB 2 bar. Conversely, the growth of the MgB 2 platelets on TiB 2 nanoparticles is only limited along [0002] MgB 2 , one of the in-plane directions at the MgB 2 /TiB 2 interface. According to [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref] , these MgB 2 platelets with a semi-circular shape are likely growing more randomly without a preferred growth direction.
[formula] 1010 MgB 2 | | 1010 Mg {0002} MgB 2 ∼ 2.0 - f rom 1210 Mg 1010 MgB 2 | | 0002 Mg {0002} MgB 2 </mo> f rom 1210 Mg1010 MgB 2 | | 1010 TiB 2 (AlB 2 ) {0002} MgB 2 ∼ 0.03 - f rom {0002} TiB 2 (AlB 2 ) 1210 MgB 2 | | 12 10 TiB 2 (AlB 2 ) {0002} MgB 2 ∼ 0.01 - f rom {0002} TiB 2 (AlB 2 ) [/formula]
The different MgB 2 morphologies, ranging from MgB 2 bars to MgB 2 platelets, are here ascribed to the different extent of the strain energy density induced at the interface to the nucleation center. Different aspect ratios of 1210 : 1010 : [0002] are determined to be about 50:5:1 for MgB 2 bars growing on Mg [fig_ref] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption [/fig_ref] , and about 15:15:1 for MgB 2 platelets growing on TiB 2 [fig_ref] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption [/fig_ref]. As the elastic-strain energy density is proportional to Yε 2 , where Y is Young's modulus of MgB 2 , and ε is the atomic misfit between MgB 2 and Mg or TiB 2 , the strain energy density can be thus roughly estimated by the related misfits, by assuming constant Y for simplicity. Along the interatomic direction 1010 , the induced strain energy density is approximated to be more than 6 times larger for MgB 2 on Mg grains (misfit of 4.2%) than for MgB 2 on TiB 2 nanoparticles (misfit of 1.7%), as listed in [fig_ref] Table 4: The [/fig_ref] [fig_ref] Table 1: The interplanar misfit between MgB 2 {0002} and possible match planes of... [/fig_ref]. Therefore, the difference in the elastic strain energy density here is considered to be the root of the different MgB 2 morphologies, and furthermore, the changed desorption kinetic performances.illustrates the complete process of the MgB 2 formation as discussed above. For MgB 2 growing on Mg grains, the further growth along 1010 MgB 2 is hampered due to a huge in-plane strain energy density, which forces itself to mainly grow in the out-ofplane direction 1210 , ending up with these parallel-lying rectangular MgB 2 bars. On the contrary, with an interatomic misfit of 1.7% between MgB 2 and TiB 2 along 1010 , this in-plane growth constraint is not relevant. The further growth of MgB 2 along the in-plane direction 1010 MgB 2 , and the out-of-plane direction 1210 MgB 2 can thus continuously proceed, leading to semi-circular shaped MgB 2 platelets. Furthermore, a critical value of the interatomic misfit between 1.7% and 4.2% may be expected, above which the in-plane growth of MgB 2 is restricted. This may result in a deceleration of the MgB 2 nucleation and growth and, therefore, a decelerated desorption kinetics for the LiBH 4 -MgH 2 composite. As the growth of MgB2 commonly leads to hexagonal plates, we speculate that the MgB2 formation occurs at the interface between MgB2 and its nucleation centers rather than on the MgB2 surface exposed to the liquefied LiBH4. This means that the formation of MgB2 is always under the control of the in-plane strain energy density. From this perspective, the highly regular rectangular shape of the MgB2 bars can be explained consistently. However, future in situ experiments are still required to confirm this speculation.
# Conclusions
Manifold TEM studies on the LiBH4-MgH2 composite with and without additives, combined with the predictions of the edge-to-edge matching model on the orientation relationship between the nucleation center and the resulting MgB2 phase, contribute to a As the growth of MgB 2 commonly leads to hexagonal plates, we speculate that the MgB 2 formation occurs at the interface between MgB 2 and its nucleation centers rather than on the MgB 2 surface exposed to the liquefied LiBH 4 . This means that the formation of MgB 2 is always under the control of the in-plane strain energy density. From this perspective, the highly regular rectangular shape of the MgB 2 bars can be explained consistently. However, future in situ experiments are still required to confirm this speculation.
# Conclusions
Manifold TEM studies on the LiBH 4 -MgH 2 composite with and without additives, combined with the predictions of the edge-to-edge matching model on the orientation relationship between the nucleation center and the resulting MgB 2 phase, contribute to a better understanding of the RHCs' kinetics. According to this, the additives deliver a large number of nucleation centers with a small interatomic misfit of 1.7% to MgB 2 , corresponding to low in-plane strain energy density. They facilitate the nucleation and growth of MgB 2 , leading to the morphology of MgB 2 semi-circular platelets. In contrast, large in-plane strain energy density is expected for the formation of MgB 2 bars on Mg, limiting the growth of MgB 2 and consequently slowing down the dehydrogenation kinetics. To further improve the kinetic performance of the LiBH 4 -MgH 2 composite, we suggest that the atomic misfit delivered by additives with respect to MgB 2 should be considered for future additive selection. It is also believed that these conclusions hold for other RHC systems interesting for hydrogen energy storage and release.
[fig] Figure 1: (a) XRD patterns of 2LiBH4-MgH2 with x mol% 3TiCl3•AlCl3 (x = 0, 0.625, and 20) after milling and after desorption; (b) desorption kinetics of 2LiBH4-MgH2 with x mol% 3TiCl3•AlCl3 (x = 0, 0.625, and 20) at 400 °C and under 4 bar H2. [/fig]
[fig] Figure 2a: presents the morphology of the decomposed LiBH4-MgH2 compositenamely, LiH and MgB2. The local electron diffraction pattern indicates the existence of MgB2 crystals with a hexagonal closed packed (hcp) structure (P6m/mm,No. 191) [/fig]
[fig] Figure 2: The results of 2LiBH4-MgH2 without additives after desorption: (a) STEM-HAADF image and the corresponding electron diffraction pattern; (b) HRTEM image with an inset of the local zoomed-out overview showing the growth direction of MgB2 bars, and the corresponding FFT; (c) volume rendering from tomographic reconstruction of MgB2 bars with an inset showing the corresponding STEM-HAADF image; (d) volume rendering of one selected MgB2 bar chosen from (c). [/fig]
[fig] Figure 3: The results of 2LiBH4-MgH2 with 20 mol% 3TiCl3•AlCl3 after desorption: (a) STEM-HAADF image and the corresponding electron diffraction pattern; (b) STEM-HAADF image of the selected area in (a) and EDXS elemental map of Mg (yellow), Ti (blue), and Al (red); (c) volume rendering from tomographic reconstruction of a MgB2 hollow tube viewed at different angles; (d) surface rendering of the cross-section of (c); (e) surface rendering of a segmented MgB2 platelet selected from (d). [/fig]
[fig] Figure: 3c exhibits a reconstructed volume rendering of MgB2 in different directions. In this case, two different MgB2 morphologies can be distinguished. Orienting along the x-axis, the parallel-lying MgB2 bars grow into a hollow tube. Inside the tube, the second morphology of platelet-like MgB2 can be identified. It can be seen in 3dthat these MgB2 platelets grow from the tube walls of MgB2 bars in a more or less random orientation. For a closer look at these MgB2 platelets, one piece was cut out from 3dand displayed as a segmented surface rendering in 3e. In contrast to the rectangular shape of the MgB2 bars, these MgB2 platelets with a lateral size of about 500 nm possess a [/fig]
[fig] Supplementary: Materials: The supporting information can be downloaded at: https://www.mdpi. com/article/10.3390/nano12111893/s1, Figure S1: The results of 2LiBH 4 -MgH 2 without additives after desorption: STEM-HAADF image acquired from the corresponding position in Figure 2a and EDX elemental map of Mg; Figure S2: The results of 2LiBH 4 -MgH 2 with 5 wt% 3TiCl 3 ·AlCl 3 after incomplete desorption: (a) STEM-HAADF image, and (b) the corresponding electron diffraction pattern; (c) STEM-HAADF image acquired from the selected area in (a) and (d) the corresponding EDX elemental map of Mg; Figure S3: The results of 2LiBH 4 -MgH 2 with 0.625 mol% 3TiCl 3 ·AlCl 3 after desorption: (a) STEM-HAADF images showing the morphology of the generated MgB 2 crystals; (b) electron diffraction pattern; (c) STEM-HAADF image acquired from the corresponding position in (a), and EDX map of Mg; Figure S4: The results of 2LiBH 4 -MgH 2 with 20 mol% 3TiCl 3 ·AlCl 3 after desorption: STEM-HAADF image acquired from the selected position in Figure 3a, and the corresponding EDX elemental map of Mg, Ti, and Al; Figure S5: The results of 2LiBH 4 -MgH 2 with 20 mol% 3TiCl 3 ·AlCl 3 after desorption: HRTEM image acquired from the position of purple agglomerates in Figure 4a, and the corresponding FFT, showing the existence of TiB 2 (and AlB 2 ); Figure S6: The results of 2LiBH 4 -MgH 2 with 20 mol% 3TiCl 3 ·AlCl 3 after desorption: (a) STEM-HAADF image showing the distribution of MgB 2 platelets; (b) electron diffraction patterns showing the crystallographic orientation of the corresponding MgB 2 platelets indicated in (a); Figure S7: Simulated superimposed diffraction patterns of MgB 2 /Mg (a,b), and MgB 2 /TiB 2 (c,d); Table S1: The interatomic misfit between 0002 MgB 2 and the possible matching directions of Mg nucleation center (%); Table S2: The interatomic misfit between 0002 MgB 2 and the possible matching directions of MB 2 (M = Ti or Al) nucleation center (%). Author Contributions: Conceptualization, O.J., C.P. and A.P.; Data curation, O.J. and A.P.; Formal analysis, O.J., Y.S., X.H., X.M., D.V.S., T.T.L., S.W. and A.P.; Funding acquisition, C.P. and A.P.; Investigation, O.J., Y.S., X.H. and X.M.; Methodology, O.J., Y.S., X.H., X.M., D.V.S., S.W., C.K., C.P. and A.P.; Project administration, C.K., C.P. and A.P.; Resources, C.P. and A.P.; Software, O.J., Y.S., X.H., X.M., D.V.S., T.T.L. and S.W.; Supervision, D.V.S., C.P. and A.P.; Validation, O.J., X.H., X.M., D.V.S., T.T.L., S.W., C.K., C.P. and A.P.; Visualization, O.J., Y.S., X.H., D.V.S., S.W., C.K., C.P. and A.P.; Writing-original draft, O.J. and Y.S.; Writing-review & editing, O.J., Y.S., X.H., X.M., D.V.S., T.T.L., S.W., C.K., C.P. and A.P.. All authors have read and agreed to the published version of the manuscript. [/fig]
[fig] Funding: This research was funded by Deutsche Forschungsgemeinschaft grant number PU 131/16-1 and PI 1488/2-1. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: The data that support the findings of this study are openly available in KITOpen under https://publikationen.bibliothek.kit.edu/1000140178 (accessed on 1 January 2022). [/fig]
[table] Table 1: The interplanar misfit between MgB 2 {0002} and possible match planes of Mg nucleation center (%). [/table]
[table] Table 3: The [/table]
[table] Table 4: The [/table]
|
10.3390/biomedicines11123186
|
CCBY
|
10741140
|
38137407
|
s2orc_pubmed_articles
|
An Overview on Lipid Droplets Accumulation as Novel Target for Acute Myeloid Leukemia Therapy
Metabolic reprogramming is a key alteration in tumorigenesis.In cancer cells, changes in metabolic fluxes are required to cope with large demands on ATP, NADPH, and NADH, as well as carbon skeletons.In particular, dysregulation in lipid metabolism ensures a great energy source for the cells and sustains cell membrane biogenesis and signaling molecules, which are necessary for tumor progression.Increased lipid uptake and synthesis results in intracellular lipid accumulation as lipid droplets (LDs), which in recent years have been considered hallmarks of malignancies.Here, we review current evidence implicating the biogenesis, composition, and functions of lipid droplets in acute myeloid leukemia (AML).This is an aggressive hematological neoplasm originating from the abnormal expansion of myeloid progenitor cells in bone marrow and blood and can be fatal within a few months without treatment.LD accumulation positively correlates with a poor prognosis in AML since it involves the activation of oncogenic signaling pathways and cross-talk between the tumor microenvironment and leukemic cells.Targeting altered LD production could represent a potential therapeutic strategy in AML.From this perspective, we discuss the main inhibitors tested in in vitro AML cell models to block LD formation, which is often associated with leukemia aggressiveness and which may find clinical application in the future.
# Introduction
Cancer cells are sustained by a complex metabolic reprogramming involving glycolysis (Warburg effect), glutaminolysis, mitochondrial biogenesis, pentose phosphate pathway, and lipid metabolism [bib_ref] Reprogramming of fatty acid metabolism in cancer, Koundouros [/bib_ref].The well-studied Warburg effect consists of the preference of cancer cells to use glucose anaerobically rather than oxidative phosphorylation as an energy source [bib_ref] The Mechanism of Warburg Effect-Induced Chemoresistance in Cancer, Liu [/bib_ref].However, in cancer cells, lipid metabolism dysregulation usually occurs with a significant increase in de novo lipogenesis, lipid droplet (LD) synthesis, β-oxidation (FAO), lipid desaturation, and uptake of exogenous fatty acids (FAs).This energy boost inside the cell supports cancer survival as well as the self-renewing of cancer stem cells (CSCs) to facilitate tumor initiation and progression and also results in chemo-and radio-resistance in several cancer types, including acute myeloid leukemia (AML) [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Lipid metabolism in cancer: New perspectives and emerging mechanisms, Broadfield [/bib_ref] [bib_ref] Lipid droplets: A cellular organelle vital in cancer cells, Jin [/bib_ref] [bib_ref] Lipid Droplet Biosynthesis Impairment through DGAT2 Inhibition Sensitizes MCF7 Breast Cancer Cells..., Nisticò [/bib_ref].Lipid aberrations are strongly implicated in tumor progression because they rapidly stimulate the production of cell membranes, lipid-derived signaling molecules, and messengers and are also a source of a significant amount of energy [bib_ref] Lipid droplet functions beyond energy storage, Welte [/bib_ref].In tumor cells, lipids accumulate in the form of droplets.These cytoplasmic organelles were first described around the middle of the last century and are now considered a hallmark of tumor cells [bib_ref] Lipid droplets in the cytoplasm of malignant cells, Apffel [/bib_ref] [bib_ref] A transplantable rat liver tumor induced by 4-dimethylaminoazobenzene, Novikoff [/bib_ref] [bib_ref] Lipid Droplets in Cancer: From Composition and Role to Imaging and Therapeutics, Antunes [/bib_ref].Many proteins are involved in supporting this energy machine, the first of which is lipoprotein lipase (LPL), which has been identified as a key regulator of lipid metabolism [bib_ref] Cellular and Biochemical Characterization of Mesenchymal Stem Cells from Killian Nasal Polyp, Mesuraca [/bib_ref].LPL assisted by CD36 exerts a role in FA uptake in a variety of tumors, including breast, prostate, and liposarcoma tumors, as well as in leukemic cells [bib_ref] Lipoprotein lipase links dietary fat to solid tumor cell proliferation, Kuemmerle [/bib_ref] [bib_ref] Analysis of LPL gene expression in patients with chronic lymphocytic leukemia, Bilous [/bib_ref].In many cases, tumor progression and invasiveness are also related to fatty acid binding protein (FABP) dysregulation.These proteins are intracellular lipid chaperones that act by binding long chain fatty acids to facilitate intracellular localization.Intriguingly, the role of FABPs was reported in the early expansion of cancer cells [bib_ref] Role of fatty acid binding proteins (FABPs) in cancer development and progression, Mckillop [/bib_ref].Moreover, the CCAAT-enhancer binding protein α (C/EBPα), critical for hematopoiesis and granulopoiesis, has a role cancer metabolic dysregulation.Specifically, multiomics analyses highlighted an increase in lipid anabolism when C/EBPα and Fms-like tyrosine kinase 3 (FLT3) were coordinately activated in in vivo experiments and in patients with FLT3-mutant acute myeloid leukemia (AML).In this context, C/EBPα promotes fatty acid (FA) biosynthesis and desaturation by regulating the fatty acid synthase (FASN)-stearoyl-CoA desaturase (SCD) axis [bib_ref] C/EBPα Confers Dependence to Fatty Acid Anabolic Pathways and Vulnerability to Lipid..., Sabatier [/bib_ref].Furthermore, the proliferator-activated receptor γ (PPARγ), the master regulator of adipocyte differentiation, was significantly implicated in inflammation and cancers.PPARγ expression has a central role in the development of a variety of tissue, including prostate, breast, colon, and bladder, and its dysregulation sustains tumor development and progression.PPARγ can be considered a valid therapeutic target in the treatment of various cancers, including AML; in particular, it has been shown that stimulation with pioglitazone, a PPARg agonist, could be highly effective for patients with non-M3 AML [bib_ref] Alteration of PPAR-GAMMA (PPARG; PPARγ) and PTEN gene expression in acute myeloid..., Esmaeili [/bib_ref].Additionally, epigenetic modifications such as DNA methylation and histone acetylation alter gene expression in cancer cells.These modifications establish silent chromatin regions controlling the expression of the genes involved in lipid metabolism.The unmethylated cytosine analogs 5-azacytidine (5-AzaC)4 and 5-Aza-2 -deoxycytidine (DAC) are DNA methyltransferase inhibitors currently used for the treatment of acute myeloid leukemia patients and myelodysplastic syndromes (MDS) [bib_ref] A comparison of azacitidine and decitabine activities in acute myeloid leukemia cell..., Hollenbach [/bib_ref] [bib_ref] The epigenetic drug 5-azacytidine interferes with cholesterol and lipid metabolism, Poirier [/bib_ref].The analogs 5-AzaC and DAC act by hypomethylating the DNA and subsequently re-activating the silenced hypermethylated tumor suppressor genes.However, 5-AzaC and DAC are also able to modulate cholesterogenic and lipid gene expression.Specifically, 5-AzaC, unlike DAC, negatively regulates sterol regulatory element-binding protein (SREBP) target genes in an independent manner from DNA methylation [bib_ref] The epigenetic drug 5-azacytidine interferes with cholesterol and lipid metabolism, Poirier [/bib_ref].
The proteolytic activation of SREBP is regulated by cholesterol levels through a negative feedback loop involving sterol-sensing endoplasmic reticulum (ER)-resident proteins.The imbalance of this mechanism was associated with cancer [bib_ref] SREBP transcription factors: Master regulators of lipid homeostasis, Eberlé [/bib_ref].UMP and cytidine treatment were able to reverse 5-AzaC's effects, underlying the mechanistic role of UMP synthase and CTP inhibition.SREBPs blockage by 5-AzaC resulted in triglyceride synthesis and lipid droplet accumulation essentially due to the reduction in the de novo synthesis of CTP-phosphocholine cytidylyltransferase activity, required for CDP-diacylglycerol production by phosphatidic acid (PA)-CTP cytidylyltransferase (CDS).In conclusion, 5-AzaC treatment could be used both for DNA hypomethylation and inhibition of CTP synthesis and SREBP signaling in MDS patients [bib_ref] A comparison of azacitidine and decitabine activities in acute myeloid leukemia cell..., Hollenbach [/bib_ref] [bib_ref] The epigenetic drug 5-azacytidine interferes with cholesterol and lipid metabolism, Poirier [/bib_ref].LD accumulation in tumor cells is not limited to energy production alone; it is contextualized in a broader perspective aimed at modulating cross-talk with other cell types (muscle, immune, endothelial, and stroma cells) in the tumor microenvironment (TME).The increasing number of LDs stimulates the potentiation of the activity of the proton pump V-ATPase, causing acidification of the tumor microenvironment.Subsequently, TME acidification induces the peripheral localization of LDs and thus the rate of cancer progression [bib_ref] Lipid droplet velocity is a microenvironmental sensor of aggressive tumors regulated by..., Nardi [/bib_ref].Nowadays, a growing number of studies highlight an interesting correlation between increased fatty acid metabolism and poor prognosis in several solid and hematopoietic malignancy, including acute myeloid leukemia [bib_ref] The Lipid Metabolic Landscape of Cancers and New Therapeutic Perspectives, Wang [/bib_ref] [bib_ref] The fatty acid elongase Elovl6 is crucial for hematopoietic stem cell engraftment..., Kiyoki [/bib_ref].In addition, LD increase is often related to chemotherapeutic resistance.In some solid tumors, the resistance to drug induced DNA damage, like etoposide, can be unleashed by the LPL-linked gene BCL2 with a well-known anti-apoptotic function.On the other hand, in a drug-resistant cell model derived from myeloid leukemia a directly proportional ratio between the increasing LD content and resistance to an aminopeptidase inhibitor, the ERK/Akt/mTOR survival pathway was targeted [bib_ref] Multifactorial resistance to aminopeptidase inhibitor prodrug CHR2863 in myeloid leukemia cells: Downregulation..., Verbrugge [/bib_ref].
Lipidomic analyses, in addition to genomics and proteomics, today offer the possibility to illuminate the entire spectrum of lipid alterations involved in LD production and in any case underlying cancer formation and progression [bib_ref] Lipidomics: Novel insight into the biochemical mechanism of lipid metabolism and dysregulation-associated..., Zhao [/bib_ref] [bib_ref] Validation of a novel shotgun proteomic workflow for the discovery of protein-protein..., Bernaudo [/bib_ref].This review discusses the nodal points related to the biochemical processes underlying the production of lipid droplets and focuses on recent research results about novel molecules that target LDs that could have therapeutic potential in the treatment of acute myeloid leukemia.
## Lipid droplets in healthy cells and cancer cells
## Physiological roles of lds
Lipid droplets (LDs) are ubiquitous dynamic organelles composed of a core of esterified neutral lipids, such as fatty acids (FAs), triacylglycerols (TAGs), cholesterol and other sterol esters, retinyl esters, and ceramides esterified into acyl ceramides, enclosed by a phospholipid monolayer enriched in several proteins and then packaged by intermediate filament vimentin [bib_ref] Lipid Droplets in Cancer, Petan [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref] [bib_ref] Ceramide Is Metabolized to Acylceramide and Stored in Lipid Droplets, Senkal [/bib_ref] [bib_ref] Lipid Droplets in Cancer: Guardians of Fat in a Stressful World, Petan [/bib_ref] [bib_ref] The why, when and how of lipid droplet diversity, Thiam [/bib_ref] [bib_ref] Aires, V. Emergence of Lipid Droplets in the Mechanisms of Carcinogenesis and..., Delmas [/bib_ref].
LDs are physiologically involved in fat storage and membrane biosynthesis, but in the last few decades, several studies have demonstrated their critical role in cell signaling, pathogen infection [bib_ref] ssRNA Virus and Host Lipid Rearrangements: Is There a Role for Lipid..., Pagliari [/bib_ref] , inflammation, and cancer [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Lipid droplets: A cellular organelle vital in cancer cells, Jin [/bib_ref] [bib_ref] Lipid Droplets in Cancer, Petan [/bib_ref] [bib_ref] Lipid Droplets in Cancer: Guardians of Fat in a Stressful World, Petan [/bib_ref].
They can vary in terms of the number, size (from 20-40 nm to 100 µm), localization, mobility in the cell, and lipid and protein composition among cells or even within the same cell, reflecting cellular states and nutrient availability [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Lipid droplets: A cellular organelle vital in cancer cells, Jin [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref].Indeed, FA exposure rapidly activates triacylglycerol (TAG) synthesis and LD biogenesis [bib_ref] Acyl-CoA synthetase 3 promotes lipid droplet biogenesis in ER microdomains, Kassan [/bib_ref] , whereas FA and glucose depletion promote the rapid mobilization and redistribution of lipid droplets in the cell, increasing their contact with mitochondria to couple lipolytic FA release from stored TAGs with mitochondrial FA intake and energy production with fatty acid oxidation (FAO) .However, a recent study has shown that, paradoxically, mitochondria-lipid droplet contact could also induce TAG synthesis and lipid droplet expansion [bib_ref] Bound to Lipid Droplets: Where Mitochondrial Dynamics Regulate Lipid Storage and Utilization, Benador [/bib_ref].
The major protein components on the LD surface are structural proteins stably associated with LDs and the ER, such as protein family PAT (perilipin-ADRP-TIP47), which is composed of five members (PLINs from 1 to 5), and the cell death-inducing DFF45-like effector (CIDE) family; membrane-trafficking proteins such as Rab10, Rab18, Rab32, and Arf1 proteins and soluble NSF attachment protein receptors (SNAREs); enzymes involved in lipid synthesis, such as DGAT2 for TAGs synthesis; and catabolism, such as adipose tissue triacylglycerol lipase (ATGL) and hormone-sensitive lipase (HSL) proteins recruited directly from the cytosol to the LD surface [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref].
Recent studies have revealed that the LD proteome exhibited approximately 150 proteins involved in lipid metabolism and signaling, as well as in membrane trafficking, redox metabolism, gene transcription, sequester histones, transcription factors (e.g., NFAT5), chaperones (e.g., Hsc70 and calreticulin), protein quality control, autophagy, ubiquitination, and immunity [bib_ref] Lipid Droplets in Cancer, Petan [/bib_ref] [bib_ref] Establishing the lipid droplet proteome: Mechanisms of lipid droplet protein targeting and..., Bersuker [/bib_ref] [bib_ref] A Proximity Labeling Strategy Provides Insights into the Composition and Dynamics of..., Bersuker [/bib_ref] [bib_ref] Fat-specific protein 27 modulates nuclear factor of activated T cells 5 and..., Ueno [/bib_ref] [bib_ref] Lysophosphatidylcholine acyltransferase 2-mediated lipid droplet production supports colorectal cancer chemoresistance, Cotte [/bib_ref].
Moreover, ultrastructural analysis has demonstrated that LDs create contact with several organelles, such as the ER, where LD formation starts; mitochondria, where FAO takes place; and peroxisomes and lysosomes, through protein-based tethering complexes [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref].
PLINs are the main proteins that promote contact with other organelles, and they are differently distributed within the whole organism and, in the subcellular localization, inside the cells.PLIN1 is the most abundant protein and is mainly present in both white and brown adipose tissue.A lack of this protein is associated with a downregulation of SREBP-1 gene expression and a consequent reduction in LD formation in mature adipocytes [bib_ref] Perilipin-mediated lipid droplet formation in adipocytes promotes sterol regulatory element-binding protein-1 processing..., Takahashi [/bib_ref].PLIN2 is ubiquitously expressed on LDs; PLIN3 has a ubiquitous expression, but its localization is primarily in the cytoplasm, unlike PLIN1 and PLIN2, and in the case of FA excess, it moves to nascent LDs, enhancing TAG synthesis and storage.PLIN4 is mainly present in adipose tissue and in skeletal and cardiac muscle, even if less expressed.PLIN5 expression is mainly detected in oxidizing tissues such as liver, heart, pancreas, and adipose tissue but was also in contact with mitochondria, enhancing the FA β-oxidation in muscle cells [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Lipid Droplets in Cancer, Petan [/bib_ref] [bib_ref] Lipid Droplets in Cancer: Guardians of Fat in a Stressful World, Petan [/bib_ref] [bib_ref] Lipid droplets at a glance, Guo [/bib_ref] [bib_ref] Perilipin 5, a lipid droplet protein adapted to mitochondrial energy utilization, Kimmel [/bib_ref] [bib_ref] Perilipin 5, a lipid droplet-associated protein, provides physical and metabolic linkage to..., Wang [/bib_ref].
LD formation starts from the ER, and its biogenesis can be represented by two models.According to the main budding model, LDs derive from the accumulation of newly synthesized neutral lipids thanks to TAG synthesis and cholesterol ester synthesis enzymes, which are deposited between the ER membrane bilayers.Then, the cytoplasmic leaflet with phospholipids and ER membrane protein buds, forming nascent LDs, is still linked to the ER [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref].At a certain lipid concentration, new LDs are released into the cytosol thanks to Arf1/COPI complexes and ER membrane proteins such as seipin (necessary for LD stabilization and growth), which may trigger bridge formation between the ER and nascent LDs with the rearrangement of membrane lipids, leading to asymmetrical budding in the cytosol [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref] [bib_ref] Membrane Asymmetry Imposes Directionality on Lipid Droplet Emergence from the ER, Chorlay [/bib_ref].
The alternative model shows that the neutral lipid accumulation inside the ER bilayer induces the formation of oil lens, which is subsequently released into the cytoplasm.After synthesis, LDs can grow thanks to lipid synthesis localized inside LDs, the transport of lipids to LDs, or fusion with other LDs [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref].
LD biogenesis is induced by an excess of non-esterified FAs and cholesterol, which are stored in LDs as neutral lipids, TAGs, and sterol esters, mainly cholesterol esters, to reduce lipotoxicity, as well as by external stimuli such as oxidative stress, infections, cytokines, and lipids [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] (Figure [fig_ref] Figure 1: Figure 1 [/fig_ref].
expression is mainly detected in oxidizing tissues such as liver, heart, pancreas, and adipose tissue but was also in contact with mitochondria, enhancing the FA β-oxidation in muscle cells [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Lipid Droplets in Cancer, Petan [/bib_ref] [bib_ref] Lipid Droplets in Cancer: Guardians of Fat in a Stressful World, Petan [/bib_ref] [bib_ref] Lipid droplets at a glance, Guo [/bib_ref] [bib_ref] Perilipin 5, a lipid droplet protein adapted to mitochondrial energy utilization, Kimmel [/bib_ref] [bib_ref] Perilipin 5, a lipid droplet-associated protein, provides physical and metabolic linkage to..., Wang [/bib_ref].
LD formation starts from the ER, and its biogenesis can be represented by two models.According to the main budding model, LDs derive from the accumulation of newly synthesized neutral lipids thanks to TAG synthesis and cholesterol ester synthesis enzymes, which are deposited between the ER membrane bilayers.Then, the cytoplasmic leaflet with phospholipids and ER membrane protein buds, forming nascent LDs, is still linked to the ER [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref].At a certain lipid concentration, new LDs are released into the cytosol thanks to Arf1/COPI complexes and ER membrane proteins such as seipin (necessary for LD stabilization and growth), which may trigger bridge formation between the ER and nascent LDs with the rearrangement of membrane lipids, leading to asymmetrical budding in the cytosol [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref] [bib_ref] Membrane Asymmetry Imposes Directionality on Lipid Droplet Emergence from the ER, Chorlay [/bib_ref].
The alternative model shows that the neutral lipid accumulation inside the ER bilayer induces the formation of oil lens, which is subsequently released into the cytoplasm.After synthesis, LDs can grow thanks to lipid synthesis localized inside LDs, the transport of lipids to LDs, or fusion with other LDs [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Dynamics and functions of lipid droplets, Olzmann [/bib_ref].
LD biogenesis is induced by an excess of non-esterified FAs and cholesterol, which are stored in LDs as neutral lipids, TAGs, and sterol esters, mainly cholesterol esters, to reduce lipotoxicity, as well as by external stimuli such as oxidative stress, infections, cytokines, and lipids [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] (Figure [fig_ref] Figure 1: Figure 1 [/fig_ref].TAGs are composed of a glycerol backbone linked to three FA chains, and their synthesis requires several steps: (i) Acyl-CoA synthetase (ACS) converts saturated and/or unsaturated FAs into fatty acyl-coenzyme A (FA-CoA) esters; (ii) glycerol kinase promotes the phosphorylation of glycerol or the formation of cytosolic glycerol-3-phosphate from dihydroxyacetone phosphate (DHAP) by glycerol-3-phosphate dehydrogenase during glycolysis (in the liver and adipose tissues), or it can be synthesized from peroxisomal conversion of the DHAP (in adipose and other tissues), which is subject to two acylation reactions by dihydroxyacetone phosphate acyltransferase and 1-acyl-dihydroxyacetone phosphate oxidoreductase (DHAP-OR); and (iii) FA-CoA is added to glycerol-3-phosphate to form 1acylglycerol-3-phosphate (MAG-P) by glycerol-3-phosphate acyltransferase.Subsequently, 1-acyl-glycerol-3-phosphate acyltransferase converts MAG-P into 1,2-diacylglycerol phosphate or phosphatidic acid (PA), which in turn can be dephosphorylated to 1,2-diacylglycerol (DAG) by phosphatidic acid phosphatase (lipin), a cytosolic Mg 2+ -dependent enzyme that catalyzes the reaction in the ER.The rate-limiting enzymes for TAG synthesis acyl-CoA:diacylglycerol acyltransferase 1 and 2 (DGAT1 and DGAT2) promotes the third esterification of DAG into TAG [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref].
Moreover, another main component of LDs is CE, which derives from the esterification between FA-CoA and cholesterol by acyl-CoA:cholesterol acyltransferase 1 and 2 (ACAT1 and ACAT2) enzyme isoforms.Notably, ACAT1 is ubiquitous, whereas ACAT2 is present in the intestine and liver in particular.Subsequently, TAGs and CEs are stored in LDs, or they can be released into the bloodstream as very-low-density lipoproteins (VLDLs) from the liver to reach other body tissues [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref].
Furthermore, based on the energetic condition, cells modulate the LD proteome in order to induce lipogenesis or lipolysis.For instance, under basal conditions, PLIN1 impairs the access of cytosolic lipase HSL to LDs and ATGL activity by binding with the cofactor perilipin-associated comparative gene identification-58 (CGI-58), which is essential to activate ATGL.Under lipogenic stimuli or an FA surplus, PLIN2, PLIN3, and PLIN4 are localized on nascent LDs at the ER to promote lipogenesis [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Perilipin A mediates the reversible binding of CGI-58 to lipid droplets in..., Subramanian [/bib_ref].
On the other hand, to provide energy in a short amount of time or in the case of β-adrenergic stimulation, LDs can be degraded by lipolysis, mobilizing TAGs (Figure [fig_ref] Figure 1: Figure 1 [/fig_ref].This process can occur through the degradation activity of cytoplasmic triglyceride lipases such as ATGL and HSL recruited to the LD surface or through lysosomal lipase activity followed by autophagy.In the first case, the active protein kinase A (PKA) phosphorylates PLIN1 and HSL, and this, in turn, promotes HSL binding to the LD surface.In parallel, CGI-58 release activates ATGL, promoting its localization to the LDs in order to stimulate TGA lipolysis thanks to its binding with GBF1 factor and with Arf/COPI complex [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Perilipin A is essential for the translocation of hormone-sensitive lipase during lipolytic..., Sztalryd [/bib_ref].
Therefore, TAGs are converted into DAGs thanks to the hydrolysis activity of ATGL; then HSL, recruited from cytosol to LDs, converts DAGs into 1-acylglycerols (MAGs); and these in turn are converted into free FAs (FFAs) and glycerol by monoacylglycerol lipase (MGL).Then, FFAs are delivered to the mitochondria, where β-oxidation takes place to produce energy, or as substrates for esterification, in order to support membrane synthesis or release signaling molecules [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] FAT SIGNALS-lipases and lipolysis in lipid metabolism and signaling, Zechner [/bib_ref] [bib_ref] Regulation of triglyceride metabolism. IV. Hormonal regulation of lipolysis in adipose tissue, Jaworski [/bib_ref].
LDs can also be enclosed in autophagosomes, which in turn are incorporated into lysosomes, forming autolysosomes, where lysosomal acid lipase (LAL) hydrolyzes TAGs and CEs, whereas proteases promote protein degradation.This mechanism is called lipophagy, which is a form of selective (macro)autophagy and is mainly activated in hepatocytes with less ATGL and HLS activity, but is still poorly understood [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Lipid Droplets in Cancer, Petan [/bib_ref] [bib_ref] Lysosome: Regulator of lipid degradation pathways, Settembre [/bib_ref].
FAs released from LDs can promote mitochondrial energy production but also act as signals that activate transcriptional factors, like peroxisome proliferator-activated receptors (PPARs), which are essential to the proper coupling of FA supply with mitochondrial biogenesis, function, and oxidative cell capacity [bib_ref] ATGL-mediated fat catabolism regulates cardiac mitochondrial function via PPAR-α and PGC-1, Haemmerle [/bib_ref].
In cancer cells, including AML cells, an excess of nonesterified FAs can derive from the upregulation of lipogenesis de novo or from the upregulation of FA uptake through cell surface receptors for plasma lipids, such as cluster of differentiation 36 (CD36) or fatty acid transport proteins (FATPs), inducing LD accumulation, which support cell proliferation and metastasis [bib_ref] Lipid Droplets in Cancer: From Composition and Role to Imaging and Therapeutics, Antunes [/bib_ref] [bib_ref] CD36 Drives Metastasis and Relapse in Acute Myeloid Leukemia, Farge [/bib_ref].
## Pathological roles of lds
Besides their physiological role, LDs exhibit several functions in cancer cells: (1) preventing lipotoxicity, including potential toxic endogenous and exogenous lipids, such as saturated and polyunsaturated FAs, CEs, and cholesterol, and storing them as inert triacylglycerols (TAGs), acylceramides, and CEs, respectively; (2) supporting energy and redox homeostasis by providing FAs for energy production but also to protect cells from ROS, thus acting as ROS scavengers and promoting NADPH synthesis and membrane biogenesis during rapid cell growth; (3) organizing FA trafficking and distribution; (4) regulating autophagy; (5) maintaining ER and membrane homeostasis, carrying out protein quality control, and protecting cells against ER stress; (6) producing bioactive lipid mediators and pro-and anti-inflammatory signaling molecules derived from polyunsaturated FAs, such as eicosanoids and specialized pro-resolving mediators; [bib_ref] Lipid Droplet Biosynthesis Impairment through DGAT2 Inhibition Sensitizes MCF7 Breast Cancer Cells..., Nisticò [/bib_ref] sequestering lipophilic compounds supporting chemoresistance; and (8) promoting epithelial mesenchymal transition (EMT) [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Lipid Droplets in Cancer: From Composition and Role to Imaging and Therapeutics, Antunes [/bib_ref] [bib_ref] Lipid Droplets in Cancer, Petan [/bib_ref] [bib_ref] Lipid Droplets in Cancer: Guardians of Fat in a Stressful World, Petan [/bib_ref].Additionally, several pieces of evidence have revealed that LDs of different cancer cells displayed an increase in proteins also involved in tumorigenesis, such as PI3K, extracellular signal-regulated kinase 1 and 2 (ERK1 and ERK2), caveolins, and cyclooxygenase 2 (COX-2) for prostaglandin E2 (PGE2) production [bib_ref] An Overview of Lipid Droplets in Cancer and Cancer Stem Cells, Tirinato [/bib_ref] [bib_ref] Caveolin-2 is targeted to lipid droplets, a new "membrane domain" in the..., Fujimoto [/bib_ref] [bib_ref] Lipid bodies are reservoirs of cyclooxygenase-2 and sites of prostaglandin-E2 synthesis in..., Accioly [/bib_ref] [bib_ref] Phosphatidylinositide 3-kinase localizes to cytoplasmic lipid bodies in human polymorphonuclear leukocytes and..., Yu [/bib_ref] [bib_ref] Lipid droplets in inflammation and cancer, Bozza [/bib_ref].
However, it was shown that LD accumulation also occurred in early apoptosis.This event was due to an increase in de novo lipid synthesis, which in turn inhibited fatty acid β-oxidation, with a consequent rapid rise in mitochondrial membrane potential and an attendant incremental release of reactive oxygen species (ROS) [bib_ref] Apoptosis-induced mitochondrial dysfunction causes cytoplasmic lipid droplet formation, Boren [/bib_ref].Therefore, LD modulation is dynamic and highly regulated and influences several pathways.
## Targeting lipid droplets in acute myeloid leukemia
Acute myeloid leukemia (AML) is the most common type of acute leukemia; in fact, the worldwide incidence is constantly increasing [bib_ref] ZNF521 Enhances MLL-AF9-Dependent Hematopoietic Stem Cell Transformation in Acute Myeloid Leukemias by..., Chiarella [/bib_ref].
It is a clonal malignancy of myeloid progenitor cells that maintain a hyperproliferative, undifferentiated state and survive by inhibiting apoptosis [bib_ref] Targeting of Mevalonate-Isoprenoid Pathway in Acute Myeloid Leukemia Cells by Bisphosphonate Drugs, Chiarella [/bib_ref] [bib_ref] Editorial to the Special Issue "Recent Advances in Biochemical Mechanisms of Acute..., Mesuraca [/bib_ref].This cell differentiation inhibition has been associated with the disruption of several cellular processes, including transcriptional, chromatin, and metabolic regulation.
The development of de novo AML is usually associated with advanced age, male sex, and exposure to chemicals or cigarette smoking; however, it can occurs secondarily to other diseases or treatment with cytotoxic agents or high doses of radiotherapy [bib_ref] Updates in infection risk and management in acute leukemia, Logan [/bib_ref].
AML patients can be stratified into three risk categories: favorable, intermediate, and adverse, based on their cytogenetic and molecular profile, as emerged from the latest guidelines from the European Leukemia Net (ELN) and World Health Organization (WHO) [bib_ref] AML classification in the year 2023: How to avoid a Babylonian confusion..., Huber [/bib_ref].
Understanding the molecular processes that lead to the genesis and progression of AML is critical to developing targeted therapies.
In the last few decades, several studies have revealed that lipid metabolism reprogramming plays a critical role in tumorigenesis and progression in various cancer types, including AML [bib_ref] Prognostic role of TPD52 in acute myeloid leukemia: A retrospective multicohort analysis, Ha [/bib_ref].Recently, researchers have explored the clinical value of lipid metabolism reprogramming in AML and developed a prognostic risk signature based on six lipid metabolism-related genes, including LDLRAP1, PNPLA6, DGKA, PLA2G4A, CBR1, and EBP.Indeed, this risk signature showed a significant correlation with clinical outcomes and immune cell infiltration in AML patients.Therefore, lipid droplets and lipid metabolism-related genes could be used as potential prognostic biomarkers and therapeutic targets in AML .
Notably, it was shown that AML patients exhibited alterations in sphingolipid metabolism and fatty acid accumulation and oxidation.Recent evidence has revealed that tumor protein D52 (TPD52), a regulator of lipid metabolism involved in fatty acid storage and lipid droplet formation, was overexpressed in several cancers, including AML [bib_ref] Prognostic role of TPD52 in acute myeloid leukemia: A retrospective multicohort analysis, Ha [/bib_ref] [bib_ref] Clinically Tractable Outcome Prediction of Non-WNT/Non-SHH Medulloblastoma Based on TPD52 IHC in..., Delaidelli [/bib_ref] [bib_ref] TPD52L2 Is a Prognostic Biomarker and Correlated With Immune Infiltration in Lung..., Zhong [/bib_ref].Indeed, this gene was overexpressed in three independent AML patient cohorts, which was associated with poor prognosis per the Kaplan-Meier curve and multivariate analysis, suggesting TPD52 as a possible biomarker for AML.
In addition, it was demonstrated that 20% of AML patients had mutations in the metabolic enzyme isocitrate dehydrogenase (IDH), which is involved in a variety of metabolic and epigenetic cellular processes, including lipid metabolism alterations, and whose alterations may differentially affect prognosis [bib_ref] The role of IDH mutations in acute myeloid leukemia, Montalban-Bravo [/bib_ref] [bib_ref] Stable Isotope Labeling Highlights Enhanced Fatty Acid and Lipid Metabolism in Human..., Stuani [/bib_ref].In particular, proteomic analysis of IDH1 mutant AML cells has revealed an upregulation of the protein involved in cholesterol and sterol biosynthesis and in fatty acid oxidation [bib_ref] Stable Isotope Labeling Highlights Enhanced Fatty Acid and Lipid Metabolism in Human..., Stuani [/bib_ref].
Torii et al. found a unique cell line, HPB-AML-I (AML-I), derived from peripheral blood mononuclear cells collected from a patient with acute myeloid leukemia (AML: M1), that exhibited several features related to preadipocytes, such as storage of LDs, and several surface antigens similar to bone marrow stromal cell antigens.Indeed, a cocktail (INC) composed of methylisobutylxanthine, hydrocortisone, and indomethacin was able to differentiate AML-I cells from adipocytes, but this effect was not reached with the adipogenic stimulator troglitazone.Conversely, PPARγ activation reduced LD content in AML-I cells.These findings suggest that there were some distinct lineages in human adipocytes and that the differentiation and regulation of human preadipocytes can be further investigated using the unique human derived-preadipocyte cell line AML-I [bib_ref] Establishment of a human preadipose cell line, HPB-AML-I: Refractory to PPARgamma-mediated adipogenic..., Torii [/bib_ref].
To date, a limited number of inhibitors of lipid metabolism and, in particular, of lipid droplet biogenesis, have been tested on cellular models of acute myeloid leukemia.
Among lipid metabolism inhibitors, galectin-12 was found to have a role in modulating LD production in acute myeloid leukemia in vitro.Galectin-12 belongs to the lectin family and plays a central role during adipogenesis, being a key regulator of C/EBPα, C/EBPβ, and PPARγ for lipid droplet formation [bib_ref] Galectin-12 is required for adipogenic signaling and adipocyte differentiation, Yang [/bib_ref].Moreover, galectin-12 expression is particularly abundant in AML samples classified as FAB subtype M3, corresponding to acute promyelocytic leukemia, as emerged from public datasets.APL cells are sustained by a chromosomal translocation involving the genes mapped on chromosomes 15 and 17 and resulting in the PML-RARα fusion protein.Specifically, galectin-12 silencing stimulated all-trans-retinoic acid (ATRA)-induced neutrophil differentiation by increasing CD11b cell-surface expression and NBT reduction, whereas lipid droplet formation turned out to be impaired.When galectin-12 knockdown NB4 cells were exposed to ATRA, lipid droplet accumulation was significantly lower compared to that of the control group.This phenotype is supported by the negative gene regulation of PPARγ, the master gene of adipocyte differentiation, as well as C/EBP transcription factors in ATRA-differentiated galectin-12 knockdown cells compared to control cells.The lipogenesis process is indeed firstly governed by C/EBPβ expression, which in turn activates C/EBPα and PPARγ expression.In addition, galectin-12 silencing in NB4 cells induces ROS production during ATRA differentiation by promoting NADPH oxidase subunit P47phox from cytoplasm to the plasma membrane.This evidence suggests that galectin-12 targeting in APL could represent an amenable strategy for treatment of the ATRA-resistant subset cells [bib_ref] Galectin-12 inhibits granulocytic differentiation of human NB4 promyelocytic leukemia cells while promoting..., Xue [/bib_ref].
On the other hand, a recent study investigated the contribution of lipid droplets to developing acquired resistance to CHR2863, an orally available hydrophobic aminopeptidase inhibitor in acute myeloid leukemia, generating CHR2863-resistant cell models.Among the results, they found that CHR2863-resistant AML cell models were able to sequester CHR2863 into LDs, with a consequent increase in LDs in cells and activation of the prosurvival Akt/mTOR pathway, which could be targeted using mTOR-targeted drugs like rapamycin to overcome CHR2863 drug resistance [bib_ref] Multifactorial resistance to aminopeptidase inhibitor prodrug CHR2863 in myeloid leukemia cells: Downregulation..., Verbrugge [/bib_ref].
Although FFAs have a lower energetic role in normal hematopoietic cells, they represent an important carbon source to constantly fuel the TCA cycle and oxidative phosphorylation in AML cells.Recently, Bosc and colleagues demonstrated in an elegant study that LD accumulation can be related to the inhibition of autophagy [bib_ref] Autophagy regulates fatty acid availability for oxidative phosphorylation through mitochondria-endoplasmic reticulum contact..., Bosc [/bib_ref].Specifically, when AML cell lines (MOLM14 and U937) as well as AML blasts and normal hematopoietic cells (peripheral blood mononuclear cells (PBMC) and CD34 + cells) were exposed to etomoxir (3 µM), a block of carnitine palmitoyl transferase 1 activity occurred and, consequently, FA degradation in the mitochondria.This metabolic alteration led to a significant decrease in ATP production and oxygen consumption rate (OCR) in AML cell lines and primary patient samples compared to primary normal hematopoietic cells [bib_ref] Autophagy regulates fatty acid availability for oxidative phosphorylation through mitochondria-endoplasmic reticulum contact..., Bosc [/bib_ref].
FFAs can be generated from lipid droplet degradation via autophagia, thereby sustaining the mitochondrial oxidative phosphorylation system (OxPHOS) and tumor lipid metabolism.Indeed, when MOLM14 and U937 were exposed to 3-methyladenine (3-MA), an inhibitor of autophagosome formation, an accumulation of lipids occurred, as emerged from flow cytometry analysis for a BODIPY probe as well as in the content of triglycerides.Similar data were obtained when targeting ATG12 with shRNA to produce inhibition of autophagosomes.Both of these approaches demonstrated a strong decrease in OCR and mitochondrial ATP production, suggesting the central role of autophagy in promoting AML cell survival through lipid catabolism [bib_ref] Autophagy regulates fatty acid availability for oxidative phosphorylation through mitochondria-endoplasmic reticulum contact..., Bosc [/bib_ref].This metabolic adaptation is not documented in normal hematopoietic cells, where lipids are not a major respiratory source.Interestingly, lipid metabolism can in turn be regulated by OxPHOS.Gene set enrichment analysis (GSEA) of AML cells exposed to metformin, a mitochondrial electron transfer chain (ETC) complex I inhibitor, revealed a downregulation of genes involved in lipid metabolism.However, the triglyceride level increased in AML cells exposed to metformin or to antimycin A, a specific ETC complex III inhibitor (AA), compared to normal hematopoietic cells, suggesting that the inhibition of the mitochondrial respiratory chain results in an accumulation of LD in AML cells without having a metabolic impact on normal hematopoietic cells.In acute myeloid leukemia cells, OxPHOS was found to be involved in controlling LD degradation via the autophagy process.In light of this, it was determined that mitochondrial ETC inhibition modulates autophagy; indeed, in metformin-or AA-treated AML cells, it a decrease in the conversion of LC3B-I to LC3B-II was observed, which is considered an indicator of autophagosome production in the presence or absence of chloroquine (chloro), used as inhibitor of lysosomal degradation.Mechanistically, LD accumulation subordinated to autophagy inhibition upon ETC inhibition was explained by the loss of the mitochondriaendoplasmic reticulum (ER) contact sites (MERCs), affecting cell proliferation in vitro and tumor growth in vivo [bib_ref] Autophagy regulates fatty acid availability for oxidative phosphorylation through mitochondria-endoplasmic reticulum contact..., Bosc [/bib_ref].This LD accumulation, associated with a rise in mitochondrial membrane potential and ROS production, probably promoted apoptosis in AML cells [bib_ref] Apoptosis-induced mitochondrial dysfunction causes cytoplasmic lipid droplet formation, Boren [/bib_ref].Therefore, LD modulation is a dynamic process that cancer cells can use to adapt to harsh conditions, such as nutrient deprivation or treatments or other external stimuli, and the relationship between mitochondrial activity and lipid homeostasis is crucial for AML cells to survive [bib_ref] Autophagy regulates fatty acid availability for oxidative phosphorylation through mitochondria-endoplasmic reticulum contact..., Bosc [/bib_ref].
Furthermore, many factors and proteins contribute to LD formation, the first of which is peroxisome proliferator-activated receptor γ (PPARγ).PPARγ is a well-recognized key regulator of adipocyte and macrophage differentiation processes; however, it also plays a central role in lipid and glucose metabolism in cancer cells [bib_ref] PPARγ and TGFβ-Major Regulators of Metabolism, Inflammation, and Fibrosis in the Lungs..., Kökény [/bib_ref] [bib_ref] Deficit in Adipose Differentiation in Mesenchymal Stem Cells Derived from Chronic Rhinosinusitis..., Chiarella [/bib_ref].
PPARγ expression was found to be upregulated in 30 AML-diagnosed patients compared to healthy patients.The abnormal expression of PPARγ induced a negative regulation of the tumor suppressor protein PTEN [bib_ref] Alteration of PPAR-GAMMA (PPARG; PPARγ) and PTEN gene expression in acute myeloid..., Esmaeili [/bib_ref].
A number of studies demonstrated the ability of PPARγ ligands to suppress proliferation in a variety of AML cell lines (HL-60, KG-1, Mono-MAC6, and THP-1) and primary myeloid leukemia cells by promoting differentiation or apoptosis.For example, the combined treatment of pioglitazone or PGJ2 with ATRA induced a block in the clonal proliferation of NB4 cells and simultaneously stimulated granulocytic maturation.Interestingly, in cells treated with PPARγ ligand and ATRA, an accumulation of lipid droplets and a significant increase in triacylglycerols were observed.In this context, PPARγ targeting could represent a strategy for cancer treatment [bib_ref] Peroxisome proliferator-activated receptor gamma ligands stimulate myeloid differentiation and lipogenensis in human..., Yasugi [/bib_ref].
Recent evidence has revealed that the levels of the bile acid chenodeoxycholic acid (CDCA) were less present in the feces and plasma of AML patients compared to healthy people but were also positively associated with gut microbiota diversity, predicting poor AML prognosis.Notably, in vitro and in vivo experiments have shown that CDCA was able to block AML progression, promoting mitochondrial dysfunction in terms of mitochondrial morphology damage, mitochondrial membrane potential reduction, high mitochondrial calcium levels, and overproduction of excessive reactive oxygen species (ROS).At the same time, this increase in ROS levels promoted p38 MAPK signaling pathway activation, with a consequent LD accumulation due to diacylglycerol O-acyltransferase 1 (DGAT1) overexpression, enhancing lipid peroxidation and apoptosis in AML cells.In parallel, it has been demonstrated that CDCA was able to impair M2 macrophage polarization, blocking AML cell proliferation as well, as result of co-cultured experiments [bib_ref] Chenodeoxycholic acid suppresses AML progression through promoting lipid peroxidation via ROS/p38 MAPK/DGAT1..., Liu [/bib_ref].
# Conclusions
The formation and progression of a tumor and the development of metastases are supported by metabolic alterations within cells, which imply the dysfunction of both glucose and lipid metabolism.In recent years, attention has been focused on the accumulation of lipid droplets, considered a great source of energy for neoplastic cells and therefore able to feed tumors with highly malignant genotypes and phenotypes.In this review, we highlighted emerging findings related to lipid droplet targeting in AML.We investigated studies in which inhibitory molecules, including galactin-12, PPARγ ligands, CDCA, CHR2863, and others, were used to modulate lipid droplet production in cellular models of AML (Figure [fig_ref] Figure 2: Figure 2 [/fig_ref].Although to date the amount of scientific evidence related to the blocking of LD biogenesis in AML is limited, it still represents an interesting challenge, especially for a translational perspective of AML treatment.
chondrial calcium levels, and overproduction of excessive reactive oxygen species (ROS).At the same time, this increase in ROS levels promoted p38 MAPK signaling pathway activation, with a consequent LD accumulation due to diacylglycerol O-acyltransferase 1 (DGAT1) overexpression, enhancing lipid peroxidation and apoptosis in AML cells.In parallel, it has been demonstrated that CDCA was able to impair M2 macrophage polarization, blocking AML cell proliferation as well, as result of co-cultured experiments [bib_ref] Chenodeoxycholic acid suppresses AML progression through promoting lipid peroxidation via ROS/p38 MAPK/DGAT1..., Liu [/bib_ref].
# Conclusions
The formation and progression of a tumor and the development of metastases are supported by metabolic alterations within cells, which imply the dysfunction of both glucose and lipid metabolism.In recent years, attention has been focused on the accumulation of lipid droplets, considered a great source of energy for neoplastic cells and therefore able to feed tumors with highly malignant genotypes and phenotypes.In this review, we highlighted emerging findings related to lipid droplet targeting in AML.We investigated studies in which inhibitory molecules, including galactin-12, PPARγ ligands, CDCA, CHR2863, and others, were used to modulate lipid droplet production in cellular models of AML (Figure [fig_ref] Figure 2: Figure 2 [/fig_ref].Although to date the amount of scientific evidence related to the blocking of LD biogenesis in AML is limited, it still represents an interesting challenge, especially for a translational perspective of AML treatment.
[fig] Figure 1: Figure 1.Schematic representation of LD production and accumulation resulting from the fine balance between lipolysis and lipogenesis.TAG: triacylglycerol, ATGL: adipose triglyceride lipase, LAL: lysosomal acid lipase, DAG: diacylglycerol, HSL: hormone-sensitive lipase, MAG: monoacylglycerol, MAGL: monoacylglycerol lipase, DGAT: diacylglycerol acyltransferase, ACAT: acyl-CoA cholesterol acyltransferase, FA: fatty acids, Chol: cholesterol. [/fig]
[fig] Figure 2: Figure 2. Schematic representation of the main molecules involved in targeting LD accumulation in AML.Author Contributions: Conceptualization, C.N., E.C., C.N. and E.C.; writing-original draft preparation, C.N. and E.C.; writing-review and editing, C.N. and E.C.; funding acquisition, C.N. and E.C.All authors have read and agreed to the published version of the manuscript. [/fig]
|
10.1186/s12986-017-0230-2
|
CCBY
|
5704430
|
29209405
|
s2orc_pubmed_articles
|
Retroconversion is a minor contributor to increases in eicosapentaenoic acid following docosahexaenoic acid feeding as determined by compound specific isotope analysis in rat liver
Dietary docosahexaenoic acid (DHA, 22:6n-3) not only increases blood and tissue levels of DHA, but also eicosapentaenoic acid (EPA,. It is generally believed that this increase is due to DHA retroconversion to EPA, however, a slower conversion of α-linolenic acid (ALA, 18:3n-3) derived EPA to downstream metabolic products (i.e. slower turnover of EPA) is equally plausible. In this study, 21-day old Long Evans rats were weaned onto an ALA only or DHA + ALA diet for 12 weeks. Afterwards, livers were collected and the natural abundance 13 C-enrichment was determined by compound specific isotope analysis (CSIA) of liver EPA by isotope ratio mass-spectrometry and compared to dietary ALA and DHA 13 C-enrichment. Isotopic signatures (per mil, ‰) for liver EPA were not different (p > 0.05) between the ALA only diet (−25.89 ± 0.39 ‰, mean ± SEM) and the DHA + ALA diet (−26.26 ± 0.40 ‰), suggesting the relative contribution from dietary ALA and DHA to liver EPA did not change. However, with DHA feeding estimates of absolute EPA contribution from ALA increased 4.4-fold (147 ± 22 to 788 ± 153 nmol/g) compared to 3.2-fold from DHA (91 ± 14 to 382 ± 13 nmol/g), respectively. In conclusion, CSIA of liver EPA in rats following 12-weeks of dietary DHA suggests that retroconversion of DHA to EPA is a relatively small contributor to increases in EPA, and that this increase in EPA is largely coming from elongation/desaturation of ALA.
# Introduction
Dietary docosahexaenoic acid (DHA, 22:6n-3) is primarily obtained from marine and seafood sources [bib_ref] Changes in consumption of omega-3 and omega-6 fatty acids in the United..., Blasbalg [/bib_ref] [bib_ref] Polyunsaturated fatty acids in the food chain in Europe, Sanders [/bib_ref] , and increased consumption of n-3 polyunsaturated fatty acids (n-3 PUFA) are known to increase blood and tissue n-3 PUFA levels. These changes in tissue n-3 fatty acids can regulate numerous biological processes including many that are related to brain function [bib_ref] Polyunsaturated fatty acids and their metabolites in brain function and disease, Bazinet [/bib_ref]. Dietary DHA not only increases blood and tissue DHA concentrations, but has repeatedly been shown to increase eicosapentaenoic acid (EPA, 20:5n-3)a metabolic precursor in the biosynthesis of DHAas well, with very recent studies in humans [bib_ref] Supplementation with high-dose docosahexaenoic acid increases the Omega-3 index more than high-dose..., Allaire [/bib_ref] , rats [bib_ref] Serum n-3 Tetracosapentaenoic acid and Tetracosahexaenoic acid increase following higher dietary alpha-Linolenic..., Metherel [/bib_ref] , mice [bib_ref] Docosahexaenoic acid blocks progression of western diet-induced nonalcoholic steatohepatitis in obese Ldlr−/−..., Lytle [/bib_ref] and pigs [bib_ref] Effect of alpha-linolenic acid and DHA intake on lipogenesis and gene expression..., De Tonnac [/bib_ref] demonstrating this response. In vitro evidence for retroconversion appears to be mediated by a round of peroxisomal βoxidation in combination with saturation by Δ3,Δ2-enoyl-CoA isomerase, Δ3,5,Δ2,4-dienoyl-CoA isomerase and 2,4-dienoyl-CoA reductase [bib_ref] Evidence for peroxisomal retroconversion of adrenic acid (22:4(n-6)) and docosahexaenoic acids (22:6(n-3))..., Hagve [/bib_ref] [bib_ref] ) dominates over elongation to tetracosahexaenoic acid, Park [/bib_ref] [bib_ref] The Zellweger syndrome: deficient conversion of docosahexaenoic acid (22:6(n-3)) to eicosapentaenoic acid..., Gronn [/bib_ref] [bib_ref] Peroxisomal retroconversion of docosahexaenoic acid (22:6(n-3)) to eicosapentaenoic acid (20:5(n-3)) studied in..., Gronn [/bib_ref] [bib_ref] Peroxisomal betaoxidation-a metabolic pathway with multiple functions, Poirier [/bib_ref] [bib_ref] Functional redundancy of mitochondrial enoyl-CoA isomerases in the oxidation of unsaturated fatty..., Van Weeghel [/bib_ref]. Recently, Park et al. [bib_ref] ) dominates over elongation to tetracosahexaenoic acid, Park [/bib_ref] demonstrated retroconversion in breast epithelial (MCF7), liver (HepG2), neuronal (SK-N-SH) and retinal (Y79) cancer cell lines, and their results suggest that although retroconversion is present in each tissue, retroconversion is higher in the non-neuronal tissue.
Although the evidence for retroconversion of DHA to EPA in vitro is clear, much of the earlier evidence for retroconversion of DHA to EPA in vivo was supported through dietary feeding studies where dietary DHA also increases blood or tissue EPA levels [bib_ref] Supplementation with an algae source of docosahexaenoic acid increases (n-3) fatty acid..., Conquer [/bib_ref] , or from oral doses of [bib_ref] Functional redundancy of mitochondrial enoyl-CoA isomerases in the oxidation of unsaturated fatty..., Van Weeghel [/bib_ref] C-DHA tracers [bib_ref] Retroconversion and metabolism of [13C]22:6n-3 in humans and rats after intake of..., Brossard [/bib_ref]. In dietary DHA feeding studies an assumption is often made that all increases in EPA are a result of retroconversion [bib_ref] Supplementation with an algae source of docosahexaenoic acid increases (n-3) fatty acid..., Conquer [/bib_ref] [bib_ref] Effects of docosahexaenoic acid supplementation on PUFA levels in red blood cells..., Schuchardt [/bib_ref] , and these studies estimate that 7-14% of DHA is retroconverted to EPA; however, the source of EPA in such feeding studies cannot be verified to be DHA or a precursor to EPA such as αlinolenic acid (ALA, 18:3n-3). In the latter case, the desaturation and elongation of ALA may be slowed at EPA such that the turnover of EPA via elongation/desaturation, eicosanoid production or even β-oxidation may be downregulated. Oral doses of 13 C-DHA indicate a much lower rate of retroconversion based on plasma appearance of 13 C-EPA, [bib_ref] Functional redundancy of mitochondrial enoyl-CoA isomerases in the oxidation of unsaturated fatty..., Van Weeghel [/bib_ref] C-docosapentaenoic acid (DPAn-3, 22:5n-3) and [bib_ref] Functional redundancy of mitochondrial enoyl-CoA isomerases in the oxidation of unsaturated fatty..., Van Weeghel [/bib_ref] -ALA to be between 0.7 and 4.3% [bib_ref] Retroconversion and metabolism of [13C]22:6n-3 in humans and rats after intake of..., Brossard [/bib_ref] [bib_ref] Disturbance in uniformly 13C-labelled DHA metabolism in elderly human subjects carrying the..., Chouinard-Watkins [/bib_ref] [bib_ref] Plasma incorporation, apparent retroconversion and betaoxidation of 13C-docosahexaenoic acid in the elderly, Plourde [/bib_ref] in humans and 9% in rats [bib_ref] Retroconversion and metabolism of [13C]22:6n-3 in humans and rats after intake of..., Brossard [/bib_ref]. Although isotopic evidence for retroconversion of DHA to EPA discussed above appears strong, it remains unclear what proportion of the increases in EPA following DHA intake in vivo, if any, may be attributed to either increased retroconversion or a decreased EPA turnover. Recently, our lab has exploited differences in natural 13 C/ 12 C abundance between dietary ALA and DHA sources combined with compound specific isotope analysis (CSIA) by isotope-ratio mass-spectrometry (IRMS) to determine the origin of mouse brain DHA [bib_ref] Compound specific isotope analysis resolves the dietary origin of docosahexaenoic acid (DHA)..., Lacombe [/bib_ref].
Materials isolated from C 3 plants such as ALA, are isotopically depleted (−23‰ to −32‰) in 13 C compared to materials isolated from C 4 plants (−10‰ to −16‰) [bib_ref] Two categories of c/c ratios for higher plants, Smith [/bib_ref]. Furthermore, the fractionation of carbon isotopes from aquatic organism results in 13 C enrichment of aquatic materials such as DHA that differs slightly (−16‰ to −22%) from both C 3 and C 4 plants [bib_ref] Compound specific isotope analysis resolves the dietary origin of docosahexaenoic acid (DHA)..., Lacombe [/bib_ref] [bib_ref] Geographical variations in the carbon isotope composition of the diet and hair..., Nakamura [/bib_ref]. These natural differences in isotopic enrichment of ALA and DHA allows for determination of the dietary source of individual fatty acids in the PUFA biosynthesis pathway. Therefore, through the determination of the natural 13 C/ 12 C liver EPA abundance levels, the purpose of the current study is to CSIA to identify the source of liver EPA following 12 weeks of an ALA only diet compared to a DHA + ALA diet. Briefly, we identified no differences in the natural 13 C/ 12 C abundance in liver EPA between ALA only and DHA + ALA diets despite nearly 5-fold higher total liver EPA concentration on the DHA + ALA diet. We further determined that ALA was the primary source for the increase in liver EPA following DHA feeding; suggesting that slower metabolism of EPAand not retroconversion from DHAmay be the primary mechanism responsible for the increased EPA. The present analysis was performed as sub-analysis from a larger DHA supplementation study that has yet to be published.
# Methods
## Animals
All experimental procedures were performed in agreement with the policies set out by the Canadian Council on Animal Care and were approved by the Animal Ethics Committee at the University of Toronto (Protocol # -20011797). The animals and data used for this study were taken as a sub-analysis of a larger animal DHA supplementation protocol. Long Evans dams with male non-littermate pups were ordered from Charles River Laboratories (St. Constant, QC, Canada). Upon arriving at the University of Toronto, the dams and pups were acclimated for 3 days and at 21-days old the pups were randomly assigned and weaned onto either a 2% ALA diet (n = 6) or a 2% ALA +2% DHA diet (n = 6) for 12 weeks. At 15-weeks old, animals were euthanized by carbon dioxide asphyxiation and liver samples were collected from the right medial lobe, flash-frozen in liquid nitrogen and stored at −80°C until analysis.
## Diets
The diets were modified from the AIN-93G custom low n-3 rodent diets (Dyets, Inc., Bethlehem, PA, USA). The diets contained 10% lipids by weight with the fat content being 32.8% safflower oil, 63.2% hydrogenated coconut oil and 4% added fatty acid ethyl ester oils. The added oils were 2% oleate ethyl ester (Nu-Chek Prep, Inc., Elysian, MN, USA) and 2% ALA ethyl ester (gift from BASF Pharma, Callanish Ltd., Isle of Lewis, UK) in the ALA diet, and in the DHA diet the added oleate ethyl ester was replaced by 2% DHA ethyl ester (gift from BASF Pharma, Callanish Ltd.). Oleate ethyl ester was included in the ALA diet to maintain the total fat content equivalent to that of the DHA diet. Each oil was determined to be >98% pure by gas chromatographyflame ionization detection (GC-FID).
## Lipid extraction
Total lipid extracts (TLE) were obtained from approximately 50 mg of liver tissue by a method modified from Folch, Lees and Sloane-Stanley [bib_ref] A simple method for the isolation and purification of total lipides from..., Folch [/bib_ref]. Briefly, liver tissue was homogenized with a Polytron benchtop homogenizer (Brinkman Instruments, Toronto, ON, Canada) in 6 mL of 2:1 chloroform methanol containing 30 mg of docosatrianoic acid (22:3n-3) ethyl ester as internal standard, and vortexed. A 0.88% potassium chloride aqueous buffer (1.75 mL) was added, and samples were inverted twice and centrifuges at 500 g for 10 min to separate the lipidcontaining chloroform phase from the aqueous phase. The chloroform lipid-containing layer was collected and stored in a new test tube.
## Transmethylation and gc-fid
An aliquot of the TLE was collected, dried under nitrogen and transmethylated via a method adapted from Morrison and Smith [bib_ref] Preparation of fatty acid methyl esters and dimethylacetals from lipids with boron..., Morrison [/bib_ref] using 1 mL 14% boron trifluoride in methanol and 0.3 mL hexane at 100°C for 1 h. Hexane and water (1 mL each) were added, vortexed and centrifuged at 500 g for 5 min. The hexane layer containing fatty acid methyl esters (FAMEs) was collected, evaporated under nitrogen, reconstituted in hexane and stored in GC vials for analysis by GC-FID as previously described [bib_ref] Compound specific isotope analysis resolves the dietary origin of docosahexaenoic acid (DHA)..., Lacombe [/bib_ref]. Peaks were identified by retention times through comparison to an external FAME standard (GLC-569, Nu-Chek Prep Inc.). After running on GC-FID, samples were re-capped and stored at −80°C for CSIA analysis by IRMS.
# Isotopic analysis
CSIA of isolated FAMEs was performed by GC-IRMS, as described previously in detail [bib_ref] Compound specific isotope analysis resolves the dietary origin of docosahexaenoic acid (DHA)..., Lacombe [/bib_ref]. Briefly, FAMEs (2 μL) were injected in splitless mode using a TriPlus RSH autosampler (Thermo Scientific, Bremen, Germany) onto the SP-2560 capillary column described earlier interfaced in a Trace 13,010 GC (Thermo Scientific). Complete resolution of analyte peaks of interest from surrounding peaks is imperative for highprecision CSIA by GC-IRMS [bib_ref] Stable isotope analysis of fatty acids by gas chromatography-isotope ratio mass spectrometry, Meier-Augenstein [/bib_ref]. Therefore, coelution of EPA with 24:0 using the oven temperature program described previously necessitated an alternate temperature program for complete resolution of EPA from surrounding peaks by GC-IRMS. Complete baseline separation of EPA was achieved with the following oven temperature program: initial temperature of 60°C with an immediate ramp of 15°C/min to 180°C and an immediate ramp to 1.5°C/min to 240°C and an 18 min hold. Column flow rate was set to 1.2 mL/min.
## Isotopic normalization
Isotopic abundance data collected by IRMS was normalized and converted to the international carbon isotope reference scale, Vienna Peedee Belemnite (VPBD), by multi-point linear normalization and reported as per mil (‰) [bib_ref] Normalization of measured stable isotopic compositions to isotope reference scales-a review, Paul [/bib_ref]. Certified calibrated 20-carbon FAME reference materials USGS70, USGS71 and USGS72 were injected at least once each during the sequence. Linear regression of measured values versus true values (−30.53 ± 0.04, −10.50 ± 0.03, and −1.54 ± 0.03 ‰ for USGS70, USGS71, and USGS72, respectively) was used to generate the normalizing equation to report δ 13 C values for all data. R 2 values for all normalizing equations were >0.9998.
## Methylation correction
Isotopic analysis by GC-IRMS provides data on CO 2 produced from the combustion of individual FAMEs, and therefore measurements include the isotopic contribution of carbon from the derivatized methyl group. To account for the contribution of the derivatized carbon, a methyl correction calculation was performed as previously described in detail [bib_ref] Compound specific isotope analysis resolves the dietary origin of docosahexaenoic acid (DHA)..., Lacombe [/bib_ref].
## Estimation of dietary source of fames
Isotopic signatures in the ALA only diet have previously been determined to −28.22 ± 0.29 ‰ for ALA and −30.44 ± 0.09 ‰ for linoleic acid (LNA, 18:2n-6) [bib_ref] Compound specific isotope analysis resolves the dietary origin of docosahexaenoic acid (DHA)..., Lacombe [/bib_ref] , and in the DHA + ALA diet we determined isotopic signatures at −22.30 ± 0.27 ‰ for DHA, −28.54 ± 0.22 ‰ for ALA and −30.77 ± 0.08 ‰ for LNA [fig_ref] Table 1: Dietary fatty acid composition and isotopic signatures [/fig_ref]. Since ALA and DHA are the only n-3 fatty acids present in the diets, we can compare these dietary signatures to those determined in the liver and estimate the proportion of a specific FAME, presently EPA, that is derived from either the ALA source or the DHA source [bib_ref] Changes in tissue lipid and fatty acid composition of farmed rainbow trout..., Hixson [/bib_ref]. The equation is as follows:
[formula] δ EPAliver 13 C ¼ x à δ ALAdiet 13 C þ 1-x ð ÞÃδ DHAdiet 13 Cð1Þ [/formula]
Where x = the proportion of liver EPA from dietary ALA, 1x = the proportion of liver EPA from dietary DHA. These proportions can then be applied to total liver EPA concentrations to estimate the absolute amounts of liver EPA derived from either ALA or DHA. The equations are as follows:
[formula] EPA ½ Total ¼ EPA ½ ALAdiet þ EPA ½ DHAdietð2Þ [/formula]
where,
[formula] EPA ½ ALAdiet ¼ x à EPA ½ Totalð3ÞEPA ½ DHAdiet ¼ 1-x ð Þ Ã EPA ½ Totalð4Þ [/formula]
Statistics All statistical analyses were performed with IBM SPSS Statistics 24 software (IBM, Armonk, New York). Differences between diets for total fatty acid concentrations and proportion of EPA from ALA and arachidonic acid (ARA, 20:4n-6) from LNA were determined by Student's t-test. Differences in concentration estimates of sources of liver EPA and ARA were assessed by two-way ANOVA followed by Student's t-test between diets and paired t-test between sources. Significance for all statistical analyses was determined at p < 0.05. All data is presented as mean ± SEM.
# Results
## Liver fatty acid concentrations
Total liver fatty acid concentrations were determined for EPA and ARA and are presented in nmol/g and μmol/g, respectively. EPA increased (p < 0.05) from 238 ± 16 nmol/g on the ALA only diet to 1169 ± 158 nmol/g on the DHA + ALA diet, an increase of nearly 4-fold. In addition, liver DPAn-3 concentrations increased (p < 0.05) from 581 ± 41 nmol/g to 979 ± 99 nmol/g on the ALA only and DHA + ALA diet, respectively (data not shown). No difference (p = 0.13) in ARA concentration was determined between the ALA diet (30.1 ± 1.2 μmol/g) and DHA + ALA (26.8 ± 1.1 μmol/g).
## Eicosapentaenoic acid and arachidonic acid isotopic signatures
After 12 weeks of dietary feeding, there was no change in the isotopic signature of liver EPA with values determined to be −25.89 ± 0.39 ‰ and −26.26 ± 0.40 ‰ for the ALA only and DHA + ALA diets, respectively . Conversely, liver ARA on the ALA only diet was less enriched (p = 0.017) in 13 C at −29.32 ± 0.04 ‰ compared to the DHA + ALA only diet at −29.07 ± 0.08 ‰ .
Contribution of dietary ALA to EPA and dietary LNA to ARA Using Eq. 1 described earlier, we estimated that there is no difference (p > 0.05) in the proportion of liver EPA that was derived from dietary ALA sources in the ALA only diet (60.6 ± 6.6%) compared to the DHA + ALA diet (63.4 ± 6.4%) . Conversely, a higher (p < 0.05) proportion of liver ARA was estimated to be derived from dietary linoleic acid (LNA, 18:2n-6) in the ALA only diet (86.7 ± 0.5%) as compared to the DHA + ALA diet (80.6 ± 0.9%) . Utilizing these proportional estimates, the absolute contribution of ALA versus DHA sources to liver EPA concentrations, and the absolute contribution of LNA or other sources to liver ARA concentration were estimated using Eq. 2 . Liver concentration of EPA derived from dietary ALA sources increased (p < 0.01) from 147 ± 22 nmol/g on the ALA only diet to 788 ± 153 nmol/g on the DHA + ALA diet. Similarly, liver EPA derived from dietary DHA sources (or more 13 C-enriched sources) was also higher (p < 0.001) on the DHA + ALA diet (382 ± 13 nmol/g) compared to the ALA only diet (91 ± 14 nmol/g). These contributions represent an apparent retroconversion rate of 4.5 ± 0.6% as a result of dietary DHA + ALA feeding. Liver concentration of ARA derived from LNA was higher (p < 0.05) in the ALA only
# Discussion
Presently, we performed a sub-analysis on a dataset obtained from a larger DHA supplementation study in Long Evans rats. Liver TLE were analyzed by GC-FID for concentration and by GC-IRMS for CSIA of EPA and ARA to determine the effects of 12 weeks of DHA feeding on the dietary contribution to liver EPA, and ultimately the contribution of apparent retroconversion of DHA to EPA. Briefly, we demonstrated that adding 2% dietary DHA to a 2% ALA diet does not change the isotopic signature of liver EPA. However, due to the nearly 4-fold increase in total liver EPA we estimated that the absolute amount of liver EPA derived from both dietary ALA and DHA are higher following DHA + ALA feeding compared to ALA feeding only. Although it is not clear what lipid fraction liver EPA is being incorporated into presently, previous studies have shown that following 14 weeks DHA supplementation in humans plasma EPA concentration increases in the phospholipid fraction with no changes in EPA concentration in plasma triacylglycerols or cholesteryl esters [bib_ref] Incorporation of n-3 fatty acids into plasma lipid fractions, and erythrocyte membranes..., Vidgren [/bib_ref]. Overall, our carbon isotopic analysis suggests that while retroconversion of dietary DHA to liver EPA may be occurring, it is a relatively small contributor to increases in EPA compared to a reduced turnover of EPA. For the purpose of identifying any potential underlying contribution to liver EPA from carbon recycling through glycolysis, we included an assessment of the carbon isotopic abundance of ARA which is a product of the shared PUFA synthesis pathway, but of the n-6 portion. This is important since both diets include the 18-carbon n-6 dietary equivalent to ALA, with 24-29% LNA [fig_ref] Table 1: Dietary fatty acid composition and isotopic signatures [/fig_ref] , but no dietary equivalent to DHA. Following 12 weeks ingestion of the DHA + ALA diet, liver ARA was significantly more enriched compared to the ALA only diet, suggesting that a more enriched source of ARA may be present, independent of retroconversion. Dietary carbohydrates present in our diets are isolated from corn, which is a C 4 plant. C 4 plants are highly enriched in 13 C with carbon isotopic signatures of between −10 and −16 ‰ compared to C 3 plants with signatures between −23 and −32 ‰ [bib_ref] Two categories of c/c ratios for higher plants, Smith [/bib_ref]. As such, acetyl-CoA produced from glycolysis can be used in the chain elongation of LNA to ARA, thereby increasing the 13 C enrichment of ARA. Alternatively, β-oxidation favoring the isotopically lighter carbons [bib_ref] Carbon isotopic evidence for different feeding patterns in two hyrax species occupying..., Deniro [/bib_ref] or carbon recycling from DHA into ARA through the acetate pool [bib_ref] Linoleate, alpha-linolenate, and docosahexaenoate recycling into saturated and monounsaturated fatty acids is..., Greiner [/bib_ref] could further explain the 13 C-enrichment of ARA following the DHA + ALA diet.
Finally, the presence of a very small amount (0.07 ± 0.0003%) of ARA in the DHA + ALA diet could influence the results; however, levels were too low to determine isotopic signatures. The potential 13 C enrichment of liver ARA from acetyl-CoA, β-oxidation or acetate is important when discussing the apparent retroconversion of DHA to EPA as similar contributions may be present for EPA, and may therefore overestimate the determined EPA contribution from DHA. In fact, liver EPA enrichment on the ALA only diet appears to indicate that only 61% of liver EPA is coming from ALA suggesting a more enriched source of carbons for EPA on the ALA only diet, and indicates a significant contribution from the acetyl-CoA pool via glycolysis or preferential βoxidation. A summary of the potential carbon contributions to EPA is provided in [fig_ref] Figure 4: Representation of theoretical carbon flow influencing 13 C enrichment of EPA in... [/fig_ref].
Despite these findings, no differences in the estimated proportion of liver EPA derived from ALA or DHA were determined between ALA only and DHA + ALA diets; however, the estimated absolute contribution to liver EPA from dietary ALA and DHA were increased by 4.4 and 3.2-fold following the DHA + ALA diet, respectively. This suggests that although there may be more EPA being produced from DHA, it does not appear to be a result of an upregulation of enzymes in the retroconversion pathway. In support, the expression of 2,4-dienoyl-CoA reductase 2, an enzyme involved in the removal of one double bond from DHA for retroconversion [bib_ref] ) dominates over elongation to tetracosahexaenoic acid, Park [/bib_ref] , shows no effect of DHA supplementation in pigs fed increasing dietary DHA [bib_ref] Effect of alpha-linolenic acid and DHA intake on lipogenesis and gene expression..., De Tonnac [/bib_ref]. However, they also showed no effect on elongation of very long-chain 5 (Elovl5), one of two enzymes (the other being Elovl2) responsible for the elongation of EPA to DPAn-3. Our IRMS data indicates that the increase in liver EPA concentration is not the result of an increase in proportional contribution from DHA retroconversion, suggesting that EPA elongation into DPAn-3 may in fact be lower suggesting a downregulation of Elovl5 and/or Elovl2. Importantly, β-oxidation and/or [bib_ref] Supplementation with high-dose docosahexaenoic acid increases the Omega-3 index more than high-dose..., Allaire [/bib_ref] slower elongation/desaturation of EPA to DHA, [bib_ref] Serum n-3 Tetracosapentaenoic acid and Tetracosahexaenoic acid increase following higher dietary alpha-Linolenic..., Metherel [/bib_ref] preferential β-oxidation of lighter 12 C isotopes from EPA and additional EPA metbaolism. ALA -α-linolenic acid, 18:3n-3; DHA docosahexaenoic acid, 22:6n-3; EPAeicosapentaenoic acid, 20:5n-3 eicosanoid production from EPA among other pathways may also be downregulated. However, although liver DPAn-3 concentrations are also significantly higher on our DHA + ALA diet, the increase of only 69% is much less pronounced than the 490% higher EPA levels on the DHA + ALA diet, and is supportive of a reduced elongation of EPA to DPAn-3. Furthermore, dietary DHA supplementation on a high-fat diet significantly reduced Elovl5 expression in rats, where Elovl2 was lower but not significantly [bib_ref] Different effects of eicosapentaenoic and docosahexaenoic acids on Atherogenic high-fat dietinduced non-alcoholic..., Suzuki-Kemuriyama [/bib_ref]. Low affinity of Elovl2 for 18-carbon PUFAs combined with high affinity for 20-carbon PUFAs particularly n-3 s [31]would implicate Elovl2 as a primary contributor to lower EPA turnover/elongation with DHA feeding. Interestingly, Elovl2 knockout mice have higher liver EPA levels compared to wild type mice [bib_ref] Elovl2 ablation demonstrates that systemic DHA is endogenously produced and is essential..., Pauter [/bib_ref] in a manner similar to that of dietary DHA shown in our study. Future studies should be aimed at assessing the role of Elovl2 and Elovl5 in increased EPA concentrations following DHA feeding.
The increase in estimated EPA via retroconversion shown presently represents 4.5 ± 0.6% of DHA being retroconverted to EPA after 12 weeks of feeding, and is lower than the 13.9 ± 2.9% calculated if all increases in EPA were assumed to be due to retroconversion. The latter calculation has been implemented previously with apparent retroconversion of DHA to EPA determined to be between 9 and 11% [bib_ref] Supplementation with an algae source of docosahexaenoic acid increases (n-3) fatty acid..., Conquer [/bib_ref] [bib_ref] Dietary docosahexaenoic acid as a source of eicosapentaenoic acid in vegetarians and..., Conquer [/bib_ref] [bib_ref] Differential eicosapentaenoic acid elevations and altered cardiovascular disease risk factor responses after..., Stark [/bib_ref] in human plasma/ serum. Our values are comparable to apparent retroconversion determinations of between 0.7 and 4.3% in human plasma 28 days following an oral dose of isotopically labeled 13 C-DHA from individuals not supplemented with DHA [bib_ref] Disturbance in uniformly 13C-labelled DHA metabolism in elderly human subjects carrying the..., Chouinard-Watkins [/bib_ref] [bib_ref] Plasma incorporation, apparent retroconversion and betaoxidation of 13C-docosahexaenoic acid in the elderly, Plourde [/bib_ref] , and 5 months of fish oil supplementation does not increase the apparent retroconversion to EPA as determined by the appearance of plasma 13 C-EPA [bib_ref] Kinetics of 13C-DHA before and during fish-oil supplementation in healthy older individuals, Plourde [/bib_ref]. Furthermore, if retroconversion was solely responsible for the higher liver EPA with DHA feeding, liver EPA would become isotopically enriched and more closely resemble the enrichment of dietary DHA (22.30 ± 0.27‰), and at 25.89 ± 0.39‰ and −26.26 ± 0.40‰ for the two diets, this is not the case.
# Conclusions
By utilizing CSIA of liver EPA by IRMS following 12weeks supplementation of an ALA only diet or a DHA + ALA diet we have determined that the primary source of increased EPA with DHA feeding is from ALA and not DHA. Although the liver is believed to be the major site of PUFA biosynthesis and retroconversion these results may be limited to liver EPA determinations only, and as an endpoint measure in a dynamic tissue may not fully represent values determined on a whole-body or specific tissue level. Based on our values, previous determinations of apparent retroconversion may be overestimated as a smaller proportion of the increase in liver EPA than previously reported may be attributed to DHA retroconversion. Moreover, the contribution of DHA to EPA may be further overestimated as additional [bib_ref] Functional redundancy of mitochondrial enoyl-CoA isomerases in the oxidation of unsaturated fatty..., Van Weeghel [/bib_ref] C enrichment via preferential β-oxidation of 12 C carbon and/or 13 C-enriched acetyl-CoA from glycolysis may be present. In conclusion, the rate of retroconversion of DHA to EPA following dietary DHA feeding appears to be a minor contributor, with the majority of EPA derived from the elongation and desaturation of ALA.
[fig] Figure 1 3, Figure 2: Liver concentrations of (a) eicosapentaenoic acid and (b) arachidonic acid following 12 weeks ingestion of 2% ALA or 2% DHA + 2% ALA diet in male Long Evans rats. *represents statistically significant liver concentration between dietary groups as determined by significant Student's t-test, p < 0.05. N = 6, mean ± SEM. ALA -α-linolenic acid, 18:3n-3; ARAarachidonic acid, 20:4n-6; DHAdocosahexaenoic acid, 22:6n-3; EPAeicosapentaenoic acid, 20:5n-Liver carbon isotope signatures for eicosapentaenoic acid and arachidonic acid in male Long Evans rats following 12 weeks ingestion of 2% ALA or 2% DHA + 2% ALA diet in male Long Evans rats. *represents significantly different liver carbon isotope signatures between dietary groups as determined by significant Student's t-test, p < 0.05. N = 6, mean ± SEM. ALA -α-linolenic acid, 18:3n-3; ARAarachidonic acid, 20:4n-6; DHAdocosahexaenoic acid, 22:6n-3; EPAeicosapentaenoic acid, 20:5n-3; VPDB -Vienna Peedee BelemniteFig. 3Estimates of (a) percent of liver EPA derived from ALA, (b) absolute concentration of liver EPA from ALA and DHA sources, (c) percent of liver ARA derived from LNA and (d) absolute concentration of liver ARA derived from LNA and other sources. * -represents significant differences between diets for proportion of liver EPA from ALA (a) and ARA from LNA (b) as determined by Student's t-test. * -represents significant differences between diets by Student's t-test and # represents significant differences between sources of EPA and ARA by paired t-test following significant interaction effect by two-way ANOVA (c and d), p < 0.05. N = 6, mean ± SEM. c Interaction = 0.036, Diet <0.0001, Source = 0.008; d Interaction = 0.001, Diet = 0.07, Source <0.0001. ALA -α-linolenic acid, 18:3n-3; ARAarachidonic acid, 20:4n-6; DHAdocosahexaenoic acid, 22:6n-3; EPA, eicosapetaenoic acid, 20:5n-3; LNAlinoleic acid, 18:2n-6 [/fig]
[fig] Figure 4: Representation of theoretical carbon flow influencing 13 C enrichment of EPA in the ALA + DHA diet. [1] Elongation and desaturation of dietary ALA (−28.22‰), [2] retroconversion of dietary DHA (−22.30‰), [3] 2 carbons via acetyl-CoA provided from glycolysis (−12 to −16‰) for elongation of ALA, [/fig]
[table] Table 1: Dietary fatty acid composition and isotopic signatures [/table]
|
10.1371/journal.pone.0149089
|
CCBY
|
4758710
|
26890307
|
s2orc_pubmed_articles
|
Selecting Optimal Random Forest Predictive Models: A Case Study on Predicting the Spatial Distribution of Seabed Hardness
Spatially continuous predictions of seabed hardness are important baseline environmental information for sustainable management of Australia's marine jurisdiction. Seabed hardness is often inferred from multibeam backscatter data with unknown accuracy and can be inferred from underwater video footage at limited locations. In this study, we classified the seabed into four classes based on two new seabed hardness classification schemes (i.e., hard90 and hard70). We developed optimal predictive models to predict seabed hardness using random forest (RF) based on the point data of hardness classes and spatially continuous multibeam data. Five feature selection (FS) methods that are variable importance (VI), averaged variable importance (AVI), knowledge informed AVI (KIAVI), Boruta and regularized RF (RRF) were tested based on predictive accuracy. Effects of highly correlated, important and unimportant predictors on the accuracy of RF predictive models were examined. Finally, spatial predictions generated using the most accurate models were visually examined and analysed. This study confirmed that: 1) hard90 and hard70 are effective seabed hardness classification schemes; 2) seabed hardness of four classes can be predicted with a high degree of accuracy; 3) the typical approach used to pre-select predictive variables by excluding highly correlated variables needs to be re-examined; 4) the identification of the important and unimportant predictors provides useful guidelines for further improving predictive models; 5) FS methods select the most accurate predictive model(s) instead of the most parsimonious ones, and AVI and Boruta are recommended for future studies; and 6) RF is an effective modelling method with high predictive accuracy for multi-level categorical data and can be applied to 'small p and large n' problems in environmental sciences. Additionally, automated computational programs for AVI need to be developed to increase its computational efficiency and caution should be taken when applying filter FS methods in selecting predictive models.Estimation of substratum composition. The seabed and associated epibenthos(Fig 1B)were recorded along underwater video transects using a forward-facing towed-video system. Selecting Optimal Random Forest Predictive Models PLOS ONE |
# Introduction
Seabed substrate data is important baseline environmental information for supporting the sustainable management of Australia's marine jurisdiction. Seabed substrate is an important factor controlling the spatial distribution of benthic marine communities as it influences the colonisation and formation of ecological communities and the abundance of benthic organisms [bib_ref] Physical surrogates for macrofaunal distribution and abundance in a tropical gulf, Post [/bib_ref] [bib_ref] Prediction of benthic biotopes an a Norwegian offshore bank using a combination..., Mortensen [/bib_ref] [bib_ref] Animal/sediment relationships in coastal deposits of the eastern English Channel, Newell [/bib_ref] [bib_ref] The distribution of sublitoral macrofauna communities in the Bristol Channel in relation..., Warwick [/bib_ref] [bib_ref] On the use of abiotic surrogates to describe marine benthic biodiversity, Mcarthur [/bib_ref] [bib_ref] The relationship between depth, substrate and ecology: a drop video study from..., Williams [/bib_ref]. Seabed hardness is an important character of seabed substrate as it may influence the nature of attachment of an organism to the seabed [bib_ref] The relationship between depth, substrate and ecology: a drop video study from..., Williams [/bib_ref]. Hard substrates provide environments that generally support sessile suspension feeders, while soft (unconsolidated) substrates generally support discrete motile invertebrates [bib_ref] On the use of abiotic surrogates to describe marine benthic biodiversity, Mcarthur [/bib_ref]. Hence, a spatially continuous measurement of seabed hardness would be a significant aid in predicting the spatial distribution of benthic marine communities and thus to marine ecosystem management. Moreover, it can also be used for the sustainable exploitation of marine resources and planning infrastructure (e.g. selection of pipeline routes).
Despite its importance, seabed hardness data is often difficult to acquire [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. It can be directly measured at (often widely-dispersed) discrete locations, inferred from underwater video footage at discrete locations over small areas [bib_ref] Fish-habitat associations on a deep reef at the edge of the Oregon..., Stein [/bib_ref] , or inferred from multibeam backscatter data [bib_ref] Acoustic seabed classification: current practice and future directions, Anderson [/bib_ref] [bib_ref] Multi-beam backscatter measurements used to infer seabed habitats, Kloser [/bib_ref] [bib_ref] A review of shallow-water mapping systems, Basu [/bib_ref]. However, there are disadvantages associated with these methods. For example, the direct measurements are only available at point locations, and the inferred data are either only available at discrete locations over small areas or their accuracy is unknown and may be affected by many factors [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref]. Therefore, predictive modelling provides an alternative approach to generate spatially continuous data of seabed hardness [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] , where seabed hardness was classified into two classes (i.e. hard/soft binary data) according to the approach by Stein et al. [bib_ref] Fish-habitat associations on a deep reef at the edge of the Oregon..., Stein [/bib_ref] that is however difficult to use for certain substratum compositions. Moreover, all mixed classes were classified into hard class [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] , so the relevant information of the mixed classes is missing for the management of associated habitats. However, no study has been conducted on predicting seabed hardness based on four classes data yet.
Predictive variables are essential to making predictions of seabed hardness. Physical properties derived from multibeam backscatter and bathymetry have proven to be useful predictors for predicting seabed hardness [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref]. Multibeam data collection can result in hardness and roughness maps which differentiate between different substrate types of the seabed. At regional and continental scales, seabed sediments are often strongly correlated to bathymetry [bib_ref] Predicting Seabed Mud Content across the Australian Margin: Comparison of Statistical and..., Li [/bib_ref] [bib_ref] Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and..., Li [/bib_ref]. At these scales, bathymetry and its derived properties are informative predictors of seabed substrate types. Moreover, recent technological advancements in sonar equipment have significantly increased the amount of biophysical data for making spatially continuous predictions of seabed hardness.
To predict seabed hardness, one of the most important decisions to make is to identify the best modelling technique that can generate a surface to truthfully represent seabed hardness. Random forest (RF) developed by Ho [bib_ref] The random subspace method for constructing decision forests, Ho [/bib_ref] and Breiman [bib_ref] Random forests, Breiman [/bib_ref] has proven to have high predictive accuracy (PA) in data mining and many other disciplines [bib_ref] Random forests for classification in ecology, Cutler [/bib_ref] [bib_ref] Gene selection and classification of microarray data using random forest, Diaz-Uriarte [/bib_ref] [bib_ref] The performance of state-of-the-art modelling techniques depends on geographical distribution of species, Marmion [/bib_ref] [bib_ref] Random forest for gene expression based cancer classification: overlooked issues. Pattern Recognition..., Okun [/bib_ref] [bib_ref] Newer classification and regression tree techniques: bagging and random forests for ecological..., Prasad [/bib_ref]. It outperformed a number of statistical modelling techniques for spatial prediction using continuous data in the marine environmental sciences [bib_ref] Predicting Seabed Mud Content across the Australian Margin: Comparison of Statistical and..., Li [/bib_ref] [bib_ref] Application of machine learning methods to spatial interpolation of environmental variables, Li [/bib_ref] [bib_ref] Can we improve the spatial predictions of seabed sediments? A case study..., Li [/bib_ref]. Hence, it was applied to predicting the spatial distribution of seabed hardness based on two classes and again achieved high PA [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. In a recent study, seabed substrate types were predicted based on four textural classes derived from relative proportions of sediment grain size and their ratio [bib_ref] A Comparison of Supervised Classification Methods for the Prediction of Substrate Type..., Stephens [/bib_ref] ; and RF was again found to be one of the most accurate methods. However, its predictions were for grain size ranging from mud to gravel, which falls into the soft class according to Li et al. [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. In addition, a Self-Organising Map and a hierarchical clustering method were jointly applied to angular backscatter response curves and produced seabed hardness classification with multiple classes, but its accuracy was less than RF and the results were not spatially continuous [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref].
Model selection is essential for identifying an optimal predictive model and various methods have been developed [bib_ref] A review of feature selection techniques in bioinformatics, Saeys [/bib_ref] [bib_ref] Performance of feature selection methods, Dougherty [/bib_ref]. However, it is often argued that model selection is less important for RF, because: 1) RF is insensitive to un-important variables, as it selects the most important variable at each node split [bib_ref] Random forest for gene expression based cancer classification: overlooked issues. Pattern Recognition..., Okun [/bib_ref] ; 2) it is of high predictive performance even when most predictive variables are noisy [bib_ref] Gene selection and classification of microarray data using random forest, Diaz-Uriarte [/bib_ref] ; and 3) its PA depends only on the number of strong features and not on the number of noisy variables if sample sizes are large (500 to 1000) [bib_ref] Analysis of a random forest method, Biau [/bib_ref]. It was found that excluding the correlated variables may improve the PA [bib_ref] Application of machine learning methods to spatial interpolation of environmental variables, Li [/bib_ref] [bib_ref] Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine..., Li [/bib_ref]. In contrast, it was observed that including some correlated variables could improve the PA [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and..., Li [/bib_ref] [bib_ref] Predictive Modelling Using Random Forest and Its Hybrid Methods with Geostatistical Techniques..., Li [/bib_ref] [bib_ref] Irrelevant Inputs and Parameter Choices: Do They Matter to Random Forest for..., Li [/bib_ref] , suggesting that correlated variables may be able to compensate for the small number of predictors generally found in the environmental sciences. These contradictory findings demonstrate that model selection is necessary for identifying an optimal predictive model for RF.
A model selection procedure for RF was developed previously by Li et al. [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] and it selects the predictors based on the variable importance produced from RF. This is a stepwise procedure using both forward and backward selection to add or eliminate predictors, which is similar to what has been proposed in recent studies [bib_ref] Variable selection using random forest, Genuer [/bib_ref] , but uses PA to determine the selection of each predictive variable. In the environmental sciences, predictive variables are often correlated, which may affect the observed variable importance for the predictors when using RF. To deal with this, an R package 'extendedForest'was developed to compensate for the shortcomings in the existing RF package by Liaw and Wiener. Furthermore, two R functions, Boruta [bib_ref] Feature Selection with the Boruta Package, Kursa [/bib_ref] and RRF (i.e. regularized RF), were developed to automatically search for the important predictors for RF. All these studies provide fundamental tools for selecting the important predictors in this study.
In this study, we aim to select the most accurate model to predict the spatial distribution of seabed hardness based on four classes of seabed hardness. To achieve this, we: 1) introduced two new classification schemes for seabed video classification; 2) tested the effects of various predictor sets on the accuracy of RF predictive models based on video classifications and seabed biophysical variables; and 3) examined the influence of five feature selection (FS) methods on the most accurate predictive model identified. Finally, the most accurate models were used to predict the spatial distribution of seabed hardness and the predictions were visually examined and compared with the predictions of hardness in two classes [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref].
# Methods
## Study region
The study region is located in the eastern Joseph Bonaparte Gulf, northern Australian marine margin [fig_ref] Fig 1: a [/fig_ref]. Four areas (A-D) in the region were used in this study [fig_ref] Fig 1: a [/fig_ref] , which were surveyed in 2009 [bib_ref] Sol4934-Post-survey Report, Heap [/bib_ref] and 2010under the permissions of Geoscience Australia and Department of the Environment, Water Heritage and the Arts. In these surveys, high-resolution multibeam bathymetry and backscatter data and co-located underwater video transects were acquired across the four areas [fig_ref] Fig 1: a [/fig_ref]. The areas comprise a spatially complex suite of geomorphic features including shallow flat-topped banks, terraces, ridges, deep valleys and plains [fig_ref] Fig 1: a [/fig_ref].
## Schemes for deriving seabed hardness from underwater video observations
The video footage was analysed based on a 15-second window for each transect to classify substratum composition. The substratum composition was visually estimated to 5% precision [bib_ref] Distribution of deep-water gorgonian corals in relation to benthic habitat features in..., Mortensen [/bib_ref] in terms of seven size-class categories of rock, boulders, cobble, rubble, gravel, sand and mud as defined by [bib_ref] A scale of grade and class terms for clastic sediments, Wentworth [/bib_ref]. While this method is subjective and may not be as precise as the regularly used point-count method, previous studies have proven it to be highly comparable and less time-consuming [bib_ref] Distribution of deep-water gorgonian corals in relation to benthic habitat features in..., Mortensen [/bib_ref]. The definitions and representative images of these categories are provided in S1 File.
We grouped substratum composition into two categories: soft and hard materials. Anything larger than gravel (i.e. rubble, cobbles, boulders and bedrock) was classified as 'hard' material, while mud, sand and gravel were classified as 'soft' material according to Stein et al. [bib_ref] Fish-habitat associations on a deep reef at the edge of the Oregon..., Stein [/bib_ref]. For a given location, the sum of total hard and total soft cover (i.e. the percentage of 'hard' and 'soft' materials) was always 100%. Additionally, the presence of epibenthic communities provides additional information to correctly classify substratum. For instance, biota/benthic organisms (i.e. sessile organisms like sponges, hard corals and octocorals) that require hard substratum for growth [bib_ref] Physical surrogates for macrofaunal distribution and abundance in a tropical gulf, Post [/bib_ref] [bib_ref] Animal/sediment relationships in coastal deposits of the eastern English Channel, Newell [/bib_ref] [bib_ref] The distribution of sublitoral macrofauna communities in the Bristol Channel in relation..., Warwick [/bib_ref] [bib_ref] Habitat complexity and bottom fauna composition at different scales on the continental..., Buhl-Mortensen [/bib_ref] [bib_ref] Fishing disturbance and marine biodiversity: role of habitat structure in simple soft-sediment..., Thrush [/bib_ref] were found in amongst soft substratum according to the video data alone.
Seabed hardness classification. Seabed substratum is usually classified based on the video footage according to the approach by Stein et al. [bib_ref] Fish-habitat associations on a deep reef at the edge of the Oregon..., Stein [/bib_ref]. However for certain substratum compositions, it is difficult to apply. For example, for a substratum composition with 40% sand, 45% pebble and 15% gravel, it is impossible to apply this approach. Therefore, to overcome such issues we developed two new systems by modifying Stein et al.'s approach as below.
On the basis of Stein et al. [bib_ref] Fish-habitat associations on a deep reef at the edge of the Oregon..., Stein [/bib_ref] , we developed a new system to classify the seabed substrate into four categories: hard, hard-soft, soft-hard and soft. If a substratum consisted of >70% hard material, it was classed as 'hard'. If it consisted of 70% and >50% hard material, it was classed as 'hard-soft'. If it consisted of <50% and 30% hard material, it was classed as 'softhard'. And if it consisted of <30% hard material, it was classed as 'soft'. This classification system is hereinafter referred to as 'hard70' in this study. This system would produce similar results as Stein et al. [bib_ref] Fish-habitat associations on a deep reef at the edge of the Oregon..., Stein [/bib_ref] , but could avoid some issues associated with the concepts of primary and secondary substrate in Stein et al. [bib_ref] Fish-habitat associations on a deep reef at the edge of the Oregon..., Stein [/bib_ref] as discussed above. Our initial assessment showed that our samples were mostly classified as 'soft', while a small number of samples were 'hard', leaving a limited number of samples for the mixed classes when hard70 was employed. Hence, another new classification system is proposed in this study. If a substratum consisted of 90% hard material, it was classed as 'hard'. If it consisted of <90% and >50% hard material, it was classed as 'hard-soft'. If it consisted of <50% and >10% hard material, it was classed as 'soft-hard'. And if it consisted of 10% hard material, it was classed as 'soft'. This system is hereinafter referred to as 'hard90'.
In these two classification systems, a substrate consisting of 50% 'hard' and 50% 'soft' materials was assumed to be non-existent. If it indeed exists in reality, we would need to re-examine the video images and decide which one has greater cover or assign a fifth class to the system(s). However, this is not the case in this study. In total, 140 samples of seabed hardness were considered in this study. Of the 140 samples, 9 and 6 samples were recorded as hard, 11 and 14 hard-soft, 6 and 9 soft-hard and 114 and 111 soft based on hard70 and hard90 systems respectively. The resultant datasets were used to predict seabed hardness, with hardness classes based on hard90 presented in [fig_ref] Fig 1: a [/fig_ref]
## Predictive variables
Following a preliminary analysis based on data availability and the relationships with seabed hardness as discussed above and in previous studies [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Acoustic seabed classification: current practice and future directions, Anderson [/bib_ref] [bib_ref] Multi-beam backscatter measurements used to infer seabed habitats, Kloser [/bib_ref] [bib_ref] A review of shallow-water mapping systems, Basu [/bib_ref] [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref] , 41 predictive variables (i.e. features) were initially selected for this study. . Surface area (surface): the ratio of the "true" surface area and its "planar" surface area, [bib_ref] Multi-beam backscatter measurements used to infer seabed habitats, Kloser [/bib_ref]. Topographic position index (tpi): a measure of difference between a cell elevation and the average of the elevation values in the surrounding cells, Backscatter (bs10 to bs36): a diffused reflection of acoustic energy due to scattering process back to the direction from which it's been generated, measured as the ratio of the acoustic energy sent to a seabed to that returned from the seabed, normalised to incidence angles between 10°and 36°, The first two variables are the coordinates of sample locations. The next eight variables are bathymetry and its derived variables. The remaining 31 variables are backscatter and its derived variables. Acquisition and processing of multibeam bathymetry, backscatter and their derived variables, and prock have been detailed in previous studies [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref] and in relevant online metadata [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Backscatter Local Moran..., Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Bathymetry Local Moran..., Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Backscatter Variance, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Backscatter Homogeneity, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Profile Curvature, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Planar Curvature, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Topographic Position Index, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Surface Area, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Relief, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Slope, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Backscatter Grids, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Bathymetry Grids, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Probability of Seabed..., Siwabessy [/bib_ref]. All these variables were available at each grid cell to a 10 m resolution in the four study areas for generating the spatial predictions of seabed hardness and are freely available [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Backscatter Local Moran..., Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Bathymetry Local Moran..., Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Backscatter Variance, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Backscatter Homogeneity, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Profile Curvature, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Planar Curvature, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Topographic Position Index, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Surface Area, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Relief, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Slope, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Backscatter Grids, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Bathymetry Grids, Siwabessy [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Probability of Seabed..., Siwabessy [/bib_ref]. These 41 variables were also available at 140 sample locations for developing models to predict seabed hardness. The dataset for developing predictive models in this paper is provided as S2 File, where the study areas and hardness are factors and all the variables are numerical.
## Selecting predictors based on correlation analysis
There were strong correlations among some predictive variables based on Spearman's rank correlation that was used due to non-linear relationships between some variables. We removed 21 backscatter (bs) variables that were perfectly correlated with other variables or with a ρ = 0.99, which is usually called a correlation-based filter FS method [bib_ref] A review of feature selection techniques in bioinformatics, Saeys [/bib_ref] [bib_ref] On the relationship between feature selection and classification accuracy, Janecek [/bib_ref]. The selection was also according to their relation with the total hard (i.e., whether they displayed a better relationship with total hard) and their correlation coefficients with other bs variables. The bs25 should have been removed according to the above selection criteria, but was retained because it was used in a previous study [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. The Pearson's correlation (r) was also derived for the remaining bs variables. The correlations of the remaining 20 variables [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref] and seabed hardness (i.e., hard total, the percentage cover of hard materials) were presented in .
Seabed hardness (total hard) was strongly correlated with prock, bathy, and bs and its derived variables; while it was weakly correlated with northing, planar.curv, profile.curv, slope and variance . The relationships among the 20 variables were further illustrated in S1 High prock values were typically associated with "hard" substrate (S1 In contrast, low prock values were associated with "soft" substrate. While a similar pattern was observed for backscatter, bathymetry showed an opposite pattern with a negative correlation . These relationships were typically non-linear. These variables could potentially be good predictors of seabed hardness.
## Application of rf
Random forest, as briefly described in [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] , is an ensemble machine learning method that combines many individual regression or classification trees in the following way: from the original sample, many bootstrap samples and portions of predictive variables are drawn, and an unpruned regression or classification tree is fit to each bootstrap sample using the sampled variables. From the complete forest, the status of the response variable is usually predicted either as an average of the predictions of all trees for regression or as the class with the majority vote for classification [bib_ref] Random forests, Breiman [/bib_ref] [bib_ref] Bias in random forest variable importance measures: Illustrations, sources and a solution, Strobl [/bib_ref]. The R function, randomForest by Liaw and Wiener, was employed to develop a model to predict the spatial distribution of seabed hardness. The default values of mtry, ntree and nodesize are often good options [bib_ref] Gene selection and classification of microarray data using random forest, Diaz-Uriarte [/bib_ref] that were also observed in marine environmental sciences [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and..., Li [/bib_ref] , so the default values were used for these parameters.
## Feature selection
The model selection was based on a procedure developed for RF in previous studies [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Predictive Modelling Using Random Forest and Its Hybrid Methods with Geostatistical Techniques..., Li [/bib_ref] [bib_ref] Predicting the spatial distribution of seabed gravel content using random forest, spatial..., Li [/bib_ref] , which involved two steps. One step was to select predictors to form a model that is often termed as feature selection, and the other was to estimate the predictive accuracy of the model formed that is addressed in the next section. To select predictive variables, we adopted the same principle used in rfcv, a cross-validation function in the randomForest package, that is, identifying and removing the least important variables based on the importance of predictive variables.
Five FS methods were used to select predictors in this study based on all 140 samples. These methods are: 1) the variable importance (VI), 2) averaged variable importance (AVI), 3) knowledge informed AVI (KIAVI), 4) Boruta and 5) RRF. The first method (i.e., VI) was based on the procedure in a previous study [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] as detailed below and was applied to hard90 data with 20 variables. For this FS method, we initially used all 20 variables to establish the full model. We then reduced the full model by gradually removing the least important variable(s) from the previous model based on the variable importance measure by RF (see Fig A in S3 File), which resulted in 22 models (see [fig_ref] Table 3: (Continued) [/fig_ref]. Two exceptions to this are that: 1) for model 18 and 19, since bs25 and bs27 are equally important, we excluded them from model 17 respectively; and 2) for model 22, we included it because prock was the most important predictor in a previous study [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. After reaching the model with minimum number of predictors (i.e. only one predictor remained), we then identified the important predictor(s) based on the PA of the models developed. The important predictor(s) were defined as follows: if their exclusion reduced the PA of the subsequent model, they were determined to be important (i.e. variance, surface and relief). We also identified the unimportant predictor(s) (i.e. bs27) that increased the PA when they were excluded. We then repeated above procedure by adding these important predictors and removing the unimportant predictor to the most accurate model so far (i.e., model14) to develop predictive models until no further improvement in the PA could be achieved.
We then applied the second model selection method (i.e., AVI). Due to the randomness associated with the importance of predictive variables generated by RF algorithm, the least important variable(s) may change with individual iterations; meanwhile, correlated variables may also affect the order of the least important variable(s); so an R package 'extendedForest'was used and repeated 100 times to generate the average values of variable importance (Fig B in S3 File) that were used to select the predictors. This approach was applied to hard90 [fig_ref] Table 3: (Continued) [/fig_ref]. A brief summary of RF modelling process for hard90 data using various FS methods and predictive variables. 1) models 1-25 based on the VI using 20 variables; 2) models 26-29 based on the AVI using 20 variables; 3) models 30-31 based on KIAVI using 20 variables; 4) models 32-43 based on the AVI using 41 variables; and 5) models 44-45 based on the Boruta and model 46 based on the RRF using 41 variables. Model.fit is the predictive accuracy (ccr) of training samples by each RF model developed. The corresponding predictor for each number is listed in [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref] using 20 variables, which led to four models (i.e. models 26 to 29; see [fig_ref] Table 3: (Continued) [/fig_ref]. We initially selected the most importance predictors when the level was 4 or 6 according to [fig_ref] Fig 3: Correct classification rate [/fig_ref] File, then added the next important predictor(s) until no further improvement in PA was gained. We then removed the least important predictor from the most accurate model to determine if further improvement was possible. We also used KIAVI (i.e. applied AVI to a combined model that was based on two most accurate models identified via VI and AVI using 20 predictive variables) to see if we could further improve the accuracy. Since AVI can deal with correlated variables, we also applied this approach to the whole dataset (i.e. using 41 variables) [fig_ref] Fig 3: Correct classification rate [/fig_ref] by adopting the rationale for adding and removing predictor(s) as stated above.
We also applied AVI to hard70 data using 20 and 41 variables. We also applied AVI to a combined model based on the most accurate model identified for hard90 and the most accurate model based on AVI approach using 20 variables for hard70 to see if we could further improve the PA, which is also a kind of KIAVI and hereinafter referred to as the 'KIAVI'.
We then used Boruta [bib_ref] Feature Selection with the Boruta Package, Kursa [/bib_ref] to search for the important predictors for hard90 and hard70 data using all 41 variables because it is an all-relevant FS algorithm. For hard90 data, the default value (i.e., 100) was used for the maximal number of importance source runs in the final round. To resolve predictive variables left as 'Tentative', we increased the number of importance source runs to 2000 and 5000 respectively, but they selected the same variables. For hard70 data, we used the default value as well as the values of 2000 and 5000 for the maximal number of importance source runs. The selected variables were then used in the randomForest function.
Lastly, we used RRFto search the important predictors for hard90 and hard70 data using 41 variables because it is also an all-relevant FS algorithm. This is hereinafter referred to as RRF approach in this study. The selected variables were then used in the randomForest function.
## Model validation
To identify the most accurate predictive model, we need to know the PA of each model formed from the above FS methods. To achieve this, we used rf.cv that validates one model with fixed predictive variables for all iterations for a given number of predictive variables [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. This function allows variations in datasets generated by cross-validation and ensures the model select relevant predictors from a list of the fixed predictive variables. Given that the response variable is categorical, the correct classification rate (ccr) [bib_ref] A review of methods for the assessment of prediction errors in conservation..., Fielding [/bib_ref] and kappa [bib_ref] A coefficient of agreement for nominal scales, Cohen [/bib_ref] were used to measure the accuracy of the predictive model and were calculated using the built-in functions in rf.cv. To assess the predictive ability of each model, we used 10-fold cross-validation. To deal with the random error associated with each 10-fold cross validation [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Predictive Modelling Using Random Forest and Its Hybrid Methods with Geostatistical Techniques..., Li [/bib_ref] [bib_ref] Predicting the spatial distribution of seabed gravel content using random forest, spatial..., Li [/bib_ref] , the cross validation procedure was repeated 100 times. The choice of this iteration number was based on findings in previous studies [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Predictive Modelling Using Random Forest and Its Hybrid Methods with Geostatistical Techniques..., Li [/bib_ref] and that the dynamics of the predictive accuracies with iterations of relevant models in this study suggested that averaged accuracies stabilised after 20-80 iterations. The final results were based on the average of 100 iterations of the cross validation.
## Model comparison and spatial predictions
Since the data of ccr and kappa were not normally distributed based on the Shapiro-Wilk normality test, with heterogeneous variance based on Fligner-Killeen test of homogeneity of variances, or both, Mann-Whitney tests were used to compare the PA in terms of ccr and kappa between the most accurate models for both hard90 and hard70 data.
Finally, the most accurate predictive models for hard90 and hard70 data were used to predict seabed hardness at each 10m grid cell in the study areas. A portion of area A (A1) that comprises a variety of seabed geomorphic features was selected to illustrate and compare the predictions.
All relevant computing work was implemented in R 2.15.2. Relevant maps were then produced using ArcGIS (ESRI 1 ArcMap TM 10.0).
# Results
## Predictive model for hard90 data
Model selection using VI for 20 predictive variables (VI & filter). In total, 25 models were developed based on the model selection approach using variable importance of 20 variables (models 1-25 in [fig_ref] Table 3: (Continued) [/fig_ref] , [fig_ref] Fig 2: Correct classification rate [/fig_ref]. Correct classification rates gradually increased from model 1 and reached a maximum mean (i.e. 87.64%) for model 14, except that the PA of models 3, 5 and 11 slightly decreased after the removal of surface, relief and variance respectively. It then began to decrease from model 15 onwards with an abrupt increase for model 18 due to the exclusion of bs27, and reached the lowest value for model 22 that contained only one predictor. After adding variance and surface to model 14, the PA was further improved and reached the highest mean value of 88.53% for model 24 [fig_ref] Fig 2: Correct classification rate [/fig_ref]. Kappa displayed a similar pattern as ccr and reached the highest mean value of 0.6449 for model 24, with the Selecting Optimal Random Forest Predictive Models exception of that the lowest value was attained for model 21 that again contained only one predictor. It showed that bs35 was the most important predictor based on ccr while prock was the most important predictor based on kappa. Overall, model 24 was more accurate than other models in terms of both of ccr and kappa. This model contained nine predictors [fig_ref] Fig 2: Correct classification rate [/fig_ref]. In addition, most models perfectly predicted the training samples. Removing bs27 from model 24 did not improve the accuracy, so it was not presented in the results.
Model selection using AVI and KIAVI for 20 predictive variables (AVI & filter and KIAVI & filter). On the basis of the AVI using 20 variables, further four models were developed (models 26-29 in [fig_ref] Table 3: (Continued) [/fig_ref] , [fig_ref] Fig 2: Correct classification rate [/fig_ref]. Correct classification rates reached the highest mean value of 88.51% for model 27 after the inclusion of bs27 and prock. It then started to decrease after adding planar.curv (i.e. model 28) and excluding prock (i.e. model [bib_ref] Performance of feature selection methods, Dougherty [/bib_ref]. It showed that adding or excluding a further variable reduced the PA of model 27. Kappa displayed an identical pattern to ccr and reached the highest mean value of 0.6304 for model 27. Combing the two best performing models identified so far (i.e. models 24 and 27) resulted in two further models (models 30 and 31 in [fig_ref] Table 3: (Continued) [/fig_ref] that did not improve the PA in comparison with model 27.
Overall, model 27, containing eight predictors, was more accurate than other models in terms both of ccr and kappa [fig_ref] Fig 2: Correct classification rate [/fig_ref].
Model selection using AVI for 41 predictive variables. On the basis of the AVI using 41 variables, a further 12 model were developed (models 32-43 in [fig_ref] Table 3: (Continued) [/fig_ref] , [fig_ref] Fig 2: Correct classification rate [/fig_ref]. The PA increased after adding bs10, planar.curv and northing, which resulted in a highly accurate model (i.e. model [bib_ref] Irrelevant Inputs and Parameter Choices: Do They Matter to Random Forest for..., Li [/bib_ref]. After adding the next three most important predictors, the PA decreased. Further tuning to model 36 by removing the next less important predictor(s) resulted in models 37 to 39. The PA of all these models was less than that of model 35. Further tuning to model 35 by removing the least important predictor (i.e. planar curv) resulted in the most accurate model 40 (with a mane ccr of 89.78%). Model 43, which used all 41 variables, was much less accurate than all other models [fig_ref] Fig 2: Correct classification rate [/fig_ref]. Kappa displayed an identical pattern as ccr and reached the highest mean value of 0.6753 for model 40. Overall, model 40 was relatively more accurate than other models in terms both of ccr and kappa and contained 15 predictors [fig_ref] Table 3: (Continued) [/fig_ref]. In addition, model 40 could not be further improved by removing either the next less important predictor (model 41) or the highly correlated predictor (model 42).
Model selection using Boruta for 41 predictive variables. On the basis of the Boruta approach, two models were developed (models 44-45 in [fig_ref] Table 3: (Continued) [/fig_ref]. Model 44 contained 30 predictors with a mean ccr of 85.79% (ranging from 85% to 86.43%) and a mean kappa of 0.5279 (ranging from 0.4897 to 0.5616). Model 45 contained 31 predictors with a mean ccr of 85.83% (ranging from 85% to 87.14%) and a mean kappa of 0.5301 (ranging from 0.4986 to 0.5845), which was slightly more accurate than model 44.
Model selection using RRF for 41 predictive variables. On the basis of the RRF approach, a further model was developed (model 46 in [fig_ref] Table 3: (Continued) [/fig_ref]. Model 46 contained 31 predictors with a mean ccr of 85.27% (ranging from 83.57% to 87.14%) and a mean kappa of 0.5078 (ranging from 0.4511 to 0.5853).
In summary, 46 models were developed for hard90 data. [fig_ref] Table 4: Confusion matrix between the observed and predicted values of four hardness classes... [/fig_ref]. When the hard-soft and soft-hard classes were merged into the hard class, the accuracies were improved, especially for the user's accuracy for the hard class . The user's accuracy was higher than the producer's accuracy for non-soft classes [fig_ref] Table 4: Confusion matrix between the observed and predicted values of four hardness classes... [/fig_ref]. Non-soft classes, particularly hard-soft, were under-predicted while the soft class was over-predicted.
## Predictive model for hard70 data
Model selection using AVI for 20 predictive variables (AVI & filter). Twenty five models were developed based on the AVI for 20 variables (models 1-25 in [fig_ref] Table 6: A brief summary of RF modelling process for hard70 data using various... [/fig_ref] , [fig_ref] Fig 3: Correct classification rate [/fig_ref]. The first twenty models were developed by removing the least important variable based on AVI. Correct classification rates reached a local maximum for model 11. Two predictors (i.e. bs21 and bathy.moran) were identified as important variables and a few predictors were identified as unimportant variables (e.g. profile.curv, bs.moran, variance).
After further adding the important variables to model 11 and removing the unimportant predictors from subsequent models, ccr increased and reached the highest mean value of 86.27% for model 24. Kappa displayed a similar pattern as ccr and reached the highest mean value of 0.4905 for model 24. Overall, model 24 with 10 predictors was more accurate than other models. In addition, most models perfectly predicted the training samples.
Model selection using AVI and KIAVI for 41 predictive variables. On the basis of the AVI for 41 variables, a further 13 models were developed (models 26-38 in [fig_ref] Table 6: A brief summary of RF modelling process for hard70 data using various... [/fig_ref] , [fig_ref] Fig 3: Correct classification rate [/fig_ref]. The AVI of model 38 for hard70 data using 41 variables showed that easting should be excluded, while on the basis of model 24 for hard70 data it should be included. Hence no further model development was conducted for hard70 data using 41 variables. The most accurate model was model 33 with a mean ccr of 86.52%. Kappa displayed an identical pattern as ccr On the basis of model 24 for hard70 and also model 40 for hard90 (i.e. the most accurate model), we further tuned model 33 by including additional predictors that were used in these two models. This resulted in a further 11 models (models 39-49 in [fig_ref] Table 6: A brief summary of RF modelling process for hard70 data using various... [/fig_ref]. Model 39 was the most accurate model in terms of kappa (i.e. 0.4973) while model 47 is the most accurate model in terms of ccr (i.e. 86.36) [fig_ref] Table 6: A brief summary of RF modelling process for hard70 data using various... [/fig_ref] and [fig_ref] Fig 3: Correct classification rate [/fig_ref]. Overall, model 33 was relatively more accurate than other models.
Model selection using Boruta for 41 predictive variables. On the basis of the Boruta approach, three models were developed (models 50-52 [fig_ref] Table 6: A brief summary of RF modelling process for hard70 data using various... [/fig_ref]. Model 50 contained 25 predictors with a mean ccr of 87.04% (ranging from 85.71% to 87.86%) and a mean kappa of 0.5328 (ranging from 0.4836 to 0.5609). Model 51 contained 27 predictors with a mean ccr of 86.85% (ranging from 84.29% to 87.86%) and a mean kappa of 0.5318 (ranging from 0.4538 to 0.5692). Model 52 contained 29 predictors with a mean ccr of 86.99% (ranging from 84.29% to 87.86%) and a mean kappa of 0.5309 (ranging from 0.4434 to 0.5609). Model 50 was slightly more accurate than model 51 and model 52.
Model selection using RRF for 41 predictive variables. On the basis of the RRF, one model was developed (model 53 in [fig_ref] Table 6: A brief summary of RF modelling process for hard70 data using various... [/fig_ref]. Model 53 contained 31 predictors with a mean ccr Agreement of the observed and predicted values of the most accurate model. The predicted values based on the most accurate model (i.e., model 50) and the observed values matched very well for the soft class in terms of both user's accuracy and producer's accuracy; and producer's accuracy was poor for non-soft classes, especially for the hard-soft class [fig_ref] Table 7: Confusion matrix between the observed and predicted values of four hardness classes... [/fig_ref]. When the hard-soft and soft-hard classes were merged into the hard class, the accuracies were improved, especially the user's accuracy for the hard class [fig_ref] Table 8: Confusion matrix between the observed and predicted values of two hardness classes... [/fig_ref]. The user's accuracy was higher than the producer's accuracy for non-soft classes [fig_ref] Table 7: Confusion matrix between the observed and predicted values of four hardness classes... [/fig_ref]. All non-soft classes were under-predicted while the soft class was over-predicted.
## Comparison of fs methods based on the most accurate predictive models identified for hard90 and hard70 data
The accuracy of full models (i.e. model 43 for hard90, and model 26 for hard70) and the most accurate models identified based on various FS methods have been summarised in and pre-selected 20 variables for both hard90 and hard70 . The most accurate models based on various FS techniques were significantly more accurate than the models using either all 41 variables or the pre-selected 20 variables in terms of both ccr and kappa for both hard90 and hard70 . The most accurate models based on the FS methods were compared in [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref] and In terms of computing efficiency as measured by the number of models developed for identifying the most accurate model for each FS method, RRF and Boruta were much higher than AVI, VI and KIAVI.
## Comparison of spatial predictions of seabed hardness based on hard90 and hard70 data
The spatial predictions of the most accurate models for hard90 and hard70 were similar, with a match rate between corresponding hardness classes as high as 92.31% (i.e. with a [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref]. Comparison of the accuracy of the most accurate models (i.e. model 40 for hard90 and model 50 for hard70) with the most accurate models based various FS techniques, and also model 40 with model 50. The differences between these comparisons based on the Mann-Whitney tests (n = 100 for each model).
# Fs method hard90
Models p-value for ccr p-value for kappa corresponding mismatch rate of 7.69%, [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref]. Hard and soft were predicted less often, while hard-soft and soft-hard were predicted in greater number based on hard90 data than hard70. Of the mismatched predictions, about 1.3% of hard predictions for hard70 were predicted as hard-soft and soft-hard for hard90, while about 5.18% of soft predictions for hard70 were predicted as hard-soft and soft-hard for hard90, leading low match rates for certain classes between hard90 and hard70. The predictions of the most accurate models for hard90 and hard70 were illustrated in [fig_ref] Fig 5: Spatial predictions of seabed hardness for a section of area A [/fig_ref] a portion of study area (A1) to visually compare the predictions of the RF predictive models. This area was chosen as an example as it contains highly contrasting geomorphic [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref]. Confusion matrix between predictions for individual classes based on hard90 and hard70 data for all study areas and for a portion of area A (A1). features. For this particular portion, the match rate is 78.07% (i.e. with a corresponding mismatch rate of 21.93%, [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref]. Of the mismatched predictions, about 5.24% of hard predictions for hard70 were predicted as hard-soft for hard90, while about 13.83% of soft predictions for hard70 were predicted as hard-soft and soft-hard for hard90, resulting in low match rates for some classes between hard90 and hard70. Their predictions captured similar major patterns, while the some hard predictions for hard70 in the high banks were predicted as hard-soft or soft-hard for hard90 in the southern portion, and some soft predictions for hard70 on the terrace in the northeast corner were predicted as soft-hard and hard-soft for hard90. The hard substrates were found mostly on banks that were associated with the highest backscatter values [fig_ref] Fig 5: Spatial predictions of seabed hardness for a section of area A [/fig_ref]. The hard-soft and soft-hard substrates were also mostly found on banks as well as on portions of terraces. In comparison, soft substrates were mostly found on valleys that were often associated with the lowest backscatter values; and portions of terraces were also predicted as soft.
## Comparison of spatial predictions: four classes vs. two classes
The spatial predictions for hard90 and hard70 were similar with the predictions based on two hardness classes [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Methodologies for multibeam seabed hardness mapping in the Timor Sea-Hardness Prediction Grids, Li [/bib_ref]. The match rates were 92.06% and 93.18% respectively when the predictions of hard, hard-soft and soft-hard were pooled into one category (i.e. hard) for hard90 and hard70. The spatial predictions for hard90 and hard70 in area A1 were similar with the predictions based on two hardness classes [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [fig_ref] Fig 5: Spatial predictions of seabed hardness for a section of area A [/fig_ref] , with the match rates of 81.53% and 89.42% respectively when the predictions of hard, hard-soft and soft-hard were combined into a single category (i.e. hard) for hard90 and hard70.
## Discussion and conclusions predictive accuracy of seabed hardness
All the models developed produced a good to perfect fit to the data [fig_ref] Table 3: (Continued) [/fig_ref] ; and their PA is also high [fig_ref] Fig 2: Correct classification rate [/fig_ref]. There is no definite relationship between the models' fit and their PA, which is consistent with the findings in our previous study [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. The PA of the most accurate models based on various FS techniques varies from 0.4852 to 0.6753 for hard90 and from 0.4116 to 0.5328 for hard70 in terms of kappa [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref]. According to , the agreement between the predicted and the tested values is good if kappa is between 0.4 and 0.75. This demonstrates that the PA of the models developed for predicting the seabed hardness is high. The PA of the most accurate models is even more notable in terms of ccr; and it varies from 84.62% to 89.78% for hard90 and from 84.09% to 87.04% for hard70. The high PA implies that: 1) seabed substrate was properly classified based on underwater video footage, and the data produced for the response variable is of high quality; 2) the hardness classification schemes developed (i.e. hard90 and hard70) are robust; and 3) the predictors used were informative. Furthermore, the PA for each model is stable and reliable because it is an averaged PA based on 100 repetitions of 10-fold cross-validation. Hence we can confirm that:
- sample size used in this study was adequate for modelling the seabed hardness;
- RF is an effective modelling method with high PA not only for presence/absence data but also for multi-level categorical data;
- robust predictive models were developed;
- seabed hardness of four classes can be predicted with a high accuracy; and
- RF can be applied to 'small p and large n' problems in environmental sciences, with the number of predictive variables as low as only one.
This study further affirms the superior performance of RF in marine environmental sciences [bib_ref] Predicting Seabed Mud Content across the Australian Margin: Comparison of Statistical and..., Li [/bib_ref] [bib_ref] Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and..., Li [/bib_ref] [bib_ref] Application of machine learning methods to spatial interpolation of environmental variables, Li [/bib_ref] [bib_ref] Can we improve the spatial predictions of seabed sediments? A case study..., Li [/bib_ref] [bib_ref] Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine..., Li [/bib_ref]. The excellent performance of RF has been attributed to a number of features associated with RF including the ability to model non-linear relationships commonly found in the environmental sciences as discussed in previous studies [bib_ref] Application of machine learning methods to spatial interpolation of environmental variables, Li [/bib_ref] [bib_ref] Can we improve the spatial predictions of seabed sediments? A case study..., Li [/bib_ref] [bib_ref] Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine..., Li [/bib_ref].
It is apparent that non-soft classes were under-predicted while the soft class was over-predicted for both hard90 and hard70 data, which could be attributed to the unbalanced sample size of the hardness classes. The hard-soft class was under-predicted and approximately 50% of observed hard-soft samples were predicted as the soft class for both datasets, which was unexpected as it is anticipated to be more similar to hard or soft-hard classes than to the soft class. Although the limitations discussed in [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] could be potential factors resulting in such phenomenon, further studies are recommended to investigate this phenomenon. Approximately 50% of the non-soft classes were predicted as the soft class for hard70 data, which could be also attributed to the limitations discussed in [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. This finding highlights the difference between hard90 and hard70, suggesting that hard90 data may have more accurately classified the seabed substrate and should be adopted in the future studies.
When the hard-soft and soft-hard classes were merged into the hard class, the user's and producer's accuracy between the observed and the predicted values was improved but still slightly lower than the previously published findings, particularly for hard70 data [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref]. This reduction in the accuracy could be attributed to that bathymetry was an important predictor in the previous studies [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref] , but it is no longer an important predictor in this study. This change is because banks and terraces are mostly located in shallow water and are most often associated with hard substrates. However, the hard class in the previous study was split into three classes in the current study, which are located at similar water depths, thus bathymetry can no longer differentiate these classes and was not found to be an important predictor in this study.
## Predictive accuracy and correlated predictive variables
The PA changes with highly correlated predictors for models developed for hard90 and hard70 data. Their influence on the PA changes with individual predictors. The phenomenon that the inclusion of highly correlated predictors (i.e. ρ0.99 and r0.95) could improve the PA was observed for many models in this study (e.g. model 1 vs. model 43 and model 17 vs. model 19 for hard90, model 1 vs. model 26 for hard70, and those models compared in [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref]. This could be explained by the fact that these correlated predictors are informative as they have high variable importance (S3 File) and their inclusion can increase the number of informative predictors selected for each individual tree in RF, thus improving the PA. It was observed that including some correlated variables improved the PA in previous studies [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and..., Li [/bib_ref] [bib_ref] Predictive Modelling Using Random Forest and Its Hybrid Methods with Geostatistical Techniques..., Li [/bib_ref] , suggesting that correlated variables may be able to compensate for the small number of predictors in environmental sciences, and more often than not we only have correlated proxy predictive variables in environmental sciences instead of causal predictors as seen in simulation studies [bib_ref] Analysis of a random forest method, Biau [/bib_ref].
In contrast, the exclusion of some highly correlated predictors may improve the PA in some other cases. It was observed for hard70 where bs27 and bs25 were highly correlated (ρ = 0.99, r = 1.00), but the exclusion of bs27 from model 4 led to a slight increase in PA for the subsequent model (i.e. model 5) containing bs25. The exclusion of bs25 from model 6 for hard70 also resulted in slight improvement in PA for model 7 that contains bs21. This phenomenon was further evidenced in [fig_ref] Fig 2: Correct classification rate [/fig_ref] that was highly correlated with bs28 and bs33 (with ρ0.99 and r = 0.98). Similar findings were observed in previous studies [bib_ref] Application of machine learning methods to spatial interpolation of environmental variables, Li [/bib_ref] [bib_ref] Predicting Seabed Mud Content across the Australian Margin II: Performance of Machine..., Li [/bib_ref] [bib_ref] Irrelevant Inputs and Parameter Choices: Do They Matter to Random Forest for..., Li [/bib_ref].
Contradictory findings were observed in previous studies in the marine environmental sciences as well as in this study regarding correlated predictors for RF. These opposite effects imply that not all highly correlated predictors should be used even if there are of high VI or excluded, and that there are no short-cuts in identifying the optimal predictive model. The extendedForestpackage can efficiently deal with the correlated variables in terms of the variable importance, but selecting predictors that can improve PA from correlated predictive variables is a challenging task. No free lunch theorems [bib_ref] No free lunch theorems for optimization, Wolpert [/bib_ref] still apply even if the predictors are highly correlated. This finding also suggests that the typical approach used in pre-selecting predictors by excluding correlated variables (i.e. r0.95 or the inflation factor 20) needs to be re-examined for identifying predictive models using machine learning methods, at least for the application of random forest in marine environmental sciences. These applications further demonstrate that feature selection is essential for identifying an optimal predictive model for RF in marine environmental sciences [bib_ref] Predicting Seabed Sand Content across the Australian Margin Using Machine Learning and..., Li [/bib_ref] [bib_ref] Application of machine learning methods to spatial interpolation of environmental variables, Li [/bib_ref].
## Predictive accuracy and important and unimportant predictors
Some important and unimportant predictors were identified based on the VI for hard90 and the AVI for hard70. They were excluded during the initial FS process, which may obviously lead to the exclusion of some important predictors and thus result in less accurate predictive models which has also been observed in previous studies [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Predicting the spatial distribution of seabed gravel content using random forest, spatial..., Li [/bib_ref]. Adding the important predictors and removing the unimportant ones could partially solve this problem. However, the accuracy of predictive models in this study showed inconsistent response patterns to the inclusion or removal of the important or unimportant predictors. Of the three important predictors identified for hard90, two of them (i.e. variance and surface) further improved the PA after adding them back while the inclusion of the remaining one (i.e. relief) reduced the PA [fig_ref] Table 3: (Continued) [/fig_ref] and [fig_ref] Fig 2: Correct classification rate [/fig_ref]. Of the two identified important variables for hard70, the inclusion of bs21 further improved the PA while the inclusion of bathy.moran reduced the PA; and of the three unimportant variables, removing profile.curv and bs.moran further improved the PA, whereas removing variance reduced the PA [fig_ref] Table 6: A brief summary of RF modelling process for hard70 data using various... [/fig_ref] and [fig_ref] Fig 3: Correct classification rate [/fig_ref]. It is apparent that their influence on PA changes with individual predictors. It was also observed that reliance upon variable importance only can lead to suboptimal predictive model and the most accurate predictive model may be overlooked in previous studies [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Predicting the spatial distribution of seabed gravel content using random forest, spatial..., Li [/bib_ref]. These findings suggest that the predictor(s) with the least variable importance should not be excluded without further testing. The detection of the important and unimportant predictors provides signals of potential candidates or establishes prior information for further improvement of the PA. However efforts are required to select relevant predictors from them to further improve the PA.
# Feature selection methods
Five FS methods, VI, AVI, KIAVI, Boruta and RRF, were tested in terms of PA in this study. They are essentially all VI-based wrapper methods according to Janecek et al. and Saeys et al. [bib_ref] A review of feature selection techniques in bioinformatics, Saeys [/bib_ref] [bib_ref] On the relationship between feature selection and classification accuracy, Janecek [/bib_ref]. These methods were applied to the full dataset and a pre-selected sub-dataset. Since the VI was obtained from RF, these methods are also embedded techniques [bib_ref] A review of feature selection techniques in bioinformatics, Saeys [/bib_ref]. To develop the final predictive models based on VI and AVI, the contribution to the PA of relevant predictors was used to determine their inclusion or elimination in the backward or forward stepwise selection. Hence, these applications cover three different FS techniques: filter, wrapper and embedded. In this study, filter and/or wrapper FS techniques were used in combination with embedded FS techniques within RF.
It is apparent that all FS techniques used in this study improved the PA except that the filter FS method significantly reduced the PA with respect to the full models. The effects of these techniques are however different for hard90 and hard70. For hard90, the model developed by applying AVI to 41 variables is significantly more accurate than other FS techniques. For hard70, the model developed by applying Boruta to 41 variables is significantly more accurate than other FS techniques. These findings highlight that: 1) these FS techniques are data sensitive and data-specific, and 2) to obtain an optimal predictive model for a given dataset, relevant FS techniques should be tested to select the most appropriate FS technique, otherwise sub-optimal models may be produced. All the most accurate models identified in [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref] are believed to be the local optimum, which were identified under the FS methods with the limited resources used. In fact, all optimal models identified based on FS methods are believed to be local optimal. A complete search for the global optimal model(s) is time consuming and at times, impossible, especially when there are a large number of predictive variables. This is because the computational requirements have a factorial increase with the number of variables, highlighting the importance of FS methods. Although the application of KIAVI did not further improved the PA, it was found that informing knowledge to VI can further improve the PA [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Predicting the spatial distribution of seabed gravel content using random forest, spatial..., Li [/bib_ref]. Again no free lunch theorems [bib_ref] No free lunch theorems for optimization, Wolpert [/bib_ref] are still effective when using FS methods. In general, AVI and Boruta show their effectiveness in searching for the most accurate predictive models and are recommended for future studies. It may be worth testing whether applying the AVI to the variables selected by Boruta could further improve the PA in future studies. However, caution should be taken when applying the filter FS method because: 1) the number of predictive variables is usually small in environmental sciences; and 2) inclusion of highly correlated predictors (with r 0.99) could improve the PA as discussed above.
Computational demand is critical in choosing a FS method. Among the five methods, the computational time decreases from VI, AVI, KIAVI, Boruta to RRF in terms of the number of models tested to find the final model. Three advantages of VI were discussed previously [bib_ref] Predictive Modelling Using Random Forest and Its Hybrid Methods with Geostatistical Techniques..., Li [/bib_ref]. KIAVI and AVI share these advantages and AVI led to the most accurate model for hard90 in this study, but apparently their computational demand is high and automated programs need to be developed to increase their computational efficiency in the future. Moreover, these FS methods are applicable to datasets with small number of features such as those tested in this study. For datasets with large number of features (e.g. from thousands to millions), dimension reduction methods may be required to reduce the computing demand. However, caution needs to be taken when non-linear correlations exist [bib_ref] Evaluation of statistical models used for predicting plant species distributions: Role of..., Austin [/bib_ref].
As discussed in the previous study [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] , the PA is the ultimate measure for selecting the predictive model. Model selection via these FS methods is based on the PA. These FS methods will produce models that are the most accurate or optimal instead of the most parsimonious as discussed above and previously [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. The traditional model selection methods such as AIC and BIC for regression models (e.g. linear model, generalised linear model) attempt to select the most parsimonious models that are not necessarily the most accurate models, especially when proxy variables are used as predictors instead of causal variables. Since the ultimate goal of predictive modelling is to identify the most accurate model(s), these FS methods are more appropriate than AIC and BIC methods in selecting predictive model(s). The principles underpinning these FS methods can be easily applied to other machine learning methods as well as regression models. Therefore, they are recommended for selecting predictive model(s) for RF and other modelling techniques in future studies.
## Hardness classification methods and prediction maps of seabed hardness
Hard90 vs hard70. Two seabed hardness classification schemes, hard90 and hard70, were proposed in this study. In comparison with hard70, the advantages of Hard90 were presumed to be that: 1) the probabilities of 'hard' and 'soft' are increased accordingly by using a threshold of 90% or 10%, which ensured the 'hard' or 'soft' substratum classification to be composed mostly of the representative substratum (i.e. 'hard' or 'soft'); and 2) number of samples for the mixed classes of hardness are increased and samples are more evenly distributed among four classes. Thus the PA was expected to be higher for models based on hard90 than on hard70, which were confirmed by the findings in this study [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref].
Less hard and soft and more mixed classes were also anticipated to be predicted for hard90 than for hard70, which are confirmed by the findings in [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref]. Hard predictions for hard70 predicted as the hard-soft class for hard90 was less than soft predictions for hard70 predicted as the soft-hard class for hard90 (1.26% vs. 3.12%); even a portion of soft predictions for hard70 (2.06%) was predicted as hard-soft for hard90. However, the majority of the predictions are similar for hard90 and hard70, with a match rate as high as 92.31% between these two schemes [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref]. By comparing the distribution of the hardness classes in the hard70 and hard90, there were three samples that shifted from hard to hard-soft, and from soft to softhard respectively, but the resultant changes in predictions are not as even as the samples.
Four classes vs. two classes. The PA for four hardness classes in this study seems less than that for two classes in the previous study [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref]. A few factors were expected to affect the PA either positively or negatively:
1. The hard class in the previous study was further divided into three classes in this study, so the PA was expected to be reduced.
2. Since the locations of the video tracks are again different to their true locations (thus to the locations of backscatter data), we should not expect to reduce the predictive error much by using the predictive variables at video locations directly.
3. The predictive variables derived in this study are expected to be more reliable than those in previous studies [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref] , because in previous studies the variables were derived for the sample locations that are usually a few meters to a few kilometres away from the video track, while in this study all variables were generated for the location of the video locations to remove the possible effects of the inaccuracy in the variables caused by the distance. Hence an accuracy improved model is expected.
4. Bathymetry was an important predictor for hard and soft in the previous study [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] but not in the current study as discussed above in 4.1. This may further explain why the PA was slightly reduced in this study.
The spatial predictions for hard90 and hard70 were similar with the predictions based on two hardness classes [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] , with match rates as high 92.06% and 93.18% respectively. They are also as high as 81.53% and 89.42% respectively for A1. These findings show that major patterns were captured in their predictions, although predictions based on hard70 as expected are more similar to the predictions based on two classes than the predictions based on hard90. This is because both the two hardness class system in the previous study [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] and hard70 scheme in this study used the same 70% and 30% thresholds to classify hard and soft classes.
Geomorphic features. The predicted maps reflect the influence of various geomorphic features such as banks, terraces, and valleys [fig_ref] Fig 1: a [/fig_ref]. The associations of the predicted hardness with geomorphic features are generally similar to those discussed in previous studies [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref]. These associations were supported by ecological studies because certain organisms expected to be found on hard [bib_ref] Seabed Habitats and Hazards of the, Przeslawski [/bib_ref] or soft substrates [bib_ref] Seabed Habitats and Hazards of the, Przeslawski [/bib_ref] were observed in the corresponding substrates in this study.
The predictions that were illustrated using a portion of study area (A1) [fig_ref] Fig 5: Spatial predictions of seabed hardness for a section of area A [/fig_ref] highlight the difference in predictions among all geomorphic features for hard90 and hard70, with a lower match rate than that for all areas (78.07% vs 92.31%) and with more mixed classes predicted for hard90 [fig_ref] Table 1: Predictive variables and their corresponding number [/fig_ref]. Such differences were mainly observed on banks and terraces [fig_ref] Fig 1: a [/fig_ref].
## Limitations and other issues
The nature of the seabed is a fundamental factor in controlling the acoustic returns (i.e. backscatter). Soft seabed substrates generally produce lower backscatter intensity; in contrast, hard seabed substrates generally produce higher backscatter intensity [bib_ref] Acoustic seabed segmentation from direct statistical clustering of entire multibeam sonar backscatter..., Hamilton [/bib_ref] [bib_ref] Correlation of side-scan backscatter intensity with grain-size distribution of shelf sediments, Goff [/bib_ref] [bib_ref] Relationships between multibeam backscatter, sediment grain size and Posidonia oceanica seagrass distribution, De Falco [/bib_ref] [bib_ref] The effects of fine-scale surface roughness and grain size on 300 kHz..., Ferrini [/bib_ref]. However, the strength of the acoustic returns can be significantly attenuated or even lost because of scattering from surface topography and the sessile benthic organisms, causing an apparent reduction in the backscatter intensity. Likewise, very strong seabed reflections can be observed due to wellsorted unconsolidated sediments in the absence of any surface topography or the presence of hard surface underneath a veneer of soft materials. The existence of these factors can affect the reliability of acoustic data and classification of video images [bib_ref] Methodologies for seabed substrate characterisation using multibeam bathymetry, backscatter and video data:..., Siwabessy [/bib_ref] , thus affecting the PA.
In addition, the limitations discussed in relation to using video footage to validate substrate grounds and to the nature of backscatter intensity in Li et al. [bib_ref] Predicting Seabed Hardness Using Random Forest in R, Li [/bib_ref] are also applicable to this study.
# Conclusions
Two seabed hardness classification schemes proposed in this study, hard90 and hard70, are effective and they all led to high PA for seabed hardness predictions. Seabed hardness is predictable and can be predicted into a spatially continuous layer with a high accuracy, especially to large areas where multibeam acoustic data exist and predictions of seabed classes are needed for marine planning and management.
Not all highly correlated predictive variables should be used or excluded. The usual approach used in pre-selecting variables by excluding correlated ones should be re-examined for identifying predictive models using machine learning methods, at least for the application of RF in the environmental sciences.
The identification of the important and unimportant predictors provides guideline for further improving the predictive models, although additional effort is required to select relevant predictors from the important and unimportant predictors.
FS is essential for identifying an optimal predictive model for RF in environmental sciences but is a challenging task as a complete search for the global optimal model(s) is time consuming and sometimes impossible. AVI and Boruta show their effectiveness in searching for the most accurate predictive models and are recommended for future studies. Automated computational programs for AVI are recommended to be developed to improve its computational efficiency in the future. However, caution should be taken when applying filter FS method in selecting optimal predictive models.
This study further affirms the superior performance of RF in marine environmental sciences. RF is an effective modelling method with high PA not only for presence/absence data and but also for multi-level categorical data. RF can be applied to 'small p and large n' problems in environmental sciences. It is recommended for generating spatially continuous predictions of categorical variables like seabed hardness when the information of relevant predictive variables is available.
## Supporting information
[fig] Fig 1: a) Location of the study region in the eastern Joseph Bonaparte Gulf, northern Australian marine margin overlaid with bathymetry; b) location of the four study areas (A, B, C, and D) in the study region and seabed hardness types (hard, hard-soft, soft-hard and soft) based on hard90 overlaid with bathymetry at video transect; and c) the geomorphic features of the four study areas. doi:10.1371/journal.pone.0149089.g001 Selecting Optimal Random Forest Predictive Models PLOS ONE | DOI:10.1371/journal.pone.0149089 February 18, 2016 [/fig]
[fig] 38: Homogeneity of backscatter (homogeneity): a measure of closeness of the distribution of elements in the Gray-Level Co-occurrence Matrix (GLCM) to the GLCM diagonal, 39. Variance of backscatter (variance): a measure of the dispersion of the values around the mean within the GLCM, 40. Local Moran I of backscatter (bs.moran): a measure of local spatial autocorrelation in backscatter, and 41. Prock: the probability of hard substrate. [/fig]
[fig] Fig 2: Correct classification rate (%) and kappa (mean: black line; minimum and maximum: dash red lines) of 43 RF models with different predictor sets based on the averages over 100 iterations of 10-fold cross validation for seabed hardness based on hard90 data; and the model with the maximum mean ccr and mean kappa (circle). a) models 1-25 based on the VI using 20 predictive variables; b) models 26-29 based on the AVI and models 30-31 based on KIAVI using 20 variables; c) models 32-43 based on the AVI using 41 variables. doi:10.1371/journal.pone.0149089.g002 [/fig]
[fig] Fig 4: The models developed from 41 variables were more accurate than the models from the [/fig]
[fig] Fig 3: Correct classification rate (%) and kappa (mean: black line; minimum and maximum: dash red lines) of 49 RF models with different predictor sets based on the averages over 100 iterations of 10-fold cross validation for seabed hardness based on hard70 data; and the model with the maximum mean ccr and mean kappa (circle). a) models 1-25 based on the AVI using 20 predictive variables; b) models 26-38 based on the AVI using 41 variables; c) models 39-49 based on KIAVI using 41 variables. doi:10.1371/journal.pone.0149089.g003 [/fig]
[fig] Fig 5: Spatial predictions of seabed hardness for a section of area A (A1): a) hard90, b) hard70, c) hardness with two classes, and d) geomorphic features. doi:10.1371/journal.pone.0149089.g005 Selecting Optimal Random Forest Predictive Models [/fig]
[fig] S1: Fig. The relationships between seabed hardness (i.e. total of hard) and 20 predictive variables. (DOCX) File. The definitions and representative images of seven size-class categories of seabed substratum composition. (DOCX) S2 File. The dataset for developing predictive models in this paper.csv. (CSV) S3 File. Variable importance (VI) for hard90. Fig A: measured by RF using randomForest package for 20 predictive variables [1]; Fig B: averaged VI from 100 iterations of RF using extendedForest package for 20 variables [2]; and Fig C: averaged VI from 100 iterations of RF using extendedForest package for 41 variables [2]. (DOCX) [/fig]
[table] Table 1: Predictive variables and their corresponding number. [/table]
[table] Table 4: Confusion matrix between the observed and predicted values of four hardness classes based on the average of 100 times of 10-fold cross validation using the most accurate predictive model (i.e., model 40) for hard90. [/table]
[table] Table 6: A brief summary of RF modelling process for hard70 data using various FS methods and pre-Boruta with the maximal number of importance source runs of 2000, 100 and 5000, and model 53 based on the RRF using 41 variables. The model fit is the predictive accuracy (ccr) of training samples by each RF model developed. The corresponding predictor for each number is listed inTable 1. [/table]
[table] Table 7: Confusion matrix between the observed and predicted values of four hardness classes based on the average of 100 times of 10-fold cross validation using the most accurate predictive model (i.e., model 50) for hard70. [/table]
[table] Table 8: Confusion matrix between the observed and predicted values of two hardness classes based on the average of 100 times of 10-fold cross validation using the most accurate predictive model (i.e., model 50) for hard70. with the most accurate models based various FS methods. The differences between these comparisons based on the Mann-Whitney tests (n = 100 for each model). [/table]
|
10.14202/vetworld.2019.1218-1224
|
CCBY
|
6755392
|
31641300
|
s2orc_pubmed_articles
|
Chemical compositions, contaminants, and residues of organic and conventional goat milk in Bogor District, Indonesia
Aim: This study aimed to compare chemical composition and contaminants (pesticide residues, antibiotic residues, and heavy metal residues) between organic and conventional goat milk in Bogor District, West Java Province, Indonesia.Materials and Methods:Milk sampling was carried out from March to August 2018 at six goat farms. The chemical quality of milk was checked using the Lactoscan Ultrasonic Milk Analyzer device. Fatty acids were analyzed using gas chromatography (GC). Pesticide residues in goat's milk were analyzed using a GC-electron capture detector (GC-ECD). Antibiotic residues were analyzed using bioassay screening test method. The lead (Pb) and arsenic (As) residues were analyzed using the Atomic Absorption Spectrophotometer (AAS).Results:The content of fat, protein, and lactose showed that there was no difference in the composition of goat's milk between organic and conventional farms. Caprylic acid (C8:0) and capric acid (C10:0) of organic goat milk are higher than conventional goat milk. Stearic acid (C18:0) and linoleic acid (C18:2) of conventional goat milk are higher than organic goat milk. The total fatty acid of organic goat milk is higher than conventional goat milk. Organochlorine pesticide residues were not detected in organic goat milk and conventional goat milk. Tetracycline antibiotic residues were found in one sample (5.56%) of organic goat milk, and macrolides residues were found in two samples (11.11%) of conventional goat milk. Pb residue in organic goat milk is 50 ppb while conventional goat milk is 80 ppb. Residue As in organic goat milk is 70 ppb while conventional goat milk is 110 ppb.Conclusion:There was no chemical composition (fat, protein, and lactose) difference between organic and conventional goat milk. Saturated fatty acid (SFA) in organic goat milk is higher than conventional goat milk. Pesticide residues are not found in both organic and conventional goat milk. Tetracycline antibiotics were found in organic goat milk and macrolide antibiotic groups found in conventional goat milk. Pb and As residues were found in both organic goat milk and conventional goat milk.
# Introduction
Milk contains many essential nutrients, so it is recommended to be consumed regularly by children. Consumers are looking for food that can improve their health. One kind of milk that can be used as food with good nutritional value is organic milk. Organic milk has a higher selling value than milk derived from the conventional farming system because it implements high requirements regarding quality in the production and management process [bib_ref] Milk quality of organic farm, Wanniatie [/bib_ref].
Nowadays, the demand for organic milk is increasing [bib_ref] Identifying significant characteristics of organic milk consumers: A CART analysis of an..., Liu [/bib_ref] [bib_ref] Relationship between production conditions and gross milk composition in ewe's and goat's..., Malissiova [/bib_ref] assuming that consuming milk from organic farms will provide different benefits compared to consuming milk from conventional farms [bib_ref] Relationship between production conditions and gross milk composition in ewe's and goat's..., Malissiova [/bib_ref]. Milk prices derived from organic farms are higher than milk originating from conventional farms because organic milk is produced environmentally friendly from livestock that does not use antibiotics, hormones, synthetic chemicals, and without genetic modification so that it has potential benefits for human health [bib_ref] Invited review : Organic and conventionally produced milk an evaluation of factors..., Schwendel [/bib_ref]. The nutritional content of organic milk differs from conventional milk, while other reports claim that there is no difference [bib_ref] Organic foods: Health and environmental advantages and disadvantages, Forman [/bib_ref] [bib_ref] Organic production enhances milk nutritional quality by shifting fatty acid composition: A..., Benbrook [/bib_ref].
Organic goat milk consumed cannot be guaranteed to be free from various contaminants. Contaminants can come from feed or the environment around the cage. Contaminants which can contaminate milk are pesticide residues, antibiotics, and heavy metals. The presence of antimicrobial residues from a public health standpoint raises a variety of problems, including, the main potential for consumers [bib_ref] Control and prevention of antibiotic residues and contaminants in sheep and goat's..., Berruga [/bib_ref]. Food chemical contamination is an extensive topic of many exogenous chemicals that may or may not be harmful to consumers. In general, contaminants can be categorized as agrochemicals (especially residues of veterinary drugs and pesticides), environmental contaminants (especially heavy metals, persistent organic pollutants, and natural poisons), and processing of contaminants (from cooking, processing, or packaging) [bib_ref] Potential impacts of climate change on veterinary medicinal residues in livestock produce:..., Cooper [/bib_ref].
Hazardous chemicals that might contaminate milk can come from the environment, namely, pesticides, antibiotics, and heavy metals. The presence of pesticide residues, antibiotics [bib_ref] Control and prevention of antibiotic residues and contaminants in sheep and goat's..., Berruga [/bib_ref] , and heavy metals [bib_ref] Pathogenic bacteria and heavy metals toxicity assessments in evaluating unpasteurized raw milk..., Iqbal [/bib_ref] when viewed from a public health perspective, among them, is the main potential for consumer health problems. Pesticides are one of the agrochemical materials used to control pests in plants and animals. Pesticides are a concern of the community because they include dangerous chemical compounds. Excessive use and not following the rules of use can lead (Pb) to agent resistance, residues in food products, and public health disorders such as poisoning, immunosuppressive, and cancer.
Antibiotics are used on farms not only for clinical purposes but also as a growth promotor in increasing livestock production. Improper usage of antibiotics can Pb to the resistance of pathogenic bacteria and contributes to the global health crisis. The presence of antimicrobial residues in milk can cause drug hypersensitivity reactions to consumers, such as dermal reactions, asthma, or anaphylaxis [bib_ref] Assessment of antibacterial drug residues in milk for consumption in Kosovo, Rama [/bib_ref].
Heavy metals are found widely in the environment and have two primary sources, namely, human activity and geological background. Heavy metals are metallic, and metalloid chemical elements have high atomic weights and specific gravity, which can be toxic to living things. Types of heavy metals in food are arsenic (As), cadmium (Cd), mercury (Hg), tin (Sn), and Pb. Heavy metal content in milk can come from plants or water consumed by livestock [bib_ref] Pollution based study of heavy metals in medicinal plants Aloe vera and..., Iqbal [/bib_ref] ; this will cause a buildup of metals in the body and will migrate to humans who consume their products. Indonesian National Standard (SNI) number 3141.1:2011requires a maximum limit of heavy metal content in milk, namely, Pb of maximum 20 ppb, maximum Hg of 30 ppb, and maximum As of 100 ppb. Pb and As are the most dangerous heavy metals that are carcinogenic and hematopoietic disorders and can cause kidney and gastrointestinal disorders [bib_ref] Pathogenic bacteria and heavy metals toxicity assessments in evaluating unpasteurized raw milk..., Iqbal [/bib_ref].
In Indonesia, most consumers prefer to drink raw goat milk because of the belief in better taste and nutritional value and are useful as health-enhancing drugs or even disease healing agents [bib_ref] Microbiological quality of raw goat milk in Bogor, Taufik [/bib_ref]. Besides being a source of nutrition, milk can also contain dangerous chemical contamination. Research on nutrition and chemical contamination in organic goat milk has never been done in Indonesia. This research aimed to compare nutritional values and chemical contamination (pesticide residues, antibiotic residues, and heavy metal residues) between organic and conventional goat milk in Bogor District, West Java Province, Indonesia.
# Materials and methods
# Ethical approval
No live animals were used in the present study. No ethical approval was needed for the current study.
## Study area
The study was conducted in Bogor District, Indonesia, consist of three location organic farming, i.e., Ciampea (-6.5866710 S'latitude, 106.6843140 W'longitude), Caringin (-6.696547 S'latitude, 10.835356 W'longitude), and Cijeruk (-6.695693 S'latitude, 106.770186 W'longititude) and another three for conventional farming, namely, Ciampea (-6.5644079 S'latitude, 106.6951570 W'longititude), Caringin (-6.730147 S'latitude, 106.834260 W'longititude), and Cijeruk (-6.6976560 S'latitude, 106.7968700 W'longititude).
## Sampling
Milk sampling was carried out from March to August 2018. Samples were taken from six goat farms (three organic farms and three conventional farms), as many as 500 mL from each farm in Bogor District, Indonesia. The samples were put into a sterile bottle and carried using ice cubes at a temperature of 4-10°C. Each milk sample was divided into five, one for chemical analysis using the Lactoscan Ultrasonic Milk Analyzer (Milkotronic, Bulgaria) device, one for determination of fatty acid profiles by gas chromatography (GC), one for organochlorine pesticide residues with GC-electron capture detector (GC-ECD), one for antibiotic residues with bioassay screening test method, and one for metal residues weight with an Atomic Absorption Spectrophotometer (GTA 120 Graphite Tube Atomizer; 200 series AA, Agilent Technologies).
## Procedures
## Milk compositions
Examination of the chemical quality of organic milk was done using parameters such as lactose content, protein content, fat content, moisture content, and nonfat dry matter. A chemical examination was carried out using the Lactoscan Ultrasonic Milk Analyzer (Milkotronic, Bulgaria) device. Fatty acids were analyzed using GC. Before the hydrolysis and esterification process was carried out, fat extraction of milk samples was done using goldfish extraction. After esterification into fatty acid methyl ester (FAME), the sample was analyzed using GC. A standard solution of 1 mg was added to 100 mL of milk sample. Then, 0.5 N NaOH-Methanol solution of 1.5 mL was put into the milk sample and then vortexed. The N 2 gas was then exhaled to remove O 2 gas, which can oxidize FAME so that the measurement results will be negative. The mixture solution was put into a water bath with a temperature of 80-100°C for 5 min so that the saponification reaction occurs. The mixture was added with 2 mL of BF 3 -Methanol for the esterification reaction, then blown again with N 2 gas and put back into the 80-100°C water bath for 5 min. The FAME mixture obtained was separated by solvent extraction (liquid-liquid extraction) by adding hexane as much as 1.5 mL and vortexed. If there was no separation, then 3 mL of saturated NaCl was added and vortexed. Then, the hexane phase was taken carefully and added anhydrous Na 2 SO 4 to bind water. The liquid formed was then taken using a pipette and injected into the GC (7890A GC System, Agilent Technologies, USA) instrument.
## Organochlorine pesticide residue test
Analysis of pesticide residues in goat milk using GC (7890A GC System, Agilent Technologies, USA) followed the QuECheRS (Quick Easy Cheap Effective Rugged Safe) method [bib_ref] Pesticide residues in foods by acetonitrile extraction and partitioning with magnesium sulfate, Lehotay [/bib_ref]. The 15 g goat milk sample was put into a 50 mL centrifuge tube (Agilent Technologies, USA), and 15 mL of acetonitrile containing 1% acetic acid was added, and QuECheRS salt (Magnesium sulfate, Sodium Acetate, QuECheRS AOAC, Agilent Technologies, USA) and ceramic (Agilent Technologies, USA) were added and then shaken for 1 min to homogenize all parts of the material. The sample was extracted by shaking the homogenizer tube for 5 min. Extraction was continued by centrifuging for 13 min at 14°C at 4000 rpm. The extract was taken as much as 6 mL and put into a dispersive SPE tube (Agilent Technologies, USA) of 15 mL with ceramic then shaken for 2 min then left for 1 min. Then, the extract was centrifuged for 10 min at 4000 rpm. As many as 1-mL supernatants were taken into the evaporation flask to evaporate to dryness. The residue was added with 1 mL of acetone and put into a vial of 1.5 mL. As many as 10 µl of the solution was ready to be injected into the GC-ECD for the detection of pesticide residues.
## Antibiotic residue test
Antibiotic residues were analyzed using triple bio screening test method, which refers to SNI No. 2782.1998. The stages of the test consist of preparation, testing, and reading the results. Preparation included preparation of agar media, media culture, buffer solution, and standard solution. Bioassay testing was aimed at four classes of antibiotics, namely, tetracycline, macrolides, aminoglycosides, and penicillin. Media culture was used Bacillus stearothermophilus ATCC 7953, yeast extract, peptone, bacto agar, and dextrose for penicillin. Media culture was used Bacillus cereus ATCC 11778, yeast extract, beef extract, peptone, and bacto agar for tetracycline. Media culture was used Bacillus subtilis ATCC 6633, beef extract, peptone, and bacto agar for aminoglycosides. Media culture was used Kocuria rhizophila (Micrococcus luteus) ATCC 9341, yeast extract, beef extract, peptone, bacto agar, and glucose for macrolide. A total of 10 mL of sample was put in a test tube.
Meanwhile, media culture was prepared by pouring 8 mL on each Petri dish. Sterile disc paper was then placed on the surface of the media culture. Each Petri dish contained five pieces of paper discs, which consist of three pieces of disc each with 75 µL samples to be analyzed, one paper dripped 75 μL of standard antibiotic solution 0.01 IU/mL as a positive control and one more paper buffer solution phosphate as a negative control. The disc paper was placed on the surface of the media culture. Petri dishes are closed and incubated at different temperatures depending on the antibiotic group. The media culture for the tetracycline group was incubated at 30±1°C and penicillin group at 55±1°C, while the macrolides and aminoglycosides at 36±1°C, for 16-18 h, respectively. Results of the test were done by observing and measuring the diameter of the zone of resistance formed around the paper disc using the calipers. The sample was positive for antibiotics if the inhibition zone was formed ≥2 mm from the edge of the paper disc. The sample was negative if the inhibitory zone formed was 0-2 mm, because the inhibition zone formed <2 mm was considered due to the presence of natural inhibitors. The diameter of the resistance zone in the positive control was 20±1 mm, while the negative control did not form an inhibitory zone.
## Heavy metal residue test
Determination of Pb and As levels of heavy metals in milk was analyzed through the extraction process using microwave high-temperature dewatering after the process of ignition, precipitation, and dissolution using HNO [bib_ref] Relationship between production conditions and gross milk composition in ewe's and goat's..., Malissiova [/bib_ref] and H 2 O 2 as oxidizers followed by measurements using an Atomic Absorption Spectrophotometer/AAS (GTA 120 Graphite Tube Atomizer; 200 series AA, Agilent Technologies, USA) with a wavelength of 283.3 nm. The work procedure carried out to test Pb and As heavy metal residues were goat milk samples weighed 0.3-0.5 g and put in a sample tube (vessel), then added 65% HNO 3 as much as 8 mL and 30% H 2 O 2 with use a 10.0 mL pipette. The test was continued by conducting destruction according to the Microwave Digestion System (Ethos One, Milestone, Italy) program for ±2 h. The results of destruction were transferred into a 50 mL measuring flask and added distilled water thinner to the limit. The dilution results were inserted into the AAS tool to measure the absorbance. Testing of Pb and As heavy metals were done by calibrating the curve by entering the Pb (Lead standard solution, Merck KGaA, Germany) and As (Arsenic standard solution, Merck KGaA, Germany) standards into the AAS system. The Pb and As standards were entered using concentrations of 0.1 µg/g, 0.2 µg/g, 0.3 µg/g, 0.4 µg/g, 0.5 µg/g, and 0.6 µg/g. After obtaining a standard curve, then the filtrate sample was entered into the AAS system. The reading of the test results was done by integrating the results of sample absorbance with the standard heavy metal calibration curve.
Available at www.veterinaryworld.org/Vol.12/August-2019/7.pdf
# Statistical analysis
Using SPSS Statistics version 21.0 (IBM, USA) , statistical analysis was performed to explore any differences between organic and conventional milk. Mann-Whitney U-test was used for a quantitative variable, while the exact descriptive test was used for quantitative. Normality of dependent variable and residual was assessed using either Kolmogorov-Smirnov test or Shapiro-Wilk test without showing deviation from normality. The result was considered significantly different statistically (p<0.05).
# Results
The test results on the chemical and physical composition of organic and conventional goat milk are presented in1. The results of statistical tests on the testing of fat, protein, and lactose content showed no differences in the composition of goat milk between organic and conventional farms. Organic and conventional goat milk fatty acids are shown in2. Caprylic acid (C8:0) and capric acid (C10:0) of organic goat milk are higher than conventional goat milk. Stearic acid (C18:0) and linoleic acid (C18:2) of conventional goat milk are higher than organic goat milk. The total fatty acid of organic goat milk is higher than conventional goat milk.
Organochlorine pesticide residues were not detected in organic goat milk and conventional goat milk. The results of analysis based on filter test on antibiotic residues showed that there was one sample (5.56%) of organic goat milk detected containing tetracycline antibiotics and two samples (11.11%) of conventional goat milk containing macrolide antibiotics. Pb residue in organic goat milk was 50 ppb, while conventional goat milk was 80 ppb. Residue As in organic goat milk is 70 ppb while conventional goat milk is 110 ppb.
# Discussion
## Chemical composition
Statistical results on the testing of fat, protein, and lactose content showed no differences in the composition of goat milk between organic and conventional farms. The result is in agreement with the results of Malissiova et al. [bib_ref] Relationship between production conditions and gross milk composition in ewe's and goat's..., Malissiova [/bib_ref] who found that there were no differences in the composition of fat, protein, and lactose in goat milk from organic and conventional farms in Greece. Feeding management or the type of feed provided provides an opportunity for differences in fat, protein, and lactose content. The type of grass/ forage given by organic goat farms is almost the same with conventional goat farms. The only difference is that in conventional farms goat given additional concentrations of tofu pulp, tempeh pulp, and date pulp. Additional concentrates only increase the quantity of goat milk, while the fat content remains lower than organic goat milk, in contrast to research by Tsiplakou et al. [bib_ref] Differences in sheep and goats milk fatty acid profile between conventional and..., Tsiplakou [/bib_ref] and Tudisco et al. [bib_ref] Influence of organic systems on milk fatty acid profile and CLA in..., Tudisco [/bib_ref].
Saturated fatty acids (SFA), namely, caprylic acid (C8:0) and capric acid (C10:0) organic goat milk statistically show differences with conventional goat milk. Organic goat milk has higher caprylic acid and capric acid than conventional goat milk. This result is different from research by Tsiplakou et al. [bib_ref] Differences in sheep and goats milk fatty acid profile between conventional and..., Tsiplakou [/bib_ref]. Caprylic acid and capric acid in milk are of exogenous Different superscript within the same row indicate a significant difference (p<0.05). [bib_ref] Asam lemak trans (trans-C18:1) dalam susu kambing, Tasee [/bib_ref] , which are highly dependent on the availability of the acyl CoA synthetase enzyme in liver tissue and mammary tissue. Caprylic acid and capric acid are short chains of SFA, which are relatively higher in organic goat milk due to differences in feed consumption patterns [bib_ref] Differences in sheep and goats milk fatty acid profile between conventional and..., Tsiplakou [/bib_ref]. Organic goats consume grass and legumes, while conventional goats in addition to forages also consume concentrates.
Stearic acid (C18:0) of organic goat milk is lower than conventional goat milk. Stearic acid originates from exogenous which can be synthesized in tissues and organs of livestock including microbes in the rumen. Stearic acid concentration is higher in conventional goat milk because the feed given is in the form of concentrates, namely, tofu pulp and tempeh pulp. The feed used as a source of long-chain SFA is soybean meal, coconut cake, vegetable oil, olive oil, and coconut oil [bib_ref] Asam lemak trans (trans-C18:1) dalam susu kambing, Tasee [/bib_ref].
Linoleic acid (C18:2) organic goat milk is lower than conventional goat milk. Trans fatty acids in goat milk are the result of microbial activity in the rumen. As one of polyunsaturated fatty acids (PUFA) metabolites, linoleic acid (C18:2) transforms into vaccenic acid (C18:1) and lastly into stearic acid (C18:0) through bio hydrogenation [bib_ref] Dietary lipids and forages interactions on cow and goat milk fatty acid..., Chilliard [/bib_ref]. The value of monounsaturated fatty acid (MUFA) and PUFA of conventional goat milk is higher than that of organic goat milk. These results differ from the results of the study [bib_ref] Differences in sheep and goats milk fatty acid profile between conventional and..., Tsiplakou [/bib_ref] [bib_ref] Influence of organic systems on milk fatty acid profile and CLA in..., Tudisco [/bib_ref] , which showed that the amount of MUFA and PUFA of organic goat milk was higher than conventional goat milk.
## Organochlorine pesticide residues
The primary sources of organochlorine contamination in milk come from grass and feed, drinking water, soil (partially digested during grazing), and air [bib_ref] Exposure of ruminants to persistent organic pollutants and potential of decontamination, Rychen [/bib_ref]. These compounds in the animal's body are easily distributed from the digestive tract and accumulate mainly in the liver, adipose tissue, and milk. Organochlorine if consumed in low doses can endanger health resulting in hormonal disorders, reduced intelligence, fertility disorders, and cancer [bib_ref] Chemosphere assessment of health risk from organochlorine xenobiotics in goat milk for..., Witczak [/bib_ref]. Indonesia limits organochlorine residue contamination in milk, namely, lindane (10 ppb), heptachlor (6 ppb), aldrin/dieldrin (6 ppb), endosulfan (4 ppb), and dichlorodiphenyltrichloroethane (20 ppb).
## Antibiotic residue
The presence of antibiotic residues in organic and conventional goat milk is due to the use of antibiotics as control, prevention, and treatment of infections [bib_ref] Assessment of antibacterial drug residues in milk for consumption in Kosovo, Rama [/bib_ref]. The use of antibiotics in the mother of goats to treat mastitis and other diseases is a common practice in dairy goat farms. The presence of antibiotic residues is usually due to the use of antibiotics from veterinarian prescriptions and insufficient knowledge about appropriate doses, route of administration, or withdrawal time [bib_ref] Evaluation of the Charm maximum residue limit beta-lactam and tetracycline test for..., Beltrán [/bib_ref]. The presence of antibiotic residues can also cause obstacles in the processing of other dairy products and will eventually cause antibiotic resistance to pathogenic bacteria so that it will become a global health crisis [bib_ref] Evaluation of a microbiological indicator test for antibiotic detection in ewe and..., Comunian [/bib_ref] [bib_ref] Identification of antibiotic residues in raw milk samples coming from the metropolitan..., Pogurschi [/bib_ref].
Tetracycline antibiotics detected in organic goat milk are used as a treatment for mastitis. Tetracycline is used in veterinary medicine as a broad-spectrum antibiotic for the treatment of aerobic and Gramnegative anaerobic bacteria, including Actinomyces, Mycoplasma, Rickettsia, and Spirochete. Tetracyclines, including chlortetracycline, are routinely used to prevent and treat mastitis in dairy cows [bib_ref] Veterinary drug residues determination in raw milk in Croatia, Bilandzic [/bib_ref]. Macrolides are usually used in the treatment of mastitis in goats [bib_ref] Control and prevention of antibiotic residues and contaminants in sheep and goat's..., Berruga [/bib_ref]. The most commonly used macrolides are erythromycin, spiramycin, and tylosin. Tylosin is used globally as a broad spectrum of antibiotics in veterinary medicine against a variety of Gram-positive and Gram-negative anaerobic and aerobic bacteria [bib_ref] Veterinary drug residues determination in raw milk in Croatia, Bilandzic [/bib_ref].
In Indonesia, penicillin, tetracycline, macrolides, and aminoglycosides are the most common antibiotics used in the treatment of mastitis among dairy goat farms. Antibiotic residue testing was carried out using screening tests based on the SNI, namely, penicillin, tetracycline, macrolides, and aminoglycosides as a standard protocol. The next research opportunity is to detect the presence of other antibiotic residues such as lincomycin, clindamycin, and pirlimycin that are widely used in dairy goat farms in throughout the world but are still rarely used in Indonesia.
Bioassay is a screening test to identify antibiotic residues that are widely used all globally and often used in Indonesia because they are easy to use, fast, and relatively inexpensive. This test has a high sensitivity indicated by the detection limit of the minimum concentration of antibiotic residue. The detection limit of beta-lactams, tetracycline, macrolides, and aminoglycosides was 0.00125 ppm, 0.03 ppm, 0.1 ppm, and 0.1 ppm, respectively. According to the SNI, a bioassay is a standard procedure to detect antibiotic residues in the dairy product. The bioassay has the potential to determine a wide range of antibiotics within a single test. The presence of antibiotic residues in the media will inhibit the growth of bacteria. The method is applied routinely in the screening of antibiotics within the milk samples [bib_ref] Microbial screening methods for detection of antibiotic residues in slaughter animals, Pikkemat [/bib_ref]. It will be better if the authors tested the positive screening samples with further confirmation method, such as liquid chromatography-tandem mass spectrometry, but such equipment is not available in the authors' laboratory. Globally, the use of antibiotics in organic farms is not allowed. If the antibiotics are used to treat the disease, the organic status of the livestock will be lost and must wait 90 days during the organic maintenance period again. Conventional farms can use antibiotics in the process of healing the disease, but the milk to be sold or consumed must be based on withdrawal time.
## Heavy metal residue
Pb residue in organic goat milk was 50 ppb, while conventional goat milk was 80 ppb. These values are above the maximum limit set in the SNI No. 3141.1 2011, which is 20 ppb. The residual Pb of goat's milk is higher than in Pakistan 30-50 ppb [bib_ref] Comparative study of heavy metals distribution in soil, forage, blood and milk, Tahir [/bib_ref] and goat's milk in Iran (7.37 ppb) [bib_ref] Vicia faba bioassay for environmental toxicity monitoring: A review, Iqbal [/bib_ref]. The high amount of Pb residue contamination in organic and conventional goat milk is caused by pollution from the environment. The increase in the number of Pb can be due to the use of phosphate fertilizers for plants, the environment around farms and highways [bib_ref] Comparative study of heavy metals distribution in soil, forage, blood and milk, Tahir [/bib_ref].
Heavy metals have been reported to be toxic (cytogenetic) to living organisms [bib_ref] Lead and cadmium concentrations in goat, cow, sheep, and buffalo milks from..., Rahimi [/bib_ref] , which are not decomposed [bib_ref] Heavy metals in vegetables and respective soils irrigated by canal, municipal waste..., Ismail [/bib_ref] and are in the environment for a long time [bib_ref] Microbial screening methods for detection of antibiotic residues in slaughter animals, Pikkemat [/bib_ref]. Impacts on human health due to exposure to heavy metals are kidney failure, osteoporosis, lung and blood cancer, bone damage, gastrointestinal, hormonal disorders, and metabolic disorders including anemia and excretory loss of enzymes and proteins [bib_ref] Heavy metals in milk: Global prevalence and health risk assessment, Ismail [/bib_ref] [bib_ref] Essential trace and toxic element concentrations in organic and conventional milk in..., Rey-Crespo [/bib_ref]. Livestock is a food source for humans; if feed and its maintenance practices cause contamination with heavy metals, it will be harmful to humans [bib_ref] Effects of dietary cobalt deficiency on performance, blood and rumen metabolites and..., Abou-Zeina [/bib_ref] [bib_ref] Milk authentication and discrimination via metal content clustering: A case of comparing..., Zain [/bib_ref]. Heavy metals in milk usually come from milk containers, processing, and contaminated water that is used for agriculture, animal feed, and the surrounding environment [bib_ref] A review on detection of heavy metal ions in water an electrochemical..., Gumpu [/bib_ref]. Heavy metals commonly found in foods are Hg, As, Cd, and Pb [bib_ref] Trace elements in sheep and goat milk samples from Apulia and Basilicata..., Miedico [/bib_ref].
As residue in organic goat milk is 0.07 mg/kg while conventional goat milk is 110 ppb. Organic goat milk has As residues which are lower than SNI No. 3141.1 2011 (100 ppb), while conventional goat milk has higher As residues. The residue of As in organic and conventional milk is higher than that reported by Rey-crespo et al. [bib_ref] Essential trace and toxic element concentrations in organic and conventional milk in..., Rey-Crespo [/bib_ref] , namely, 1.048 ppb and 0.921 ppb, respectively, in Spain. These results are also higher than the As levels in goat milk in Italy (5 ppb) [bib_ref] Trace elements in sheep and goat milk samples from Apulia and Basilicata..., Miedico [/bib_ref] but lower than in Pakistan (403 ppb) [bib_ref] Uptake of heavy metal residues from sewerage sludge in the milk of..., Aslam [/bib_ref]. As is the most toxic metal found in the food chain and is related to cancer cases in humans [bib_ref] Veterinary drug residues determination in raw milk in Croatia, Bilandzic [/bib_ref]. The residual content of As in livestock is shown to be an indicator of As content in the soil [bib_ref] Cattle as biomonitors of soil arsenic, copper, and zinc concentrations in Galicia, Alonso [/bib_ref].
# Conclusion
The results indicated that there were no differences in fat, protein, and lactose levels between organic and conventional goat milk. The fatty acid profile produced that caprylic acid, capric acid, and the amount of SFA organic goat milk were significantly different from conventional goat milk. Pesticide residues are not found in both organic and conventional goat milk. Tetracycline antibiotics were found in organic goat milk and macrolide antibiotic groups found in conventional goat milk. Pb residue in organic goat milk and conventional goat's milk and As residue in conventional goat's milk were higher than SNI No. 3141.1. 2011.
[table] Table - 1: : Chemical and physical composition of organic and conventional goat milk. [/table]
[table] Table - 2: : Composition of fatty acid in organic and conventional goat milk (mean±SD). MUFA=C16:1+C18:1; PUFA=C18:2+C18:3; UFA=MUFA+PUFA. SFA=Saturated fatty acid, MUFA=Monounsaturated fatty acids, PUFA=Polyunsaturated fatty acids, UFA=Unsaturated fatty acid Available at www.veterinaryworld.org/Vol.12/August-2019/7.pdf origin, for example, feed and endogenous [/table]
[table] Table - 3: : Organochlorine pesticide residues in organic goat milk and conventional goat milk.Table-4:Antibiotic and heavy metal residues in organic and conventional goat milk. [/table]
|
10.1038/srep24713
|
CCBY
|
4838821
|
27097529
|
s2orc_pubmed_articles
|
Neural circuitry involved in quitting after repeated failures: role of the cingulate and temporal parietal junction
The more times people fail the more likely they are to give up, however little is known about the neural mechanisms underlying this impact of repeated failure on decision making. Here we have used a visual shape discrimination task with computer-controlled feedback combined with functional magnetic resonance imaging (fMRI) to investigate the neural circuits involved. The behavioral task confirmed that the more times subjects experienced failure the more likely they were to give up, with three successive failures being the key threshold and the majority of subjects reaching the point where they decided to quit and try a new stimulus set after three or four failures. The fMRI analysis revealed activity changes in frontal, parietal, temporal, limbic and striatal regions, especially anterior cingulate cortex (ACC), posterior cingulate cortex (PCC) and temporal parietal junction (TPJ) associated with the number of previous failures experienced. Furthermore, their parameter estimates were predictive of subjects' quitting rate. Thus, subjects reach the point where they decide to quit after three/four failures and this is associated with differential changes in brain regions involved in error monitoring and reward which regulate both failure detection and changes in decision-making strategy.
Reinforcement learning (RL) is a process whereby the estimation of state/action values is improved through trial and error so that behavior becomes more advantageous [bib_ref] Reinforcement learning: computing the temporal difference of values via distinct corticostriatal pathways, Morita [/bib_ref]. The detection and evaluation of behavioral feedback are thus of critical importance to allow adjustment of subsequent choices efficiently [bib_ref] Temporal filtering of reward signals in the dorsal anterior cingulate cortex during..., Seo [/bib_ref] [bib_ref] Optimal decision making and the anterior cingulate cortex, Kennerley [/bib_ref] [bib_ref] An approximately Bayesian delta-rule model explains the dynamics of belief updating in..., Nassar [/bib_ref] [bib_ref] Rational regulation of learning dynamics by pupil-linked arousal systems, Nassar [/bib_ref] [bib_ref] Modulation of feedback related activity in the rostral anterior cingulate cortex during..., Amiez [/bib_ref]. For example, failure (negative feedback) is common and can be used to shift from one set of stimulus-response translation rules to another [bib_ref] Neurocognitive mechanisms of cognitive control: the role of prefrontal cortex in action..., Ridderinkhof [/bib_ref]. In many situations optimal behavior requires the selection of actions based on the expected value that is derived from an individual's recent history of failures [bib_ref] Midbrain dopamine neurons encode a quantitative reward prediction error signal, Bayer [/bib_ref] [bib_ref] Dopaminergic prediction errors persevere in the nucleus accumbens core during negative reinforcement, Hollon [/bib_ref]. Thus, feedback on failure is important for making subsequent choices and plays a key role in influencing behavioral decision making. For example, learned helplessness is suggested to occur when an individual's desire to learn is destroyed because of unpleasant and uncontrollable events.
To behave adaptively, animals and humans need the capacity for reinforcement learning to internally monitor responses and to evaluate external reinforcement or feedback [bib_ref] Individual differences in reinforcement learning: behavioral, electrophysiological, and neuroimaging correlates, Santesso [/bib_ref] , and learn by trial and error to act in a manner that maximizes reward and minimizes punishment [bib_ref] Computational models of reinforcement learning: the role of dopamine as a reward..., Samson [/bib_ref] [bib_ref] Reinforcement learning can account for associative and perceptual learning on a visual-decision..., Law [/bib_ref]. This process applies to many different situations and appears to be supported by subcortical and cortical neural networks [bib_ref] Surprise signals in anterior cingulate cortex: neuronal encoding of unsigned reward prediction..., Hayden [/bib_ref] [bib_ref] Lateral habenula as a source of negative reward signals in dopamine neurons, Matsumoto [/bib_ref] [bib_ref] Functional organization of the medial frontal cortex, Rushworth [/bib_ref] [bib_ref] Neural basis of reinforcement learning and decision making, Lee [/bib_ref] [bib_ref] Reinforcement learning in the brain, Niv [/bib_ref]. Widespread neural systems are involved in encoding the value of outcomes received and responses related to monitoring or evaluation of a previously executed action. The main regions implicated include the anterior cingulate cortex (ACC), medial prefrontal cortex (MPFC), dorsolateral prefrontal cortex (DLPFC), amygdala, ventral striatum (VStr), posterior cingulate cortex (PCC) and temporal parietal junction (TPJ) [bib_ref] Decoding the neural substrates of reward-related decision making with functional MRI, Hampton [/bib_ref] [bib_ref] Right temporoparietal junction activation by a salient contextual cue facilitates target discrimination, Geng [/bib_ref] [bib_ref] Posterior cingulate cortex: adapting behavior to a changing world, Pearson [/bib_ref] [bib_ref] Neural Correlates of Reinforcement Learning and Social Preferences in Competitive Bidding, Van Den Bos [/bib_ref]. However, previous studies have focused on the effects of a single feedback given on a trial by trial basis 4,21-23 rather than on how decision making, and associated neural mechanisms, are impacted by the more common experience of repeated failure during the performance of a task.
The "three strikes" phenomenon is well known and has often been reported in research related to decision making or sports psychology [bib_ref] The rule of three: How the third event signals the emergence of..., Carlson [/bib_ref] [bib_ref] The hot hand exists in volleyball and is used for allocation decisions, Raab [/bib_ref]. This phenomenon relates to the so called "hot hand" belief in sport where a player is considered to have a higher chance of making a shot after two or three successful shots than after two or three misses (resulting in "streaks"). However, in all these previous papers it is clear that the paradigms used resulted in subjects having an increased expectation of success after each failure. The current study aimed to identify more precisely what cognitive rules operate and which neural networks are involved in monitoring performance and predicting choice in order to regulate decision making in the face of repeated failures, and where each failure is not associated with a greater expectation of subsequent success. We used a novel feedback visual discrimination paradigm [bib_ref] Role of prefrontal and anterior cingulate regions in decision-making processes shared by..., Fleck [/bib_ref] with computer-controlled feedback, where the subject could either decide to continue (try again) or abort (quit and try a new discrimination) after failing on a trial. In the study we focused on the critical number of sequential trials on which subjects failed before deciding to quit and try a new stimulus set. Importantly, the paradigm deliberately minimized the potential for modulatory influences of situations where subjects could either achieve high gains (such as in gambling situations) or had other strong motivations (such as providing solutions to problems of significant intrinsic interest) for persisting with solving a task which might have resulted in an increased and highly variable tolerance of repeated failure. We hypothesized that subjects would reach a threshold after a specific number of failures when they would be more likely to change from persisting with the current stimulus set to giving up and trying a new stimulus set. We further hypothesized that the frontal and limbic brain regions specifically involved in decision making and controlling the effects of repeated failure would be involved, and exhibit changes in activity and functional connectivity associated with behavioral performance.
# Results
Relationship between negative feedback and quitting rate. Trial-by-trial analysis revealed that whether subjects decided to continue or quit discrimination of a specific stimulus set was influenced by the number of times they received negative feedback. We used a logistic regression model to predict whether subjects continued or gave up after receiving different numbers of negative feedbacks. This simplified model used one factor (the number of negative feedbacks) and could be expressed as
[formula] = = + ⁎ ⁎ P y quit exp Am F exp AM F ( ) ( )/1 ( )(1) [/formula]
with the following functions: Logit(P) = Am*F, where F is the number of negative feedbacks, Am is the regression coefficient of negative feedback. The indices of the model were: Cox & Snell R 2 = 0.286, Nagelkerke R 2 = 0.451, accuracy of prediction = 84.2%. Based on the model, we drew a ROC curve with an area under the curve of 0.8 < S = 0.869 < 0.9, suggesting that the model diagnosis was good (S.E. = 0.008, p < 0.001, 95% confidence interval is 0.854-0.884). Using Hosmer-Lemeshow analysis to test goodness of fit we found that there was no difference between the prediction result and initial data, which also indicated that the model was good (χ 2 < 0.001, df = 3, p > 0.05). These results showed that whether to continue or to quit was indeed influenced by the number of times subjects received negative feedback. Descriptive statistics showed that for the largest proportion of trials (91.33%) the decision to quit occurred between 1 and 4 failures, with most subjects doing so after 3 or 4 failures. Only 7.04% of quitting decisions occurred after 5 failures, 1.52% after 6 and 0.11% after 7 (see [fig_ref] Figure 1: Behavioral results [/fig_ref]. In the subsequent analyses we therefore combined trials where a 4 th , 5 th , 6 th or 7 th negative feedback was received. With respect to quitting rate, a repeated-measures ANOVA revealed a main effect of the number of negative feedbacks, F [bib_ref] Neural differentiation of expected reward and risk in human subcortical structures, Preuschoff [/bib_ref] = 206, p < 0.001. Post-hoc multiple Bonferroni-corrected comparison analyses indicated that there were significant differences between the each of the different numbers of negative feedbacks received (M1 = 0.57%, SD1 = 1.26%, M2 = 3.98%, SD2 = 7.2%, M3 = 26.61%, SD3 = 20.48%, M4 = 77.44%, SD4 = 9.62%, all p < 0.001), other than between the first and second time (p = 0.170). These results showed that subjects reached the point where they started to change from continuing to quitting after the third time of receiving negative feedback. To further explore the relationship between the number of times negative feedback was received and quitting rate (percentage of quitting the item), we conducted a linear regression (y = ax + b) and a quadratic polynomial curve (y = ax 2 + bx + c) separately. The linear regression function was y = 25.32x − 36.16, R 2 = 0.74 (see [fig_ref] Figure 1: Behavioral results [/fig_ref]. The fit of the quadratic polynomial function was also good: y = 11.86x 2 − 33.96x + 23.12, R 2 = 0.9989, 95% confidence intervals (see [fig_ref] Figure 1: Behavioral results [/fig_ref]. To illustrate this finding, [fig_ref] Figure 1: Behavioral results [/fig_ref] ,e show the quadratic polynomial function for quitting rate in individual trials as a function of the number of negative feedbacks in two representative subjects. Since the quadratic polynomial function was superior to the linear regression one we used it in all subsequent analyses. Overall, these results suggested that quitting rate was predicted by the number of negative feedbacks received and reached the point where subjects started to decide to quit rather than continue after a third failure with most quitting after 4 or more failures.
An analysis of confidence levels after each failure showed that they decreased following each successive failure (see supplementary information and [fig_ref] Figure 1: Behavioral results [/fig_ref]. Importantly this confirmed that in our paradigm subjects did not experience an increased expectation of subsequent success after each failure.
Neural correlates of negative feedback. To investigate which brain regions showed activity changes associated with negative feedback in the task, we first performed a whole-brain analyses on fMRI data collected during the feedback stage. Since there were too few trials where subjects had more than 4 failures to perform an effective analysis of fMRI data we only analyzed neural changes following 1, 2, 3 and 4+ failures. This analysis revealed that a number of middle and superior frontal, temporal and occipital cortical regions and the ventral striatum showed increased activity whereas medial and dorsolateral frontal regions showed decreased activity. These neural activity changes were significantly associated with the number of times negative feedback was received (see and [fig_ref] Figure 2: Whole-brain analysis results [/fig_ref]. A number of frontal regions also showed activation differences in the contrast between positive and negative feedback and [fig_ref] Figure 2: Whole-brain analysis results [/fig_ref]. Greater activation changes during the receipt of negative feedback that were associated with the subsequent decision to quit as opposed to continue only occurred in dACC/MPFC and [fig_ref] Figure 2: Whole-brain analysis results [/fig_ref] , whereas in other frontal and temporal regions greater activation changes associated with the decision to quit rather than to continue only occurred during the actual choice stage and [fig_ref] Figure 2: Whole-brain analysis results [/fig_ref].
Neural correlates of number of negative feedbacks received and quitting rate. To explore how activity changes in the different brain regions related to the number of times negative feedback was received and quitting rate, we used the same quadratic polynomial function as for analysis of the behavioral data. Repeated-measures ANOVA was used to test whether the parameter estimates of each target region were significantly different for each of the four frequencies of negative feedback (i.e. whether the pattern of each target region was similar to that observed in the behavioral results). The results showed there were main effects of negative feedback in dACC, PCC and TPJ [F We also explored how activity changes in dACC, PCC and TPJ related to the number of times negative feedback was received and quitting rate. First, we found that parameter estimates of dACC, PCC and TPJ fitted the same quadratic polynomial function (R 2 = 0.98, R 2 = 0.93, R 2 = 0.99, p < 0.05 in all cases). We next explored the relationship between parameter estimates of dACC, PCC and TPJ and quitting rate and obtained similar results (R 2 = 0.93, R 2 = 0.94, R 2 = 0.89, p < 0.05 in all cases, [fig_ref] Figure 3: ROI results [/fig_ref].
For completeness we also investigated relationships between activity changes and number of feedbacks received and quitting rate using a linear regression. Results were broadly similar to those obtained using a quadratic polynomial function (see Supplementary information and [fig_ref] Figure 2: Whole-brain analysis results [/fig_ref].
Interactions between neural systems. Having identified dACC, PCC and TPJ as the main regions associated with the number of times negative feedback was received, we next aimed to determine whether any of their functional connections showed a similar association. We therefore performed a set of generalized form of context-dependent psychophysiological interactions (gPPI) analyses using the dACC, PCC and TPJ as seed ROIs.
We contrasted the level of functional connectivity during the feedback stage with different numbers of times negative feedback was received. The analyses showed that there was significantly increased functional connectivity after ≥ 3 failures (f3 + f4) versus 1-2 failures (f1 + f2) for dACC-MTG (− 45, 2, − 26), TPJ-hippocampus
# Discussion
Using fMRI and a novel decision task paradigm the current study has elucidated how repeated negative feedback (failure) affects people's subsequent decision-making behavior. Our findings show that the more times subjects experience failure the more likely they are to give up, with three failures being the key threshold for subjects to start quitting and the majority of them doing so after a fourth failure. The brain areas most closely associated with the number of times negative feedback was received, and predictive of whether subjects gave up, were the dACC in particular but also the PCC and TPJ.
In the current study, we found activity and functional connectivity changes in the dACC, PCC and TPJ were associated with receiving negative feedback, which is consistent with previous studies [bib_ref] Decoding the neural substrates of reward-related decision making with functional MRI, Hampton [/bib_ref] [bib_ref] Updating beliefs for a decision: neural correlates of uncertainty and underconfidence, Stern [/bib_ref]. Neural activity changes have also been described which are related to this effect of negative feedback, such as encoding the value of outcomes [bib_ref] Abstract reward and punishment representations in the human orbitofrontal cortex, O'doherty [/bib_ref] , expectation of the value of the available actions [bib_ref] Neural correlates of decision variables in parietal cortex, Platt [/bib_ref] and monitoring or evaluation of previous actions 30 . These changes have been found in diverse brain regions including the ACC, MPFC, DLPFC, amygdala, and striatum [bib_ref] Decoding the neural substrates of reward-related decision making with functional MRI, Hampton [/bib_ref]. Although behavioral decisions are likely to depend on information computed in this network of brain regions, it is not yet known which components are of specific importance for decision making which guides the selection of action [bib_ref] Decoding the neural substrates of reward-related decision making with functional MRI, Hampton [/bib_ref]. Our finding from the contrast between the decision to quit or continue that the ACC is activated during receipt of negative feedback, rather than during the actual choice stage, suggests it is particularly involved in error detection and monitoring of behavioral responses.
Our behavioral results showed that the more times subjects experienced failure the more likely they were to give up. Three failures appeared to be point at which there was a clear increase in quitting behavior, indicating that this was a key threshold. However, the greatest proportion of subjects decided to quit after four failures. The relationship between the number of times negative feedback was received and quitting rate fitted a quadratic polynomial function. Based on this we extracted the parameter estimates from regions of interest that fitted the same function. The "three strikes" phenomenon is well known and relates to the "hot hand" belief [bib_ref] The rule of three: How the third event signals the emergence of..., Carlson [/bib_ref] [bib_ref] The hot hand exists in volleyball and is used for allocation decisions, Raab [/bib_ref]. However, in all previous studies it is clear that the paradigms used resulted in subjects having increased expectation of success after each failure. In our study, we also asked subjects to report their confidence level after each round and this showed that their confidence in success decreased with successive failures and confidence level was negatively correlated with the number of negative feedbacks received (see [fig_ref] Figure 1: Behavioral results [/fig_ref]. Thus in this paradigm subjects had a progressively higher expectation of failure (not success) after each failed trial.
In our study changes in BOLD activity were only found during the feedback stage and not during the choice one and thus primarily reflect changes occurring following decision making. We also found that dACC activation could predict behavioral choice when subjects received a second and third negative feedback (see [fig_ref] Figure 2: Whole-brain analysis results [/fig_ref]. The ACC, as a component of cognitive control network (CCN), engages multiple processes, including task switching, response inhibition, error detection, response conflict and working memory [bib_ref] Frontal cortex and reward-guided learning and decision-making, Rushworth [/bib_ref] [bib_ref] Lesion mapping of cognitive control and value-based decision making in the prefrontal..., Gläscher [/bib_ref] [bib_ref] Distinct brain networks for adaptive and stable task control in humans, Dosenbach [/bib_ref] [bib_ref] Functional connectivity in the cognitive control network and the default mode network..., Alexopoulos [/bib_ref] [bib_ref] Resting-state functional MRI in depression unmasks increased connectivity between networks via the..., Sheline [/bib_ref] [bib_ref] Sleep deprivation increases dorsal nexus connectivity to the dorsolateral prefrontal cortex in..., Bosch [/bib_ref]. On the other hand, the ACC may play a central role in decision making due to its involvement in both learning and using extended action-outcome histories to optimize choice behavior [bib_ref] Optimal decision making and the anterior cingulate cortex, Kennerley [/bib_ref] [bib_ref] Probing human and monkey anterior cingulate cortex in variable environments, Walton [/bib_ref] [bib_ref] The neural basis of human error processing: reinforcement learning, dopamine, and the..., Holroyd [/bib_ref]. Reinforcement learning theory postulates that the ACC is involved in learning after obtaining an outcome that is worse than expected, i.e., a reward prediction error [bib_ref] The neural basis of human error processing: reinforcement learning, dopamine, and the..., Holroyd [/bib_ref] [bib_ref] Learned predictions of error likelihood in the anterior cingulate cortex, Brown [/bib_ref]. Importantly, in our current study activity in dACC could reliably predict whether subjects continued or gave up. The ACC has also been implicated in reward-action associations, average expected values and negative reward prediction errors [bib_ref] Optimal decision making and the anterior cingulate cortex, Kennerley [/bib_ref] [bib_ref] Anterior cingulate error-related activity is modulated by predicted reward, Amiez [/bib_ref] [bib_ref] Neuronal correlates of goal-based motor selection in the prefrontal cortex, Matsumoto [/bib_ref] [bib_ref] The anterior cingulate and error avoidance, Magno [/bib_ref] [bib_ref] Role for cingulate motor area cells in voluntary movement selection based on..., Shima [/bib_ref]. It is likely that the key role of the ACC in these processes is due in part to its position within the reward system, and also on its use of outcome information for action value adjustments and behavioral regulation [bib_ref] Behavioral shifts and action valuation in the anterior cingulate cortex, Quilodran [/bib_ref]. In humans, ACC lesions promote response slowing and variability 45 but the ability to learn from feedback is spared [bib_ref] Impairments of attention after cingulotomy, Cohen [/bib_ref] [bib_ref] Motivation of extended behaviors by anterior cingulate cortex, Holroyd [/bib_ref]. Thus, the function we can attribute to the dACC activity changes we have found may be not only to evaluate feedback but also to participate in monitoring the different steps of the task at hand to optimize action adaptation and valuation [bib_ref] Behavioral shifts and action valuation in the anterior cingulate cortex, Quilodran [/bib_ref].
The PCC is also an important region linking reward processing, attention, memory and motor control systems, and mediates the integration of variables such as reward, uncertainty, errors and option switching across multiple trials [bib_ref] Decision salience signals in posterior cingulate cortex, Heilbronner [/bib_ref]. Recent nonhuman primate work provides evidence for a more active role in the control of cognition through signaling an environmental change and the need to alter behavior [bib_ref] Posterior cingulate cortex: adapting behavior to a changing world, Pearson [/bib_ref]. In this probabilistically rewarded choice task, monkeys' behavior followed a win-stay or lose-shift heuristic. After choosing the risky option, monkeys were more likely to choose it again if they received a larger reward but switched to the safe option if they received a smaller one. Firing rates of PCC neurons were correspondingly higher following smaller rewards than large ones, and variability in responses predicted the likelihood monkeys would switch their choice on the next trial [bib_ref] Posterior cingulate cortex: adapting behavior to a changing world, Pearson [/bib_ref]. Thus the PCC, as well as dACC, is a key node responsible for environmental change detection and subsequent alterations in behavioral strategy.
The right TPJ is also activated by unexpected stimulus events of behavioral relevance [bib_ref] Right temporoparietal junction activation by a salient contextual cue facilitates target discrimination, Geng [/bib_ref]. The TPJ is involved in stimulus-driven representation of task-relevant information that can be used to engage an appropriate behavioral response [bib_ref] Right temporoparietal junction activation by a salient contextual cue facilitates target discrimination, Geng [/bib_ref] [bib_ref] Across-study and within-subject functional connectivity of a right temporo-parietal junction subregion involved..., Jakobs [/bib_ref]. Although implicated in stimulus detection, the TPJ serves critical social cognitive and regulatory roles subserving higher order cognitive processes (e.g. perspective taking, mentalizing, theory of mind) [bib_ref] A nexus model of the temporal-parietal junction, Carter [/bib_ref] [bib_ref] Neural circuitry underlying affective response to peer feedback in adolescence, Guyer [/bib_ref]. In summary therefore, dACC, PCC and TPJ are all associated with valuation, reward learning and decision-making functions.
In addition, gPPI results showed increased dACC-MTG/TPJ -Hippocampus/PCC-Hippocampus/PCC-IPL functional connectivity after three or more failures compared to after only one or two. Attention-demanding, goal-directed behavior is mediated not only by distributed patterns of cerebral activation but, remarkably, also by concurrent suppression of activity in a distinct set of brain areas, such as fronto-parietal circuits (dACC/PCC/ IPL/TPJ) [bib_ref] Across-study and within-subject functional connectivity of a right temporo-parietal junction subregion involved..., Jakobs [/bib_ref] [bib_ref] Development-related dynamics in a top-down control network, Akhrif [/bib_ref] [bib_ref] Transient suppression of broadband gamma power in the default-mode network is correlated..., Ossandón [/bib_ref]. The PCC forms strong, reciprocal connections with the medial temporal lobe, especially the parahippocampal gyrus, which is important for associative learning and episodic memory [bib_ref] Posterior cingulate cortex: adapting behavior to a changing world, Pearson [/bib_ref]. We also observed greater functional connectivity between right TPJ and a number of attentional and decision-making areas, including the right inferior frontal gyrus, MPFC and parahippocampal gyrus [bib_ref] Right temporoparietal junction activation by a salient contextual cue facilitates target discrimination, Geng [/bib_ref].
Von der Gablentz et al. [bib_ref] Performance monitoring and behavioral adaptation during task switching: An fMRI study, Von Der Gablentz [/bib_ref] used a modified Eriksen-Flanker task that required the participants to derive the correct stimulus-response association based on a feedback given after each flanker stimulus. Participants had to continuously monitor and adapt their performance as the stimulus-response contingency switched. The results showed that the switch was associated with activation of the precuneus, the cingulate cortex, the insula and brainstem regions. This brainstem system appears to interact with the cortical network and seems to be essential for performance monitoring and behavioral adaptation. One key difference is that in the reversal paradigm, stimulus-response contingencies were switched. In our study, we have used a visual shape discrimination task with computer-controlled feedback, where the subject could either decide to continue (try again) or abort (quit and try a new discrimination) after failing on a trial. We focused on the critical number of sequential trials on which subjects failed before deciding to quit and try a new stimulus set. The brain areas most closely associated with the number of times negative feedback was received, and predictive of whether subjects give up, are the dACC, PCC and TPJ, especially the dACC. It appears that in the human brain computation of both error monitoring and reward outcome to guide decision making is associated with integrated changes in frontal, parietal, temporal and limbic circuitry. While, the two tasks have in common that the same stimulus display is shown repeatedly, so that participants can also effectively "try again" in the reversal task, an important difference is that in the reversal task repeated failures are not guaranteed to occur in immediate succession.
According to principles of reinforcement learning, our results can be interpreted as a reward prediction error signal sent from posterior regions (PCC, TPJ) to more anterior circuitry (dACC) involved in behavioral adaptation, with the dACC monitoring how behavior is adjusted in the future [bib_ref] Neurocognitive mechanisms of cognitive control: the role of prefrontal cortex in action..., Ridderinkhof [/bib_ref] [bib_ref] Cortical electrophysiological network dynamics of feedback learning, Cohen [/bib_ref] [bib_ref] The neural basis of human error processing: reinforcement learning, dopamine, and the..., Holroyd [/bib_ref] [bib_ref] Motivation of extended behaviors by anterior cingulate cortex, Holroyd [/bib_ref]. However, some limitations in our current study need to be taken into account. A major source of difficulty in interpreting our findings concerns the nature of stay/switch decision making. In our paradigm making the decision to switch was easier than deciding to continue and uncertainty has a clear influence on decision making (1-2 feedbacks = my decision to stay is clear, 3-4 feedbacks = greater uncertainty). Thus it is difficult to determine from our paradigm and results what influence uncertainty may have played. In addition, the paradigm deliberately made it hard for subjects to learn anything about the stimuli in order to investigate pure negative feedback effects and exclude learning, and increasing expectations of success, as potential confounding factors. As such the paradigm may be less ecologically relevant compared with one in which a behavioral strategy can at least implicitly be learned and adapted to in accordance with feedback received.
In summary, the present study has provided new insights into both patterns of behavior and changes in associated neural circuitry during the experience of repeated failure in a visual discrimination task. Behaviorally, the more times subjects experienced failure the more likely they were to give up, with three failures being the key threshold for starting to adopt quitting and the majority of subjects actually quitting after receiving a fourth negative feedback. Changes in the activity and functional connectivity of dACC, PCC and TPJ were particularly associated with the number of previous failures and also predictive of subjects giving up. Both the pattern of behavior and that of activity changes in these three regions fitted a quadratic polynomial regression model. The current findings therefore provide new evidence for an association between the experience of repeated failures and adaptive changes in decision making. It appears that in the human brain the decision to quit after three/four failures is associated with integrated changes in frontal, parietal, temporal and limbic circuitry, which serve to compute both error monitoring and reward outcome to guide decision making.
# Methods
Participants. 20 healthy right-handed undergraduate Han Chinese subjects with no history of neurological or psychiatric disorder (M age = 22.0, SD = 2.07; 9 males) were recruited to participate in the experiment. All subjects provided written informed consent and were paid according to their performance in the experiment. The study was approved by the ethics committee of Southwest University (China) and the Institutional Human Participants Review Board of the Southwest University Imaging Center for Brain Research. The methods were carried out in accordance with approved guidelines.
Experimental procedures and stimuli. Before the fMRI session, subjects participated in 5 training trials after being given detailed instructions. In the task, subjects were asked to play a game where they were required to determine from ten stimuli comprising two different colored areas which had the largest overall area and with Scientific RepoRts | 6:24713 | DOI: 10.1038/srep24713 varying degrees of difficulty. After a decision round involving four (3 hard and 1 easy) out of the ten possible stimuli (8 hard and 2 easy) the subjects received feedback (succeed or fail) and when the feedback was negative they were given an alternative to either continue with same stimulus set again or to quit and try a new stimulus set. In this task, there are actually 112 different permutations for each stimulus set (i.e. C 1 2 *C 3 8 = 2*8*7*6/3* 2 = 2*56 = 112). The subject is therefore unable to reach the conclusion that they cannot solve the condition after only 2 or 3 failures. Subjects could continue to try each stimulus set as many times as they wanted and so could receive as many negative feedbacks as they were prepared to accept. The monetary compensation subjects received depended on the total number of points they gained from stimulus sets where they successfully passed the discrimination test (Money (RMB) = points/1000). The visual discrimination task paradigm was written in E-prime 2.0 (http://www.pstnet.com/eprime.cfm, Psychology Software Tools, USA) and the specific task flow is shown in .
One hundred and sixty pictures created by Computer Aided Design software (CAD) were used as stimuli in the perception task. Each picture (18 mm*10 mm) was subdivided into two irregular parts which were randomly colored either orange or blue and separated by a jagged line. To help avoid the subjects realizing that they were sometimes receiving false feedback and to keep them motivated by making sure the task was not too difficult, we used two different levels of difficulty ('easy' or 'hard'), although subjects were not informed about this before performing the task. Stimuli were divided into those which were 'easy' (32 pictures where subjects achieved 95% ± 5% accuracy and the area size difference was > 30%) or 'hard' (128 pictures where subjects achieved 55% ± 5% and the area difference was < 15%) based on the findings of a prior experiment. Stimuli were divided into those which were 'easy' (32 pictures where subjects achieved 95% ± 5% accuracy and the area size difference was > 30%) or 'hard' (128 pictures where subjects achieved 55% ± 5% and the area difference was < 15%) based on the findings of a prior experiment. In a pilot study, 22 healthy right-handed undergraduate Han Chinese subjects (11 females and 11 males, M age = 20.7 years old) performed the study in order to assess the difficulty of the stimuli. Each picture was presented four times, making 640 pictures in total. In addition, subjects received false feedback (succeed or fail) calculated in terms of probability using permutation and combination functions (see . For example, where the probability of a correct feedback was (C 3 4 + C 4 4 )* (1/2) 4 = 5/16 = 31% then the number of times false correct feedback was given was 31%*64 = 20 (Permutation A). Where the subjects were wrong on the first occasion, and then succeeded on the next attempt, the probability was (C 0 4 + C 1 4 + C 2 4 )*(1/2) 4 * [bib_ref] Optimal decision making and the anterior cingulate cortex, Kennerley [/bib_ref] 4 + C 4 4 )* (1/2) 4 = 11/16 *5/16 = 21%. Thus the number of times false feedback was given for this permutation was 21%*64 = 13 (Permutation B), and so on for Permutations C, D, E, F and G. Using this approach we could systematically manipulate the number of failures each subject experienced and no subjects reported being aware of receiving false feedback.
Each trial had four stages: judgment, rating, feedback and choice. At the beginning of the trial a fixation cross was displayed in the center of the screen for 1000 ms. Next, the ten pictures in a specific stimulus set were shown Figure 5. Visual shape discrimination task. Each trial in the task started with, a fixation point displayed in the center of the screen for 1000 ms. Next, ten pictures (each picture was divided into two colored areas) in a specific stimulus set, and the potential reward value, were shown for 1000 ms. Subjects were required to make judgments on which of the two colored areas was larger for four pictures, which were randomly selected from the ten pictures in the stimulus set (judgment stage). For each picture, subjects were given a maximum of 3 s to make their judgment and after the last decision was made the display screen went blank for the remainder of the 12 s. After subjects had made all decisions they gave their confidence level on their accuracy on a scale of 1 to 5 (rating stage -4s). Subjects were told that if they accurately identified which colored part was larger in at least 3/4 stimuli then they would be informed on the screen after 3s (1-5 s jittered interval) that they had passed, otherwise they would be told that they had failed (feedback stage). After a further 3s interval (1-5s jittered) if the subjects passed they were informed they had gained the points on offer (3s) and then moved on to the start of another stimulus set. If subjects failed they were informed that they had won 0 points and how many times they had failed on that specific stimulus set (number displayed on the top of the screen). Subjects were then given the choice to either try again with the same stimulus set or to quit and try a new one (choice stage).
Scientific RepoRts | 6:24713 | DOI: 10.1038/srep24713 (easy or hard discrimination difficulty) and the potential reward value (100-120 points in increments of 5) for 1000 ms. Here, the reward was just to motivate subjects to take part in the experiment seriously. The range of reward values was very small (100-120) and was varied slightly just to help increase subjects' attention. Subjects were then required to make judgments on a random selection of four out of the ten pictures in the stimulus set for which color had the largest overall area (judgment stage). In each case the subjects were given a maximum of 3 s to make their judgment and after the last decision was made the display screen went blank for the remainder of the 12 s. After subjects had made all their decisions they gave their confidence level on their accuracy on a scale of 1 (no confidence) to 5 (very high confidence) (rating stage -4s). Subjects were told that if they accurately identified which colored part was larger in at least 3/4 stimuli (i.e. 3/4 or 4/4) then they would be informed on the screen after 3s (1-5 s jittered interval) that they had passed, otherwise they would be told that they had failed (feedback stage). In fact, as described above, all feedback given was computer controlled and calculated in terms of probability, making it hard for subjects to learn anything about the picture sets from either their successes or failures. After a further 3s interval (1-5s jittered), if the subjects had passed they were informed that they had gained the points on offer (3s) and then moved on to the start of another stimulus set. If the subjects had failed they were informed that they had won 0 points and how many times they had failed on that specific stimulus set (number displayed on the top of the screen). Subjects were then given the choice to either try again with the same stimulus set or to quit and try a new one (choice stage). In all cases where decisions were made the response time to press the button was recorded. Overall the task included 160 different stimulus sets presented in a random sequence of three difficult and one easy. Since we wanted to reveal subjects' normal decision making strategies no limit was placed on the number of times they could fail and choose to continue with a specific stimulus set. Mean monetary compensation subjects received was 60 RMB. fMRI data acquisition. During each fMRI scan an Echo-Planar imaging (EPI) sequence was used for data collection, and T2*-weighted images were recorded per run using a Siemens TRIO 3.0T MRI scanner [repetition time (TR) = 2000 ms; echo time (TE) = 30 ms; flip angle = 90°; FoV = 220 × 220 mm 2 ; matrix size = 64 × 64; 32 interleaved 3 mm-thick slices; in-plane resolution = 3.4 × 3.4 mm 2 ; inter slice skip = 0.99 mm]. Additionally, anatomical images (256 × 256 × 176) with 1 × 1 × 1 mm 3 resolution were obtained by a T1-weighted three-dimensional magnetization prepared rapid gradient echo (MPRAGE) sequence (inversion time = 900 ms; TR = 1900 ms; TE = 2.52 ms; flip angle = 9°). fMRI data analyses. Statistical analyses were performed using the general linear model in SPM8 [bib_ref] Statistical parametric maps in functional imaging: a general linear approach, Friston [/bib_ref]. Slice timing was used to correct slice order, the data was realigned to estimate and modify the six parameters of head movement and the first three images were discarded to achieve magnet-steady images. These images were then normalized to MNI space in 3 × 3 × 3 mm 3 voxel sizes. The normalized data were spatially smoothed with a Gaussian kernel; the full width at half maximum (FWHM) was specified as 8 × 8 × 8 mm 3 . After pre-processing, the ten regressors from each run [i.e., judgment stage (stimulus/judgment), rating stage (rating), feedback stage (correct, 1, 2, 3, 4 or more failures -correct, f1/f2/f3/f4), choice stage (continue/quit)] were modeled to create the design matrix. The regressors were then convolved with the Canonical Hemodynamic Response Function and the six realignment parameters for each subject were also included as confounding factors.
Whole-brain analyses. Our results primarily relate to feedback processing We investigated brain regions whose activation was associated with the process of negative feedback using one-sample t-tests. Moreover, we also investigated the contrast between positive (correct and get points) and negative feedback (f1 + f2 + f3 + f4). In addition, the contrast between the decision to stay or shift after negative feedback during both the negative feedback and choice stages was conducted. The image threshold for fMRI data significance was set to, p < 0.05 FWE corrected, with clusters of ≥10 contiguous voxels. . The number of task outcome permutations. Every permutation type was included in the sequence and followed a random probability principle. "X" mean negative feedback presented. "√ " mean correct. The random possibility of Permutation type A is .
[formula] D X X X X X √ 6 C X X X √ 6 B X √ 13 A √ 20+ = = ⁎ (C C ) ( [/formula]
## Region of interest analyses (roi).
We used the MarsBar toolbox [bib_ref] Region of interest analysis using the MarsBar toolbox for SPM 99, Brett [/bib_ref] for use with SPM8 to perform ROI analyses to further characterize patterns of activation which matched behavioral decision making performance. Brain activation was extracted from 8 mm diameter spheres centered on co-ordinates identified in previous research. ROIs included the MPFC (0, 47, 21), dACC (0, [bib_ref] Role of prefrontal and anterior cingulate regions in decision-making processes shared by..., Fleck [/bib_ref] [bib_ref] Neural correlates of decision variables in parietal cortex, Platt [/bib_ref] , insula (− 36, 23, 0), amygdala (23, − 11, − 27), DLPFC [bib_ref] Learned predictions of error likelihood in the anterior cingulate cortex, Brown [/bib_ref] [bib_ref] Transient suppression of broadband gamma power in the default-mode network is correlated..., Ossandón [/bib_ref] [bib_ref] Right temporoparietal junction activation by a salient contextual cue facilitates target discrimination, Geng [/bib_ref] , VStr (− 24, 2, 7/27, − 4, 10), PCC (− 9, − 22, 43), precuneus , TPJ(66, − 49, 1) and MTG (− 51, − 13, − 14) 6,18,57 using MarsBar software. Repeated-measures ANOVA was used to test whether the parameter estimates for each of the target regions were significantly different across the four frequencies (1-4 times) of negative feedback. Quadratic polynomial function analysis was conducted to explore the relationship between ROI parameter estimates and quitting rate at different frequencies of negative feedback.
Functional connectivity analyses. Based on behavioral results, we could infer that subjects tended to continue with the same stimulus set after up to 2 failures (f1 + f2) but would quit after ≥ 3 failures (f3 + f4). Functional connectivity strengths were then analyzed to investigate changes between regions associated with behavioral choice (i.e. continue or quit). We measured functional connectivity using gPPI analysis [bib_ref] Statistical parametric maps in functional imaging: a general linear approach, Friston [/bib_ref] [bib_ref] A generalized form of context-dependent psychophysiological interactions (gPPI): a comparison to standard..., Mclaren [/bib_ref] [bib_ref] Tools of the trade: psychophysiological interactions and functional connectivity, O'reilly [/bib_ref]. For the gPPI analysis, we extracted the de-convolved time-course of each seed region (dACC, PCC, TPJ) in each subject, based on an 8 mm radius sphere centered on the peak-activation voxel from the ROI. We calculated the product of this activation time-course and the vector of the psychological variable of interest to create the psychophysiological interaction term. New SPMs were computed for each subject, including the interaction term, the physiological variable (i.e. the ROI activation time course) and the psychological variable as regressors. We then identified areas where activation was predicted by the psychophysiological interaction term, with ROI activity and the psychological regressor treated as confound variables. These analyses were carried out separately for both f1 + f2 and f3 + f4 conditions. Individual PPI SPMs were entered into a random-effects group analysis contrasting connectivity patterns between f1 + f2 and f3 + f4 conditions with a one-sample t-test, threshold at p < 0.001, small volume corrected, with a minimum cluster size of 10 voxels.
[fig] Figure 1: Behavioral results. (a) The coverage of choosing to continue. (b,c) A linear regression (b) and a quadratic polynomial (c) function between the number of negative feedbacks and quitting rate. (d,e) The relationship between the number of negative feedbacks and quitting rate using a quadratic polynomial function in two representative subjects. [/fig]
[fig] − 30 ,: − 19, − 11), PCC-hippocampus/amygdala (− 21, − 10, − 11) and PCC-inferior parietal lobule (IPL, − 57, − 43, 43/69, − 37, 28) (all small volume corrected, p < 0.001; see Fig. 4). [/fig]
[fig] Figure 2: Whole-brain analysis results. (a) Neuroanatomical location of brain regions during the feedback stage. (b,c) Brain regions exhibiting significant increased activation in the contrast between positive and negative feedback (b) and between quitting and continuing with a stimulus set (c) during the feedback stage, and the latter contrast during the choice stage (d). FWE, p < 0.05 corrected and >10 voxels. [/fig]
[fig] Figure 3: ROI results. (a-c) Brain activations were extracted from an 8 mm diameter spherical dACC, PCC, TPJ ROIs. The relationship between parameter estimates of dACC (a), PCC (b), TPJ (c) and the number of negative feedbacks or quitting rate is a quadratic polynomial function. [/fig]
[fig] Figure 4: gPPI results. (a) The dACC seed had increased functional connectivity with MTG for the f3 + f4 condition. (b) The PCC seed had increased functional connectivity with bilateral IPL and hippocampus/ amygdala for f3 + f4 condition. (c) The TPJ seed had increased connectivity with hippocampus for f3 + f4 condition (p < 0.001, small volume corrected).Scientific RepoRts | 6:24713 | DOI: 10.1038/srep24713 [/fig]
|
10.3390/ijerph18062999
|
CCBY
|
7999735
|
33803985
|
s2orc_pubmed_articles
|
Is Stress in Contact Centers Inevitable?
# Introduction
Political, social, technological, and economic changes have brought new paradigms and challenges to the organizational world, and it is crucial that organizations adapt in order to remain competitive in the market. CC have emerged in this context and may be characterized as an industry with a high rate of growth, at both international and national levels [bib_ref] Training opportunities and employee exhaustion in call centres: Mediation by psychological contract..., Chambel [/bib_ref]. CC have become the main form of contact and interaction with customers [bib_ref] Dual processes at work in a call centre: An application of the..., Bakker [/bib_ref] , whereby communication is made through a number of channels (i.e., emails, telephone calls), reducing costs for organizations and improving customer service [bib_ref] The efficiency and quality dilemma: What drives South african call centre management..., Banks [/bib_ref]. CC therefore offer various services, ranging from problem solving (and dealing with complaints) to providing additional information [bib_ref] Customer experience and profitability: An application of the empathy rating index (ERIC)..., Lywood [/bib_ref] and fostering customer loyalty [bib_ref] Service quality in call centres: Implications for customer loyalty, Dean [/bib_ref] , as there is closer proximity between the organization and its customers [bib_ref] An Empirical Study of Turnover Intentions in Call Centre Industry of Pakistan, Khan [/bib_ref] in so far as there are no geographical barriers and round the clock service is offered [bib_ref] Call centers, Holman [/bib_ref].
According to the Portuguese Association of Contact Centers (APCC), there are over 80,000 employees in Portugal. However, although CC contribute to a personalized and higher quality customer service, they are associated with high levels of employee turnover, absenteeism, stress and burnout [bib_ref] The effects of performance monitoring on emotional labor and well-being in call..., Holman [/bib_ref]. These consequences are the result of monotonous and repetitive duties, comparable to a modern form of Taylorism [bib_ref] What is typical for call centre jobs? Job characteristics and service interactions..., Zapf [/bib_ref] , where employees have little autonomy over their work and tasks [bib_ref] Differences in call centre agents' perception of their job characteristics, physical work..., Miller [/bib_ref] which are neither complex nor challenging. This simplification of work results from a structural division where, in most cases, an employee only performs front-office (answering calls) or back-office (administrative tasks) duties. The lack of rewards is compounded by low pay and a high workload which can have a negative impact on employees' stress and well-being [bib_ref] Agent Recruitment Planning in Knowledge-Intensive Call Centers, Bordoli [/bib_ref]. It should also be noted that the scripts used by these employees, with detailed instructions that structure and organize their intervention [bib_ref] Work relationships in telephone call centres: Understanding emotional exhaustion and employee withdrawal, Deery [/bib_ref] , enable a high level of control by the organization.
Nevertheless, some research has shown that Human Resources practices can be developed [bib_ref] Reducing burnout in call centers through HR practices, Castanheira [/bib_ref] [bib_ref] Worker participation in diverse settings: Does the form affect the outcome, and..., Batt [/bib_ref] [bib_ref] Beyond bureaucracy? Work organization in call centers, Frenkel [/bib_ref] which, combined with the establishment of positive relationships with the supervisor (leaders) and co-workers (peers) [bib_ref] Demands, control, and support: A meta-analytic review of work characteristics interrelationships, Luchman [/bib_ref] can mitigate these negative effects on employees' stress and well-being.
In this study, the aim is to ascertain the extent to which the different back-office and front-office duties differ both in terms of the psychosocial work environment, and the levels of stress and well-being experienced by the employees. Considering the nature of the work of front-office and back-office employees, job characteristics, social support, HR practices, work-life conflict, workplace attitudes and well-being and general well-being were compared in both duties. Additionally, this study seeks to ascertain whether these characteristics are identical across all the companies, or whether there may be differences in the tasks performed or in their management. To this end, the employees of a total sample of 15 CC companies were studied, thus making it possible to estimate the proportion of the psychosocial work environment conditions, stress and well-being attributable to the characteristics of the organization. Hence, a further aim is to clarify whether the characteristics inherent to the work of CC are inevitable or whether they depend on the duties performed (i.e., back-office and front-office) or organizational context characteristics. Since there is no theoretical development according to which it might be possible to establish potentially significant differences, this study is of an exploratory nature, seeking solely to ascertain whether there is consistency between the job, context and well-being characteristics of back and front-office employees in fifteen distinct companies.
Consequently, this study contributes to the construction of a healthier working environment, in an area characterized by constant growth.
## Theoretical framework
## Job characteristics
According to the JD-C (Job Demands-Control) model, the demands and control (i.e., autonomy) job characteristics can explain the stress and well-being of employees, and situations of stress occur when the job is characterized by high demands and low control [bib_ref] The very best of the millennium": Longitudinal research and the demand-control (support)..., De Lange [/bib_ref] [bib_ref] Job demands-control model: A summary of current issues and recommendations for future..., Karasek [/bib_ref] [bib_ref] Job Demands-Control Model: A Summary of Current Issues and Recommendations for Future..., Kain [/bib_ref].
Previous studies have shown that CC are associated with high demands related to the way employees are constantly monitored and evaluated [bib_ref] What is typical for call centre jobs? Job characteristics and service interactions..., Zapf [/bib_ref]. The evaluation of employees is mostly based on quantitative criteria, which consider several factors such as the number of calls, their duration and also the number of calls on hold. Therefore, this excessive workload, resulting from HR systems' pressure to meet the pre-established goals of the organization and the constant monitoringmake CC work highly demanding.
As for autonomy, CC employees are generally considered to have low control and to be dependent on the planning and organization of the tasks they perform [bib_ref] Call centers, Holman [/bib_ref] [bib_ref] Differences in call centre agents' perception of their job characteristics, physical work..., Miller [/bib_ref] [bib_ref] Motivation through the design of work: Test of a theory, Richard [/bib_ref] : In addition to the fact that employees do not control when or with whom they speak, they are obliged to follow scripts that organize and structure their intervention [bib_ref] Work relationships in telephone call centres: Understanding emotional exhaustion and employee withdrawal, Deery [/bib_ref].
Thus, and with recourse to the job characteristics model of Karasek [bib_ref] Job demands-control model: A summary of current issues and recommendations for future..., Karasek [/bib_ref] [bib_ref] Job Demands-Control Model: A Summary of Current Issues and Recommendations for Future..., Kain [/bib_ref] , CC are predicted to be environments characterized by high demands and low control (i.e., autonomy) [2,3,10].
## Social support
Social support has been recognized as a way of mitigating stress and reducing the negative effects of high demands and low control situations [bib_ref] Work relationships in telephone call centres: Understanding emotional exhaustion and employee withdrawal, Deery [/bib_ref] [bib_ref] Perceived organizational support as a moderator of emotional labor/outcomes relationships, Duke [/bib_ref] [bib_ref] The Relationship between Job Demands and Key Performance Indicators: Moderating Effects of..., Dwyer [/bib_ref] [bib_ref] Exploring the relationship between social support and job burnout among correctional staff, Lambert [/bib_ref] and the JD-C(S) (Job Demands-Control-Support) model identifies social support as a third dimension that can influence stress at work [bib_ref] Job demands-control model: A summary of current issues and recommendations for future..., Karasek [/bib_ref] [bib_ref] Job Demands-Control Model: A Summary of Current Issues and Recommendations for Future..., Kain [/bib_ref].
Social support occurs as a result of the fact that social relationships and interactions at work act as resources to combat job demands [bib_ref] The impact of interpersonal environment on burnout and organizational commitment, Leiter [/bib_ref] , as employees receive the information they require and develop different coping strategies that can be used in their daily lives [bib_ref] Demands, control, and support: A meta-analytic review of work characteristics interrelationships, Luchman [/bib_ref].
Since employees with access to more resources are better able to effectively respond to any demands that may arise [bib_ref] Dual processes at work in a call centre: An application of the..., Bakker [/bib_ref] [bib_ref] Moderating effects of supervisor support, monetary rewards, and career paths on the..., Choi [/bib_ref] , the CC that promote social support may be predicted to promote employees' resources to combat the demands of the task and therefore promote their experience of well-being and reduce their levels of stress.
## Hr practices
In the same vein, the HR practices adopted by the organization can also contribute to the well-being of employees [bib_ref] Worker participation in diverse settings: Does the form affect the outcome, and..., Batt [/bib_ref]. Employees' perception of job characteristics (i.e., demands, control) are influenced by the Human Resource (HR) practices in place [bib_ref] Training opportunities and employee exhaustion in call centres: Mediation by psychological contract..., Chambel [/bib_ref] which may lead to increased or decreased levels of stress. Several Human Resource Management (HRM) models do not focus on employee performance alone [bib_ref] Editorial introduction: Progressing our understanding of the mediating variables linking HRM, employee..., Boxall [/bib_ref] , such as the HIM (High Involvement Management) Model, which highlights the importance of empowering employees through power, information, knowledge and rewards, while equally ensuring their performance and well-being [bib_ref] Research and theory on high-performance work systems: Progressing the high-involvement stream, Boxall [/bib_ref].
The challenge of HRM in relation to CC lies in establishing a balance between HR control practices, geared towards standardizing the work, and HR practices compatible with the HIM model that seek to reduce employees' stress [bib_ref] High-involvement work processes, work intensification and employee well-being, Boxall [/bib_ref] [bib_ref] Employee involvement management practices, work stress, and depression in employees of a..., Mackie [/bib_ref]. Therefore, it is important to study a broad range of HR practices and analyze their impact on employees' stress and well-being [bib_ref] Employee well-being and the HRM-performance relationship: A review of quantitative studies, Van De Voorde [/bib_ref] [bib_ref] Desegregating HRM: A review and synthesis of micro and macro human resource..., Wright [/bib_ref]. To this end, the recruitment and selection process, the welcoming and integration process, training opportunities, rewards and the performance evaluation process were deemed HR measures that can directly influence CC employees' levels of stress.
## Recruitment and selection
One way to contribute to the well-being of employees is by recruiting people with a suitable profile for the job and the organization. Thus, it is important to invest in the recruitment and selection process, since it can influence the quality of the work, the interpersonal relationships of employees and also the services provided by the organization [bib_ref] A meta-analytic examination of realistic job preview effectiveness: A test of three..., Meglino [/bib_ref].
In the context of CC, and since these employees perform unchallenging, monotonous and repetitive tasks, the recruitment focus should not be on seeking highly skilled employees, but rather on identifying individuals with soft skills, thus giving priority to behavioral/social skills over technical skills [bib_ref] Ignorant theory and knowledgeable workers: Interrogating the connections between knowledge, skills and..., Thompson [/bib_ref]. On the other hand, Chapman and Webster [bib_ref] Integrating applicant reactions into the critical contact framework of recruiting, Chapman [/bib_ref] also highlight the fundamental role of recruiters in this context, since they can positively influence employees' perception of the job characteristics. In addition to the responsibility of recruiting people with a suitable profile for the job and the organization (facilitating their integration), recruitment also contributes to employees' adjustment of expectations and behaviors by clarifying their role in the organization [bib_ref] Trust and breach of the psychological contract, Robinson [/bib_ref].
## Welcoming and integration process
The welcoming and integration process is important, as it reduces the levels of stress of the new employees, thus providing them with a positive experience at the beginning of their new job [bib_ref] Three Must-Have Onboarding Elements for New and Relocated Employees, Krasman [/bib_ref] and contributing to their well-being [bib_ref] Employee commitment and well-being: A critical review, theoretical framework and research agenda, Meyer [/bib_ref]. This process fosters the construction and development of the employee-organization relationship, facilitating the sharing of information, internal communication, and team cohesion [bib_ref] Is managing the work-family interface worthwhile? Benefits for employee health and performance, Van Steenbergen [/bib_ref] , while also accelerating the new employee's adaptation, since it stimulates the acquisition of knowledge regarding the culture, values and goals of the organization.
## Training
Training promotes personal development [bib_ref] Learning in the twenty-first-century workplace, Noe [/bib_ref] which contributes to the development of employees' personal resources (i.e., self-control) [bib_ref] Psychological capital and employee performance: A latent growth modeling approach, Peterson [/bib_ref]. Thus, through a set of duly planned learning experiences, individuals acquire new knowledge and technical skills that can facilitate the execution of their tasks, which in turn reduces job demands.
In fact, training can be a strategy used by HRM to alleviate stress, as the sharing of knowledge and strategies are tools which better prepare employees to respond to job demands [bib_ref] Human resource management and performance in UK call centres, Wood [/bib_ref].
## Compensation and rewards
Several CC have adopted a variable salary component for all the employees who meet the pre-established objectives and goals. Batt [bib_ref] Managing customer services: Human resources practices, quit rates and sales growth, Batt [/bib_ref] has identified this incentive compensation as an HR practice that is compatible with the HIM model, which equally values employee performance and well-being [bib_ref] Research and theory on high-performance work systems: Progressing the high-involvement stream, Boxall [/bib_ref]. This HR practice aligns the interests of the organization with the interests of all its employees, contributing to job satisfaction and the well-being of employees [bib_ref] Job satisfaction: A Literature Review, Aziri [/bib_ref] who feel that their effort is being rewarded [bib_ref] Self-Determination theory in work organizations: The state of a science, Deci [/bib_ref].
## Performance assessment
Performance assessment consists of the continuous monitoring process of employees' behavior and performance, which enables an assessment of how efficiently they perform their duties [bib_ref] Performance appraisal, Fletcher [/bib_ref]. CC use high levels of monitoring with quantitative and qualitative criteria by which employees are assessed [bib_ref] Employee stress and health complaints in jobs with and without electronic performance..., Smith [/bib_ref] : These criteria include not only customer satisfaction, but also the number of calls made, their duration, and the number of calls on hold [bib_ref] The efficiency and quality dilemma: What drives South african call centre management..., Banks [/bib_ref]. This practice ensures standardization of the job but can have a negative impact on employees' well-being [bib_ref] The effects of performance monitoring on emotional labor and well-being in call..., Holman [/bib_ref]. In fact, according to Deery, Iverson and Walsh [bib_ref] Work relationships in telephone call centres: Understanding emotional exhaustion and employee withdrawal, Deery [/bib_ref] , high monitoring levels are for several reasons associated with increased levels of stress among employees. Firstly, stress may result from high demands which may lead to a role conflict [bib_ref] The Effect of Industrial and Internal Factors to the Firm's Performance, Isalmi [/bib_ref] , as employees are expected to establish a positive relationship with customers. However, on the other hand, they are also obliged to meet quantitative criteria (i.e., quantity and speed of calls), leading to an intensified workload [bib_ref] Call centers, Holman [/bib_ref]. Moreover, there may be additional pressure to meet the pre-established goals, since in most cases CC employees work according to an incentive compensation system [bib_ref] Ethical issues in electronic performance monitoring: A consideration of deontological and teleological..., Alder [/bib_ref]. Secondly, the high degree of monitoring reduces employees' autonomy, as they are obliged to follow scripts that structure and organize their interaction [bib_ref] An assembly line in the head': Work and employee relations in the..., Taylor [/bib_ref]. Finally, besides a heavy workload and limited autonomy, the constant monitoring to which they are subject also implies high emotional regulation on the part of these employees, as their performance is also assessed through customer satisfaction [bib_ref] Reducing burnout in call centers through HR practices, Castanheira [/bib_ref]. Employees therefore use the few resources they have to combat the additional stress they experience as a result of being observed, instead of focusing on providing a quality service [bib_ref] Stress management interventions: Improving subjective psychological well-being in the workplace, Holman [/bib_ref].
However, Grant and Higgins [bib_ref] Computerized performance monitors: Factors affecting acceptance, Grant [/bib_ref] maintain that performance assessment can have a positive impact on the well-being of employees. According to these authors, monitoring can be a means of identifying training needs, thus promoting the development of employees' new skills and knowledge, which, as previously mentioned, is associated with the reduction of stress levels.
In this regard, the effects of monitoring and its impact on employees' well-being depend on how the performance assessment data is used.
## Work-life conflict
According to the role conflict theory, an individual's resources are finite and decrease according to the roles they play. Thus, and based on the resource scarcity (i.e., time, energy) hypothesis, role conflict arises when the demands of each domain are incompatible, and the individual is obliged to choose where to apply these resources [bib_ref] The effects of work demands and resources on work-to-family conflict and facilitation, Voydanoff [/bib_ref]. As resources are finite, when individuals participate in one domain (i.e., work), this implies an investment of their resources (i.e., time and energy), and consequently their participation in other domains is compromised [bib_ref] Sources of conflict between work and family roles, Greenhaus [/bib_ref] [bib_ref] Investigating the effect of role conflict and role ambiguity on employees' job..., Soltani [/bib_ref].
The job characteristics of CC are associated with high levels of stress, impairing employees' participation in other fields.
## Attitudes and well-being at work
## Organizational commitment
Organizational commitment is a psychological state which defines the employee's level of identification with the organization and its objectives [bib_ref] The Measurement of Organizational Commitment, Monday [/bib_ref] As an attitude, organizational commitment reflects the bond which links employees to the organization for which they work [bib_ref] Organizational commitment and psychological attachment: The effects of compliance, identification, and internalization..., O'reilly [/bib_ref]. According to Meyer and Allen [bib_ref] A three-component conceptualization of organizational commitment, Meyer [/bib_ref] , this bond may be represented in different ways (i.e., affective, normative and continuity commitment), which condition the behavior of employees. Affective commitment is negatively associated with stress [bib_ref] Commitment to organizations and occupations: Extension and test of a three-component conceptualization, Meyer [/bib_ref] , as employees develop a positive emotional relationship with the organization and regard its goals as being compatible with their own [bib_ref] Job attitudes, Judge [/bib_ref] , leading to a reduction in the ambiguity of their role [bib_ref] Affective, continuance, and normative commitment to the organization: A meta-analysis of antecedents,..., Meyer [/bib_ref] which enhances their well-being.
It has been acknowledged that CC are associated with high levels of stress and therefore affective commitment may be used as a resource [bib_ref] Work-personal life conflict and burnout in contact centers: The moderating role of..., Geraldes [/bib_ref] to combat stress. The studies of Schmidt [bib_ref] Organizational commitment: A further moderator in the relationship between work stress and..., Schmidt [/bib_ref] conclude that, due to the job characteristics of CC, employees with an emotional connection to the organization display lower burnout levels than their co-workers.
## Work engagement
Work Engagement is a stable and persistent psychological state which reflects the wellbeing and motivation of employees at work [bib_ref] The measurement of engagement and burnout: A two sample confirmatory factor analytic..., Schaufeli [/bib_ref]. According to Schaufeli and Bakker [bib_ref] Job demands, job resources, and their relationship with burnout and engagement: A..., Schaufeli [/bib_ref] , it is through work engagement that employees' energy levels are expressed, reflected in their effort and persistence in the face of difficulties (vigor), their enthusiasm, pride and inspiration (dedication) and also the intrinsic pleasure and concentration associated with the performance of their duties (absorption). Thus, it may be concluded that a workforce with high engagement may constitute a competitive advantage [bib_ref] Van den Heuvel, M. Leader-member exchange, work engagement, and job performance, Breevaart [/bib_ref] , since this variable is positively associated with job satisfaction [bib_ref] Work engagement: A quantitative review and test of its relations with task..., Christian [/bib_ref] , general well-being [bib_ref] Van den Heuvel, M. Leader-member exchange, work engagement, and job performance, Breevaart [/bib_ref] and is, consequently, negatively related to stress [bib_ref] Job demands, job resources, and their relationship with burnout and engagement: A..., Schaufeli [/bib_ref].
However, engagement depends on the resources (social, physical and organizational characteristics) obtained by individuals and used in the work context [bib_ref] A cross-national study of work engagement as a mediator between job resources..., Salanova [/bib_ref]. Therefore, the organization should provide the resources required by all its employees in order to promote their intrinsic satisfaction and enhance their well-being.
## Burnout
According to Maslach and Leiter, burnout is a means of identifying stress in the workplace, reflected in the employee who has not been able to adapt to the duties/organization. This may be operationalized as a prolonged response to emotional and interpersonal stressors at work and may be analyzed through two core dimensions: Exhaustion and cynicism [bib_ref] The measurement of engagement and burnout: A two sample confirmatory factor analytic..., Schaufeli [/bib_ref]. Exhaustion refers to feelings of extreme fatigue, emotional overload, and a lack of energy and emotional resources to perform one's work. Cynicism consists of adopting negative, cold, and distant attitudes towards work.
Maslach, and Leiterhave identified a number of burnout risk factors such as excessive workload, lack of control, and low pay. Thus, low control, a high workload and low pay contribute to the onset of burnout in CCs. Moreover, burnout is negatively associated with employee satisfaction and well-being, and positively related to stress. Thus, one of the challenges faced by organizations is that of adopting measures that contribute to the reduction of burnout.
## General well-being
According to Johnson, Cooper and Cartwright [bib_ref] The experience of work-related stress across occupations, Johnson [/bib_ref] , there is a correlation between job satisfaction and the physical and psychological well-being of employees. Thus, it is important to analyze dimensions such as job characteristics, social support and the HR practices adopted by the organization in the context of CC, as these variables can explain and predict the satisfaction and general well-being of employees [bib_ref] Differences in call centre agents' perception of their job characteristics, physical work..., Miller [/bib_ref].
Considering job design and the job characteristics model [bib_ref] Motivation through the design of work: Test of a theory, Richard [/bib_ref] , it may be said that work in CC is monotonous and demanding and employees have a low level of autonomy. Therefore, low job control, high job demands, and the limited diversity of tasks have a negative impact on employees' satisfaction and are also associated with high levels of stress. Although there is little flexibility in terms of monitoring and job design in CC, several studies have pointed to a solution being found in the HR practices adopted by the organization [bib_ref] Call centers, Holman [/bib_ref] and the promotion of social support, as both these features can mitigate the effects of stressors. Such is also the case with organizations that implement measures to foster a work-life balance, as they increase employee satisfaction and, consequently, contribute to their general well-being [bib_ref] When work and family are allies: A theory of work-family enrichment, Greenhaus [/bib_ref].
# Method
## Procedure and sample
The data collection for this study was carried out as part of a research project conducted within the scope of a partnership with the Portuguese Association of Contact Centers (APCC), with the purpose of identifying and diagnosing psychosocial risks at work in the context of CC. To such end, associated companies were contacted by APCC management to participate in the study. The employees of the CC companies who agreed to participate were notified by HR of the objectives of the study and were invited to take part in the study. Through the SurveyMonkey platform, a link was generated which directed participants to an online survey. Finally, the employees were informed that their participation was voluntary, confidential and anonymous.
A convenience sample was obtained, corresponding to a total of 2232 employees from 15 different CC companies, with a response rate of over 70% (ranging from 71% to 81% among the companies, corresponding to 32-432 respondents per company).
However, due to a lack of responses to some of the assessed scales, only 1440 participants were considered for the study. The characteristics of the sample are presented in [fig_ref] Table 1: Sample Demographics' [/fig_ref] , in which the characteristics of the whole sample of front-office and back-office employees are presented.
## Measures
Job Characteristics. Job demands and control (i.e., autonomy) were measured by means of the Job Content Questionnaire (JCQ) [bib_ref] The Job Content Questionnaire (JCQ): An instrument for internationally comparative assessments of..., Karasek [/bib_ref] , as the Portuguese version had already been used in previous studies [bib_ref] Work-to-family enrichment and employees' well-being: High performance work system and job characteristics, Carvalho [/bib_ref]. Therefore, using a Likert scale of 1 (I totally disagree) to 5 (I totally agree), the participants responded to a questionnaire composed of 7 items that analyzed job demands (e.g., I have too much work to do) and 4 items referring to the level of autonomy they had at work (e.g., I have control over what happens in my work). Thus, high scores in these two scales indicate high demands and high autonomy, respectively. The two scales have a good rate of internal consistency, as Cronbach's α was always above 0.7(0.88 and 0.88 for demands and 0.84 and 0.85 for autonomy, for front-office and back-office employees, respectively). The JCQ was also used to measure social supportthrough 5 items that analyzed supervisor support (e.g., My supervisor is concerned about the well-being of his/her employees) and 6 items regarding peer support (e.g., The people I work with help in the accomplishment of tasks). The participants assessed the extent to which they agreed with each statement using a Likert scale of 1 (I totally disagree) to 7 (I totally agree). High scores correspond to a high level of supervisor support and peer support. Internal consistency rates were 0.88 and 0.89 for supervisor support and 0.86 and 0.87 for peer support, for front-office and back-office employees, respectively.
Human resources practices. Human resources practices were analyzed with recourse to an adaptation of the scale used by Chambel, Castanheira, and Sobral [bib_ref] Temporary agency versus permanent workers: A multigroup analysis of human resource management,..., Chambel [/bib_ref] , based on the scales of Lepak and Snell [bib_ref] Examining the human resource architecture: The relationships among human capital, employment, and..., Lepak [/bib_ref] , Slattery, Selvarajan, and Anderson [bib_ref] Influences of new employee development practices on temporary employee work-related attitudes, Slattery [/bib_ref] , Takeuchi, Lepak and Wang [bib_ref] An empirical examination of the mechanisms mediating between highperformance work systems and..., Takeuchi [/bib_ref] and Chambel and Castanheira (2012) [bib_ref] Training opportunities and employee exhaustion in call centres: Mediation by psychological contract..., Chambel [/bib_ref]. The questionnaire consisted of a total of 22 items which analyzed the various human resources practices adopted by the organization to which the participants responded using a Likert scale of 1 (I totally disagree) to 7 (I totally agree). Recruitment was measured by 4 items (e.g., When I was recruited by this company my specific knowledge was analyzed), presenting a Cronbach's α of 0.80 and 0.83 for front-office and back-office employees, respectively. The welcoming and integration process consisted of 4 items (e.g., When I started working in this company I had initial support from my supervisor) and Cronbach's α was 0.83 for front-office and 0.81 for back-office employees. Training was analyzed by means of 5 items (e.g., With the training/experience I have received I can easily change roles within this company), with a Cronbach's α of 0.91 for front-office and 0.90 for back-office employees. Performance assessment was measured by 4 items (e.g., The performance assessment criteria are clear in this company), with a Cronbach's α of 0.88 for front-office employees and 0.91 for back-office employees. Finally, compensation was analyzed by means of 5 items (e.g., In this company, the criteria for assigning the variable component of the salary are clear), with a Cronbach's α of 0.88 and 0.90 for front-office and back-office employees, respectively.
High scores in these dimensions indicate that employees had a more positive perception of the HR practices in place.
Work-life conflict. The work-life conflict was measured through the Portuguese version of the scale of Keeney, Boyd and Sinha [bib_ref] From "work-family" to "work-life": Broadening our conceptualization and measurement, Keeney [/bib_ref] , used by Chambel, Carvalho, and Cesário. It considers 8 work-related domains: Health, family, home management, friendship, education, love relationships, leisure, and community involvement. However, the latter domain was not considered for this study since people in Portugal do not have a high and systematic involvement in community activities. The interference of work in one's personal life may occur in two distinct dimensions, namely time (e.g., Work takes the time that I would like to spend with my family away from me) and stress (e.g., Due to all the pressures of work, I am sometimes too stressed to engage in family activities). Each was measured by 7 items and participants had to assess the extent to which they agreed with each statement using a Likert scale of 1 (I totally disagree) to 5 (I totally agree). Cronbach's α for time was 0.93 for front-office and back-office employees, and for stress 0.94 and 0.95 for front-office and back-office employees, respectively.
Affective organizational commitment. Affective organizational commitment was measured through the Portuguese version of Meyer, Allen and Smith's [bib_ref] Commitment to organizations and occupations: Extension and test of a three-component conceptualization, Meyer [/bib_ref] scale used in the study of Chambel and Castanheira [bib_ref] Reducing burnout in call centers through HR practices, Castanheira [/bib_ref]. The scale is composed of 6 items (e.g., This company has a high personal meaning to me) that were answered using a Likert scale of 1 (I totally disagree) to 7 (I totally agree). This scale also presented good internal consistency, since Cronbach's α was 0.88 for front-office employees and 0.90 for back-office employees.
Well-being at work. Well-being at work was measured by work engagement and burnout. Work engagement was analyzed using the Portuguese version of the Schaufeli, Bakker and Salanova [bib_ref] The measurement of work engagement with a short questionnaire: A cross-national study, Schaufeli [/bib_ref] scale, used previously by Chambel et al. [bib_ref] Temporary agency versus permanent workers: A multigroup analysis of human resource management,..., Chambel [/bib_ref]. This version consisted of 3 items to measure vigor (e.g., In my work I feel full of energy); 3 items for dedication (e.g., I am enthusiastic about my work) and 3 items for absorption (e.g., I am immersed in my work). Participants' responses were measured using a Likert scale of 1 (Never) to 7 (Everyday) and high scores indicate high levels of work engagement. Considering the front-office and back-office employees, Cronbach's α were 0.94 and 0.95 respectively. Burnout was measured by means of the Portuguese version of the Maslach, Jackson and Leiterscale, used previously by Chambel and Castanheira [bib_ref] Reducing burnout in call centers through HR practices, Castanheira [/bib_ref]. This scale is composed of 5 items that analyze exhaustion at work (e.g., I feel exhausted by my work) and 5 items related to cynicism (e.g., I have lost enthusiasm for my work), both measured on a Likert scale ranging from 1 (Never) to 7 (Everyday). As with engagement, high scores indicate high burnout levels. For exhaustion, Cronbach's α was 0.92 for both front-office and back-office employees, while for cynicism, Cronbach's α was 0.85 and 0.84, respectively.
General well-being. General well-being was measured through an adapted version of the General Health Questionnaire (GHQ-12. This questionnaire is composed of 12 items (e.g., Have you been feeling sad and depressed?) and the participants responded using a Likert scale of 1 (Not at all) to 4 (Much more than usual). The scale was subdivided into two dimensions in order to analyze employees' stress and well-being, not in a professional context, but on a general level (i.e., in a free context). Regarding internal consistency, Cronbach's α showed no differences between front-office and back-office employees, standing at 0.84 in the stress sub-scale and 0.87 in the well-being sub-scale.
# Data analysis
The data analysis was performed through the IBM Statistical Package for the Social Sciences (SPSS 25.0, IBM, New York, NY, USA) program.
In order to characterize the sample, a descriptive analysis of variables such as gender, age, marital status, qualifications, work shift and tenure of the respondents, was conducted for the whole sample and for both front-office and back-office duties. A descriptive analysis of the instruments used was then carried out, which made it possible to calculate the main measures of central tendency and dispersion of each of the studied variables. The Student t-test was performed to verify whether the means of the two groups, both front-office and back-office, were statistically different. The Cronbach's alpha of each scale was also calculated to analyze the internal consistency.
Finally, the intra-class correlation coefficient (ICC) was calculated to evaluate the amount of variation in the responses at the individual level for each scale that can be explained by the variability among the 15 CC companies. The intraclass correlation (ICC) was calculated to assess the amount of variance in individual-level responses for each variable that can be explained by variability among the fifteen organizations:
[formula] ICC = (msb − msw)/(msb + ((ng − 1) msw)) (1) [/formula]
where msb is the between-group mean square, msw is the within-group mean square, and ng is the group size, [bib_ref] Within-group agreement, non-independence, and reliability: Implications for data aggregation and analysis, Bliese [/bib_ref]. The higher the ICC value, the higher the proportion of total variance in a subscale is explained by organizational membership. When evaluating the ICC, values exceeding 0.05 are considered relevant for aggregation of individual-level data to a higher organizational level, and 0.20 is considered to be a high level. Thus, it is possible to identify which of the organization's characteristics influence the psychosocial work environment and the stress and well-being of its employees. [fig_ref] Table 2: Mean, Standard Deviation, and Student t-test of the Variables under Study, according... [/fig_ref] shows the mean (M) and standard-deviation (SD) of the variables in the sample under study and by means of the Student t-test, the comparison of means between the back-office and front-office groups may be observed. On the basis of this comparison, it was possible to verify that the employees of these two groups have a similar perception in several of the factors considered, showing that the latter are independent of the duties. Thus, it was possible to observe that employees in both groups have a moderately high perception of job demands and feel that there is moderate supervisor and peer support. As far as HR practices are concerned, employees have a slightly positive perception of the integration and recruitment processes, as well as the training and assessment carried out by the organization. Employees show a weak affective commitment to the organization, relatively low work engagement, relatively high cynicism in the exercise of their professional activity and weak general well-being. However, it was possible to observe some significant differences between the two groups. Front-office employees show lower values in the perception of autonomy, higher values in compensation, work-life conflict (stress dimension) and exhaustion, but lower for general stress. Thus, and although the conditions are similar, the results appear to indicate that front-office and back-office duties influence the perception of some job characteristics and the environment and, consequently, their own well-being. The Intra-class Correlation Coefficient (ICC) values of the variables analyzed in this study may be observed in [fig_ref] Table 3: Intra-class correlation coefficient of the variables under study, by occupation [/fig_ref] , through which the proportion of variance explained by the organization for each duties, front-office and back-office, may be verified. As the number of participants in each organization differed, the Bonferroni test was conducted for each variance analyzed. As far as back-office employees are concerned, some of the job characteristics (job demands, autonomy, supervisor and peer support) of the work environment (recruitment, integration, training, performance assessment, compensation, and work-life conflict-stress dimension) and employees' attitudes and stress and wellbeing (affective organizational commitment, work engagement, cynicism and general well-being-stress dimension), are observed to present significant differences among the various companies (ICC values ≥ 0.05). Thus, these dimensions are dependent on the company, suggesting that back-office work may vary according to the company in which the employee works. On the other hand, exhaustion, work-life conflict (time dimension) and general well-being (well-being dimension) appear to be common to all companies, since they do not present significant differences in variance (ICC values < 0.05). As regards front-office duties, significant differences in variance in job characteristics (job demands, autonomy, peer support), work environment (recruitment, integration, training, performance assessment, compensation, and work-life conflict-stress and time dimension) and attitudes and stress and well-being (affective organizational commitment, cynicism and general well-being-stress dimension) may be observed. However, this is not the case for supervisor support, work engagement, exhaustion and general well-being, -well-being dimension), since no significant differences in variance are observed and, therefore they are common to all companies.
# Results
Thus, it may be concluded that feelings of exhaustion and general well-being in CC appear to be independent of the duties performed or of the companies in which employees develop their professional activity. On the other hand, the remaining job characteristics, namely those related to environment, attitudes and stress and well-being depend either on the duties performed or the company's characteristics.
# Discussion
This study sought to ascertain whether the characteristics of CC work are inevitable or whether they depend on the duties performed, namely front-office or back-office, and the company's characteristics. It was possible to observe significant differences between the two functional groups: Front-office employees appear to have a more negative perception of autonomy and a greater perception of work-life conflict (stress dimension), consequently presenting worse levels of exhaustion. On the other hand, when comparing the results of fifteen different companies, job characteristics, environment, and levels of stress and wellbeing of the employees show significant differences, indicating that these characteristics are not inevitable in CC but rather depend on each company's management strategy.
Regarding job characteristics, as expected, the CC context was found to be characterized by high demands and low control, resulting in high stress and low well-being levels [bib_ref] The very best of the millennium": Longitudinal research and the demand-control (support)..., De Lange [/bib_ref] [bib_ref] Job demands-control model: A summary of current issues and recommendations for future..., Karasek [/bib_ref] [bib_ref] Job Demands-Control Model: A Summary of Current Issues and Recommendations for Future..., Kain [/bib_ref]. However, front-office employees perceived less autonomy compared to back-office employees, in line with the assumption that the use of scripts that organize and structure the making of calls in the case of front-office dutieshas negative repercussions for control in terms of the planning and organization of the tasks performed [bib_ref] Call centers, Holman [/bib_ref] [bib_ref] Differences in call centre agents' perception of their job characteristics, physical work..., Miller [/bib_ref]. When comparing the professionals of these two groups, front-office employees presented higher levels of stress [bib_ref] A three-component conceptualization of organizational commitment, Meyer [/bib_ref] , namely exhaustion, which, in line with the role conflict theory, had an impact on a higher perception of work-family conflict (stress dimension).
On the other hand, this study managed to demonstrate innovatively that stressful characteristics [bib_ref] Job Demands-Control Model: A Summary of Current Issues and Recommendations for Future..., Kain [/bib_ref] (high demands and low control) are not inevitable in the context of CC, as the data suggests variability among the companies and both back-office and front-office duties, showing that it is possible to reduce the workload and increase the autonomy of employees by redesigning these duties. In fact, the data of this study suggest that employees' stress and well-being levels may differ [bib_ref] Dual processes at work in a call centre: An application of the..., Bakker [/bib_ref] [bib_ref] The very best of the millennium": Longitudinal research and the demand-control (support)..., De Lange [/bib_ref] as the quantitative requirements that are associated with monitoring and performance assessment [bib_ref] Employee stress and health complaints in jobs with and without electronic performance..., Smith [/bib_ref] and employees' autonomy, giving them some freedom to plan and organize the tasks they perform [bib_ref] Call centers, Holman [/bib_ref] [bib_ref] Differences in call centre agents' perception of their job characteristics, physical work..., Miller [/bib_ref] , may also vary depending on their occupation and company.
Regarding social support on the part of the supervisor, this study identified a similarity among all the companies for employees with front-office duties. It suggests that the need to monitor and assess customer service may favor the standardization of supervisory duties among different companies. Given the knowledge that social support increases the resources required by employees to deal with high demand situations, this study highlights the need to promote the ability of supervisors to offer adequate social support in the context of CC [bib_ref] Work relationships in telephone call centres: Understanding emotional exhaustion and employee withdrawal, Deery [/bib_ref].
On the other hand, the HR practices analyzed were considered to depend on the company. If these practices are considered fundamental to explain the results obtained in the context of CC [bib_ref] Call centers, Holman [/bib_ref] , but also to explain the perception of the job characteristics themselves [bib_ref] Reducing burnout in call centers through HR practices, Castanheira [/bib_ref] , then the following measures are sorely needed: Investment in appropriate recruitment and selection processes adapted to the duties [bib_ref] Trust and breach of the psychological contract, Robinson [/bib_ref] ; investment in welcoming and integration programs that foster the creation of positive interpersonal relationships among employees [bib_ref] Is managing the work-family interface worthwhile? Benefits for employee health and performance, Van Steenbergen [/bib_ref] ; provision of specific and planned training in order to increase employees' resources; adoption of a remuneration model that is compatible with the HIM model [bib_ref] Managing customer services: Human resources practices, quit rates and sales growth, Batt [/bib_ref] ; and the use of performance assessment as a diagnostic tool which aims to identify features requiring improvement [bib_ref] Computerized performance monitors: Factors affecting acceptance, Grant [/bib_ref].
As for work-life conflict, stress was considered to vary depending on the company and is, therefore, an avoidable variable. This suggests that a discrepancy between the demands of the domain in which the employee participates and the resources to which this professional has access [bib_ref] Investigating the effect of role conflict and role ambiguity on employees' job..., Soltani [/bib_ref] is not observed in all companies. Hence, and although employees have a relatively neutral perception of the work-life conflict, organizations should take measures to promote a balance between these two domains.
Regarding attitudes at work, affective organizational commitment was considered to vary depending on the company, and significant variance was observed in both frontoffice and back-office groups. Affective organizational commitment reflects the bond employees experience with their organization [bib_ref] Organizational commitment and psychological attachment: The effects of compliance, identification, and internalization..., O'reilly [/bib_ref] and may be used as a tool to combat their stress [bib_ref] Commitment to organizations and occupations: Extension and test of a three-component conceptualization, Meyer [/bib_ref]. As this positive attitude depends on the organizational context, namely the human resources management practices in place [bib_ref] High-involvement work processes, work intensification and employee well-being, Boxall [/bib_ref] , and as the perception of these practices differs depending on the company in question, differences in attitude were also expected. However, and since the data suggests that participants have a neutral perception of affective commitment, it is imperative to focus on developing a positive emotional relationship between employees and their organizations [bib_ref] Commitment to organizations and occupations: Extension and test of a three-component conceptualization, Meyer [/bib_ref] in order to promote their well-being.
As for well-being at work, differences in relation to burnout were observed between the two core dimensions of this chronic stress at work syndrome: Exhaustion appears to be cross-cutting and independent of the company and back-office or front-office duties; cynicism, conversely, appears to be dependent on the company for both back-office and front-office duties. This difference may be justified if the development of the burnout syndrome, as posited by the Conservation of Resources theory (COR, [bib_ref] Social and psychological resources and adaptation, Hobfoll [/bib_ref] , is taken into account. According to this theory, employees invest strongly in the acquisition of resources to meet the excessive demands with which they are confronted during their professional activity, resulting in a feeling of high exhaustion which characterizes a stress situation. Thus, working in a CC may be considered a highly demanding situation conducive to stress (i.e., exhaustion), as employees tend to perceive a loss of resources, the threat of resource loss or to invest in resources to face these demands. However, in order to cope with these same demands, employees use coping strategies, which may or may not trigger distancing responses, i.e., cynicism. If the context does not provide resources to protect employees from this sense of loss or threat of loss, they tend to drain their energy resources, and consequently, to protect themselves they will adopt an attitude of detachment which will result in cynicism, a characteristic of the burnout syndrome. However, if the context provides resources (e.g., control, social support, human resource practices that respond to employees' needs), this stress situation may not become a burnout situation, as individuals do not need to adopt this cynical distance to deal with such situations of loss or threat of loss of resources.
With regard to work engagement, this positive psychological state appears to depend on the company for employees with back-office duties; however, it is cross-cutting and independent of the organization for front-office duties. Considering that work engagement is mainly dependent on the resources available to employees in the accomplishment of their tasks [bib_ref] Job demands, job resources, and their relationship with burnout and engagement: A..., Schaufeli [/bib_ref] , in the case of back-office duties, there appear to be situations where the availability of resources varies, thus leading to variable degrees of work engagement, while in front-office duties the differences in resources are not sufficient to reflect differences in this indicator of well-being at work. In line with the results obtained, supervisor support was also found to be a resource that appeared not to differ among companies for employees with front-office duties. Since this resource has an extrinsic motivational role for being instrumental in the acquisition of work objectives, but also intrinsic for being able to satisfy the basic psychological need for relationships [bib_ref] A critical review of the job demands-resources model: Implications for improving work..., Schaufeli [/bib_ref] , it can play a central role in the development of these employees' work engagement.
Finally, and analyzing general well-being, it is possible to observe that general stress is dependent on the company and for both back-office and front-office duties. This result shows that organizations can implement a number of strategies to mitigate employees' stress. Furthermore, according to the burnout literature [bib_ref] Burnout and risk of cardiovascular disease: Evidence, possible causal paths, and promising..., Melamed [/bib_ref] and the health impairment process proposed by the Job Demand-Resource Model [bib_ref] Job demands, job resources, and their relationship with burnout and engagement: A..., Schaufeli [/bib_ref] , burnout leads to different health problems outside the work context. As previously mentioned, although the levels of exhaustion (i.e., stress) are common to the different companies, their association with cynicism (i.e., burnout) is dependent on the company, hence the levels of stress outside the work context may also be dependent on the company.
In conclusion, this study appears to corroborate the idea that working in a CC implies experiencing stress in the accomplishment of one's work, as the levels of exhaustion are independent of back-office and front-office duties and of the company in which one works. However, the development of burnout (i.e., cynicism) and general ill-being does not appear to be inevitable as the emergence of these conditions depends on the company in which the duties are performed. Moreover, job characteristics, peer support, HR practices, work-life conflict (in the stress dimension) and affective organizational commitment also appear to depend on the company. Thus, this study suggests that through a structured and planned intervention at the organizational level, it is possible to promote a healthier work environment that will foster the well-being of employees.
## Limitations and future implications
A number of limitations need to be addressed regarding the present study. First, the method used for data collection may have skewed the results, since a self-assessment questionnaire was used, which was disclosed internally by each company's HR department. Thus, and despite the anonymity of the responses, they may contain some level of social desirability. Secondly, as this is a cross-sectional study, the data refers to a single point in time, which does not allow for the establishment of cause-effect relationships. Therefore, it was only possible to make inferences with regard to the assessment of positive or negative relationships between the studied variables. It would be interesting to conduct a longitudinal study in order to analyze any further developments in the perception of employees regarding the factors studied in this research, as well as to monitor intervention plans that may have been applied. Thirdly, the sample consisted only of employees from Portuguese CC, and it was not possible to generalize the results to other countries or other sectors of activity. In the future, it would be important to compare the results obtained in this study with those of other countries and/or other sectors that are dominated by customer service, such as the hotel industry or trade employees. Finally, the quantitative analysis used in this study does not allow for a comprehensive vision of employees' experience in a contact center context. In the future, it would be interesting to conduct a qualitative study that could examine the meaning of the employees' history in this context, analyzing their experiences (e.g., peer support climate, supervisor support strategies, learning of day-to-day problem-solving strategies) and respective repercussions for their well-being and health.
# Conclusions
This study provides evidence that the specific nature of the duties performed by Contact Center employees, i.e., front-office and back-office duties, has an impact on how they perceive their job characteristics and environment and, consequently, on their wellbeing. In addition, it highlights that although the exhaustion and general well-being of CC are independent of their duties and common to all employees regardless of the company in which they work, job characteristics, psychosocial environment, and the levels of affective organizational commitment, cynicism, and general stress of CC employees depend on the company in which they work.
Author Contributions: D.G.-C. was involved in the design, writing and original draft preparation of this paper. M.J.C. was involved in the data collection, design supervision, writing, methodology and formal analysis. V.S.C. was involved in the reviewing and editing process. All authors have read and agreed to the published version of the manuscript.
Funding: Financial Support of the Foundation of Science and Technology granted to the Research Center for Psychological Science of University of Lisbon (CICPSI)-UIDB/04527/2020.
## Institutional review board statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee of Faculty of Psychology, University of Lisbon (protocol "The management of the work-personal (family) life border in the case of temporary workers" approved on 20 November 2019).
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
# Data availability statement:
The data presented in this study are available on request from the corresponding author.
## Conflicts of interest:
The authors declare no conflict of interest.
[table] Table 1: Sample Demographics'. [/table]
[table] Table 2: Mean, Standard Deviation, and Student t-test of the Variables under Study, according to Occupation. The t value was calculated by the mean difference between back-office and front-office employees. ** According to the Levene Test, equal variances were deemed not assumed. [/table]
[table] Table 3: Intra-class correlation coefficient of the variables under study, by occupation. [/table]
|
10.1038/s41598-021-89628-z
|
CCBY
|
8160131
|
34045512
|
s2orc_pubmed_articles
|
Transcript levels of spindle and kinetochore-associated complex 1/3 as prognostic biomarkers correlated with immune infiltrates in hepatocellular carcinoma
The spindle and kinetochore-associated protein complex (Ska) is an essential component in chromosome segregation. It comprises three proteins (Ska1, Ska2, and Ska3) with theorized roles in chromosomal instability and tumor development, and its overexpression has been widely reported in a variety of tumors. However, the prognostic significance and immune infiltration of Ska proteins in hepatocellular carcinoma (HCC) are not completely understood. The bioinformatics tools Oncomine, UALCAN, gene expression profiling interactive analysis 2 (GEPIA2), cBioPortal, GeneMANIA, Metascape, and TIMER were used to analyze differential expression, prognostic value, genetic alteration, and immune cell infiltration of the Ska protein complex in HCC patients. We found that the mRNA expression of the Ska complex was markedly upregulated in HCC. High expression of the Ska complex is closely correlated with tumor stage, patient race, tumor grade, and TP53 mutation status. In addition, high expression of the Ska complex was significantly correlated with poor diseasefree survival, while the high expression levels of Ska1 and Ska3 were associated with shorter overall survival. The biological functions of the Ska complex in HCC primarily involve the amplification of signals from kinetochores, the mitotic spindle, and (via a MAD2 invasive signal) unattached kinetochores. Furthermore, the expression of the complex was positively correlated with tumorinfiltrating cells. These results may provide new insights into the development of immunotherapeutic targets and prognostic biomarkers for HCC.Hepatocellular carcinoma (HCC), the most common type of primary liver cancer, is one of the most common cancers and ranks second among causes of cancer-related deaths 1 . Hepatitis virus infection, alcohol consumption, obesity, and aflatoxin are considered risk factors for HCC 2,3 . In recent years, the treatment of liver cancer has been greatly developed, including arterial chemoembolization, hepatectomy, radiotherapy, and targeted therapy. However, due to frequent late-stage diagnosis, recurrence, and metastasis, the overall 5-year survival rate (7%) remains poor 4 . Many studies have explored the role of immune infiltration-related mechanisms in HCC, in search of specific targets for immunotherapy 5 . Thymocyte selection-associated high mobility group box protein (TOX) and P-selectin glycoprotein ligand-1 (PSGL-1) modulate the tumor microenvironment by depleting CD8 + T cells, and they hold promise as targets for tumor immunotherapy 6,7 . Due to the poor treatment results and low survival rate of HCC, it is particularly important to identify reliable predictive biological targets for early diagnosis and to improve the prognosis of patients through immunotherapy.
Mitotic abnormalities are a common feature of most tumors, and the separation of chromosomes during mitosis is mainly driven by kinetochores attached to specific regions of spindle microtubules [bib_ref] A Kinesin-5, Cin8, recruits protein phosphatase 1 to kinetochores and regulates chromosome..., Suzuki [/bib_ref]. The spindle and kinetochore-associated (Ska) complex is composed of three protein subunits: Ska1, Ska2, and Ska3, which are necessary for the stabilization of kinetochore-spindle microtubule attachment during mitosis [bib_ref] Kinetochore recruitment of the spindle and kinetochore-associated (Ska) complex is regulated by..., Lange [/bib_ref] [bib_ref] Multitasking Ska in chromosome segregation: its distinct pools might specify various functions, Zhang [/bib_ref]. Many studies have shown that dysregulation of the SKA family of genes is associated with a variety of cancers. For example, upregulation of SKA1 expression in esophageal squamous cell carcinoma tissues is associated with tumor differentiation and pathological tumor node metastasis (TNM) stage. Esophageal cancer patients with high expression of SKA1 have a poorer prognosis than patients with low expression [bib_ref] SKA1 overexpression is associated with the prognosis of esophageal squamous cell carcinoma..., Hu [/bib_ref]. SKA2 is significantly upregulated in breast cancer tissues and is associated with TNM stage and lymph node metastasis. High expression of SKA2 promotes invasion and metastasis of breast cancer cells via epithelial-mesenchymal transition (EMT) [bib_ref] SKA2 mediates invasion and metastasis in human breast cancer via EMT, Ren [/bib_ref]. SKA3 expression is also increased in cervical cancer tissues, and cervical cancer patients with high SKA3 expression have a poor prognosis. SKA3 overexpression promotes cervical cancer cell proliferation and migration and accelerates tumor growth [bib_ref] SKA3 promotes cell proliferation and migration in cervical cancer by activating the..., Hu [/bib_ref]. In pancreatic cancer, high expression of SKA1 and SKA3 is associated with poor prognosis and immune cell infiltration. Furthermore, SKA1 has emerged as a prognostic indicator associated with tumor cell infiltration and holds promise as a therapeutic target in adrenocortical carcinoma [bib_ref] Identification of tumor-infiltrating immune cells and prognostic validation of tumor-infiltrating mast cells..., Tian [/bib_ref].
In recent years, the SKA gene family has been increasingly studied in HCC, and previous studies have shown that these genes are highly expressed in HCC [bib_ref] SKA1 overexpression is associated with poor prognosis in hepatocellular carcinoma, Chen [/bib_ref] [bib_ref] SKA2 promotes proliferation and invasion of hepatocellular carcinoma cells via activating the..., Wang [/bib_ref] [bib_ref] SKA3 Promotes tumor growth by regulating CDK2/P53 phosphorylation in hepatocellular carcinoma, Hou [/bib_ref]. However, the potential importance of SKA genes in this disease, especially in prognostic development and immune infiltration, has not been comprehensively elucidated.
The role of the SKA gene family in HCC has been explored with gene sequencing and the use of various bioinformatics databases. In this study, several databases were used for data mining of HCC patients, aiming to systematically and comprehensively explore the gene expression, prognostic value, immune correlation, and potential function of SKA genes in HCC patients. Our study may reveal the molecular mechanisms involved in the expression and regulation of the Ska complex and the development of HCC and could provide reliable targets for HCC diagnosis and treatment.
# Materials and methods
Oncomine database. The Oncomine database (www. oncom ine. org) is a publicly accessible online database that provides an analysis of genome-wide expression with a range of cancer microarray information [bib_ref] ONCOMINE: a cancer microarray database and integrated data-mining platform, Rhodes [/bib_ref]. The expression data of SKA genes in diverse cancer types were obtained from Oncomine. In this study, a Student's t-test was performed on this data with the significance threshold set as follows: P value = 0.05; fold change = 2; gene rank: 10%; data type: mRNA.
UALCAN. UALCAN (http:// ualcan. path. uab. edu/ analy sis. html) is an interactive web resource for in-depth analysis of cancer data from The Cancer Genome Atlas (TCGA) database [bib_ref] UALCAN: a portal for facilitating tumor subgroup gene expression and survival analyses, Chandrashekar [/bib_ref]. It was used to analyze the expression of SKA1-3 in both normal and cancerous tissues. Student's t-test was used to generate P values. The P value cutoff was set at 0.05. GEPIA2. GEPIA2 (http:// gepia2. cancer-pku. cn/) is a website for analyzing the RNA sequencing expression data of 9736 tumors and 8587 normal samples from the TCGA and Genotype-Tissue Expression (GTEx) projects, using a standard processing pipeline [bib_ref] GEPIA2: an enhanced web server for large-scale expression profiling and interactive analysis, Tang [/bib_ref]. In this study, we explored the expression differences of SKA genes in HCC tissues and normal tissues, the analysis of pathological stages, and the related prognostic analysis using the "Single Gene Analysis" module of GEPIA. Student's t-test was used with a critical value for the P value of 0.05. cBioPortal. cBioPortal (www. cbiop ortal. org), an open online tool, can visualize and analyze multidimensional cancer genomics [bib_ref] The cBio cancer genomics portal: an open platform for exploring multidimensional cancer..., Cerami [/bib_ref] [bib_ref] Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal, Gao [/bib_ref]. Based on the TCGA database, genetic alterations and the summary of gene types were analyzed using cBioPortal, as well as the relationship between gene mutations and the prognosis of HCC patients. Statistical significance was set at P < 0.05.
GeneMANIA. GeneMANIA (http:// www. genem ania. org) is a flexible web interface that can generate hypotheses about gene function, analyze gene lists, and prioritize genes for functional assays [bib_ref] The GeneMANIA prediction server: biological network integration for gene prioritization and predicting..., Warde-Farley [/bib_ref]. Using GeneMA-NIA, it was possible to identify the relationships between the Ska proteins and their interactive genes.
Metascape. Metascape (http:// metas cape. org) is a web-based portal that provides gene annotation, enrichment analysis, and a protein-protein interaction (PPI) network based on over 40 independent knowledgebases [bib_ref] Metascape provides a biologist-oriented resource for the analysis of systems-level datasets, Zhou [/bib_ref]. To verify the enrichment of the Ska proteins and the SKA genes, the "Express Analysis" module of Metascape was used.
TIMER. TIMER (http:// timer. cistr ome. org/) is a freely available web server for investigating the infiltration of diverse immune cells and their clinical impact [bib_ref] TIMER2.0 for analysis of tumor-infiltrating immune cells, Li [/bib_ref]. The Ska proteins were submitted to the "Gene module" of TIMER and their correlation with immune cells (B cells, CD4 + T cells, CD8 + T cells, macrophages, neutrophils, and dendritic cells) was explored.
# Results
Abnormal expression of SKA genes in HCC patients. To explore the expression of SKA genes in HCC patients, we first quantified the mRNA expression level using the Oncomine database, which showed that SKA1 expression was dramatically elevated in HCC tissues compared with normal tissues [fig_ref] Figure 1: mRNA expression levels of SKAs in different types of human cancer [/fig_ref]. Specifically, using Chen's dataset, the analysis showed that SKA1 was overexpressed 2.534-fold in liver hepatocellular carcinoma www.nature.com/scientificreports/ (LIHC) specimens. Using the Wurmbach liver dataset [fig_ref] Table 1: Transcription expression of SKAs family members between LIHC and normal liver tissues [/fig_ref] , SKA1 was overexpressed 1.701-fold. UALCAN was also used to analyze the expression of SKA genes in HCC and normal tissues, and the results showed that the expression of SKA1 (P = 1.62e−12), SKA2 (P = 1.62e−12), and SKA3 (P < 1e−12) increased significantly in HCC tissues . We also used the GEPIA2 database to verify the expression of SKA genes in HCC tissues. Consistent with the previous results, the protein expression levels of SKA1, SKA2, and SKA3 were significantly higher in HCC patients . In summary, all three genes in the SKA family were significantly upregulated in HCC patients.
Clinicopathological parameters of SKA genes in HCC patients. We explored the relationship between the SKA gene expression level and the clinical characteristics of patients with HCC, including tumor stage, patient race, tumor grade, and TP53 mutation status. As shown in [fig_ref] Figure 3: Correlation between mRNA expression levels of SKAs and clinicopathological parameters of HCC... [/fig_ref] , SKA gene expression was significantly associated with the stage of HCC, with increased expression correlating with a higher stage. The expression of SKA family genes was significantly upregulated in stages 1, 2, 3, and 4 compared with normal liver tissue, while there was no significant difference between normal tissue and stage 4 HCC, possibly due to the small number of stage 4 cases (n = 6). Among these, there were also significant differences in the expression of SKA genes between stages 1 and 2 and between stages 2 and 3. We also examined the relationship between SKA expression levels and the race of the patients, which showed that Caucasian, African American, and Asian patients all presented with significantly higher expression levels compared with normal tissues, and that Asian patients showed higher expression differences than Caucasian patients [fig_ref] Figure 3: Correlation between mRNA expression levels of SKAs and clinicopathological parameters of HCC... [/fig_ref]. In terms of tumor grade, an increasing grade correlated with increased SKA gene expression [fig_ref] Figure 3: Correlation between mRNA expression levels of SKAs and clinicopathological parameters of HCC... [/fig_ref]. Furthermore, there was significant variability in the expression of SKA genes between levels. Finally, we also examined the relationship between the expression of SKA genes and the mutation status of TP53, which showed that expression of the Ska protein complex was significantly upregulated in patients with TP53 mutations compared with patients without mutations [fig_ref] Figure 3: Correlation between mRNA expression levels of SKAs and clinicopathological parameters of HCC... [/fig_ref]. In general, SKA gene expression levels were significantly correlated with tumor stage, tumor grade, and TP53 mutation status, and Asian patients presented a higher relative increase in the expression of SKA genes than Caucasian and African American patients.
## Prognostic value of ska genes in hcc patients.
To determine the relationship between differential expression of SKA genes and prognosis of HCC patients, we first analyzed the correlation between differential expression and overall survival using GEPIA2. This showed that patients with high expression of SKA1 (P = 0.0023) and SKA2 (P = 0.00042) were primarily associated with shorter overall survival [fig_ref] Figure 4: The prognostic value of mRNA expression level of SKAs in HCC patients [/fig_ref]. The relationship between the differential expression of SKA genes and disease-free survival was also evaluated. We found that patients with high expression of SKA1 (P = 0.00037), SKA2 (P = 0.024), and SKA3 (P = 0.0027) were significantly associated with shorter disease-free survival [fig_ref] Figure 4: The prognostic value of mRNA expression level of SKAs in HCC patients [/fig_ref]. These results suggest that SKA gene expression plays a crucial role in the prognosis of patients with HCC and may become a reliable predictor of survival in these patients.
Frequency changes of SKA genes in HCC patients. We used the cBioPortal database to analyze frequency changes in SKA genes in HCC patients. Fifty-nine (17%) patients had significant alterations in the SKA genes, including missense mutations, amplifications, deep deletions, and transcriptional upregulation. Specifically, the percentages of gene alterations in SKA1, SKA2, and SKA3 were 4%, 10%, and 8%, respectively [fig_ref] Figure 5: Alteration frequency of SKAs and their prognostic value in HCC patients [/fig_ref]. We further explored the impact of gene alterations in the SKA family on the prognosis of patients with HCC, which showed that patients with genetically altered HCC had shorter overall survival than patients with unchanged SKA genes (P = 0.0294). However, there was no relationship between the SKA family gene alterations and disease-free survival (P = 0.0963) in HCC patients [fig_ref] Figure 5: Alteration frequency of SKAs and their prognostic value in HCC patients [/fig_ref].
Co-expression and enrichment of SKA genes in HCC patients. To further explore the role of the SKA genes in HCC patients, we analyzed their regulatory network and their functionally similar genes using the GeneMANIA database. The results showed 20 genes with the strongest correlation, which were nudixhydrolase 5 (NUDT5), kinetochore protein SPC24 (SPC24), kinetochore-associated protein DSN1 homolog (DSN1), kinetochore protein NDC80 homolog (NDC80), SS18-like protein 1 (SS18L1), centromere protein E (CENPE), mitotic checkpoint serine/threonine-protein kinase BUB1 (BUB1), aurora kinase B (AURKB), kinetochore protein SPC25 (SPC25), centromere protein U (CENPU), centromere protein K (CENPK), centromere protein M (CENPM), protein MIS12 homolog (MIS12), DNA excision repair protein ERCC-6-like (ERCC6L), baculoviral IAP repeat containing protein 5 (BIRC5), kinetochore protein Nuf2 (NUF2), shugoshin 1 (SGO1), centromere protein A (CENPA), kinesin family member 18A (KIF18A), ZWILCH kinetochore protein (ZWILCH) [fig_ref] Figure 6: Gene-gene network of SKAs in HCC patients [/fig_ref]. Subsequently, we used Metascape to explore the biological functions of SKA genes and the aforementioned coexpressed genes. The top nine most abundant terms are shown in [fig_ref] Figure 7: The enrichment analysis of SKAs and the 20 co-expressed genes in HCC... [/fig_ref] : amplification of signals from the kinetochores, cell division, chromosome segregation, PID PLK1 pathway, kinetochore organization, microtubule cytoskeleton organization, NDC80 kinetochore complex, microtubule polymerization or depolymerization, and meiotic nuclear division. In addition, we constructed a network map of the enriched terms [fig_ref] Figure 7: The enrichment analysis of SKAs and the 20 co-expressed genes in HCC... [/fig_ref]. To further analyze the relationship between SKA genes and HCC, PPI network maps were constructed, and MCODE component analysis was also performed [fig_ref] Figure 7: The enrichment analysis of SKAs and the 20 co-expressed genes in HCC... [/fig_ref]. The most significantly different MCODE components were extracted from the PPI network graph, and the results showed that there were two significantly different components. MCODE1 was associated with amplification of signals from unattached kinetochores via a MAD2 inhibitory signal, amplification of signals from the kinetochores, and the mitotic spindle checkpoint. MCODE2 was associated with the CENP-H-I complex, CENP-A NAC-CAD complex, and CEN complex. [fig_ref] Figure 8: The relationship between SKAs and tumor immunological features of HCC patients [/fig_ref].
# Discussion
Ska1, Ska2 and Ska3, which constitute the major components of the Ska complex, play a key role in the normal segregation of chromosomes during mitosis. Chromosomal malformation, a peculiar phenomenon of tumors, leads to genomic instability, thereby promoting tumor occurrence and development [bib_ref] Using telomeric chromosomal aberrations to evaluate clastogen-induced genomic instability in mammalian cells, Bolzan [/bib_ref]. An increasing number of studies have demonstrated that Ska proteins play an important role in tumorigenesis, cancer cell proliferation, and apoptosis [bib_ref] SKA3 promotes cell growth in breast cancer by inhibiting PLK-1 protein degradation, Ruan [/bib_ref] [bib_ref] High expression of spindle and kinetochore-associated protein 1 predicts early recurrence and..., Jiang [/bib_ref]. In recent years, immunotherapy has drawn increasing attention in the treatment of cancer; however, the prognostic value and immune infiltration of the Ska proteins in HCC have not been comprehensively explored. This study demonstrated that the Ska proteins and the SKA genes were abnormally highly expressed in HCC, suggesting a link between the dysregulation of these genes and HCC. We further examined the relationship between the expression of SKA genes and the clinical characteristics of patients with HCC. The results showed www.nature.com/scientificreports/ that a higher expression of SKA genes was significantly correlated with tumor stage and pathological grade and that this expression also differed with the ethnicity of the patients, with a greater increase in expression levels in Asian than in Caucasian patients. In addition, there was a correlation between Ska protein complex expression and TP53 mutations. TP53 mutations are the most common mutations in HCC, and they lead to the downregulation of the immune response and differential expression of immune-related genes in HCC [bib_ref] Development and validation of a TP53-associated immune prognostic model for hepatocellular carcinoma, Long [/bib_ref]. It has been shown that receptor activity-modifying protein 3 (RAMP3) in HCC patients may reduce the detrimental effect of TP53 mutations on survival [bib_ref] RAMP3 is a prognostic indicator of liver cancer and might reduce the..., Fang [/bib_ref]. Furthermore, in HCC patients, high expression of SKA1 and SKA2 was notably associated with shorter overall survival, while high expression of the Ska protein complex was markedly associated with shorter disease-free survival. To date, many studies have confirmed that SKA genes play an important role in HCC. One study examined 166 HCC and paired adjacent normal tissues and found that SKA1 was highly expressed in HCC and correlated with tumor size and TNM stage [bib_ref] SKA1 overexpression is associated with poor prognosis in hepatocellular carcinoma, Chen [/bib_ref]. Another study found that LINC00339 could [bib_ref] LINC00339 promotes growth and invasiveness of hepatocellular carcinoma by the miR-1182/SKA1 pathway, Xiao [/bib_ref]. It has also been confirmed that SKA2 can accelerate HCC progression by upregulating Wnt/β-catenin signaling [bib_ref] Spindle and kinetochore-associated protein 2 facilitates the proliferation and invasion of hepatocellular..., Jiang [/bib_ref]. Our findings regarding SKA gene expression in HCC are in agreement with those of a previous study. Therefore, we speculated that individual SKA genes or SKA family genes could serve as potential prognostic biomarkers for patients with HCC. However, the effects of SKA genes on the development, metastasis, cell proliferation, and apoptosis of HCC have not been comprehensively studied. The occurrence and development of HCC is complex and multifaceted, and genetic changes play a role in this process [bib_ref] Integrative analysis reveals novel driver genes and molecular subclasses of hepatocellular carcinoma, Yang [/bib_ref]. Therefore, we explored the molecular characteristics of SKA genes in HCC. In HCC, the differential expression of SKA genes often undergo genetic changes, which are relevant to the overall survival rate. The most significant genetic change was the increase in mRNA expression.
Previous studies have revealed that Ska proteins function in other diseases, mainly by regulating chromosome segregation [bib_ref] Integrating pathology, chromosomal instability and mutations for risk stratification in early-stage endometrioid..., Li [/bib_ref] [bib_ref] Ensemble-level organization of human kinetochores and evidence for distinct tension and attachment..., Roscioli [/bib_ref]. In this study, we explored the core genes underlying the function of Ska proteins, some of which have been identified as regulators of these proteins. Redli et al. [bib_ref] The Ska complex promotes Aurora B activity to ensure chromosome biorientation, Redli [/bib_ref] reported that Ska proteins promote AURKB activity to limit their own microtubule and mitochondria association and ensure that kinetochore-microtubule (KT-MT) dynamics and stability fall within an optimal bi-directional equilibrium range. Sivakuma et al. [bib_ref] Phosphatase-regulated recruitment of the spindle-and kinetochore-associated (Ska) complex to kinetochores, Sivakumar [/bib_ref] demonstrated that NUF2 binding to Ska1 promotes the recruitment of the Ska complex to kinetochores, reducing metaphase arrest upon chromosome segregation. This implicates NUF2 as a potential regulator of Ska1. NDC80 has also been shown to influence the recruitment of the Ska complex [bib_ref] Mechanism of Ska recruitment by Ndc80 complexes to kinetochores, Janczyk [/bib_ref]. In our study, we performed functional enrichment analysis to understand the biological functions of the SKA genes. The results suggest that these genes are mainly involved in the amplification of signals from the kinetochores and mitochondrial spindle checkpoint, and further amplification of signals from unattached kinetochores occurs via a MAD2 inhibitory signal. Our results concur with those of a previous report 42 that high expression of the mitotic checkpoint protein MAD2 in the mammary gland of mice resulted in mitotic checkpoint hyperactivation, mitotic arrest, and retarded tumor growth. This suggests that SKA genes play a crucial role in the development and progression of tumors.
Accumulating evidence suggests that immune cell infiltration can influence tumorigenesis and recurrence and serve as an important determinant of immunotherapy response and clinical outcome [bib_ref] The prognostic landscape of tumor-infiltrating immune cell and immunomodulators in lung cancer, Liu [/bib_ref]. Similarly, the immune microenvironment of tumors plays an important role in HCC [bib_ref] Deviations of the immune cell landscape between healthy liver and hepatocellular carcinoma, Rohr-Udilova [/bib_ref]. CD8 + cytotoxic T lymphocytes can specifically identify major histocompatibility complex (MHC) antigens, which are widely used in tumor-targeted therapies [bib_ref] RIG-I-like helicases induce immunogenic cell death of pancreatic cancer cells and sensitize..., Duewell [/bib_ref]. One study showed that an elevated ratio of CD4+/CD8+ T cells was associated with a favorable prognosis in HCC [bib_ref] High-intensity focused ultrasound ablation combined with transcatheter arterial chemoembolization improves long-term efficacy..., Sun [/bib_ref]. It has also been shown that the co-expression of PD-1 and T-cell immunoglobulin and tyrosine inhibitory motif domain (TIGIT) in CD4+ and CD8+ T cells of HCC patients was significantly increased and negatively correlated with the overall survival and disease-free survival of patients [bib_ref] PD-1(+) TIGIT(+) CD8(+) T cells are associated with pathogenesis and progression of..., Liu [/bib_ref]. Our study suggests that the expression levels of SKA genes may be significantly associated with the infiltration of immune cells (B cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells). This suggests that SKA genes not only respond to the prognosis of HCC but also reflect the immune status of the disease and can provide new insights into HCC immunotherapy. www.nature.com/scientificreports/ In addition, it must be acknowledged that our study has some limitations. First, we obtained data from several different databases, and it was difficult to guarantee that the data were consistent. Moreover, experimental validation of these data has not been performed at the time of writing, but this will be performed in our future work.
# Conclusions
We found that SKA gene expression levels were highly elevated in HCC. Ska1 and Ska3 could be considered potential prognostic markers. SKA gene expression was also significantly associated with the infiltration of immune cells (B cells, CD4 + T cells, macrophages, neutrophils, and dendritic cells), which indicated that Ska proteins may regulate the development of HCC by influencing the immune microenvironment. Inhibition of Ska protein expression and function, potentially in combination with immunotherapies, could represent a promising treatment strategy for patients with HCC.
[fig] Figure 1: mRNA expression levels of SKAs in different types of human cancer (Oncomine). Red indicates high expression. Blue indicates low expression. P < 0.05, fold-change > 2 and gene rank = 10% were considered statistically significant. Numbers in each cell represent the data set meeting the threshold. SKA spindle and kinetochore-associated complex.Immune cell infiltration of Ska proteins in HCC patients.Immune infiltration is associated with the development of cancer. In the present study, the TIMER database was used to analyze the correlation between SKA gene expression (and therefore Ska protein expression) and immune infiltration.All three SKA genes showed positive correlation between expression and immune cell infiltration. The expression of SKA1 was positively correlated with B cells (Rho = 0.38, P = 2.90e−13), CD4 + T cells (Rho = 0.206, P = 1.16e−4), macrophages (Rho = 0.304, P = 8.41e−9), neutrophils (Rho = 0.412, P = 8.32e−3), and dendritic cells (Rho = 0.477, P = 4.93e−21) (Fig. 8a). SKA2 was also positively correlated with B cells (Rho = 0.241, P = 5.97e−6), CD8 + T cells (Rho = 0.135, P = 1.23e−2), CD4 + T cells (Rho = 0.205, P = 1.3e−4), macrophages (Rho = 0.31, P = 34.29e−9), neutrophils (Rho = 0.237, P = 8.47e−6), and dendritic cells (Rho = 0.336, P = 1.47e−10) (Fig. 8b). Similarly, there was a positive correlation between the expression of SKA3 and the infiltration of B cells (Rho = 0.42, P = 3.66e−12), CD4 + T cells (Rho = 0.237, P = 8.34e−6), macrophages (Rho = 0.298, P = 1.67e−8), neutrophils (Rho = 0.146, P = 6.43e−3), and dendritic cells (Rho = 0.498, P = 4.73e−23) [/fig]
[fig] Figure 3: Correlation between mRNA expression levels of SKAs and clinicopathological parameters of HCC patients. (UALCAN). (a) Correlation between mRNA expression levels of SKAs and tumor stages of HCC patients. (b) Correlation between mRNA expression levels of SKAs and race of HCC patients. (c) Correlation between mRNA expression levels of SKAs and tumor grade of HCC patients. (d) Correlation between mRNA expression levels of SKAs and TP53 mutation status of HCC patients. [/fig]
[fig] Figure 4: The prognostic value of mRNA expression level of SKAs in HCC patients (GEPIA2). (a) The relationship between SKAs expression and OS in HCC patients. (b) The relationship between SKAs expression and DFS in HCC patients. OS overall survival; DFS disease-free survival. [/fig]
[fig] Figure 5: Alteration frequency of SKAs and their prognostic value in HCC patients (cBioPortal). (a,b) Summary of alterations in the SKAs in HCC patients. (c,d) K-M plots curve of OS and DFS in HCC patients with/without the SKAs alterations. [/fig]
[fig] Figure 6: Gene-gene network of SKAs in HCC patients (GeneMANIA). GeneMANIA database identified 20 genes most associated with SKAs. [/fig]
[fig] Figure 7: The enrichment analysis of SKAs and the 20 co-expressed genes in HCC patients (Metascape). (a) Bar chart of the first nine enriched terms for SKAs and the 20 co-expressed genes. (b) Net graph of enriched terms, Different colors represent different cluster ID. (c,e) PPI network and MCODE components identified. Scientific Reports | (2021) 11:11165 | https://doi.org/10.1038/s41598-021-89628-z [/fig]
[fig] Figure 8: The relationship between SKAs and tumor immunological features of HCC patients. (a) The relationship between SKA1 and immune infiltrating cells. (b) The relationship between SKA2 and immune infiltrating cells. (c) The relationship between SKA3 and immune infiltrating cells. [/fig]
[table] Table 1: Transcription expression of SKAs family members between LIHC and normal liver tissues (Oncomine). [/table]
|
10.1210/en.2014-1554
|
CCBY
|
4298327
|
25456067
|
s2orc_pubmed_articles
|
Unliganded Thyroid Hormone Receptor α Regulates Developmental Timing via Gene Repression in Xenopus tropicalis
Thyroid hormone (TH) receptor (TR) expression begins early in development in all vertebrates when circulating TH levels are absent or minimal, yet few developmental roles for unliganded TRs have been established. Unliganded TRs are expected to repress TH-response genes, increase tissue responsivity to TH, and regulate the timing of developmental events. Here we examined the role of unliganded TR␣ in gene repression and development in Xenopus tropicalis. We used transcription activator-like effector nuclease gene disruption technology to generate founder animals with mutations in the TR␣ gene and bred them to produce F1 offspring with a normal phenotype and a mutant phenotype, characterized by precocious hind limb development. Offspring with a normal phenotype had zero or one disrupted TR␣ alleles, and tadpoles with the mutant hind limb phenotype had two truncated TR␣ alleles with frame shift mutations between the two zinc fingers followed by 40 -50 mutant amino acids and then an out-of-frame stop codon. We examined TH-response gene expression and early larval development with and without exogenous TH in F1 offspring. As hypothesized, mutant phenotype tadpoles had increased expression of TH-response genes in the absence of TH and impaired induction of these same genes after exogenous TH treatment, compared with normal phenotype animals. Also, mutant hind limb phenotype animals had reduced hind limb and gill responsivity to exogenous TH. Similar results in methimazole-treated tadpoles showed that increased TH-response gene expression and precocious development were not due to early production of TH. These results indicate that unliganded TR␣ delays developmental progression by repressing TH-response genes.(Endocrinology 156: 735-744, 2015)
The lethal effect of paired box transcription factor 8 Ϫ/Ϫ mutation (lack of a thyroid gland and TH) compared with the nonlethal effect of TR␣/TR null mutations in mice points to the developmental actions of TR isoforms in the absence of TH [bib_ref] Athyroid Pax8Ϫ/Ϫ mice cannot be rescued by the inactivation of thyroid hormone..., Mittag [/bib_ref]. Three specific examples of TR␣ action in the absence of TH in cerebellum [bib_ref] Deletion of the thyroid hormone receptor ␣1 prevents the structural alterations of..., Morte [/bib_ref] , heart [bib_ref] Thyroid hormone receptor ␣ is a molecular switch of cardiac function between..., Mai [/bib_ref] , and cochlea (7) development have been identified using TR␣-deficient mice. These studies showed evidence for TR␣-mediated repression of neuron migration or ion channel expression before TH availability, although altered organ function was not observed for cochlea under euthyroid conditions. The paucity of phenotypes attributable to unliganded TR may be due to the presence of low levels of TH from maternal or fetal sources that may conceal potential effects of unliganded TR. Deleterious effects of unliganded TR may be uncovered in pathological hypothyroid conditions, as shown for the cochlea. A useful model to reveal potential effects of unliganded TR in development in both euthyroid and hypothyroid conditions is frog metamorphosis because of the dramatic and wellstudied dependence of developmental events on TH and importantly because maternal sources of TH are not present in premetamorphic tadpoles prior to endogenous TH production [bib_ref] Unliganded thyroid hormone receptor regulates metamorphic timing via the recruitment of histone..., Shi [/bib_ref].
First proposed when frog TRs were cloned and then elaborated later , a dual function model for the role of TR in postembryonic development states the following: 1) TRs act to repress genes involved in developmental progression in the absence of TH to allow larval growth, and 2) upon TH release into circulation, TRs bind TH and induce the previously repressed genes to initiate metamorphosis. Strong evidence exists for the second part of this model, but unequivocal support for the role of unliganded TR in gene regulation and metamorphosis has been lacking.
The dual function model emerged from knowledge of the ligand-dependent action of TRs and the developmental profiles of TR expression and circulating TH levels [bib_ref] Molecular and developmental analyses of thyroid hormone receptor function in Xenopus laevis,..., Buchholz [/bib_ref]. As in other vertebrates, two gene loci for TR, TR␣ and TR, are found in frogs [bib_ref] Xenopus laevis ␣ and  thyroid hormone receptors, Yaoita [/bib_ref]. TRs are nuclear receptors that mediate the TH signal by regulating expression of TH-response genes, depending on the presence or absence of hormone [bib_ref] Repression of transcription mediated at a thyroid hormone response element by the..., Sap [/bib_ref] [bib_ref] Protein encoded by v-erbA functions as a thyroid-hormone receptor antagonist, Damm [/bib_ref]. In the absence of TH, TRs bind promoter or enhancer regions of regulated genes and recruit corepressors, nuclear receptor corepressor (NCoR) and silencing mediator of retinoid and thyroid receptors, which deacetylate histones and repress gene expression, and in the presence of TH, a conformational shift in TR favors binding coactivators, including as CBP/p300, coactivator-associated arginine methyltransferase 1, protein arginine methyltransferase, and steroid receptor coactivators, leading to histone acetylation and methylation, altered chromatin structure, and gene induction [bib_ref] Molecular and developmental analyses of thyroid hormone receptor function in Xenopus laevis,..., Buchholz [/bib_ref] [bib_ref] Dual functions of thyroid hormone receptors in vertebrate development : the roles..., Shi [/bib_ref] [bib_ref] Mechanisms of thyroid hormone receptor action during development: lessons from amphibian studies, Grimaldi [/bib_ref]. One or both TR isoforms are expressed in virtually all tissues in the tadpole [bib_ref] Developmental and regional expression of thyroid hormone receptor genes during Xenopus metamorphosis, Kawahara [/bib_ref] , and TR␣ expression in particular begins very early in development before the start of feeding [bib_ref] Early metamorphic competence of Xenopus larvae, Tata [/bib_ref]. TR expression levels are very low in premetamorphosis and are highly induced by TH [bib_ref] A correlation of thyroid hormone receptor gene expression with amphibian metamorphosis, Yaoita [/bib_ref]. Detectable TH in circulation is coincident with the initiation of metamorphosis and reaches a peak at metamorphic climax [bib_ref] La triiodothyronine: hormone de la métamorphose des amphibiens, Leloup [/bib_ref]. Importantly, TH is not detected in circulation until 3 weeks after feeding begins, indicating a substantial amount of time that TRs exist in the unliganded condition.
Extensive indirect mechanistic evidence supports widespread effects of unliganded TR in the premetamorphic tadpole. The initiation of expression of TR correlates with the decreased expression in the TH-direct response genes stromelysin-3 (ST3) [bib_ref] Transcriptional activation of the matrix metalloproteinase gene stromelysin-3 coincides with thyroid hormone-induced..., Patterton [/bib_ref] and sonic hedgehog [bib_ref] Xenopus sonic hedgehog as a potential morphogen during embryogenesis and thyroid hormone-dependent..., Stolow [/bib_ref]. Also, overexpression of TR/RXR in embryos by mRNA injection precociously repressed these genes to some extent [bib_ref] Both thyroid hormone and 9-cis retinoic acid receptors are required to efficiently..., Puzianowska-Kuznicka [/bib_ref]. Quantitative chromatin immunoprecipitation analysis in vivo showed that high TR binding to the TR promoter is constitutive [bib_ref] Gene-specific changes in promoter occupancy by thyroid hormone receptor during frog metamorphosis...., Buchholz [/bib_ref] [bib_ref] Specific histone lysine 4 methylation patterns define TR-binding capacity and differentiate direct..., Bilesimo [/bib_ref] , and NCoR is recruited there in a TH-dependent manner [bib_ref] A dominant-negative thyroid hormone receptor blocks amphibian metamorphosis by retaining corepressors at..., Buchholz [/bib_ref] [bib_ref] Nuclear receptor corepressor recruitment by unliganded thyroid hormone receptor in gene repression..., Sachs [/bib_ref]. In functional studies involving NCoR, tail muscle injection of a dominant negative NCoR that blocks endogenous NCoR from binding nuclear receptors increased transcription from coinjected TR␣ and a reporter plasmid (TRE-tk-luciferase) [bib_ref] Nuclear receptor corepressor recruitment by unliganded thyroid hormone receptor in gene repression..., Sachs [/bib_ref]. Also, transgenic overexpression of a dominant-negative NCoR significantly accelerated metamorphic progression [bib_ref] A role of unliganded thyroid hormone receptor in postembryonic development in Xenopus..., Sato [/bib_ref]. On the other hand, TR binding to the TH/basic leucine zipper domain (bZIP) promoter is low or not detectable in the absence of T 3 [bib_ref] Gene-specific changes in promoter occupancy by thyroid hormone receptor during frog metamorphosis...., Buchholz [/bib_ref] [bib_ref] Specific histone lysine 4 methylation patterns define TR-binding capacity and differentiate direct..., Bilesimo [/bib_ref] , yet NCoR is recruited to TH/bZIP in premetamorphosis, even though TR is not well bound [bib_ref] Nuclear receptor corepressor recruitment by unliganded thyroid hormone receptor in gene repression..., Sachs [/bib_ref]. Also, in the transgenic overexpression of a dominant negative NCoR, TR and TH/bZIP showed no significant derepression, and ST3 and sonic hedgehog showed derepression but not at all time points [bib_ref] A role of unliganded thyroid hormone receptor in postembryonic development in Xenopus..., Sato [/bib_ref]. Thus, available evidence for a role for unliganded TRs in premetamorphosis is based on expression correlations, nonendogenous conditions, and results with NCoR that may not be due to direct interactions with TR. Nevertheless, a role for unliganded TR has not been ruled out. Direct experimental evidence in vivo on endogenous gene regulation and development is required to unequivocally demonstrate a role for unliganded receptor in frogs and reveal potential developmental pathologies in the hypothyroid condition.
In addition to the repression of metamorphic genes and tissue competence, the premetamorphic TR may also contribute to tissue sensitivity/responsivity. Sensitivity refers to a tissue's competence to respond to a given TH concentration, and responsivity refers to a tissue's degree or magnitude of the effect of TH on the tissue at a given TH concentration and time point. Sensitivity and responsivity can be interrelated because of potential similar underlying mechanisms, eg, the expression level of TR. Indeed, TR levels correlate with sensitivity/responsivity to TH among larval tissues [bib_ref] Tadpole competence and tissue-specific temporal regulation of amphibian metamorphosis: roles of thyroid..., Shi [/bib_ref]. Furthermore, the overexpression of TR increases cell sensitivity and responsivity to TH [bib_ref] Higher thyroid hormone receptor expression correlates with short larval periods in spadefoot..., Hollar [/bib_ref] [bib_ref] Regulation of thyroid hormone sensitivity by differential expression of the thyroid hormone..., Nakajima [/bib_ref]. However, the experimental reduction of tissue sensitivity by the reduction of TR expression levels has not previously been accomplished to determine the requirement for TR␣ in tissue sensitivity or responsivity. Here, as presaged previously in the quote above, we made use of the recently developed gene disruption technology, transcription activator-like effector nucleases (TALENs) [bib_ref] High efficiency TALENs enable F0 functional analysis by targeted gene disruption in..., Suzuki [/bib_ref] , to disrupt TR␣ and provide direct and dramatic evidence of the role of unliganded TR␣ in tissue responsivity, gene repression, and developmental timing.
# Materials and methods
## Tr␣ talens and mrna synthesis
TALEN plasmids were assembled using the Platinum Gate TALEN kit (Addgene; catalog number 1000000043) as described previously [bib_ref] Repeating pattern of non-RVD variations in DNA-binding modules enhances TALEN activity, Sakuma [/bib_ref]. The ptCMV-153/47-VR-HD and ptCMV-153/47-VR-NG vectors were used as destination vectors for the left and right TALENs, respectively. Activity of constructed TALEN plasmids were validated by the human cell-based single-strand annealing assay as previously described [bib_ref] Efficient TALEN construction and evaluation methods for human cell and animal applications, Sakuma [/bib_ref]. TALEN mRNAs were synthesized from plasmid DNA encoding TR␣ TALENs (right and left arms) linearized with XmaI followed by lithium chloride precipitation using T7 mMESSAGE mMACHINE (Life Technologies). To monitor embryo injections, mCherry mRNA from KpnI-linearized CS108-mCherry (a gift from Dr M. Khokha) was also prepared with T7 and lithium chloride precipitation. A mixture of TR␣ TALEN mRNAs (400 pg for the left arm, 400 pg or the right arm), and mCherry mRNAs (25 pg) were injected into each embryo.
## Animals and microinjection
To induce breeding and obtain offspring, female and male Xenopus tropicalis from the laboratory colony were primed with 20 U of human chorionic gonadotropin (Sigma-Aldrich) 14 -16 hours before being boosted with 200 U. Freshly fertilized eggs from mated breeding adults in reconstituted reverse osmosis water were collected and dejellied for 5-10 minutes in 3% L-cysteine (Sigma-Aldrich) in 0.1ϫ modified Barth's solution (MBS)and transferred to 3% Ficoll in 0.1ϫ MBS. Dejellied embryos were injected with mRNA into one cell at the two-cell stage. After 3-5 hours, the surviving embryos were transferred to 0.01ϫ MBS. After 2 days, tadpoles hatched and were sorted under a fluorescence dissecting microscope into left side and right side groups based on mCherry fluorescence. Hatched tadpoles were fed finely ground frog brittle (Nasco) twice daily and reared at 26°C with daily water changes. Tadpoles were staged according to Nieuwkoop and Faber (NF). The sex of the tadpoles used is unknown because gonad differentiation begins at NF stage 54 in Xenopus, prior to most of our experiments. Some F1 offspring were reared in 0, 2, or 10 nM T 3 (Sigma) without feeding or 1 mM methimazole (Sigma) with feeding, all with daily water changes. Fluorescence with an RFP filter set and bright-field images were taken using a Leica MZ 16F fluorescence dissection microscope and a Leica DFC420 digital camera. The use of animals in the study was approved by the University of Cincinnati Institutional Animal Care and Use Committee (protocol number 06-10-03-01).
# Genetic analysis
Genomic DNA was prepared from the whole body or tail tip using Quick-gDNA Miniprep (Zymo Research). Up to 500 ng of genomic DNA was used as template for PCR to amplify the region surrounding the TR␣ TALEN target site (Takara Taq; Takara Bio Inc) that contained 0.2 M of forward (5Ј-GGTTTCTTYCGCCGCACCA) (Y stands for C or T) and reverse (5Ј-ATCCATTGCCATGCCAACGG) primers with the following reaction conditions: 94°C for 5 minutes, 33 cycles at 94°C for 30 seconds, 55°C for 30 seconds, and 72°C for 30 seconds. PCR products were spin column purified (Qiagen), cloned into TOPO T vectors (Invitrogen), and sequenced.
## Quantitative real-time pcr (qpcr)
Tadpole snout-to-vent length, hind limb length, and NF stage of hind limb were measured, and the whole body was snap frozen and stored at Ϫ80°C until RNA extraction (TriZol; Invitrogen). One mi-crogram of total RNA was used for cDNA synthesis (high-capacity cDNA reverse transcriptase kit; Applied Biosystems), and one microliter of cDNA was used for each 20-L qPCR (TaqMan 2ϫ universal PCR master mix; Applied Biosystems) with a qPCR primer/probe set (Applied Biosystems) on a 7300 Real Time PCR system with default reaction conditions (50°C for 2 min and 95°C for 10 min and then 40 cycles of 95°C for 10 sec and 60°C for 1 min) (Applied Biosystems). Primer/probe sets designed to span a large intron were used for ST3 (forward: 5Ј-GTCAACCAGGTGGAAAATGAGAGTA, reverse: 5Ј-CACGAATTGTAGACACTGCATCAAA, probe: 5Ј-CATG-CATCAGGCTCTGC), Krüppel-like factor 9 (KLF9) (forward: 5Ј-CCTTAAAGCCCATTACAGAGTCCAT, reverse: 5Ј-GCAG TCAGGCCACGTACA, probe: 5Ј-ACAGGTGAACGCCCTTTT), TR (forward: 5Ј-CAAGAGTTGTTGATTTTGCCAAAAA, reverse: 5Ј-ACATGATCTCCATACAACAGCCTTT, probe: 5Ј-CT-GCCATGTGAAGACC), TR␣ (forward: 5Ј-CCACTGGAAA-CAGCGTAGGA, reverse: 5Ј-CATGGGAGACTGCCCGATAT, probe: 5Ј-CTTCCGGCAGAAACT), and rpL8 (forward: 5Ј-CACAATCCTGAAAC-CAAGAAAACCA, reverse: 5Ј-CCACAC-CACGGACACGT, probe: 5Ј-AAGGC-CAAGAGAAACT). Two technical replicates were used per sample or notemplate control. The ddCt methodwas used to analyze the qPCR results. A one-way ANOVA was performed to test for significant differences among treatments and/or genotypes (JMP statistical software; SAS Institute).
# Results
## Talen design and construction
To create TR␣ knockout frogs, TALENs were designed to target the DNA binding domain of the TR␣ gene of X. tropicalis [fig_ref] Figure 1: Diagram of TR␣ TALEN design [/fig_ref]. The TALEN target site is located in between two zinc finger domains followed by the D domain (hinge region) and the E/F domain (ligand binding domain) of TR␣ [bib_ref] The D domain of the thyroid hormone receptor a1 specifies positive and..., Lee [/bib_ref] [bib_ref] Structural determinants of nuclear receptor assembly on DNA direct, Rastinejad [/bib_ref]. The single-strand annealing assay was carried out to evaluate TALEN targeting efficiency, and a TR␣ TALEN set was confirmed to have high disruption activity (data not shown). TR␣ TALEN mRNA was injected into fertilized eggs and showed a high survival rate.
## Phenotypes in f0 tr␣ mutant animals
To visualize the potential phenotypic effects of TR␣ mutations, we injected an mRNA mixture of the two TR␣ TALEN arms and mCherry into one cell of two cell-stage embryos to disrupt TR␣ on the left or right side of the tadpole body. Embryos were sorted on the third or fourth day (beginning of feeding) into the left side or right side groups based on mCherry expression [fig_ref] Figure 2: TR␣ TALEN founder developmental phenotypes [/fig_ref]. We observed phenotypic differences between the injected and noninjected sides of the body after 2 weeks when hind limb buds appeared at NF stage 46 -48 [fig_ref] Figure 2: TR␣ TALEN founder developmental phenotypes [/fig_ref]. The hind limb on the injected side developed faster than one on the noninjected side in 71% (99 of 139 from several injection experiments) of the injected embryos. In the remaining 29%, no difference in the hind limb was observed. When the noninjected side reached NF stage 49, the injected side averaged 1-1.5 NF stages higher [fig_ref] Figure 2: TR␣ TALEN founder developmental phenotypes [/fig_ref]. By the beginning of metamorphic climax 6 -8 weeks after injection (NF stage 58 -60), most of these F0 animals came to have similar hind limbs on both sides of the body, although some F0 frogs maintained asymmetric hind limbs (data now shown). Stage differences in forelimb development mirrored differences in hind limbs between injected and noninjected sides, but forelimbs on the mutant side emerged from the operculum later than on the noninjected side (data not shown). In addition to limbs, distinctive differences were also found in skin development at later stages of development between noninjected and injected side with a frequency of 20% (7 of 35 surviving tadpoles not used in experiments) [fig_ref] Figure 2: TR␣ TALEN founder developmental phenotypes [/fig_ref]. Specifically, at NF stage 56 -57, the nostril, upper jaw, cranium, trunk, hind limb, and forelimb within the operculum on the injected side began to show precocious thicker and more pigmented skin compared with the noninjected side. The skin phenotype frequency was less than for the hind limb because it is less distinctive than standardized morphological limb staging criteria of Nieuwkoop and Faber.
## Production of f1 offspring and mutation analysis
To generate nonmosaic tadpoles with both alleles of TR␣ disrupted, we reared 17 female and 10 male F0 founders that had asymmetric hind limbs as tadpoles. These founders were bred one or two times to each other to obtain a total of 11 clutches (Supplemental [fig_ref] Table 1: Amino Acid Sequence Alignment at TR␣ TALEN Target Site [/fig_ref]. As expected based on results from the F0 tadpoles, we observed two classes of individuals based on their hind limb phenotype when the tadpoles were 1-2 weeks old, in which some offspring from each of the 11 clutches had more developed fore/hind limbs than expected, given their small body size, when examined near the start of feeding [fig_ref] Figure 3: Hind limb phenotype in F1 offspring [/fig_ref]. The frequency of this mutant hind limb phenotype ranged from 2.0% to 40% across all clutches (Supplemental Table 1). The same phenotype was seen in tadpoles reared since feeding (before development of follicles in thyroid gland) in 1 mM methimazole that blocks TH synthesis (data not shown). From three F1 clutches, we sequenced the TR␣ TALEN target recognition site region from three individuals from each clutch with a normal and mutant phenotype (comprising a total of 79 successful sequencing runs after TOPO cloning) (Supplemental . We found that in nine of nine cases, individuals with the normal hind limb phenotype had at least one wild-type TR␣ allele, and in nine of nine cases, individuals with the mutant phenotype had two disrupted TR␣ alleles. We found no in-frame mutations in mutant phenotype individuals, but we found one in-frame deletion (six base pairs) in one allele of a normal hind limb individual [fig_ref] Table 1: Amino Acid Sequence Alignment at TR␣ TALEN Target Site [/fig_ref]. All mutations occurred at the TALEN target site between the two zinc fingers of the DNA binding domain, and the frame-shift mutations would abolish the second zinc finger and replace it with 40 -50 mutant amino acids followed by an out-of-frame stop codon [fig_ref] Table 1: Amino Acid Sequence Alignment at TR␣ TALEN Target Site [/fig_ref]. These results [fig_ref] Table 1: Amino Acid Sequence Alignment at TR␣ TALEN Target Site [/fig_ref] ] was isolated to analyze mRNA expression before TH circulation begins at NF stages 54 -55. HL phenotype animals had higher expression levels of TR (A), ST3 (B), and KLF9 (C) mRNA than normal animals. D, The TALEN-induced mutation in TR␣ had no effect on its mRNA expression levels, which were not significantly different in normal and hind limb phenotype animals (n ϭ 4 -6); bars show expression levels relative to the housekeeping gene rpL8, and error bars represent SD. Significance levels for a one-way ANOVA were: *, P Ͻ .05; **, P Ͻ 0.01; ***, P Ͻ 0.001. suggest that the mutant phenotype is distinctly recognizable and occurs only upon disruption of both alleles of TR␣. Also, we detected five to seven TR␣ mutant alleles from a single F0 breeding pair, indicating substantial mosaicism present in the germline of founder animals. Presumably, other tissues in founders were mosaic, justifying the need to produce F1 offspring for confirmation of the mutant phenotype and for further analysis of the effects of the TR␣ mutation.
## The absence of tr␣ releases gene repression and decreases tissue responsivity
The dual-function model hypothesizes that the lack of TR␣ would derepress TH-response genes in premetamorphosis in the absence of T 3 . We measured the mRNA levels of three TH-response genes, TR, ST3, and KLF9, in sibling normal or mutant phenotype tadpoles. Tadpoles from four clutches were collected, in which mutants had hind limbs at NF stage 50 -51 and normal tadpoles had hind limbs at NF stage 48 -49 and all had similar snout vent length as in [fig_ref] Figure 3: Hind limb phenotype in F1 offspring [/fig_ref]. Mutant phenotype tadpoles expressed higher levels of TR, ST3, and KLF9 mRNA than normal phenotype tadpoles in the absence of exogenous TH [fig_ref] Figure 4: Derepression of TH-response genes in hind limb phenotype tadpoles [/fig_ref] , A-C). We also measured TR␣ mRNA levels using primers to exons in the ligand binding domain and found no difference between normal and mutant phenotype individuals in TR␣ mRNA levels [fig_ref] Figure 4: Derepression of TH-response genes in hind limb phenotype tadpoles [/fig_ref].
In addition to the role for unliganded TR predicted by the dual-function model, another role is to render tissues sensitive and/or more responsive to the TH signal. To address this issue, we measured gene expression and observed morphological changes in sibling normal and hind limb phenotype tadpoles at NF stage 48 -49 and 50 -51, respectively, treated with exogenous T 3 . As predicted, levels of TR, ST3, and KLF9 mRNA were higher in normal compared with hind limb phenotype animals after treatment with exogenous T 3 at both concentrations used (2 and 10 nM) [fig_ref] Figure 5: Impaired induction of TH-response genes in hind limb phenotype tadpoles [/fig_ref]. Interestingly, T 3 increased the expression levels of these TH-response genes 2-fold or more in hind limb phenotype animals compared with the 0 nM T 3 treatment, suggesting functional TR in early larval stages. However, given that we did not sequence verify every individual, we cannot rule out residual TR␣ activity from a hypomorphic allele.
To compare tissue sensitivity/ responsivity at the morphological level, we compared responses of hind limb and gill in normal and mutant phenotype sibling tadpoles at the same size and age (age matched) as well as with larger, older normal tadpoles of the same hind limb stage (stage matched) after 7 days of exogenous T 3 treatment. Hind limbs of the age-matched normal tadpoles advanced in development from NF stage 48 to NF stage 51 in the presence of T 3 but not in the absence of T 3 . Hind limbs of the stage-matched normal tadpoles advanced in development from NF stage 53 to NF stage 54 in the absence of T 3 and to NF stage 57 in the presence of T 3 , bottom row). Hind limbs in the mutant phenotype tadpoles advanced from NF stage 53 to NF stage 55 without exogenous T 3 , which was one stage higher than the stage-matched control, and exhibited very little re- . Reduced hind limb responsivity to TH in hind limb phenotype tadpoles. Normal and hind limb (HL) mutant phenotype tadpoles from the same clutch were treated with 0 or 10 nM T 3 for 7 days. Hind limbs in mutant phenotype tadpoles (middle row) were compared with normal tadpoles of the same age (top row, age matched) or at the same stage (bottom row, stage matched). In the absence of T 3 , hind limbs advanced up to two stages in 7 days in normal and mutant phenotype tadpoles, (compare day 0-T 3 and day 7-T 3 ). In the presence of T 3 (compare day 0-T 3 and day 7ϩT 3 ), hind limbs in normal tadpoles advanced from NF stage 48 to NF stage 51 (top row) or from NF stage 53 to NF stage 57 (bottom row), whereas hind limbs of mutant phenotype tadpoles advanced from NF stages 53 to NF stage 55. Importantly, the effect of T 3 in mutant hind limbs (no difference in stage but slight increase in hind limb length) was greatly reduced compared with the effect in normal tadpoles of the same stage (compare day 7-T 3 and day 7ϩT 3 ). Hind limb images were photographed with the same magnification. This experiment was repeated with a sample size of four to six with similar results. doi: 10.1210/en.2014-1554 endo.endojournals.org sponse to T 3 , ie, the length of the hind limb increased slightly due to T 3 treatment but toe differentiation was not induced , middle row). For the gills, 7 days of treatment with 10 nM T 3 caused complete gill resorption in age-matched and stage-matched normal tadpoles at NF stage 48 and NF stage 53, but the process was not quite complete in the mutant phenotype tadpoles [fig_ref] Figure 7: Figure 7 [/fig_ref]. In the absence of T 3 , gill morphology did not change during the 7-day treatment period in normal or mutant phenotype tadpoles (data not shown).
# Discussion
Two nonmutually exclusive roles of unliganded TR␣ in the absence of TH during premetamorphosis have been proposed: 1) repress TH target genes required to initiate metamorphosis and 2) regulate tissue sensitivity/responsivity to TH underlying the timing of developmental events. Here we used TALEN gene disruption technology to disrupt TR␣ and provide direct evidence for both of these roles in frog development. We found higher expression of TH-response genes in the absence of TH, decreased responsivity to induction by TH, and precocious initiation of developmental progression in tadpoles with both TR␣ alleles disrupted. We targeted a region in between the two zinc fingers in the DNA binding domain of TR␣, and the disruption was sufficiently efficient to produce a high frequency of tadpoles with mutant phenotypes in the injected F0 individuals. However, the range in degree of hind limb phenotypes in founder animals suggested mosaic gene disruption among cells and/or production of mutant TR␣ proteins of altered function rather than null mutations. The 11 clutches of F1 offspring from paired founders exhibited variable frequencies of germline transmission (2.0%-40%), suggesting mosaicism among germ cells in gonads of founder animals. Nevertheless, a consistent hind limb phenotype within F1 offspring (discussed in more detail below) was observed. This mutant phenotype perfectly coincided with frame-shift mutations between the two zinc fingers and subsequent mutant amino acids and outof-frame stop codon, which are not known to leave TR␣ with any residual activity. Mutations in the second zinc finger abrogate effects on transcription [bib_ref] Second zinc finger mutants of thyroid hormone receptor selectively preserve DNA binding..., Nagaya [/bib_ref] and heterodimerization is required for efficient binding to DNA [bib_ref] Heterodimerization preferences of thyroid hormone receptor alpha isoforms, Nagaya [/bib_ref]. The truncations in our study eliminate the second zinc finger and the dimerization domains, suggesting complete loss of DNA binding and gene regulation. In mammals, TR␣1 is the bona fide receptor, with TR␣2 and TR␣3 as the known splice variants (41), TR␣2 and TR␣3 do not bind TH and may act as weak dominant negatives. Such variants, if present in frogs, would be disrupted in our tadpoles. Also, in mammals, TRa⌬1 and TRa⌬2 are generated from an internal promoter with a transcription start site at an exon 3Ј of our TALEN-induced mutations and cannot bind DNA or TH and have limited tissue distribution [bib_ref] Identification of transcripts initiated from an internal promoter in the c-erbA␣ locus..., Chassande [/bib_ref]. These isoforms, if present in frogs, would not be expected to have a significant impact on external morphology or whole-body measurements of gene expression and thus would not alter the conclusions here about TR␣mediated gene regulation. We cannot rule out off-target effects of the TR␣ TALENs or some transactivation activity of the A/B domain of TR␣ that would remain intact, but we note that the phenotypes observed to date fully comply with expectations regarding TR␣ mutations.
TH induces metamorphosis by up-regulating TH-response genes, including TR, ST3, and KLF9 measured herein [bib_ref] A correlation of thyroid hormone receptor gene expression with amphibian metamorphosis, Yaoita [/bib_ref] [bib_ref] Transcriptional activation of the matrix metalloproteinase gene stromelysin-3 coincides with thyroid hormone-induced..., Patterton [/bib_ref] [bib_ref] Basic transcription element binding protein is a thyroid hormone-regulated transcription factor expressed..., Hoopfer [/bib_ref]. The dual-function model predicts that Reduced gill responsivity to TH in hind limb phenotype tadpoles. Normal and hind limb (HL) mutant phenotype tadpoles were the same individuals as in . Gills in mutant phenotype tadpoles (middle row) were compared with normal tadpoles of the same age (top row, age matched) or at the same stage (bottom row, stage matched). In the presence of T 3 , gill resorption in normal tadpoles (top and bottom rows) occurred to a greater extent than in mutant phenotype tadpoles (middle row) (the clear area seen in mutant phenotype tadpoles is larger than in normal tadpoles, near white brackets). Gill images were photographed with the same magnification. This experiment was repeated with a sample size of four to six with similar results.
these same genes will be repressed by TR in the absence of TH to delay metamorphosis until endogenous TH reaches circulation [bib_ref] Molecular and developmental analyses of thyroid hormone receptor function in Xenopus laevis,..., Buchholz [/bib_ref]. We found increased expression of these genes in mutant phenotype animals in the absence of TH before metamorphosis. At the level of morphology, precocious limb development in early larvae is consistent with derepressed levels of TH-response genes responsible for metamorphic initiation. The same phenotype after methimazole treatment rules out the possibility that TR␣ disruption caused a precocious increase in TH production to explain the increase in TH-response gene expression and precocious hind limb development. Thus, our data suggest that TR␣ represses TH-response genes during premetamorphosis to delay hind limb development prior to TH in circulation. In addition to derepression, mutant phenotype animals revealed the second predicted role for unliganded TR␣, decreased sensitivity/responsivity to TH. In the absence of TR␣, gene induction by TH is expected to be absent or weak due to the low TR expression levels in premetamorphic tadpoles. As predicted, the expression levels of TR, ST3, and KLF9 in mutant phenotype tadpoles were significantly lower than those in normal tadpoles treated with exogenous T 3 , indicating impaired gene induction in the mutant phenotype tadpoles.
Even though all tissues of the tadpole require TH for transformation, we observed precocious development only in the hind limb in mutant phenotype tadpoles. In general, hind limb buds come out at an early stage (NF stage 46) and grow and develop slowly until NF stage 54 when hind limb growth becomes dependent on TH and the tadpole has increased its body size 3-fold. Hind limbs are known to express high levels of TR␣ early (NF stage 52), compared with other organs [bib_ref] Tadpole competence and tissue-specific temporal regulation of amphibian metamorphosis: roles of thyroid..., Shi [/bib_ref] [bib_ref] Expression of type II iodothyronine deiodinase marks the time that a tissue..., Cai [/bib_ref]. In addition, TR is detectable at later stages in the hind limb by immunocytochemistry, restricted to regions of cartilage-forming cells [bib_ref] An immunocytochemical analysis of the expression of thyroid hormone receptor ␣ and..., Fairclough [/bib_ref]. Thus, it is consistent that the absence of TR␣ would be first and best revealed in the hind limb. The gills retained to a great extent the ability to resorb after T 3 treatment in mutant phenotype tadpoles, indicating TR␣ may not play as dominant a role in gill compared with hind limbs. Indeed, the TR-selective agonist GC-1 caused complete gill resorption but less hind limb growth and elongation compared with T 3 , whereas a TR␣-selective agonist CO23 induced hind limb growth preferentially over gill resorption [bib_ref] Induction of larval tissue resorption in Xenopus laevis tadpoles by the thyroid..., Furlow [/bib_ref] [bib_ref] Design and characterization of a thyroid hormone receptor ␣ (TR␣)-specific agonist, Ocasio [/bib_ref]. The remaining low responsivity to TH in hind limb and substantial responsivity in gill in the absence of TR␣ is most likely due to premetamorphic TR expression and/or nongenomic actions of TH.
The unliganded function of TR has been shown in the cerebellum, heart, and under hypothyroid conditions in the cochlea in mammals (5-7). As in tadpoles, the expres-sion of TR before the onset of thyroid gland in mammalian fetuses has engendered the idea that TR functions in the unliganded state. Although TH may pass through the placenta, it is unlikely that all or most TRs in the fetus are occupied by TH [bib_ref] Do unliganded thyroid hormone receptors have physiological functions?, Chassande [/bib_ref]. The lack of known widespread actions of unliganded TR in mammals may be because low levels of TH are present in the fetus such that effects of unliganded TR␣ may appear only in hypothyroid conditions as in the cochlea. Because frogs exist for a long period of time in the absence of TH (3 wk of premetamorphosis), TH-dependent development in TR␣-disrupted frogs may be a good model to reveal the effects of unliganded TR difficult to uncover in the mammalian model.
[fig] Figure 1: Diagram of TR␣ TALEN design. A, Binding regions for left and right arms of TR␣ TALEN (bold nucleotides) flank a spacer region in which a double-stranded break (black triangle) catalyzed by FokI nuclease domains is repaired by nonhomologous end joining. Zinc finger domains, amino acids, and exon boundaries (E4, E5, E6) are indicated. B, TR␣ is composed of nine exons (E1-9), and the targeted region is within exon 5 (E5), part of the coding region for the DNA binding domain (DBD). Black bars in E1 and E9 represent 5Ј and 3Ј untranslated regions, respectively. A/B, C, D, E/F, nuclear receptor protein domains, LBD, ligand binding domain. [/fig]
[fig] Figure 2: TR␣ TALEN founder developmental phenotypes. A, mCherry mRNA was coinjected with TR␣ TALEN mRNA into a single blastomere of two-cell stage embryos. Ventral view of the injected tadpole shows mCherry expression only on the left side of its body. B and C, Hind limb and skin phenotype are shown. B, Ventral view of the lower abdomen and proximal part of tail shows different developmental progression in the hind limbs between uninjected and injected sides. White arrows point to side views of the hind limb on each side to show the alteration of not just limb length but stage of limb development. C, Skin on the injected side of a NF stage 57 tadpole shows advanced iridophore development (gold skin color, black triangles). Thicker skin obscures the ultimobranchial body visible only on the uninjected side (white triangle). The dotted line delineates the uninjected and injected sides. D, Histogram of hind limb stage difference in TR␣ TALEN founders. When the stage of hind limb development on the uninjected side was NF stage 49, the difference in stage compared with the injected side was determined and plotted in the histogram. The hind limb on the injected side was always more advanced, typically by 1 to 1.5 stages (n ϭ 42 injected animals from the same clutch). Tadpoles that lacked a stage difference were not quantified. HL, hind limb. [/fig]
[fig] Figure 3: Hind limb phenotype in F1 offspring. Representative sibling offspring from a pair of TR␣ TALEN founders were imaged at feeding stage (upper panel). The developmental difference in hind limbs is shown in the lower panels. The hind limbs are bracketed. HL, hind limb. [/fig]
[fig] Figure 4: Derepression of TH-response genes in hind limb phenotype tadpoles. Total RNA from whole bodies of F1 normal or hind limb (HL) phenotype tadpoles at NF stage 48 -49 and NF stage 50 -51, respectively, was isolated from four clutches [clutches6,7,9, and 11 (Supplemental [/fig]
[fig] Figure 5: Impaired induction of TH-response genes in hind limb phenotype tadpoles. Total RNA from whole bodies of sibling F1 normal or hind limb (HL) phenotype tadpoles at NF stages 48 -49 and NF stages 50 -51, respectively, was isolated after treatment with 0, 2, or 10 nM T 3 for 24 hours. The mRNA expression levels of TR (A), ST3 (B), and KLF9 (C) were significantly higher in normal tadpoles treated with 2 or 10 nM T 3 . Uninduced levels were increased for TR and KLF9 in HL phenotype tadpoles, as inFigure 4. n ϭ 4 -6, bars show expression levels relative to the housekeeping gene rpL8, and error bars represent standard deviation. One-way ANOVA showed significant differences for most comparisons, P Ͻ .05; *, Ͻ 0.01; **, Ͻ 0.001; ***. [/fig]
[fig] Figure 7: Figure 7. Reduced gill responsivity to TH in hind limb phenotype tadpoles. Normal and hind limb (HL) mutant phenotype tadpoles were the same individuals as in Figure 6. Gills in mutant phenotype tadpoles (middle row) were compared with normal tadpoles of the same age (top row, age matched) or at the same stage (bottom row, stage matched). In the presence of T 3 , gill resorption in normal tadpoles (top and bottom rows) occurred to a greater extent than in mutant phenotype tadpoles (middle row) (the clear area seen in mutant phenotype tadpoles is larger than in normal tadpoles, near white brackets). Gill images were photographed with the same magnification. This experiment was repeated with a sample size of four to six with similar results. [/fig]
[table] Table 1: Amino Acid Sequence Alignment at TR␣ TALEN Target Site [/table]
|
10.3390/cells10092442
|
CCBY
|
8467979
|
34572091
|
s2orc_pubmed_articles
|
Establishment and Characterization of a Novel Gill Cell Line, LG-1, from Atlantic Lumpfish (Cyclopterus lumpus L.)
Citation: Sindre, H.; Gjessing, M.C.; Fosse, J.H.; Hermansen, L.C.; Böckerman, I.; Amundsen, M.M.; Dahle, M.K.; Solhaug, A. Establishment and Characterization of a Novel Gill Cell Line, LG-1, from Atlantic Lumpfish (Cyclopterus lumpus L.). Cells 2021, 10, 2442. https://
# Introduction
Infestation with salmon lice (Lepeophtheirus salmonis) is one of the biggest challenges in farming of Atlantic salmon (Salmo salar L.). For years, bath treatment with anti-parasitic pharmaceuticals was the predominant treatment strategy. However, in the course of the past decades, a biological approach using cleaner fish feeding on lice from infested salmon has become more and more common. At present, Atlantic lumpfish (Cyclopterus lumpus L.) is the most commonly used cleaner fish species in Norwegian aquaculture, with about 43 million lumpfish used in 2019 . Lumpfish is considered particularly suitable for this use, as they continue feeding at low temperatures and have a relatively short life cycle that allows them to be introduced into salmon farms already 4 months after hatching [bib_ref] The effect of temperature and fish size on growth of juvenile lumpfish..., Nytrø [/bib_ref].
The extensive use of cleaner fish in salmon farming is in its infancy and is facing several challenges. A large proportion of cleaner fish die in the course of the production cycle. Yet, we have a limited understanding of the causes of this mortality, in contrast to the knowledge we have of other farmed fish, like the salmon. The farming conditions are designed for salmon production, and even if environmental enrichments are provided, the conditions represent a suboptimal environment for lumpfish, with health and welfare issues as a consequence. Increased knowledge is needed to be able to provide the best conditions for salmon and lumpfish cohabitation. The most severe welfare issues in cleaner fish today are related to handling stress and infectious diseases. The extensive cohabitation of two fish species at high densities is a biosecurity risk factor and facilitates the transmission of potential pathogens. Cross-species infections may pave the way for more virulent variants by increasing the chance of exposure and adaptation [bib_ref] Cross-Species Virus Transmission and the Emergence of New Epidemic Diseases. Microbiol, Parrish [/bib_ref] [bib_ref] Cleaner fish in aquaculture: Review on diseases and vaccination, Erkinharju [/bib_ref]. Outbreaks of the notifiable disease viral haemorrhagic septicaemia (VHS) have been reported in farmed lumpfish in Iceland [bib_ref] Outbreak of viral haemorrhagic septicaemia (VHS) in lumpfish (Cyclopterus lumpus) in Iceland..., Guðmundsdóttir [/bib_ref] , and a ranavirus similar to epizootic hematopoietic necrosis virus (EHNV) has been detected in lumpfish in Ireland, the Faroe Islands, Scotland and Iceland [bib_ref] Characterization of ranaviruses isolated from lumpfish Cyclopterus lumpus L. in the North..., Stagg [/bib_ref]. Cyclopterus lumpus virus (CLuV) was described in Norway in 2017 [bib_ref] New virus of the family Flaviviridae detected in lumpfish (Cyclopterus lumpus), Skoge [/bib_ref] and is associated with disease, severe liver lesions and high mortalities in lumpfish. A broad range of parasites have been identified in lumpfish, including both endo-and ectoparasites [bib_ref] Cleaner fish in aquaculture: Review on diseases and vaccination, Erkinharju [/bib_ref]. One of the latter is Paramoeba perurans, the causative agent of amoebic gill disease, leading to similar gill lesions as reported for Atlantic salmon [bib_ref] Cyclopterus lumpus L.) develop amoebic gill disease (AGD) after experimental challenge with..., Haugland [/bib_ref]. Altogether, it is crucial to gain a better understanding of which pathogens can be harboured on cleaner fish in order to design screening programs to plot the infection status of both species.
Fish cell lines are valuable tools for in vitro fish research and may serve as models to mimic complex in vivo biology. A repertoire of cell lines from the species and organ of interest allows controlled and reproducible experimental conditions and may replace or reduce the number of experimental animals in line with the "3Rs" principles [bib_ref] Russell and Burch's 3Rs then and now: The need for clarity in..., Tannenbaum [/bib_ref]. Fish cell lines serve as tools to study host cell-pathogen interactions, and for the isolation and characterization of viruses [bib_ref] Applications and needs of fish and shellfish cell culture for disease control..., Villena [/bib_ref]. Furthermore, they have been extensively used to study cellular responses and the prediction of acute toxicity as a part of the hazard assessment of chemicals [bib_ref] Repeatability and Reproducibility of the RTgill-W1 Cell Line Assay for Predicting Fish..., Fischer [/bib_ref] [bib_ref] Cell culture-based biosensing techniques for detecting toxicity in water, Tan [/bib_ref]. They can also be used to identify biomarkers as indicators of environmental pollutants, infection and disease [bib_ref] In vitro bioassays for detecting dioxin-like activity-Application potentials and limits of detection,..., Eichbaum [/bib_ref].
The main function of gills is respiration. However, the gills also carry out other important processes, including osmoregulation, the excretion of nitrogenous waste and immunological functions. Hence, gill diseases may lead to compromise on several physiological levels. Because most gill diseases are not notifiable, the true extent and economic loss in aquaculture connected to compromised gill function is not known [bib_ref] Complex Gill Disease: An Emerging Syndrome in Farmed Atlantic Salmon (Salmo salar..., Herrero [/bib_ref]. Only a few cell lines from gills have been established and published. One of them is the widely used gill epithelial cell line from rainbow trout (Oncorhynchus mykiss), RTgill-W1. Another gill epithelial cell line from Atlantic salmon, ASG10, was recently established in our lab [bib_ref] Development and characterization of two cell lines from gills of Atlantic salmon, Gjessing [/bib_ref]. Salmonids have a long evolutionary distance to lumpfish [bib_ref] Genome evolution and biodiversity in teleost fish, Volff [/bib_ref] ; hence, cell lines from Atlantic salmon and rainbow trout are not reliable tools to study aspects of lumpfish biology. Apart from a fibroblastic cell line from lumpfish fin, characterized in 1977 [bib_ref] New Cell Line from the Marine Lumpfish, Cyclopterus lumpus, Li [/bib_ref] , no cell lines from lumpfish have been reported.
In this study, we have addressed this gap and report the development and characterization of a novel gill cell line from lumpfish (LG-1). In the long term, our ambition is that LG-1 may serve as a tool to gain more insight into lumpfish biology and generate knowledge that can lead to a more robust lumpfish population in aquaculture.
# Materials and methods
## Animal husbandry and ethical considerations
Gills for the development of primary cells were obtained from an adult farmed lumpfish cultivated at a Norwegian commercial lumpfish farm. The fish was euthanized with an overdose of MS222 (Sigma-Aldrich, St-Louis, MO, USA. Two whole gill arches were then removed and incubated in medium consisting of Leibovitz-15 (L-15, Lonza, Basel, Switzerland) culture media with 10,000 units/mL penicillin, 10.0 mg/mL streptomycin (1% Pen/Strep; Lonza), 1 µg/mL Amphotericin B (Thermo Fisher Scientific, Waltham, MA, USA) and 20% fetal bovine serum (FBS superior, Biochrom, Cambridge, UK) and processed as described in Section 2.2 on the same day the tissue was harvested. For transmission electron microscopy, another fish was euthanized and the gills were processed as described in Section 2.7.
## Development of primary cells
The gill arches were incubated twice for 10 min in a sterile 50 mL tube containing 40 mL Hanks' balanced salt solution (HBSS, Lonza) supplemented with 0.05 mg/mL gentamycin (Lonza) with constant gentle rotation (30-60 rpm). The gill arch was then placed in a sterile petri dish and the cartilage of the arch was removed. A few droplets of supplemented L-15 medium: L-15 (Lonza) with an addition of 0.1 mM of non-essential amino acid, (Lonza), 1 mM sodium-pyruvate (Lonza), 0.01 mg/mL insulin, 0.01 mg/mL transferrin and 0.01 µg/mL selenium (ITS, Lonza), 4 mM L-glutamine (Lonza), 0.05 mg/mL gentamycin (Lonza), 0.03 mM 2-mercaptoethanol (Gibco ™ , Thermo Fisher) and 20% FBS, were added to the gill filaments. The filaments were cut by a scalpel in explants of 1-2 mm and about 2-4 explants were plated in 25 cm 2 cell bind flasks (Corning, New York, NY, USA) with 1 mL of supplemented L-15 medium to allow the tissue pieces to adhere. After one day, when the explants had adhered to the surface of the flasks, 2 mL of supplemented L-15 medium were gently added. Then, 2 days later, adherent cells were confirmed under the micropscope and the medium was changed to remove gill residuals and dying cells. The cells were then grown further at 15 - C for 4 weeks until a confluent monolayer was established [fig_ref] Figure 1: Lumpfish [/fig_ref]. The cells were then detached as followed: the cells were washed once with 5 mL PBS and 0.5 mL 0.25% trypsin/EDTA (Lonza) were added to the cells. After 10-15 min in room temperature, the cells detached and were passaged 1:2 to new cell culture flasks in 5 mL supplemented L-15 medium. The procedure was repeated several times until stable dividing cell cultures were established. In passage 7, the cell culture was tested negative for mycoplasma (Myco Alert; Lonza).
## Routine maintenance
The cells were maintained in complete cell culture medium: Leibovitz's L-15 Medium, GlutaMAX™ Supplement (Gibco™, Thermo Fisher) with an addition of 10% FBS, 100 units potassium penicillin and 100 µg streptomycin sulfate (1% pen/strep, Lonza). The cells were cultured at 20 - C, which gives a proliferation rate that allows for sub-cultivation 1:2 every second week. This is done as followed: the cells were washed with 10 mL PBS, and 2 mL 0.25% trypsin/EDTA were added to the cells. After 10-15 min in room temperature, the cells detach, then 10 mL of complete cell culture medium were added. 6 mL of the cell suspension, containing about 2.5 × 10 6 cells, were then transferred to a new 75 cm 2 flask. One confluent flask (75 cm 2 ) contains about 5.0 × 10 6 cells. For experiments, passage number 15-40 was used. No differences were observed between early and late passage numbers with regard to proliferation or morphology. The cells were seeded on standard plastic cell culture plates, 100,000 cells/cm 2 , resulting in a confluent cell layer the next day. The LG-1 cells can successfully be stored by cryopreservation. Here, 5 × 10 6 cells were suspended in 2.5 mL FBS. This cell suspension was then transferred to 1.8 mL Cryo Tube ™ Vials (Nunc ™ , Thermo Fisher), 0.5 mL/tube, and then 0.5 mL freezing medium (L-15 medium, 1% Pen/Strep, 10% FBS, 20% DMSO) were gently added to each tube. The cells were then placed in a Mr Frosty freezing container (Nalgene™, Thermo Fisher) at −80 - C, which allows for a gradual reduction in temperature. After 2 days, the cells were transferred to liquid N 2 for continued storage (−196 - C). For shorter storage periods, the cells (a confluent flask) can also be stored in the refrigerator at 4 - C. At this temperature, some cells will die, but after a short period at 15-20 - C, the remaining cells will start proliferating again.
## Species identification
RNA was extracted using RNeasy Mini kit (Qiagen, Hilden, Germany), following the manufacturers protocol. Briefly, 600 µL RLT lysis buffer (Qiagen) and 6 µL βmercaptoethanol (<99%) were used to lyse LG-1 cells in a 75 cm 2 cell culture flask. The lysate was homogenized using a 1 mL syringe. The quantity and purity of the extracted RNA were measured using a NanoDrop™ 2000 spectrophotometer (Thermo Scientific). For transcription analysis, cDNA was synthesized using 1000 ng of total RNA using a QuantiTect Reverse Transcription kit (Qiagen) with gDNA elimination, according to the manufacturer's instruction. The samples were incubated for 30 min at 42 - C to activate reverse transcription and then for three min at 95 - C to inactivate the reaction. After synthesis, the samples were frozen and stored at −20 - C. The qPCR was performed in duplicates with 5 ng of cDNA input in a total volume of 10 µL per reaction using SsoAdvanced™ Universal SYBR ® Green Supermix (Bio-Rad, Hercules, CA, USA). The thermal program was set to 95 - C for 30 s, 39 cycles of 95 - C for 15 s and 60 - C for 30 s. Primers for Atlantic lumpfish (Cyclopterus lumpus L.) interleukin 6 (IL-6) were designed using NCBI primer Blast™, and levels of Atlantic lumpfish (Cyclopterus lumpus L.) interleukin 6 (IL-6) and elongation factor 1 α (EF1α)were assessed using 10 µM primers. Primers targeting Atlantic salmon Salmo salar) Elongation factor 1α (ss-EF1a) [bib_ref] Intramuscular vaccination of Atlantic lumpfish (Cyclopterus lumpus L.) induces inflammatory reactions and..., Erkinharju [/bib_ref] were used as a negative control. Amplicon length of each qPCR product was controlled using a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) along with an associated kit, Agilent DNA 1000, according to the manufacturer's protocol. The primers used are shown in [fig_ref] Table 1: Primer over view [/fig_ref].
## Proliferation
The cells were seeded in a 96 well plate, 50,000 cells/cm 2 , and cultured for 1-14 days at different temperatures (20, 16, 10, 4 - C). For quantification, the cells were stained with DRAQ5 (Thermo Fisher; nuclear staining, 1:500) for 30 min at room temperature and the cell number in a specific area of the well was counted by the spectramax i3x plate reader equipped with a microscopic module (MiniMax300Imaging Cytometer, Molecular Devices, San Jose, CA, USA).
## Morphology
The cells were seeded in 6-well plates (100,000 cells/cm 2 ). At confluence, the cells were stained with Calcein-AM (1 µM; Sigma) and visualized by light and fluorescence microscopy (Axio Observer A1, Zeiss, Jena, Germany).
## Transmission electron microscopy (tem)
Lumpfish gills (from 2.1) were cut into 1 mm 2 under fixative (2% paraformaldehyde/1.25% glutaraldehyde/0.1 M cacodylate buffer) and stored at 4 - C until further processing, as described for the ASG10 cells [bib_ref] Development and characterization of two cell lines from gills of Atlantic salmon, Gjessing [/bib_ref]. The LG-1 cells were seeded on transwell inserts (Costar™ 0.4 µm polyester membrane, Sigma-Aldrich) or in a culture flask, 100,000 cells/cm 2 . After 5 days, the cells were washed once with PBS, the cells on the membrane were fixed directly on the membrane and the cells in the flask were scraped and pelleted by centrifugation (500× g, 10 min) prior to fixation. Fixation were done in 2% paraformaldehyde/1.25% glutaraldehyde/0.1 M cacodylate buffer for 15 min at room temperature, washed with 0.1 M sodium cacodylate buffer, embedded in 3% low-melting agarose and post-fixed in 1% osmium tetroxide in 0.1 M sodium cacodylate buffer for 1 h. Subsequently, the cells were washed thoroughly in 0.1 M sodium cacodylate buffer, dehydrated with 10 min steps in ascending ethanol series (50-100%) and embedded in LR White resin (London Resin Company, EMS, Agar Scientific, Stansted, UK). Ultrathin sections were obtained using a Leica EM UC6 Ultramicrotome (Leica, Wetzlar, Germany). The sections were stained with 4% uranyl acetate and 1% potassium permanganate for 10 min and examined and photographed using a FEI Morgagni 268 transmission electron microscope (FEI, Hillsboro, OR, USA). Contrast and white balance were adjusted in Adobe Photoshop software (Adobe systems, San Jose, CA, USA).
## Periodic acid-schiff (pas), alcian blue ph 1 staining and immunostaining for chloride cells
The cells were fixed for 10 min in 4% paraformaldehyde at room temperature, washed with PBS, stained with PAS and Alcian blue for the detection of mucus cells as described inand immunostained for chloride cells as, as previously described In short, the cells were washed gently in PBS, incubated for 20 min in Tris-buffered saline (TBS) with 2.5% bovine serum albumin (BSA) for prevention of non-specific binding, incubated at 60 min with primary antibody directed against a conserved region of Na + /K + ATPase subunit (a5, Developmental Studied Hybridoma Bank, University of Iowa, Iowa City, IA, USA and diluted 1:100 in TBS with 2.5% BSA. A biotinylated rabbit anti mouse secondary antibody and an alkaline phosphatase conjugated streptavidin system was used to visualize the binding. Paraffin embedded gills were used as positive controls and inspected under a Leica microsope.
## Immunostaining for cell markers
The cells were plated on Millicell EZ slides (Merck, Darmstadt, Germany). At confluence, the cells were washed once in PBS and fixed with 4% paraformaldehyde for 15 min at room temperature. For F-actin (phalloidin) staining, the cells were permeabilized with saponin (0.05%) for 10 min at room temperature and stained with phalloidin Alexa fluor 555 (#8953, 1:300, Thermo Fisher) for 30 min. The cells were then washed 3 times with PBS and coverslips mounted with prolong mounting medium. Confocal fluorescence microscopy was performed using a Zeiss LSM710 microscope. For recombinant mouse antipan cytokeratin (clone [C-11] (ab7753), 1:1000, Abcam, Cambridge, UK), purified mouse anti-E-Cadherin (clone 36/E-Cadherin, 1:1000, BD Biosciences, Franklin Lakes, NJ, USA), mouse anti-human ZO-1 (#33-9100, clone ZO1-1A12, 1:100, Thermo Fisher) and rabbit polyclonal anti-human von Willebrand factor (DAKO A0082, 1:1000, Agilent) staining, the cells were seeded and fixed as described above and permeabilized in ice cold methanol (100%). The cells were then washed 3 times with PBS and blocked in 5% BSA for 60 min followed by incubation with primary antibodies overnight at 4 - C. The cells were then rinsed 3 times and incubated with secondary antibody conjugated with Alexa Fluor 488 (Molecular probes, Thermo Fisher) for 2 h at room temperature. The cells were then washed 3 times with PBS, the nuclei were stained with DAPI (1:1000) and coverslips were mounted with a prolong mounting medium. Confocal fluorescence microscopy was performed using a Zeiss LSM710 microscope.
## Transepithelial/transendothelial electrical resistance (ter)
The cells were seeded on transwell inserts (Costar 0.4 µm polyester membrane; 100,000 cells/cm 2 ) and TER was automatically measured every 6 h, using CellZscope E (Nano Analytics, Münster, Germany).
## Cyp1a activity
The induction of CYP1A is measured by the EROD-assay in the presence of NADPH (β-nicotinamide adenine dinucleotide phosphate). CYP1A converts the artificial substrate EROD to resorufin, which can be measured via fluorescence spectroscopy. The assay was done as described in [bib_ref] Improving the in vitro ethoxyresorufin-O-deethylase (EROD) assay with RTL-W1 by metabolic normalization..., Heinrich [/bib_ref] with minor modifications. Briefly, the cells were seeded in black 96 well plates (100,000 cells/cm 2 ) and incubated for 24 h in culture medium. The cells were then incubated with the CYP1A inducer, beta-naphtaflavone (BNF; 1-100 nM; Sigma-Aldrich) for 24 h. The next day, the culture medium was replaced with 200 µL EROD assay media (DMEM w/o phenol red; Gibco™, Thermo Fisher, 10% FBS, 8 µM 7-Ethoxyresorufin; Sigma-Aldrich). After 30 min of incubation, resorufin fluorescence (Ex: 530 nm/Em: 580 nm) was quantified using a plate reader (Spectramax i3x plate reader, San Jose, CA, USA).
## Testing for susceptibility to fish viruses
To test for permissiveness to different viruses, LG-1 were incubated with viruses, as listed in [fig_ref] Table 2: Overview of virus isolates and primary antibodies [/fig_ref] [bib_ref] Outbreak of viral haemorrhagic septicaemia (VHS) in seawater-farmed rainbow trout in Norway..., Dale [/bib_ref] [bib_ref] Pathogenicity of nodavirus strains from striped jack Pseudocaranx dentex and Atlantic halibut..., Totland [/bib_ref]. In short, the cells were seeded in 96 well plates (Corning ® ) using complete cell culture medium containing 15% FBS and grown to approx. 80-90%. The growth medium was then removed and 50 µL of complete growth medium without FBS containing about 50 TCID 50 of each virus were added. Following 3-4 h of incubation at 15 - C, 150 µL of complete growth medium containing 10% FBS were added to each well. In addition, negative controls without the virus were included. The plates were incubated at 15 - C for 7 days for IHNV, IPNV, betanodavirus, ISAV and SAV1, 2 and 3. The incubation of the plate inoculated with VHSV was terminated at full CPE 4 days post infection. The cells and supernatants from wells inoculated with CLuV, were harvested 7, 14 and 21 days post-inoculation and tested for virus propagation by RT-qPCR, as described in [bib_ref] New virus of the family Flaviviridae detected in lumpfish (Cyclopterus lumpus), Skoge [/bib_ref]. For the other viruses, the medium was removed from all wells and IFAT was performed as described previously [bib_ref] Isolation in cell culture of nodavirus from farmed Atlantic halibut Hippoglossus hippoglossus..., Dannevig [/bib_ref]. In short, the cells were fixed with 80% acetone for 20 min, dried and stained with specific antibodies against the individual viruses as listed in [fig_ref] Table 2: Overview of virus isolates and primary antibodies [/fig_ref] [bib_ref] Characterization and mapping of monoclonal antibodies against the Sleeping disease virus, an..., Moriette [/bib_ref] [bib_ref] Experimental infection of Atlantic halibut Hippoglossus hippoglossus with nodavirus: Tissue distribution and..., Grove [/bib_ref] , and biotin labelled goat anti-mouse Ig (E 0433, Dako) and FITC-labelled streptavidin (11-4317, AH diagnostics, Tilst, Denmark) were used for the secondary and tertiary step, respectively. For the staining of betanodavirus, FITC-labeled goat anti-rabbit IgG (4030-02, Southern Biotech, Birmingham, AL, USA) was used as secondary antibody. The nuclei were visualized by staining with propidium iodide (Sigma-Aldrich). The stained preparations were evaluated by wide field inverted fluorescence microscopy (Leica DMIL).
For further investigation into susceptibility to SAV1, nodavirus, IHNV and VHSV, LG-1 and OIE-recommended cell lines for each virus type were seeded in 25 cm 2 flasks. At about 80%-confluence LG-1 was inoculated with 500 TCID 50 of SAV1, fish nodavirus, IHNV and VHSV, whereas OIE-recommended CHSE-214 (ATCC CRL-1681), E-11 (ECACC 01110916), EPC (ATCC CRL-2872) and BF-2 (ATCC CCL-91) were inoculated with the same amounts of SAV1, nodavirus, IHNV and VHSV, respectively. Following 14 days of incubation at 15 - C (except for E-11, which was incubated at 20 - C), the supernatant was harvested. The virus titration of cell supernatants following virus propagation was performed in 10-fold dilution series in 6 parallel wells with an incubation period of 7 days. The cells were then fixed and IFAT performed, as described above. A TCID 50 titer was then calculated according to the Spearman-Kärber method [bib_ref] Beitrag zur kollektiven Behandlung pharmakologischer Reihenversuche, Kärber [/bib_ref].
# Statistical analysis
The data analyses were performed using Sigma Plot version 12.0 (Systat Software, San Jose, CA, USA). Statistical significance (p < 0.05) was assessed using 1-way-ANOVA, followed by Dunnetts post-test. For analysis of cell proliferation data in 3.2, Normality test (Sharpiro-Wiik) failed. Kruskal-Wallis One Way Analysis of Variance on Ranks followed by Dunnetts post-test was therefore used.
# Results
## Development and maintenance of the lg-1 cell line
A primary cell line (LG-1) was established from lumpfish gill explants as illustrated in [fig_ref] Figure 1: Lumpfish [/fig_ref]. The species origin of the LG-1 cell line was determined by RT-qPCR using species-specific primers targeting the housekeeping mRNA encoding Elongation factor 1a from lumpfish (LumpEF1a) and Atlantic salmon (ssEF-1a) as negative control, as well as lumpfish interleukin 6 (LumpIL-6), a cytokine gene expressed by epithelial and endothelial cells [bib_ref] Expression and function of interleukin-6 in epithelial cells, Krueger [/bib_ref] [bib_ref] Interleukin-6 gene expression in human endothelial cells: RNA start sites, multiple IL-6..., May [/bib_ref]. Both lumpfish mRNAs, but not the Atlantic salmon housekeeping mRNA, were detected in the sample. The amplicon products were confirmed to be of correct length, according to the target sequences [fig_ref] Figure 2: Figure 2 [/fig_ref].
# Statistical analysis
The data analyses were performed using Sigma Plot version 12.0 (Systat Software, San Jose, CA, USA). Statistical significance (p < 0.05) was assessed using 1-way-ANOVA, followed by Dunnetts post-test. For analysis of cell proliferation data in 3.2, Normality test (Sharpiro-Wiik) failed. Kruskal-Wallis One Way Analysis of Variance on Ranks followed by Dunnetts post-test was therefore used.
# Results
## Development and maintenance of the lg-1 cell line
A primary cell line (LG-1) was established from lumpfish gill explants as illustrated in [fig_ref] Figure 1: Lumpfish [/fig_ref]. The species origin of the LG-1 cell line was determined by RT-qPCR using species-specific primers targeting the housekeeping mRNA encoding Elongation factor 1a from lumpfish (LumpEF1a) and Atlantic salmon (ssEF-1a) as negative control, as well as lumpfish interleukin 6 (LumpIL-6), a cytokine gene expressed by epithelial and endothelial cells [bib_ref] Expression and function of interleukin-6 in epithelial cells, Krueger [/bib_ref] [bib_ref] Interleukin-6 gene expression in human endothelial cells: RNA start sites, multiple IL-6..., May [/bib_ref]. Both lumpfish mRNAs, but not the Atlantic salmon housekeeping mRNA, were detected in the sample. The amplicon products were confirmed to be of correct length, according to the target sequences [fig_ref] Figure 2: Figure 2 [/fig_ref].
## Proliferation
To define optimal growth conditions for the LG-1 cells, we evaluated cell proliferation at different temperatures. We seeded cells at a density (50,000 cells/cm 2 ) that resulted in approximately 50% confluence after 1 day. After 1, 7 and 14 days, the cells were stained with the nuclear stain DRAQ5 and counted by a plate reader equipped with an imager module [fig_ref] Figure 3: Cell proliferation [/fig_ref]. When grown at 20 °C, cell numbers doubled within 14 days. At 16 °C, the cells grew at similar rates as at 20 °C for the first 7 days but then reduced their proliferation rate. At 10 °C, no change in cell numbers was observed at any time point, and at 4 °C, cell numbers declined, probably due to cell death, from days 1 to 7, but remained constant between days 7 and 14. Control of lumpfish origin. The origin of the LG-1 cells was controlled by RT-qPCR on total RNA using lumpfish-specific primers targeting IL-6 and EF1α, and Atlantic salmon specific primers targeting EF1α were used as a negative control. The correct amplicon length was confirmed by capillary electrophoresis (Bioanalyzer). Lump: Lumpfish. Ss: Salmo salar.
## Proliferation
To define optimal growth conditions for the LG-1 cells, we evaluated cell proliferation at different temperatures. We seeded cells at a density (50,000 cells/cm 2 ) that resulted in approximately 50% confluence after 1 day. After 1, 7 and 14 days, the cells were stained with the nuclear stain DRAQ5 and counted by a plate reader equipped with an imager module [fig_ref] Figure 3: Cell proliferation [/fig_ref]. When grown at 20 - C, cell numbers doubled within 14 days. At 16 - C, the cells grew at similar rates as at 20 - C for the first 7 days but then reduced their proliferation rate. At 10 - C, no change in cell numbers was observed at any time point, and at 4 - C, cell numbers declined, probably due to cell death, from days 1 to 7, but remained constant between days 7 and 14.
## Proliferation
To define optimal growth conditions for the LG-1 cells, we evaluated cell proliferation at different temperatures. We seeded cells at a density (50,000 cells/cm 2 ) that resulted in approximately 50% confluence after 1 day. After 1, 7 and 14 days, the cells were stained with the nuclear stain DRAQ5 and counted by a plate reader equipped with an imager module [fig_ref] Figure 3: Cell proliferation [/fig_ref]. When grown at 20 °C, cell numbers doubled within 14 days. At 16 °C, the cells grew at similar rates as at 20 °C for the first 7 days but then reduced their proliferation rate. At 10 °C, no change in cell numbers was observed at any time point, and at 4 °C, cell numbers declined, probably due to cell death, from days 1 to 7, but remained constant between days 7 and 14.
## Cell morphology
When examined by light microscopy, the adherent LG-1 cell population appeared homogenous with a flat, polygonal and stretched-out, almost transparent, appearance [fig_ref] Figure 4: Cell morphology of the established LG-1 cell line [/fig_ref]. Flow cytometry supported the observation of a homogenous cell population, as the cells clustered on a side scatter forward scatter plot, suggesting cells of similar size and complexity (data not shown). Staining the cells with the live cell stain Calcein-AM made it easier to assess their morphology [fig_ref] Figure 4: Cell morphology of the established LG-1 cell line [/fig_ref]. When reaching confluence, the cells stopped proliferating (data not shown), demonstrating that LG-1 cells are responsive to contact inhibition. Confluent cultures remained viable at 20 - C for at least 3 weeks (data not shown). PAS and Alcian blue staining suggested that no goblet cells were present in the LG-1 cultures [fig_ref] Figure 5: LG-1 cells stained with PAS [/fig_ref]. Similarly, staining for ATPase gave no signal, suggesting that chloride cells are not present in the cell population (data not shown).
Cells 2021, 10, x FOR PEER REVIEW 9 of 17 [fig_ref] Figure 3: Cell proliferation [/fig_ref]. Cell proliferation. The cells were plated at different densities and grown for 1, 7 and 14 days (d) at different temperatures, as indicated. The data are representative of three independent experiments and are expressed as mean ± SD of four parallel incubations. * indicates significantly different from day 1.
## Cell morphology
When examined by light microscopy, the adherent LG-1 cell population appeared homogenous with a flat, polygonal and stretched-out, almost transparent, appearance [fig_ref] Figure 4: Cell morphology of the established LG-1 cell line [/fig_ref]. Flow cytometry supported the observation of a homogenous cell population, as the cells clustered on a side scatter forward scatter plot, suggesting cells of similar size and complexity (data not shown). Staining the cells with the live cell stain Calcein-AM made it easier to assess their morphology [fig_ref] Figure 4: Cell morphology of the established LG-1 cell line [/fig_ref]. When reaching confluence, the cells stopped proliferating (data not shown), demonstrating that LG-1 cells are responsive to contact inhibition. Confluent cultures remained viable at 20 °C for at least 3 weeks (data not shown). PAS and Alcian blue staining suggested that no goblet cells were present in the LG-1 cultures [fig_ref] Figure 5: LG-1 cells stained with PAS [/fig_ref]. Similarly, staining for ATPase gave no signal, suggesting that chloride cells are not present in the cell population (data not shown).
## Cell morphology
When examined by light microscopy, the adherent LG-1 cell population appeared homogenous with a flat, polygonal and stretched-out, almost transparent, appearance [fig_ref] Figure 4: Cell morphology of the established LG-1 cell line [/fig_ref]. Flow cytometry supported the observation of a homogenous cell population, as the cells clustered on a side scatter forward scatter plot, suggesting cells of similar size and complexity (data not shown). Staining the cells with the live cell stain Calcein-AM made it easier to assess their morphology [fig_ref] Figure 4: Cell morphology of the established LG-1 cell line [/fig_ref]. When reaching confluence, the cells stopped proliferating (data not shown), demonstrating that LG-1 cells are responsive to contact inhibition. Confluent cultures remained viable at 20 °C for at least 3 weeks (data not shown). PAS and Alcian blue staining suggested that no goblet cells were present in the LG-1 cultures [fig_ref] Figure 5: LG-1 cells stained with PAS [/fig_ref]. Similarly, staining for ATPase gave no signal, suggesting that chloride cells are not present in the cell population (data not shown). Transmission electron microscopy was performed to further evaluate the morphological features of the LG-1 cells [fig_ref] Figure 5: LG-1 cells stained with PAS [/fig_ref]. To our knowledge, the ultrastructural features of lumpfish gill cells have not been described, and we therefore included whole gill tissue as a reference [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. Resembling observations in rainbow trout, the surface of lumpfish gill lamella was covered by thin, elongated pavement epithelial cells with a ruffled surface [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. Pillar endothelial cells covered the lumen of the gill lacuna [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. We also observed red blood cells [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref] , a mitochondrial-rich cell, often referred to as chloride cells [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref] , and cells compatible with goblet cells (data not shown). The TEM analysis of LG-1 confirmed the flattened shape observed by wide field microscopy [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. In line with the results from specific stains described above, the features of LG-1 cells were clearly distinct from the characteristic chloride cells and the mucus-containing goblet cells observed in tissue sections. Interestingly, some cells formed surface protrusions [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. Altogether, the TEM observations are compatible with properties of the flat pavement epithelium observed in the whole lumpfish gill, but a pillar endothelial cell type cannot be excluded. Moreover, a few desmosome-like structures were seen [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref] , further indicating that the LG-1 cell line is of epithelial or endothelial origin. Transmission electron microscopy was performed to further evaluate the morphological features of the LG-1 cells [fig_ref] Figure 5: LG-1 cells stained with PAS [/fig_ref]. To our knowledge, the ultrastructural features of lumpfish gill cells have not been described, and we therefore included whole gill tissue as a reference [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. Resembling observations in rainbow trout, the surface of lumpfish gill lamella was covered by thin, elongated pavement epithelial cells with a ruffled surface [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. Pillar endothelial cells covered the lumen of the gill lacuna [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. We also observed red blood cells [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref] , a mitochondrial-rich cell, often referred to as chloride cells [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref] , and cells compatible with goblet cells (data not shown). The TEM analysis of LG-1 confirmed the flattened shape observed by wide field microscopy [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. In line with the results from specific stains described above, the features of LG-1 cells were clearly distinct from the characteristic chloride cells and the mucus-containing goblet cells observed in tissue sections. Interestingly, some cells formed surface protrusions [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref]. Altogether, the TEM observations are compatible with properties of the flat pavement epithelium observed in the whole lumpfish gill, but a pillar endothelial cell type cannot be excluded. Moreover, a few desmosome-like structures were seen [fig_ref] Figure 6: Ultrastructure of whole lumpfish gill [/fig_ref] , further indicating that the LG-1 cell line is of epithelial or endothelial origin. Phalloidin, which stains filamentous actin, revealed stress fibers and circumferential actin filaments [fig_ref] Figure 7: Cytoskeletal filament profile of the LG-1 cells [/fig_ref]. The cells also exhibited positive staining for cytokeratin, further supporting an epithelial or endothelial cell type [fig_ref] Figure 7: Cytoskeletal filament profile of the LG-1 cells [/fig_ref]. No specific signal was observed when staining for the E-cadherin, ZO-1 or the endothelial-specific von Willebrand factor (data not shown). Notably, these antibodies are raised against human peptides and while they, in our experience, cross-react with epitopes of Atlantic salmon proteins, their ability to recognize lumpfish proteins remains unknown. As a result, the significance of these negative findings remains unclear. Phalloidin, which stains filamentous actin, revealed stress fibers and circumferential actin filaments [fig_ref] Figure 7: Cytoskeletal filament profile of the LG-1 cells [/fig_ref]. The cells also exhibited positive staining for cytokeratin, further supporting an epithelial or endothelial cell type [fig_ref] Figure 7: Cytoskeletal filament profile of the LG-1 cells [/fig_ref]. No specific signal was observed when staining for the E-cadherin, ZO-1 or the endothelial-specific von Willebrand factor (data not shown). Notably, these antibodies are raised against human peptides and while they, in our experience, cross-react with epitopes of Atlantic salmon proteins, their ability to recognize lumpfish proteins remains unknown. As a result, the significance of these negative findings remains unclear. Cells 2021, 10, x FOR PEER REVIEW 11 of 17
## Generation of ter
The formation of a selective permeable barrier is an important functional characteristic of both epithelial and endothelial cells. Transepithelial/transendothelial electrical resistance (TER) is a widely accepted quantitative measurement of the integrity of a cellular monolayer and tight junction dynamics [bib_ref] TEER measurement techniques for in vitro barrier model systems, Srinivasan [/bib_ref]. To evaluate the ability of the LG-1 cell line to generate TER, the cells were seeded on transwell membranes and TER was measured every 6 h over a period of 4 days (96 h). The TER increased over a period of 48 h when it reached its peak at about 16 TER/cm 2 [fig_ref] Figure 8: The cells were plated on transwell membranes and TER values measured every... [/fig_ref]. After this, the TER gradually decreased. When cells were cultured for a longer period, the TER continued to decline, reaching approximately 8 TER/cm 2 at 256 h (10 days, data not shown).
## Generation of ter
The formation of a selective permeable barrier is an important functional characteristic of both epithelial and endothelial cells. Transepithelial/transendothelial electrical resistance (TER) is a widely accepted quantitative measurement of the integrity of a cellular monolayer and tight junction dynamics [bib_ref] TEER measurement techniques for in vitro barrier model systems, Srinivasan [/bib_ref]. To evaluate the ability of the LG-1 cell line to generate TER, the cells were seeded on transwell membranes and TER was measured every 6 h over a period of 4 days (96 h). The TER increased over a period of 48 h when it reached its peak at about 16 TER/cm 2 [fig_ref] Figure 8: The cells were plated on transwell membranes and TER values measured every... [/fig_ref]. After this, the TER gradually decreased. When cells were cultured for a longer period, the TER continued to decline, reaching approximately 8 TER/cm 2 at 256 h (10 days, data not shown).
## Generation of ter
The formation of a selective permeable barrier is an important functional characteristic of both epithelial and endothelial cells. Transepithelial/transendothelial electrical resistance (TER) is a widely accepted quantitative measurement of the integrity of a cellular monolayer and tight junction dynamics [bib_ref] TEER measurement techniques for in vitro barrier model systems, Srinivasan [/bib_ref]. To evaluate the ability of the LG-1 cell line to generate TER, the cells were seeded on transwell membranes and TER was measured every 6 h over a period of 4 days (96 h). The TER increased over a period of 48 h when it reached its peak at about 16 TER/cm 2 [fig_ref] Figure 8: The cells were plated on transwell membranes and TER values measured every... [/fig_ref]. After this, the TER gradually decreased. When cells were cultured for a longer period, the TER continued to decline, reaching approximately 8 TER/cm 2 at 256 h (10 days, data not shown).
## Cyp1a induction
Both gill epithelial cells and pillar endothelial cells are able to upregulate and activate CYP1A, a key enzyme in oxidative metabolism, in response to the stimulation of the aryl hydrocarbon receptor. To investigate the ability of the LG-1 cells to induce CYP1A activity, the cells were treated with beta-napthoflavone (BNF; 24 h), an aryl hydrocarbon receptor agonist and a known inducer of CYP1A expression [bib_ref] Improving the in vitro ethoxyresorufin-O-deethylase (EROD) assay with RTL-W1 by metabolic normalization..., Heinrich [/bib_ref]. The subsequent CYP1A activity was measured by using the EROD assay, and, here, the BNF-exposed LG-1 cells had a significant higher CYP1A activity [fig_ref] Figure 9: CYP1A induction [/fig_ref]. Cells 2021, 10, x FOR PEER REVIEW 12 of 17
## Cyp1a induction
Both gill epithelial cells and pillar endothelial cells are able to upregulate and activate CYP1A, a key enzyme in oxidative metabolism, in response to the stimulation of the aryl hydrocarbon receptor. To investigate the ability of the LG-1 cells to induce CYP1A activity, the cells were treated with beta-napthoflavone (BNF; 24 h), an aryl hydrocarbon receptor agonist and a known inducer of CYP1A expression [bib_ref] Improving the in vitro ethoxyresorufin-O-deethylase (EROD) assay with RTL-W1 by metabolic normalization..., Heinrich [/bib_ref]. The subsequent CYP1A activity was measured by using the EROD assay, and, here, the BNF-exposed LG-1 cells had a significant higher CYP1A activity [fig_ref] Figure 9: CYP1A induction [/fig_ref].
## Susceptibility to fish viruses
The LG-1 cells were inoculated with a panel of viruses associated with severe disease in farmed fish. Following 4-7 days of infection, the cells were fixed and an IFAT performed to visualise virus uptake and propagation in the cells.
LG-1 cells were not permissive to IPNV (sp serotype) or lumpfish flavivirus (data not shown). After infection with ISAV, SAV2 or SAV3, single cells stained positive for viral antigens, indicating either uptake of the virus or limited viral protein expression in individual cells (demonstrated for ISAV in [fig_ref] Figure 1: Lumpfish [/fig_ref]. However, there was no observed CPE nor an increase in viral load over time and no further transmission of the virus after prolonged incubation (>21 days), indicating that LG-1 cells do not support the propagation of these viruses. In contrast, LG-1 cells were susceptible to infection by SAV1, fish nodavirus (BFNNV genotype), IHNV genogroup M and VHSV genogroup III, resulting in typical cytopathic effect (CPE) and the distinct cellular expression of viral antigens [fig_ref] Figure 1: Lumpfish [/fig_ref]. For VHSV, CPE developed very rapid and complete CPE was observed within a few days after inoculation. [fig_ref] Figure 1: Lumpfish [/fig_ref]. The susceptibility of LG-1 to SAV1, fish nodavirus, IHNV and VHSV was investigated further by comparing the viral titers obtained in LG-1 to titers obtained in OIE recommended standard cell lines used for propagation and diagnostic purposes (CHSE-214, E-11, EPC and BF-2, respectively). The propagation of VHSV in LG-1 cells resulted in a slightly higher titer than in BF-2, 10 6.8 TCID50/mL and 10 6.3 TCID50/mL, respectively. For the other viruses, titers obtained in LG-1 cells were >2 log lower than in standard cell lines.
## Susceptibility to fish viruses
The LG-1 cells were inoculated with a panel of viruses associated with severe disease in farmed fish. Following 4-7 days of infection, the cells were fixed and an IFAT performed to visualise virus uptake and propagation in the cells.
LG-1 cells were not permissive to IPNV (sp serotype) or lumpfish flavivirus (data not shown). After infection with ISAV, SAV2 or SAV3, single cells stained positive for viral antigens, indicating either uptake of the virus or limited viral protein expression in individual cells (demonstrated for ISAV in [fig_ref] Figure 1: Lumpfish [/fig_ref]. However, there was no observed CPE nor an increase in viral load over time and no further transmission of the virus after prolonged incubation (>21 days), indicating that LG-1 cells do not support the propagation of these viruses. In contrast, LG-1 cells were susceptible to infection by SAV1, fish nodavirus (BFNNV genotype), IHNV genogroup M and VHSV genogroup III, resulting in typical cytopathic effect (CPE) and the distinct cellular expression of viral antigens [fig_ref] Figure 1: Lumpfish [/fig_ref]. For VHSV, CPE developed very rapid and complete CPE was observed within a few days after inoculation. [fig_ref] Figure 1: Lumpfish [/fig_ref]. The susceptibility of LG-1 to SAV1, fish nodavirus, IHNV and VHSV was investigated further by comparing the viral titers obtained in LG-1 to titers obtained in OIE recommended standard cell lines used for propagation and diagnostic purposes (CHSE-214, E-11, EPC and BF-2, respectively). The propagation of VHSV in LG-1 cells resulted in a slightly higher titer than in BF-2, 10 6.8 TCID 50 /mL and 10 6.3 TCID 50 /mL, respectively. For the other viruses, titers obtained in LG-1 cells were >2 log lower than in standard cell lines. Cells 2021, 10, x FOR PEER REVIEW 13 of 17 [fig_ref] Table 2: Overview of virus isolates and primary antibodies [/fig_ref]. Microscope: Leica DMIL.
# Discussion
Here, we document the establishment of a stable gill cell line (LG-1), maintained for 40 passages and with confirmed lumpfish origin. The cell line was defined as an epithelial or endothelial cell type, based on several findings. First, the morphological features of LG-1 cells resemble those of squamous epithelium, as they are adherent, homogenous and have a flat, stretched-out, almost transparent appearance. Furthermore, the thin and delicate nature of LG-1 cells corresponds to our observations of pavement epithelial cells in this species and is consistent with the main function of the gills, namely respiration. Second, the LG-1 cells form a confluent monolayer with a cobble-stone appearance, and cell division is inhibited upon cell-cell contact/contact-inhibition. Third, LG-1 cells form close contacts, with structures consistent with desmosomes, the anatomical hallmark of barrier cells such as epithelial and endothelial cells, and they generate trans-epithelial resistance [fig_ref] Table 2: Overview of virus isolates and primary antibodies [/fig_ref]. Microscope: Leica DMIL.
# Discussion
Here, we document the establishment of a stable gill cell line (LG-1), maintained for 40 passages and with confirmed lumpfish origin. The cell line was defined as an epithelial or endothelial cell type, based on several findings. First, the morphological features of LG-1 cells resemble those of squamous epithelium, as they are adherent, homogenous and have a flat, stretched-out, almost transparent appearance. Furthermore, the thin and delicate nature of LG-1 cells corresponds to our observations of pavement epithelial cells in this species and is consistent with the main function of the gills, namely respiration. Second, the LG-1 cells form a confluent monolayer with a cobble-stone appearance, and cell division is inhibited upon cell-cell contact/contact-inhibition. Third, LG-1 cells form close contacts, with structures consistent with desmosomes, the anatomical hallmark of barrier cells such as epithelial and endothelial cells, and they generate trans-epithelial resistance (TER). Interestingly, the TEM images also clearly showed that some LG-1 cells generated long cellular protrusions. Epithelial and endothelial cells also typically express IL-6 [bib_ref] Expression and function of interleukin-6 in epithelial cells, Krueger [/bib_ref] [bib_ref] Interleukin-6 gene expression in human endothelial cells: RNA start sites, multiple IL-6..., May [/bib_ref] , as shown here using IL-6 qPCR.
Predicting acute toxicity in fish is important, and fish cell lines represent an attractive alternative method to using live fish. The gills are one of the major extra-hepatic sites of metabolic activity. Accordingly, the RTgill-W1 has been extensively used in research over the last decade, demonstrating a close correlation between RTgill-W1 cytotoxicity and acute fish toxicity [bib_ref] Repeatability and Reproducibility of the RTgill-W1 Cell Line Assay for Predicting Fish..., Fischer [/bib_ref] [bib_ref] Predicting fish acute toxicity using a fish gill cell line-based toxicity assay, Tanneberger [/bib_ref]. The detoxification capabilities of the gills includes the phase I enzyme CYP1A, whose activity can be measured by the EROD assay. CYP1A activity is often used as a sensitive biomarker of exposure to organic and inorganic environmental pollutants with a planar structure, such as 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD), Benzo(a)pyrene (B(a)P), polychlorinated biphenyls (PCBs) and Polybrominated diphenyl ethers (PBDEs). To detect these chemicals, the EROD assay is used by several cellular ex vivo and in vitro systems, such as gill filaments [bib_ref] Evaluation of the gill filamentbased EROD assay in African sharptooth catfish (Clarias..., Mdegela [/bib_ref] , primary fish liver and gill cell cultures [bib_ref] EROD activity and cytochrome P4501A induction in liver and gills of Senegal..., Oliva [/bib_ref] [bib_ref] Xenobiotic and steroid biotransformation activities in rainbow trout gill epithelial cells in..., Leguen [/bib_ref] and the rainbow trout liver cell line RTL-W1 [bib_ref] Comparison of Toxic Equivalent Factors for Selected Dioxin and Furan Congeners Derived..., Clemons [/bib_ref]. Interestingly, the LG-1 cells can induce CYP1A activity and may thus serve as a tool to study biotransformation or as a biosensor. In conclusion, the LG-1 cell line may contribute to toxicity testing; however, its sensitivity towards known toxins should first be evaluated.
The isolation and propagation of infectious agents in cell culture are needed to allow for the identification and characterization of emerging viruses and their interaction with host cells. Cellular factors that regulate the permissiveness to infection and disease, including surface proteins used for viral attachment and intrinsic antiviral responses, generally differ between species. An example is the manifestation of amoebic gill disease, which in lumpfish is different from that in salmon, with a slower development of characteristic pathology [bib_ref] Cyclopterus lumpus L.) develop amoebic gill disease (AGD) after experimental challenge with..., Haugland [/bib_ref] , suggesting species-specific strategies to combat Paramoeba perurans. This difference may be related to aspects of the environment or agent, but putative species differences can now be investigated using gill cell lines from the different species. Hence, the LG-1 cell line is an important addition to the current diagnostic toolbox in aquaculture.
Many aquatic viruses replicate in fish gill epithelium and may pass through this thin cell layer, leading to systemic infection. LG-1 cells supported the growth of SAV1, the causative agent of pancreas disease in Atlantic salmon. SAV exists in six genotypes. Genotype SAV3 so far is exclusively found in Norway, marine SAV2 is found in both the UK and Norway and SAV1 has not been detected in Norway but is endemic to the UK and Ireland. While SAV1 was propagated in LG-1, the cells were not permissive to either SAV2 or SAV3. This susceptibility of lumpfish to SAV1 is novel and highly interesting, as the marine reservoirs for the different SAV genotypes have not been identified. Although some SAV genotypes have been found in marine fish species like the common dab [bib_ref] Matejusova, I. Common dab, Limanda limanda (L.), as a natural carrier of..., Simons [/bib_ref] [bib_ref] Detection of salmonid alphavirus RNA in Celtic and Irish Sea flatfish, Mccleary [/bib_ref] , pancreas disease is only described in salmonids [bib_ref] Alphavirus infections in salmonids-a review, Mcloughlin [/bib_ref].
Although nodavirus has not been detected in either farmed or wild-caught lumpfish so far, a previous challenge trial has shown that lumpfish can be experimentally infected. The fact that LG-1 supported the growth of a betanodavirus field isolate from Norwegian halibut indicates that gill cells from lumpfish express surface proteins that mediate nodavirus infection. In Norway, this virus has already been detected in connection with disease in farmed cod, halibut and turbot [bib_ref] Isolation in cell culture of nodavirus from farmed Atlantic halibut Hippoglossus hippoglossus..., Dannevig [/bib_ref] [bib_ref] Outbreaks of viral nervous necrosis in juvenile and adult farmed Atlantic cod,..., Hellberg [/bib_ref] [bib_ref] Nodavirus in farmed Atlantic cod Gadus morhua in Norway, Patel [/bib_ref] [bib_ref] Viral encephalopathy and retinopathy (VER) in Atlantic salmon Salmo salar after intraperitoneal..., Korsnes [/bib_ref]. The introduction of nodavirus into lumpfish farms may be of concern, as the virus is known to cause persistent infections and disease in susceptible species. However, salmonids are not considered natural hosts for nodavirus infections, although the virus can be introduced experimentally [bib_ref] Viral encephalopathy and retinopathy (VER) in Atlantic salmon Salmo salar after intraperitoneal..., Korsnes [/bib_ref]. Hence, the risk of natural transmission of the virus and associated disease to farmed salmon would probably be low, although the long-term cohabitation of species may increase possible transfer and adaptation, as noted earlier.
VHS is a notifiable disease in both EU and OIE, and the introduction of VHSV into Norwegian aquaculture may seriously impact both fish health and trade. The LG-1 cells supported the efficient propagation of a field isolate of VHSV genogroup III from an outbreak of VHS of presumed marine origin in a rainbow trout farm in Norway [bib_ref] Evaluation of the gill filamentbased EROD assay in African sharptooth catfish (Clarias..., Mdegela [/bib_ref].
Outbreaks of VHS in farmed cleaner fish, including lumpfish, have already been reported in other countries [bib_ref] Outbreak of viral haemorrhagic septicaemia (VHS) in lumpfish (Cyclopterus lumpus) in Iceland..., Guðmundsdóttir [/bib_ref] [bib_ref] A mortality event in wrasse species (Labridae) associated with the presence of..., Munro [/bib_ref] , and the screening of wild fish populations has demonstrated that VHSV is present in marine fish reservoirs both in Scotland, Norway, Denmark and the Baltic Seas [bib_ref] High prevalence of viral haemorrhagic septicaemia virus (VHSV) in Norwegian spring-spawning herring, Johansen [/bib_ref] [bib_ref] Isolation of viral haemorrhagic septicaemia virus (VHSV) from wild marine fish species..., Mortensen [/bib_ref] [bib_ref] Isolation and identification of Viral Haemorrhagic Septicaemia (VHS) viruses from cod Gadus..., Smail [/bib_ref]. So far, disease caused by the virus has only been reported in rainbow trout in Norway, but close contact between salmon and lumpfish may facilitate an adaptation of virus and "host-jumping", a strategy associated with many other RNA viruses like coronaviruses and influenza viruses [bib_ref] Jumping species-a mechanism for coronavirus persistence and survival, Menachery [/bib_ref] [bib_ref] Enabling the 'host jump': Structural determinants of receptor-binding specificity in influenza A..., Shi [/bib_ref].
In conclusion, LG-1 provides a new tool to study gill epithelial cell function as well as the detection, propagation and characterization of viruses from lumpfish and other marine species. Furthermore, the use of LG-1 as a biosensor for toxins and xenobiotic activity of CYP1A appears promising.
[fig] Figure 1: Lumpfish (A) and explants (B) from lumpfish gills, with gill filaments (left, indicated by arrows), an extensive growth of cells from the explant (middle and right part) and a confluent layer of cells (4 days after seeding). The thickness of the gill filament indicated by the arrows is about 200 µm. Microscope: Leica DMIL. [/fig]
[fig] Figure 2: Figure 2. Control of lumpfish origin. The origin of the LG-1 cells was controlled by RT-qPCR on total RNA using lumpfish-specific primers targeting IL-6 and EF1α, and Atlantic salmon specific primers targeting EF1α were used as a negative control. The correct amplicon length was confirmed by capillary electrophoresis (Bioanalyzer). Lump: Lumpfish. Ss: Salmo salar. [/fig]
[fig] Figure 3: Cell proliferation. The cells were plated at different densities and grown for 1, 7 and 14 days (d) at different temperatures, as indicated. The data are representative of three independent experiments and are expressed as mean ± SD of four parallel incubations. * indicates significantly different from day 1. [/fig]
[fig] Figure 4: Cell morphology of the established LG-1 cell line. (A) Light microscopy. (B) The cells were stained with the live cell stain Calcein-AM and visualized by fluorescence microscopy. Scale bar = 100 µm Microscope: Leica DMIL. [/fig]
[fig] Figure 5: LG-1 cells stained with PAS (A) and Alcian blue (B) to clarify the presence of mucus cells (bright pink or blue, respectively). Microscope: Leica DMIL. [/fig]
[fig] Figure 6: Ultrastructure of whole lumpfish gill (A) and LG-1 cells (B-F). (A) TEM of gill showed pavement epithelial cells (PE), pillar endothelial cells (EC), red blood cells (RBC) and chloride cells (CC). (B-F) TEM of LG-1 cells showed (B) elongated shape, (C) surface protrusions and (D-F) structures compatible with desmosomes. [/fig]
[fig] Figure 7: Cytoskeletal filament profile of the LG-1 cells. F-actin (A; phalloidin; red) and cytokeratin (B; green) were stained and visualized by confocal microscopy. Scale bar = 50 µm. [/fig]
[fig] Figure 8: The cells were plated on transwell membranes and TER values measured every 6 h. The results represent mean ± SEM of 3 independent experiments. * indicates significantly different from 0 h. [/fig]
[fig] Figure 9: CYP1A induction. The cells were treated with BNF for 24 h and their catalytic activity was measured by the EROD assay. The results represent mean ± SEM of 3 independent experiments. * indicates significantly different from control. [/fig]
[table] Table 1: Primer over view; species identification. [/table]
[table] Table 2: Overview of virus isolates and primary antibodies. [/table]
|
10.3390/molecules23092342
|
CCBY
|
6225133
|
30216992
|
s2orc_pubmed_articles
|
Anti-Inflammatory and Skin-Moisturizing Effects of a Flavonoid Glycoside Extracted from the Aquatic Plant Nymphoides indica in Human Keratinocytes
Nymphoides indica, an aquatic plant, is used as folk medicine in some countries. Our previous study demonstrated that the methanol extract of N. indica inhibited the activity of tyrosinases, tyrosine related protein (TRP)1 and TRP2, and microphthalmia-associated transcription factor, as well as the activity of protein kinase A, by effectively inhibiting cyclic adenosine monophosphate. Although the biological activities of N. indica extract have been reported, there are no reports on the skin bioactivity of the main compound(s) on human keratinocytes. This study investigated the anti-inflammatory and moisturizing effects of quercetin 3,7-dimethyl ether 4 -glucoside (QDG) isolated from N. indica. In brief, ultraviolet B irradiated keratinocytes were pretreated with different concentrations of QDG, and the effects of QDG on various inflammatory markers were determined. QDG significantly inhibited inflammation-related cytokines and chemokines and enhanced the activation of skin barrier factors. Additionally, QDG also attenuated phosphorylation inhibition of the upstream cytokines and nuclear factor-κB expression. These results suggest that QDG isolated from N. indica may serve as a potential source of bioactive substances for chronic inflammatory skin diseases.
# Introduction
Keratinocytes express and release inflammatory mediators in response to skin inflammation, and pro-inflammatory cytokines during the progression phase of the inflammatory process [bib_ref] The inflammatory response of keratinocytes and its modulation by Vitamin D: The..., Miodovnik [/bib_ref]. Therefore, keratinocytes play an important role in the pathogenesis of inflammatory skin diseases such as atopic and contact dermatitis [bib_ref] Resident skin cells in psoriasis: A special look at the pathogenetic functions..., Albanesi [/bib_ref]. Epidermal cells can also produce substantial amounts of cytokines, constitutively or following activation, strongly supporting the skin's function as an immune organ. The cytoplasm of keratinocytes in all the epidermal layers contains the pro-inflammatory cytokine interleukin (IL)-1, and the passively released IL-1 induces the expression of other cytokines such as IL-6, IL-8, IL-10, and IL-13. IL-8 is a chemokine with strong chemotactic effects on polymorphonuclear neutrophils and lymphocytes. Importantly, keratinocytes are known to express IL-8 after stimulation with IL-1β, interferon (IFN)-γ, or tumor necrosis factor (TNF)-α [bib_ref] Chemical induction of interleukin-8, a pro-inflammatory chemokine, in human epidermal keratinocyte cultures..., Wilmer [/bib_ref].
Additionally, thymus-and activation-regulated chemokine (TARC) and macrophage-derived chemokine (MDC) can play an important role in the development of chronic inflammatory diseases. Keratinocytes and other skin-resident cells produce cytokines that regulate intercellular communication. Thus, limited cytokine expression can contribute to dysfunctional barriers observed in chronic
# Results and discussion
## Cell migration
We confirmed the anti-inflammatory activity in the HaCaT cells of the N. indica extract prior to these experiments. As a result, COX-2 protein expression was inhibited by 25%, 38%, and 63% in a concentration-dependent manner at the concentrations of 5, 10, and 20 µg/mL of the N. indica extract. In addition, the anti-inflammatory activity of the ethyl acetate fraction (80% at 20 µg/mL) was confirmed by measuring the anti-inflammatory activity of the solvent fraction (data not shown). Therefore, the QDG of this study was isolated from the ethyl acetate fraction and the anti-inflammatory effect of UVB in the HaCaT cells was examined.
The keratinocytes of the skin play an important role in maintaining the homeostasis of the skin by producing various cytokines and growth factors involved in immune and inflammatory reactions and cell proliferation [bib_ref] Studies on the antimicrobial effect of collected bee venom using electric shock..., Han [/bib_ref]. In this study, the effects of QDG on the migration ability of HaCaT cells were investigated utilizing a wound-healing assay. HaCaT cells, uniformly grown in a monolayer, were scratched with a yellow tip and all the cells in the solid line were removed. The QDG concentration of the keratinocyte layer was determined by the MTT assay and was determined to be 1, 5, and 10 µg/mL (data not shown). Jang et al. [bib_ref] Effect of electrospun non-woven mats of dibutyryl chitin/poly(lactic acid) blends on wound..., Jang [/bib_ref] reported dibutyryl chitin activity similar to the highest concentration of dibutyryl chitin, 100 µg/mL, and QDG 10 µg/mL, compared with the cell migration of 25, 50, and 100 µg/mL of keratinocytes. QDG was able to confirm the superior cell migration ability. Results indicate that the control group cells showed some migration ability, and the QDG-treated group exhibited a dose-dependent increase in migration. This effect was more pronounced at 10 µg/mL of QDG [fig_ref] Figure 1: Chemical structure of quercetin 3,7-dimethyl ether 4′-glucoside [/fig_ref]. Thus, it can be suggested that QDG provides anti-inflammatory effects by increasing the cell migration ability of keratinocytes.
Molecules 2018, 23, x 3 of 13 concentration of the keratinocyte layer was determined by the MTT assay and was determined to be 1, 5, and 10 μg/mL (data not shown). Jang et al. [bib_ref] Effect of electrospun non-woven mats of dibutyryl chitin/poly(lactic acid) blends on wound..., Jang [/bib_ref] reported dibutyryl chitin activity similar to the highest concentration of dibutyryl chitin, 100 μg/mL, and QDG 10 μg/mL, compared with the cell migration of 25, 50, and 100 μg/mL of keratinocytes. QDG was able to confirm the superior cell migration ability. Results indicate that the control group cells showed some migration ability, and the QDG-treated group exhibited a dose-dependent increase in migration. This effect was more pronounced at 10 μg/mL of QDG [fig_ref] Figure 1: Chemical structure of quercetin 3,7-dimethyl ether 4′-glucoside [/fig_ref]. Thus, it can be suggested that QDG provides antiinflammatory effects by increasing the cell migration ability of keratinocytes.
## Qdg's inhibitory effect on cytokine production
Cytokines function as signaling peptides regulating cell intercourse and providing control of the tissue-specific cell homing. In the skin, chemokines are secreted by the resident cell. Chemokines and cytokines participate in the induction and maintenance of inflammation in the skin [bib_ref] Chemokines and cytokines network in the pathogenesis of the inflammatory skin diseases:..., Nedoszytko [/bib_ref]. To further understand QDG's control of the activation of HaCaT cells, we studied its effects on proinflammatory cytokines. In the present study, we particularly evaluated the activation of TNF-α, IL-1β, IL-6, and IL-8. Interestingly, QDG dose-dependently suppressed the expression of TNF-α, IL-1β, IL-6, and IL-8. Furthermore, at a dose of 10 μg/mL, QDG significantly inhibited IL-1β, IL-6, and IL-8 [fig_ref] Figure 2: Effect of QDG treatment on cytokine expression in HaCaT cells [/fig_ref]. Jeong et al. [bib_ref] Esculetin from Fraxinus rhynchophylla attenuates atopic skin inflammation by inhibiting the expression..., Jeong [/bib_ref] reported that IL-1β, IL-6, and IL-8 inhibited the cytokine-inhibitory activity of esculetin in HaCaT cells. In particular, QDG showed better IL-1β inhibitory activity. These results demonstrate the potential usefulness of QDG to treat skin inflammation.
## Qdg's inhibitory effect on cytokine production
Cytokines function as signaling peptides regulating cell intercourse and providing control of the tissue-specific cell homing. In the skin, chemokines are secreted by the resident cell. Chemokines and cytokines participate in the induction and maintenance of inflammation in the skin [bib_ref] Chemokines and cytokines network in the pathogenesis of the inflammatory skin diseases:..., Nedoszytko [/bib_ref]. To further understand QDG's control of the activation of HaCaT cells, we studied its effects on pro-inflammatory cytokines. In the present study, we particularly evaluated the activation of TNF-α, IL-1β, IL-6, and IL-8. Interestingly, QDG dose-dependently suppressed the expression of TNF-α, IL-1β, IL-6, and IL-8. Furthermore, at a dose of 10 µg/mL, QDG significantly inhibited IL-1β, IL-6, and IL-8 [fig_ref] Figure 2: Effect of QDG treatment on cytokine expression in HaCaT cells [/fig_ref]. Jeong et al. [bib_ref] Esculetin from Fraxinus rhynchophylla attenuates atopic skin inflammation by inhibiting the expression..., Jeong [/bib_ref] reported that IL-1β, IL-6, and IL-8 inhibited the cytokine-inhibitory activity of esculetin in HaCaT cells. In particular, QDG showed better IL-1β inhibitory activity. These results demonstrate the potential usefulness of QDG to treat skin inflammation.
## Qdg's inhibitory effect on chemokine production
Chronic inflammatory skin diseases such as atopic and contact dermatitis occur due to loss of skin barrier function and inability to control the T helper type 2 (Th2)/T helper type 1 (Th1) immune balance [bib_ref] Th2 cytokines act on S100/A11 to downregulate keratinocyte differentiation, Howell [/bib_ref] [bib_ref] Loricrin and involucrin expression is down-regulated by Th2 cytokines through STAT-6, Kim [/bib_ref]. Environmental factors, such as ultraviolet light, are an important factor in inflammatory diseases, with an increase in chemokines and cytokines. Therefore, we explored the effect of QDG on Th2 immune modulation, as well as its effect on the expression of TARC and MDC, members of the CC chemokine subfamily, expressed by the keratinocytes. QDG inhibited UVBoverexpressed MDC and TARC expression in a concentration-dependent manner. Especially at an MDC concentration of 10 μg/mL, the inhibition rate was over 40% higher than that of the control, and it was confirmed that the inhibitory activity was better than that of EGCG [fig_ref] Figure 3: Effect of QDG treatment on macrophage-derived chemokine [/fig_ref]. TARC and MDC selectively control the refection and migration of Th2 lymphocytes to inflammatory sites and are considered major factors in the pathogenesis of inflammatory diseases, such as atopic dermatitis [bib_ref] Both Th2 and Th1 chemokines (TARC/CCL17, MDC/CCL22, and Mig/CXCL9) are elevated in..., Shimada [/bib_ref] [bib_ref] Serum thymus and activation-regulated chemokine, macrophage-derived chemokine and eotaxin as markers of..., Jahnz-Rozyk [/bib_ref]. Thus, the inhibitory effects of QDG on the expression of these chemokines reveal its potential for the treatment of inflammatory diseases.
## Qdg's inhibitory effect on chemokine production
Chronic inflammatory skin diseases such as atopic and contact dermatitis occur due to loss of skin barrier function and inability to control the T helper type 2 (Th2)/T helper type 1 (Th1) immune balance [bib_ref] Th2 cytokines act on S100/A11 to downregulate keratinocyte differentiation, Howell [/bib_ref] [bib_ref] Loricrin and involucrin expression is down-regulated by Th2 cytokines through STAT-6, Kim [/bib_ref]. Environmental factors, such as ultraviolet light, are an important factor in inflammatory diseases, with an increase in chemokines and cytokines. Therefore, we explored the effect of QDG on Th2 immune modulation, as well as its effect on the expression of TARC and MDC, members of the CC chemokine subfamily, expressed by the keratinocytes. QDG inhibited UVB-overexpressed MDC and TARC expression in a concentration-dependent manner. Especially at an MDC concentration of 10 µg/mL, the inhibition rate was over 40% higher than that of the control, and it was confirmed that the inhibitory activity was better than that of EGCG [fig_ref] Figure 3: Effect of QDG treatment on macrophage-derived chemokine [/fig_ref]. TARC and MDC selectively control the refection and migration of Th2 lymphocytes to inflammatory sites and are considered major factors in the pathogenesis of inflammatory diseases, such as atopic dermatitis [bib_ref] Both Th2 and Th1 chemokines (TARC/CCL17, MDC/CCL22, and Mig/CXCL9) are elevated in..., Shimada [/bib_ref] [bib_ref] Serum thymus and activation-regulated chemokine, macrophage-derived chemokine and eotaxin as markers of..., Jahnz-Rozyk [/bib_ref]. Thus, the inhibitory effects of QDG on the expression of these chemokines reveal its potential for the treatment of inflammatory diseases.
## Qdg's inhibitory effect on chemokine production
Chronic inflammatory skin diseases such as atopic and contact dermatitis occur due to loss of skin barrier function and inability to control the T helper type 2 (Th2)/T helper type 1 (Th1) immune balance [bib_ref] Th2 cytokines act on S100/A11 to downregulate keratinocyte differentiation, Howell [/bib_ref] [bib_ref] Loricrin and involucrin expression is down-regulated by Th2 cytokines through STAT-6, Kim [/bib_ref]. Environmental factors, such as ultraviolet light, are an important factor in inflammatory diseases, with an increase in chemokines and cytokines. Therefore, we explored the effect of QDG on Th2 immune modulation, as well as its effect on the expression of TARC and MDC, members of the CC chemokine subfamily, expressed by the keratinocytes. QDG inhibited UVBoverexpressed MDC and TARC expression in a concentration-dependent manner. Especially at an MDC concentration of 10 μg/mL, the inhibition rate was over 40% higher than that of the control, and it was confirmed that the inhibitory activity was better than that of EGCG [fig_ref] Figure 3: Effect of QDG treatment on macrophage-derived chemokine [/fig_ref]. TARC and MDC selectively control the refection and migration of Th2 lymphocytes to inflammatory sites and are considered major factors in the pathogenesis of inflammatory diseases, such as atopic dermatitis [bib_ref] Both Th2 and Th1 chemokines (TARC/CCL17, MDC/CCL22, and Mig/CXCL9) are elevated in..., Shimada [/bib_ref] [bib_ref] Serum thymus and activation-regulated chemokine, macrophage-derived chemokine and eotaxin as markers of..., Jahnz-Rozyk [/bib_ref]. Thus, the inhibitory effects of QDG on the expression of these chemokines reveal its potential for the treatment of inflammatory diseases.
## Qdg's effect on the skin barrier and hyaluronic acid synthase production
The epidermal skin barrier plays a significant role in the susceptibility and severity of chronic inflammatory diseases, such as atopic dermatitis [bib_ref] Reevaluation of the normal epidermal calcium gradient, and analysis of calcium levels..., Leinonen [/bib_ref] [bib_ref] Impaired trafficking of the desmoplakins in cultured Darier's disease keratinocytes, Dhitavat [/bib_ref]. The differentiation of HaCaT cells and the subsequent formation of the skin barrier are a tightly regulated process, often triggered by calcium sensitization and release from the endoplasmic reticulum [bib_ref] Modulations in epidermal calcium regulate the expression of differentiation-specific markers, Elias [/bib_ref]. This study evaluated the effects of QDG on skin barrier peptide expression and hyaluronic acid production. QDG significantly increased the production of filaggrin, involucrin, loricrin, and hyaluronic acid synthase-1 (HAS-1) with reduced expression rates. In particular, QDG increased the expression of filaggrin, involucrin, loricrin, and HAS-1 by 78%, 85%, 93%, and 95%, respectively, at a final concentration of 10 µg/mL. Interestingly, QDG treatment dose-dependently upregulated the expression of filaggrin, involucrin, loricrin, and HAS-1 levels [fig_ref] Figure 4: Effect of QDG treatment on skin barrier and hyaluronic acid synthase expressions... [/fig_ref]. Kim et al. [bib_ref] The skin protective effects of compound K, a metabolite of ginsenoside Rb1..., Kim [/bib_ref] reported that compound K enhances the expression of filaggrin and HAS-1 mRNA in the HaCaT cells, and QDG is superior to compound K in the reported skin protection effect. These results suggest that QDG is essential for retaining water and maintaining intercellular space and plays an important role in skin moisture retention by stimulating the expression of genes that facilitate transportation of ions and nutrients.
## Qdg's effect on the skin barrier and hyaluronic acid synthase production
The epidermal skin barrier plays a significant role in the susceptibility and severity of chronic inflammatory diseases, such as atopic dermatitis [bib_ref] Reevaluation of the normal epidermal calcium gradient, and analysis of calcium levels..., Leinonen [/bib_ref] [bib_ref] Impaired trafficking of the desmoplakins in cultured Darier's disease keratinocytes, Dhitavat [/bib_ref]. The differentiation of HaCaT cells and the subsequent formation of the skin barrier are a tightly regulated process, often triggered by calcium sensitization and release from the endoplasmic reticulum [bib_ref] Modulations in epidermal calcium regulate the expression of differentiation-specific markers, Elias [/bib_ref]. This study evaluated the effects of QDG on skin barrier peptide expression and hyaluronic acid production. QDG significantly increased the production of filaggrin, involucrin, loricrin, and hyaluronic acid synthase-1 (HAS-1) with reduced expression rates. In particular, QDG increased the expression of filaggrin, involucrin, loricrin, and HAS-1 by 78%, 85%, 93%, and 95%, respectively, at a final concentration of 10 μg/mL. Interestingly, QDG treatment dose-dependently upregulated the expression of filaggrin, involucrin, loricrin, and HAS-1 levels [fig_ref] Figure 4: Effect of QDG treatment on skin barrier and hyaluronic acid synthase expressions... [/fig_ref]. Kim et al. [bib_ref] The skin protective effects of compound K, a metabolite of ginsenoside Rb1..., Kim [/bib_ref] reported that compound K enhances the expression of filaggrin and HAS-1 mRNA in the HaCaT cells, and QDG is superior to compound K in the reported skin protection effect. These results suggest that QDG is essential for retaining water and maintaining intercellular space and plays an important role in skin moisture retention by stimulating the expression of genes that facilitate transportation of ions and nutrients. HaCaT cells were treated with different concentrations of QDG (1, 5, and 10 μg/mL) after irradiation with 20 mJ/cm 2 UVB. After 6 h, cells were harvested and relative mRNA levels were determined. Histogram shows the densitometry for the skin barrier proteins and hyaluronic acid synthase mRNA normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Each value represents mean ± SD for the three individual experiments. Nor: No treatment group (0 h), Cont: 20 mJ/cm 2 UVB treatment group, QDG: QDG treatment group. n = 3, * = p < 0.001 and ** = p < 0.0001 compared with the control group.
## Phosphorylation of p38/jnk/erk/iκb
Among the inflammatory response intracellular signaling pathways, MAPKs are the wellknown signaling pathways involved in the inflammatory response [bib_ref] Pathogenesis and therapy of psoriasis, Lowes [/bib_ref] [bib_ref] New insights in the immunologic basis of psoriasis, Nograles [/bib_ref] [bib_ref] Psoriasis: Evolution of pathogenic concepts and new therapies through phases of translational..., Guttman-Yassky [/bib_ref]. Activated MAPKs, in response to cell stimulation, induce the expression of target genes by activating other kinases or transcription factors, such as p38, c-Jun N-terminal kinase (JNK), and ERK. p38 is a central regulator of the inflammatory response regulating IL-6, IL-8, and TNF-α production and the expression of nitric oxide and metalloproteinase [bib_ref] The biology of p38 kinase: A central role in inflammation, Schieven [/bib_ref]. JNK regulates the inflammatory response through c-Jun phosphorylation and increased activator protein (AP)-1 [bib_ref] MAPKs and their relevance to arthritis and inflammation, Thalhamer [/bib_ref]. ERK is extensively activated by stimulating factors, and activated ERK induces the translocation of NF-κB into the nucleus, known to be the main mechanism of inflammatory expression, and regulates inflammatory expression [bib_ref] Ligustilide prevents LPS-induced iNOS expression in RAW 264.7 macrophages by preventing ROS..., Chiou [/bib_ref]. These transcription factors are closely associated with chemokine and cytokine production in UVBinduced HaCaT cells. The results showed that QDG significantly inhibited p38, JNK, and ERK phosphorylation by 57%, 47%, and 35%, respectively, at a concentration of 10 μg/mL. In addition, QDG further upregulated the expression level of inhibitory kappa B alpha (IκBα) and inhibited the
## Phosphorylation of p38/jnk/erk/iκb
Among the inflammatory response intracellular signaling pathways, MAPKs are the well-known signaling pathways involved in the inflammatory response [bib_ref] Pathogenesis and therapy of psoriasis, Lowes [/bib_ref] [bib_ref] New insights in the immunologic basis of psoriasis, Nograles [/bib_ref] [bib_ref] Psoriasis: Evolution of pathogenic concepts and new therapies through phases of translational..., Guttman-Yassky [/bib_ref]. Activated MAPKs, in response to cell stimulation, induce the expression of target genes by activating other kinases or transcription factors, such as p38, c-Jun N-terminal kinase (JNK), and ERK. p38 is a central regulator of the inflammatory response regulating IL-6, IL-8, and TNF-α production and the expression of nitric oxide and metalloproteinase [bib_ref] The biology of p38 kinase: A central role in inflammation, Schieven [/bib_ref]. JNK regulates the inflammatory response through c-Jun phosphorylation and increased activator protein (AP)-1 [bib_ref] MAPKs and their relevance to arthritis and inflammation, Thalhamer [/bib_ref]. ERK is extensively activated by stimulating factors, and activated ERK induces the translocation of NF-κB into the nucleus, known to be the main mechanism of inflammatory expression, and regulates inflammatory expression [bib_ref] Ligustilide prevents LPS-induced iNOS expression in RAW 264.7 macrophages by preventing ROS..., Chiou [/bib_ref]. These transcription factors are closely associated with chemokine and cytokine production in UVB-induced HaCaT cells. The results showed that QDG significantly inhibited p38, JNK, and ERK phosphorylation by 57%, 47%, and 35%, respectively, at a concentration of 10 µg/mL. In addition, QDG further upregulated the expression level of inhibitory kappa B alpha (IκBα) and inhibited the phosphorylation of IκBα [fig_ref] Figure 5: Effects of QDG treatment on mitogen-activated protein kinase [/fig_ref]. Within the cell, increased expression of p-IκBα results in the degradation of IκBα, an endogenous inhibitor that prevents NF-κB translocation. These results confirmed that QDG exerts its anti-inflammatory effect by controlling expression of inflammatory factor through p38, JNK, and ERK.
Molecules 2018, 23, x 6 of 13 phosphorylation of IκBα [fig_ref] Figure 5: Effects of QDG treatment on mitogen-activated protein kinase [/fig_ref]. Within the cell, increased expression of p-IκBα results in the degradation of IκBα, an endogenous inhibitor that prevents NF-κB translocation. These results confirmed that QDG exerts its anti-inflammatory effect by controlling expression of inflammatory factor through p38, JNK, and ERK.
## Signaling pathways leading to the activation of nf-κb
The NF-κB signaling pathways have been implicated in the development and progression of chronic inflammation disease. Chronic inflammation disease, such as atopic dermatitis is associated with the activation and expression of pro-inflammatory cytokines [bib_ref] Suppression of thymus-and activation-regulated chemokine (TARC/CCL17) production by 1,2,3,4,6-pentaO-galloyl-b-D-glucose via blockade of..., Ju [/bib_ref] [bib_ref] Reactive oxygen species are involved in the IFN-c-stimulated production of Th2 chemokines..., Qi [/bib_ref]. The translocation of NF-κB from the cytoplasm to the nucleus is a molecular event associated with inflammation. This process results in the transcription of pro-inflammatory genes that contribute to the progression of the inflammation disease [bib_ref] Mechanism of thymusand activation-regulated chemokine (TARC)/CCL17 production and its modulation by roxithromycin, Komine [/bib_ref] [bib_ref] The adenylyl cyclase-cAMP system suppresses TARC/CCL17 and MDC/ CCL22 production through p38..., Qi [/bib_ref]. Therefore, we investigated the effect of QDG on UVB-induced NF-κB translocation. QDG showed 83%, 65%, and 57% NF-κB protein expression in 1, 5, and 10 μg/mL concentration, respectively [fig_ref] Figure 6: Effect of QDG treatment on NF-κB protein expression in HaCaT cells [/fig_ref]. QDG showed stronger inhibitory activity when compared to only UVB-irradiated group, a potent pharmacological inhibitor of NF-κB translocation into the nucleus [fig_ref] Figure 6: Effect of QDG treatment on NF-κB protein expression in HaCaT cells [/fig_ref]. Interestingly, compounds derived from natural products such as curcumin, capsaicin, resveratrol, and green tea polyphenols have been shown to be potent inhibitors of the NF-κB pathway by inhibiting IKK activity [bib_ref] A Beginner's guide to NF-κB signaling pathways, Delhalle [/bib_ref] [bib_ref] NF-kappaB regulation in the immune system, Li [/bib_ref]. Since QDG could be shown to inhibit NF-κB activation, it can be assumed that QDG affects IKK and thus affects the translocation of NF-κB from cytoplasm into the nucleus. Therefore, QDG is considered similar to the way the previously reported Rhizoma coptidis extract affects the NF-κB pathway in HaCaT [bib_ref] Differential effect of Rhizoma coptidis and its main alkaloid compound berberine on..., Enk [/bib_ref]. This approach has been suggested as an indirect method to control inflammatory disease. These results show that QDG activates molecular events that prevent the translocation of NF-κB.
## Signaling pathways leading to the activation of nf-κb
The NF-κB signaling pathways have been implicated in the development and progression of chronic inflammation disease. Chronic inflammation disease, such as atopic dermatitis is associated with the activation and expression of pro-inflammatory cytokines [bib_ref] Suppression of thymus-and activation-regulated chemokine (TARC/CCL17) production by 1,2,3,4,6-pentaO-galloyl-b-D-glucose via blockade of..., Ju [/bib_ref] [bib_ref] Reactive oxygen species are involved in the IFN-c-stimulated production of Th2 chemokines..., Qi [/bib_ref]. The translocation of NF-κB from the cytoplasm to the nucleus is a molecular event associated with inflammation. This process results in the transcription of pro-inflammatory genes that contribute to the progression of the inflammation disease [bib_ref] Mechanism of thymusand activation-regulated chemokine (TARC)/CCL17 production and its modulation by roxithromycin, Komine [/bib_ref] [bib_ref] The adenylyl cyclase-cAMP system suppresses TARC/CCL17 and MDC/ CCL22 production through p38..., Qi [/bib_ref]. Therefore, we investigated the effect of QDG on UVB-induced NF-κB translocation. QDG showed 83%, 65%, and 57% NF-κB protein expression in 1, 5, and 10 µg/mL concentration, respectively [fig_ref] Figure 6: Effect of QDG treatment on NF-κB protein expression in HaCaT cells [/fig_ref]. QDG showed stronger inhibitory activity when compared to only UVB-irradiated group, a potent pharmacological inhibitor of NF-κB translocation into the nucleus [fig_ref] Figure 6: Effect of QDG treatment on NF-κB protein expression in HaCaT cells [/fig_ref]. Interestingly, compounds derived from natural products such as curcumin, capsaicin, resveratrol, and green tea polyphenols have been shown to be potent inhibitors of the NF-κB pathway by inhibiting IKK activity [bib_ref] A Beginner's guide to NF-κB signaling pathways, Delhalle [/bib_ref] [bib_ref] NF-kappaB regulation in the immune system, Li [/bib_ref]. Since QDG could be shown to inhibit NF-κB activation, it can be assumed that QDG affects IKK and thus affects the translocation of NF-κB from cytoplasm into the nucleus. Therefore, QDG is considered similar to the way the previously reported Rhizoma coptidis extract affects the NF-κB pathway in HaCaT [bib_ref] Differential effect of Rhizoma coptidis and its main alkaloid compound berberine on..., Enk [/bib_ref]. This approach has been suggested as an indirect method to control inflammatory disease. These results show that QDG activates molecular events that prevent the translocation of NF-κB.
# Materials and methods
## General procedures
Column chromatography was conducted using 70-230 mesh silica gel (Merck, Darmstadt, Germany). Watchers ® Silica gel Si 60 (70-230 mesh) was used for column chromatography (Isu Industry Co., Seocho, Korea). TLC analysis was carried out on precoated silica gel 60 F254 plates (Merck). Detection of spots on the TLC plate was performed by observation under a UV lamp (Spectroline, model CM-24A, Spectronics Corp., New York, NY, USA) or by spraying 10% aqueous H2SO4 on the developed plate followed by heating. Prep LC was performed with a YMC LC-Forte/R (YMC, Kyoto, Japan). NMR spectra were recorded on a Bruker Ascend 400 and Avance 500 (Bruker, Rheinstetten, Germany). High-resolution electrospray ionization mass spectrometry (HR-ESI-MS) was carried out using a SYNAPT G2 electrospray mass spectrometer (Waters, Elstree, UK) at the Korean Basic Science Institute, Seoul, Korea.
## Reagents
Dulbecco's Modified Eagle Medium (DMEM), fetal bovine serum (FBS), and streptomycinpenicillin were purchased from GIBCO (Grand Island, NY, USA). The recombinant human IL-6, 8 recombinant human TNF-α, recombinant human IFN-γ, Quantikine ELISA kits for Macrophage Derived Chemokine (MDC), Thymus Activation Regulated Chemokine (TARC) were purchased from R&D Systems . The primary antibodies for involucrin, loricrin, filaggrin p38, JNK, and ERK were purchased form Abcam (Cambridge, UK). Antibodies against nuclear factor κB (NF-κB), inhibitory kappa B alpha (IκBα), and phosphorylated IκBα were
# Materials and methods
## General procedures
Column chromatography was conducted using 70-230 mesh silica gel (Merck, Darmstadt, Germany). Watchers ® Silica gel Si 60 (70-230 mesh) was used for column chromatography (Isu Industry Co., Seocho, Korea). TLC analysis was carried out on precoated silica gel 60 F254 plates (Merck). Detection of spots on the TLC plate was performed by observation under a UV lamp (Spectroline, model CM-24A, Spectronics Corp., New York, NY, USA) or by spraying 10% aqueous H 2 SO 4 on the developed plate followed by heating. Prep LC was performed with a YMC LC-Forte/R (YMC, Kyoto, Japan). NMR spectra were recorded on a Bruker Ascend 400 and Avance 500 (Bruker, Rheinstetten, Germany). High-resolution electrospray ionization mass spectrometry (HR-ESI-MS) was carried out using a SYNAPT G2 electrospray mass spectrometer (Waters, Elstree, UK) at the Korean Basic Science Institute, Seoul, Korea. against nuclear factor κB (NF-κB), inhibitory kappa B alpha (IκBα), and phosphorylated IκBα were purchased from Cell Signaling (Beverly, MA, USA). Mouse monoclonal anti-β-actin was obtained from Sigma Aldrich (St. Louis, MO, USA). Polyvinylidene difluoride (PVDF) membrane, Tetramethylethylenediamine (TEMED), Sodium Dodecyl Sulfate (SDS) and acrylamide were purchased from Bio-rad (Hercules, CA, USA). Nuclear and cytoplasmic extraction reagents and first strand cDNA synthesis kit were obtained from Thermo Scientific (Rockford, IL, USA). All other chemical reagents were of the highest pure analytical grade commercially available.
## Reagents
# Plant material
Whole plants of Nymphoides indica (L.) Kuntze were purchased from Agricultural Corporation Lotus Green Co., Ltd. in Gwangju, Korea in May, 2016. The plant taxa were identified using the DNA barcoding system, by comparing the sequences obtained either with public databases (NBCI GenBank) and/or with a database made for this purpose from the National Institute of Biological Resources (NIBR), Korea (voucher No. NIBRVP0000592689).
## Isolation and structure determination of compound 1
The air-dried samples (700 g) were cut into pieces and extracted for three days with 95% methyl alcohol (MeOH) (18 L × 3) at room temperature. The combined crude extracts were concentrated under reduced pressure to yield a MeOH extract (159.8 g) and dissolved in distilled water (H 2 O). The suspended extract was partitioned using n-hexane, methylene chloride (MC), ethyl acetate (EtOAc), and n-butanol to yield layers of 14.5, 2.4, 3.7, and 12.0 g, respectively. A portion of EtOAc fraction (3.0 g) was subjected to silica gel column chromatography with gradient mixtures of EtOAc and MeOH (9:1-4:1) to give seven fractions (Fr. E1-Fr. E7). Fr. E5 was subjected to preparative reverse-phase LC (YMC Actus Triart C18 column; 250 × 20 mm, S-5 µm, 12 mm; flow rate, 10.0 mL/min; 30% acetonitrile in H 2 O for 60 min; UV detection at 254 nm) to afford compounds 1 (167 mg) (t R = 45.0 min) [fig_ref] Figure 1: Chemical structure of quercetin 3,7-dimethyl ether 4′-glucoside [/fig_ref]. purchased from Cell Signaling (Beverly, MA, USA). Mouse monoclonal anti-β-actin was obtained from Sigma Aldrich (St. Louis, MO, USA). Polyvinylidene difluoride (PVDF) membrane, Tetramethylethylenediamine (TEMED), Sodium Dodecyl Sulfate (SDS) and acrylamide were purchased from Bio-rad (Hercules, CA, USA). Nuclear and cytoplasmic extraction reagents and first strand cDNA synthesis kit were obtained from Thermo Scientific (Rockford, IL, USA). All other chemical reagents were of the highest pure analytical grade commercially available.
# Plant material
Whole plants of Nymphoides indica (L.) Kuntze were purchased from Agricultural Corporation Lotus Green Co., Ltd. in Gwangju, Korea in May, 2016. The plant taxa were identified using the DNA barcoding system, by comparing the sequences obtained either with public databases (NBCI GenBank) and/or with a database made for this purpose from the National Institute of Biological Resources (NIBR), Korea (voucher no. NIBRVP0000592689).
## Isolation and structure determination of compound 1
The air-dried samples (700 g) were cut into pieces and extracted for three days with 95% methyl alcohol (MeOH) (18 L × 3) at room temperature. The combined crude extracts were concentrated under reduced pressure to yield a MeOH extract (159.8 g) and dissolved in distilled water (H2O). The suspended extract was partitioned using n-hexane, methylene chloride (MC), ethyl acetate (EtOAc), and n-butanol to yield layers of 14.5, 2.4, 3.7, and 12.0 g, respectively. A portion of EtOAc fraction (3.0 g) was subjected to silica gel column chromatography with gradient mixtures of EtOAc and MeOH (9:1-4:1) to give seven fractions (Fr. E1-Fr. E7). Fr. E5 was subjected to preparative reverse-phase LC (YMC Actus Triart C18 column; 250 × 20 mm, S-5 μm, 12 mm; flow rate, 10.0 mL/min; 30% acetonitrile in H2O for 60 min; UV detection at 254 nm) to afford compounds 1 (167 mg) (tR = 45.0 min) [fig_ref] Figure 1: Chemical structure of quercetin 3,7-dimethyl ether 4′-glucoside [/fig_ref].
## Cell culture and uvb irradiation
Immortalized human keratinocytes (HaCaT) were purchased from the American Type Culture collection (Manassas, VA, USA). The cells were cultured in high-glucose DMEM containing 10% FBS, 1% streptomycin-penicillin at 37 - C in a 5% CO 2 humidified atmosphere. The cells were exposed to UVB radiation using an UV irradiation system (BIO-LINK Crosslinker, WA, Wembley, Australia) delivering the 280-320 nm wavelength range, with maximum emission at 312 nm. Seeded cells were rinsed with PBS and then exposed to 20 mJ/cm 2 of UVB.
## Cell migration
HaCaT cells were incubated, at 5 × 10 5 cells/mL for 24 h, in a cell culture incubator. Next, the cell monolayers were scratched with a 200-µL yellow tip and washed once with phosphate-buffered saline (PBS). Next, cell monolayers were treated with different concentrations of QDG (1, 5, and 10 µg/mL) and cultured in a CO 2 incubator for 24 h. Cell motility was assessed 24 h later, using a photomicroscope, and the scratched area was measured. Measurements were taken to determine the distance traveled, in the 24 h period, by measuring the scratched area in the photographed pictures.
## Immunoassays for cytokines and chemokines
HaCaT cells (1 × 10 5 cells/300 µL or 5 × 10 5 cells/400 µL for the cytokine or chemokine assay, respectively) were grown in a 24-well plate and treated with UVB (20 mJ/cm 2 ). After centrifugation at 412× g for 10 min, the amounts of TNF-α, IL-1β, IL-6, IL-8, MDC and TARC in the culture supernatant were analyzed using the corresponding enzyme-linked immunosorbent assay (ELISA) kits, according to the manufacturer's instructions. The absorbance was measured at 450 nm using a microplate reader (Magellan; Tecan Ltd, Salzburg, Austria).
## Measurement of skin barrier peptide and hyaluronic acid
HaCaT cells were seeded in six-well plates, at a density of 1 × 10 5 cells/well. The cells were starved for 24 h, after which they were stimulated with 1, 5, and 10 µg/mL of QDG for 24 h. Supernatants were collected and ELISA kits utilized to measure relative filaggrin, loricrin, and HA production, according to the manufacturer's instruction.
## Preparation of cytosolic and nuclear extracts
HaCaT cells (5 × 10 6 cells/mL) were treated with LPS for 30 min, at 37 - C. Keratinocyte cytosolic and nuclear extracts were prepared as previously described [bib_ref] Suppressive effects of ethanol extract of Aralia elata on UVB-induced oxidative stress..., Kwak [/bib_ref]. Keratinocytes were harvested by centrifugation at 412× g for 10 min and washed twice with PBS. The cells were suspended in 400 µL of lysis buffer (10 µM KCl, 1.5 µM MgCl 2 , 0.1 µM EDTA, 0.1 µM EGTA, 1 µM dithiothreitol, 0.5 µM PMSF, 1 µM sodium orthovanadate, 2 µg/mL aprotinin, 2 µg/mL leupeptin, and 10 mM Hepes-KOH, pH 7.8) and were allowed to swell on ice for 15 min. Next, 25 µL of a 10% Nonidet NP-40 solution (final concentration: approximately 0.6%) were added, and the tubes were vigorously vortexed for 10 s. The homogenates were centrifuged at 12,000× g for 10 min at 4 - C. The supernatants were stored as cytoplasmic extracts and kept at −70 - C. The nuclear pellets were re-suspended in 50 µL of an ice-cold hypertonic solution containing 5% glycerol and 0.4 M NaCl lysis buffer. Furthermore, the tubes were incubated on ice for 30 min and then centrifuged at 12,000× g for 15 min at 4 - C. The supernatants were collected as nuclear extracts and stored at −70 - C. Protein concentrations were determined using the Bradford method according to the manufacturer's instructions (Bio-Rad Laboratories).
## Western blot assay
HaCaT cells were collected on ice, washed three times with ice-cold PBS, and treated with a homogenizing buffer containing protease inhibitor cocktail (Roche Diagnostics, Indianapolis, IN, USA). After brief sonication, the cell lysates were centrifuged at 12,000 rpm for 10 min, and supernatants were collected. Next, the protein concentrations were determined using Bradford protein assay reagent (Bio-Rad Laboratories). Twenty micrograms of the protein were separated on a 7.5-10% SDS gel and then transferred to a PVDF membrane, which was then probed with specific primary antibodies overnight with gentle shaking, followed by incubation with secondary antibodies for 1 h. Blots were developed using enhanced chemiluminescence (Amersham Biosciences, Little Chalfont, Buckinghamshire, UK) and quantified using a Gel-pro analyzer (Media Cybernetics Inc., Rockville, MD, USA).
## Immunofluorescence
HaCaT cells were aliquoted in an eight-well Lab-Tek chamber (Nalge-Nunc, Madison, WI, USA) with 1 × 10 3 cells and allowed to grow for 24 h after QDG treatment. Next, they were washed with cold PBS three times and 95% Triton X-100 was added for 10 min. After washing with PBS, 1% of bovine serum albumin was added, and the cells were incubated for 1 h. Next, the c-fos primary antibody (1:100) was added, and the cells were incubated at 4 - C overnight. In the next step, cells were treated with a secondary antibody, Alexa 488-conjugated goat anti-mouse immunoglobulin G (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA), and fluorescein isothiocyanate (1:1000). Stained cells were then mounted on a slide after washing with PBS and observed by a fluorescent microscope for NF-κB activity.
# Statistical analysis
Analysis of variance was performed in SPSS (SPSS Inc., Chicago, IL, USA). All data are expressed as mean ± SD, and statistically significant differences between experimental and control values were analyzed by one-way ANOVA followed by a t-test. * p-value < 0.001, ** p-value < 0.0001 was considered statistically significant.
# Conclusions
We investigated the effects of QDG, a Nymphoides indica component, against the activity of inflammation-related factors in UVB-irradiated keratinocytes. QDG inhibited TNF-α, IL-1β, IL-6, and IL-8 in a dose-dependent manner and also inhibited the expression of TARC and MDC induced by UVB irradiation. In the UVB-exposed group, the expression of filaggrin, involucrin, loricrin, and HAS-1 was decreased compared with normal cells. We also investigated whether the anti-inflammatory effects of QDG are due to its modulation of NF-κB. QDG treatment in keratinocytes not only reduced the phosphorylation of p38, JNK, and ERK in a concentration-dependent manner, but also decreased NF-κB levels and the generation of chronic inflammatory disease factors due to UVB irradiation. According to the report, glycolic acid induced by UVB stimulation inhibits overexpressed factors by UVB-mediated NF-κB signaling in the HaCaT cells, suggesting that it inhibits inflammation [bib_ref] Topical application of glycolic acid suppresses the UVB induced IL-6, IL-8, MCP-1..., Tang [/bib_ref]. Our study also suggests that QDG has an anti-inflammatory effect on the overexpression of inflammatory factors by UVB stimulation. Taken together, these findings suggest that QDG may be useful for the treatment of chronic inflammatory skin diseases.
[fig] Figure 1: Chemical structure of quercetin 3,7-dimethyl ether 4′-glucoside (QDG) (A) and increased cell proliferation and migration activities of QDG-treated human keratinocytes (HaCaT) cells (B). HaCaT cells were scratched using a yellow tip. Migration levels of HaCaT cells were observed using an optical microscope and photographs were obtained. HaCaT cells were treated with different concentrations of QDG (1, 5, and 10 μg/mL) for 24 h. QDG treatment leads to an increase in migration of HaCaT cells. Nor: No treatment cell group (0 h), Cont: 20 mJ/cm 2 ultraviolet B (UVB) treatment cell group, QDG: QDG treatment group. [/fig]
[fig] Figure 2: Effect of QDG treatment on cytokine expression in HaCaT cells. HaCaT cells were treated with different concentrations of QDG (1, 5, and 10 μg/mL) after irradiation with 20 mJ/cm 2 UVB. After 24 h, cytokine expression was determined in the cell supernatant according to the kit manual. Each value represents mean ± SD for the three individual experiments. Nor: No treatment cell group (0 h), Cont: 20 mJ/cm 2 UVB treatment cell group, QDG = QDG treatment group, EGCG = positive control. n = 3, * = p < 0.001 and ** = p < 0.0001 compared with the control group. [/fig]
[fig] Figure 3: Effect of QDG treatment on macrophage-derived chemokine (MDC) and thymus-and activation-regulated chemokine (TARC) expression in HaCaT cells. HaCaT cells were treated with different concentrations of QDG (1, 5, and 10 μg/mL) after irradiation with 20 mJ/cm 2 UVB. After 24 h, chemokine expression was determined in the cell supernatant according to the kit manual. Each value represents mean ± SD for the three individual experiments. Nor: No treatment cell group (0 h), Cont: 20 mJ/cm 2 UVB treatment cell group, QDG = QDG treatment group, EGCG = positive control. n = 3. * = p < 0.001 and ** = p < 0.0001 compared with the control group. [/fig]
[fig] Figure 4: Effect of QDG treatment on skin barrier and hyaluronic acid synthase expressions in HaCaT cells. [/fig]
[fig] Figure 5: Effects of QDG treatment on mitogen-activated protein kinase (MAPK) phosphorylation expression in HaCaT cells. HaCaT cells were treated with different concentrations of QDG (1, 5, and 10 μg/mL) after irradiation with 20 mJ/cm 2 UVB. After 30 min, cells were harvested and relative protein levels were determined. Histogram shows the densitometry of phosphorylated-p38, -c-Jun Nterminal kinase (JNK), extracellular signal-regulated kinase (ERK), and inhibitory kappa B alpha (IκBα) proteins normalized to GAPDH. Each value represents mean ± SD for the three individual experiments. Nor: No treatment group (0 h), Cont: 20 mJ/cm 2 UVB treatment group, QDG: QDG treated group. n = 3, * = p < 0.001 and ** = p < 0.0001 compared with the control group. [/fig]
[fig] Figure 6: Effect of QDG treatment on NF-κB protein expression in HaCaT cells. HaCaT cells were treated with different concentrations of QDG (1, 5, and 10 μg/mL) after irradiation with 20 mJ/cm 2 UVB. After 6 h, cells were harvested, and (A) protein and (B) NF-κB-FITC levels were determined. Histogram shows the densitometry of NF-κB protein normalized to glyceraldehyde 3-phosphate dehydrogenase. Each value represents mean ± SD for the three individual experiments. Nor: No treatment group (0 h), Cont: 20 mJ/cm 2 UVB treatment group, QDG = QDG treatment group. n = 3, * = p < 0.001 and ** = p < 0.0001 compared with the control group. [/fig]
[fig] Figure 7: Separation procedure of methanol extract from Nymphoides indica. [/fig]
[fig] Author: Contributions: Y.A.K., D.H.K., and B.J.P. conceived the idea and designed the experiments. Y.A.K., D.H.K., C.B.P., and T.S.P. conducted the experiments and analyzed the data. Y.A.K. and D.H.K. wrote the manuscript, handled the required revisions, and supervised the process. B.J.P. discussed the study. [/fig]
[fig] Funding: This study was supported by a grant from the Korea Healthcare Technology R&D Project, Ministry of Health & Welfare, Republic of Korea (HN15C0103). [/fig]
[bib_ref] Studies on the antimicrobial effect of collected bee venom using electric shock..., Han [/bib_ref]
|
10.3390/s23156801
|
CCBY
|
10422573
|
37571585
|
s2orc_pubmed_articles
|
Escalator Foundation Bolt Loosening Fault Recognition Based on Empirical Wavelet Transform and Multi-Scale Gray-Gradient Co-Occurrence Matrix
An escalator is an essential large-scale public transport equipment; once it fails, this inevitably affects the operation of the escalator and even leads to safety concerns, or perhaps accidents. As an important structural part of the escalator, the foundation of the main engine can cause the operation of the escalator to become abnormal when its fixing bolts become loose. Aiming to reduce the difficulty of extracting the fault features of the footing bolt when it loosens, a fault feature extraction method is proposed in this paper based on empirical wavelet transform (EWT) and the gray-gradient co-occurrence matrix (GGCM). Firstly, the Teager energy operator and multi-scale peak determination are used to improve the spectral partitioning ability of EWT, and the improved EWT is used to decompose the original foundation vibration signal into a series of empirical mode functions (EMFs). Then, the gray-gradient co-occurrence matrix of each EMF is constructed, and six texture features of the gray-gradient co-occurrence matrix are calculated as the fault feature vectors of this EMF. Finally, the fault features of all EMFs are fused, and the degree of the loosening of the escalator foundation bolt is identified using the fused multi-scale feature vector and BiLSTM. The experimental results show that the proposed method based on EWT and GGCM feature extraction can diagnose the loosening degree of foundation bolts more effectively and has a certain engineering application value.
# Introduction
Escalators have become an important transportation tool in modern city life. Escalators are commonly used in large shopping malls, railway stations and subway stations, and other large buildings, transporting a large number of people and equipment every day. And, the safety of escalators is closely related to the safety of the public. Once an escalator malfunction occurs, it may affect the operation of the escalator to malfunction, in mild cases, or cause serious accidents, which can cause great damage to people's lives as well as the economy [bib_ref] Study on the cause and prevention of subway escalator passenger injury accident, Liu [/bib_ref] [bib_ref] High frequency demodulation technique for instantaneous angular speed estimation, Bonnardot [/bib_ref]. Due to frequent overtime and overload operation, loosening of the anchor bolts of the escalator's mainframe footing occurs from time to time. As the key part of the escalator, the loosening of the fixing bolts of the mainframe feet leads to periodic shocks when the escalator is operating, which causes the vibration of the escalator coupling system to intensify, in turn affecting the stability of the escalator's operation, thus endangering the operational safety of the escalator in serious cases. Therefore, the condition monitoring and fault diagnosis of escalator bracket fixing bolts is essential. The vibration signals of the footing of the escalator are usually nonlinear and non-stationary [bib_ref] Research on measuring device and quantifiable risk assessment method based on FMEA..., Ma [/bib_ref]. Moreover, due to the complex operating environment, the collected vibration signals often contain a large amount of noise and interference signals, which makes it difficult to effectively extract the features of the early footing bolt loosening fault. How to effectively suppress noise interference and extract effective bolt loosening fault features from nonlinear and Sensors 2023, [bib_ref] An expert system for EMI data classification based on complex Bispectrum representation..., Mitiche [/bib_ref] non-stationary footing vibration signals is a key problem that must be solved in escalator condition monitoring. In view of the nonlinear and non-stationary nature of mechanical fault signals and the characteristics of weak early fault characteristics and susceptibility to noise interference, numerous vibration signal fault diagnosis methods have been proposed, such as short-time Fourier transform (STFT), wavelet transform (WT), empirical mode decomposition (EMD), variational mode decomposition (VMD), etc. However, the window function of STFT cannot be adaptively adjusted according to the frequency of the signal itself, which affects the accuracy of fault diagnosis [bib_ref] Extended Empical Wavelet Transformation: Application to structural updating, Karimpour [/bib_ref] [bib_ref] Sensitive Sub-band Selection Criteria for Empirical Wavelet Transform to Detect Bearing Fault..., Sharma [/bib_ref]. The selection of wavelet basis functions and decomposition layers of WT lacks adaptivity [bib_ref] Handling Data Heterogeneity in Electricity Load Disaggregation via Optimized Complete Ensemble Empirical..., Chui [/bib_ref] [bib_ref] Time-Varying Parameter Identification of Bridges Subject to Moving Vehicles Using Ridge Extraction..., Li [/bib_ref]. Empirical mode decomposition (EMD) [bib_ref] Fault diagnosis of rolling bearing based on empirical mode decomposition and improved..., Sun [/bib_ref] has made a significant breakthrough in vibration signal fault information extraction, but EMD suffers from serious mode aliasing phenomena and endpoint effects and lacks the necessary theoretical foundation [bib_ref] An adaptive empirical mode decomposition and stochastic resonance system in high efficient..., Wang [/bib_ref] [bib_ref] Detection of SSVEP based on empirical mode decomposition and power spectrum peaks..., Antelis [/bib_ref]. In order to surmount the shortcomings of EMD methods, many improved EMD algorithms have been proposed, such as local mean decomposition (LMD) [bib_ref] Application of adaptive complementary ensemble local mean decomposition in underwater acoustic signal..., Lu [/bib_ref] , local characteristic-scale decomposition (LCD) [bib_ref] Fault diagnosis of automaton based on local characteristic-scale decomposition and individual feature..., Du [/bib_ref] , ensemble EMD (EEMD), etc. However, these methods cannot completely solve the limitations of EMD. Variational mode decomposition (VMD) [bib_ref] Applicationof variational mode decomposition optimized with improved whale optimization algorithm in bearing..., Wang [/bib_ref] overcomes the deficiencies of EMD and LMD. Signal decomposition is transformed into a variational problem to determine the center frequency and bandwidth of the component signals by seeking the optimal solution of the variational problem so as to achieve the effective separation of the component signals. VMD has a sound theoretical basis and can better suppress modal aliasing. However, the combination of parameters and the number of decompositions of the penalty factor need to be determined before decomposition because different combinations of parameters and the number of decompositions can affect the decomposition accuracy of the signal, which poses great difficulties to the accurate decomposition of the signal [bib_ref] A multiverse optimization based colour image segmentation using variational mode decomposition, Chouksey [/bib_ref] [bib_ref] Research on a NovelImproved Adaptive Variational Mode Decomposition Method in Rotor Fault..., Yan [/bib_ref].
Recently, Gilles [bib_ref] Seismic Reservoir Delineation via Hankel Transform Based Enhanced Empirical Wavelet Transform, Li [/bib_ref] proposed empirical wavelet transform (EWT) to perform the adaptive decomposition of non-stationary signals. The main idea is to extract each mode component of the signal by constructing a series of suitable band-pass filter banks.
Compared with EMD and VMD methods, EWT has a reliable mathematical theory foundation, and the filter frequency band is self-adaptive according to the signal spectrum, which avoids mode aliasing, and the end effect is better. At the same time, because EWT is not decomposed in an iterative way, the decomposition speed is very fast. Due to the above advantages, EWT has been widely used in the identification of the fault information of rolling bearings [bib_ref] A double impulsiveness measurement indices-bilaterally driven empirical wavelet transform and its application..., Ding [/bib_ref] [bib_ref] Feature extraction method based on adaptive and concise empirical wavelet transform and..., Zhang [/bib_ref] and fan bearings. In this paper, EWT is introduced into the processing of the foundation vibration signals and is used to perform the multi-scale decomposition of foundation vibration signals. In order to better suppress the influence of noise and non-correlated vibration on EWT band division, the EWT is improved by using the Teager energy operator and multi-scale peak location so that it can separate different modes of the bolt loosen fault signal more effectively.
After decomposing the machine feet vibration signal into a group empirical mode function (EMF) by EWT, how to accurately extract the fault features from each mode component is another key issue in the identification of bolt loosening fault. The EMFs decomposed by EWT contain rich feature information, and different features represent different physical meanings. Choosing a suitable feature extraction method can significantly improve the recognition accuracy. The gray-gradient co-occurrence matrix (GGCM), based on bispectrum analysis, is a very effective fault feature extraction method. The gray-gradient co-occurrence matrix contains all the phase and texture information of processed signals and can better suppress the influence of Gaussian noise [bib_ref] An expert system for EMI data classification based on complex Bispectrum representation..., Mitiche [/bib_ref]. Liu et al. [bib_ref] Surface crack characterization using laser nonlinear ultrasonics based on the bispectrum, Liu [/bib_ref] applied the GGCM to the detection of microcracks under mixed frequency excitation. Wang et al. [bib_ref] Fault Characteristic Extraction by Fractional Lower-Order Bispectrum Methods, Wang [/bib_ref] used the fractional GGCM of bispectrum analysis to identify fault characteristics, and the results showed that the GGCM can effectively extract the fault features from small cracks. Xu et al. [bib_ref] A phase linearisation-based modulation signal bispectrum for analysing cyclostationary bearing signals. Struct...., Xu [/bib_ref] applied the GGCM based on the bispectrum analysis method to identify and analyze the signals of bearing; the experimental results in this study verify the effectiveness of the GGCM based on the bispectrum method in microcrack feature extraction.
Based on the above analysis, a feature extraction method for escalator foundation bolt loosening faults based on EWT and the gray-gradient co-occurrence matrix is proposed in this paper; the method ensures the identification of foundation bolt loosening faults and determines their degree of looseness. Firstly, in order to reduce the influence of noise and irrelevant components, the Teager energy operator and multi-scale peak location method are used to determine the EWT spectral segmentation boundary, and the improved EWT is used to decompose the foundation vibration signal into a group of empirical mode functions (EMFs). Then, in order to avoid complex parameter settings and accurately extract the fault features of loose foundation bolts, the bispectrum and grayscale gradient co-occurrence matrix of each EMF are calculated, and six fault features are extracted through the GGCM of each EMF. Finally, the fault features of all EMFs are fused into an 18-dimensional fault feature vector, and bidirectional long short-term memory (BiLSTM) is used to identify the loosening status of foundation bolts. Experimental analysis is conducted using the measured vibration signals of the escalator foundation, and the results show that the proposed method can effectively identify the looseness faults of the foundation bolts and determine the degree of bolt looseness.
The remainder is structured as follows. Section 2 introduces the basic theory of EWT and Bi-LSTM. The fault features of foundation bolt loosening are extracted using the GGCM based on bispectrum in Section 3. Section 4 constructs the bolt loosening diagnosis model based on EWT and the multi-scale GGCM. An experiment verification and performance comparisons are shown to demonstrate the effectiveness and validity of the proposed method in Section 5. Finally, a conclusion is drawn in Section 6.
## Theories of ewt and bi-lstm
## Ewt and spectrum division
## Ewt
Empirical wavelet transform (EWT) is constructed on the basis of wavelet theory. EWT consists of two important steps: First, the adaptive partitioning of the signal spectrum; then, the signal is decomposed using an orthogonal wavelet filter bank to obtain a modal component signal with tight support characteristics. It is assumed that the spectrum range of the vibration signal after Fourier transform is [0, π]. By dividing the whole spectrum into N segments, each segment of the spectrum is denoted as Λ n = [ω n−1 , ω n ] n = 1, 2, . . . , N. ω 0 and ω N are the left and right boundaries of the spectrum division, respectively. Then, the whole spectrum of the signal can be represented as ∪ N n=1 Λ n = [0, π]. EWT is a bandpass filter defined on each spectrum Λ n . According to the theory of the wavelet, the scaling function ϕ(x) and wavelet function ψ(x) of EWT are defined in the frequency domain as follows [bib_ref] Seismic Reservoir Delineation via Hankel Transform Based Enhanced Empirical Wavelet Transform, Li [/bib_ref] :φ
[formula] n (ω) = 1; |ω| ≤ (1 − λ)ω n cos π 2 β 1 2λω n ( ω −(1 − λ)ω n ) (1 − λ)ω n ≤ |ω| ≤ (1 + λ)ω n 0; others ; (1) ψ n (ω) = 1; (1 + λ)ω n ≤ |ω| ≤ (1 − λ)ω n+1 cos π 2 β 1 2λω n+1 ( ω −(1 − λ)ω n+1 ) ; (1 − λ)ω n+1 ≤ |ω| ≤ (1 + λ)ω n+1 sin π 2 β 1 2λω n ( ω −(1 − λ)ω n ) (1 − λ)ω n ≤ |ω| ≤ (1 + λ)ω n+1 0; others(2) [/formula]
where β n = y 4 (35 − 84y + 70y 2 − 20y 3 ), 0 < λ < 1, and λ < min n ω n+1 −ω n ω n+1 +ω n . The decomposition of EWT is similar to that of classical wavelets. The decomposed detail and approximation coefficients can be defined as:
[formula] W ε g (n, x) =< g(x), ψ n (x) > = g(τ)ψ n (τ − x)dτ = F −1 [ĝ(ω)ψ n (ω)] (3) W ε g (0, x) =< g(x), φ 1 (x) > = g(τ)φ 1 (τ − x)dτ = F −1 [ĝ(ω)φ 1 (ω)](4) [/formula]
where F −1 (·) denotes the inverse Fourier transform, andφ 1 (ω) andψ n (ω) are obtained by Equations (1) and [bib_ref] High frequency demodulation technique for instantaneous angular speed estimation, Bonnardot [/bib_ref]. In empirical wavelet transform, the reconstructing formula is:
[formula] g(x) = W ε g (0, x)φ 1 (x) + N ∑ n=1 W ε g (n, x)ψ n (x) = F −1 [Ŵ ε g (0, ω)φ 1 (ω) + N ∑ n=1Ŵ ε f (n, ω)ψ n (ω)](5) [/formula]
According to Formulas (3) and (4), the empirical mode function (EMF) formula after EWT decomposition can be calculated using the following formulas:
[formula] g 0 (x) = W ε g (0, x)φ 1 (t) g n (x) = W ε g (n, x)ψ n (x) n = 1, 2 . . . , N(6) [/formula]
A series of empirical mode functions (EMFs) can be obtained after the machine feet vibration signal is decomposed using EWT. Moreover, bispectrum analysis is performed for each EMF to extract the fault eigenvectors for each mode.
## Spectrum division improvement in ewt
According to the basic concept of EWT, whether the component signals obtained by EWT are single or not largely depends on the adaptive division of the Fourier spectrum. If the boundary is detected by the basic "locmaxmin" method in EWT, the detection boundary is the minimum value between the two maximum values. Therefore, if the signal is severely affected by noise interference, the "locmaxmin" method can easily cause some frequency bands to be too finely divided while others are not able to be reasonably divided. For the fault signal of escalator foot looseness, the sideband of fault vibration will be easily divided into different kinds of components, thereby resulting in unreasonable spectrum segmentation and the unfavorable demodulation analysis of fault features. In this paper, in order to better separate the fault signals of different modes, the Teager energy operator is used to concentrate the energy for the Fourier spectrum. Meanwhile, the abovementioned method can reduce the influence of noise and irrelevant components. Multi-scale peak search and location is utilized to identify the spectrum segmentation boundary of EWT adaptively while the mode decomposition ability is enhanced. The concrete steps based on the Teager energy operator and multi-scale peak location (TMPL-EWT) are as follows:
Step 1: Utilize Teager energy operator to concentrate energy on the spectrum. For discrete-time signals x(n), the Teager energy operator is defined as [bib_ref] Advanced Phase-Difference Estimation Algorithm Using Cross Teager-Kai ser Energy Operator for DERs..., Lee [/bib_ref] :
[formula] Ψ(x(n)) = [x(n)] 2 − x(n − 1)x(n + 1)(7) [/formula]
Let the Fourier transform amplitude of the footing vibration signal bef (n). Then, the Teager energy operator transformation result is:
[formula] Ψ(f (n)) = [f (n)] 2 −f (n − 1)f (n + 1)(8) [/formula]
It is more reliable to determine the spectrum segmentation boundary from the perspective of spectrum energy than solely from the perspective of amplitude; at the same time, the interference of noise on segmentation is greatly decreased.
Step 2: Multi-scale peak localization In order to accurately locate the position of the "peak" in the energy spectrum Ψ(f (n)), this paper proposes a multi-scale peak search and location method whose basic idea is to use a set of window functions with different widths to smooth the data. Because the local maximum on different scales is related, the accurate identification of the peak value can be conducted by the relevance.
For a given N × 1 dimensional input sequence s, the peak discriminant criterion is defined as Ω, which is a N × 1 dimension vector and its initialization is Ω = {0} N×1 .
(i) Detect all local maximum points of s, form a set Γ, and make the first update to the discriminant criterion Ω, so that: s(k) → Ω(k), ∀k ∈ Γ .
[formula] Ω(k) = Ω(k) + s(k)·m, ∀k ∈ Γ m(9) [/formula]
(iv) After m scale iterations, the position of the spectral "peak" can be determined by Ω. For a given number P of peak detections, the first P points with the highest values are taken in Ω.
Step 3: EWT spectrum division The midpoint of the selected adjacent peak positions is used as the empirical wavelet spectrum segmentation boundary. Corresponding empirical wavelet filter banks are established, and corresponding modal components can be extracted.
To verify the effectiveness of this method, a set of simulation fault signals are established for experiment analysis. The simulation signal is as follows:
[formula] s 1 (t) = (1 + 0.7 sin(20πt) cos(200πt + 0.7 sin(20πt)) [/formula]
s 2 (t) = (0.9 + 1.8 sin(30πt) sin(600πt + 0.5 cos(30πt)) s 3 (t) = (1 + sin(40πt)) cos(1000πt + 0.8 sin(40πt))
[formula] s(t) = s 1 (t) + s 2 (t) + s 3 (t) + n(t)(10) [/formula]
where n(t) denotes white noises with zero mean value. The simulation signal s(t) consists of three mode signals whose center frequencies are 100 Hz, 300 Hz, and 500 Hz, respectively. After adding in zero-mean Gaussian white noise with a signal-to-noise ratio of dB to the simulation signal, the time domain and frequency domain information of the simulation signal are presented in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , where the sampling frequency is set as f s = 2000 Hz, and the sampling time is set as 1 s.
The number of components of the simulation signal is three. Considering the influence of noise, the modal component estimator is set as N = 4 here. The results of directly using the EWT algorithm to perform the boundary detection and spectrum segmentation on the simulation signal are presented in [fig_ref] Figure 2: Comparison of spectrum division results of EWT [/fig_ref]. Due to the large spectral amplitude of component two, when using "Locmaxmin" for boundary detection, the detected boundaries are all concentrated in component two with the phenomenon of over decomposition. Moreover, part of the modulation signal from component three and component two is mixed in the same mode. Therefore, that is not conducive to the demodulation analysis. The results of using the TMPL-EWT algorithm to perform boundary detection and spectrum segmentation on the simulation signal are shown in [fig_ref] Figure 2: Comparison of spectrum division results of EWT [/fig_ref]. It can be seen that the TMPL-EWT method effectively detects the resonance frequency bands excited by three intrinsic frequencies, thereby avoiding frequency band breakage caused by the improper segmentation of the original EWT, resulting in the achievement of a better separation of fault information.
[formula] ( ) ( ) ( ) ( ) ( ) s t t t t s t t t t s t t t t [/formula]
s t s t s t s t n t π π π π π π π π π = +
[formula] + = + + = + + = + + + (10) [/formula]
where ( ) n t denotes white noises with zero mean value. The simulation signal ( ) s t consists of three mode signals whose center frequencies are 100 Hz, 300 Hz, and 500 Hz, respectively. After adding in zero-mean Gaussian white noise with a signal-to-noise ratio of dB to the simulation signal, the time domain and frequency domain information of the simulation signal are presented in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , where the sampling frequency is set as s f = 2000 Hz, and the sampling time is set as 1 s. The number of components of the simulation signal is three. Considering the influence of noise, the modal component estimator is set
[formula] as 4 N = [/formula]
here. The results of directly using the EWT algorithm to perform the boundary detection and spectrum segmentation on the simulation signal are presented in [fig_ref] Figure 2: Comparison of spectrum division results of EWT [/fig_ref]. Due to the large spectral amplitude of component two, when using "Locmaxmin" for boundary detection, the detected boundaries are all concentrated in component two with the phenomenon of over decomposition. Moreover, part of the modulation signal from component three and component two is mixed in the same mode. Therefore, that is not conducive to the demodulation analysis. The results of using the TMPL-EWT algorithm to perform boundary detection and spectrum segmentation on the simulation signal are shown in [fig_ref] Figure 2: Comparison of spectrum division results of EWT [/fig_ref]. It can be seen that the TMPL-EWT method effectively detects the resonance frequency bands excited by three
## Bi-lstm
## Lstm
Long short-term memory (LSTM) is an improved model based on the recurrent ral network (RNN). The structure of LSTM is shown in [fig_ref] Figure 3: LSTM network structure diagram [/fig_ref] , which contains four gate structures: input gate i , forget gate f , control gate c , and output gate o . The input gate i determines which new information will be stored in the new cell state, and the calculation of the input gate at moment t is defined as:
[formula] i t = σ(W i ·[h t−1 , x t ] + b i )(11) [/formula]
The forget gate f determines which information should be ignored from prior memory, and the forgetting gate at moment t is calculated and as defined as:
[formula] f t = σ(W f ·[h t−1 , x t ] + b f )(12) [/formula]
The control gate updates when the control unit status changes from c t−1 to c t according to Equations [bib_ref] An adaptive empirical mode decomposition and stochastic resonance system in high efficient..., Wang [/bib_ref] and [bib_ref] Detection of SSVEP based on empirical mode decomposition and power spectrum peaks..., Antelis [/bib_ref].
[formula] c t = tanh(W c - [h t−1 , x t ] + b c ) (13) c t = f t - c t−1 + i t - c t(14) [/formula]
The output gate o generates the output and updates hidden vector h t−1 . The control process of the output gate is defined as:
[formula] o t = σ(W o - [h t−1 , x t ] + b o )(15)h t = o t - tanh(c t )(16) [/formula]
In Equations (11) state, respectively; σ(·) and tanh(·) are the transfer functions; - represents the inner vector product; and the symbol - denotes multiplication by elements.
## Bi-lstm
Bidirectional recurrent neural networks (BRNNs) form a bidirectional network structure by adding a backpropagation layer to a recurrent neural network in order to use contextual information simultaneously. In the bidirectional network structure, the RNN units are replaced by LSTM units to form a bidirectional LSTM (Bi-LSTM). The structure of the Bi-LSTM network is given in [fig_ref] Figure 4: BI-LSTM network structure diagram [/fig_ref] , in which two independent LSTM networks are included to propagate information forwards and backwards, respectively. The Bi-LSTM network has the advantages of both the RNN and LSTM network. Bi-LSTM overcomes the decline problem which the RNN had. Through a parallel forward propagation network and backward propagation network, an estimate of the impact of forward and backward events on current events can effectively be made.
## Multi-scale fault feature extraction based on ggcm
# Bispectrum analysis theory
Bispectrum can effectively suppress Gaussian noise with a high resolution and can obtain the signal amplitude, phase, energy, and other related information. Bispectrum analysis is simple to calculate, but it still contains all the feature information of the higherorder spectrum. Therefore, in this paper, bispectrum analysis is used to extract information about the fault characteristics in the machine feet vibration signal. The steps of bispectrum calculation using the direct method are as follows [bib_ref] An expert system for EMI data classification based on complex Bispectrum representation..., Mitiche [/bib_ref] [bib_ref] Surface crack characterization using laser nonlinear ultrasonics based on the bispectrum, Liu [/bib_ref] :
(1) The vibration signal to be analyzed is divided into K segments, with each segment containing M samples. So, the signal after segmentation is:
[formula] ( ) ( ) ( ) ( ) ( ) { (0), (1), , ( 1)} k k k k n x x x M = − x , 1,2..., k K =(17) [/formula]
(2) For the k th segment of data ( ) ( ) k n x , calculate its discrete Fourier transform.
(3) Calculate the third-order autocorrelation coefficients of DFT.
[formula] 1 1 ( ) 1 ( , ) ( ) L L k b X i λ λ λ = + [/formula]
## Multi-scale fault feature extraction based on ggcm
# Bispectrum analysis theory
Bispectrum can effectively suppress Gaussian noise with a high resolution and can obtain the signal amplitude, phase, energy, and other related information. Bispectrum analysis is simple to calculate, but it still contains all the feature information of the higherorder spectrum. Therefore, in this paper, bispectrum analysis is used to extract information about the fault characteristics in the machine feet vibration signal. The steps of bispectrum calculation using the direct method are as follows [bib_ref] An expert system for EMI data classification based on complex Bispectrum representation..., Mitiche [/bib_ref] [bib_ref] Surface crack characterization using laser nonlinear ultrasonics based on the bispectrum, Liu [/bib_ref] :
(1) The vibration signal to be analyzed is divided into K segments, with each segment containing M samples. So, the signal after segmentation is:
[formula] x (k) (n) = x (k) (0), x (k) (1), . . . , x (k) (M − 1) , k = 1, 2 . . . , K(17) [/formula]
(2) For the kth segment of data x (k) (n), calculate its discrete Fourier transform. (3) Calculate the third-order autocorrelation coefficients of DFT.
[formula] X (k) (λ) = 1 M M−1 ∑ n=0 x (k) (n) exp(−j 2πnλ M ), λ = 0, 1 . . . , M/2; k = 1, 2 . . . , K(18)b k (λ 1 , λ 2 ) = 1 ∆ 2 0 L 1 ∑ i 1 =−L 1 L 1 ∑ i 2 =−L 1 X (k) (λ 1 + i 1 )· ·X (k) (λ 2 + i 2 )·X (k) (−λ 1 − λ 2 − i 1 − i 2 )(19) [/formula]
where ∆ 0 = f s /N 0 , M = (2L 1 + 1)N 0 , and f s denotes the sampling frequency. (4) Calculate the bispectrum estimation of the vibration signal.
[formula] B(ω 1 , ω 2 ) = 1 K K ∑ k=1 b k (ω 1 , ω 2 )(20) [/formula]
where ω 1 = 2π f s N 0 λ 1 and ω 2 = 2π f s N 0 λ 2 . It can be seen from the definition that the bispectrum is a complex spectrum with two frequency variables, ω 1 and ω 2 . Bispectrum has 12 symmetrical regions in the frequency plane composed of ω 1 and ω 2 . Only the bispectrum values in the main region need to be calculated, and then, all the bispectrum values in the (ω 1 , ω 2 ) plane can be calculated according to its symmetry.
## Bispectrum analysis of ewt for escalator foundation vibration signal
With regard to the complexity of the environment, the actual vibration signal collected from the foundation of the escalator usually contains interfering noise. Bispectrum analysis can only effectively suppress Gaussian noise, but it is powerless against non-Gaussian noise. Therefore, EWT is used to remove the effect of non-Gaussian noise from the signal before performing bispectrum analysis. After the decomposition of the machine feet vibration signal using EWT, a series of EMFs are obtained. Because of the high noise frequency, the noise-containing signal is often concentrated in the highest frequency modal component EMF1 after EWT decomposition [bib_ref] A double impulsiveness measurement indices-bilaterally driven empirical wavelet transform and its application..., Ding [/bib_ref] [bib_ref] Feature extraction method based on adaptive and concise empirical wavelet transform and..., Zhang [/bib_ref] , while the fault feature information is contained in the remaining low-frequency modal components. So, this paper discards the first layer of empirical mode function EMF1 and only performs bispectrum analysis on the remaining mode functions to extract the fault feature information hidden in the mode components.
The sampling frequency of the vibration signal of the escalator base foot is 2000 Hz. The signals of normal, loosening one lap, loosening two laps, and loosening three laps of fixed bolts are collected, respectively, and some of the collected signals are intercepted as shown in [fig_ref] Figure 5: Original machine feet vibration signal [/fig_ref]. EWT decomposition is carried out for each of the four signals, and the number of frequency band intervals for EWT decomposition is taken as four. The decomposed EMFs of the machine feet vibration signal with loosening one lap are shown in [fig_ref] Figure 6: Results of EWT decomposition [/fig_ref] (the EMFs of the vibration signal of other cases are similar to this, but due to space limitations, they are not shown one by one).
The bispectrum analysis of mode function 2 (EMF2), mode function 3 (EMF3), and mode function 4 (EMF4) of four kinds of signals is carried out, respectively. [fig_ref] Figure 7: Contour [/fig_ref] shows the two-dimensional contour plot of bispectrum analysis of the machine feet vibration signal when the fixed bolt is normal. [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] show the two-dimensional contour plot of bispectrum analysis of the footing vibration signal when the fixed bolt is loosening for one, two, and three laps, respectively. From [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , it can be seen that the twodimensional spectrum of the escalator footing bolt shows obvious differences under normal working conditions and loose operating conditions. Compared with EMF2 and EMF3, EMF4 contains the least noise and interference signals, and EMF4 also concentrates the majority of the energy of the foundation vibration. Observing the bispectrum of EMF4 [fig_ref] Figure 5: Original machine feet vibration signal [/fig_ref] , it can be seen that when the foundation bolt is normal, the energy of the bispectrum of the foundation vibration signal converges towards the center. As the bolts gradually loosen, the energy of the bispectrum of the foundation vibration signal gradually expands outward. As the degree of bolt loosening increases, the frequency extension range of the bispectrum also gradually increases.
## Fault feature extraction based on gray-gradient co-occurrence matrix
In order to automatically identify whether the bolt is loose and confirm the degree of loosening, the feature information contained in the bispectrum coefficients needs to be extracted. The two-dimensional contour plot of the bispectrum analysis contains the basic feature information of the vibration signal, while its gradient plot depicts the edge and abrupt change information of the contour map. If both the two-dimensional contour plots and gradient plots are combined for feature extraction, more accurate fault characteristics can be obtained. Based on the fact that the grey-gradient co-occurrence matrix (GGCM) [bib_ref] Weed and Corn Seedling Detection in Field Based on Multi Feature Fusion..., Chen [/bib_ref] can describe both the gray information and the gradient information in the image, this paper constructs a grey-gradient co-occurrence matrix for the two-dimensional contour plot of the bispectrum analysis of vibration signals. The GGCM is used to extract the characteristic parameters of the bispectrum analysis contour map as the fault characteristics when the bolt is loose. majority of the energy of the foundation vibration. Observing the bispectrum of EMF4 [fig_ref] Figure 5: Original machine feet vibration signal [/fig_ref] , it can be seen that when the foundation bolt is normal, the energy of the bispectrum of the foundation vibration signal converges towards the center. As the bolts gradually loosen, the energy of the bispectrum of the foundation vibration signal gradually expands outward. As the degree of bolt loosening increases, the frequency extension range of the bispectrum also gradually increases.
## Fault feature extraction based on gray-gradient co-occurrence matrix
In order to automatically identify whether the bolt is loose and confirm the degree of loosening, the feature information contained in the bispectrum coefficients needs to be extracted. The two-dimensional contour plot of the bispectrum analysis contains the basic feature information of the vibration signal, while its gradient plot depicts the edge and abrupt change information of the contour map. If both the two-dimensional contour plots and gradient plots are combined for feature extraction, more accurate fault characteristics can be obtained. Based on the fact that the grey-gradient co-occurrence matrix (GGCM) [bib_ref] Weed and Corn Seedling Detection in Field Based on Multi Feature Fusion..., Chen [/bib_ref] can describe both the gray information and the gradient information in the image, this paper constructs a grey-gradient co-occurrence matrix for the two-dimensional contour plot of the bispectrum analysis of vibration signals. The GGCM is used to extract the characteristic parameters of the bispectrum analysis contour map as the fault characteristics when the bolt is loose. majority of the energy of the foundation vibration. Observing the bispectrum of EMF4 [fig_ref] Figure 5: Original machine feet vibration signal [/fig_ref] , it can be seen that when the foundation bolt is normal, the energy of the bispectrum of the foundation vibration signal converges towards the center. As the bolts gradually loosen, the energy of the bispectrum of the foundation vibration signal gradually expands outward. As the degree of bolt loosening increases, the frequency extension range of the bispectrum also gradually increases.
## Fault feature extraction based on gray-gradient co-occurrence matrix
In order to automatically identify whether the bolt is loose and confirm the degree of loosening, the feature information contained in the bispectrum coefficients needs to be extracted. The two-dimensional contour plot of the bispectrum analysis contains the basic feature information of the vibration signal, while its gradient plot depicts the edge and abrupt change information of the contour map. If both the two-dimensional contour plots and gradient plots are combined for feature extraction, more accurate fault characteristics can be obtained. Based on the fact that the grey-gradient co-occurrence matrix (GGCM) [bib_ref] Weed and Corn Seedling Detection in Field Based on Multi Feature Fusion..., Chen [/bib_ref] can describe both the gray information and the gradient information in the image, this paper constructs a grey-gradient co-occurrence matrix for the two-dimensional contour plot of the bispectrum analysis of vibration signals. The GGCM is used to extract the characteristic parameters of the bispectrum analysis contour map as the fault characteristics when the bolt is loose.
## Normalization of gray matrix and gradient matrix
After the gray processing of the bispectrum analysis two-dimensional contour plot, the gray matrix f (m, n) is obtained. Because the pixel values of the gray matrix are between [0, 255], there is no need for normalization. In this section, only the gradient matrix of the two-dimensional contour plot is normalized. Let g(m, n) be the pixel value of point (m, n) in the gradient matrix.
[formula] g(p, q) = g 2 x (p, q) + g 2 y (p, q)(21) [/formula]
where:
[formula] g x (p, q) = f (p + 1, q − 1) + 2 f (p + 1, q) + f (p + 1, q + 1) − f (p − 1, q − 1) − 2 f (p − 1, q) − f (p − 1, q + 1) g y (p, q) = f (p − 1, q + 1) + 2 f (p, q + 1) + f (p + 1, q + 1) − f (p − 1, q + 1) − 2 f (p, q − 1) − f (p + 1, q − 1) p = 1, 2, . [/formula]
. . , P, q = 1, 2, . . . , Q, P, Q represent the number of rows and columns of the gray matrix.
Let g max be the maximum value in the gradient image and L g be the expected maximum gradient value after normalization. Then, the normalized gradient matrix is: G(p, q) = I NT[g(p, q)·L g /g max ] + 1In this paper, L g = 64. After normalization, two normalization matrices are obtained: gray normalization matrix F(p, q) = f (p, q) and gradient normalization matrix G(p, q).
## Generation of ggcm
The gray-gradient co-occurrence matrix based on bispectrum analysis is C = c(x, y). Then, its element c(x, y) is defined as the total number of pairs of image points with pixel values x in the normalized grey matrix F(p, q) and pixel values y in the normalized gradient matrix G(p, q). c(x, y) is equal to the number of pairs of image points that make x = F(p, q) and y = G(p, q). In order to facilitate texture feature extraction of the GGCM, it is necessary to normalize it. Let the element value of the normalized GGCM beĉ(x, y), then we have:ĉ
[formula] (x, y) = c(x, y) 255 ∑ x=0 63 ∑ y=0 c(x, y)(23) [/formula]
## Texture feature extraction with gray-gradient co-occurrence matrix
More feature information can be extracted from the GGCM of bispectrum analysis [bib_ref] Fault Characteristic Extraction by Fractional Lower-Order Bispectrum Methods, Wang [/bib_ref]. But if too many feature parameters are selected, this can lead to excessive computational effort and affect the accuracy of the fault identification. Therefore, in this section, Small-gradient dominance:
[formula] T 1 = 255 ∑ x=0 63 ∑ y=0ĉ (x, y) y 2 / 255 ∑ x=0 63 ∑ y=0ĉ (x, y)(24) [/formula]
2.
Inhomogeneity of the grey distribution:
[formula] T 2 = 255 ∑ x=0 [ 63 ∑ y=0ĉ (x, y)] 2 / 255 ∑ x=0 63 ∑ y=0ĉ (x, y)(25) [/formula]
3.
Inhomogeneity of the gradient distribution:
[formula] T 3 = 63 ∑ y=0 [ 225 ∑ x=0ĉ (x, y)] 2 / 255 ∑ x=0 63 ∑ y=0ĉ (x, y)(26) [/formula]
4. Grey entropy:
[formula] T 4 = − 255 ∑ x=0 [ 63 ∑ y=0ĉ (x, y)] 2 · log[ 63 ∑ y=0ĉ (x, y)](27) [/formula]
5. Gradient entropy:
[formula] T 5 = − 63 ∑ y=0 [ 255 ∑ x=0ĉ (x, y)] 2 · log[ 63 ∑ y=0ĉ (x, y)](28) [/formula]
6.
Mixed entropy:
[formula] T 6 = 255 ∑ x=0 63 ∑ y=0ĉ (x, y) log[ĉ(x, y)](29) [/formula]
The results of randomly selecting the data from normal state, loosening one lap, loosening two laps, and loosening three laps and calculating the normalized values of EMF2, EMF3, and EMF4 texture features after EWT decomposition are shown in [fig_ref] Table 1: Normalized values of the 6 texture features of EMFs [/fig_ref]. By analyzing the data in [fig_ref] Table 1: Normalized values of the 6 texture features of EMFs [/fig_ref] , it can be found that there are significant differences in the values of texture feature parameters with different degrees of looseness. In order to analyze the experimental results more intuitively, the values of small gradient advantage, non-uniformity of gray distribution, non-uniformity of gradient distribution, gray entropy, gradient entropy, and entropy of mixing are analyzed as a line chart, and the results are shown in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref]. By analyzing [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , it can be found that there is a better discrimination between the small gradient dominance values of EMF2 and EMF4 components in different states when the foot of the escalator is in a normal, loose state of one turn, loose state of one turns, and loose state of three turns. However, there is a certain degree of overlap between the small gradient advantage values of EMF3 components. It can be seen from [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] that when the foundation bolts are loosened for one circle, two circles, and three circles, respectively, there is a clear distinction between the uneven gray distribution of EMF2, EMF3, and EMF4. For signals in a normal state, the uneven distribution of EMF3 and EMF4 gray-scale is also well differentiated from the fault signal, but there is a certain degree of overlap between the uneven distribution of EMF2 gray-scale in normal state and the fault signal. It can be seen from [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] that when the footing is loosened for one circle, two circles, and three circles, respectively, there is obvious discrimination between the uneven gray distribution of EMF2, EMF3, and EMF4. When the footing is in different states, there is an obvious separation between the gradient distribution nonuniformity feature vectors, and there is no aliasing.
It can also be seen from [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] ,e,f that when the footing is in different states, the eigenvectors composed of gray entropy, gradient entropy, and entropy of mixing of each modal component also have high discrimination, and there is no aliasing between feature vectors. Therefore, combining [fig_ref] Table 1: Normalized values of the 6 texture features of EMFs [/fig_ref] and [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , it can be seen that the six feature values extracted through the gray co-occurrence matrix are effective and can effectively distinguish the fault status of the footing.
After obtaining the six normalized texture feature values of the k -th modal compo- By analyzing [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , it can be found that there is a better discrimination between the small gradient dominance values of EMF2 and EMF4 components in different states when the foot of the escalator is in a normal, loose state of one turn, loose state of one turns, and loose state of three turns. However, there is a certain degree of overlap between the small gradient advantage values of EMF3 components. It can be seen from [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] that when the foundation bolts are loosened for one circle, two circles, and three circles, respectively, there is a clear distinction between the uneven gray distribution of EMF2, EMF3, and EMF4. For signals in a normal state, the uneven distribution of EMF3 and EMF4 gray-scale is also well differentiated from the fault signal, but there is a certain degree of overlap between the uneven distribution of EMF2 gray-scale in normal state and the fault signal. It can be seen from [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] that when the footing is loosened for one circle, two circles, and three circles, respectively, there is obvious discrimination between the uneven gray distribution of EMF2, EMF3, and EMF4. When the footing is in different states, there is an obvious separation between the gradient distribution nonuniformity feature vectors, and there is no aliasing.
It can also be seen from [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] -f that when the footing is in different states, the eigenvectors composed of gray entropy, gradient entropy, and entropy of mixing of each modal component also have high discrimination, and there is no aliasing between feature vectors. Therefore, combining [fig_ref] Table 1: Normalized values of the 6 texture features of EMFs [/fig_ref] and [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , it can be seen that the six feature values extracted through the gray co-occurrence matrix are effective and can effectively distinguish the fault status of the footing.
After obtaining the six normalized texture feature values of the k-th modal component EMF k (k = 2, 3, 4) of the foundation vibration signal, the texture features (T ) of T × is used as an input, and a bidirectional long short-term network (Bi LSTM) is used as a classifier to identify different degrees of fundamental motion faults. We randomly select a set of data from normal state, loose one circle, loose two circles, and loose three circles. After EWT decomposition, the texture features of the three modes are fused, and the fused results are shown in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref]. From [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , it can be seen that apart from individual features, the fused feature vectors of different states have a relatively clear degree of differentiation. The fused 18dimensional feature vectors can effectively diagnose and recognize the loose footings in one, two, three, and normal states.
## Bolt loosening fault diagnosis model based on ewt and ggcm
In this section, bidirectional LSTM is used to demonstrate fault identification for the loosening of an anchor bolt of an escalator. Based on the multi-scale GGCM feature and Bi-LSTM, the process of anchor bolt loosening fault diagnosis is shown in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref]. The specific steps of bolt loosening fault identification are as follows:
Step 1: EWT is used to decompose each training sample into four levels. And, the last three order empirical mode functions (EMFs) are retained;
Step 2: Bispectrum analysis is performed on the last three order EMFs (EMF2, EMF3, EMF4) of each training sample. The gray matrix and gradient matrix are constructed by the two-dimensional contour plot of bispectrum analysis, and the GGCM is constructed;
Step 3: Six texture features are extracted from the GGCM of EMFk (k = 2, 3, 4) to form the 18-dimensional fault feature vector of training samples.
Step 4: The Bi-LSTM network is trained with training data. And, an anchor bolt loosening fault diagnosis model is established. From [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , it can be seen that apart from individual features, the fused feature vectors of different states have a relatively clear degree of differentiation. The fused 18-dimensional feature vectors can effectively diagnose and recognize the loose footings in one, two, three, and normal states.
## Bolt loosening fault diagnosis model based on ewt and ggcm
In this section, bidirectional LSTM is used to demonstrate fault identification for the loosening of an anchor bolt of an escalator. Based on the multi-scale GGCM feature and Bi-LSTM, the process of anchor bolt loosening fault diagnosis is shown in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref]. The specific steps of bolt loosening fault identification are as follows:
Step 1: EWT is used to decompose each training sample into four levels. And, the last three order empirical mode functions (EMFs) are retained;
Step 2: Bispectrum analysis is performed on the last three order EMFs (EMF 2 , EMF 3 , EMF 4 ) of each training sample. The gray matrix and gradient matrix are constructed by the two-dimensional contour plot of bispectrum analysis, and the GGCM is constructed;
Step 3: Six texture features are extracted from the GGCM of EMF k (k = 2, 3, 4) to form the 18-dimensional fault feature vector of training samples.
Step 4: The Bi-LSTM network is trained with training data. And, an anchor bolt loosening fault diagnosis model is established.
## Analysis of experimental results
## Collection of experimental data and evaluation
## Analysis of experimental results
## Collection of experimental data and evaluation index
## Experimental data
In this section, simulation experiments are conducted on a Schindler S9700-30 escalator model with a vibration sensor sampling frequency of 2000 Hz and an acquisition time of 8 s. The vibration sensor is installed on the anchor bolt position of the escalator through the thread.
By adjusting the loosening degree of the anchor bolt, a total of 800 sets of experimental data were collected, including 200 groups of data when the anchor bolt was normal, 200 sets of data when the fixed bolt was loosening one lap, 200 sets of data when the fixed bolt was loosening two laps, and 200 groups of data when the fixed bolt was loosening three laps. According to the escalator product characteristics, the frequency range of the anchor bolt vibration is determined to be between 1 and 5 khz. Therefore, a low-pass filter is used to filter the high-frequency part to avoid the interference signal being introduced into the next level. The filter circuit design uses an active filter circuit to filter out the spurious signal and then amplifies the signal at the same time. During the collection process, a marker line is first drawn on the mainframe anchor bolt not only to record the initial position of the mainframe anchor bolt but also to avoid excessive or inappropriate loosening of the mainframe anchor bolt. After drawing the marker line, the mainframe foot screw is loosening by one, two, and three laps, respectively, against the marker line and confirmed by measurement that the mainframe anchor bolt is loosened in place.
## Evaluation index
To evaluate the performance of the proposed method, accuracy, precision (P), recall (R), and F1-score were selected as the evaluation indices [bib_ref] Rolling bearing fault diagnosis method based on SSAE and softmax classifier with..., Wang [/bib_ref]. Accuracy represents the ratio of the number of correctly classified samples to the total number of samples, and its calculation formula is:
[formula] Accuracy = TP + TN TP + TN + FN + FP(31) [/formula]
Precision (P) represents the proportion of samples that are truly positive, and its calculation formula is:
[formula] Precision = TP TP + FP(32) [/formula]
Recall rate (R), as one of the important indicators for model performance evaluation, reflects the proportion of all positive samples detected by the model to all true positive samples, and its calculation formula is:
[formula] Recall = TP TP + TN(33) [/formula]
F1-score is a comprehensive evaluation indicator that can balance the precision and recall of the model and better reflects the robustness of the model; its calculation formula is:
[formula] F1 − score = 2PR P + R(34) [/formula]
where TP is the number of positive samples predicted by the model as positive, and TN is the number of negative samples predicted by the model as negative; FN represents the number of negative samples that were incorrectly identified as positive samples, while FP represents the number of negative samples that were incorrectly identified as positive samples.
## Comparative analysis of different fault feature extraction methods
EWT decomposition was carried out on the experimentally collected vibration data of the anchor bolt under normal and loosening conditions. EMF2, EMF3, and EMF4 were subjected to bispectrum analysis. And, six texture features were extracted for each layer of EMF using a grey-gradient co-generation matrix to form an 18-dimensional fault feature vector, which was input into the Bi-LSTM network to identify the degree of loosening of the anchor bolt. For the vibration signal when the bolt is normal and when the bolt is loose, the corresponding output is "0" for a normal bolt, "1" for loosening 1 lap, "2" for loosening 2 laps, and "3" for loosening 3 laps. In Bi-LSTM, the number of neurons in the input layer, output layer, and hidden layer are 30, 1, and 20, respectively. The internal parameters of the LSTM were trained using Adam's algorithm with a learning rate of 0.001, a training number of 1000, and a training target of 0.0001. The parameters of the reverse layer network were the same as those of the forward layer. The fault identification accuracy of different methods was analyzed using a 10-fold cross-validation method. In the 10-fold cross-validation, the eigenvector set of the base vibration signal is divided into ten parts on average: nine of them are taken as training data, and one is taken as test data for the classification experiment. Finally, the average of the 10 experimental results is used as an estimate of the accuracy of the method. [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] and [fig_ref] Table 2: Classification results obtained by four methods [/fig_ref] show the identification accuracy of four methods for the degree of the looseness of foundation bolts when using ten-fold cross validation. It can be seen that the four indexes of the GGCM-BiLSTM method are all the lowest because the direct GGCM method does not perform multi-scale decomposition on the original vibration signal and directly extracts fault features, which affects the accuracy of fault feature extraction. The four evaluation indices of the EMD-GGCM-BiLSTM method are significantly higher than those of the GGCM-BiLSTM method. This is because EMD performs modal separation on the original vibration signal, which to some extent suppresses noise and can more accurately extract fault features. The EEMD-GGCM-BiLSTM method shows a certain improvement in overall recognition performance compared to the EMD-GGCM-BiLSTM method, but this improvement is not significant. The proposed EWT-GGCM-BiLSTM model achieved better results in terms of the four performance indices. Compared with the EEMD-GGCM-BiLSTM method, which showed a better performance, the accuracy of the proposed method improved by about 2.63%, precision improved by about 2 62%, recall increased by approximately 8.50%, and F1 score increased by approximately 5.49%. The experimental results in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] and [fig_ref] Table 2: Classification results obtained by four methods [/fig_ref] indicate that the recognition framework based on EWT and multi-scale GGCM feature extraction can effectively identify the degree of looseness of escalator foundation bolts. In order to further verify the test results of the proposed model for normal (0), loose 1 circle (1), loose 2 circle (2), and loose 3 circle (3) of the foundation bolt, the confusion matrix shown in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] was utilized, where the horizontal coordinate represents the real label, and the vertical coordinate represents the prediction result. It can be seen that among all the diagnostic models, the performance of the proposed model in diagnosing bolt loosening faults is far superior to other comparative models, which also verifies the rationality of the proposed model. The reasons for the better identification results of the proposed method are as follows: (1) EWT uses bandpass filter banks to extract each mode In order to further verify the test results of the proposed model for normal (0), loose 1 circle (1), loose 2 circle (2), and loose 3 circle (3) of the foundation bolt, the confusion matrix shown in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] was utilized, where the horizontal coordinate represents the real label, and the vertical coordinate represents the prediction result. It can be seen that among all the diagnostic models, the performance of the proposed model in diagnosing bolt loosening faults is far superior to other comparative models, which also verifies the rationality of the proposed model. The reasons for the better identification results of the proposed method are as follows: (1) EWT uses bandpass filter banks to extract each mode component, effectively avoiding mode aliasing, and can separate the fault information of different scales as far as possible. (2) The multi-scale GGCM feature based on EWT can focus on the main vibration features of the loosening of the foundation bolt and suppress the interference of noise and irrelevant background signals on the extracted features. The above experimental results show that high-quality features are the key to improving the identification accuracy of the foundation bolt loosening, and the results of the confusion matrix also verify the rationality of the proposed method. [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref]. Confusion matrix from ten-fold CV with four methods.
## Comparative with machine learning algorithms
In order to further compare the performance of the proposed method, the propos method is compared with common machine learning methods, including support vec machine (SVM), random forest (RF), and naive Bayes classifier (NBC). The kernel funct [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref]. Confusion matrix from ten-fold CV with four methods.
## Comparative with machine learning algorithms
In order to further compare the performance of the proposed method, the proposed method is compared with common machine learning methods, including support vector machine (SVM), random forest (RF), and naive Bayes classifier (NBC). The kernel function of SVM uses the Gaussian radial kernel function, kernel width, and penalty factor determined via grid search. In the random forest model, the values of three important parameters (the number of decision trees, the maximum depth of decision trees, and the minimum sample size that each decision tree can partition) are also selected via grid search. In the naive Bayes classifier model, the prior probability distribution is set to Gaussian distribution. The 10-fold cross validation method is used to analyze the recognition accuracy of different methods, and the average of the ten results is then taken as the final result. The accuracy, precision, recall, and F1-score of the four methods are shown in [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] and [fig_ref] Table 3: Classification results obtained by four ML methods [/fig_ref]. From [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] and [fig_ref] Table 3: Classification results obtained by four ML methods [/fig_ref] , it can be seen that the proposed EMT-GGCM-BiLSTM model has the best recognition performance, with an accuracy of 99.38%, precision of 98.99%, recall of 98.50%, and F1-score of 98.75%. Compared with the SVM method, the proposed method showed an improvement in accuracy of 1.51%, precision of 2.54%, recall of 3.50%, and F1 of 3.03%. Compared with the SVM method, the accuracy, precision, recall, and F1-score of the proposed method improved by 1.51%, 2.54%, 3.50%, and 3.03%, respectively. Compared with the RF method, the accuracy, precision, recall, and F1-score of the proposed method improved by 2.38%, 3.16%, 6.50%, and 4.87%, respectively. Compared with the NBC method, the accuracy, precision, recall, and F1-score of the proposed method improved by 3.75%, 7.11%, 8.00%, and 7.57%, respectively. The above experimental results indicate that the recognition performance of the NBC method is relatively low, and the recognition performance of SVM and RF is basically the same, while the recognition performance of the proposed method is better than that of SVM and RF methods. Therefore, compared with the classical shallow machine learning model, the BiLSTM method based on the deep learning model has certain advantages, with a higher classification accuracy and generalization performance.To further demonstrate the ability of different methods to identify the looseness of foundation bolts, [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] shows the confusion matrix of the identification results of the four methods. As can be seen from [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , it is relatively easy to identify the normal condition and one loosening circle, but the error rate of identifying two loosening circles and three loosening circles is relatively high, which may be due to the similar fault feature of two loosening circles and three loosening circles. Although the proposed method also has the above situation, the overall identification effect is the highest, indicating that the BiLSTM method used can effectively improve the identification ability of different fault feature vectors. From [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] and [fig_ref] Table 3: Classification results obtained by four ML methods [/fig_ref] , it can be seen that the proposed EMT-GGCM-BiLSTM model has the best recognition performance, with an accuracy of 99.38%, precision of 98.99%, recall of 98.50%, and F1-score of 98.75%. Compared with the SVM method, the proposed method showed an improvement in accuracy of 1.51%, precision of 2.54%, recall of 3.50%, and F1 of 3.03%. Compared with the SVM method, the accuracy, precision, recall, and F1-score of the proposed method improved by 1.51%, 2.54%, 3.50%, and 3.03%, respectively. Compared with the RF method, the accuracy, precision, recall, and F1-score of the proposed method improved by 2.38%, 3.16%, 6.50%, and 4.87%, respectively. Compared with the NBC method, the accuracy, precision, recall, and F1-score of the proposed method improved by 3.75%, 7.11%, 8.00%, and 7.57%, respectively. The above experimental results indicate that the recognition performance of the NBC method is relatively low, and the recognition performance of SVM and RF is basically the same, while the recognition performance of the proposed method is better than that of SVM and RF methods. Therefore, compared with the classical shallow machine learning model, the BiLSTM method based on the deep learning model has certain advantages, with a higher classification accuracy and generalization performance.To further demonstrate the ability of different methods to identify the looseness of foundation bolts, [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] shows the confusion matrix of the identification results of the four methods. As can be seen from [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref] , it is relatively easy to identify the normal condition and one loosening circle, but the error rate of identifying two loosening circles and three loosening circles is relatively high, which may be due to the similar fault feature of two loosening circles and three loosening circles. Although the proposed method also has the above situation, the overall identification effect is the highest, indicating that the BiLSTM method used can effectively improve the identification ability of different fault feature vectors. [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref]. Confusion matrix from ten-fold CV with four ML methods.
# Conclusions
A diagnostic method based on EWT and multi-scale GGCM feature extraction is proposed for detecting the looseness fault of escalator foundation bolts, and the effectiveness of the proposed method is verified using measured signals.
(1) The Teager energy operator is used to concentrate spectral energy on the spectrum, which greatly reduces the erroneous influence of noise and irrelevant components on spectrum segmentation. The multi-scale "peak" localization method based on the Teager energy operator can detect the segmentation boundary of EWT adaptively, avoid the wrong decision caused by a single "pseudo-value point", and make the spectrum segmentation more appropriate. (2) The feature extraction method based on the multi-scale GGCM does not require too many parameter settings, and the feature vector extracted from the multi-scale GGCM can effectively diagnose the loose footing fault. This method is simple, [fig_ref] Figure 1: Simulated signal and FFT spectrum [/fig_ref]. Confusion matrix from ten-fold CV with four ML methods.
# Conclusions
A diagnostic method based on EWT and multi-scale GGCM feature extraction is proposed for detecting the looseness fault of escalator foundation bolts, and the effectiveness of the proposed method is verified using measured signals.
(1) The Teager energy operator is used to concentrate spectral energy on the spectrum, which greatly reduces the erroneous influence of noise and irrelevant components on spectrum segmentation. The multi-scale "peak" localization method based on the Teager energy operator can detect the segmentation boundary of EWT adaptively, avoid the wrong decision caused by a single "pseudo-value point", and make the spectrum segmentation more appropriate. (2) The feature extraction method based on the multi-scale GGCM does not require too many parameter settings, and the feature vector extracted from the multi-scale GGCM can effectively diagnose the loose footing fault. This method is simple, intuitive, and accurate and overcomes the difficulties of fault feature extraction using the traditional method. (3) The fault recognition method by jointing multi-scale GGCM feature extraction and the BiLSTM classifier has a high recognition accuracy and certain degree of universality, which can be used to identify and warn of foundation bolts loosening faults during the operation of escalators.
In the fault identification of escalator base bolt loosening, the suppression of vibration signal interference noise and the improvement in classifier performance have important effects on the fault identification accuracy. How to use multi-objective optimization algorithm [bib_ref] Multi objective optimization of sound transmission across laminated composite cylindrical shell lined..., Talebitooti [/bib_ref] to more effectively suppress interference noise and how to use the attention mechanism to further improve the performance of the BiLSTM classifier are the two research areas that we hope to study in the future.
Author Contributions: Conceptualization, X.E. and W.W.; methodology, X.E. and W.W.; validation, X.E. and W.W.; data curation, X.E. and W.W.; writing-original draft prepation, X.E.; writing-review and editing, X.E., W.W. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China (Nos. 61671338), and "the 14th Five Year plan" Hubei Provincial advantaged characteristic disciplines (groups) project of Wuhan University of Science and Technology (2023C0204).
[fig] Figure 1: Simulated signal and FFT spectrum. [/fig]
[fig] s 2023 ,: 22, x FOR PEER REVIEW 7 of 25 intrinsic frequencies, thereby avoiding frequency band breakage caused by the improper segmentation of the original EWT, resulting in the achievement of a better separation of fault information. [/fig]
[fig] Figure 2, Figure 2: Comparison of spectrum division results of EWT. Comparison of spectrum division results of EWT. -term memory (LSTM) is an improved model based on the recurrent neural network (RNN). The structure of LSTM is shown inFigure 3, which contains four main gate structures: input gate i, forget gate f , control gate c, and output gate o. [/fig]
[fig] Figure 2: Comparison of spectrum division results of EWT. [/fig]
[fig] Figure 3: LSTM network structure diagram. [/fig]
[fig] Figure 4: BI-LSTM network structure diagram. [/fig]
[fig] Figure 5: Original machine feet vibration signal. [/fig]
[fig] Figure 6: Results of EWT decomposition. analysis of mode function 2 (EMF2), mode function 3 (EMF3), and mode function 4 (EMF4) of four kinds of signals is carried out, respectively.Figure 7 [/fig]
[fig] Figure 7: Contour [/fig]
[fig] Figure 8, Figure 8 25, Figure 9, Figure 9: Contour Contour of loosening 1 lap: (a) EMF 2 ; (b) EMF 3 ; (c) EMF 4 . Sensors 2023, 22, x FOR PEER REVIEW 13 of Contour of loosening 2 laps (a) Contour of loosening 2 laps (a) EMF 2 ; (b) EMF 3 ; (c) EMF 4 . [/fig]
[fig] Figure 9: Contour [/fig]
[table] Table 1: Normalized values of the 6 texture features of EMFs. [/table]
[table] Table 2: Classification results obtained by four methods. [/table]
[table] Table 3: Classification results obtained by four ML methods. [/table]
|
10.1186/s40510-016-0158-5
|
CCBY
|
5276803
|
28133715
|
s2orc_pubmed_articles
|
Factors affecting dental biofilm in patients wearing fixed orthodontic appliances
Background: The aim of this study is to investigate the amount and the distribution of biofilm in patients wearing fixed appliances and its relation with age, gender, frequency of tooth brushing, and patient motivation. Methods: The sample comprised 52 patients (15.5 ± 3.6 years old, 30 females and 22 males) wearing fixed orthodontic appliances. Dental biofilm was assessed using a modified plaque index (PI). A questionnaire was used to collect patient's information, including gender, age, treatment motivation, and frequency of tooth brushing. Results: Gingival (PI score = 0.9 ± 0.7), mesial (0.8 ± 0.6), and distal (0.8 ± 0.5) areas accumulated more biofilm than occlusal areas (0.3 ± 0.3) (P < 0.038). The maxillary lateral incisors (1.1 ± 0.8) and maxillary canines (1.0 ± 0.8) had more biofilm than other teeth (P < 0.05). The maxillary arch (0.8 ± 0.7) had significantly more biofilm than mandibular arch (0.6 ± 0.6) (P = 0.042). No significant difference was found between the right side (0.7 ± 0.7) and left side (0.7 ± 0.6) (P = 0.627). Less biofilm was found in females (0.6 ± 0.5), adults (0.3 ± 0.3), and "self-motivated" patients (0.3 ± 0.3), compared with males (0.9 ± 0.5), children (0.8 ± 0.6), and "family-motivated" patients (1.1 ± 0.5) (P < 0.001). The amount of biofilm was associated with self-report of the frequency of daily tooth brushing (P < 0.001). Conclusions: Patients wearing fixed orthodontic appliances have the highest biofilm accumulation on the maxillary lateral incisors and maxillary canines, particularly in the gingival area and areas behind arch wires. Less biofilm was observed in female and adult patients and in those who were self-motivated and brushed their teeth more often.
# Background
Biofilm formation around fixed orthodontic appliances can cause important side effects. This includes white spot lesions (WSLs) and, in severe cases, tooth decay, with a negative impact on patient's quality of life [bib_ref] Orthodontic treatment with fixed appliances and biofilm formation-a potential public health threat?, Ren [/bib_ref] [bib_ref] Risk factors for incidence and severity of white spot lesions during treatment..., Chapman [/bib_ref]. Although many auxiliary dental products such as interdental brushes, specialized toothbrushes, and mouth rinses are commercially available, the prevalence of WSLs still remains as high as 72.9% [bib_ref] Incidence of caries lesions among patients treated with comprehensive orthodontics, Richter [/bib_ref]. This is because the placement of fixed orthodontic appliances severely impedes tooth brushing, makes conventional oral hygiene procedures more difficult, and provides areas of low salivary flow that allow bacterial adhesion and biofilm formation [bib_ref] Esthetic comparison of white-spot lesion treatment modalities using spectrometry and fluorescence, Yuan [/bib_ref] [bib_ref] Oral bacterial adhesion forces to biomaterial surfaces constituting the bracket-adhesiveenamel junction in..., Mei [/bib_ref].
The introduction of fixed appliances into the oral cavity not only promotes the amount of biofilm formation but also increases the level of acidogenic bacteria inside the biofilm, resulting in a higher cariogenic challenge around orthodontic brackets and bands [bib_ref] Effect of orthodontic appliances on oral microbiota-6 month follow-up, Topaloglu-Ak [/bib_ref] [bib_ref] White-spot lesions and gingivitis microbiotas in orthodontic patients, Tanner [/bib_ref] [bib_ref] Clinical and microbiological findings at sites treated with orthodontic fixed appliances in..., Kim [/bib_ref]. If patients cannot maintain good oral hygiene during orthodontic treatment, the acid produced by dental biofilms will eventually lead to enamel demineralization and WSLs. Though some superficial soft WSLs can be remineralized, most will persist after the removal of the fixed appliances [bib_ref] Esthetic comparison of white-spot lesion treatment modalities using spectrometry and fluorescence, Yuan [/bib_ref].
The distribution of dental biofilm seen in patients wearing fixed orthodontic appliances may, with time, reflect the eventual distribution of WSLs [bib_ref] Risk factors for incidence and severity of white spot lesions during treatment..., Chapman [/bib_ref] [bib_ref] Distribution of plaque and gingivitis and associated factors in 3-to 5-year-old Brazilian..., Feldens [/bib_ref]. It has been confirmed that the presence of biofilm on the tooth surface is a predictive factor for the development of carious lesions in children [bib_ref] Early plaque accumulation-a sign for caries risk in young children, Alaluusua [/bib_ref]. The distribution of biofilm is significantly related to the distribution of gingivitis, and the greater the accumulation of biofilm, the higher the gingival bleeding index [bib_ref] Distribution of plaque and gingivitis and associated factors in 3-to 5-year-old Brazilian..., Feldens [/bib_ref] [bib_ref] Experimental gingivitis in man, Loe [/bib_ref]. For better WSL prevention in orthodontics, it is important to understand which factors influence the amount of dental biofilm and its distribution pattern in patients with fixed appliances.
The aims of this study are to investigate the amount and distribution of dental biofilm in patients wearing fixed appliances and to evaluate its association with age, gender, motivation to treatment, and frequency of tooth brushing.
# Methods
## Study design and participants
The study was designed as a cross-sectional study. A convenience sample of 127 orthodontic patients were screened at the Department of Orthodontics of the University of Otago, and 52 eligible patients (30 females and 22 males, mean age = 15.5 ± 3.6 years) were finally included in the study on the basis of sample size used in previous published researches [bib_ref] Frequency of mechanical removal of plaque as it relates to gingival inflammation:..., Pinto [/bib_ref] [bib_ref] Quantity and distribution of plaque in orthodontic patients treated with molar bands, Erbe [/bib_ref]. The patients were selected according to the following inclusion criteria: wearing fixed orthodontic appliances, at least 20 natural teeth, and a willingness to participate in the study; and to the following exclusion criteria: wearing lingual fixed appliances, extensive dental restorations, active periodontal disease, systemic diseases, or the use of medication which may influence periodontal health. This study was approved by the Human Ethics Committee of the University of Otago (13/106).
## Dental biofilm assessment
All consenting patients were examined for the status of biofilm formation using the modified Silness and Löe plaque index (PI) and a periodontal probe [bib_ref] Quantifying plaque during orthodontic treatment, Al-Anezi [/bib_ref]. Each tooth was divided into four areas in relation to the bracket: G = gingival; M = mesial; D = distal; and O = occlusal [fig_ref] Figure 1: Biofilm formation on the four areas of a tooth in relation to... [/fig_ref].
Plaque was then scored in each area based on the original Silness and Löe plaque index [bib_ref] Quantifying plaque during orthodontic treatment, Al-Anezi [/bib_ref] [bib_ref] The gingival index, the plaque index and the retention index systems, Loe [/bib_ref]. The PI scores for all teeth except the second and third molars were measured and recorded [bib_ref] Effect of visual method vs plaque disclosure in enhancing oral hygiene in..., Peng [/bib_ref]. All measurements were taken by two calibrated dental investigators (J.C. and C.W.). The agreement between examiners was assessed using kappa statistics (Kappa value = 0.84).
A questionnaire was used to collect patient information. This included gender, age, motivation to undergo orthodontic treatment (self-motivated, family-motivated, self-and family-motivated), and self-reported frequency of tooth brushing (times per day).
# Statistical analysis
PI scores were reported as means, standard deviations, and 95% confidence intervals (95% CI) [bib_ref] Frequency of mechanical removal of plaque as it relates to gingival inflammation:..., Pinto [/bib_ref]. The data were analyzed using a mixed model analysis. The plaque index was considered as the response variable, whereas age, gender, motivation, frequency of tooth brushing, and site were entered into the models as covariates (fixed effect). A random term for study participant was also entered. The significance level was set at P < 0.05. All analyses were performed using SPSS software (version 17.0 for Windows; SPSS, Chicago, IL).
# Results
Biofilm formation on the four areas of a tooth in relation to the bracket was significantly different (P < 0.001). The gingival (PI score = 0.9 ± 0.7), mesial (0.8 ± 0.6), and distal (0.8 ± 0.5) areas accumulated greater amount of biofilm than the occlusal area (0.3 ± 0.3) (P ≤ 0.038) [fig_ref] Figure 1: Biofilm formation on the four areas of a tooth in relation to... [/fig_ref]. No significant difference was found among the gingival, mesial, and distal areas (P ≥ 0.132). Biofilm formation on each tooth is summarized in [fig_ref] Table 1: Plaque index [/fig_ref] and [fig_ref] Figure 2: Mean levels of biofilm formation on each tooth as indicated by the... [/fig_ref]. The highest level of biofilm formation was found on the maxillary lateral incisors and maxillary canines, amounting to 1.1 ± 0.8 and 1.0 ± 0.8, respectively (P ≤ 0.043). These areas accumulated almost three times more biofilm than the mandibular premolars (0.4 ± 0.3), which had the lowest level of biofilm accumulation (P = 0.036) [fig_ref] Figure 2: Mean levels of biofilm formation on each tooth as indicated by the... [/fig_ref].
The maxillary arch (0.8 ± 0.7) had significantly more dental biofilm than the mandibular arch (0.6 ± 0.6) (P = 0.042), whereas no significant difference was found between the right side (0.7 ± 0.7) and left side (0.7 ± 0.6) (P = 0.627) . The amount of biofilm formation did not differ significantly between quadrants (P ≥ 0.057) [fig_ref] Table 1: Plaque index [/fig_ref].
Male patients (0.9 ± 0.5) had significantly more biofilm than female patients (0.6 ± 0.5) (P < 0.001), whereas children (<18 years old) (0.8 ± 0.6) had significantly more biofilm than adults (≥18 years old) (0.3 ± 0.3) (P < 0.001) . Furthermore, patients who were "family-motivated" to orthodontic treatment had more biofilm accumulation (1.1 ± 0.5) than patients who were "selfand family-motivated" (0.7 ± 0.6) and "self-motivated" (0.3 ± 0.3) (P < 0.001) . In addition, strong association was found between the frequency of tooth brushing and the amount of biofilm formation (P < 0.001), which showed a negative gradient [fig_ref] Figure 5: Biofilm formation in patients with different daily tooth brushing frequencies [/fig_ref].
# Discussion
Biofilm-related side effects during orthodontic treatment are common and may severely impact the quality of treatment outcome as well as patient's quality of life. Hence, it is important for clinicians to gain insights into where dental biofilm accumulates. Understanding the locations at risk for biofilm formation can aid orthodontists and patients when introducing preventive strategies to minimize the development of WSLs.
The distribution of dental biofilm in non-orthodontic samples has been previously investigated [bib_ref] Patterns of de novo plaque formation in the human dentition, Furuichi [/bib_ref]. Molars were found to accumulate more biofilm than anterior teeth, and the mandibular dentition harbored more biofilm than the maxillary dentition [bib_ref] Patterns of de novo plaque formation in the human dentition, Furuichi [/bib_ref]. These observations are in contrast with the pattern of biofilm distribution found in our orthodontic patients, where the upper lateral incisors and upper canines generally had more biofilm accumulation than the upper and lower premolars and the maxillary arch accumulated more biofilm than the mandibular arch. The maxillary lateral incisors and canines accumulated the highest level of dental biofilm, which may be because these teeth are located at the corners of the mouth and receive relatively less tooth brushing strokes during daily oral hygiene. Furthermore, the orthodontic hooks and elastics are usually attached to these areas, making it more difficult to clean [bib_ref] Prevention and treatment of demineralisation during fixed appliance therapy: a review of..., Chambers [/bib_ref]. The site-specific distribution of biofilm in patients wearing fixed appliances could help to explain, at least in part, the considerable variation in the distribution of WSLs seen between individuals and at different sites within the same mouth [bib_ref] Risk factors for incidence and severity of white spot lesions during treatment..., Chapman [/bib_ref] [bib_ref] The incidence of caries and white spot lesions in orthodontically treated adolescents..., Hadler-Olsen [/bib_ref]. Though most subjects generally used their right hand to brush their teeth, in this study, no significant difference was found in the biofilm formation between the left and right sides, consistently with previous findings [bib_ref] Distribution of plaque and gingivitis and associated factors in 3-to 5-year-old Brazilian..., Feldens [/bib_ref] [bib_ref] The effect of toothbrushing frequency, toothbrushing hand, sex and social class on..., Addy [/bib_ref]. The gingival, distal, and mesial areas, in relation to the brackets, attracted more biofilm than the occlusal areas, which was mostly due to the interference of arch wires and ligating devices on tooth brushing. There is also relatively less selfcleaning from natural chewing in these areas [bib_ref] Biofilm retention by 3 methods of ligation on orthodontic brackets: a microbiologic..., Garcez [/bib_ref].
The pattern of biofilm distribution seen in our orthodontic patients is also consistent with previous observations that WSLs occur 2.5 times more frequently in the maxillary arch than in the mandibular arch, with the most prevalent areas being the maxillary lateral incisors, maxillary canines, molars, and mandibular canines [bib_ref] Orthodontic treatment with fixed appliances and biofilm formation-a potential public health threat?, Ren [/bib_ref] [bib_ref] Risk factors for incidence and severity of white spot lesions during treatment..., Chapman [/bib_ref] [bib_ref] White spot lesions during orthodontic treatment: mechanisms and fluoride preventive aspects, Ogaard [/bib_ref].
The plaque index (PI) has been widely used for assessing the level of biofilm formation [bib_ref] The gingival index, the plaque index and the retention index systems, Loe [/bib_ref]. However, it was originally designed for normal populations in the absence of fixed appliances. Therefore, in order to obtain more valid and discriminatory PI measurements, we used the modified PI system in the study, which acknowledges the impact of brackets on biofilm distribution and has greater categorical discrimination than the original Silness and Löe index [bib_ref] Quantifying plaque during orthodontic treatment, Al-Anezi [/bib_ref] [bib_ref] The gingival index, the plaque index and the retention index systems, Loe [/bib_ref].
In our study, self-motivated patients were found to be more cooperative with the clinician's instructions than Comparison of biofilm formation by sex, age group, and arch. Males had significantly more biofilm than females (P < 0.001). Children had significantly more biofilm than adults (P < 0.001). The maxillary arch had significantly more biofilm than the mandibular arch (P = 0.042) Biofilm formation in patients with different motivations to undergo orthodontic treatment. "Family-motivated" patients had more biofilm formation, followed by subjects who were "selfand family-motivated," and "self-motivated" (P < 0.001) self-and family-motivated and family-motivated categories, resulting in a significant difference in biofilm formation among the three categories of patients in the study. This indicates that motivation to undergo orthodontic treatment is important when attempting to predict patient cooperation [bib_ref] Patient compliance with oral hygiene regimens: a behavioural self-regulation analysis with implications..., Ramsay [/bib_ref] [bib_ref] Orthodontic treatment motivation and cooperation: a cross-sectional analysis of adolescent patients' and..., Daniels [/bib_ref]. A lack of cooperation has a significantly negative effect on the results and duration of treatment, as well as on oral hygiene maintenance.
Our findings suggest that adults and females had less biofilm than children and girls. This is consistent with previous observations that adults and females adhere better to clinician's instructions and consequently may maintain better oral hygiene [bib_ref] Distribution of plaque and gingivitis and associated factors in 3-to 5-year-old Brazilian..., Feldens [/bib_ref] [bib_ref] Effect of visual method vs plaque disclosure in enhancing oral hygiene in..., Peng [/bib_ref] [bib_ref] Patient compliance with oral hygiene regimens: a behavioural self-regulation analysis with implications..., Ramsay [/bib_ref]. Orthodontists should therefore take age and gender differences into account when delivering oral hygiene measures.
As expected, our study found that the higher the selfreported frequency of tooth brushing, the less the biofilm formation [bib_ref] Toothbrushing frequency as it relates to plaque development and gingival health, Lang [/bib_ref] [bib_ref] Impacts of toothbrushing frequency on periodontal findings in a group of elderly..., Vysniauskaite [/bib_ref]. Despite the inherent limitation of selfreported assessments, this finding emphasizes the need for sufficient tooth brushing in patients with fixed appliances, especially on teeth more at risk of biofilm formation such as the maxillary lateral incisors and maxillary canines.
One of the limitations of the study is the crosssectional nature, which makes it difficult to make a causal inference. Our sample size was relatively small, and this represents a limitation of the present study. Additionally, some patients were extraction cases, for example premolar extractions. This may have reduced the power of some of our statistical comparisons. Although the differences of biofilm formation among different variables were mostly found statistically significant in the study, the clinical significance may be not. However, the differences of biofilm formation become more pronounced in selected individuals [bib_ref] Distribution of plaque and gingivitis and associated factors in 3-to 5-year-old Brazilian..., Feldens [/bib_ref] [bib_ref] Frequency of mechanical removal of plaque as it relates to gingival inflammation:..., Pinto [/bib_ref]. Furthermore, the differences in wires, loops, ligature, and auxiliary devices used in patients may also influence the generalizability of our results. Future studies with larger sample sizes and homogeneous fixed appliances are needed. Researchers in the future should aim to design more effective and specific methods to improve oral hygiene in patients with fixed appliances, particularly on areas of teeth that accumulate more biofilm.
# Conclusions
Patients wearing fixed orthodontic appliances have the highest biofilm accumulation on the maxillary lateral incisors and maxillary canines, particularly in the gingival area and areas behind arch wires. Less biofilm was observed in female and adult patients and in those who were self-motivated and reported brushing their teeth more often.
[fig] Figure 1: Biofilm formation on the four areas of a tooth in relation to the bracket (G gingival, M mesial, D distal, O occlusal). The occlusal area accumulated the least amount of biofilm compared with the gingival, mesial, and distal areas (P < 0.038). No significant difference was found among the gingival, mesial, and distal areas (P > 0.132) [/fig]
[fig] Figure 2: Mean levels of biofilm formation on each tooth as indicated by the plaque index and a color-coded map [/fig]
[fig] Figure 5: Biofilm formation in patients with different daily tooth brushing frequencies. A significantly association was found between the frequency of daily tooth brushing and the amount of biofilm formation (P < 0.001). Note the pronounced gradient showing less biofilm in patients reporting higher frequency of tooth brushing [/fig]
[table] Table 1: Plaque index (PI) score of each tooth [/table]
|
10.1002/mnfr.201600685
|
CCBY
|
5516247
|
28101967
|
s2orc_pubmed_articles
|
Host‐related factors explaining interindividual variability of carotenoid bioavailability and tissue concentrations in humans
# Introduction
Carotenoids are natural pigments with a C-30 or C-40 backbone. They can be produced by most plants, bacteria, and fungi, but not by animals or humans, making diet their sole source. Carotenoids have recently been investigated with much interest, as their dietary intake and endogenous concentrations have been associated with a reduced risk of several chronic diseases. For example, carotenoid intake has been positively associated with a reduced risk of cancer [1], type 2 diabetes mellitus (T2D) [bib_ref] Intake of fruit, vegetables, and antioxidants and risk of type 2 diabetes:..., Hamer [/bib_ref] , cardiovascular diseases [bib_ref] Dietary carotenoids are associated with cardiovascular disease risk biomarkers mediated by serum..., Wang [/bib_ref] , and asthma [bib_ref] Airway and circulating levels of carotenoids in asthma and healthy controls, Wood [/bib_ref] , while plasma carotene concentration was shown to be significantly associated with reduced total mortality [bib_ref] Plasma carotene and alpha-tocopherol in relation to 10-y all-cause and cause-specific mortality..., Buijsse [/bib_ref]. In addition, some carotenoids, including ␣-, -carotene and -cryptoxanthin [fig_ref] Figure 1: Predominant carotenoids in our diet, common metabolites and nomenclature [/fig_ref] , are vitamin A precursors, constituting the predominant source of vitamin A in most developing countries (up to 90%as well as in Western countries especially with respect to vegetarians. Recently, it has also been suggested that cis-carotenoids are even more beneficial for the prevention of atherosclerosis and T2D than their all-trans isomers [bib_ref] Prevention of atherosclerosis progression by 9-cis-beta-carotene rich alga Dunaliella in apoE-deficient mice, Harari [/bib_ref] [bib_ref] The inhibition of macrophage foam cell formation by 9-cis beta-carotene is driven..., Zolberg Relevy [/bib_ref]. Finally, it is now acknowledged that lutein and zeaxanthin play a role in vision by improving contrast sensitivity and visual acuity [bib_ref] Lutein and zeaxanthin supplementation and association with visual function in age-related macular..., Liu [/bib_ref] and participate in the prevention of age-related macular degeneration [bib_ref] Lutein and zeaxanthin intake and the risk of age-related macular degeneration: a..., Ma [/bib_ref].
However, several intervention trials with carotenoid supplements have not supported these beneficial associations, or even found negative health effects [bib_ref] Mortality in randomized trials of antioxidant supplements for primary and secondary prevention:..., Bjelakovic [/bib_ref] [bib_ref] Chemoprevention of lung cancer: the beta-Carotene and Retinol Efficacy Trial (CARET) in..., Omenn [/bib_ref]. To explain this discrepancy, it is hypothesized that the food matrix (missreceptor substrate 1; ISX, intestine specific homeobox; KD, equilibrium dissociation constant; LCAT, lecithin-cholesterol acyltransferase; LDLR, low density lipoprotein receptor; LIPC, lipase C, hepatic type; LIPF, gastric lipase; LPL, lipoprotein lipase; LRAT, lecithin-retinol acyltransferase; LRP1, low density lipoprotein receptor-related protein 1; LXR, liver X receptor; MC4R, melanocortin 4 receptor; MTTP/MTP, microsomal triglyceride transfer protein/gene; NF-B, nuclear factor kappa-B; NRF2/NFE2L2, nuclear factor (erythroid-derived 2) like 2; NPC1L1, NPC1 like intracellular cholesterol transporter 1; PGA3/4/5, pepsinogen3/4/5; PGC, progastricsin; PGC1␣, peroxisome proliferator-activated receptor gamma coactivator 1alpha; PKD1L2, polycystin 1 like 2; PLRP2, pancreatic lipaserelated protein-2; PNLIP, pancreatic lipase; PPAR, peroxisome proliferator-activated receptor; PXR, pregnane X receptor; RAR, retinoic acid receptor; RBP1/3/4, retinol binding protein 1/3/4; RPE65, retinal pigment epithelium specific protein 65kDa; RSD, relative standard deviation (RSD = SD/mean), equal to CV (coefficient of variation); RXR, retinoid X receptor; RXRA, retinoid X receptor alpha; SAR1B, secretion associated Ras related GT-Pase 1B; SR-BI/SCARB1, scavenger receptor class B member 1, protein/gene; SHP, short heterodimer partner; SNP, single nucleotide polymorphism; SETD7, SET domain containing lysine methyltransferase 7; SLC27A6, solute carrier family 27 (fatty acid transporter), member 6; SOD2, superoxide dismutase 2, mitochondrial; StARD3, StAR related lipid transfer domain containing 3; STRA6, stimulated by retinoic acid gene 6 protein homolog; T2D, type II diabetes mellitus; TCF7L2, transcription factor 7 like 2; TRL, triacylglycerol-rich lipoprotein fraction; WT, wild-type ing synergistic effects, e.g. with other antioxidants such as polyphenols), a larger array of natural occurring carotenoids compared to single carotenoids in high amounts, presentation in the form of carotenoid supplements (powder, solid matrix) and continuous intake in case of supplements, influence absorption, distribution, metabolism and excretion (ADME), and therefore also their bioactivity. However, it has also been emphasized that ADME-related factors, including digestion and matrix release, solubilisation in mixed micelles, epithelial uptake in the (small) intestine, and further biodistribution, all prerequisites for exerting potential biological effects, can be different between individuals. This likely results in variable blood/tissue concentrations [bib_ref] Dual isotope test for assessing beta-carotene cleavage to vitamin A in humans, Hickenbottom [/bib_ref] [bib_ref] Interindividual variability of lutein bioavailability in healthy men: characterization, genetic variants involved,..., Borel [/bib_ref] [bib_ref] Compartmental and noncompartmental modeling of (1)(3)C-lycopene absorption, isomerization, and distribution kinetics in..., Moran [/bib_ref]. However, blood plasma/serum alone may not constitute the best indicator to assess carotenoid status, and additional methods, such as isotopic labelling, similar as for retinoids [bib_ref] Usefulness of vitamin A isotope methods for status assessment: from deficiency through..., Tanumihardjo [/bib_ref] or easy accessible compartments such as white blood cells [bib_ref] Dietary beta-carotene is taken up by blood plasma and leukocytes in dogs, Chew [/bib_ref] or buccal cells [bib_ref] Carotenoids in human buccal mucosa cells after 4 wk of supplementation with..., Paetau [/bib_ref] , may allow for more insights regarding endogenous carotenoid levels, carotenoid compartments, and turnover [bib_ref] Isotopic labeling and LC-APCI-MS quantification for investigating absorption of carotenoids and phylloquinone..., Kurilich [/bib_ref].
This intra-and interindividual variability can be attributed, in addition to dietary habits [bib_ref] Bioavailabilty of non-provitamin A carotenoids, Bohn [/bib_ref] [bib_ref] Factors affecting intestinal absorption of highly lipophilic food microconstituents (fat-soluble vitamins, carotenoids..., Borel [/bib_ref] , to host-related factors including disease state [bib_ref] Relationship between markers of HIV-1 disease progression and serum beta-carotene concentrations in..., Baeten [/bib_ref] , possibly physical activity [bib_ref] Relationship of alcohol use, physical activity and dietary habits with serum carotenoids,..., Kitamura [/bib_ref] [bib_ref] Associations of plasma carotenoids with risk factors and biomarkers related to cardiovascular..., Wang [/bib_ref] , being overweight/obese [bib_ref] Associations of plasma carotenoids with risk factors and biomarkers related to cardiovascular..., Wang [/bib_ref] , alcohol use [bib_ref] Relationship of alcohol use, physical activity and dietary habits with serum carotenoids,..., Kitamura [/bib_ref] [bib_ref] Moderate consumption of beer, red wine and spirits has counteracting effects on..., Van Der Gaag [/bib_ref] [bib_ref] Relationship between alcohol and nutrient intakes and blood C 2017 The Authors...., Walmsley [/bib_ref] , smoking habits [bib_ref] Associations of plasma carotenoids with risk factors and biomarkers related to cardiovascular..., Wang [/bib_ref] , drug intake [bib_ref] Effects of simvastatin on carotenoid status in plasma, Ryden [/bib_ref] , age [bib_ref] Correlates of serum lycopene in older women, Casso [/bib_ref] , and genetic aspects [bib_ref] Genetic variations involved in interindividual variability in carotenoid status, Borel [/bib_ref] [bib_ref] A combination of single-nucleotide polymorphisms is associated with interindividual variability in dietary..., Borel [/bib_ref]. However, the underlying mechanisms for this variability -e.g. lower bioaccessibility, reduced . Overview of host (non-dietary) factors proposed to influence (in addition to genetic make-up and malabsorption diseases of the GI) intra-and interindividual differences regarding carotenoid ADME Factor Type of study Carotenoids investigated and variability Reference Age Observational, n = 400 adults (males, females)
Younger age correlated with lower serum carotenoids [bib_ref] Human serum carotenoid concentrations are related to physiologic and lifestyle factors, Brady [/bib_ref] Observational, n = 946 postmenopausal women
Lower serum lycopene levels associated with higher age [bib_ref] Correlates of serum lycopene in older women, Casso [/bib_ref] Observational, n = 12500 Lower serum -carotene levels with older age [bib_ref] Factors influencing blood concentration of retinol, alphatocopherol, vitamin C, and beta-carotene in..., Faure [/bib_ref] Review
Higher plasma carotenoid levels with older age [bib_ref] The vitamin status and its adequacy in the elderly: an international overview, Haller [/bib_ref] Alcohol Observational, n = 2895 women No consistent effect of alcohol consumption on plasma levels of ␣-carotene, -carotene, -cryptoxanthin, and lutein-zeaxanthin [bib_ref] Associations of plasma carotenoids with risk factors and biomarkers related to cardiovascular..., Wang [/bib_ref] Observational, n = 194 men Negative correlations of plasma levels of lycopene, -carotene, -cryptoxanthin (not lutein) with units of alcohol/day, Spearman's rank correlation: -0.27 to -0.51 [bib_ref] Relationship of alcohol use, physical activity and dietary habits with serum carotenoids,..., Kitamura [/bib_ref] Observational, n = 1198 subjects
Higher alcohol consumption related to higher plasma lycopene (ca. 20%), no effect on ␣-and -carotene, lutein, and -cryptoxanthin [bib_ref] Relationship between alcohol and nutrient intakes and blood C 2017 The Authors...., Walmsley [/bib_ref] Intervention, n = 12 healthy men Consumption of wine, beer, or spirits for 3 weeks reduced plasma -carotene by 15%, no effect on lycopene, lutein, zeaxanthin, -cryptoxanthin, and ␣-carotene [bib_ref] Moderate consumption of beer, red wine and spirits has counteracting effects on..., Van Der Gaag [/bib_ref] Observational, n = 400 adults (male, female)
Higher alcohol consumption correlated with lower serum carotenoids [bib_ref] Human serum carotenoid concentrations are related to physiologic and lifestyle factors, Brady [/bib_ref] Observational, n = 12500 adults (male, female)
Lower -carotene levels with alcohol consumption [bib_ref] Factors influencing blood concentration of retinol, alphatocopherol, vitamin C, and beta-carotene in..., Faure [/bib_ref] Asthma Observational, women with (n = 84) & without asthma (n = 47)
Higher plasma total-carotenoids in women with asthma [bib_ref] Circulating antioxidant profile of pregnant women with asthma, Mclernon [/bib_ref] Body weight, BMI Observational, n = 2895 women Obese women had lower plasma levels of ␣-carotene, -carotene, -cryptoxanthin, and lutein-zeaxanthin, by ca. 10%, compared to normal-weight women. Plasma lycopene was higher by 10% [bib_ref] Associations of plasma carotenoids with risk factors and biomarkers related to cardiovascular..., Wang [/bib_ref] Observational, n = 194 men Negative correlation of BMI with serum lutein but not lycopene, -carotene, -cryptoxanthin, R: -0.12 (Spearman rank correlation) [bib_ref] Relationship of alcohol use, physical activity and dietary habits with serum carotenoids,..., Kitamura [/bib_ref] Observational, n = 400 adults (males, females)
Higher BMI associated with lower ␣-and -carotene and xanthophyll serum levels [bib_ref] Human serum carotenoid concentrations are related to physiologic and lifestyle factors, Brady [/bib_ref] Observational, n = 946 postmenopausal women
Higher BMI correlated with lower plasma lycopene levels [bib_ref] Correlates of serum lycopene in older women, Casso [/bib_ref] Observational, n = 600 healthy adults
Higher abdominal obesity related to lower serum carotenoid levels (␣-,-carotene, canthaxanthin [bib_ref] Association of abdominal obesity with decreased serum levels of carotenoids in a..., Suzuki [/bib_ref] Observational, n = 55 women Similar total adipocyte -carotene content in lean and obese, -carotene concentration reduced in obese [bib_ref] The concentration of -carotene in human adipocytes, but not the wholebody adipocyte..., Östh [/bib_ref] Gender Observational, n = 12,500 adults (male, female)
Women had higher serum -carotene levels than men [bib_ref] Factors influencing blood concentration of retinol, alphatocopherol, vitamin C, and beta-carotene in..., Faure [/bib_ref] Helicobacter pylori infection Observational, n = 49 anemic patients (male, female)
Lower gastric mucosal -carotene reported with increased H. pylori infection (though no effect on plasma -carotene levels) [bib_ref] Vitamin E and carotenoids in gastric biopsies: the relation to plasma concentrations..., Sanderson [/bib_ref] HIV Observational, n = 1669 women Lower serum -carotene levels in HIV subjects [bib_ref] HIV and other predictors of serum beta-carotene and retinol in pregnancy: a..., Friis [/bib_ref] Hyperthyroidism Observational, n = 36 patients Lower serum -carotene in subjects with hyperthyroidism compared to hypo-and euthyroidismC [bib_ref] The influence of iron and zinc supplementation on the bioavailability of provitamin..., Kana-Sop [/bib_ref] Observational, n = 400 women HIV positive
Lower serum -carotene associated with markers of disease progression, univariate regression, R: -0.083-0.244 [bib_ref] Relationship between markers of HIV-1 disease progression and serum beta-carotene concentrations in..., Baeten [/bib_ref] Observational, n = 1665 men and women, healthy and diabetic 20% lower plasma -carotene levels in diabetes subjects compared to healthy ones [bib_ref] Diabetes mellitus and serum carotenoids: findings from the Third National Health and..., Ford [/bib_ref] Blood lipids, cholesterol
Observational, n = 400 adults (male, female)
Higher non-HDL cholesterol associated with lower serum carotenoids [bib_ref] Human serum carotenoid concentrations are related to physiologic and lifestyle factors, Brady [/bib_ref] Observational, n = 12,500 (male, female)
Higher total cholesterol and lower triglycerides associated with higher -carotene in serum [bib_ref] Factors influencing blood concentration of retinol, alphatocopherol, vitamin C, and beta-carotene in..., Faure [/bib_ref] Drug intake Intervention, n = 8 volunteers Intake of simvastatin (lipid-lowering drug), 40 mg/day for 8 weeks, reduced plasma levels of carotenes (lycopene, ␣-and -carotene) and xanthophylls (-cryptoxanthin, lutein), by 5 and 21%, respectively [bib_ref] Effects of simvastatin on carotenoid status in plasma, Ryden [/bib_ref] Intervention trial, n = 6 patients (1 male, 5 female)
Intake of orlistat (lipid-lowering drug) decreased levels of ␣-, and -carotene in plasma [bib_ref] Effects of orlistat therapy on plasma concentrations of oxygenated and hydrocarbon carotenoids, Sundl [/bib_ref] Intervention trial, n = 228 obese subjects (male, female)
Intake of orlistat (lipid-lowering drug) decreased -carotene levels in plasma [bib_ref] One-year treatment of obesity: a randomized, double-blind, placebo-controlled, multicentre study of orlistat,..., Finer [/bib_ref] Malaria Observational, n = 100 malaria and 50 control children (boys, girls)
Lower serum concentration of all major carotenoids compared to control [bib_ref] Plasma alphatocopherol, retinol, and carotenoids in children with falciparum malaria, Das [/bib_ref] Menstrual cycle Intervention trial, n = 9 women Lower plasma carotenoids during early than late follicular phase [bib_ref] Effect of menstrual cycle phase on the concentration of individual carotenoids in..., Forman [/bib_ref] Microbiota Observational, n = 25 subjects (males, females)
Collinsella spp. were reduced in subjects with atherosclerosis. These subjects had lower -carotene in serum and the metagenome showed lower phytoene-dehydrogenase. [bib_ref] Symptomatic atherosclerosis is associated with an altered gut metagenome, Karlsson [/bib_ref] Physical activity Observational, n = 2895 women Exercising women (>1 time/week) had higher levels of ␣-carotene, -carotene, -cryptoxanthin, and lutein-zeaxanthin, by ca. 5-10%, compared to normal-weight individuals without exercise. No effect on lycopene [bib_ref] Associations of plasma carotenoids with risk factors and biomarkers related to cardiovascular..., Wang [/bib_ref] Observational, n = 194 men Positive correlation of plasma levels of -cryptoxanthin and lutein with physical activity, rho: 0.12 to 0.17 [bib_ref] Relationship of alcohol use, physical activity and dietary habits with serum carotenoids,..., Kitamura [/bib_ref] Race absorption, altered tissue distribution, turnover, and excretion as well as possible interactions of individual carotenoids on absorption and bio-activation of other carotenoids [bib_ref] Carotenoid interactions, Van Den Berg [/bib_ref] [bib_ref] Plasma carotenoid response to chronic intake of selected foods and beta-carotene supplements..., Micozzi [/bib_ref] , are only poorly understood. These factors can result in huge variability of carotenoid absorption and circulating plasma levels [fig_ref] Table 2: C 2017 The Authors [/fig_ref]. In a double tracer study [bib_ref] Dual isotope test for assessing beta-carotene cleavage to vitamin A in humans, Hickenbottom [/bib_ref] with D 6 -carotene (37 mol), lowest AUC (area under the plasmaconcentration-time curve, mol h/L) versus highest AUC were found to be 0.01 and 30.00, respectively. Major host factors influencing carotenoid ADME patterns are likely to include:
(i) Factors influencing carotenoid release from the food matrix, and their transition from lipid droplets to mixed micelles, i.e. factors impacting bioaccessibility. This includes genes responsible for the expression of digestive enzymes (e.g. gastric lipase, cholesterol esterase, pancreatic lipase, etc.), and bile acid formation aiding in carotenoid micellization [bib_ref] Bioavailabilty of non-provitamin A carotenoids, Bohn [/bib_ref] [bib_ref] In vitro models for studying secondary plant metabolite digestion and bioaccessibility, Alminger [/bib_ref] ; (ii) Factors altering carotenoid uptake into (or efflux out of) the intestinal epithelium. This encompasses uptake/efflux transporters such as scavenger receptor class B member 1 (SR-BI), cluster of differentiation 36 (CD36), and Niemann-Pick C1 like intracellular cholesterol transporter 1 (NPC1L1), and perhaps other ATP-binding cassette (ABC) proteins such as ABCG5/G8, or ABCA1 [bib_ref] Genetic variations involved in interindividual variability in carotenoid status, Borel [/bib_ref] [bib_ref] Proteins involved in uptake, intracellular transport and basolateral secretion of fat-soluble vitamins..., Reboul [/bib_ref] , but also small intestinal surface available for absorption; (iii) Factors contributing to intracellular cleavage, especially BCO1/2 (-carotene oxygenase 1/2), responsible for centric/asymmetric cleavage of carotenoids, respectively, producing a variety of retinoids and potential endogenous occurring apocarotenoids [bib_ref] Enzymology of the carotenoid cleavage dioxygenases: reaction mechanisms, inhibition and biochemical roles, Harrison [/bib_ref] [bib_ref] Complex interactions between dietary and genetic factors impact lycopene metabolism and distribution, Moran [/bib_ref] [bib_ref] Two carotenoid oxygenases contribute to mammalian provitamin A metabolism, Amengual [/bib_ref] ; (iv) Factors that impinge on carotenoid intracellular transport in the gut epithelium, i.e. lecithin-retinol acyltransferase (LRAT) for retinol (and perhaps other -carotene cleavage products), and maybe intestinal fatty acid binding protein (FABP2/I-FABP); (v) Factors altering the secretion of carotenoid-containing chylomicrons into the lymphatic system, such as apolipoprotein B 48 (APOB48), APOAIV, SAR1B (secretion associated Ras related GTPase1B), and microsomal triglyceride transfer protein (MTTP) [bib_ref] Complex interactions between dietary and genetic factors impact lycopene metabolism and distribution, Moran [/bib_ref] ; (vi) Factors influencing carotenoid transport in the blood plasma/serum, such as their further distribution into lipoproteins, e.g. lipoprotein lipase (LPL), APOA-I, APOB, APOE and perhaps APOC3, and low density lipoprotein receptor (LDLR) [bib_ref] Influence of apolipoprotein E genotype on fat-soluble plasma antioxidants in Spanish children, Ortega [/bib_ref] ; and the liver such as by hepatic lipase (LIPC); (vii) Deposition of carotenoids in "target tissues", e.g. the macula (lutein, zeaxanthin), influenced by SR-BI, glutathione S-transferase P1 isoform (GSTP1), StAR-related lipid transfer domain protein 3 (StARD3), BCO1, cholesterol transporters (SR-BI, ABCA1, ABCG5/8), retinal pigment epithelial-specific protein (RPE65), elongation of very long chain fatty acids like 2 (ELOVL2), and those involved in visual pigment metabolism [bib_ref] Genetic determinants of macular pigments in women of the Carotenoids in Age-Related..., Meyers [/bib_ref]. However, also deposition in adipocytes, involving e.g. LDLR, could play a role [bib_ref] Carotenoids and their conversion products in the control of adipocyte function, adiposity..., Bonet [/bib_ref] ; (viii) Any factor associated with carotenoid catabolism and excretion (in addition to BCO1/2), possibly those involving cytochrome P450 enzymes or the aryl-hydrocarbonreceptor [bib_ref] Ah receptor-dependent CYP1A induction by two carotenoids, canthaxanthin and beta-apo-8'-carotenal, with no..., Gradelet [/bib_ref] ; (ix) The microbiota. This may alter e.g. patterns and concentrations of secondary bile acids [bib_ref] Gut microbiota regulates bile acid metabolism by reducing the levels of tauro-beta-muricholic..., Sayin [/bib_ref] , or possibly carotenoid absorption or degradation patterns [bib_ref] The bioavailability of alpha-and beta-carotene is affected by gut microflora in the..., Grolier [/bib_ref] , though it is not sure if a significant fraction of carotenoids can be absorbed from the colon [bib_ref] Mind the gap-deficits in our knowledge of aspects impacting the bioavailability of..., Bohn [/bib_ref]. (x) Effects of individual carotenoids on the absorption, binding, transport and bioactivation of other carotenoids as well as selective absorption, binding, transport and bioactivation of individual carotenoids [bib_ref] Carotenoid interactions, Van Den Berg [/bib_ref]. (xi) Any factor associated with vitamin A/retinoid storage and metabolism.
Thus, as individual responses can depend on many varying factors, it is paramount to understand these and their influence on the biological variability of carotenoid ADME. In this review, it is aimed to highlight known host-related factors that predispose for variations in carotenoid metabolism, such as genetic factors (e.g. single nucleotide polymorphisms (SNPs)), though additional ones (disease state, body weight, smoking, physical activity etc.) are also briefly reviewed. The manuscript structure is oriented around the metabolic path of carotenoids, from digestion (chapter 2) to intestinal absorption (chapter 3) and further transport to the liver (chapter 4) and distribution to target tissues (chapter 5) to storage and excretion related pathways (chapter 6). Searched databased included Pubmed and Scopus, for all years, in English language, employing the following search terms (abstract and title) to start with: "Human* AND (lutein OR lycopene OR xanthophyll OR carotene*) AND (bioavailab* OR pharmacokinetic* OR kinetic* OR absorption OR postprandial OR metabol* OR microb* OR microflora OR biliary OR enterhohepatic* OR chylomicron OR plasma OR tissue OR metabolism OR enterocyte OR lipoproteins OR transporters OR Single nucleotide polymorphism* OR genetic varia* OR SNP OR cleavage OR enzym* OR intestine) AND (intra* OR inter*) NOT (druginteraction OR in-vitro)", though additional literature following the primary search results were surveyed.
## Host factors influencing digestion
aspects -from matrix release to bioaccessibility
## General aspects and oral phase of digestion
Bioavailability of carotenoids depends on their bioaccessibility, i.e. the release from the food matrix and subsequent availability for absorption. As carotenoids are apolar, with octanol/water partition coefficients of 8-12 [bib_ref] Effects of physicochemical properties of carotenoids on their bioaccessibility, intestinal cell uptake,..., Sy [/bib_ref] , their incorporation into mixed micelles is necessary prior to their cellular uptake, which is assumed to take place predominantly in the small intestine. Mastication during oral digestion results in enhanced surface area and the breakdown into smaller particles. In addition, saliva appears to contain some lipase activity [bib_ref] Lingual lipase activity in the orosensory detection of fat by humans, Kulkarni [/bib_ref] , though not necessarily lingual lipase (a triacylglycerol-lipase, EC 3.1.1.3) [bib_ref] Intestinal lipid absorption and transport, Phan [/bib_ref] [bib_ref] The role of lipolysis in human orosensory fat perception, Voigt [/bib_ref] [fig_ref] Table 3: Host factors influencing carotenoid release from food matrix and bioaccessibility [/fig_ref]. As exposure in the oral cavity is rather short (usually less than 1 min), the enzymatic effect on carotenoid bioavailability is presumably small, though smaller particle size has been related to improved carotenoid bioavailability [bib_ref] A comprehensive overview on the micro-and nano-technological encapsulation advances for enhancing the..., Soukoulis [/bib_ref]. To our knowledge, no mutagenesis on or polymorphisms with effects on oral lipases and lipid digestion has been reported to date. As salivary alpha-amylase (EC 3.2.1.1) participates in the break-down of starch, food matrices rich in both starch and carotenoids, such as sweet potato, may be influenced by alterations in alpha-amylase levels. It has been reported that in populations traditionally exposed to high levels of starch, more copies of the salivary amylase gene (AMY1) and higher enzyme levels were found [bib_ref] Diet and the evolution of human amylase gene copy number variation, Perry [/bib_ref] , though its influence on the digestion of carotenoids has never been investigated.
## Gastric phase of digestion
In the stomach, the primary digestion enzymes include pepsin (3.4.23.1) and gastric lipase (3.1.1.3), though orally secreted lipases may still be active. In addition, a small amount of phospholipids [bib_ref] A standardised static in vitro digestion method suitable for food -an international..., Minekus [/bib_ref] is released from the mucus layer [bib_ref] Secretion of gastric mucus phospholipids in response to beta-adrenergic G proteincoupled receptor..., Slomiany [/bib_ref] , aiding in the emulsification of lipophilic constituents. The pH may have an influence, as low pH can result in the degradation of epoxy-carotenoids (e.g. violaxanthin, neoxanthin), resulting in epoxide-furanoid transitions [bib_ref] Divalent minerals decrease micellarization and uptake of carotenoids and digestion products into..., Biehler [/bib_ref]. Human gastric pH is influenced mostly by meal, with a complex meal increasing the pH from initially 2 to 3-5, though interindividual differences in fasting pH exist [bib_ref] In vitro models for studying secondary plant metabolite digestion and bioaccessibility, Alminger [/bib_ref].
A few native foods are rich in both proteins and carotenoids, including egg yolk, salmon, and some types of cheese, and protein digestion could contribute to the release of carotenoids. In addition, (partly) digested proteins may aid in emulsifying carotenoids [bib_ref] A comprehensive overview on the micro-and nano-technological encapsulation advances for enhancing the..., Soukoulis [/bib_ref]. Expression of pepsin has been reported to depend on the pepsinogen genes PGA3, PGA4, PGA5, and progastricsin (PGC) [bib_ref] Transcription regulation of human and porcine pepsinogen A, Pals [/bib_ref]. However, varying the amount of pepsin in in vitro trials did not appear to have measurable effects on carotenoid bioaccessibility from leafy vegetables [bib_ref] Dietary and host-related factors influencing carotenoid bioaccessibility from spinach (Spinacia oleracea), Biehler [/bib_ref] , and at least for such and similar sources, variations in pepsin are not expected to contribute to plasma level variability. Similarly, using a test meal composed of meat, carrots, spinach and tomato paste, gastric digestion (in vitro) had no significant influence on carotenoid bioaccessibility [bib_ref] Development of an in vitro digestion method to assess carotenoid bioavailability from..., Garrett [/bib_ref] , suggesting rather small effects on carotenoid bioavailability at this step, though in these trials, gastric lipase was not involved. By contrast, Periago et al. [bib_ref] Detection of key factors affecting lycopene in vitro accessibility, Periago [/bib_ref] reported a positive effect of pepsin on lycopene micellization from a puree in vitro. It is possible that for this very apolar carotenoid, protein degradation products added to the emulsifying effect, or aided in matrix breakdown.
The genes related to the production and secretion of mucus containing phospholipids, which could aid in the emulsification process, are not clearly identified. Concentration variations of phospholipids between 0.03 and 0.6 mM have been reported, [bib_ref] Designing a dynamic dissolution method: a review of instrumental options and corresponding..., Culen [/bib_ref] and may be expected to have some influence on carotenoid micellization. However, their influence and strengths of effect are unknown and would also be superseded by dietary phospholipids, which are expected to play a more important role. This would be true especially following ingestion of lipid-rich meals (a mean intake of 2-8 g/d of phosphatidylcholine has been reported [bib_ref] Health effects of dietary phospholipids, Kullenberg [/bib_ref] , which would translate into ca. 8 mM (if taken within 1 out of 3 major meals per day, and dissolved in 1 L gastric fluid).
Gastric lipase, encoded by the LIPF (lipase F, gastric type) gene [bib_ref] Comparative studies of mammalian acid lipases: Evidence for a new gene family..., Holmes [/bib_ref] and secreted by gastric chief cells, can digest up to 25% of the ingested lipids [bib_ref] In vitro models for studying secondary plant metabolite digestion and bioaccessibility, Alminger [/bib_ref]. It thus could be expected to influence the accumulation of carotenoids in lipid droplets, and their degradation, important for the following transition of carotenoids from lipid droplets to mixed micelles. This occurs mostly in the small intestine. Unfortunately, gastric lipase cannot, at present, be studied in vitro, due to the unavailability of human gastric lipase. Other sources, such as those from fungi, have different cleavage kinetics, differing in their pH optimum and also the type sequence of cleavage [bib_ref] Relevant pH and lipase for in vitro models of gastric digestion, Sams [/bib_ref]. Rabbit lipase would be an interesting option, but is not commercially available. Some, such as lipase from the fungus Rhizopus oryzae have been tested (cleavage optimum pH 5-9), though no significant improvement in bioaccessibility was found [bib_ref] Selective factors governing in vitro beta-carotene bioaccessibility: negative influence of low filtration..., Corte-Real [/bib_ref].
## Small intestinal phase of digestion
The most crucial step influencing carotenoid bioacessibility is the small intestinal phase. Here, micellization occurs or is completed, following the secretion of bile salts, in addition to pancreatic lipase, and additional enzymes (pancreatic amylase, nucleosidases, trypsinogen, chymotrypsinogen, carboxypeptidase, elastases, phospholipases, and carboxyl ester lipase). Bile salts aid in the emulsification process and formation and stability of the mixed micelles, while pancreatic lipase produces free fatty acids and monoglycerides, fostering emulsification. Thus, it can be expected that modifications of both bile-acid and pancreatic lipase secretions have strong effects on the micellization of carotenoids, a pre-requisite for their diffusion to the unstirred water layer prior to absorption. This has been confirmed by several in vitro studies, where micellization and resulting bioaccessibility was very much compromised when either bile salts or pancreatic lipase were missing. Without bile, bioaccessibility of total carotenoids fell to 30%, and without pancreatic lipase or both, to below 5% of their original value [bib_ref] Dietary and host-related factors influencing carotenoid bioaccessibility from spinach (Spinacia oleracea), Biehler [/bib_ref]. Similar strong effect were found by Garret et al. [bib_ref] Development of an in vitro digestion method to assess carotenoid bioavailability from..., Garrett [/bib_ref] , where total carotenoid micellization dropped below 5% of the original values without bile salts. The effect of pancreatin was less drastic (reduction by approximately 50%), possibly due to differences between test meals. In order to study factors influencing lycopene bioaccessibility, tomato puree was digested under various conditions, testing among other factors gastric pH, gastric digestion time, pepsin concentration, intestinal pH, pancreatin concentration, bile salt concentration, colipase addition and intestinal digestion time [bib_ref] Detection of key factors affecting lycopene in vitro accessibility, Periago [/bib_ref]. It was found that only pepsin positively influenced micellization, while olive oil had a slightly negative effect, likely due to entrapment of lycopene by non-hydrolysed olive oil.
Following intragastric in vivo administration of carotenoid rich meals, duodenal fluid was aspirated, and micellization determined [bib_ref] Processing of vegetable-borne carotenoids in the human stomach and duodenum, Tyssandier [/bib_ref]. Variability between subjects' micellization efficacy (fractional bioaccessibility) was considerably lower compared to plasma or triacylglycerol-rich lipoprotein (TRL) carotenoid variability following interventions, being 20, 23, and 32%, respectively for -carotene, lutein and lycopene, vs. typically 50-80% for plasma, though variations between studies can be considerable [fig_ref] Table 3: Host factors influencing carotenoid release from food matrix and bioaccessibility [/fig_ref]. This may point out that, although enzyme or bile salt concentrations surely play a role in interindividual variation, an additional and about equal portion of variability is added during and after absorption. Bile acid production by the liver is governed by a variety of genes, involving for instance bile acid synthetic enzyme (CYP7A1), activators of CYP7A1 expression such as HNF4␣ (hepatocyte nuclear factor 4 alpha, encoded by HNF4A), and PGC1␣ (encoded by PPARGC1A), repressors of CYP7A1 (farnesoid X receptor (FXR, encoded by NR1H4)), short heterodimer partner (SHP, encoded by NR0B2), G protein pathway suppressor 2 (GPS 2, encoded by GPS2), pregnane X receptor (PXR, encoded by NR1I2), fibroblast growth factor 19 (FGF19; encoded by FGF19), fibroblast growth factor receptor 4 (FGFR4; encoded by FGFR4), klotho B (encoded by KLB), and forkhead box O1 (FOXO1; encoded by FOXO1) [bib_ref] Association of genes involved in bile acid synthesis with the progression of..., Inamine [/bib_ref] , however, their role in carotenoid absorption and tissue variability has not been examined.
At least three lipases are secreted from the pancreas, including pancreatic triglyceride lipase (encoded by PN-LIP), which is the most abundant lipase (producing sn2monoacylglycerol and free fatty acids), but also two homologues, pancreatic lipase-related proteins 1 (not apparently active regarding lipolysis) and 2 (PLRP1 and PLRP2) [bib_ref] Properties and function of pancreatic lipase related protein 2, Lowe [/bib_ref]. PLRP2 possess a broader substrate specificity, also cleaving, unlike PNLIP, phospholipids and galactolipids. The frequency of a SNP in the PLRP2 gene (rs4751995) has been associated with populations historically consuming a diet rich in cereals, and may have repercussions on lipid digestion [bib_ref] Colloquium paper: human adaptations to diet, subsistence, and ecoregion are due to..., Hancock [/bib_ref].
Though pancreatic triglyceride lipase activity is usually reduced by bile-salts, this effect is offset by colipase, also secreted by the pancreas [bib_ref] Intestinal lipid absorption, Iqbal [/bib_ref]. Formation of colipase preprotein is regulated by the CLPS gene, and mice deficient for CLPS showed lower survival and weight gain on a high-fat diet, suggesting the inability to cope with lipids on a high fat diet [bib_ref] Decreased postnatal survival and altered body weight regulation in procolipase-deficient mice, D'agostino [/bib_ref]. A polymorphism for the gene encoding procolipase has been related to lipid metabolism and diabetes risk [bib_ref] A polymorphism in the gene encoding procolipase produces a colipase, Arg92Cys, with..., D'silva [/bib_ref] , and would be an interesting candidate also regarding carotenoid metabolism. In a recent study, a SNP in PNLIP (rs11197742) was found in a combination of SNPs associated with chylomicron secretion of lycopene [bib_ref] Lycopene bioavailability is associated with a combination of genetic variants, Borel [/bib_ref] , although its contribution was rather low and did not reach statistical significance when investigated individually (p = 0.086). Several SNPs in PNLIP have been reported in children, and the latter was related to altered plasma lipoprotein and total cholesterol concentrations [bib_ref] Polymorphisms in PNLIP, encoding pancreatic lipase, and associations with metabolic traits, Hegele [/bib_ref] , as well as with lycopene bioavailability [fig_ref] Table 4: List of SNPs known, or speculated, to influence carotenoid metabolism C 2017... [/fig_ref].
Carboxyl-ester lipase (CEL), also termed cholesterolesterase, typically cleaves cholesterol esters in the gut, and its ability to cleave carotenoid esters, such as of lutein, present in many leafy vegetables, has been controversially discussed [bib_ref] Carotenol fatty acid esters: easy substrates for digestive enzymes? Comp, Breithaupt [/bib_ref]. At least five types of CEL are known, though human carboxylesterases CES1 and CES2 may play the most important role during digestion. These are situated on the gut mucosa (brush border enzymes), and have shown to cleave carotenoid esters [bib_ref] Hydrolysis of zeaxanthin esters by carboxyl ester lipase during digestion facilitates micellarization..., Chitchumroonchokchai [/bib_ref]. Its origin (pancreatic vs. enterocyte) remains somewhat unclear. However, this cleavage is expected to influence bioavailability, as the more apolar esters are characterized by lower micellization efficiency and absorption than the cleaved carotenoids [bib_ref] Methods for assessing aspects of carotenoid bioavailability, Biehler [/bib_ref]. In fact, in plasma and circulating chylomicrons, free xanthophylls are almost exclusively found, suggesting that cleavage is in fact quite complete [bib_ref] Plasma appearance of unesterified astaxanthin geometrical E/Z and optical R/S isomers in..., Coral-Hinostroza [/bib_ref] , though reduced absorption of the esters could play a role. A number of SNPs have been described in humans for CES1 and CES2 [bib_ref] Twelve novel single nucleotide polymorphisms in the CES2 gene encoding human carboxylesterase..., Kim [/bib_ref] , though not in relation to carotenoid or lipophilic phytochemical/micronutrient metabolism.
Certain diseases such as pancreatitis may also result in lower secretion of digestion enzymes [bib_ref] Pancreatic enzymes: secretion and luminal nutrient digestion in health and disease, Layer [/bib_ref]. Also during older age reduction of lipid absorption has been reported, perhaps also due to reduced epithelial surface [bib_ref] The intestinal epithelial barrier: does it become impaired with age?, Meier [/bib_ref] , which may thus be expected to correlate with lower carotenoid absorption, as suggested by some, though not all studies .
## Host factors determining aspects of intestinal absorption
## Factors influencing cellular uptake and cleavage
Following their extraction from the food matrix and incorporation, at least in part, into mixed micelles, carotenoids are taken up by enterocytes. This process is not only passive, as previously thought [bib_ref] beta-carotene intestinal absorption: bile, fatty acid, pH, and flow rate effects on..., Hollander [/bib_ref] , and several apical membrane proteins have been shown to facilitate carotenoid uptake [bib_ref] Proteins involved in uptake, intracellular transport and basolateral secretion of fat-soluble vitamins..., Reboul [/bib_ref]. SR-BI, encoded by SCARB1, is involved in the uptake of carotene [bib_ref] CD36 and SR-BI are involved in cellular uptake of provitamin A carotenoids..., Borel [/bib_ref] [bib_ref] Class B scavenger receptor-mediated intestinal absorption of dietary beta-carotene and cholesterol, Van Bennekum [/bib_ref] , lutein [bib_ref] Carotenoid transport is decreased and expression of the lipid transporters SR-BI, NPC1L1,..., During [/bib_ref] and lycopene [bib_ref] Lycopene absorption in human intestinal cells and in mice involves scavenger receptor..., Moussa [/bib_ref]. CD36 facilitates -carotene [bib_ref] CD36 and SR-BI are involved in cellular uptake of provitamin A carotenoids..., Borel [/bib_ref] uptake and could facilitate lycopene uptake [bib_ref] CD36 is involved in lycopene and lutein uptake by adipocytes and adipose..., Moussa [/bib_ref] , while NPC1L1 participates in the uptake of lutein [bib_ref] Involvement of cholesterol membrane transporter Niemann-Pick C1-like 1 in the intestinal absorption..., Sato [/bib_ref]. All of these proteins have SNPs in their encoding genes associated with carotenoid plasma concentrations [fig_ref] Table 4: List of SNPs known, or speculated, to influence carotenoid metabolism C 2017... [/fig_ref] , and their contribution to carotenoid uptake has been confirmed in cellular models (e.g. human Caco-2 cell line), but also in models employing transfected kidney (HEK) cells. After enterocyte uptake, carotenoids can be metabolized by BCO1 [bib_ref] The human enzyme that converts dietary provitamin A carotenoids to vitamin A..., Dela Sena [/bib_ref] and BCO2 [bib_ref] Two carotenoid oxygenases contribute to mammalian provitamin A metabolism, Amengual [/bib_ref]. BCO1 catalyses the oxidative cleavage of provitamin A carotenoids (chiefly -carotene, ␣-carotene, cryptoxanthin), apo-carotenals, and lycopene, but not that of lutein [bib_ref] Substrate specificity of purified recombinant human beta-carotene 15,15'-oxygenase (BCO1), Dela Sena [/bib_ref]. BCO1 is presumably the main cleaving-enzyme for -carotene [bib_ref] Provitamin A metabolism and functions in mammalian biology, Von Lintig [/bib_ref]. Lycopene was suggested to be predominantly cleaved by BCO2 [bib_ref] Lycopenoids: are lycopene metabolites bioactive?, Lindshield [/bib_ref] , while recently lycopene cleavage by BCO1 was also reported [bib_ref] Purified recombinant human -carotene 15-15 -oxygenase 1 (BCO1) cleaves -apocarotenals and lycopene, Sena [/bib_ref]. However, until now no lycopene derived BCO1-products were determined [bib_ref] substrate specificity of purified recombinant chicken betacarotene 9',10'-oxygenase (BCO2), Sena [/bib_ref] [bib_ref] Mammalian carotenoid-oxygenases: key players for carotenoid function and homeostasis, Lobo [/bib_ref] and were only postulated [bib_ref] Lycopenederived bioactive retinoic acid receptors/retinoid-X receptors-activating metabolites may be relevant for lycopene's..., Aydemir [/bib_ref] [bib_ref] Lycopene-induced nuclear hormone receptor signaling in inflammation and lipid metabolism via still..., Rühl [/bib_ref]. BCO2 has also been shown to be involved in lutein metabolism [bib_ref] A mitochondrial enzyme degrades carotenoids and protects against oxidative stress, Amengual [/bib_ref]. Most -carotene conversion (>70%) is thought to occur in the intestine; by using stable isotope techniques it was estimated that about 20-30% occurs after absorption [bib_ref] Shortterm (intestinal) and long-term (postintestinal) conversion of beta-carotene to retinol in adults..., Tang [/bib_ref] , contributing to overall vitamin A homeostasis. In addition, a controlled temporal and spatial conversion of carotenoids to bioactive retinoids is also of physiological importance, indicated by a specific pattern of BCO1 expression in various tissues [bib_ref] Evidence for compartmentalization of mammalian carotenoid metabolism, Palczewski [/bib_ref]. This expression is linked to RAR-mediated signaling [bib_ref] Two carotenoid oxygenases contribute to mammalian provitamin A metabolism, Amengual [/bib_ref] [bib_ref] Beta-carotene reduces body adiposity of mice via BCMO1, Amengual [/bib_ref].
The involvement of several proteins in the intestinal absorption of carotenoids (apical uptake) suggests that variations in the genes encoding these proteins could modulate carotenoid absorption efficiency. This has been confirmed in an association study by Borel et al. [bib_ref] Human plasma levels of vitamin E and carotenoids are associated with genetic..., Borel [/bib_ref] where the influence [bib_ref] A combination of single-nucleotide polymorphisms is associated with interindividual variability in dietary..., Borel [/bib_ref] [bib_ref] Common variation in the beta-carotene 15,15'-monooxygenase 1 gene affects circulating levels of..., Ferrucci [/bib_ref] rs4889286
[formula] -carotene d) [297] rs12934922 -carotene d) [297] rs4889293 ␣-carotene d) [297] rs4889286 ␣-carotene d) [297] rs12918164 -cryptoxanthin d) [297] rs4889293 -cryptoxanthin d) [297] rs56389940 [/formula]
Lutein/zeaxanthin d) [bib_ref] beta-Carotene 15,15'-monooxygenase 1 single nucleotide polymorphisms in relation to plasma carotenoid and..., Hendrickson [/bib_ref] rs10048138
Lutein/zeaxanthin d) of candidate SNPs of genes involved in lipid metabolism on the fasting blood concentration of several carotenoids was investigated. More specifically, SNPs in SCARB1 were associated with -carotene but not with lycopene concentrations. These SNPs explained differences in -carotene plasma concentrations by up to 50%. Several additional SNPs have meanwhile been identified, including several in BCO1 in genomewide association studies [bib_ref] Genetic variations involved in interindividual variability in carotenoid status, Borel [/bib_ref] [bib_ref] Common variation in the beta-carotene 15,15'-monooxygenase 1 gene affects circulating levels of..., Ferrucci [/bib_ref] [bib_ref] Imputation of variants from the 1000 Genomes Project modestly improves known associations..., Wood [/bib_ref]. Three recent studies have reported associations of combinations of SNPs involved in interindividual variability of the bioavailability of lutein [bib_ref] Interindividual variability of lutein bioavailability in healthy men: characterization, genetic variants involved,..., Borel [/bib_ref] , lycopene [bib_ref] Lycopene bioavailability is associated with a combination of genetic variants, Borel [/bib_ref] and -carotene [bib_ref] A combination of single-nucleotide polymorphisms is associated with interindividual variability in dietary..., Borel [/bib_ref] , employing a candidate gene approach in postprandial studies. In these, plasma-TRL carotenoids, representing newly absorbed carotenoids, were measured in healthy male adults. These combinations were associated with 73, 72, and 69% of the interindividual variability of the bioavailability of lutein, lycopene and -carotene, respectively. While some SNPs were located in genes expressed in other tissues or were closely involved in plasma-TRL metabolism, others were involved with carotenoid transport or metabolism at the enterocyte level. These included ABCA1, ABCG5, BCMO1, CD36, ELOVL2, and ISX (intestine specific homeobox). Interestingly, one SNP in ELOVL2 (rs9468304) was very strongly associated with all three phenotypes, possibly due to the inhibitory effect of eicosapentaenoic acid, which is further elongated to docosapentaenoic acid and docosahexaenoic acid by ELOVL2, on carotenoid absorption, as has been shown with -carotene [bib_ref] Eicosapentaenoic acid inhibits intestinal beta-carotene absorption by downregulation of lipid transporter expression..., Mashurabad [/bib_ref].
## Influence of nutritional status
Host vitamin A status has been linked with -carotene absorption variability. Lobo et al. [bib_ref] ISX is a retinoic acid-sensitive gatekeeper that controls intestinal beta,beta-carotene absorption and..., Lobo [/bib_ref] demonstrated that the intestinal transcription factor ISX acts as a repressor of SCARBI and BCO1 expression following retinoic acid induction. This mechanism is thought to serve as a negative feedback loop regulating retinal and further retinoic acid, retinyl esters and retinol status through modulation of provitamin A carotenoid absorption and cleavage efficiencies. Interestingly, the same team has reported the existence of an SNP in the ISX binding site in the BCO1 promoter (rs6564851) which was associated with decreased conversion rates by 50% and increased fasting blood levels of -carotene [bib_ref] Genetics and diet regulate vitamin A production via the homeobox transcription factor..., Lobo [/bib_ref]. Though the mechanisms are not fully elucidated, low iron status was suggested to interact with retinol homeostasis, resulting in decreased mobilization of liver vitamin A and thus low serum concentrations [bib_ref] Kinetic analysis shows that iron deficiency decreases liver vitamin A mobilization in..., Jang [/bib_ref] , possibly involving altered BCO1 activity [bib_ref] Betacarotene 15,15'-dioxygenase activity is responsive to copper and iron concentrations in rat..., During [/bib_ref]. Also a low zinc status appears to reduce -carotene absorption from the gut [bib_ref] Low zinc intake decreases the lymphatic output of retinol in rats infused..., Noh [/bib_ref] , perhaps as phospholipase A2 can bind zinc and may be more active. These effects were confirmed in human studies, where supplementation with iron and zinc following a vitamin A deficient diet improved retinol and carotenoid plasma appearance, respectively [bib_ref] The influence of iron and zinc supplementation on the bioavailability of provitamin..., Kana-Sop [/bib_ref]. Also low protein status appears to hinder conversion of -carotene to vitamin A, contributing to carotenoid variability [bib_ref] Molecular and dietary regulation of beta,beta-carotene 15,15'-monooxygenase 1 (BCMO1), Lietz [/bib_ref].
BCO1 and BCO2 were also described to be controlled by peroxisome proliferator-activated receptor (PPAR) -retinoid X receptor (RXR) mediated signaling [bib_ref] Beta,beta-carotene decreases peroxisome proliferator receptor gamma activity and reduces lipid storage capacity..., Lobo [/bib_ref] [bib_ref] Dietary lycopene downregulates carotenoid 15,15'-monooxygenase and PPAR-gamma in selected rat tissues, Zaripheh [/bib_ref]. The endogenous ligands of the PPARs ␣, /␦ and ␥ are ranging from free fatty acids to various eicosanoids such as prostaglandins, leukotrienes and mono-hydroxylated fatty acids [bib_ref] Endogenous ligands for nuclear receptors: digging deeper, Schupp [/bib_ref]. The RXR was described to be activated by 9-cis-retinoic acid (9CRA) [bib_ref] 9-cis-retinoic acid is a natural antagonist for the retinoic acid receptor response..., Carlberg [/bib_ref] , as well as the newly found endogenous relevant ligand 9-cis-13, 14-dihydro-retinoic acid/9CDHRA [bib_ref] 9-cis-13,14-dihydroretinoic acid is an endogenous retinoid acting as RXR ligand in mice, Rühl [/bib_ref]. It is debated whether 9CRA occurs endogenously [bib_ref] An endogenous mammalian retinoid X receptor ligand, at last! Chem, De Lera [/bib_ref]. Currently, 9CRA is considered mainly as a ligand that is present after high non-physiological and non-nutritional relevant vitamin A intake, leaving 9CDHRA as the principal endogenous and the nutritional relevant RXR ligand. PPAR ligands are mainly food derived [bib_ref] Endogenous ligands for nuclear receptors: digging deeper, Schupp [/bib_ref] , while the nutritional precursors of the endogenous RXR ligand 9CDHRA were not yet identified. The PPAR-regulatory pathway of BCO1/2 expression and further carotenoid bioactivation is thus controlled by the amount and fractional distribution of lipids present in the food matrix. In addition to genomic regulation of BCO1/2 expression, carotenoid cleavage can also be modulated by inhibitory effects of lutein on -carotene cleavage [bib_ref] Effect of lutein on beta-carotene absorption and cleavage, Van Den Berg [/bib_ref]. This indicated that not just the individual carotenoid concentration is of relevance to bioactivation towards retinoic acid and further transcriptomic regulation, but also the concentration of carotenoids inhibiting this metabolic step, as well as their concentration relative to -carotene. The consequences of BCO1/2 mediated regulation of retinoic acid synthesis and further transcriptional signaling by additional factors and its consequences for our health will be discussed later (chapter 7), highlighting the special importance of BCO1/2 on explaining interindividual variability, likely related to the beneficial health effects of carotenoids.
## Colonic fermentation as an interindividual source
To date, it is unclear to what extent the microbiota contributes to carotenoid metabolism, and whether carotenoids/ their metabolites can be taken up in the colon. It is known that a large proportion of carotenoids reaches the colon, as only 5-50% are absorbed in the small intestine. It is also known that carotenoids are partly bioaccessible in the colon [bib_ref] Bioaccessibility of beta-carotene, lutein, and lycopene from fruits and vegetables, Goni [/bib_ref]. However, only 10-50% of the carotenoids remain intact after fermentation, while the remainder reacts to unknown compounds [bib_ref] Bioaccessibility of beta-carotene, lutein, and lycopene from fruits and vegetables, Goni [/bib_ref] [bib_ref] Carotenoid and polyphenol bioaccessibility and cellular uptake from plum and cabbage varieties, Kaulmann [/bib_ref] [bib_ref] Determination of beta-carotene and lutein available from green leafy vegetables by an..., Serrano [/bib_ref]. This was supported by carotenoid standards as the only fermentation source in vitro, as >98% losses for -carotene and zeaxanthin were reported [bib_ref] Determination of beta-carotene and lutein available from green leafy vegetables by an..., Serrano [/bib_ref]. Very little is known on carotenoid interaction with the microbiota [bib_ref] Bioactivity of carotenoids -chasms of knowledge, Bohn [/bib_ref]. Unlike polyphenols, which are heavily metabolized, no carotenoid degradation products/bacterial metabolites have been identified. In general, bacteria in the colon are able to deglycosylate, hydrolyse, deglucuronidate, demethylate, and cause ring-fission in some molecules, among other [bib_ref] Mind the gap-deficits in our knowledge of aspects impacting the bioavailability of..., Bohn [/bib_ref] [bib_ref] Bioavailability of dietary polyphenols and gut microbiota metabolism: antimicrobial properties, Marin [/bib_ref]. However, in germ-free rats, higher carotenoid utilization (of ␣and -carotene) as measured by their liver levels, has been reported compared to rats with intact microbiota [bib_ref] The bioavailability of alpha-and beta-carotene is affected by gut microflora in the..., Grolier [/bib_ref]. It was suggested that indirect effects, such as decreased intestinal transit time and an altered bile pool in the absence of bacteria could have played a role, though a reduced level of bacterial breakdown products and more remaining native compounds could have been involved. In support of a potential absorption of carotenoids in the colon, a study in mice found BCO1 to be expressed in many cells including mucosal, glandular cells in the stomach, small intestine, and the colon [bib_ref] Cell type-specific expression of beta-carotene 9',10'-monooxygenase in human tissues, Lindqvist [/bib_ref]. BCO2 is known to be expressed in almost all cell types known to express BCO1. However, BCO2 was not found in the colon, suggesting that only symmetric cleavage of carotenoids may happen in the mucosal cells in the colon.
In a previous study, -carotene uptake into human exfoliated epithelial cells of the colon, separated from feces, has been demonstrated [bib_ref] In vitro uptake of betacarotene by human exfoliated colonic epithelial cells, Gireesh [/bib_ref]. Following the consumption of -carotene rich spirulina, the concentration of -carotene in the cells increased approximately 3-fold, demonstrating colonic cellular presence. However, this may have occurred not necessarily through direct cellular uptake via colonocytes, as carotenoids could have been absorbed via the small intestine and then distributed via the circulatory system to the colonocytes. Furthermore, the same constituents known to enhance carotenoid bioavailability, namely bile salts, emulsifiers such as lecithin [bib_ref] Phospholipids affect the intestinal absorption of carotenoids in mice, Baskaran [/bib_ref] [bib_ref] Lysophosphatidylcholine enhances carotenoid uptake from mixed micelles by Caco-2 human intestinal cells, Sugawara [/bib_ref] , enhanced colonic cellular uptake. Though carotenoids can be taken up by colonic derived Caco-2 cells, direct colonic uptake is not easy to prove, and studies so far have not suggested a strong correlation between dietary intake of carotenoids and colon concentrations [bib_ref] Relationships between serum and colon concentrations of carotenoids and fatty acids in..., Sen [/bib_ref]. Oshima et al. [bib_ref] Absorption and distribution of lycopene in rat colon, Oshima [/bib_ref] investigated colonic absorption and distribution of lycopene in rats with or without a colostomy at mid colon that diverted the fecal stream but without resection of the distal colon. In rats given intragastric treatment, lycopene was found in the mucosa in the proximal colon and in the distal colon, also of the colostomized rats, whose distal colon was isolated from the faecal stream, indicating that lycopene may be transported via the blood into the colon. Moreover, lycopene reached the liver to an appreciable extent even when administered into the isolated distal colon, indicating that absorption is possible from the distal colon in rats.
Taken together, these results indicate that carotenoid absorption from the colon could be relevant and contribute to interindividual variation in carotenoid bioavailability, depending on the food matrix and microbiota. Furthermore, as faecal transplants have shown to be able to trigger obesity, at least in animal models [bib_ref] The new era of treatment for obesity and metabolic disorders: evidence and..., Jayasinghe [/bib_ref] , and obese subjects having generally lower concentrations of circulating carotenoids , a potential direct or indirect link between the microbiota and carotenoid tissue levels may exist. In a study with atherosclerotic subjects, patients showed a metagenome with reduced phytoene-dehydrogenase and lower -carotene serum levels compared to healthy controls, which was associated with a higher level of Collinsella spp. in diseased subjects [bib_ref] Symptomatic atherosclerosis is associated with an altered gut metagenome, Karlsson [/bib_ref] , highlighting the potential role of the microbiota.
## Diseases and medical intervention effecting the intestine and colon
Any condition reducing the intestinal mucosal surface area can be expected to reduce carotenoid absorption. As most studies do not directly measure carotenoid absorption efficiency but rather look at blood carotenoid levels (or a plasma fraction), it is important to distinguish between direct effects on carotenoid absorption (i.e. through reduced mucosal surface area or limited transport capacity) and indirect effects (through dietary adaptations, e.g. high fiber or low fat diet). This is usually achieved by controlling for carotenoid dietary intake. A study with 20 Crohn's disease patients reported lower fasting blood carotenoid concentrations, independent of dietary intake [bib_ref] Fasting plasma carotenoids concentrations in Crohn's and pancreatic cancer patients compared to..., Drai [/bib_ref] , suggesting that malabsorption affected carotenoid uptake, though increased turnover rate and colonic losses via e.g. bleeding could not be excluded. Similar results were obtained by Geerling et al. [bib_ref] Comprehensive nutritional status in patients with long-standing Crohn disease currently in remission, Geerling [/bib_ref] in a study with 32 Crohn's disease patients and Genser et al. [bib_ref] Status of lipidsoluble antioxidants and TRAP in patients with Crohn's disease and..., Genser [/bib_ref] with 24 patients. Crohn's disease usually affects the ileum but only three of the 20 patients in the study had ileal inflammation, indicating the importance of the colonic mucosal integrity for carotenoid absorption. Patients undergoing bariatric surgery (Roux-en-Y gastric bypass and biliopancreatic diversion) also displayed lower blood carotenoid levels [bib_ref] Depletion of serum carotenoid and other fat-soluble vitamin concentrations following obesity surgery, Granado-Lorencio [/bib_ref]. Since fruit and vegetable consumption was apparently normal, the effect was attributed to malabsorption due to reduced mucosal surface area and also due to limited capacity of transport related to decreased lipoprotein concentration. Also reduced gastric digestion (via gastric lipase, or mechanic dispersion), could have played a role, as could have biliopancreatic diversion, affecting bile and pancreatic enzyme concentrations in the gut. In another study, subjects with Celiac disease and Crohn's disease (n = 22) showed significantly 37% decreased levels of macular carotenoids compared to controls (n = 25 [bib_ref] Macular and serum carotenoid concentrations in patients with malabsorption syndromes, Ward [/bib_ref].
Short bowel syndrome, usually due to large resections of the small intestine to treat pathologies such as Crohn's disease or gastrointestinal tumors, have also been associated with carotenoid malabsorption. Edes et al. [bib_ref] Essential fatty acid sufficiency does not preclude fatsoluble-vitamin deficiency in short-bowel syndrome, Edes [/bib_ref] reported undetectable -carotene blood levels following supplementation, despite adequate fat absorption, in a patient with extensive small intestinal resection (serum vitamin A levels appeared normal). Perhaps carotenoid absorption occurred in a more limited section of the intestine, or absorbed -carotene was fully converted to vitamin A. Luo et al. [bib_ref] Prospective analysis of serum carotenoids, vitamin A, and tocopherols in adults with..., Luo [/bib_ref] reported no increase in blood carotenoid levels in subjects with short bowel syndrome undergoing intestinal rehabilitation, despite a 12-week-long supplementation with -carotene, lutein and lycopene. This was attributed to low fat absorption (about 30 versus >95% in healthy subjects) in these patients. However, no estimates of the contribution of the colon to the observed differences in absorption efficiencies were reported. Therefore, it is uncertain if it is the disease affecting the lower gut, the limited length of residual ileum, the presence or absence of the colon, the patient's lifestyle, or a combination that results in low plasma carotenoids. Similar low levels were observed in 63 patients with total gastrectomy [bib_ref] Metabolic and surgical aspects of total gastrectomy (author's transl), Husemann [/bib_ref] , possibly due to duodenal bypass and short interposition of a small intestine loop.
Intestinal parasites and bacterial overgrowth can also damage mucosal cells and result in increased permeability and decreased absorption of nutrients. In Indonesian children receiving red sweet potato, serum retinol concentrations increased to a greater extent when children infected with intestinal helminths were dewormed, than when the intensity of infection was high [bib_ref] Serum retinol concentrations in children are affected by food sources of beta-carotene,..., Jalal [/bib_ref] , though the effect may have been also due to improved fat absorption. In tropical countries, also enteropathies, resulting in inflamed epithelium and reduced surface available for absorption, are likely to contribute to low carotenoid and vitamin A status [bib_ref] Tropical malabsorption, Ramakrishna [/bib_ref].
## Host factors influencing intracellular
transport and transport to the liver
## Intracellular transport within the enterocyte
After their uptake at the apical side of the enterocyte by membrane proteins, which are involved in the uptake of other liposoluble micronutrients, e.g. vitamin E/D [bib_ref] Vitamin D intestinal absorption is not a simple passive diffusion: evidences for..., Reboul [/bib_ref] , carotenoids have to cross the aqueous environment of the cell to reach its basolateral side. As carotenoids are very hydrophobic [bib_ref] Factors affecting intestinal absorption of highly lipophilic food microconstituents (fat-soluble vitamins, carotenoids..., Borel [/bib_ref] it is assumed that they need to be associated with intracellular proteins to move through this medium [bib_ref] Proteins involved in uptake, intracellular transport and basolateral secretion of fat-soluble vitamins..., Reboul [/bib_ref]. Though candidate proteins have been suggested, limited evidence of their involvement is available yet. A first one is human retinal lutein-binding protein [bib_ref] Purification and partial characterization of a lutein-binding protein from human retina, Bhosale [/bib_ref] , as it shows a good cross-reactivity with antibodies raised against carotenoid-binding protein, which has been shown to transport carotenoids in the midgut cytosol of the silkworm Bombyx mori [bib_ref] A CD36-related transmembrane protein is coordinated with an intracellular lipid-binding protein in..., Sakudoh [/bib_ref]. However, its expression in the enterocyte should be verified. Other candidates could be the enterocyte FABPs (FABP2/I-FABP and FABP1/L-FABP) that allow the transport of various lipids. Finally, it can be hypothesized that the main enzyme responsible for carotenoid cleavage in the enterocyte, i.e. BCO1 [bib_ref] Two carotenoid oxygenases contribute to mammalian provitamin A metabolism, Amengual [/bib_ref] [bib_ref] Mammalian carotenoid-oxygenases: key players for carotenoid function and homeostasis, Lobo [/bib_ref] , could also be involved, as it attracts and binds carotenoids for further cleavage, and it may also function as a non-identified but predicted selective carotenoid-transporter. The involvement of some of these candidate proteins in carotenoid transport within the enterocyte is supported by studies that have observed associations between SNPs in genes encoding these proteins and carotenoid status or bioavailability. This is the case for FABP and lycopene [bib_ref] Human plasma levels of vitamin E and carotenoids are associated with genetic..., Borel [/bib_ref] and BCO1 and -carotene [bib_ref] A combination of single-nucleotide polymorphisms is associated with interindividual variability in dietary..., Borel [/bib_ref] , though this second association can also be due to the catalytic activity of this protein. Functional studies employing cell cultures or transgenic mice should be performed to identify the respective proteins. Nevertheless, it can be hypothesized that variations in genes encoding proteins involved in the transport of carotenoids within the enterocyte contribute to the observed interindividual variability in carotenoid bioavailability.
The previously described interaction of lutein and carotene was not investigated further in detail, but it was predicted also to be of relevance regarding mutual interferences during absorption [bib_ref] Carotenoid interactions, Van Den Berg [/bib_ref] [bib_ref] Effect of lutein on beta-carotene absorption and cleavage, Van Den Berg [/bib_ref]. A different fractional absorption efficacy was also suggested for cis-isomers of lycopene [bib_ref] Bioavailabilty of non-provitamin A carotenoids, Bohn [/bib_ref] [bib_ref] Lycopene from heat-induced cis-isomer-rich tomato sauce is more bioavailable than from all-trans-rich..., Unlu [/bib_ref] [bib_ref] Carotenoid absorption in humans consuming tomato sauces obtained from tangerine or high-beta-carotene..., Unlu [/bib_ref] [bib_ref] In vitro micellarization and intestinal cell uptake of cis isomers of lycopene..., Failla [/bib_ref]. Unfortunately, for lutein and -carotene as well as for lycopene and -carotene cis-isomers, the mechanism of this altered transport efficiency was not examined further, but it appears to have an important physiological importance due to the different and possibly augmented health beneficial effects of especially 9-cis--carotene versus all-trans--carotene, at least in respect to atherosclerosis [bib_ref] 9-cis beta-carotene-rich powder of the alga Dunaliella bardawil increases plasma HDL-cholesterol in..., Shaish [/bib_ref].
## Secretion at the basolateral and apical side of the enterocyte
During the postprandial period following the intake of a meal providing carotenoids, the latter are recovered in chylomicrons and their remnants, circulating in the blood [bib_ref] Interindividual variability of lutein bioavailability in healthy men: characterization, genetic variants involved,..., Borel [/bib_ref] [bib_ref] A combination of single-nucleotide polymorphisms is associated with interindividual variability in dietary..., Borel [/bib_ref] [bib_ref] Lycopene bioavailability is associated with a combination of genetic variants, Borel [/bib_ref]. This allows physiologists to conclude that carotenoids are incorporated into chylomicrons within the enterocyte, then secreted into the lymph, and finally transported to the blood. Two observations support this paradigm. First, studies on Caco-2 cell monolayers, an acknowledged model of the human intestinal epithelium, have shown that carotenoids added to the apical side of these cells are recovered in the lipoprotein chylomicron-rich fraction secreted at the basolateral side [bib_ref] Carotenoid uptake and secretion by CaCo-2 cells: betacarotene isomer selectivity and carotenoid..., During [/bib_ref] [bib_ref] Mechanisms of provitamin A (carotenoid) and vitamin A (retinol) transport into and..., During [/bib_ref]. Second, clinical studies have shown associations between SNPs in MTP, involved in chylomicron formation within the enterocyte, and APOB (the main chylomicron apoprotein), and carotenoid bioavailability [bib_ref] Interindividual variability of lutein bioavailability in healthy men: characterization, genetic variants involved,..., Borel [/bib_ref] [bib_ref] A combination of single-nucleotide polymorphisms is associated with interindividual variability in dietary..., Borel [/bib_ref] [bib_ref] Lycopene bioavailability is associated with a combination of genetic variants, Borel [/bib_ref]. Secretion via chylomicrons implies that polymorphisms of genes involved in chylomicron formation, such as those involved in cholesterol biosynthesis, may potentially have a role in explaining inter-individual variation in carotenoid uptake or processing, as has been suggested for patients with hypercholesterolemia [bib_ref] Uptake of chylomicron remnant retinyl esters in human leukocytes in vivo, Skrede [/bib_ref].
Although it is acknowledged that a significant fraction of newly absorbed carotenoids is secreted by the enterocyte via chylomicrons, it should be noted that another fraction is metabolized within the intestinal cell. The size of this fraction depends on several factors such as the carotenoid species and the vitamin A status, affecting provitamin A carotenoid absorption and cleavage [bib_ref] Genetics and diet regulate vitamin A production via the homeobox transcription factor..., Lobo [/bib_ref]. As stated above, BCO1 and BCO2 are responsible for this mechanism. Their action results in several carotenoid metabolites, e.g. retinal, apo-carotenals etc. [bib_ref] Naturally occurring eccentric cleavage products of provitamin A beta-carotene function as antagonists..., Eroglu [/bib_ref] , which may not share a fate similar to that of the parent molecules, and thus are not necessarily incorporated into chylomicrons. As at least some of these metabolites are water soluble (logP-values around 5, such as for retinoic acid -4.4, http://www.drugbank.ca/drugs/DB00982), it can be hypothesized that they may be secreted to the portal vein and then reach the liver.
Another pathway involved in carotenoid secretion at the basolateral side of the enterocyte may be via APOA1. This involves the membrane protein ABCA1, responsible for the lipid transfer from this membrane to APOA1/HDL in the lymph. Though it was shown that ABCA1 is not involved in the efflux of carotenoids to HDL at the basolateral side of Caco-2 cells [bib_ref] Mechanisms of provitamin A (carotenoid) and vitamin A (retinol) transport into and..., During [/bib_ref] , a recent study demonstrated that a fraction of carotenoids, at least the xanthophylls, is transferred via ABCA1 to APOA1, not directly to HDL [bib_ref] Effect of compounds affecting ABCA1 expression and CETP activity on the HDL..., Niesor [/bib_ref].
Thus, the complex mechanisms that are involved in the secretion of carotenoids, and of their metabolites at the basolateral side of the enterocyte involve several genes and are likely to be modulated by genetic variations affecting the expression or activity of the proteins encoded by these genes. It was thus hypothesized that SNPs in these genes correlate with interindividual variability of carotenoid bioavailability. This hypothesis was supported by results of three recent human clinical studies. These have shown that SNPs in MTP and in APOB, involved in the APOB dependent pathway, as well as SNPs in ABCA1, involved in the APOA1 dependent pathway, are associated with lutein [bib_ref] Interindividual variability of lutein bioavailability in healthy men: characterization, genetic variants involved,..., Borel [/bib_ref] , lycopene [bib_ref] Lycopene bioavailability is associated with a combination of genetic variants, Borel [/bib_ref] , and -carotene [bib_ref] A combination of single-nucleotide polymorphisms is associated with interindividual variability in dietary..., Borel [/bib_ref] bioavailability. SNPs in APOB were associated with -carotene concentrations while SNPs in apolipoprotein A4 (APOA4) and APOB were associated with lycopene concentrations [bib_ref] Human plasma levels of vitamin E and carotenoids are associated with genetic..., Borel [/bib_ref]. These SNPs explained differences in e.g. -carotene plasma concentrations by up to 50%.
Finally, carotenoids may also be re-excreted via the apical side into the gut lumen. Results from a human intervention trial (with tomato puree) suggested that the ABCB1 gene plays a key role in lycopene transport, possibly by effluxing a fraction of the absorbed lycopene back into the intestinal lumen [bib_ref] Lycopene bioavailability is associated with a combination of genetic variants, Borel [/bib_ref]. This hypothesis needs to be examined further.
## Postprandial chylomicron transport and blood plasma appearance
It is believed that most newly-absorbed carotenoids are postprandially secreted in chylomicrons, and that the role of chylomicrons, among other, is to carry carotenoids and their lipophilic metabolites from the intestine to the liver. During their transport, chylomicron triglycerides undergo hydrolysis by LPL, resulting in the generation of smaller chylomicrons termed chylomicron remnants. After their uptake by the liver, a fraction of carotenoids appears to be stored in the liver, another one is metabolized (e.g. into vitamin A for the provitamin A carotenoids). The remaining fraction is re-secreted into the blood within VLDL. VLDL, via their metabolism into LDL, are thought to be responsible for the further tissue distribution of carotenoids. Due to their hydrophobicity, it is thought that carotenoids stay located within the core of the chylomicron(remnant)s during their transport in blood [bib_ref] Carotenoids in biological emulsions: solubility, surface-tocore distribution, and release from lipid droplets, Borel [/bib_ref]. Thus, it is hypothesized that chylomicron carotenoids i) are not significantly transferred to other circulating lipoproteins (VLDL, LDL, HDL), and ii) they are not significantly transferred to tissues. However, an in vitro study has suggested that this assumption needs to be revisited because an exchange of carotenoids between VLDL and HDL was found [bib_ref] Carotenoids, mostly the xanthophylls, exchange between plasma lipoproteins, Tyssandier [/bib_ref].
Although it is possible that some chylomicron carotenoids can be transferred to other lipoprotein classes or to tissues during lipoprotein metabolism, it is assumed that this transfer is rather limited. Thus, the postprandial blood metabolism of carotenoids embedded in chylomicrons is closely related to lipoprotein metabolism. The metabolism of chylomicrons involves several proteins, starting with the apolipoproteins that are associated with these lipoparticles during their synthesis, i.e. APOB48 and APOA1, followed by the apoproteins that are transferred from other lipoprotein classes during chylomicron blood transport, e.g. APOE, and ending with enzymes that transfer or hydrolyse chylomicron lipids, e.g. cholesterol ester transfer protein (CETP) and [bib_ref] Carotenoids, mostly the xanthophylls, exchange between plasma lipoproteins, Tyssandier [/bib_ref].
Any variability of affinity of the above-described transporters/proteins involved in chylomicron metabolism would alter carotenoid kinetics. However, only few human studies have examined these, including studies on lutein and -carotene [bib_ref] Intestinal absorption, serum clearance, and interactions between lutein and betacarotene when administered..., Kostic [/bib_ref] , lycopene [bib_ref] A physiological pharmacokinetic model describing the disposition of lycopene in healthy men, Diwadkar-Navsariwala [/bib_ref] , and also retinyl esters [bib_ref] Uptake of chylomicron remnant retinyl esters in human leukocytes in vivo, Skrede [/bib_ref]. In the latter study, a 7-compartment model demonstrated a saturable absorption process, in support of the uptake mostly via transporters. Variability of absorption was similar over the range of dosing (10-120 mg), with a relative standard deviation (RSD) of ca. 50%. In an intervention study by Kostic et al. [bib_ref] Intestinal absorption, serum clearance, and interactions between lutein and betacarotene when administered..., Kostic [/bib_ref] , adult subjects were given single equimolar doses (0.5 mol/kg body weight) of lutein and/or -carotene solubilized in oil. Absorption had an RSD of 43 and 68%, respectively. A single peak of mean serum lutein concentration at 16 h was found, while for -carotene a small initial peak appeared at 6 h, and a second peak at around 32 h. The first peak was assumed to be chylomicron-borne, the second peak was believed to represent newly absorbed -carotene from the liver circulating as VLDL/HDL [bib_ref] Distribution of orally administered beta-carotene among lipoproteins in healthy men, Johnson [/bib_ref] , whereas the intermediate peak for lutein was unexplained. This suggests different mechanisms for the distribution of the two carotenoids, leading to a different time-course of serum peaks, in line with an altered transfer between lipoproteins compared to carotenes.
It is unclear whether any differences in serum carotenoids described in the literature are related to any of the above proteins involved in uptake, transport and chylomicron metabolism, but it can be hypothesized. A variety of apolipoprotein polymorphisms were studied regarding concentrations of several carotenoids in children (n = 447), in a sample of the Stanislas Study. Lower concentrations of lutein/zeaxanthin (19%), -cryptoxanthin (51%), ␣-carotene (55%) and -carotene (47%) were found in children expressing the S447X allele versus the S447S allele of the LPL gene [bib_ref] The lipoprotein lipase serine 447 stop polymorphism is associated with altered serum..., Herbeth [/bib_ref] , though no other correlations were found. In another study [bib_ref] Human fasting plasma concentrations of vitamin E and carotenoids, and their association..., Borel [/bib_ref] , human fasting concentrations of ␣and -carotene were associated with genetic variants in FABP and LIPC, while ␣and ␥tocopherol were influenced also by APOC3 (a component of LDL), CETP, and MTP (required for lipoprotein assembly), indicating that these may be involved also in carotenoid metabolism. In an earlier study, serum concentrations of carotenoids [fig_ref] Table 4: List of SNPs known, or speculated, to influence carotenoid metabolism C 2017... [/fig_ref] were associated with SNPs in APOB and APOA4 [bib_ref] Human plasma levels of vitamin E and carotenoids are associated with genetic..., Borel [/bib_ref].
In addition to these proteins, other factors may play a role . Brady et al. investigated the association between serum carotenoids and physiological and life-style factors. Lower serum levels of several carotenes and xanthophylls were associated with being male (perhaps related to lower fruit/vegetable intake), smoker, of younger age, having lower non-HDL cholesterol, higher alcohol consumption and higher body mass; only serum lycopene was not associ-ated with these factors but with age [bib_ref] Human serum carotenoid concentrations are related to physiologic and lifestyle factors, Brady [/bib_ref]. Age also showed to be significantly associated with chylomicron response of lycopene [bib_ref] Comparison of the postprandial chylomicron carotenoid responses in young and older subjects, Cardinault [/bib_ref] , but not with other carotenoids. However, the underlying mechanisms of these associations are unclear. It can be speculated that all factors are related to dietary pattern, though a higher body mass and a higher amount of adipose tissue may result in increased carotenoid storage in adipocytes, while smoking may increase the turnover of carotenoids due to enhanced oxidative stress , [bib_ref] Associations of plasma carotenoids with risk factors and biomarkers related to cardiovascular..., Wang [/bib_ref]. Similarly, in the SU.VI.MAX study (n>12 000 participants), it was found that -carotene plasma levels correlated (negatively) with smoking status, blood triglycerides, alcohol consumption and age. Again, females had higher -carotene serum levels than men . Menstrual cycle also showed to influence plasma carotenoids. In an intervention trial with nine women consuming standardized diets for two cycles, carotenoid plasma concentration was usually lower in the earlier follicular phase compared to the late follicular phase and in part higher than in the luteal phase, possibly due to hormonal influences on the blood concentration of lipoproteins as carotenoid carriers [bib_ref] Effect of menstrual cycle phase on the concentration of individual carotenoids in..., Forman [/bib_ref]. In a larger study, higher serum retinol levels were associated with higher serum estradiol and testosterone levels during the menstrual cycle [bib_ref] Serum antioxidants are associated with serum reproductive hormones and ovulation among healthy..., Mumford [/bib_ref].
## Further transport and biodistribution to potential target tissues
# Introduction
Carotenoids are transported in the blood stream associated with lipoproteins, where carotenes dominate carotenoid pattern in the LDL fraction and xanthophylls are almost equally distributed between LDL and HDL [bib_ref] Plasma LDL and HDL subspecies are heterogenous in particle content of tocopherols..., Goulinet [/bib_ref]. Especially the potential exchange of xanthophylls between lipoproteins is important in this context and may depend on the activity of CETP and LCAT [bib_ref] Carotenoids, mostly the xanthophylls, exchange between plasma lipoproteins, Tyssandier [/bib_ref]. Consequently, changes of the lipoprotein pattern, due to external or host-related factors, may modulate tissue distribution of carotenoids. At the site of the target tissue, selective uptake systems may be operative to accumulate particular carotenoids, which are further transported to specific cells of the tissue; or within a cell, directed to subcellular compartments. Uptake might be hindered by tissue barriers (e.g. the blood-brain barrier), permeable only for certain compounds, though the lipophilic carotenoids would be expected to pass. Also, due to their lipophilicity, their volume of distribution (V D ) in the body is quite large [bib_ref] Pharmacokinetics and tissue distribution of orally administered lycopene in male dogs, Korytko [/bib_ref] , and plasma concentrations will only to some extent reflect tissue levels. Thus, plasma concentrations are expected to be influenced if the V D is altered, which may explain lower circulating carotenoid levels in obese subjects [bib_ref] Isotopic labeling and LC-APCI-MS quantification for investigating absorption of carotenoids and phylloquinone..., Kurilich [/bib_ref] [bib_ref] Isotopic tracer techniques for studying the bioavailability and bioefficacy of dietary carotenoids,..., Van [/bib_ref]. Unfortunately, only little is known about host related factors such as genetic makeup (e.g. SNPs) or other individual determinants and their impact on carotenoid tissue distribution.
## Liver
Data from animal and human studies provide evidence that the hepatic tissue is a major site of carotenoid accumulation and metabolism [bib_ref] Carotenoid composition, concentrations, and relationships in various human organs, Kaplan [/bib_ref] [bib_ref] Concentrations of selected carotenoids and vitamin A in human liver, kidney and..., Schmitz [/bib_ref] [bib_ref] Retinyl ester (vitamin A ester) and carotenoid composition in human liver, Tanumihardjo [/bib_ref] [bib_ref] cis-trans isomers of lycopene and beta-carotene in human serum and tissues, Stahl [/bib_ref]. -Carotene and lycopene e.g. are found in the nmol range per gram wet tissue, however, individual values widely vary . Carotenoids travel with lipoproteins and the liver is a central hub for lipoprotein assembling and release. Hepatic endocytosis of chylomicron remnants, which contain newly absorbed carotenoids, depends on the interaction of the APOE protein with the membrane receptor LRP1 or to some extent with the LDLreceptor. Also involved in the uptake mechanism is hepatic LPL. Carotenoids remain in the liver for storage, alternatively, they are secreted with VLDLs, which are further processed to LDLs. Mechanisms involved in the coordination of storage, cleavage and secretion are however not known yet. It is likely that the regulation of the specific cleaving enzymes plays a major role in carotenoid plasma levels.
As noted in earlier publications, tissues such as the liver (or testes and adrenals) which possess a large number of LDL receptors, exhibit high levels of carotenoids. On the other hand, lipids from circulating HDL are taken up by this organ, too. The central role of the liver in lipid metabolism makes it likely that individual differences (polymorphisms) in proteins affecting this process can influence carotenoid distribution. SNPs in LDL receptors may play a role, as they are critical for the endocytosis of the remnant chylomicron particle into liver hepatocytes, influenced by the binding of APOE to the surface of chylomicrons, and may therefore be suspected to play a role in carotenoid distribution. Though carotenoid levels in plasma have been associated with genetic polymorphisms in genes related to lipid metabolism [bib_ref] Human plasma levels of vitamin E and carotenoids are associated with genetic..., Borel [/bib_ref] , an impact on tissue distribution or uptake has not been proven so far in humans. By contrast, mice expressing APOE4 as compared to APOE3 had lower levels of -carotene in the bloodstream and lower levels of -carotene and lutein in adipose tissue [bib_ref] Dietary betacarotene and lutein metabolism is modulated by the APOE genotype, Huebbe [/bib_ref] , while hepatic expression of BCO1/2 was significantly higher, suggesting a correlation of both factors.
In addition to the conversion of provitamin A carotenoids into retinal by BCO1, the liver is also a central tissue for xenobiotic metabolism, mediated by an array of phase I/II enzymes. Especially the metabolism of carotenoids by cytochrome P450-dependent monooxygenases has been topic of research, and an active hepatic P450 dependent metabolism was shown for several carotenoids [bib_ref] A 3-hydroxy beta-end group in xanthophylls is preferentially oxidized to a 3-oxo..., Nagao [/bib_ref] [bib_ref] Identification, quantification, and relative concentrations of carotenoids and their metabolites in human..., Khachik [/bib_ref]. Several metabolites have been detected [bib_ref] Metabolism and CYP-inducer properties of astaxanthin in man and primary human hepatocytes, Kistler [/bib_ref] and thus a great deal of P450 related genes expressed in the liver including CYPs 3A4, 2C9, 2C8, 2E1, and 1A2, and to a lesser extent 2A6, 2D6, 2B6, 2C19, plus the extrahepatically expressed CYPs 2J2, 1A1, and 1B1, are expected to potentially influence carotenoid tissue levels [bib_ref] Cytochrome P450 enzymes in drug metabolism: regulation of gene expression, enzyme activities,..., Zanger [/bib_ref]. Phase I and II enzymes are usually inducible and respond to internal and external challenges (stress) a host is exposed to. Thus, in addition to genetic factors, also external factors can influence the extent of metabolism and metabolic pattern. A number of different polymorphisms involved in phase I/II enzymes in humans have already been revealed. They individually affect the metabolism of drugs or endogenous compounds. Little is known with respect to carotenoids, but it was shown that a polymorphism regarding CYP26B1 (rs2241057) influences the degradation of retinoic acid, and is likely related to the risk of Crohn's disease [bib_ref] Polymorphism in the retinoic acid metabolizing enzyme CYP26B1 and the development of..., Fransen [/bib_ref] and atherosclerosis [bib_ref] A CYP26B1 polymorphism enhances retinoic acid catabolism and may aggravate atherosclerosis, Krivospitskaya [/bib_ref].
## Adipose tissue
Adipose tissue and in particular the lipid fraction of adipocytes is an important site of carotenoid accumulation [bib_ref] Carotenoid composition, concentrations, and relationships in various human organs, Kaplan [/bib_ref] [bib_ref] Carotenoids in human blood and tissues, Parker [/bib_ref] [bib_ref] Purified low-density lipoprotein and bovine serum albumin efficiency to internalise lycopene into..., Gouranton [/bib_ref]. Vitamin A is stored in adipose tissue primarily as unesterified retinol [bib_ref] Retinoids and retinoid-binding protein expression in rat adipocytes, Tsutsumi [/bib_ref] [bib_ref] Reduced adiponectin expression after high-fat diet is associated with selective up-regulation of..., Landrier [/bib_ref]. Concentrations of carotenoids in adipose tissue have been reported by several groups [bib_ref] Individual carotenoid concentrations in adipose tissue and plasma as biomarkers of dietary..., El-Sohemy [/bib_ref] [bib_ref] Site-specific concentrations of carotenoids in adipose tissue: relations with dietary and serum..., Chung [/bib_ref] [bib_ref] Adipose tissue as a medium for epidemiologic exposure assessment, Kohlmeier [/bib_ref]. According to these studies, concentrations of -carotene, -cryptoxanthin, lycopene and lutein/zeaxanthin were comparable in variation and concentration to their plasma concentration. Although concentrations of carotenoids per g tissue are higher in some other organs, adipose tissue contains the highest total amounts, and is assumed to contribute to carotenoid storage. Unfortunately, knowledge is scarce on the mechanisms involved in the regulation of carotenoid uptake/release in this tissue.
LPL, expressed in adipose-and other tissues, is the primary enzyme responsible for triacylglycerol lipolysis, provided by chylomicron-and VLDL transport vehicles for carotenoids, and implicated in fatty acid uptake. Thus, it may also play a role in adipocyte carotenoid uptake. Cell culture studies suggest that CD36 is involved in the uptake of lycopene and lutein by adipocytes [bib_ref] CD36 is involved in lycopene and lutein uptake by adipocytes and adipose..., Moussa [/bib_ref]. Hormone sensitive lipases are implicated in the release of retinol from storage tissue due to cleavage of retinyl esters, and may aid in releasing retinol by hydrolyzing triglycerides of the intracellular fat droplets [bib_ref] Vitamin A metabolism: an update, D'ambrosio [/bib_ref]. However, their impact on carotenoid release from adipose compartments is unclear. Retinol binding protein 4 (RBP4) is synthesized by adipocytes as a signaling molecule. It was shown to coordinate bidirectional retinol uptake in adipose tissue together with its membrane receptor STRA6 (stimulated by retinoic acid gene 6) [bib_ref] Retinol-binding protein 4 and its membrane receptor STRA6 control adipogenesis by regulating..., Muenzner [/bib_ref]. Retinol-loaded holo-RBP4 blocked adipocyte differentiation by activating RAR␣, while retinol-free apo-RBP4 triggered retinol efflux, resulting in reduced cellular retinoids and RAR␣ mediated transcription and enhanced adipogenesis [bib_ref] Retinol-binding protein 4 and its membrane receptor STRA6 control adipogenesis by regulating..., Muenzner [/bib_ref].
Several host related factors have been identified in EU-RAMIC (European multicenter case-control study on antioxidants, myocardial infarction and breast cancer), scrutinizing correlations of carotenoid levels in adipose tissue [bib_ref] Predictors of adipose tissue carotenoid and retinol levels in nine countries. The..., Virtanen [/bib_ref]. In another trial, obesity was associated with carotenoid levels in adipose tissue, though it is not quite clear whether this could also be due to decreased dietary intake. A correlation was also observed between higher alcohol intake and lower levels of -carotene and lycopene in adipose tissue of men and women, respectively. Although the concentration of -carotene in the adipocytes of obese subjects was lower compared to non-obese, the total amount of -carotene in all lipid stores was similar [bib_ref] The concentration of -carotene in human adipocytes, but not the wholebody adipocyte..., Östh [/bib_ref]. Whether this implies a regulation of total lipid body stores is still controversial. Regarding alcohol consumption, lower intake, altered liver-metabolism, decreased small intestinal uptake, or enhanced consumption due to oxidative stress may play a role. In this context, the activity of -carotene metabolizing enzymes is likely important and genetic differences but also alcohol intake is expected to affect a balanced distribution in adipose tissues.
Mice in which BCO1 is deleted and receiving marginal vitamin A sufficient diet with -carotene, accumulate -carotene in adipose tissue [bib_ref] Beta-carotene reduces body adiposity of mice via BCMO1, Amengual [/bib_ref]. This illustrates the role of adipose tissue as a storage tissue for lipophilic compounds such as carotenoids. Recently, in yellow rabbits, a triplet deletion was identified in the BCO2 gene, resulting in the absence of an asparagine in BCO2. This was suggested to cause accumulation of carotenoids in adipose tissue [bib_ref] A novel AAT-deletion mutation in the coding sequence of the BCO2 gene..., Strychalski [/bib_ref]. This agrees with an earlier finding in sheep, where a BCO2 mutation was found to be tightly associated with white adipose tissue carotenoid accumulation [bib_ref] A nonsense mutation in the betacarotene oxygenase 2 (BCO2) gene is tightly..., Vage [/bib_ref] , and in bovines where BCO2 mutations were associated with carotenoid accumulation in adipose tissue and milk [bib_ref] Mutation in bovine beta-carotene oxygenase 2 affects milk color, Berry [/bib_ref] [bib_ref] Genetic variation in the beta, beta-carotene-9', 10'-dioxygenase gene and association with fat..., Tian [/bib_ref]. These observations were lately confirmed by BCO2 inactivation in sheep using CRISPR/Cas9 technology, resulting in yellow fat, establishing a causal relationship between BCO2 activity and carotenoid accumulation in white adipose tissue [bib_ref] Biallelic beta-carotene oxygenase 2 knockout results in yellow fat in sheep via..., Niu [/bib_ref]. Together, these findings not only illustrate the importance of white adipose tissue as a carotenoid storage organ, but also show whole body physiological regulation of carotenoid homeostasis, posing another layer of complexity on understanding inter-individual variation (see also . This is exemplified by sex specific responses resulting from -carotene accumulation in white adipose tissue, which showed that 4970 genes were affected in WT female mice, while only 407 were affected in male mice [bib_ref] Gene expression response of mouse lung, liver and white adipose tissue to..., Van Helden [/bib_ref] , with the majority of the commonly affected genes [fig_ref] Figure 1: Predominant carotenoids in our diet, common metabolites and nomenclature [/fig_ref] showing a strong negative, rather than positive, correlation of expression between males and females. This negative correlation was also seen in BCO1 knockout mice, although the number of genes affected were more similar between the two sexes (1522 gene in females and 1202 in males, 33 overlapping) [bib_ref] Organ specificity of beta-carotene induced lung gene-expression changes in Bcmo1−/− mice, Van Helden [/bib_ref]. In both WT and BCO1 knockout mice, only a minority of genes is commonly affected by -carotene in females and males. Strikingly, the opposite regulation of genes in response to -carotene exposure was also prominent in the lung of BCO1 knockout mice, but this was not seen in WT mice in this tissue [bib_ref] Beta-carotene affects gene expression in lungs of male and female Bcmo1 (−/−)..., Van Helden [/bib_ref]. On the other hand, WT liver showed a strong positive correlation of -carotene responsive genes between males and females [bib_ref] Organ specificity of beta-carotene induced lung gene-expression changes in Bcmo1−/− mice, Van Helden [/bib_ref].
Adipose tissue is distributed over various depots in the body. Functional differences between depots exist and visceral adipose tissue is especially associated with adverse health effects. In a study aiming to identify differences between adipose tissue depots, it was found that many of the genes differentially expressed in subcutaneous and visceral adipose tissue were regulated by retinoic acid [bib_ref] Intra-and interindividual variation in gene expression in human adipose tissue, Van Beek [/bib_ref]. The master regulator of adipogenesis, PPARG is functionally active as an obligatory dimer with RXRA, linking adipogenesis with vitamin A metabolism. Retinoic acids are generated from retinaldehyde in adipose tissue by aldehyde dehydrogenase 1 (ALDH1), though this is discussed controversially [bib_ref] Reduced adiponectin expression after high-fat diet is associated with selective up-regulation of..., Landrier [/bib_ref]. Female mice with inactivated ALDH1A1 were resistant to high-fat diet-induced visceral adipose formation. This was not seen in male mice, while subcutaneous adipose tissue was reduced to the same extent in males and females [bib_ref] Autocrine function of aldehyde dehydrogenase 1 as a determinant of diet-and sex-specific..., Yasmeen [/bib_ref]. Together, this underlines a role for vitamin A metabolism in differential adipose tissue (i.e. visceral versus. subcutaneous) formation. It has been suggested that estrogen mediated suppression of ALDH1A2/3 mRNA expression is involved in differential retinoic acid formation between males and females [bib_ref] Aldehyde dehydrogenase 1A1: friend or foe to female metabolism?, Petrosino [/bib_ref]. Although major gaps in our knowledge exist, genetic variation in uptake, storage and processing in adipose tissue of various carotenoids may influence adipose tissue distribution and functionality and associated health outcomes, and, if so, will likely do this in a sex dependent manner. Effects of carotenoids on adipose tissue biology have been reviewed recently, while effects of genetic ablation of genes encoding various retinoid metabolism enzymes, including BCO1 and ALDH1A1, but also RBP1, RBP3 and retinol saturase (RESTAT) on adiposity in mice are reviewed elsewhere [bib_ref] Intra-and interindividual variation in gene expression in human adipose tissue, Van Beek [/bib_ref].
## Skin
The carotenoid pattern in human skin comprises carotenes and xanthophylls. Plasma levels of lycopene and less notably -carotene are correlated with their respective concentration in the skin [bib_ref] Significant correlations of dermal total carotenoids and dermal lycopene with their respective..., Scarmo [/bib_ref]. However, no such correlation was observed for lutein, zeaxanthin, and -cryptoxanthin, and thus skin measurements may not be representative of total carotenoid exposure or status. Carotenoids are not equally distributed in the different skin areas. Highest levels occur in skin of the forehead and in the palms of the hands and lower levels in dorsal skin, inside of the arm or back of hand. LDL-receptors are expressed in human skin [bib_ref] Low density lipoprotein receptor-binding activity in human tissues: quantitative importance of hepatic..., Rudling [/bib_ref] and may play a role regarding selective uptake. Skin can be divided into epidermis and dermis with underlying subcutaneous adipose tissue. Blood vessels reach the dermis but not the epidermis, and different ways of how carotenoids may be transported and distributed to and within the different layers of our skin have been discussed [bib_ref] Skin bioavailability of dietary vitamin E, carotenoids, polyphenols, vitamin C, zinc and..., Richelle [/bib_ref]. Subcutaneous tissue is a storage compartment for carotenoids and part of the balanced distribution system of carotenoids in adipose depots. Thus, host factors already mentioned above are expected to also play a role in carotenoid uptake and storage in the skin. There is some evidence that other host-related factors influence carotenoid skin levels [bib_ref] One-year study on the variation of carotenoid antioxidant substances in living human..., Darvin [/bib_ref]. On the long term, a carotenoid-rich diet may increase carotenoid skin levels. However, smoking and alcohol intake induced a rapid decrease in carotenoid levels of the skin. Skin lycopene levels are sensitive to UV-irradiation [bib_ref] Skin lycopene is destroyed preferentially over betacarotene during ultraviolet irradiation in humans, Ribaya-Mercado [/bib_ref]. Upon irradiation in vivo, lycopene concentration in the skin is significantly lowered. This effect is less pronounced with -carotene. Therefore, individual preferences regarding sun or UV exposure (tanning beds) would affect skin carotenoid concentrations.
## Macula lutea
The macula lutea is a small yellow area of the retina. It is the region of maximum visual acuity and its yellow color is due to xanthophylls, mainly lutein, zeaxanthin and mesozeaxanathin, located in the cone axons of the Henle fiber layer. It was shown that macular pigment density was positively correlated with serum concentrations of lutein and zeaxanthin, and inversely correlated with serum oxidized lowdensity lipoprotein [bib_ref] Association of macular pigment optical density with serum concentration of oxidized low-density..., Nagai [/bib_ref]. Consequently, host related oxidative stress conditions would impact the supply of the retina with oxo-carotenoids. Absolute levels and the patterns of lutein and zeaxanthin differ within an individual retina sample [bib_ref] Analysis of zeaxanthin distribution within individual human retinas, Landrum [/bib_ref]. Large interindividual differences have also been described [bib_ref] HPLC measurement of ocular carotenoid levels in human donor eyes in the..., Bhosale [/bib_ref]. In the center of the macula lutea, levels of lutein and zeaxanthin have been reported, with 2.4 and 3.4 pmol/mm², respectively. Much lower concentrations are found in peripheral areas; medial 0.22 pmol lutein/mm² and 0.14 pmol zeaxanthin/mm². Levels further decrease in outer circles while the ratio of lutein:zeaxanthin increases. Since carotenes are not present in the macula lutea, selective mechanisms of uptake must be operative.
It is likely that host related factors (individual differences) have an impact on the density of the macula pigment. In a 6 months intervention study with a lutein-rich supplement the impact of genetic variances in four genes (ABCG8, BCO1, CD36, and NPC1L1) on lutein plasma levels and macular pigment optical density was evaluated [bib_ref] Genetic variants in BCMO1 and CD36 are associated with plasma lutein concentrations..., Borel [/bib_ref]. The results provide evidence that (as with plasma) retina levels, i.e. macula pigment optical density of lutein are affected by SNPs of CD36 and BCO1. The TT variant at the BCO1 rs7501331 locus was associated with a higher macula pigment optical density. Study subjects with GG at the CD36 locus rs1761667 had a higher macula pigment optical density compared to those with an A allele, although the underlying mechanisms remain to be established. Compounds delivered to the retina must pass the blood-retinal barrier, provided by tight junctions between endothelial cells. This barrier is sensitive to inflammatory and oxidative damage often associated with hyperglycemia [bib_ref] Hyperglycemia as a Risk Factor of Ischemic Stroke, Zhang [/bib_ref]. HDLs as the transport vehicles of lutein and zeaxanthin have been discussed to play an important role in the transport of macular carotenoids [bib_ref] Xanthophylls are preferentially taken up compared with beta-carotene by retinal cells via..., During [/bib_ref]. There is evidence that SR-BI, a tissue receptor for HDL, plays a role in the delivery of carotenoids to this tissue [bib_ref] Dietary sources of lutein and zeaxanthin carotenoids and their role in eye..., Abdel-Aal [/bib_ref].
It was further suggested that the delivery of macular carotenoids involves the inter-photoreceptor-retinoid [bib_ref] The interphotoreceptor retinoid binding (IRBP) is essential for normal retinoid processing in..., Parker [/bib_ref]. GSTP1 has been identified as the macular binding protein for zeaxanthin/meso-zeaxanthin in humans [bib_ref] Surface plasmon resonance (SPR) studies on the interactions of carotenoids and their..., Vachali [/bib_ref]. It was shown that StARD3 acts as the lutein binding protein in the human macula [bib_ref] Identification of StARD3 as a lutein-binding protein in the macula of the..., Li [/bib_ref]. In-vitro studies have proven the selectivity of both proteins with respect to either zeaxanthin or lutein. Equilibrium dissociation constants (KD values) for the complex of GSTP1 with zeaxanthin/meso-zeaxanthin are in the range of 0.14 -0.19 M while for lutein/-carotene they are >6-fold higher. Contrarily, StARD3 exhibited a high affinity to lutein (KD ca. 0.59 M), compared to zeaxanthin/mesozeaxanthin (KD-values ca. 1.6 M).
An alternative mechanism contributing to the selective enrichment of the macula carotenoids in primates has been suggested [bib_ref] Inactivity of human beta,beta-carotene-9',10'-dioxygenase (BCO2) underlies retinal accumulation of the human macular..., Li [/bib_ref]. The affinity of the human xanthophyll metabolizing BCO2 for lutein, zeaxanthin, and meso-zeaxanthin is 10-40 fold lower than the affinity observed in mice, who do not accumulate these carotenoids in the retina. Thus, it has been speculated that ineffective cleavage of xanthophylls contributes to their accumulation in the macula lutea. BCO2 knockout mice, unlike WT mice, accumulate zeaxanthin in their retinas. Thus, genetic variances in the respective enzymes or transport proteins likely affect the accumulation, distribution, and metabolism of oxo-carotenoids in the macula lutea.
## Tissues relevant for cardiovascular diseases
Besides the adipose tissue, other tissues such as the pancreas and various cells of the vascular system, including endothelial cells and macrophages, are important targets of related health beneficial effects of carotenoids. In an older study, carotenoid levels in the pancreas were found to be comparable (1.8-4.5 g/g wet weight [bib_ref] Carotenoids in man, Blankenhorn [/bib_ref] or 4.5-95 g/g [bib_ref] Distribution of carotene and vitamin A in liver, pancreas and body fat..., Dagadu [/bib_ref] to those in adipose tissue levels. The beneficial and risk preventive effects of carotenoids on atherosclerosis development were suggested to be mediated in macrophages and the endothelial cells of the blood vessels [bib_ref] Serum carotenoids and atherosclerosis. The Rotterdam Study, Klipstein-Grobusch [/bib_ref] [bib_ref] The effect of carotenoids on the expression of cell surface adhesion molecules..., Martin [/bib_ref]. Unfortunately, a direct mechanistic connection of carotenoid levels in white blood cells and especially monocytes/macrophages, as well as endothelial cells and further induced biological effects was never examined and confirmed.
As carotenoids were described in relation to the prevention of T2D, a potential target organ is the pancreas as a major regulator for insulin and glucagon production and secretion. Recently, RAR-and RXR-mediated signaling pathways were positively and negatively correlated with insulin and glucagon secretion [bib_ref] Retinoids in the pancreas, Brun [/bib_ref] , respectively. As pro-vitamin A carotenoids are the major precursor for the physiological and nutritional ligands of these two receptor subclasses [bib_ref] Carotenoid metabolism in mammals, including man: formation, occurrence, and function of apocarotenoids, Eroglu [/bib_ref] , a further genetic regulation related to carotenoids appears plausible, but has not yet fully confirmed and described. This plausible correlation to T2D, starting from beneficial regulatory pathways of retinoids as carotenoid metabolites via RAR-and RXR-mediated signaling pathways seems also to be highly dependent on carotenoid accumulation for substrate availability [bib_ref] Modulation of plasma alltrans retinoic acid concentrations by the consumption of carotenoid-rich..., Rühl [/bib_ref].
## Other tissues
Carotenoids occur in almost all human tissues. However, some were in the focus of research. Based on epidemiological studies, it was proposed that a frequent intake of tomato products rich in lycopene is associated with a decreased risk for prostate cancer [bib_ref] Tomatobased randomized controlled trial in prostate cancer patients: Effect on PSA, Paur [/bib_ref] , though it is unclear whether lycopene is selectively taken up in the prostate and which mechanisms may be relevant in this context. Levels of lycopene in prostate tissue have been reported around 1.7 ng/mg tissue following supplementation [bib_ref] Low prostate concentration of lycopene is associated with development of prostate cancer..., Mariani [/bib_ref] ; similar to levels in adipose and liver .
In the human brain carotenes and xanthophylls were detected and the latter accounted for about 70% of total carotenoids, with lutein and zeaxanthin dominating [bib_ref] Carotenoid, tocopherol, and retinol concentrations in elderly human brain, Craft [/bib_ref]. Levels were different in different brain areas and ranged from 2.8 to 11.8 pmol/g for lutein, 1.8 to 9.2 pmol/g zeaxanthin, and 7.6 to 15.2 pmol/g -carotene. A supplementation study with rhesus monkeys has shown that the brain levels of lutein and zeaxanthin are significantly related to their levels in the macula lutea [bib_ref] Macular lutein and zeaxanthin are related to brain lutein and zeaxanthin in..., Vishwanathan [/bib_ref] , and therefore, macular pigment density may be used as a surrogate biomarker of lutein and zeaxanthin in primate brain tissue. Analyses of human brain tissues revealed a relationship between StARD3 levels and the concentration of lutein [bib_ref] Relationship between concentrations of lutein and StARD3 among pediatric and geriatric human..., Tanprasertsuk [/bib_ref] , with strongest correlations observed in infant (versus. adult or centenarian) brains.
Colostrum contains significant amounts of ␣and carotene, lycopene, -cryptoxanthin, canthaxanthin, lutein and zeaxanthin, which are responsible for the yellow color. With ongoing lactation the content of carotenoids in mature milk declines, and the carotenoid pattern changes [bib_ref] Effect of the stage of lactation in humans on carotenoid levels in..., Schweigert [/bib_ref] , being correlated with lower lipid content of breast milk [bib_ref] Longitudinal survey of carotenoids in human milk from urban cohorts in China,..., Lipkie [/bib_ref]. This could suggest that a temporal specific mechanism is involved in the transfer of carotenoids to human milk, though via which mechanism is not understood. On the other hand, breast adipose tissue carotenoid concentrations from tumor patients were significantly related to serum concentrations, and were highest for -cryptoxanthin (3.5 mol/kg), -carotene (2.3 mol/kg) and lutein (1.8 mol/kg), not indicating a considerable change of patterns [bib_ref] Correlation between carotenoid concentrations in serum and normal breast adipose tissue of..., Yeum [/bib_ref].
Kidney concentrations, as other organ carotenoid levels, appear more variable than plasma concentrations, and have been reported to be in the range of 0.2-12.7 nmol/g for total carotenoids [bib_ref] Concentrations of selected carotenoids and vitamin A in human liver, kidney and..., Schmitz [/bib_ref]. However, a significant correlation for total carotenoids was found between liver, kidney and lung, thus additional discriminations of patterns could not be concluded. Again, host related factors especially those regarding the genetic variations of the proteins mention above might have impact on carotenoid distribution in these tissues.
## Storage and turnover aspects
Very little is known on the metabolism of especially nonprovitamin A carotenoids in xenobiotic pathways and on the extent of degradation and elimination of the parent compounds [bib_ref] Distribution of [14C]canthaxanthin and [14C]lycopene in rats and monkeys, Mathews-Roth [/bib_ref]. Lutein and -carotene may interact with each other during postprandial serum clearance when administered together. Both mutually enhancing and inhibiting actions have been observed in the limited number of volunteers studied. Interindividual differences in the response were studied after a high dose (0.5 mol/kg body weight) of either or both carotenoids followed by blood sampling [bib_ref] Intestinal absorption, serum clearance, and interactions between lutein and betacarotene when administered..., Kostic [/bib_ref] [bib_ref] Pharmacokinetics of beta-carotene and canthaxanthin after ingestion of individual and combined doses..., White [/bib_ref]. On average, the volunteers showed faster postprandial serum elimination (up to 120 h) of both carotenoids when given together, while subsequent elimination (up to 32 days) was unaffected by the other carotenoid. In line with this trial, the loss of liver vitamin A in rats dosed with -carotene was not affected by concomitant dosing with lutein; however, the initial storage was enhanced by smaller lutein doses and inhibited by larger doses [bib_ref] Effects of different amounts of lutein, squalene, phytol and related substances on..., High [/bib_ref]. Effects during -carotene absorption appear more likely to explain the inhibitory actions of lutein, whereas the enhanced vitamin A storage following -carotene dosing by low concomitant doses of lutein is more difficult to explain, and may involve interactions other than at the step of absorption. However, it is apparent from this study that (at least in rats) lutein does not affect subsequent loss of hepatic or renal stores of vitamin A. Whether a similar phenomenon on initial retinol storage exists in humans, which could partially explain interindividual variation in body stores of carotenoids is not known. Carotenoid kinetic aspects were determined in depletion studies of 70-80 days in females with and without isotope dilution by Burri et al. [bib_ref] Serum carotenoid depletion follows first-order kinetics in healthy adult women fed naturally..., Burri [/bib_ref]. Following a carotenoid controlled diet, carotenoids were measured in blood plasma of 19 healthy adults, and half-lives recorded. Lutein had the longest half-life (76 ± 17 days), followed by ␣-carotene, -cryptoxanthin, zeaxanthin, -carotene and lycopene (45 ± 7, 40 ± 5, 38 ± 7, 37 ± 5, and 27 ± 3 days, respectively), with lutein and lycopene differing significantly in half-lives from other carotenoids. Other studies based on postprandial designs have reported other half-lives, such as 2-3 days for lycopene and 5-7 days for -carotene [bib_ref] Clinical pharmacokinetics of antioxidants and their impact on systemic oxidative stress, Schwedhelm [/bib_ref] , likely to reflect plasma exchange with deeper compartments, while the longer reported half-lives would reflect losses from those deeper compartments. In the study by Burri et al., all carotenoids followed similar first order kinetic rates, indicating a small variation in plasma kinetics but clear differences in carotenoid half-lives. Concentrations of all carotenoids were highly correlated. The differences in carotenoid halflives were unexplained, though differences in degradation by acting as antioxidants or transfer to deeper compartments were suspected. Half-lives were unrelated to physical or demographic characteristics (BMI, energy metabolism, cholesterol and triglyceride levels, ethnicity, or age). However, the number of subjects was small and participants in this study rather homogenous. Thus, it is difficult to judge the effect of half-lives on interindividual carotenoid variations regarding blood levels.
In a study by Shvetsov et al. [bib_ref] Intraindividual variability in serum micronutrients: effects on reliability of estimated parameters, Shvetsov [/bib_ref] , the shorter halflife of lycopene was considered as an explanation for the higher intraindividual variability of plasma lycopene compared to other carotenoids (lutein, -carotene). In their study, plasma carotenoids of 381 women were measured repeatedly (4 times) with 4 month intervals, and intraindividual variability was approximately half of that of interindividual variability, except for lycopene, where it was equal. Similar higher intraindividual variability for lycopene was also found by Cooney et al. [bib_ref] Seasonal variations in plasma micronutrients and antioxidants, Cooney [/bib_ref]. The source of intra-individual variation is unclear. Seasonal variations appeared to be low (below 3% in the study by Shvetsov, partly due to the constant climate of Hawaii). When adjusted for age, race, alcohol drinking, and tobacco smoking, intraclass correlation coefficients were 0.69, 0.45, and 0.74 for total plasma lutein, lycopene, and -carotene, respectively. Dietary factors, age, gender, ethnicity, geographic location, and season were employed as main factors to explain intraindividual variability. A slightly better correlation with increased age was also reported, for reasons unknown. Further studies need to investigate the potential effect of diet, life-style and additional factors on carotenoid depletion rates.
Not much is known regarding carotenoid excretion pathways. Khachick et al. reported on polar metabolites of lycopene in humans [bib_ref] Identification, quantification, and relative concentrations of carotenoids and their metabolites in human..., Khachik [/bib_ref] , approximately 5% of an isotopically labelled -carotene ( 14 C) dose was excreted in urine (70% in feces) during 12 d [bib_ref] Long-term kinetic study of beta-carotene, using accelerator mass spectrometry in an adult..., Dueker [/bib_ref]. The minor fraction excreted in urine likely represents polar metabolites. In rat hepatocytes, it was shown that astaxanthin could be metabolized into 3-hydroxy-4-oxo--ionol, 3-hydroxy-4-oxo--ionone, and their reduced forms, 3-hydroxy-4-oxo-7,8-dihydro--ionol and 3-hydroxy-4oxo-7,8-dihydro--ionone, and that this was associated with induction of the cytochrome P450 enzyme (CYP3A4 as well as of CYP2B6) [bib_ref] Metabolism and CYP-inducer properties of astaxanthin in man and primary human hepatocytes, Kistler [/bib_ref]. Thus, alterations in P450 enzyme activity may influence carotenoid metabolism. In human keratinocyte cells, it was shown that various CYP enzymes metabolized all-trans retinoic acid and cis-isomers into water soluble products [bib_ref] A novel human cytochrome P450, CYP26C1, involved in metabolism of 9-cis and..., Taimi [/bib_ref].
## Disease conditions altering carotenoid turnover
Several clinical conditions related to carotenoid status affect carotenoid concentrations in human plasma, in addition to those involved in hampering carotenoid uptake (see chapter 3.4). However, it is often difficult to assess whether this happens by interference with carotenoid uptake, excretion and/or metabolism. In patients with renal failure, plasma -carotene levels increased. In the same study, decreased plasma levels of -carotene were observed in subjects with liver cirrhosis. Similarly, a study measuring carotenoids in tissues by needle biopsies showed much lower hepatic levels at all stages of liver disease [bib_ref] Differential depletion of carotenoids and tocopherol in liver disease, Leo [/bib_ref]. Another study examined -carotene plasma concentration in 53 Filipino children with cholestatic liver disease and found decreased concentrations in 45 patients [bib_ref] beta-Carotene deficiency in cholestatic liver disease of childhood is caused by beta-carotene..., Socha [/bib_ref]. This suggests that liver diseases interfere with storage and excretion of carotenoids. In the same study, six children received a single dose of 10 mg/kg body weight of -carotene. No increased plasma levels were detectable in 5/6 children, pointing to a main effect of the disease on -carotene absorption rather than further metabolism (re-distribution or excretion). This was explained by missing bile salt secretion and reduced solubilisation and cellular uptake. However, infections (e.g. helminths) could not be excluded as an additional factor.
The liver, via the bile, also plays an important part in the excretion of carotenoids back into the gut. As many transporters expressed in the intestine are also present in liver cells [bib_ref] SR-BI, CD36, and caveolin-1 contribute positively to cholesterol efflux in hepatic cells, Truong [/bib_ref] , it can be assumed that genetic differences in e.g. SCARB1 and CD36 do also influence biliary excretion, however, this constitutes a gap in our knowledge. In a study with the purpose of comparing plasma and biliary concentrations of carotenoids among controls and patients with biliary and pancreatic diseases, both plasma and bile concentrations of -carotene were significantly decreased in patients with bile duct stones, impairing biliary excretion. Moreover, the plasma/bile ratio was maintained as well as the correlation between them, and plasma -carotene decreased even more in patients with complete biliary obstruction, probably reflecting malabsorption due to limited carotenoid solubilisation in the gut. A tight correlation between plasma and bile -carotene still persisted in patients with pancreatic disease, confirming the role of plasma -carotene in determining bile concentrations [bib_ref] Carotenoids and tocopherols in various hepatobiliary conditions, Leo [/bib_ref]. The authors concluded that carotenoids undergo, at least in part, biliary excretion, that biliary concentrations reflect plasma levels in both normal and pathologic states, and that a decreased biliary excretion does not increase plasma concentrations. The study thus mainly highlights the importance of the bile for carotenoid absorption and the consequent tight link between bile and plasma levels.
Obesity was associated with circulating plasma carotenoids in several studies , however, the amount of carotenoids in the total adipose tissue has been found constant among obese and normal-weight subjects [bib_ref] The concentration of -carotene in human adipocytes, but not the wholebody adipocyte..., Östh [/bib_ref]. It is possible that a higher amount of adipose tissue with its high affinity to store carotenoids merely reduces the release of carotenoids into the bloodstream or enhances their uptake from the circulatory system, possibly via LDL receptors. As obesity is related to chronic inflammation, it cannot be excluded that upregulation of nuclear factor kappa-B (NF-B) and nuclear factor erythroid-derived 2 like 2 (NRF2) [bib_ref] Carotenoids, inflammation, and oxidative stress-implications of cellular signaling pathways and relation to..., Kaulmann [/bib_ref] is somehow related to a higher degradation rate of carotenoids, but this remains hypothetical. On the other hand, it is possible that the body may adapt to increased oxidative stress by upregulating circulating plasma antioxidants. For example, in a study with asthmatic women, higher levels of total carotenoids were found compared to non-asthmatic control women [bib_ref] Circulating antioxidant profile of pregnant women with asthma, Mclernon [/bib_ref].
An enhanced conversion of -carotene to retinol was suggested to explain significantly lower (by approximately 50%) serum levels of -carotene in hyperthyroid patients compared to those with hypo-and euthyroidism. However, the detailed mechanism remains unclear. Hypothyroidism also led to increased (2-fold) -carotene absorption in this study, explaining the yellowing of the skin in these patients. Thyroid hormones may thus alter -carotene absorption, however an effect on its distribution cannot be ruled out.
## Effect of vitamin a status and lifestyle on carotenoid status
It has been demonstrated that vitamin A status modulates -carotene absorption and cleavage (see chapter 3.2). A study examined the ability of deuterated retinol-dilution to detect changes in the body pool size and status of vitamin A and the effect on the bioconversion of carotenoids to vitamin A [bib_ref] Bioconversion of plant carotenoids to vitamin A in Filipino school-aged children varies..., Ribaya-Mercado [/bib_ref]. Changes were detected in the body pool size after 3 days. The bioconversion of dietary mixed plant food carotenoids varied inversely with vitamin A status, and improvements in status after intervention were strongly affected by total body stores of vitamin A, which can be explained by feed-back mechanism of vitamin A status and BCO1, highlighting the relation of circulating pro-vitamin A carotenoids and vitamin A status. Chronic alcohol consumption may perturb vitamin A and carotenoid metabolism. A study in rats given vitamin A or -carotene examined the effect of chronic alcohol consumption on vitamin A status and found a decrease in hepatic vitamin A storage, which was not due to malabsorption of either retinyl acetate or -carotene, nor to altered activities of several enzymes involved in ethanol or vitamin A metabolism [bib_ref] Effect of chronic alcohol consumption and moderate fat diet on vitamin A..., Grummer [/bib_ref]. However, other studies show inconsistent findings; studies suggested inhibition of BCO1/2 by ethanol [bib_ref] High dose lycopene supplementation increases hepatic cytochrome P4502E1 protein and inflammation in..., Veeramachaneni [/bib_ref].
## Bco1/2 aspects with respect to carotenoid tissue levels
As for the intestine, disruption of BCO1 expression is well known to reduce vitamin A and increase -carotene concentration in tissues, identifying BCO1 as the major enzyme for vitamin A production and for carotenoid cleavage [bib_ref] CMO1 deficiency abolishes vitamin A production from beta-carotene and alters lipid metabolism..., Hessel [/bib_ref]. The role of carotenoid oxygenases involved in the cleavage and storage of carotenoids is confirmed by other studies. For example, expression of BCO1 was documented by RNA blotting and immunostaining methods in a wide selection of human tissues [bib_ref] Cell type-specific expression of beta-carotene 9',10'-monooxygenase in human tissues, Lindqvist [/bib_ref] [bib_ref] Biochemical properties of purified recombinant human beta-carotene 15,15'-monooxygenase, Lindqvist [/bib_ref]. In the study by Lindquist et al. in mice, BCO1 was expressed in virtually all tissues, and the same is assumed for humans. An animal study examined the effect of knockout BCO1/2 in mice given a controlled diet primarily providing -carotene [bib_ref] Two carotenoid oxygenases contribute to mammalian provitamin A metabolism, Amengual [/bib_ref]. Accumulated levels of -carotene in serum, liver and lungs in BCO1 (−/−) and BCO1 (−/−) /BCO2 (−/−) mice were found. BCO1 (−/−) mice showed 100 fold higher concentrations of -carotene in tissues compared to wildtype and BCO2 (−/−) mice, confirming the role of BCO1 as the major -carotene-metabolizing enzyme. This does not negate a role for BCO2, since BCO2 inactivation in sheep resulted in carotenoid accumulation in adipose tissue [bib_ref] Biallelic beta-carotene oxygenase 2 knockout results in yellow fat in sheep via..., Niu [/bib_ref].
Another study in mice investigated the effect of gene expression induced by -carotene supplementation, knockout of BCO1, and differences in gender on -carotene levels in lungs, liver and inguinal white adipose tissue [bib_ref] Gene expression response of mouse lung, liver and white adipose tissue to..., Van Helden [/bib_ref]. Lungs were mainly affected by knockout, liver by knockout and gender, while the white adipose tissue was mainly affected by gender. Hardly any -carotene affected genes were in common in the three tissues, suggesting that changes in gene expression are primarily determined by tissue and gender.
-Carotene exposure increases -carotene concentration in the lung, but also the concentrations of retinol and retinyl esters [bib_ref] Knockout of the Bcmo1 gene results in an inflammatory response in female..., Van Helden [/bib_ref]. Inactivation of BCO1 increases -carotene concentrations, and decreases retinyl ester concentrations in males and females, while retinol concentration are only decreased in females [bib_ref] Downregulation of Fzd6 and Cthrc1 and upregulation of olfactory receptors and protocadherins..., Van Helden [/bib_ref]. Differential regulation of genes involved in vitamin A metabolism in the lung upon -carotene exposure, for example LRAT (conversion of retinol in retinyl esters) and ADH7 (conversion of retinol into retinal) suggest that polymorphisms in these genes can also have a role in interindividual responses, which could be relevant because these genes can functionally determine retinoid sufficiency [bib_ref] Knockout of the Bcmo1 gene results in an inflammatory response in female..., Van Helden [/bib_ref].
## Bioactivation -from carotenoids to nuclear hormone receptor ligands and further induced transcriptional signaling
The mechanisms of action of carotenoids regarding beneficial health effects comprises two major pathways: a) functioning as direct antioxidants via various pathways [bib_ref] Carotenoid-radical interactions, Krinsky [/bib_ref] , which has recently been discussed controversially due of lack of sufficiently high concentrations to transmit these effects endogenously [bib_ref] Carotenoids, inflammation, and oxidative stress-implications of cellular signaling pathways and relation to..., Kaulmann [/bib_ref] and b) functioning as precursors of various oxidative cleavage products. These can further interact with various nuclear hormone receptors such as the RAR, RXR, PPARs, LXRs, FXR, NF-B, and NRF2 [bib_ref] Carotenoids, inflammation, and oxidative stress-implications of cellular signaling pathways and relation to..., Kaulmann [/bib_ref] [bib_ref] Carotenoids and apocarotenoids in cellular signaling related to cancer: A review, Sharoni [/bib_ref]. This conversion was mainly demonstrated in cellular and mouse studies, while in humans merely individual steps of this activation cascade were confirmed [bib_ref] Lycopenederived bioactive retinoic acid receptors/retinoid-X receptors-activating metabolites may be relevant for lycopene's..., Aydemir [/bib_ref] [bib_ref] Lycopene-induced nuclear hormone receptor signaling in inflammation and lipid metabolism via still..., Rühl [/bib_ref] [bib_ref] Modulation of plasma alltrans retinoic acid concentrations by the consumption of carotenoid-rich..., Rühl [/bib_ref] [bib_ref] Overview of retinoid metabolism and function, Blomhoff [/bib_ref] [bib_ref] Lycopene supplementation restores vitamin A deficiency in mice and possesses thereby partial..., Aydemir [/bib_ref]. Assuming that that carotenoids function mainly via precursors of nuclear hormone receptors implies that not the serum and organ levels of native carotenoids are of major importance, but mainly the conversion of these endogenous present carotenoids to oxidative cleavage products. This again highlights the importance of the two carotenoid oxygenases BCO1/2. When focussing on health related effects of carotenoids, the focus should be placed on correlat-ing carotenoid intake with resulting endogenous carotenoid metabolite levels and their effects on nuclear hormone receptor mediated signaling, not merely on the concentrations of the native carotenoids, however, data on these are still scant. In this regard, the retinoic acid receptors (RARs), with the isotypes RAR␣,  and ␥, are activated mainly by the provitamin A carotenoid metabolite all-trans-apo-15´-carotenoic acid (all-trans retinoic acid [bib_ref] Nuclear receptor that identifies a novel retinoic acid response pathway, Mangelsdorf [/bib_ref]. In addition, non-well examined pathways with low affinity activation ligands such as all-trans-4-oxo-retinoic acid [bib_ref] The retinoid ligand 4-oxo-retinoic acid is a highly active modulator of positional..., Pijnappel [/bib_ref] , all-trans-3,4-didehydro retinoic acid [bib_ref] Retinoic acid receptors and retinoid X receptors: interactions with endogenous retinoic acids, Allenby [/bib_ref] and all-trans-13,14-dihydroretinoic acid activation were described [bib_ref] Activation of retinoic acid receptors by dihydroretinoids, Moise [/bib_ref]. The activation of the RAR as well as the RXR (RXR␣,  and ␥) by 9CRA seems to be of non-physiological relevance, due to low or even non-existing transcriptional activation of both RARs and RXRs by relevant endogenous levels [bib_ref] 9-cis-13,14-dihydroretinoic acid is an endogenous retinoid acting as RXR ligand in mice, Rühl [/bib_ref] [bib_ref] An endogenous mammalian retinoid X receptor ligand, at last! Chem, De Lera [/bib_ref].
Recently, in addition to -carotene functioning as the major precursors for nuclear hormone receptor ligands for RARs and RXRs, the acyclic carotenoid lycopene was described to transmit further RAR-mediated signaling [bib_ref] Lycopenederived bioactive retinoic acid receptors/retinoid-X receptors-activating metabolites may be relevant for lycopene's..., Aydemir [/bib_ref] [bib_ref] Lycopene induces retinoic acid receptor transcriptional activation in mice, Aydemir [/bib_ref]. Alternatively, RXR-and PPAR-mediated signaling was postulated to be mediated via lycopene oxidative metabolites. The detailed mechanisms including the actively involved lycopene oxidative metabolites remains still controversial but it is likely involving apo-15´-lycopenoids for further RXRmediated signaling [bib_ref] Lycopenederived bioactive retinoic acid receptors/retinoid-X receptors-activating metabolites may be relevant for lycopene's..., Aydemir [/bib_ref] and apo-10´-lycopenoids for further PPAR-mediated signaling [bib_ref] Lycopene-induced nuclear hormone receptor signaling in inflammation and lipid metabolism via still..., Rühl [/bib_ref]. Future research should focus on correlating carotenoid intake with carotenoid levels in the individual, further metabolic conversion to bioactive oxidative carotenoid metabolites, and the identification and quantification of marker genes related to beneficial health effects.
# Conclusion and perspective
Interindividual carotenoid variability following intervention studies with dietary carotenoids has been investigated in a number of body compartments, including digesta, chylomicrons, blood, skin, and the retina. Additional variation has been observed in observational studies . Variability in the carotenoid bioaccessible fraction is approximately half of that of plasma concentrations, approximately 20-30%. This may indicate that about half the variability may be explained by factors influencing carotenoid bioaccessibility, namely digestion enzyme concentrations, bile salts, and intestinal transit time. These are influenced by diseases (e.g. short bowel syndrome), the microbiota, and also gene expression related to enzymes (gastric lipase, PNLIP, CEL, CLPS). Factors influencing absorption also include parasites, such as hookworms, i.e. factors decreasing absorptive surface. Carotenoid absorption itself may be influenced by uptake transporters and associated SNPs (e.g. in the CD36, NPC1L1, and SCARB1 genes). Intracellular transport and chylomicron or also HDL secretion (ABCA1) have been associated with ABCG5/8, FABP2, ELOVL2, INSIG2, SLC27A6, and MTP, and intracellular cleavage with BCO1, likely involving BCO2 for some carotenoids. Further biodistribution, affecting plasma levels, likely include LPL, LIPC, CETP, and APOA1, APOA4, APOE and APOB. Tissue incorporation is influenced by all these preceding processes, in addition to specific uptake transporters such as GSTP1, StARD3, and RPE65, in case of the retina. Many other SNPs, involved in inflammatory processes and certain diseases have been shown to correlate with carotenoid tissue levels [fig_ref] Table 4: List of SNPs known, or speculated, to influence carotenoid metabolism C 2017... [/fig_ref] , though their exact role and contribution to variability remains to be elucidated. Hormones and gender play a role, as do possibly age and percentage of adipose tissue, though again the mechanisms are not comprehended. Several diseases have also been reported to influence carotenoid turnover and excretion, including hyperthyroidism and diseases of the liver and kidney, though the underlying mechanisms are not understood. While thus many of the potential candidates explaining interindividual variability, especially concerning cellular uptake and cleavage in the epithelium, have been determined, much less is known on their actual contribution to interindividual variability, and less is known on factors effecting intraindividual variability, which appears to be approximately half of interindividual variability according to some studies [fig_ref] Table 3: Host factors influencing carotenoid release from food matrix and bioaccessibility [/fig_ref]. Influences of season and diet are most likely to contribute. In addition, we tried to emphasize connections between the individual carotenoid levels in the organisms and their bioactivation, resulting in oxidative carotenoid cleavage metabolites, mainly retinoids, and further mediation of transcriptional signaling of health related marker genes.
In the future, studies should aim at identifying additional SNPs related to carotenoid ADME parameters, to increase our knowledge on the contribution of genetic variations to interindividual variability. This may include SNPs in genes encoding for digestion enzymes and proteins involved in further tissue distribution, which so far have received limited attention. Furthermore, the connection between SNPs and health related marker genes should be in the focus of research to scrutinize carotenoid health protective effects. In addition, epigenetic factors and the microbiota are areas which until to date have been mostly overlooked. These will reveal new insights into explaining the variability of carotenoid concentrations in human tissues, but perhaps also explain varying biological responses to dietary intervention, opening the door to personalized nutrition and "food to health" strategies employing carotenoids.
[fig] Figure 1: Predominant carotenoids in our diet, common metabolites and nomenclature. [/fig]
[fig] C 2017 The: Authors. Molecular Nutrition & Food Research published by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.mnf-journal.com [/fig]
[table] 2017: The Authors. Molecular Nutrition & Food Research published by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim [/table]
[table] Table 2: C 2017 The Authors. Molecular Nutrition & Food Research published by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.mnf-journal.com Studies investigating the variability of carotenoids in blood and target tissues, following intervention trials and observational studies [/table]
[table] Table 3: Host factors influencing carotenoid release from food matrix and bioaccessibility [/table]
[table] Table 4: List of SNPs known, or speculated, to influence carotenoid metabolism C 2017 The Authors. Molecular Nutrition & Food Research published by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim [/table]
|
10.1038/ncomms10941
|
CCBY
|
4799369
|
26979973
|
s2orc_pubmed_articles
|
Photoacoustics of single laser-trapped nanodroplets for the direct observation of nanofocusing in aerosol photokinetics
Photochemistry taking place in atmospheric aerosol droplets has a significant impact on the Earth's climate. Nanofocusing of electromagnetic radiation inside aerosols plays a crucial role in their absorption behaviour, since the radiation flux inside the droplet strongly affects the activation rate of photochemically active species. However, size-dependent nanofocusing effects in the photokinetics of small aerosols have escaped direct observation due to the inability to measure absorption signatures from single droplets. Here we show that photoacoustic measurements on optically trapped single nanodroplets provide a direct, broadly applicable method to measure absorption with attolitre sensitivity. We demonstrate for a model aerosol that the photolysis is accelerated by an order of magnitude in the sub-micron to micron size range, compared with larger droplets. The versatility of our technique promises broad applicability to absorption studies of aerosol particles, such as atmospheric aerosols where quantitative photokinetic data are critical for climate predictions.
U nderstanding fundamental processes that govern the reaction dynamics of gas phase, aerosol and cloud processes is crucial for the advancement of global atmospheric chemistry modelling [bib_ref] Polluted dust promotes new particle formation and growth, Nie [/bib_ref] [bib_ref] Heterogeneous photochemistry in the atmosphere, George [/bib_ref] [bib_ref] Infrared-driven unimolecular reaction of CH 3 CHOO Criegee intermediates to OH radical..., Liu [/bib_ref] [bib_ref] Photochemical kinetics of pyruvic acid in aqueous solution, Harris [/bib_ref] [bib_ref] Chlorine chronicles, Finlayson-Pitts [/bib_ref] [bib_ref] Alternative pathway for atmospheric particles growth, Monge [/bib_ref] [bib_ref] Evolution of organic aerosols in the atmosphere, Jimenez [/bib_ref] [bib_ref] Relative humidity dependence of light absorption by mineral dust after long-range atmospheric..., Lack [/bib_ref] [bib_ref] Reactions at interfaces as a source of sulfate formation in sea-salt particles, Laskin [/bib_ref] [bib_ref] Photolysis of sulfuric acid vapor by visible solar radiation, Vaida [/bib_ref] [bib_ref] Strong radiative heating due to the mixing state of black carbon in..., Jacobson [/bib_ref] [bib_ref] Influences of cloud photochemical processes on tropospheric ozone, Lelieveld [/bib_ref] [bib_ref] Nitric acid cloud formation in the cold Antarctic stratosphere: a major cause..., Crutzen [/bib_ref] [bib_ref] On the blue colour of the sky, and the polarization of light, Tyndall [/bib_ref]. Much of the chemistry occurring in the Earth's atmosphere is driven by sunlight. Photochemical reactions, in which aerosol particles or droplets act as the active reaction medium, can be highly complex because they are influenced by optical phenomena, transport properties and surface effects 2 . Optical phenomena play a fundamental role in light-initiated particle processes since the radiation flux within the particles determines the activation rate of the photochemically active species.
Focusing of electromagnetic radiation inside small particles leads to an enhancement of the overall light intensity, compared with the intensity of the incident radiation and to structuring and localization of the internal optical fields [bib_ref] Spatial distribution of the internal and near-field intensities of large cylindrical and..., Benincasa [/bib_ref] [bib_ref] Cavity enhanced droplet spectroscopy: principles, perspectives and prospects, Symes [/bib_ref] [bib_ref] Optical cavity resonances in water micro-droplets: implications for shortwave cloud forcing, Cappa [/bib_ref] [bib_ref] Laboratory-measured optical properties of inorganic and organic aerosols at relative humidities up..., Brem [/bib_ref] [bib_ref] Vibron and phonon hybridization in dielectric nanostructures, Preston [/bib_ref] [bib_ref] From plasmon spectra of metallic to vibron spectra of dielectric nanoparticles, Preston [/bib_ref] [bib_ref] Mapping nanoscale absorption of femtosecond laser pulses using plasma explosion imaging, Hickstein [/bib_ref] [bib_ref] Electron mean free path from angle-dependent photoelectron spectroscopy of aerosol particles, Goldmann [/bib_ref]. These phenomena depend strongly on the particle size, the particle composition and the wavelength of electromagnetic radiation. The fundamental influence of the enhanced electromagnetic energy density on the rate of photochemical reactions in micro-and nanodroplets has been recognized and calculations have provided limited evidence for enhanced photochemical rates [bib_ref] Actinic flux and photolysis in water droplets: Mie calculations and geometrical optics..., Mayer [/bib_ref] [bib_ref] Effect of optical resonances on photochemical reactions in microdroplets, Ray [/bib_ref] [bib_ref] Modelling of radiation quantities and photolysis frequencies in the aqueous phase in..., Ruggaber [/bib_ref] [bib_ref] Electromagnetic energy within dielectric spheres, Bott [/bib_ref]. Experimental results remain inconclusive concerning the influence of light enhancement on the kinetics, mainly because direct observation of the actual photoactive step was not possible [bib_ref] The use of chemical actinometry for the evaluation of the light absorption..., Penconi [/bib_ref] [bib_ref] Enhanced photolysis in aerosols: evidence for important surface effects, Nissenson [/bib_ref] [bib_ref] Quantum yields of CO 2 and SO 2 formation from 193 nm..., Mills [/bib_ref] [bib_ref] A laser trapping-spectroscopy study on the photocyanation of perylene across a single..., Kitagawa [/bib_ref] [bib_ref] Enhanced fluorescence yields through cavity quantum-electrodynamic effects in microdroplets, Barnes [/bib_ref] [bib_ref] A study of aerosol chemical reactions by optical resonance spectroscopy, Taflin [/bib_ref] [bib_ref] Photochemical polymerization of acrylamide aerosol particles, Ward [/bib_ref]. The observation of size-dependent effects in ensembles of aerosol or emulsion droplets is often hindered because the droplet size distribution cannot be varied and determined with the necessary accuracy. However, even single-droplet techniques have so far not provided size-dependent photolysis rates because the direct measurement of the population decay of the photoactive substance was not possible.
Elastic light scattering is sensitive enough to allow measurements on single sub-micron-sized droplets, but the information content is not specific enough to extract size-dependent rates. Raman spectroscopy, by contrast, could provide specific information but it comes with the disadvantage of low sensitivity (long averaging times), which would make its application to study processes in single submicrometre droplets where nanofocusing becomes important very challenging. Single-droplet fluorescence studies require a fluorescing compound, which strongly restricts its applicability. Furthermore, the fluorescence depletion is not always a reliable measure of the population decay of the photoactive species because of varying quenching efficiencies. The recently presented cavity ring-down studies on single droplets provide information on the extinction but not directly on the droplet absorption [bib_ref] Albedo measurements and optical sizing of single aerosol particles, Sanford [/bib_ref] [bib_ref] Optical extinction efficiency measurements on fine and accumulation mode aerosol using single..., Cotterell [/bib_ref]. Even in combination with light scattering measurements, the determination of rates in nanodroplets is likely prohibited by the uncertainty of the derived absorption.
This study reports the direct observation of light nanofocusing on the photokinetics in nanometre-to micron-sized droplets in the ultraviolet/visible (UV/VIS) range. To this end, we introduce single-droplet photoacoustic (PA) absorption spectroscopy, allowing the direct detection of the population decay of the photoactive substance. PA spectroscopy has been successfully used for the investigation of ensembles of aerosol particles [bib_ref] Photoacoustic spectroscopy for analytical measurements, Haisch [/bib_ref] [bib_ref] A wide spectral range photoacoustic aerosol absorption spectrometer, Haisch [/bib_ref] [bib_ref] Aircraft instrument for comprehensive characterization of aerosol optical properties, part 2: black..., Lack [/bib_ref] [bib_ref] Photoacoustic optical properties at UV, VIS, and near IR wavelengths for laboratory..., Gyawali [/bib_ref] , but its applicability to single aerosol particle studies has been controversial and has not previously been realised experimentally. Here we demonstrate the feasibility of single-droplet PA spectroscopy in combination with laser trapping, and provide direct experimental evidence for the size-dependence of the photolysis rate in model aerosol droplets due to nanofocusing effects. The results are compared with simulations using classical cavity electrodynamics.
# Results
Principle of single-droplet PAs. The two experimental set-ups, using a microphone and a quartz tuning fork, respectively, for resonant single-droplet PA measurements, are sketched in ,b (see Methods). For the droplet absorption experiments, we use a l ¼ 445 nm excitation laser (Nichia laser diode NDB7112E) of variable power (0.3-40 mW) modulated at the resonance frequency of the PA-resonator and the tuning fork, respectively. The resonance frequency and the Q-factor of the PA-resonator and the tuning fork are 3.97 kHz and B8.9, and 32.7 kHz and B8,000, respectively . The power of the excitation laser is recorded by a power meter after passing the PA cell and the tuning fork, respectively . The amplified PA signals are averaged over either 30 ms or 200 ms. For single-particle trapping, we use a counter-propagating optical tweezer built from a continuous laser beam of l ¼ 660 nm of B200 mW (Laser Quantum, Opus 660) . Such multiple beam optical traps allow trapping of sub-micron droplets, and combine the advantage of a comparatively simple set-up with high trapping stability and tight particle confinement (o100 nm) [bib_ref] Stability of aerosol droplets in Bessel beam optical traps under constant and..., David [/bib_ref] [bib_ref] Dynamics of submicron aerosol droplets in a robust optical trap formed by..., Thanopulos [/bib_ref] [bib_ref] Measurement of the instantaneous velocity of a brownian particle, Li [/bib_ref]. Droplets are trapped by gradient forces pointing towards the trap centre for all translational degrees of freedom. Single droplets are captured in the trap centre from a plume of aerosol generated by a nebulizer (see Methods). The droplet size is determined from laser light elastically scattered by the droplet [bib_ref] Stability of aerosol droplets in Bessel beam optical traps under constant and..., David [/bib_ref] (see Methods, . shows an example for light scattering measurements for droplet sizing. In the microphone set-up , the trap centre is located in the middle of the PA-resonator above the microphone [bib_ref] Development and characterization of a mobile photoacoustic sensor for on-line soot emission..., Beck [/bib_ref]. The trapping and excitation lasers enter and exit the cell through wideband, anti-reflective windows coated for the respective wavelengths. The CMOS camera for particle imaging and light scattering measurements is placed perpendicular to the excitation and trapping laser. The aerosol inlet and outlet are on the side of the PA cell outside the resonator.
In the tuning fork set-up , the droplet is trapped between the tines of the fork with collinearly aligned excitation laser and trapping laser beams. The CMOS camera for particle imaging and light scattering measurements is placed opposite the tines of the fork. ,d shows images of a single droplet trapped in between the tines and of a droplet ensemble flowing through the tines, respectively. The principal attractiveness of the tuning fork derives from its high detection sensitivity (very high Q-factors) and low sensitivity to environmental acoustic noise [bib_ref] Applications of quartz tuning forks in spectroscopic gas sensing, Kosterev [/bib_ref]. In our set-up, we mainly profit from the ease of combining it with laser trapping and light scattering measurements, as well as from the fact that it is chemically inert.
PA response of a single droplet. provides typical noise levels, background signals and a proof for single-droplet detection. illustrates the noise level and the background signal for the empty trap with the trapping laser on. With the excitation laser off ( À 5 soto0 s), the background signal (average) and noise level (1 s.d.) are B2.2±1.2 mV. Once the excitation laser is turned on at time t ¼ 0 s, a background signal of BS ¼ 5.3 mV with a noise level of NL ¼ 1.7 mV is recorded. The background signal is caused by excitation laser light scattered from the cell walls and hitting the microphone. Blocking the trapping laser (that is, disabling the trap) leaves the noise, as well as the background unchanged. shows the same as but with a single VIS441/tetraethylene glycol (TEG) solution droplet in the trap. The PA signal reaches a maximum (S max ) just after the excitation laser is turned on and then decreases exponentially as the VIS441 absorber undergoes photolysis. In , the trap is disabled at ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10941 t ¼ 7 s which leads to the immediate loss of the droplet and to a decrease of the PA signal to the background signal BS. This proves that the signal between 0 soto7 s indeed comes from a single droplet. The signal to noise ratio S NL ¼ S max À BS NL depends on the power P of the excitation laser, the concentration of the solution and the droplet size. shows exemplary experimental data for solution droplets of different size excited with different laser powers P. With the tuning fork set-up, we find an improvement in the S NL ratio of a factor of B3 compared with the microphone set-up. Note that the PA signal is caused by absorptive heating of the droplet and subsequent heat transfer to the surroundings. Evaporation of the solvent does not occur.
Detection limit. The minimum absorbance detectable with the single-particle PA set-ups at a given power of the excitation laser can be estimated from the PA signal and the single-droplet absorbance assuming that the noise level NL is the detection limit (see Methods, Calculation of droplet absorbance). As an example, we use the PA signal at t ¼ 0 (S max ) of the 530 nm droplet shown in (S max À BS) ¼ 33.3 mV corresponds to an equivalent absorbance of A ¼ 1.8 Â 10 À 5 . From the measured NL of the 530 nm droplet of NL ¼ 1.7 mV, we derive a minimum absorbance A min ¼ 9 Â 10 À 7 detectable with the microphone set-up. The improvement in S NL by a factor of B3 for the tuning fork set-up reduces the detection limit to A min B3 Â 10 À 7 , a minimum detectable absorption coefficient of a min ¼ 0.0074 Â 10 À 6 m À 1 or a minimum detectable absorption cross-section C abs,min ¼ 0.0037 mm 2 (laser power of 2.8 mW and averaging time of 200 ms) (see Methods equations (2)-(4)). The equivalent particle radius of 146 nm corresponds to a probe volume of 13 al. This far exceeds the performance of typical spectrometers (A min B10 À 3 -10 À 4 ), and is at least comparable to the most sensitive laser spectroscopic absorption measurements for macroscopic probe volumes. Note that in our set-up, this sensitivity is achieved for small (attolitre) probe volumes and very short (oo1 s) measurement times. Both can be further improved by increasing the laser power. Size-dependent photokinetics. The photokinetics in small droplets do not follow simple pseudo first order kinetics because the light intensity distribution inside the droplets is time dependent; that is, because of the concentration dependence of the nanofocusing. Therefore, we use the same initial concentration for all experiments so that the first half-life can be used as a measure for the size-dependence of the photokinetics. With our PA set-up, we directly measure the decay in absorption resulting from the population decay of the photoactive substance. Diffusion is so fast in the droplets (B10 mm 2 s À 1 ) that concentration gradients cannot build up and homogeneous concentrations for the solute can be assumed at all times. A model for the droplet photokinetics under these conditions is provided in Methods (Calculation of droplet photokinetics).
The model prediction (full blue line in shows a strong droplet size-dependence with a maximum in the inverse first halflife 1 t 1=2 at a droplet radius of B0.5 mm. Pronounced increases in 1 t 1=2 and fluctuations due to resonances are observed over the droplet size range from B50 nm to B1.2 mm. In this size range, the increase of the rate is caused by the enhancement of the internal electromagnetic field intensity through focusing of the light inside the droplet. shows the distribution of the light intensity inside a B0.5 mm droplet at t ¼ 0 s. The enhancement of the overall intensity and the local variation of the intensity are pronounced. The inverse half-life at a droplet radius of B0.5 mm is increased by a factor of B2.5 compared with the infinitely small droplet limit. The kinetics in these small droplets is no longer influenced by nanofocusing inside the droplet or light scattering by the droplet as visualized by the internal field intensity in . The inverse half-life of larger droplets (46 mm) exhibits only a weak size-dependence but decreases continuously (towards zero for infinitely large particles). The rate for these large droplets is determined by the balance between the decay of the absorber and the rise of the decay rate with time. As the photolysis proceeds, the penetration depth of the light and hence the internal field intensity increases. As in the case of very small droplets, nanofocusing is not important for very large droplets. Large droplets essentially represent the behaviour of thin bulk films with the same effective thickness as the droplet. The 1 t 1=2 increases by a factor of B10 for a 0.5 mm compared with a 13 mm droplet, which implies a substantial increase in the rate of sub-micron-size droplets relative to bulk. The dashed red line in simulates the behaviour of a hypothetical droplet excluding the influence of nanofocusing and light scattering but still accounting for the droplet-size-dependent absorption (see Methods, Calculation of droplet photokinetics). This curve represents bulk behaviour. The comparison with the full blue line clearly shows the pronounced influence of light focusing on the rate in the sub-micron to micron size range.
The statistically evaluated experimental first half-lives (black circles in are determined from time-dependent PA measurements (see Methods, Statistical analysis). The experimental results clearly follow the size-trend predicted by the model (full blue line). The pronounced maximum of 1 t 1=2 at a droplet size of B0.5 mm is clearly visible even though the data scatter notably below B1 mm mainly because of the uncertainty in the droplet size determination (Methods, Droplet Sizing). Our experimental data show somewhat higher values of the inverse half-life for larger droplets than the model prediction. Deviations from the model assumptions including modified PA response in large droplets 47 could potentially account for this. We have recently introduced a broad-band scattering method for accurate sizing of submicrometre particles, which will allow us in future to significantly reduce the size uncertainty in the submicrometre range (unpublished data). However, already at the current level of accuracy, the data in provide the first direct observation of the strong influence of nanofocusing of light on the photokinetics in droplets.
# Discussion
The experimental results in confirm a strong size-dependence of the rate of photochemical reactions in droplets. This optical phenomenon shows the most pronounced effect in the submicrometre to micrometre droplet size range for electromagnetic radiation in the UV to VIS range, that is, for the relevant frequency range in atmospheric processes. Classical cavity electrodynamics provides a semi-quantitative description of the kinetics for our ideal model system. The photokinetics of our model system is representative of typical atmospheric aerosols; that is, of typical optical properties of these particles. For example, similar quantitative results are predicted for aqueous droplets . The acceleration of the kinetics we find in the visible range is predicted to be even more pronounced in the UV range of the solar spectrum . Many aerosol particles are non-spherical. However, for particles with different shapes but the same volumes one finds quantitatively similar nanofocusing effects as for droplets. Nanofocusing also affects surface reactions since the strong intensity enhancement in forward direction shown for the internal field in extends to the external field near the surface (not shown). The ability to measure and thus quantify the kinetics of the light-induced step in photochemical reactions in aerosol particles is of fundamental importance for atmospheric chemistry, where chemical processes are largely driven by sunlight. The diverse and complex processes (for example, transport and surface phenomena) in atmospheric aerosol particles require direct measurement methods as the one introduced here because simple models are of limited applicability.
The introduction of single-droplet PA spectroscopy in the present study finally makes the direct observation of the photoactive step possible. Single-droplet PA was previously deemed not feasible because of sensitivity and background issues. Here we demonstrate the viability of this new method and its very high sensitivity (C abs,min ¼ 0.0037 mm 2 ) enabling studies even of single nanodroplets (10 al). PA spectroscopy provides a general absorption method that can be used in any frequency range. The combination with laser trapping lets us follow the evolution of individual droplets under controlled conditions over extended periods of time (up to several days). This versatility enables fundamental studies on many different droplet systems relevant to atmospheric and technical processes. The investigation of droplet photokinetics is just one example where this new broadly applicable single-droplet method can make an important contribution.
# Methods
PA measurements with microphone. The PA cell is made of brass and consists of a longitudinal PA-resonator (length 40 mm, diameter 4 mm), which is connected to two buffer volumes with acoustical baffles for sound insulation [bib_ref] Development and characterization of a mobile photoacoustic sensor for on-line soot emission..., Beck [/bib_ref]. A sensitive microphone (EK 23029, Knowles) is used with a custom-made preamplifier. The output signal is recorded by a lock-in amplifier (Stanford, SR 830).
PA measurements with tuning fork. The distance between the two tines of the tuning fork (Q 32.768 kHz TC 38, AURIS) is 300 mm. Each tine has a width of 600 mm, a thickness of 340 mm and a length of 3.8 mm. The quartz tuning fork acts as the resonant acoustic transducer, which generates an electric signal on resonant excitation by an acoustic wave due to the piezoelectric effect [bib_ref] Applications of quartz tuning forks in spectroscopic gas sensing, Kosterev [/bib_ref]. The signal recording is identical to the microphone set-up except for the more precise reference frequency adapted to the higher Q-factor.
Aerosol generation and materials. To study photokinetics in single droplets, solutions of the photoactive dye VIS441 (Cyanine dye with formula NaC 17 H 25 N 3 O 5 S 3 and molar mass 470, QCR solutions) in TEG solvent (ACROS organics, 99.5%) are nebulized with a medical nebulizer (Pari, PARI Boy SX). A concentration of 4.55 gl À 1 VIS441 in TEG is used. For measurements on pure solvent droplets, pure TEG is nebulized. The shows an UV/VIS spectrum of a bulk solution of VIS441 in TEG and of pure TEG solvent, respectively. Droplet sizing. The particle size is determined from excitation laser light scattered elastically by the droplet. The scattered light intensity is collected for scattering angles between 76.5˚and 103.5˚and focused onto a CMOS camera (Thorlabs, DCC1645C, 1280 Â 1024 pixels) using a camera objective (Super Carenar, focal length ¼ 50 mm, f-number ¼ 1.7). The particle size is retrieved by fitting calculated phase functions to experimental ones using Mie theory [bib_ref] Stability of aerosol droplets in Bessel beam optical traps under constant and..., David [/bib_ref]. The sizing of sub-micron-sized droplets is difficult because only few fringes are left in the scattering pattern (for example, . Larger particles exhibit brighter scattering images and many more fringes , which makes sizing easier. We estimate uncertainties in the droplet radius of about half the wavelength.
Calculation of droplet absorbance. The PA signal S is assumed to be proportional to the power P abs absorbed by the droplet, which is located at the centre of a Gaussian excitation laser beam (beam waist radius of 87 mm and cross-section q L ¼ 11,889 mm 2 ) with incident power P
[formula] S / P abs ¼ I 0 Á C abs ¼ P Á C abs q Lð1Þ [/formula]
I 0 is the intensity incident on the droplet and C abs is the droplet's absorption cross-section. The equivalent absorbance A due to absorption is given by
[formula] A ¼ ln P P À P abs ¼ ln q L q L À C abs % C abs q Lð2Þ [/formula]
For a single droplet in the PA cell, the equivalent absorption coefficient is given by,
[formula] a ¼ C abs V resð3Þ [/formula]
where V res ¼ 0.5 cm 3 is the volume of the PA-resonator. The absorption cross-section of the droplet is calculated from the Mie theory 44 with a refractive index of the surroundings equal to 1:
[formula] C abs ¼ 2p e Á m Á o 2 X j 2j þ 1 ð Þ Re a j x; m ð Þþb j x; m ð Þ È É À a j x; m ð Þ 2 À b j x; m ð Þ 2 " #ð4Þ [/formula]
Here a n and b n are the scattering coefficients, x ¼ 2pa l is the size parameter, a is the droplet radius, l its wavelength of light in vacuum, o is the angular frequency of the light, e and m are the permittivity and the permeability, respectively, of the droplet, and m ¼ n þ ik is the droplet's complex index of refraction at the wavelengths of the excitation laser (l ¼ 445 nm). The latter is determined from UV/VIS absorption and refractometric measurements of VIS441/TEG bulk solutions and a pure TEG solution using Kramers-Kronig inversion. The refractive index of the VIS441/TEG solution for a dye concentration 4.55 gl À 1 and the pure TEG solvent are n þ ik ¼ 1.463 þ i Á 0.0062 and n s þ ik s ¼ 1.460 þ i Á 0.0000, respectively. The refractive index of the VIS441/TEG solution (dye concentration 4.55 g l À 1 ) in the UV/VIS range is provided in the . For other dye concentrations (see photokinetics), it is assumed to depend linearly on the dye concentration.
Calculation of droplet photokinetics. The droplet photokinetics is described by the following rate equation p. Here r denotes the location within the droplet and I is the local field intensity. Both I and s depend on the complex index of refraction, which in turn depends on the number density N, so that the rate law is no longer pseudo first order. The power absorbed by the droplet is given by the rate of absorption integrated over the droplet's volume V
[formula] dN dt ¼ À p ÁIðrÞP abs t ð Þ ¼ I 0 Á C abs N ð Þ ¼ Àhn Á Z p À 1 dN dt dVð6Þ [/formula]
Assuming fast diffusion, that is, N 6 ¼ NðrÞ, we obtain:
[formula] dN dt ¼ À f Á V À 1 C abs ðNÞ ð 7Þ [/formula]
where f ¼ p Á I 0 =hn is the product of incident photon flux and reaction probability. Equation (7) is integrated using a 4th order Runge-Kutta method with the time-dependent PA signal given by equation (1). The corresponding inverse first half-lives of the PA signal as a function of droplet radius are shown as a full blue line in .
To illustrate the effect of nanofocusing, we compare the above model (full blue line in with a model that neglects the influence of nanofocusing (dashed red line in . This model is obtained from equations (5) and (6) C abs ¼ N t ð Þ Z I r ð Þ I 0 s r ð ÞdV ð8Þ by inserting the small particle limit 44 for s
[formula] sV À 1 ¼ p l Im À 18 m 2 þ 2 & 'ð9Þ [/formula]
and a Beer-Lambert expression for the intensity distribution within the particle
[formula] IðrÞ ¼ I 0 Á exp À 4pk l 'ðrÞ & 'ð10Þ [/formula]
where 'ðrÞ ¼ r cos y þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi a 2 À r sin y ð Þ 2 q is the absorption path length at distance r from the centre of the particle and at polar angle y relative to the incident beam direction.
Statistical analysis of experimental photolysis data. To account for the uncertainties both in the particle radii and in the decay half-lives of the experimental data set , we perform a two-step maximum likelihood analysis. First, the distribution of particle radii D a ð Þ ¼ P g i ðaÞ is analysed assuming a normally distributed error for the size determination,
[formula] g i a ð Þ ¼ exp À ða À a i Þ 2 =2s 2 a È É s a ffiffiffiffiffi 2p pð11Þ [/formula]
with a constant s.d. of s a ¼ 220 nm. The local extrema in D at a k divide the size range into sections with a lower and an upper half for each cluster of data, which are combined into a single section for isolated data points. For each section, we finally obtain the most probable values for particle radius and the inverse half-life as weighted averages over the particles with weights given by,
[formula] w i;k ¼ s À 2 t;i Z ak ak À 1 g i ðaÞda= X i s À 2 t;i Z ak ak À 1 g i ðaÞdað12Þ [/formula]
This implies normally distributed errors for the experimental inverse half-lives with s.d. s t,i ranging from about 10% for the most accurate measurements to about 50% for measurements with S NL oo10 (typically small particles). The error bars in were obtained by s.e. propagation.
[fig] Figure 2 |Figure 1 |: Typical noise levels and background signals for the microphone set-up. (a) For the empty trap, identical noise levels and background signals are recorded for a pure TEG solvent droplet in the trap (data not shown). (b) With a VIS441/TEG solution droplet in the trap. At t ¼ 0 s, the excitation laser (445 nm) is switched on and at t ¼ 7 s the trapping laser is switched off. Sketch of the experimental single-droplet PA set-ups. (a) Microphone set-up with PA-resonator, excitation laser, trapping laser and light scattering measurements. The colours in the PA cell indicate that the acoustic mode has its maximum amplitude (red) in the vicinity of the microphone and a value close to zero (green) in the region of the acoustical baffles. (b) Tuning fork set-up. (c) Snapshot of a single droplet trapped between the tines of the tuning fork (view from top). Note that the droplet (B1 mm) is much smaller than the detection volume between the tines (B0.3 Â 0.34 Â 2 mm). (d) Snapshot of an ensemble of droplets flowing in between the tines from left to right. (e) Light scattering image as recorded by the CMOS camera (left) with experimental and calculated phase function (right) for a droplet with a radius of a ¼ 530 nm. [/fig]
[fig] Figure 3 |Figure 4 |: Exemplary PA signals of VIS441/TEG solution droplets as a function of time. The decay of the signal is caused by photolysis of the solute. The experimental data (blue dots) are recorded for different droplet radii a and different power P of the excitation laser. (a,b) Recorded with the microphone set-up. (c,d) Recorded with the tuning fork setup. The red lines are fits to the experimental data providing experimental first half-lives t Size-dependent photokinetics. (a) Inverse first half-lives 1 t 1=2 as a function of the droplet radius for a laser power of 1 mW. Black circles: statistically evaluated experimental data. Error bars show 1 s.d. Full blue line: model prediction including nanofocusing and scattering effects. The calculations are for a quantum yield of 7 Â 10 À 6 . Dashed red line: model prediction for a hypothetical bulk limit, that is, excluding contributions from nanofocusing and scattering. Distribution of the light intensity inside droplets at t ¼ 0 s for (b) a 0.5 mm droplet and (c) a 20 nm droplet. The colour scheme is relative to an incident light intensity of 1. [/fig]
|
10.1172/jci.insight.126533
|
CCBY
|
6542649
|
30998505
|
s2orc_pubmed_articles
|
Blocking IL-10 receptor signaling ameliorates Mycobacterium tuberculosis infection during influenza-induced exacerbation
## Lipid extraktion
500µl of lung homogenates were transferred into an ice-cooled 2.0 ml tube covered with aluminum foil. Afterwards, 0.5 µl of butyl hydroxyl-toluene in methanol (BHT; 1.0 mg/ml) followed by 25µl of internal standard mix containing PGE2-d9, RvD2-d5, LTB4-d4, 5-HETE-d8, AA-d11 at a concentration of 500 fmol/L was added. Subsequently, 320 µL of ice-cold chloroform was added together with 640 µL of a freshly prepared ice-cold MeOH/HAc (97+3; v/v) solution to the mixture.
The mixture was incubated for 30 min at room temperature under slight stirring. For phase separation 320 µL of ice-cold water was added and the mixture was incubated for another 30 min under constant shaking. At last, the samples were centrifuged at 15,000 x g for 15 min at 4 °C to improve phase separation. Afterwards the organic phase was collected and transferred to an ice-cooled sample tube.
The aqueous was three more times extracted with 320 µL of ice-cold CHCl 3 . Organic phases were combined and dried under a slight stream of nitrogen. Lipid extracts were dissolved in 40 µL of solvent A and 2 µL were injected for the LC-MS 2 analysis.
## Liquid chromatographytandem mass spectrometry (lc-ms 2 )
Chromatographic separation was performed at a flow rate of 10 µl/min with a Luna C18 (2)
[fig] Figure S1: IAV coinfection 30 days after Mtb infection impairs bacterial control in the lung. (A) Schematic time-line of experimental set-up. (B) C57BL/6 mice were infected via aerosol with a low dose of Mtb H37Rv and 30 days later coinfected i.n. with 10 4 PFU IAV (A/HH/05/09 H1N1). At indicated time points after Mtb infection, lung bacterial burden was determined (n= 8-9 per group, data pooled from two independent experiments). Each data point represents one mouse. **p ≤ 0.01 determined by unpaired t-test. [/fig]
[fig] Figure S2: Anti-IL10R treatment does not restore Mtb-specific T cell responses. C57BL/6 mice were infected via aerosol with a low dose of Mtb H37Rv and 12 days later coinfected with 10 4 PFU IAV (A/HH/05/09 H1N1). 200 µg anti-IL10R antibody (1B1.3) or isotype control (polyclonal rat serum IgG) were administered i.p. on day 5 after IAV infection (day 17 Mtb). Lungs were analyzed by flow cytometry on day 21 Mtb for (A, B) I-A b ESAT-6 4-17 -specifc CD4 + T cells (n= 3-5 per group, one experiment), and (C) IFN-γ producing CD4 + T cells upon ex vivo stimulation with ESAT-6 1-20 peptide (n= 3-5 per group, one experiment). Each data point represents one mouse. *p ≤0.05; **p ≤ 0.01, determined by one-way ANOVA followed by Tukey's multiple-comparison test. [/fig]
[fig] Figure S3: Flow cytometry gating strategy for determination of I-A b ESAT-6 4-17 + T cells. Representative flow cytometry blots for gating of I-A b ESAT-6 4-17 + CD4 + T cells isolated from lungs and dLNs. Gating strategy depicted by an example of cells isolated from the lung of an IAV-infected mouse and Mtb-infected mouse on day 21 after Mtb infection (day 9 IAV infection). [/fig]
[fig] Figure S4: Flow cytometry gating strategy for determination of TNF + and IFN-γ + CD4 + T cells.Representative flow cytometry blots for gating of TNF and IFN-γ + CD4 + T cells isolated from lungs and dLNs. Cells from lungs of Mtb-infected mice on day 21 after Mtb infection are representatively shown. [/fig]
[fig] Figure S5: Flow cytometry gating strategy for determination of proliferated CD4 + T cells in dLNs. Representative flow cytometry blots for detection of proliferated CFSE-labeled p25TCRtg-CD4 + T cells isolated from dLNs. Proportion of proliferated cells was determined as difference of % total CFSE + -% unproliferated cells of CFSE + cells. Gating strategy depicted by an example of cells isolated from the dLN of an Mtb-infected mouse and IAV-infected mouse (negative control) shown for day 21 after Mtb infection (day 9 IAV infection). [/fig]
[fig] Figure S6: Flow cytometry gating strategy for determination of Foxp3 + and Foxp3 + IL-10 + CD4 + T cells. Representative flow cytometry blots for gating of Foxp3 + and Foxp3 + IL-10 + CD4 + T cells isolated from lungs. Plots are from lung cells of an Mtb-IAV coinfected mouse on day 18 after Mtb infection. [/fig]
[fig] Figure S7: Flow cytometry gating strategy for determination of IL-10 producers among myeloid cells. Representative flow cytometry blots for gating of IL-10 + cells among neutrophils, monocytederived dendritic cells (moDCs) and CD11b + DCs. Exclusion of autofluorescent cells (AFs; w/o AFs) and neutrophils (w/o neutrophils) along the gating. Plots are from lung cells of an Mtb-IAV coinfected mouse on day 18 after Mtb infection. [/fig]
[fig] Figure S8: Flow cytometry gating strategy for determination of IL-10 producers among lymphoid cells. Representative flow cytometry blots for gating of IL-10 + cells among NK cells, CD4 + and CD8 + T cells. Plots are from lung cells of an Mtb-IAV coinfected mouse on day 18 after Mtb infection. [/fig]
[fig] Figure S9: Flow cytometry gating strategy for determination of H1N1-NP366 + (dextramer + ) CD8 + T cells. Representative flow cytometry blots for gating of H1N1-NP366 dextramer + CD8 + T cells. Plots are from lung cells of an Mtb-and an Mtb-IAV coinfected mouse on day 21 after Mtb infection (9 days after IAV infection). [/fig]
[table] Table S1: Selected parameters for the characterization of LM during an LC-MS 2 run. [/table]
|
10.1038/s41598-023-48777-z
|
CCBY
|
10710453
|
38071213
|
s2orc_pubmed_articles
|
Synergistic effect of cryptotanshinone and temozolomide treatment against human glioblastoma cells
Glioblastoma multiforme (GBM) is a complex disease to treat owing to its profound chemoresistance.Therefore, we evaluated the combined effect and therapeutic efficacy of temozolomide (TMZ), a potent alkylating agent and the current gold standard therapy for GBM, and cryptotanshinone (CTS), which inhibits glioma cell proliferation in GBM cells.Using LN229 and U87-MG human GBM cells in a short-term stimulation in vitro model, the cytotoxic and anti-proliferative effects of single and combined treatment with 4 μM CTS and 200 μM TMZ were investigated.Furthermore, cell viability, DNA damage, apoptosis rate, and signal transducer and activator of transcription 3 (STAT3) protein were measured using cytotoxic assay, comet assay, flow cytometry, and western blotting analysis, respectively.The two drugs' synergistic interaction was validated using the synergy score.We found that the anti-proliferative effects of combination therapy using the two drugs were greater than that of each agent used alone (CTS or TMZ).Western blot analysis indicated that treatment of GBM cells with CTS combined with TMZ more significantly decreased the expression of MGMT and STAT3, than that with TMZ alone.Combined treatment with CTS and TMZ might be an effective option to overcome the chemoresistance of GBM cells in a long-term treatment strategy.Gliomas are the most common primary brain tumors, and over half of these are glioblastoma tumors, which are the most malignant 1 .A combination of surgery, radiotherapy, and/or chemotherapy is the primary strategy for glioblastoma multiforme (GBM) therapy 2 .Temozolomide (TMZ) is a first-choice alkylating agent and the gold standard therapy for GBM because of its ability to penetrate the blood-brain barrier, weaker adverse side effects, and higher effectiveness in extending the life span of patients3,4.The anti-tumor effect of TMZ depends on its ability to deliver a methyl group to purine bases of DNA (N7-guanine, N3-guanine, O6-guanine, and N3-adenine), which can induce cell cycle arrest at the G2/M phase and eventually lead to apoptosis[5][6][7].However, TMZ is ineffective against approximately 50% of primary or recurrent GBMs 8 .TMZ resistance is associated with the expression levels of DNA alkylating proteins and repair enzymes9,10.The endogenous enzyme O-6-methylguanine DNA methyltransferase (MGMT) binds to the alkyl compound on the sixth oxygen atom of DNA guanine and transfers the alkyl compound to the 145th cysteine active site of MGMT.Guanine, which causes DNA alkylation, is reduced, thus attenuating the killing effect of TMZ on tumor cells[11][12][13].The absence of a mismatch repair (MMR) system is another critical mechanism that generates resistance in GBM14,15.Many other molecular mechanisms, including DNA repair systems, abnormal signaling pathways, autophagy, epigenetic modifications, microRNAs, and the production of extracellular vesicles, contributing to TMZ resistance have been revealed in recent years[16][17][18][19].Although many of these mechanisms have been elucidated, the efficacy of TMZ has not improved.Thus, novel drugs and therapeutic protocols for the combined treatment of TMZresistant glioma cells are urgently required.
A convenient and efficient method for discovering and developing new chemotherapy drugs is studying the anti-cancer properties of Chinese herbal medicines and their derivatives.Cryptotanshinone (CTS) (IUPAC name: (R)-1,2,6,7,8,9-hexahydro-1,6,6-trimethyl-phenanthro(1,2-b) furan-10,11-dione) (Fig. [fig_ref] Figure 1: Figure 1 [/fig_ref] is a biologically active diterpenoid quinone compound extracted from fat-soluble constituents in the roots and rhizomes of Radix Salvia miltiorrhiza.CTS possesses various bioactivities, such as antibiotic [bib_ref] From bioprofiling and characterization to bioquantification of natural antibiotics by direct bioautography..., Jamshidi-Aidji [/bib_ref] [bib_ref] Synergistic interactions of cryptotanshinone and aminoglycoside antibiotics against Staphylococcus aureus in vitro, Teng [/bib_ref] , anti-inflammatory [bib_ref] Cryptotanshinone specifically suppresses NLRP3 inflammasome activation and protects against inflammasomemediated diseases, Liu [/bib_ref] , anti-aging, and anti-tumor [bib_ref] Retraction Note to: A new synthetic derivative of cryptotanshinone KYZ3 as STAT3..., Zhang [/bib_ref] [bib_ref] Cryptotanshinone inhibits human glioma cell proliferation in vitro and in vivo through..., Lu [/bib_ref] [bib_ref] A system bioinformatics approach predicts the molecular mechanism underlying the course of..., Jiaojiao [/bib_ref] properties.The anti-cancer properties of CTS have drawn the attention of many researchers in recent years.Previous studies on the molecular mechanisms associated with CTS intervention have revealed that the JAK2/STAT3, PI3K/AKT, and cell cycle pathways are involved in the inhibitory and pro-apoptotic effects of this compound in different tumor cell lines [bib_ref] A review of the biological activity and pharmacology of cryptotanshinone, an important..., Li [/bib_ref].CTS can inhibit the Tyr705 phosphorylation of signal transducer and activator of transcription 3 (STAT3) by binding to its SH2 domain, promoting the upregulated activity of the SHP-2 protein tyrosine phosphatase.It further blocks STAT3 dimerization and transcriptional activity after entering the nucleus, thus inhibiting the proliferation of tumor cells [bib_ref] The autophagy-independent role of BECN1 in colorectal cancer metastasis through regulating STAT3..., Hu [/bib_ref] [bib_ref] Asialoglycoprotein receptor 1 functions as a tumor suppressor in liver cancer via..., Zhu [/bib_ref].However, the inhibition of STAT3 phosphorylation is independent of its effect on JAK2 phosphorylation [bib_ref] Cryptotanshinone inhibits human glioma cell proliferation in vitro and in vivo through..., Lu [/bib_ref].As an MGMT transcription factor, the activation and nucleation of Tyr705 in STAT3 are critical for DNA damage repair in tumor cells.Knocking down STAT3 can effectively reduce MGMT expression [bib_ref] Inhibition of STAT3 enhances the radiosensitizing effect of temozolomide in glioblastoma cells..., Han [/bib_ref] [bib_ref] STAT3 inhibition overcomes temozolomide resistance in glioblastoma by downregulating MGMT expression, Kohsaka [/bib_ref] [bib_ref] On-target JAK2/STAT3 inhibition slows disease progression in orthotopic xenografts of human glioblastoma..., Stechishin [/bib_ref] [bib_ref] A STAT3-based gene signature stratifies glioma patients for targeted therapy, Tan [/bib_ref].Therefore, screening compounds that directly target STAT3 and inhibit MGMT expression are of considerable significance for improving the efficacy of TMZ chemotherapy in glioma patients.
Currently, the anti-glioma effects of TMZ in combination with CTS have not been reported.Therefore, we aimed to address the synergistic sensitization effect of CTS and TMZ on GBM cells.The MGMT and STAT3 signaling pathways were evaluated to investigate the mechanism underlying the synergistic effect of CTS and TMZ through effective combination treatment.
## Characterization of the anti-glioma cells effect of cts and tmz co-treatment
CTS has not yet been reported elsewhere for its combined effect with TMZ; therefore, it was selected for further investigation.The cell viability of 7-by-7 drug dose was detected by CCK-8 kit, and the inhibition rate of each concentration of drug on cell viability was calculated (the data is ordered with the first drug in columns and the second in rows).And then, synergy score of the two drugs combination was obtained by Combenifit software to evaluate the cell activity inhibition rate at different drug concentrations.As shown in Fig. [fig_ref] Figure 2: Figure 2 [/fig_ref] , LN229 and U87-MG cells were exposed to various 7-by-7 matrix concentrations of CTS (0.25-16 μM) and TMZ (25-1600 μM) for 72 h.The half-maximal inhibitory concentrations (IC50) of CTS against LN229 and U87-MG cells were 3.49 μM and 3.41 μM, respectively, and the IC50 of TMZ were 421 μM and 468 μM, respectively.In addition, classical synergy models, including the highest single agent (HSA) and Bliss reference models, were applied to determine HSA and Bliss values, which are readouts for synergistic inhibition and depict the difference between the expected inhibition and observed inhibition.Most HSA and Bliss values were above 0, indicating synergistic interactions between the two drugs (Fig. [fig_ref] Figure 2: Figure 2 [/fig_ref].In addition, the expected versus observed dose-response surface plot of the 7-by-7 matrix viability data helped verify the dosage of the synergistic effect of the two agents against LN229 and U87-MG GBM cells (Fig. [fig_ref] Figure 2: Figure 2 [/fig_ref].
## Combination treatment of gbm cells with cts and tmz suppresses proliferation
To further confirm whether CTS combined with TMZ has a synergistic inhibitory effect on glioma cells, LN229 and U87-MG cells were treated with a series of TMZ concentrations, with or without CTS treatment, for 72 h, and the IC50 value of TMZ was assessed using a cell viability assay.As shown in Fig. [fig_ref] Figure 3: Figure 3 [/fig_ref] , compared with the control group, the IC50 value of TMZ decreased from 418 to 197 μM following the combination treatment of TMZ and CTS.Subsequently, we evaluated the long-term inhibitory effect of TMZ combined with CTS on GBM cells using colony and 3D spheroid formation assays.The combination of TMZ and CTS substantially suppressed colony formation in LN229 and U87-MG cells (Fig. [fig_ref] Figure 3: Figure 3 [/fig_ref] and 3D spheroid formation (Fig. [fig_ref] Figure 3: Figure 3 [/fig_ref] compared with that of either treatment alone.
## Combined treatment with cts and tmz synergistically induced dna alkylation damage and apoptosis
Next, we determined the combined effect of TMZ and CTS treatment on GBM cells.We treated LN229 cells with vehicle, 4 μM CTS, 200 μM TMZ, or both CTS and TMZ for 72 h and detected the proportion of cells with DNA damage in each group.As shown in Fig. [fig_ref] Figure 4: Figure 4 [/fig_ref] , the significantly elongated tailings ratio of the combination treatment relative to either the control or single agent treatments suggests that more fragments of DNA damage were present after treatment with both CTS and TMZ.The inhibition of STAT3 activation (p-STAT3) reduces MGMT expression [bib_ref] STAT3 inhibition overcomes temozolomide resistance in glioblastoma by downregulating MGMT expression, Kohsaka [/bib_ref].As the nucleosome histone H2AX is rapidly phosphorylated at Serine 139 (γ-H2AX) at the double-strand break (DSB) site, γ-H2AX formation is commonly used to quantify DSBs.To confirm whether www.nature.com/scientificreports/ the combination of CTS and TMZ enhances the toxic effect of these drugs on tumor cells, western blot analysis was performed to detect the expression of these proteins.As shown in Fig. [fig_ref] Figure 4: Figure 4 [/fig_ref] , p-STAT3 (Tyr705) expression was negligible with CTS treatment alone, whereas the expression of γ-H2AX was slightly increased.However, after TMZ treatment, MGMT protein expression was significantly upregulated.Meanwhile, compared with TMZ single-agent treatment, MGMT expression was significantly decreased, and γ-H2AX expression was significantly increased with the CTS and TMZ combined treatment.TMZ-induced DNA damage may be blocked by STAT3, which is inhibited by CTS during the activation of MGMT expression.Finally, we observed the effect of combination therapy on apoptosis.As shown in Fig. [fig_ref] Figure 4: Figure 4 [/fig_ref] , apoptosis was significantly induced by CTS and TMZ combined treatment and compared with the TMZ single-agent.Therefore, CTS combined with TMZ induced a higher ratio of cell apoptosis.
## Cts reverses tmz resistance by targeting stat3 to suppress mgmt expression in tmz-resistant cell models
To confirm the promotional effect of CTS on TMZ-induced cytotoxicity, we conducted further studies using TMZ-resistant (TMZ-R) cell models.As mentioned previously, TMZ-resistant cell models were obtained by low-dose stimulation, and qRT-PCR and western blot analyses demonstrated that MGMT mRNA and protein expression was significantly increased (Fig. [fig_ref] Figure 5: Figure 5 [/fig_ref].Next, we compared the inhibitory effect of CTS combined with TMZ on the growth of wild-type and TMZ-resistant cells through cell viability experiments.We observed that CTS exhibited superior inhibition effect on both TMZ-resistant and wild-type cells.The combination of CTS and TMZ significantly enhanced the inhibiting effect of TMZ on TMZ-resistant cells, whereas the inhibition of TMZ on U251-MG_TMZ-R and U87-MG_TMZ-R cells dramatically reduced at the same concentration when compared to wild-type cells (Fig. [fig_ref] Figure 5: Figure 5 [/fig_ref].Next, we observed the DNA damage effects of CTS combined with TMZ on GBM cells.As shown in Fig. [fig_ref] Figure 5: Figure 5 [/fig_ref] , CTS treatment alone had no effect on DNA damage in U251-or U87-TMZ-R cells.With 200 µM TMZ treatment, DNA damage in U251 cells increased significantly, whereas that in U251-TMZ-R cells did not.Finally, western blot analysis was used to detect the expression of γ-H2AX and MGMT in wild-type and TMZ-resistant cells treated with CTS and TMZ.As shown in Fig. [fig_ref] Figure 5: Figure 5 [/fig_ref] , MGMT background expression in U251-TMZ-R cells was relatively high compared with that in U251-MG cells.TMZ stimulation further increased the MGMT protein expression.Conversely, MGMT expression in TMZ-resistant cells treated with TMZ and CTS was inhibited, which promoted an increase in TMZ-induced DNA damage.
# Discussion
Human GBM is one of the most common, aggressive, and malignant brain tumors of the central nervous system in adults.As the first-line option, TMZ-based chemotherapy is a routine and crucial therapy for GBM, followed by standard surgical excision.However, glioma cells eventually metastasize and develop chemoresistance, which is an obstacle to the therapeutic efficacy of TMZ in GBM.Therefore, the molecular mechanisms underlying the suppression of TMZ resistance must be elucidated to identify and develop potential novel molecular targets for GBM therapy.In the present study, we demonstrated that CTS treatment suppressed STAT3 activity and arrested GBM cell proliferation in a dose-and time-dependent manner.Given that STAT3 is a crucial target for a number of biological processes, inhibition of STAT3 activity mediated by CTS does not completely prevent cell proliferation.We investigated the possibility of enhancing the anti-glioma efficacy of TMZ by combination therapy, considering the significance of TMZ in glioma chemotherapy and the fact that CTS reduces tumor cell proliferation.Based on the synergy score evaluation, we identified the optimal concentrations of the antiglioma drugs CTS and TMZ.In this study, we demonstrated for the first time that CTS has a combined effect with TMZ, which effectively inhibits colony formation and tumor spheroid proliferation.We further confirmed that CTS, in combination with TMZ, had an enhanced therapeutic effect on LN229 and U87-MG GBM cells compared with that of CTS or TMZ treatment alone.In addition, CTS combined with TMZ effectively reduced the required dosage of TMZ.MGMT is a DNA repair protein that widely exists in organisms from bacteria to mammals, and can repair DNA alkylation damage caused by TMZ, leading to TMZ resistance in GBM cells.The heavy dependence of cancer cells on MGMT-mediated DNA repair makes the regulation of MGMT expression an attractive target for cancer therapy.In this study, DNA fragmentation was analyzed using a comet assay.The significantly elongated tailings ratio of combination treatment relative to control or single-agent treatment suggests an increased abundance of DNA-damaged fragments following treatment with both TMZ and CTS.Furthermore, the accumulation of DNA damage caused by the inhibition of MGMT expression led to the apoptosis of tumor cells.We also observed that low-dose CTS treatment resulted in a better inhibition efficiency of TMZ-R cells than in the control group.However, TMZ-R cells reversed their insensitivity to TMZ when co-treated with CTS.Our work validates the mechanism by which CTS reverses TMZ resistance and induces cell DNA damage in TMZ-R GBM cells by inhibiting STAT3 activation to decrease MGMT expression and enhance TMZ-induced MGMT-dependent DNA damage.
In conclusion, we found that CTS exhibited a time-and dose-dependent inhibition of glioma cell proliferation, and that the synergistic interaction between CTS and TMZ increased the curative effects of TMZ.The primary mechanism could be that CTS suppresses MGMT expression by inhibiting STAT3 activation, thus enhancing apoptosis induced on by TMZ alkylation.Although our findings point to the possible advantages of CTS and TMZ combination therapy, the comprehensive molecular modification of the STAT3 signaling pathway by CTS must be investigated to explore its application in other malignant diseases.
# Materials and methods
## Reagent preparation and storage
CTS and TMZ were purchased from MedChemExpress (MCE, New Jersey, USA).CTS was dissolved in 100% DMSO to prepare a stock solution at a concentration of 10 mM; TMZ was dissolved in 100% DMSO to prepare a stock solution at a concentration of 100 mM.Both were equipped and wrapped in a brown centrifugal tube for protection against light and stored at − 20 °C.
## Cell line and cell culture
The two typical human GBM cell lines LN229 and U87-MG were obtained from the Institute of Biochemistry and Cell Biology, Chinese Academy of Science.Cells were maintained in Dulbecco's Modified Eagle Medium (glucose 4.5 g/L; Gibco, NY, USA) with 10% Fetal Bovine Serum (Gibco, NY, USA) and penicillin (100 U/mL)streptomycin (100 μg/mL) in a humidified incubator at 37 °C with 5% CO 2 and authenticated using a short tandem repeat (STR) assay (Genetic Testing Biotechnology, Jiangsu, China).
## Cell viability assay
Cell viability was determined using Cell Counting Kit 8 (CCK-8) from Beyotime (Beyotime, Shanghai, China), stored at 4 °C.Following being seeded at a population density of 2 × 10 3 in 96-well plates for 24 h, the cells were treated for 72 h with either CTS, TMZ, or CTS + TMZ.Then, the cells were treated for 1 h at 37 °C with 10 µL of CCK-8.Utilizing a 450 nm absorbance measurement, the quantity of viable cells was ascertained.
## Establishment of tmz-resistant gbm cell line
The U251-MG and U87-MG TMZ-resistant cell line was developed by using the TMZ concentration gradient progressive methods as previous described [bib_ref] Inhibition of cyclin E1 overcomes temozolomide resistance in glioblastoma by Mcl-1 degradation, Liang [/bib_ref].Briefly, 1 × 10 5 /mL of U251-MG or U87-MG cells were inoculated in TMZ-free culture medium for 24 h until they got to the logarithmic phase.Then the culture medium was replaced with that containing low concentration (50 µM) TMZ for 72 h.Then the culture medium containing drugs and dead cells was discarded.Cells were collected and re-inoculated in TMZ-free culture medium to get recovered before the next TMZ treatment.After the cells got adjusted to the present TMZ treatment, the concentration of TMZ was increased in turn until the cells survived well and developed resistance to the 400 µM TMZ treatment.
## Quantitative real-time polymerase chain reaction (qrt-pcr)
Cellular RNA was isolated using TRIzol reagent (Invitrogen, CA, USA) and reverse transcribed to complementary DNA (cDNA) using a cDNA Synthesis Kit (Vazyme, Nanjing, China).For QPCR, we used GAPDH gene Forward primer 5′-GTC TCC TCT GAC TTC AAC AGCG-3′, GAPDH gene reverse primer 5′-ACC ACC CTG TTG CTG TAG CCAA-3′; MGMT gene forward primer 5′-ACC GTT TGC GAC TTG GTA CTT-3′, MGMT gene reverse primer 5′-GGA GCT TTA TTT CGT GCA GACC-3′; STAT3 gene forward primer 5′-CCT GCT AAA ATC AGG GGT CC-3′, Vol:.( 1234567890
## Colony formation assay and 3d spheroid formation assay
For colony formation assay, cells were treated with CTS, TMZ, or CTS + TMZ for 72 h following their seeding in 12-well plates at a population density of 200 per well for 4 days.Colonies were fixed with methanol and stained with crystal violet, following the manufacturer's protocol.The plating efficiency was determined by counting the number of colonies and was calculated as follows:
For the 3D spheroid formation assay, 200 cells per well were seeded in ultra-low-binding 96-well plates with a round bottom (Costar, MA, USA).After nine days of treatment with either CTS, TMZ, or CTS + TMZ, digital images of the spheroids were captured using a phase-contrast microscope (Leica, Weztlar, Germany).The volumes of the 3D spheroids were calculated using the formula "(V = 0.5 × Length × Width 2 )" based on the major and minor axial lengths (Length: major axial length; Width: minor axial length).
## Comet assay
For comet assay, cells were treated with CTS, TMZ, or CTS + TMZ for 72 h following their seeding in 6-well plates at a population density of 2 × 10 5 for 24 h.Then, cells were harvested, washed, and resuspended in phosphatebuffered saline (PBS) at a concentration of 1 × 10 6 cells/m.Subsequently, low melting agarose was dissolved in the cell suspension to generate a 0.75% agarose mixture.100 μL normal melting agarose and 100 μL cell-agarose mixtures were separately added to a fully frosted slide at 10 min intervals to form two layers and were incubated at 4 °C for 30 min.After removing the coverslip, the slide was incubated in lysis buffer (10 mM Tris, pH 10.0, 2.5 M NaCl, 0.1 M EDTA2Na, 10% DMSO, and 1% Triton X-100) at 4 °C overnight.The slide was subjected to electrophoresis in Mini Horizontal Cells (Bio-Rad, CA, USA) filled with ice-cold alkaline electrophoresis buffer (0.3 M NaOH, 1 mM EDTA) and ran at 300 mA and 25 V for 30 min.After soaking in neutralization buffer (0.4 M Tris-HCl, pH 7.5) for 5 min, the slide was counter-stained with 5 mg/mL of the nuclear stain DAPI.Images were acquired using a Leica Thunder microscope and analyzed using the Comet Score software (CASP, CASP-Lab).
## Flow cytometry
For flow cytometry assay, cells were treated with CTS, TMZ, or CTS + TMZ for 72 h following their seeding in 6-well plates at a population density of 2 × 10 5 for 24 h.Then, cells were harvested and washed with PBS.A total of 2 × 10 5 cells were added to 100 μL binding buffer to create a cell suspension, followed by the addition of 5 μL Annexin V and 10 μL propidium iodide (PI) in sequence.Once mixed, this preparation was incubated for 25 min at 4 °C.After the further addition of 200 μL of binding buffer, flow cytometry was performed, and cell apoptosis was determined.The extent of apoptosis was measured using an Annexin V-FITC apoptosis detection kit (Beyotime, Shanghai, China) and analyzed using flow cytometry software (Beckman, USA).
# Western blotting analysis
Following being seeded at a population density of 3 × 10 5 in 6-well plates for 24 h, the cells were treated for 72 h with either CTS, TMZ, or CTS + TMZ.Cells were lysed and homogenized in radioimmunoprecipitation assay (RIPA) buffer, and protein concentrations were determined using a BCA Protein Assay Kit (Thermo Fisher, CA, USA).Protein samples were separated using 10% sodium dodecyl sulfate-polyacrylamide gels and transferred to polyvinylidene fluoride membranes (Merck Millipore, MA, USA).The membranes were blocked with 3% bovine serum albumin for 1 h.The blots were incubated with specific primary antibodies overnight at 4 °C and then incubated with corresponding horseradish peroxidase-conjugated secondary antibodies for 1 h.The primary antibodies used were anti-phospho-STAT3 (Y705) (1:5000; catalog no.76315), anti-STAT3 (1:2000; catalog no.68153), anti-MGMT (1:2000; catalog no.39253), anti-H2AX (1:5000; catalog no.124781), anti-γH2AX (1:5000; catalog no.81299) (Abcam, MA, USA), and anti-GAPDH (1:4000; catalog no.10494-I-AP) (Proteintech, Chicago, USA).Finally, the protein bands were visualized using an ECL kit (Beyotime, Shanghai, China), and the density of the immunoreactive bands was analyzed using Imager software (Tanon, Shanghai, China).
# Statistical analysis
Statistical analysis was performed with Graph-Pad Prism7 (version 7.0; GraphPad Software, Inc., San Diego, CA, USA).The unpaired two-group comparison and multiple comparisons were made with the Student's t-test or one-way ANOVA, respectively.All data are presented as the mean ± standard deviation (mean ± SD).Differences between means were considered statistically significant when P < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001).The synergistic effect analysis of the two drug combinations was performed using Combinefit 2.02 software (Combinefit, Inc., San Diego, CA, USA).
[fig] Figure 1: Figure 1.CTS treatment inhibits cell growth by suppressing STAT3 activity.(A) Chemical structure of CTS.(B,C) LN229 and U87-MG human GBM cells were treated with a series of concentrations of CTS for 24 or 48 h, and cell viability was determined with a CCK-8 assay.(D,E) Western blot analysis showing protein expression in LN229 and U87-MG cells using the indicated antibodies.(F,G) LN229 and U87-MG cells were treated with vehicle control or 4 μM CTS for different periods, and cell viability was determined with a CCK-8 assay.(H,I) LN229 and U87-MG cells were treated with vehicle control or 4 μM CTS for different periods, and protein expression was detected with the indicated antibodies (***P < 0.001, **P < 0.01, *P < 0.05). [/fig]
[fig] Figure 2: Figure 2. Characterization of the anti-glioma activity of a combination treatment of CTS and TMZ.(A,B) The IC50 value of CTS and TMZ as single-agent treatments of LN229 and U87-MG human GBM cells was verified with a CCK-8 assay.(C,D) Synergy plots generated by Combenefit depicting the interaction between CTS and TMZ.An analysis of interactions provided HSA and Bliss values (n = 3, technical replicates), indicating synergistic efficacy as calculated from expected and observed growth inhibition.HSA and Bliss values > 0 indicate synergistic effects.(E,F) Single and combinatorial titration of CTS and TMZ in a three-day growth assay of LN229 and U87-MG cell lines (***P < 0.001, **P < 0.01, *P < 0.05). [/fig]
[fig] Figure 3: Figure 3. Low-dose CTS and TMZ combination treatment suppresses cell proliferation.(A,B) IC50 values of TMZ treatment for 72 h with or without CTS of LN229 and U87-MG GBM human cells.(C,D) Representative images (upper panels) and quantification (lower panels) of the colony formation assays of LN229 and U87-MG cells treated with the indicated concentrations of TMZ and CTS.(E,F) Representative images (left panels) and quantification (right panels) of 3D spheroids formation assays of LN229 and U87-MG cells treated with the indicated concentrations of TMZ and CTS (***P < 0.001, **P < 0.01, *P < 0.05). [/fig]
[fig] Figure 4: Figure 4. Combining CTS and TMZ treatments synergistically induce oxidative DNA damage and apoptosis.(A) Representative images (left panels) and quantification (right panels) of comet assays of LN229 human GBM cells treated with the indicated concentrations of TMZ and CTS.(B) The combination of CTS and TMZ treatments synergistically induced DNA damage, as confirmed by western blot analysis.(C) Representative images (left panels) and quantification (right panels) of apoptosis following the combination of CTS and TMZ treatments in LN229 cells (***P < 0.001, **P < 0.01, *P < 0.05). [/fig]
[fig] Figure 5: Figure 5. Combined treatment of CTS and TMZ reversed TMZ resistance of human GBM cells.(A) Validation of TMZ-resistant cell models by qRT-PCR.(B) Validation of TMZ-resistance cell models by western blot analysis.(C,D) Cell viability of each treatment with the indicated concentrations of TMZ and CTS in TMZresistant cell models.(E) Representative images (upper panels) and quantification (lower panels) of comet assays in TMZ-resistant cells treated with the indicated concentrations of TMZ and CTS.(F) The combination of CTS and TMZ treatments synergistically induced DNA damage and reversed TMZ resistance, as confirmed by western blot analysis (***P < 0.001, **P < 0.01, *P < 0.05). [/fig]
|
10.3389/fcell.2021.705196
|
CCBY
|
8215651
|
34164407
|
s2orc_pubmed_articles
|
Editorial: Emerging Roles of TRP Channels in Brain Pathology
Editorial on the Research TopicEmerging Roles of TRP Channels in Brain PathologyThe mammalian transient receptor potential (TRP) ion channel superfamily comprises six subfamilies, TRPC (canonical), TRPV (vanilloid), TRPM (melastatin), TRPA (ankyrin), TRPP (polycystin), and TRPML (mucolipin)(Ramsey et al., 2006;Venkatachalam and Montell, 2007). TRP channels are tetrameric and each subunit contains intracellular N-and C-termini and six membrane-spanning segments, with the fifth and sixth segments and the re-entrant loop between them forming the ion-conducting pore (Cao, 2020). They function as non-selective cation channels, with prominent Ca 2+ permeability for most of them, and are activated by diverse physical, chemical and biological stimuli. Their Ca 2+ permeability, poly-modal activation and wide expression place these channels in a vital position mediating Ca 2+ signalling in a range of physiological processes. Not surprisingly, accumulating evidence supports an important role for the TRP channels in the pathogenesis of numerous diseases(Nilius et al., 2007). Many TRP channels are expressed in the brain. This Research Topic, including 14 review and original research articles, offers critical and new insights into the role of TRP channels, particularly the Ca 2+ -permeable ones, in multiple brain pathologies.TRPC IN ISCHEMIC BRAIN DAMAGEBrain is highly vulnerable to damage by ischemia, if severe or lasting, and reperfusion after transient ischemia. Jeon et al. have critically evaluated the literature and also presented their recent study regarding the role of TRPC channels in ischemic brain damage. A complex role for the TRPC channels is emerging from studies, using middle carotid artery occlusion followed by reperfusion (MCAO/R), an in vivo model of ischemic stroke, and neuronal death induced by oxygen and glucose deprivation followed by reoxygenation (OGD/R), an in vitro model of ischemia/reperfusion, in combination with using transgenic knockout (KO) mice and neurons or brain slices derived from KO mice or using pharmacological interventions. TRPC channels, including TRPC3, TRPC4, TRPC6, and TRPC7, play a significant role in mediating ischemic brain damage. TRPC1 was proposed to act as a protective mechanism against ischemic brain damage via inhibiting generation of reactive oxygen species (ROS). However, Jeon et al. showed that TRPC1-KO attenuated OGD/R-induced neuronal death.Neuroinflammation is another mechanism for ischemic brain damage and many other brain pathologies. Liu et al. showed that inhibition or depletion of TRPC6 suppressed OGD/R-induced Ca 2+ response, production of interleukin (IL)-1β and IL-6, two neurotoxic proinflammatory cytokines, caspase-3 activation and apoptosis in astrocytes. Inhibition of TRPC6 also attenuated
## Editorial on the research topic
## Emerging roles of trp channels in brain pathology
The mammalian transient receptor potential (TRP) ion channel superfamily comprises six subfamilies, TRPC (canonical), TRPV (vanilloid), TRPM (melastatin), TRPA (ankyrin), TRPP (polycystin), and TRPML (mucolipin) . TRP channels are tetrameric and each subunit contains intracellular N-and C-termini and six membrane-spanning segments, with the fifth and sixth segments and the re-entrant loop between them forming the ion-conducting pore. They function as non-selective cation channels, with prominent Ca 2+ permeability for most of them, and are activated by diverse physical, chemical and biological stimuli. Their Ca 2+ permeability, poly-modal activation and wide expression place these channels in a vital position mediating Ca 2+ signalling in a range of physiological processes. Not surprisingly, accumulating evidence supports an important role for the TRP channels in the pathogenesis of numerous diseases. Many TRP channels are expressed in the brain. This Research Topic, including 14 review and original research articles, offers critical and new insights into the role of TRP channels, particularly the Ca 2+ -permeable ones, in multiple brain pathologies.
## Trpc in ischemic brain damage
Brain is highly vulnerable to damage by ischemia, if severe or lasting, and reperfusion after transient ischemia. Jeon et al. have critically evaluated the literature and also presented their recent study regarding the role of TRPC channels in ischemic brain damage. A complex role for the TRPC channels is emerging from studies, using middle carotid artery occlusion followed by reperfusion (MCAO/R), an in vivo model of ischemic stroke, and neuronal death induced by oxygen and glucose deprivation followed by reoxygenation (OGD/R), an in vitro model of ischemia/reperfusion, in combination with using transgenic knockout (KO) mice and neurons or brain slices derived from KO mice or using pharmacological interventions. TRPC channels, including TRPC3, TRPC4, TRPC6, and TRPC7, play a significant role in mediating ischemic brain damage. TRPC1 was proposed to act as a protective mechanism against ischemic brain damage via inhibiting generation of reactive oxygen species (ROS). However, Jeon et al. showed that TRPC1-KO attenuated OGD/R-induced neuronal death.
Neuroinflammation is another mechanism for ischemic brain damage and many other brain pathologies. Liu et al. showed that inhibition or depletion of TRPC6 suppressed OGD/R-induced Ca 2+ response, production of interleukin (IL)-1β and IL-6, two neurotoxic proinflammatory cytokines, caspase-3 activation and apoptosis in astrocytes. Inhibition of TRPC6 also attenuated MCAO/R-induced caspase-3 activation, elevated level of IL-1β and IL-6 in the peri-infarct areas, and infarction in mice. These results support a critical role for TRPC6 in astrocytes in mediating neuroinflammation and ischemic brain damage.
## Trp in neurodegenerative and neurological diseases
Alzheimer's disease (AD), Parkinson's diseases (PD), Huntington's disease, amyotrophic lateral sclerosis and epilepsy represent the most common neurodegenerative and neurological diseases. Extensive research efforts have been devoted to exploring the TRP channels in the pathogenesis of these conditions. Lee et al. have provided a concise overview of the potential involvement of the TRP channels in AD (TRPC1, TRPC6, TRPV, TRPM2, TRPM7, and TRPML1), PD (TRPC1, TRPC4, TRPC5, TRPM7, and TRPML1), Huntington's disease (TRPC1 and TRPC5), amyotrophic lateral sclerosis (TRPC4, TRPM2, TRPM3, TRPM7, and TRPML1) and epilepsy (TRPC4, TRPC5, TRPV4, TRPM7, and TRPA1). They further elaborated diverse TRP-mediated Ca 2+ -dependent downstream signalling pathways to the associated pathologies. Vaidya and Sharma have drawn their attention to the TRP channels expressed in substantial nigra pars compacta and other brain areas affected in PD. TRPC1 was shown to protect PD-related neuronal death. Intriguingly, both activation and inhibition of TRPV1, reported by different studies, improved motor function through regulation of distinctive molecular and cellular mechanisms. Oxidative stress, due to accumulation of high levels of ROS, is a pathological factor as well as a conspicuous pathological feature in PD. Vaidya and Sharma also discussed the role of TRPM2 and TRPM7, both known to be sensitive to activation by ROS, in mediating PD-related neuronal death.
## Trpv in bbb dysfunction and related brain pathologies
Endothelial cells play a key role in forming blood brain barrier (BBB), and BBB dysfunction results in infiltration of peripheral immune cells into the brain to exacerbate neuroinflammation and thereby ischemic stroke and neurodegenerative diseases. TRPV4 is known to regulate endothelial barrier function via mediating Ca 2+ influx into endothelial cells. Consistently, Rosenkranz et al. demonstrated that pharmacological inhibition of TRPV4 in mouse brain microvascular endothelial cells improved endothelial barrier function. Such an effect was however obliterated in endothelial cells after exposure to interferon-γ and tumour necrosis factor-α that down-regulated the TRPV4 expression. In mice, TRPV4-KO failed to prevent BBB dysfunction and associated experimental autoimmune encephalomyelitis, a model of multiple sclerosis, or MCAO/R-induced brain damage, suggesting loss of TRPV4mediated regulation of BBB function under inflammation. Luo et al. examined and revealed a different profile of TRPV1-4 expression in brain endothelial cells of rat and human origins, ranking from high to low: TRPV4 > TRPV2 > TRPV3 > TRPV1 in rat endothelial cells, and TRPV2 >> TRPV4 > TRPV1 > TRPV3 in human endothelial cells. Thus, TRPV4 and TRPV2 represent the predominant TRPV in human and rat brain endothelial cells, respectively. Such species difference may complicate testing drugs using rat models of BBB dysfunction and related brain pathologies.
## Trpm3 in developmental disorders
TRPM3 is activated by endogenous neurosteroid pregnenolone sulphate. TRPM3 expression is documented in several brain regions, including hippocampus and choroid plexus, and in cerebellar Purkinje neurons and oligodendrocytes. Held and Tóth have reviewed the potential role of TRPM3 in brain physiological and pathophysiological functions, and highlighted genetic alterations in the TRPM3 gene in patients with development and intellectual disabilities. Particularly interesting is the identification of de novo gain-of-function mutations in patients with developmental and epileptic encephalopathies. Deletions in the TRPM3 gene were found in patients with Kabuki syndrome, a multisystem disorder including intellectual disability, and in autism patients. However, it remains unclear how alteration of TRPM3 affects brain development and other functions.
## Trp in psychiatric disorders
Psychiatric disorders such as depression and anxiety disorders are caused by diverse and complex factors, both genetic and environmental. As proposed in neurodegenerative diseases, oxidative stress is a critical pathological factor, and Ca 2+ signalling is disrupted, in psychiatric disorders. Nakao et al. have overviewed the redox-sensitive Ca 2+ -permeable TRP channels in neurons and glial cells, including TRPC4, TRPC5, TRPV1, TRPM2, and TRPA1, in regulating cell functions. They have provided a concise summary of the studies that showed using a battery of behaviour tests that anxiety-like behaviours in mice were attenuated by genetic deletion (TRPC4, TRPC5, TRPV1, TRPM2, or TRPA1) or pharmacological inhibition (TRPC4/TRPC5). They also discussed the possible mechanisms by which these channels mediate oxidative stressinduced Ca 2+ signalling and subsequent alterations in neuronal connectivity, synaptic plasticity and glial cell function, leading to psychiatric disorders.
## Trp in chronic pain and its inhibition by resolvins
Chronic pain occurs as a result of neuronal tissue inflammation and/or damage. Increased expression and/or activation of TRP channels in nociceptive neurons can enhance the excitability of nociceptive neurons and thereby intensify the pain signals. Resolvins are a class of lipid mediators generated during the resolution phase of acute inflammation. Roh et al. discussed the studies showing analgesic effects of different resolvins against inflammatory pain, through activating distinct cognate receptors to inhibit nociceptive TRP channels, including TRPV1, TRPV3, TRPV4, and TRPA1. The role of such inhibitory mechanisms in neuropathic pain is unknown.
## Trp in brain tumours
Malignant glioma including glioblastoma is the most common group of primary brain tumours and exhibits resistance to treatment and high recurrence. Chinigò et al. have conveyed the literature informing the expression of TRP channels and their potential role in brain tumours. The expression in glioma tissues and cells was shown to be upregulated (TRPC1, TRPC6, TRPV4, TRPM2, TRPM7, TRPM8, and TRPML2), or down-regulated (TRPV1, TRPV2, and TRPML1). Such alterations in some cases exhibited correlations with glioma progression and overall patient survival. Moreover, some of these channels (TRPC1, TRPC6, TRPV4, TRPM7, TRPM8, and TRPML2) appear protumorigenic and, conversely, others (TRPV1, TRPV2, TRPM2, TRPM3, TRPA1, and TRPML1) are anti-tumorigenic. Chinigò et al. have proposed diverse signalling pathways, downstream of TRP channel-mediated elevation in intracellular Ca 2+ , in the regulation of glioma cell proliferation, migration and invasion as well as cell death.
## Trp in regulating ca 2+ and h + homeostasis and interacting with other signalling mechanisms
Disruption in Ca 2+ homeostasis has been alluded by multiple articles as an important mechanism in the pathogenesis of brain pathologies. Plasma membrane Ca 2+ -ATPase drives Ca 2+ efflux and H + influx and thus plays a vital role in modulating intracellular Ca 2+ and H + homeostasis. Hwang et al. have reviewed the role in neurodegenerative diseases of the plasma membrane Ca 2+ -ATPase in concerted actions with Ca 2+permeable TRP channels in altering Ca 2+ and H + homeostasis in neuronal and glial cells.
Hermann et al. investigated contribution of dinucleotide nicotinic acid adenine phosphate (NAADP)-induced Ca 2+ release in mouse hippocampal neurons to the Ca 2+ response evoked by glutamate, the major excitatory neurotransmitter in the brain. They proposed that NAADP induces Ca 2+ release from intracellular acidic stores by activating TRPML1 and two-pore channels, and also from ER by activating ryanodine receptors, and showed that inhibition of NAADP signalling reduced glutamate-evoked Ca 2+ response, particularly the sustained component. The study demonstrated substantial contribution of NAADP to glutamate-induced Ca 2+ signalling in hippocampal neurons.
The activity of small Rho GTPases is critical for actin-based cytoskeletal remodelling. Ca 2+ is an intracellular signal upstream of Rho GTPases. Lavanderos et al. have introduced the role of Rho GTPases in the regulation of axon growth, dendritic spine development, synapse formation, glial cell migration, and endothelial permeability, which are vital in sustaining brain structure and function. They further presented the potential Rho GTPase-dependent mechanisms that mediate Ca 2+ -permeable TRP channels in various brain conditions, including AD (TRPC6 and TRPV1), ischemic bran damage (TRPM7), glioma and neuroblastoma (TRPC6, TRPV1, TRPM7, and TRPM8).
Manchanda et al. extended their previous work to study TRPV1 in regulating the generation of endocannabinoids. Treatment of human embryonic kidney 293 cells expressing TRPV1 with capsaicin increased the levels of 2-acyl glycerols and reduced the levels of N-acyl ethanolamines, depending upon temperature. These results support the interesting notion that the TRPV1 and the receptor for endocannabinoids cross-talk in response to changes in temperature.
Collectively, the insightful evaluation of the literature and the new information presented in this Research Topic have evolved a better understanding of the TRP channels in the pathogenesis of multiple brain pathologies, and also raised many outstanding questions that need further researches in order to gain comprehensive and mechanistic insights into these pathological conditions and provide the proof of concept of targeting the TRP channels and/or related mechanisms as potential therapeutic strategies.
Frontiers in Cell and Developmental Biology | www.frontiersin.org
June 2021 | Volume 9 | Article 705196
AUTHOR CONTRIBUTIONSAll authors have contributed to the writing, the revision of this Editorial Article, and approved it for publication.
|
10.1007/s40980-021-00102-w
|
CCBY
|
10438900
|
37600470
|
s2orc_pubmed_articles
|
Conflict and Climate Factors and the Risk of Child Acute Malnutrition Among Children Aged 24–59 Months: A Comparative Analysis of Kenya, Nigeria, and Uganda
Acute malnutrition affects a sizeable number of young children around the world, with serious repercussions for mortality and morbidity. Among the top priorities in addressing this problem are to anticipate which children tend to be susceptible and where and when crises of high prevalence rates would be likely to arise. In this article, we highlight the potential role of conflict and climate conditions as risk factors for acute malnutrition, while also assessing other vulnerabilities at the individual-and household-levels. Existing research reflects these features selectively, whereas we incorporate all the features into the same study. The empirical analysis relies on integration of health, conflict, and environmental data at multiple scales of observation to focuses on how local conflict and climate factors relate to an individual child's health. The centerpiece of the analysis is data from the Demographic and Health Surveys conducted in several different cross-sectional waves covering 2003-2016 in Kenya, Nigeria, and Uganda. The results obtained from multi-level statistical models indicate that in Kenya and Nigeria, conflict is associated with lower weight-forheight scores among children, even after accounting for individual-level and climate factors. In Nigeria and Kenya, conflict lagged 1-3 months and occurring within the growing season tends to reduce WHZ scores. In Uganda, however, weight-for-height scores are primarily associated with individual-level and household-level conditions and demonstrate little association with conflict or climate factors. The findings are valuable to guide humanitarian policymakers and practitioners in effective and efficient targeting of attention, interventions, and resources that lessen burdens of acute malnutrition in countries prone to conflict and climate shocks. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article'
# Introduction
Worldwide, nearly half of the deaths of children below 5 years of age result from malnutrition . Of particular concern is acute malnutrition, characterized by a sudden, rapid decrease in caloric intake resulting in a reduced weight-for-height z-score (WHZ) or in wasting (WHZ < −2) (UNICEF/WHO/World Bank, 2019). While low WHZ is broadly associated with food insecurity and insufficient caloric and nutritional intake, research has shown that conflict and climate conditions may increase the risk of malnutrition by acting through factors (e.g., reduced agricultural production, reduced access to markets and health/humanitarian aid) more directly related to food security [bib_ref] Drought, conflict and children's undernutrition in Ethiopia 2000-2013: A meta-analysis, Delbiso [/bib_ref] [bib_ref] Systematic review of current efforts to quantify the impacts of climate change..., Phalkey [/bib_ref] [bib_ref] Characterizing the contribution of high temperatures to child undernourishment in Sub-Saharan Africa, Baker [/bib_ref] [bib_ref] Food security and conflict: Empirical challenges and future opportunities for research and..., Martin-Shields [/bib_ref]. Researchers generally evaluate how food security and malnutrition are affected by conflict conditions (e.g., [bib_ref] Armed conflicts and food insecurity: Evidence from Boko Haram's attacks, George [/bib_ref] [bib_ref] The impact of the Boko Haram insurgency in Northeast Nigeria on childhood..., Dunn [/bib_ref] or by climate conditions (e.g., [bib_ref] Multilevel and spatial analyses of childhood malnutrition in Uganda: Examining individual and..., Amegbor [/bib_ref] [bib_ref] Characterizing the contribution of high temperatures to child undernourishment in Sub-Saharan Africa, Baker [/bib_ref] [bib_ref] A multi-country assessment of factors related to smallholder food security in varying..., Niles [/bib_ref] [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref] and rarely directly consider situations where conflict events and climate extremes occur together [bib_ref] Drought, conflict and children's undernutrition in Ethiopia 2000-2013: A meta-analysis, Delbiso [/bib_ref] [bib_ref] Malnutrition and conflict in Eastern Africa: Impacts of resource variability on human..., Rowhani [/bib_ref] provide notable exceptions).
These distinct areas of research use a range of different methods and datasets to explore the idea that climate or conflict factors disrupt the food system and may lead to an increase in household-level food insecurity and malnutrition. However, it is not clear what happens to food security and child health when a poor agricultural season occurs at the same time and in about the same place as a violent conflict event [bib_ref] The impact of armed civil conflict on household welfare and policy, Justino [/bib_ref]. In other words, when violent conflict occurs during a poor growing season does that strain the food system and increase food insecurity even further than if either a drought or a conflict occurred alone? Furthermore, can we gain insight into the spatial and temporal variation in acute childhood malnutrition-an outcome of food insecurity and an important underlying component of mortality-when we consider the relationship between WHZ, armed conflict, and local climate factors together?
In this analysis, we evaluate the relationship between individual-level WHZ and recent community-level climate conditions, growing season conditions, and regional-level conflict conditions. In an effort to advance understanding of how these relationships may vary depending on the setting, we compare results across three sub-Saharan African countries with unique histories of conflict: Nigeria, Kenya, and Uganda. The countries share important characteristics relevant to this analysis, including relatively high and persistent rates of acute malnutrition, as well as spatially varying land use and climate and conflict conditions. We conduct a comparative analysis across and between countries to allow for discussion of both the general trends that might explain variation in WHZ while exploring context-specific relationships between conflict, climate and child health.
To conduct this analysis, we use several rounds of individual-level, spatially referenced cross-sectional survey data from recently collected Demographic and Health Surveys (DHS), providing large and nationally representative samples of WHZ. Violent conflict data come from the Uppsala Conflict Data Program (UCDP) and capture sub-national, regional-level lethal events in the context of violent conflict involving organized armed actors (state and/or non-state forces). Climate data come from several different sources capable of capturing relatively fine-scale variation in conditions over time and space. Regression analyses that consider a range of individual-and household-level factors in addition to the climate and conflict conditions are estimated for each country.
# Background
## Acute malnutrition
Acute malnutrition occurs because of sudden and rapid change in nutrition and caloric intake and can be identified through wasting or a reduction in a child's WHZ. According to recent global estimates, 17 million children under age 5 were categorized with severe acute malnutrition and another 32 million were categorized with moderate acute malnutrition (UNICEF/WHO/World Bank, 2019). 1 Differing from other forms of malnutrition (e.g., stunting), treatments exist that can move children out of the wasted categorization and into a healthy weight for their height (UNICEF/WHO/World Bank, 2019). Because wasting increases mortality risk, but can also be successfully treated, researchers and policymakers continue to work to identify different risk factors associated with wasting to support effective interventions. Through empirical research and policy efforts, a number of broadly relevant and widely used frameworks have emerged to help identify key areas for interventions [bib_ref] Severe childhood malnutrition, Bhutta [/bib_ref]. In particular, the popular framework of the United Nations Children's Fund (UNICEF) organizes factors into three categories: (1) immediate causes involving dietary intake and disease at an individual level, (2) underlying causes related to food security, care practices, and hygiene environments at household or community levels, and (3) basic causes such as political and economic conditions, land use, and education, at sub-national regional, country or international levels (UNICEF/WHO/World Bank, 2019).
While the UNICEF framework reflects an integration of different factors across scales, existing research has typically focused either on more micro-level factors (as in categories 1 and 2) or on macro-level factors (as in category 3). Projects that focus primarily on the macro-level patterns risk over-generalizing and may miss important within-group heterogeneity [bib_ref] Virtue and vulnerability: Discourses on women, gender and climate change, Arora-Jonsson [/bib_ref] [bib_ref] Climate change through a poverty lens, Hallegatte [/bib_ref] , while those, often smaller-scale, studies that focus primarily on individual-level variation may struggle to identify the underlying and basic factors associated with child health and to situate their findings beyond the specific research setting [bib_ref] Path analyses of risk factors for linear growth faltering in four prospective..., Prado [/bib_ref] [bib_ref] Systematic review of current efforts to quantify the impacts of climate change..., Phalkey [/bib_ref]. A rapidly emerging body of literature aims to consider the individual within context, with attention to how contextual factors measured at different spatial scales may be useful for understanding the complex and multi-scalar processes underlying individual-, community-and regional-level health variation (e.g., [bib_ref] Multilevel and spatial analyses of childhood malnutrition in Uganda: Examining individual and..., Amegbor [/bib_ref] [bib_ref] Drought, conflict and children's undernutrition in Ethiopia 2000-2013: A meta-analysis, Delbiso [/bib_ref] [bib_ref] Malnutrition and conflict in Eastern Africa: Impacts of resource variability on human..., Rowhani [/bib_ref] [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref]. This paper seeks to contribute to this emerging literature.
## Conflict events, climate conditions and child health
Conflict and climate conditions are regularly linked to food security, nutrition, and health outcomes, including wasting, using the UNICEF and other related frameworks [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref] [bib_ref] Systematic review of current efforts to quantify the impacts of climate change..., Phalkey [/bib_ref] [bib_ref] Empirical studies of factors associated with child malnutrition: Highlighting the evidence about..., Brown [/bib_ref] [bib_ref] The impact of armed civil conflict on household welfare and policy, Justino [/bib_ref]. The specific pathways and linkages explored when evaluating these different external factors and outcomes vary because of different data, analytic goals, and disciplinary perspectives. In general, researchers propose that adverse climate conditions reduce cropped and harvested area and also reduce agricultural yields. Conflict conditions may create significant challenges when individuals and families try to buy or sell food and livestock from local markets [bib_ref] The impact of armed civil conflict on household welfare and policy, Justino [/bib_ref]. In terms of specific impacts on child health, researchers find that warmer, dryer climate conditions in sub-Saharan Africa are generally associated with adverse child health outcomes in terms of wasting, stunting, or related nutrition outcomes [bib_ref] Mapping the effects of drought on child stunting, Cooper [/bib_ref] [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref] [bib_ref] Child health outcomes in sub-Saharan Africa: A comparison of changes in climate..., Davenport [/bib_ref]. Empirical studies have detected greater rates of chronic malnutrition among children in settings affected by armed conflict in several African countries, including Burundi [bib_ref] Health and civil war in rural Burundi, Bundervoet [/bib_ref] , Ethiopia and Eritrea [bib_ref] Wars and child health: Evidence from the Eritrean-Ethiopian conflict, Akresh [/bib_ref] , Cote d'Ivoire [bib_ref] Armed conflict, household victimization, and child health in Côte d'Ivoire, Minoiu [/bib_ref] , and Rwanda [bib_ref] Civil war, crop failure, and child stunting in Rwanda, Akresh [/bib_ref] and increased risk of acute malnutrition in Nigeria [bib_ref] The impact of the Boko Haram insurgency in Northeast Nigeria on childhood..., Dunn [/bib_ref] and Somalia [bib_ref] Conflict in Somalia: Impact on child undernutrition, Kinyoki [/bib_ref]. Research of armed conflict and agricultural labor demands shows that separately these factors may adversely impact the care environment with a negative impact on food and nutrition outcomes [bib_ref] The mental health of children affected by armed conflict: Protective processes and..., Betancourt [/bib_ref] [bib_ref] Development consequences of armed conflict, Gates [/bib_ref] [bib_ref] Armed conflict and birth weight: Evidence from the al-Aqsa Intifada, Mansour [/bib_ref] [bib_ref] Stunted from the start: Early life weather conditions and child undernutrition in..., Randell [/bib_ref] [bib_ref] Climatic conditions and infant care: Implications for child nutrition in rural Ethiopia, Randell [/bib_ref].
Depending on the type of data used and the specific research design, considering placebased factors related to conflict and climate helps to situate individuals within their unique contexts, as in the UNICEF structure, while also allowing for discussion of relationships between variables [bib_ref] Multilevel and spatial analyses of childhood malnutrition in Uganda: Examining individual and..., Amegbor [/bib_ref] [bib_ref] Path analyses of risk factors for linear growth faltering in four prospective..., Prado [/bib_ref]. Integration across scales implies that not all households or individuals exposed to the same contexts or macro-level conditions will necessarily experience the same health outcomes. This approach also suggests that how factors are defined and measured in each level of the UNICEF framework may also be dependent on the place and time [bib_ref] The basic determinants of malnutrition: Resources, structures, ideas and power, Harris [/bib_ref] [bib_ref] Multilevel and spatial analyses of childhood malnutrition in Uganda: Examining individual and..., Amegbor [/bib_ref] [bib_ref] Malnutrition and conflict in Eastern Africa: Impacts of resource variability on human..., Rowhani [/bib_ref]. What this means in practice is that the way conflict and climate factors relate to individual-level WHZ and wasting outcomes may vary over space and time and may be exacerbated or mitigated by dynamic conditions (e.g., disease or available resources).
One of the ongoing discussions in climate-health and conflict-health research is using appropriate spatial and temporal aggregation strategies. In fact, related research uses vastly different aggregation approaches when considering these exposures (see [bib_ref] Systematic review of current efforts to quantify the impacts of climate change..., Phalkey [/bib_ref] [bib_ref] Climate-conflict research: Some reflections on the way forward, Buhaug [/bib_ref] [bib_ref] Top 10 research priorities in spatial lifecourse epidemiology, Jia [/bib_ref] [bib_ref] Malnutrition and conflict in Eastern Africa: Impacts of resource variability on human..., Rowhani [/bib_ref] [bib_ref] Drought, conflict and children's undernutrition in Ethiopia 2000-2013: A meta-analysis, Delbiso [/bib_ref]. Therefore, ambiguity still exists about the ideal time frame for considering climate or conflict impacts on child health. Recent efforts in a small selection of countries have highlighted that for food security outcomes, locally occurring slow-onset events like drought will generally begin to modify household consumption and impact some nutrition outcomes within about 5-11 months after the growing season. For rapid-onset disasters (like earthquakes), the timing and spatial footprint of the impact will vary based on magnitude of the event and the impact on local infrastructure and the presence of factors like humanitarian aid. The impact from violent conflict on individuallevel nutrition may occur without much of a temporal lag but will depend on local and individual-level vulnerabilities. 2 In all cases, timing, spatial extent, and local vulnerabilities are important to consider to advance the scientific understanding of the processes underlying acute malnutrition and to help identify the key windows for humanitarian interventions to reduce morbidity and mortality among those with the greatest needs.
Here we consider climate factors at the scale of the community (10 km)-a finer-scale of analysis that acknowledges that spatial heterogeneity in rainfall, temperature and growing season conditions in the three countries. Much of the related research uses significantly coarser data, approximately 5-10 times coarser than our climate measures (e.g., [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref]. We also employ a relatively extreme measure of conflict: violent armed conflict associated with a lethal event at the regional-level scale (subnational administrative level 1). Other approaches used in the literature vary dramatically and include considering all reported events, regardless of magnitude, using binary presence-absence indicators (e.g., [bib_ref] Drought, conflict and children's undernutrition in Ethiopia 2000-2013: A meta-analysis, Delbiso [/bib_ref] [bib_ref] The impact of the Boko Haram insurgency in Northeast Nigeria on childhood..., Dunn [/bib_ref] versus those approaches that focus only on significant levels of battle-related deaths (> 25 per annually) [bib_ref] Malnutrition and conflict in Eastern Africa: Impacts of resource variability on human..., Rowhani [/bib_ref]. We choose a measure of conflict that falls somewhere between these two extremes-at least one fatality from a violent event-and count the number of these events within the region. The reason for this measure is because armed conflict leading to a lethal event may be more likely to change behaviors and access to agricultural plots and markets, and result in reduced food availability or access. Using an event count, rather than a binary (yes/no) indicator may also allow us to consider the intensity of exposure to conflict in an area as well.
In terms of the temporal dimension, most variables are collected at the time of survey for each individual child (between ages 24-59 months of age). Given the potential for temporal lags between conflict events or the growing season and malnutrition outcomes, we match children to local conditions with attention to specific time frames. WHZ is sensitive to very recent changes and children are able to regain weight lost, we focus on conflict conditions that occur relatively close in time to the survey date (within 6 months). We further explore the conflict-WHZ relationship using two different time lags-1-3 month and 4-6 month lags-of the count of events of violent conflict within the region. The idea is that impacts on acute malnutrition of violent conflict could appear very soon after the event or after a short delay if food and health services are interrupted for several months. We also consider the quality of the most recent growing season as part of the potential for a sub-optimal growing season to increase food insecurity. Finally, because climate conditions like flooding and high temperatures can have immediate impacts on WHZ [bib_ref] Drought, conflict and children's undernutrition in Ethiopia 2000-2013: A meta-analysis, Delbiso [/bib_ref] [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref] , we include average climate conditions of the 3 months before the survey. Our approach aims to advance discussion of the individual-level vulnerabilities, specifically with how recent conditions relate to acute malnutrition, in a context of climate change and in the presence of violent conflict. Including factors at different scales and time periods provides insight into the questions of which sub-national geographic areas-and within those areas, which specific households and children-are more likely to experience acute malnutrition.
## Setting
In this study, we include three African countries: Kenya, Nigeria, and Uganda. In all three countries, child malnutrition, food security, and land use vary spatially [bib_ref] Trends in socio-economic inequalities in child undernutrition: Evidence from Nigeria demographic and..., Akombi [/bib_ref] [bib_ref] Multilevel and spatial analyses of childhood malnutrition in Uganda: Examining individual and..., Amegbor [/bib_ref] [bib_ref] Child malnutrition and climate in Sub-Saharan Africa: An analysis of recent trends..., Grace [/bib_ref]. Each country has a history of armed conflict, exhibits a food system significantly dependent on locally grown, rain-fed crops, and faces persistent child malnutrition. In Nigeria, the southern regions are hot and wet most of the year with a long wet-season, whereas regions in the north having an extended dry season from October to April. In Kenya, the long and short rains characterize the highly populated temperate subtropical climate in the southwest highlands, and a much dryer climate in the north and east. Similar to Kenya, Uganda has two rainy seasons, but is in general semi-humid with some regions receiving up to 2200 mm of rain (see FEWS NET livelihood zone descriptions for more details for within-country variation).
Conflict also varies spatially and temporally within each country-with the armed conflict events observed in this study outcomes of different political events, state-sponsored violence and within-country dynamics between groups (see [bib_ref] Climate change, violent conflict and local institutions in Kenya's drylands, Adano [/bib_ref] Kenya has been comparatively peaceful with some recurrent conflict events in the North and Northeast . In Uganda-especially the North and Eastern areas of the country-significant violent conflict has been present since the 1960s up through at least 2010 [bib_ref] Struggling over land in post-conflict Uganda, Kandel [/bib_ref]. In Nigeria, the year 2000 brought a shift in political leadership and increasing tension between religious groups resulting in increasing violent conflict events across much of the country with Boko Haram becoming more violent towards civilians beginning about 2010 [bib_ref] Networks of violence: Predicting conflict in Nigeria, Dorff [/bib_ref] [bib_ref] The impact of the Boko Haram insurgency in Northeast Nigeria on childhood..., Dunn [/bib_ref].
# Methods and measures
## Data
In the analysis, we use a combination of cross-sectional and time-varying data drawn from multiple sources, summarized in . The DHS provides cross-sectional data on anthropometric measures for children and other indicators of health, demographic, and socio-economic characteristics of individuals, families, households, and communities. We use multiple waves of the DHS that were conducted in [bib_ref] Armed conflict, household victimization, and child health in Côte d'Ivoire, Minoiu [/bib_ref] and . We include all children surveyed between the ages of 24-59 months. The reason we use this age range is because the DHS contains the most detailed information on child health for children under 60 months old. We exclude the youngest children (under 24 months) even though wasting levels may be relatively high within this age group. We exclude this age group because breastfeeding behaviors may vary significantly with notable impacts on child health [bib_ref] Climatic conditions and infant care: Implications for child nutrition in rural Ethiopia, Randell [/bib_ref]. The DHS information on breastfeeding does not contain sufficient detail to evaluate frequency, duration, and other care factors vital to growth of children under 24 months.
The Uppsala Conflict Data Program's Georeferenced Event Dataset (UCDP-GED) Version 19.1 [bib_ref] Introducing the UCDP georeferenced event dataset, Sundberg [/bib_ref] supplies the data on conflict. UCDP-GED comprehensively captures events of organized armed violence that occurred during conflicts active around the world since 1989. An event is "an incident where armed force was used by an organized actor against another organized actor, or against civilians, resulting in at least one direct death at a specific location and a specific date". The dataset includes the georeferenced location and specific date of each event, the type of the conflict (state-based conflict, non-state conflict, or one-sided violence), the actors involved, the severity (measured in terms of battle-related fatalities), and other descriptive information. In the related literature on health and conflict previously cited, a range of different conflict measures are employed and explored for different purposes and with no clear consensus on measures (see [bib_ref] How conflict affects land use: Agricultural activity in areas seized by the..., Eklund [/bib_ref] [bib_ref] Food abundance and violent conflict in Africa, Koren [/bib_ref]. We specifically focus on armed conflict and lethal events as a relatively narrow and severe form of conflict (versus cattle raids or urban demonstrations, for example) with the idea that this type of conflict may be the most disruptive to local agriculture and food systems. Spatial details on the location of conflict events are varied, in part because of the data collection process. Ultimately, the finest reliable spatial detail for the data is the region-level (subnational administrative level 1). Please see a longer description in the "Appendix" about the spatial detail of the conflict data.
The Climate Hazards InfraRed Precipitation with Stations (CHIRPS) dataset provides daily precipitation at a 0.05° × 0.05° degree spatial resolution (~ 5 km 2 ) around the world . US and international agencies (e.g., Famine Early Warning System Network (FEWS NET)) routinely use the CHIRPS data set to monitor drought and food insecurity. Also, the recently developed Climate Hazards InfraRed Temperature with Stations (CHIRTSmax) dataset provides daily maximum temperature with the same spatial resolution and global coverage [bib_ref] A high-resolution 1983-2016 T max climate data record based on infrared temperatures..., Funk [/bib_ref] [bib_ref] Development and validation of the CHIRTS-daily quasi-global high-resolution daily temperature data set, Verdin [/bib_ref]. This resource, which leverages remotely sensed infrared radiation temperatures to estimate two-meter air temperatures, is demonstrated to have excellent measurement performance, most notably in data-sparse regions such as sub-Saharan Africa (e.g., see usage notes in [bib_ref] A high-resolution 1983-2016 T max climate data record based on infrared temperatures..., Funk [/bib_ref] [bib_ref] Global urban population exposure to extreme heat, Tuholske [/bib_ref]. We rely on CHIRPS and CHIRTSmax to derive spatially and temporally varying indicators capturing climate conditions near to the date of interview.
We also use annual seasonal maximum of the Normalized Difference Vegetation Index (NDVI) from NASA's Moderate Resolution Imaging Spectroradiometer (MODIS). This fine-grained measure of vegetation vigor allows for a direct measure of above ground biomass as related to agro-climatic growing conditions during the primary growing season (as indicated by FEWS NET's growing season calendars). NDVI describes the state of vegetation and has been shown to correlate well with deviations in crop yield, though performance is variable by region [bib_ref] Recent trends in agricultural production of Africa based on AVHRR NDVI time..., Vrieling [/bib_ref]. A common practice is to use community-level aggregate NDVI as a proxy for local-scale crop health and agricultural production of food [bib_ref] Real-time prediction of crop yields from MODIS relative vegetation health: A continent-wide..., Petersen [/bib_ref] [bib_ref] In-season prediction of corn grain yield potential using normalized difference vegetation index, Teal [/bib_ref] [bib_ref] A slow rainy season onset is a reliable harbinger of drought in..., Shukla [/bib_ref].
We use livelihood zone data from the Famine Early Warning System (FEWS NET). These zones capture general food and income characteristics of an area (http://fews.net/). The FEWS NET zones provide more detail (on specific types of crops and livestock, for example) than is necessary for this project and we instead, aggregate the zones into broad categories. The categories themselves and zone aggregations vary somewhat by country, in our case the Uganda zone includes more information on "urban" areas than the other two countries, for example, but in general they differentiate between those households with a greater dependence on cropping (agriculturalists), from those that are more likely dependent on livestock (pastoralists), from those engaged with both livestock and crops (agropastoralists). The FEWS NET zone reports also provide information on the key growing season months for each country used to calculate seasonal maximum NDVI.
Aggregated livelihood zones and growing season calendars have been used in studies of children's health and climate in the developing world (e.g., [bib_ref] Climate, birth weight, and agricultural livelihoods in Kenya and Mali, Bakhtsiyarava [/bib_ref] [bib_ref] Investigating the linkages between pregnancy outcomes and climate in sub-Saharan Africa, Davenport [/bib_ref] [bib_ref] Climatic conditions and infant care: Implications for child nutrition in rural Ethiopia, Randell [/bib_ref]. While not everyone within a particular zone is equally engaged in the same strategy to procure food or income, the data provide a way of gauging differences among geographic areas that may condition the impact and timing of climatological and conflict factors on economic and food production strategies.
## Spatially merging data
The DHS is spatially referenced at the level of the survey cluster. Global Positioning System (GPS) coordinates are collected and publicly released for locations of clusters where surveys are administered. Each cluster consists of a number of households that participated in the survey, spread across one or more populated places within a geographic area. To maintain confidentiality, the location of the survey cluster is randomly displaced. Most coordinates for clusters in rural locations are displaced by up to 5 km in any direction; a random sample of 1% of rural clusters are displaced by up to 10 km. For urban clusters, the displacement is up to 2 km. DHS recommends that researchers average any environmental data over a 5-10 km buffer around the coordinates of each DHS rural cluster, with the expectation that the specific community where households were sampled will fall somewhere within this buffer [bib_ref] Influence of demographic and health survey point displacements on raster-based analyses, Perez-Heydrich [/bib_ref]. The three environmental indicators -rainfall, temperature, and NDVI -are calculated by averaging across the area (using a 10 km buffer) around each offset DHS survey cluster GPS point. As previously, noted, considering climate/environmental variability at the level of the DHS community allows for comparisons across communities [bib_ref] Stunted from the start: Early life weather conditions and child undernutrition in..., Randell [/bib_ref] and differs from some of the more coarse-scale analysis, providing greater insight into the "on-the-ground" conditions [bib_ref] Global urban population exposure to extreme heat, Tuholske [/bib_ref]. The cluster's offset GPS point is also matched to regional and livelihood-level boundaries to merge with the data on conflict and livelihood zones. For merging with administrative regions or livelihood zones, we use whichever region or zone contains the publicly released point. In [fig_ref] Figure 1: Fig. 1. [/fig_ref] , we show maps of the DHS clusters and include regional boundaries for each country. [fig_ref] Table 2: Summary information of variables included in the analysis * For Uganda, We... [/fig_ref] summarizes the data, by country, for all the variables used in our analysis. Variables are grouped by the level of measurement in the multi-scalar structure of the analysis. For each variable, we report means or percentages as appropriate. The dependent variable in our analysis is a child's WHZ from the DHS data.
## Measures
We also include a range of independent variables that map onto the levels of the UNICEF framework. Our choice of variables is based on findings from related research that have identified specific factors as associated with acute malnutrition, and are organized using the UNICEF framework's categories of basic, underlying, or immediate causes (e.g., [bib_ref] Stunted from the start: Early life weather conditions and child undernutrition in..., Randell [/bib_ref]
## Key variables in the analysis (associated with basic causes)-we
include multiple indicators reflecting the count of conflict events, all of which are derived using the UCDP-GED data. To address variability in the spatial precision of the measurement of event locations, we aggregate counts at the level of the region. Aggregations are performed with reference to the most recent regional boundary data obtained from the Integrated Public Use Microdata Series-International (IPUMS-I) for the relevant boundaries used in the UCDP-GED for Kenya 2009 (provinces), Nigeria 2010 (states), and Uganda 2002 (districts). We consider counts of armed conflict events during the 1-3 month and 4-6 month intervals before the month of the DHS data collection for each child. Empirical analyses of conflict impacts use widely varying conflict measures (see [bib_ref] Malnutrition and conflict in Eastern Africa: Impacts of resource variability on human..., Rowhani [/bib_ref] [bib_ref] Drought, conflict and children's undernutrition in Ethiopia 2000-2013: A meta-analysis, Delbiso [/bib_ref] [bib_ref] Climate-conflict research: Some reflections on the way forward, Buhaug [/bib_ref]. While there is no consistent time period of conflict exposure and child health in the literature, some researchers consider cumulative effects, impacts within the last 6-12 months or the most recent month to the survey. Building on the discussion incomparing rapid onset and slow onset events, we choose these two different (1-3 month and 4-6 month), yet relatively proximate, time frames to help uncover the temporal process that connects conflict to child health shortly after exposure to conflicts.
To measure climatological variation that may have a direct association with child health, we include two variables that measure the average total precipitation and the average maximum temperature (derived from CHIRPS and CHIRTSmax) during the three months prior to the data collection in a given survey cluster (see [bib_ref] Stunted from the start: Early life weather conditions and child undernutrition in..., Randell [/bib_ref] [bib_ref] Malnutrition and conflict in Eastern Africa: Impacts of resource variability on human..., Rowhani [/bib_ref] [bib_ref] Linking climate change and health outcomes: Examining the relationship between temperature, precipitation..., Grace [/bib_ref].
As a gauge of agricultural productivity, we include the maximum NDVI for the most recent completed primary growing season in each country (e.g., [bib_ref] Climate, birth weight, and agricultural livelihoods in Kenya and Mali, Bakhtsiyarava [/bib_ref]. Growing season information is obtained from the FEWS NET country-specific growing season calendars. NDVI values range from −1 (no vegetation/greenness) to 1 (high indication of greenness/vegetation). In general, more vegetation (higher NDVI values) are associated with more a better growing season and more agricultural productivity [bib_ref] A slow rainy season onset is a reliable harbinger of drought in..., Shukla [/bib_ref]. In addition, we include dummy variables that flag whether or not any conflict during the periods measured by the lagged variables coincided with the growing season.
## Other independent variables associated with underlying causes-we
include several variables to control for potential underlying causes vulnerability to acute malnutrition. The child's age, sex, and birth order (first-, second-, third-born, etc.) are routinely included in studies of child health and food security because they capture biologically and culturally relevant factors related to growth, feeding, and caretaking. We also include child's birth month and birth year. Consistent with related research we also include maternal age and educational attainment [bib_ref] Path analyses of risk factors for linear growth faltering in four prospective..., Prado [/bib_ref] [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref]. Another set of variables measures characteristics of households. Household toilet facilities and water sources are associated with malnutrition in some settings possibly through pathogens in fecal matter and contaminated water [bib_ref] The association between acute malnutrition and water, sanitation, and hygiene among children..., Van Cooten [/bib_ref]. The flooring material in the house where a child currently lives (either unfinished or finished) provides a simple indicator to approximate a household's socio-economic status, primarily by differentiating the most poor (those with unfinished flooring) from the less poor (those with finished flooring). Household electricity-status may capture food storage potential as well as general socio-economic status and is also included in the analysis. Flooring has the advantage of tending to be less correlated with urban or rural residence than other factors that can capture aspects of socio-economic status such as electricity use or water access [bib_ref] Child health outcomes in sub-Saharan Africa: A comparison of changes in climate..., Davenport [/bib_ref] [bib_ref] The impact of the Boko Haram insurgency in Northeast Nigeria on childhood..., Dunn [/bib_ref]. The household-level variables are relatively static, compared to other measures of socio-economic status that focus on different types of assets or especially on income. Thus, the value of each of these variables when the survey was administered is likely to reflect conditions in at least the recent past, potentially dating back to when the children in the household were born.
We also include characteristics of communities (i.e., survey clusters). Whether the location is urban or rural (a classification determined with reference to criteria specific to each country) and the type of livelihood zone (agricultural, agropastoral, fishing, pastoral, or urban, from FEWS NET) defines a surrounding context that can influence food security and health outcomes. These indicators capture broader temporal trends and spatial patterns that can likewise affect food security and health outcomes, beyond what is reflected in other variables. We include region of residence to control for regional-level time invariant factors.
## Other independent variables associated with immediate causes-
Finally, we consider whether or not each child experienced fever and/or diarrhea within the 2 weeks preceding the administration of the survey, using the data collection in the DHS. Both of these variables are known to be proximate risk factors for acute malnutrition [bib_ref] Measuring local determinants of acute malnutrition in Chad: A case-control study, Tesfai [/bib_ref].
# Statistical analysis
We estimate a suite of regression models to explore the relationships between conflict, climate and acute malnutrition. The primary model is a linear mixed-effects regression model estimating the continuous WHZ score for each child in relation to the number of conflict events in the child's region of residence. The model also includes immediate and underlying factors that can impact child WHZ as described above.
The data has a hierarchical structure 3 -there are often several children with the same mother and there are several households within the same DHS cluster. To account for the shared characteristics between individuals, we include a random effect for the mother and for the DHS cluster [bib_ref] Within-mother analysis of seasonal patterns in health at birth, Currie [/bib_ref].
The model can be specified as follows:
Y ijk = β 0 + β 1 conflict jk + β 2 conflict_grs jk + β 3 temp jk + β 4 precip jk + β 5 NDV I jk + β n X z + w j + u k + e ijk In the equation, Y is a continuous WHZ score for child i from mother j in DHS cluster k. Conflict is a count variable counting violent events (within an administrative region) during either a) 1-3 months before the survey or b) 4-6 months before the survey. We estimate separate models for the number of conflict events during the 1-3-month period before the survey (Model 1) and 4-6-month period before the survey (Model 2). 4 Conflict_grs is a dummy variable-a value of 1 indicates that when conflict occurred it occurred during the growing season and a value of 0 indicates the conflict did not occur during the growing season. 5 Parameters β 3 , β 4 , and β 5 are the terms for the maximum average monthly temperature three months before the survey, average monthly precipitation three months before the survey, and maximum NDVI during the previous growing season, respectively. Keep in mind that these variables are dependent on the interview date (of the mother) and the cluster location. The model also controls for child-, mother-, household-, and community-level variables (X z ) described in the Measures section, as well as fixed effects for the month and year of birth and region of residence. W j is a random effect for woman, and accounts for multiple children per woman. The DHS cluster-level random effect is specified by u k . Fixed effects account for unobserved (time invariant) factors.
# Results
We first describe the overall results by country below. We then summarize the patterns that we observe in terms of timing of conflict and spatial variation when considering the results together.
## Kenya
The frequency of armed conflict events within the administrative region where a child lives during the 1-3 month lagged period is not significantly associated with a child's WHZ score as shown in Model 1 in [fig_ref] Table 3: Climate and conflict continuous WHZ score regression results for children aged 24-59... [/fig_ref]. However, in the case that the timing of the specific conflict occurred during the growing season, we observe a significant negative relationship between the exposure and WHZ score. Thus, when considering conflict events during the 1-3 month period, the count of events itself was not significant, however if conflict occurred during the growing season, the mean WHZ score is 0.06 standard deviations lower than if the conflict did not occur during the growing season. In the event that conflict occurred 4-6 months before the survey month, Model 2, we observe a significant and negative association with child health but no linear association with regard to growing season timing of this later lag. An additional conflict event is then associated with a 0.01 decrease in mean WHZ score. In terms of climate and environmental conditions, poor growing conditions, as indicated by lower amounts of precipitation, higher maximum temperatures, and a lower vegetation (NDVI) index, are negatively associated with WHZ scores. These relationships remain consistent across the two models. Household-and mother-level patterns suggest that, in general, children living in households with more resources (e.g., finished flooring, electricity, higher maternal education) have, on average, higher WHZ scores than children in less well-resourced households. Child-specific factors also play an important role, with younger children who are of a higher birth order (in other words, children with more siblings), tending to have lower WHZ scores. Factors associated with recent illness-diarrhea and fever-are also negatively associated with WHZ.
## Nigeria
According to the Models for Nigeria [fig_ref] Table 4: Climate and Conflict Continuous WHZ score regression results for children aged 24-59... [/fig_ref] , children who live in regions of Nigeria that experienced higher counts armed conflict within 1-3 months or 4-6 months before the survey are more likely to have lower WHZ scores. In Model 1, conflict occurring during the growing season is associated with lower WHZ scores-a reduction of 0.84. The results also show that greater amounts of precipitation and higher temperatures are associated with reduced WHZ scores. While elevated temperatures are commonly associated with worse health outcomes [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref] , rainfall is often positively associated with health outcomes, although not consistently [bib_ref] Stunted from the start: Early life weather conditions and child undernutrition in..., Randell [/bib_ref] [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref]. A possible explanation for these relationships is that hot and wet conditions may be associated with disease-especially malaria. Therefore, it is possible that the recent weather conditions are indicative of increased risk of malaria among children. In Kenya, by contrast, malaria is not consistently endemic across the country and precipitation conditions are more likely associated with high quality growing seasons and higher food availability.
In terms of household-, maternal-and child-level factors, a child with a recent fever in Nigeria is significantly more likely to have a worse WHZ score than a child who has not had a recent fever. We also observe a general increase in WHZ for female versus male children. Other factors associated with household or community resources are not significantly associated with the outcome variable-possibly suggesting that these community-and regional-level factors are important for understanding variation in risk of acute malnutrition in Nigeria.
## Uganda
As reported in , no statistically significant association was detected between the frequency of recent conflict events and WHZ scores among children in Uganda, even when conflict occurred during the growing season. As mentioned in the Methods section -we also estimated models where we combined the counts of conflict over the prior 6 months and where we used an NDVI × conflict interaction effect (see "Appendix"). We do not present all of these results-but they all generally produced estimates yielding similar insight-there is no significant relationship between conflict, environmental factors and WHZ. None of the environmental factors were significantly associated with a child's WHZ score. Beyond living in a pastoral region, where children, on average, have WHZ scores that are around 0.40 standard deviations lower than children living in agricultural areas, the only factors with significant associations are select immediate and underlying causes. Specifically, children who are younger, female, experienced a recent fever, live in households with piped water (versus relying on surface water or sources that may include water trucks and purchased water) and no electricity are more likely to have lower WHZ scores than their counterparts.
## Similarities and differences across the three countries
Across the three countries and during similar time frames, we observe very different relationships factors associated with variation in WHZ. The relationship between counts of conflict in the administrative region of residence, and WHZ varies significantly across the settings. We note that there is a significant negative relationship between the count of events and WHZ score in Nigeria versus non-significant correlations in the other two countries. Variability, however, is relatively higher in Uganda than in Kenya. The environmental factors of interest also vary by country with Kenya and Nigeria reporting significant impacts on WHZ from rainfall and temperature (albeit the coefficients are very small and the direction of the relationship varies for rainfall). NDVI is positively and significantly related to WHZ scores in Kenya and not significant in the other two countries. In general, despite their geographic proximity, results from Uganda and Kenya, appear quite different from each other in terms of the significance of household-, maternal-, and child-level factors with very little obvious shared relationships between the conflict and environmental factors. Considering the regression results together, therefore, suggests that exposure to conflict is related to child health but only as it acts through other downstream factors and suggests that there are a number of unmeasured factors that differentiate the contexts. In other words, there are additional place-based factors that must be considered that shape the way that conflict relates to health outcomes.
Among the other factors included in the analysis, there are very few similarities across the countries in terms of significance, or magnitude and direction of factors. For example, a child's age and sex are significant across all three models (although in Kenya, sex is not significant). However, in Nigeria older female children have higher WHZ scores whereas in Uganda, younger male children have higher WHZ scores. In Kenya, younger children, regardless of sex, have lower WHZ scores.
# Discussion
Unlike other forms of malnutrition, with strategic interventions, acutely malnourished children can recover and avoid serious illness and death. To combat acute malnutrition and lessen the adverse impacts of this food insecurity outcome, researchers and policymakers continue to work to uncover the processes associated with acute malnutrition and variation in WHZ. It is therefore especially important to consider the timing of exposure and how different contexts shape the relationship between exposure and outcomes.
Broadly speaking, our results suggest varied relationships between conflict, climate and food security across the three countries. In other words, the linkages between climate and conflict and acute child health outcomes related to food insecurity and malnutrition, are not consistent across settings. At the country-level, however, there are clear indications for Kenya and Nigeria, that targeting interventions based on climate and conflict conditions could help to improve child health and reduce wasting. Developing effective interventions will likely require ongoing monitoring and evaluation at a sub-national-level, using detailed spatially relevant data that can capture a range of local factors.
Our analyses employed fairly fine-scale measures of climate conditions associated with child health-rainfall and temperature conditions around the time of survey-as well as with vegetation measures associated with local agricultural production (seasonal NDVI-a measure of growing season vegetation). As compared to related research (e.g., [bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref] [bib_ref] The impact of the Boko Haram insurgency in Northeast Nigeria on childhood..., Dunn [/bib_ref] , these climate and vegetation measures employ more spatial variation. Despite these highly detailed data, however, the relationship between the variables and the outcome are not consistent across settings. We interpret this finding as suggesting that there are local strategies that individuals, households or communities employ to secure their children's health in the context of specific climate and food production conditions. These strategies are potentially situated in historic and place-based cultural practices associated with risk management that are not measured in the standard health surveys used here yet are key components of vulnerability and variability in child health outcomes in a context of climate change. Locally varying factors, unmeasured in this analysis, related to roads, humanitarian aid, population mobility, among other factors may also play an important role. Future research exploring these dimensions may help to isolate these different relationships [bib_ref] The impact of the Boko Haram insurgency in Northeast Nigeria on childhood..., Dunn [/bib_ref].
Along the same lines, the relationship between armed conflict and acute malnutrition is inconsistent across countries and time periods. While this finding mirrors related research that fails to find a consistent linkage between conflict and food security, our approach suggests that even when survey data from the same source as well as constructing identical measures of conflict across settings, the relationship between conflict and acute malnutrition is not consistent across countries. While some research has suggested that there may be some variation in the relevant timing of the events on child health outcomes and food security, the results here suggest that there is not a consistent temporal linkage between events and child health outcomes. We attempted to model conflict using a relatively extreme definition of conflict, one which may be most likely to impact food systems and change behaviors [bib_ref] The post-war public health effects of civil conflict, Ghobarah [/bib_ref] , nonetheless, in the case of Uganda we see no relationship between conflict conditions and child health. Whereas in Nigeria we see a significant association between conflict and child health, with WHZ scores decreasing greater when conflict events occurred during the growing season. In Kenya, however, the relationship is somewhat different, with a significant association between conflict and child health in the case that a child is exposed to conflict during the 3 months preceding the survey and if this period occurs during the growing season.
We are unable to evaluate how the historical presence of conflict has shaped the ways that communities cope with armed events. Kenya has been comparatively peaceful compared to the other countries, Uganda, has become more peaceful after an extended civil war, while Nigeria has just entered into a situation of escalating violence. Is it possible that Ugandans are accustomed to violence in a way that reduces the potential for violent events to dramatically uproot behaviors and experiences? The data available and our analytic approach was unable to uncover the ways that historical experiences shape current realities.
Additionally, the analysis could examine conflict in a more spatially specific manner (using finer-scale data on conflict locations and durations), similar to what is possible with climate. Doing so might help reinforce our ability to determine if individual children and households in specific neighborhoods have been exposed to observed conflict conditions. A limitation of the available data, however, is that precision in the measurement of spatial locations of conflict events is reliable only to the regional administrative level. Some conflict is actually dispersed to that extent, affecting larger areas within a country. Other conflict events are simply measured imprecisely, given available information. Whether conflict is dispersed within a region, or merely occurred within a region but was not necessarily nearby a specific household, a potential exists for residents of the region to be exposed to the implications.
We also note that there are important limitations with the health survey data included in this research. One limitation is that we focus solely on living children and living respondents (mothers)-perhaps our sample includes those who are the most robust as the frailest have not survived to be included in the survey. This kind of bias would lead to results showing that the effects of conflict are less severe than they actually are. Additional research using spatially detailed data on maternal and infant/child mortality would help to uncover the potential for selection bias. Overall, our research highlights the need for improved data on armed conflict that can be used within the context of individual-level studies on nutrition and food security. It also highlights the need for more qualitative investigations that consider household coping strategies reflecting the multiple stresses that families may face in a context of climate change.
## Description of spatial detail in conflict data
The data set codes the location of conflicts with varying degrees of spatial precision. In a share of cases events are pinpointed to particular population centers (e.g., village, town, city) or even specific neighborhoods, transportation interchanges, or physical landmarks. Other sizeable shares of cases are associated with larger geographic areas such as constituent regions of countries. For example, in the entire dataset, covering 123 countries from 1989 to 2020, 88% of the events are coded to the first-order administrative division. 42% of these events are at, or in close vicinity (25 km) of, a specific point location. Among this 42%, 23% of the points in the exact location. The countries included in our study report similar statistics.
This variation in spatial precision is due to two main reasons resulting from the datagenerating process. First, certain conflict events occur with wider geographic scopes of activity, extending over areas within or across countries, rather than at point locations. The geographic coding of these events to areas is therefore factually accurate, just not spatially precise. Second, available information about events is not uniform in terms of providing complete details relevant to establishing locations with maximum spatial precision. Events may be coded with less spatial precision because the information did not permit a greater degree of spatial precision. Thus, the geographic coding of these events is accurate within the limits of the information, but not necessarily fully accurate factually.
Ultimately, the common denominator for most cases is that the user will know the locations of conflict events with certainty with respect to first-order administrative divisions at a sub-national level. Thus, this degree of spatial resolution tends to be reliable for analyses using the conflict event data where geography matters. Analysis with any more precise degrees of spatial resolution will require (1) making dubious assumptions about locations of a substantial share of cases (e.g., assigning events with imprecise spatial codings to centroids or major population centers within known areas, thereby inserting artificial precision of measurement), (2) imputing locations for those cases (thereby increasing uncertainty in measurement), or (3) dropping the cases altogether (thereby introducing incompleteness and risking bias in measurement). All of these circumstances may affect analytical results and findings (Appendix [fig_ref] Table 6: Interacting Conflict Counts and Recent Growing Season Condition to Analyze WHZ [/fig_ref]. All independent variables found in [fig_ref] Table 3: Climate and conflict continuous WHZ score regression results for children aged 24-59... [/fig_ref] , 5 are included in these models. We present the main independent variables of interest in this abbreviated tabl DHS clusters and regional boundaries (first-order administrative division) for Kenya, Nigeria, and Uganda (in order) Table 1 Overview of data sources Dataset Uses Spatial resolution of the raw data
## Temporal resolution in analysis
The Demographic and Health Survey (DHS): Individual-level health and sociodemographic characteristics
## <0.01
Conflict period overlaps with growing season Climate and Conflict Continuous WHZ score regression results for children aged 24-59 months in Uganda To account for correlation among siblings, mother is included as a random effect
[fig] Figure 1: Fig. 1. [/fig]
[table] Table 6: Interacting Conflict Counts and Recent Growing Season Condition to Analyze WHZ [/table]
[table] Table 2: Summary information of variables included in the analysis * For Uganda, We use the livelihood zone coding for urban versus rural. Spat Demogr. Author manuscript; available in PMC 2023 August 18. For Kenya and Nigeria we use the DHS coding for urban Spat Demogr. Author manuscript; available in PMC 2023 August 18. [/table]
[table] Table 3: Climate and conflict continuous WHZ score regression results for children aged 24-59 months in Kenya Coefficient Confidence Interval p-value Coefficient Confidence Interval p-value Spat Demogr. Author manuscript; available in PMC 2023 August 18. Coefficient Confidence Interval p-value Coefficient Confidence Interval p-value [/table]
[table] Table 4: Climate and Conflict Continuous WHZ score regression results for children aged 24-59 months in NigeriaCoefficient Confidence Interval p-value Coefficient Confidence Interval p-value [/table]
[bib_ref] Climate variability and child nutrition: Findings from sub-Saharan Africa, Thiede [/bib_ref]
|
10.3390/ijerph16122158
|
CCBY
|
6617153
|
31216746
|
s2orc_pubmed_articles
|
High Prevalence of Respiratory Symptoms among Particleboard Workers in Ethiopia: A Cross-Sectional Study
Work in the wood industry might be associated with respiratory health problems. The production of particleboard used for furniture making and construction is increasing in many countries, and cause dust, endotoxin and formaldehyde exposure of the workers. The aim of the study was to assess the prevalence of respiratory symptoms and to measure lung function among Ethiopian particleboard workers using Eucalyptus trees as the raw material. In total 147 workers, 74 from particleboard production and 73 controls, participated in the study. Mean wood dust in the particleboard factories was measured to be above recommended limit values. Particleboard workers had a mean age of 28 years and the controls were 25 years. They had been working for 4 and 2 years, respectively. Lung function test was done using spirometry following American Thoracic Society (ATS) recommendations. Respiratory symptoms were collected using a standard questionnaire of ATS. Particleboard workers had higher prevalence of wheezing, cough, cough with sputum production, phlegm, and shortness of breath compared to controls. Lung function status was similar in the two groups. The symptoms might be related to the work in the factories. Longitudinal studies are recommended to explore the chronic impact of work in particleboard factories on respiratory health.
# Introduction
Wood dust is a complex substance and one of the hazards generated from processing of various wood types for a wide range of applications. Workers exposed to wood dust may develop different respiratory health problems [bib_ref] Functional disorders of the lung and symptoms of respiratory disease associated with..., Neghab [/bib_ref] [bib_ref] Respiratory symptoms and lung function in never-smoking male workers exposed to hardwood..., Bislimovska [/bib_ref] including reduced lung function [bib_ref] Rubberwood dust and lung function among Thai furniture factory workers, Thetkathuek [/bib_ref] [bib_ref] Respiratory and skin effects of exposure to wood dust from the rubber..., Sripaiboonkij [/bib_ref]. An endotoxin component in the cell walls of gram-negative bacteria [bib_ref] Health effects due to endotoxin inhalation (review), Liebers [/bib_ref] might be present as a part of organic dust and may induce inflammatory responses in the airways [bib_ref] Health effects due to endotoxin inhalation (review), Liebers [/bib_ref] [bib_ref] Respiratory health and breath condensate acidity in sawmill workers, Ljubičiććalušić [/bib_ref] [bib_ref] Respiratory health effects of exposure to low levels of airborne endotoxin-A systematic..., Farokhi [/bib_ref]. In addition, formaldehyde that is added to the urea resin for gluing of wood products is associated with respiratory health problems [bib_ref] Respiratory symptoms due to occupational exposure to formaldehyde and MDF dust in..., Thetkathuek [/bib_ref] [bib_ref] Respiratory effects due to occupational exposure to formaldehyde: Systematic review with meta-analysis, Mathur [/bib_ref] and decrements in lung function [bib_ref] Respiratory health of plywood workers occupationally exposed to formaldehyde, Malaka [/bib_ref]. A range of biologically active compounds like quinones, terpenes, stilbenes, phenols, tannins, and flavonoids might also be released during wood processing.
In Ethiopia the manufacturing sector, comprising wood, metal, food, textile, leather and construction industries, accounts for 6.9% of the national work force. Eucalyptus, an evergreen hardwood tree, is the main raw material for production of particleboard in Ethiopia and is used for furniture like office tables and shelves, for interior wall partitioningand construction. The Eucalyptus tree is cheap, locally available, has rapid growth, and is adaptable to a range of climates and soil types. This makes it a promising source of inputs as the native wood species are diminishing due to deforestation. Despite the increasing production of furniture in Ethiopia little is known about safety measures and occupational health in these workplaces [bib_ref] Occupational health and safety in Ethiopia: A review of situational analysis and..., Kumie [/bib_ref]. In previous studies from the wood industry including particleboard production, the workers have been exposed to other types of trees [bib_ref] Rubberwood dust and lung function among Thai furniture factory workers, Thetkathuek [/bib_ref] [bib_ref] Respiratory symptoms and lung function in relation to wood dust and monoterpene..., Lofstedt [/bib_ref]. More knowledge on the respiratory health of the particleboard workers exposed to dust from the Eucalyptus tree is needed to evaluate the need of occupational preventive measures in Ethiopia and other developing countries.
The international literature about the prevalence of respiratory health symptoms among wood workers varies greatly. For example, the prevalence of respiratory symptoms reported in Thailand and Iran varies from 15.5% to 41% [bib_ref] Functional disorders of the lung and symptoms of respiratory disease associated with..., Neghab [/bib_ref] [bib_ref] Respiratory and skin effects of exposure to wood dust from the rubber..., Sripaiboonkij [/bib_ref]. Decrement in lung function among wood workers is reported in studies done in Thailand, Pakistan, Iran and Sweden [bib_ref] Functional disorders of the lung and symptoms of respiratory disease associated with..., Neghab [/bib_ref] [bib_ref] Rubberwood dust and lung function among Thai furniture factory workers, Thetkathuek [/bib_ref] [bib_ref] Respiratory and skin effects of exposure to wood dust from the rubber..., Sripaiboonkij [/bib_ref] [bib_ref] Respiratory symptoms and lung function in relation to wood dust and monoterpene..., Lofstedt [/bib_ref] [bib_ref] Lung function in Pakistani wood workers, Meo [/bib_ref] , while other studies done in Poland and Denmark do not show any effect on lung function [bib_ref] Lung function: Occupational exposure to wood dust, Baran [/bib_ref] [bib_ref] Cross-shift and longitudinal changes in FEV1 among wood dust exposed workers, Jacobsen [/bib_ref]. Thus, the international literature is not conclusive regarding the respiratory health for wood workers, and a study among particleboard workers using Eucalyptus trees as raw materials in Ethiopia is needed.
The aim of this study was to assess the prevalence of respiratory symptoms and to measure lung function among particleboard factory workers of Ethiopia and compare the findings with a control group from water bottling factories with low exposure to dust. The findings might help to fill the research gaps on respiratory health among particleboard workers which can be applied to prevent respiratory disease at these workplaces.
# Materials and methods
## Study design and period
A cross-sectional study was performed to assess respiratory symptoms and to measure lung function among workers from two of the largest particleboard factories in Ethiopia which use Eucalyptus trees as raw material. One of the factories was established in 2005, is situated in northern Ethiopia and has 663 workers. The other factory was established in 2002 and is located in southern Ethiopia and has 249 workers. The particleboard factories are found in urban areas and selected both from North and South. A control group was established of workers employed in a water bottling factory, with a total of 339 workers from northern Ethiopia. The controls were selected from a factory considered to have low dust concentration in the work environment. The study period of this paper was from May 2017 to August 2017.
## Exposure measurements
Personal inhalable dust was sampled in the breathing zone of the workers using a conductive plastic inhalable conical sampler (CIS, JS Holdings, Stevenage, UK) [bib_ref] Rovell-Rixx, D.C. The performance of personal inhalable dust samplers in wood-products industry..., Tatum [/bib_ref] mounted with a 37 mm glass-fiber (GFA) filter (Whatman International Ltd, Maidstone, UK) using an air flow of 3.5 L/min Side Kick Casella (SKC) pump for 2 to 4 hours sampling duration per shift. In total, 76 workers in particleboard production were selected for repeated sampling of inhalable dust (n = 152). From the control group, i.e., the water bottling factory, 8 repeated samples were taken (n= 16). Inhalable dust was analyzed using gravimetric method in a room with controlled climatic conditions (22 - C, 45% relative humidity; desiccation ≥24 h) with an analytical balance with 0.1 µg readability (Mettler-Toledo Ltd, Greifensee, Switzerland) and the concentration was estimated in mg/m 3 . Endotoxin was analyzed using the Kinetic Amoebocyte Lysate test (Kinetic-QCL endotoxin kit, BioWhittaker, Walkersville, MA, USA) and the concentration was estimated in EU/m 3 .
The geometric mean, arithmetic mean and range of inhalable dust for particleboard workers was 4.66 mg/m 3 , 9.17 mg/m 3 , and 0.47-184 mg/m 3 , respectively. For the control group the figures were 0.21 mg/m 3 , 0.24 mg/m 3 , and 0.02-0.4 mg/m 3 , respectively.
The geometric mean, arithmetic mean and range of endotoxin for particleboard workers was 62.2 EU/m 3 , 245.6 EU/m 3 , and 0.9-9202.2 EU/m 3 , respectively; while for the control group it was 0.66 mg/m 3 , 0.75 EU/m 3 , and 0.3-2.3 EU/m 3 , respectively. The concentration of formaldehyde measured with Dräger-Tubes by color tubes in the particleboard factories using "worst-case" sampling strategy ranged from <0.2 ppm to 5 ppm.
## Study population and sample size
The required number of participants for this study was calculated using OpenEpi software (http: //www.openepi.com/SampleSize/SSMean.htm) sample mean difference by taking into consideration forced expiratory volume in one second (FEV 1 ) as main output of interest with 95% power, 95% confidence interval and 5% level of significance. The FEV 1 for exposed (3.77 L/s, SD = 0.99) and control group (4.29 L/s, SD = 0.86) was taken from a previously study done among particleboard workers in Ethiopia. The estimated number of participants needed was 166 workers (83 from exposed and 83 from control groups).
## Data collection
To plan the study, the factories and their leadership were visited. After obtaining permission to perform the study, we asked the management to provide a list of workers in each work shift during the first phase of data collection as stated in a previously published paper [bib_ref] Knowledge, attitude and practice related to chemical hazards and personal protective equipment..., Asgedom [/bib_ref] aimed to assess workers knowledge, attitude and practice regarding chemical hazards and personal protective equipment. Before the actual data collection, randomly selected participants were informed about the objective of the study, its relevance, and how to perform the interview and lung function measurements and asked for written consent to participate in the study.
## Respiratory symptom assessment
The interview of respiratory symptom assessment was done in a quiet and private place by six trained bachelor environmental health professionals.
Information on respiratory symptoms was collected using a validated standard questionnaire from the American Thoracic Society (ATS). The information collected were data on sex (M/F), age (years), education (highest grade completed), uses biomass fuel as sources of energy at home(Y/N), availability of separate kitchen (Y/N), number of service years in the present industry, occupational history in dusty working environment, and smoking (Y/N). The workers were also asked about their history of past respiratory illness (Y/N). If they had experienced any diseases, they were asked to tell what type it had been.
Questions about respiratory symptoms were asked like this; whether the workers in the last 12 months have/had: cough (Y/N), cough with sputum production (Y/N), phlegm (Y/N), episode of cough and phlegm (Y/N), wheezing (Y/N), shortness of breath (Y/N). The interviewed participants were also observed if they were using personal protective equipment (PPE) mainly face mask during the study.
The interview was based on questions prepared in English and translated to Amharic by a translator, and then translated back from Amharic to English by another translator, to check the consistency. Pretesting of the questionnaire was done on 5% of the sample population in one of the factories before the main study. The data collection tool was modified to suite the Ethiopian context. Information such as marital status and race were excluded from the questionnaire. Additional information about the use of biomass fuel as source of energy for cooking, availability of separate kitchen at home were added to the data collection tool.
## Lung function test
Prior to performing the lung function measurements, the participants ID, age, sex, standing height (m) and weight (kg) were measured as recommended by the American Thoracic Society [bib_ref] Standardisation of spirometry, Miller [/bib_ref] , and combined with the interview described in 2.4.1. Lung function test was done in sitting position using Spirometry (Minispir light with Winspiro software, Medical International Research (MIR), Rome, Italy) connected to a Laptop following American Thoracic Society guidelines [bib_ref] Standardisation of spirometry, Miller [/bib_ref]. The spirometry measurements were performed prior to the morning shift that started at 6:00 a.m. The lung function measurements were done by the trained researcher until the trial generated three acceptable maneuvers. From the three records of lung function test, the best trial was kept and used for the data analysis. The lung function parameters considered were FVC, FEV 1 , FEV 1 /FVC ratio and FEF 25-75% . The FEV 1 /FVC ratio < 70% was the cutoff point for air flow limitations as stated by Global Initiative for Chronic Obstructive Lung Disease.
## Data management and analysis
Collected data were checked for completeness and missing values at the end of each day of data collection. Data was exported from EpiData version 3.1 (EpiData Association, Odense, Denmark) to the statistical package SPSS, version 25 (International Business Machines Corporation (IBM), Armonk, NY, USA) for analysis. Lung function parameters were normally distributed. Descriptive statistics were used to summarize demographic and anthropometric data. Pearson chi-square or Fisher's exact tests (if the expected value was less than 5) were used to test for difference in categorical variables, while an Independent t-test was used to compare means of continuous variables between exposed and controls. Poisson regression analysis was used to determine the prevalence ratio of cough between particleboard workers and water bottling workers while adjusting for educational status, previous work in dusty environment, age and availability of separate kitchen. Prevalence ratio (PR) was chosen instead of prevalence odds ratio (POR) due to the high respiratory symptom prevalence in this study [bib_ref] Prevalence odds ratio versus prevalence ratio: Choice comes with consequences, Tamhane [/bib_ref].
Multiple linear regression was applied to analyze differences in lung function between the particleboard workers and the controls while adjusting for age, height, previous respiratory illness, availability of separate kitchen and use of biomass fuel as source of energy.
# Ethical approval
Ethical clearance for the study was obtained from the Regional Committee for Medical and Health were informed about the purpose of the study, confidentiality of their information, duration of the interview, lung function measurement procedure and the possibility to withdraw from the study at any time they wanted. Written consent both from each of the study participants and consent from factory management was assured before data collection.
The questionnaire and spirometry results were stored with only ID numbers of the participants, not their names. The data were locked in a safe place accessible only to the researcher to keep every person's information confidential.
# Results
## Characteristics of the study population
From 166 people who were invited, 157 workers (94.5%) (83 particleboard workers and 74 water bottling workers) participated in the study. The remaining 5.5% did not want to participate in the study. Among the 157 who participated, one person from the particleboard workers could not properly perform the lung function measurement and we therefore discarded his lung function data. The majority, 147 (93.6%), of the participants in the study were males and used in the final analysis. Due to their low number, the females (5.7%)-eight participants from exposed and one participant from the water bottling factory-were excluded from the further analyses since gender differences are recognized for respiratory symptoms as well as for lung function. The exposed group was older than the controls (28 vs. 25 years; p = 0.006) and had more service years than the controls (4 vs. 2 years; p < 0.001). The exposed groups were also more educated than controls [fig_ref] Table 1: Demographic and anthropometric characteristics of 74 particleboard workers [/fig_ref] and had higher body weight (63 vs. 56 kg; p < 0.001). The exposed and the control groups had the same average height (1.70 m). All of the study participants were neither smokers nor using proper personal protective devices such as face masks that can protect them from dust and other chemical exposures. The majority (64%) of the exposed group had separate kitchens and only 30% used biomass fuel for cooking.
Some respondents in the exposed group had a history of previous illness such as bronchitis (n = 12), asthma (7), pneumonia (3) and tuberculosis (3), but such illnesses were not reported in the control group. In total, 18 (24%) participants had previous disease, of which 4 had more than one diagnosis.
## Respiratory symptoms
The prevalence of all recorded respiratory symptoms was significantly higher among the exposed (range 24-45%) than the controls (2.7-15%) [fig_ref] Table 2: Prevalence of respiratory symptoms among 74 particleboard workers [/fig_ref]. A separate analysis was performed, excluding the participants from the particleboard factories who had previous respiratory diseases (n = 18). The results were still the same, except for cough which did not show any significant difference between the groups when these 18 persons were excluded (result not shown).
The prevalence ratio of cough among the exposed group was 1.56 (95% CI; 0.61, 3.95) compared to the controls when adjusted for education status, previous work in dusty environment, past history of respiratory illness, age, use of biomass fuel for cooking and availability of separate kitchen.
## Lung function
Lung function (FEV 1 /FVC) between the exposed group and the controls was significantly different (p = 0.004) when no adjustments were made . . A comparison of lung function status among particleboard workers (n = 74) and controls (n = 73) in Ethiopia, both using t-test and multivariate regression; adjusting for age, height, previous respiratory illness, availability of separate kitchen and use of biomass fuel as source of energy. However, in multiple regression models the difference in lung function between exposed and control groups was not significant when adjusting for age, height, previous respiratory illness, availability of separate kitchen and use of biomass fuel as source of energy . All participants had FEV 1 /FVC values > 70%, indicating that none of the workers had airflow limitation. The same result was obtained when 18 participants with previous respiratory diseases were excluded from the analysis, both for the crude comparison of the groups and the regression analysis (result not shown).
## Lung function parameters
# Discussion
Workers in the particleboard manufacturing factories in Ethiopia had significantly higher prevalence of respiratory symptoms compared to water bottling workers (controls). The lung function values were not significantly different between the two groups when adjusting for age, height, previous respiratory illness, availability of separate kitchen and use of biomass fuel as source of energy.
The present study showed a higher prevalence of cough among particleboard workers than the controls. This finding agrees with studies done among wood workers in Tanzania, Macedonia, Iran and Sweden [bib_ref] Functional disorders of the lung and symptoms of respiratory disease associated with..., Neghab [/bib_ref] [bib_ref] Respiratory symptoms and lung function in never-smoking male workers exposed to hardwood..., Bislimovska [/bib_ref] [bib_ref] Respiratory symptoms and lung function in relation to wood dust and monoterpene..., Lofstedt [/bib_ref] [bib_ref] Respiratory symptoms and dust exposure among male workers in small-scale wood industries..., Rongo [/bib_ref]. The increased prevalence of phlegm and wheezing among particleboard workers in our study was also in compliance with the findings among parquet manufacturing workers in Macedonia and sawmill workers in Iran [bib_ref] Functional disorders of the lung and symptoms of respiratory disease associated with..., Neghab [/bib_ref] [bib_ref] Respiratory symptoms and lung function in never-smoking male workers exposed to hardwood..., Bislimovska [/bib_ref]. Furniture manufacturing workers in Thailand had an increased risk of wheezing and breathlessness compared to office workers [bib_ref] Respiratory and skin effects of exposure to wood dust from the rubber..., Sripaiboonkij [/bib_ref] [bib_ref] Respiratory symptoms due to occupational exposure to formaldehyde and MDF dust in..., Thetkathuek [/bib_ref] which is also consistent with our finding.
A high prevalence of past history of respiratory illness (24%) was reported among particleboard workers but not in controls. However, the causes of past respiratory illnesses were not investigated. It may emanate from wood working activities or other dusty environments as some of the particleboard workers had worked in other dusty environment than the controls but can also be due to other unknown reasons.
Lung function was not significantly different between the particleboard workers and the control group. Our finding agrees with findings in Iran and Pakistan [bib_ref] Functional disorders of the lung and symptoms of respiratory disease associated with..., Neghab [/bib_ref] [bib_ref] Lung function in Pakistani wood workers, Meo [/bib_ref] which showed an insignificant difference in lung function between exposed and control groups. In our finding, all workers had a FEV 1 /FVC > 70%. This is similar with findings in Iran, Macedonia, Poland and Pakistan which showed that the mean FEV 1 /FVC ratio was higher than 70% among the study participants [bib_ref] Functional disorders of the lung and symptoms of respiratory disease associated with..., Neghab [/bib_ref] [bib_ref] Respiratory symptoms and lung function in never-smoking male workers exposed to hardwood..., Bislimovska [/bib_ref] [bib_ref] Lung function in Pakistani wood workers, Meo [/bib_ref] [bib_ref] Lung function: Occupational exposure to wood dust, Baran [/bib_ref]. However, the result of our study is in contrast with studies done among Danish furniture workers which shows a reduced lung function [bib_ref] Cross-shift and longitudinal changes in FEV1 among wood dust exposed workers, Jacobsen [/bib_ref]. According to the recommendation of Global Initiative for Chronic Obstructive Lung Disease (GOLD) the ratio of FEV 1 /FVC < 70% confirms the presence of persistent airflow limitation.
We also evaluated the FEF values in this study, as low FEF 25-75% value might be associated with asthma [bib_ref] Forced expiratory flow at 25-75% as a marker for airway hyper responsiveness..., Raji [/bib_ref]. The dust in the particleboard factory is made of organic particles as it comes from the Eucalyptus tree, and organic dust is known to cause asthmatic conditions [bib_ref] Asthma and other respiratory symptoms in New Zealand pine processing sawmill workers, Douwes [/bib_ref]. However, the reason for the lack of reduced lung function variables in the particleboard factories in this study, might be that the workers had been working in the factories for very few years. The present study showed a high level of dust exposure. The geometric mean level of dust was 4.66 mg/m 3 , which is higher than the recommended limit values for inhalable wood dust of 1 mg/m 3. Endotoxin was also documented in the particleboard factories, although the exposure levels were below the recommended health based standard of 90 EU/m 3 set by the Dutch experts. Formaldehyde was also measured in the particleboard factories, with a wide range of exposure (<0.2-5 ppm). However, the exposure time of the workers in the particleboard factories might be too short for the development of chronic lung diseases detectable by spirometry. The average service time for dust and chemicals for these workers was short, only 4 years.
To our knowledge this is the first study done among workers manufacturing particleboard from Eucalyptus to assess the prevalence of respiratory symptoms and to measure lung function status. We selected a control group from a water bottling factory, not from the general population to reduce bias that can be attributed due to baseline characteristics such as socio-demographic and economic differences. Another strength of the study is the high response rate of the participants. Furthermore, multiple regression and poisson regression was applied to control for possible confounders for lung function and respiratory symptoms. The lung function measurements were done using calibrated and sensitive portable spirometry equipment following American Thoracic Society recommendation, which should increase the validity of these results. The data collection was performed addressing one-by-one workers in a place without others listening, to reduce any possible information bias. However, there might still be a bias present, as the workers in the particleboard factory may have reported more symptoms due to their wish for an improved work environment. The size of such a bias is unknown to us. Also, the workers may have caused a recall bias in the respiratory symptom assessment, as symptoms might not be easy to remember.
Our results were based on a cross sectional study and share the limitation of this study design. The study is not conclusive concerning any cause-effect relationship between inhalable wood dust, endotoxin and formaldehyde exposure in the particleboard factories and respiratory symptoms. Longitudinal studies are needed to confirm a possible relationship between these factors. There could also be other unknown predictors present, which are not addressed in the present study. Also, the study might suffer from a healthy worker effect and young age of the workers which may affect the validity of the result regarding the lung function parameters. Studies where the workers have longer service time would be of interest.
The particleboard workers that did not use proper personal protective equipment (PPE) mainly face mask during work. The absence of proper PPE in this working environment with high dust exposure may cause respiratory health problems of the workers in the future. With the present knowledge about the high dust levels in these factories, respiratory health protection is recommended among the workers, to avoid the development of any adverse health effects due to the dust exposure.
The finding is limited to male particleboard workers due to the low number of female production workers. Therefore, the finding is valid only for male particleboard workers with similar work settings in developing countries.
# Conclusions
Particleboard workers in Ethiopia, exposed to wood dust, endotoxin and formaldehyde, had higher prevalence of respiratory symptoms than the controls, i.e., water bottling workers. However, lung function did not appear to be different among particleboard workers and controls. A longitudinal study is recommended to explore the cumulative impact of dust, endotoxin and formaldehyde exposure on respiratory health of particleboard workers. However, we also recommend respiratory health protection of workers with high dust exposure levels, as this type of protection was not used in the factories.
[table] Table 1: Demographic and anthropometric characteristics of 74 particleboard workers (Exposed) and 73 water bottling workers (Controls) in Ethiopia. [/table]
[table] Table 2: Prevalence of respiratory symptoms among 74 particleboard workers (Exposed) and 73 water bottling workers (Controls) in Ethiopia. [/table]
|
10.1371/journal.pone.0289478
|
CCBY
|
10399790
|
37535609
|
s2orc_pubmed_articles
|
Episodic future thinking in type 2 diabetes: Further development and validation of the Health Information Thinking control for clinical trials
Episodic Future Thinking (EFT) reduces delay discounting and may have the potential as a clinical tool to increase the likelihood of health-promoting behaviors. However, evaluations of EFT in clinical settings require control conditions that match the effort and frequency of cue generation, as well as participants' expectations of improvement. The Health Information Thinking (HIT) control addresses these issues, but how this control affects delay discounting in individuals with diabetes and obesity when utilizing diabetesmanagement specific health-information vignettes is unknown. Moreover, little research has explored whether EFT reduces delay discounting in individuals with type 2 diabetes. To this end, we examined the impact of EFT, HIT, and a secondary no-cue control condition (NCC; assessments as usual) on delay discounting in 434 adults with self-reported type 2 diabetes and obesity recruited using Amazon Mechanical Turk. After completing an initial screening questionnaire, eligible participants reported demographics, then were randomized to EFT, HIT, or NCC conditions. Following the generation of seven EFT or HIT cues, participants assigned to EFT or HIT conditions completed a delay discounting task while imagining EFT or HIT cues; no-cue participants completed the task without cues. EFT participants demonstrated significantly lower delay discounting levels than HIT or NCC participants; no differences in delay discounting between HIT and NCC participants were observed. These results suggest that engaging in EFT, but not diabetes-specific HIT, results in lower delay discounting in adults with type 2 diabetes and obesity. This provides further evidence for the appropriateness of the HIT control for clinical trials examining the effect of EFT on delay discounting in adults with self-reported type 2 diabetes.
[formula] a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 [/formula]
# Introduction
Episodic future thinking (EFT), the vivid imagining of personal future events, has been shown to reliably reduce delay discounting, the devaluation of future rewards [bib_ref] Episodic future thinking, Atance [/bib_ref] [bib_ref] Experimental reductions of delay discounting and impulsive choice: A systematic review and..., Rung [/bib_ref]. Higher delay discounting rates are associated with poorer glycemic control and diabetes-related self-care [bib_ref] Time preferences, diabetes self-management behaviours and outcomes: a systematic review, Madsen [/bib_ref]. Importantly, a change in delay discounting over one year was predictive of increased HbA1c (glycated hemoglobin, a measure of retrospective blood sugar within the last 3 months) in adults with prediabetes, while controlling for demographic variables and body mass index (BMI; [bib_ref] Role of delay discounting in predicting change in HBA1c for individuals with..., Epstein [/bib_ref]. In recent years, researchers have examined the viability of EFT as an intervention to decrease delay discounting and increase engagement in health-promoting behaviors. In basic laboratory arrangements, EFT has been shown to reduce caloric consumption [bib_ref] The Future Is Now: Reducing Impulsivity and Energy Intake Using Episodic Future..., Daniel [/bib_ref] [bib_ref] Episodic Future Thinking reduces delay discounting and energy intake in children, Daniel [/bib_ref] , calories purchased in an online grocery shopping task [bib_ref] Episodic future thinking and grocery shopping online, Hollis-Hansen [/bib_ref] , and behavioral economic demand for fast food [bib_ref] Bleak Present, Bright Future: Online Episodic Future Thinking, Scarcity, Delay Discounting, and..., Sze [/bib_ref]. EFT has been similarly efficacious in naturalistic settings, reducing caloric consumption [bib_ref] Episodic future thinking reduces eating in a food court, O'neill [/bib_ref] and calories mothers purchase while grocery shopping. Finally, in clinical settings, EFT has been shown to reduce alcohol consumption [bib_ref] Future thinking to decrease real-world drinking in alcohol use disorder: Repairing reinforcer..., Athamneh [/bib_ref] and increase medication adherence in adults with type 2 diabetes mellitus (T2DM;; however, at least one other translational effort to incorporate EFT as a component of a clinical intervention has been less successful [bib_ref] Effects of 6-month episodic future thinking training on delay discounting, weight loss..., Epstein [/bib_ref].
Regardless of the translation setting or phase, studies examining EFT's effects typically include a control condition for comparison. A common control condition in basic laboratory settings is episodic-recent thinking (ERT), the vivid remembering of recent personal events (i.e., within the past few days). While ERT is an appropriate control in the laboratory or naturalistic settings, repeated (i.e., daily or multiple times per day) engagement in ERT over weeks or months is impractical in clinical settings. To keep cues related to recent events, ERT cues would need to be generated often; matching EFT cue generation to this schedule would be prohibitive. Additionally, control conditions in clinical experiments are more likely to promote adherence to the intervention and minimize demand characteristics if they equate expectation of improvement between groups. Thus, participants should view the control condition as clinically relevant or potentially helpful as the active participants view the potential helpfulness of the intervention. However, participants engaging in ERT do not report the belief that ERT will increase the likelihood of engagement in health-promoting behaviors or decrease delay discounting, which may limit adherence in clinical trials.
To address these limitations, Rung and Epsteindeveloped Health Information Thinking (HIT), a novel control condition well suited for long-term investigations on the effects of EFT on delay discounting and other health behaviors in clinical settings. In an online experiment, Rung and Epsteindemonstrated that participants (a general sample of 254 Amazon Mechanical Turk workers] engaging in EFT had significantly lower delay discounting than participants engaging in ERT or HIT. In the HIT control, participants read and responded to health-information vignettes and generated HIT cues by describing their reactions. HIT participants also rated the information on several dimensions, including how much they liked learning the information, the importance of understanding the information, how exciting it was to learn the information, and how useful it was to learn the information. This approach results in personalized summaries of and reactions to health information, which mirrors common survey-guided EFT cue generation methods. Additionally, the health-information vignettes may be tailored for a given clinical condition, increasing the likelihood that individuals with said condition will view HIT cues as a relevant intervention component. Thus, the HIT condition is likely to control for participants' expectation of improvement and allows for matching cue generation and regeneration schedules, making HIT a potentially powerful control condition for long-term clinical studies examining the effect of EFT on delay discounting and health behaviors.
Although Rung and Epsteinshowed that the HIT condition did not significantly impact delay discounting, the authors recruited participants from the general population and provided health information about a broad range of topics (e.g., effects of alcohol on sleep, recognizing symptoms of depression, understanding nutrition labels). In contrast, studies exploring the clinical effects of EFT have primarily focused on specific health behaviors in individual clinical populations (e.g., diet and physical activity in patients with prediabetes). It is unknown if a modified form of the HIT condition in which the health information is relevant to the prevention or treatment of a specific clinical target and population (e.g., weight loss and glycemic control in patients with T2DM) would yield different conclusions. Specifically, individuals with T2DM and obesity who generate HIT cues based on T2DM-related health-information vignettes (e.g., self-monitoring food, aerobic activity, nutrition labels) may react to these cues with increased motivation to engage in health-promoting behaviors, potentially spurring prospection and reducing delay discounting.
To further develop HIT as a clinical control for this population, the present study describes an online experiment in which participants with self-reported T2DM and obesity were recruited using Amazon Mechanical Turk and randomized to EFT, HIT, or a no-cue control (NCC) condition. EFT and HIT participants generated cues using a self-guided survey, while NCC participants did not generate cues. Afterward, participants completed a delay discounting task while instructed to vividly imagine or consider their EFT or HIT cues; NCC participants completed the delay discounting task without these instructions. We hypothesized that participants who were instructed to engage in EFT during the delay discounting task would have lower levels of delay discounting (i.e., more frequent selection of larger later rewards) compared to both HIT and NCC participants.
# Materials and methods
## Participants
We recruited 434 participants with self-reported obesity and T2DM using Amazon Mechanical Turk, from February 10 th , 2022, to November 17 th , 2022. The qualification criteria in Amazon Mechanical Turk specified that participants must be located in the US, have completed at least 100 previous human-intelligence tasks, and have an approval rating of at least 99%. To prevent individuals outside of the US from accessing the eligibility screening block or attempting to screen multiple times, we applied three sequential screening criteria to all responses. First, we used IP address screening to prevent participants using a virtual proxy network (VPN) from accessing the screening block [bib_ref] The shape of and solutions to the MTurk quality crisis, Kennedy [/bib_ref]. Secondly, we prevented participants with IP addresses associated with countries outside of the US from accessing the screening block. Thirdly, we prevented participants with IP addresses that had already attempted or completed the survey from accessing the survey a second time. Participants who were not flagged for using a VPN, an IP address outside of the US, or a duplicate IP address were then shown the eligibility screening block. To be eligible, participants must have self-reported height and weight corresponding with a BMI � 30 and self-reported a diagnosis of T2DM. To increase confidence in the truthfulness of self-reported T2DM, we asked participants to indicate any medical conditions (diagnosed by a medical professional) using two separate medical diagnoses questions. Each diagnosis question included the same 13 answer choices (i.e., various diseases and medical conditions; see survey file hosted on GitHub [https://github.com/ jeremiahmbrown/public-remedi-hit-pilot] for complete list) presented in random order. After answering the first diagnosis question, participants answered three other unrelated questions (i.e., height, weight, and age) before answering the second diagnosis question on a separate page. To continue to the informed consent block, the selected answer options in both diagnoses questions had to match, and participants could not have selected more than 4 total diagnoses on either diagnoses questions. Additionally, we prevented participants from continuing if they selected any of the following answer choices on either diagnosis question: Type 1 diabetes, chronic obstructive pulmonary disease (COPD), colorectal cancer, breast cancer, or none of the illnesses in this list. The eligibility criteria regarding BMI, T2DM, and other medical diagnoses were assessed at the same stage in the survey; violating one or more of the criteria rendered participants ineligible and ended the survey. [fig_ref] Fig 1: Participant flow diagram [/fig_ref] a flow diagram of participant enrollment, eligibility, dropout, and completion.
## Plos one
We determined the target sample size using an a priori power analysis. 396 participants (132 per group) provide 95% power to detect a medium effect size difference (Cohen's d = 0.5) in delay discounting between individual pairwise comparisons (e.g., EFT vs. HIT; EFT vs. NCC) following an omnibus ANOVA. We assumed a medium effect size based on values observed by Rung and Epstein. Higher attrition rates were observed in the EFT group than in the HIT and NCC control groups; we discontinued recruitment after obtaining 434 completed responses, resulting in 172 completed responses in the NCC control group, 142 in the HIT group, and 120 in the EFT group.
## Procedure
This research was approved by the Virginia Tech IRB and performed in accordance with the 1964 Declaration of Helsinki. Participants completed the experiment online using Qualtrics survey software. After assessing eligibility as described in the Participants Section, participants provided informed consent by reading an IRB-approved consent information sheet and indicating their agreement; acquiring written informed consent was waived by the IRB. Participants then completed a demographic questionnaire before being randomized to EFT, HIT, or NCC conditions. EFT and HIT participants then generated seven cues according to their condition (participants assigned to the NCC condition did not generate cues) before completing an adjusting-amount delay discounting task [bib_ref] Cross-Cultural Comparisons of Discounting Delayed and Probabilistic Rewards, Du [/bib_ref]. After completing the delay discounting task, participants were thanked for their time and compensated $2.50 via Amazon Mechanical Turk. Additionally, participants could earn a $2.50 bonus payment if participants passed three of four attention checks embedded in the adjusting-amount delay discounting task and generated cues longer than 50 characters (if randomized to the EFT or HIT conditions).
Demographics questionnaire. Participants answered questions regarding their age, income, gender, and other sociodemographic characteristics. Additionally, participants reported their most recent HbA1c reading and indicated their current thoughts and actions regarding losing weight and better managing diabetes via a contemplation ladder [bib_ref] The Contemplation Ladder: Validation of a Measure of Readiness to Consider Smoking..., Biener [/bib_ref]. The ladder ranged between zero and ten, with zero representing "I have no thoughts or plans to lose weight and better manage my blood sugar" and ten representing "I am taking actions to lose weight and better manage my blood sugar".
EFT cue generation. Participants randomized to the EFT condition generated EFT cues using a self-guided cue generation survey (see Brown & Stein, for details on this method). Participants first identified personally meaningful possible future events using single sentence descriptions (e.g., "In one month, I am celebrating my wife's birthday"), then rated those events on four dimensions: how much they enjoy the event, how important the event is, how exciting the event is, and how vividly they can imagine experiencing the event. Afterward, participants generated additional vivid, episodic details describing that event, creating a future thinking cue. For example, a completed EFT cue may read, In one month, I am celebrating my wife's birthday. We are having a nice dinner with friends and family at a local restaurant. I am asking her about her year and what she is looking forward to this upcoming year. I am happy to be surrounded by friends and family.
Participants generated seven EFT cues across seven future time frames: one month, three months, six months, one year, three years, five years, and ten years. These time frames matched the time frames used in the adjusting-amount delay discounting task.
HIT cue generation. Participants randomized to the HIT condition generated seven cues based on T2DM-related health-information vignettes; specifically: aerobic physical activity, ultra-processed foods, self-monitoring, glycemic index, energy density, variety (of food), and strengthening exercises. Participants first read the vignette, then described (using a single sentence) one specific piece of information learned while reading. Afterward, participants rated their description on four dimensions: how much they liked learning the information, how important it was to learn the information, how exciting it was to learn the information, and how useful it was to learn the information. This process was repeated for the remaining vignettes. To complete the HIT cues, participants reread the vignettes and provided additional details to their original descriptions. Participants were prompted to include details regarding how the information fits into their existing knowledge, what the information made them think about, who or what the information might be useful for, and how the information made them feel. For example, a complete HIT cue generated after reading the vignette about strengthening exercises may read, I learned that strength training can help improve blood sugar level and decrease fat mass, along with several other benefits. I already knew many of the benefits of strength training, but this reinforced several concepts. The information made me think about my own strength training schedule. The information is useful for anyone who may not know how or why to start strength training. The information reinforced my love for strength training exercises Examples of HIT cues generated from other health-information vignettes can be found in the supplementary materials.
Delay discounting task. Participants completed an adjusting-amount delay discounting task for a $1000 (USD) gain, including seven-time frames displayed in random order: one month, three months, six months, one year, three years, five years, and ten years [bib_ref] Cross-Cultural Comparisons of Discounting Delayed and Probabilistic Rewards, Du [/bib_ref]. Participants indicated their preferences between a smaller monetary reward available immediately or a larger one available after a delay. Depending on participants' choices, the amount of the smaller reward was titrated upwards or downwards after each trial; participants made a total of six choices within each block. Four attention checks were embedded into the task, asking participants to choose between receiving $500 now and $1000 now; these checks were the final choices during the one-month, six-months, three-years, and ten-years blocks. During the discounting task, EFT participants were instructed to engage in EFT, while HIT participants were instructed to engage in HIT; NCC participants completed the delay discounting task without additional instructions.
EFT-cued delay discounting. To elicit EFT during the discounting task, EFT participants were shown one EFT cue before beginning one block of choice trials in the discounting task. Additionally, the single-sentence descriptions for each cue were presented during each trial within a block. Participants were instructed to indicate their preferred option while imagining their single-sentence description. EFT cue time frames were matched with the delay discounting time frames (i.e., participants engaged with their one-month EFT cue while completing the one-month block in the discounting task).
HIT-cued delay discounting. To elicit HIT during the discounting task, HIT participants were shown one complete HIT cue before beginning one block of choice trials in the discounting task. Additionally, the single sentence description of the specific piece of information first identified in the HIT cue generation process (e.g., "I learned that muscle strengthening is recommended for those suffering from diabetes") was presented during each trial within the block. Participants were instructed to indicate their preferred option while imagining their single-sentence description. As HIT cues do not involve future time frames, cues were matched based on the order of appearance; the HIT cue generated first was paired with the first future time frame in the discounting task (i.e., participants were exposed to aerobic physical activity cues while completing the one month block).
# Data analysis
Data were accessed continually for research purposes from February 2022 to November 2022 to monitor study progress and for interim analyses, and from November 2022 to April 2023 to conduct final analyses reported in this manuscript. All data analyses were completed using R (version 4.2.1;and RStudio. The tidyverse package was used to wrangle, clean, and shape data; the gtsummary package was used to create [fig_ref] Table 1: Demographics of participants who completed all study procedures, overall and by group... [/fig_ref] ; the ggplot2 package was used to create all figures [bib_ref] Reproducible Summary Tables with the gtsummary, Sjoberg [/bib_ref] [bib_ref] Welcome to the Tidyverse, Wickham [/bib_ref]. We compared EFT and HIT participant's average ratings of cues to test for group differences in liking/enjoyment, importance, and excitement using three t-tests (i.e., one for each cue characteristic; ratings of usefulness generated by HIT participants and ratings of vividness generated by EFT ratings could not be compared). We controlled for type I error rate using Bonferroni correction (α = .0167). To quantify delay discounting, ordinal area under the curve (AUC) values were calculated using the discAUC package [bib_ref] An alternative approach to calculating Area-Under-the-Curve (AUC) in delay discounting research, Borges [/bib_ref]. Higher values of AUC reflect lower levels of delay discounting. To assess if delay discounting varied between EFT, HIT, and NCC groups, we performed a one-way ANOVA with Tukey HSD post hoc comparisons. As a sensitivity analysis, we performed these tests twice: once including only ordinal AUC values from participants who correctly answered three out of four attention checks during the delay discounting task, and again including all ordinal AUC values (see supporting information). To test for differences in delay discounting attention check passing rates The income variable has been recoded as a continuous variable; see note in data analysis section. 1 Median (IQR); n / N (%) between groups, we used a logistic regression to predict attention check passing status as a function of group. To test for differences in study attrition between groups, we used Fisher's exact test. We used a significance level of α = 0.05 to determine statistical significance for all tests, unless otherwise stated. Income was self-reported using an ordinal scale (e.g., $10,000-19,999); in order to display income data in [fig_ref] Table 1: Demographics of participants who completed all study procedures, overall and by group... [/fig_ref] , income was recoded as a continuous variable by assigning participants a value in the center of their selected category (e.g., participants who reported $10,000 -$19,999 were recoded as $14,999.50). Deidentified raw data and R code used to generate results are available on the authors' Github page (https://github.com/ jeremiahmbrown/public-remedi-hit-pilot).
# Results
Demographics [fig_ref] Table 1: Demographics of participants who completed all study procedures, overall and by group... [/fig_ref] depicts demographic information of participants who completed the experiment; groups were not significantly different in regards to age, BMI, income, contemplation ladder score, self-reported HbA1c, and gender.
## Cue ratings
Cue ratings for the liking/enjoyment characteristic were significantly higher for
## Attrition
As depicted in the flow diagram [fig_ref] Fig 1: Participant flow diagram [/fig_ref] , many eligible participants discontinued after randomization (i.e., during cue generation or the discounting task that followed) and could, therefore, not be included in the analysis of delay discounting. Of the 174 randomized EFT participants, 120 (69%) completed the study (54 voluntarily withdrew after randomization). Of the 175 randomized HIT participants, 142 (81%) completed the study (33 voluntarily withdrew after randomization). Of the 175 randomized NCC participants, 172 (98%) completed the study (3 voluntarily withdrew after randomization). Fisher's exact test was used to determine if the proportion of participants randomized to each group compared to participants who completed the study in each group were different. Based on the p-value obtained from the test (p < .001), we reject the null hypothesis (i.e., that the number of participants who voluntarily withdrew vs. completed the study after randomization were the same between groups). Pairwise comparisons using Fisher's exact test revealed significant differences in attrition rates between all groups (EFT vs. HIT: p = .002; EFT vs. NCC: p < .001; HIT vs. NCC: p = < .001).
# Discussion
In this online experiment, we made two contributions to the literature: 1) the HIT control (using cues generated in response to T2DM related health-information) does not result in lower delay discounting compared to a no-cue control condition in adults with self-reported obesity and T2DM, and 2) brief engagement in EFT does result in lower delay discounting compared to the HIT control and a no-cue control in adults with self-reported obesity and T2DM. We will discuss both contributions, study limitations, and future directions.
Health Information Thinking appears to be a suitable control for clinical trials involving repeated engagement in EFT as an intervention to reduce delay discounting. In this clinical sample (adults with self-reported obesity and T2DM), engaging in HIT using cues generated
## Plos one
from T2DM-related health-information vignettes did not result in lower delay discounting relative to participants engaging in EFT. Participants engaging in EFT exhibited significantly greater ordinal AUC (i.e., lower delay discounting) than HIT and NCC conditions. No differences in ordinal AUC were observed between HIT and NCC participants, suggesting that HIT is an inert intervention concerning delay discounting. These results are consistent with Rung and Epstein, who showed that the HIT and ERT control conditions produced comparable levels of delay discounting. These findings have important implications; namely, that HIT does not appear to influence delay discounting, thus making it appropriate as a control condition in clinical studies designed to reduce delay discounting and influence health behavior. Likewise, the HIT condition is structurally similar to the EFT intervention, as the number of vignettes and cues can be matched to the future events used in EFT. This provides a strong control for effort and time in clinical studies, as participants in both the active intervention and control groups would be scheduled to complete cue generation at approximately the same frequency.
These results suggest that EFT may be an efficacious intervention to reduce delay discounting in adults with obesity and T2DM. To date, only one multiple-baseline small-n study has examined the effects of EFT on medication adherence and delay discounting in adults with T2DM. This prior study reported a large effect size difference in discounting between baseline and EFT conditions (ES = 1.2), although the small sample size (n = 4) prevented inferential statistics. Thus, the present study's findings are important to the literature examining EFT as a clinical intervention for individuals with T2DM. However, as these findings are limited to acute (i.e, single exposure) EFT and delay discounting, future research should examine the effects of repeated (i.e., multiple exposures) EFT on delay discounting, health behaviors, and health outcomes.
Although the present study is the first large-scale examination of EFT in adults with selfreported T2DM, other studies have shown that acute and repeated EFT reliably reduces or results in lower delay discounting compared to control conditions in adults with prediabetes [bib_ref] Effects of 6-month episodic future thinking training on delay discounting, weight loss..., Epstein [/bib_ref] [bib_ref] Bleak present, bright future: II. Combined effects of episodic future thinking and..., Stein [/bib_ref] , a condition of elevated blood glucose that increases the risk of developing T2DM [bib_ref] Prediabetes: A high-risk state for developing diabetes, Tabá K [/bib_ref]. While EFT did not impact other health behavior outcomes in the aforementioned studies of adults with prediabetes, other work in adults with overweight or obesity has been promising [bib_ref] The Future Is Now: Reducing Impulsivity and Energy Intake Using Episodic Future..., Daniel [/bib_ref] [bib_ref] Bleak Present, Bright Future: Online Episodic Future Thinking, Scarcity, Delay Discounting, and..., Sze [/bib_ref] [bib_ref] Episodic future thinking reduces eating in a food court, O'neill [/bib_ref] [bib_ref] Mothers' DASH diet adherence and food purchases after week-long episodic future thinking..., Hollis-Hansen [/bib_ref]. Thus, more work is required to explore potential moderators of the efficacy of EFT (e.g., HbA1c, education, BMI) in regard to changing health behaviors.
This study is not without limitations that should be addressed in future research. First, while we are reasonably confident that our screening strategies limited the number of participants who completed the study but do not have a diagnosis of T2DM or obesity, it is likely some unknown number who did not meet our qualifications were able to meet the screening criteria by chance. However, this number is likely to be low, as we implemented emerging best practices for online research utilizing Amazon Mechanical Turk [bib_ref] The shape of and solutions to the MTurk quality crisis, Kennedy [/bib_ref] , and further developed methods to limit the ability to misrepresent a T2DM diagnosis in screening. As a result, less than 3% of participants could continue past the screening questionnaire. Additionally, previous research supports the validity of self-reported T2DM via interview methods, although it is unknown if these results generalize to self-reported T2DM via an online survey [bib_ref] Validity of self-reported diabetes in health interview surveys for measuring social inequalities..., Espelt [/bib_ref] [bib_ref] Validity and Reliability of Self-reported Diabetes in the Atherosclerosis Risk in Communities..., Schneider [/bib_ref] [bib_ref] Validity of self-reported diabetes among middle-aged and older Chinese adults: the China..., Yuan [/bib_ref]. Furthermore, the prevalence of self-reported T2DM in MTurkers aged 40-49 (measured in 2015) was not different than the prevalence of T2DM in a representative sample of Americans (i.e., the 2013 Behavioral Risk Factor Surveillance System; [bib_ref] Crowdsourced Health Data: Comparability to a US National Survey, Yank [/bib_ref].
Second, we observed significantly higher attrition rates in participants randomized to the EFT group compared to participants randomized to HIT or NCC groups; this difference may be due to the increased effort required to generate EFT cues and engage in EFT during the delay discounting task. Importantly, no significant group differences in demographic variables related to delay discounting were observed (i.e., age, BMI, income, education, HbA1c), which suggests that the observed differences in attrition did not introduce bias related to any measured participant characteristic. We note, however, this differential attrition may be less likely to occur in a sample of motivated participants enrolled in a clinical trial using behavioral interventions and EFT to manage T2DM better.
Thirdly, one other potential benefit of HIT was elucidated by Rung and Epsteinbut was not explored in the present study; namely, the ability to control for participants' expectation of improvement. Future research should measure participants' expectations of improvement associated with EFT and HIT conditions. Fourthly, Rung and Epsteinand the present study were unable to examine the effects of HIT (acute or repeated) on other health behaviors (i.e., food consumption, physical activity engagement); future research is required to examine these outcomes in clinical samples and to compare expectations of improvement in the context of a clinical trial. Indeed, our group is currently using the HIT control condition in a clinical trial examining the effects of repeated engagement in EFT (embedded within a behavioral intervention) on change in delay discounting, body weight, and HbA1c in adults with type 2 diabetes and obesity (see https://clinicaltrials.gov/ct2/show/NCT05280925). While data collection is ongoing, observing significant group differences in behavioral outcomes would suggest that the HIT control is inert with regard to delay discounting, health behaviors, and health outcomes. Similarly, observing no differences in participants' treatment acceptability ratings and perceived helpfulness would suggest that expectations of improvement are not different between EFT and HIT.
Fifth, we observed significantly different average participant ratings of EFT and HIT cues on excitement and liking/enjoyment characteristics, although no differences in ratings of importance were observed. While differences in cue ratings between active and control groups are not ideal, EFT vs. HIT cue generation are, by design, different exercises resulting in qualitatively different written descriptions (i.e., a narrative about personal engagement in possible future events vs. a personal reaction to health information). Therefore, differences in EFT and HIT cues on ratings of excitement and liking/enjoyment may be related to the expected differences in content between EFT vs. HIT. Nonetheless, when applied in a long-term RCT involving repeated, daily engagement in EFT or HIT, these content differences may reduce adherence to HIT and, therefore, its suitability as a control condition. Future research should examine possible differences in intervention adherence between EFT and HIT in long-term RCTs.
In conclusion, the present experiment increases our confidence in using the HIT control for clinical examinations of the effect of EFT on delay discounting. The HIT control cues can be generated using health-information vignettes directly relevant to the clinical sample; engaging in HIT using these cues does not result in lower delay discounting relative to a standard delay discounting task. HIT control is a promising candidate for long-term clinical trials involving EFT as an intervention to reduce delay discounting and change health behaviors.
## Supporting information
[fig] Fig 1: Participant flow diagram. Flow diagram depicting screening of MTurkers, assessment of eligibility, randomization, and study completion. Vol. With. = Voluntary withdrawal (i.e., did not continue responding to the survey). https://doi.org/10.1371/journal.pone.0289478.g001 [/fig]
[fig] Fig 2: Group-level mean and individual indifference points. Mean and individual indifference points of participants who passed at least three of four attention checks during the discounting task. Panel A depicts mean indifference points as a function of delay by group assignment, plotted on an ordinal scale. Panels B, C, and D depict mean indifference points within each group (bold lines); transparent lines depict individual subject discounting curves. Error bars represent standard error. https://doi.org/10.1371/journal.pone.0289478.g002 [/fig]
[fig] Fig 3: Ordinal AUC by group. Notched boxplot of grouped ordinal AUC values of participants who passed at least three of four attention checks during the discounting task. Notches represent the median ± 1.58 * IQR/ p n. https://doi.org/10.1371/journal.pone.0289478.g003 [/fig]
[fig] S1: Fig. Group-level mean and individual indifference points, all participants. Mean and individual indifference points of participants, including all participants. Panel A depicts mean indifference points as a function of delay by group assignment, plotted on an ordinal scale. Panels B, C, and D depict mean indifference points within each group (bold lines); transparent lines depict individual subject discounting curves. Error bars represent standard error. (TIFF) S2 Fig. Ordinal AUC by group, all participants. Notched boxplot of grouped ordinal AUC values, including all participants. Notches represent the median ± 1.58 * IQR/ p n. (TIFF) S3 Fig. Mean and SE of characteristic ratings for EFT and HIT cues. (PNG) [/fig]
[table] Table 1: Demographics of participants who completed all study procedures, overall and by group assignment. [/table]
[table] Table: Demographic table of study completers vs. non-completers.(PDF) S1 Appendix. Sensitivity analysis of AUC comparisons. (PDF) [/table]
|
10.3389/fmolb.2022.825706
|
CCBY
|
8920986
|
35300111
|
s2orc_pubmed_articles
|
Carbon Dioxide and the Carbamate Post-Translational Modification
# Introduction
Since its discovery in gas exhaled from the lung in 1757 [bib_ref] A Century of Pulmonary Gas Exchange, West [/bib_ref] [bib_ref] Joseph Black, Carbon Dioxide, Latent Heat, and the Beginnings of the Discovery..., West [/bib_ref] carbon dioxide (CO 2 ) has been recognised as a critical component of biological processes throughout the biosphere. Its contribution to the essential physiological processes of metabolism, photosynthesis, chemosensing, and cellular homeostasis [bib_ref] Mechanisms and Consequences of Oxygen and Carbon Dioxide Sensing in Mammals, Cummins [/bib_ref] has meant that organisms across the three domains of life had evolved mechanisms to sense, transport, and respond to CO 2 [bib_ref] Carbon Dioxide-Sensing in Organisms and its Implications for Human Disease, Cummins [/bib_ref]. Although a great deal of knowledge exists about the physiological processes where CO 2 is produced or consumed, less is known about the direct mechanisms of CO 2 interactions with biomolecules.
One way in which CO 2 has been shown to interact with protein directly is through carbamylation of neutral N-terminal -amino or lysine ε-amino groups . This carbamate post-translational modification (PTM) is critical to regulating oxygen-binding in haemoglobin and the activation of the CO 2 -fixing enzyme RuBisCO. It has been suggested that protein carbamylation could form the basis of a widespread mechanism for biological regulation [bib_ref] CO2 Adducts of Certain Amino Acids, Peptides, and Sperm Whale Myoglobin Studied..., Morrow [/bib_ref] [bib_ref] Carbon Dioxide and Carbamate Formation: the Makings of a Biochemical Control System, Lorimer [/bib_ref]. Computational studies predict that carbamates may be found in more than 1.3% of large proteins.
In this review, we bring together contextual examples of carbamylation and explore recent computational and experimental approaches with the potential to uncover the distribution of protein carbamylation within proteomes.
## Co 2 and haemoglobin
The linked processes of ventilation and metabolism are essential to survival in higher animals. Following ventilation, gas transfer enables oxygen to be provided promptly to the cells and their mitochondria, which is used in the final process of aerobic cellular respiration (oxidative phosphorylation). Oxidative phosphorylation coordinates with the tricarboxylic acid (TCA) cycle to form adenosine triphosphate (ATP) and the metabolites required by the organism for survival, producing CO 2 as a waste product [bib_ref] Mitochondrial TCA Cycle Metabolites Control Physiology and Disease, Martínez-Reyes [/bib_ref]. CO 2 exits the cells to the bloodstream and is transported to the lungs for excretion, thus contributing to pH homeostasis [bib_ref] Mechanisms and Consequences of Oxygen and Carbon Dioxide Sensing in Mammals, Cummins [/bib_ref].
The link between ventilation and metabolism was first made in 1777 by Antoine Lavoisier with the observation that "Eminently respirable air [oxygen] that enters the lung, leaves it in the form of chalky aeroform acids [carbon dioxide]. . . in almost equal volume. . .." [bib_ref] Joseph Black, Carbon Dioxide, Latent Heat, and the Beginnings of the Discovery..., West [/bib_ref]. At this time, it was presumed that the process of metabolism ("slow combustion") was performed within the lung [bib_ref] Joseph Black, Carbon Dioxide, Latent Heat, and the Beginnings of the Discovery..., West [/bib_ref]. By the mid to late 1800s it was clear that O 2 was transported via the blood to tissues (where metabolism occurred) by the formation of a loose, dissociable interaction with haemoglobin (oxyhaemoglobin) [bib_ref] The Nature of Oxyhaemoglobin, with a Note on its Molecular Weight, Barcroft [/bib_ref] [bib_ref] Hemoglobin Expression in Nonerythroid Cells: Novel or Ubiquitous?, Saha [/bib_ref] , and CO 2 was returned to the lung by similar means [bib_ref] Electrodes for Blood pO2 and pCO2 Determination, Severinghaus [/bib_ref] [bib_ref] A Historical Perspective on Protein Crystallization from 1840 to the Present Day, Giegé [/bib_ref].
In 1904 Bohr, Hasselbalch and Krogh measured haemoglobin oxygenation in canine blood and described the sigmoidal (rather than hyperbolic) nature of the oxyhaemoglobin dissociation curve [bib_ref] Ueber einen in biologischer Beziehung wichtigen Einfluss, den die Kohlensäurespannung des Blutes..., Bohr [/bib_ref]. This experiment demonstrated that increasing pCO 2 resulted in a lowered affinity of haemoglobin for O 2 (known as the Bohr effect). Conversely, Christiensen et al. described that increasing pO 2 resulted in a decreased affinity of haemoglobin for CO 2 (known as the Haldane effect). The Bohr and Haldane effects were reversible, and observed in various mammalian systems [bib_ref] The Absorption and Dissociation of Carbon Dioxide by Human Blood, Christiansen [/bib_ref] [bib_ref] Three Classical Papers in Respiratory Physiology by Christian Bohr (1855-1911) Whose Work..., West [/bib_ref]. Combined, the Bohr-Haldane effect results in haemoglobin being an efficient O 2 transporter from the lungs to tissues and CO 2 from the tissues to the lungs [bib_ref] Is Cooperative Oxygen Binding by Hemoglobin Really Understood?, Eaton [/bib_ref]. The nature of the oxyhaemoglobin dissociation curve led to the hypothesis that multiple O 2 binding sites on haemoglobin acted cooperatively [bib_ref] The Nature of Oxyhaemoglobin, with a Note on its Molecular Weight, Barcroft [/bib_ref] [bib_ref] The Combinations of Haemoglobin with Oxygen and with Carbon Monoxide. I, Hill [/bib_ref].
Haemoglobins belongs to a large family of proteins with members distributed across all three domains of life. The first structures (myoglobin and equine haemoglobin) were determined by X-ray crystallography in the 1950s [bib_ref] Structure of Haemoglobin: A Three-Dimensional Fourier Synthesis at 5.5-Å. Resolution, Obtained by..., Perutz [/bib_ref] [bib_ref] Structure of Haemoglobin : An X-ray Examination of Reduced Horse Haemoglobin, Perutz [/bib_ref]. Human adult haemoglobin is a tetramer consisting of two α and two β subunits similar in structure and size. The α and β subunits are formed of seven and eight helixes, respectively (A-H), joined by non-helical segments. Each subunit binds a heme group consisting of a porphyrin ring that coordinate a Fe 2+ ion (capable of binding to O 2 ) by four nitrogen atoms at its centre. The oxygenated and deoxygenated haemoglobin quaternary structures differ. The gap between two polypeptide chains in the haemoglobin molecule narrows when O 2 binds to the Fe 2+ [bib_ref] 1.25 Å Resolution Crystal Structures of Human Haemoglobin in the Oxy, Deoxy..., Park [/bib_ref] [bib_ref] Structure of Haemoglobin: A Three-Dimensional Fourier Synthesis at 5.5-Å. Resolution, Obtained by..., Perutz [/bib_ref] [bib_ref] Structure of Haemoglobin : An X-ray Examination of Reduced Horse Haemoglobin, Perutz [/bib_ref]. The binding of the first O 2 to the haemoglobin subunit enhances the ability of subsequent O 2 molecules to bind to the remaining subunits [bib_ref] A Critical Study of the Direct Method of Measuring the Osmotic Pressure..., Adair [/bib_ref] [bib_ref] The Oxygen Equilibrium of Hemoglobin and its Structural Interpretation, Pauling [/bib_ref]. This knowledge was used to develop the Monod Wyman and Changeux "two-state concerted" model for allostery where deoxygenated haemoglobin exists in a tense (T) state (with relatively low O 2 affinity). When Fe 2+ binds O 2, there is a movement of Fe 2+ into the heme plane, which triggers a transition to the relaxed (R) state. In this R state, the remaining binding sites are more exposed and have an increased O 2 affinity [bib_ref] Tertiary and Quaternary Structural Basis of Oxygen Affinity in Human Hemoglobin as..., Bringas [/bib_ref]. Additional allosteric sites on haemoglobin were available for binding allosteric modulators, including H + , CO 2 , 2,3diphosphoglycerate (2,3-DPG) and Cl − [bib_ref] Stereochemistry of Cooperative Effects in Haemoglobin: Haem-Haem Interaction and the Problem of..., Perutz [/bib_ref] [bib_ref] Allosteric Proteins: Lessons to Be Learned from the Hemoglobin Intermediates, Perrella [/bib_ref] [bib_ref] New Look at Hemoglobin Allostery, Yuan [/bib_ref].
In the 1930s, CO 2 was shown to combine rapidly and reversibly with haemoglobin to form carbaminohaemoglobin [bib_ref] The Direct Chemical Estimation of Carbamino Compounds of CO2with Haemoglobin, Ferguson [/bib_ref] [bib_ref] The Rôle of the Carbamino Compounds in the Transport of CO 2..., Stadie [/bib_ref] [bib_ref] The Carbamate Equilibrium, Stadie [/bib_ref]. It was suggested that carbaminohaemoglobin formation was possible at multiple amino sites on haemoglobin [bib_ref] The Carbamate Equilibrium, Stadie [/bib_ref]. Confirmation of carbamate formation at the N-terminal of the valines of each of the four human deoxyhaemoglobin chains was performed using cyanate-based blocking, which inhibited the uptake of CO 2 by haemoglobin [bib_ref] Inhibition of CO2 Combination and Reduction of the Bohr Effect in Haemoglobin..., Kilmartin [/bib_ref] and confirmed that CO 2 -binding was O 2 -linked (at constant pCO 2 and pH, deoxyhemoglobin forms more haemoglobin-CO 2 than haemoglobin-O 2 ). The CO 2 binding site was demonstrated to occur at the Val-1β site and linked to the O 2binding state of the β-chain [bib_ref] Quantitative Determination of Carbamino Adducts of Alpha and Beta Chains in Human..., Matthew [/bib_ref]. [bib_ref] Identification of the High and Low Affinity CO2-binding Sites of Human haemoglobinIdentification..., Perrella [/bib_ref] found that the adduct formed on the β chain is more prominent than the α chain. Physiologically, the reaction between the α-amino group and CO 2 stabilises the protein's deoxygenated form. [bib_ref] CO2 Adducts of Certain Amino Acids, Peptides, and Sperm Whale Myoglobin Studied..., Morrow [/bib_ref] suggested that carbamino formation may be a general and functionally important phenomenon throughout biology and not limited to haemoglobin [bib_ref] CO2 Adducts of Certain Amino Acids, Peptides, and Sperm Whale Myoglobin Studied..., Morrow [/bib_ref].
one of the most abundant on the planet [bib_ref] The Most Abundant Protein in the World, Ellis [/bib_ref]. Initially discovered in the 1940s by Wildman and Bonner [bib_ref] Along the Trail from Fraction I Protein to Rubisco (Ribulose Bisphosphate Carboxylase-Oxygenase), Wildman [/bib_ref] , these enzymes are represented across the three domains of life and grouped as structurally distinct operational forms based on protein sequence and secondary and tertiary structure [bib_ref] The Transition between the Open and Closed States of Rubisco Is Triggered..., Duff [/bib_ref] [bib_ref] Comparison of the crystal Structures of L2 and L8S8 Rubisco Suggests a..., Schneider [/bib_ref] [bib_ref] RUBISCO: Structure and Mechanism, Schneider [/bib_ref]. Forms I, II and III catalyse the carboxylation (and oxygenation) of ribulose 1,5bisphosphate, while form IV contains RuBisCO-like proteins (RLP), which although sequentially and structurally similar, perform distinct biological functions [bib_ref] Unusual Ribulose 1,5-Bisphosphate Carboxylase/Oxygenase of Anoxic Archaea, Watson [/bib_ref] [bib_ref] Distinct Form I, II, III, and IV Rubisco Proteins from the Three..., Tabita [/bib_ref] [bib_ref] A RuBisCO-Mediated Carbon Metabolic Pathway in Methanogenic Archaea, Kono [/bib_ref].
RuBisCO I is found in plants, cyanobacteria, algae, and some proteobacteria and is responsible for the vast majority of atmospheric CO 2 fixation through the Calvin-Benson-Bassham (CBB) reductive pentose phosphate pathway [bib_ref] Structure and Function of Rubisco, Andersson [/bib_ref] ; the dark reactions of photosynthesis). Most commonly, Form I is a hexadecamer of eight large and eight small subunits (L8S8). There are four Form I subtypes: A and B in cyanobacteria, eukaryotic algae, and higher plants; C and D are found in nongreen algae and phototropic bacteria [bib_ref] Structural Mechanism of RuBisCO Activation by Carbamylation of the Active Site Lysine, Stec [/bib_ref]. It has been demonstrated that although operational RuBisCO forms differ in overall structure, all share a similar catalytic subunit dimer [bib_ref] The Transition between the Open and Closed States of Rubisco Is Triggered..., Duff [/bib_ref] [bib_ref] Comparison of the crystal Structures of L2 and L8S8 Rubisco Suggests a..., Schneider [/bib_ref] [bib_ref] Function, Structure, and Evolution of the RubisCO-like Proteins and Their RubisCO Homologs...., Tabita [/bib_ref] [bib_ref] Distinct Form I, II, III, and IV Rubisco Proteins from the Three..., Tabita [/bib_ref] , formed from the interaction of the large subunits (the C-terminus of the first interacting with the N-terminus of the second to form two active sites with residues from both). It is thought that RuBisCO remains in the open state when receiving substates and eliminating products. Still, when catalytic events occur, it is in a closed state, essentially sequestering the active site from the bulk solvent [bib_ref] The Transition between the Open and Closed States of Rubisco Is Triggered..., Duff [/bib_ref].
RuBisCO requires the catalytic site to be activated [bib_ref] Function, Structure, and Evolution of the RubisCO-like Proteins and Their RubisCO Homologs...., Tabita [/bib_ref] [bib_ref] Distinct Form I, II, III, and IV Rubisco Proteins from the Three..., Tabita [/bib_ref] by sequential carbamylation of a specific lysine side chain followed by Mg 2+ binding [bib_ref] Structural Mechanism of RuBisCO Activation by Carbamylation of the Active Site Lysine, Stec [/bib_ref] [fig_ref] FIGURE 2 |: The RuBisCO active site showing the K201 carbamate [/fig_ref]. Once activated, RuBisCO's role within the CBB is in catalysing the initial stage of carboxylation of ribulose 1,5-bisphosphate (RuBP) and cleavage to form two molecules of 3phosphoglycerate (3-PGA) (one of which is used in RuBP regeneration, the other diverted to sustain biosynthesis of sugars and other high energy compounds). However, once activated, RuBisCO also catalyses a competitive oxygenation reaction, leading to the formation of one molecule of 3-PGA and one molecule of 2-phosphoglycolate (2-PG), which inhibits central carbon metabolism [bib_ref] The Mechanism of Rubisco-Catalysed Oxygenation, Tcherkez [/bib_ref]. RuBisCO is a relatively inefficient enzyme with a low turnover rate [bib_ref] Rubisco Is Not Really So Bad, Bathellier [/bib_ref]. The turnover rate is further reduced by competition from the oxygenation reaction with the potential to reduce carbon fixation by up to 50% [bib_ref] Structural Framework for Catalysis and Regulation in Ribulose-1,5-Bisphosphate Carboxylase/oxygenase, Andersson [/bib_ref] [bib_ref] Catalysis and Regulation in Rubisco, Andersson [/bib_ref]. Therefore, optimisation of the carboxylation action of RuBisCO has been explored to improve crop efficiency and climate-resilient photosynthesis [bib_ref] Structure and Function of Rubisco, Andersson [/bib_ref] [bib_ref] A Short History of RubisCO: the Rise and Fall (?) of Nature's..., Erb [/bib_ref] [bib_ref] Structural and Functional Analyses of Rubisco from Arctic Diatom Species Reveal Unusual..., Valegård [/bib_ref].
Identification of the RuBisCO catalytic site and its activation were, in most part, uncovered through kinetic and physical investigations in the 1960s and 70s. It became apparent that the sequential addition of CO 2 [bib_ref] The Activation of Ribulose-1,5-Bisphosphate Carboxylase by Carbon Dioxide and Magnesium Ions. Equilibria,..., Lorimer [/bib_ref] and Mg 2+ [bib_ref] Electron Paramagnetic Resonance, 1H, and 13C Nuclear Magnetic Resonance Studies of the..., Miziorko [/bib_ref] [bib_ref] The Activation of Ribulose-1,5-Bisphosphate Carboxylase by Carbon Dioxide and Magnesium Ions. Equilibria,..., Lorimer [/bib_ref] was required for activation. It was further demonstrated that this occurred through CO 2 .Mg 2+ .Enzyme complex formation [bib_ref] Electron Paramagnetic Resonance, 1H, and 13C Nuclear Magnetic Resonance Studies of the..., Miziorko [/bib_ref]. Further experiments demonstrated that the CO 2 involved in the activation process was distinct from the substrate CO 2 used in the subsequent carboxylase reaction [bib_ref] Ribulose-1,5-biphosphate Carboxylase. Evidence in Support of the Existence of Distinct CO2 Activator..., Miziorko [/bib_ref] [bib_ref] Carbamate Formation on the .epsilon.-amino Group of a Lysyl Residue as the..., Lorimer [/bib_ref] and that the catalytic site was located on the RuBisCO large subunit [bib_ref] Structure and Function of Chloroplast Proteins, Nishimura [/bib_ref] [bib_ref] Carbamate Formation on the .epsilon.-amino Group of a Lysyl Residue as the..., Lorimer [/bib_ref] [bib_ref] Function, Structure, and Evolution of the RubisCO-like Proteins and Their RubisCO Homologs...., Tabita [/bib_ref]. The hypothesis that activation was through carbamate formation on the ε-amino group of a lysine residue [bib_ref] The Activation of Ribulose-1,5-Bisphosphate Carboxylase by Carbon Dioxide and Magnesium Ions. Equilibria,..., Lorimer [/bib_ref] was confirmed by [bib_ref] Carbamate Formation on the .epsilon.-amino Group of a Lysyl Residue as the..., Lorimer [/bib_ref] , building on the work of (Ephraim [bib_ref] The Chemical Structure of Some Diamine Carbamates1, Katchalski [/bib_ref] and [bib_ref] Studies on the Active Site of the Enzyme Ribulosediphosphate Carboxylase, Akoyunoglou [/bib_ref]. This finding was followed by identifying the carbamylation site on Lys-201close to the Mg 2+ co-factor binding site [bib_ref] The Sites for Catalysis and Activation of Ribulosebisphosphate Carboxylase Share a Common..., Pierce [/bib_ref]. Based on the observed carbamylation of Haemoglobin and RuBisCO, George Lorimer proposed carbamate modification of N-terminal α-amino and lysine εamino groups as the basis of a widespread mechanism for CO 2 detection. However, as observed by Professor Lorimer, there was no available method for trapping these carbamates on protein to enable their identification. The carbamate modification was, therefore, largely forgotten.
## Discovery of other carbamates
Carbamates have been observed on specific lysine side chains in the crystal structures of several proteins, e.g. class D β-lactamase, urease [bib_ref] Substitution of the Urease Active Site Carbamate by Dithiocarbamate and Vanadate, Yamaguchi [/bib_ref] , alanine racemase [bib_ref] Structure of a Michaelis Complex Analogue: Propionate Binds in the Substrate Carboxylate..., Morollo [/bib_ref] , and transcarboxylase 5 S [bib_ref] Transcarboxylase 5S Structures: Assembly and Catalytic Mechanism of a Multienzyme Complex Subunit, Hall [/bib_ref]. Lysine carbamylation can be critical to protein/ enzyme activity. Examples include having a direct role in catalytic activity (class D β-lactamase,, being a cocatalytic determinant, which bridges metal ions, (e.g. urease, [bib_ref] Substitution of the Urease Active Site Carbamate by Dithiocarbamate and Vanadate, Yamaguchi [/bib_ref] , or in promoting (through hydrogen bonding) active site rearrangements for protein function (e.g. alanine racemase; [bib_ref] Structure of a Michaelis Complex Analogue: Propionate Binds in the Substrate Carboxylate..., Morollo [/bib_ref]. [fig_ref] FIGURE 3 |: Defined role of the carbamate PTM in a selection of CO 2binding... [/fig_ref] summarises the carbamate functional role in haemoglobin, RuBisCO and other proteins discussed below.
Class D β-lactamase β-lactam antibiotics are essential in treating bacterial infection, inhibiting bacterial growth by acylating an active-site serine in essential penicillin-binding proteins and preventing the crosslinking of peptide chains to form peptidoglycan. β-Lactamases enable resistance to β-lactam antibiotics. These enzymes hydrolyse compounds containing a β-lactam ring and are classified based on sequence motifs and mechanism of hydrolysis (Class A-D). Class D OXA enzymes are the most diverse and least well-understood β-lactamases [bib_ref] Insights into Class D β-Lactamases Are Revealed by the Crystal Structure of..., Maveyraud [/bib_ref] [bib_ref] β-Lactamases and β-Lactamase Inhibitors in the 21st Century, Tooke [/bib_ref]. The class is widely dispersed, found in gram-negative and gram-positive bacterial species and mobilised horizontally predominantly by plasmids and integrons [bib_ref] Past and Present Perspectives on β-Lactamases, Bush [/bib_ref] [bib_ref] β-Lactamases and β-Lactamase Inhibitors in the 21st Century, Tooke [/bib_ref]. In 2000, the first crystal structures of OXA-10 of Pseudomonas aeruginosa were published [bib_ref] Insights into Class D β-Lactamases Are Revealed by the Crystal Structure of..., Maveyraud [/bib_ref] [bib_ref] Crystal Structure of the Class D β-lactamase OXA-10, Strynadka [/bib_ref]. [bib_ref] Insights into Class D β-Lactamases Are Revealed by the Crystal Structure of..., Maveyraud [/bib_ref] discovered that OXA-10 is a dimeric beta-lactamase. However, the overall topology of OXA-10 class D β-lactamase is like class A, although amino acid sequence identity is low. Significant structural differences were found in the active site of OXA-10 compared to corresponding regions in class A (and C βlactamases). In its native state, OXA-10 Lys-70 is carbamylated. It was suggested that this carbamylation offered a possible relationship between enzyme activation by CO 2 and anion inhibition [bib_ref] Insights into Class D β-Lactamases Are Revealed by the Crystal Structure of..., Maveyraud [/bib_ref].used activity assays and fluorescence-based approaches to confirm that Lys-70 carbamylation is reversible and that OXA-10 β-lactamase depends on Lys-70 for enzyme acylation and deacylation steps in catalysis. This finding was supported by the inability of Lys-70 mutants to support deacylation. It was suggested that the hydrophobicity of the OXA-10 active site lowered the Lys-70 pK a favouring carbamylation [bib_ref] Comparison of -lactamases of Classes A and D: 1.5-A Crystallographic Structure of..., Sun [/bib_ref].
## Urease
Urease plays an essential role in nitrogen metabolism in archaea, bacteria, fungi, plants and invertebrates [bib_ref] Jack Bean Urease (EC 3.5.1.5). Metalloenzyme. Simple Biological Role for Nickel, Dixon [/bib_ref] [bib_ref] Ureases I. Functional, Catalytic and Kinetic Properties: A, Krajewska [/bib_ref] [bib_ref] The Emerging Role of Urease as a General Microbial Virulence Factor, Rutherford [/bib_ref]. It catalyses urea hydrolysis to form carbonic acid (H 2 CO 3 ) and ammonia (NH 3 ) and is important in human disease, plant metabolism, and agricultural ammonia emissions [bib_ref] Ureases I. Functional, Catalytic and Kinetic Properties: A, Krajewska [/bib_ref] [bib_ref] Inhibition of Urease Activity by Different Compounds Provides Insight into the Modulation..., Svane [/bib_ref]. In bacteria, ureases' active site and primary structure are well studied in organisms such as Proteus mirabilis, Helicobacter pylori, and Klebsiella aerogenes [bib_ref] The crystal Structure of Urease from Klebsiella Aerogenes, Jabri [/bib_ref] [bib_ref] Structures of the Klebsiella Aerogenes Urease Apoenzyme and Two Active-Site Mutants, Jabri [/bib_ref]. The mechanism of urease activation is CO 2 dependent in vitro [bib_ref] Chemical Rescue of Klebsiella Aerogenes Urease Variants Lacking the Carbamylated-Lysine Nickel Ligand, Pearson [/bib_ref]. This dependence has been explained by the active site containing two Ni 2+ ions, which are bridged (stabilised) by the carboxyl group of a carbamylated lysine residue (Lys-217) that is essential for urease activation [bib_ref] Structures of the Klebsiella Aerogenes Urease Apoenzyme and Two Active-Site Mutants, Jabri [/bib_ref] [bib_ref] Requirement of Carbon Dioxide for In Vitro Assembly of the Urease Nickel..., Park [/bib_ref].
## Computational approaches for the identification of carbamate sites
To enhance protein carbamate discovery, Jimenez-Morales et al.
(2014) developed a model to predict the uncarboxylated and carboxylated status of lysine residues in proteins. They initially used a training set of 251 proteins (identified by X-ray crystallography) which contained at least one protein subunit with a carboxylated lysine residue (KCX) to investigate the characteristics of the carbamate microenvironment when compared to uncarboxylated (LYS) sites. They observed that the crucial feature of the KCX microenvironment "was the large numbers of packed atoms, water molecules and ions found in proximity to the KCX site chain." Additionally, all KCX sites were buried (inaccessible from the surface), with residues converging structurally at the KCX site dispersed along the protein's primary sequence (so no sequence motif was associated with lysine carboxylation). The KCX residue was often in contact with positively charged ions with mononuclear or bi-nuclear interactions with divalent ions (Zn 2+ , Mg 2+ , Co 2+ , Fe 2+ , Ni 2+ , Mn 2+ ). His and Asp residues were present in all analyses of KCX containing metal-binding sites. No more than one KCX site was observed on any single protein chain. The RuBisCO carbamylated lysine (used in the training set) is typical of a structure where the carbamate binds Mg 2+ coordinated by Asp and Glue residues [fig_ref] FIGURE 2 |: The RuBisCO active site showing the K201 carbamate [/fig_ref]. This information was used to develop a naïve Bayesian model (to predict potential KCX and LYS sites (Predictor of Lysine Carboxylation: PreLysCarb)). The 251-protein data set was used for training/testing (along with sub-sets with redundancy reduction implemented at 40 and 90% sequence identity). They carried out "leave-one-out cross-validation tests" on the three data sets. At 90% sequence identity PreLysCar correctly classified 54/62 KCX sites (87% sensitivity) and 4255/4259 LYS sites (99.7% specificity). Investigating false-positive rates in highresolution protein structures, the model indicated that 11 to 19/ 575 proteins were incorrectly predicted to have a KCX residue (false positive rate of between 1.9 and 3.3%). When PreLysCar was applied to a subset of solved protein structures from the PDB (structures greater than 200 residues, solved by X-ray crystallography, containing 14,261 protein chains after 90% redundancy removal), it predicted that at least 1.3% of proteins with more than 200 amino acids in the PDB could potentially be subject to spontaneous lysine carboxylation. As the model has been trained using previously identified stable carbamates, which are predominantly buried, it may only represent a subset of possible KCX sites [bib_ref] Conformational Changes and Channel Gating Induced by CO 2 Binding to Connexin26, Brotherton [/bib_ref].
What is clear from this analysis is that no consensus sequence enables easy CO 2 -binding site prediction as has proven so successful for phosphorylation sites, for example. This observation is borne out by analysing the growing numbers of experimentally observed carbamates (see following sections) with no clear primary consensus sequence. Future machine learning approaches as experimental data sets increase in size will, therefore, be essential as predictive tools.
## Mass spectrometry approaches for the identification of carbamate sites
Mass spectrometry-based proteomics is a powerful tool for discovering and exploring PTMs. Although carbamylation of neutral N-terminal α-amino or lysine ε-amino groups may be a "general and functionally important phenomenon throughout biology" [bib_ref] CO2 Adducts of Certain Amino Acids, Peptides, and Sperm Whale Myoglobin Studied..., Morrow [/bib_ref] and "form the basis for a widespread mechanism of biological regulation" [bib_ref] Carbon Dioxide and Carbamate Formation: the Makings of a Biochemical Control System, Lorimer [/bib_ref] , the transient and readily reversible nature of this PTM all but renders impossible the use of mass spectrometry-based proteomics workflows (even when soft ionisation techniques are utilised; [bib_ref] Carbamino Group Formation with Peptides and Proteins Studied by Mass Spectrometry, Terrier [/bib_ref]. This ready reversibility limits our discovery of the distribution and contribution of carbamylation to biological functions.
Linthwaite et al., 2018 developed a chemical mechanism of covalently "trapping" carbamates under physiologically relevant conditions to overcome this limitation. In this technique, the triethyloxonium ion (TEO; a crystalline salt soluble under aqueous conditions) stabilises pre-formed carbamate modifications (formed through incubation of proteins or cellular lysates with CO 2 ) by transferring an ethyl group from the oxonium ion to the negatively charged carbamate, forming a covalent bond which can withstand downstream protease digestion and enables CO 2 -binding protein identification by HPLC-ESI-MS/MS analysis. This approach has been validated for non-buried exchangeable carbamate binding sites (where presumably the carbamate is labile and the bound CO 2 exchangeable into the bulk solvent) using individual amino acids and single proteins (previously known to be carbamylated; [bib_ref] The Identification of Carbon Dioxide Mediated Protein post-translational Modifications, Linthwaite [/bib_ref]. This approach has recently contributed to studies of CO 2 interaction with connexin 26 and ubiquitin (Sections 6.1 and 6.2) and shows potential in carbamate discovery through screening cell lysates (Section 6.3).
## Co 2 and connexin 26
Connexins (Cx) have an essential role in intercellular and extracellular communication and, therefore, homeostasis in multicellular organisms. These transmembrane proteins form hexameric connexons or hemichannels (HCs) in the plasma membrane, allowing low molecular weight molecules to transfer across the plasma membrane [bib_ref] Connexin26 Hemichannels with a Mutation that Causes KID Syndrome in Humans Lack..., Meigh [/bib_ref]. In closely apposed membranes, two HCs can dock together to form homomeric or heteromeric gap junctions (GJs), which mediate intercellular communication through the transfer of low molecular weight molecules (<1.1.5 kD), e.g. ions, metabolites and second messengers, between cells [bib_ref] Gap Junctions and Connexin-Interacting Proteins, Giepmans [/bib_ref] [bib_ref] Connexin26 Hemichannels with a Mutation that Causes KID Syndrome in Humans Lack..., Meigh [/bib_ref] [bib_ref] Connexins and Their Channels in Inflammation, Willebrords [/bib_ref]. GJ intercellular communication can be regulated by pH (closed by acidification), transmembrane voltage (opened by voltage potential greater than −20 mV) and calcium concentration (opened by removal of extracellular Ca 2+ ). PTMs such as S-nitrosylation, sumoylation and phosphorylation can directly regulate GJ opening [bib_ref] Connexins and Their Channels in Inflammation, Willebrords [/bib_ref].
Connexin 26 (Cx26) HCs and GJs are found in cells throughout the body, including in the cochlear epithelial network of the ear, keratinocytes of the skin, alveolar epithelium of the lungs, epithelial cells of the GI tract, and chemosensory areas of the brain [bib_ref] Targeted Ablation of Connexin26 in the Inner Ear Epithelial Gap Junction Network..., Cohen-Salmon [/bib_ref] [bib_ref] CO2-dependent Opening of Connexin 26 and Related β Connexins, Huckstepp [/bib_ref] [bib_ref] Connexin Hemichannel-Mediated CO2-dependent Release of ATP in the Medulla Oblongata Contributes to..., Huckstepp [/bib_ref] [bib_ref] Connexin26 Hemichannels with a Mutation that Causes KID Syndrome in Humans Lack..., Meigh [/bib_ref] [bib_ref] Connexins and Their Channels in Inflammation, Willebrords [/bib_ref] [bib_ref] Human Diseases Associated with Connexin Mutations, Srinivas [/bib_ref] [bib_ref] Connexin26 Mediates CO2-dependent Regulation of Breathing via Glial Cells of the Medulla..., Van De Wiel [/bib_ref]. Mutations in the gene encoding Cx26 (GJB2) are relatively common (frequency of carriers~2-4% of the human population) and are linked to nine non-syndromic and syndromic deafness disorders (SDD; [bib_ref] Cx26 Keratitis Ichthyosis Deafness Syndrome Mutations Trigger Alternative Splicing of Cx26 to..., Cook [/bib_ref] [bib_ref] Human Diseases Associated with Connexin Mutations, Srinivas [/bib_ref]. SDDs are associated with visual impairment and dermatological abnormalities, and in some instances, Keratitis-Ichthyosis-Deafness syndrome is underpinned by Cx26-A88V and Cx26-G45E missense mutations [bib_ref] Connexin26 Hemichannels with a Mutation that Causes KID Syndrome in Humans Lack..., Meigh [/bib_ref]. [bib_ref] Connexin Hemichannel-Mediated CO2-dependent Release of ATP in the Medulla Oblongata Contributes to..., Huckstepp [/bib_ref] [bib_ref] Regulation of Ventral Surface CO2/H+-sensitive Neurons by Purinergic Signalling, Wenker [/bib_ref] suggested that Cx26 has a role in mediating the central CO 2 -dependant drive to breathe with HCs enabling ATP release from the medulla oblongata in the absence of extracellular acidification. Exploring the potential molecular mechanism, [bib_ref] CO2-dependent Opening of Connexin 26 and Related β Connexins, Huckstepp [/bib_ref] found that increasing pCO 2 at fixed pH opens Cx26 HCs (and HCs of two related beta-connexins, Cx30 and Cx32).compared the amino acid sequences of Cx26, Cx30 and Cx32 with Cx31 (a connexin that has no sensitivity to pCO 2 ) and hypothesised a carbamylation motif present in Cx26, 30, 32 that was absent from Cx31. Using the existing Cx26 crystal structure [bib_ref] Structure of the Connexin 26 gap junction Channel at 3.5 Å Resolution, Maeda [/bib_ref] , the authors noted that the carbamylation "motif" contained K125 at the end of a subunit's alpha helix where K125 is oriented towards R104 on a neighbouring subunit of the hexamer. Therefore, they hypothesised that if K125 was carbamylated, it could feasibly form a salt bridge ("carbamate bridge") with R104, linking the subunits, and preventing hemichannel closure. By inserting the identified carbamylation "motif" into Cx31, they demonstrated that it was sufficient to form a CO 2 -sensitive hemichannel. If the K125 residue was substituted for an amino acid that could not be carbamylated, CO 2 sensitivity was lost. Using Cx26, they confirmed that K125 and R104 were essential for forming the carbamate bridge. Meigh et al., 2015 used a similar mutationbased approach to explore the potential for alternative mechanisms of bridge formation between residues 125 and 104 in adjacent hexamer subunits. They found they could convert the CO 2 -sensitive hemichannels to a NO/NO 2− sensitive hemichannel using Cx26-K125C with Cx26-R104 or a redox-sensitive hemichannel using the combination of Cx26-K125C and Cx26-R104C, thus suggesting that distinct mechanisms of bridging involving residues 125 and 104 on adjacent hexamer subunits of Cx26 was possible [bib_ref] Rational Design of New NO and Redox Sensitivity into Connexin26 Hemichannels, Meigh [/bib_ref].
Using a combination of carbamate trapping, high-resolution cryo-EM and classification of particles, [bib_ref] Conformational Changes and Channel Gating Induced by CO 2 Binding to Connexin26, Brotherton [/bib_ref] proposed that under physiologically relevant high pCO 2 conditions (90 mmHg), a carbamate was formed on Lys-125, and additionally at two other positions Lys-108 and Lys-122, but not under physiologically relevant low pCO 2 conditions (20 mmHg). They suggested that the shared environment within the cytoplasmic TM2 and TM3 regions of the Cx26 mobile loop favoured CO 2 modification. Classification of particles indicated that the positions of TM2 and TM3 could influence the conformation of the N-terminal helix and found that under high pCO 2 conditions, the N-terminus was more defined than under low pCO 2 conditions. The authors hypothesised that gating is mediated by the movement of the N-terminal helix and its ability to plug the channel under physiologically relevant conditions [bib_ref] Conformational Changes and Channel Gating Induced by CO 2 Binding to Connexin26, Brotherton [/bib_ref].
Recent work byhas shown in contrast to the opening of HCs at high physiological CO 2 that intact Cx26 GJs connecting HeLa cells are closed when pCO 2 is increased from 35 to 55 mmHg. This closing effect is dependent on the same residues (K125 and R104) involved in the CO 2 -dependant opening of Cx26 hemichannels. Further, the action is also directly attributed to a change in pCO 2 rather than to changes in pH. They explained the contrasting action based on the free energy difference between CO 2 -bound and unbound states for HC and GJ . Specifically, it was energetically more favourable for HCs to bind CO 2 in the open state and less energetically unfavourable to close. In contrast, it was energetically more favourable for the gap junction to bind CO 2 in the closed state and then energetically unfavourable to open. The docking of two connexons as an HC provided a close interaction that constrained the conformation of hemichannels docked in a gap junction. Niijar et al., 2020 proposed the need for more information about the structures of Cx26 as free HCs and as components of GJs to investigate conformational differences and provide a greater understanding of the differential modulation of hemichannels and gap junctions by CO 2 . What is clear from these studies is that CO 2 can have a signalling role via carbamate formation.
## Co 2 and ubiquitin
Ubiquitin (Ub) is a highly conserved protein found in all eukaryotic cells. It plays a crucial role in regulating protein activity and degradation through Ub protein covalent conjugation to a lysine side chain on target proteins [bib_ref] The Emerging Complexity of Protein Ubiquitination, Komander [/bib_ref]. Ubiquitylation conjugation occurs through the sequential activity of Ub-activating enzymes (E1), Ubconjugating enzymes (E2), and Ub ligases (E3) [bib_ref] The Increasing Complexity of the Ubiquitin Code, Yau [/bib_ref] [bib_ref] Mass Spectrometry Techniques for Studying the Ubiquitin System, Heap [/bib_ref] and can be reversed by deubiquitylating enzymes (DUBs). Ub can be conjugated into poly-Ub chains through the N-terminus (M1) or at any of the seven conserved lysine residues (K6, K11, K27, K29, K33, K48, and K63) forming single, mixed or branched Ub-chains, resulting in a diverse range of Ub-chain structures [bib_ref] Ubiquitin Modifications, Swatek [/bib_ref] [bib_ref] The Increasing Complexity of the Ubiquitin Code, Yau [/bib_ref] [bib_ref] Mass Spectrometry Techniques for Studying the Ubiquitin System, Heap [/bib_ref]. Distinct Ub-chains are involved in different biological functions [bib_ref] Mass Spectrometry Techniques for Studying the Ubiquitin System, Heap [/bib_ref]. The most abundant and well-studied of these are K48 single linked Ubchains which target proteins for proteasomal degradation [bib_ref] Mass Spectrometry Techniques for Studying the Ubiquitin System, Heap [/bib_ref] , and K63 single linked Ub-chains involved in cell signalling, trafficking and lysosomal degradation [bib_ref] Ubiquitin Acetylation Inhibits Polyubiquitin Chain Elongation, Ohtake [/bib_ref] [bib_ref] Ubiquitin Modifications, Swatek [/bib_ref]. Roles of other less abundant single and mixed linked Ub-chains include regulation of enzymatic activity (K33), cell cycle regulation (K11), inflammation and immune response (M1, K63/M1, K48) and post-replication repair (K6 and K33) [bib_ref] The Emerging Complexity of Protein Ubiquitination, Komander [/bib_ref].
Ub activity can be further regulated by additional PTMs, e.g. acetylation, phosphorylation and SUMOylation [bib_ref] The Evolving World of Ubiquitin: Transformed Polyubiquitin Chains, Morimoto [/bib_ref] [bib_ref] Uncovering the SUMOylation and Ubiquitylation Crosstalk in Human Cells Using Sequential Peptide..., Lamoliatte [/bib_ref] [bib_ref] Programmed Ubiquitin Acetylation Using Genetic Code Expansion Reveals Altered Ubiquitination Patterns, Lacoursiere [/bib_ref]. For example, all Ub lysine residues (except K29) are acetylated under differing cellular conditions [bib_ref] Lysine Acetylation Targets Protein Complexes and Co-regulates Major Cellular Functions, Choudhary [/bib_ref] [bib_ref] Quantitative Proteomic Atlas of Ubiquitination and Acetylation in the DNA Damage Response, Elia [/bib_ref] [bib_ref] Programmed Ubiquitin Acetylation Using Genetic Code Expansion Reveals Altered Ubiquitination Patterns, Lacoursiere [/bib_ref]. [bib_ref] Ubiquitin Acetylation Inhibits Polyubiquitin Chain Elongation, Ohtake [/bib_ref] demonstrated that when endogenous Ub is acetylated at K6 and K48, this does not affect the ability of Ub to conjugate with substrate protein but inhibits the elongation of K11, K48 and K63-linked Ub-chains by several E2 enzymes (by neutralising the lysine residue positive charge involved in non-covalent interaction of Ub with specific E2s). This acetylation results in the accumulation of monoubiquitylated substrates in the cell. Using a SILAC based approach, [bib_ref] Ubiquitin Acetylation Inhibits Polyubiquitin Chain Elongation, Ohtake [/bib_ref] confirmed that acetylation of K6 and K48 was linked to the enrichment of chromosome or chromatin related factors, including histone H2B, and that monoubiquitylated H2B was stabilised by the expression of acetylated Ub [bib_ref] Ubiquitin Acetylation Inhibits Polyubiquitin Chain Elongation, Ohtake [/bib_ref] [bib_ref] The Evolving World of Ubiquitin: Transformed Polyubiquitin Chains, Morimoto [/bib_ref]. It has recently been suggested that Ub PTMs add an extra layer of complexity to the 'ubiquitin code' and that this extends beyond currently identified PTMs [bib_ref] The Emerging Complexity of Ubiquitin Architecture, Ohtake [/bib_ref]. recently explored the potential modification of human Ub by CO 2 . They demonstrated that Ub K33 and K48 could be carbamylated under physiological conditions by CO 2 (25 mM CO 2 /HCO 3 at pH 7.4), using the "trapping" method outlined in [bib_ref] The Identification of Carbon Dioxide Mediated Protein post-translational Modifications, Linthwaite [/bib_ref]. Ub carbamylation was confirmed independently by 13 C-NMR, identifying K6 and K63 as additional carbamylation sites. Using in vitro conjugation assays, found that di-Ub formation at K48 was significantly decreased when the CO 2 concentration was increased from the normal physiological reference range (1.8-2.3 mM dissolved CO 2 ) to the hypercapnic range (>2.3 mM CO 2 ).
The effect of elevated CO 2 on Ub-dependant processes was explored and focused on the regulation of nuclear factor κB (NF-κB) [bib_ref] Diverse Roles of the Ubiquitin System in NF-Κb Activation, Iwai [/bib_ref]. NF-κB is a transcription factor central to inflammation, innate and acquired immune responses, nervous system function, and cell survival [bib_ref] Diverse Roles of the Ubiquitin System in NF-Κb Activation, Iwai [/bib_ref] [bib_ref] What Is Nuclear Factor Kappa B (NF-Κb) Doing in and to the..., Albensi [/bib_ref] [bib_ref] The Role of Ubiquitination in NF-kB Signaling during Virus Infection, Song [/bib_ref]. Under elevated pCO 2 , NF-kB mediated transcription is suppressed [bib_ref] NF-κB Links CO2 Sensing to Innate Immunity and Inflammation in Mammalian Cells, Cummins [/bib_ref]. NF-κB is inactive in resting cells, bound in the cytoplasm to the inhibitor of NF-KBs (IκBs) proteins, blocking its transport to the nucleus. There are two known pathways of NF-κB activation, the canonical and non-canonical pathways. In the canonical pathway, the IκB kinase complex (IKK1, IKK2 and NF-κB essential modulator (NEM)) is activated upon interaction with, e.g., inflammatory cytokines or Toll-like receptor (TLR) ligands. This activation results in specific phosphorylation of IκB Ser residues within IκBs. Phosphorylated IκBs are recognised by the Ub ligase complex SCF βTrCPs to generate K48-linked poly-Ub chains and are subsequently degraded by the proteasome. NF-κB is released from IκBs and translocates to the nucleus, where it binds to the DNA consensus sequence of a target gene. TNF-receptor family proteins activate the non-canonical pathway. This activation results in the stabilisation of NIK and its phosphorylation of IKKα. This phosphorylated IKKα induces further phosphorylation of NF-κB2/p100, which forms a complex with RelB. Once phosphorylated NF-κB/p100 is ubiquitinated by SCF βTrCPs E3s generating K48-linked poly-Ub chains. This ubiquitinylation is followed by the partial degradation to p52 by the proteasome. The resulting RelB/p52 heterodimer is translocated to the nucleus [bib_ref] Diverse Roles of the Ubiquitin System in NF-Κb Activation, Iwai [/bib_ref].
When cells carrying an NF-κB dependent GFP reporter were exposed to increasing TNFα concentrations under normal (5% (v/v) CO 2 in air) and elevated CO 2 conditions (10% (v/v) CO 2 in air), NF-kB dependent GFP reporter activity was significantly decreased under elevated CO 2 . Transfection of the same cells with plasmids encoding wild-type Ub, mutant K48R Ub, mutant K63R Ub, or an empty vector was used to address the hypothesis that overexpression of K48R Ub would alter the relative response of the NF-kB pathway to elevated CO 2 . Cells transfected with an empty vector, Wt Ub or K63R Ub, showed an unaltered CO 2 response. In contrast, when cells were transfected with K48R Ub, the effect of increased CO 2 on the inhibition of the NF-κB reporter was ablated, suggesting that Ub K48 may be the target for CO 2 in the NF-KB-dependent transcriptional response to hypercapnia in human cells.
## Mass spectrometry-based approaches for discovery of carbamate sites in the proteome
The TEO-based carbamate trapping method combined with HPLC-MS/MS has also been used to discover carbamylated proteins in whole-cell lysates. A screen of an Arabidopsis thaliana cell lysate, which corresponded to 6% of the total proteome (3614 proteins/ 25,000 proteins), identified eight CO 2 -binding sites (i.e. lipid-transfer protein (Lys-K65), Rubisco Large Subunit (Lys-185), Peroxidase (Lys-262 and Lys-268), FBA1 (Lys-293), eukaryotic aspartyl protease family protein (Lys-251), PSBQA (Lys-109), and Fe Superoxide dismutase 1 (Lys-208) [bib_ref] The Identification of Carbon Dioxide Mediated Protein post-translational Modifications, Linthwaite [/bib_ref]. Similarly, a screen of an Escherichia coli lysates corresponding to 14% of the total proteome (294/4300 proteins) identified six CO 2 -binding sites [bib_ref] A Methodology for Carbamate posttranslational Modification Discovery and its Application in Escherichia..., Linthwaite [/bib_ref]. In this instance, these included proteins involved in cellular processes identified as responsive to CO 2, i.e. assisting in the refolding of stressdenatured proteins, i.e. 60 kDa chaperone, carbamylated at Lys-34; [bib_ref] Proteome-wide Analysis of Chaperonin-dependent Protein Folding in Escherichia coli, Kerner [/bib_ref] , preventing denaturation of DNA under extreme conditions (histone-like DNA HU-α, Lys-67; [bib_ref] The HU Regulon Is Composed of Genes Responding to Anaerobiosis, Acid Stress,..., Oberto [/bib_ref] , and proteins not previously identified in cellular processes responsive to CO 2, i.e. glutamine-binding periplasmic protein (Lys-127), ribose import binding protein RbsB (Lys-45 and Lys-285), and tryptophanase (Lys-121). In the future, this approach could be optimised for rapid screening of proteomes for non-buried exchangeable CO 2 -binding sites by including protein and/or peptide fractionation steps to increase proteome coverage and through the development of mechanisms of carbamate enrichment.
# Conclusion
We can draw some clear conclusions from the studies presented. First, carbamylation is a non-enzymatic PTM that can also be FIGURE 4 | Trapping a protein carbamate with TEO. TEO transfers an ethyl group (red) to the anionic carbamate derived from CO 2 (blue) and protein primary amine (green).
Frontiers in Molecular Biosciences | www.frontiersin.org
March 2022 | Volume 9 | Article 825706 readily reversible. Second, the biological consequences of the carbamate PTM can vary discretely with changing CO 2 . Third, the carbamate PTM is more widespread among proteomes than suspected. Several future challenges arise from these conclusions. First, the extent of carbamylation in a proteome is unknown. More extensive proteomics analyses can address this challenge. However, such studies might benefit from developing enrichment methods and parallel computational approaches for carbamate prediction. Second, once the complement of carbamates in a proteome is known, how do we identify those that might have a functional role in CO 2 sensing and signalling instead of forming through standard physicochemical mechanisms and being functionally neutral? Third, is carbamylation the sole mechanism for CO 2 detection? If not, what other means exist, and how do we identify them? It is clear; there is plenty more to do in CO 2 detection.
# Author contributions
LB and MC conceived, designed and wrote the review manuscript.
# Funding
This review was supported by the Biotechnology and Biological Sciences Research Council grant no. BB/S015132/1.
[fig] FIGURE 2 |: The RuBisCO active site showing the K201 carbamate. The figure shows the active site of Arabidopsis thaliana RuBisCO (PDB: 5IU0) with two large subunit chains shown in green and cyan and bound transition-state analogue 2-carboxy-D-arabinitol-1,5-bisphosphate (2-CAB [/fig]
[fig] FIGURE 3 |: Defined role of the carbamate PTM in a selection of CO 2binding proteins. Frontiers in Molecular Biosciences | www.frontiersin.org March 2022 | Volume 9 | Article 825706 [/fig]
|
10.1039/d0sc03889j
|
CCBY
|
8179348
|
34163978
|
s2orc_pubmed_articles
|
Activating mechanosensitive channels embedded in droplet interface bilayers using membrane asymmetry†
Droplet microcompartments linked by lipid bilayers show great promise in the construction of synthetic minimal tissues. Central to controlling the flow of information in these systems are membrane proteins, which can gate in response to specific stimuli in order to control the molecular flux between membrane separated compartments. This has been demonstrated with droplet interface bilayers (DIBs) using several different membrane proteins combined with electrical, mechanical, and/or chemical activators. Here we report the activation of the bacterial mechanosensitive channel of large conductance (MscL) in a dioleoylphosphatidylcholine:dioleoylphosphatidylglycerol DIB by controlling membrane asymmetry. We show using electrical measurements that the incorporation of lysophosphatidylcholine (LPC) into one of the bilayer leaflets triggers MscL gating in a concentration-dependent manner, with partial and full activation observed at 10 and 15 mol% LPC respectively. Our findings could inspire the design of new minimal tissues where flux pathways are dynamically defined by lipid composition.
# Introduction
The eld of bottom-up synthetic biology aims to reconstitute the form, function and behaviour of biological organisms from self-assembled chemical systems.To this end, different pathways have been explored to create compartmentalised biomimetic microstructures capable of supporting functions such as chemical synthesis, 6-8 environment sensing,information transductionand motility.One route involves the use of lipid monolayer-stabilised water-in-oil (w/o) droplets, where contact between two droplets leads to the spontaneous selfassembly of a lipid bilayer at the interface [fig_ref] Figure 1: Probing MscL function in asymmetric droplet-interface bilayers [/fig_ref]. These structures, which are intended to mimic natural lipid membranes found in biology, are known as droplet interface bilayers . DIBs offer several advantages over conventional planar bilayer systems (such as black lipid membranes (BLMs) or aperture suspended bilayers, including increased stability,compartmentalisation of droplet content and the ability to support droplet volumes spanning three orders of magnitude from ml to pl.Unlike other approaches, the DIB platform also offers the option to assemble bilayer networks by connecting additional lipid-monolayer coated droplets in series, enabling the construction of minimal articial tissues.By supplying different lipids to each droplet compartment,DIBs also offer a route for controlling membrane asymmetry, a feature that is ubiquitous in native biological membranesand is essential in facilitating core biological functions such as apoptosis and phagocytosis.This range of features combined with the ability to take quantitative measurements using either uorescence microscopy or by electrophysiology (supported by hydrogel-coated silver/silver chloride electrodes inserted into each droplet) have led to the use of DIBs in a number of different studies concerning the properties of lipid membranes, 24-26 the permeability of drug candidates/agrochemical compoundsand the incorporation of membrane proteins.By using the DIB as an environment for the reconstitution of membrane proteins, the interplay between the protein and lipid bilayer can be interrogated ex vivo independently of the complex interaction networks found in cell biology. To-date multiple classes of membrane proteins have been reconstituted, including bacterial outer membrane proteins, 21 ion channels 31 and pore-forming peptide oligomers.In these systems, the inserted protein is typically activated by stimuli such as transmembrane potential,tension 36,37 and/or pH,offering the user full control over the molecular ux across the membranea quality that can also be nely tuned through the use of mutants. Membrane asymmetry has also been shown to affect protein function, with an early study by Hwang et al. demonstrating the effect of charge asymmetry on the spontaneous gating probability of OmpG.The bacterial mechanosensitive channel of large conductance (MscL) is a membrane protein of interest due to its relatively large ($3 nm, $3.5 nS) unselective pore size in the openstate. G22C F93W mutants enable the chemical activationand spectroscopic detection of MscL, whereas the V23T 41 and G22S 42 gain-of-function (GOF) mutants offer lower tension thresholds for activation ($6 mN m À1 ). Although both GOF mutants have been reconstituted and activated in DIBs 41-45 with greater ease than their wild-type counterpart, they still require individual droplets to be prepared or manipulated by the user, meaning that ux pathways cannot be dened in real-time. This becomes a problem when constructing droplet networks, especially given that network architecture and composition can be used to dene the ow of molecular information throughout the tissue.To this end, recent work has focused on using external stimuli such as lightor temperature 47 to dynamically control network activation, however this has only been achieved for the water-soluble alpha-toxin alpha haemolysin and not for waterinsoluble membrane proteins, highlighting the need to develop methods that offer new, orthogonal ways to dene information ow across a bilayer network.
Here we show for the rst time in a droplet system that the activation of MscL can be achieved using bilayer asymmetry, i.e. in the absence of any channel activators or applied external pressure. We achieve this by incorporating 1-oleoyl-2-hydroxysn-glycero-3-phosphocholine (LysoPC/LPC) into one of the bilayer leaets [fig_ref] Figure 1: Probing MscL function in asymmetric droplet-interface bilayers [/fig_ref] to generate an asymmetric change in the lateral pressure prole that has been shown previously to activate the MscL channel,and identify protein activity using single-channel electrical measurements [fig_ref] Figure 1: Probing MscL function in asymmetric droplet-interface bilayers [/fig_ref]. Our method to control the full gating of the G22C F93W loss-offunction (LOF) channel in a DIB system using membrane patterning could be applied to other mutants and serve as a new strategy to control molecular ux in droplet networks, helping to design and build minimal tissues capable of increasingly complex information processing.
# Results
## Activity of mscl g22c f93w in lipid vesicles
MscL G22C F93W was expressed in BL21(DE3) E. coli and puried via cobalt-immobilized affinity chromatography as in previous work [fig_ref] Figure 2: LPC triggers MscL activation in lipid vesicles [/fig_ref]. The activity of the expressed MscL channel was tested through reconstitution into 1 : 1 (mol : mol) 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) : 1,2-dioleoylsn-glycero-3-phospho-(1 0 -rac-glycerol) (DOPG) lipid vesicles containing a self-quenching (50 mM) concentration of entrapped calcein.MscL-vesicles were produced via thin-lm hydration, extrusion and detergent mediated reconstitution, before separating unencapsulated calcein from the produced vesicles via size-exclusion chromatography to form vesicles $100 nm in diameter [fig_ref] Figure 3: Membrane asymmetry gates MscL G22C F93W channels reconstituted into DIBs [/fig_ref].
MscL channel activity was then assayed via three different activation methods in parallel: (i) chemical modication of vesicle structure through the addition of LPC [fig_ref] Figure 2: LPC triggers MscL activation in lipid vesicles [/fig_ref] (ii) chemical modication of vesicle structure through the enzymatic activity of secretory phospholipase A2 (sPLA2) 55 and (iii) chemical modication of the MscL channel through addition of the channel activator [2-(trimethylammonium)ethyl]methanethiosulfonate (MTSET). An LPC concentration gradient was added to vesicles AE MscL, and the calcein ux response was monitored over 5 hours [fig_ref] Figure 2: LPC triggers MscL activation in lipid vesicles [/fig_ref]. Vesicles containing the MscL channel displayed calcein ux that increased as a function of increasing LPC concentration, indicating that the lysolipid could successfully activate the channel. When the same LPC concentrations were added to vesicles lacking the MscL channel, negligible calcein ux occurred over the 300 minute experimental lifetime [fig_ref] Figure 2: LPC triggers MscL activation in lipid vesicles [/fig_ref] and S4 †), indicating that the channel is essential for triggered release in response to LPC, and that the expressed MscL protein is active.
Based on these ux measurements, the threshold concentration of LPC necessary for channel activation lies between 1- 5 mol% LPC. This value agrees well with previous investigations of the effect of LPC on MscL reconstituted into vesicles,where less than 10 mol% LPC was necessary for channel activation.To further conrm channel activity, enzymatic and chemical activation of the reconstituted channel was undertaken using sPLA2 and MTSET respectively . Again, the presence of the MscL channel enabled calcein ux responses to enzymatic and chemical (channel labelling) stimuli, whilst negligible release was observed for vesicles lacking the channel, indicating high activity of the channel.
## Lpc can activate mscl channels reconstituted into dibs via membrane asymmetry
Aer conrming the activity of expressed MscL, the effect of LPC on MscL-functionalized DIBs was studied. MscL was reconstituted into vesicles with lower charge better suited for DIB formation (95 : 5 DOPC : DOPG vesicle composition), and used as the droplet-stabilising lipid source (lipid-in) for w/o droplets (1 ml). Droplets were manipulated manually on the tips of agar-coated silver/silver chloride electrodes to assemble DIBs and the membrane was probed using droplet electrophysiology [fig_ref] Figure 1: Probing MscL function in asymmetric droplet-interface bilayers [/fig_ref] and S1 †) exactly as described in our previous work, applying a À100 mV potential across the bilayer.To generate asymmetric DIBs, the rst droplet was formed containing PC : PG : LPC vesicles with increasing LPC concentrations (5, 10 and 15 mol%), whilst a second droplet was formed containing PC:PG vesicles AE MscL [fig_ref] Figure 3: Membrane asymmetry gates MscL G22C F93W channels reconstituted into DIBs [/fig_ref] , B and S9 † for 15, 10 and 5 mol% LPC traces respectively). Asymmetric DIBs incorporating MscL showed LPC-dependent gating behaviour, with 10-fold greater gating events occurring in MscL DIBs containing 15 mol% vs. 10 mol% respectively. We observed negligible gating events in DIBs containing 5 mol% LPC (#events ¼ 42/6/1 for 15/10/5% LPC respectively). These events could be clustered into seven gating states, where each state is dened as a function of the percentage of full channel conductance (S%). Six of these states align with previously proposed conductance states for the channel 57,58 and . A histogram analysis approach was used to characterise the step changes of all traces (see Note S1 for further information †).
The majority of events in our 15% LPC results occurred at S4.5 $14 pA (0.14 nS, n ¼ 23) and S6.6 $21 pA (0.21 nS, n ¼ 4). We additionally observed events occurring at S13 $42 pA (0.42 nS, n ¼ 3), which we assign to the sub-conducting state of MscL observed in our previous work.At times, we saw combinations of such states leading to the observation of events $55 pA [fig_ref] Figure 3: Membrane asymmetry gates MscL G22C F93W channels reconstituted into DIBs [/fig_ref] , which could represent the gating of two separate MscL proteins or a change in the conformational state of a single channel. Statistical testing between these three states observed in 15% LPC DIBs conrms that the events can be clustered into three populations in this manner (p < 0.001), as well as conrming that the S4.5 sub-conductance state was observed in both 10 and 15% LPC DIBs respectively (p < 0.001). As only the lowest two sub-conductance states are observed at 10% LPC, we conclude that 10% LPC is able to only partially gate the channel.
As mentioned above, the observed gating events $42 pA correlate well with our previous work 56 which is established as S12, a subconductance state $12% open conductance.Similarly, the low sub-conductance state of S6.6 has been previously observed in MscL gating studies.We note that subconductance states were maintained for extended lengths in our recorded 15% LPC traces , indicating that the combination of MscL and asymmetric LPC may be useful for sustained ux across a DIB. Interestingly, the most frequent subconductance state observed here (S4.5) has not been observed in previous MscL electrophysiology experiments. Its appearance here is likely due to the combination of using a lossof-function channel mutant under high potential (À100 mV), which has been shown to increase sampling of low conductance states.The asymmetric incorporation of 15% LPC in MscL DIBs also led to the occupation of higher gating states [fig_ref] Figure 4: Large DpA MscL gating events are observed in the presence of 15... [/fig_ref]. The states shown in [fig_ref] Figure 4: Large DpA MscL gating events are observed in the presence of 15... [/fig_ref] -C are attributed to the higher subconductance states of the channel, showing good agreement with previously established conductances for these states and . The open state of the channel (S100 $3.2 nS, n ¼ 3) was observed to occur through these larger subconductance states [fig_ref] Figure 4: Large DpA MscL gating events are observed in the presence of 15... [/fig_ref] , S13 and S14 †), indicative that asymmetry can drive full opening of the MscL channel. We note that LPC ip-op has previously been shown to be negligible on similar timescales to those within our system.Here, LPC ip-op appears insignicant as MscL gating occurs throughout the $30 min experimental timescale [fig_ref] Figure 3: Membrane asymmetry gates MscL G22C F93W channels reconstituted into DIBs [/fig_ref].
If DIBs were produced containing MscL but without LPC, negligible gating events occurred, conrming that the protein is in the closed state in the absence of LPC [fig_ref] Figure 3: Membrane asymmetry gates MscL G22C F93W channels reconstituted into DIBs [/fig_ref]. This agrees with our previous work, where gating of reconstituted MscL in symmetric DIBs was not observed.Similarly, if DIBs were produced without MscL but LPC asymmetry was maintained [fig_ref] Figure 3: Membrane asymmetry gates MscL G22C F93W channels reconstituted into DIBs [/fig_ref] , no gating events were detected, indicating that gating could not be attributed to transient pore formation in the DIB caused by the presence of non-bilayer forming LPC.To probe the role of LPC asymmetry further, control experiments with symmetric DIBs containing 15 mol% LPC were performed. This composition attenuated bilayer formation resulting in poor DIB success rates (n ¼ 1/6). When formed, the recordings indicated bilayer instability such that protein activity could not be delineated from the trace [fig_ref] Figure 1: Probing MscL function in asymmetric droplet-interface bilayers [/fig_ref]. All unltered traces for our complete data set can be observed in -S12 † in the ESI. †
The increased likelihood for occupying sub-conductance states was not observed in previous work where patch-clamp electrophysiology was used to analyse the effect of LPC on MscL.We attribute this to differences in experimental setup between excised bilayer patches in patch-clamp and the DIB membrane utilised in droplet electrophysiology, the high applied potential of À100 mV leading to increased occupancy of subconductance states 57 and the LOF mutant used in the study (WT vs. G22C F93W here). Whilst further optimisation is necessary to generate more digital activation behaviour, our electrophysiology experiments indicate that LPC can be successfully employed to gate MscL reconstituted into asymmetric DIB membranes.
# Discussion
We have presented the rst evidence that bilayer asymmetry alone can be used to gate MscL channels reconstituted into DIBs. Two aspects of our electrophysiology results indicate that LPC-gating of MscL could represent a powerful tool in controlling ux through DIB networks. Firstly, we see extended gating of the MscL channel in LPC bilayers on the minute timescale [fig_ref] Figure 3: Membrane asymmetry gates MscL G22C F93W channels reconstituted into DIBs [/fig_ref] and C, S13 and S14 †) which is ideal for sustained ux across the membrane. This is activated without externally applied tension, and hence could be used without the presence of any external mechanical actuation. The lower subconductance states observed here correlate with our previous electrophysiology work on the same mutant (triggered via MTSET-labelling of the channel)and this has been shown to be sufficient for the ux of calcein (MW ¼ 622.6 Da) across MscL DIBs 62 and gating of solutes $6.6 kDa through MscL reconstituted into lipid vesicles.We can therefore infer that LPCgated MscL DIBs should enable molecular ux with a MWCO between 0.6 and 6.6 kDa (and potentially higher than this based on our higher conductance gating events which generate the $3 nm diameter of the open MscL pore 64 ).
Secondly, the ux we have observed is actuated purely by asymmetry in the lipid bilayer and could be patterned during network assembly, or tuned dynamically using optical tweezers as demonstrated recently using vesicles.Furthermore, LPC ip-op could be used to build pre-dened ux patterns into the network, as ip-op to a symmetric bilayer should close the MscL channel.The slow kinetics of this process 67 could be potentially altered if combined with LPC-chelators such as BSA that have been shown to deactivate MscL channels upon removal of LPC from the membrane.The activation mechanism of MscL in response to asymmetric membrane incorporation of LPC has been previously indicated to occur via asymmetric changes in the lateral pressure prole of the membrane, and not via-stretch-induced tension in the membrane.In our DIB setup, the global curvature of the membrane is insignicant from the perspective of a single channel, whilst all our experiments are conducted without applied pressure or mechanical stimulation. We therefore attribute the gating observed here to an asymmetric curvature stress induced by LPC in the DIB perturbing the lateral pressure prole at the water-lipid interface. Such a change in membrane lateral pressure reduces the gating energy for the channel, enabling spontaneous gating without applied pressure.MD simulations have indicated that the asymmetric presence of LPC can generate areas of high local curvature in the membrane that contribute to channel activation.This would be energetically less likely to occur in DIBs compared to vesicles, as DIBs possess $30-60-fold higher surface tension ($1-2 mN To gain further insight to our system the monolayer surface tension of aqueous droplets stabilised with PC:PG:LPC vesicles containing 0-15 mol% LPC was quantied using droplet shape analysis [fig_ref] Figure 1: Probing MscL function in asymmetric droplet-interface bilayers [/fig_ref]. A linear decrease in surface tension was observed from 0-15 mol%, with the surface tension difference between a droplet containing 0 and 15% LPC found to decrease by 0.74 mN m À1 , from 1.83 AE 0.23 mN m À1 for 0 mol% LPC to 1.09 AE 0.71 mN m À1 for 15 mol% LPC. This tension decrease is expected considering that LPC acts as a surfactant, and measured tensions are similar to those of previously studied asymmetric DIBs (1-2 mN m À1 (ref.). Indeed, a similar decrease in tension in response to LPC has been shown previously for phosphatidic acid-stabilised droplets.These results appear to match well with the predicted effect of asymmetric LPC generating a pressure differential across the leaets of a DOPC membrane.This further indicates that the primary driver of channel activation observed here is leaet asymmetry and not simply a high bilayer surface tension, as DIB surface tensions are an order of magnitude lower than the activation tension of the channel (>12-14 mN m À1 (ref.).
We note that the threshold activation concentration of LPC required for MscL gating differs between the vesicle and DIB model systems: gating occurs from 5 mol% LPC added to MscL vesicles, whilst 10 mol% LPC is necessary when the channel is reconstituted into DIBs. We hypothesise that the increased activation threshold may be due to LPC partitioning between the DIB and droplet monolayers or into the bulk hexadecane.These mechanisms would reduce the effective LPC concentration in the bilayer to minimise the destabilising effect of the lyso-lipid 73 and hence minimise the free energy of the system.
Although we see full gating of the MscL channel in our work, the most frequently observed events appeared to be gating via sub-conductance states. This could be due to LPC partitioning, higher applied voltage or may reect the high activation energy of the mutant used here.The sidedness of MscL insertion may also play a role: MscL is reconstituted without preference into vesicles (and hence into the DIB), and the presence of both orientations in the same membrane may affect which channels are gated by LPC asymmetry as well as channel gating probability. Interestingly, the amphipath 2,2,2-triuoroethanol was recently shown to activate MscL when added to either leaet of a patch-clamp setup,indicating that LPC may also be able to activate both channel orientations in the DIB. Further investigation could be conducted by using protocols which reconstitute MscL in a single orientation into the DIB membrane as well as testing the LPC activation mechanism with wild-type or GOF channel mutants with lower gating energies.Indeed, MscL V23T has shown a higher probability for the open state when reconstituted in DIBs in response to mechanical actuation,and channel activation should be feasible by using LPC to create asymmetric bilayers.
Recently, MscL V23T was shown to possess voltagedependent gating in asymmetric DPhPC:DOPhPC DIBs under mechanical stimulation when negatively hyperpolarized.This was achievable for the V23T mutant due to a dominant dielectric effect compared to the WT channel on account of increased pore solvation. The G22C F93W channel possesses increased pore hydrophobicity (and hence decreased pore solvation 75 ) compared to the WT, reducing the dielectric component of the channel in the presence of an electric eld compared to both V23T and WT. Charge asymmetry is therefore signicantly unlikely to gate either the WT or G22C MscL channels without applied tension, but such asymmetries may affect the onset of MscL gating to an LPC asymmetry. This may be useful as a tool to further differentiate channel function in droplet networks.
# Conclusions
In conclusion, we have shown that LPC-induced asymmetry in DIB membranes can be utilised as a tool for the activation of incorporated mechanosensitive channels in a sustained manner. This work furthers recent application of the MscL protein as a model mechanosensitive channel 'part' for use in bottom-up synthetic biology, 10,76,77 extending its utility in droplet-interface bilayer networks. The combination of responsive protein channels and bilayer asymmetry shown here points a way towards using lipid composition to dene network functiona still underexplored parameter in the design of droplet systems and minimal tissues.
## Conflicts of interest
There are no conicts to declare.
[fig] Figure 1: Probing MscL function in asymmetric droplet-interface bilayers (DIBs) using electrophysiology. (A) DIBs are formed by bringing together two lipid monolayer stabilised water-in-oil (w/o) droplets. As the two monolayers are brought into contact, a lipid bilayer spontaneously forms at the interface. Asymmetric bilayers can be easily assembled by supplying droplets with different lipids. (B) MscL can be incorporated by supplying vesicles containing reconstituted protein to the system, while inserting hydrogel-coated silver/silver chloride electrodes into each droplet facilitates electrical measurements of the membrane using electrophysiology. MscL is activated by the presence of LPC in one of the bilayer leaflets. (C) If symmetric bilayers are generated without LPC, reconstituted MscL channels remain shut, preventing flux of content across the DIB. [/fig]
[fig] Figure 2: LPC triggers MscL activation in lipid vesicles. (A) MscL is reconstituted into calcein-loaded lipid vesicles via detergent-mediated reconstitution. By encapsulating calcein at high concentrations, its fluorescence is fully quenched. Asymmetric insertion of LPC into the outer leaflet of the vesicle membrane is facilitated by the addition of LPC micelles to solution. If MscL is activated by LPC insertion, calcein will diffuse through the channel out of the vesicle where it will become diluted and generate a fluorescent signal as it becomes unquenched. (B) Calcein flux (%) over 5 hours from MscL-vesicles upon the addition of increasing mol% LPC to solution. LPC-dependent calcein flux can be observed indicating successful MscL activation by LPC. (C) MscL is critical for calcein flux. If calcein flux for vesicles AE MscL is assessed 100 minutes after LPC addition, negligible flux is observed at all LPC concentrations for vesicles lacking the channel. Error bars for (B) and (C) represent 1 SD (n ¼ 3). [/fig]
[fig] Figure 3: Membrane asymmetry gates MscL G22C F93W channels reconstituted into DIBs. (A) Asymmetric MscL DIBs containing 15 mol% LPC are formed by incubating LPC-containing vesicles in the left droplet of the DIB (PC : PG : LPC 80 : 5 : 15) and MscL-containing vesicles in the right droplet (PC : PG : LPC 95 : 5 : 0). Each 1 ml droplet additionally contained 20 mM HEPES, 100 mM KCl pH 7.4 buffer. The activation of sub conducting states was observed, clustered into three main gating events $14, 21 and 42 pA that are attributed to sub-conducting states of the MscL channel (n ¼ 4). Further events for 15 mol% LPC discussed in Fig. 4. (B) Reducing the LPC concentration in the asymmetric DIB attenuates channel function. An $10-fold drop in gating events are observed when LPC is reduced to 10 mol%, and no fully open states are recorded for the MscL channel (n ¼ 3). (C) If a symmetric MscL DIB is generated without LPC, no gating events are observed indicating that the MscL channel is closed (n ¼ 2). (D) If asymmetric LPC DIBs are prepared without MscL, no gating events are observed, indicating that in (A) flux across the bilayer is controlled by the response of MscL to membrane asymmetry (n ¼ 3). Cartoons illustrate lipid and protein compositions. Black traces show examples of full, 50 kHz recorded trace for each composition with featured zoom in to indicate channel activity. Zoom in data shown with an applied low pass filter at 2.5 kHz. All traces recorded at À100 mV. The percentage standard error of the mean is denoted by % SE. [/fig]
[fig] Figure 4: Large DpA MscL gating events are observed in the presence of 15 mol% LysoPC. Divided into 4 clusters, large gating events are also observed in asymmetric DIBs formed with 15 mol% LysoPC and MscL. All data show example traces with mean and SE of grouped gating events presented (n ¼ 3) (A-C). Representative gating events at 84, 135 and 252 pA are observed. (D) Full opening of MscL at 323 pA passing through distinct sub states as indicated by sustained regions in the total pA change. Cartoons illustrate lipid and protein compositions. The percentage standard error of the mean is denoted by % SE. m À1 (ref. 26 ) vs. $30 mN m À1 (ref. 70 )). Such an effect may decrease the probability of channel activation in DIBs compared to vesicles assuming local curvatures are truly necessary. Comparative electrophysiology experiments on the same MscL channel in both asymmetric DIBs and patch-clamp setups may help further elucidate the mechanism of LPC-activation of MscL. [/fig]
|
10.1371/journal.pone.0003848
|
CCBY
|
2585478
|
19050761
|
s2orc_pubmed_articles
|
IQGAP1-Dependent Signaling Pathway Regulates Endothelial Cell Proliferation and Angiogenesis
Background: Vascular endothelial growth factor receptor-2 (VEGFR-2) signaling is an obligate requirement for normal development and pathological angiogenesis such as cancer and age-related macular degeneration. Although autophosphorylation of tyrosine 1173 (Y1173) of VEGFR-2 is considered a focal point for its angiogenic signal relay, however, the mechanism of phosphorylation of Y1173, signaling proteins that are recruited to this residue and their role in angiogenesis is not fully understood.Methodology/Principal Findings: In this study we demonstrate that c-Src kinase directly through its Src homology 2 (SH2) domain and indirectly via c-Cbl binds to phospho-Y1057 of VEGFR-2. Activation of c-Src kinase by a positive feedback mechanism phosphorylates VEGFR-2 at multi-docking site, Y1173. c-Src also catalyzes tyrosine phosphorylation of IQGAP1 and acts as an adaptor to bridge IQGAP1 to VEGFR-2. In turn, IQGAP1 activates b-Raf and mediates proliferation of endothelial cells. Silencing expression of IQGAP1 and b-Raf revealed that their activity is essential for VEGF to stimulate angiogenesis in an in vivo angiogenesis model of chicken chorioallantoic membrane (CAM).Conclusions/Significance: Angiogenesis contributes to the pathology of numerous human diseases ranging from cancer to age-related macular degeneration. Determining molecular mechanism of tyrosine phosphorylation of VEGFR-2 and identification of molecules that are relaying its angiogenic signaling may identify novel targets for therapeutic intervention against angiogenesis-associated diseases. Our study shows that recruitment and activation of c-Src by VEGFR-2 plays a pivotal role in relaying angiogenic signaling of VEGFR-2; it phosphorylates VEGFR-2 at Y1173, facilitates association and activation of IQGAP1 and other signaling proteins to VEGFR-2. IQGAP1-dependent signaling, in part, is critically required for endothelial cell proliferation, a key step in angiogenesis. Thus, Y1057 of VEGFR-2 serves to regulate VEGFR-2 function in a combinatorial manner by supporting both diversity of recruitment of angiogenic signaling proteins to VEGFR-2, and its ability to promote angiogenesis.
# Introduction
Activation of vascular endothelial growth factor receptor-2 (VEGFR-2 also called FLK-1/KDR) plays a pivotal role in mediating growth of blood vessels, angiogenesis [bib_ref] Angiogenesis in life, disease and medicine, Carmeliet [/bib_ref]. VEGF stimulation of VEGFR-2 provokes pleiotropic responses in endothelial cells including endothelial cell growth, survival, differentiation, migration, and tube formation [bib_ref] VEGF receptor signalling-in control of vascular function, Olsson [/bib_ref]. Autophosphorylation of tyrosine 1173 (Y1173) of VEGFR-2 is a focal point for activation of angiogenic signal relay of VEGFR-2, as it has emerged as a multi-functional docking site involved in the recruitment of multiple signaling proteins including the p85 regulatory subunit of phosphatidylinositol 3-kinase (p85/PI3-K) [bib_ref] Identification of tyrosine residues in vascular endothelial growth factor receptor-2/FLK-1 involved in..., Dayanir [/bib_ref] [bib_ref] A mechanosensory complex that mediates the endothelial cell response to fluid shear..., Tzima [/bib_ref] , phospholipase Cc1 (PLCc1) [bib_ref] A single autophosphorylation site on KDR/Flk-1 is essential for VEGF-A-dependent activation of..., Takahashi [/bib_ref] [bib_ref] Recruitment and activation of phospholipase Cgamma1 by vascular endothelial growth factor receptor-2..., Meyer [/bib_ref] , Shb [bib_ref] The adaptor protein shb binds to tyrosine 1175 in vascular endothelial growth..., Holmqvist [/bib_ref] , and the Shcrelated adaptor protein, Sck [bib_ref] Sck interacts with KDR and Flt-1 via its SH2 domain, Igarashi [/bib_ref] [bib_ref] The Shcrelated adaptor protein, Sck, forms a complex with the vascular-endothelialgrowth-factor receptor..., Warner [/bib_ref] [bib_ref] Sck is expressed in endothelial cells and participates in vascular endothelial growth..., Ratcliffe [/bib_ref] and the transmission of a range of biological signals to coordinate endothelial cell function and angiogenesis. The critical and direct role of Y1173 in enabling VEGFR-2 to promote angiogenesis was recently revealed by a gene targeting study where mice homozygous for the mutant VEGFR-2 Y1173F knock-in allele died between E8.5 and E9.5 without any organized blood vessels [bib_ref] Essential role of Flk-1 (VEGF receptor 2) tyrosine residue 1173 in vasculogenesis..., Sakurai [/bib_ref] similar to the VEGFR-2 null mice [bib_ref] Failure of blood-island formation and vasculogenesis in Flk-1-deficient mice, Shalaby [/bib_ref]. More recent studies indicate that in addition to Y1173, the kinase domain tyrosine, Y1057 also takes part in VEGFR-2 function and is involved in VEGFR-2-mediated cell proliferation [bib_ref] Tyrosine residues 951 and 1059 of vascular endothelial growth factor receptor-2 (KDR)..., Zeng [/bib_ref]. Y1057 along with Y1173 also is engaged in the recruitment and activation of the ubiquitin E3 ligase, c-Cbl. Phospho-Y1057 contributes to the recruitment of c-Cbl to VEGFR-2 by directly associating with c-Cbl, while phospho-Y1173 indirectly via PLCc1 participates in its recruitment to VEGFR-2 [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref] , resulting in the ubiquitination of PLCc1 and inhibition of angiogenesis in vitro [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref].
The recent study indicates that IQGAP1 associates with VEGFR-2 and its activity is required for proliferation of endothelial cells in vitro [bib_ref] IQGAP1, a novel vascular endothelial growth factor receptor binding protein, is involved..., Yamaoka-Tojo [/bib_ref]. It remains, however, unclear how IQGAP1 interacts with VEGFR-2 since it lacks phospho-tyrosine binding domains such as SH2 and PTB and how its activity is regulated by VEGFR-2. IQGAP1 contains multiple proteininteraction domains including calponin homology domain (CHD), poly-proline-binding domain (WW), calmodulin-binding domain (IQ) and rasGTPase-activating protein (GAP)-related domain (GRD) and tyrosine and serine/threonine phosphorylation sites [bib_ref] IQGAP proteins are integral components of cytoskeletal regulation, Briggs [/bib_ref] and is involved in multiple cellular functions including calcium/calmodulin signaling, MAPK signaling and regulation of cytoskeletal structure, cell adhesion and cell motility [bib_ref] IQGAP1 modulates activation of B-Raf, Ren [/bib_ref] [bib_ref] IQGAP1 binds ERK2 and modulates its activity, Roy [/bib_ref] [bib_ref] IQGAP proteins are integral components of cytoskeletal regulation, Briggs [/bib_ref].
In this study, we uncovered a surprising correlation between phosphorylation of Y1057 and Y1173 and that phosphorylation of Y1057 plays a multitasking role in mediating VEGFR-2-directed angiogenic signaling events in endothelial cells. Phospho-Y1057 recruits c-Src kinase to VEGFR-2, and in part catalyzes phosphorylation of Y1173 via Src kinases. c-Src also bridges IQGAP1 to VEGFR-2 and directly phosphorylates IQGAP1 and promotes endothelial cell proliferation, a key step in angiogenesis.
# Results
## Identification of tyrosine 1057 of vegfr-2 as a binding site for src kinases
Although stimulation of VEGFR-2 in endothelial cells is known to promote activation of Src kinases [bib_ref] The presence of a single tyrosine residue at the carboxyl domain of..., Meyer [/bib_ref] and Src activation is linked to endothelial cell permeability, survival and angiogenesis [bib_ref] Selective requirement for Src kinases during VEGF-induced angiogenesis and vascular permeability, Eliceiri [/bib_ref] [bib_ref] Role of Raf in vascular protection from distinct apoptotic stimuli, Alavi [/bib_ref] , the nature of its association with VEGFR-2, in particular, the putative phospho-tyrosine residue on VEGFR-2 involved in its recruitment is not. Our survey of several primary endothelial cells including, HUVEC, HMVE and BAEC cells showed that c-Src is ubiquitously tyrosine phosphorylated in response to VEGF stimulation [fig_ref] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2 [/fig_ref] and associates with VEGFR-2 in a ligand-dependent manner in these primary endothelial cells [fig_ref] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2 [/fig_ref]. Cell lysates obtained from HEK-293 (human embryonic kidney-293) cells which do not express VEGFR-2 was negative for VEGFR-2 and was used as a negative control [fig_ref] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2 [/fig_ref]. The cytoplasmic region of VEGFR-2 contains multiple tyrosine phosphorylation sites [bib_ref] VEGFR-1 and VEGFR-2: two non-identical twins with a unique physiognomy, Rahimi [/bib_ref] , albeit none of these tyrosine autophosphorylation sites shows obvious sequence similarities to the preferred Src SH2 recognition motif, pYEEI [bib_ref] SH2 domain specificity and activity modified by a single residue, Marengere [/bib_ref]. To analyze association of c-Src with VEGFR-2 we utilized the previously characterized VEGFR-2 chimera [bib_ref] Receptor chimeras indicate that the vascular endothelial growth factor receptor-1 (VEGFR-1) modulates..., Rahimi [/bib_ref] herein called CKR which is created by replacing the extracellular domain of VEGFR-2 with that of human Colony stimulating factor-1 receptor (CSF1R) and expressed in PAE (porcine aortic endothelial) cells as an experimental system. In our previous studies we have extensively characterized CKR with VEGFR-2 in a comparative manner. CKR in term of its ability undergo downregulation, activation, recruitment of signaling proteins and to stimulate cellular responses is identical to VEGFR-2 [bib_ref] Identification of tyrosine residues in vascular endothelial growth factor receptor-2/FLK-1 involved in..., Dayanir [/bib_ref] [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref] [bib_ref] Receptor chimeras indicate that the vascular endothelial growth factor receptor-1 (VEGFR-1) modulates..., Rahimi [/bib_ref] [bib_ref] The carboxyl terminus of VEGFR-2 is required for PKC-mediated down-regulation, Singh [/bib_ref] and data presented in this manuscript. We used this strategy to avoid cross-talk between VEGFR-2 and other VEGF receptors, in particular, VEGFR-1. Our initial observation demonstrated that mutation of tyrosines 799, 820, 925 to phenylalanine (F) and deletion of C-terminus of VEGFR-2 coupled with mutation of tyrosines 1173 and 1212 has no effect on the ability of VEGFR-2 to associate with the SH2 domain of c-Src [fig_ref] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2 [/fig_ref]. In contrast, mutation of Y1052 to glutamic acid (E) and particularly Y1057, located in the kinase domain, severely reduced the binding of VEGFR-2 with SH2 domain of c-Src [fig_ref] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2 [/fig_ref]. GST alone does not associate with CKR . Consistent with its reduced binding capability to E1052, the result showed that E1052 only partially reduces the ability of VEGFR-2 to phosphorylate c-Src [fig_ref] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2 [/fig_ref]. In contrast, the potential of mutant Y1057 (E1057/ CKR) to stimulate Src tyrosine phosphorylation was severely compromised and was near to undetectable [fig_ref] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2 [/fig_ref]. Since both Y1052 and Y1057 are located in the kinase domain we replaced these residues to glutamic acid (E) rather than phenylalanine (F). This type of mutations has been suggested to be useful for characterization of sensitive tyrosine sites on the RTKs (receptor tyrosine kinases) when their mutation to phenylalanine interferes with receptor activation [bib_ref] The presence of a single tyrosine residue at the carboxyl domain of..., Meyer [/bib_ref]. However, unlike these RTKs, single mutation of Y1052 and Y1057 either to E or F did not interfere with activation of VEGFR-2 in the context of CKR (Figures S1 and S2). Furthermore, to make sure that our observation in CKR also is true in the context of non-chimeric VEGFR-2, we mutated Y1057 to F in the full length VEGFR-2, expressed in PAE cells and similarly analyzed for its ability to bind SH2 domain of c-Src. The result showed that VEGFR-2 associates with SH2-Src in VEGF-dependent manner and mutation of Y1057 inhibited association of VEGFR-2 with c-Src [fig_ref] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2 [/fig_ref]. Altogether, the data show that Y1057 of VEGFR-2 is a major c-Src binding site on VEGFR-2 and its presence is critically required for VEGFR-2 to activate c-Src.
c-Cbl associates with c-Src and participates in the interaction of c-Src with VEGFR-2 A recent study indicates that tyrosine 1057 of VEGFR-2 also serves as a binding site for c-Cbl by directly interacting with its tyrosine kinase binding (TKB) domain [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref] , raising an interesting possibility as to whether c-Cbl and c-Src competitively bind to Y1057 or c-Cbl acts to bridge c-Src to VEGFR-2 since c-Cbl is known to interact with c-Src via its SH3 domain [bib_ref] Identification and functional characterization of an Src homology domain 3 domain-binding site..., Sanjay [/bib_ref]. To test these two distinct possibilities, we used PAE cells co-expressing CKR either with wild type c-Cbl or Cbl-N, where the TKB domain is deleted [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref] and analyzed the ability of GST-SH2 domain of c-Src to interact with VEGFR-2. Over-expression of wild type c-Cbl enhanced association of SH2 domain of c-Src with VEGFR-2, where disabling the binding of c-Cbl with VEGFR-2 had no effect on the binding of SH2-Src to VEGFR-2. Quantification of the blots (three independent experiments) showed that indeed interaction of VEGFR-2 with GST-SH2-Src is more than doubled when c-Cbl is over-expressed, where as the deletion of TKB domain almost entirely is inhibited this effect of c-Cbl. Over-expression of either c-Cbl or c-Cbl-N had no effect on the protein levels of CKR or on its tyrosine phosphorylation, 2D). Expression of c-Cbl and c-Cbl-N also is shown. Since up-regulation of c-Cbl by over-expression enhanced association of c-Src with VEGFR-2, we further tested whether silencing expression of c-Cbl by siRNA would reduce interaction of SH2-Src with VEGFR-2. As shown, silencing the expression of c-Cbl in PAE cells reduced the binding of GST-SH2 Src with VEGFR-2. Quantification of the blots (three independent experiments) is also shown which further support the hypothesis that endogenous c-Cbl plays a role in bridging c-Src to VEGFR-2. The silencing effect of c-Cbl-siRNA on the c-Cbl protein level also is shown. It showed be noted that c-Cbl is highly expressed in these cells and siRNA only partially reduced its expression. Altogether, the data show that c-Cbl, in part, facilitates association of Src with VEGFR-2. To firmly establish a direct binding of SH2 domain of c-Src with phospho-Y1057 of VEGFR-2 we synthesized a phosphorylated peptide corresponding to Y1057 and tested its binding ability to GST-SH2 domain of c-Src in an in vitro dot blot assay. The result shows that SH2 domain of Src also directly interacts with phosphorylated Y1057 in a c-Cbl independent manner, suggesting that the interaction of c-Src with Y1057 of VEGFR-2 is established by a direct binding involving its SH2 domain and an indirect interaction involving c-Cbl. [bib_ref] VEGF receptor signalling-in control of vascular function, Olsson [/bib_ref] or stimulated with VEGF (+) for 10 minutes, lysed, immunoprecipitated with c-Src antibody and immunoblotted with an anti-VEGFR-2 antibody (C). Whole cell lysates (WCL) from the same group was immunoblotted with an anti-VEGFR-2 as a control (D). PAE cells expressing wild type chimeric VEGFR-2 (CKR) and tyrosine mutant CKRs, F799/CKR, F820/CKR, F925/CKR and carboxyl terminus deleted CKR coupled with mutation of Y1173 and Y1212 denoted as DCKR/F1173/F1212 was stimulated with CSF-1 for 10 minutes. Cells were lysed, and whole cell lysates were incubated with purified GST-SH2-Src protein. The GST-SH2-Src bound proteins were subjected to Western blot analysis using an anti-VEGFR-2 antibody (E). Whole cell lysates from the same groups was blotted with an anti-VEGFR-2 antibody (F). PAE cells expressing CKR, E1052/CKR, and E1057/CKR were either unstimulated or stimulated with CSF-1 for 10 minutes and cells were lysed and cell lysates was subjected to GST-SH2-Src pull-down assay (G). An immunoblot of whole cell lysates also were probed with an anti-VEGFR-2 antibody (F,H), an anti-phospho-Src (pY416) antibody (I) or an anti-Src antibody (J). PAE cells expressing VEGFR-2 or F1057/VEGFR-2 were prepared and subjected to pull-down analysis as panel E (K) or whole cell lysates was blotted with an anti-VEGFR-2 antibody (L). The results shown in A-L are representative of three separate experiments. doi:10.1371/journal.pone.0003848.g001
## Tyrosine 1173 of vegfr-2 is phosphorylated by src kinases
Tyrosine 1173 functions as a multitasking residue on VEGFR-2 [3-7,10] and is a bona fide mediator of angiogenic signaling of VEGFR-2 in vivo [bib_ref] Essential role of Flk-1 (VEGF receptor 2) tyrosine residue 1173 in vasculogenesis..., Sakurai [/bib_ref]. Hence we investigated a possible crosstalk between the kinase domain tyrosines (i.e., Y1052 and Y1057) and Y1173 and the potential role of Src kinases in this crosstalk. For this aim, we tested phosphorylation of Y1173 in the background of E1052/CKR and E1057/CKR expressed in PAE cells. The result showed that mutation of 1057 markedly reduced ligand-dependent phosphorylation of Y1173 as detected by an anti-phospho-Y1173 specific antibody . Mutation of Y1052 only slightly reduced phosphorylation of Y1173, suggesting that Y1057 but not Y1052, is specifically involved in modulation of phosphorylation of Y1173 . However, mutation of either Y1052 or Y1057 had no major effect on the phosphorylation Y1212 . Quantification of phosphorylation of Y1173 and Y1212 obtained from three separate independent experiments also is shown . Since Src kinase binds to Y1057 of VEGFR-2, we tested potential role of Src in the phosphorylation of Y1173. Our analysis showed that co-expression of v-Src with non-chimeric wild-type VEGFR-2 in a transient transfection of HEK-293 cells also increases phosphorylation of Y1173 , where phosphorylation of Y1052, Y1057 and Y1212 were not effected . The quantification of phosphorylation of Y1173, Y1052, Y1057 and Y1212 also is shown . Of note, coexpression of v-Src with E1057 mutant VEGFR-2 also rescued the reduced phosphorylation of Y1173 further suggesting that c-Src is involved in catalyzing phosphorylation of Y1173 of VEGFR-2 (data not shown). . c-Cbl-dependent and independent association of c-Src with VEGFR-2. Serum-starved PAE cells expressing chimeric VEGFR-2 (CKR) or co-expressing CKR with c-Cbl and CKR with Cbl-N were either unstimulated (2) or stimulated with CSF-1 (+) for 10 minutes. Cells were lysed and subjected to an in vitro pull-down assay using purified GST-SH2-Src protein (A). Quantification of blots of SH2-Src binding to VEGFR-2 of three separate experiments is shown. Each bar represents the mean 6SD of triplicate experiments. *P,0.01 (B). A parallel immunoblot of whole cell lysates (WCL) was probed with an anti-VEGFR-2 antibody (C), with an anti-phosphotyrosine antibody (D) and with an anti-c-Cbl antibody (E). PAE cells coexpressing chimeric VEGFR-2 (CKR) with a control siRNA or Cbl-siRNA were stimulated with CSF-1 for 10 minutes and cells were lysed and prepared for in vitro pull-down assay as panel A (F). A parallel immunoblot of whole cell lysates were probed with an anti-c-Cbl antibody for confirmation of c-Cbl knockdown (H) and anti-Hsp70 antibody as a control for protein loading (I). To detect a direct interaction between the SH2 domain of c-Src and Y1057 of VEGFR-2, the indicated quantities of purified recombinant GST-SH2-Src (top) and GST control (bottom row) were dot blotted as described in the Materials and Methods and detected by an anti-pY1057 VEGFR-2 antibody (J). The graph represents the average binding of SH2-Src to VEGFR-2 in the absence or presence of c-Cbl-siRNA (6SD) of three separate experiments. Quantification of blots of SH2-Src binding to VEGFR-2 of three separate experiments is shown. *P,0.01 (G). All the experiments repeated at least three times. A summary of the proposed interaction of c-Src with VEGFR-2 is shown (K). doi:10.1371/journal.pone.0003848.g002 . c-Src kinase activity is required for maximal phosphorylation of Y1173 of VEGFR-2. Serum-starved PAE cells expressing wild type CKR, E1052/CKR and E1057/CKR were either unstimulated (2) or stimulated with CSF-1 (+) for 10 minutes. Whole cell lysates (WCL) were subjected to Western blot analysis using an anti-phosphoY1173 specific-VEGFR-2 antibody (A), an anti-phosphoY1212 specific-VEGFR-2 antibody (B) or an anti-VEGFR-2 antibody (C). The graph represents the mean phosphorylation of Y1173 and Y1212 of VEGFR-2 (6SD) of three separate experiments. *P,0.01 (D). HEK-293 cells transiently expressing VEGFR-2 alone or co-expressing v-Src were stimulated with VEGF and immunoblotted with anti-pY1173 (E), anti-pY1052 (F), anti-pY1057 (G), anti-pY1212 (H), anti-c-Src (I) and anti-VEGFR-2 (J) antibodies. The blots were also quantified using Image J program (developed by NIH) and represents average tyrosine phosphorylation sites on VEGFR-2 of three separate experiments (6SD) from three separate experiments (K). SYF cells and SYF cells expressing chimeric VEGFR-2 (CKR) were either unstimulated (2) or stimulated with CSF-1 (+) for 10 minutes and cells lysates were prepared and immunoblotted with anti-pY1173 (L), anti-pY1052 (M) and anti-VEGFR-2 (N) antibodies. PAE cells expressing CKR were treated with different concentration of SU6656 for 30 minutes and stimulated with CSF-1 for 10 minutes and cell lysates were prepared as panel A and immunoblotted with anti-pY1173 (O), anti-pY1052 (P) and anti-VEGFR-2 (Q) antibodies. Phosphorylation of Y1173 and Y1052 in response to SU6656 treatment was quantified as panel D and is presented as the mean (6SD) of three separate independent experiments (R). A summary of the proposed model of phosphorylation Y1173 of VEGFR-2 by c-Src and VEGFR-2 itself is shown (S). doi:10.1371/journal.pone.0003848.g003
To further link Src kinase activity to phosphorylation of Y1173 of VEGFR-2, we tested the phosphorylation of Y1173 of VEGFR-2 in the triple knockout SYF (c-Src, c-Yes and c-Fyn) cells [bib_ref] Src family kinases are required for integrin but not PDGFR signal transduction, Klinghoffer [/bib_ref]. The data showed that stimulation of SYF cells ectopically expressing chimeric VEGFR-2 (CKR) with ligand increases phosphorylation of Y1052. In contrast, phosphorylation of VEGFR-2 at Y1173 was almost undetectable and 3M). Only a long exposure of the film detected a weak phosphorylation of Y1173 (data not shown), suggesting that in addition to Src kinases, VEGFR-2 itself also catalyzes phosphorylation of Y1173 of VEGFR-2. Since in the recent years various Src inhibitors were used for therapeutic purposes in cancer and anti-angiogenesis treatments [bib_ref] Src family kinases in tumor progression and metastasis, Summy [/bib_ref] , we tested the effect of Srcspecific inhibitor, SU6656 to inhibit phosphorylation of VEGFR-2 at Y1173. The result showed that SU6656 selectively inhibits phosphorylation of VEGFR-2 at Y1173 in a dose-dependent manner but had no effect on the phosphorylation of Y1052 , 3P). The quantification of inhibition of phosphorylation of Y1173 of VEGFR-2 by SU6656 also is shown . Taken together, the data demonstrate that Src kinases upon activation by VEGFR-2 phosphorylate Y1173 of VEGFR-2 .
## Role of c-src in vegfr-2 mediated endothelial cell tube formation and proliferation
Since activation of Src family kinases is linked to angiogenesis [bib_ref] Selective requirement for Src kinases during VEGF-induced angiogenesis and vascular permeability, Eliceiri [/bib_ref] [bib_ref] Role of Raf in vascular protection from distinct apoptotic stimuli, Alavi [/bib_ref] we initially analyzed the ability E1052/CKR and E1057/ CKR to stimulate endothelial tubulogenesis in vitro. Our analysis showed that mutation of Y1052 had no apparent effect on the ability of CKR to promote tube formation of PAE cells. However, mutation of Y1057 significantly reduced the ability of CKR to stimulate tube formation of PAE cells [fig_ref] Figure 4: Role of c-Src in VEGFR-2-dependent endothelial cell tubulogenesis and proliferation [/fig_ref]. Quantification of tube formation of PAE cells in response to CKR, E1052/CKR and E1057/CKR activation is shown [fig_ref] Figure 4: Role of c-Src in VEGFR-2-dependent endothelial cell tubulogenesis and proliferation [/fig_ref]. To test whether c-Src activation is required for tube formation of PAE cells, we used previously established PAE cells co-expressing CKR with either wild type c-Src or dominantly negative acting kinase inactive Src [bib_ref] The presence of a single tyrosine residue at the carboxyl domain of..., Meyer [/bib_ref]. The result showed that over-expression of c-Src or dominant negative c-Src in PAE cells do not alter the ability of VEGFR-2 to stimulate tube formation [fig_ref] Figure 4: Role of c-Src in VEGFR-2-dependent endothelial cell tubulogenesis and proliferation [/fig_ref] , indicating that perhaps c-Src activity is not required for VEGFR-2 to stimulate tube formation of endothelial cells. In contrast, our analysis showed that over-expression of c-Src significantly enhances the ability of VEGFR-2 to stimulate proliferation of PAE cells and over-expression of dominant negative Src (Src kinase-dead) inhibits the VEGFR-2 driven proliferation of PAE cells [fig_ref] Figure 4: Role of c-Src in VEGFR-2-dependent endothelial cell tubulogenesis and proliferation [/fig_ref]. To test the role of c-Src in endothelial cell proliferation further we knocked down c-Src expression in HUVE and HMVE cells by siRNA strategy and tested their VEGFdependent proliferation. The result showed that silencing the expression of c-Src in endothelial cells significantly inhibits the ability of VEGFR-2 to stimulate proliferation of these primary endothelial cells [fig_ref] Figure 4: Role of c-Src in VEGFR-2-dependent endothelial cell tubulogenesis and proliferation [/fig_ref]. Interestingly, HMVE cells were more sensitive to silencing of c-Src than HUVE cells since silencing of Src expression almost totally inhibited VEGF- Identification of IQGAP1 as a novel Src kinase substrate IQGAP1 is a scaffold protein and participates in signaling cascades mediated by a diverse group of cell surface receptors [bib_ref] IQGAP proteins are integral components of cytoskeletal regulation, Briggs [/bib_ref] , including VEGFR-2 [bib_ref] IQGAP1, a novel vascular endothelial growth factor receptor binding protein, is involved..., Yamaoka-Tojo [/bib_ref]. In addition, we recently have identified IQGAP1 as a binding partner of VEGFR-2 by a proteomic approach (our unpublished data), raising the likelihood for the involvement of Y1057 in the recruitment and tyrosine phosphorylation of IQGAP1 by VEGFR-2. Our initial observation showed that IQGAP1 is tyrosine phosphorylated in PAE cells by VEGFR-2 and mutation of Y1057 reduces the ability of VEGFR-2 to stimulate its tyrosine phosphorylation . IQGAP1 also is tyrosine phosphorylated by other RTKs including, ErbB1/EGFR-1 and PDGFRb ectopically expressed in PAE cells (our unpublished data), suggesting that IQGAP1 serves as a common substrate for RTKs. In addition, our data show that overexpression of c-Src in PAE cells highly enhanced tyrosine phosphorylation of IQGAP1. In a sharp contrast, over-expression of a dominant negatively acting c-Src inhibited tyrosine phosphorylation of IQGAP1 .
To further establish role of c-Src in phosphorylation of IQGAP1 we show that in SYF knockout cells where Src family kinases (Yes, Fyn and Src) are absent, IQGAP1 is not tyrosine phosphorylated in response to activation of VEGFR-2, where re-introduction of c-Src rescued phosphorylation of IQGAP1 by VEGFR-2 . Our further studies showed that in an in vitro kinase assay recombinant c-Src protein phosphorylates GST-fusion IQGAP1 (data not shown). Taken together, the data support the hypothesis that IQGAP1 directly associates with c-Src and undergoes c-Srcdependent tyrosine phosphorylation. Hence to further establish as to whether c-Src interacts with IQGAP1 we tested the ability of SH2 and SH3 domains of c-Src to associate with IQGAP1 in an in vitro pull-down assay. The result showed that SH2 domain of c-Src interacts with IQGAP1 independent of its tyrosine phosphorylation since SH2 domain of c-Src forms a complex with IQGAP1 prior to stimulation of cells with ligand . Additional analysis using PAE cells null for chimeric VEGFR-2 revealed that indeed association of SH2 domain of Src with IQGAP1 is independent of VEGFR-2 (data not shown). In contrast to the SH2 domain, the SH3 domain of c-Src failed to interact with IQGAP1 with or without stimulation of VEGFR-2 . The data suggest that SH2 domain mediates the interaction of c-Src with IQGAP1 by a novel mechanism that is independent of tyrosine phosphorylation of IQGAP1.
## Iqgap1 signaling pathway is required for vegf-induced angiogenesis
To establish the biological importance of IQGAP1 in VEGFR-2 signaling, we initially examined its role in proliferation of endothelial cells. To this end, we employed siRNA-mediated knockdown strategy and analyzed proliferation of endothelial cells in response to VEGF stimulation. IQGAP1 siRNA significantly reduced expression of IQGAP1 in HMVE cells . We further analyzed these cells for their ability to undergo proliferation in response to VEGF. Silencing the expression of IQGAP1 in primary endothelial cells, but not control siRNA, significantly inhibited VEGF-dependent cell proliferation , suggesting that VEGFR-2/c-Src orchestrated tyrosine phosphorylation of IQGAP1 serves an important role in endothelial cell proliferation. Our recent study indicates that IQGAP1 is capable of modulating activity of b-Raf [bib_ref] IQGAP1 modulates activation of B-Raf, Ren [/bib_ref] , to test whether IQGAP1 activity is also required for b-Raf phosphorylation in endothelial cells in response to VEGFR-2 activation we tested phosphorylation of b-Raf where expression of IQGAP1 is knocked down by siRNA. The data show that stimulation of endothelial cells with VEGF promotes phosphorylation of b-Raf at Ser 455 and silencing expression of IQGAP1 notably reduced its phosphorylation . Quantification of phosphorylation of Ser 455 of b-Raf based on the three independent experiments also is shown . The reduced phosphorylation of b-Raf in IQGAP1 siRNA treated cells was not due to the differential amount of protein since relatively equal amount of b-Raf protein is present in each group . The expression of IQGAP1 in cells treated with control siRNA or IQGAP1 siRNA also is shown . In short, the data indicate that VEGFR-2-dependent endothelial cell proliferation, in part, is established by Src, IQGAP1and b-Raf axis .
To further address the physiological importance of c-Src, IQGAP1and b-Raf in angiogenesis in an in vivo setting, we used chicken chorioallantoic membrane (CAM) angiogenesis assay where their expression were knocked down by siRNA strategy. The result showed that targeting IQGAP1, c-Src and b-Raf individually by siRNA suppresses the ability of VEGF to stimulate angiogenesis , where control siRNA had no negative effect on angiogenesis . Furthermore, quantification of CAM assay showed that VEGF treatment of CAM induced robust angiogenic response where IQGAP1-siRNA, c-Src-siRNA and b-Raf-siRNA each decreased VEGF-induced angiogenesis . As noted the siRNA-mediated inhibition of VEGFinduced angiogenesis was near to baseline angiogenesis underscoring the importance of this pathway for VEGF-induced angiogenesis. The ability of these specific siRNAs to knockdown expression of c-Src, IQGAP1 and b-Raf also are analyzed . The same membranes were re-probed for PLCc1 as a loading control . In sum, our data demonstrate that IQGAP1, c-Src and b-Raf pathway plays a critical role both in vitro and in vivo in VEGF-mediated angiogenesis.
# Discussion
In this report, we have identified a novel angiogenic pathway emanating from phospho-Y1057 of VEGFR-2. Phosphorylation of Y1057 plays a combinatorial role in VEGFR-2 signaling; it recruits c-Src kinase to VEGFR-2, regulates phosphorylation of multi-docking Y1173 site, mediates VEGFR-2-dependent proliferation of endothelial cells and angiogenesis through IQGAP1 and b-Raf. One of the most interesting aspects of this study is the discovery of an unexpected role of Src family kinases in the phosphorylation of Y1173 of VEGFR-2. Y1173 of VEGFR-2 is considered to be a bona fide mediator of angiogenic signaling of VEGFR-2 in vivo since its mutation abrogates its ability to stimulate angiogenesis during mouse embryonic development [bib_ref] Recruitment and activation of phospholipase Cgamma1 by vascular endothelial growth factor receptor-2..., Meyer [/bib_ref] [bib_ref] Essential role of Flk-1 (VEGF receptor 2) tyrosine residue 1173 in vasculogenesis..., Sakurai [/bib_ref] and Y1173 functions as a multi-docking residue on VEGFR-2 that recruits p85/PI3kinase, PLCc1, Shb, and Sck [bib_ref] Identification of tyrosine residues in vascular endothelial growth factor receptor-2/FLK-1 involved in..., Dayanir [/bib_ref] [bib_ref] A mechanosensory complex that mediates the endothelial cell response to fluid shear..., Tzima [/bib_ref] [bib_ref] A single autophosphorylation site on KDR/Flk-1 is essential for VEGF-A-dependent activation of..., Takahashi [/bib_ref] [bib_ref] Recruitment and activation of phospholipase Cgamma1 by vascular endothelial growth factor receptor-2..., Meyer [/bib_ref] [bib_ref] The adaptor protein shb binds to tyrosine 1175 in vascular endothelial growth..., Holmqvist [/bib_ref] [bib_ref] Sck is expressed in endothelial cells and participates in vascular endothelial growth..., Ratcliffe [/bib_ref]. Thus, phosphorylation of Y1173 by Src kinases represents one pathway by which the intensity and extent of Y1173-dependent signaling of VEGFR-2 is controlled. Conversely, inhibition of Src kinases could skew the binding of these signaling proteins to VEGFR-2 and their subsequent activation by VEGFR-2.
A recent study indicates that c-Cbl directly interacts with Y1057 of VEGFR-2 [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref]. Interestingly, not only does c-Src bind directly to Y1057, but c-Cbl provides an additional mechanism for c-Src to interact with VEGFR-2 by serving as an ''adaptor'' which in turn may modulate its activation by VEGFR-2. Activation of c-Src is linked to vascular permeability [bib_ref] Selective requirement for Src kinases during VEGF-induced angiogenesis and vascular permeability, Eliceiri [/bib_ref] , endothelial cell survival and angiogenesis [bib_ref] Selective requirement for Src kinases during VEGF-induced angiogenesis and vascular permeability, Eliceiri [/bib_ref] [bib_ref] Role of Raf in vascular protection from distinct apoptotic stimuli, Alavi [/bib_ref]. Our findings show that although c-Src activity is dispensable for VEGFR-2-dependent tube formation of endothelial cells, its activity is highly critical for VEGF-stimulated endothelial cell proliferation and angiogenesis in CAM model of angiogenesis. The mechanism by which c-Src contributes to VEGFR-2-dependent cell proliferation is linked, in part to its ability to phosphorylate Y1173, which is known to regulate VEGFR-2-mediated endothelial cell proliferation [bib_ref] A single autophosphorylation site on KDR/Flk-1 is essential for VEGF-A-dependent activation of..., Takahashi [/bib_ref] [bib_ref] Recruitment and activation of phospholipase Cgamma1 by vascular endothelial growth factor receptor-2..., Meyer [/bib_ref] [bib_ref] The adaptor protein shb binds to tyrosine 1175 in vascular endothelial growth..., Holmqvist [/bib_ref]. c-Src also associates with IQGAP1 in endothelial cells and catalyzes tyrosine phosphorylation of IQGAP1 which acts as a downstream target of VEGFR-2 and c-Src to stimulate endothelial cell proliferation. A recent work has demonstrated that c-Src regulates activation of Raf-1 (also known as c-Raf) in endothelial cells [bib_ref] Role of Raf in vascular protection from distinct apoptotic stimuli, Alavi [/bib_ref]. c-Src also associates directly with another Raf family protein, b-Raf, leading to activation of MAPK pathway in a Ras-independent manner [bib_ref] IQGAP1 modulates activation of B-Raf, Ren [/bib_ref] [bib_ref] IQGAP proteins are integral components of cytoskeletal regulation, Briggs [/bib_ref]. Our demonstration that IQGAP1 is activated by VEGFR-2 further supports earlier studies that shows VEGFR-2 activates MAPK pathway in Ras-independent manner [bib_ref] IQGAP1 binds ERK2 and modulates its activity, Roy [/bib_ref].
In conclusion, in this study we have identified and defined a molecular switch by which VEGFR-2 regulates endothelial cell proliferation and stimulates angiogenesis by identifying Y1057 of VEGFR-2 that undergoes phosphorylation and orchestrates . Role of IQGAP1 in VEGFR-2 signaling and endothelial cell Proliferation. Serum-starved PAE cells expressing CKR (chimeric VEGFR-2) and E1057/CKR were either unstimulated (2) or stimulated with CSF-1 (+) for 10 minutes. Whole cell lysates were subjected to immunoprecipitation with IQGAP1 and Western blot analysis using an anti-phosphotyrosine antibody (A) or anti-IQGAP1 antibody (B). PAE cells co-expressing either wild type c-Src or a dominant negative form of c-Src (KD-Src) were stimulated and subjected to Western blot analysis as panel A using an anti-phosphotyrosine antibody (C). The same membrane striped off the antibody and re-blotted with an anti-IQGAP1 antibody for protein levels (D). SYF cells co-expressing CKR with c-Src or expressing CKR alone were prepared as above and blotted with anti-phosphotyrosine (E), anti-IQGAP1 (F), and anti-VEGFR-2 (G). Cell lysates derived from PAE cells expressing CKR were either precipitated with GST, GST-SH2-Src, (H) or GST-SH3-Src (I) and blotted for IQGAP1 using an anti-IQGAP1 antibody. Human primary dermal vascular endothelial (HMVE) cells were stimulated with VEGF for 10 minutes, cells were lysed and immunoprecipitated with either control antibody (Ctrl.IgG) or c-Src antibody and blotted with anti-IQGAP1 (J). Whole cell lysates was blotted with anti-Src antibody (K). HMVE cells expressing either control siRNA (Ctrl.siRNA) or IQGAP1 siRNA were lysed and blotted for IQGAP1 using an anti-IQGAP1 antibody (L). HMVE cells expressing either control siRNA or IQGAP1 siRNA were subjected to proliferation assay as described in Method section. Error bars represent mean 6 SD of triplicate samples. *p,0.001 as compared to proliferation of HMVE cells in response to 100 gg VEGF-A (M). HMVE cells expressing either control siRNA or IQGAP1-siRNA were unstimulated (2) or stimulated (+) with VEGF for 10 minutes and cells were lysed and immunoprecipitated with an anti-b-Raf antibody and blotted with an anti-phospho-Ser445 antibody (N) or with anti-b-Raf antibody (O). Whole cell lysates was blotted for IQGAP1 (P). Also shown is the graph of the mean phosphorylation of b-Raf where the expression of IQGAP1 is silenced by siRNA. It represents (6SD) of three separate experiments. *P,0.01 (Q). Summary of the proposed VEGFR-2dependent signal relay in endothelial cell proliferation also is shown (R). doi:10.1371/journal.pone.0003848.g005 recruitment and activation of c-Src and modulates endothelial cell proliferation via IQGAP1-dependent signaling pathway.
# Materials and methods
## Reagents, antibodies, sirnas and vectors
Recombinant human VEGF-A and colony stimulating factor-1 (CSF-1) were purchased from R&D Systems . Mouse monoclonal anti-phosphotyrosine antibody 4G10 and anti-IQGAP1 antibody were purchased from Upstate Biotechnology (Lake Placid, NY). Rabbit anti-phospho-VEGFR-2 (pY1052) and rabbit polyclonal anti-phospho-VEGFR-2 (pY1057) were made in collaboration with Upstate Biotechnology (Lake Placid, NY). Rabbit polyclonal anti-VEGFR-2 sera were raised against either a glutathione S-transferase-VEGFR-2 kinase insert domain fusion protein (1410) or a GST-VEGFR-2 carboxyl-terminus fusion protein (1412) [bib_ref] VEGFR-1 and VEGFR-2: two non-identical twins with a unique physiognomy, Rahimi [/bib_ref] [bib_ref] The carboxyl terminus of VEGFR-2 is required for PKC-mediated down-regulation, Singh [/bib_ref]. Rabbit monoclonal anti-phospho-VEGFR-2 (pY1173) was purchased from Cell Signaling Technology (Beverly, MA). Rabbit polyclonal anti-PLCc1 was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). The oligonucleotide inserted into pSUPER.retro.puro vector used to produce c-Cbl siRNA was described elsewhere [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref] , and other oligonucleotide siRNA used in this study are described in supplementary Data S1.
## Cell culture, cell lines and retroviral infection
Porcine aortic endothelial (PAE) cells were grown in 10% FBS cells and lack endogenous expression of VEGFR-2 [bib_ref] Receptor chimeras indicate that the vascular endothelial growth factor receptor-1 (VEGFR-1) modulates..., Rahimi [/bib_ref]. Primary human dermal microvascular endothelial cells (HMVE cells) were purchased from (Cell Applications, Inc. San Diego) and human umbilical vein endothelial (HUVE) cells, bovine vascular endothelial cells (BAE) cells and human embryonic kidney cells (HEK-293) cells were purchased from ATCC and were grown in 10% FBS plus growth factor supplements and penicillin/ streptomycin. SYF cells are triple knockout (Src, Yes and Fyn) fibroblast cells and they were used as described [bib_ref] The presence of a single tyrosine residue at the carboxyl domain of..., Meyer [/bib_ref]. Creation of the VEGFR-2 chimera (CKR), in which the extracellular domain of VEGFR-2 is replaced with that of the human CSF-1R, has been previously described [bib_ref] VEGFR-1 and VEGFR-2: two non-identical twins with a unique physiognomy, Rahimi [/bib_ref]. The oligonucleotide siRNA for c-Src was 59-AAACTCCCCTTG-CTCATGTAC-39. The oligonucleotide siRNA used for IQGAP1was cloned into pSUPER.retro.puro vector as previously described [bib_ref] IQGAP1 modulates activation of B-Raf, Ren [/bib_ref]. Avian siRNAs generated for c-Src, IQGAP1 and b-Raf are the following; siRNAs of avian c-Src . IQGAP1, Src and b-Raf activity is required for angiogenesis in vivo: 9-day old chicken chorioallantoic membrane (CAM) (n = 10 CAM for each experimental group) were prepared as described in Materials and Method section and exposed to filter paper disks saturated with siRNAs of either IQGAP1, Src, b-Raf or scrambled control siRNA for 24 hours and then incubated in the presence or in absence VEGF for an additional 72 hours. Pictures were taken by Leica Camera under microscope and a representative pictures from each group are shown (A). Blood vessels were quantified by Leica imaging software system. Each bar represents the mean 6SD of 10 CAM (n = 10). *P,0.05 versus control siRNA plus VEGF (B). CAM tissue was prepared and subjected to Western blot analysis using an anti-IQGAP1 antibody (C), anti-c-Src antibody (E), and anti-b-Raf antibody (G). The same membranes were re-blotted with an anti-PLCc1 antibody as a control for protein loading (D, F, H). doi:10.1371/journal.pone.0003848.g006
were; CAUUGCCAAGGUCAGCGAU UU and AUGACGC-CAC AGCGCAGGC UU, siRNA for avian b-raf were; UUGGCUGGGACACUGACAU UU and AGGAUAGGAU-CUGGAUCAU UU; siRNA for avian IQGAP1 was; ACUCUG-CAAGCCUUACA GAUU. Creation of mutant VEGFR-2s and other related methods are described in supplementary Data S1.
In vitro angiogenesis/tubulogenesis assay Angiogenesis assay was performed essentially as described [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref] [bib_ref] ) c-Cbl is downstream of c-Src in a signalling pathway necessary for..., Tanaka [/bib_ref]. Briefly, PAE cells were suspended in DMEM containing 1% fetal bovine serum and 0.24% (w/v) carboxymethylcellulose (4000 centipoise) in non-adherent round-bottom 96-well plates. After 24 h, all cells formed one single spheroid per well. Spheroids were cultured for additional 48 h before using them in the in vitro angiogenesis assay in the manner previously described [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref] [bib_ref] ) c-Cbl is downstream of c-Src in a signalling pathway necessary for..., Tanaka [/bib_ref]. Sprouting and tubulogenesis was observed under an inverted phase-contrast microscope and pictures were taken using a Leica digitial camera system. For statistical analysis purposes four sprouting colonies were used for each experimental condition and their sprouting was quantified using image J program (NIH).
## Immunoprecipitation, immunoblotting and in vitro pulldown assay
Cells were prepared for immunoprecipitation as described [bib_ref] A critical role for the E3-ligase activity of c-Cbl in VEGFR-2-mediated PLCgamma1..., Singh [/bib_ref]. Briefly, cells were grown in 10-cm culture dishes until 80-90% confluent and after serum starvation, cells were left either resting or stimulated with appropriate ligands as indicated in figure legends. Cells were lysed and normalized whole cell lysates were subject to immunoprecipitation by incubating with appropriate antibodies. Immunocomplexes were captured by incubating with either Protein-A Sepharose or Protein-G-Agarose beads and immunoprecipitated protein were subjected to immunoblotting analysis. Additional experimental details are provided in Supplementary Data S1. In some occasions the blots were scanned and they were subsequently quantified using Image J program (NIH).
# Statistical analysis
For statistical purposes the blots of three independent experiments were scanned and blots were quantified using NIH Image J program. Comparison of the different parameters for the blots was determined by repeated measures analysis of variance (ANOVA). Significant differences were assigned using Kruskal-Wallis post hoc test. The criterion for significance for all tests was set at p , 0.05. Specific software was used to assist in the data analysis (GraphPadPrism v4.0b, GraphPad Software, San Diego, USA).
## Gst-pull down assay
In vitro GST fusion protein binding experiments were performed as described [bib_ref] Substitution of C-terminus of VEGFR-2 with VEGFR-1 promotes VEGFR-1 activation and endothelial..., Meyer [/bib_ref]. To this end, equal numbers of endothelial cells were grown to 90% confluency prior to serum starvation for overnight. Unstimulated or ligand-stimulated cells were lysed in ice-cold lysis buffer supplemented with 2 mM Na 3 VO 4 and a protease inhibitor cocktail. Equal amounts of the appropriate immobilized GST fusion proteins were incubated with normalized whole cell lysates by rocking for 3 h at 4uC. The beads were washed four times in the presence of protease inhibitors and Na 3 VO4, and proteins were eluted and analyzed in Western blot analysis using appropriate antibody.
## Dot blot assay
Purified recombinant GST, GST-SH2-c-Src and GST-SH3-c-Src were eluted from glutathione-Sepharose beads and subjected to dot blot. Further information regarding this assay is provided in Supplementary Data S1.
## Cell proliferation assay
Cell proliferation was evaluated by direct cell counting. Endothelial cells were seeded at a density of 2610 4 cells/well in 24-well plates and cultured overnight; the cells were then incubated in serum-free medium for 12 hours. Cells were stimulated with recombinant human CSF-1 or VEGF-A at different concentrations as indicated in figure legends, and after 48 hours they were washed with PBS, harvested by mild trypsinization, and counted with a hematocytometer. Experiments were performed in triplicate and 3 separate experiments were performed and where appropriate, values were presented as means of 6 SD.
## Chicken chorioallantoic membrane (cam) angiogenesis assay
Pathogen free grade fertilized chick embryos (purchased from SPAFAS, Preston, CT) were incubated for 10 days (37uC with 70% humidity). A hole was made with a drill over the air sac at the end of the egg as described [bib_ref] Selective requirement for Src kinases during VEGF-induced angiogenesis and vascular permeability, Eliceiri [/bib_ref]. A small window was cut in the egg shell over the dropped CAM, in order to place growth factor and siRNA on the CAM [bib_ref] Selective requirement for Src kinases during VEGF-induced angiogenesis and vascular permeability, Eliceiri [/bib_ref] [bib_ref] Role of Raf in vascular protection from distinct apoptotic stimuli, Alavi [/bib_ref]. Filter disks were soaked in cortisone acetate, VEGF (200 gg), siRNA and plus transfection agent and the disk filters were applied to the CAM of 9 day chick embryos and incubated at 37uC for 72 hours. 10 CAM were used per each group. Pictures were taken by Leica Camera under microscope from three different area randomly and these areas were further used to quantify blood vessels using Leica imaging software system. Statistical Analysis: Data represent mean 6SD of 10 CAM which includes three different fields from each CAM was used unless otherwise stated. Statistical significance was calculated by ANOVA soft ware program (prism v4.0b), considering p,0.05 as statistically significant. Tissue preparation for Western blotting: CAM tissue was also harvested for Western blotting. For this purpose CAM tissue was snap frozen in liquid nitrogen and samples were homogenized in a modified RIPA buffer. Five CAMs from each experimental group were homogenized in 1 ml of the RIPA buffer. Equal amounts of protein were subjected to Western blot analysis using desired antibody as described in the figure legends.
## Supporting information
Data S1
Found at: doi:10.1371/journal.pone.0003848.s001 (0.03 MB DOC) [fig_ref] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2 [/fig_ref] Effect of mutation of tyrosines 1052 and 1057 to phenylalanine: Equal number of serum-starved PAE cells expressing wild type CKR, F1052/CKR, F1057/CKR and F1052/F1057/CKR were either unstimulated (2) or stimulated with CSF-1 (+) for 10 minutes. Total cell lysates were subjected to Western blot analysis using an anti-phosphoY1052 specific-VEGFR-2 antibody (A), an anti-phosphoY1057 specific anti-VEGFR-2 antibody (B), an anti-phosphotyrosine antibody (C), and an anti-VEGFR-2 antibody (D). Equal number of serumstarved PAE cells expressing wild type CKR, F1052/CKR, F1057/CKR and F1052/F1057/CKR without stimulation were lysed and immunoprecipitated with an anti-VEGFR-2 antibody. The immunoprecipitated proteins were subjected to an in vitro kinase assay with increasing concentrations of ATP as indicated and described in Materials and Methods and immunoblotted with an anti-phosphotyrosine antibody (E). The same membranes were re-blotted with an anti-VEGFR-2 antibody for proteins levels (F). Found at: doi:10.1371/journal.pone.0003848.s002 (0.22 MB TIF)Effect of mutation of Tyrosines 1052 and 1057 to glutamic acid. Partial amino acid sequence of mouse VEGFR-2 containing Y1052 and Y1057 and the schematic presentation of VEGFR-2 are shown (A). Equal number of serum-starved PAE cells expressing wild type CKR, E1052/CKR, E1057/CKR and E1052/E1057/CKR were either unstimulated (2) or stimulated with CSF-1 (+) for 10 minutes. Total cell lysates were subjected to Western blot analysis using an anti-phosphoY1052 specific-VEGFR-2 antibody (B), an anti-phosphoY1057 specific anti-VEGFR-2 antibody (C), an anti-phosphotyrosine antibody (D), and an anti-VEGFR-2 antibody (E). Equal number of serumstarved PAE cells expressing wild type CKR, E1052/CKR, E1057/CKR and E1052/E1057/CKR without stimulation were lysed and immunoprecipitated with an anti-VEGFR-2 antibody. The immunoprecipitated proteins were subjected to an in vitro kinase assay with increasing concentrations of ATP as indicated and described in Materials and Methods and immunoblotted with an anti-phosphotyrosine antibody (F). The same membranes were re-blotted with an anti-VEGFR-2 antibody for proteins levels (G). The image analysis program was used to quantify the data. The tyrosine phosphorylation values were normalized over total protein (H). Arbitrary unit (AU). Found at: doi:10.1371/journal.pone.0003848.s003 (0.24 MB TIF) GST-Src but not GST binds to VEGFR-2: Serumstarved PAE cells expressing CKR were either unstimulated (2) or stimulated (+) with CSF-1 for 10 minutes, lysed, and whole cell lysates were incubated with purified GST alone or GST-SH2-Src protein. The GST-SH2-Src bound proteins were subjected to Western blot analysis using an anti-VEGFR-2 antibody (A). Whole cell lysates were blotted with anti-VEGFR-2 antibody as a control (B).
[fig] Figure 1: Tyrosine 1057 mediates recruitment of c-Src to and its activation by VEGFR-2. HUVEC, HMVE and BAE cells were stimulated with VEGF-A (100 gg/ml) for indicated times. Whole cell lysates (WCL) were immunoblotted with an anti-phospho-Src (pY416) antibody (A) or an anti-Src antibody (B). HUVEC, HMVE and HEK-293 cells were either unstimulated [/fig]
[fig] Figure: 2K summarizes our observation regarding interaction of c-Src with VEGFR-2 and role of c-Cbl in this process. [/fig]
|
10.1038/s41598-020-77910-5
|
CCBY
|
7718245
|
33277559
|
s2orc_pubmed_articles
|
Serotonin receptor 5-HT7 in Drosophila mushroom body neurons mediates larval appetitive olfactory learning
Serotonin (5-HT) and dopamine are critical neuromodulators known to regulate a range of behaviors in invertebrates and mammals, such as learning and memory. Effects of both serotonin and dopamine are mediated largely through their downstream G-protein coupled receptors through cAMP-PKA signaling. While the role of dopamine in olfactory learning in Drosophila is well described, the function of serotonin and its downstream receptors on Drosophila olfactory learning remain largely unexplored. In this study we show that the output of serotonergic neurons, possibly through points of synaptic contacts on the mushroom body (MB), is essential for training during olfactory associative learning in Drosophila larvae. Additionally, we demonstrate that the regulation of olfactory associative learning by serotonin is mediated by its downstream receptor (d5-HT7) in a cAMP-dependent manner. We show that d5-HT7 expression specifically in the MB, an anatomical structure essential for olfactory learning in Drosophila, is critical for olfactory associative learning. Importantly our work shows that spatiotemporal restriction of d5-HT7 expression to the MB is sufficient to rescue olfactory learning deficits in a d5-HT7 null larvae. In summary, our results establish a critical, and previously unknown, role of d5-HT7 in olfactory learning.OPEN
The biogenic amine serotonin (5-HT) plays a crucial role in learning and memory both in invertebrates and mammals [bib_ref] The molecular biology of memory storage: a dialogue between genes and synapses, Kandel [/bib_ref]. Serotonin is involved in rodent learning and memory through the modulation of synaptic plasticity in the hippocampus (e.g. LTP) 2,3 . In the marine mollusk Aplysia, noxious stimuli are known to increase serotonergic activity in the central nervous system (CNS), which enhances gill withdrawal, a simple reflex for studying short-term learning [bib_ref] Depletion of serotonin in the nervous system of Aplysia reduces the behavioral..., Glanzman [/bib_ref]. While the role of biogenic amine dopamine in regulating learning and memory has been well studied [bib_ref] Functional architecture of reward learning in mushroom body extrinsic neurons of larval..., Saumweber [/bib_ref] [bib_ref] Eight different types of dopaminergic neurons innervate the Drosophila mushroom body neuropil:..., Mao [/bib_ref] , the role of serotonin in Drosophila learning remains largely unexplored [bib_ref] Cellular and circuit mechanisms of olfactory associative learning in Drosophila, Boto [/bib_ref]. The physiological effects of biogenic amines are mediated through its G-protein coupled receptors, although the molecular identity of serotonin receptors mediating learning and memory in Drosophila remains to be elucidated.
The fruit fly Drosophila melanogaster has been a favored genetic model to study cellular and molecular mechanisms underlying learning and memory [bib_ref] Traces of Drosophila memory, Davis [/bib_ref] [bib_ref] Olfaction and olfactory learning in Drosophila: recent progress, Fiala [/bib_ref] [bib_ref] Drosophila olfactory memory: single genes to complex neural circuits, Keene [/bib_ref] , due to the relatively simple brain layout and well characterized neuronal architecture [bib_ref] The complete connectome of a learning and memory centre in an insect..., Eichler [/bib_ref] [bib_ref] Neuronal assemblies of the Drosophila mushroom body, Tanaka [/bib_ref] [bib_ref] The neuronal architecture of the mushroom body provides a logic for associative..., Aso [/bib_ref]. It has been established that Drosophila olfactory associative learning is mediated by biogenic amines including dopamine and octopamine [bib_ref] Induction of cAMP response element-binding protein-dependent medium-term memory by appetitive gustatory reinforcement..., Honjo [/bib_ref] [bib_ref] Distinctive neuronal networks and biochemical pathways for appetitive and aversive memory in..., Honjo [/bib_ref] [bib_ref] A subset of dopamine neurons signals reward for odour memory in Drosophila, Liu [/bib_ref] [bib_ref] Drosophila larvae establish appetitive olfactory memories via mushroom body neurons of embryonic..., Pauls [/bib_ref] [bib_ref] Reward signaling in a recurrent circuit of dopaminergic neurons and peptidergic Kenyon..., Lyutova [/bib_ref] [bib_ref] The role of dopamine in Drosophila larval classical olfactory conditioning, Selcho [/bib_ref]. In addition to dopamine and octopamine, some studies have also implicated another biogenic amine serotonin, in certain forms of learning and memory in Drosophila (e.g. aversive place memory) [bib_ref] Serotonin is necessary for place memory in Drosophila, Sitaraman [/bib_ref] , anesthesia-resistant memory [bib_ref] Serotonin-mushroom body circuit modulating the formation of anesthesia-resistant memory in Drosophila, Lee [/bib_ref] and long-term memory [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref]. However, the role of serotonin in Drosophila short-term olfactory learning remains controversial, with some studies indicating that serotonin is dispensable for olfactory learning [bib_ref] The serotonergic central nervous system of the Drosophila larva: anatomy and behavioral..., Huser [/bib_ref]. Most of the effect of biogenic amines on behaviors including learning and memory are mediated through their downstream receptors, which are primarily GPCRs that modulate downstream cAMP-PKA signaling, a critical pathway for learning and memory [bib_ref] Drosophila olfactory memory: single genes to complex neural circuits, Keene [/bib_ref] [bib_ref] Global and local missions of cAMP signaling in neural plasticity, learning, and..., Lee [/bib_ref]. Additionally, the cellular cAMP regulators rutabaga1 and dunce1 which are essential for short-term and long-term olfactory learning, are preferentially expressed in the Drosophila mushroom body (MB) [bib_ref] Traces of Drosophila memory, Davis [/bib_ref] [bib_ref] Localization of a short-term memory in Drosophila, Zars [/bib_ref] [bib_ref] Tissue-specific expression of a type I adenylyl cyclase rescues the rutabaga mutant..., Zars [/bib_ref] , a brain region critical for olfactory learning in Drosophila analogous to the mammalian amygdala [bib_ref] Olfactory perception: receptors, cells, and circuits, Su [/bib_ref]. Recent studies have also shown an increase in intracellular cAMP levels in the mushroom body with stimuli mimicking olfactory appetitive learning [bib_ref] Cyclic AMP-dependent plasticity underlies rapid changes in odor coding associated with reward..., Louis [/bib_ref] [bib_ref] D1 dopamine receptor dDA1 is required in the mushroom body neurons for..., Kim [/bib_ref] [bib_ref] Appetitive learning requires the alpha1-like octopamine receptor OAMB in the Drosophila mushroom..., Kim [/bib_ref] [bib_ref] Concerted actions of octopamine and dopamine receptors drive olfactory learning, Sabandal [/bib_ref] [bib_ref] An octopamine-mushroom body circuit modulates the formation of anesthesiaresistant memory in Drosophila, Wu [/bib_ref] [bib_ref] A GABAergic feedback shapes dopaminergic input on the Drosophila mushroom body to..., Pavlowsky [/bib_ref] [bib_ref] Pre-and postsynaptic role of dopamine D2 receptor DD2R in Drosophila olfactory associative..., Qi [/bib_ref] [bib_ref] Dopamine receptor Dop1R2 stabilizes appetitive olfactory memory through the Raf/MAPK Pathway in..., Sun [/bib_ref] , the identity of serotonin receptors essential for Drosophila olfactory learning remain poorly understood.
In Drosophila, there are four serotonin receptors (d5-HT1A & B, d5-HT2, d5-HT7) [bib_ref] Distribution of serotonin (5-HT) and its receptors in the insect brain with..., Blenau [/bib_ref]. Of these, only the d5-HT7 receptor is known to increase intracellular cAMP [bib_ref] Cloning and characterization of a Drosophila serotonin receptor that activates adenylate cyclase, Witz [/bib_ref]. Another serotonin receptor 5-HT (apAC1) was identified and characterized in Aplysia which increases intracellular cAMP levels and is involved in learningrelated heterosynaptic facilitation [bib_ref] Identification of a serotonin receptor coupled to adenylyl cyclase involved in learning-related..., Lee [/bib_ref]. Several pharmacological studies have indicated that 5-HT7 receptor mediates various forms of learning and memory in mammals including hippocampus-associated spatial memory [bib_ref] 5-HT7 receptors as modulators of neuronal excitability, synaptic transmission and plasticity: physiological..., Ciranna [/bib_ref] [bib_ref] The 5-HT(7) receptor in learning and memory, Roberts [/bib_ref]. We thus hypothesize that the expression of 5-HT receptors, particularly those positively coupled to cAMP in the mushroom body, may be essential for learning and memory in Drosophila.
In this study we show that the output of serotonergic neurons through points of synaptic contacts on the mushroom body (MB) is required for olfactory associative learning in Drosophila larvae. Additionally, we demonstrate that the regulation of olfactory associative learning by serotonin is mediated by its downstream receptor (d5-HT7), in a cAMP-dependent manner. We show that d5-HT7 expression specifically in the mushroom body (MB) is critical for olfactory associative learning. Additionally, we also show that spatio-temporal restriction of d5-HT7 expression to the MB is sufficient to rescue olfactory learning deficits in a d5-HT7 null larvae, thus demonstrating a clear requirement for this receptor in olfactory learning.
# Results
Serotonin is required for olfactory appetitive learning in Drosophila larvae. Serotonergic neurons project on to various regions of the adult Drosophila brain, including the mushroom body (MB) [bib_ref] Serotonin-mushroom body circuit modulating the formation of anesthesia-resistant memory in Drosophila, Lee [/bib_ref] [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref] [bib_ref] Serotonin is critical for rewarded olfactory short-term memory in Drosophila, Sitaraman [/bib_ref] [bib_ref] Localization of the contacts between Kenyon cells and aminergic neurons in the..., Pech [/bib_ref] , a structure involved in olfactory learning [bib_ref] Traces of Drosophila memory, Davis [/bib_ref].
While serotonergic neurons project to the mushroom body in the adult Drosophila brain, recent studies indicate that serotonergic neurons do not project to the larval mushroom body [bib_ref] The serotonergic central nervous system of the Drosophila larva: anatomy and behavioral..., Huser [/bib_ref]. Most of these neuron tracing studies are based on GFP expression driven by Gal4 drivers which are not ideally suited for tracing of synaptic contacts. Additionally, the role of serotonin in Drosophila olfactory learning has remained controversial [bib_ref] Serotonin is necessary for place memory in Drosophila, Sitaraman [/bib_ref] [bib_ref] Serotonin-mushroom body circuit modulating the formation of anesthesia-resistant memory in Drosophila, Lee [/bib_ref] [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref] [bib_ref] The serotonergic central nervous system of the Drosophila larva: anatomy and behavioral..., Huser [/bib_ref] [bib_ref] Serotonin is critical for rewarded olfactory short-term memory in Drosophila, Sitaraman [/bib_ref]. To conclusively define the role of serotonin in Drosophila larval olfactory learning, we first asked if serotonergic neurons make synaptic contacts on to the mushroom body in Drosophila third instar larvae. To address this, we utilized the GFP Reconstitution Across Synaptic Partners (GRASP) [bib_ref] GFP Reconstitution Across Synaptic Partners (GRASP) defines cell contacts and synapses in..., Feinberg [/bib_ref] [bib_ref] Motor control in a Drosophila taste circuit, Gordon [/bib_ref] method to examine if serotonergic neurons project on to the mushroom body. We selectively labelled points of reconstituted GFP between serotonergic neurons and the larval mushroom body lobes (Kenyon cell axons) by crossing the MB247-LexA; Trh-Gal4 strain with the LexAop-rCD2.RFP; UAS-CD4-spGFP1-10, LexAop-CD4-spGFP11 strain. We detected distinct reconstituted GFP signal on both the vertical and medial lobe of the mushroom body by using a monoclonal anti-GFP antibody [fig_ref] Figure 1: Output of serotonergic neurons is essential for olfactory appetitive learning [/fig_ref] while the mushroom body was labelled by RFP in a LexAop dependent manner. The reconstituted splitGFP signals are mainly observed in the upper vertical lobe (UVL) compartment and the shaft (SHA) compartment of the medial lobe [fig_ref] Figure 1: Output of serotonergic neurons is essential for olfactory appetitive learning [/fig_ref]. Both these regions of the mushroom body have been implicated in Drosophila larval learning [bib_ref] Functional architecture of reward learning in mushroom body extrinsic neurons of larval..., Saumweber [/bib_ref] [bib_ref] Distinctive neuronal networks and biochemical pathways for appetitive and aversive memory in..., Honjo [/bib_ref]. Interestingly, this innervation pattern is similar to that seen for the MBIN-e2 neuron reported by Eichler et al. [bib_ref] The complete connectome of a learning and memory centre in an insect..., Eichler [/bib_ref] , which is negative for neurotransmitters dopamine or octopamine. Thus, our results clearly demonstrate that points of synaptic contacts exist between serotonergic neurons and the mushroom body of Drosophila larvae, with innervation patterns unlike any dopaminergic or octopaminergic neuron.
Next, we wanted to determine if the output of serotonergic neurons is involved in olfactory learning in the Drosophila larvae. To address this we used the simple olfactory learning assay developed by Furukubo-Tokunaga et al. [bib_ref] Induction of cAMP response element-binding protein-dependent medium-term memory by appetitive gustatory reinforcement..., Honjo [/bib_ref] [bib_ref] Distinctive neuronal networks and biochemical pathways for appetitive and aversive memory in..., Honjo [/bib_ref] in combination with the Gal4-UAS binary expression system [bib_ref] Targeted gene expression as a means of altering cell fates and generating..., Brand [/bib_ref]. As a validation of the larval learning assay, we show that the two wild type strains (Canton-S and w 1118 ) could successfully associate pentyl-acetate (odor) with sucrose when compared to odor and water. However, two well characterized olfactory learning mutants, dunce1 and rutabaga1 were unable to successfully associate odor with sucrose resulting in impaired olfactory learning [fig_ref] Figure 1: Output of serotonergic neurons is essential for olfactory appetitive learning [/fig_ref]. Additionally, both naïve olfactory response and gustatory response to 1 M SUC were similar to wild-type controls . To determine the contribution of serotonin on olfactory learning, we where serotonergic neurons make synaptic contacts on to the vertical lobe (i) and medial lobe (ii). Scale bar = 10 μm. (B) (Top) A schematic showing the typical larval learning regime. (Bottom) Wild-type strains w 1118 (n = 7) and Canton-S (n = 6) show normal olfactory associative learning when the odor pentyl-acetate was paired with a sugar reward. The learning mutants dunce1 (n = 6) and rutabaga1 (n = 6) fail to show normal olfactory learning. (C) (Top) A schematic of the strategy used to specifically study the role of 5-HT neurons in olfactory learning. (Bottom) The output of serotonergic neurons was blocked in a temperature sensitive manner by expressing the shibire protein (shi ts1 ) under the TpH-Gal4 driver. TpH-Gal4; UAS-Shi ts1 larvae trained at 32 °C and tested at 25 °C show impaired olfactory learning (n = 6) while larvae trained and tested at 25 °C/25 °C (n = 6) or at 25 °C/32 °C (n = 5) show normal olfactory learning. The TpH-Gal4 (n = 7) or the UAS-Shi ts1 (n = 7) transgenes alone displayed normal olfactory learning when trained at 32 °C and tested 25 °C respectively. (D) The output of the serotonergic neurons was selectively activated during the training phase by expressing channelrhodopsin 2 (ChR2) in these neurons. Blue light was used to activate the ChR2. TpH-Gal4; UAS-ChR2 larvae were exposed to odor (pentyl acetate) in the presence of blue light (BL) but no sucrose. Larvae trained in the presence of BL for 10 min (n = 6) show significantly higher RI values compared to larvae trained with odor and distilled water alone (n = 6). Student t-test: ** p < 0.01; ***p < 0.001. www.nature.com/scientificreports/ specifically expressed the temperature sensitive mutant of dynamin (Shibire), in the serotonergic (5-HT) neurons of the Drosophila larvae using a serotonin specific Gal4 driver (TpH-Gal4) with expression pattern similar to Trh-Gal4 in Drosophila larvae [bib_ref] The serotonergic central nervous system of the Drosophila larva: anatomy and behavioral..., Huser [/bib_ref]. The shibire mutants show a temperature dependent blockage of pre-synaptic neurotransmitter release at 32 °C but exhibit normal physiological function at 25 °C [bib_ref] Conditional modification of behavior in Drosophila by targeted expression of a temperature-sensitive..., Kitamoto [/bib_ref]. By training and testing the TpH-Gal4; UAS-Shi ts1 larvae at 25 °C/32 °C and at 32 °C/25 °C respectively, we specifically blocked neurotransmitter release during memory recall and the memory acquisition phases. Inhibiting the output of 5-HT neurons, during the training phase led to impaired olfactory learning [fig_ref] Figure 1: Output of serotonergic neurons is essential for olfactory appetitive learning [/fig_ref]. However, inhibiting the output of 5-HT neurons exclusively during the testing phase did not impair learning. In contrast, larvae with either the TpH-Gal4 or UAS-Shi ts1 transgenes alone displayed normal learning when trained at 32 °C and tested at 25 °C.
To further define the role of 5-HT neurons in olfactory learning, we used optogenetics [bib_ref] Identifying behavioral circuits in Drosophila melanogaster: moving targets in a flying insect, Griffith [/bib_ref] [bib_ref] Light-induced activation of distinct modulatory neurons triggers appetitive or aversive learning in..., Schroll [/bib_ref] to selectively activate the serotonergic neurons during learning. To this end, light-sensitive channelrhodopsin 2 (ChR2) was selectively expressed in 5-HT neurons (Tph-Gal4 driver) of Drosophila larvae. These larvae were then exposed to blue light (BL, 470 nm) in the presence of an odor (pentyl acetate) during the training phase. No sugar reward was included during training. Larvae exposed to BL for 10 min in the presence of odor and water, but no sucrose show a significant increase in learning [fig_ref] Figure 1: Output of serotonergic neurons is essential for olfactory appetitive learning [/fig_ref]. However, larvae trained with odor and water in absence of blue light showed no associative learning. These results demonstrate that serotonin conveys the unconditioned stimulus (US) and its release during odor association is sufficient for appetitive learning in Drosophila.
## Drosophila 5-ht7 receptor (d5-ht7) is required for olfactory appetitive learning. our results
suggest that the output of the serotonergic neurons is essential for olfactory appetitive learning, possibly through synaptic contacts between the serotonergic neurons and the mushroom body. However, the identity of the downstream receptor which mediates the effect of serotonin on larval learning remains unknown. In vivo imaging of the adult Drosophila during olfactory appetitive learning has shown elevation of cAMP levels in the mushroom body [bib_ref] Cyclic AMP-dependent plasticity underlies rapid changes in odor coding associated with reward..., Louis [/bib_ref]. Further, classical genetics studies have suggested a role for cAMP in mushroom body during olfactory learning [bib_ref] Localization of a short-term memory in Drosophila, Zars [/bib_ref] [bib_ref] Tissue-specific expression of a type I adenylyl cyclase rescues the rutabaga mutant..., Zars [/bib_ref]. There are three subtypes of serotonin receptors in Drosophila which are GPCRs and they either increase or decrease cAMP levels through Gαs or Gαi/o coupled activation or inhibition of the enzyme adenylate cyclase. In addition to behavioral studies, alteration in cAMP-PKA signaling can regulate synaptic strength in neurons 1 , thereby acting as molecular mechanism essential for learning.
We thus asked if application of serotonin alters synaptic transmission in Drosophila using primary neuronal culture-based assays [bib_ref] Fast excitatory synaptic transmission mediated by nicotinic acetylcholine receptors in Drosophila neurons, Lee [/bib_ref]. Our experiments showed that focal application of 20 µM 5-HT for 30 s on wild-type (w 1118 ) neurons resulted in a significant increase in the frequency of excitatory cholinergic postsynaptic currents (EPSCs, [fig_ref] Figure 2: Expression of d5-HT7 receptor is essential for olfactory learning and excitatory synaptic... [/fig_ref]. Inhibition of PKA (a downstream target of cAMP), by the inclusion of H-89 (a chemical inhibitor of PKA) in the bath solution blocked the effect of 5-HT on the frequency of cholinergic EPSCs [fig_ref] Figure 2: Expression of d5-HT7 receptor is essential for olfactory learning and excitatory synaptic... [/fig_ref].
Amongst the various serotonin receptors, only the d5-HT7 receptor is positively coupled to cAMP signaling [bib_ref] Cloning and characterization of a Drosophila serotonin receptor that activates adenylate cyclase, Witz [/bib_ref] [bib_ref] Cell-specific manipulation of second messengers; a toolbox for integrative physiology in Drosophila, Kerr [/bib_ref] , indicating that this receptor may be mediating the observed increase in cAMP levels upon serotonin application. To determine the role of d5-HT7 on synaptic transmission, we identified and characterized a mutant strain with a transposable element insertion in the intron for the d5-HT7 receptor (Bloomington Stock #24705). We confirmed that this mutant lacks a functional copy of the d5-HT7 transcript by PCR analysis [fig_ref] Figure 2: Expression of d5-HT7 receptor is essential for olfactory learning and excitatory synaptic... [/fig_ref] and henceforth refer to this strain as a d5-HT7 null. The effect of serotonin on synaptic transmission was absent in neurons from the d5-HT7 null mutant [fig_ref] Figure 2: Expression of d5-HT7 receptor is essential for olfactory learning and excitatory synaptic... [/fig_ref] with no significant difference in the baseline EPSC frequency (before serotonin application) between wild-type and d5-HT7 null mutant larvae. These results demonstrate that serotonin-induced increase in cAMP levels in neurons is primarily mediated through the d5-HT7 receptor.
To determine the impact of d5-HT7 receptor loss on olfactory learning, we carried out olfactory learning assays with the d5-HT7R null mutant and another independent d5-HT7 deficiency strain. Our experiments showed that both these strains are deficient in olfactory associative learning [fig_ref] Figure 2: Expression of d5-HT7 receptor is essential for olfactory learning and excitatory synaptic... [/fig_ref] compared to the wild-type strain (w 1118 ). We did not observe any significant defects in response to naïve odor or the gustatory response to 1 M SUC in the d5-HT7 null mutant . To further confirm that this deficit in learning is due to the loss of receptor expression in the larval brain, we used a pan neuronal Gal4 strain (1407-Gal4) to express d5-HT7-RNAi in the larval brain. Our results show that the down regulation of d5-HT7 expression results in impaired olfactory learning [fig_ref] Figure 2: Expression of d5-HT7 receptor is essential for olfactory learning and excitatory synaptic... [/fig_ref]. These experiments using the receptor null, deficiency and the knockdown RNAi strains collectively indicate that the d5-HT7 receptor expression in the larval brain is essential for olfactory learning. . Sensory response index. Naïve olfactory and gustatory response was carried out as described in methods. Sensory responses reported as Mean±SEM. No significant difference (p < 0.5) was observed between genotypes when compared using the Student's T-test or the Mann-Whitney U test. N = 5 for all genotypes. . Student t-test: ***p < 0.001. (F) The d5-HT7 receptor null mutant (n = 6) and the d5-HT7 deficiency strain (n = 5) fail to show any learning after training in contrast to wild-type (w 1118 ) larvae which show a significant increase in response index after being trained with sucrose and odor compared to sucrose and distilled water (n = 7). The pan-neuronal expression of UAS-d5-HT7-RNAi by the 1407-Gal4 strain leads to impaired appetitive olfactory learning (n = 5). receptor expression in the brain is required for Drosophila larval olfactory learning. However, it is not clear if the d5-HT7 receptor is indeed expressed in the larval mushroom body. In this context, it has been shown that the d5-HT7 receptor is primarily expressed in regions outside the mushroom body [bib_ref] The serotonin 5-HT7Dro receptor is expressed in the brain of Drosophila, and..., Becnel [/bib_ref] , and that this receptor is not required for Drosophila olfactory learning [bib_ref] Anatomy and behavioral function of serotonin receptors in Drosophila melanogaster larvae, Huser [/bib_ref]. To conclusively determine if the d5-HT7 receptor is expressed in the MB neurons, we isolated MB neurons and probed for the presence of d5-HT7 transcripts. We used fluorescent-activated cell sorting (FACS) to isolate GFP positive MB neurons 53 from the 201Y-Gal4; UAS-mCD8::GFP strain [fig_ref] Figure 3: Genetic manipulation of d5-HT7 receptor expression in the larval MBs alters olfactory... [/fig_ref]. This Gal4 strain expresses GFP in the MB with very little expression in the larval brain outside of the MB [bib_ref] Drosophila larvae establish appetitive olfactory memories via mushroom body neurons of embryonic..., Pauls [/bib_ref]. GFP + and GFPneurons obtained from three independent sorts were pooled; the mRNA was isolated and converted to cDNA. An RT-PCR analysis for d5-HT7 receptor expression in both GFP + and GFP − cells show the presence of receptor transcripts in both the populations [fig_ref] Figure 3: Genetic manipulation of d5-HT7 receptor expression in the larval MBs alters olfactory... [/fig_ref]. The housekeeping gene RP49 also showed expression in both sorted cell populations. These data thus convincingly demonstrate that the d5-HT7 receptor transcript is expressed in the MB neurons of Drosophila larvae. receptor expression in the MB neurons is essential for olfactory associative learning, we identified the two 201Y-Gal4 and 30Y-Gal4 strains showing Gal4 expression specifically in Drosophila larvae MB neurons [bib_ref] Induction of cAMP response element-binding protein-dependent medium-term memory by appetitive gustatory reinforcement..., Honjo [/bib_ref] [bib_ref] Distinctive neuronal networks and biochemical pathways for appetitive and aversive memory in..., Honjo [/bib_ref] [bib_ref] Drosophila larvae establish appetitive olfactory memories via mushroom body neurons of embryonic..., Pauls [/bib_ref] [bib_ref] Unc-51/ATG1 controls axonal and dendritic development via kinesin-mediated vesicle transport in the..., Mochizuki [/bib_ref]. Moreover, these Gal4 strains also show GFP expression in distinct subsets of larval MB neurons [bib_ref] Drosophila larvae establish appetitive olfactory memories via mushroom body neurons of embryonic..., Pauls [/bib_ref] , making them ideal to explore the functional consequence of downregulating d5-HT7 receptor expression in the entire gamut of MB neurons.
## Scientific reports
To specifically examine the contribution of d5-HT7 receptor expression in the MB neurons on olfactory learning, we used the Gal4-UAS system 45 to down-regulate d5-HT7 receptor (UAS-d5-HT7-RNAi) expression in the MB neurons. Both the 201Y-Gal4 and the 30Y-Gal4 strains when crossed with UAS-d5-HT7-RNAi resulted in a significant decrease in olfactory learning [fig_ref] Figure 3: Genetic manipulation of d5-HT7 receptor expression in the larval MBs alters olfactory... [/fig_ref]. However, no such defect in olfactory learning was seen in the Gal4 only, UAS-5HT7-RNAi only and in a strain in which RNAi was driven by 5-HT7 promotor (Bloomington strain# 46627). This demonstrates that the downregulation of the d5-HT7 receptor specifically in the MB neurons impairs olfactory appetitive learning.
Further, we generated a 201Y-Gal4; UAS-d5-HT7 homozygous strain, to over-express this receptor in the MB neurons. This strain shows a significantly higher response index (RI) when trained with odor and sucrose compared to similarly trained 201Y-Gal4/ + larvae [fig_ref] Figure 3: Genetic manipulation of d5-HT7 receptor expression in the larval MBs alters olfactory... [/fig_ref]. Thus, our results show that both the downregulation and over-expression of the d5-HT7 receptor in the MB neurons alters olfactory learning in Drosophila.
## Spatio-temporal restriction of expression in the mb neurons alone is sufficient for learning.
Our results suggest a possible role of the d5-HT7 receptor expression in the MB neurons for olfactory associative learning which is contrary to studies showing that ablation of d5-HT7 receptor expression does not alter olfactory learning [bib_ref] Anatomy and behavioral function of serotonin receptors in Drosophila melanogaster larvae, Huser [/bib_ref] and that receptor expression outside the MB neurons is required for olfactory learning [bib_ref] The serotonin 5-HT7Dro receptor is expressed in the brain of Drosophila, and..., Becnel [/bib_ref]. Thus, our results may either be due to developmental effects of receptor down-regulation or due to the non-specific effects of Gal4 expression outside of the MB neurons. To rule out both these possibilities, we carried out a series of experiments to spatio-temporally restrict receptor expression in the MB neurons.
The Gal80 ts1 /Gal4 expression system allows for spatio-temporal regulation of downstream gene expression [bib_ref] Spatiotemporal gene expression targeting with the TARGET and gene-switch systems in Drosophila, Mcguire [/bib_ref]. We thus used this system to restrict d5-HT7-RNAi expression in the MB in a temperature sensitive manner. At 19 °C, Gal80 ts1 inhibits Gal4 however upon transfer to 30 °C, it denatures Gal80 ts1 resulting in Gal4 expression and UAS mediated downregulation of the d5-HT7 receptor [fig_ref] Figure 4: Learning defects caused by the MB specific down-regulation of d5-HT7 can be... [/fig_ref]. We crossed the 201Y-Gal4;UAS-d5-HT7-RNAi strain to the tub-P-Gal80 ts1 ;TM6 strain, which expresses Gal80 ts1 under the tubulin promoter. The corresponding F1 genotype larvae (201Y-Gal4/tub-P-Gal80 ts1 ; UAS-d5-HT7-RNAi/TM6) were able to associate odor with sucrose normally when reared at 19 °C throughout. However, when the same larvae were reared at 19 °C and then moved to 30 °C for 16 h before training, they failed to associate odor with the sucrose reward [fig_ref] Figure 4: Learning defects caused by the MB specific down-regulation of d5-HT7 can be... [/fig_ref]. Thus, 16 h of UAS-d5-HT7-RNAi expression in the MB neurons was sufficient to impair olfactory learning and thus rules out a major developmental defect due to impaired receptor expression in the MB neurons.
The 201Y-Gal4 strain used to down-regulate the receptor expression in the MB neurons is also known to be expressed in a small subset of neurons outside the MB [bib_ref] Drosophila larvae establish appetitive olfactory memories via mushroom body neurons of embryonic..., Pauls [/bib_ref]. This may lead to receptor expression outside the MB which may contribute to the observed defects in learning. To rule out this possibility we expressed a MB specific Gal80 (MB-Gal80) [bib_ref] Mosaic analysis with a repressible cell marker for studies of gene function..., Lee [/bib_ref] in the 201Y-Gal4; UAS-d5-HT7-RNAi strain. By blocking d5-HT7R-RNAi expression specifically in the MB neurons of the learning deficient 201Y-Gal4; UAS-d5-HT7-RNAi larvae, we show a complete rescue of the learning deficit in these larvae [fig_ref] Figure 4: Learning defects caused by the MB specific down-regulation of d5-HT7 can be... [/fig_ref]. The response index of these larvae was comparable to wild-type larvae. This shows that d5-HT7 receptor expression in the MB neurons alone is essential for olfactory learning.
We next sought to determine if the over-expression of the d5-HT7 receptor in a d5-HT7 mutant background could rescue the defects in olfactory learning. Expression of a wild-type copy of the d5-HT7 gene, in the MB neurons of the d5-HT7 null mutant (201Y-Gal4/UAS-d5-HT7; d5-HT7 null/d5-HT7 null), results in a complete rescue of learning and shows response index similar to wild-type larvae [fig_ref] Figure 4: Learning defects caused by the MB specific down-regulation of d5-HT7 can be... [/fig_ref]. In contrast, neither 201Y-Gal4; d5-HT7 null or UAS-d5-HT7; d5-HT7 null shows impaired learning like the receptor null mutants [fig_ref] Figure 4: Learning defects caused by the MB specific down-regulation of d5-HT7 can be... [/fig_ref].
In conclusion, we used three independent genetic approaches to show that the expression of the d5-HT7 receptor specifically in the MB neurons is sufficient for olfactory associative learning in Drosophila larvae.
# Discussion
Serotonergic neurons project to various anatomical structures in the adult Drosophila brain including the mushroom body (MB), a structure critical for olfactory learning in the flies [bib_ref] Serotonin-mushroom body circuit modulating the formation of anesthesia-resistant memory in Drosophila, Lee [/bib_ref] [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref] [bib_ref] Localization of the contacts between Kenyon cells and aminergic neurons in the..., Pech [/bib_ref]. Although recent studies suggest that the output of serotonergic neurons is essential for certain forms of olfactory learning [bib_ref] Serotonin is necessary for place memory in Drosophila, Sitaraman [/bib_ref] [bib_ref] Serotonin-mushroom body circuit modulating the formation of anesthesia-resistant memory in Drosophila, Lee [/bib_ref] [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref] [bib_ref] Serotonin is critical for rewarded olfactory short-term memory in Drosophila, Sitaraman [/bib_ref] , the evidence for serotonin's role in olfactory appetitive learning in Drosophila remains rather sparse when compared to the extensive evidence for dopamine's role in olfactory learning (reviewed in [bib_ref] Cellular and circuit mechanisms of olfactory associative learning in Drosophila, Boto [/bib_ref].
In this study we demonstrate that the output of serotonergic neurons is essential for olfactory appetitive learning and demonstrate that its downstream receptor, the d5-HT7 (Drosophila 5-HT7 homolog) is critical for learning in Drosophila larvae. We conclusively show that the expression of this receptor in the MB neurons alone is essential for olfactory appetitive learning. Additionally, we show that d5-HT7 receptor increases intracellular cAMP levels in neurons upon the application of serotonin. In context of recent evidence suggesting that elevation of cAMP in MB neurons is associated with olfactory appetitive learning 28 , our findings strongly suggest that changes in cAMP signaling mediated by the d5-HT7 receptors forms the underlying cellular basis for olfactory appetitive learning.
A detailed anatomical characterization of MB extrinsic serotonergic neurons by GRASP reveals that serotonergic neurons form close synaptic contacts on to the lobes of the MB neurons in adult Drosophila [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref] [bib_ref] Localization of the contacts between Kenyon cells and aminergic neurons in the..., Pech [/bib_ref] www.nature.com/scientificreports/ neurons 23 . By using the GRASP technique, we convincingly demonstrate that serotonergic neurons form synaptic contacts on both the vertical and medial lobes of the mushroom body [fig_ref] Figure 1: Output of serotonergic neurons is essential for olfactory appetitive learning [/fig_ref]. The innervation pattern of serotonergic is morphologically distinct from, and rather sparse, when compared to the innervation patterns of dopaminergic neurons in Drosophila larvae [bib_ref] Functional architecture of reward learning in mushroom body extrinsic neurons of larval..., Saumweber [/bib_ref] [bib_ref] The complete connectome of a learning and memory centre in an insect..., Eichler [/bib_ref] suggesting the identification of a novel circuit component in Drosophila larvae. Additionally, we show that the output of the serotonergic neurons is essential during the acquisition phase of olfactory appetitive learning [fig_ref] Figure 1: Output of serotonergic neurons is essential for olfactory appetitive learning [/fig_ref]. Further, optogenetic activation of serotonergic neurons, we show that depolarizing these neurons during training induces olfactory appetitive learning even in the absence of sucrose [fig_ref] Figure 1: Output of serotonergic neurons is essential for olfactory appetitive learning [/fig_ref]. Our results show that the synaptic output of the serotonergic neurons is essential for conveying the unconditioned stimulus (sucrose) during the memory acquisition phase. Our data is in contrast to the widely held view that the output of the dopaminergic neurons are exclusively responsible for conveying the unconditioned stimulus during appetitive olfactory learning [bib_ref] Reward signaling in a recurrent circuit of dopaminergic neurons and peptidergic Kenyon..., Lyutova [/bib_ref] [bib_ref] The role of dopamine in Drosophila larval classical olfactory conditioning, Selcho [/bib_ref] [bib_ref] Light-induced activation of distinct modulatory neurons triggers appetitive or aversive learning in..., Schroll [/bib_ref] [bib_ref] Four individually identified paired dopamine neurons signal reward in larval Drosophila, Rohwedder [/bib_ref]. However, our data is consistent with reports which suggest that other biogenic amines besides dopamine like octopamine, is also capable of conveying the unconditioned stimulus in larval olfactory learning [bib_ref] Distinctive neuronal networks and biochemical pathways for appetitive and aversive memory in..., Honjo [/bib_ref] www.nature.com/scientificreports/ While the identity and neurotransmitter expression profile of the neuron projecting on to the MB neurons in our study remains unknown, our result suggest that the biogenic amines other than dopamine, like octopamine and serotonin, may provide additional layers of regulation for olfactory appetitive learning in Drosophila larvae. This may occur either through the same neuron co-expressing both dopamine and serotonin as was observed in the adult Drosophila brain 60 or through serotonergic neurons communicating at the synaptic level with downstream dopaminergic neurons projecting on to the Drosophila MB lobes [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref]. Whether either one of these scenarios is essential for olfactory appetitive learning in Drosophila larvae remains a subject for future investigation. Elevation of cAMP levels in the MB neurons have recently been shown to be essential for both short-term 28 and long-term olfactory learning in Drosophila 22 . Additionally, several recent studies show an in vivo elevation of cAMP-PKA levels in the Drosophila mushroom body under conditions mimicking a classic olfactory learning paradigm [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref] [bib_ref] Cyclic AMP-dependent plasticity underlies rapid changes in odor coding associated with reward..., Louis [/bib_ref] [bib_ref] Dopaminergic modulation of cAMP drives nonlinear plasticity across the Drosophila mushroom body..., Boto [/bib_ref] [bib_ref] PKA dynamics in a Drosophila learning center: coincidence detection by rutabaga adenylyl..., Gervasi [/bib_ref] [bib_ref] Dynamics of learning-related cAMP signaling and stimulus integration in the Drosophila olfactory..., Tomchik [/bib_ref]. While Drosophila express five mammalian homologs of serotonin receptors, only one amongst them, the d5-HT7 receptor, is known to positively activate cAMP signaling [bib_ref] Distribution of serotonin (5-HT) and its receptors in the insect brain with..., Blenau [/bib_ref]. This led us to examine if the activation of the d5-HT7 receptor downstream of serotonergic neurons, particularly in the MB neurons contributes to olfactory appetitive learning. By pan-neuronal suppression of this receptor in genetic mutants we demonstrate that d5-HT7 receptor expression in the larval brain is essential for olfactory appetitive learning in Drosophila larvae [fig_ref] Figure 2: Expression of d5-HT7 receptor is essential for olfactory learning and excitatory synaptic... [/fig_ref]. Although the expression of the d5-HT7 receptor has been cataloged in the MB neurons of several invertebrates 36 , current evidence suggests that this receptor is not expressed in the Drosophila MB [bib_ref] The serotonergic central nervous system of the Drosophila larva: anatomy and behavioral..., Huser [/bib_ref] [bib_ref] The serotonin 5-HT7Dro receptor is expressed in the brain of Drosophila, and..., Becnel [/bib_ref] [bib_ref] A single pair of neurons links sleep to memory consolidation in Drosophila..., Haynes [/bib_ref]. In fact, some studies suggest that the d5-HT7 receptor is not essential for olfactory learning in Drosophila 52 . We probed for receptor expression in larval MB neurons by FACS sorting to probe for receptor expression. Our results convincingly demonstrate that d5-HT7 receptor cDNA is expressed in GFP positive MB neurons [fig_ref] Figure 3: Genetic manipulation of d5-HT7 receptor expression in the larval MBs alters olfactory... [/fig_ref]. Additionally, we show that down-regulating d5-HT7 receptor expression specifically in the MB neurons results in impaired olfactory appetitive learning [fig_ref] Figure 3: Genetic manipulation of d5-HT7 receptor expression in the larval MBs alters olfactory... [/fig_ref]. While MB restricted expression of a serotonin receptor (d5-HT1A) and the d5-HT2A receptor have been implicated in anesthesia-resistant aversive memory 21 and long-term memory [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref] , neither of these receptors are positively coupled to cAMP signaling. Additionally, while MB limited expression of d5-HT1A receptor, which is negatively coupled cAMP signaling, is essential for sleep consolidation [bib_ref] A sleep-promoting role for the Drosophila serotonin receptor 1A, Yuan [/bib_ref] , there is no evidence to suggest any role for the d5-HT7 receptor in Drosophila olfactory learning via expression in the mushroom body. Our results demonstrate a clear role for a serotonin receptor positively coupled to cAMP in Drosophila olfactory learning, consistent with elevation of cAMP levels in the MB neurons during olfactory appetitive learning 7,28 . Our results are also consistent with studies which suggest a role for octopamine receptors (which activate cAMP levels) in olfactory learning through expression in the MB neurons [bib_ref] Appetitive learning requires the alpha1-like octopamine receptor OAMB in the Drosophila mushroom..., Kim [/bib_ref] [bib_ref] Concerted actions of octopamine and dopamine receptors drive olfactory learning, Sabandal [/bib_ref] [bib_ref] Layered reward signalling through octopamine and dopamine in Drosophila, Burke [/bib_ref]. While we show strong evidence for the involvement of the d5-HT7 receptor in olfactory learning, we cannot completely rule out the possible crosstalk between different serotonin receptors in olfactory appetitive learning, possibly though expression in the MB neurons. This scenario remains a distinct possibility given the recent study highlighting the role of two distinct dopamine receptors in the Drosophila MB, which differentially modulate the temporal integration of odor cue with the reinforcement signal [bib_ref] Distinct dopamine receptor pathways underlie the temporal sensitivity of associative learning, Handler [/bib_ref].
Previous studies report that d5-HT7 receptor is expressed in several regions of the Drosophila brain including the central complex, large-field R neurons of the ellipsoid body, antennal lobes but is not expressed in the MB neurons [bib_ref] The serotonergic central nervous system of the Drosophila larva: anatomy and behavioral..., Huser [/bib_ref] [bib_ref] The serotonin 5-HT7Dro receptor is expressed in the brain of Drosophila, and..., Becnel [/bib_ref]. Using FACS sorting of Drosophila larval MB neurons we show that the d5-HT7 receptor transcript is expressed in the MB neurons [fig_ref] Figure 3: Genetic manipulation of d5-HT7 receptor expression in the larval MBs alters olfactory... [/fig_ref]. While our data suggests a clear role for the d5-HT7 receptor in olfactory appetitive learning, its role has been challenged in Drosophila larvae [bib_ref] Anatomy and behavioral function of serotonin receptors in Drosophila melanogaster larvae, Huser [/bib_ref]. In the study by Huser and colleagues, the authors demonstrate that ablation of d5-HT7 receptor expressing neurons has no effect on olfactory appetitive learning. The effect of knocking down early in development does not rule out a development contribution of the d5-HT7 receptor in olfactory learning. To rule out any development effect of d5-HT7 receptor knockdown on larval learning, we used the temperature sensitive Gal4/Gal80 system 55 to down-regulate d5-HT7 receptor expression for a short duration before training [fig_ref] Figure 4: Learning defects caused by the MB specific down-regulation of d5-HT7 can be... [/fig_ref]. This allowed us to rule out any effect of receptor downregulation on olfactory learning prior to the 3rd instar larval stage. Additionally, since the 201Y-Gal4 strain we used in our study to knockdown d5-HT7 receptor is known to express outside the MB neurons 17 , we ruled out any effect of receptor expression outside the MB neurons on larval learning. We were also able to restore normal olfactory learning by MB specific expression of d5-HT7 in receptor mutant larvae [fig_ref] Figure 4: Learning defects caused by the MB specific down-regulation of d5-HT7 can be... [/fig_ref]. Taken together, our results suggest that the expression of d5-HT7 receptor in the MB neurons but not outside of the MB, is essential for olfactory appetitive learning in Drosophila larvae. Since increase in cAMP levels have been observed by 2-Photon imaging in the Drosophila brain during appetitive learning 28 , we believe the d5-HT7 receptor by being positively coupled to cAMP, may play an essential role in olfactory learning. Thus, in addition to a handful of biogenic amine receptors (positively coupled to cAMP) whose expression in the MB neurons is essential for olfactory learning [bib_ref] Appetitive learning requires the alpha1-like octopamine receptor OAMB in the Drosophila mushroom..., Kim [/bib_ref] [bib_ref] Concerted actions of octopamine and dopamine receptors drive olfactory learning, Sabandal [/bib_ref] [bib_ref] An octopamine-mushroom body circuit modulates the formation of anesthesiaresistant memory in Drosophila, Wu [/bib_ref] [bib_ref] Layered reward signalling through octopamine and dopamine in Drosophila, Burke [/bib_ref] , our data provides evidence for the d5-HT7 as a key mediator of olfactory learning.
With recent advances in 2-photon imaging, an increasing number of studies have reported that replicating conditions encountered during olfactory learning by pairing of the CS with a biogenic amine like dopamine or octopamine (known to convey the US), results in changes in cAMP-PKA levels in the MB neurons [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref] [bib_ref] Cyclic AMP-dependent plasticity underlies rapid changes in odor coding associated with reward..., Louis [/bib_ref] [bib_ref] Dopaminergic modulation of cAMP drives nonlinear plasticity across the Drosophila mushroom body..., Boto [/bib_ref] [bib_ref] PKA dynamics in a Drosophila learning center: coincidence detection by rutabaga adenylyl..., Gervasi [/bib_ref] [bib_ref] Dynamics of learning-related cAMP signaling and stimulus integration in the Drosophila olfactory..., Tomchik [/bib_ref]. Additionally, enzymes regulating the cAMP-PKA signaling pathway like dunce1 and rutabaga1 has been identified as a key coincidence detectors during this process [bib_ref] Dunce phosphodiesterase acts as a checkpoint for Drosophila long-term memory in a..., Scheunemann [/bib_ref] [bib_ref] PKA dynamics in a Drosophila learning center: coincidence detection by rutabaga adenylyl..., Gervasi [/bib_ref]. These results emphasize the essential role of cAMP signaling as an essential co-incidence detector in the Drosophila MB neurons during olfactory learning. While activation of both dopamine and octopamine neurons and pairing with CS resulted in alteration of cAMP levels in the MB neurons (reviewed in 7 ), whether activation of serotonin neurons can do the same in olfactory learning is not known. By utilizing Drosophila primary neuronal cultures, which have been used to document changes in cAMP signaling on synaptic transmission [bib_ref] Fast excitatory synaptic transmission mediated by nicotinic acetylcholine receptors in Drosophila neurons, Lee [/bib_ref] [bib_ref] Suppression of inhibitory GABAergic transmission by cAMP signaling pathway: alterations in learning..., Ganguly [/bib_ref] [bib_ref] Suppression of excitatory cholinergic synaptic transmission by Drosophila dopamine D1-like receptors, Yuan [/bib_ref] , we show that application of serotonin causes an increase in excitatory cholinergic transmission through an increase in cAMP levels. We also show that this effect of serotonin is mediated through the d5-HT7 receptor [fig_ref] Figure 2: Expression of d5-HT7 receptor is essential for olfactory learning and excitatory synaptic... [/fig_ref]. Our results are consistent with studies showing an increase in cAMP levels during olfactory learning [bib_ref] Cyclic AMP-dependent plasticity underlies rapid changes in odor coding associated with reward..., Louis [/bib_ref]. Although the effect of serotonin on in vivo cAMP levels in
## Scientific reports
| (2020) 10:21267 | https://doi.org/10.1038/s41598-020-77910-5 www.nature.com/scientificreports/ MB neurons during conditions mimicking olfactory learning is yet to be explored, our data strongly suggests a role for the d5-HT7 receptor in coincidence detection via elevation of cAMP in the mushroom body neurons.
# Materials and methods
Fly strains. Wild-type flies used in all experiments were the w 1118 (a "Cantonized" white eye stock) and the Canton-S genotype (Wt). The Tph-Gal4 line was a kind gift from Dr. J. Kim at KIAST, Korea. The UAS-5HT7R line was gifted by Dr. Julian Dow at Univ. of Glasgow 50 and the UAS-Shi ts1 line was a kind gift from Dr. T Kitamoto at Univ. of Iowa 46 . MB-Gal80 strain was a gift from Dr. W. Joiner (UCSD). UAS-ChR2 was obtained from Dr. Barry Condron at Univ. of Virginia. The MB247-LexA(III) strain was a gift from Dr. Marcus Gallio (Northwestern University). All homozygous lines used in this study were made by standard genetic crosses using the w 1118 ; CyO/Sco; TM2/TM6B double balancer line. For P-element mobilization of the UAS-d5-HT7 gene from the II to the III chromosome, standard genetic cross scheme was used with a strain carrying the Δ 2-3 transposase enzyme in the w 1118 background. All the other lines used were from the Bloomington Stock Center, details of which are encapsulated below. Larval learning assay. The larval learning assay was adapted from the one described earlier [bib_ref] Induction of cAMP response element-binding protein-dependent medium-term memory by appetitive gustatory reinforcement..., Honjo [/bib_ref] [bib_ref] Distinctive neuronal networks and biochemical pathways for appetitive and aversive memory in..., Honjo [/bib_ref]. Briefly, 3rd
instar larvae, 84-87 h after egg laying on fly food were separated from the fly food by using 15% glucose solution.
Parental flies not more than 15 days old were used for egg laying. Since the density of the larvae was less than that of the fly food, the larvae floated to the top and were then transferred onto a sieve. The 500 μm sieve (Newark Wire Cloth Company) retained the larvae on top of the mesh whereas small bits of food and the glucose solution passed through. The larvae were washed several times with water to remove any traces of glucose.
For the training stage about 100 washed larvae were taken on a freshly made 2.5% agarose plate (8.5 cm in diameter) containing 1 mL of 1 M sucrose solution spread as a thin film on top. The inner surface of the lid of the plate had a filter paper disc (Whatmann filter) spotted with 10 μL of undiluted odor (Pentyl acetate or Propionic acid). The lid was then put on top of the agar plate. For control experiments, a similar agar plate was spread with a thin film of 1 mL of ddH 2 O instead of the sucrose solution. About 100 larvae were put on this control plate in the same order as described above for the training plate. The larvae were allowed to associate the odor with the sucrose and distilled water for 30 min at 24.5 °C in a uniformly lit fume hood with a white table top background.
For the testing phase of the experiment, 50-100 trained larvae were lined along the center of a fresh Petri dish containing 2.5% agarose. This dish had two filter discs, each placed on the top of a 1.5 mL Eppendorf tube cap pressed down into the agar at a distance of 0.7 cm from each edge of the plate. Each cap along with the filter disc formed the center of a 3 cm semi-circle on both sides of the center line. 2.5 μL of the odor used for training was then spotted on one of the filter discs and the lid was closed. The larvae were allowed to move towards the odor and non-odor side and the response index (R.I) were calculated after 3 min as described below:
Naïve olfactory response and gustatory response to 1 M sucrose was carried out as described earlier [bib_ref] Induction of cAMP response element-binding protein-dependent medium-term memory by appetitive gustatory reinforcement..., Honjo [/bib_ref].
Optogenetic larval learning assay. The optogenetic larval learning assays were performed using protocols similar to those described in studies from our group [bib_ref] Optogenetic rescue of locomotor dysfunction and dopaminergic degeneration caused by Alpha-Synuclein and..., Qi [/bib_ref]. Briefly, larvae were exposed to blue light (BL; intensity, 1 mW/mm 2 ), either during testing or prior to testing on plates containing 1 mM all-trans-retinal (Sigma-Aldrich). Light intensity was measured using a Sanwa Mobiken laser power meter. To activate chan-nelrhodopsin2, a blue light source LED (Luxeon Rebel Color LEDs, 07040 PB000-D, wavelength 470 nm) with a power supply (GW Instek, Laboratory DC power supply Model GPS-1830D) was used. The intensity of the blue light is 25 mW.
Immunohistochemistry. For staining the Drosophila larval brains, the brains were dissected out using sharp forceps in ice cold dissecting saline solution. The brains were then fixed in 4% paraformaldehyde for 30 min on ice and then washed thoroughly with PBS. The brains were then permeabilized in 10 mM PBSTX (PBS + 0.1% Triton X-100) and 5% Normal Goat Serum (Sigma-Aldrich, MO) for one hour at room temperature on a rocker. Primary antibody incubation was overnight at 4 °C. The following day, after three 30 min washes at room temperature in PBSTX, secondary antibody incubation was carried out for 2 h at room temperature (adult brains) or overnight at 4 °C (larval brains). The secondary antibody was diluted in the blocking solution. The brains were then washed thoroughly before mounting and imaging.
Confocal imaging. Larval brain images were taken using a Zeiss Laser Scanning Microscope 510 (Carl Zeiss, Inc., USA). For the larval mushroom body images, 1.0 μm z-sections were taken at 100 × magnification and then stacked to create a maximum intensity projection image using the Zeiss LSM Zen 2008 software.
Reverse transcription PCR analysis. Exon1-Exon2 boundary spanning primers were designed for the 5-HT7R gene homolog in Drosophila melanogaster. Primers for the RP49 housekeeping gene were used as a positive control. The Drosophila mRNA sequence was used in PrimerBlast (NCBI) to design the primers. Primers for the 5-HT7R homolog gene (Forward Primer: TCG TTG ACC CAG TTC CCG ACGA; Reverse Primer: CCA GTG CTG ACC TGC TGC CC) and RP49 primers (Forward Primer: GAG AAC GCA GGC GAC CGT TG; Reverse primer: TGA CCA TCC GCC CAG CAT AC) were obtained from Eurofins MWG Operon Inc. USA. These sets of primers were designed to yield a product size of 334 base pairs (5-HT7R) and 390 base pairs (RP49) with the isolated mRNA. Exon-exon spanning primers allowed elimination of products amplified from contaminating genomic DNA, as this would give a larger sized product than that expected with only mRNA. The mRNA was isolated using the RNAeasy Mini Kit (Qiagen Inc., CA) by following the protocol for animal tissue samples described in the product manual. 16-18 3rd instar larvae were isolated for each genotype and crushed in 300 μL www.nature.com/scientificreports/ RLT buffer using a pestle in a 1.5 mL tube. The lysate was passed through the QiaShredder (Qiagen Inc., CA) for homogenization and then an isolation protocol was followed as described in the product manual. After RNA isolation, the RNA was quantified using Nanodrop 1000 (Thermo Scientific, DE). Any contaminating genomic DNA was removed using the DNA-free kit (Ambion Inc./Applied Biosystems, TX) as per the manufacturer's protocol. After genomic DNA removal, the RNA was re-quantified and converted to cDNA using the Omniscript RT Kit (Qiagen Inc., CA) for 1 μg of RNA in a final reaction volume of 20 μL as per the manufacturer's protocol. PCR was performed using a PCR kit (Clonetech Laboratories Inc., WI), for a final volume of 30 μL. 10 mM stock solution of each primer for both sets and 1 μL of mRNA were added to the reaction mixture in both cases. Annealing temperature for 5-HT7R and RP49 primers was 65 °C and 60 °C respectively. A standard 30 cycle PCR reaction was set up on a PTC 150 Mini Cycler (MJ Research/Bio-Rad Laboratories., CA) for both the primer sets.
For the mRNA isolation from FACS sorted GFP + and GFP− cells, a total of one million GFP + cells were taken by pooling together cells obtained from 3 separate sorts. The mRNA was isolated using the protocol described for animal cells in the Qiagen RNAeasy Micro Kit (Qiagen Inc., CA). The mRNA was converted to cDNA using the AffinityScript qPCR cDNA synthesis Kit (Stratagene, CA) for a final volume of 40 μL as per the manufacturer's protocol. PCR was performed using a PCR kit (Takara Bio Inc., Japan) for a final volume of 20 μL. A 10 mM stock solution of each primer was used for both genes. The annealing temperature for 5-HT7R and RP49 primers was 65 °C and 60 °C respectively. A 40 cycle PCR reaction was set up on a PTC 150 Mini Cycler (MJ Research/ Bio-Rad Laboratories., Hercules CA) for both primer sets.
## Fluorescence activated cell sorting (facs) of mushroom body neurons.
To obtain a single cell suspension for FACS sorting, 3rd instar larvae from the 201Y-Gal4; UAS-mCD8::GFP homozygous strain (84-87 hafteregg-laying) were collected. The brains were dissected and the ventral nerve cords removed for 100 such larvae for each sort experiment. About 25 w 1118 (wild-type) larval brains were dissected simultaneously and served as the negative control for the FACS sorting. The brains were dissected and placed in ice cold dissecting saline solution before processing. The brains were washed with 0.1% PBS solution twice and spun down in 1.5 mL Eppendorf tubes. A pestle was used to homogenize the brains before treatment with 0.25% Trypsin-EDTA (Sigma-Aldrich, MO) for 4-5 min at 37 °C. Fetal Bovine Serum (Invitrogen, USA) to a final concentration of 4% was added to stop the action of Trypsin-EDTA. The suspension was passed through a 70 μm filter (Cata-log# 35 2350, BD Biosciences, CA) and re-suspended in 0.1% PBS before sorting. For the sorting procedure, a pre-sort was done on BDAria FACS sorter with wild-type cells to set the gates for sorting GFP + and GFP − cells. Following this, GFP +/GFP − cells were sorted from the 201Y-Gal4; UAS-mCD8::GFP strain. The sorted cells were flash frozen in liquid nitrogen and stored at − 80 °C until mRNA isolation. Data analysis. Individual PSCs were analyzed using the Minianalysis detection software (Synaptosoft, Decatur, GA, USA) with threshold criteria for individual events of 7.5 pA amplitude and 7 pF charge transfer for cholinergic PSCs [bib_ref] Fast excitatory synaptic transmission mediated by nicotinic acetylcholine receptors in Drosophila neurons, Lee [/bib_ref]. Only events with a fast rising and slowly decaying phase were used for analysis. For each acquired trace of data, all events were detected, and the frequency was calculated every 5 s. If the PSC frequency declined continuously over 20 s before drug application, a rundown of PSCs was suspected, and the data were not analyzed further. For the frequency analysis, only PSCs (before drug application) with a stable frequency longer than 20 s were used to calculate the control average frequency. This calculated control average frequency was then normalized as 100%. The frequency of the entire recording was compared with the control average frequency, and then the percent frequency was calculated and plotted before, during and after drug application. PSCs were analyzed in the same way and plotted in Origin 7.0 (OriginLab, MA) and Graph Pad Prism (La Jolla, CA).
All summary graphs show the means ± SEMs. Statistical comparisons with Student's t test yielded *p < 0.05; **p < 0.01, and ***p < 0.001 and with Two-Way ANOVA for comparison between groups yielded #p < 0.01 and ##p < 0.001. All analysis was carried out using Graph Pad Prism (La Jolla, CA). www.nature.com/scientificreports/
[fig] Figure 1: Output of serotonergic neurons is essential for olfactory appetitive learning. (A) Reconstituted splitGFP fluorescence (GRASP) showing regions where projections of serotonergic (5-HT) neurons (green) make synaptic contacts on to the larval mushroom body (MB) marked by the MB247-LexA signal (red). (A′) Enlarged view of both vertical and medial lobes from A, showing points of synaptic contact [/fig]
[fig] Figure 2: Expression of d5-HT7 receptor is essential for olfactory learning and excitatory synaptic plasticity. (A) Patch-clamp recording from wild-type (w 1118 ) neurons in culture showing spontaneous cholinergic EPSCs (sEPSCs). Application of 20 µM 5-HT (for 30 s) shows an increase in the frequency of cholinergic EPSCs. (B) The application of 20 µM 5-HT (for 30 s) in the presence of a PKA inhibitor (50 µM H-89, in the bath solution) completely blocks the increase in the frequency of cholinergic EPSC. (C) Molecular confirmation of d5-HT7 null mutant. (Top) A diagrammatic representation of the d5-HT7 receptor gene in Drosophila showing the location of the P element insertion and the exon-exon spanning primers for the d5-HT7 receptor gene (arrows). (Bottom) Gel image showing the PCR products from the d5-HT7 null mutant and wild-type (w 1118 ) larvae. No PCR product was detected in the d5-HT7 null strain. RP49 gene served as a control. (D) In contrast to wild-type neurons, neurons cultured from the d5-HT7 null mutant show no increase in the frequency of cholinergic EPSCs by 20 µM 5-HT (for 30 s). (E) A cumulative plot showing the effect of 20 µM 5-HT on the normalized frequency of cholinergic EPSCs on wild-type (w 1118 ) neurons (n = 9), wild-type (w 1118 ) neurons in the presence of 50 µM H-89 (n = 5) and on the d5-HT7 null neurons (n = 7) [/fig]
[fig] Figure 3: Genetic manipulation of d5-HT7 receptor expression in the larval MBs alters olfactory learning. (A) Fluorescence activated cell sorting (FACS) was utilized to confirm expression of d5-HT7 receptor in the larval mushroom body. FACS plots showing the gates used to isolate GFP + and GFP− neurons from larval brain of 201Y-GAL4; UAS-mCD8::GFP strain. This strain expresses GFP specifically in the MB neurons of 3rd instar larvae. Average frequency of GFP + cells obtained from three independent sorts was 5.15%. (B) mRNA isolated from GFP + (~ 90,000 cells) and GFP− (~ 600,000) cells sorted in (A) was used as a template for RT-PCR analysis using primers specific to RP49 and d5-HT7 receptor. The gel image shows that MB/GFP + neurons express the d5-HT7 receptor. (C) MB specific Gal4 drivers (201Y and 30Y) were crossed to UAS-d5-HT7-RNAi strain. Both the 201Y-Gal4; UAS-d5-HT7-RNAi (n = 7) and 30Y-Gal4; UAS-d5-HT7-RNAi (n = 6) strains showed impaired learning. However, the strains carrying only the 201Y-Gal4 (n = 7), 30Y-Gal4 (n = 7), UAS-d5-HT7-RNAi (n = 6) transgenes and a strain expressing UAS-d5-HT7-RNAi under 5-HT7-Gal4 (n = 5) show normal olfactory associative learning. The overexpression of d5-HT7 receptor under the MB specific Gal4 driver (201Y-Gal4; UAS-d5-HT7) resulted in higher associative learning when compared to the 201Y-Gal4/ + larvae. Student t-test: **p < 0.01; ***p < 0.001. Two-way ANOVA for comparison across genotypes: #p < 0.01; ##p < 0.001. in Drosophila MB neurons regulates olfactory learning. To establish if d5-HT7 [/fig]
[fig] Figure 4: Learning defects caused by the MB specific down-regulation of d5-HT7 can be rapidly induced and are fully reversible. (A) Schematic representation of the Gal80 ts1 /Gal4 expression system used for spatiotemporal regulation of gene expression. At 19 °C, Gal80 ts1 inhibits Gal4 however transfer to 30 °C denatured Gal80 ts1 resulting in Gal4 expression. (B) Schematic representation of the training/testing paradigms used to assess olfactory associative learning during spatio-temporally restricted expression of d5-HT7 in the MB neurons. (B′) 201Y-Gal4/tub-P-Gal80 ts1 ; UAS-d5-HT7-RNAi/TM6 larvae grown at 19 °C throughout development show normal learning (n = 6), but when these larvae were subjected to heat shock at 30 ºC for 16 h prior to training, it resulted in impaired learning (n = 6). (C) A mushroom body specific Gal80 strain (MB-Gal80) is crossed with a 201Y-Gal4; UAS-d5-HT7-RNAi homozygous strain and then subjected to learning paradigm. Gal80 inhibits Gal4 expression in the MB neurons and restores normal olfactory learning (n = 7). The expression of a wild-type copy of the d5-HT7 receptor gene in the MB neurons of the d5-HT7 receptor null larvae (201YGal4/UAS-d5-HT7; d5-HT7 null/d5-HT7 null) completely rescues olfactory associative learning (n = 7). However, the 201Y-Gal4/201Y-Gal4; d5-HT7 null/d5-HT7 null strain larvae (n = 5) and UAS-d5-HT7/ UAS-d5-HT7; d5-HT7 null/d5-HT7 null strain larvae (n = 5) with the Gal4 and the UAS transgenes in d5-HT7 null mutant background, show impaired olfactory learning. Student t-test: *p < 0.05; ***p < 0.001.Scientific Reports| (2020) 10:21267 | https://doi.org/10.1038/s41598-020-77910-5 [/fig]
[fig] Response: Index(R.I) = No. of larvae in the 3 cm ring with odor − No. of larvae in the 3 cm ring without odor Total number of larvae in both the rings Scientific Reports | (2020) 10:21267 | https://doi.org/10.1038/s41598-020-77910-5 [/fig]
|
10.3390/s19184016
|
CCBY
|
6767287
|
31533366
|
s2orc_pubmed_articles
|
Fluorescence Characteristics of Dissolved Organic Matter (DOM) in Percolation Water and Lateral Seepage Affected by Soil Solution (S-S) in a Lysimeter Test
The composition and structure of dissolved organic matter (DOM) are sensitive indicators that guide the water infiltration process in soil. The DOM chemical composition in seepage affects river water quality and changes soil organic matter (SOM). In this lysimeter test study, fluorescence spectra and optical indices were used to examine the interaction between the percolation water (P-W) and leachate water (L-W) DOMs affected by the soil solution (S-S). The L-W DOM had a higher aromaticity (SUVA 254 ), average molecular weight (S 275-295 ) and terrestrial source (fluorescence index (FI)), but fewer autochthonous sources (biological index (BIX)) than the P-W DOM. Organic carbon standardization (OCS) and protein-(PLF), fulvic-(FLF) and humic-like fluorescence (HLF) intensity showed that L-W DOM increased 44%, 55% and 81%, respectively, compared to the P-W DOM. The linear regression slopes between OCS FLF and PLF were 0.62, 1.74 and 1.79 for P-W, L-W and S-S, respectively. The slopes between OCS HLF and PLF were 0.15, 0.58 and 0.64 for P-W, L-W and S-S, respectively. The P-W DOM was in contact with the soil litter layer, where S-S labile lignin phenolic compounds released and dissolved into the L-W DOM. This increased its aromaticity, and extent of humification. soil plant litter lignin, which contains phenolic compounds[4,6,7]. The shallow seepage may increase the dissolved organic carbon (DOC) concentration and change the DOM chemical composition and structure. The discharged water DOM results in degraded river water quality and causes SOM loss[7,8].DOM and SOM are heterogeneous and complex organic mixtures. Many advanced techniques have been applied to the qualitative and quantitative DOM and SOM chemical composition and structure, such as NMR, FTIR, HPLC and Py-GC-MS. Advanced technical analysis requires complex pretreatment processes and is time consuming[9][10][11][12][13]. Spectral methodology, UV-Vis and fluorescence spectroscopy are rapid, non-destructive and sensitive detection methods that are widely used to determine measured chemical compositions and structures of various DOM and SOM[11,[14][15][16][17].UV-Vis indices SUVA 254 and S 275-295 show DOM and SOM aromaticity and average molecular weight, respectively[15,17,18]. The fluorescence index (FI) distinguishes terrestrial source contribution[17,19,20]. The biological Iindex (BIX) distinguishes the contribution of autochthonous sources[17,21]. A fluorescence excitation/emission matrix (EEM) analyzes DOM species and water quality characteristics with fluorescent intensity and intensity ratios[22][23][24][25]. The three regions of protein-(PLF), fulvic-(FLF) and humic-like fluorescence (HLF) represent the wavelength ranges of intensities in EEM, respectively. The chemical composition of DOM and SOM were examined by the fluorescence intensity and ratio change in the three regions[26,27].Previous studies have used optical indicators to explore DOM differences and mainly focused on surface water and deep groundwater[6,[28][29][30]. However, research on the percolation water (P-W) DOM in shallow infiltration and lateral seepage affected by the SOM chemical composition and structure is still lacking.In this study, using a medium-scale lysimeter, optical indices (UV-Vis and fluorescence) and fluorescence spectroscopy were used to examine the chemical structure and composition relationship between P-W and leachate water (L-W) DOM as well as extracted soil organic solution (S-S). The aromaticity, average molecular weight and the terrestrial and autochthonous contributions were used to examine the P-W and L-W DOM as well as the S-S with optical indicators SUVA 254 , S 275-295 , FI and BIX, respectively. In addition, the fluorescent intensity differences between PLF, FLF and HLF in P-W, L-W and S-S was examined with intensity mean and linear regression. The chemical structure and composition differences between P-W and L-W DOM and how they were affected by the S-S were discussed.
# Introduction
Dissolved organic matter (DOM) is an important component of water; it changes drinking water quality and reduces water treatment efficiency [bib_ref] Energy resources and global development, Chow [/bib_ref]. Surface water DOM infiltration into soil is affected by the natural DOM characteristics such as chemical structure, functional group content or molecular size [bib_ref] Controls of bioavailability and biodegradability of dissolved organic matter in soils, Marschner [/bib_ref]. Furthermore, DOM transport is controlled by both physicochemical and biological processes in soil that serve to retain and transform the DOM and release soil organic matter (SOM) from different soil fractions [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] The dynamic exchange of dissolved organic matter percolating through six diverse soils, Scott [/bib_ref].
The hypothesis of DOM infiltration into the deep groundwater aquifer assumes that there are two processes controlling the DOM chemical composition and structure. One is that DOM hydrophobic components adsorb on soil minerals and/or DOM partitions into the SOM. The other process is that the DOM metabolism by microorganisms subsequently produces small molecular weight, hydrophilic and autochthonous DOM and the metabolites of microorganisms [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] Riverine export of aged carbon driven by flow path depth and residence..., Barnes [/bib_ref]. Therefore, the DOM in the deep groundwater aquifer has mainly biological sources. However, the shallow infiltration of lateral seepage in river banks and drainage channels may yield a preferential release of dissolved DOM containing Sample pH EC μS/cm DO mg/L OM% P-W 7.61 ± 0.42 1840 ± 240 1.44 ± 0.16 -* L-W 7.90 ± 0.11 110.3 ± 5.8 --S-S 8.24 ± 0.08 --2.92 ± 0.37 *Did not measure; a sand = 45.8%, silt = 33.1% and clay = 21.1% (loam soil)
## Soil collection and extraction
Before each percolation test, an auger (inner diameter, 5 cm) was used to collect soil samples to a soil depth of 60 cm. Each of the soil columns was uniformly mixed to produce a soil sample. Two soil samples were taken from each lysimeter in each test; a total of 16 soil samples was collected in the four experiments. The hole was refilled with the adjacent farm soil.
The soil samples were air-dried and large particles were removed. The soil samples were then ground and sieved at 2.0 mm (sieve #10). The sieved soil sample was washed with 0.1 N HCl acid to remove alkaline earth metals and carbonate. The residual soil sample was mixed with a 0.1 N NaOH solution at a ratio of 1/20 (soil/water) and extracted with a reciprocating shaker (150 rpm, 24 h). The soil suspension was centrifuged at 3200 g for 20 min and the supernatant was passed through a 0.45 μm membrane (Pall) to collect the soil solution (S-S). The S-S solution was stored at 4 °C. The DOC concentration, UV-Vis and fluorescence spectra were measured within three days after the S-S was extracted.
## Uv/vis and fluorescence spectrophotometer measurement
The filtered bulk P-W and L-W samples and the S-S samples were diluted with ultrapure water to a concentration of 5 mg-C/L to measure UV/Vis absorbance with a UV/Vis spectrophotometer
## Water samples collection
There were three types of samples in this study. The P-W sample was well water pumped from 35 m and 15 m deep groundwater away from the test site. The L-W sample was collected from the drainage pipe outlet as shown in [fig_ref] Figure 1: The configuration and placement of the experimental lysimeter [/fig_ref]. The S-S sample was extracted from the collected soils in the test site. Water quality parameters (pH, dissolved oxygen (DO), EC) of P-W and L-W were measured as well as the soil basic properties, pH and soil organic matter content. [fig_ref] Table 1: Water quality of P-W and L-W and soil basic properties a [/fig_ref] lists the water quality of P-W and L-W and soil basic properties. The water infiltration tests were carried out four times. In each test, two P-W samples were collected (one for each lysimeter) for a total of eight P-W samples. Additionally, in each test, eight L-W samples (four samples for each lysimeter) were collected for a total of 32 L-W samples. The DOC concentration, UV-Vis and fluorescence spectra of the water samples were measured after filtration with a membrane (<0.45 µm).
## Soil collection and extraction
Before each percolation test, an auger (inner diameter, 5 cm) was used to collect soil samples to a soil depth of 60 cm. Each of the soil columns was uniformly mixed to produce a soil sample. Two soil samples were taken from each lysimeter in each test; a total of 16 soil samples was collected in the four experiments. The hole was refilled with the adjacent farm soil.
The soil samples were air-dried and large particles were removed. The soil samples were then ground and sieved at 2.0 mm (sieve #10). The sieved soil sample was washed with 0.1 N HCl acid to remove alkaline earth metals and carbonate. The residual soil sample was mixed with a 0.1 N NaOH solution at a ratio of 1/20 (soil/water) and extracted with a reciprocating shaker (150 rpm, 24 h). The soil suspension was centrifuged at 3200 g for 20 min and the supernatant was passed through a 0.45 µm membrane (Pall) to collect the soil solution (S-S). The S-S solution was stored at 4 - C. The DOC concentration, UV-Vis and fluorescence spectra were measured within three days after the S-S was extracted.
## Uv/vis and fluorescence spectrophotometer measurement
The filtered bulk P-W and L-W samples and the S-S samples were diluted with ultrapure water to a concentration of 5 mg-C/L to measure UV/Vis absorbance with a UV/Vis spectrophotometer (Hitachi, U-2900). The absorbance at 700-800 nm was used as the background; the absorbance of the sample was subtracted from the background value [bib_ref] Absorption spectral slopes and slope ratios as indicators of molecular weight, source,..., Helms [/bib_ref]. The samples were scanned at UV/Vis wavelengths 800-200 nm, and the scanning rate was 800 nm/min. The samples were acidified to a pH of less than 3 with 3 M sulfuric acid and analyzed by a fluorescence spectrometer (Hitachi, F-7000). Ultrapure water was used as a blank, and the measured data were subtracted from the blank. The UV/Vis absorbance at 254 nm was <0.2; therefore, the inner filter effect was ignored [bib_ref] Properties of fluorescent dissolved organic matter in the Gironde Estuary, Huguet [/bib_ref] [bib_ref] Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic material..., Mcknight [/bib_ref]. Excitation wavelengths were from 200 to 450 nm at 5 nm increments and emission wavelengths were from 250 to 550 nm at 2 nm increments. The scanning rate was 2400 nm/min.
## Optical index
The UV/Vis specific ultraviolet absorbance at 254 nm (SUVA 254 , L/mg-C/m) is calculated as Equation (1) [bib_ref] Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition..., Weishaar [/bib_ref]. The UV-Vis spectral slopes (S 275-295 ) is the nonlinear fit of an exponential function to the absorption spectrum over the wavelength range between 275 and 295 nm [bib_ref] Absorption spectral slopes and slope ratios as indicators of molecular weight, source,..., Helms [/bib_ref]. The fluorescence index (FI) is calculated as Equation (2) [bib_ref] Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic material..., Mcknight [/bib_ref]. The biological index (BIX) is calculated as Equation (3) [bib_ref] Properties of fluorescent dissolved organic matter in the Gironde Estuary, Huguet [/bib_ref]. [bib_ref] Fluorescence excitation-emission matrix characterization of some sewage-impacted rivers, Baker [/bib_ref] [bib_ref] Microbial transformation of dissolved leaf litter organic matter and its effects on..., Hur [/bib_ref] [bib_ref] Using synchronous fluorescence technique as a water quality monitoring tool for an..., Hur [/bib_ref]. The PLF, FLF and HLF values are the sum of the fluorescence intensities of the measured solution in the design regions. The PLF, FLF and HLF organic carbon standardized (OCS) are the fluorescence intensity divided by DOC concentration for P-W, L-W and S-S samples.
[formula] SUVA 254 = A 254 /[DOC] × 100 (1) FI = F λem=450 nm F λem=500 nm , λ ex=370 nm (2) BIX = F λem=380 nm F λem=430 nm , λ ex=310 nm(3) [/formula]
## Statistic
All statistical analyses were performed using S-Plus V 6.2 statistical software. Statistical significance was accepted at p < 0.05. Owing to the different size and lack of homogeneity of variance, the data sets were analyzed using a nonparametric test. Kruskal-Wallis tests were performed to compare the differences among the three data sets. Differences in the two data sets were analyzed using the Mann-Whitney U test. The EEM and UV/Vis indicators calculation and the EEM analysis and plots followed the R script developed by Lapworth and Kinniburgh [bib_ref] An R script for visualising and analysing fluorescence excitation-emission matrices (EEMs), Lapworth [/bib_ref]. This followed the needs of the experiments to modify the R software (2.13.2 V) script.
# Results and discussion
## Doc concentrations of dom and s-s
The DOC concentrations of P-W, L-W and S-S are shown in [fig_ref] Table 2: Disolved organic matter [/fig_ref]. The average concentration of alkaline-extracted soil organic matter (S-S) (680 mg/kg) was higher than water extracted soil organic matter (WEOM) in agricultural soil (56-81 mg/kg). The DOC concentration of L-W DOM was significantly higher than P-W DOM (p < 0.001). In a soil column (30 cm high) infiltration test [bib_ref] Composition of dissolved organic matter in groundwater, Longnecker [/bib_ref] , the DOC concentrations were 0.80-1.01 mg/L and 1.14-2.38 mg/L in percolation and leachate water, respectively. The leachate DOC concentrations were higher than the percolation water DOC concentrations and the results were similar to the lysimeter test. However, the DOC concentration in groundwater was very low [bib_ref] Understanding groundwater, surface water, and hyporheic zone biogeochemical processes in a Chalk..., Lapworth [/bib_ref] and similar to P-W in the lysimeter test. In addition, the DOC concentrations of deep groundwater were much lower than DOC concentration in rainwater and surface water; the deep groundwater DOC concentration was significantly reduced compared to the surface water DOC concentration. For example, in the Yasu River watershed, Mostofa et al. [bib_ref] Dynamics and characteristics of fluorescent dissolved organic matter in the groundwater, river..., Mostofa [/bib_ref] reported that the DOC concentrations were 1.31 ± 1.06 and 1.90 ± 0.74 mg/L for groundwater and river water, respectively. The DOC concentration of groundwater decreased. Shen et al.reported that the DOC concentration of surface water was 9.47 ± 6.29 mg/L; the groundwater DOC concentration was significantly decreased, ranging from 0.82 to 1.19 mg/L.
## Optical indices
Optical methods have been widely applied to analyze the chemical composition and structure of DOM and S-S [bib_ref] An overview of the methods used in the characterisation of natural organic..., Matilainen [/bib_ref] [bib_ref] Absorption spectral slopes and slope ratios as indicators of molecular weight, source,..., Helms [/bib_ref] [bib_ref] Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition..., Weishaar [/bib_ref] [bib_ref] Characterization of dissolved organic matter in cave and spring waters using UV-Vis..., Birdwell [/bib_ref] [bib_ref] Lipid biomarkers and spectroscopic indices for identifying organic matter sources in aquatic..., Derrien [/bib_ref] [bib_ref] Differences in N loading affect DOM dynamics during typhoon events in a..., Yeh [/bib_ref] [bib_ref] Use of fluorescence quenching method to measure sorption constants of phenolic xenoestrogens..., Yeh [/bib_ref]. In this study, UV-Vis indices SUVA 254 and spectral slope S 275-295 and fluorescence indices FI and BIX were used to examine the chemical structure and composition of P-W, L-W DOM and S-S. [fig_ref] Table 2: Disolved organic matter [/fig_ref] lists the optical index values. The SUVA 254 index is positively correlated with the aromatic content [bib_ref] An overview of the methods used in the characterisation of natural organic..., Matilainen [/bib_ref] [bib_ref] Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition..., Weishaar [/bib_ref]. The quantitative order of average SUVA 254 values was S-S > LW > PW, and the SUVA 254 of S-S was significantly higher than the P-W value (p = 0.009). The SUVA 254 value (<3 L/mg-C/m) indicated that the DOM solution mainly consisted of hydrophilic compounds, while the SUVA 254 value (>4 L/mg-C/m) DOM solution mainly contained hydrophobic compounds [bib_ref] An overview of the methods used in the characterisation of natural organic..., Matilainen [/bib_ref]. The average SUVA 254 values for the three solutions were 2.68, 3.37 and 3.84 L/mg-C/m for P-W, L-W and S-S, respectively. The SUVA 254 values > 4 were 25, 13 and 38% for P-W, L-W and S-S, respectively, and the SUVA 254 values < 3 were 63, 19 and 12% for P-W, L-W and S-S, respectively. The SUVA 254 values suggested that the P-W DOM contained mainly hydrophilic compounds, L-W DOM and S-S contained partially hydrophilic and partially hydrophobic compounds [bib_ref] An overview of the methods used in the characterisation of natural organic..., Matilainen [/bib_ref] [bib_ref] Enhanced coagulation: US requirements and a broader view, Edzwald [/bib_ref].
The UV-Vis spectral slope S 275-295 is inversely proportional to the average molecular weight [bib_ref] Absorption spectral slopes and slope ratios as indicators of molecular weight, source,..., Helms [/bib_ref] [bib_ref] Correlations between aromaticity of dissolved organic matter and trace metal concentrations in..., Kikuchi [/bib_ref] , and the magnitude order of S 275-295 was P-W > L-W > S-S, which suggested L-W had a lower average molecular weight than S-S, but had a higher average molecular weight than P-W.
Surface water SUVA 254 values typically ranged from 1.0 L/mg-C/m to 6.0 L/mg-C/m [bib_ref] Optical properties of dissolved organic matter (DOM): Effects of biological and photolytic..., Hansen [/bib_ref]. Surface water had S 275-295 values ranging from 0.014 to 0.018 nm −1 [bib_ref] Absorption spectral slopes and slope ratios as indicators of molecular weight, source,..., Helms [/bib_ref] , and the terrestrial systems had values ranging from 0.012 to 0.023 nm −1 [bib_ref] Dissolved organic carbon and chromophoric dissolved organic matter properties of rivers in..., Spencer [/bib_ref]. The DOM values of SUVA 254 and S 275-295 were comparable to those for surface water and deep groundwater, but the trend was reversed [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] Comparative study of dissolved organic matter from groundwater and surface water in..., Chen [/bib_ref].
Shen et al.investigated rainfall water and groundwater DOM. The results showed that the SUVA 254 values were 2.88 ± 0.38 and 1.54 ± 0.09 L/mg-C/m, and the S 275-295 values were 0.015 ± 0.001 and 0.021 ± 0.001 nm −1 for surface water and groundwater DOM, respectively.
The hypothesis for surface water percolation into deep groundwater assumes that the DOM hydrophobic compounds sorb onto the minerals and partition into the soil organic matter [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] Riverine export of aged carbon driven by flow path depth and residence..., Barnes [/bib_ref]. Therefore, the deep percolation DOM contain a lower aromatic abundance and average molecular weight than the surface water DOM. The hypothesis is in contrast to the results of P-W and L-W in the lysimeter test.
FI provides an indicator for distinguishing DOM derived from terrestrial and microbial sources [bib_ref] Characterization of dissolved organic matter in cave and spring waters using UV-Vis..., Birdwell [/bib_ref]. The low FI value (<1.4) had a strong terrestrial source contribution but the high FI value (>1.9) had a weak terrestrial source [bib_ref] Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic material..., Mcknight [/bib_ref]. The BIX is an indicator for the presence of autochthonous material where a high value (>1.0) corresponds to a recently produced DOM of autochthonous origin. Low BIX value (<0.6) suggested an allochthonous origin [bib_ref] Characterization of dissolved organic matter in cave and spring waters using UV-Vis..., Birdwell [/bib_ref] [bib_ref] Properties of fluorescent dissolved organic matter in the Gironde Estuary, Huguet [/bib_ref]. The range of BIX and FI values of P-W and L-W showed that the DOM mainly contained biological or aquatic bacterial origin. S-S contained an intermediate autochthonous component.
The FI and BIX values of S-S were similar to the soil WEOM [bib_ref] Spectroscopic characteristics of dissolved organic matter in afforestation forest soil of Miyun..., Gao [/bib_ref]. The FI and BIX values of P-W and L-W DOM were similar to the values of groundwater FI and BIX in rural areas [bib_ref] Spectral fingerprints of groundwater organic matter in rural areas, Pavelescu [/bib_ref]. The quantitative order was P-W > L-W > S-S for both FI and BIX values. S-S had the highest terrestrial and allochthonous sources but P-W had the highest aquatic and autochthonous sources. FI and BIX values of P-W DOM were significantly larger than L-W DOM (p < 0.01), suggesting that L-W DOM contained more terrestrial and less new microbial sources than P-W DOM [bib_ref] Characterization of dissolved organic matter in cave and spring waters using UV-Vis..., Birdwell [/bib_ref] [bib_ref] Properties of fluorescent dissolved organic matter in the Gironde Estuary, Huguet [/bib_ref].
## Fluorescence eem
The fluorescence excitation/emission matrices (EEMs), based on the Ex/Em maxima of the fluorescence peaks, are a useful tool for distinguishing different types and sources of natural waters and soil organic matter [bib_ref] The impact of glacier runoff on the biodegradability and biochemical composition of..., Fellman [/bib_ref] [bib_ref] Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter, Chen [/bib_ref] [bib_ref] Use of fluorescence quenching method to measure sorption constants of phenolic xenoestrogens..., Yeh [/bib_ref] [bib_ref] Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy, Coble [/bib_ref]. [fig_ref] Figure 2: The excitation/emission matrix [/fig_ref] -c shows the fluorescence EEM of the three solutions that show P-W and L-W DOM had three major peaks (Peak B, Peak A and Peak C), and S-S had a fourth peak (Peak M). The Peak C excitation/emission wavelength was centered at Ex/Em = 320 − 325/400 − 425 nm, which is attributed to the UVC humic-like substances from terrestrial sources. The Peak A wavelength was centered at Ex/Em = 230 − 235/400 − 410 nm, which is attributed to the recently generated UVA humic-like substances. The Peak B wavelength was centered at Ex/Em = 220 − 230/300 − 310 nm, which is attributed to the tyrosine protein-like material. The Peak M wavelength was centered at the Ex/Em = 270/425 nm, which is classified as a marine humic-like substance, autochthonous production and a low molecular weight substance [bib_ref] Fluorescence analysis of dissolved organic matter in natural, waste and polluted waters-A..., Hudson [/bib_ref] [bib_ref] Characterization of dissolved organic matter in cave and spring waters using UV-Vis..., Birdwell [/bib_ref] [bib_ref] The impact of glacier runoff on the biodegradability and biochemical composition of..., Fellman [/bib_ref] [bib_ref] Higher reactivity of allochthonous vs. autochthonous DOC sources in a shallow lake, Catalán [/bib_ref] [bib_ref] Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy, Coble [/bib_ref]. It noteworthy that the Peak C Em wavelength in the S-S and L-W was 10-15 nm right-shifted more than the P-W Em wavelength.
Sensors 2019, 19, x FOR PEER REVIEW had a weak terrestrial source [bib_ref] Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic material..., Mcknight [/bib_ref]. The BIX is an indicator for the presence of autochthonous material where a high value (>1.0) corresponds to a recently produced DOM of autochthonous origin. Low BIX value (<0.6) suggested an allochthonous origin [bib_ref] Characterization of dissolved organic matter in cave and spring waters using UV-Vis..., Birdwell [/bib_ref] [bib_ref] Properties of fluorescent dissolved organic matter in the Gironde Estuary, Huguet [/bib_ref]. The range of BIX and FI values of P-W and L-W showed that the DOM mainly contained biological or aquatic bacterial origin. S-S contained an intermediate autochthonous component.
The FI and BIX values of S-S were similar to the soil WEOM [bib_ref] Spectroscopic characteristics of dissolved organic matter in afforestation forest soil of Miyun..., Gao [/bib_ref]. The FI and BIX values of P-W and L-W DOM were similar to the values of groundwater FI and BIX in rural areas [bib_ref] Spectral fingerprints of groundwater organic matter in rural areas, Pavelescu [/bib_ref]. The quantitative order was P-W > L-W > S-S for both FI and BIX values. S-S had the highest terrestrial and allochthonous sources but P-W had the highest aquatic and autochthonous sources. FI and BIX values of P-W DOM were significantly larger than L-W DOM (p < 0.01), suggesting that L-W DOM contained more terrestrial and less new microbial sources than P-W DOM [bib_ref] Characterization of dissolved organic matter in cave and spring waters using UV-Vis..., Birdwell [/bib_ref] [bib_ref] Properties of fluorescent dissolved organic matter in the Gironde Estuary, Huguet [/bib_ref].
## Fluorescence eem
The fluorescence excitation/emission matrices (EEMs), based on the Ex/Em maxima of the fluorescence peaks, are a useful tool for distinguishing different types and sources of natural waters and soil organic matter [bib_ref] The impact of glacier runoff on the biodegradability and biochemical composition of..., Fellman [/bib_ref] [bib_ref] Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter, Chen [/bib_ref] [bib_ref] Use of fluorescence quenching method to measure sorption constants of phenolic xenoestrogens..., Yeh [/bib_ref] [bib_ref] Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy, Coble [/bib_ref]. [fig_ref] Figure 2: The excitation/emission matrix [/fig_ref] -c shows the fluorescence EEM of the three solutions that show P-W and L-W DOM had three major peaks (Peak B, Peak A and Peak C), and S-S had a fourth peak (Peak M). The Peak C excitation/emission wavelength was centered at Ex/Em = 320 − 325/400 − 425 nm, which is attributed to the UVC humic-like substances from terrestrial sources. The Peak A wavelength was centered at Ex/Em = 230 − 235/400 − 410 nm, which is attributed to the recently generated UVA humic-like substances. The Peak B wavelength was centered at Ex/Em = 220 − 230/300 − 310 nm, which is attributed to the tyrosine protein-like material. The Peak M wavelength was centered at the Ex/Em = 270/425 nm, which is classified as a marine humic-like substance, autochthonous production and a low molecular weight substance [bib_ref] Fluorescence analysis of dissolved organic matter in natural, waste and polluted waters-A..., Hudson [/bib_ref] [bib_ref] Characterization of dissolved organic matter in cave and spring waters using UV-Vis..., Birdwell [/bib_ref] [bib_ref] The impact of glacier runoff on the biodegradability and biochemical composition of..., Fellman [/bib_ref] [bib_ref] Higher reactivity of allochthonous vs. autochthonous DOC sources in a shallow lake, Catalán [/bib_ref] [bib_ref] Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy, Coble [/bib_ref]. It noteworthy that the Peak C Em wavelength in the S-S and L-W was 10-15 nm right-shifted more than the P-W Em wavelength.
The right shift of the Em wavelength represented a more conjugated structure and aromaticity in the L-W than P-W DOM, which can be attributed to the release and dissolution of the phenolic compounds from litter lignin in the S-S [bib_ref] The dynamic exchange of dissolved organic matter percolating through six diverse soils, Scott [/bib_ref]. The Peak C Em right shift was similar to that reported by Mostofa et al. [bib_ref] Dynamics and characteristics of fluorescent dissolved organic matter in the groundwater, river..., Mostofa [/bib_ref]. They were observed in the Yasu River watershed, Japan study, where the groundwater DOM (Ex/Em = 320 ± 9/425 ± 5 nm) and river water DOM (Ex/Em = 340 ± 5/432 ± 4 nm) were in contrast to the lysimeter results. The Yasu River water had a longer emission wavelength than the groundwater. The decrease in the Yasu River groundwater Em wavelength was attributed to the phenolic compounds of the river water DOM, which was adsorbed by soil minerals during river water infiltration. However, in the lysimeter study, the L-W DOM Peak C increased emission wavelength was attributed the SHS desorption of phenolic compounds that were dissolved into the L-W. The right shift of the Em wavelength represented a more conjugated structure and aromaticity in the L-W than P-W DOM, which can be attributed to the release and dissolution of the phenolic compounds from litter lignin in the S-S [bib_ref] The dynamic exchange of dissolved organic matter percolating through six diverse soils, Scott [/bib_ref]. The Peak C Em right shift was similar to that reported by Mostofa et al. [bib_ref] Dynamics and characteristics of fluorescent dissolved organic matter in the groundwater, river..., Mostofa [/bib_ref]. They were observed in the Yasu River watershed, Japan study, where the groundwater DOM (Ex/Em = 320 ± 9/425 ± 5 nm) and river water DOM (Ex/Em = 340 ± 5/432 ± 4 nm) were in contrast to the lysimeter results. The Yasu River water had a longer emission wavelength than the groundwater. The decrease in the Yasu River groundwater Em wavelength was attributed to the phenolic compounds of the river water DOM, which was adsorbed by soil minerals during river water infiltration. However, in the lysimeter study, the L-W DOM Peak C increased emission wavelength was attributed the SHS desorption of phenolic compounds that were dissolved into the L-W.
## Protein-(plf), fulvic-(flf) and humic-like (hlf) fluorescence
The DOM and S-S compositions are a heterogeneous and complex mixture. The 3D fluorescence EEM provides the composition and structure information of fluorescent molecules. Different wavelength regions represent different fluorescent substances [bib_ref] Characterization of dissolved organic matter in cave and spring waters using UV-Vis..., Birdwell [/bib_ref] [bib_ref] The impact of glacier runoff on the biodegradability and biochemical composition of..., Fellman [/bib_ref] [bib_ref] Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter, Chen [/bib_ref] [bib_ref] Higher reactivity of allochthonous vs. autochthonous DOC sources in a shallow lake, Catalán [/bib_ref] [bib_ref] Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy, Coble [/bib_ref]. The values of the three EEM fluorescent regions PLF, FLF and HLF [fig_ref] Figure 2: The excitation/emission matrix [/fig_ref] represent the fluorescence intensities of protein-, fulvic acidand humic acid-like substances, respectively [bib_ref] Fluorescence excitation-emission matrix characterization of some sewage-impacted rivers, Baker [/bib_ref] [bib_ref] Microbial transformation of dissolved leaf litter organic matter and its effects on..., Hur [/bib_ref]. Changes in fluorescence intensity and ratios of PLF, FLF and HLF were used to track the DOM sources and composition changes [bib_ref] Fluorescence excitation-emission matrix characterization of some sewage-impacted rivers, Baker [/bib_ref] [bib_ref] Spectroscopic characterization of dissolved organic matter and its interactions with metals in..., Park [/bib_ref] [bib_ref] Tracing groundwater flow and sources of organic carbon in sandstone aquifers using..., Lapworth [/bib_ref]. [fig_ref] Table 2: Disolved organic matter [/fig_ref] lists the total PFL, FLF and HLF fluorescent intensity of the three solutions as well as the OCS PFL, FLF and HLF (au/mg-C) fluorescent intensity. The total PLF, FLF and HLF fluorescent intensities were in the order of S-S > L-W > P-W, respectively. S-S contained the most abundant fluorescent substances, while the L-W DOM had a higher fluorescent intensity than the P-W DOM, which can be attributed to the fluorescent substances' dissolution from S-S. Compared with P-W DOM, the L-W DOM average fluorescent intensities increased 115%, 141% and 181% for total PLF, FLF and HLF, respectively. The increased rates of total FLF and HLF were greater than the rate of the total PLF. [fig_ref] Figure 3: The PLF, FLF and HLF percentages for P-W, L-W and S-S [/fig_ref] shows the fluorescent distribution fractions of PLF, FLF and HLF for P-W, L-W and S-S, respectively. The DOM and S-S had the highest FLF percentage and the lowest HLF percentage. The S-S had the lowest PLF percentage but had the highest HLF percentage. The L-W increased the HLF fraction compared to the P-W (p = 0.079), which was attributed to the dissolution of the HLF from S-S.
Sensors 2019, 19, x FOR PEER REVIEW slopes of the linear regression were 0.62, 1.74 and 1.79 between OCS FLF and OCS PLF for P-W, L-W and S-S, respectively. The slopes of the linear regression were 0.15, 0.58 and 0.64 between OCS HLF and OCS PLF for P-W, L-W and S-S, respectively. The linear slopes of L-W were close to the S-S slopes and both slopes were greater than the P-W slope. The tendency of the linear regression slopes was different from the average values due to the large variation of PLF in P-W .
The linear regression slopes revealed that the chemical composition and structure of L-W were similar to S-S and greater than P-W DOM. The regression results were consistent with the results of the optic characteristics. For instance, the L-W DOM increased in aromaticity, average molecular weight and terrestrial source but decreased in the autochthonous sources. Additionally, the OCS FLF and HLF fluorescence intensity significantly increased compared to the P-W DOM. However, the average ratios of FLF/PLF and HLF/PLF did not reflect that trend. Hence, it is appropriate to use the linear regression slope instead of the average value when data have a high variation such as in the lysimeter case. . Total and organic carbon standardization (OCS) (in parentheses) fluorescent intensity of percolation water (P-W), leachate water (L-W) and soil solution (S-S) for the EEM three regions protein-(PLF), fulvic-(FLF) and humic-like fluorescence (HLF). The magnitude order of OCS PFL and OCS FLF fluorescence intensities was L-W > P-W > S-S, but the magnitude order of OCS HFL fluorescence intensity was L-W > S-S = P-W . The OCS FLF and HLF fluorescence intensity of L-W was significantly greater than P-W (p < 0.001). Compared with the OCS P-W DOM, the OCS L-W DOM fluorescent intensity increased following the order HLF (81%) > FLF (55%) > PLF (44%). The total and OCS fluorescent intensities of L-W DOM were greater than P-W DOM for three fluorescent regions and increasing ratios of HLF and FLF were greater than PLF, which demonstrated that the L-W composition had higher humification substances than P-W. . Total and organic carbon standardization (OCS) (in parentheses) fluorescent intensity of percolation water (P-W), leachate water (L-W) and soil solution (S-S) for the EEM three regions protein-(PLF), fulvic-(FLF) and humic-like fluorescence (HLF). Lapworth et al. [bib_ref] Tracing groundwater flow and sources of organic carbon in sandstone aquifers using..., Lapworth [/bib_ref] reported that the FLF/PLF ratios of DOM for two groundwater aquifers were 0.33-2.0 and 2.5-5.0, respectively. These two groundwater DOM were affected by sheep waste sources. The other DOM was affected by a terrestrial and microbial source mixture. The agricultural animal waste had high a PLF intensity; therefore, the FLF/PLF ratio was lower than the source of the terrestrial and microbial source mixture. Naden et al. used FLF/PLF to distinguish DOM composition of different water bodies. The ratios of FLF/PLF of forest seepage, reservoir water and urban river DOM were 2.13, 0.90 and 1.45, respectively. The forest soil seepage contained more phenolic compounds due to soil dissolution, thus, the FLF/PLF ratio was higher than the other two DOM. The reservoir water had freshly produced DOM by algae and aquatic organisms and had the lowest FLF/PLF ratio. Baker [bib_ref] Fluorescence excitation-emission matrix characterization of some sewage-impacted rivers, Baker [/bib_ref] explored the FLF/PLF ratios of different river sections ranging from 0.65 to 3.3. The ratios of upstream and small stream were 1.0-3.33 (one exception 0.65), and the downstream ratios of a major river section were 0.83-1.43. The low ratios of the downstream section implied it received STP effluent with a high concentration of PLF, which reduced the FLF/PLF ratio. [fig_ref] Figure 4: The OCS FLF/PLF and OCS HLF/PLF ratios for P-W, L-W and S-S [/fig_ref] shows the average ratios of FLF/PLF were 1.71 ± 0.78, 1.50 ± 0.18 and 1.96 ± 0.18 for P-W, L-W and S-S, respectively. The ratio of S-S was significantly greater than L-W and P-W (p < 0.001). The average ratios of HLF/PLF were 0.50 ± 0.24, 0.51 ± 0.07 and 0.90 ± 0.10 for P-W, L-W and S-S, respectively. The ratio of S-S was significantly greater than L-W and P-W (p < 0.001). The ratios of FLF/PLF and HLF/PLF between L-W and P-W were insignificantly different (p = 0.48 and p = 0.96, respectively). However, the coefficients of variance (VC) were 46% and 47% for the FLF/PLF and HLF/PLF ratios of P-W, respectively. The high VC values may not reflect the overall trend of the ratios. The average values were further compared to the slopes of the linear regression. [fig_ref] Figure 5: The linear regression of OCS FLF to OCS PLF [/fig_ref] ,b shows the linear regression between OCS FLF and OCS PLF as well as between OCS HLF and OCS PLF. The results showed that P-W, L-W and S-S had a significant linear relationship (except P-W OCS HLF and PLF linear regression had p = 0.046, the remaining five linear regressions had p < 0.01). The slopes of the linear regression were 0.62, 1.74 and 1.79 between OCS FLF and OCS PLF for P-W, L-W and S-S, respectively. The slopes of the linear regression were 0.15, 0.58 and 0.64 between OCS HLF and OCS PLF for P-W, L-W and S-S, respectively. The linear slopes of L-W were close to the S-S slopes and both slopes were greater than the P-W slope. The tendency of the linear regression slopes was different from the average values due to the large variation of PLF in P-W .
## Plf (au) flf (au) hlf (au)
## Plf (au) flf (au) hlf (au)
## Percolation mechanisms in the lysimeter
The UV-Vis and fluorescence index results showed that the S-S had the highest extent of humification in the L-W and P-W DOM (the HLF fluorescence intensity and highest ratios of FLF/PLF and HLF/PLF). In addition, S-S had a slightly higher aromaticity, average molecular weight, terrestrial origin and allochthonous origin than the P-W and L-W DOM. In comparison with the optical characteristics of P-W DOM, L-W DOM increased DOC concentration, aromaticity (SUVA254), average molecular weight (S275-295) and terrestrial source contribution (FI decreases) but decreased the autochthonous DOM (BIX decreased). The Em wavelength of L-W DOM Peak C was 10 nm rightshifted compared to the P-W DOM Peak C. In addition, the OCS FLF and HLF fluorescence intensity of L-W DOM was significantly greater than the P-W DOM intensity.
Generally, comparing surface water, groundwater DOC concentration decreased and the DOM microbial source increased [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] Riverine export of aged carbon driven by flow path depth and residence..., Barnes [/bib_ref]. The trend was in contrast to the L-W and P-W DOM optical characteristics in the lysimeter study. The hypothesis of deep vertical infiltration of surface water DOM was affected by the adsorption of soil minerals (aluminosilcate clay minerals and metal oxides/hydroxides in soil) and retained specific DOM functional groups [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] The dynamic exchange of dissolved organic matter percolating through six diverse soils, Scott [/bib_ref] [bib_ref] Riverine export of aged carbon driven by flow path depth and residence..., Barnes [/bib_ref]. DOM hydrophobicity component was partitioned into soil organic matter. Furthermore, the biodegradation of protein-like DOM by soil microorganisms produced the microbial metabolism. DOM hydrophobicity substances stayed on the mineral surface and were partitioned into soil organic matter, and the undecomposed protein-like DOM and soluble microbial byproducts infiltrated into the deep underground [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] Riverine export of aged carbon driven by flow path depth and residence..., Barnes [/bib_ref]. The linear regression slopes revealed that the chemical composition and structure of L-W were similar to S-S and greater than P-W DOM. The regression results were consistent with the results of the optic characteristics. For instance, the L-W DOM increased in aromaticity, average molecular weight and terrestrial source but decreased in the autochthonous sources. Additionally, the OCS FLF and HLF fluorescence intensity significantly increased compared to the P-W DOM. However, the average ratios of FLF/PLF and HLF/PLF did not reflect that trend. Hence, it is appropriate to use the linear regression slope instead of the average value when data have a high variation such as in the lysimeter case.
## Percolation mechanisms in the lysimeter
The UV-Vis and fluorescence index results showed that the S-S had the highest extent of humification in the L-W and P-W DOM (the HLF fluorescence intensity and highest ratios of FLF/PLF and HLF/PLF). In addition, S-S had a slightly higher aromaticity, average molecular weight, terrestrial origin and allochthonous origin than the P-W and L-W DOM. In comparison with the optical characteristics of P-W DOM, L-W DOM increased DOC concentration, aromaticity (SUVA 254 ), average molecular weight (S 275-295 ) and terrestrial source contribution (FI decreases) but decreased the autochthonous DOM (BIX decreased). The Em wavelength of L-W DOM Peak C was 10 nm right-shifted compared to the P-W DOM Peak C. In addition, the OCS FLF and HLF fluorescence intensity of L-W DOM was significantly greater than the P-W DOM intensity.
Generally, comparing surface water, groundwater DOC concentration decreased and the DOM microbial source increased [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] Riverine export of aged carbon driven by flow path depth and residence..., Barnes [/bib_ref]. The trend was in contrast to the L-W and P-W DOM optical characteristics in the lysimeter study. The hypothesis of deep vertical infiltration of surface water DOM was affected by the adsorption of soil minerals (aluminosilcate clay minerals and metal oxides/hydroxides in soil) and retained specific DOM functional groups [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] The dynamic exchange of dissolved organic matter percolating through six diverse soils, Scott [/bib_ref] [bib_ref] Riverine export of aged carbon driven by flow path depth and residence..., Barnes [/bib_ref]. DOM hydrophobicity component was partitioned into soil organic matter. Furthermore, the biodegradation of protein-like DOM by soil microorganisms produced the microbial metabolism. DOM hydrophobicity substances stayed on the mineral surface and were partitioned into soil organic matter, and the undecomposed protein-like DOM and soluble microbial byproducts infiltrated into the deep underground [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] Riverine export of aged carbon driven by flow path depth and residence..., Barnes [/bib_ref].
In this study, the increased DOC concentration and change in chemical composition and structure of the L-W DOM were attributed to the desorption of the S-S component during the P-W infiltration process. The results were in contrast to the results from the infiltration DOM into the deep groundwater because in the lysimeter test, percolation water was a shallow, fast and short-flowing lateral flow. The infiltration water DOM had a short contact time with the soil, and in the percolation process, the lignin phenolic compounds in the soil litter layer were desorbed and dissolved in the infiltration water DOM. With a short retention time, the desorption of S-S was superior to the DOM decomposition by microorganisms, which resulted in a higher DOC concentration of L-W than P-W. Rapid flow of infiltration water reduced soil adsorption/co-precipitation and metabolic breakdown by microorganisms, resulting in litter-derived DOM transportation. The L-W DOM Em wavelength right-shifted, showing that the L-W DOM had a more conjugated structure and aromatics, which attributed to S-S terrestrial and litter lignin phenolic compounds and dissolved humic acid-like substances [bib_ref] Cycling downwards-Dissolved organic matter in soils, Kaiser [/bib_ref] [bib_ref] Composition of dissolved organic matter in groundwater, Longnecker [/bib_ref]. The L-W OCS PLF, FLF and HLF fluorescence intensity increased, and the increase rate of fluorescence intensity was HLF > FLF > PLF. The average ratios of FLF/PLF and HLF/PLF showed that the P-W and L-W ratios were similar; however, the L-W linear regression slopes were similar to the S-S slopes and greater than the P-W slopes. These optical characteristics suggested that the shallow infiltration lateral seepage L-W DOM was affected by the S-S desorption mechanism. Since the soil depth of the lysimeter was 1.0 m and the collected drainage depth was 0.3 and 0.6 m each, it was a shallow lateral flow. The optical characteristics showed that the L-W DOM chemical composition and structural changes were caused by desorption of S-S. There is a significant difference between the shallow lateral flow and the deep vertical infiltration of groundwater DOM and S-S.
# Conclusions
In this study, optical characteristics were used to investigate the chemical composition and structure of the DOM in the shallow infiltration lateral flow affected by soil S-S. The S-S had higher aromaticity, average molecular weight and terrestrial sources than the DOM. The L-W DOM had higher aromatic abundance (SUVA 254 ) and average molecular weight (S 275-295 ) than the P-W DOM. In addition, the L-W DOM had more terrestrial (Fl) but less autochthonous content (BIX) than the P-W DOM. The change in the chemical composition and structure in the L-W DOM was due to the desorption of S-S compounds during the infiltration process. The average ratios of FLF/PLF and HLF/PLF were insignificantly different between the P-W and L-W DOMs. However, linear regression showed that the slopes of the L-W DOM were much larger than the P-W DOM slopes, and the slopes of the L-W DOM were similar to the slopes of the S-S. The DOM composition and structure resulted from shallow P-W infiltration, which was different from the deep groundwater infiltration. The change in the L-W DOM in the shallow soil infiltration was substantial and was significantly affected by the S-S desorption.
# Funding:
We sincerely thank Taiwan Chia-Nan Irrigation Association for financial support.
[fig] Figure 1: The configuration and placement of the experimental lysimeter. [/fig]
[fig] Figure 2: The excitation/emission matrix (EEM) plots of P-W, L-W and S-S, respectively. (a) P-W, (b) L-W, (c) S-S. [/fig]
[fig] Figure 3: The PLF, FLF and HLF percentages for P-W, L-W and S-S. [/fig]
[fig] Figure 4: The OCS FLF/PLF and OCS HLF/PLF ratios for P-W, L-W and S-S. [/fig]
[fig] Figure 5: The linear regression of OCS FLF to OCS PLF (a) and OCS HLF to OCS PLF (b). [/fig]
[fig] Author: Contributions: Methodology and software: Y.-L.Y. and T.-P.C.; Investigation and analysis: W.-S.H. and T.-P.C.; Writing-Original draft preparation, review, and editing: Y.-L.Y. and T.-C.C. [/fig]
[table] Table 1: Water quality of P-W and L-W and soil basic properties a . * Did not measure; a sand = 45.8%, silt = 33.1% and clay = 21.1% (loam soil) [/table]
[table] Table 2: Disolved organic matter (DOC) concentration and optical indices of P-W, L-W and S-S solutions. [/table]
|