Profiling ocular surface responses to preserved and non‐preserved topical glaucoma medications: A 2‐year randomized evaluation study

Use of topical glaucoma medications has been reported to cause ocular surface (OS) discomfort and inflammation. This study explores the profile of inflammatory cytokines and OS symptoms induced in response to preserved and non‐preserved drops.


| INTRODUCTION
Primary open-angle glaucoma (POAG) is the leading cause of irreversible blindness worldwide. [1][2][3] It affects approximately 73.6 million people globally with an estimated increase to 111.8 million in 2040 including 3.7% of United Kingdom and 2.6% of Oceania population aged between 48 and 92 years. 1,[3][4][5] Anti-glaucoma topical medications are widely used to control intraocular pressure (IOP) in POAG patients and are usually the first line treatment recommended. 1,2,4 Evidence suggests that patients receiving long-term topical anti-glaucoma drops may develop ocular surface (OS) disease (conjunctival hyperaemia, dry-eye like syndrome and irritability) or experience aggravation of pre-existing OS disease. 6 Preservatives are used to prolong the shelf life of glaucoma medications and some evidence suggested that their detergent properties increased the penetration of the active ingredients of the glaucoma drops. 6,7 Benzalkonium chloride (BAK) is a quaternary ammonium compound that is the most widely used preservative in topical ophthalmic medication. 6,7 It has been shown to have toxic effects on OS epithelia, inducing low-grade inflammation 7-10 and may pose a risk for failure of glaucoma surgery. 11,12 Polyquad (PQ) is polyquaternium-1 belonging to the family of polycationic polymers and is commonly used in contact-lens solution and topical ophthalmic medications. 6 Experimental studies have demonstrated that PQ does not induce cytotoxic effects on the OS. 13,14 However, the long-term effects of PQ-preserved topical glaucoma medications on OS epithelia and patient experience remains unknown.
The aim of this study was to evaluate the inflammatory effects of three different preserved glaucoma drop preparations on the OS of treatment naïve patients. This involved the profiling of inflammatory cytokines on the OS over a 24-month period and their correlation to changes in OS discomfort as measured with the OS disease index (OSDI) patient reporting instrument.   (Supplementary Table S1). Informed consent was obtained from each patient prior to enrolment.
A single-centre, pragmatic randomized controlled study design was adopted to explore the effect of preserved (bimatoprost 0.01% and travoprost 0.04%) and non-preserved (latanoprost 0.005%, timolol 0.5% dorzolamide 2%) topical anti-glaucoma medications on the expression of cytokines at the OS. Treatment-naïve glaucoma patients requiring topical hypotensive medication were enrolled and randomized into three groups, receiving either BAK, PQ or preservative-free (PF) drops. A computer-based randomization process provided by the Clinical Trials Unit of the University of Nottingham was used to determine study group allocation. Study visits were scheduled at baseline (prior to commencement of medication), 1, 3, 6, 12 and 24 months following therapy initiation. There are no published longitudinal studies that demonstrated inflammatory effects of the preserved and PF drops on the patient OS. Therefore, as a guide to estimate the sample size, we utilized the outcomes of our previous impression cytology (IC) studies 15-17 that were unrelated to glaucoma drops. It was estimated that n = 8 samples/group will have 85% power at alpha = .05 to detect twofold difference (SD = 0.5) in cytokine levels between the groups. As a precautionary step and considering the possibility of attrition of samples and patient drop out during the study period, we collected samples from total 36 patient with 12 enrolled in each treatment group.
At each visit, IC and tear samples were taken, and patients completed the OSDI questionnaire. Where monotherapy was insufficient to control IOP, drops with the same class of preservative were added for further IOP lowering. For PQ allocated patients only two PQ-preserved options exist (travoprost 0.004% monotherapy or travoprost 0.004%/timolol 0.5% combined therapy) if dual therapy was insufficient then the third medication added was PF.

| OSDI assessment
The standard OSDI questionnaire consisting of 12 questions for assessment of symptoms, functional limitations and discomfort related to dry eye was used. 18,19 The questionnaire was administered by trained ophthalmic staff at the start of study and then at each follow-up visits. The baseline questionnaire was elicited prior to study group allocation. In this questionnaire, each question is graded on a scale of 0 to 4, with 0 indicating 'none of the time' and 4 indicating 'all the time'. OSDI was calculated as: (sum of scores × 25)/(total number of questions answered). An OSDI score between 0 and 100 was obtained. Scores below 12 reflected normal OS health and above 33 indicated severe dry-eye condition.

| Sample collection
Samples were collected using a microcapillary glass tube (5 μL, Drummond Scientific Co., PA) from canthal and inferior fornices 20 and stored at −80 C. Conjunctival epithelium was collected using IC as previously reported. 15 Briefly, cellulose-ester discs of pore-size 0.45 μm (Millipore Corporation, Massachusetts) and diameter of 13 mm were cut into halves and applied to the upper and lower bulbar conjunctiva under topical anaesthesia (proxymetacaine hydrochloride 0.5%) for 10 to 15 seconds. The discs were gently peeled off and transferred into RNA lysis buffer (buffer RLT; Qiagen, UK).

| Total RNA isolation and cDNA synthesis
Total RNA was isolated from the IC samples using RNeasy mini kit (Qiagen) as per manufacturer's instructions. Then, 150 ng of total RNA was then reverse transcribed into cDNA using QuantiTect RT kit (Qiagen).

| Quantitative real-time polymerase chain reaction
Quantitative polymerase chain reaction (QPCR) was performed to measure the relative fold-change of genes of interest using TaqMan probe assays (Supplementary  Table S2; Applied Biosystems, UK). Each reaction was prepared to 20 μL final volume containing 5 μL of diluted cDNA (1 in 5 with nuclease-free [NF] water), 10 μL of TaqMan gene expression master mix, 1 μL of gene-of-interest, 1 μL of endogenous gene (hypoxanthine-guanine phosphoribosyltransferase-1) and 3 μL of NF water. Data were acquired on Mx3005p real-time PCR instrument (Stratagene/Agilent Technologies, UK) and analysed using delta-delta CT method as previously described. 21

| Cytokine measurement
Tear cytokines were measured using a BD-cytometric bead array (CBA) human inflammatory cytokines kit (BD Biosciences, UK) as previously reported. 22 The tear samples were diluted 1 in 25 to a final volume of 50 μL. To diluted tear samples, 50 μL of each capture beads (phycoerythrin [PE] conjugated) and PE-detection reagents was added. The mixture (150 μL) was incubated in the dark for 3 hours and then washed (×2 with wash buffer) via centrifugation. The resulting pellet was resuspended in 300 μL of wash buffer for beads analysis using BD-LSR II flow cytometer (BD Biosciences). The acquired data were further analysed using the FCAP array software version 3.0 (BD Biosciences) and plotted using the Prism 7.0 software (GraphPad Software Inc., California).

| Statistical analysis
QPCR and CBA data were statistically analysed using the Prism 7.0 (GraphPad Software Inc.). Two-way analysis of variance (ANOVA) was performed to evaluate the statistical differences between PF vs PQ, PF vs BAK and PQ vs BAK, respectively. As our aim was to compare the effect of preservatives with PF drops, we only reported the statistical data for the PF vs PQ and PF vs BAK groups.

| Masking
Masking of patients and clinicians to treatment was not possible. However, IC and tear samples were masked to group allocation and the analysis of biomarkers was undertaken in a masked fashion.

| Demographics and patient samples
A total of 36 patients were recruited based on the study criteria and randomized in to three groups: 12 in each of the PF, PQ group and BAK group. One participant from the BAK group withdrew after baseline sample collection. Abbreviations: BAK, benzalkonium chloride; PF, preservative-free; PQ, polyquad.

| Impression cytology
Here, 11 of the 35 IC samples could not be used due to low RNA quantity and/or RNA quality. Cytokine gene expression analysis using QPCR was performed on the following IC samples (n = 7 in PF, n = 8 in PQ and n = 9 in BAK).

| Tear samples
It was not possible to analyse 8 of the 35 tear samples due to low collection volume (samples <2 μL were not used). BD-CBA assay was only performed on complete patient samples (collection at all time-points) in each group (n = 7 in PF, n = 8 in PQ and n = 9 in BAK). Demographic information of patients whose samples (n = 24 in total had complete samples for all time-points) were analysed in this study are provided in Table 1. Supplementary Table S3 reports the IOP (mmHg), visual acuity (VA) and visual field mean deviation (MD) of the three patients' groups at baseline and 24 month. There was a reduction in mean IOP at 24 month in all groups compared to baseline. Mean group VA and mean MD in all groups remained unchanged at 24 month. All patients were started on single drop treatment (latanoprost in PF, F I G U R E 1 Gene expression analysis of cytokines in conjunctival impression cytology (IC) samples. Based on 24 patients with complete data collection. Relative fold change (RFC; Y-axis) of IL-6, IL-8, IL-1β, TNF-α, IL-10 and IL-12a in IC samples that are collected at different durations (in months; X-axis) from patients on anti-glaucoma medications that are preservative-free (PF) and preserved with polyquad (PQ) and benzalkonium chloride (BAK), respectively. RFC is represented as mean ± SD. Two-way analysis of variance (ANOVA) was performed to assess the statistical significance between PF vs PQ and PF vs BAK groups with alpha level set at P = .05 travoprost in PQ and bimatoprost in BAK) (Supplementary Table S3). In the PF group, all participants remained on a single drop throughout the study, one patient was started on an ocular lubricant (OL) at 12 month. In the PQ group, four of the eight patients started dual drop therapy (travoprost + timolol) beginning at 3 month. From 12 month onwards, two patients were on three drops with the third added drop (dorzolamide) being a PF formulation. In addition, four of the patients on single drop in PQ group also received an OL. In the BAK group, two patients were moved to dual drop therapy (bimatoprost + timolol) starting 12 month. None of the patients in BAK group were commenced on an OL throughout the study. Refer the Consolidated Standards of Reporting Trials (CONSORT) flow diagram (supplementary Figure S1) and the CONSORT checklist (supplementary Table S4) for further information.

| Interleukin 6
In the BAK group, IL-6 mRNA showed more than a twofold increase in 7/9 samples at all time-points compared to baseline, but this did not reach statistical significance compared to PF and PQ groups. On the other hand, the IL-6 mRNA levels in the PF and PQ groups were modestly increased at all time-points with no significant difference noted compared to baseline.

| Interleukin 8
In the BAK group, IL-8 mRNA (9/9 samples) was increased more than fourfold starting at 1 month with significant elevated levels noted at 3 month (4.76-fold increase, P = .0380) and 6 month (5.18-fold increase, P = .0445). Upon reaching 12 month, IL-8 was slightly reduced, but still remained elevated (9/9 samples) compared to PF and PQ groups. Unlike the BAK group, IL-8 mRNA levels did not change significantly in PF and PQ groups at any time-point compared to baseline.

| Interleukin 1β
In the BAK group, IL-1β mRNA increased in a timedependent manner from 3 month onwards with significantly increased level noted at 6 month (6.03-fold increase, P = .0023; 9/9 samples), and remained at a similar high level until 24 month compared to baseline and all time-points in PF and PQ groups. In the PQ group, there was a 2.92-fold increase in IL-1β mRNA levels at 24 months. There was no change noted in the PF group at all time-points.

| Interleukin 10
IL-10 mRNA expression demonstrated a 1.5-fold increase in PF group starting 1 month and remained unchanged until 24 month (4/7 samples). In the PQ group, IL-10 showed reduced expression until 12 month with slight elevation noted by 24 month (in 7/8 samples). In the BAK group, IL-10 mRNA (in 5/9 samples) was modestly increased in a linear fashion starting at 3 months but failed to reach statistical significance compared to baseline and all time-points in PF and PQ groups.

| TNF-α and IL-12a
TNF-α and IL-12a mRNA did not change significantly in either of the treatment groups.

| Quantification of cytokines in tear samples
The baseline level of IL-6, IL-8 and IL-1β (83.34 ± 55.63 pg/mL, 1119.40 ± 674.68 pg/mL and 2.06 ± 5.65 pg/mL, respectively) in all patients randomized into three groups were not significantly different ( Figure 2). The number of patients who showed cytokines level above the detection limit of the BD-CBA assay kit has been detailed in Supplementary Table S5.

| Interleukin 6
In the BAK group, IL-6 showed elevated levels starting at 6 month time-point with significantly increased levels noted at 24 month (161.06 ± 73.91 pg/mL; P = .0368; 9/9 samples) compared to PF group (41.69 ± 36.12 pg/mL; 5/ 7 samples). There was also a mild non-significant elevation in IL-6 levels in the PQ group beginning at 6 months, which remained at a similar level until 24 month timepoint.

| Other cytokines
IL-10, IL-12p70 (a heterodimer encoded by IL-12a gene) and TNF-α were detected at inconsistent low-levels in up to two of eight patient samples (refer Supplementary  Table S5 for details regarding number of patients that showed cytokines above the detection limit of BD-CBA assay kit).

| OSDI evaluation
The mean OSDI score was evaluated for the patients whose samples were used for cytokine estimation. All patients randomized to the PF, PQ and BAK groups showed similar OSDI scores at the baseline (mean OSDI = 5.22) and there was no significant difference between the groups (Figure 3).
For the PF group, the mean OSDI score was less than 12 for 6/7 patients at all time-points.

F I G U R E 3 Ocular surface disease index (OSDI). Box plot
represents the OSDI score for n = 24 patients with complete data collection. OSDI was calculated based on questionnaire completed by patients, randomized to preservative-free (PF) and preserved [polyquad (PQ) and benzalkonium chloride (BAK)] anti-glaucoma medications, at baseline (BL) and at each subsequent visit (1, 3, 6, 12 and 24 months; X-axis). Data are represented as mean (solid line in the box) with minimum and maximum OSDI values indicated as vertical bars. Two-way analysis of variance (ANOVA) was performed to assess the statistical significance between PF vs PQ and PF vs BAK groups with alpha level set at P = .05. * denotes P < .05 and *** denotes P < .001 F I G U R E 2 Quantification of inflammatory cytokines in tear samples. Based on 24 patients with complete data collection. Cytokines (IL-6, IL-8, IL-1β, TNF-α, IL-10 and IL-12p70) were measured in tear samples that are collected at different durations (in months; X-axis) from patients on anti-glaucoma medications that are preservative-free (PF) and preserved with polyquad (PQ) and benzalkonium chloride (BAK), respectively. Data for TNF-α, IL-10 and IL-12p70 were not shown due to insignificantly low levels. Number of patients that showed cytokines above detection limit were provided in Supplementary Figure S4. Quantity of cytokines (in pg/mL) is represented as mean ± SD. Two-way analysis of variance (ANOVA) was performed to assess the statistical significance between PF vs PQ and PF vs BAK groups with alpha level set at P = .05 For the PQ group, the mean OSDI score was more than 12 starting at 6 months in 4/8 patients, with 1 of 8 patients was scored more than 20 at 24 months.
For the BAK group, the mean OSDI score was more than 20 at 12 months (5/9 patients; P < .0001), with 3 of 9 patients was scored more than 30 at 24 months (P < .0001) compared to PF group.
As shown in Table 2, the linear correlation between OSDI and cytokines (OSDI vs IC-cytokines and OSDI vs tear-cytokines) was measured using the Pearson correlation test method. We have noted a significant correlation between OSDI and IC/IL-1β (r = .832, R 2 = .692 and P = .040), OSDI and IC/IL-10 (r = .925, R 2 = .856 and P = .008) and OSDI and tear/IL-1β (r = .899, R 2 = .808 and P = .014). However, the correlation between OSDI and tear level for TNF-α, IL-10 and IL-12p70 was not determined since these were only detected in insignificant levels (in up to two of eight samples).

| DISCUSSION
Topical drops are the primary treatment for patients with glaucoma. Evidence suggests that the presence of BAK preservative in drops may lead to ocular irritation, tearfilm instability, chronic inflammation, subconjunctival fibrosis and increase the risk for failure of glaucoma filtration surgery. 8,9,23 BAK-preserved drops have been implicated in the development of subclinical OS inflammation through increased levels of inflammatory mediators. 10,24,25 Unlike previous reports, which are often cross sectional, in this study, we have undertaken a three-way evaluation of different preservative effects on treatmentnaïve patients and profiled the temporal effect of these on cytokines in IC and tear samples.
Cytokines and chemokines are known to be secreted from a variety of cell types in response to infectious or inflammatory stimuli. These proteins may then trigger activation of intracellular signalling, enhancing cell proliferation/differentiation or lead to programmed cell death. 26 IL-1β and TNF-α are known to induce other cytokines/chemokines and profibrotic growth factors leading to sterile inflammation 27 and tissue fibrosis. 28 IL-6 and IL-8 (CXCL8) are pro-inflammatory cytokine with the ability to amplify autoimmune responses via the activation of infiltrating neutrophils and T-cells. 26 IC provides an alternative to biopsy or scraping to obtain conjunctival cells for cell-surface markers analysis using flow cytometry 25 and gene expression studies of host defence genes. 29 Normal conjunctival epithelium collected using the IC technique has been shown to express constitutive mRNA levels of IL-1, IL-6, IL-8 and TNF-α. 30 In response to inflammatory stimuli, these have been produced both at mRNA and protein level. 30 Elevated expression of IL-1β, IL-6 and IL-8 have been reported in the conjunctival IC samples of patients with Sjogren's syndrome keratoconjunctivitis sicca 31 and dryeye condition. 32 Conjunctival IC samples from patients on BAK-preserved anti-glaucoma drops show elevated expression of human leucocyte antigen-DR, IgE, IL-8 and two chemokine receptors (CCR4 and CCR5) 10,24,25 and increased levels of T helper (Th)-1 inflammatory cytokines have been demonstrated in tear samples of patients receiving preserved anti-glaucoma medications for more than 6 months duration. 33 BAK also induces cytokine production from trabecular meshwork cells in laboratory studies. 34 Our study demonstrates that mRNA for cytokines are constitutively expressed in conjunctival epithelium samples at baseline. An explanation for increased levels of IL-6, IL-8 and IL-1β in IC in the tear samples of patients randomized to the BAK group during the study could be that these cytokines are sensitive to any inflammatory stimuli and their homeostatic balance is key for maintenance of immune-privilege status of OS. 35 Although we have demonstrated the matching trend of both mRNA and protein expression of cytokines in samples from BAK group, it is likely that conjunctival epithelium may not be the only source for tear cytokines. Other possible sites for cytokine secretion in response to BAK are corneal epithelium, 36 tenon's fibroblast 37 and limbal-resident immune cells. 38 A previous study has demonstrated that the levels of tear IL-1β and TNF-α increased in patients on multiple BAK-containing drops compared to those on single drop therapy, that is, concentrating effect of BAK on ocular inflammation. 33 Here, all our patients randomized to the BAK group were on a single drop and have shown timedependent increase in IL-6, IL-8 and IL-1β levels.
The advantages of PF over preserved medication for glaucoma treatment have been widely reported. 6,39,40 Similarly, as evident from OSDI scoring, our patient group on PF and PQ containing drops have not complained of OS discomfort which can be correlated with reduced levels of inflammatory cytokines. On the other hand, those randomized to BAKpreserved drops had a high OSDI score, which strongly correlated to elevated levels of IL-1β. Recent phase IIIb clinical trials have demonstrated that switching patients from BAK-preserved latanoprost to PFtafluprost have improved OS tolerability and patient compliance within 12 weeks. 40 PQ has been previously shown to be safe and did not induce cytotoxic/inflammatory effects on OS epithelium in in vitro or in vivo studies. 13,14 Samples from our patient cohort using PQ-preserved drops have shown moderately high levels of IL-1β and IL-6 from 12 month onwards. This is also mimicked in the trends noted in the OSDI results. Our results suggest that chronic, long-term use of PQ preserved drops could induce delayed subclinical OS inflammation.
One criticism could be that some of the effects may be related to other non-preservative components of the drops. Previous in vitro studies have successfully compared the inflammatory and cytotoxic effects of widely used prostaglandin analogues: latanoprost, travoprost and bimatoprost. 41,42 Notably, none of these three analogues was shown to induce inflammatory markers in cultured conjunctival epithelial cells. Similarly, our study has also demonstrated that the addition of PF-timolol between 12 and 24 month duration as a second active drug has no significant effect on the cytokine levels; even in PF-latanoprost group. This is consistent with previous studies that demonstrated non-toxic effects of PF-timolol and PF-latanoprost on OS epithelium compared to BAKpreserved timolol and latanoprost drops, respectively. 30,39 Clinical studies have shown that patients on bimatoprost frequently develop ocular hyperemia. [43][44][45][46] This effect of bimatoprost 0.03% on OS has been confirmed to be due to vasodilation mediated via over-production of nitric oxide synthase, but not due to the induction of inflammatory cytokines. 43,47 Our observation that BAK is responsible for the increased expression of cytokines is consistent with numerous studies that evaluated the inflammatory markers in tears of patients treated with preservative containing and PF drops. [8][9][10]25,33,38 These published observations though cannot exclude the possibility of an unknown inflammatory effect from bimatoprost. The magnitude of effect if any exists in our BAK patient group is likely to be negligible and it does not detract from the observations or conclusions of this report. However, a future study is warranted to validate and compare the OS responses in patients on PF-bimatoprost 0.01% vs BAKbimatoprost 0.01%, PF-latanoprost 0.005% vs BAKlatanoprost 0.005% and PF-travoprost 0.04% vs BAKtravoprost 0.04%.
The strengths of this study are as follows: 1. A first of its kind, prospective evaluation of the changes in OS markers of inflammation in treatmentnaïve patients that were randomized to three preserved options available for glaucoma drops. 2. Baseline sampling of treatment-naïve patients for comparative analysis of cytokines over a period of 24 months have successfully reduced the effect of confounding variables. BL samples showed no significant differences in the baseline cytokine levels across the three groups. 3. Statistical correlation between the OS inflammatory changes and clinical symptoms as measured by the OSDI tool was successfully demonstrated.
The weakness is that we have recruited and analysed relatively small numbers (due to poor sample quality and/or quantity). Although even with this small sample size, we have demonstrated a clear effect on inflammatory markers in IC and tear samples. The use of different prostaglandin analogues in each of the preservative groups introduces slight uncertainty into the possibility that some of the inflammatory effects in the BAK group may be related to active ingredient, bimatoprost.
In conclusion, our results have demonstrated that patients on BAK preparation showed increased OS inflammation compared to those on PF and PQ-preserved preparations. The inflammatory effect of BAK was detectable starting 3 months and sustained for the duration of the 24 month follow-up. The strong correlation seen between OSDI scores and increased levels of cytokines in BAK group supports the conclusion that BAK induces ocular discomfort in patients. [7][8][9][10] There is a suggestion that PQ is also capable of producing delayed OS inflammatory response, which may lead to OS discomfort.