24(S)-Hydroxycholesterol protects the ex vivo rat retina from injury by elevated hydrostatic pressure

In the central nervous system, 24(S)-hydroxycholesterol (24(S)-HC) is an oxysterol synthesized from cholesterol by cholesterol 24-hydroxylase (CYP46A1) encoded by the cyp46a1 gene. In the present study using a rat ex vivo glaucoma model, we found that retinal 24(S)-HC synthesis is facilitated by pressure elevation. Moreover, we found that 24(S)-HC is neuroprotective against pressure mediated retinal degeneration. Quantitative real-time RT-PCR, ELISA, and immunohistochemistry revealed that elevated pressure facilitated the expression of cyp46a1 and CYP46A1. Immunohistochemically, the enhanced expression of CYP46A1 was mainly observed in retinal ganglion cells (RGC). LC-MS/MS revealed that 24(S)-HC levels increased in a pressure-dependent manner. Axonal injury and apoptotic RGC death induced by 75 mmHg high pressure was ameliorated by exogenously administered 1 μM 24(S)-HC. In contrast, voriconazole, a CYP46A1 inhibitor, was severely toxic even at normobaric pressure. Under normobaric conditions, 30 μM 24(S)-HC was required to prevent the voriconazole-mediated retinal damage. Taken together, our findings indicate that 24(S)-HC is facilitated by elevated pressure and plays a neuroprotective role under glaucomatous conditions, while voriconazole, an antifungal drug, is retinotoxic. 24(S)-HC and related compounds may serve as potential therapeutic targets for protecting glaucomatous eyes from pressure-induced injuries.


Effects of pressure loading on endogenous levels of cholesterol and 24(S)-HC. Cholesterol
hydroxylation is important for the maintenance of cholesterol homeostasis in the retina, where the conversion to 24(S)-HC catalyzed by CYP46A1 represents the major mechanism of cholesterol elimination ( Fig. 2A). We thus measured levels of cholesterol (per wet retinal weight), the substrate of CYP46A1, in rat ex vivo eyecups, and found a pressure-dependent decrease in cholesterol at 35 mmHg (p < 0.05) and 75 mmHg (p < 0.001) compared to 10 mmHg ( Fig. 2B and Table S2B). Consistently, liquid chromatography-tandem mass spectrometry (LC-MS/ MS) analysis revealed that 24(S)-HC (ng/g retinal protein), the product of enzymatic activities of CYP46A1, increased in hyperbaric conditions (p < 0.05 at 35 mmHg and p < 0.05 at 75 mmHg, compared to 10 mmHg) in units of ng/g retinal protein ( Fig. 2C and Table S2C) and ng per wet retinal weight ( Fig. 2D and Table S2D-1). Administration of 1 μ M voriconazole, an inhibitor of CYP46A1 32 showed no significant influence on 24(S)-HC concentrations, while 10 μ M voriconazole significantly diminished 24(S)-HC levels at each pressure ( Fig. 2D and Table S2D-2,S2D-3).
Because we previously found that high pressure increases local levels of the neurosteroid, allopregnanolone (AlloP), and AlloP is protective against high pressure 33 , we also examined the effects of pressure on AlloP levels with or without 10 μ M voriconazole. Three eyes were examined by LC-MS/MS at each pressure, and AlloP levels are expressed as ng per wet retinal weight (g). In the presence of voriconazole, LC-MS/MS analysis revealed pressure-dependent increases of AlloP at 35 mmHg (p < 0.01) and 75 mmHg (p < 0.001) compared to 10 mmHg, indicating that the decrement in 24(S)-HC levels with voriconazole has some specificity and does not involve another major endogenous cholesterol-derived product ( Fig. 2E and Table S2F). In the absence of voriconazole, AlloP levels at 35 mm Hg did not rise as high as they do in the presence of voriconazole ( Fig. 2F and Table S2F). In contrast, the LC-MS/MS analysis revealed a significant increase of AlloP at 75 mm Hg compared to 10 or 35 mm Hg. These results suggest that AlloP production may be saturated by the severe stress of 75 mmHg (so this level does not change with voriconazole), but that the milder stress of 35 mmHg only elevates AlloP levels when cholesterol metabolism to 24(S)-HC is blocked.

24(S)-HC preserves retinal histology in the presence of high pressure.
Consistent with our previous reports 13,14 , retinas incubated at 10 mmHg (Fig. 3A) or 35 mmHg (Fig. 3B) exhibited no remarkable changes in morphology. However, retinas incubated at 75 mmHg showed axonal swelling in the nerve fiber layer (NFL), though the other retinal layers remained intact except for the INL and IPL where small vacuoles were present ( Fig. 3C and Fig. S3-1). In the presence of 1 μ M 24(S)-HC administered for the 24 hour incubation period, retinas exhibited no remarkable changes at any pressure ( Fig. 3D-F). Importantly, exogenous 24(S)-HC prevented the axonal swelling typically observed at 75 mmHg.
A quantitative assessment of structural changes induced by pressure elevation in the absence and presence of 24(S)-HC is summarized in Table 1 (also see Table 1 Source data). The nerve fiber layer thickness (NFLT), neuronal damage score (NDS), and density of damaged cells in the GCL in retinas incubated at 75 mmHg were significantly increased compared to those in control retinas incubated at 10 mm Hg (p < 0.0001). By contrast, administration of 1 μ M 24(S)-HC resulted in no significant changes in the NFLT, NDS, or density of damaged cells in the GCL in retinas incubated at 75 mmHg compared to normobaric controls.
A cholesterol 24-hydroxylase inhibitor is severely retinotoxic at low pressure. In contrast to 24(S)-HC, we found that voriconazole was highly neurotoxic. Although administration of 1 μ M voriconazole showed no remarkable changes ( Fig. S3-3  Interestingly, the retinal damage induced by voriconazole was most prominent at 10 mmHg and least prominent at the highest pressure (Fig. 3G). A quantitative assessment of structural changes induced by pressure elevation in the presence of voriconazole is summarized in Table 1 (also see Table 1 Source data). Administration of 10 μ M voriconazole produced significant increases in the NDS and density of damaged cells in the GCL compared to the retinas incubated without voriconazole at each pressure.  A quantitative assessment of structural changes induced by voriconazole and 24(S)-HC at 10 mmHg is summarized in Table 1 (also see Table 1 Source data). Administration of 1 μ M 24(S)-HC resulted in no neuroprotection, with increased NDS and density of damaged cells in the GCL in the presence of 10 μ M voriconazole at 10 mmHg compared to those in the control retinas incubated without voriconazole at 10 mmHg. At 30 μ M, however, co-administration of 24(S)-HC resulted in no significant changes in the NDS and density of damaged cells in the GCL in retinas incubated with 10 μ M voriconazole at 10 mmHg compared to control retinas incubated at 10 mmHg in the absence of either agent. An intermediate degree of protection was observed with 10 μ M 24(S)-HC. The NFLT did not show significant changes under any of these conditions.
Because voriconazole-induced damage has histological features similar to excitotoxicity, we also examined the effects of glutamate receptor antagonists. In retinas incubated with 10 μ M voriconazole at 10 mmHg, administration of either 1 μ M dizocilpine (MK801), a NMDA type glutamate receptor antagonist ( A quantitative assessment of structural changes induced by voriconazole and glutamate receptor antagonists at 10 mmHg is summarized in Table 2 (also see Table 2 Source data). In the presence of either 1 μ M MK801 alone or 1 μ M GYKI alone, retinas incubated with 10 μ M voriconazole at 10 mmHg showed significant increases in the NDS (p < 0.0001) and density of damaged cells in the GCL (p < 0.0001) compared to control retinas incubated without voriconazole at 10 mmHg. In contrast, the combination of 1 μ M MK801 and 1 μ M GYKI with voriconazole showed no remarkable changes in the NDS and density of damaged cells in the GCL at 10 mmHg. In whole mounted retinas, RGC damage induced by pressure elevation was visualized as reduced numbers of cells that were positive for NeuN ( Fig. 5A-C). Figure 5A,A′ illustrate examples of confocal images of NeuN-labeled RGCs that were obtained from a control eye incubated at 10 mmHg. Pressure elevation (75 mmHg) reduced the number of cells that were positive for NeuN (Fig. 5B,B′). The confocal images in Fig. 5C,C′ illustrate the neuroprotective effect of 24(S)-HC (1 μ M) on RGC survival in hyperbaric conditions. As shown in Fig. 5D,D′, RGC numbers are significantly different from control images consistent with the damage induced by 10 μ M voriconazole at 10 mmHg. However, less disruption of the RGCs occurred when 30 μ M 24(S)-HC was applied with 10 μ M voriconazole (Fig. 5E,E′). Figure 5F,G show summaries of RGC survival (also see Table S5F,G).

Pressure-induced apoptosis and neuroprotection with 24(S)-HC.
At 10 mmHg, a small number of TUNEL-positive cells are observed only in the ONL (Fig. 6A). Exposure to elevated pressure induced apoptosis that was apparent in the GCL and to a lesser extent in the INL (Fig. 6B). The number of TUNEL-positive cells was reduced when 1 μ M 24(S)-HC was administered (Fig. 6C). The graph in Fig. 6D shows the number of apoptotic cells in the retina in each condition (Table S6D).    Discussion 24(S)-HC is synthesized from cholesterol by CYP46A1, a neuronal specific enzyme localized to the endoplasmic reticulum 22 . In the rat retina, the enzyme is expressed in the GCL and INL, where the cell bodies of neurons reside, but not in the IPL or OPL that contain axons and synapses 23 . Similar results were shown in mice: the enzyme is expressed in the GCL and partially in the INL 22 . Consistent with these prior studies, we found that the expression of CYP46A1 is exclusively limited to the GCL and INL under normobaric conditions (Fig. 2C). However, in the hyperbaric condition, enzyme expression is not only facilitated, but is also more widely distributed over several retinal layers including the IPL and OPL (Fig. 2D,E), suggesting that the distribution, as well as the strength of the expression, is susceptible to modulation by stressors such as elevated pressure. In the human retina, the distribution of 24-hydroxylase is broader than in rodents 24 . During pressure elevation in the retina or an AACA in vivo, 24(S)-HC levels have been reported to increase. In rats, elevation of intraocular pressure in vivo stimulates CYP46A1 within 3 days followed by sustained increases in 24(S)-HC levels 26 . Consistent with this, we observed that expression of the cyp46a1 gene and CYP46A1 protein ( Fig. 2A,B), as well as 24(S)-HC production (Fig. 3A) are enhanced by hyperbaric conditions. Furthermore, we found that enhanced immunofluorescence staining against CYP46A1 is observed over several retinal layers (Fig. 3B,C). In our ex vivo glaucoma model, retinas were exposed to hyperbaric conditions for only 24 hours and showed significant elevations in 24(S)-HC levels, suggesting that the induction of the enzyme during AACA can be relatively rapid. As a consequence of up-regulation of CYP46A1, the present study demonstrated a lower concentration of cholesterol and higher concentration of 24(S)-HC in the pressure-loaded retina compared to the retina incubated in the normobaric condition. Based on changes in these metabolite and substrate concentrations, it appears that retinal cholesterol turnover is increased by upregulation of CYP46A1 under hyperbaric conditions.
Is the induction of CYP46A1 during AACA physiological (neuroprotective) or pathological (neurotoxic)? Based on prior studies, it appears that the product of CYP46A1, 24(S)-HC, can have varying effects depending on experimental conditions [34][35][36] . In the present study, the finding that exogenous administration of 1 μ M 24(S)-HC prevents pressure-induced axonal injury and apoptotic RGC death indicates that 24(S)-HC has neuroprotective actions rather than pathological effects. This finding also suggests that facilitated expression of cholesterol 24-hydroxylase triggered by AACA is not a pathological sequela but perhaps a physiological and/or homeostatic mechanism. At this point, however, we do not know whether the preservation of retinal morphology translates into improvement in retinal physiology.
Our results are consistent with a model proposed by Sodero and colleagues 37,38 in which activation of glutamate receptors during stress and aging in the hippocampus promotes translocation of CYP46A1 to the plasma membrane, resulting in cholesterol loss and subsequent stimulation of neuronal survival pathways. Enhanced CYP46A1 expression has also been found to be neuroprotective in in vivo models of neurodegenerative illnesses 39 . Together, these findings suggest that decreases in cholesterol via CYP46A1, coupled with increases in levels of 24(S)-HC and perhaps other cholesterol-derived modulators, help to promote neuronal survival under stressful conditions.
To determine whether 24(S)-HC is important for preserving retinal integrity, we also examined the effects of voriconazole. This drug is used clinically as an antifungal agent and inhibits CYP46A1 with relatively high potency 40 . Although the safety of intravitreous injection of voriconazole on retinal function has been described in rabbits 41 , voriconazole transiently impairs bipolar cell function in monkeys 42 and is associated with retinal dysfunction in rats following repeated systemic administration 43 . Recently, the effects of repeated doses of voriconazole on the vision of healthy human subjects was investigated in a double-blind, placebo-controlled study 44 . This latter study showed that voriconazole reduced scotopic maximal a-and b-wave amplitudes, oscillatory potential amplitude and the 30-Hz photopic flicker response amplitude compared with placebo, while also impairing color vision discrimination. In the present study (Fig. 3G), to our surprise, voriconazole proved to be severely retinotoxic with edematous changes in the IPL and bull's eye formation in the INL, well-known characteristics of excitotoxic retinal damage 45,46 Indeed, this severe retinal damage was prevented by a combination of antagonists blocking non-NMDA and NMDA ionotropic receptors (Fig. 4F), supporting the idea that the damage involves activation of both types of glutamate receptor. We previously reported that ischemic degeneration of isolated rat retinas is mediated by activation of both types of ionotropic glutamate receptor 47     induced by voriconazole has histological features that are similar to the damage induced by retinal ischemia. Our findings indicate that voriconazole has significant retinotoxic potential and suggest the need for caution when using the drug under conditions that impair the integrity of the blood-retinal barrier. Amelioration of voriconazole-induced damage by 24(S)-HC supports the hypothesis that voriconazole induces retinal degeneration via inhibition of CYP46A1. However, contrary to our expectation, full protection against voriconazole required 30 μ M 24(S)-HC, a concentration higher than needed to block the effects of high pressure. Importantly, 24(S)-HC is the major brain metabolite of cholesterol and is present at endogenous levels in the tens of micromolar concentration range. Thus, even 30 μ M 24(S)-HC is considered to be within the physiological range in human brain homogenates 48 . It may seem odd that 10 μ M 24(S)-HC offers only partial protection against voriconazole (Fig. 4A and Table 2), taking into account that 1 μ M 24(S)-HC fully prevents axonal swelling at 75 mmHg (Fig. 3F). This finding could reflect the possibility that a reservoir of the oxysterol is depleted by voriconazole because even at 10 mmHg, 24(S)-HC levels are reduced below basal levels by voriconazole (Fig. 2C). Substantial amounts of exogenous 24(S)-HC could thus be needed to replenish basal levels in the presence of voriconazole and diminished endogenous oxysterol production.
The ability of 24(S)-HC to protect retinas against the effects of high pressure and voriconazole is also paradoxical in light of recent studies showing that this oxysterol is a positive allosteric modulator of NMDA receptors in hippocampus, and thus might be expected to worsen excitotoxicity 49,50 . Nonetheless, as noted above, prior studies have found that 24(S)-HC can play complex roles under pathological conditions 51 , and another endogenous oxysterol, 25-hydroxycholesterol, is a silent modulator of 24(S)-HC, dampening 24(S)-HC effects on NMDA receptors while having little intrinsic effect on its own 52 . Additionally, cholestane-3β ,5α ,6β -triol, also an endogenous oxysterol, is a negative allosteric modulator of NMDA receptors, further complicating interpretation of the role of oxysterols under various physiological and pathological conditions 53 . These observations make it important for future studies to determine the various oxysterols that are generated in the retina under different conditions and how these oxysterols interact.
Interestingly, the retinal damage induced by voriconazole was more prominent at 10 mmHg than at 35 or 75 mmHg. This observation cannot be explained by facilitated synthesis of 24(S)-HC at 75 mmHg, because voriconazole inhibits the production of 24(S)-HC at all pressures studied. However, this phenomenon could be explained by the robust rise in AlloP, a GABA-enhancing neurosteroid, at elevated pressures (Fig. 2E,F). We previously reported that exogenously administered AlloP attenuates the development of axonal swelling at high pressures, while blockade of AlloP synthesis at 75 mm Hg results in excitotoxic retinal damage with features akin to voriconazole at low pressure 30 . Based on these findings, it is likely that both 24(S)-HC and AlloP are cholesterol-derived modulators that help to protect the retina endogenously and can potentially be exploited for therapeutic purposes. However, 24(S)-HC appears to be more important for maintaining retinal integrity under normobaric conditions, because at a normobaric pressure blockade of AlloP production does not induce retinal damage 30 , while inhibiting 24(S)-HC synthesis results in severe excitotoxicity (Fig. 3G).
Taken together, our findings indicate that enhanced synthesis of 24(S)-HC has important roles in maintaining retinal integrity under hyperbaric conditions, helping to protect the retina from pressure-induced damage. Thus, 24(S)-HC and related compounds may serve as potential therapeutic targets to protect glaucomatous eyes from pressure-induced injuries.

Materials and Methods
Protocols for animal use were approved by the Akita Graduate University Animal Studies Committee in accordance with the guidelines of the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research.

Rat ex vivo Eyecup Preparation.
Rat ex vivo eyecups were prepared from 28-32 day-old male Sprague-Dawley rats (Charles River Laboratories International Inc., Wilmington, MA) as previously described 13,14 . The cornea was excised circumferentially with microscissors and the lens and vitreous were removed. The empty eyecup was placed on a flat cutting surface and immersed in ice-cold aCSF. The retina was not detached from the sclera. During experiments, several eyecups (3)(4)(5)(6)(7)(8) were placed at the bottom of a 100 ml glass beaker filled with aCSF containing (in mM): 124 NaCl, 5 KCl, 2 MgSO 4 , 2 CaCl 2 , 1.25 NaH 2 PO 4 , 22 NaHCO 3 , and 10 glucose, and incubated at 30 °C for 24 hours using a closed pressure-loading system (Fig. 1A). In this model, the glass beaker was carefully placed at the bottom of an acrylic pressure chamber (2,000 ml) with pH maintained at 7.35 to 7.40. A 95% O 2 -5% CO 2 gas mixture was delivered via plastic tubing and an air filter (Cat#SLGP033RS, Merck Millipore, Billerica, MA), with the tubing terminating 1 cm above the bottom of the beaker. Gas flow was regulated with an infusion valve and a control dial on the lid of the pressure chamber. The 95% O 2 -5% CO 2 gas mixture was infused until the pressure reading given by a manometer reached the desired level. The pressure was then locked in place by adjusting the control dial of an effusion valve, and monitored continuously for 24 h. After maintaining the chamber at the set pressure ( Quantitative real-time RT-PCR. We quantified cyp46a1 mRNA expression in pressure-loaded eyecup specimens incubated at 10, 35, and 75 mmHg for 24 h based on previously reported methods 14 . At the end of each experiment, the retina of the empty eyecup was detached from sclera and immersed in RNAlater solution (Qiagen, Hilden, Germany). In the present study, six independent experiments were performed for each condition. All PCR reactions were repeated in duplicate, and the average values were used for statistical analysis. The RNA expression levels were normalized to S16 ribosomal protein mRNA (rps16) expression (see Supporting data of Fig. 1B. Validation of internal control).
The list of primers used in the present study is summarized as followings; Gene cyp46A1. GenBank accession number NM_001108723. Atlanta, GA). According to the manufacturer's instructions, the absorbance was detected at 450 nm and a standard curve was delineated based on the absorbance of standards. The protein concentration of retinal samples was determined by the Bradford method (Bio-Rad Laboratories, Hercules, CA) using the assay solution and serum-globulin as the standard. For quantification of immunohistochemical data, images of each section (5 sections per animal) were captured. Digital images were analyzed, and the average intensity of immunofluorescence was measured using Image-Pro Plus software (Media Cybernetics, Rockville, MD).

Cholesterol measurements.
After pressure loading, the retina was detached from sclera. Total lipids were extracted from the retina according to Folch's method with chloroform/methanol 54 , and quantified using Cholestest ® (Sekisui Medical Corp. Tokyo, Japan).

LC-MS/MS. 24(S)-HS. The measurement of 24(S)-HC was based on previously reported methods,
with some modification 55 . Briefly, the retina of the empty eyecup was detached from sclera at the end of each experiment. After a rat retina was homogenized in distilled water, 24-hydroxy-cholesterol-d 7 was added to the suspension as an internal standard. Butylated hydroxytoluene and 1 N potassium hydroxide were added to the suspension, and then saponified at 37 °C for 1 h. After saponification, distilled water was added, and 24S-hydroxycholesterol was extracted with hexane. The extract was evaporated to dryness, and the residue was picolynoyl-esterified and subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS). For measurement of 24(S)-HC in retina, an API-4000 triple stage quadrupole mass spectrometer equipped with positive electrospray ionization (ESI) (AB Sciex, Mass, USA) connected to Nexera ultra high performance liquid chromatography systems (Shimadzu, Kyoto, Japan) was employed. The column was a Hypersil GOLD column (150 × 2.1 mm, 3 μ m, Thermo Fisher Scientific, MA, USA) used at 40 °C. The mobile phase consisting of 0.1% acetic acid (solvent A) and acetonitrile-methanol (50:50, v/v) (solvent B) was used with a gradient elution. For quantification of 24S-hydroxycholesterol, transition of m/z 635.4/512.0 and m/z 642.4/519.5 were selected for 24S-hydroxycholesterol and 24S-hydroxycholesterol-d 7 , respectively. Results were expressed as 24(S)-HC levels (ng) per total retinal weight (g).
AlloP. Three eyes were examined by LC-MS/MS in each condition. The measurement of AlloP was based on previously reported methods 30 . Light Microscopy. At the end of each experiment, eyecup preparations were fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer overnight at 4 °C. The fixed eyecups were rinsed in 0.1 M phosphate buffer and placed in 1% buffered osmium tetroxide for 60 minutes. The eyecups were dehydrated with an ethanol dilution series, embedded in epoxy resin (Epon 812, TAAB Laboratories, Aldermaston, UK) and cut into 1 μ m thick semi-thin sections. The tissue was then stained with toluidine blue and evaluated by light microscopy.

Data Analysis.
In histological studies, we examined the middle portion of the retina, greater than 1,200 μ m away from the center of the optic disc along the inner limiting membrane (ILM) according to previously described methods 13,14 . The nerve fiber layer thickness (NFLT) was measured by light microscopy along 5-6 lines perpendicular to the pigment epithelium at a distance of 15 μ m from each other around 1,200 μ m away from the center of the optic disc ( Fig. S3-2). The average NFLT was determined in 10 different light micrographs taken from 3 to 5 eyecup samples in each condition, divided by total retinal thickness, and mean ± standard deviation (SD) was analyzed and compared with control.
The severity of neuronal damage was assessed by light microscopy in ten fields from each experiment using a neuronal damage score (NDS) as previously described 13 . The NDS was determined in 10 different light micrographs taken from 3 to 5 eyecup samples in each condition.
The density of degenerated cells in the GCL was determined by counting 10 fields of 250 μ m length at 10 different locations in light micrographs taken from the block of the middle retinal part 950 to 1450 μ m away from the center of the optic disc.
These morphometrical parameters were assessed by three raters, who remained unaware of the experimental condition. Upon completion of data assessment, significance of individual differences among raters was evaluated using five randomly selected samples in each morphometric parameter by one-way analysis of variance (one-way ANOVA) followed by a post-hoc test. There were no significant differences among the raters in any of the morphometric measurements.
Preparation of whole mounted retinas. The anterior part of the eye was removed by making an incision along the entire limbus. After incubation in the closed pressure system, retinas from five eyes in each group were processed for immunostaining as "whole mounted" retinas. The retina was carefully detached from the eye by making cuts along the ora serrata and optic nerve. Whole retinas were then flat-mounted, pinned out in an acrylic plate with the RGC layer facing upward using stainless steel pins, and fixed in 4% paraformaldehyde-0.1 M phosphate buffer overnight at 4 °C. After the samples were fixed, the tissue was rinsed with PBS three times. To block nonspecific binding, the tissue was incubated in 2% BSA in PBS containing 0.5% Triton X-100. The whole mounted retinas were incubated in the rabbit anti-NeuN polyclonal antibody solution (Cat#ab104225, Abcam) (1:100) by gently shaking at 4 °C, overnight. After rinsing 3 times using PBS, the retina was incubated in FITC-conjugated secondary antibody (goat anti-rabbit IgG (H + L)) (Cat#81-6111, Zymed Laboratories Inc) (1:300). The retina tissue was then rinsed 3 times with PBS and mounted on glass slides using 50% PBS and 50% glycerol.
The retinal flat-mounts were imaged throughout the GCL in each of the four defined retinal quadrants 4 mm from the optic nerve head using a confocal microscope. Each quadrant was analyzed using a 1 mm 2 frame, and counted using Image-Pro Plus software. The density of NeuN positive RGCs per square millimeter was averaged and compared in experimental and control retinas 56 . RGC counts were analyzed using Image-Pro Plus software.

Apoptosis.
To visualize apoptotic cells, we used the DeadEndTM Colorimetric TUNEL System (Promega, Madison, WI) to determine the apoptotic cells according to the manufacturer's instructions. The nuclei were counterstained with DAPI. Five retinal sections were randomly selected per each condition. After the length of each section was measured (Image-Pro Plus software), the TUNEL-positive cells were counted in the whole section length (between ora serrata to ora serrata). The number of apoptotic cells was expressed per 200 μm of retinal section 57 .