Neuroprotective effects of exogenous erythropoietin in Wistar rats by downregulating apoptotic factors to attenuate N-methyl-D-aspartate-mediated retinal ganglion cells death

The aim of this study was to investigate whether exogenous erythropoietin (EPO) administration attenuates N-methyl-D-aspartate (NMDA)-mediated excitotoxic retinal damage in Wistar rats. The survival rate of retinal ganglion cells (RGCs) were investigated by flat mount analysis and flow cytometry. A total of 125 male Wistar rats were randomly assigned to five groups: negative control, NMDA80 (i.e., 80 nmoles NMDA intravitreally injected), NMDA80 + 10ng EPO, NMDA80 + 50ng EPO, and NMDA80 + 250ng EPO. The NMDA80 + 50ng EPO treatment group was used to evaluate various administrated points (pre-/co-/post- administration of NMDA80). Meanwhile, the transferase dUTP Nick-End Labeling (TUNEL) assay of RGCs, the inner plexiform layer (IPL) thickness and the apoptotic signal transduction pathways of μ-calpain, Bax, and caspase 9 were assessed simultaneously using an immunohistochemical method (IHC). When EPO was co-administered with NMDA80, attenuated cell death occurred through the downregulation of the apoptotic indicators: μ-calpain was activated first (peak at ~18hrs), followed by Bax and caspase 9 (peak at ~40hrs). Furthermore, the images of retinal cross sections have clearly demonstrated that thickness of the inner plexiform layer (IPL) was significantly recovered at 40 hours after receiving intravitreal injection with NMDA80 and 50ng EPO. Exogenous EPO may protect RGCs and bipolar cell axon terminals in IPL by downregulating apoptotic factors to attenuate NMDA-mediated excitotoxic retinal damage.


Introduction
Glaucoma is one of the major causes of irreversible blindness worldwide [1,2]. It is a group of optic neuropathies characterized by the loss of retinal ganglion cells (RGCs) [3]. Even  appearance of transferase dUTP Nick-End Labeling (TUNEL)-positive cells. It can even prevent mitochondrial depolarization. However, there were very few reports on the preventative effects of EPO on RGC death caused by glaucoma or chronic neurodegenerative diseases [39]. A previous study evaluated the effects of EPO on RGCs in episcleral vessel cautery-induced rats with glaucoma using 200ng EPO, which found no significant decrease in the RGC of the EPO-treated group [40][41][42]. However, little is known about how EPO signal transduction occurs in vivo and the practical usefulness of EPO in the prevention of chronic purely apoptotic neuronal cell death, which contributes to vision loss in glaucoma and the progression of neurodegenerative diseases. The objective of this study was to establish NMDA-mediated excitotoxic retinal damage in Wistar rats, and then to determine the number of RGC in GCL and the thickness of IPL at different time after receiving NMDA80 and 50ng EPO. In addition, the downregulation of the apoptotic signal transduction pathways of μ-calpain, Bax, and caspase 9 may involve the crosstalk in NMDA-mediated neurotoxic rats with EPO treatment.

Animal model
A total of 125 male Wistar rats (Taiwan National Laboratory Animal Center, Taipei, Taiwan) weighing between 225 and 250 g (8 weeks of age), were group housed four per cage in a temperature-controlled environment (21-22˚C) under a 12-h light-dark cycle. Food (Purina Rodent Chow) and water were provided ad libitum. We provided environmental enrichment items (e.g., wooden blocks, chew sticks, and other wood toys) to support their behavioral needs. All studies were conducted in accordance with the guidelines of the Association for Research in Vision and Ophthalmology Statement on the Use of Animals in Ophthalmic and Vision Research. The study protocol was approved by the Institutional Animal Care and Use Committee of National Defense Medical Center (Permit number: IACUC-07-175 and IACUC-08-209).

Drug treatment
NMDA (E5627, Sigma, St. Louis, MO)-mediated retinal neurotoxicity was established as reported previously by Nivison-Smith et al [19]. The rats were briefly anesthetized with a mixture of ketamine hydrochloride (50 mg/kg, Nang Kuang Pharmaceutical, Tainan, Taiwan, Republic of China) and xylazine hydrochloride (13.3 mg/kg, X1126, Sigma, St. Louis, MO). After the topical administration of 0.5% proparacaine hydrochloride (Alcon Lab, Fort Worth, TX), an intravitreal injection of 2 μL 40 mM NMDA (M3262, Sigma, St. Louise, MO) in BSS PLUS solution (Alcon Lab, Fort Worth, TX,) was placed into the right eye of each rat using a 30-gage needle (Hamilton, Reno, NV,). The solution was administered into the sclera approximately 1 mm behind the limbus. The BSS PLUS solution was used as a vehicle control and the NMDA and different doses of EPO were used in the therapeutic treatment groups. Two of them were injected into the left eye. We co-injected different doses of EPO with NMDA for the left eye. A group of male Wistar rats were randomly assigned to five groups: negative control, NMDA80 (i.e., 80 nmoles NMDA intravitreally injected), NMDA80 + 10ng EPO, NMDA80 + 50ng EPO, and NMDA80 + 250ng EPO. The NMDA80 + 50ng EPO treatment group was selected to evaluate various administrated points (pre-/co-/post-administration of NMDA80). All solutions were administered into the sclera approximately 1 mm behind the limbus.

Retrograde labeling of retinal ganglion cells
RGCs were labeled with FG (Fluorogold) (39286, Sigma, St. Louis, MO) by injecting the FG solution into the superior colliculi using a stereotaxic device (Stoelting, Wood Dale, IL) as described previously [43]. Seven days after NMDA induction, the rats were quickly anesthetized with the carbon dioxide (CO 2 ). The expected time to unconsciousness was usually within 2-3 minutes to ensure that the endpoint is humane. Each rat was observed for lack of respiration and fixed and dilated pupil before the eyes were enucleated. The skin over the cranium was then incised to expose the scalp. Two vertical holes 1 mm in diameter were drilled on both sides of the skull with a dentist drill 6 mm posterior to the bregma and 1.5 mm lateral to the midline. Two microliters of 3% FG solution was delivered using a micropipette 3.8, 4.0, and 4.2 mm beneath the bone surface.

Retinal flat mount imaging and FG-labeled RGC counting
Seven days after the NMDA induction, the rats were euthanized with carbon dioxide (CO 2 ). The expected time to unconsciousness was usually within 2-3 minutes to ensure that the endpoint is humane. Each rat was observed for lack of respiration and fixed and dilated pupil before the eyes were immediately enucleated. The retinas were dissected and fixed in 4% paraformaldehyde (16005, Sigma, St. Louis, MO) for 1 hour. After washing with phosphate buffered saline (PBS), the retinal flat mount samples were placed as four-radial incisions and were prepared on slides in PBS with 10% glycerol (G7893, Sigma, St. Louis, MO). The retinal slides were examined under a fluorescence microscope (Olympus BX-50, Olympus Optical, Tokyo) using UV excitation (330-385 nm) and a barrier filter (420 nm) to determine the RGC cell number. Digital images were taken using a CCD camera (SPOT, Diagnostic Instruments, Sterling Heights, MI). Each retina was visually divided into four quadrants (i.e., superior, inferior, nasal and temporal). The quadrants were further divided into central (0.8-1.2 mm from the optic disk), middle (1.8-2.2 mm from the optic disk), and peripheral regions (0.8-1.2 mm from the retinal border). Within each region, the cells falling in two fields (200 × 200 μm 2 in size) were counted. Both manual and automatic counts were used to count RGCs. For automatic counts, the digital images were processed using Image J (http://rsbweb.nih.gov/nih-image/, U.S. National Institutes of Health, Bethesda, MD) and the RGB (red-green-blue) images were converted to 8-bit grayscale for binary counting [22]. The RGCs were classified into three groups based on soma sizes described previously (small 9.4 μm, medium 9.4-12.6 μm, and large 12.6 μm) [30]. FGlabeled cells were counted and the mean number of RGCs per square millimeter was calculated.

Flow cytometry
Retinal cells were evaluated using a flow cytometer (Calibur; Becton Dickinson, San Jose, CA). The parameters of the cytometer were optimized with the following settings, which were based on preliminary studies: the FSC (forward scatter) value used for reflecting cell volume/size and the SSC (side scatter) value used for reflecting fluorescence intensity and its internal complexity. FL1, FL2, and FL3 reflect the green (FITC, fluorescein), orange (PE, phycoerythrin-R), and red (peridinin-chlorophyll protein (PerCP), peridinin-chlorophyll) labeled fluorescence intensities, respectively. These) were applied with a high flow rate (60 μl/min). The compensation values were determined from CaliBRITE three-color kit (BD Biosciences, Franklin Lakes, NJ). The cell sizes were estimated from the FSC using a calibration curve, which was established by 6, 10, 15, and 20 μm diameters of polystyrene microspheres (Polysciences, Warrington, PA). Ten-thousand cells per sample were counted automatically; and incubated cells without the fluorescence-conjugated antibody were used as an unstained blank. The expression of FG intensity and Thy1.1 were evaluated as geometric means of the FL2 and FL3 fluorescence intensities, respectively. The fluorescence intensity data was analyzed using CellQuest Pro (Becton Dickinson, San Jose, CA) and FCS Express V3 software (De Novo Software, Los Angeles, CA). The cell viability can be used to evaluate the health of cells due to the dye reacting with cellular amines. In necrotic cells, the reactive dye can enter the cell via the compromised membranes and react with the free amines in the interior and on the surface of the cell, resulting in an intense fluorescent staining in unhealthy or dead cells. Cell-surface staining of amines in the viable cells will result in relatively dim staining that can easily be distinguished from the bright staining of dead cells.

Histological analysis
After enucleation, the eyes were immersed overnight at 4˚C in a fixative solution consisting of 4% paraformaldehyde in PBS (pH 7.4). This was followed by dehydration with gradientincreased ethanol (75-to 100%) and embedded in paraffin. The retina cross-section had a thickness of 4 μm was prepared and stained with hematoxylin and eosin. The morphology analysis was conducted using an optical microscope. The cell numbers of the ganglion cell layer (GCL) were counted 1.0 to 1.5 mm to the margin of the optic nerve head. Five samples were taken with an interval of 100 μm, and the mean thickness of the inner plexiform layer was analyzed. The thickness of the outer nuclear layer was used as the internal standard.

Immunohistochemistry, IHC
The histosection of the retina was attached on gelatin-coated glass slides. The slides were dewaxed with xylene twice, rehydrated with a decreasing-gradient ethanol solution (100 to 80%), then rinsed three times with PBS. Citrate buffer (0.01M pH 6) was used to maintain boiling for 20 minutes and rinsed three times with PBS. BSA 0.1% was used as a blocking agent to prevent nonspecific binding of antigens and antibodies for 10 minutes. The primary antibodies of Bax (1:100), caspase-9 (1:100), and μ-calpain (1:100) were reacted for 1 hour individually. After washing with PBS, animal serum was used for a post-primary blocking agent for 15 mins. The Novolink polymer (anti-mouse/rabbit IgG-Poly-HRP) was reacted with the primary antibody for 15 minutes, 3, 3'-diaminobenzidine (DAB chromogen), and 0.02% hematoxylin to stain the nucleus. The coverslips were washed three times with PBS, the samples were embedded with Fluoromount-G and then examined under an optical microscope. GCL and positive cells were counted 1.0 to 1.5 mm to the margin of the optic nerve head. The distribution of EPOR was examined, the primary antibodies of goat anti-EPOR (1:50) and mouse anti-Thy-1 (1:50) were stained overnight. These were then washed thrice with PBS and the secondary antibody of Alexa Fluor 488-labeled donkey anti-goat IgG (1:200) was incubated for 30 minutes. After the washing, Alexa Fluor 594-labeled goat anti-mouse IgG was reacted and washed three times with PBS. DAPI was used to detect nuclei in fluorescence.

Statistical analysis
All data was expressed as median. The data was analyzed using Mann-Whitney U-test in SPSS v22 software (IBM, USA). Differences with P-values less than 0.05 were considered statistically significant.

EPO has neuroprotective function against NMDA on RGCs after 7 days
EPO dose-response by retinal flat mount analysis. The retrograde RGC labeling was performed with FluoroGold (FG) and applied to the surface of the optic nerve. The survival rate of RGC was analyzed using retinal flat mount analysis and individually rescued with the following different dose of EPO (0, 10, 50, or 250 ng) ( Table 1). After seven days, NMDAmediated neurotoxicity was effective in triggering RGC death and the survival rates of total, medium, and small RGCs were reduced to 57.85 ± 5.69% (p < 0.001), 62.72 ± 6.83% (p < 0.001), and 52.05 ± 7.28% (p < 0.001) (n = 22). NMDA-mediated neurotoxicity reduced the survival rate of RGC compared to the control groups (which had a 100% survival rate). Large RGCs had a better survival rate due to a low response rate from NMDA-mediated neurotoxicity. Under 10 ng EPO treatment, the survival rate of RGC after NMDA-mediated neurotoxicity was not significantly improved. Under 50 ng of EPO promoted a significantly higher survival rate of the total, medium, and small RGCs after NMDA-mediated neurotoxicity, which were 84.09 ± 8.04% (p < 0.01), 104.53 ± 9.64% (p < 0.001), and 76.22 ± 12.79% (p < 0.01) (n = 7) of the non-EPO treatment, respectively. When the dosage was titrated to 250 ng EPO, it only produced a slightly higher neuroprotective function for medium RGCs and the survival rate was 82.87 ± 12.46% (p < 0.05) (n = 7) compared to the NMDA-treated 50 ng EPO treatment. However, this was not statistically significant. The large RGCs are less responsive to co-administration of NMDA and EPO (p > 0.05). Overall, 50 ng EPO had a better neuroprotective function on medium and small RGCs, where there was an overall increase in the survival rate of RCG. EPO pre/co/post-treatment by retinal flat mount analysis. The neuroprotective function of EPO against NMDA-mediated neurotoxicity pretreatment, co-treatment, and posttreatment were evaluated (Fig 1). The survival rates of total, large, medium, and small RGCs were evaluated by flat mount analysis in NMDA-mediated neurotoxicity at doses of 80 nmoles. The administration of 50 ng EPO 4 or 8 hours pre-or post-NMDA induction was conducted. The co-administration of NMDA and EPO was conducted as the control group (100% survival rate, normalized).
For the post-treatment, EPO was applied 4 hours after NMDA induction, the survival rates of total, large, medium, and small RGCs were 82.11 ± 7.85%, 102.34 ± 9.85%, 83.87 ± 9.18, and 77.18 ± 13.62% (n = 8), respectively. When the EPO was applied 8 hours after NMDA induction, the survival rates of total, large, medium, and small RGCs were 41.93 ± 3.84%, 97.25 ± 8.56%, 46.50 ± 9.23%, and 38.69 ± 10.80% (n = 8), respectively. EPO application 4 hours after NMDA induction resulted in a significantly higher survival rate for total (p<0.001), medium (p < 0.01), and small (p < 0.05) RGCs compared to the EPO post-treatment group at 8 hours. In addition, in the same interval of pretreatment and post-treatment, the EPO-pretreated 8 hours NMDA induction had a higher survival rate for total RGCs (p<0.05) compared to the group for EPO post-treated 8 hours after NMDA induction (Fig 1). When the NMDA and EPO were co-administered at the same time, the survival rates of the total, medium, and small RGCs was at the highest compared with the pre-and post-treatment at 4 or 8 hours. The results also illustrated that the large RGCs had much lower response rates pre-and post-treatment, but had more resistance to NMDA toxicity, as indicated by the retinal flat mount analysis.

The time course trials of EPO recovered NMDA damage in RGCs
EPO rescued the survival rate of RGCs. TUNEL assay is a method used for detecting apoptotic DNA fragmentation. It is widely used to identify and quantify apoptotic cells. In the results of the TUNEL assay in the Wistar rats (Fig 3A and 3B Recent studies have shown that segmented IPL thickness was associated with glaucomatous damage [12,42]. In this study, the histological results stained with hematoxylin and eosin (H&E) were examined for the neuroprotective effect of EPO. At the times intervals of 0, 6, 18, 30, 42, 66, and 90 hours, respectively, the RGCs counts in the GCL and the ratio of IPL/outer nuclear layer (ONL) were measured (Fig 4A). The results showed that EPO could protect the cells in the GCL and it may reverse the damage to bipolar cell axon terminals in IPL induced by NMDA. The 42, 66, and 90 hours after intravitreal injection, the median of RGCs counts in GCL was significantly lower by NMDA induction compared to the control [56.5 vs. 85.5 (p < 0.05, n = 4), 57.5 vs. 86.5 (p < 0.05, n = 4), and 52 vs. 84.5 cells/mm (p < 0.05, n = 4)] (Fig 4B). The median ratio of IPL/ONL was also significantly lower at 66, and 90 hours after receiving NMDA80, compared to the control group [0.56 vs. 0.92(p < 0.05, n = 4), and 0.53 vs. 0.91 (p < 0.005, n = 4)]. However, the 42, 66, and 90 hours after receiving intravitreal injection with 80 nmoles NMDA and 50 ng EPO, the median of RGCs counts were increased significantly to 78.5 (p < 0.05, n = 4), 94 (p < 0.05, n = 4), and 82 cells/mm (p < 0.05, n = 4) and the median ratio of IPL/ONL also increased significantly to 0.73 (p < 0.05, n = 4) after 90 hours (Fig 4C).
EPO protected the RGCs by downregulating apoptotic factors. Immunohistochemical staining using an antibody raised against EPOR protein (green color) revealed an intense immunoreactivity in the RGC layer of the retina. Thy-1 (red color) is a surface glycoprotein uniquely expressed on the RGC of retinas. We found an intense immunoreactivity in the RGC layer of Wistar rats. DAPI (blue color) was used for immunohistofluorescence staining. These staining results demonstrated that almost all RGCs were immunoreactive for EPOR (Fig 5).
In order to analyze the signal transduction of RGCs after EPO injection, the apoptotic signals of μ-calpain, caspase 9, and Bax were measured by immunohistofluorescence and immunohistochemistry (Fig 6, S1, S2 and S3 Figs).
The median percentage of μ-calpain-positive cells in GCL were 14.66 and 3.98 (p < 0.05, n = 4) after 18 hours for the NMDA-mediated groups and the control groups (BSS) (Fig 6  upper), respectively. Compared to the NMDA-mediated groups, the median percentage of μcalpain-positive cells in GCL was lower to 4.64 ± 0.64% (p < 0.05, n = 4) at 18 hours after combining 80 nmoles NMDA and 50 ng EPO treatment. The median percentage of caspase 9-positive cells in GCL were 3.59 and 17.83 (p < 0.05, n = 4) after 42 hours for the control (BSS) and the NMDA-mediated groups (Fig 6 middle), respectively. The median percentage of caspase 9-positive cells in GCL were 1.76 and 12.48 (p < 0.05, n = 4) after 66 hours for the control (BSS) and the NMDA-mediated groups. The median percentage of caspase 9-positive cells in  GCL were 1.19 and 7.38 (p < 0.05, n = 4) after 90 hours for the control and NMDA-mediated groups. Compared to the NMDA-mediated groups, the median percentage of caspase 9-positive cells in GCL were 5.8 (p < 0.05), 2.93 (p < 0.05) and 0.86 (p < 0.05) after 42, 66 and 90 hours, respectively, after combining 80 nmoles NMDA and 50 ng EPO treatment.
The median percentage of Bax-positive cells in GCL were 12.12 and 25.28 (p < 0.05, n = 4) after 42 hours for the control (BSS) and NMDA-mediated groups (Fig 6 lower), respectively. The median percentage of Bax-positive cells in GCL were 8.14 and 13.33 (p < 0.05, n = 4) after 66 hours for the control and NMDA-mediated groups, respectively. Compared to the NMDAmediated groups, the median percentage of Bax-positive cells in GCL were lower in the 80 nmoles NMDA and 50 ng EPO treatment group, with 12.19 (p < 0.05, n = 4) and 10.24

PLOS ONE
(p < 0.05, n = 4) after 42 and 66 hours, respectively. At 90 hours, the median percentage of Bax-positive cells in GCL was higher in NMDA-mediated group, compared to the controls and NMDA80-+50 ng EPO treatment; however, this did not reach statistically significant. The present study revealed that erythropoietin (EPO) protects RGCs against NMDA-mediated apoptosis in Wistar rat models. It blocked the generation of pro-apoptotic proteins, such as μcalpain, Bax, and caspase 9.

Discussion
Our study has shown that NMDA was able to initialize the apoptosis of RGCs. The EPO treatment can then reverse the NMDA-mediated neurotoxicity on the survival rate of small and medium RGCs by retinal flat mount analysis in an experimental model in vivo (Table 1). The large RGCs were more resistant to NMDA-mediated neurotoxicity. Thus, the EPO treatment (10 ng, 50 ng, or 250 ng) did not significantly alter the survival rate. A difference in response between small and large RGCs was also observed in studies of brain-derived neurotrophic factors [44], and in our previous studies in vitro [30]. These provide support for the notion that large RGCs are more resistant to the NMDA-mediated toxicity. We also found that the effect of EPO was dose-dependent. Under low dosage (10 ng), there was no evident protective effect across small, medium, or large RGCs. When the dose of EPO increased to 50 ng, the protective effect was significant for the medium and small RGCs. For the EPO dosage of 250 ng, the protective effect on the medium RGCs remained but was absent for the small RGCs. Under the higher EPO dosage (250 ng), its potential neuroprotective effects and underlying mechanism in small RGCs was unclear. Andrews et al found that high doses of EPO can have thrombotic toxicities on rats [45]. Therefore, high EPO dosage may have had thrombotic toxicities, causing ischemia and damage to the small RGCs.
We found that with EPO treatment 4 hours after NMDA induction (post-treatment) had higher survival rates for total, medium, and small RGCs compared to the EPO control treatment 8 hours after NMDA induction (Fig 1). Hence, earlier EPO post-treatment significantly enhanced survival rates, due to the stability and intervention of EPO. In addition, NMDA requires time to induce neurotoxicity on RGC. After 8 hours, the EPO pretreatment group had a higher survival rate for total RGCs than the post-treated group for the post-treatment. This could be due to irreversible, long-lasting NMDA-mediated neurotoxicity on these RGCs after 8 hours, where EPO did not produce a significant effect. Furthermore, NMDA-mediated a significantly higher apoptotic signal (μ-calpain and Bax) than BSS after 6 hours (Fig 6) and EPO requires a certain amount of time to activate the signal transduction pathway that reduces the apoptosis of RGCs. Therefore, the neuroprotection effect of EPO is more apparent when it is presented prior to the NMDA treatment. However, when we administered NMDA and EPO simultaneously, the survival rates of total, medium, and small RGCs were higher compared to the pretreatment groups after 4 or 8 hours, suggesting that EPO may degrade after several hours. Therefore, EPO was administered too early before the NMDA-mediated neurotoxicity, the concentration of EPO would decline and limiting its neuroprotective effect.
The flow cytometry showed that under NMDA-mediated neurotoxicity, EPO increased the survival rate of small and medium RGCs but not in large RGCs. It also recovered the health of the soma and axon in large, medium, and small RGCs. This result was consistent with our previous retinal flat mount analysis (Fig 2). This suggests that large RGCs were more resistant to NMDA toxicity. In other words, glaucoma unequally causes more damage to smaller RGCs compared to larger RGCs. This result was also consistent with our previous study using retinal cell culture method in vitro [30]. The responses between small and large RGCs were inconsistent across different studies. Mey et al showed that cutting down the optic nerve in rats causes the apoptosis of the small RGCs, which was consistent with our study [46]. However, Quigley et al showed that chronic glaucoma will selectively damage the large RGCs in monkeys and humans [47][48][49]. The inconsistency in the results may be attributable to the different experimental conditions and design, such as different animal species and different types of damage to RGCs.
The results of the retinal cross-section analysis by TUNEL assay showed that EPO rescued the NMDA-mediated damaged cells 30 to 90 hours after NMDA injection (Fig 3). Furthermore, through H&E staining we found that EPO could protect the cells in GCLs and it may decrease the damage to bipolar cell axon terminals of IPL at 66 to 90 hours after NMDA toxicity was initiated. These findings were consistent with the flow cytometry and retinal flat mount results, which further confirmed that EPO could protect the ganglion cell body, but also protect bipolar cell axon and reverse the damage to IPL by NMDA The immunohistofluorescence staining in vivo demonstrated that EPOR was expressed intensely in the RGCs of Wistar rats. This proved that EPO could act on RGCs via EPOR ( Fig  5). Therefore, EPO may show its neuroprotection effect through binding with EPOR on the ganglion cells, further the activation of the signal transduction pathway in order to inhibit apoptosis of the ganglion cell.
In order to explore the effects of the upregulation of EPO on neuroprotection signal transduction in rat model in vivo, the RGCs apoptosis associated factor such as μ-calpain, caspase 9, and Bax were investigated. The median percentage of μ-calpain-positive cells were increased significantly in the GCL at 6 to 18 hours after NMDA injection, whereas the median percentage of Bax-positive cells were significantly increased in the GCL after 6 hours and 42-66 hours after NMDA injection (Fig 6, S1, S2 and S3 Figs). The median percentage of caspase 9-positive cells were significantly increased in the GCL 42 to 90 hours after NMDA injection. This suggest that during the signal transduction of NMDA-mediated neurotoxicity, μ-calpain was activated first, followed by Bax, and finally caspase 9. The percentage of Bax-positive cells also increased significantly in the GCL at 6 hours after NMDA injection. Yet, when the results after 6, 42, and 66 hours were compared, the median percentage of Bax-positive cells was the highest after 42 hours, where Bax was activated after μ-calpain. This signal transduction cascade of NMDA-mediated apoptosis was consistent with previous studies [30]. Hartwick et al have shown that the activation of the NMDA receptors will cause the death of RGCs, increase the flow of Ca 2+ , and activates μ-calpain in after~12 minutes [50]. In glaucoma patients, Wax et al found that caspase 8 was activated in the first stage (beginning stage) and caspase 9 was activated in the second stage (decisional stage) during apoptosis of RGCs [51]. EPO inhibits the apoptotic pathway to attenuate NMDA-mediated excitotoxic retinal damage. The significant increase of μ-calpain after 6 to 18 hours post-NMDA treatment (Fig 6) and the poor performance of post-treatment on RGC survival (80 nmoles NMDA, 50 ng EPO) imply that μ-calpain was crucial in determining the survival rate of RGC with EPO treatment. The posttreatment did not produce an effect due to the delayed onset time and the continue rise of μcalpain. EPO should be administered before the significant rise in μ-calpain, especially for post-treatment application (e.g., 80 nmoles NMDA+50 ng EPO after 4 hours) for better outcomes. Schuettauf et al found that intravitreal injection of NMDA will cause cell apoptosis through extrinsic and intrinsic pathways [52]. On the extrinsic pathway, procaspase 8 will be activated to caspase 8, which will subsequently activate caspase 3. On the intrinsic pathway, the Bax will migrate to the mitochondria, where it promotes the mitochondria to release cytochrome C. The cytochrome C will activate caspase 9 which will then activate caspase 3. Caspase 3 will further activate caspase-activated DNase, causing damage to the DNA. Our study confirmed that EPO presents the neuroprotection effect by downregulating the activity of μ-calpain, Bax, and caspase 9, which results in the reduction of apoptosis of the ganglion cell through binding EPOR on the surface of ganglion cells. According to our study and previous studies, we postulated the possible pathway of neuroprotection of EPO in the apoptotic pathway (Fig 7).
Our immunohistochemical results at different time intervals revealed that NMDA requires couple hours to induce neurotoxicity on RGC, and that EPO also requires couple hours to intervene the apoptosis cascade. Thus, it is imperative to urgently find an efficient delivery method for EPO to recover the apoptosis. Nevertheless, EPO can modulate programed cell death and may be a good potential therapeutic candidate for neuroprotection. This can potentially be used to treat glaucoma.

Conclusions
Our results demonstrated that NMDA induces neurotoxicity in the majority of small and medium RGCs, and EPO treatment can protect RGC in GCL and bipolar cell axon to maintain the thickness of IPL from NMDA-mediated toxicity. In addition, EPO expressed a dosedependent effect where increasing the EPO dose significantly increases the protective effect in medium and small RGCs. Furthermore, EPO requires a certain amount of time to activate the signal transduction pathway to reduce the apoptosis of the ganglion cell. Thus, the EPO expresses better neuroprotection effect if it is administered earlier once the NMDA toxicity begins, and even before the toxicity is induced. Our findings suggest that there are EPORs on the RGCs. Hence, the EPO may show its neuroprotection effect through binding EPOR on the ganglion cells, which activates the signal transduction pathway to inhibit apoptosis. Furthermore, we observed that in this signal transduction pathway μ-calpain is activated first, followed by Bax, and then caspase 9. We also confirmed that EPO produces the neuroprotection effect by downregulating the activity of μ-calpain, Bax, and caspase 9, and facilitating the reduction of apoptosis in the ganglion cells of Wistar rats. Future studies will focus on the neuroprotective effects of EPO on human ganglion cells to explore its potential application for glaucoma patients.