Free
Biochemistry and Molecular Biology  |   March 2013
Alterations in the Tear Proteome of Dry Eye Patients—A Matter of the Clinical Phenotype
Author Notes
  • From Experimental Ophthalmology, Department of Ophthalmology, University Medical Center of the Johannes Gutenberg University, Mainz, Germany. 
  • Corresponding author: Franz H. Grus, Experimental Ophthalmology, Department of Ophthalmology, University Medical Center of the Johannes Gutenberg University, Langenbeckstrasse 1, 55131 Mainz, Germany; grus@eye-research.org
Investigative Ophthalmology & Visual Science March 2013, Vol.54, 2385-2392. doi:10.1167/iovs.11-8751
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Nils Boehm, Sebastian Funke, Michaela Wiegand, Nelli Wehrwein, Norbert Pfeiffer, Franz H. Grus; Alterations in the Tear Proteome of Dry Eye Patients—A Matter of the Clinical Phenotype. Invest. Ophthalmol. Vis. Sci. 2013;54(3):2385-2392. doi: 10.1167/iovs.11-8751.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose.: Previous studies demonstrated alterations in the tear proteome of dry eye patients. The aim of the present study was to analyze tear protein patterns of dry eye patients considering different clinical phenotypes in order to examine their influence on tear film protein composition.

Methods.: We applied a surface-enhanced laser desorption/ionization-time-of-flight (SELDI-TOF)/matrix-assisted laser desorption/ionization-time-of-flight (MALDI-TOF)/TOF mass spectrometry (MS)–based strategy to detect/identify candidate biomarkers. Tear samples of 169 patients, enrolled in two independent studies, were analyzed. Patients were subdivided into healthy controls (CTRL: N = 39), aqueous-deficient dry eye (DRYaq: N = 40), lipid-deficient dry eye (DRYlip: N = 40), and a combination of the two (DRYaqlip: N = 40).

Results.: We uncovered six peptide/protein markers matching the stringent criteria applied for selection of reliable markers (P < 5.0E-03 in both studies). For example, proline-rich protein 4 was found to be diminished in DRYaq and DRYaqlip patients when compared to healthy subjects. Mammaglobin B and lipophilin A were found to be increased in these patients, as well as calgranulin S100A8. Remarkably, DRYlip patients revealed only slight alterations; these patients strongly deviated from the DRYaq or DRYaqlip group. With regard to classification of patients, we achieved discrimination from healthy subjects with a sensitivity and specificity ≈100% for DRYaq and DRYaqlip patients (receiver operating characteristic curve [ROC curve]: area under the curve [AUC] = 1) through use of the six-biomarker set.

Conclusions.: This study demonstrates that different clinical phenotypes of dry eye are reflected by specific alterations of the tear film proteome. Especially a deficiency of the aqueous phase of the tear film seems to strongly influence the expression patterns of several proteins.

Introduction
Dry eye disease is defined as “a multifactorial disease of the tears and ocular surface that results in symptoms and discomfort, visual disturbance, and tear instability with potential damage to the ocular surface. It is accompanied by increased osmolarity of the tear and inflammation of the ocular surface.” 1 Patients suffering from dry eye typically show symptoms like irritation, photophobia, burning, or a general discomfort. Moreover, they have an elevated risk of corneal infection potentially resulting in irreversible tissue damage. 2 The prevalence of dry eye syndrome is rapidly increasing; and in Western countries, more than 6% of the population over the age of 40 suffer from dry eye and more than 15% of people aged 65 years and older. 3 In Asian populations, an incidence of even 21% to 50% has been detected, 46 thus reflecting the varying vulnerabilities for dry eye in different ethnic groups. Two main subclasses contributing to dry eye have been identified: deficiency of the aqueous phase (tear deficiency) and alteration of the lipid layer composition. 2 The deficiency of the lipid layer leads to an increased evaporation of fluid (evaporative dry eye), thus considerably increasing the tear turnover rate as well as the osmolarity of tears. 7 The diagnosis of dry eye syndrome, and especially the subclassification of dry eye patients, remains complicated. Several factors, for example tear osmolarity, 810 the concentration of cytokines, 1113 and the occurrence of protein glycosylation patterns, 14 have been a focus as tools for diagnostic purposes in dry eye. Also, alterations of the tear proteome have been the subject of several studies in the context of dry eye, as is also the case for other ocular surface–affecting diseases, such as Sjögren's syndrome, 15 meibomian gland disease (MGD), 16 pterygium, 17 allergy, 18 and diabetes. 19  
With respect to dry eye, several proteins have been found to be altered in comparison to those in healthy subjects. For example, lactoferrin, lyzozyme, proline-rich protein 4 (PRR4), and calgranulin S100A8 were detected in increased or decreased amounts in patients suffering from dry eye. 2023 Comparison of protein patterns obtained from capillary tear collection or collection of tears by Schirmer strip was also the subject of several studies, with more or less congruency between the two sampling methods. 24,25  
However, despite the multiple proteomic studies related to this topic, it is still uncertain how the protein compositions of tears from patients suffering from different dry eye subtypes deviate from each other. 
The objective of this study was to detect and confirm alterations of the protein expression profiles in the tears of patients suffering from different subtypes of dry eye syndrome, and to examine their suitability for deployment as a diagnostic tool. The idea behind this classification into subtypes of dry eye is to look not only for general dry eye markers, but also for proteins that could differentiate between the different subtypes of dry eye and therefore, for example, help in the choice of an appropriate, personalized treatment strategy. 
Methods
Patients
Procurement of samples was performed in accordance with the Declaration of Helsinki on biomedical research involving human subjects and with the consent of the ethics committee of the Landesärztekammer of Rhineland-Palatinate. Informed consent was obtained from all study subjects. 
Study I (Discovery Study).
Forty-nine patients were included in this study. Nine patients were categorized as control subjects (CTRL: 5 females, 4 males). Dry eye patients were classified as patients with aqueous-deficient dry eye (DRYaq: n = 10; 5 females, 5 males), patients with changes in the lipid phase (DRYlip: n = 10; 6 females, 4 males), and patients with a combination of the two (DRYaqlip: n = 10; 5 females, 5 males). 
Study II (Validation Study).
One hundred twenty patients were subjects of this study, subdivided into 30 healthy control subjects (12 females, 18 males), 30 patients suffering from DRYaq (16 females, 14 males), 30 patients suffering from DRYlip (15 females, 15 males), and 30 patients suffering from DRYaqlip (19 females, 11 males). 
Tear samples (Schirmer strip) were collected using the same criteria for classification of patients in both studies. The ophthalmic examination included a detailed slit-lamp inspection, a Schirmer test with anesthesia (BST), and measurement of the tear breakup time (TBUT). The LIPCOFs (lid-parallel conjunctival folds) were also used for classification of dry eye patients. 26 Further, in both studies, fluorescein staining and lissamine green staining were performed. With slight changes, a score was built based on clinical parameters as described by Foulks and Bron. 27 Each patient was asked for his or her subjective symptoms, such as burning, itching, foreign body sensation, or dryness. Findings were combined into a patient-specific score revealing the patient's complaints of pain and discomfort. Study subjects with a BST result > 10 mm and a TBUT > 10 seconds were classified as healthy control (CTRL). If the BST and/or the TBUT were considered pathologic, the subject was classified as a dry eye patient. Patients with a BST < 10 mm and a normal TBUT were classified as DRYaq, since a reduced aqueous phase will predominantly result in a lowered BST but not in a pathologic TBUT. 1,28 Patients with a TBUT < 10 seconds and a BST > 10 mm were considered DRYlip patients, since a dysfunctional, thinned lipid layer goes along with a decreased TBUT but not necessarily with a pathological BST. 29 Patients revealing a BST < 10 mm as well as a TBUT < 10 seconds were classified as subjects suffering from a combined pathogenesis. A more detailed description of criteria selected for patient classification can be found in a previous publication (see Supplementary Material), as well in Boehm et al. 13  
Excluded from this study were patients suffering from Sjögren's syndrome, diabetes mellitus, allergy to local anesthetics, contact lens wearers, and patients who had undergone ophthalmic surgery during the last 6 months. Further, the use of any systemic drug suspected to have an influence on the tear production or systemic inflammatory processes led to exclusion of subjects (e.g., beta-blockers, selective serotonin reuptake inhibitors, oral contraceptives, postmenopausal estrogen therapy, local sympathomimetic drugs, nonsteroidal antirheumatic drugs [NSAR analgesics, such as ibuprofen, diclofenac, Voltaren, aspirin, Novalgin, Arcoxia], steroidal antirheumatic drugs, antihistamines). 
Sample Preparation
For extraction of sample proteins, the first 10 mm of the Schirmer strips was eluted overnight in 500 μL HPLC-grade H2O with 0.1% n-dodecyl-ß-maltoside and 0.1% trifluoroacetic acid. 
Study I (Discovery Study).
Fractionation of sample proteins was performed using the ProteinChip Biomarker System (Ciphergen Biosystems, Inc., Fremont, CA). All binding and washing steps were carried out on an automatic liquid handling station (BIOMEK 2000; Beckman, Fullerton, CA). Twenty microliters of the eluted tear proteins was incubated on two different chromatographic surfaces: a weak cation exchange surface (CM-10) and a hydrophobic chip surface for reversed-phase chemistry (H50). All steps were performed according to the standard protocols of the manufacturer. Binding buffers were 20 mM sodium–acetate buffer, pH 5 (CM-10), and 5% acetonitrile/0.1% trifluoroacetic acid (H50). As energy-absorbing molecule (EAM) we used saturated 3,5-dimethoxy-4-hydroxycinnamic acid (50% acetonitrile, 0.5% trifluoroacetic acid). Each sample was analyzed in duplicate on separate arrays. 
Study II (Validation Study).
The preparation of tear samples and their incubation on ProteinChip were performed as described above. 
Data Acquisition via Surface-Enhanced Laser Desorption/Ionization-Time-of-Flight (SELDI-TOF) Mass Spectrometry (MS)
ProteinChip Arrays were analyzed on a PBS-IIc ProteinChip Reader using ProteinChip Software version 3.2 (Ciphergen Biosystems, Inc.). Each array was read with two different protocols. Protocol A was optimized for high-molecular-weight proteins (>10 kDa) by averaging the signals of 195 laser shots from each spot with a laser intensity of 190, a deflector setting of 3000 Da, and a mass range of 3000 to 200,000 Da. To optimize the measurement of low-molecular-weight proteins (<10 kDa), the laser intensity was set at 180 and the deflector at 1500 Da. 
Data Analysis
Data obtained from SELDI-TOF MS were transferred to CiphergenExpress 2.1 database software (Ciphergen Biosystems, Inc.). Workup and analysis of raw data, for example normalization, peak detection, and creation of peak cluster lists, were performed as previously described. 23 The cluster lists, containing the normalized peak intensity values for each single patient, were transferred to Statistica (V8; StatSoft, Tulsa, OK). Based on peak intensities, statistical analysis was performed using methods such as ANOVA and t-tests. Further, we performed an analysis based on artificial neural networks (ANN) for determination of classification power of potential biomarkers. Therefore, data sets were randomly split into two parts with even numbers of patients per group. One half was used for training of the ANN and the second half for testing the trained ANN with regard to its classification power. No samples included in the training data set were used for classification purposes. Results were visualized by plotting sensitivity against specificity (receiver operating characteristic curve, ROC curve). A detailed description of methods applied for statistical analysis can be found in previous publications from our group. 23,30,31  
Identification of Candidate Biomarkers
We applied a previously shown workflow for identification of candidate biomarkers (protein/peptide peaks from SELDI-TOF spectra). 17,23,32 In order to enable identification of each of the demonstrated molecular weights/proteins, we pooled five samples of patients from the group revealing the highest level of the specific protein. In the first step, these tear sample pools were fractionated using Weak Cation Exchange (WCX), Strong Cation Exchange (SCX), and C18 chromatography; generated fractions were subsequently run on tricine SDS-PAGE (16%; Invitrogen, Carlsbad, CA). Molecular weight regions of interest were sliced from gel, and intact proteins/peptides were eluted from gel slices using 50% formic acid, 15% acetonitrile, 25% isopropanol, and 10% HPLC-grade water. An aliquot of each elution was measured on ProteinChip to check for the sole presence of target proteins/peptides. If several peaks were detectable, samples were further fractionated (CLINPROT magnetic beads; Bruker Daltonics, Bremen, Germany). For protein digestion via trypsin (10 ng/μL in 50 mM NH4HCO3 for 12 hours; Promega, Fitchburg, WI), only those fractions that contain the target molecular mass were used. Digest solutions were spotted onto anchor chip targets (Bruker Daltonics). As EAM we used α-cyano-4-hydroxycinnamic acid (2 mg/mL, 50% ACN, 0.2% trifluoroacetic acid). Mass spectrometry analysis was performed using an Ultraflex II (Bruker Daltonics). MS-mode acquisition (1000–4000 Da) consisted of 650 laser shots averaged from five sample positions. The top 10 peaks were used for subsequent tandem mass spectrometry (MS/MS) analysis. Peptide fragmentation was performed using collision-induced dissociation, and 850 laser shots from five sample positions were summed up for each parent ion. Data processing of raw spectra and protein identification were performed using Bruker software (flexAnalysis 2.4 and BioTools 3.1; Bruker Daltonics) and Mascot (Matrix Science Limited, London, UK). Mascot search was conducted using a MS tolerance of 100 parts per million (PPM), a MS/MS tolerance of 0.7 Da, one allowed miss cleavage, standard scoring, and a significance threshold of P < 0.05 for protein/peptide identification. Database search was performed using SwissProt (Geneva, Switzerland) release 57.12. All proteins that exceeded the significance threshold and additionally had at least two assigned peptides were counted as identified, whereas one of the peptide needs to be identified with a Mascot score > 40. 
Results
In the present study we analyzed the tear protein patterns of patients suffering from different subtypes of dry eye syndrome. With the guiding principle of developing reliable biomarkers for the classification, and especially subclassification, of dry eye patients, we performed two studies with independently enrolled patient cohorts. Comprehensive clinical documentation, including determination of the TBUT, LIPCOF, lissamine green, and fluorescein vital staining, was performed for all study subjects. Additionally, subjective symptoms were gathered via a questionnaire. The results of the clinical examination are summarized in Tables 1 and 2
Table 1
 
Clinical Data of Patients Enrolled in Study II
Table 1
 
Clinical Data of Patients Enrolled in Study II
Age ± SD BST ± SD, mm LIPCOF ± SD TBUT ± SD, s Fluorescein ± SD, Oxford Scale Lissamine Green ± SD, Oxford Scale Score for Complaints of Pain and Discomfort DEWS Score ± SD
CTRL 51.21 ± 16.61 21.06 ± 5.91 1.61 ± 0.83 14.82 ± 0.58 NS NS NS NS
DRYaq 52.83 ± 17.20 6.69 ± 2.71 2.57 ± 0.82 14.51 ± 1.15 1.14 ± 1.03 1.26 ± 1.07 9.40 ± 3.14 1.69 ± 0.37
DRYlip 53.09 ± 15.30 18.81 ± 6.11 2.60 ± 0.90 9.02 ± 1.88 1.04 ± 1.00 1.12 ± 0.98 16.84 ± 4.47 1.54 ± 0.40
DRYaqlip 55.41 ± 14.72 5.00 ± 2.94 2.71 ± 0.78 8.20 ± 2.16 1.54 ± 0.95 1.67 ± 1.02 19.54 ± 3.85 2.13 ± 0.47
Table 2
 
Clinical Data of Patients Enrolled in Study II: Post Hoc Values: Age
Table 2
 
Clinical Data of Patients Enrolled in Study II: Post Hoc Values: Age
CTRL DRYaq DRYlip DRYaqlip
CTRL 0.975 0.956 0.670
DRYaq 0.975 0.999 0.894
DRYlip 0.956 0.999 0.908
DRYaqlip 0.670 0.894 0.908
For all parameters, significant differences were detected between healthy and dry eye subjects (P < 1.0E-05). Some of the parameters also revealed differences between the dry eye subgroups—an outcome due to the criteria selected for subclassification of dry eye patients. However, it could be observed that patients suffering from a combined pathogenesis, with a deficiency in the aqueous and the lipid layer (DRYaqlip), exhibited the strongest clinical phenotypes (mean [ME] ± SD: BST = 5 ± 2.94; LIPCOF = 2.71 ± 0.78; TBUT = 8.2 ± 2.16). Also, the complaints of pain and discomforts in these patients were significantly greater when compared to those of other dry eye subgroups (P < 1.0E-05). Therefore, DRYaqlip patients revealed the highest Dry Eye WorkShop (DEWS) severity code (ME ± SD: 2.13 ± 0.47). 
In terms of biomarker profiling, tear proteins of all patients, obtained by elution of Schirmer strips, were analyzed using the SELDI-TOF technology. For each patient, complex protein and peptide patterns could be generated; and overall, more than 300 different protein/peptide clusters (molecular weights) were identified. Each cluster could be consistently detected in all study groups. Concerning the search for potential biomarkers, only peaks smaller than 20 kDa were included, a restriction due to the mass spectrometer's sensitivity and resolution optimum in this mass range. To ensure a robust conclusion regarding the suitability of protein patterns as a diagnostic tool for dry eye, as well as for the subclassification of dry eye patients, we considered only those markers that revealed a highly significant difference between one of the dry eye subgroups and the CTRL group (P < 5.0E-03) and that could be further validated in study II taking the same stringent significance level as a basis. Moreover, these markers needed to be successfully identified by matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) MS/MS for further consideration. Overall, we discovered six biomarkers matching these criteria (Figs. 1, 2; see Supplementary Material and Supplementary Tables S1A–F, S2, and Fig. S2). Markers at 4043 Da and 4078 Da were identified as PRR4, whereby the 4043-Da marker presents the oxidized form of the previously detected PRR4 fragment at 4027 Da.23 However, some of the assigned peptides revealed low Mascot scores (see Supplementary Material and Supplementary Table S2), but each of the proteins was identified by a peptide revealing a Mascot score > 40. Further, the calculated molecular weights of assigned proteins (e.g., protein S100 or beta-2 microglobulin) given by the database differed considerably from the observed molecular weights. These differences may be due to posttranslational modifications or truncated forms of the observed proteins. 
Figure 1. 
 
Shown are box plots for detected biomarkers at 4043 Da and 4078 Da. Both markers were identified as proline-rich protein 4. Diagrams on the left exhibit data from study I; those on the right are from study II. The y-axis represents the relative, normalized intensities. Boxes represent ME ± SE, rectangles the ME ± SD.
Figure 1. 
 
Shown are box plots for detected biomarkers at 4043 Da and 4078 Da. Both markers were identified as proline-rich protein 4. Diagrams on the left exhibit data from study I; those on the right are from study II. The y-axis represents the relative, normalized intensities. Boxes represent ME ± SE, rectangles the ME ± SD.
Figure 2. 
 
Shown are box plots for detected biomarkers at 10,117 Da (mammaglobin B), 10,234 Da (lipophilin A), 10,866 Da (calgranulin S100A8), and 12,729 Da (beta-2 microglobulin precursor). Diagrams on the left exhibit data from study I; those on the right are from study II. The y-axis represents the relative, normalized intensities. Boxes represent ME ± SE, rectangles the ME ± SD.
Figure 2. 
 
Shown are box plots for detected biomarkers at 10,117 Da (mammaglobin B), 10,234 Da (lipophilin A), 10,866 Da (calgranulin S100A8), and 12,729 Da (beta-2 microglobulin precursor). Diagrams on the left exhibit data from study I; those on the right are from study II. The y-axis represents the relative, normalized intensities. Boxes represent ME ± SE, rectangles the ME ± SD.
In comparison to healthy subjects, both markers were found to be significantly decreased in DRYaq (mean fold decrease: 3.08; P < 3.48E-03) and DRYaqlip (mean fold decrease: 3.41; P < 2.45E-03) patients, but not in subjects suffering from DRYlip (P < 2.20E-01). Results of study II confirmed these observations, revealing a more pronounced group difference between DRYaq/DRYaqlip and healthy subjects (both P < 3.0E-05). 
In contrast to these low-molecular-weight markers, we also found proteins that are considerably increased in the tears of dry eye patients. These include two members of the secretoglobin family—mammaglobin B (10,117 Da) and lipophilin A (10,234 Da)—as well as calgranulin S100A8 (10,866 Da) and the beta-2 microglobulin precursor (12,729 Da). In comparison to the CTRL group, mammaglobin B and lipophilin A revealed a significant deviation for DRYaq (P < 5.0E-03; P < 1.42E-04) and DRYaqlip (P < 2.47E-4; P < 4.81E-04) patients, but not for DRYlip (P < 1.0E+0; P < 9.92E-01) subjects—a finding that is similar to results for the PRR4 markers. In terms of mammaglobin B, the highest increase could be detected for DRYaqlip patients (mean fold increase: 5.5); and in the case of lipophilin A, DRYaq patients revealed the strongest alteration (mean fold increase: 7.53). The analysis of tear samples from independently enrolled patients of study II confirmed these tendencies and statistical findings (see Supplementary Tables S1C, S1D). In study I, calgranulin S100A8 was found to be significantly increased in tears of DRYaq and DRYaqlip patients (P < 4.58E-04; P < 1.29E-04), but also in subjects suffering from DRYlip (mean fold increase: 5.75; P < 8.53E-03). However, in study II, a significant alteration of calgranulin S100A8 could be detected only for DRYaq and DRYaqlip patients (see Supplementary Material and Supplementary Table S1E). Beta-2 microglobulin precursor revealed an increased protein level in DRYaq (P < 2.64E-04) and DRYaqlip (P < 1.29E-04) patients but not in DRYlip subjects enrolled in study I. Study II confirmed these results (see Supplementary Material and Supplementary Table S1F). Interestingly, the quantities of most of these markers exhibit a strong difference between DRYaq/DRYaqlip patients and patients suffering solely from deficiency of the lipid layer (DRYlip), but none of them revealed a statistical difference between DRYaq patients and those with a combined pathogenesis (DRYaqlip). Detailed statistical data for all markers have been published previously (see Supplementary Material and Supplementary Tables S1A–F). 
Besides the strong intergroup differences, we detected a slight correlation between the protein amounts of uncovered candidate biomarkers and the estimated BST values, thus further corroborating the suggested influence of the aqueous state on the protein composition of the tear film. For the PRR4/nasopharyngeal carcinoma associated proline-rich 4 (NCAPP4) markers, a positive correlation with the BST could be detected (4043 Da: r = 0.45; 4078 Da: r = 0.64), and a negative correlation for mammaglobin B (10,117 Da: r = −0.54), lipophilin A (10,234 Da: r = −0.5), calgranulin S100A8 (10,886 Da: r = −0.5), and beta-2 microglobulin precursor (12,729 Da: r = −0.58). However, correlations are a bit lower if CTRL subjects are removed from the calculation (4043 Da: r = 0.46; 4078 Da: r = 0.64; 10,117 Da: r = −0.49; 10,234 Da: r = −0.419; 10,866 Da: r = −0.47; 12,729 Da: r = −0.52), and no correlations are observed in individual subgroups after outlier elimination (see Supplementary Material and Supplementary Table S3). 
In order to test the suitability of discovered biomarkers as a diagnostic tool for dry eye, we performed a multivariate analysis of discriminance as well as data analysis through the use of ANN. Analysis of discriminance provides an indication of how different single patients are from each other on the basis of their protein patterns, meaning their biomarker profiles. An appropriate parameter to visualize these differences is given by the so-called canonical roots, classifying each patient in a discriminant space. The closer together the roots of the different patients are, the more similar their protein patterns. Figure 3 reveals the canonical roots for patients analyzed in studies I and II, taking the measured values for the six discovered biomarkers as a basis. It can be observed that DRYaq and DRYaqlip patients, regardless of whether they were enrolled in study I or study II, are the most distant from the control group, followed by DRYlip patients. 
Figure 3. 
 
Depicted are the canonical roots for patients enrolled in studies I and II. They demonstrate the similarities/differences between the protein patterns of single individuals. Blue circles: CTRL; red boxes: DRYaq; green boxes: DRYlip; black boxes: DRYaqlip.
Figure 3. 
 
Depicted are the canonical roots for patients enrolled in studies I and II. They demonstrate the similarities/differences between the protein patterns of single individuals. Blue circles: CTRL; red boxes: DRYaq; green boxes: DRYlip; black boxes: DRYaqlip.
Further, the output data of ANN, trained with the biomarker panel of a subset of enrolled patients, were used to assess the sensitivity and specificity achieved for the classification and subclassification of dry eye patients. Regardless of the different dry eye subtypes, dry eye patients could be differentiated from healthy subjects with a sensitivity and specificity > 80% (Fig. 4A). This finding is due to the similar protein patterns of DRYlip and CTRL patients, lowering the sensitivity and specificity for the detection of dry eye patients in general. In contrast, DRYaq and DRYaqlip patients could be perfectly discriminated from healthy subjects by use of the discovered biomarker panel (sensitivity and specificity 100%; Figs. 4E, 4F). Also, strong discrimination between DRYaqlip and DRYlip patients could be achieved using the marker proteins (sensitivity 90%; specificity 80%; Fig. 4C). For discrimination of DRYaq and DRYlip, a 100% correct classification could be achieved (Fig. 4D). DRYaq and DRYaqlip patients could be discriminated with a sensitivity and specificity of only >70%, reflecting the similar tear protein patterns of these patient groups (Fig. 4B). 
Figure 4. 
 
ROC curves depicting the classification power of discovered biomarkers.
Figure 4. 
 
ROC curves depicting the classification power of discovered biomarkers.
Discussion
In this study, we could demonstrate for the first time that different clinical phenotypes of dry eye patients are associated with specific tear protein patterns, and that patients suffering from an aqueous deficiency considerably deviate from those with a deficit of the lipid layer. 
Today, proteome analysis of body fluids by use of mass spectrometry techniques is becoming more and more important with regard to the detection and development of reliable disease markers and their utilization for diagnostic purposes. Especially in the case of tears, mass spectrometry represents a powerful tool for pattern recognition and comparative quantification of proteins due to the relatively low complexity of the tear proteome. In the present study we could confirm the so far detected deviations of protein patterns from dry eye patients when compared to healthy subjects by use of SELDI-TOF MS. Specific candidate markers were found to be significantly decreased (PRR4) or increased (e.g., calgranulin S100A8) in dry eye patients—alterations that have already been observed in previous studies by our group and others. 20,23 The fact that these markers could be detected in different laboratories by use of different methods as well as their confirmation in large patient cohorts and independent validation studies, as demonstrated in the present work, strongly substantiates their reliability and their usefulness in terms of clinical diagnostics. However, up to now most studies have analyzed the protein patterns of dry eye patients in general, without consideration of the different pathological or etiological aspects of dry eye disease. 20,23 Recently, two studies focused on dry eye subgroups, evaporative dry eye 33 and MGD, 29 in comparison to healthy subjects, but none have compared the tear protein expression levels of patients suffering from different dry eye subtypes with each other. As demonstrated in the present study, especially a deficiency of the aqueous phase of the tear film seems to strongly influence the expression patterns of several proteins as observed for the DRYaq and DRYaqlip groups. Thereby both increased (e.g., mammaglobin B, lipophilin A) and decreased protein levels (PRR4 fragments at 4043 Da and 4078 Da) could be detected in these patients when compared to healthy or DRYlip patients. In contrast, an impaired integrity of the lipid layer, caused by meibomian gland disease for example, has only a minor impact on the protein composition of the tear film. For all candidate markers detected in this study, merely slight deviations from measured protein intensities of the CTRL group could be observed in DRYlip patients, and most of them revealed no statistical significant difference when compared to healthy subjects. With regard to the more than 90 proteins reported to be secreted by meibomian glands, 16 this finding is remarkable, since it is expected that inadequate secretion will lead to significant alterations of the tear film proteome. The latest findings from Tong et al. revealed an altered protein concentration of calgranulin S100A8 in MGD patients 29 —a result that coincides with findings from study I of this work but that could not be confirmed in study II. However, the “discrimination pattern” generated by discriminant analysis via canonical roots for the different dry eye subgroups in the present study suggests the assumption that only slight differences between healthy and DRYlip patients occur. The discrimination accuracy of 100% for differentiation of study subjects from the DRYlip and the DRYaq groups, achieved by use of the developed biomarker panel, further reflects the discrepancy between these dry eye subgroups. Therefore, we suppose that the underlying pathomechanisms of these patients are based on proteome-independent events and parameters, which have only a minor, potentially epiphenomenal influence on the protein composition of tears. Another explanation may be that applied methods and technologies are insufficient and not sensitive enough to detect potentially occurring deviations in the tear proteome of DRYlip patients. 
In this study we detected several strong candidates for biomarkers enabling the classification of dry eye patients with differing underlying disease mechanism. However, in terms of protein biomarker profiling, besides the diagnostic potential of detected markers, their biological functions and especially their impact on and relevance to disease pathologies are also of great interest. Furthermore, their characteristics under medical treatment represent important information concerning the monitoring of treatment success based on an objective biochemical parameter. For the glycoprotein PRR4, which has been described as a product of the lacrimal gland, 34 a role in the protection of the ocular surface was supposed. 35 However, definite, experimentally based proof for this assertion is still missing. Members of the secretoglobin family, which were found to be increased in tears of dry eye patients as well as in those with contact lens–related dry eye, 32,36 were associated with anti-inflammatory and immunomodulatory processes. 37 Likewise, calgranulin A, a member of the calcium-binding S100 protein family, seems to be involved in immunological processes, especially in association with neutrophilic granulocytes. 38,39 Nevertheless, the actual biological function of the different candidate protein markers detected in tear film–associated diseases is not yet understood. Important information for the understanding of disease-specific pathomechanisms, such as potential protein or lipid interaction partners of markers, is still missing, as well as their expression characteristics under medical treatment. With regard to an optimized diagnosis and treatment strategy for patients suffering from dry eye, this information is urgently needed, and a more detailed characterization of these fundamental processes needs to be the subject of future studies. In summary, the present study demonstrates that dry eye patients with different etiologies distinctly deviate from each other with respect to the quantities of specific proteins of the tear film proteome. 
Supplementary Materials
References
DEWS. Research in dry eye: report of the Research Subcommittee of the International Dry Eye WorkShop (2007). Ocul Surf . 2007; 5: 179–193. [CrossRef] [PubMed]
Lemp MA. Report of the National Eye Institute/Industry workshop on Clinical Trials in Dry Eyes. CLAO J . 1995; 21: 221–232. [PubMed]
Schein OD Munoz B Tielsch JM Bandeen-Roche K West K. Prevalence of dry eye among the elderly. Am J Ophthalmol . 1997; 124: 723–728. [CrossRef] [PubMed]
Jie Y Xu L Wu YY Jonas JB. Prevalence of dry eye among adult Chinese in the Beijing Eye Study. Eye (Lond) . 2009; 23: 688–693. [CrossRef] [PubMed]
Guo B Lu P Chen X Zhang R Chen W. Prevalence of dry eye disease in Mongolians at high altitude in China: the Henan eye study. Ophthalmic Epidemiol . 2010; 17: 234–241. [CrossRef] [PubMed]
Lee AJ Lee J Saw SM Prevalence and risk factors associated with dry eye symptoms: a population based study in Indonesia. Br J Ophthalmol . 2002; 86: 1347–1351. [CrossRef] [PubMed]
Khanal S Tomlinson A Diaper CJ. Tear physiology of aqueous deficiency and evaporative dry eye. Optom Vis Sci . 2009; 86: 1235–1240. [CrossRef] [PubMed]
Sullivan BD Whitmer D Nichols KK An objective approach to dry eye disease severity. Invest Ophthalmol Vis Sci . 2010; 51: 6125–6130. [CrossRef] [PubMed]
Versura P Profazio V Campos EC. Performance of tear osmolarity compared to previous diagnostic tests for dry eye diseases. Curr Eye Res . 2010; 35: 553–564. [CrossRef] [PubMed]
Suzuki M Massingale ML Ye F Tear osmolarity as a biomarker for dry eye disease severity. Invest Ophthalmol Vis Sci . 2010; 51: 4557–4561. [CrossRef] [PubMed]
Lam H Bleiden L de Paiva CS Tear cytokine profiles in dysfunctional tear syndrome. Am J Ophthalmol . 2009; 147: 198–205, e1. [CrossRef] [PubMed]
Massingale ML Li X Vallabhajosyula M Analysis of inflammatory cytokines in the tears of dry eye patients. Cornea . 2009; 28: 1023–1027. [CrossRef] [PubMed]
Boehm N Riechardt AI Wiegand M Pfeiffer N Grus FH. Pro-inflammatory cytokine profiling of tears from dry-eye patients by means of antibody-microarrays. Invest Ophthalmol Vis Sci . 2011; 52: 7725–7730. [CrossRef] [PubMed]
Lei Z Beuerman RW Chew AP Quantitative analysis of N-linked glycoproteins in tear fluid of climatic droplet keratopathy by glycopeptide capture and iTRAQ. J Proteome Res . 2009; 8: 1192–2003.
Tomosugi N Kitagawa K Takahashi N Sugai I Ishikawa S. Diagnostic potential of tear proteomic patterns in Sjogren's syndrome. J Proteome Res . 2005; 4: 820–825. [CrossRef] [PubMed]
Tsai PS Evans JE Green KM Proteomic analysis of human meibomian gland secretions. Br J Ophthalmol . 2006; 90: 372–377. [CrossRef] [PubMed]
Zhou L Beuerman RW Ang LP Elevation of human alpha-defensins and S100 calcium-binding proteins A8 and A9 in tear fluid of patients with pterygium. Invest Ophthalmol Vis Sci . 2009; 50: 2077–2086. [CrossRef] [PubMed]
Li K Liu X Chen Z Huang Q Wu K. Quantification of tear proteins and sPLA2-IIa alteration in patients with allergic conjunctivitis. Mol Vis . 2010; 16: 2084–2091. [CrossRef] [PubMed]
Herber S Grus FH Sabuncuo P Augustin AJ. Two-dimensional analysis of tear protein patterns of diabetic patients. Electrophoresis . 2001; 22: 1838–1844. [CrossRef] [PubMed]
Zhou L Beuerman RW Chan CM Identification of tear fluid biomarkers in dry eye syndrome using iTRAQ quantitative proteomics. J Proteome Res . 2009; 8: 4889–4905. [CrossRef] [PubMed]
Ohashi Y Ishida R Kojima T Abnormal protein profiles in tears with dry eye syndrome. Am J Ophthalmol . 2003; 136: 291–299. [CrossRef] [PubMed]
Danjo Y Lee M Horimoto T Hamano K. Ocular surface damage and tear lactoferrin in dry eye syndrome. Acta Ophthalmol (Copenh) . 1994;72: 433–437.
Grus FH Podust VN Bruns K SELDI-TOF-MS ProteinChip array profiling of tears from patients with dry eye. Invest Ophthalmol Vis Sci . 2005; 46: 863–876. [CrossRef] [PubMed]
Green-Church KB Nichols KK Kleinholz NM Zhang L Nichols JJ. Investigation of the human tear film proteome using multiple proteomic approaches. Mol Vis . 2008; 14: 456–470. [PubMed]
Saijyothi AV Angayarkanni N Syama C Two dimensional electrophoretic analysis of human tears: collection method in dry eye syndrome. Electrophoresis . 2010; 31: 3420–3427. [CrossRef] [PubMed]
Hoh H Schirra F Kienecker C Ruprecht KW. Lid-parallel conjunctival folds are a sure diagnostic sign of dry eye [in German]. Ophthalmologe . 1995; 92: 802–808. [PubMed]
Foulks GN Bron AJ. Meibomian gland dysfunction: a clinical scheme for description, diagnosis, classification, and grading. Ocul Surf . 2003; 1: 107–126. [CrossRef] [PubMed]
DEWS. The definition and classification of dry eye disease: report of the Definition and Classification Subcommittee of the International Dry Eye WorkShop (2007). Ocul Surf . 2007; 5: 75–92. [CrossRef] [PubMed]
Tong L Zhou L Beuerman RW Zhao SZ Li XR. Association of tear proteins with Meibomian gland disease and dry eye symptoms. Br J Ophthalmol . 2010; 95: 848–852. [CrossRef] [PubMed]
Grus FH Augustin AJ. Analysis of tear protein patterns by a neural network as a diagnostical tool for the detection of dry eyes. Electrophoresis . 1999; 20: 875–880. [CrossRef] [PubMed]
Grus FH Zimmermann CW. Identification and classification of autoantibody repertoires (Western blots) with a pattern recognition algorithm by an artificial neural network. Electrophoresis . 1997; 18: 1120–1125. [CrossRef] [PubMed]
Kramann C Boehm N Lorenz K Effect of contact lenses on the protein composition in tear film: a ProteinChip study. Graefes Arch Clin Exp Ophthalmol . 2011; 249: 233–243. [CrossRef] [PubMed]
Versura P Nanni P Bavelloni A Tear proteomics in evaporative dry eye disease. Eye (Lond) . 2010; 24: 1396–1402. [CrossRef] [PubMed]
Dickinson DP Thiesse M. A major human lacrimal gland mRNA encodes a new proline-rich protein family member. Invest Ophthalmol Vis Sci . 1995; 36: 2020–2031. [PubMed]
Fung KY Morris C Sathe S Sack R Duncan MW. Characterization of the in vivo forms of lacrimal-specific proline-rich proteins in human tear fluid. Proteomics . 2004; 4: 3953–3959. [CrossRef] [PubMed]
Nichols JJ Green-Church KB. Mass spectrometry-based proteomic analyses in contact lens-related dry eye. Cornea . 2009; 28: 1109–1117. [CrossRef] [PubMed]
Lehrer RI Xu G Abduragimov A Lipophilin, a novel heterodimeric protein of human tears. FEBS Lett . 1998; 432: 163–167. [CrossRef] [PubMed]
Kerkhoff C Klempt C Sorg M. Novel insights into structure and function of MRP8 (S100A8) and MRP14 (S100A9). Biochim Biophys Acta . 1998; 1448: 200–211. [CrossRef] [PubMed]
Roth J Vogl T Sorg C Sunderkotter C. Phagocyte-specific S100 proteins: a novel group of proinflammatory molecules. Trends Immunol . 2003; 24: 155–158. [CrossRef] [PubMed]
Footnotes
 Supported in part by Alcon Laboratories. The funding organization participated in the design of the study. The authors alone are responsible for the content and writing of the paper.
Footnotes
 Disclosure: N. Boehm, None; S. Funke, None; M. Wiegand, None; N. Wehrwein, None; N. Pfeiffer, None; F.H. Grus, None
Figure 1. 
 
Shown are box plots for detected biomarkers at 4043 Da and 4078 Da. Both markers were identified as proline-rich protein 4. Diagrams on the left exhibit data from study I; those on the right are from study II. The y-axis represents the relative, normalized intensities. Boxes represent ME ± SE, rectangles the ME ± SD.
Figure 1. 
 
Shown are box plots for detected biomarkers at 4043 Da and 4078 Da. Both markers were identified as proline-rich protein 4. Diagrams on the left exhibit data from study I; those on the right are from study II. The y-axis represents the relative, normalized intensities. Boxes represent ME ± SE, rectangles the ME ± SD.
Figure 2. 
 
Shown are box plots for detected biomarkers at 10,117 Da (mammaglobin B), 10,234 Da (lipophilin A), 10,866 Da (calgranulin S100A8), and 12,729 Da (beta-2 microglobulin precursor). Diagrams on the left exhibit data from study I; those on the right are from study II. The y-axis represents the relative, normalized intensities. Boxes represent ME ± SE, rectangles the ME ± SD.
Figure 2. 
 
Shown are box plots for detected biomarkers at 10,117 Da (mammaglobin B), 10,234 Da (lipophilin A), 10,866 Da (calgranulin S100A8), and 12,729 Da (beta-2 microglobulin precursor). Diagrams on the left exhibit data from study I; those on the right are from study II. The y-axis represents the relative, normalized intensities. Boxes represent ME ± SE, rectangles the ME ± SD.
Figure 3. 
 
Depicted are the canonical roots for patients enrolled in studies I and II. They demonstrate the similarities/differences between the protein patterns of single individuals. Blue circles: CTRL; red boxes: DRYaq; green boxes: DRYlip; black boxes: DRYaqlip.
Figure 3. 
 
Depicted are the canonical roots for patients enrolled in studies I and II. They demonstrate the similarities/differences between the protein patterns of single individuals. Blue circles: CTRL; red boxes: DRYaq; green boxes: DRYlip; black boxes: DRYaqlip.
Figure 4. 
 
ROC curves depicting the classification power of discovered biomarkers.
Figure 4. 
 
ROC curves depicting the classification power of discovered biomarkers.
Table 1
 
Clinical Data of Patients Enrolled in Study II
Table 1
 
Clinical Data of Patients Enrolled in Study II
Age ± SD BST ± SD, mm LIPCOF ± SD TBUT ± SD, s Fluorescein ± SD, Oxford Scale Lissamine Green ± SD, Oxford Scale Score for Complaints of Pain and Discomfort DEWS Score ± SD
CTRL 51.21 ± 16.61 21.06 ± 5.91 1.61 ± 0.83 14.82 ± 0.58 NS NS NS NS
DRYaq 52.83 ± 17.20 6.69 ± 2.71 2.57 ± 0.82 14.51 ± 1.15 1.14 ± 1.03 1.26 ± 1.07 9.40 ± 3.14 1.69 ± 0.37
DRYlip 53.09 ± 15.30 18.81 ± 6.11 2.60 ± 0.90 9.02 ± 1.88 1.04 ± 1.00 1.12 ± 0.98 16.84 ± 4.47 1.54 ± 0.40
DRYaqlip 55.41 ± 14.72 5.00 ± 2.94 2.71 ± 0.78 8.20 ± 2.16 1.54 ± 0.95 1.67 ± 1.02 19.54 ± 3.85 2.13 ± 0.47
Table 2
 
Clinical Data of Patients Enrolled in Study II: Post Hoc Values: Age
Table 2
 
Clinical Data of Patients Enrolled in Study II: Post Hoc Values: Age
CTRL DRYaq DRYlip DRYaqlip
CTRL 0.975 0.956 0.670
DRYaq 0.975 0.999 0.894
DRYlip 0.956 0.999 0.908
DRYaqlip 0.670 0.894 0.908
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×