Association of Tear Cytokine Concentrations with Symptoms and Signs of Dry Eye Disease: Baseline Data from the Dry Eye Assessment and Management (DREAM) Study

Abstract Purpose To describe tear concentrations of IL-1β, Il-6, IL-8, IL-10, IL-17A, IFNγ and TNFα in tears, collected by microcapillaries, and their correlation with symptoms and signs in subjects with dry eye disease (DED) in the DREAM Study. Methods Cytokine levels of patients with moderate to severe DED were determined using a magnetic bead assay. Scores for Ocular Surface Disease Index, corneal and conjunctival staining, tear break-up time (TBUT), and Schirmer’s test were obtained using standardized procedures. Associations of cytokines with each other and signs/symptoms were assessed with Spearman correlation coefficients (r). Results Assay results from 131 patient samples from 10 sites with tear volumes ≥ 4 ul were analyzed. Cytokine concentrations did not correlate with each other in a generally acknowledged pro-inflammatory/anti-inflammatory pattern, such as proinflammatory IL-17A and IFNγ were not inversely correlated to anti-inflammatory cytokine IL-10, and cytokines did not correlate with DED symptoms. Lower corneal staining was correlated with higher concentrations of IL-17A (r= −0.24, p = 0.006), IL-10 (r= −0.25, p = 0.005) and IFNγ (r= −0.33, p = 0.0001). Higher concentrations of IFNγ were associated with lower conjunctival staining (r= −0.18, p = 0.03). Higher concentrations of IL-17A were associated with higher TBUT scores (r = 0.19 p = 0.02). Conclusions Cytokines IL-10, IL-17A and IFNγ were highly correlated with each other but weakly correlated with some DED signs. No key cytokines or definitive expression patterns were identified in this study of moderate to severe DED patients. Further studies addressing various biases, including methodological and sampling biases, and standardization of methodology for inter-laboratory consistency are needed to confirm and establish pathological and clinical relevance of tear cytokines in DED.


Introduction
Dry eye disease (DED) is a multifactorial condition affecting 40% of adults over 40 years of age and represents a great economic burden worldwide. [1][2][3] The pathology of DED involves tear film instability and hyperosmolarity, ocular surface damage, and neurosensory abnormalities. 1 Clinical diagnosis consists of quantification of these pathological signs in the form of tear breakup time (TBUT), tear volume (Schirmer's test), tear osmolarity, corneal fluorescein staining score and conjunctival lissamine green staining score. 4,5 However, these tests are largely subjective in nature, further compounded by poor correlation of these signs with patient reported symptoms. [6][7][8] Inflammation is a well-described component of DED and is proposed as a core mechanism of the disease pathogenesis. [9][10][11] This is supported by multiple studies in animal models of DED and patients where changes have been observed in multiple inflammatory cell types on the ocular surface and inflammatory mediators in tears. [12][13][14][15][16][17][18] Consequently, all of these have the potential to serve as biomarkers with objectively measurable metrics. [19][20][21][22][23] Cytokines and chemokines are particularly attractive biomarker candidates due to their central and wide-ranging roles that include pro-and anti-inflammatory effects. [24][25][26] Tears in DED represent an effective source for cytokines due to their association with the disease organ and the relative ease of collection and tear collection is minimally invasive. 27,28 The development of highly sensitive assay systems that require very small tear volumes have strengthened the rationale for exploring tear cytokines as potential minimally invasive objective metrics for DED. [29][30][31][32] Studies have consistently demonstrated measurable amounts of cytokines in tears from DED patients, including IL-1a, IL-1b, IL-6, IL-8, IL-10, L-13, IL-17A, IFNc, and TNFa. 12,16,31,[33][34][35][36][37][38][39] However, there is no consensus on which cytokine or combinations thereof, can serve as biomarkers of DED. The challenges to synthesizing the results from different studies include, (a) variations in methods and time of tear sample acquisition, (b) differences in tear assays and data analysis and (c) lack of consistent correlation between symptoms and signs, and (d) variability of patient populations studied. 22,30,33,40 In this study, we measured the concentration of IL-1b, Il-6, IL-8, IL-10, IL-17A, IFNc and TNFa in tears of DED patients participating in the multi-site Dry Eye Assessment and Management (DREAM#) study (clinicaltrials.gov ID NCT02128763). 41 We selected the cytokines on the basis of published research on DED available at the time the study was initiated and also the ability to evaluate all cytokines at the same time with Luminex technology. We assessed the correlations at baseline among the cytokine concentrations and the correlations of cytokine concentrations with clinical symptoms and signs to explore their potential as clinically relevant biomarkers of DED.

Patients and study design
A total of 535 moderate to severe DED patients were recruited from 27 clinical sites throughout the United States for the DREAM study, a randomized clinical trial to assess the effect of omega-3 supplementation on DED. Patient inclusion and exclusion criteria, as well as standardized procedures for examination and tear collection, have been detailed in previous publications. 41,42 As shown in Figure 1, among 535 randomized participants, 244 participants were from sites without sample storage capability (-80 degree C freezers) and therefore were not targeted for tear collection. Samples collected from patients before treatment with omega-3 supplement or placebo were designated as 'baseline'. Among 291 participants enrolled from the 10 sites that had storage capability, 233 provided a baseline tear sample. The study was conducted in compliance with the tenets of the Declaration of Helsinki for the protection of human subjects in medical research, was approved by the institutional review board for each site, and was authorized by FDA through an Investigational New Drug (IND) application.

Patient assessment
Patients completed the Ocular Surface Disease Index (OSDI) questionnaire for symptoms and study-certified clinicians measured signs that included corneal staining with fluorescein dye, conjunctival staining with lissamine green dye, TBUT, and Schirmer's tear test with anesthesia. OSDI questionnaire consisted of 12 questions with ratings from 0-4 with scoring to transform the responses to a 0 to 100 scale, with higher score indicating more symptoms. TBUT was measured 30 s after instillation of 5 ml, 2% fluorescein-containing solution. The time between the last blink and the appearance of the first discontinuity in the fluoresceinstained tear film was noted and repeated twice more. Corneal fluorescein staining was graded using the cobalt blue filter of a slit lamp approximately 2.5 min after fluorescein instillation. Staining was scored using the National Eye Institute [NEI]/industry-recommended guidelines (0-15). Conjunctival staining was graded after 1-2 min of placing 5 ml of 1% lissamine green dye into the lower conjunctival sac. Grading was done for the nasal-bulbar and temporalbulbar conjunctiva using a modified version of the NEI/ industry-recommended guidelinesthe temporal and nasal section of each eye was graded on a scale of 0 to 3 (0: no staining, 3: severe staining) for a total possible score of 6 in each eye. Assessments for signs were done first in the right eye and then in the left eye. Schirmer's test was performed, following administration of preservative-free topical anaesthetic, by inserting the strip at the junction of the lower lid for 5 min.

Tear sample collection, shipment and storage
Tear samples were collected according to a previously published standard operating procedure (SOP). 31 Tears were collected prior to any exam or imaging procedures and at least two hours since the use of any eye drops. The study protocol provided clear guidelines for proper collection and storage of tears by personnel at clinical sites. In brief, basal non-stimulated tears from the right and left eye from an individual patient were collected using microcapillary tubes at the lateral canthus without anaesthesia and minimum contact between the capillary tube and the surrounding tissues. Time of day for tear collection was not controlled.  the same collection tube, and immediately stored at minus eighty degrees Celsius. As most DED patients have very low levels of basal tear production, tears were pooled to ensure the availability of a minimum of 4 ml volume to carry out the multiplex assay. 31 As the inter-eye Pearson's correlation coefficient was high for conjunctival staining score (0.61), corneal staining score (0.77), TBUT (0.65) and Schirmer's test (0.77) and % of conjunctival epithelial cells expressing inflammatory marker HLA-DR (0.87), 43 the pooling of tears was not expected to affect the overall individual patient cytokine profile. Frozen tear samples were shipped on dry ice to the ocular biomarker research lab at the Icahn School of Medicine at Mount Sinai and stored at minus eighty degrees Celsius. All assays were performed within 12 months of collection.

Tear sample processing by the Central laboratory
Tear samples in collecting tubes were first centrifuged at 14,000 rpm for 10 min at room temperature. Up to a maximum of 10 ml of pooled tear sample (right and left eye) from each patient was transferred to a new tube and brought up to a total volume of 50 ml with buffer provided in the cytokine assay kits (next section). Based on our preliminary work, results of tear samples with volumes less than 4 ml were not included in the analysis because of concern of validity of the measured cytokine values. 31 Magnetic bead-based 96-well plate assay, data acquisition and analysis Concentrations of IL-1a, IL-1b, IL-6, IL-8, IL-10, L-13, IL-17A, IFNc, and TNFa were measured by MILLIPLEX-MAG kit (High Sensitivity Human Cytokine Kit, Cat # HSCYTMAG-28SK, Millipore Corporation, Billerica, MA 01821) based on the LuminexV R xMAP V R technology, as per the manufacturer's manual and our laboratory standard operating procedures (SOP). 31 Strict quality control measures were followed to ensure accuracy and repeatability of data output as well as comparable inter-plate performance.
In order to ensure consistency between assays, three sets of quality control (QC) samples were used for every run. Two sets were manufacturer provided samples, QC1 and QC2, with known concentrations of study cytokines within the lower (20-50 pg/mL) and the higher (100-500 pg/mL) concentration range respectively. The third quality control sample, QC3, consisted of two subsets (QC3A and QC3B) of pooled tears collected from two groups of healthy volunteers and aliquoted in 10 ul amounts frozen at minus eighty degrees Celcius and stored to last the entire duration of the study. For the assay, a 96-well plate was pre-wetted using the kit provided washing buffer at room temperature for 10 min. 50 ul of standards, controls or diluted tear samples were added to designated wells in the plate, followed by 25 ul of mixed beads to each well. The plates were sealed with a light-proof sealer and incubated at 4 C for 16-18 h on an orbital shaker. Following three washings, 50 ul of detection antibodies was added to each well and incubated at room temperature for 1 h, followed by addition of 50 ul Streptavidin-Phycoerythrin. Following incubation for 30 min at room temperature, the assay beads were resuspended in 150 lL of Sheath Fluid and read on a Luminex FLEXMAP 3 D system (EMD Millipore Corporation, USA) with xPONENT software. Concentrations of cytokines were calculated using MILLIPLEX Analyst 5.1 software and reported as pg/ml of tear. Coefficient of variation (CV) and % of recovery rate of all the standards and controls for all assays fell within the acceptable value: intra-CV was less than 10%, inter-CV was less than 20% and % recovery was within 80-120%. The kit specified minimal detectible cytokine concentrations was as follows: IL-1b, 0.12 pg/ml; IL-6 0.13 pg/ ml; IL-8, 0.12 pg/ml; IL-10, 0.58 pg/ml; IL-17A, 0.28 pg/ml; IFNc, 0.32 pg/ml and TNFa, 0.15 pg/ml. Samples with cytokine levels below minimal detectible concentrations were assigned the value of '0'pg/ml.

Statistical analyses
Statistical analysis was performed only on data from 131 patients with at least 4ul of available tear sample, the minimum required for accurate results based on a previous study. 31 Distribution of cytokines were assessed by histograms and summarized using median (1st quartile, 3rd quartile), and the percent of patients with zero cytokine values due to their skewed distribution. As cytokines were measured from pooled tear samples from the right and left eye for each individual patient, scores of dry eye signs for the right and left eye of each patient were averaged. Correlations between cytokine concentration and patient symptoms and signs were assessed using the non-parametric Spearman correlation coefficient and the associated p-value.
To better describe the association of tear cytokines with dry eye signs and symptoms, patients were grouped into 4 groups based on quartiles of each cytokine concentrations measured from this study, and mean scores of dry eye signs and symptoms in each quartile were calculated. Scatterplots with super-imposed with locally estimated scatterplot smoothing (LOESS) curves were used to visualize the relation of each cytokine concentration with dry eye signs and symptoms. For all tests, p < 0.05 was considered to be statistically significant.

Sample collectability
Among 291 patients approached for collecting tear samples for baseline cytokine assessment, no tear sample was collected in 58 patients despite attempting to collect from both eyes ( Figure 1). In addition, cytokines could not be measured in 43 patients due to insufficient tears (<0.1 ul), and in 59 patients cytokines could be measured but might not be valid because tear volume !4 ul is needed to provide valid tear cytokine concentrations based on our previous research. 31 Only cytokine results of tear samples (OD and OS combined) that had !4 ul were included in our statistical analysis (Table 1). Patients with low tear volume (<4 ul) were older and had worse dry eye signs. Therefore, cytokine analysis of pooled tears (OD and OS) was completed on 131 subjects from the DREAM trial and each sample was analyzed for IL-1a, IL-1b, IL-6, IL-8, IL-10, L-13, IL-17A, IFNc, and TNFa. The mean volume of tears collected from both eyes of these 131 patients was 7.9 ul with SD 2.2 ul (range: 4 to 10 ul). The mean (SD) volume of tears collected from both eyes of patients from each of 10 clinical sites ranged from 5.5 (1.8) ul to 9.3 (1.9) ul.

Patient demographic and clinical characteristics
Patient demographic and clinical characteristics at the baseline visit of 131 patients that provided sufficient tears for cytokine measures are summarized in

Tear cytokine concentrations
There was a wide range of tear cytokine concentrations detected, with many cytokines being below minimal detectible concentration (assigned the value '0') in a substantial proportion of patients. Cytokines IL-17A, IL-1b, IL-10, and TNFa had the highest percent of patients with "0" value (21.4%, 21.5%, 27.7%, and 53.9%, respectively, Table 3). The overall median concentration and the 1 st and 3 rd quartiles of the 7 cytokines assayed are summarized in Table 3.

Tear cytokine concentrations and their correlation with DED symptoms and signs
Correlations between cytokine concentration and dry eye symptoms and signs symptoms are summarized in Table 5. No significant correlation was observed between OSDI scores and concentration of any of the seven cytokines. Weak but statistically significant correlations were found between three DED signs and concentration of some cytokines. Greater conjunctival staining score was negatively correlated with IFNc (r¼ À0.18, p ¼ 0.03). Greater Corneal  3.4 (1.5) b Schirmer's Test, mm/5 min, mean (SD) 12.8 (8.1) a 2 patients had more than one of the listed systemic diseases, so the total did not add to 131 subjects. b average of the 2 eyes since pooled tears were used for analysis staining negatively correlated with IL-17A (r¼ À0.24, p ¼ 0.006), IL-10 (r¼ À0.25, p ¼ 0.005, Online Supplement Figure 1) and IFNc (r¼ À0.33, p ¼ 0.0001, Online Supplement Figure 2). Longer TBUT positively correlated with the IL-17A (r ¼ 0.19, p ¼ 0.02). There were no significant correlations with Schirmer scores.
To further explore the relationship between cytokine levels and the scores for symptoms and signs, we grouped patients based on quartiles of the expression levels of each cytokine and calculated the mean scores for symptoms and signs in Table 6. While the mean values for the symptom score and signs were not the same across the quartiles of cytokine levels, they generally did not increase or decrease monotonically. For example, although the Spearman correlation coefficient in Table 5 for the association of IL-17A with corneal staining was À0.24 (p ¼ 0.006), the mean score (4.3) was higher for 2 nd quartile 2 (IL-17A levels of >1 to 8 pg/ml) as compared to mean score (3.3) of the first quartile 1(IL-17A levels of 0 to 1.6 pg/ml).

Discussion
Analysis of tear samples from the baseline visit of subjects with DED in the DREAM trial was completed to better understand immune mechanisms that are associated with disease severity.
Results of tear analysis collected from 131 well-characterized moderate to severe DED DREAM subjects at baseline demonstrated these key findings: a) many of the inflammatory cytokines tested had zero or very low concentration in the tears: IL-17A, IL-1b, IL-10, and TNFa, (b) patients with tear volume <4ul and, therefore not included in cytokine analysis, were significantly older and had more severe signs than those with greater tear volume, (c) cytokines did not correlate with each other in a generally acknowledged proinflammatory/anti-inflammatory pattern, (d) concentrations of the cytokines analyzed in this study did not correlate to patient DED symptoms, (e) some cytokines correlated with DED signs, but not in the expected directioni.e. higher concentration of inflammatory cytokines associated with greater ocular surface staining or shorter TBUT and/or shorter wetting of Schirmer's strips.    Tear cytokines have been reported in past studies 12,16,33,37,38,44 and these studies and the animal DED studies were used to select the cytokines to be tested in the DREAM trial, maximizing those that could be tested at the same time from each sample. While the cytokines were detected in patient samples in varying degrees (Table 3), we were surprised by the lack of correlation among pro or antiinflammatory cytokines. Our results may be a reflection of the heterogeneity of DED, the inability to obtain tears from those with more severe signs of DED, that these markers are not the best reflection of inflammatory response of the ocular surface in DED and/or reflect collection issues including diurnal variation, and reflex tears, though both of these collection issues were attempted to be controlled with standardized collection methods among the 10 sites that collected tears.
Past reports on tear cytokines have shown mixed results. McDonnell et al. study 34 with analysis of 217 tear samples, showed no difference between controls and subjects with mild to moderate DED for levels of the cytokines IL-1b, IL-6, IL-17A, IFNc, TNFa, IL-2, IL-4, IL-8, and IL-13, among others. These results are in contrast with Roda et al. 39 a meta-analysis of tear cytokine studies in DED showed overall that higher levels of Il-1B, Il-6, Il-8, Il-10, IFNc, TNFa in DED compared to tears from controls. Key issues that were noted in this review were the lack of consistency of cytokines findings between studies and no guidelines for cutoff values to differentiate abnormal from normal tear cytokine tear concentrations. The lack of a standardized approach to tear analysis was emphasized by the observation that of the 118 studies that met the most basic search criteria, only 13 met the inclusion criteria which included (1) study type: case-control; (2) population: patients having DED of any etiology; (3) purpose: measurement of pro-inflammatory mediator concentrations in tears; (4) outcome variables for qualitative synthesis: to report the different concentrations of tear cytokines between non-DED control subjects and DED patients.
We continue to search for markers that correlate with symptoms, given DREAM results and past studies that have not shown a correlation between symptoms and standard DED signs (vital dye staining, TBUT, Schirmer). It was hoped that tear cytokines might provide an objective minimally invasive metric that correlated with symptoms. Results did not reveal any cytokine or collection of cytokines that correlated with DED symptoms, as measured by OSDI. A few tear cytokines correlated with DED signs, but mainly unexpectedly in a negative direction-higher inflammatory cytokine concentrations were associated with less severe DED signs.
There was a wide variation in the concentration of the 7 cytokines studied in DREAM. While IL-8, IL-6 and IFNc were present at detectible levels in a high percentage of patient samples (99.2%, 96.9% and 93.9% respectively), TNFa, IL-10, IL-1b, and IL-17A were detected respectively only in 46.1%, 72.3%, 78.5%, and 78.6% of the samples. This is in keeping with the wide variability that has been observed in studies with tears from DED patients [14][15][16]32,33  and different concentrations found in DED and normal. 12,15,33,37,38,44 We did not find correlations of cytokines that could describe mechanisms of disease, such as increased proinflammatory cytokines, IL-1b, IL-6, IL-17A, IFNc and TNFa. For instance, IL-8, considered a proinflammatory cytokine, showed negative correlation, though weak, with the proinflammatory cytokines IL-17A, and IFNc. Past studies have shown increased concentration of IL-10, along with proinflammatory cytokines, in DED as compared to normal. 16,36,37,44 Strong support for the coexistence of pro-and anti-inflammatory cytokines in the present report comes from the study by Pinto-Fraga et al. studying the distribution of cytokines among DED patients. 34 They show that the levels of the proinflammatory cytokine IL-8 (p ¼ 0.04) were higher in the severe DED group, whereas the levels of proinflammatory cytokine IFNc (p ¼ 0.002), and anti-inflammatory cytokine IL-10 (p ¼ 0.04) were higher in the moderate DED group.
Though it is generally acknowledged that DED is a heterogenous disease and the different phenotypes may have differing pathogenesis, including variation in the inflammatory response, multiple processing issues may contribute to difficulties in interpreting tear cytokine levels and comparison among studies, basal vs. reflex tears, 47 effects of collection method, 48 diurnal variations, 46,49 low tear volume in DED 3,4 , lack of defined cut-off between "normal" and DED cytokine levels. 40 DREAM followed our published standard operating method for tear cytokine analysis. 31 However, there are not standard operating procedures in place that would better allow comparison of results among studies, and often details are not fully described in publications. 22 Strengths of the DREAM multi-site study of tear cytokines in subjects with moderate to severe DED are: standardized exam and sample collection, all samples analyzed by a central laboratory and one of the largest number of DED subjects evaluated to date for tear analysis. However, issues that could impact interpretation of results include inability to obtain processable samples from 53% of subjects due to low tear volume, especially those with more severe signs, diurnal variations in tear cytokine levels and/or reflex tearing.
In summary, tear analysis for IL-1b, Il-6, IL-8, IL-10, IL-17A, IFNc and TNFa in DREAM DED subjects did not demonstrate markers indicative of symptoms or signs of DED or show a universal marker for inflammation in these subjects with moderate to severe DED. Future research will be needed, including searching for other markers that reflect mechanisms of ocular surface disease in DED and adding new approaches to improve collection from all subjects, including those with more severe signs, and methods that are generalizable to all laboratories to analyze the available small volumes that are typically collected from the ocular surface in DED. Future work will need to be done with different groups, including normal subjects, to better understand tear cytokines and their relationship to eye pathology. This report highlights the promise and challenges of tear analysis to discover biomarkers and elucidate disease mechanisms. The goal remains to demonstrate the validity of minimally invasive objective metrics that define/classify DED and correlate with signs and symptoms.