The Mechanisms of Neuroprotection by Topical Rho Kinase Inhibition in Experimental Mouse Glaucoma and Optic Neuropathy

Abstract

Purpose: The purpose of this study was to delineate the neuroprotective mechanisms of topical 2% ripasudil (Rip), a Rho kinase (ROCK) inhibitor.

Methods: In 340 mice, scheduled 2% Rip or balanced salt solution (BSS) saline drops were intermittently, unilaterally delivered. Intracameral microbead glaucoma (GL) injection increased intraocular pressure (IOP) from 1 day to 6 weeks (6W), whereas other mice underwent optic nerve (ON) crush. Retinal ganglion cell (RGC) loss was assessed using retinal wholemount anti-RNA Binding Protein with Multiple Splicing (RBPMS) labeling and ON axon counts. Axonal transport was quantified with β-amyloid precursor protein (APP) immunolocalization. Micro-Western (Wes) analysis quantified protein expression. Immunofluorescent expression of ROCK pathway molecules, quantitative astrocyte structural changes, and ON biomechanical strains (explanted eyes) were evaluated. ROCK activity assays were conducted in separate ON regions.

Results: At 6W GL, mean RGC axon loss was 6.6 ± 13.3% in Rip and 36.3 ± 30.9% in BSS (P = 0.04, n = 10/group). RGC soma loss after crush was lower with Rip (68.6 ± 8.2%) than BSS (80.5 ± 5.7%, P = 0.006, n = 10/group). After 6W GL, RGC soma loss was lower with Rip (34 ± 5.0%) than BSS (51 ± 8.1%, P = 0.03, n = 10/group). Axonal transport of APP within the unmyelinated ON (UON) was unaffected by Rip. Maximum principal mechanical strains increased similarly in Rip and BSS-treated mice. Retinal ROCK 1 and 2 activity was reduced by Rip in GL eyes. The pROCK2/ROCK2 protein ratio rose in the retina of BSS GL eyes, but not in Rip GL eyes.

Conclusions: Topical Rip reduced RGC loss in GL and ON crush, with suppression of ROCK signaling in the retina and ON. The neuroprotection mechanisms appear to involve effects on both RGC and astrocyte responses to IOP elevation.

Rho kinase (ROCK) inhibitors were originally proposed by Epstein and colleagues as intraocular pressure lowering (IOP) agents due to their ability to alter trabecular meshwork cell cytoskeletal architecture and to induce significant IOP reduction in preclinical models.¹ To date, two topical ROCK inhibitors have been approved as IOP lowering medications, one in the United States (netarsudil) and one in Japan (fasudil).² Topical netarsudil treatment can cause conjunctival erythema, subconjunctival hemorrhage, and corneal pigment deposition which limits adherence and leads to a drop of cessation.³ Off target effects could, however, provide additional benefit in glaucoma therapy in the form of IOP-independent neuroprotection. 

Whereas neuroprotective effects of ROCK inhibitors have been reported in glaucoma models, the cellular target(s) have not been clearly defined. In addition to lowering IOP, the ROCK inhibitor, ripasudil, administered orally⁴ or topically^5,6 reduces the loss of retinal ganglion cells (RGCs) after optic nerve (ON) crush. ROCK inhibition with fasudil, netarsudil, H-1152, and C3 transferase was potentially neuroprotective in ON injury experiments in mice,⁷ which further validated ROCK signaling as a viable neuroprotective target.^6,8–11 Although ROCK inhibitor levels may not reach therapeutic levels when delivered topically in large eyes, in mouse eyes, topically applied ROCK inhibitors produced concentrations that were orders of magnitude higher than the half-maximal inhibitory concentration (IC₅₀).^12,13 In fact, RGCs, microglia, astrocytes, vascular cells, and scleral stromal cells have each been proposed as potentially important targets in ROCK inhibitor-mediated neuroprotection.^6,14–15 

ROCK 1 and 2 are cytosolic serine/threonine kinases that integrate signals from cell surface receptors, and upon activation, control essential cellular processes.¹⁶ ROCKs are activated through interactions with Rho GTPases. Through their downstream kinase activity, they regulate fundamental cellular processes including stress fiber formation, cell migration, and cell proliferation.¹⁷ Because ROCK signaling plays a central role in these basic cell functions, it is necessary for normal development.¹⁸ Pharmacologic inhibition of ROCK activity has widespread effects on multiple different organ systems.¹⁹ ROCK activity additionally drives diverse pathologies and pharmacologic ROCK inhibition has shown promise in fibrosis prevention, cancer treatment, neuroprotection, and neuroregeneration.^6,14,20–23 Due to the pervasive effects of pharmacologic ROCK inhibition on many cell types, the mechanisms of the therapeutic effect of ROCK inhibitors in different disease states is not fully known. 

We investigated the effects of topical ripasudil in acute and chronic experimental glaucoma and ON crush in mice with the goal of further elucidating its neuroprotective mechanism(s). In addition to evaluating RGC survival with IOP elevation and ON crush, we examined effects of ripasudil treatment on optic nerve head (ONH) biomechanics by assessing axonal transport of amyloid precursor protein (APP) and strain response of the astrocytic lamina (AL) with globe inflation studies. We additionally examined ONH and ON ROCK activity using enzyme activity assays, micro-Western (Wes) blot analysis of key molecules in the mechanosensation pathways, and ON head immunolabeling to delineate ROCK targets in these disease models in a cell type specific manner. 

Methods

Animals

We utilized 2 to 5 month old mice (N = 340) treated in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, using protocols approved and monitored by the Johns Hopkins University School of Medicine Animal Care and Use Committee. Two strains were used: CD1 mice (Charles River, Wilmington, MA, USA) and mice ubiquitously expressing green fluorescent protein, using hemizygotes from the cross of C57BL/6J with C57BL/6-Tg (CAG-EGFP)1Osb/J (B6-eGFP, Jackson strain #3291; The Jackson Laboratory, Bar Harbor, ME, USA). The studies carried out with each mouse strain are shown in Supplementary Table S1. 

Topical Ripasudil Treatment

Animals received eye drops of 2% ripasudil, pH adjusted to 7.3 by addition of NaOH (ripasudil hydrochloride; Sigma Aldrich, St. Louis, MO, USA) in balanced salt solution (BSS), BSS alone eye drops, or were bilaterally untreated (naïve). Pilot experiments with unbuffered 2% ripasudil led to corneal damage. For experimental microbead glaucoma (GL) mice, one 5 µL drop (either ripasudil [Rip] or BSS) was delivered unilaterally per day (100 µg per day) for 3 days prior to and on the day of intracameral microbead injection, and for 3 more days. Tissue was collected at 3 days or 6 weeks after microbead injection. For experimental GL mice followed for 6 weeks, additional drops were given 7 and 14 days after microbead injection. Thus, 7 doses of ripasudil or BSS were given to 3 day GL mice (3D GL) and 9 doses to 6 week GL mice. For animals receiving ON crush, drops were given daily for 3 days before crush, on the day of crush, daily for 3 days after crush, and at 7 and 14 days after crush (total = 9 doses). 

Animal Procedures

Two ON damage models were used, either unilateral intracameral microbead injection to induce experimental glaucoma with IOP elevation (GL)^24,25 (modified from Sappington et al.)²⁶ or ON crush. For microbead injection, the mice were anesthetized with intraperitoneal ketamine (50 mg/kg), xylazine (10 mg/kg), and acepromazine (2 mg/kg) and topical 0.5% proparacaine hydrochloride eye drops (Akorn Inc., Buffalo Grove, IL, USA). One eye underwent intracameral injection of a 50:50 mixture of 6 µm and 1 µm diameter microbeads (4 µL;, Polybead Microspheres; Polysciences, Inc., Warrington, PA, USA), followed by 1 µL of viscoelastic compound (10 mg/mL sodium hyaluronate; Healon; Advanced Medical Optics Inc., Santa Ana, CA, USA). Tissue collected at 1 or 3 days after microbead injection underwent immunohistologic assessment and axonal transport block analysis, Wes analysis and ROCK activity assay, while tissue collected at 6 weeks in the GL model were used for assessment of RGC loss in the ON and retina, and for immunohistologic analyses. 

For the ON crush model, in 20 mice, the orbital ON was compressed for 1 second with cross-action forceps cephalad to the entry of the central retinal artery into the ON. Tissue was collected after 2 weeks. Fellow, untreated eyes and five bilaterally untreated mice were used as controls for RGC count comparison. 

For IOP measurement, the mice were anesthetized with either an intraperitoneal injection of ketamine, xylazine, and acepromazine (dose above) prior to microbead injection and at sacrifice, or isoflurane for IOP reading and eye drop delivery, using a RC2-Rodent Circuit Controller (VetEquip, Inc., Pleasanton, CA, USA), delivering 2.5% of isoflurane in oxygen, at 500 cc/minute for inhalation. The TonoLab (Icare Finland Oy, Inc., Vantaa, Finland) was used for IOP measurements. The IOP was measured consistently in the morning at the following time points: on the first day before drops were given, 4 hours after first drop administration, and then daily for 6 days. IOPs were also measured on days 7 and 14 in the crush and 6-week GL experiments. 

Tissue was collected for histology at 1 of 4 time points: 1 day, 3 days, or 6 weeks after GL, or 14 days after ON crush, under intraperitoneal anesthesia (as above), followed by transcardial perfusion (4% paraformaldehyde in 0.1 M Sorensen's phosphate buffer, Na₃PO₄, pH = 7.2) and concurrent exsanguination. Following euthanasia, the eyes and nerves were enucleated and dissected for one of the following: cryopreservation (histology), embedding in epoxy resin (axon counts), kept in buffer until wholemount staining (histology). Mice for Wes and ROCK activity assay (1 day or 3 days after GL) were euthanized by intraperitoneal anesthesia overdose (as above) before the eyes were enucleated and collected tissues were immediately frozen at –80°C. 

Quantification of RGC Axons and Somas

For quantification of axons in the myelinated optic nerve (MON) in GL mice, after fixation by perfusion, the ON was removed 1.5 mm behind the globe, post-fixed in 1% osmium tetroxide (OsO₄), dehydrated in ascending alcohol concentrations, and embedded in epoxy resin at 60°C for 48 hours. One µm thick cross-sections of the MON were stained with 1% toluidine blue. For each nerve, low and high-power digital images were taken using a Cool Snap camera with Metamorph software (Molecular Devices, Sunnyvale, CA, USA). Five 40 × 40 µm fields were acquired, corresponding to a 9% sample of the total nerve area. Masked observers deleted non-axonal elements from each image, generating an axon density from the software. The average axon density/mm² was multiplied by the individual nerve area to estimate total axon number. Axon loss was calculated by comparing GL eyes to the mean axon number in pooled, fellow eye ONs of each mouse group. 

For quantification of RGCs in retinal whole mounts, retinas were fixed with 4% paraformaldehyde, then incubated for 5 days at 4°C in antibody to RNA Binding Protein with Multiple Splicing (RBPMS), then for 1 hour at room temperature in secondary antibody (Supplementary Table S2). Nuclei were stained with DAPI. Samples were imaged with a Zeiss Apotome Confocal Microscope (Zeiss MicroImaging LLC, Thornwood, NY, USA) using a 10 × objective at a resolution of 2048 × 2048 pixels. Manual quantification of RGCs positive for RBPMS was performed in masked fashion within a 2000 × 1000 µm area within the superior retinal quadrant, corresponding to a 10% sample of the total retinal area using FIJI software (ImageJ; National Institutes of Health, Bethesda, MD, USA). 

Analysis of Axonal Transport

To assess axonal transport of β-APP, we studied mice in 4 groups: 3D GL with either ripasudil or BSS eye drop treatment, fellow eyes, and bilaterally untreated naïve eyes, using a published protocol.^27,28 In brief, longitudinal cryosections of mice were labeled with antibody to APP, imaged by laser scanning microscopy, as described above, then imported into FIJI. The polygon selection tool outlined four regions of interest; retina, prelamina, unmyelinated optic nerve (UON), and myelinated optic nerve (MON), as published. Images were exported in grayscale format into MATLAB to generate pixel intensity values by region. We calculated the mean, median, 97.5th and 99th percentile values of the APP label intensity. The 97.5th and 99th percentile control values were used as benchmarks for comparison to the distribution of the 3D GL eyes. Based on these thresholds, we generated the following parameters: mean intensity of the pixels brighter than the 97.5th percentile of the control and mean intensity of the pixels brighter than the 99th percentile of the controls.²⁷ The pixel intensity values formed right skewed distributions and nonparametric Mann Whitney tests were used to compare values among regions and treatments. 

Immunohistologic Study

Fixed tissues were placed in increasing concentrations of sucrose solution, then cryopreserved in a mixture of sucrose and TissueTek optimum cutting temperature (OCT) compound solution (Sakura Finetek USA, Inc., Torrance, CA, USA). Eyes were cryo-sectioned at 8 µm thickness for immunofluorescent staining with antibodies listed in Supplementary Table S2 and imaging on a Zeiss 710 or 880 confocal laser scanning microscope (Carl Zeiss Microscopy, White Plains, NY, USA) with either Plan-Apochromat 20×/0.8 M27 or 40× objective lenses to produce either 8-bit or 16-bit images, using independent channel acquisition with a combination of excitation settings: 488 nm, 568 nm, 647 nm, and 405 nm for DAPI. For anti-p-ROCK1 analysis, a masked observer graded slides by the intensity of the p-ROCK1 label in the UON among the groups: Rip GL, BSS GL, and bilaterally untreated eyes. Averages of the label intensity were compared between groups. 

Micro-Western Blot Analysis

Automated Western blot analysis with the ProteinSimple Wes instrument (Bio-Techne, San Jose, CA, USA) was used for protein quantification, due to the small amounts of protein in mouse tissues. Protein quantification was assessed in bilaterally untreated naïve controls, ripasudil only, and 3D GL eyes treated with BSS or ripasudil drops. Eyes were enucleated and ON sheath was removed. ON was separated into three regions: UON, myelin transition zone (MTZ), and MON (the MTZ was discarded). Tissues collected (retina, UON, and MON) were frozen and kept at –80°C until use. The GL treated eyes were initially placed into two groups, a higher IOP group and a lower IOP group, based on peak IOP and largest GL-to-fellow-eye IOP difference. For most comparisons, there were no statistically significant differences between these groups, and they were considered together. Six samples from each region were pooled. Samples were homogenized in protein extraction lysis buffer (T-PER; Pierce Biotechnology, Waltham, MA, USA) supplemented with protease inhibitor (Sigma Aldrich, St. Louis, MO, USA), 1 mM phenylmethylsulfonyl fluoride, 1 mM Na₃VO₄, and 1 mM NaF at 0°C. The lysate was agitated for 1 hour at 4°C, followed by centrifugation at 15,000 × g for 15 minutes at 4°C before the protein containing supernatant was removed and aliquoted. Protein concentration was measured using the Pierce BCA kit (Thermo Fisher, Waltham, MA, USA). Protein concentration for each well was adjusted to 1.2 µg/µL using 0.1 × buffer solution (calculated using provided 10 × sample buffer from Wes Separation Capillary Cartridges for 12–230 kDa kit (#SM-W004, Bio-Techne). Samples were processed using the Wes protocol using the high dynamic range (HDR) detection profile for chemiluminescence. Data were comprised of the area under the immunoreactivity curve for each molecule, as identified by its specific antibody at the appropriate migration location using the Compass for SW software (Bio-Techne, San Jose, CA, USA). Data were normalized to the median values for GAPDH from three or more matched samples. The data allowed assessment of the comparative amount of phosphorylated to total protein, the effect of the GL model, and the effect of ripasudil treatment. We tested the following molecules in Wes analysis, GAPDH, ROCK1, ROCK2, pROCK2, cofilin, p-cofilin, RhoA, and p38 MAPK, using antibodies listed in Supplementary Table S2. 

Rho Kinase Activity Assay

A commercial ROCK activity assay (#STA-416; Cell Biolabs, San Diego, CA, USA) was used in bilaterally untreated naïve mice, ripasudil only treated eyes, and in GL eyes concurrently treated with BSS or 2% ripasudil drops. Mice were treated for 3 days prior to GL, on the day of GL induction and 2 hours before euthanization, which occurred 24 hours after GL injection in this group. Micro-dissected specimens were prepared from the retina, UON, and MON (after clearing and discarding the meninges, peripapillary sclera, and MTZ). Samples were placed into Eppendorf tubes, frozen at –80°C, then tested after thawing with ROCK substrate, MYPT1, labeled with primary antibody to p-MYPT1, and labeled with horseradish peroxidase conjugated secondary antibody. Samples were treated with substrate solution to assess phosphorylation by colorimetric immunoassay. Absorbance of each well was completed on a spectrophotometer (BioTek Synergy H1 microplate reader with Gen5 software version 3.08; Agilent Technologies Inc., Santa Clara, CA, USA) using 450 nm wavelength. The positive control was purified ROCK2, with which a titration curve was constructed. A second order polynomial line of best fit was calculated for the titration curve. Absorbance values were input to calculate the protein concentration for each well. Statistical comparisons were made to bilaterally untreated naïve eyes. 

Inflation Testing for Strain Estimation

Inflation tests of explanted eyes were performed according to a published method²⁹ comparing images at 10 mm Hg to those at 30 mm Hg with the Zeiss 710 laser scanning microscope and Coherent Chameleon Ultra II multiphoton laser (Coherent Laser Group, Santa Clara, CA, USA). In brief, the following B6-eGFP mouse groups were tested: naïve, ripasudil drops alone, and 3D GL (ripasudil or BSS drops). Mice were euthanized as described above under general anesthesia, the eyes were explanted and the ONs were removed just posterior to the sclera. The cornea of the intact eye globe was glued to a plastic holder and the anterior chamber was cannulated with a needle connected to a BSS reservoir whose height was used to regulate IOP during inflation testing. The holder and eye were maintained in phosphate-buffered saline at room temperature throughout the experiment. Two images were acquired with the 710 microscope with 20x W Plan Apochromat objective, numerical aperture 1.0 at IOP = 10 mm Hg and one at IOP = 30 mm Hg, with 15 minutes of equilibration at each pressure, using a two-photon (TPF) laser source, excitation wavelength of 895 nm, and collection of the emission signal with 500 to 550 nm band-pass filter. Z-stacks of 40 to 100 µm spaced every 1 µm were collected with an estimated total x- and y- resolution of 0.415 µm/pixel. Image post-processing was performed with a deconvolution algorithm (Huygens Essential, Scientific Volume Imaging, The Netherlands). Digital volume correlation (DVC)³⁰ was used to compute the displacement field from the post-processed images. The maximum principal strain, E_max, in the x-y plane and the maximum shear strain, γ_max, in the x-y plane were computed. 

ON Astrocyte Structure in Cross-Section

In naïve mice and in 3D and 6W GL mice (total n = 37), cross-sections were cut from retina to MON and sections containing the UON were labeled with anti-glial fibrillary acidic protein (GFAP) and DAPI. Using methods previously published,³¹ these sections were analyzed and quantified for astrocyte structural parameters and for nuclear cluster area and density. Four non-GL eyes were treated with 2% ripasudil drops only and 8 eyes were bilaterally untreated. Two cryosections were analyzed from each UON using the C-Apochromat 40x/1.2 NA W corrected, FCS objective, 2 × 2 tiling with unidirectional scanning, and 12% tile overlap. Images were converted in FIJI and the imaged channels were split. A custom program in MATLAB (version 9.14.0.2254940 (R2023a) Update 2) was used to skeletonize astrocyte process to measure thickness, length, and connectivity using data from the captured images. The analysis was performed for the overall ON cross-section and for three regions of the ON: the outer rim, a mid-peripheral zone, and the central nerve. 

Statistical Methods

Data were compared as mean ± standard deviation or median values. Paired or unpaired t tests were used for normally distributed data and Mann Whitney tests were used for nonparametric data. One-way ANOVAs were used to compare more than two groups with post hoc Tukey’s honestly significant difference (HSD) test. The significance level was P ≤ 0.05. 

Results

Intraocular Pressure

Topical ripasudil lowered IOP at 4 hours after administration in non-GL eyes by 6.1 ± 0.95 mm Hg, a 41% reduction from baseline (P ≤ 0.001, t-test; Supplementary Fig. S1). However, by 24 hours after ripasudil administration, IOP was not significantly different from untreated fellow eyes at any of the 3 days measured (all P > 0.5). On multiple measures, after bead injection, the ripasudil treated eyes’ IOP was not statistically significant from BSS treated eyes, at either 4 hours or 24 hours after drops were administered (see Supplementary Figs. S1–S4). The IOP differences between GL or fellow eyes were not significantly different at any time point between ripasudil and BSS groups in the mice followed for 3 days or 6 weeks. 

IOPs in ON crush eyes (regardless of the eye drop treatment) were significantly lower than baseline at 1 and 2 days after crush compared to bilaterally untreated naïve eyes (see Supplementary Fig. S4; all P ≤ 0.001, ANOVA). All eyes returned to mean baseline IOP by 1 week after crush (see Supplementary Fig. S4). As in GL eyes, 4 hours after ripasudil drops in ON crush eyes, mean IOP was lower by 2.6 mm Hg from baseline (P = 0.04) but not different than BSS treated crush eyes (P = 0.46). Treatment with ripasudil drops alone did not significantly decrease mean IOP at 24 hours after delivery in ON crush eyes. 

Axial Length and Width

Axial length and width increased 3 days after GL compared to fellow eyes. In BSS treated eyes, axial length and width increased by 11% each (length 3.3 ± 0.1 mm to 3.7 ± 0.1 mm, and width 3.3 ± 0.1 mm to 3.7 ± 0.1 mm, both P ≤ 0.001, difference from controls, t-test). The ripasudil treated eyes also enlarged: 14% in length and 9% in width (length 3.4 ± 0.1 mm to 3.8 ± 0.2 mm, and width 3.4 ± 0.1 mm to 3.7 ± 0.2 mm, differences from controls, both P ≤ 0.001), with significantly greater length increase in ripasudil treated than BSS treated (P ≤ 0.02). 

In eyes followed for 6 weeks after bead injection, the BSS treated GL eyes had increased length by 12% and width by 10% (length 3.5 ± 0.1 to 3.9 ± 0.2 mm, and width 3.4 ± 0.1 to 3.8 ± 0.2 mm, differences from fellow eyes, both P ≤ 0.001, ANOVA). The ripasudil treated GL eyes also had significant length and width increases, 10% and 8%, compared to controls (length 3.5 ± 0.1 to 3.8 ± 0.2 mm, and width 3.4 ± 0.1 to 3.7 ± 0.2 mm, P ≤ 0.01 and P ≤ 0.04), and these mean increases in length were less than those of the BSS group (P = 0.02 and P = 0.05, Mann Whitney test, respectively). 

Axonal Transport Blockade

Mice in this analysis were studied 3 days after GL. Two groups of GL eyes were assessed: those that received ripasudil drops and those that received BSS drops, with each compared to their fellow eyes. The median APP brightness in the fellow eyes was highest in the retina and lowest in the MON (P = 0.057, ANOVA, n = 10 eyes per region; Fig. 1, Supplementary Table S3). In the UON, the fellow eyes of ripasudil-treated mice had a median APP brightness that was not significantly different from naïve controls, but significantly higher than BSS fellow eyes (P = 0.023, ANOVA). We have previously shown that the mean brightness >97.5 percentile and >99 percentile are measures indicating abnormally bright, local areas of APP accumulation.²⁷ In both ripasudil and BSS GL eyes, these two measures in the UON were significantly higher than their fellow eyes (all P ≤ 0.004, n = 10 per group). There was no significant difference between ripasudil and BSS GL eyes in these local accumulation measures. Nor were there significant differences in the other 3 regions in these parameters, except for higher mean >97.5 percentile for ripasudil-treated GL eyes than their fellow eyes in the prelaminar ON (P = 0.0006). 

RGC Axon Loss and RGC Soma Loss With GL and ON Crush

The mean axon number in bilaterally naïve ON was 53,103 ± 12,909. Six weeks after GL, eyes treated with only ripasudil had no significant loss in mean axon number, with a value only 8.3% below pooled fellow eye control ON and 6.6% below naïve controls (P = 0.10 and P = 0.46, respectively, n = 10 per group; Table 1). The BSS-treated GL eyes lost 36.3 ± 30.9% of axons compared to pooled fellow eyes and naïve controls (P = 0.006 and P = 0.009, respectively, n = 10 per group). The difference in mean axon loss was significantly greater in BSS than in ripasudil-treated mice (P = 0.04). Both the BSS and ripasudil treated groups had significant elevations of IOP after bead injection for the first 5 of 6 measured days (all P < 0.01; see Supplementary Fig. S1). The mean IOP elevation in GL eyes compared to fellow eyes did not significantly differ between BSS and ripasudil-treated eyes (P ≥ 0.19). Visual confirmation showed protective effects of Ripasudil (Supplementary Figs. S5A–S5C). 

Table 1.

 

Axon Loss for 6-Week Experimental Glaucoma (GL) With BSS or Ripasudil (Rip) Drop Treatment

 

Data are mean ± standard deviation. Eye drop treatments were either BSS (BSS GL) or ripasudil (Rip GL). Percent axon loss is compared to pooled bilaterally naïve eyes. ⁺P value 0.10, ^* P value 0.02 from t-tests. 

Counts of RGC bodies labeled for RBPMS in retinal whole mounts were calculated as RGC density (representative images; see Supplementary Figs. S5D–S5F). Six-week GL eyes treated with ripasudil had 34% lower density than contralateral eyes, whereas BSS GL eyes had 51% lower density (group difference, P = 0.03, Wilcoxon rank sum test; Fig. 2). Ripasudil drop treatment alone did not change RGC density compared to naïve mice (P = 0.72). ON crush samples treated with ripasudil had 68% lower RGC density than contralateral eyes, whereas BSS treated crush nerves had 80% lower density (difference between ripasudil and BSS loss, P = 0.006, Wilcoxon rank sum test; see Fig. 2). 

Figure 1.

 

APP analysis using median brightness and mean >97.5 percentile pixel brightness. (A) Median APP brightness and (B) mean >97.5 percentile by region for each treatment group. APP median brightness was highest in retina and declined sequentially from prelamina to UON to MON. Median brightness in prelamina and UON was maintained at a higher level in the fellow eyes of ripasudil-treated GL mice than in fellow eyes of BSS-treated GL mice. The mean brightness values >97.5 percentile were higher in GL eyes than fellow eyes in both ripasudil and BSS treated eyes in UON, indicating axonal transport blockade in both groups. Naïve = bilaterally untreated naïve, RF = ripasudil fellow eye, R GL = ripasudil GL eye, BSS F = BSS fellow eye, BSS G = BSS GL eye. P values, ANOVA, * P ≤ 0.05, *** P ≤ 0.001.

Figure 2.

 

Percent RGC cell density loss by RPBMS labeling for 2W ON crush and 6W GL groups. Percent RCG soma loss with eyes treated with either ripasudil or saline drops with 6W GL or 2W crush. Group means with standard errors. * P ≤ 0.05, *** P ≤ 0.006, Wilcoxon rank sum test compared to contralateral eyes.

Rho A Kinase Activity Assay

At 2 hours after eye drop delivery, the ripasudil drops alone group had lower ROCK activity compared to bilaterally untreated eyes in the UON and MON (P = 0.05 and P = 0.01 respectively; Fig. 3) and a trend toward a reduction in the retina (P = 0.09). In the retina, 24 hours after GL (2 hours after the last ripasudil drop treatment), the ROCK activity for the ripasudil GL group was significantly lower than bilaterally untreated eyes (P = 0.01). By contrast, the BSS GL group activity was significantly higher in the retina (P = 0.03). In the UON, the BSS GL group was significantly higher than bilaterally untreated naïve eyes (P = 0.01), whereas ripasudil GL eyes were not different from untreated naïve eyes. The activity was significantly lower in ripasudil GL than in BSS GL groups in retina and UON (retina P = 0.004, UON P = 0.03, and MON P = 0.61; Supplementary Table S4). 

Figure 3.

 

ROCK activity 24 hours after GL. Legend: Ratio comparisons with naïve eyes to two GL groups and ripasudil alone treated eyes. N = 10, t-test, * P ≤ 0.05, ** P ≤ 0.01 for comparison of each group to bilaterally untreated control eyes (equal to 1).

Biomechanical Strain Data

The ex vivo strain response of the AL with GL under inflation from 10 to 30 mm Hg was compared among the 4 groups: bilaterally untreated, ripasudil drops alone, and 3D GL with either ripasudil or BSS drop treatment (Fig. 4). The maximum principal strain, E_max, and the maximum principal shear (γ_max) both significantly increased in the ripasudil GL group compared to bilaterally untreated naïve eyes (P = 0.0003 and P = 0.0004, respectively; Table 2). Additionally, the increases in E_max and γ_max in the BSS GL group over bilaterally naïve eyes were also significant (P = 0.020 and P = 0.035, respectively). There was somewhat greater E_max and γ_max strains in ripasudil drop alone eyes over bilaterally naïve eyes, but the difference did not reach significance (P = 0.084 and P = 0.095, respectively). When the ripasudil GL group data were compared to the ripasudil drop alone group, the differences in E_max and γ_max were not significant. Two eyes in the ripasudil GL group and one eye in the BSS GL group were excluded as outliers as they did not inflate. 

Figure 4.

 

Strain responses of astrocytic lamina in eGFP mice comparing 4 groups; naïve, ripasudil alone, ripasudil GL, and BSS GL. Strain analysis (E_max and γ_max) for various treatment groups and control samples; bilaterally untreated naïve, ripasudil only, and GL samples BSS + 3D GL, Ripasudil + 3D GL. Significant difference between bilaterally naïve both GL groups for both strains.

Table 2.

 

E_max and γ_max Strains in Bilaterally Untreated, Ripasudil Alone, and Both BSS GL and Ripasudil GL Groups

 

Astrocyte and ON Structural Changes

Quantification of astrocyte structure in the UON in GFAP-labeled cross-sections were performed in naïve, ripasudil drop alone treated, and in the two GL groups, quantifying multiple structural parameters (Supplementary Table S5, representative images generated in Supplementary Fig. S6). For the GL groups, the analysis was conducted at both 3 days and 6 weeks after GL. The ON area was significantly greater than bilaterally untreated naïve eyes at 3D GL, but was similar to control in specimens at 6 weeks (Supplementary Table S6; P ≤ 0.05). Rip GL and BSS GL groups did not significantly differ in the ON area. 

The fraction of the ON occupied by astrocyte processes was 0.50 in bilaterally untreated naïve eyes and 0.51 in ripasudil drop treated eyes, and neither this nor any other parameter differed between bilaterally untreated naïve eyes and ripasudil drop treated eyes. In control eyes, there was a significantly greater astrocyte area fraction in the rim region (0.58) than centrally (P = 0.01). 

The astrocyte area fraction, or thick area, declined to 0.46 and 0.42 in the ripasudil GL and BSS GL groups, respectively, 3D after bead injection (differences of GL groups from bilaterally untreated naïve group, P = 0.046 and P = 0.048, respectively). Both the GL groups returned to values not significantly different from controls at 6 weeks. The mean width, or bwidth, of astrocyte processes (beams) was similar in the control and ripasudil drop treated groups (0.94 µm), and was unchanged after 3D GL in both BSS and ripasudil GL groups. Interestingly, 6 weeks after bead injection, the Rip GL group had a higher beam width (mean = 1.01 µm) than either 6-week BSS GL or control (both P = 0.004). The mean number of pores (consisting of axons and non-GFAP labeled tissues) was 10,349/ON in bilaterally untreated controls and significantly decreased in ripasudil 3D GL (7943, P = 0.023 versus controls), but was not significantly different from controls in BSS 3D GL. 

The most substantial changes in astrocyte histology were in the number and aspect ratio of nuclei, a ratio of width to height which describes cell shape. Lower aspect ratio indicates a rounder rather than an oval nuclei configuration in cross-section. Ripasudil drop treatment alone did not alter nucleus area, number or aspect ratio compared to bilaterally untreated controls. The mean number of astrocyte nuclei/ON section increased from 48 in controls to 128 in Rip 3D GL and to 153 in BSS 3D GL (differences from control, P = 0.003 and P = 0.0001, respectively; see Supplementary Fig. S6, Supplementary Table S6). Similarly, the mean density of nuclei more than doubled in both the 3D GL groups over control (both P < 0.0001). At 6 weeks after bead injection, the Rip GL group had no difference from controls in nuclear count (P = 0.56), but the BSS GL group retained an elevated number of nuclei (P = 0.015 versus control). The mean aspect ratio of nuclei was significantly lower than control in BSS GL at both 3 days and 6 weeks (P = 0.0006 and P = 0.02, respectively), but not significantly different from controls in ripasudil GL eyes (P > 0.05). 

Protein Analysis of Molecules in the ROCK Pathway

Molecules participating in the ROCK pathway were studied by immunofluorescent labeling and selected molecules were quantified by Wes analysis. Wes data were calculated for 4 groups of eyes: bilaterally untreated, ripasudil drop alone, and both 3D GL groups (ripasudil and BSS drop-treated). The quantification was performed in three micro-dissected regions: the retina, UON, and MON. Values were normalized to the GAPDH levels in each region and treatment group. Data are presented as the ratio of treated groups to each other and to bilaterally untreated naïve eyes (n = 6 samples per group). 

Astrocyte cell bodies and processes exhibited prominent labeling for GFAP, actin, and RhoA, especially in the prelaminar and UON astrocytes. The 3D GL samples showed a brighter visualization of astrocyte somas and processes labeling for GFAP, actin, and RhoA in all ON regions, with the brightest label at the central ON. RhoA showed fainter labeling in the RGC layer of the retina than the ON (Fig. 5). In Wes analysis, the RhoA level was reduced by nearly 50% with ripasudil drop treatment alone in the UON (ratio to naïve = 0.54 ± 0.08, P = 0.03), but not in the retina (0.73 ± 0.17, P = 0.21) or MON (1.06 ± 0.24, P = 0.56; Supplementary Table S7, Fig. 6). Neither ripasudil the GL nor the BSS GL groups had a quantitative increase in RhoA in the UON. By contrast, there was a significant increase in RhoA levels in the MON in both ripasudil and BSS 3D GL groups (ratios 1.84 and 1.87, P = 0.0002 and P = 0.0001, respectively). 

Figure 5.

 

Labeling of ROCK pathway molecules in controls, BSS GL, and ripasudil GL eyes. Protein immunofluorescence for various treatment groups and control samples; bilaterally untreated naïve (A–E) and GL samples: BSS + 3D GL (F–J), Rip + 3D GL (K–O). Scale bar = 100 um.

Anti-ROCK1 antibodies labeled astrocytic bodies throughout the ON (see Supplementary Fig. S7), and the retina as well as Müller cells in the retina, but not RGC bodies or retinal nerve fiber layer (see Fig. 5). In both the ripasudil and BSS 3D GL eyes, there was no visible change in ROCK1 immunolabeling pattern or intensity in semiquantitative grading by a masked observer on a scale from 0 (dimmest) to 3 (brightest). The median label intensity was 1.5 for both GL groups and 2.0 for control eyes. Phosphorylated ROCK1 (p-ROCK1) labeling of astrocytes in UON and MON was greater in both groups of 3D GL samples than in controls. However, there was a lower brightness distribution of p-ROCK1 label intensity in the ripasudil GL group than BSS GL (Fig. 7; median 1 [ripasudil] versus 2 [BSS], P values versus control = 0.04 for BSS GL, and P = 0.06 for ripasudil GL). In Wes analysis, the ROCK1 protein levels were unchanged from untreated controls by ripasudil drop treatment alone in all three regions (all P > 0.12). Likewise, ROCK1 protein levels in the two 3D GL groups were not significantly greater than untreated control values in both UON and MON (see Supplementary Table S7, Fig. 6). Unfortunately, there are no effective anti-p-ROCK1 antibodies for Wes analysis. 

Figure 6.

 

Protein quantification by Wes analysis. Protein analysis for various treatment and control groups; bilaterally untreated naïve, ripasudil (cross hatched bars) only, and 3-day pressure elevated samples treated with either ripasudil (orange) or BSS (grey) drops. Comparison of the ratios of ROCK species (A–C): proteins RhoA, ROCK1, ROCK2, and pROCK2, as well as downstream proteins (D–F): cofilin, p-cofilin, and p38 MAPK, normalized to GAPDH level in the retina (A, D), UON (B, E) and MON (C, F), and presented as a ratio to the bilaterally untreated samples. All samples were compared to the bilaterally naïve samples with a ratio of 1. * P ≤ 0.05, † P ≤ 0.01, ‡ P ≤ 0.001, § P ≤ 0.0001, t-test.

Figure 7.

 

Semiquantitative grading of p-ROCK1 images from UON in ripasudil GL and BSS GL. Semiquantitative grading of the intensity of the label for pROCK1.

ROCK2 antibodies labeled RGC bodies and the plexiform layer (inner and outer) in the retina and axon bundles but not glial columns in the ON (see Supplementary Fig. S7). ROCK2 was more prominently labeled in RGCs than ON astrocytes (see Fig. 5). By Wes analysis, ROCK2 levels with ripasudil alone treatment were unchanged in UON, but significantly increased in MON (1.52 ± 0.37, P = 0.05). ROCK2 level in UON was significantly above control in ripasudil-treated 3D GL eyes, but not in BSS 3D GL eyes. ROCK2 levels in the MON were three times higher than naïve in both GL groups. The p-ROCK2 ratio to controls was significantly higher in UON of the ripasudil GL eyes (P = 0.006), but not in BSS UON GL versus naïve (see Fig. 6). With calculation of the ratio of p-ROCK2 to ROCK2 (Table 3), the BSS 3D GL ratio was substantially higher in the retina than both the ripasudil 3D GL and control ratios (3.62 BSS GL versus 1.38 in controls and 1.22 in ripasudil GL, P = 0.01). The p-ROCK2/ROCK2 ratio in ripasudil 3D GL eyes was not different from control in retina, UON, or MON. 

Table 3.

 

Ratios of P-Cofilin to Cofilin and P-ROCK2 to ROCK2 by Wes Analysis

 

Cofilin, an actin-stabilization protein, was labeled in both axons and astrocytes, but p-cofilin localization was predominately in astrocytes in the UON in control eyes (Supplementary Fig. S8). In Wes analysis, ripasudil drop treatment alone dramatically reduced cofilin levels in UON (ratio to control = 0.32 ± 0.06, P = 0.02), but not in the retina or MON. Ripasudil drops alone did not change p-cofilin level in retina, UON, or MON. The ripasudil 3D GL group level of cofilin was significantly greater than control (2.14 ± 0.81, P = 0.03), whereas the BSS GL UON and both MON group data were not significantly altered. In the UON, p-cofilin level was not changed in either 3D GL group, whereas in the MON, both ripasudil and BSS treated 3D GL groups had increases of p-cofilin (P = 0.05 and P = 0.02, respectively; see Fig. 6). The p-cofilin/cofilin ratio (see Table 3) was not changed from controls in either 3D GL group in the retina or UON. In MON, the ratio was dramatically higher than control in both 3D GL groups (P = 0.02 and P = 0.01). 

Antibodies for p38 MAPK produced labeling of astrocytes more prominently than RGC axon bundles, but was identified in retinal RGC layer and sclera. In 3D GL eyes, p38 MAPK labeling of astrocytes in the UON increased (see Supplementary Fig. S8). Immunofluorescent labeling for phosphorylated-p38 MAPK was greater in GL specimens than controls in both astrocytes and axons (see Supplementary Fig. S8). By Wes analysis, the ripasudil drops alone group had a significant reduction in p38 MAPK in the UON (P = 0.003). Protein levels of p38 MAPK levels were significantly increased compared to controls in UON ripasudil GL (P = 0.04), but not BSS GL (P = 0.25), and also were significantly increased in both MON GL groups (ripasudil GL, P = 0.009 and BSS GL, P = 0.01; see Fig. 6). Currently, there is no available Wes-appropriate antibody for p-p38 MAPK. 

Discussion

We confirm that topical treatment with a ROCK inhibitor, ripasudil, improves RGC survival in rodent models of increased IOP and ON crush,^4–6 although the effect was less significant in the crush model. Whereas there were transient IOP lowering effects of ripasudil drops, the effect did not alter the IOP increases in the bead injection model. The ON crush model has been used as an efficient method to test for protective effects from mechanical RGC axon injury. We and others have demonstrated that when both increased IOP and crush model damage are mitigated by an intervention, as in treatment with intravitreal tozacertib,³² the pathogenetic mechanism may depend less on the mechanical events at the ONH and more on the RGC response. In a different example, oral treatment with losartan is protective in experimental mouse GL, but not in ON crush.³³ This can be interpreted as a beneficial effect of losartan which solely depends on the mechanical steps from increased IOP to axon injury and are not protective of forceps crush of the ON axons. The present protective effect of ripasudil in both GL (RGC and astrocyte dependent) and crush (RGC dependent) models would suggest that it acted predominately to improve RGC survival by effects on the RGCs and supporting glial cells. Another ROCK inhibitor, fasudil, reduced death in hippocampal neurons via the c-Jun N-terminal kinase (JNK) signal pathway.³⁴ However, it will be important to review all the evidence presented here because some effects occurred at the ONH and in astrocytes, suggesting that multiple mechanisms should be considered. 

First, we have considerable evidence that the topical application of ripasudil affected events in the ROCK pathway in the posterior globe. Our data found that treatment with ripasudil drops alone reduced ROCK activity in our assays in both the UON and the MON. In addition, topical ripasudil reduced the protein levels in the UON of 3 molecules in the ROCK pathway, RhoA itself, p38 MAPK and cofilin. This demonstrates that ripasudil, as delivered here, reached the mouse retina, UON, and MON. The effects were both at the site of injury in the GL model, the UON, and in the retina. In previous experimentation, we showed that 12-hour GL increased ROCK activity in mouse sclera.¹⁴ 

Second, several aspects of our data suggest that the neuroprotection by topical ripasudil was due to its effects on RGCs and their axons. Both our RBPMS counts and our nerve counts show that topical ripasudil treatment was neuroprotective, causing fewer RCGs and axons to die in our injury models. ROCK activity in our assays was reduced in the ripasudil-treated 3D GL group, whereas the BSS GL group had increased activity 1 day after GL in both the retina and UON. This assay measures both ROCK1 and ROCK2 activity. The ratio of pROCK2 to ROCK2 is an important measure of its activation,³⁵ and this ratio rose in the retina of BSS GL, but was unchanged in all regions in ripasudil 3D GL. This suggests that the reduction of ROCK2 activation plays an important role in the beneficial effect. Because our histological studies show that ROCK2 is more prominent in RGCs and their axons, this supports the hypothesis that ROCK inhibition acts on RGCs in ripasudil neuroprotection. 

Third, additional findings in this study do not support the hypothesis that the protective effect of ripasudil in experimental GL is on mechanical elements of the posterior eye. Axonal transport block, assessed by APP quantitative movement, was similar in ripasudil- and BSS-treated eyes with IOP increase. Transport block has been definitively shown to result from the effects of IOP increase.³⁶ In addition, the increase in mechanical strains of explanted mouse eyes after 3D GL was not reduced by ripasudil treatment compared to BSS-treated 3D GL eyes. Interestingly, ripasudil treatment alone led to greater strain estimates. One interpretation of these data is that pretreatment with ripasudil in eyes that then undergo GL in our protocol reduces the short-term mechanical effect of the GL in the ripasudil GL eyes, leading to less injury to axons. 

A group of other results differed between ripasudil-treated and BSS-treated GL eyes, suggesting that ROCK inhibition beneficially affected astrocyte response in the GL model. Immunofluorescent labeling of ROCK1 was predominately in astrocytes in the ON (see Supplementary Fig. S7). The p-ROCK1 increased in BSS 3D GL, but its increase was significantly less in ripasudil 3D GL immunofluorescent labeling. The reduction in ROCK1 activation by ripasudil in astrocytes might contribute to its protective effect. Further evidence in this regard was also seen. Structural analyses after 3D GL found that astrocyte beams were wider in the ripasudil than the BSS group, whereas the area occupied by pores between astrocyte processes were smaller in ripasudil GL eyes. In addition, the number of cell nuclei in the UON increased with 3D GL in both GL groups and the nuclear number returned to control levels only in the BSS group by 6 weeks. We have previously shown that the dominant cell type in the UON of mouse is the astrocyte, and while microglia are present, the majority of dividing cells by Ki67 labeling are astrocytes.²⁵ Taken together, these results suggest that the mechanism of the ripasudil protective effect is multifactorial. 

We confirmed and extended several aspects of the mouse GL model. Biomechanical strains have been shown to increase significantly 3 days after GL in explanted mouse eyes.³⁷ Our BSS-treated GL eyes here had similar strain increases to previously published data. Our past investigations show that the strains return to control levels by 6 weeks after bead injection. Second, we found that there were significant changes in the MON that were not evident in the retina and UON. These included increased protein levels of p38 MAPK and a greater ratio of p-cofilin/cofilin. It is documented that ROCK phosphorylation of LIMK leads to phosphorylation of cofilin,³⁸ suppressing some of its normal functions and, in general, stabilizing the cytoskeleton by the production of stress fibers.³⁹ However, these general interactions can be cell-type specific and further research is needed to corroborate this conclusion. We found increased p-cofilin in the MON, particularly in BSS GL. We also detected a smaller ratio of p-cofilin to cofilin in ripasudil GL samples than BSS GL in both UON and MON. Whereas cofilin labels both axons and astrocytes, p-cofilin was identified in ON astrocytes only in GL samples. 

Local activation of RhoA leads to an increase in actin polymerization and myosin activity in the region of activation and exogenous RhoA activation increases cell contractility.⁴⁰ ROCK activation by RhoA also induces myosin II activation by direct phosphorylation of myosin regulatory light chain (MLC),⁴¹ stabilizing the cytoskeleton by promoting F-actin polymerization.⁴² ROCK1 increases formation of thick actin stress fibers, whereas ROCK2 regulates the localization of intracellular myosin complexes. Both ROCK isoforms are involved in myofibroblast mechanosensing of extracellular matrix stiffness.⁴³ The effects of ROCK inhibitors are both cell-type and experimental method dependent.^44,45 A potential effect of topical ROCK inhibitors on ONH astrocytes is their modification of membrane-linked mechanical signaling in astrocytes. Astrocytes in the mouse GL model undergo alterations in membrane linked, junctional complexes.⁴⁶ Extracellular molecules acting at the astrocytic membrane include transforming growth factor β (TGFβ).⁴⁷ Stimulation of TGFβ receptors leads to phosphorylation of RhoA by its kinase, inducing myosin II activation by phosphorylating myosin regulatory light chain, participating in F-actin polymerization. TGFβ-signaling increases in glaucomatous ONH,^48–50 human cultured astrocytes exhibit altered morphology after TGFβ treatment⁵¹ and experimental mouse GL leads to increased TGF-β signaling in the UON.⁵² Our proteomic analysis of rat glaucoma sclera found that molecules in the RhoA pathway were altered.⁴⁸ Others have conducted investigations of the effects of ROCK inhibitors in experimental GL. Tehrani and co-workers used acute, 8 hour IOP elevation in a rat model to test effects of systemic ROCK inhibitor, fasudil, on UON astrocyte structure.⁵³ Fasudil permitted maintenance of ON actin bundles compared to control IOP elevation, while reducing short-term axonal damage. Nishijima and colleagues⁵ reported that topical ripasudil suppressed ON crush-induced phosphorylation of p38 MAPK and cofilin. Levels of ROCK2 transiently increased in retinal immunofluorescence after rat ON transection; however, contrary to our studies in mice, the investigators found no ROCK2 labeling in control rat retina.⁵⁴ Recent research combining IOP elevation and ON transection in rats suggests complex interactions between astrocyte responses and the presence/absence of axons.⁵⁵ Finally, combined delivery of topical ripasudil and brimonidine was shown to enhance RGC protection after ON crush by modulating a number of retinal signaling pathways.⁵⁶ 

There are several limitations in our research. Many of our experiments were conducted from 1 to 3 days after IOP increase. This period is a critical one for damage to RGCs in mouse GL models. It is likely that later events would provide additional or even different outcomes from those at this time. We did not test ROCK inhibitors other than ripasudil and do not know if they share the features documented here. The 2% concentration of ripasudil (2 mg/100 µL) was chosen as the maximum practical dose, as it is the solubility limit for the molecule. However, given the small size of mice and the relative thinness of their cornea and sclera, topical delivery may provide a much greater effect than would be possible in a larger animal or humans with topical delivery. Furthermore, there are known effects in fellow eyes of unilateral rodent treatment models, involving systemic absorption of topical agents, and trans-neural effects.⁵⁷ Thus, in many of the experiments here, we used bilaterally untreated naïve mouse eyes as controls. We did not detect typical side effects of ROCK inhibitor drops that are seen in human, such as subconjunctival hemorrhage. The duration of our experiments may be too short to produce such effects. We have used CD1 mice for testing of outcomes of microbead GL, as this strain produces more robust RGC loss than C57BL/6J mice.²⁴ Testing of some protein levels is restricted by nonavailability of antibodies that are compatible with the proprietary Wes technology. 

Conclusions

In summary, we have confirmed that delivery of a topical ROCK inhibitor in two mouse models of ON injury is protective for RGC loss. The mechanisms of benefit appear to involve effects on both RGC and astrocytes. ROCK inhibitors may be a future neuroprotective treatment in addition to their IOP-lowering effects, and could be tested in human clinical trials. 

Acknowledgments

Supported in part by PHS research grants EY 02120 (Dr. Quigley) and EY 01765 (Wilmer Institute Core), all from the National Eye Institute, National Institutes of Health; Research to Prevent Blindness, Inc.; the A. Edward Maumenee Professorship; the Allan and Shelley Holt Rising Professorship; and unrestricted support from Mary Bartkus and William T. Forrester. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. 

Disclosure: S.E. Quillen, None; E.C. Kimball, None; K.A. Ritter-Gordy, None; L. Du, None; Z. Yuan, None; M.E. Pease, None; S. Madhoun, None; T.D. Nguyen, None; T.V. Johnson, None; H.A. Quigley, None; I.F. Pitha, None