Identification of Tear Fluid Biomarkers in Dry Eye Syndrome Using iTRAQ Quantitative Proteomics Lei Zhou,†,‡ Roger W. Beuerman,*,†,‡,O Choi Mun Chan,§ Shao Zhen Zhao,⊥ Xiao Rong Li,⊥ He Yang,| Louis Tong,†,§ Shouping Liu,† Michael E. Stern,∇ and Donald Tan†,‡,§ Singapore Eye Research Institute, Singapore, Department of Ophthalmology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore National Eye Centre, Singapore, Singapore Immunology Network, A*STAR, Singapore, Tianjin Medical University Eye Center, Tianjin, China, and Allergan Pharmaceuticals, Irvine, California 92612 Received August 14, 2008
The proteins found in tears have an important role in the maintenance of the ocular surface and changes in the quality and quantity of tear components reflect changes in the health of the ocular surface. In this study, we have used quantitative proteomics, iTRAQ technology coupled with 2D-nanoLC-nanoESI-MS/MS and with a statistical model to uncover proteins that are significantly and reliably changed in the tears of dry eye patients in an effort to reveal potential biomarker candidates. Fifty-six patients with dry eye and 40 healthy subjects were recruited for this study. In total, 93 tear proteins were identified with a ProtScore g2 (g99% confidence). Associated with dry eye were 6 up-regulated proteins, R-enolase, R-1-acid glycoprotein 1, S100 A8 (calgranulin A), S100 A9 (calgranulin B), S100 A4 and S100 A11 (calgizzarin) and 4 down-regulated proteins, prolactin-inducible protein (PIP), lipocalin-1, lactoferrin and lysozyme. Receiver operating curves (ROC) were evaluated for individual biomarker candidates and a biomarker panel. With the use of a 4-protein biomarker panel, the diagnostic accuracy for dry eye was 96% (sensitivity, 91.0%; specificity, 90.0%). Two biomarker candidates (R-enolase and S100 A4) generated from iTRAQ experiments were successfully verified using an ELISA assay. The levels of these 10 tear proteins reflect aqueous secretion deficiency by lacrimal gland, inflammatory status of the ocular surface. The clinical classification of the severity of the dry eye condition was successfully correlated to the proteomics by using three proteins that are associated with inflammation, R1-acid glycoprotein 1, S100 A8 and S100 A9. The nine tear protein biomarker candidates (except R1-acid glycoprotein 1) were also verified using an independent age-matched patient sample set. This study demonstrated that iTRAQ technology combined with 2D-nanoLC-nanoESI-MS/MS quantitative proteomics is a powerful tool for biomarker discovery. Keywords: Dry eye • biomarker • tear proteomics • iTRAQ • quantitative proteomics
Introduction Dry eye syndrome can be defined as an abnormal tear film from either tear secretory deficiency or excessive tear evaporation and it is associated with various amounts of patient discomfort.1 Dry eye syndrome is a worldwide clinical problem which affects millions of people with a prevalence estimated to be as high as 11-22% of the general population.2 However, in Asia, the incidence of dry eye appears to be greater than in the West and the incidence has been found to be in the range of 20-24% in * Correspondence: Prof. Roger W Beuerman, Singapore Eye Research Institute, 11 Third Hospital Avenue, Singapore 168751, E-mail: [email protected]
mac.com. † Singapore Eye Research Institute. ‡ National University of Singapore. § Singapore National Eye Centre. | Singapore Immunology Network. ⊥ Tianjin Medical University Eye Center. ∇ Allergan Pharmaceuticals. O SPR in Neuroscience and Behavioural Disorders, Duke-NUS Graduate Medical School, Singapore. 10.1021/pr900686s CCC: $40.75
2009 American Chemical Society
different Asian countries.3,4 It is more common in people over 55 years of age and in females.5 The prevalence is also significantly higher in visual display terminal users and contact lens wearers.6 Other environmental factors7 such as workplace environment (indoor/outdoor), climate (altitude and humidity), exposure to chemicals (volatiles, air pollution) and dust may also have an influence on the prevalence of dry eye syndrome. Normal tear flow is regulated by a complex relationship between the ocular surface and components of the central and peripheral nervous systems which control secretory activity by the lacrimal gland.8,9 The causes of dry eye syndrome are diverse; however, a common component is dysfunction in what is called the “functional unit”, a neural loop that controls tear flow from the lacrimal gland onto the surface of the eye resulting in ocular surface tear deficiency and inflammation.10 Usually, dry eye can be classified into two major categories: tear secretion deficiency and excessive tear evaporation.7 Dysfunction of the lipid layer of the tear film leads to excessive evaporation of tears. Journal of Proteome Research 2009, 8, 4889–4905 4889 Published on Web 08/25/2009
research articles Diagnosis of dry eye has been difficult and the development of new phamacological therapies is hampered by the lack of objective tests for response outcomes.2,11,12 Diagnosis of dry eye syndrome or as it is known formally, keratitis sicca, is typically based on subjective symptoms, Schirmer’s test (evaluating the quantity of tear fluid), tear breakup time (evaluating the quality of tear film) and other clinical tests including scoring the staining of the cornea and conjunctiva epithelium by disclosing dyes. Studies have confirmed a poor correlation between clinical tests and subjective symptoms and even between different clinical tests.2,11 Tears are a complex fluid mixture of proteins, lipids, salts, mucin and other small organic molecules.13 The aqueous secretion of lacrimal gland origin is the major source of tear proteins and fluid. Contributions from ocular surface tissues secretions and the vasculature are noted as well. Proteomic analysis of tear fluids of individuals is still in its infancy as the typical sample volume of tear fluid is in the order of microliters. However, early exploration of tear proteomics by several groups including ours has provided some promising results in which changes in tear proteomics are correlated with certain conditions such as ocular surface wounding14-16 and inflammatory disease.17-19 Many technologies have been employed for analyzing tear proteins including 1D-gel,20 2D-gel,19,21,22 LC-MS,14 SELDI,15-18 MALDI,23,24 LC-MS/MS,25-27 LC-MALDI25 and protein arrays.28 Our objective was to correlate the changes of tear protein profiles with dry eye and potentially use them as diagnostic biomarkers. In this study, a quantitative proteomics method, that is, iTRAQ technology29-31 with two-dimensional nanoLCnano-ESI-MS/MS, and a statistical model was used to develop candidate biomarkers and a potential biomarker panel for application to dry eye disease.
Experimental Procedures Chemicals. HPLC grade acetonitrile, water, acetic acid and methanol were purchased from Fisher Scientific (Pittsburgh, PA). Formic acid, trifluoroacetic acid, ammonium bicarbonate, ammonium acetate and phosphate-buffered saline (PBS) were purchased from Sigma (St. Louis, MO). Patients. In total, 56 patients with dry eye and 40 healthy subjects were recruited for the whole study (Table 1). For the discovery stage using iTRAQ, there were 28 patients (7 males and 21 females; average age, 60; age range, 40-84) diagnosed with dry eye syndrome (routine patients seen in the dry eye clinic at Singapore National Eye Center by Drs. C. M. Chan and D. Tan) and 20 control subjects (7 males and 13 females; average age, 37; age range, 21-53). For verification, 28 patients with dry eye and 20 healthy subjects were obtained from an independent study site, Tianjin Medical University Eye Center, China by Drs. S. Z. Zhao and X. R. Li: (a) verification using ELISA, 8 patients with dry eye (1 male and 7 females, average age: 50) and 8 controls (4 males and 4 females, average age: 26); (b) dry eye group with age-matched control group, 15 patients with dry eye (7 males and 8 females; average age, 26; age range, 19-33; 1 sample used here was from the ELISA dry eye group) and 18 age-matched controls (8 males and 10 females; average age, 24; age range, 18-32; 6 samples used here were from ELISA control group); (c) additional 8 patients with dry eye from older age group (1 males and 7 females; average age, 55; age range, 45-72; 2 sample used here was from the ELISA dry eye group). Informed consent was obtained from all participating subjects and the procedure was approved by 4890
Journal of Proteome Research • Vol. 8, No. 11, 2009
Zhou et al. the Institutional Review Board of the Singapore Eye Research Institute and Tianjin Medical University Eye Center and to the tenets of the Declaration of Helsinki. Clinical examinations included subjective symptoms, Schirmer’s test type I (without anesthesia), tear breakup time (TBUT), and other general ophthalmic examinations such as visual acuity and lid margin and Meibomian gland status. Typical verbal complaints originated with only the patients of the dry eye group and included burning, itching and stinging, foreign body sensation, sense of dryness, blurring of vision, photophobia, pain and heavy or tired eyes. Patients were classified as dry eye based on the Schirmer’s test, TBUT, and subjective symptoms. Dry eye was defined as having dry eye symptoms with at least one abnormal finding out of three objective clinical tests (Schirmer type I test e10 mm at 5 min without anesthetic, tear breakup time e10 s, or corneal fluorescein staining higher than 2 using the Oxford scheme). The controls had no dry eye symptoms, Schirmer test >15 mm at 5 min without anesthetic, tear breakup time >10 s, no corneal fluorescein staining or sensations of discomfort. The details of patients’ clinical information are given in Table 1. None of the patients in the dry eye group had symptoms of Sjorgen’s Syndrome, which is an autoimmune disease.18 Tear Collection and Tear Protein Elution. Tear fluid for patients was collected using the standard clinical Schirmer’s strip. After collection, Schirmer’s strips were immediately frozen at -80 °C until analyzed. The first 10 mm of the Schirmer’s strip was cut into small pieces and soaked in 150 µL of phosphate-buffered saline (PBS) for 3 h to elute tear proteins. Total tear protein concentration of each sample was measured using a Micro BCA Protein Assay Kit (Pierce Biotechnology, Inc.). Study Design and iTRAQ Sample Preparation. Since iTRAQ technology allows labeling of four samples simultaneously, in the discovery stage, 2 control samples (C1 and C2) and 2 dry eye samples (DE1 and DE2) were used in each set (total 14 sets from 28 dry eye samples, control samples may be used more than once, i.e., used in pooled control and used as an individual control, see Figure 1 for the iTRAQ experimental design). In these 14 sets, individual control samples were used in 9 sets (Set 6 ∼ Set 14, Set 6 was analyzed twice) and pooled control samples (one pooled from 5 controls and one pooled from 3 controls) were used in another 5 sets of experiments (Set 1 ∼ Set 5, Set 4 was analyzed twice). Each iTRAQ experiment used 30 µg of protein from each sample and the iTRAQ procedure was followed according to the protocol provided by Applied Biosystems (Foster City, CA). Briefly, 20 µL of Dissolution Buffer (triethylammonium bicarbonate) and 1 µL of Denaturant (2% SDS) from the iTRAQ reagent kit were added to the samples after which 2 µL of Reducing Reagent [tris-(2-carboxyethyl)phosphine] was added and the samples were incubated at 60 °C for 1.5 h. Cysteine Blocking Reagent, 1 µL (methyl methanethiosulfonate), was added and the mixture was incubated at room temperature for 20 min. The samples were digested at 37 °C overnight with trypsin (included in the iTRAQ kit). iTRAQ reagents 114 and 115 were added to the control samples, while iTRAQ reagents 116 and 117 were added to the dry eye samples. The samples were then incubated at room temperature for 3 h. The contents of each iTRAQ reagent labeled sample were combined and dried using a SpeedVac after which 10 µL of loading buffer (0.1% formic acid, 2% acetonitrile (ACN) in water) was added to reconstitute the sample before running the 2D nano-LC-nano-ESI-MS/MS analysis.
Quantitative Biomarkers for Dry Eye a
Table 1. Characteristics of Patients and Controls sex
Control (n ) 20) Dry eye (n ) 28) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 Control (n ) 18) Dry eye (n ) 23) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Control (n ) 8)c Dry eye (n ) 8)d 1 2 3 4 5
Schirmer’s test (length, mm)
1. Discovery 7 males, 13 females 7 males, 21 females F F M F F F F F F F F F F F M F M F F F M F M F F M M F 8 males, 10 females 8 males, 15 females F F F M F F F F M M M M F F M F M F F F M F F 4 males, 4 females 1 males, 7 females F F F F F
Average: 37 Range: 21-53 Average: 60 Range: 40-84 59 63 49 74 44 71 58 66 84 60 45 68 62 45 46 77 64 68 50 49 60 70 84 62 56 40 45 47
3 8 5 7 5 1 4 4 8 4 14b 2 4 8 5 4 5 7 2 2 3 7 9 7 14b 4 7 35b
2. Verification Average: 24 >15 Range: 18-32 Average: 36 Range: 19-72 72 6 58 12 57 1 55 1 54 3 50 6 50 1 45 13b 33 8 33 17b 32 1 31 2 30 5 27 10 26 4 25 5 24 5 23 1 23 6 22 2 22 5 19 4 19 7 3. Verification by ELISA Average: 26 >15 Range: 18-32 Average: 50 Range: 19-72 51 1 45 1 66 1 45 5 55 2
tear breakup time (s)
3 2 7 1 4 4 10 4 4 10 6 4 3 2 2 1 2 1 1 1 3 3 4 1 2 3 1 1
N N N N Y Y Y N N N N N N N N N N N N N N N N N N Y N Y
1 6 2 2 4 5 3 5 3 6 2.6 3 4 4 10 3 1.6 1.5 4 4.4 6 2.3 2.6 >10
3 1.8 4 3.5 2.7
Y Y Y Y Y N Y Y N Y Y N N Y Y Y Y Y Y N N Y Y N
Y Y Y Y Y
a In verification, patients nos. 9∼23 (n ) 15, average age ) 26) were used for dry eye group comparing with age-matched control group (n ) 18, average age ) 24) experiment. b May be reflex tearing. c Six samples are from 2. Verification, Control group. d The other 3 samples used for ELISA are from 2. Verification, Dry eye group, patients nos. 1, 5, and 10.
In a second series of iTRAQ experiments for assessing age effects, 18 control tear samples were pooled to form a global control and labeled with iTRAQ reagent 114 (Figure 1). iTRAQ 115, 116, and 117 were used to label three individual dry eye samples (total 8 sets of iTRAQ experiments). The advantage of using global pooled controls is that it formed a common reference for comparison. The other sample preparation procedures were exactly the same as the above.
Two-Dimensional Nano-LC-nano-ESI-MS/MS Analysis. Twodimensional nanoLC (DIONEX, LC Packings, Sunnyvale, CA) coupled with nano-ESI-MS/MS (Applied Biosystems, Q-Star XL, MDS Sciex, Concord, Ontario, Canada) was used for the analysis. The two-dimensional LC separation of peptides used in this study was a configuration which consisted of a strong cation exchange (SCX) followed by reverse phase (RP) chromatography. The SCX column used in the first dimension was Journal of Proteome Research • Vol. 8, No. 11, 2009 4891
Figure 1. Experimental design. (1) Discovery phase and (2) verification phase.
from DIONEX, LC Packings (300 µm i.d. × 10 cm porosity 10S SCX). Elution of the peptide mixture was performed using 10 steps of salt plug (20 µL injection of 10, 20, 30, 40, 50, 75, 100, 250, 500, and 1000 mM ammonium acetate) all at a flow rate of 30 µL/min and using a loading solvent of 0.1% formic acid/ ACN (95:5, v/v). The RP column used in the second dimension was a 10 cm × 75 µm i.d. microcapillary LC column self-packed from a self-pack PicoFrit (360 µm o.d., 75 µm i.d., 50 cm, New Objectives, Woburn, MA). This capillary column was packed with Luna C18, 3 µm, 100 Å from Phenomenex (Torrance, CA) using a homemade column packing device. The capillary column had an integrated spray tip (15 µm opening) which could be directly coupled with the nanospray interface (Protana, Odense, Denmark) into ABI’s Q-TOF mass spectrometer. After SCX separation, the sample was loaded onto a trapping cartridge (C18, 0.3 × 5 mm, from DIONEX, LC Packings) from a Famos autosampler (DIONEX, LC Packings) at 30 µL/min for desalting. After a 5 min wash with ACN/water (2/98, v/v with 0.1% formic acid), the system was switched (Switchos, DIONEX, LC Packings) into line with the RP analytical capillary column. With the use of an Ultimate solvent delivery system (DIONEX, LC Packings), a linear gradient of ACN (0.1% formic acid) from 20% to 95% over 85 min at flow rate of ∼300 nL/min was used to analyze the tryptic digests. Key parameter settings for the nanospray and other instrumentation were as follows: ionspray voltage (IS) ) 2200 V, curtain gas (CUR) ) 20, declustering potential (DP) ) 60 V, focusing potential (FP) ) 265 V, collision gas setting (CAD) ) 5 for nitrogen gas, DP2 ) 15. All data was acquired using information-dependent acquisition (IDA) mode with Analyst QS software (Applied Biosystems). For IDA parameters, 1 s TOF MS survey scan in the mass range of 300-1200 were followed by two product ion scans of 3 s each in the mass range of 100-1500. The “enhance all” function was used in the IDA experiments. Switching criteria were set to ions greater than m/z ) 350 and smaller than m/z ) 1200 with charge state of 2-4 and an abundance threshold of >20 counts. Former target ions were excluded for 60 s. IDA collision energy (CE) parameters script was used for automatically controlling the CE. Data Analysis. Data analysis for the iTRAQ experiments was performed using ProQUANT 1.0, together with ProGroup View 1.0 (Applied Biosystems) and searched against the IPI (International Protein Index) human database v3.15. The mass tolerance set for peptide identification in ProQUANT searches were 0.15 Da for MS and 0.10 Da for MS/MS, respectively. The cutoff for the confidence settings was at 75. Other search parameters include MMTS (methyl methanethiosulfonate) as the cysteine fixed modification, 1 missed trypsin cleavage site, 4892
Journal of Proteome Research • Vol. 8, No. 11, 2009
Zhou et al. oxidation of methionine in the zone modifications and custom amino acids with iTRAQ modification to lysine and tyrosine. The results generated by ProQUANT were then analyzed and summarized by Pro Group Viewer 1.0 (Applied Biosystems) to produce a Pro Group Report. The protein identification criteria were as follows: (a) proteins with ProtScore >2 (>99% confidence) were accepted; (b) proteins with ProtScore ) 2 () 99% confidence), typically one peptide with confidence of 99%, was accepted with manually verified MS/MS spectra. For differential expression, a bias correction factor was applied for correcting possible pipetting error during the combination of differentially labeled samples. Relative quantification of proteins using iTRAQ technology is based on the ratio of peak areas of m/z 114, 115, 116, and 117 from MS/MS spectra. For example, since 2 controls (C1 and C2) were tagged with 114 and 115 and 2 dry eye samples (DE1 and DE2) were tagged with 116 and 117, the relative quantitative result can be expressed as DE1:C1, DE2:C1, and C2:C1 using ratios. For quantitative results, in addition to reporting ratios, Pro Group also gives a p-value and Error Factor (EF). A smaller p-value (on a scale of 0 to 1) suggests more confidence about altered expression levels (nonunity ratios, for example, diseased sample vs control sample) resulting from a real biological difference. The EF expresses the 95% uncertainty range for a reported ratio. The true protein ratio is expected to be found between the (reported ratio)/(EF) and the (reported ratio) × (EF) 95% of the time16 (also refers to ProQUANT manual by Applied Biosystems). Statistical Analysis. To obtain an average of ratios for a particular protein from 14 sets of iTRAQ data, we used a weighted average calculation16 by involving EF (ProQUANT manual by Applied Biosystems). First, we converted the ratios to log space [Log(ratio)]. Then, after determining the log of the EF, the inverse of the log of the EF was used as the weight. The weighted average in log space was calculated using the following formula: Weighted average (Log space) ) Sum [log(ratio) × weight]/Sum (weight), where, weight ) 1/log EF. Finally, after converting back out of Log space, the “weighted average of the ratios” was found. Weighted standard deviations14 were also calculated using the method provided by ProQUANT manual (Applied Biosystems). Briefly, Weighted standard deviation ) Unweighted standard deviation/b0.5, where Unweighted standard deviation is the standard deviation of log ratios; b is the effective base, b ) (Sum(weight))2/Sum (Weight2), Weight ) 1/log EF. The final value will be converted out of log space. The ratios were divided into two groups, that is, dry eye group (DE1:C1, DE2:C1) and control group (C2:C1). Student t tests (independent) were performed to calculate the p-values to evaluate whether the changes between the dry eye group and the control group were significant (p < 0.05 was considered to be statistically significant). Proteins with significant changes in ratios were taken as biomarker candidates. We first used different classifiers (Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), and Factor discriminant Analysis (FDA)) involving all the candidate biomarkers and any combinations of them. However, the validation accuracy was less than 80%. To obtain better separation, we developed a two-step approach. First, we performed the LDA classification using one candidate at a time. The candidate with the best validation results was used as the biomarker in the first step. No misclassification was allowed in this step. The cutoff value for this step was then the median
Quantitative Biomarkers for Dry Eye value between the first misclassification and the last perfect classification. For the remaining samples, a linear classifier with the rest of candidate biomarkers and any combinations of them was used to perform the classification in the second step. To evaluate the best cutoff value for this step, the ROC curve was used. In this study, MATLAB (http://www.mathworks.com) was used to perform all computations. Chi square tests or binary logistic regression was performed to evaluate association between a categorical dependent variable (for example, a specific tear protein level above or below mean level) and a categorical independent variable (for example, age above or below mean age). Statistical significance was set at the level of alpha ) 0.05. Quantification of r-Enolase and S100 A4 by ELISA. The kit for the R-enolase ELISA was purchased from USCNLIFE Science & Technology Company (Missouri City, TX) and the S100 A4 ELISA kit was purchased from CycLex Co., Ltd., (Nagano, Japan). Tear samples of dry eye patients (n ) 8, 1 male and 7 females, average age: 50) and controls (n ) 8, 4 males and 4 females, average age: 26) were obtained from an independent study site, Tianjin Medical University Eye Center, China, by Drs. S. Z. Zhao and X. R. Li. All procedures, clinical assessments and criteria were the same as those in the section Patients. The assay was performed in duplicates according to the manufacturer’s instructions.
Results Tear Protein Identification. In total, 93 proteins were identified in human tear fluid from 14 sets of iTRAQ experiments. Among them, 63 proteins were identified using highly stringent criteria for protein identification (ProtScore > 2, >99% confidence level. Under this stringent threshold, typically no less than two peptides with high confidence were used for protein identification) and 30 proteins were identified with ProtScore ) 2 () 99% confidence) and combined with manual inspection of the MS/MS spectra (Table 2). In ProGroup View, each peptide can contribute no more than 2.0 to the ProtScore (equivalent to a confidence level not exceeding 99.0%). Relative Quantitation of Tear Proteins from Dry Eye Patients and Normal Controls by iTRAQ Technology Uncovers Significant Differences. Overall, statistical analysis showed that a total of 10 proteins were differentially expressed between the dry eye group and the normal control group with 6 proteins up-regulated and 4 proteins down-regulated in dry eye patients (Figure 2). The ratios, p-values and EF for the above 10 potential dry eye biomarkers from these 14 sets are listed in Supplemental Table 1. Up-regulated proteins included R-enolase, R-1-acid glycoprotein 1, S100 A8 (calgranulin A), S100 A9 (calgranulin B), S100 A4 and S100 A11 (calgizzarin) and the down-regulated proteins included prolactin-inducible protein (PIP), lipocalin-1, lactoferrin and lysozyme (Figure 3). Table 3 summarizes weighted average ratios, weighted standard deviations and p-values of these 10 proteins for the control group and dry eye group. Figures 3-5 show three representative MS/MS spectra for prolactin-inducible protein, R-enolase and R-1-acid glycoprotein 1, which give both identification and relative quantitation. Figure 3A shows the MS/MS spectrum of one triply charged peptide ion, TYLISSIPLQGAFNYJ. at m/z ) 701.70 (J represents the iTRAQ-modified lysine residue) which was found to be a sequence from prolactin-inducible protein. Both b-ion series and y-ion series were observed in the range of m/z ) 100-1200. Magnified MS/MS spectrum of 111.0 to 120.0 gave four iTRAQ
research articles reporter ions at m/z )114.1, 115.1, 116.1, and 117.1. Typically, iTRAQ 114 and 115 were used to label 2 control samples and iTRAQ 116 and 117 were used to label 2 dry eye samples. The relative quantitation of protein is based on the ratio of the peak areas of reporter ion m/z )114.1, 115.1, 116.1, and 117.1. From Figure 3B, down-regulation of prolactin-inducible protein in dry eye tears was clearly observed. Figures 4A and 5A show the MS/MS spectra of one triply charged peptide ion (SGETEDFIADLWGLCTGQIJ at m/z ) 877.40) which originated from R-enolase and one quadruply charged peptide ion (YVGGQEHFAHLLILR at m/z ) 475.00) which originated from R-1-acid glycoprotein 1. Similarly, zoom-in MS/MS spectra (Figures 4B and 5B) revealed overexpression of R-enolase and R-1-acid glycoprotein 1 in the tears of patients diagnosed with dry eye. ROC Curve for Potential Dry Eye Putative Biomarkers. Initially, a ROC curve was generated for each biomarker candidate individually. ROC curve plots true positive rate (sensitivity) versus false positive rate (1 - specificity) of various cutoff value for DE:C ratios. The area under the ROC curve was also calculated and provides information on accuracy. The ROC curve is useful to compare the performance of different tests. The best individual biomarker candidate among up-regulated proteins was R-enolase with an accuracy of 85% (sensitivity of 77.0% and specificity of 88.5%, if the cutoff value for the DE:C ratio equals to 1.2, Figure 6A), while prolactin-inducible protein is most useful among the down-regulated proteins with an accuracy of 81% (sensitivity of 71.7% and specificity of 87.1%, and a cutoff value for the DE:C ratio equal to 0.8, Figure 6B). The biomarker panel contained four proteins: R-enolase (ae), prolactin-inducible protein (PIP), lipocalin-1 (Lipo) and calgranulin B (CalB). R-Enolase was the sole biomarker in the first step. The cutoff value in this step (no misclassification) for (DE: C) is 1.70. For those samples which did not meet this criterion, the other three biomarkers were used with the following formula for further identification of dry eye cases: y ) log2 ratio(CalB) + log2 ratio(PIP) + 0.8 log2 ratio(Lipo) The cutoff value for y ranges from -0.4 to -3, depending on the permitted cutoff value for the false discovery rate. With this approach, a high sensitivity of 91% and specificity of 90% could be obtained, when the cutoff values for individual 4 proteins were as follows: 1.7 for R-enolase, and -0.4 to -3.0 for y value (Figure 6C). It should be noted that such performance could not be achieved if those biomarkers were used individually or if a one-step approach with a panel of biomarkers was used. However, using this approach, the accuracy (area under the ROC) for correctly classifying a sample as a dry eye patient was increased from, for example, 85% to 96% (Figure 6C). Verification of r-Enolase and S100 A4 by ELISA. To verify the iTRAQ findings, we measured R-enolase and S100 A4 concentrations in tears from an independent sample group of dry eye patients and controls using ELISA. ELISA results showed that the tear concentrations of R-enolase (t test, p < 0.038) and S100 A4 (t test, p < 0.027) were significantly higher in the dry eye group compared to the control group (Figure 7), which confirmed the iTRAQ results. The concentrations of R-enolase in tears from dry eye patients and controls were 25.16 ( 15.75 ng/mg total protein (n ) 8) and 10.40 ( 9.20 ng/mg total protein (n ) 8), respectively. The concentrations of S100 A4 in tears from dry eye patients and controls were Journal of Proteome Research • Vol. 8, No. 11, 2009 4893
Zhou et al.
Table 2. Summary of the Proteins Identified in Human Tear Fluid no.
1 2 3 4 5 6 7 8 9 10 11 12
IPI00000816 IPI00007427 IPI00022434 IPI00022429 IPI00020091 IPI00550991 IPI00553177 IPI00465248 IPI00218918 IPI00455315 IPI00008580 IPI00024284
13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
IPI00004656 IPI00027463 IPI00013895 IPI00007047 IPI00027462 IPI00021828 IPI00032294 IPI00305477 IPI00020487 IPI00479359 IPI00247167 IPI00027497 IPI00219757 IPI00641737 IPI00025512 IPI00022488 IPI00644694 IPI00642363 IPI00550702 IPI00647471 IPI00642193 IPI00641082 IPI00552445 IPI00549291 IPI00178926 IPI00382577 IPI00025023 IPI00298860 IPI00027444 IPI00465431 IPI00009650 IPI00001468 IPI00019038 IPI00026126 IPI00299547
48 49 50 51 52 53 54 55 56 57
IPI00419585 IPI00000874 IPI00004573 IPI00022974 IPI00009682 IPI00027019 IPI00032313 IPI00022463 IPI00164623 IPI00099110
ProtScore >2 (plus a few from ProtScore ) 2 that contain 2 peptides) 14-3-3 protein epsilon AGR2 (Anterior gradient protein 2 homologue) ALB protein (Serum albumin)a Alpha-1-acid glycoprotein 1 precursor Alpha-1-acid glycoprotein 2 precursor Alpha-1-antichymotrypsin precursor Alpha-1-antitrypsin precursor Alpha-enolase Annexin A1 Annexin A2a Antileukoproteinase 1 precursor Basement membrane-specific heparan sulfate proteoglycan core protein precursor Beta-2-microglobulin precursor Calcyclin (S100 calcium-binding protein A6) Calgizzarin (S100 calcium-binding protein A11)a Calgranulin A (S100 calcium-binding protein A8) Calgranulin B (S100 calcium-binding protein A9) Cystatin B Cystatin S precursor Cystatin SN precursor Extracellular glycoprotein lacritin precursor Ezrin F-box associated region domain containing protein Glucose-6-phosphate isomerase Glutathione S-transferase P Haptoglobina Heat-shock protein beta-1 Hemopexin precursor (Beta-1B-glycoprotein) Ig alpha-1 chain C regiona Ig alpha-2 chain C regiona Ig gamma-1 chain C regiona Ig gamma-2 chain C regiona Ig gamma-4 chain C regiona Ig kappa chain C regiona Ig lambda chain C regionsa IGHM proteina Immunoglobulin J chain Kappa 1 light chain variable region Lactoperoxidase precursor Lactotransferrin precursor Leukocyte elastase inhibitor LGALS3 protein (Galectin-3)a Lipocalin-1 precursor Lipophilin-A precursor Lysozyme C precursor Mammaglobin-B precursor (Lipophilin-C) Neutrophil gelatinase-associated lipocalin precursor (Lipocalin-2) Peptidyl-prolyl cis-trans isomerase Aa Peroxiredoxin-1a Polymeric-immunoglobulin receptor precursor Prolactin-inducible protein precursor Proline-rich protein 1 precursor Proline-rich protein 4 precursor S100 calcium-binding protein A4 Serotransferrin precursor Similar to Complement C3 precursor Splice Isoform 1 of Deleted in malignant brain tumors 1 protein precursora Splice Isoform 1 of Protein-glutamine gamma-glutamyltransferase 2a Submaxillary gland androgen-regulated protein 3 homologue A precursor (Proline-rich protein 5) Submaxillary gland androgen-regulated protein 3 homologue B
Journal of Proteome Research • Vol. 8, No. 11, 2009
no. of peptides identified
2 2 47 2 2 1b 9 9 2 2 2 4 2 3 2 9 11 2 7 2 7 2 2 2 3 7 2 5 13 10 9 4 2 6 8 3 6 2 5 57 2 2 14 3 15 8 2 2 3 16 11 5 7 3 17 4 3 1b 2 2
Quantitative Biomarkers for Dry Eye Table 2. Continued no.
no. of peptides identified
61 62 63
IPI00216298 IPI00299729 IPI00166729
precursor (Proline-rich protein 3) Thioredoxin Transcobalamin I precursor Zinc-alpha-2-glycoprotein precursor
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
IPI00013890 IPI00032220 IPI00329801 IPI00165421 IPI00020632 IPI00020599 IPI00017601 IPI00643853 IPI00400826 IPI00411704 IPI00023673 IPI00016862 IPI00657682 IPI00002469 IPI00217966 IPI00293276 IPI00020008 IPI00005721 IPI00220301 IPI00219446 IPI00026962 IPI00607698 IPI00013876 IPI00555956 IPI00218914 IPI00017526 IPI00063827 IPI00003865 IPI00002818 IPI00032328
ProtScore ) 2 14-3-3 protein sigma Angiotensinogen precursor Annexin A5 Antithrombin III Argininosuccinate synthase Calreticulin precursor Ceruloplasmin precursor Chloride intracellular channel 1 Clusterin precursor Eukaryotic translation initiation factor 5Aa Galectin-3 binding protein precursor Glutathione reductase, mitochondrial precursor Glutathione S-transferase A1 Host cell factor 2 Lactate dehydrogenase Aa Macrophage migration inhibitory factor NEDD8 precursor (Ubiquitin-like protein Nedd8) Neutrophil defensin 1 Peroxiredoxin-6 Phosphatidylethanolamine-binding proteina Phospholipase A2, membrane associated precursor PKM2 protein (Pyruvate kinase isozymes M1/M2)a Probable DNA dC f dU editing enzyme APOBEC-3A Proteasome subunit beta type 4 precursora Retinal dehydrogenase 1 S-100P protein Splice Isoform 1 of Abhydrolase domain-containing protein 14B Splice Isoform 1 of Heat shock cognate 71 kDa proteina Splice Isoform 1 of Kallikrein-11 precursor Splice Isoform HMW of Kininogen-1 precursora
2 2 19 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
a The protein had either multiple splice forms or had multiple variable entries in the database which the peptides identified could not distinguish between the different forms. Hence, only one particular form was reported here. b Only one of 2 peptides were found to have a good MS/MS match when the spectra were searched using Mascot.
175.29 ( 65.37 ng/mg total protein (n ) 5) and 79.09 ( 55.98 ng/mg total protein (n ) 6), respectively. The ratios between the dry eye group and control group for R-enolase and S100 A4 were 2.42 and 2.22, which were 1.74 and 2.34 from iTRAQ results (Table 3). A Protein Panel Can Classify Mild, Moderate, and Severe Dry Eye. On the basis of the results of one of the clinical tests, that is, Tear Breakup Time (TBUT), we further classified dry eye patients into three groups: mild (TBUT ) 5-10 s.), moderate (TBUT ) 2-5 s.) and severe (TBUT < 2 s.). Interestingly, we found only R-1-acid glycoprotein 1, S100 A8 and S100 A9 were useful to differentiate the severity of dry eye among the above 10 potential biomarkers (Figure 8). For example, an even distribution was observed for R-enolase. However, higher levels of R-1-acid glycoprotein 1, S100 A8 and S100 A9 were associated with severity of dry eye (shorter tear breakup time). Verification Using Age-Matched Independent Patient Sample Set. Ten biomarker candidate protein significantly associated with dry eye were verified in an independent set of human tear samples (15 patients with dry eye and 18 agematched controls). Tear samples from 18 controls were pooled to form a global control and used as a common reference to compare with dry eye tear samples. The average ratios and SD of these biomarker candidates are listed in Table 4. The table
showed that 9 out of 10 biomarker candidates presented with similar results (t test showed significant changes between dry eye group and age-matched control group, p < 0.05) as those in the discovery phase. We did not detect a signal for R-1-acid glycoprotein 1 in many samples probably because this protein is easily degraded. Our results also showed that the tear protein changes (9 out of 10 biomarker candidates) comparing patients with dry eye and controls from two different study sites (Singapore and Tianjin, China) were very similar (Tables 3 and 4). Correlation of 9 Tear Biomarker Candidates with Age. The iTRAQ results of 9 tear biomarker candidates from 23 patients with dry eye (all from a study site in Tianjin, China, 8 males and 15 females, age range: 19-72) were used to assess the age effect. Logistic regression analysis was performed to determine whether individual protein levels are associated with age (above or below mean of 35 years). The results (Table 5) show that higher S100A8 (p ) 0.037) and S100A9 (p ) 0.011) levels are significantly associated with older age. Higher levels of S100A8 (above mean) were significantly associated (p < 0.001) with more significant dry eye (presence of corneal staining or fluorescein tear breakup time of 3 s or less) and this association was still significant after adjustment with age (p ) 0.021). Similarly, higher S100A9 levels were also Journal of Proteome Research • Vol. 8, No. 11, 2009 4895
Zhou et al.
Figure 2. (A-J) Comparison of dry eye group (ratio, DE:C, 116:114 and 117:114) and control group (ratio, C2:C1, 115:114) for 10 potential dry eye biomarkers. Student t tests (independent) were performed and p-values are indicated on each graph. 4896
Journal of Proteome Research • Vol. 8, No. 11, 2009
Quantitative Biomarkers for Dry Eye
Figure 3. Protein identification and relative quantification using iTRAQ. (A) MS/MS spectrum of a triply charged peptide ion, TYLISSIPLQGAFNYJ, at m/z ) 701.10 (J represents the iTRAQ-modified lysine residue) which is originated from prolactin-inducible protein. (B) Magnified MS/MS spectrum of 111.0 to 120.0 gives signal intensity of four iTRAQ reporter ions at m/z ) 114.1, 115.1, 116.1, and 117.1. iTRAQ 114 and 115 were used to label 2 control samples and iTRAQ 116 and 117 were used to label 2 dry eye samples. Relative quantitation is based on the ratios of peak areas of reporter ions.
associated with significant dry eye (p < 0.001), which was also significant after adjustment for age (p < 0.001).
Discussion The objective of this study was to identify and quantify the differences in tear proteins between dry eye patients and a
control group of subjects with a high level of confidence (g99%) using 2D nanoLC-nanoESI-MS/MS with iTRAQ. 2D LC separation combined with iTRAQ technology allowed us to identify over 90 tear proteins. From our observations, using iTRAQ compared with not using iTRAQ in a 2D shotgun experiment, more b-ions were obtained in MS/MS spectra and thus there Journal of Proteome Research • Vol. 8, No. 11, 2009 4897
Zhou et al.
Table 3. Summary of Weighted Average Ratios (DE:C and C2:C1), Weighted Standard Deviations and p-values of Potential Biomarkers for Control Group and Dry Eye Groupa protein name
control (C2:C1) (weighted average ratio)
control (C2:C1) (weighted SD)
dry eye (DE:C) (weighted average ratio)
dry eye (DE:C) (weighted SD)
t test, p-value
Alpha-enolase S100 A8 Calgranulin A S100 A9 Calgranulin B Alpha-1-acid glycoprotein S100 A11 Calgizzarin S100 A4 Calcium-Binding protein Lactotransferrin Von Ebner’s gland protein Prolactin-inducible protein Lysozyme C
0.936 0.951 0.868 1.255 0.987 0.851 0.926 1.019 0.995 0.930
1.065 1.110 1.212 1.115 1.095 1.721 1.089 1.118 1.065 1.063
1.742 1.358 1.408 2.423 1.654 2.342 0.792 0.720 0.593 0.698
1.154 1.180 1.296 1.352 1.215 2.268 1.198 1.138 1.103 1.140
0.0004 0.0306 0.0281 0.0296 0.0406 0.0175 0.0238 0.0119 0.0001 0.0034
The true average ratio is expected to be found in the range of “weighted average ratio/weighted SD” to “weighted average ratio × weighted SD”.
is better coverage of both b-ions and y-ions (Figures 3A, 4A, and 5A). Typically, the QSTAR mass spectrometer tends to record more y-ions in a MS/MS spectrum. Therefore, in addition to providing quantitative information, iTRAQ can also enhance the efficiency of MS/MS fragmentation33 and thus increase the confidence for protein identification. The causes for dry eye syndrome are diverse; however, fundamentally, it is due to an insufficiency of tears to maintain the ocular surface and particularly the cornea. The cornea is the most important optical element of the eye and dry eye decreases good vision as well as the quality of life of the patient.7 Dry eye affects millions of people and is a significant clinical problem in Asia as well as the U.S. and Europe. A poor correlation exists between clinical tests and subjective symptoms of dry eye which makes it difficult to diagnosis and evaluate therapeutic efficacy for new drug development. In this study, tears were collected using Schirmer strips, a routine clinical test to evaluate the quantity of the tear secretions in a dry eye clinic. Tear proteins were then eluted from the strip by phosphate buffered saline (PBS) using the same protocol for each sample. Previous work by Grus et al.17 and our experience using this method34 showed that the recovery of total tear proteins is reproducible and acceptable. Total tear protein concentrations were also measured for each sample after elution. For relative protein quantitation using iTRAQ, the sample loadings for 2 controls and 2 diseased samples were the same (30 µg each). One drawback of the iTRAQ approach mentioned in previous studies30,35 was that because 4 parallel samples were digested and labeled separately, sample handing errors could be introduced. Therefore, in our data processing, a bias correction factor was applied for data processing in Pro QUANT to eliminate such errors (ProQUANT manual by Applied Biosystems). Because iTRAQ allows labeling of four samples simultaneously (114, 115, 116 and 117), in our study design, two controls and two disease samples were used in each iTRAQ set. For controls, pooled samples (from either 3 or 5 controls) as well as individual controls were used. As expected, smaller variations were typically seen when using pooled controls compared to using individual controls. Figure 9 gives examples of comparing pooled controls with individual controls using two representative tear proteins (R-enolase and prolactininducible protein). Other tear proteins showed similar trends. The advantage of using pooled controls was the decrease in variance; however, some loss of information of individual data occurs. 4898
Journal of Proteome Research • Vol. 8, No. 11, 2009
Our results were generated by comparing with the first control sample (labeled by 114 iTRAQ reagent) and ratios of 115/114 (C2:C1) formed control group and ratios of 116/114 (DE1: C1) and 117/114 (DE2:C1) formed the dry eye group. From Pro Group View report, the proteins with altered expression indicated by ratios (DE:C) and p-value (P < 0.05) were selected as potential biomarkers for further assessment. The student t test was used to further confirm the significant changes between control group (C2:C1) and diseased group (DE1:C1 and DE2:C1) for those selected as biomarker candidates. This step assessed the altered expression of those potential tear protein biomarkers across 14 sample sets. ROC analysis was used to compare the performance of individual biomarker candidates and biomarker panel. We have shown that using a protein panel can increase the diagnostic accuracy (Figure 6C). However, the use of additional proteins in the panel would increase the parameters to establish the diagnosis model and hence needs bigger sample size. In this study, a 4-protein panel was found to provide a diagnostic accuracy of better than 95%. Proteins Up-Regulated in the Tears of Dry Eye Patients. The level of R-enolase in the tears was found to correctly identify a patient as belonging to the dry eye group in 85% of the time. R-Enolase is a key glycolytic enzyme expressed abundantly in most cells. However, more recently it was also found to be a cell surface protein,36 on hematopoietic cells (neutrophils, B cells, T Cells, monocytes), epithelial cells, and neuronal cells. In addition to its innate glycolytic function, R-enolase has been recognized as a multifunctional protein. Very recent studies indicate that it may play an important role in several disease processes, for example, in autoimmune disorders, cancer, systemic fungal disease and dental diseases.36 R1-Acid glycoprotein 1 (AGP) is a heavily glycosylated protein (45%) with a molecular weight of 41-43 kDa37 and belongs to the lipocalin family. The synthesis is controlled by glucocorticoids, interleukin-1 and interleukin-6. Its anti-inflammatory effects have been noted.37 S100 A8 and A9 belong to the S100 calcium-binding family of proteins. There is growing evidence showing that S100 A8, A9 and A12 form a new group of pro-inflammatory proteins.38 In inflammatory diseases such as rheumatoid arthritis, the overexpression of both S100 A8 and S100 A9 are typically secreted and seen at sites of inflammation and may be possible biomarkers for the disease. The overexpression of S100 A8 in dry eye tears has been reported previously.17 S100 A4, also a member of the S100 calcium binding protein family, has been cited for its role in cell shape changes. More
Quantitative Biomarkers for Dry Eye
Figure 4. (A) MS/MS spectrum of one triply charged peptide ion (SGETEDFIADLWGLCTGQIJ at m/z ) 877.40) originated from R-enolase and (B) relative quantification for R-enolase between dry eye samples and control samples. Journal of Proteome Research • Vol. 8, No. 11, 2009 4899
Zhou et al.
Figure 5. (A) MS/MS spectrum of one quadruply charged peptide ion (YVGGQEHFAHLLILR at m/z ) 475.00) originated from R-1-acid glycoprotein 1. (B) Relative quantification for R-1-acid glycoprotein 1 between dry eye samples and control samples. 4900
Journal of Proteome Research • Vol. 8, No. 11, 2009
Quantitative Biomarkers for Dry Eye
Figure 7. ELISA assay for R-enolase and S100 A4 in tears from dry eye patients and controls. The differential expression levels of R-enolase and S100 A4 found using iTRAQ were verified with the ELISA assay. Table 4. Verification Using an Independent Age-Matched Patient Sample Seta
R-enolase S100 A4 S100 A8 S100 A9 S100 A11 Lactoferrin Lysozyme C Prolactin-inducible protein Lipocalin-1
average ratio (DE vs control) n ) 15
t test, p-value
1.688 1.786 1.822 1.916 1.472 0.803 0.856 0.746
1.543 1.426 1.413 1.481 1.423 1.236 1.224 1.306
0.0033 0.0000 0.0000 0.0001 0.0001 0.0004 0.0185 0.0011
a Average ratios (DE vs Control), standard deviation (SD) of 9 potential dry eye tear biomarkers and p-value. The true average ratio is expected to be found in the range of “average ratio/SD” to “average ratio × SD”.
Figure 6. (A) ROC curve when using R-enolase as the only biomarker. The accuracy (area under the ROC curve) is 85%. (B) ROC curve when using prolactin-inducible protein as the only biomarker. The accuracy (area under the ROC curve) is 81%. (C) ROC curve when using a 4-protein biomarker panel. The accuracy (area under the ROC curve) is 96%. Red trace: ROC curve when using R-enolase as the only biomarker. Blue trace: ROC when using a 4-protein biomarker panel.
recently,39 it was shown that S100 A4 is capable of stimulating corneal neovascularization in vivo. S100 A4 also appears to take part in the homeostasis of growth, with apparent involvement also in growth factor signal transduction and apoptotic cell death. There is considerable evidence that S100 A4 expression alters the adhesive properties of cells, possibly by remodelling Journal of Proteome Research • Vol. 8, No. 11, 2009 4901
Zhou et al.
Figure 8. Ratios (DE:C) were reorganized according to tear breakup time (TBUT) of each patient for (A) R-enolase, (B) R-1-acid glycoprotein 1, (C) S100 A9 and (D) S100 A8. Different color codes were used to distinguish three subgroups: (1) yellow, mild (TBUT ) 5-10 s.); (2) orange, moderate (TBUT ) 2-5 s.); (3) red, severe (TBUT < 2 s.). 4902
Journal of Proteome Research • Vol. 8, No. 11, 2009
Quantitative Biomarkers for Dry Eye Table 5. Logistic Regression Analyses with Age (above 35 years) as Dependent Variable and Individual Protein (Rows) Ratios as the Independent Variables
S100 A4 S100 A8 S100 A9 S100 A11 enolase Lactoferrin Lysozyme PIP Lipocalin-1
95% confidence interval
0.575 0.037 0.011 0.855 0.771 0.275 0.083 0.065 0.531
1.53 3.137 3.394 1.28 1.15 0.061 0.013 0.000 0.969
(0.346, 6.77) (1.073, 9.19) (1.319, 8.73) (0.09, 18.31) (0.448, 2.955) (0.001, 8.69) (0, 2.118) (0, 1.68) (0.022, 1630)
Figure 9. Comparison of variation of pooled control ratio (C2: C1) versus individual control ratio (C2:C1) for two representative tear protein biomarkers. Larger variations were observed in the individual control group.
the extracellular matrix and promoting a redeployment of adhesion-mediating macromolecules occurring in the extracellular matrix.40 A recent paper by Rivard et al.41 showed that expression of S100A4 is markedly up-regulated by osmotic stress and involved in the renal osmoadaptive response. In the case of dry eye in humans, tear hyperosmolarity42 may stimulate the changes in S100 proteins. Finally, S100 A11, another member of the S100 calcium binding protein family, has been shown to be involved in apoptosis.43 Proteins Down-Regulated in the Tears of Dry Eye Patients. Lactoferrin and lysozyme are two abundant human tear proteins with a role in antibacterial protection of the ocular surface.13 In previous studies, down-regulation of these two tear proteins was observed in patients with dry eye.44,17 In fact, it has been found that dry eye patients are more likely to contact infectious diseases of the ocular surface.45 Lipocalin-1, also called tear specific prealbumin (TSPA) or tear lipocalin,46 is a major tear protein and acts as the principal
Figure 10. Potential tear biomarkers of dry eye and their associations with the mechanisms of dry eye.
lipid binding protein. It is regarded as a general protection factor of epithelial cell surfaces. Prolactin-inducible protein (PIP) is typically expressed in several exocrine tissues, such as the lacrimal, salivary, and sweat glands, and may also be associated with breast cancer.47 A very recent study showed that it was down-regulated in the tears of patients with blepharitis, an eyelid infection.19 Clinically, it is very important to differentiate severity (i.e., mild, moderate and severe) of dry eye to effectively use different treatment strategies. In this paper, our preliminary results have shown that it is possible to classify mild, moderate and severe dry eye by using a group of inflammation-associated proteins, R1-acid glycoprotein 1, S100 A8 and S100 A9. Independent Verification of Biomarker Candidates. We verified two biomarker candidates, that is, R-enolase and S100 A4, using commercially available ELISA kits. The ELISA results corroborated the proteomic results obtained from experiments using iTRAQ. The iTRAQ results for S100 A8, lactoferrin and lysozyme were also compared with data from previous studies.17,44 Using the ELISA assay for the verification step is labor-intensive and sometimes sample volume is an issue, in particular with tear fluid as only a few microliters are obtained for an individual patient. An alternative way for protein/peptide quantification is to use multiple reaction monitoring (MRM) MS-based technology.48,49 With this technology, it is possible to quantify many proteins/peptides with higher sensitivity in a single LC-MS/MS run. The changes of tear proteins which are associated with dry eye were found to be very similar in patients from two different locations (Singapore and Tianjin, China). Logistic regression analysis was also performed to assess if there is any correlation between patient age and protein levels. Results showed that the correlation of patient age with levels of S100 A8 and A9 is significant (p-value: 0.037 and 0.011, respectively). However, the association for dry eye was still significant after adjustment for age (S100 A8, p < 0.021; S100 A9, p < 0.001). No significant association was found for the rest of the tear biomarker candidates. A previous study reported that there was a negative correlation between age and some tear protein levels, for example, lysozyme and lactoferrin, probably due to the decline of the secretion function of the lacrimal gland in older people.50 However, we did see significant changes in lysozyme and lactoferrin between dry eye group and control group from our age-matched experiment (Table 4). Our age correlation assessment did not show significant association between age and the levels of lysozyme and lactoferrin, maybe because of the relative small sample size.
Conclusions In summary (Figure 10), based on their biological functions, the above 10 biomarker candidates could be divided into Journal of Proteome Research • Vol. 8, No. 11, 2009 4903
research articles different groups. R1-Acid glycoprotein 1, S100 A8 and S100 A9 are associated with inflammatory responses. S100 A4 and S100 A11 may be related to osmotic stress and apoptosis, respectively. These five proteins may reflect the status of cells on the ocular surface. Another 4 down-regulated proteins (lactoferrin, lysozyme, lipocalin-1 and prolactin-inducible protein) are secreted proteins from the lacrimal gland and may suggest secretory deficiency that characterizes aqueous deficient dry eye. The relationship of R-enolase with dry eye is not clear. Elevated R-enolase and S100 A4 tear levels in dry eye patients were further verified by ELISA. The results were also verified using an independent age-matched patient sample set.
Acknowledgment. Authors would like to thank Koh Siew Kwan, Foo Yong Hwee and Rachel Tseng for the technical support. This work was supported by grants NMRC/0808/2003, NMRC/CPG/002/2003, NMRC/0982/2005 and NMRC/1206/2009 from National Medical Research Council (NMRC), Singapore, and an unrestricted grant from Allergan, Irvine, CA. Supporting Information Available: Supplementary Table 1 is a summary of ratios, p-values and Error Factors (EF) from 14 sets of iTRAQ experiments (Set 4 and Set 6 were analyzed twice). There was only one peptide fragment found for those data which only have ratios with no p-values and EF. This material is available free of charge via the Internet at http://pubs.acs.org. References (1) Dogru, M.; Tsubota, K. New insights into the diagnosis and treatment of dry eye. Ocul. Surf. 2004, 2, 59–75. (2) Schein, O. D.; Munoz, B.; Tielsch, J. M.; Bandeen-Roche, K.; West, S. Prevalence of dry eye among the elderly. Am. J. Ophthalmol. 1997, 124 (6), 723–8. (3) Jie, Y.; Xu, L.; Wu, Y. Y.; Jonas, J. B. Prevalence of dry eye among adult Chinese in the Beijing eye study. Eye 2009, 23 (3), 688–93. (4) Lee, A. J.; Lee, J.; Saw, S. M.; Gazzard, G.; Koh, D.; Widjaja, D.; Tan, D. T. Prevalence and risk factors associated with dry eye symptoms: a population based study in Indonesia. Br. J. Ophthalmol. 2002, 86 (12), 1347–51. (5) Sullivan, D. A. Tearful relationships? Sex, hormones, the lacrimal gland, and aqueous-deficient dry eye. Ocul. Surf. 2004, 2, 92–123. (6) Nichols, J. J.; Sinnott, L. T. Tear film, contact lens, and patientrelated factors associated with contact lens-related dry eye. Invest. Ophthalmol. Visual Sci. 2006, 47 (4), 1319–28. (7) Epidemiology DEWS Subcommittee. The epidemiology of dry eye disease: report of the Epidemiology Subcommittee of the International Dry Eye WorkShop (2007). Ocul Surf. 2007, 5 (2), 93–107. (8) Stern, M. E.; Beuerman, R. W.; Fox, R. I.; Gao, J.; Mircheff, A. K.; Pflugfelder, S. C. The pathology of dry eye: the interaction between the ocular surface and lacrimal glands. Cornea 1998, 17 (6), 584– 9. (9) Beuerman, R. W.; Mircheff, A.; Pflugfelder, S. C.; Stern, M. S. The Lacrimal Functional Unit. In Dry Eye and Ocular Surface Disorders; Pflugfelder, S. C., Beuerman, R. W., Stern, M. S., Eds; Marcel Dekker: New York, 2004; pp 11-39. (10) Stern, M. E.; Gao, J.; Siemasko, K. F.; Beuerman, R. W.; Pflugfelder, S. C. The role of the lacrimal functional unit in the pathophysiology of dry eye. Exp. Eye Res. 2004, 78 (3), 409–16. (11) Lemp, M. A. Report of the National Eye Institute/Industry workshop on clinical trials in dry eyes. CLAO J. 1995, 21 (4), 221–32. (12) Ousler, G. W.; Gomes, P. J.; Welch, D.; Abelson, M. B. Methodologies for the study of ocular surface disease. Ocul. Surf. 2005, 3, 143–154. (13) Donald, R. K.; Jennifer, C.; Michael, D.; Jean-Pierre, G.; George, S.; Alan, T. The Tear Film; Butterworth-Heinemann: Oxford, 2002; pp 51-81. (14) Zhou, L.; Beuerman, R. W.; Barathi, A.; Tan, D. Analysis of rabbit tear proteins by high-pressure liquid chromatography/electrospray ionization mass spectrometry. Rapid Commun. Mass Spectrom. 2003, 17 (5), 401–12.
Journal of Proteome Research • Vol. 8, No. 11, 2009
Zhou et al. (15) Zhou, L.; Huang, L. Q.; Beuerman, R. W.; Grigg, M. E.; Li, S. F.; Chew, F. T.; Ang, L.; Stern, M. E.; Tan, D. Proteomic analysis of human tears: defensin expression after ocular surface surgery. J. Proteome Res. 2004, 3 (3), 410–6. (16) Zhou, L.; Beuerman, R. W.; Huang, L.; Barathi, A.; Foo, Y. H.; Li, S. F.; Chew, F. T.; Tan, D. Proteomic analysis of rabbit tear fluid: Defensin levels after an experimental corneal wound are correlated to wound closure. Proteomics. 2007, 7 (17), 3194–206. (17) Grus, F. H.; Podust, V. N.; Bruns, K.; Lackner, K.; Fu, S.; Dalmasso, E. A.; Wirthlin, A.; Pfeiffer, N. SELDI-TOF-MS ProteinChip array profiling of tears from patients with dry eye. Invest. Ophthalmol. Visual Sci. 2005, 46 (3), 863–76. (18) Tomosugi, N.; Kitagawa, K.; Takahashi, N.; Sugai, S.; Ishikawa, I. Diagnostic potential of tear proteomic patterns in Sjogren’s syndrome. J. Proteome Res. 2005, 4 (3), 820–5. (19) Koo, B. S.; Lee, D. Y.; Ha, H. S.; Kim, J. C.; Kim, C. W. Comparative analysis of the tear protein expression in blepharitis patients using two-dimensional electrophoresis. J. Proteome Res. 2005, 4 (3), 719– 24. (20) Kuizenga, A.; van Haeringen, N. J.; Kijlstra, A. SDS-Minigel electrophoresis of human tears. Effect of sample treatment on protein patterns. Invest. Ophthalmol. Visual Sci. 1991, 32 (2), 381–6. (21) Molloy, M. P.; Bolis, S.; Herbert, B. R.; Ou, K.; Tyler, M. I.; van Dyk, D. D.; Willcox, M. D.; Gooley, A. A.; Williams, K. L.; Morris, C. A.; Walsh, B. J. Establishment of the human reflex tear twodimensional polyacrylamide gel electrophoresis reference map: new proteins of potential diagnostic value. Electrophoresis 1997, 18 (15), 2811–5. (22) Grus, F. H.; Sabuncuo, P.; Augustin, A. J. Analysis of tear protein patterns of dry-eye patients using fluorescent staining dyes and two-dimensional quantification algorithms. Electrophoresis 2001, 22 (9), 1845–50. (23) Varnell, R. J.; Maitchouk, D. Y.; Beuerman, R. W.; Carlton, J. E.; Haag, A. Small-volume analysis of rabbit tears and effects of a corneal wound on tear protein spectra. Adv. Exp. Med. Biol. 1998, 438, 659–64. (24) Fung, K. Y.; Morris, C.; Sathe, S.; Sack, R.; Duncan, M. W. Characterization of the in vivo forms of lacrimal-specific prolinerich proteins in human tear fluid. Proteomics 2004, 4 (12), 3953– 9. (25) Li, N.; Wang, N.; Zheng, J.; Liu, X. M.; Lever, O. W.; Erickson, P. M.; Li, L. Characterization of human tear proteome using multiple proteomic analysis techniques. J. Proteome Res. 2005, 4 (6), 2052– 61. (26) Zhou, L.; Beuerman, R. W.; Foo, Y.; Liu, S.; Ang, L. P.; Tan, D. T. H. Characterisation of human tear proteins using highresolution mass spectrometry. Ann. Acad. Med. Singapore 2006, 35 (6), 400–7. (27) de Souza, G. A.; Godoy, L. M.; Mann, M. Identification of 491 proteins in the tear fluid proteome reveals a large number of proteases and protease inhibitors. Genome Biol. 2006, 7 (8), R72. (28) Sack, R. A.; Conradi, L.; Krumholz, D.; Beaton, A.; Sathe, S.; Morris, C. Membrane array characterization of 80 chemokines, cytokines, and growth factors in open- and closed-eye tears: angiogenin and other defense system constituents. Invest. Ophthalmol. Visual Sci. 2005, 46 (4), 1228–38. (29) Ross, P. L.; Huang, Y. N.; Marchese, J. N.; Williamson, B.; Parker, K.; Hattan, S.; Khainovski, N.; Pillai, S.; Dey, S.; Daniels, S.; Purkayastha, S.; Juhasz, P.; Martin, S.; Bartlet-Jones, M.; He, F.; Jacobson, A.; Pappin, D. J. Multiplexed protein quantitation in saccharomyces cerevisiae using amine-reactive isobaric tagging reagents. Mol. Cell. Proteomics 2004, 3 (12), 1154–1169. (30) DeSouza, L.; Diehl, G.; Rodrigues, M. J.; Guo, J.; Romaschin, A. D.; Colgan, T. J.; Siu, K. W. Search for cancer markers from endometrial tissues using differentially labeled tags iTRAQ and cICAT with multidimensional liquid chromatography and tandem mass spectrometry. J. Proteome Res. 2005, 4 (2), 377–86. (31) Chen, X.; Walker, A. K.; Strahler, J. R.; Simon, E. S.; TomanicekVolk, S. L.; Nelson, B. B.; Hurley, M. C.; Ernst, S. A.; Williams, J. A.; Andrews, P. C. Organellar proteomics: Analysis of pancreatic zymogen granule membranes. Mol. Cell. Proteomics 2006, 5 (2), 306–12. (32) Zhang, Y.; Wolf-Yadlin, A.; Ross, P. L.; Pappin, D. J.; Rush, J.; Lauffenburger, D. A.; White, F. M. Time-resolved mass spectrometry of tyrosine phosphorylation sites in the epidermal growth factor receptor signaling network reveals dynamic modules. Mol. Cell. Proteomics 2005, 4 (9), 1240–50. (33) Hardt, M.; Witkowska, H. E.; Webb, S.; Thomas, L. R.; Dixon, S. E.; Hall, S. C.; Fisher, S. J. Assessing the effects of diurnal variation on the composition of human parotid saliva: quantitative analysis
Quantitative Biomarkers for Dry Eye
of native peptides using iTRAQ reagents. Anal. Chem. 2005, 77 (15), 4947–54. Beuerman, R. W.; Zhou, L.; Prema, P.; Chan, C. M.; Ang, L. P. K.; Angayarkanni, N.; Foo, Y. H.; Liu, S. P.; Tan, D. T. H. Comprehensive characterization of human tear proteome using nanoliquid chromatography-QTOF tandem mass spectrometry and quantitative proteomics (iTRAQ). Association for Research in Vision and Ophthalmology, annual meeting 2006, Fort Lauderdale, FL, April 30-May 4, 2006; e-abstract, http://www.arvo.org/ eweb/DynamicPage.aspx?site)am&WebCode)AbstractSearch. Wu, W. W.; Wang, G.; Baek, S. J.; Shen, R. F. Comparative study of three proteomic quantitative methods, DIGE, cICAT, and iTRAQ, using 2D gel- or LC-MALDI TOF/TOF. J. Proteome Res. 2006, 5 (3), 651–8. Pancholi, V. Multifunctional alpha-enolase: its role in diseases. Cell. Mol. Life Sci. 2001, 58 (7), 902–20. Hochepied, T.; Berger, F. G.; Baumann, H.; Libert, C. Alpha(1)acid glycoprotein: an acute phase protein with inflammatory and immunomodulating properties. Cytokine Growth Factor Rev. 2003, 14 (1), 25–34. Roth, J.; Vogl, T.; Sorg, C.; Sunderkotter, C. Phagocyte-specific S100 proteins: a novel group of proinflammatory molecules. Trends Immunol. 2003, 24 (4), 155–8. Ryan, D. G.; Taliana, L.; Sun, L.; Wei, Z. G.; Masur, S. K.; Lavker, R. M. Involvement of S100A4 in stromal fibroblasts of the regenerating cornea. Invest. Ophthalmol. Visual Sci. 2003, 44 (10), 4255–62. Sherbet, G. V.; Lakshmi, M. S. S100A4 (MTS1) calcium binding protein in cancer growth, invasion and metastasis. Anticancer Res. 1998, 18 (4A), 2415–21. Rivard, C. J.; Brown, L. M.; Almeida, N. E.; Maunsbach, A. B.; Pihakaski-Maunsbach, K.; Andres-Hernando, A.; Capasso, J. M.; Berl, T. Expression of the calcium-binding protein S100A4 is markedly up-regulated by osmotic stress and is involved in the renal osmoadaptive response. J. Biol. Chem. 2007, 282 (9), 6644– 52.
(42) No authors listed. 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 (2), 75–92. (43) Kanamori, T.; Takakura, K.; Mandai, M.; Kariya, M.; Fukuhara, K.; Sakaguchi, M.; Huh, N. H.; Saito, K.; Sakurai, T.; Fujita, J.; Fujii, S. Increased expression of calcium-binding protein S100 in human uterine smooth muscle tumours. Mol. Hum. Reprod. 2004, 10 (10), 735–42, Epub 2004 Aug 20. (44) Ohashi, Y.; Ishida, R.; Kojima, T.; Goto, E.; Matsumoto, Y.; Watanabe, K.; Ishida, N.; Nakata, K.; Takeuchi, T.; Tsubota, K. Abnormal protein profiles in tears with dry eye syndrome. Am. J. Ophthalmol. 2003, 136 (2), 291–9. (45) Das, S.; Constantinou, M.; Ong, T.; Taylor, H. R. Microbial keratitis following corneal transplantation. Clin. Exp. Ophthalmol. 2007, 35 (5), 427–31. (46) Redl, B. Human tear lipocalin. Biochim. Biophys. Acta 2000, 1482 (1-2), 241–8. (47) Clark, J. W.; Snell, L.; Shiu, R. P.; Orr, F. W.; Maitre, N.; Vary, C. P.; Cole, D. J.; Watson, P. H. The potential role for prolactin-inducible protein (PIP) as a marker of human breast cancer micrometastasis. Br. J. Cancer 1999, 81 (6), 1002–8. (48) Anderson, L.; Hunter, C. L. Quantitative mass spectrometric multiple reaction monitoring assays for major plasma proteins. Mol. Cell. Proteomics 2006, 5 (4), 573–88. (49) Keshishian, H.; Addona, T.; Burgess, M.; Kuhn, E.; Carr, S. A. Quantitative, multiplexed assays for low abundance proteins in plasma by targeted mass spectrometry and stable isotope dilution. Mol. Cell. Proteomics 2007, 6 (12), 2212–29. (50) McGill, J. I.; Liakos, G. M.; Goulding, N.; Seal, D. V. Normal tear protein profiles and age-related changes. Br. J. Ophthalmol. 1984, 68 (5), 316–20.
Journal of Proteome Research • Vol. 8, No. 11, 2009 4905