S100A Family Proteins and Mesothelin in Tear Proteome Are Associated With Retinal Vein Occlusion

Tear samples were collected from 88 subjects and analyzed using absolute quantitative and comparative proteomic approach. We found a large proportion (505 proteins) of tear proteome between healthy donors and subjects with retinal vein occlusion (RVO). Comparative proteomic analysis revealed 30 proteins (p<0.05) signicantly differed in their quantitative property. Among them S100A6 (3.7 fmoles/ng, p<0.001), S100A8 (0.68 fmoles/ng, p<0.001), and S100A9 (2.06 fmoles/ng, p<0.001) are the most overrepresented proteins. Mesothelin was found as tear-specic protein with signicant increase (1.08 fmoles/ng versus 0.54 fmoles/ng in the control, p<0.001) in the RVO group. The selected altered proteins were combined to reconstruct the customized map of protein-protein interactions with the burden of quantitating property and the context of RVO-related association. The customized interactions map (FDR<0.01) emerged inammation and impartment of retinal hemostasis as the main RVO-associated processes. The semantic analysis of customized map encouraged the prevalence of core biological processes encompassing dysregulation of mitochondrial organization and utilization of topologically incorrect folded proteins as a consequence of oxidative stress and inammation caused by the retinal ischemic condition. Signicantly differed proteins (S100A6, S100A8, S100A9, MSL, B2M) were applied for the ROC plotting with AUC varied from 0.772 to 0.952 suggesting their association with the CRVO.


Introduction
Retinal vein occlusion (RVO) is a frequent ocular disease that is fraught with a risk of blindness or severe decrease of visual acuity since the entire retina is affected especially in older patients 1,2 . While the progression of RVO, the retinal is likely to suffer ischemic condition upstream of the occlusion, entailing to the increase of in ammatory proteins 3 .
Tear uid proteomics assays con rmed the correlation between chronic eye diseases and clinical effects of aging. Turning to the panoply of results, it has been hypothesized that the tear proteome is featured by the age-associated gradual increase of in ammatory proteins and immune response cytokines. Several cytokines, including interleukin-6 and VEGF (vascular endothelial growth factor), are known to be associated with RVO but still, it contributes little in understanding of RVO pathology. The objection is limited to the high interindividual variability of cytokines and sensitivity to the epigenetic background 4,5 . A comparative tear proteomic study of 30 RVO patients with age-related macular degeneration demonstrated signi cant enrichment by complement factors (C3 and C9), clusterin (CLU), S100A9, and S100A8 proteins 6 .
A de ciency of glucose-6-phosphate dehydrogenase (G6PD) serum level has been correlated with a risk of CRVO due to the increased vulnerability to oxidative stress 7 . The S100A12 has also been considered as a contributor to the in ammation response at the ischemic condition of RVO. At least in the animal model of RVO, a ve-fold increase of S100A12 has been found in retinal ganglion cells and their dendritic processes and con rmed by the Western blot analysis 8 . A relative abundance of S100A family proteins suggests an indication of the ongoing in ammation following the RVO and the consequent ischemic condition 9 . However, neither of these studies yielded a rational understanding of the molecular basis of CRVO.
Another side of eye-related proteomic investigation is focused on the mitochondria-associated proteins. A study of proteome after laser-induced retinal artery occlusion (RAO) relevant to clinical manifestation gured out an increased abundance of Bax protein, and simultaneous elevation of cytosolic cytochrome c and caspase-9 suggesting a severe and rapid escalation of oxidative stress accompanied by mitochondrial dysfunction and pro-apoptotic events 10 . Dysfunction of mitochondrial ATP generation was found in the proteome of retinal pigment epithelium obtained from patients with age-related macular degeneration, and in in vitro models of human embryonic stem cells [11][12][13] . Notwithstanding, dysregulation of mitochondria functioning re ects the consequence of RVO but does not determine the root of pathology.
Proteomic study of eye diseases is a challenging task low stability of such biomaterial. While modern approaches can achieve a size of proteome accounting up to 5,000 proteins, this information can be not highly promising and carry no information regarding the discovery of biomarkers since routine proteomic assay requires a large dataset to improve results reproducibility. One of the critical issues is samples collection and storage standardization requirement; another challenge is the high sensitivity of proteomics to individual variability. The latter causes poor identi cation of well-known RVO associated markers (such as ICAM1), whereas other proteins of particular interest are mostly characterized by in ammatory and immune-response activities 14 .
This study purposed to nd the answer to whether is it possible to distinguish patients with CRVO by their proteome, and whether is it possible to pick up certain key proteins that clue the treatment strategy or disease onset? To make a better understanding of molecular pathways associated with CRVO, we conducted a quantitative proteomic measurement and compiled it with the reconstruction of the diseasetargeted protein-protein interactions mapping.

Results
The proteomic analysis discovered 710 and 682 protein identi cations on the control and examined CRVO group, respectively, thus the proteome dimensionality between two groups is equilibrium condition. The shared part of the proteome comprised of 505 protein identi cations, or more than 70% of individual proteome size. Despite the proportional size of aggregated proteomes within groups, the distribution of individual proteomes was unequal and made 275 proteins (median) in the control group and 154 proteins (median) in the CRVO group ( Figure 1A). At the same time, group-speci c proteins were found to be the least frequent and consisted of 0.37 (median value) and 0.15 amongst subjects in the control and CRVO groups, correspondingly ( Figure 1B) while the median frequency of protein in the aggregated proteome consisted 0.5 and 0.23 in the control and CRVO group, respectively. Quantitative analysis demonstrated a similar distribution of measured proteins in both studied groups. It was found, that proteins are distributed along with four orders of magnitude covering the dynamic range of concentrations between 0.01 and 100 fmoles/ng. The majority of proteins were attributed to mediumcopied (between 0.1 and 10 fmoles/ng), whereas a small portion was divorced between low-copied (0.01-0.1 fmoles/ng) and high-copied (more than 10 fmoles/mL) in both studied groups (Figure 2A). The least abundant protein was found TMSB10 (0.0132 fmoles/mL) in the CRVO group and ITPR1 (0.0339 fmoles/mL) in the control group; the most abundant proteins were CDC14C (158 fmoles/mL) and LTF (108 fmoles/mL) in the OCRV and control groups, respectively.
The obtained result of concentration-dependent proteins normal distribution ( Figure 2A) and similar proteomes dimensionality suggests uniform in-depth discovery and proportional quantitative loading of proteins between studied groups.
The principal component analysis did not demonstrate a satis ed separation between the examined group of patients and the control group. The groups under consideration almost completely overlapped with a variance of PC1 made 12.84% and PC2 made 8.4% ( Figure 2C).
To unveil the difference between groups of study, we selected 137 mutual proteins with a frequency exceeding 0.6 in both control and CRVO groups. Of them, only 31 proteins passed the signi cance cut-off (p <0.05), and this fraction comprised of 9 down-regulated (including three Ig-heavy and two Ig-light chains) and 22 up-regulated proteins in the CRVO patients compared to healthy donors (Table 2 and Figure 2B). Biological processes analysis (signi cance cut-off p <0.001) showed ( Figure 3) that the majority of proteins are not highly-specialized and covers the regulation of cellular response to misfolded proteins (p =2.31e-04), cellular response to topologically incorrect protein (p =4.77e-04), and positive regulation of mitochondrion organization (p =9.91e-04). Molecular functions mapping a rmed the prevalence of cell stress-responsive proteins since the main proportion of proteins are characterized by ubiquitin (p =6.96e-06) and ubiquitin-like ligase binding activity (p =1.49e-04). Table 1 Quantitative assessment of the shared part of proteome (p <0.05). Due to the frequency percolation (more than 0.6 value within each group) has been applied prior to the selection of signi cant proteins, the estimated concentration is designed as a median value with an inter quartile (q1-q3) range (IQR). The protein concentration is estimated as an amount of certain protein (in fmoles) per one ng of total protein on-column loading. The total amount of protein on-column loading was 3 (three) µg per a sample.  tissue distribution and multifunctional attributes evidence that the large proportion of signi cantly differed proteins are poorly specialized, but may play a pivotal role in different biological processes, particularly, in immune response and hemostasis maintenance.
We conducted protein-protein interaction analysis in support of Cytoscape using the STRING-based and disease-based model to reconstruct a proteins network tightly associated with the retinal vein occlusion as a target pathology. For this purpose, the nal network was percolated to eliminate missed interaction and those proteins falling low molecular specialization. Eventually, only proteins with a high con dence score (at least 0.7), falling the criteria of implication in retinal vein occlusion (onset or progression) and being attributed to eye tissue, tear gland, or tear were selected to build the nal proposed network (Figure 4).
Despite the small proportion of the examined proteins, almost a complete set (Table 1) participated in the network reconstruction with high con dence (more than 0.7 con dence score). The designed network can be divided into four large clusters, of which three are with tissue-speci c attributes (eye, tear, and tear gland) and one general cluster with multifunctional proteins but implicated in occlusion progression ( Figure 4). The latter includes proteins of extracellular matrix, blood, peroxisomes, and proteins involved in metabolic modulation. However, proteins of tissue-speci c clusters are produced by meibomian glands, lacrimal glands, and conjunctival goblet cells and might indicate age-or disease-related changes in the tear proteome.
To determine the signi cance of proteins implicated in oxidative stress and immune response escalation, we obtained AUC measures using the ROC plot. We selected ve proteins (S100A6, S100A8, S100A9, MSLN, and B2M) that meet the criteria of closest association with CRVO and ischemic condition and featured with the most signi cant alteration between the control group and subjects with CRVO ( Figure 5).
The obtained results demonstrated satis ed AUC for each selected protein. The AUC measure varied from the least estimated to S100A6 (AUC=0.772) to a maximum value for S100A9 (AUC=0.952). It should be noticed, the most contributing proteins (S100A8, S100A9, and B2) belong to secreted proteins of the immune response and matrix-mediated regulation of cell migration, apoptosis regulation, and MHC-I representation. Despite the great speci city and sensitivity of the ROC curve ( Figure 5), we avoid combining these proteins into the biomarkers pattern since much more examination is required to establish the proper role of the pattern in CRVO pathogenesis.

Discussion
Besides age-related changes in the proteome, retinal vein occlusion (RVO) is accompanied by speci c alterations in the matrix and hemostasis-associated proteins. The pivotal role of monitored changes is impaired blood out ow, so the retina suffers ischemic conditions upstream of the occlusion 15  were not reported previously as RVO-associated, but most of them are matrix-related (including laminin subunits, integrin, and actinin isoforms). Based on this evidence, the authors concluded that RVO is tightly associated with the matrix remodeling process as well as integrin and focal adhesion signaling 18 .
The study of animal laser-induced RVO revealed, that it could be distinguished from the control samples based on the proteomic data. The main input into discrimination made pro-in ammatory proteins, of which S100A12 was the most profoundly increased 8 . Our data also suggests that ischemic conditions in RVO patients might be associated with in ammation and mediated by immune-related proteins since S100A family proteins and B2M (β2-microglobulin) ( Table 1) are signi cantly up-regulated in the RVO patients. These proteins construct a separate cluster on the protein-protein interaction map (S100A6, S100A8, and S100A9; Figure 4) and are tightly joined with matrix-related (ACTB) and metal-binding proteins (TF). Previous comparative studies of CRVO and non-ischemic ocular disease patients revealed that proin ammatory proteins level directly correlates with the severity of retinal ischemia 19 . If S100A6 is a prolactin receptor-associated protein mediating glucocorticoids secretion, while S100A9 and S100A8 both are regulators of cells migration acting in the defending mechanisms against in ammation, oxidative stress, and innate immune stimulation, one may conclude that these proteins indicate local in ammation as a consequence of CRVO ischemic condition.
The S100 family proteins were rigorously revied in the context of eye disease as promising markers patterns with a high predictive potency and disease monitoring value 20 . These proteins have been repeatedly screened in the tear and showed an ability to discriminate between dry eye disease and CRVO due to controversial regulation (upregulated in CRVO) 8, 21,22 . These secreted proteins are typically well correlated with a signi cantly suppressed MHC-I peptide representation B2M protein (Table 1) and extracellular matrix proteins abundantly observed in our study, including PRR4, ACTB, MSLN, and CST3 (Table 1) indicating active immune response regarding chemokines activation. Interestingly, but it should be noti ed, that MSLN (mesothelin) has not been ever reported in tear proteome at either eye-related disease.
The protein is generally mentioned as a target in the chemotherapy of different oncophenotypes, and mostly regarding solid tumors 23 .
The nding of MSLN in the tear was unexpected, however, if consider the beguiling idea of the increased level of extracellular matrix compounds as a biomarker for CRVO and glaucoma [24][25][26] , it seems possible the presence of MSLN due to its importance in cell adhesion. In consistence with our observation, ischemic condition at CRVO induce the expression of ocular-speci c pro-brotic TGF-β2 factor, that su ciently promotes the growth of extracellular matrix 26,27 . Therefore, MSLN was established among up-regulated proteins in this study (Table 1). However, it should be admitted the matrix proteins are highly sensitive, so their alteration is generally a consequence of oxidative stress and in ammation, whereas the primary sign is depicted in hemostasis and mitochondria maintenance dysregulation. It has been reported, that patients with glaucoma and CRVO are characterized by irregular levels of GPX, SOD, and MDA, and typically present a higher level of these proteins even after surgery 28,29 .
The association of oxidative stress and macular degeneration with damaging and dysfunction of mitochondrial 30 . Moreover, with aging mitochondria become progressively more incomplete, thus CRVO can be considered as an age-related disease with a pivotal role in increasing ROS output 31 . In this study, the oxidative stress consequence is re ected in the decrease of mitochondria and mitochondria inner membrane organization processes (Figure 3), possibly indicating the prevalence of apoptosis and celldamaging. In oppose, processes related to cell response of topologically incorrect proteins and chaperon-cofactor proteins refolding and de novo proteins folding are substantially up-regulated in the proteome of patients with CRVO ( Figure 3) along with chaperon proteins (Table 1).
While all retinal cells rely on ATP as a fuel source and the photoreceptors are the largest consumers requiring regular mitochondria functioning, hypothesized, that such cautioning processes (Figure 3) underline consequences of the impaired retinal hemostasis and accompanied tissue-damaging caused by the increased pro-in ammatory activity. Furthermore, stimulated mitochondrial oxidative phosphorylation contributes signi cantly to the increased ATP and, consequently, ROS generation and may cause protein folding and refolding processes in the endoplasmic reticulum 32 and stimulates the production of in ammatory cytokines 33 . Notwithstanding, only a few reports can be found regarding the relation of oxidative stress with CRVO 28 , but much more reports review the relation with diabetic retinopathy [34][35][36] .
If considering oxidative stress as the paramount background event of CRVO, it is essential to unfurl the role of ROS scavengers and the related transport system as a possible victim of the boosted ROS generation and cytokines production. We found the local increase of transthyretin (Table1) also known as transthyretin (TTR) in the proteome of CRVO patients with a 0.88 frequency. The protein is mainly synthesized in the liver and is also found to be expressed in retinal pigment epithelial cells 37 . Apart from the pivotal role of TTR in the transport of thyroxine to the brain, circulating TTR also carries a retinol-binding protein (RBP) transporting and stabilizing retinol as an essential antioxidant agent. Due to high toxic potency, retinol is primarily stored in form of fatty acid esters. Vitamin A can e ciently neutralize thiol-radicals and stabilizes peroxide radicals being an important epigenetic factor preventing and reducing the risk of oxidative stress and in ammatory-related diseases 38,39 . The de ciency of TTR may aggravate the risk of oxidative stress and enhance local impairment of circulation and angiogenic processes, which is particularly established for cardiovascular and nondegenerative diseases 40 . On the other side of this matter, excessive circulating TTR provides a favorable environment for the transthyretin amyloidosis being involved in numerous ophthalmological pathologies 37 .
Despite most of the studies intend to assemble proteins in a panel of putative biomarkers, we are reluctant to build the decision-making pattern of CRVO biomarkers. Instead, we focused on the ve most signi cantly altered proteins credibly associated with the pathogenesis of CRVO and capable of indicating the ongoing in ammation caused by the ischemic condition. We submitted S100A6, S100A8, S100A9, MSLN, and B2M proteins to estimate the area under the curve ( Figure 5). The obtained measures of ROC for each protein satis es speci city and sensitivity expectations. Although some of these proteins are repeatedly mentioned in the context of CRVO severity and onset, their combination might be a more proper item for consideration to gure out the very beginning of CRVO and to monitor the bases of ocular diseases treatment.

Conclusion
Tear proteome results have been reported in this study. Although we found a large proportion (505 proteins) of proteome shared between studied groups, results of the statistical analysis demonstrated no signi cant difference between the control group and patients with CRVO on the level of quantitative loading. In total, a cohort of 30 proteins has been attributed to up-or down-regulated in the CRVO group mostly providing the information about immune response, in ammatory and retinal hemostasis impairments processes. The profound decrease of B2M and immediate increase of S100A family proteins with the abundance of matrixassociated secreted proteins underlines the in uence of local oxidative stress on cell migration and cytokines stimulated in ammatory reaction as a consequence of the progressing ischemic condition. In this study, we identi ed mesothelin (MSLN), which has never been reported in tear proteome before but mostly touched in regard to targeted therapy of solid tumors. We assumed the attribution of MSLN to matrix proteins set due to its importance in cell adhesion regulation and stimulated by the ocular-speci c probrotic TGF-β2 factor.
Mitochondria maintenance proteins are the most sensitive indicators in response to escalated oxidative stress. It has been determined that processes associated with mitochondria biogenesis and mitochondria inner membrane formation (as a mitochondria repopulation) were signi cantly suppressed in the tear proteome of CRVO patients, whereas processes associated with proteins metabolism (de novo folding, chaperon cofactor-depended refolding, cellular response to topologically incorrect proteins) were prevalent in the biological processes' hierarchy. Besides, we have established an increased level of TTR in the tear. We suggested a de cient level of retinol-binding protein as a possible reason for the increased circulating TTR.
Furthermore, due to TTR is generally increased in oxidative stress conditions (caused by the ischemic condition in certain cases), we supposed that the excess of TTR aggravated by the de cient of retinolbinding protein can be a risk factor of transthyretin amyloidosis as a possible dire consequence of the impaired retinal hemostasis and occlusion.
In summary, we assume that there are several promising proteins capable to highlight new insight into understating and evaluating the CRVO pathology. But the utility of these proteins in a part or whole is questionable until multiple examinations are being provided. and no personal data can be disclosed with this paper. The detail of clinical records is presented in Table 2.

Samples collection and preparation
To minimize the ocular surface irritation, tears samples were obtained from the inferior temporal tear meniscus without anesthesia and collected using the calibrated volumetric glass microcapillary (10-µl nominal calibrated volume; Blaubrand, Germany). Samples were stored at −80°C and transported in dry ice for proteomic analysis. Totally, up to 100 µL of each tear sample was collected.
The Samples were separated using an Acquity H-Class UPLC system (Waters, the UK) and loaded in a volume of 3 µL (totally 3 µg of protein fraction on-column loading) onto an Acquity™ UPLC BEH C18 (2.1 × 50 mm, 1.7 µm particle size; Waters, the UK) column heated to 50°C with the pre-installed in-line 0.2 µm lter. at a ow rate of 0.3 mL/min. Peptides were separated in a gradient of mobile phase A (water) and mobile phase B (acetonitrile) both supplied with 0.1% formic acid and 0.015% tri uoracetic acid using the following

Statistical data analysis
Proteins that meet the criterion of unicity by the presence of at least one unique (proteotypic) peptide apart of isoform-speci c peptides, were extracted for the statistical analysis following the proteins identi cation search. The identi ed group-speci c proteomes were assessed for similarity and the shared part of proteome was extracted for the principal component analysis (PCA). Protein intensity was estimated as the normalized summed peptide intensities belonged to the certain protein and the resulting matrix of intensities was used for the quanti cation based on the calculation of intensity medians for each protein within studied groups. Wilcoxon test with a threshold of p < 0.05 was applied to groups of study to revealed outliers and signi cant differences in a quantitative loading. A measure of protein abundance was represented as a median value fold changes (FC) ratio toward the control group and calculated based on the absolute concentration sampled from the UPS-2-based quantitative analysis. Proteins with a frequency of at least 0.6 within each studied group, and beyond the log-scaled fold change cut-off of FC >1 or FC <−1 at a signi cance p-values of less than 0.05 (pairwise t-test test) were considered as signi cant in the quantitative property. Signi cantly altered proteins were submitted for functional and pathways analysis at a q-value threshold less than q < 0.01 using the Cytoscape (  Distribution of proteins quantitative property within the combined proteome of control and CRVO groups. The largest proportion (up to 78%) is covered by medium-copied proteins between 0.1-10 fmoles/ng. The rest fraction can be split between low-copied (0.01-0.1 fmoles/ng) proteins (up to 8%) and high-copied (more than 10 fmoles/ng) proteins accounted up to 14% of the tear proteome size (A). Volcano-plot scattering of 31 signi cantly (p <0.05, Wilcoxon test) differed proteins (22 upregulated (log FC > 1) and 9 downregulated (log FC < -1) proteins) between the control group and CRV subjects (B). Principal

Page 21/24
components analysis revealed no signi cant difference between groups of study with the contribution of PC1=12.84% and PC2=8.4% (C).

Figure 3
Distribution and prevalence of the GO biological processes terms ( ltered at p <0.001) associated with the set of signi cantly differed proteins found between the control and CRVO groups. The semantic analysis is based on the Resnik algorithm and measures the similarity between the enriched ontology terms after the elimination of redundant terms. The de ned enrichment measure (circles size) de nes relative abundance and the direction of a certain biological process within the study group (CRVO) relative to the control groups. The majority of processes attributed to cell cycle and mitochondria organization are signi cantly depleted and abolished, while the regulation of improperly folded proteins and stress response reaction are organized in the enhanced acting cluster.

Figure 4
The reconstructed map of protein-protein interactions between the signi cantly differed proteins. The map is reconstructed in the context of retinal vein occlusion pathogenesis, tissue (source) speci city (such as tear gland, tear, and eye), and attributed with a quantitative loading of the measured proteins (color-scaled bar). The node height level indicates the con dence of proteins interaction and normalized to eye tissue speci city. Circles color indicated the prevalent biological processes contributing to the CRVO progression and taken by mapped proteins (vesicle-mediated transport (magenta); defense response (cyan); retina hemostasis (green), regulation of immune system (orange), cell migration, and extracellular matrix organization (blue)).