Proteomics analysis of human tears from aqueous-deficient and evaporative dry eye patients

Despite the high global prevalence of dry eye syndrome (DES), the fundamental processes underlying this pathology remain largely unexplored. Therefore, this study endeavoured to investigate in-depth the tear proteome of DES patients employing the mass spectrometry (MS)-based proteomic strategies. Eighty patients were recruited and subdivided into three major DES subgroups, which are the aqueous-deficient (DRYaq), evaporative (DRYlip) and a combination of the two (DRYaqlip), as well as healthy subjects (CTRL). Discovery proteomics strategy was employed to identify large number of significantly differentially expressed tear proteins in DRYlip vs. CTRL, DRYaq vs. CTRL and DRYaqlip vs. CTRL with 22, 58 and 67 proteins, respectively. Biological functional analysis demonstrated for the first time that various metabolic processes were highly expressed in DRYaq and DRYaqlip, which might modulate various other known processes, especially the inflammatory and immune processes. Targeted proteomics strategy verified that 13 major proteins were differentially expressed in specific DES subgroups, comprising of PRR4, ZG16B, SCGB2A1, DMBT1, PROL1, LACRT, ALDH3A1, ENO1, TF, S100A8, S100A9, PEBP1 and ORM1. In conclusion, this study had explored in-depth the pathology of DES by unravelling various new fundamental processes and the major proteins responsible for the maintenance of tear film stability.

(LCN1) and prolactin-inducible protein (PIP) across all DES sub-groups, their finding demonstrated a number of proteins that were significantly differentially expressed in subgroups of DES. Other DES studies have investigated Sjögren's syndrome (SS) dry eye, which is a systemic autoimmune disease characterized by compromised lacrimal and salivary gland functions that causes severe dry eye and dry mouth 6,25 .
The aforementioned DES studies only reported limited number of differentially expressed proteins for the pathology, largely due to the limitations in the methodology employed. These studies had employed various mass spectrometry (MS)-based proteomic strategies, which have become a progressively powerful technology for the identification and quantifications of tear proteins in different ocular pathologies 5,20,21,23,[26][27][28][29][30][31][32][33][34][35] . Recently, protein quantification with isotopic labels (e.g. iTRAQ) has been commonly utilized for tear protein studies due to their quantitative accuracy, coverage and robustness 23,24 . However, despite their usefulness, they inherently require extra preparation steps that consequently increase the complexity and often prohibit their use for biological questions that require the detection of subtle yet vital proteome changes, mostly the low abundant tear proteins. On the contrary, label-free quantification (LFQ) is the simplest and most economical strategy that can be utilized for in-depth analysis of the differentially expressed proteins in large sample size with high precision 36 . Additionally, the development and improvement of the existing algorithms and software tools, especially, MaxQuant software suite, have provided an effective solution for LFQ analysis, one that achieves state-of-the-art quantification accuracy and coverage 37,38 .
Therefore, in-depth investigation of the differentially expressed proteins in tear proteome of DES patients will potentially contribute to better understanding of the fundamental processes of this pathology. Hence, the main objective of the present study was to identify the differentially expressed proteins in the major DES subgroups, which are responsible for the maintenance of tear film stability. The tear proteome profiles of DES subgroups were analysed employing the bottom-up label-free MS-based proteomics strategies. The ultimate outcome of this study is envisaged to unravel the intricate regulation profiles of specific proteins responsible for the maintenance of tear film stability and pin-point specific alteration(s) in the different DES subgroups.

Results
Identification of the differentially expressed tear proteins from DES patients. The representative tear protein profiles of DES subgroups and CTRL resolved in 1DE gel are illustrated in Fig. 1a. Notable differences were observed at band three, four, seven and ten, which represent serum albumin (ALB), Ig alpha-1 chain C (IGHA1), PRR4 and mammaglobin-B (SCGB2A1), respectively. A total of 200 proteins were detected by the discovery approach as presented in Supplementary data 1A. In order to reveal the differentially expressed proteins in the DES subgroups compared to CTRL, LFQ values of the identified proteins extracted from MaxQuant analysis were used for statistical analysis utilizing the Perseus software. Basically, there is good technical reproducibility of the entire experimental workflow, which reveals Pearson correlation of 0.96 ± 0.01 for DRYaq and 0.97 ± 0.01 for CTRL, DRYlip and DRYaqlip as shown in Fig. 1b. On average, correlation between DRYlip vs. CTRL was 0.96 ± 0.01, and slightly lower correlations were observed between the DRYaq vs. CTRL (0.89 ± 0.01) and DRYaqlip vs. CTRL (0.87 ± 0.01) groups. Hence, this demonstrated high similarities between the proteome of CTRL and DRYlip, as well as between DRYaq and DRYaqlip. In addition, this analysis indicates that reproducible data were generated from the pooled tear samples, which enabled further statistical analysis. A heat map with unsupervised hierarchical clustering of the data was generated and resulted in two major clusters, which are cluster 1 comprising CTRL and DRYlip, and cluster 2 comprising DRYaq and DRYaqlip, as shown in Fig. 1c. In order to identify a subset of proteins that significantly differentiate the DES subgroups compared to CTRL, a two samples t-test (P < 0.01) was performed. The summary of the significantly differentially expressed proteins in the different DES subgroups compared to CTRL is as tabulated in Supplementary data 1B. The total number of proteins that were significantly differentially expressed in the DRYlip vs. CTRL was 22 proteins, 58 proteins in the DRYaq vs. CTRL and 67 proteins in the DRYaqlip vs. CTRL. Next, intensity-based absolute quantification (iBAQ) values of the identified proteins extracted from MaxQuant analysis were used to elucidate the abundance of the identified proteins between DES subgroups and CTRL. Figure 1d shows that only top 20 abundant proteins from the total of 200 proteins identified make up almost 95.52 ± 0.32%, 96.36 ± 0.65%, 91.0 ± 0.13% and 92.55 ± 0.56% of the total abundance in CTRL, DRYlip, DRYaq and DRYaqlip, respectively (complete data in Supplementary data 1A). Moreover, LYZ, LCN1, lactotransferrin (LTF) and PIP make up as much as 60% of the total abundance of the CTRL tears. Generally, the aforementioned iBAQ analysis revealed that most of the significantly differentially expressed proteins were classified as low abundant proteins (lower than 0.5%) in this study, as shown in Supplementary data 1B.
Functional classification and protein interaction network analysis of the significantly altered proteins. To determine which biological processes were most affected in specific DES subgroups, the over-represented gene ontology biological process (GOBP) terms associated with the differentially expressed proteins were analysed employing the Database for Annotation, Visualization and Integrated Discovery (DAVID) tool. Several major biological process categories were observed to be decreased in DRYlip vs. CTRL, mainly inflammatory response (4%), however they were identified with lower percentage, as shown in Fig. 2a. On the contrary, 4 major biological processes categories were observed to be decreased in DRYaq vs. CTRL, especially immune responses (60%), as shown in Fig. 2b. Consequently, 18 major biological processes were found to be increased in this group and large percentage of them are involved in inflammatory response (51%), catabolic process (49%), cell death (38%), response to wounding (35%), defence response (32%), metabolic process (32%) and apoptosis (19%). Six major biological processes were found to be decreased in DRYaqlip vs. CTRL, especially immune response (60%) and defence response (32%), as shown in Fig. 2c. Meanwhile, as many as 30 biological processes were found to be increased in this group and most highly ranked were involved in cell death (83%), metabolic process (66%), inflammatory response (49%), catabolic process (46%) and apoptosis (41%).
Scientific RepoRts | 6:29629 | DOI: 10.1038/srep29629 The complete list of the over-represented GOBP terms of the differentially expressed proteins in DES subgroups compared to CTRL can be found in Supplementary data 1C. The characteristic of the differentially expressed proteins were further analysed, the over-represented gene ontology cellular component (GOCC) terms, molecule types and protein-protein interaction (PPI) networks were generated using Ingenuity Pathways Analysis. The over-represented GOCC terms of the significantly decreased proteins in DRYlip vs. CTRL were 5 extracellular space proteins and 9 cytoplasm proteins as well as increased in other 5 extracellular space proteins as shown in Fig. 3a. Meanwhile, the most significantly decreased proteins in DRYaq vs. CTRL were 8 extracellular space proteins and 12 cytoplasm proteins as well as drastic increased in other 12 extracellular space proteins and as many as 19 cytoplasm proteins as shown in Fig. 3b. Almost similar profiles were observed in DRYaqlip vs. CTRL, decreased of 13 extracellular space proteins and 12 cytoplasm proteins as well as increased in other 10 extracellular space proteins and 24 cytoplasm proteins were identified as shown in Fig. 3c. The molecular characteristics analysis of the differentially expressed proteins showed that many of these proteins are enzyme (23 proteins), transporter (12 proteins), kinase (2 proteins) peptidase (3 proteins), growth factor (1 protein), transmembrane receptor (1 protein), phosphatase (1 protein) and other (36 proteins), as shown in Supplementary data 1B. Generally, only numerous proteins of enzyme, transporter and peptidase were observed to be differentially expressed in DRYlip vs. CTRL, as shown in Fig. 4a. On the contrary, large preponderances of enzyme (12 proteins) and transporter (8 proteins) proteins were observed to be increased in DRYaq vs. CTRL, as shown in Fig. 4b. Similarly, as many as 16 enzymes and 7 transporter proteins were observed to be increased in DRYaqlip vs. CTRL, as shown in Fig. 4c. Besides, many of the differentially expressed proteins in all the DES subgroups were annotated as other, many of them of unknown characteristics, namely, PRR4 and zymogen granule protein 16 homolog B (ZG16B). The summary of the PPI networks of the differentially expressed proteins according to their GOCC terms and molecule  . In DRYaqlip vs. CTRL, the proteins with the highest number of direct PPI were 14-3-3 protein zeta/delta (YWHAZ) (23), protein S100-A9 (S100A9) (9), ALB (8), annexin A1 (ANXA1) (7), LYZ (6) and PIGR (6). Generally, the most significantly altered PPI networks and their related top diseases and functions for the DRYlip vs. CTRL, DRYaq vs. CTRL and DRYaqlip vs. CTRL were "dermatological diseases and conditions, developmental disorders", "cellular movement, immune cell trafficking, metabolic disease" and "antimicrobial response, inflammatory response, humoral immune response", respectively, as tabulated in Table 1.
Verification of the major differentially expressed tear proteins from DES patients. Finally, a rapid and robust approach via in-solution digestion of the pooled tear samples and accurate inclusion mass screening (AIMS) analysis was utilized to verify the major differentially expressed tear proteins in DES subgroups compared to the CTRL. Table 2 shows the detailed AIMS analysis of the signature peptides based on representative precursor ions for specific proteins (complete data in Supplementary data 1E). This analysis ascertained the differentially expressed profiles of 13 proteins in the DES subgroups compared to CTRL. Among the 13 differentially expressed proteins identified, PRR4, ZG16B and proline-rich protein 1 (PROL1) were found to be significantly decreased in both DRYaq and DRYaqlip subgroups but only slightly decreased in abundance in the DRYlip subgroup. SCGB2A1 and deleted in malignant brain tumors 1 protein (DMBT1) were found to be significantly decreased in both DRYaq and DRYaqlip subgroups. Extracellular glycoprotein lacritin (LACRT) was found significantly decreased only in DRYaqlip subgroup. On the contrary, S100A8, S100A9 were found to be significantly increased in both DRYaq and DRYaqlip subgroups but only slightly increased in abundance in the DRYlip subgroup. Alpha-enolase (ENO1), serotransferrin (TF), phosphatidylethanolamine-binding protein 1 (PEBP1) and alpha-1-acid glycoprotein 1 (ORM1) were found to be significantly increased in both DRYaq and DRYaqlip subgroups. Aldehyde dehydrogenase, dimeric NADP-preferring (ALDH3A1) was found significantly increased only in DRYaqlip subgroup. Two signature peptides for PRR4 were utilized in the AIMS analysis because these two peptides are essential for exact identification and quantification of this protein 35 .

Discussion
This study had unravelled as many as 79 differentially expressed tear proteins in the DES subgroups compared to CTRL based on the LFQ analysis. Similar expression profiles of the 37 of proteins from this list were already demonstrated in other studies associated with dry eye with/without systemic diseases, as tabulated in Supplementary data 1F, thereby, corroborating with the present results. This study also identified 42 novel proteins associated with DES, and this supports the relevance of the quantitative tear proteomics to identify novel differentially expressed proteins for specific DES.The functional classification and protein interaction network analysis of the significantly altered proteins suggest that the development of DES is a complicated process involving proteins of multiple biological functions. In this study, although 22 tear proteins were differentially expressed in DRYlip vs. CTRL, they were only differentially expressed in a lesser degree and most of them are involved in the dermatological diseases and conditions. On the contrary, the differentially expressed proteins in DRYaq vs. CTRL and DRYaqlip vs. CTRL demonstrated the involvement of mostly extracellular space and cytoplasm proteins, which consisted largely of enzymes and transporter protein types. Basically, the PPI networks analysis demonstrated the involvement of various biological processes and diseases, especially the cellular movement, immune cell trafficking, antimicrobial response, inflammatory response, humoral immune response, metabolic disease and neurological disease. Based on the outcomes of this study, it could be postulated that the pathological mechanisms underlying DRYaq and DRYaqlip are driven by tear film instability caused largely by the decrement of specific extracellular space and cytoplasm proteins, which mainly induces the activation of inflammatory, immune defence, cell death and apoptosis mechanisms. These host-defence mechanisms act in concert to maintain homeostasis of the tear film following the disease insult 1 . These diverse functions are bio-energetically expensive and require precise control of cellular metabolic pathways 39 . Collectively, these results reveal for the first time that high numbers of metabolic enzymes are expressed, preferentially to fuel the cell fate decisions and effector functions of the aforementioned metabolic processes.
In this study, for the first time, the iBAQ analysis demonstrated that the top 20 abundant proteins are accountable for approximately 90% of the total abundance of the tear proteome. This analysis also revealed that most of the significantly differentially expressed proteins were classified as low abundant proteins. For that reason, only 13 major proteins were successfully verified employing the AIMS strategy in DES subgroups compared to CTRL, which composed of PRR4, ZG16B, SCGB2A1, DMBT1, PROL1, LACRT, ALDH3A1, ENO1, TF, S100A8, S100A9, PEBP1 and ORM1. Drastic decrement in abundance of truncated PRR4 in the DRYaq subgroup has been demonstrated previously 6,21,24,26 6,7,25 . It is noteworthy that similar decrement patterns of PRR4 were documented in the tears of patients with other systemic diseases such as in diabetic proliferative retinopathy, thyroid-associated orbitopathy and multiple sclerosis [40][41][42] . Interestingly, although the decreased expression profiles of PRR4 is well documented to be associated with DES and the attribution to the functional relevance of the lacrimal gland, the exact biological functions of this protein in tears remains to be characterized. In addition, our recent study had characterized PRR4 as a complex protein with multiple isoforms and post-translational modifications, which provides an fundamental information for determination of the functional consequences in the disease state 35 42 . However, both studies did not verify the ZG16B expression profiles employing orthogonal methods, mainly owing to the unavailability of specific functional antibodies. The precise biological function of ZG16B is also largely unknown in tears. Up till now, ZG16B is known as a secretory lectin protein that is proposed to play a regulatory role in intestinal goblet cells and in the pancreatic acinar cells to stimulate "flushing out" of the granule content during exocytosis 43,44 . Furthermore, this secreted protein was also assumed to play an important role in the maintenance of the inflammatory state in cancer tissue [45][46][47][48][49] . It is important to highlight here that similar to the findings of this study, both ZG16B and PRR4 were found to be distinctly decreased in the tear samples of patients with multiple sclerosis, a progressive neurological disability processes 42 . On the contrary, in our recent study, both PRR4 and ZG16B were found to be significantly increased in abundance in reflex tears, a stimulus by neurological processes 50 . Taken together, the decrement of PRR4 and ZG16B in tears of DES patients could be hypothesized to be associated with the impairment of the neurological processes in the lacrimal gland. Next, similar to the findings of the present study, Srinivasan et al. also documented decrement of PROL1 in the MSDE 24 . The exact function of PROL1 is also still largely unknown. Recently, Dufour et al. reported that the immunoreactive opiorphin (QRFSR-peptide), a mature secretory peptide product of the PROL1 gene, is secreted primarily by lacrimal gland tears at the highest physiological rates (~200 ng/ml) in healthy volunteers 51 . It has been suggested that the potential role of opiorphin is in modulating lacrimal fluid homeostasis by increasing enkephalin bioavailability in case of certain causes of epiphora. Opiorphin has also been proposed to play paracrine and/or autocrine roles in the lacrimal system and at the ocular surface. Similar decrement in abundance of SCGB2A1 in DRYaq group was also confirmed by Soria et al. and Srinivasan et al. 22,24 . Although the exact physiological function of SCGB2A1 is still not known, it is hypothesized to play a crucial role as an anti-inflammatory agent owing to its association as a member of the uteroglobin family 52 . Next, DMBT1 was found for the first time to be significantly decreased in abundance in DRYaq and DRYaqlip subgroups in this study. It is well documented that this protein plays critical roles in the epithelial differentiation, cellular immune defence and mucosal defence system [53][54][55] . Decrement abundance of LACRT was also documented in various studies associated with DES, namely in the DRYaq, SS, SJS, RA, MDE, MSDE and DRY_CL 6,20,24,25 . Recent studies demonstrated that this secretory glycoprotein might play an imperative function in secretion and renewal of ocular surface and lacrimal epithelia 56,57 . In this study, protein S100A8 and S100A9 were found to be increased in abundance in all the DES subgroups, and they discriminated best the DRYaq and DRYaqlip from the CTRL. Decrement in abundance of S100A8 and S100A9 were widely documented in various studies associated with DRYaq, DRYaqlip and SS [21][22][23]25,26 . These proteins are identified to play a part in inflammatory processes and Zhou et al. demonstrated that higher expressions of these proteins were related with increased signs of dryness 23 . Zhou et al. also demonstrated that the increased abundance of ENO1 in the tears is correlated to the DES 23 . ENO1 is an important glycolytic enzyme, hitherto several studies demonstrated their potential roles in several disease progressions, namely, in cancer and autoimmune disorders [58][59][60][61] . The similar increment of ORM1 protein was also documented in DRYaqlip and SS groups 23,25 . ORM1, is a heavily glycosylated protein and categorized as a member of the immunocalin family that modulates inflammatory and immune responses 62 . Meanwhile, the similar increment of TF was only documented by Li et al. in DRY_SS group 25 . TF is an iron binding transport protein and this characteristic resulted in anti-microbial properties 63,64 . PEBP1 was found for the first time significantly increased in abundance in DRYaq and DRYaqlip groups in this study. PEBP1 is an inhibitory modulator of Raf kinase protein and G-protein coupled receptor (GPCR) signalling cascade, as well as an activator of nuclear factor κ B (NF-κ B) 65,66 . Therefore, PEBP1 basically represents a novel effector of signal transduction pathways that modulate apoptosis, motility, therapeutic resistance, genomic integrity and cellular growth 65 . Besides, the disruption of PEBP1 has been reported to be related with a diverse range of diseases, namely, pancreatitis, cancer and Alzheimer's disease, making it a potential target for disease therapy 65 . Increment in abundance of ALDH3A1 was also documented in several studies of DRY_SS, MDE but has never been verified 24,25 . ALDH3A1 is a corneal crystallin protein, highly expressed in epithelial cells and stromal keratocytes 67 . ALDH3A1 plays an active metabolism  ALB, ANXA1, ANXA5, CFH,  GC, GSN, HP, IGHG1, IGHG2,  MDH1, ORM1, ORM2, PFN1,  PPIA, S00A8, S100A9, SERPINA1,  TF, TTR   ACTB, CLU, HSPA1A/HSPA1B,  HSPG2, IGHA1, KRT1, KRT9,  S100A9, ZG16B   71  28  Cellular movement, immune cell  trafficking, metabolic  of toxic aldehydes in the cornea 68 . An essential role for ALDH3A1 in the cornea is suggested by observation that diseased corneas, especially cataract development, are associated with decreased ALDH3A1 catalytic activity 69,70 .
The rest of the 66 differentially expressed proteins were not successfully verified, largely due to the low abundance of these proteins in tears and limitations of the AIMS strategy. It is well-recognized that in the sequence of new biomarker identification, the validity of potential candidate biomarker panels must be tested thoroughly in individual samples. However, biomarker validation is a very costly and time-consuming task, primarily due to the complexity of multiplexed assays, as well as technological challenges in identifying and quantifying low abundant proteins 71,72 . Hence, in the current study, the AIMS analysis employed was limited to the pooled tear samples to screen for and verify major differentially expressed proteins in DES samples. However, since this is the first study that extensively identified and verified the cluster of differentially expressed proteins in tears, which could be potential biomarker candidate(s) for the different subgroups of DES. The important findings emerging from this this study with the use of pooled samples are envisioned to provide fundamental information and will be a highly useful platform and reference point for future studies utilizing individual samples to dissect 'personalized' expression of these proteins. This will be especially useful for analysis of individual tear samples of larger cohorts of DES patients in the clinical settings to demonstrate the discriminative power (e.g. sensitivity and specificity).
In conclusion, this study had unravelled the intricate regulation profiles of specific cluster of proteins responsible for the maintenance of tear film stability in the different DES subgroups. Besides, we have unravelled numerous proteins that participate in metabolic processes, which might provide new insights for modulating various biological processes, especially the inflammatory and immune defence disorders of the ocular surface. The outcomes of the identification of these proteins, when extrapolated to clinical application, can provide invaluable hints on development of specific diagnostic tool for clinical tests and are of great importance for the prognostic usage for improved clinical management of the disease in the future.

Materials and Methods
Tear sampling. The study was performed in strict adherence to the guidelines of the 1964 Declaration of Helsinki and all experimental protocols were approved by the "Landesärztekammer Rheinland-Pfalz" ethics committee. All participants were informed of the possible risks, the goal of the study and privacy policy, and an informed consent was signed according to the recommendations of the Declaration of Helsinki for investigation with human subjects. In this study, tear proteins of 80 patients were included and assigned into DRYlip, DRYaq, DRYaqlip and CTRL. Each group comprises 20 subjects equally divided to male (M) and female (F), age between 21 to 79 years old. The classification of patients was carried out according to the guidelines of the Tear Film and Ocular Surface Society (TFOS, Ocular Surface, 2007) at the Department of Ophthalmology at the University Medical Center of the Johannes Gutenberg University, Mainz, following specific inclusion and exclusion criteria of basic secretory test (BST), tear breakup time (TBUT), clinical parameters after Bron and Foulks Score and Lid-parallel Conjunctival Folds (LIPCOF) and extensively asking for the anamnesis and symptoms using Ocular Surface Disease Index (OSDI) questionnaire. Table 3 summarizes the clinical evaluation parameters for the classification DES subgroups and CTRL patients (complete description in Supplementary data 1G). From each individual, tear samples have been taken by a BST, a clinical standard test, using Schirmer's strip. Schirmer's strips have been placed into the lower eye lid for a few minutes, after having prepared the eye with mild anesthetic eye drops (Novesine). The eye drops were used in this study to assess the basal tear secretion from the patients as critically as possible. The samples were immediately stored at − 80 °C for subsequent analysis. Tear proteins were extracted from the Schirmer's strips by soaking each strip in 500 μ l phosphate-buffered saline (PBS) for 3 hours at 4 °C to elute the tear proteins. Subsequently, the total protein concentrations in the collected tear samples were

Discovery study: LFQ analysis via 1DE & LC-ESI-MS/MS strategy.
Label-free quantification of peptides via 1DE & LC-ESI-MS/MS strategy was employed to identify the change in protein abundance in the discovery data and to generate a list of the differentially expressed proteins from specific DES subgroups for subsequent verification. The tear samples for each assigned group (N = 20) were pooled equally (2.5 μ g per individual) to a total of 50 μ g with three replicates. The reason for equal amount of protein collection and pooling from each patient within a group was to normalize the difference between subjects and to reduce individual variation. The pooled tear samples of each DES subgroups were subjected to 1DE (50 μ g/well and sliced into 10 bands), trypsin digested and the extracted peptides were purified utilizing methods previously described 35,50 .
The LC-ESI-LTQ-Orbitrap MS system employed and the mass spectrometric settings utilized for this study has been described in detail elsewhere 35,50 . The acquired continuum MS spectra were analysed by MaxQuant computational proteomics platform version 1.4.1.2 and its built-in Andromeda search engine for peptide and protein identification, with LFQ and iBAQ algorithm enabled 38,[73][74][75] . The tandem MS spectra were searched against Uniprot Human database (date, 19/March/2014) using settings as described in detail elsewhere 50 . The false discovery rate (FDR) for protein identification was set to 0.01 with ≥ 6 amino acid residues and "unique plus razor peptides" were selected to be included for LFQ and iBAQ analysis 73 . The output of the generated "proteingroups. txt" data from the MaxQuant analysis was utilized for Pearson correlation, clustering and statistical analysis using Perseus software. Unsupervised hierarchical clustering of the LFQ values was performed based on Euclidean distances on the Z-scored between mean values. For statistical analysis, two-samples t-test-based statistics with P < 0.01 was applied on Log2 transformed LFQ values and the minimum number of values "in at least one group" is 3 to assert proteins regulation as significant for the specific groups 38,50 . The iBAQ values, which were calculated by dividing the summed peptide intensities for a given protein with the number of theoretically observable tryptic peptides, act as a degree of protein abundance were converted to percentage and employed to compare between proteins in each group.
Verification Study: Targeted MS via AIMS strategy. The identified candidate biomarkers from the discovery stage were verified employing a targeted form of MS strategy called AIMS 71,76 . The main aim of employing the AIMS strategy for verification in the present study is to determine the major differentially expressed proteins to discriminate the specific DES subgroups. The tear samples for each assigned group (N = 20) were pooled equally (0.5 μ g per individual) to a total of 10 μ g with quadruplicate and subjected to in-solution digestion. The pooled tear samples were digested with sequence grade-modified trypsin [protease: protein ratio of 1:20 (w/w)] for 16 hours at 37 °C with 50 μ l of 50 mM NH 4 HCO 3 in 10% acetonitrile. The digested samples were purified on ZipTip C18 columns and the eluates were concentrated to dryness in the SpeedVac concentrator and dissolved with 10 μ l of 0.1% TFA solution prior to targeted MS analysis 35,50 . For targeted MS analysis, the LC-ESI-MS/MS system was operated as described in detail previously with several parameters adjusted to target only a specific set of peptides 35,50 . The selection of the experimentally identified peptides list that annotated to specific protein was carried out manually from the "msms.txt" file resulting from MaxQuant analysis governed by these criterions, contains no missed cleavages, fully tryptic, no modifications and peptide only assume charge + 2. The created peptide inclusion list comprising m/z and charge was assigned to the instrument acquisition software before data acquisition. The in-solution digested tear samples consisted of very complex mixtures especially the high abundant proteins. Hence, to yield optimum results, two LC gradients of 60 min (as described elsewhere 35 ) and 120 min were utilized. The gradient for 120 min per sample is as follows: 0-  Table 3. Summary of the clinical parameters for the classification of DES subgroups and CTRL patients.
'peptides.txt' file from MaxQuant were used to calculate the sum absolute intensity of the signature peptides for each proteins, and were transferred to Statistica (v8, StatSoft, Tulsa, OK) for t-test analysis (independent, by groups, P < 0.05).
Functional annotation and pathways analysis. Proteins determined to be differentially expressed as described based on the data in our LFQ experiments were tabulated in Excel and their gene names were used for functional annotation and pathways analysis. First, DAVID tool (version 6.7) (http://david.abcc.ncifcrf.gov/ home.jsp) was used for interpreting the GOBP terms of the differentially-expressed proteins 77,78 . The protein list was uploaded into DAVID and searched for enrichment for GOBP term and the results were filtered based on threshold count ≥ 2 and P values < 0.05. Ingenuity Pathways Analysis software (IPA, Ingenuity QIAGEN Redwood City, CA) (www.qiagen.com/ingenuity) was used for interpreting the GOCC terms, molecule types and PPI networks as well as top diseases and functions associated with the differentially-expressed proteins. Top canonical pathways involving the differentially expressed proteins were presented, along with a p-value calculated using Fisher's exact test. The molecular interactions networks between proteins associated with top diseases and functions were reported. Proteins are displayed with their corresponding gene names and represented as nodes, whereas protein-protein interactions based on direct associations (experimentally observed) between two nodes are represented with an edge (line). Nodes are presented using different shapes to represent the functional protein class and node colour indicates decreased (green) or increased (red) abundance.