Intravitreal Injection of Erythropoietin Protects both Retinal Vascular and Neuronal Cells in Early Diabetes

Abstract

purpose. To explore and evaluate the protective effect of erythropoietin (EPO) on retinal cells of chemically induced diabetic rats after EPO was injected intravitreally at the onset of diabetes.

methods. Diabetes was induced in Sprague-Dawley rats by intraperitoneal injection of streptozotocin (STZ). At the onset of diabetes, a single intravitreal injection of EPO (0.05–200 ng/eye) was performed. In the following 6 weeks, the blood retinal barrier (BRB) was evaluated by Evans blue permeation (EBP). Retinal cell death in different layers was determined by terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) staining. The retinal thickness and cell counts were examined at the light microscopic level. Electron microscopy (EM) was used to scrutinize retinal vascular and neuronal injury. Neurosensory retinas of normal and diabetic rats were used as the sources of reverse transcription–polymerase chain reaction (RT-PCR) and Western blot for the detection of EPO, EPO receptor (EpoR), and products of the extracellular signal-regulated kinase (ERK) and the signal transducers and activators of transcription 5 (STAT5) pathways. The distribution of EpoR in retinal layers was demonstrated by immunohistochemistry (IHC).

results. In the diabetic rats, BRB breakdown was detected soon after the onset of diabetes, peaked at 2 weeks, and reached a plateau at 2 to 4 weeks. The number of TUNEL-positive cells increased in the neurosensory retina, especially, the outer nuclear layer (ONL) at 1 week after diabetes onset and reached a peak at 4 to 6 weeks. The retinal thickness and the number of cells in the ONL were reduced significantly. EM observations demonstrated vascular and photoreceptor cell death starting soon after the onset of diabetes. All these changes were largely prevented by EPO treatment. Upregulation of EpoR in the neurosensory retina was detected at both the transcriptional and protein levels 4 to 8 weeks after the onset of diabetes, whereas, the endogenous EPO levels of neurosensory retinas were essentially unchanged during the same period observed. In EPO-treated diabetic groups, EpoR expression remained at upregulated levels. Within 2 weeks of the onset of diabetes, activation of the ERK but not the STAT5 pathway was detected in the diabetic retina treated with EPO.

conclusions. These data demonstrate that apoptosis is an major contributor to neuronal cell death in the early course of diabetic retinopathy (DR). The upregulation of EpoR may be a compensatory response of retinal cells and tissue to diabetic stresses. The EPO/EpoR system as a maintenance–survival mechanism of retinal neurons responds to the insults of early diabetes other than ischemia. The protective function of EPO/EpoR at the least acts through the EpoR-mediated ERK pathway. Exogenous EPO administration by intravitreal injection in early diabetes may prevent retinal cell death and protect the BRB function. Therefore, this is a novel approach for treatment of early DR.

Diabetic retinopathy (DR) is the leading cause of blindness in patients 20 to 70 years of age. ¹ The direct cause of DR is still largely unknown. The notion that it is solely a microvascular complication in diabetes has been challenged in recent years. ^{2 3 4} Evidence exists that all classes of cells within the retina are involved in a multiplicity of disease processes in the early stages of diabetes. ^{3 4} The early clinical features of DR, such as breakdown of the blood–retinal barrier (BRB) and various visual deficits support this concept in humans. ^{5 6 7 8} Because of gradual but accelerating deterioration, treatment should be implemented before DR progresses to the irreversible stage. Unfortunately, there are no known treatments specifically for mild to moderate DR. ^{9 10}  

Growing evidence suggests that erythropoietin (EPO) has both neuroprotective and vascular protective functions. ^{11 12 13 14} EPO and EPO receptor (EpoR) are expressed in the human retina ¹⁵ and central nervous system. EPO promotes neural outgrowth from retinal ganglion cells in a dose-dependent manner and preserves their survival after axotomy. ^{16 17 18} Hypoxia-induced retinal EPO expression appears to protect retinal neurons from transient global ischemic and reperfusion injury through an antiapoptotic pathway. ¹⁹ The neurotrophic effect of EPO in the retina extends beyond damage from ischemia and axotomy. Systemic EPO administration protects retinal photoreceptors from light-induced apoptotic pathways in retinal degeneration models through a speculated interaction of EPO with EpoR in the photoreceptor inner segment. ^{20 21} An inhibited production of systemic EPO has been clinically observed in early diabetic nephropathy that results in anemia that is associated with an aggravated course of DR. ^{22 23} Intravenous administration of EPO to treat azotemia-induced anemia in diabetic patients demonstrated a beneficial effect on macular edema and improved visual outcome. ²⁴  

In human eyes with proliferative diabetic retinopathy (PDR), elevated EPO levels were detected in the vitreous, suggesting that EPO may be produced as an endogenous neuroprotectant against ischemia. ²⁵ Meanwhile, a recent report showed that elevated EPO in the vitreous in PDR may act as an independent angiogenic factor leading to hypoxia-induced neovascularization. ²⁶ Recently, high vitreous EPO levels were also observed in patients with diabetic macular edema (DME), in which ischemia is not a predominant event. ²⁷ Apparently, the biological significance of EPO in the nonischemic retina is not completely understood. The present work was designed to study early diabetic retinas lest it be involved in angiogenic states in PDR. The STZ-induced diabetic rats were used because this animal model represents only cellular processes characteristic of human nonproliferative diabetic retinopathy (NPDR). Particularly, within 8 weeks after diabetes onset, the retina is not ischemic. By using this model, the EPO/EpoR network was studied at both the mRNA and protein levels. The upregulation of EpoR in the retina of STZ-induced diabetic rats has prompted us to explore whether exogenous EPO as a cytoprotectant, could prevent cellular injury in the early diabetic retina. The EpoR-mediated signaling pathway, ERK, was also studied. Since systemic treatment of retinopathy with EPO may be limited by EPO’s erythropoietic and angiogenic properties, ^{20 26} intravitreal injection was the preferred approach. The protective effects of EPO on the BRB and vascular and neuronal cells were clearly demonstrated. 

Materials and Methods

Reagents

Evans blue (30 mg/mL), streptozotocin (STZ; pH 4.5), and EPO (r-Hu-EPO, 0.1 μg/μL in normal saline, i.e., 153 U/mg) were purchased from Sigma-Aldrich (Beijing Superior Chemicals and Instruments Co., Ltd., Beijing, China). Cell-viability assay and EPO kits (In situ Cell Death Detection Kit and EPO ELISA Kit) were purchased from Roche China, Ltd. (Shanghai, China). Anti-EPO antibody (H-162) and anti-EpoR antibodies (H-194, M-20) were purchased from Santa Cruz Biotechnology (Gene Company Ltd., Shanghai, China). Phospho-STAT5 (Tyr694) antibody, STAT5 antibody, phospho-p44/42 MAP kinase (Thr202/Tyr204) antibody, and p44/42 MAP kinase antibody were purchased from Cell Signaling Technology (Genetimes Technology, Inc. Shanghai, China). Biotinylated goat anti-rabbit IgG (BA1003) and SABC kits were purchased from Boster (Wuhan, China). 

RNA Isolation and Determination of Gene Expression

Total RNA was extracted from the retinal samples. The RT product (1 μL) was then amplified by PCR. The specific primers were designed on computer with commercial (Primer Premier ver. 5.0) purchased from Shanghai DNA Biotechnology Corp., Ltd. (Shanghai, China). The primers for EpoR were 5′-CTGGGAGGAAGCGGCGAACT-3′ (sense) and 5′-CGGTGGTAGCGAGGAGAT-3′ (antisense), and the size of the amplified fragment was 213 bp. The primers for EPO were 5′-CTCCAATCTTTGTGGCATCT-3′ (sense) and 5′-GGCTTCGTGACCCTCTGT-3′ (antisense), and the size of the amplified fragment was 134 bp. PCR products for β-actin were used as a positive control and internal standard. The primers for β-actin were 5′-GTAAAGACCTCTATGCCAACA-3′ (sense) and 5′-GGACTCATCGTACTCCTGCT-3′ (antisense). The size of the amplified fragment was 227 bp. Amplification conditions included an initial denaturation at 95°C for 5 minutes, followed by 35 cycles of denaturation at 94°C for 30 seconds, annealing at 55°C for 40 seconds, an extension at 72°C for 30 seconds, and a final extension at 72°C for 10 minutes. PCR products were electrophoretically separated on 2% agarose gel in 1× TBE buffer. The optical densities of EPO and EpoR were determined by computer (Quantity One software; Bio-Rad, Hercules, CA). The densitometric values were normalized by β-actin. 

Western Immunoblot Analysis for EPO/EpoR, ERK1/2, and STAT5

Individual retinas from experimental and control rats (4 single retinas from 4 rats selected randomly per group) were isolated and homogenized in ice-cold radioimmune precipitation assay (RIPA) buffer containing 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% sodium dodecyl sulfate (SDS), 1% Nonidet P-40, and 1% sodium deoxycholate, for Western blot analysis. RIPA buffer enables efficient retinal tissue lysis and protein solubilization while avoiding protein degradation and interference with immunoreactivity. This buffer was supplemented with the protease inhibitor PMSF (Shenergy Bicolor Bioscience Technology Company, Shanghai, China). After 15 minutes’ incubation on ice, the extracts were clarified by centrifugation at 12,000g for 15 minutes at 4°C and stored at −70°C. Protein concentrations were determined by protein assay kit (Bio-Rad). Equal amounts of protein were resolved in SDS-polyacrylamide gels and transferred electrophoretically onto a nitrocellulose membrane (Bio-Rad). The membranes were blocked for 30 minutes: for EPO and EpoR detection with 5% nonfat milk; for ERK detection with 1× gelatin; and for STAT5 detection with 5% BSA, respectively. The membranes after blocking were incubated overnight with anti-EpoR antibody (1:500; M-20; Santa Cruz Biotechnology), anti-EPO antibody (1:500; H-162; Santa Cruz Biotechnology), phospho-STAT5 (Tyr694) antibody (1:1000, Cell Signaling), phospho-p44/42 MAP kinase (Thr202/Tyr204) antibody (1:1000; Cell Signaling Technology) or anti-β-actin antibody (1:4000; Sigma-Aldrich). After they were washed with TBST, the membranes were incubated for 1 hour with horseradish peroxidase–conjugated anti-rabbit or anti-mouse antiserum in TBST and 5% nonfat milk. The membranes were washed three times with TBST, and proteins were visualized by enhanced chemiluminescence. After detection with the phospho-specific antibody of STAT5 or ERK, the blot was then stripped and reprobed successively with STAT5 antibody or p44/42 MAP kinase antibody (1:1000; Cell Signaling), and the optical density of each band was determined (Quantity One software; Bio-Rad). The densitometric values for the proteins of interest were normalized for protein loading with β-actin, and the resultant values compared statistically by Student’s t-test. 

Immunohistochemistry for EpoR in Retinal Layers

Rats were killed with deep anesthesia. Slash marks were made on an enucleated eye at the 3- and 9-o’clock positions on the limbus for orientation. For preparation of cryostat sections, the eyes were fixed in PBS-buffered 4% paraformaldehyde for 24 hours and then were opened along the ora serrata, and the posterior eyecups were dehydrated through a gradient concentration of sucrose from 10% to 30%. After dehydration, the eyecups were embedded in optimal cutting temperature (OCT) compound (Tissue Tek; Sakura Finetek, Tokyo, Japan) for sectioning. Serial sections (10 μm) were cut on a cryostat microtome. The sections were thawed, washed twice in PBS for 5 minutes, and incubated with 0.3% H₂O₂ for 30 minutes to block endogenous peroxidase. After the sections were washed with PBS, the sections were incubated with blocking solution (10% normal goat serum in PBS) for 30 minutes at room temperature followed by overnight incubation with polyclonal rabbit anti-EpoR antibody (1:100; H-194:sc-5624; Santa Cruz Biotechnology) diluted in PBS at 4°C. A sample without the primary antibody was used as a negative control. The following day, sections were washed in PBS three times for 5 minutes and incubated with biotinylated goat anti-rabbit IgG (BA1003; Boster) for 20 minutes at room temperature. After washing in PBS, the sections were incubated with avidin-horseradish peroxidase complex (SABC Kit; Boster) for 20 minutes at room temperature. The sections were washed again in PBS, and staining was developed by 3,3′-diaminobenzidine (DAB) without nuclear staining. After reaction was terminated by water and the sections were washed, dehydrated, passed through xylene, and coverslipped. The results were evaluated by light microscopy (LM). 

Experimental Animals and Intravitreal EPO Treatment

Male Sprague-Dawley rats of ∼150 g body weight (BW; Slaccas, SIBS, Shanghai, China) were used. The animals were treated in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. For diabetes induction, a single STZ intraperitoneal injection (60 mg/kg BW in citric buffer) was performed after the rats had been fasted for 24 hours. The control rats received an equal volume of citric buffer. All animals were maintained in a 12-hour alternating light–dark cycle, and allowed to eat and drink ad libitum. Animals receiving STZ were declared diabetic when their blood glucose exceeded 250 mg/dL for three consecutive days. The rats were excluded from the experiment if they failed to develop diabetes. The diabetic rats were randomly divided into two groups: EPO treated and nontreated. Their BW was recorded twice a week, and 4 units of NPH insulin were administered subcutaneously once a week to prevent ketosis. The rats were killed at 1, 2, 4, and 6 weeks after the onset of diabetes. Intravitreal injection of EPO was performed within 2 hours after administration of STZ, with a 30-gauge, 0.5-in. needle (BD Biosciences, Franklin Lakes, NJ) on a microsyringe (Hamilton, Reno, NV), using a temporal approach, 2 mm posterior and parallel to the limbus. EPO, ranging from 0.05 to 200 ng per eye, was dissolved in an equal volume of 2 μL. Sham injections (2 μL normal saline) were performed to both nondiabetic control rats as well as the untreated diabetic rats. The rats recovered spontaneously from the anesthesia and then were sent back to the animal room with food and water ad libitum. 

Examination of BRB Permeability

BRB permeability was evaluated according to the method of Xu et al. ²⁸ with some modifications. Briefly, the rats were anesthetized with intraperitoneal 2% pentobarbital sodium (50 mg/kg BW). The left iliac artery and vein were cannulated with a catheter (Insyte; BD Biosciences). The Evans blue solution (30 mg/mL) was injected through the left iliac vein over 10 seconds. After 2 minutes, 0.1 mL blood was drawn from the left iliac artery. An equal volume of the blood was then drawn at 15-minute intervals up to 2 hours after the injection to obtain the time-averaged Evans blue plasma concentration. After the dye had circulated for 2 hours, the chest cavity was opened, and rats were perfused via left ventricle with 1% paraformaldehyde in 0.05 M citric acid (pH 3.5) at 37°C. Immediately after the perfusion, both eyes were enucleated. The retina was carefully separated and dried at 37°C (Speed-Vac; GMI, Ramsey, MN). The two retinas from the same animal were pooled. After the dry weight was determined, the retinas were incubated in 300 μL formamide for 18 hours at 70°C. The extract was centrifuged through a 30,000 NMWL (nominal molecular weight limit) centrifuge filter (Microcon; Millipore, Bedford, MA) at 3,000g, 4°C, for 45 minutes. The volume of the filtrate was measured. The blood samples were centrifuged at 10,000g and diluted to 1:5,000. The samples (60 μL) were used for triplicate measurements with a spectrophotometer (DU800; Beckman, Fullerton, CA) at a 5-second interval. A background-subtracted absorbance was determined by measuring each sample at both 620 nm, the maximum absorbance for Evans blue, and 740 nm, the wavelength of minimum absorbance. The concentration of the dye in the blood samples and in the retinal extracts was calculated from a standard curve of Evans blue in formamide. The result of Evans blue permeation (EBP) of the retina, as a function of BRB permeability, was calculated with the following equation and expressed in (μL plasma × g retinal dry wt⁻¹ · h⁻¹).  

\[\frac{\mathrm{Evans\ blue\ ({\mu}g)/pooled\ retinal\ dry\ weight\ (g)}}{\mathrm{Time-averaged\ Evans\ blue\ concentration\ ({\mu}g/{\mu}L)\ {\times}\ circulation\ time\ (h)}}\]

 

Sample Preparation for Morphologic Studies

Rats were killed with deep anesthesia. Cutting marks were made on an enucleated eye at the 3- and 9-o’clock positions of the limbus for orientation. For cryostat sectioning, the eyes were fixed in PBS-buffered 4% paraformaldehyde for 24 hours and then were opened along the ora serrata, and the posterior eyecups were dehydrated through a gradient concentration of sucrose from 10% to 30%. After dehydration, the eyecups were embedded in OCT compound (Tissue Tek; Sakura) for sectioning. 

The eyes were oriented by the optic nerve head and pre-enucleation cuts, and sections of the retina were prepared along the nasal-temple plane of the eye. Serial sections that passed through the optic nerve head were analyzed. For measurement of retinal thickness, cell counting, and TUNEL staining, 10 μm-thick sections were used. 

Measurement of the Changes in Retinal Thickness and Cell Counts

The cryosectioned retinas were stained with hematoxylin and eosin (HE). The thickness of the different retinal layers was measured under 200× magnification, including the (1) outer limiting membrane to inner limiting membrane (OLM-ILM); (2) outer limiting membrane to ganglion cell layer (OLM-GCL); (3) outer nuclear layer and outer plexiform layer (ONL-OPL); (4) inner nuclear layer (INL); and (5) inner plexiform layer (IPL). Two measurements were taken on each section, at the two reference lines, which were 1 mm away from the optic nerve on both the nasal and temporal sides. Data from five to six rats (both eyecups) were averaged for comparison, to avoid any potential anatomic variations in different topographic regions. The number of cells in the ONL, INL, and GCL were counted in the same region as the thickness measured, under the 1000× magnification. All the cell nuclei within a fixed 25-μm column, centered with the 1 mm reference lines, were counted. The cell density was then expressed as the cell count per millimeter width of retina in the different layers. 

In Situ Detection of Cell Death in the Retina by TUNEL Assay

The TUNEL assay was performed with a kit according to the manufacturer’s instructions (In Situ Cell Death Detection Kit; Roche China, Ltd.). Positive controls were retinal sections that had been treated with grade I DNase-I for 10 minutes at room temperature, before the labeling procedure. Negative controls were the retinal sections treated with 10 μL label solution but incubated in the absence of terminal transferase. The sections, rinsed three times with PBS after incubation, were analyzed by fluorescence microscope (Nikon, Yokohama, Japan), with an excitation wavelength in the range of 450 to 490 nm. 

Electron Microscopy

The enucleated eyes were dissected along the equators and immersed in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4) for 1 day. The eye balls were then fixed in 1% osmium tetroxide, dehydrated through a graded ethanol series, and embedded in Epon 618 (TAAB Laboratories Equipment, Berks, UK). Ultrathin sections of the posterior region of the retina were stained with lead citrate and uranyl acetate, and then examined by transmission electron microscope (CM-120; Philips, Eindhoven, the Netherlands) at 100 kV. 

Statistics

Data are expressed as the mean ± SE. The statistical analyses were performed by using Student’s t-test. P ≤ 0.05 was considered statistically significant. 

Results

Establishment of the Diabetic Rat Model

After diabetes onset, over a 1- to 4-week period, serum glucose levels in the diabetic group increased 4- to 6-fold compared with the control condition (Fig. 1A) . The BW of the diabetic rats decreased 7.7% after 1 week and then remained relatively constant between 2 and 4 weeks, whereas the BW of the nondiabetic control animals increased 74.3% by 4 weeks (Fig. 1B) . After 4 weeks, the BW of the diabetic rats was 70% of that of the age-matched nondiabetic control group. EPO injected intravitreally (50 ng/eye) had no effect on glucose level and BW of animals in the diabetic group. 

EpoR Upregulation and the Alteration of the ERK Pathway with the Progress of Diabetes

The expression of EPO/EpoR was studied by using RT-PCR, Western blot, and IHC methods. There was no significant change in both mRNA and expression of the EPO protein between normal and 9-week diabetic groups (Figs. 2A 2B) . In contrast, a small but significant increase in EpoR expression at both the mRNA and protein levels was detected in the diabetic retinas with the progress of diabetes (Figs. 2A 2B) . In comparison with the normal control, more neurons in different layers of the diabetic retinas displayed EpoR immunoreactivity (Fig. 2C) . The Western blot result for ERK showed that phospho-ERK/ERK was reduced at 2 weeks after diabetes onset, when compared with the normal control. This ratio returns to the normal level in the EPO-treated groups (Fig. 2D) . 

In the same time frame as the detection of the ERK pathway, there was no significant difference in phospho-STAT5 protein expressed by normal, diabetic, and EPO-treated diabetic groups (data not shown). To corroborate this result, we stimulated the EpoR-positive R28 and RGC5 cell lines ^{29 30} with a range of EPO solutions (0.5–250 U/mL), and then tested by Western blot for phospho-STAT5 in parallel. Upregulation of phospho-STAT5 in the RGC5 cell line, but not R28, was detected after stimulation with a high concentration (250 U/mL) of EPO (data not shown). 

BRB Breakdown in the Early Stage of Diabetes and Protection by Intravitreal EPO

BRB permeability was assessed with Evans blue quantification. ^{28 31} To exclude a possible increase in BRB permeability due to the increase in unbound Evans blue in the acidic condition ³¹ (e.g., under the condition of diabetic ketoacidosis), serum pH was monitored in the diabetic rats, with and without insulin treatment. In comparison with the blood pH in normal control rats (7.44 ± 0.02, n = 5), the pH of EPO-treated diabetic rats, with and without insulin therapy were 7.41 ± 0.10 (n = 6) and 7.43 ± 0.01 (n = 6), respectively. Therefore, the constant serum pH in the diabetic group ruled out the possible influence of pH on the quantification of Evans blue. Also, as shown in Figure 3 , the insulin therapy did not interfere with the protective effect of EPO on vascular permeability at 2 weeks after the onset of diabetes. 

EBP was measured from 1 to 4 weeks after diabetes onset. Figure 4shows that the EBP in STZ-diabetic rats increased 56.8% after 1 week (n = 8, P < 0.05) in comparison with that in the nondiabetic rats (n = 10), and reached a peak after 2 weeks (62.9% increase, n = 7, P < 0.05). EBP remained at elevated levels between 2 and 4 weeks (Fig. 4 , n = 8, P < 0.05). In Figure 4 , in the first 2 weeks, all three dosages of EPO (5, 20, and 50 ng/eye) exhibited similar, effective protection of the BRB. After 3 weeks, the lowest dose, 5 ng/eye, of EPO treatment demonstrated loss of its protective effect on the BRB (n = 4; P > 0.05). At 4 weeks after the onset of diabetes, 50 ng/eye still showed significant protection of the BRB against diabetic insult (n = 8; P < 0.05). When greater dosages of EPO were used (up to 200 ng/eye), a dose dependency of EBP was observed (Fig. 5)at 2 weeks after onset of diabetes. At the highest concentration (200 ng/eye), the protective role of EPO on the integrity of BRB appeared to be less effective, but still maintained EBP significantly lower than that in the untreated diabetic group (P < 0.05). The optimal effect of EPO for BRB protection was obtained at 50 to 200 ng/eye level (Fig. 5) . 

The data indicate that a single intravitreal injection of EPO at the proper dosage can preserve BRB function for at least 4 weeks after the onset of diabetes. In Figure 4 , surprisingly, the EBP in EPO-treated diabetic groups was significantly lower than in the group of nondiabetic control (sham injection group) at the 2-week interval (n = 8; P < 0.05). It is possible that protection against EBP by sham intravitreal injection was also prevented by EPO treatment. 

Protection of the Diabetic Retina by EPO from Reduction in Thickness and Loss of Neurons in the Different Layers

Morphometric examination of HE-stained retinal cryostat sections demonstrated significantly reduced total retinal thickness in the diabetic rats 1 week after diabetes onset in comparison with that in the nondiabetic control rats (Fig. 6) . Primarily, the reduction in thickness occurred in the ONL of the diabetic rats up to 4 weeks (n = 72; P < 0.001). By the end of the fourth week after diabetes induction, there were significantly fewer cells in the ONL (84.61%, n = 52; P < 0.001), and marginal reductions in the number of cells in the INL (98.21%, n = 54; P > 0.05), when the retinal sections of diabetic and nondiabetic control eyes were compared (Fig. 7) . For diabetic rats that were treated with intravitreal injection of EPO (50 ng/eye), the total retinal thickness and number of cells in the ONL were not different from the nondiabetic control for the entire 4 weeks after the onset of diabetes (Figs. 6 7) . There was no significant difference in the thickness of the ONL and the number of cells in the nondiabetic control animals during the observation period (n = 52, P > 0.05). 

Protection of Retinal Neurons from Cell Death by EPO in Diabetes

To characterize the cell death in the GCL, INL and ONL of the diabetic and nondiabetic rats, we used TUNEL analysis. ³² In the diabetic untreated eyes, TUNEL-positive cells significantly increased in the ONL from 1 week and reached a plateau at 4 weeks after induction of diabetes (Fig. 8) . The number of TUNEL-positive cells observed in the GCL and INL of the diabetic rats was less than that in the ONL (Fig. 8) . After intravitreal injection of EPO (50 ng/eye), essentially no TUNEL-positive cells were detected in the GCL, INL, and ONL from 1 to 3 weeks. However, some TUNEL-positive cells were noted in the ONL at 4 (Fig. 8)and 6 (data not shown) weeks in the EPO-treated diabetic eyes, but the number was significantly lower than that in untreated eyes of the diabetic rats (Fig. 8) . 

EM Evidence of a Protective Effect of EPO on Retinal Vascular Cells and Neurons from Diabetic Insult

The representative ultrastructural changes in the rat retinas are illustrated in Figure 9 . The cell and tissue alterations in different retinal layers were detected at 7 to 10 days after the onset of diabetes in comparison with that in the nondiabetic retinas. Prominent changes in the microvascular endothelial cells (MECs) were discerned, including apoptotic nuclear condensations accompanied by paravascular edema and exudates (Fig. 9 , en-D). In the retinal pigment epithelium (RPE), disruption of Bruch’s membrane (BM) and the accumulation of collagen were observed in the diabetic eyes, resulting in a widened space between the BM and RPE (Fig. 9 , rpe-D). The structure of the rod and cone layers was severely damaged with widened intradisc spaces left by the dissolved segments (Fig. 9 , r/c-D). Striking changes in the ONL were the appearance of pyknosis, which correlated with the data showing cell loss and increased TUNEL-positive cells in the ONL, as described earlier (Fig. 9 , onl-D). At the same time point, the diabetic rats treated with EPO showed either no change or only minor changes (Fig. 9 , en-E, rpe-E, r/c-E, and onl-E), in comparison with the nondiabetic control group (Fig. 9 , en-C, rpe-C, r/c-C, and onl-C). 

Discussion

Diabetic retinopathy, by its classic definition, is recognized to be a microangiopathy occurring with diabetes. The microangiopathy is characterized by BRB breakdown, capillary basement membrane thickening, loss of pericytes, and the development of acellular and occluded capillaries. ^{33 34} The STZ-induced diabetic rat model displays morphologic and functional changes in the retinal vasculature similar to those observed in the early stage of human DR. ³⁵ Increased apoptosis of microvascular pericytes and endothelial cells has been well established as an important model of the diabetic retina, in both humans and rats. ^{36 37} The apoptotic processes in retinal vascular cells are not just a late event of diabetes. It has been reported that in the STZ-diabetic rat model, retinal microvascular endothelial cell apoptosis increases significantly 9 to 14 days after the induction of diabetes, compared with that in the nondiabetic retina. ^{38 39}  

Various metabolic and biochemical abnormalities in diabetic retina have prompted researchers to explore the mechanisms and treatment of DR. Although progress has been made in the study of the aldose reductase pathway, ⁴⁰ the diacyglycerol-protein kinase C pathway, ⁴¹ upregulation of the transcription factor NF-κB, ^{42 43} and the increased formation of advanced glycation end products, ⁴² the results of clinical trials based on these theories are still unsatisfactory. In our previous report, upregulation of caspase-3 mRNA (formerly called CPP32) by pericytes of the diabetic retina, with or without clinical retinopathy, indicated that vascular pericytes were at a preapoptotic state in diabetic retina. ³⁶ A recent report demonstrated that the ganglion cells in diabetic retina, with or without retinopathy, also show intense immunoreactivity for apoptosis-promoting molecules, including caspase-3. ² This important finding has strengthened the concept that both neurons and vascular cells are in a proapoptotic state in diabetic retina. Nevertheless, the marked occurrence of accelerated apoptosis in both neuronal and vascular cell populations of retina in early diabetes, ⁴⁴ though the apoptotic process is a final common pathway of various metabolic derangements, appears to be forgotten as a potential therapeutic target. As a matter of fact, in an STZ-diabetic rat model, retinal microvascular endothelial cell and neuronal apoptosis significantly increases compared with the nondiabetic retina. ^{38 39 45} Therefore, the present study was undertaken to explore whether vascular and neuronal apoptosis in diabetic retina can be prevented in the early stage. 

In the present study, we investigated the alterations of BRB function and cellular injury in both the inner and outer BRB from 1 day to 4 weeks after the onset of diabetes. The results demonstrated the breakdown of the BRB as early as 1 week after the onset of diabetes (Fig. 4) . BRB breakdown at such early stage in diabetic rats also has been reported by others. ⁴⁶ Vascular endothelial cell apoptosis was demonstrated by EM in this study as early as 7 to 10 days (Fig. 9 , en-D). The RPE cells, per se, did not show significant injury at this time, but ruptures in the adjacent BM and a widened space between the RPE and BM were observed (Fig. 9 , rpe-D). A functional defect in permeability of the RPE has been detected at 4 weeks by others in STZ diabetic rats. ⁴⁷  

Increasing evidence shows that retinal neuronal disease progresses along with alterations of other cells (such as the glia) very early in diabetes. ^{3 4 45} Some authors suggested that diabetic neuronal disease may occur before microangiopathy. ⁴⁸ In the present study, apoptotic neurons were detected by TUNEL staining, mainly in the ONL, with small a number in the INL and GCL (Fig. 8) , as early as 1 week after the onset of diabetes. The increased TUNEL staining is presumably attributable to apoptosis and is not likely to be due to ischemic necrosis, because at this stage no capillary occlusion or other signs of necrosis were observed. This apoptotic alteration of neurons, particularly photoreceptors, was also corroborated by EM (Fig. 9) . Our findings suggest that both vascular cells and neural cells die by apoptosis in early diabetes. The parallel time course of microangiopathy and neuronopathy (different from diabetic peripheral neuropathy) indicates that a pervasive diabetic insult is the trigger of the apoptotic process in the early stages. Therefore, the early molecular responses of these retinal cells to diabetic insults and whether the clue obtained from the early molecular events can be used to protect retinal cells were the targets of the present study. 

EPO is considered to act in a dual way, as a neuroprotective factor by inducing angiogenesis. ⁴⁹ To gain insight into the possible participation of the EPO/EpoR system in early diabetic retinas, the present study using rats with less than an 8-week history of diabetes, demonstrated the status of the EPO/EpoR system in diabetic retinas when ischemia is not a concern. In the present model, the upregulation of EpoR mRNA and protein in the neurosensory retina and increased immunoexpression of EpoR of neurons in different retinal layers were first reported (Fig. 2) . This fact suggests that in response to the stress of diabetes, other than ischemia, in the early stages retinal neurons send out a signal that protection is needed. Although at the tested period, EPO expression was not changed, it has been known that EpoR expression is a strong indicator of both cell proliferation and apoptosis in a developmental and compensatory lung growth model. ⁵⁰ Therefore, using exogenous EPO was justified to observe any response of deranged retinal cells in terms of function and morphology. The expression difference between EPO and EpoR in the early diabetic retina has inspired us to characterize EPO/EpoR profiles at different stages of DR in future studies. 

In the experimental STZ-diabetic rat model, the intravitreal injection of EPO demonstrated a dose-dependent inhibition of the breakdown of the BRB that followed a U-shaped response (Fig. 5) , characteristic of cytokines. ⁵¹ It is interesting to note that in EPO-treated groups leakage levels determined by Evans blue were even lower than that in the control (sham injection) at the second week (Fig. 4) . This suggests that the injection alone leads to an increase in Evans blue leakage. This trauma-induced leakage recovered spontaneously in the control eyes after a period of 2 weeks. This suggests that the protective effect of EPO against the breakdown of the BRB is related to the known antiinflammatory effect of EPO. ^{52 53} The molecular mechanism by which EPO protects BRB against its elevated permeability in diabetes is currently under investigation. The mechanism is probably via a local interaction between EPO and its receptor-expressing cells rather than systemic factors. 

The present data provide the morphologic evidence of neuron injury that can be used to explain the visual dysfunction in early human diabetes, as reported previously. ^{3 6 7 8} To explore whether EPO can prevent or delay neuronal apoptosis, intravitreal injection of EPO was performed on the day of diabetes onset. Apoptotic neurons in ONL were essentially undetectable in EPO-treated diabetic eyes up to 4 weeks (Fig. 8) . The apoptosis 6-week diabetic rats treated with EPO was significantly reduced but detectable (data not shown). In addition, the thickness of the ONL in EPO treatment groups up to 4 weeks was essentially unchanged (Fig. 6) . A slightly reduced ONL thickness was observed after 6 weeks of diabetes (data not shown). These findings suggest that the therapeutic efficacy of a single intravitreal injection of EPO may persist for at least 4 to 6 weeks, even though the vitreous EPO concentration returned to baseline within 3 days (data not shown). 

The present study is the first report that an intravitreal injection of EPO protects the function and integrity of BRB and prevents the vascular–neuronal cell apoptosis in the early diabetic retina. It has been proposed that the apoptotic process in EpoR expressing cells could be aborted by binding of EPO rapidly. ^{51 54 55} In an in vitro study, apoptosis of neurons was stopped by exposure to EPO, and a 5-minute EPO treatment protected the neurons from apoptosis as effectively as constant exposure of the cells to EPO for 8 hours. ⁵⁶ In the spinal cord injury model, a single dose of intraperitoneal EPO injection immediately after injury produced a recovery that was as effective as that of animals treated with multiple doses of EPO. ⁵⁷ Nevertheless, the mechanism by which a single intravitreal injection of EPO produces an effect for 4 to 6 weeks in diabetic rat eyes is currently unknown since the elimination of half-life from the vitreous of EPO was observed to be 24 to 36 hours (data not shown) and a return to baseline took 72 hours. The mechanism of the cytoprotective effect of EPO appears to be different from that of the erythropoietic function, which requires sustained levels of EPO to stimulate erythroid precursors in the bone marrow. ⁵⁸ It has been proposed that a distinct EPO receptor other than that expressed by erythroid precursors specifically mediates the antiapoptotic tissue protection. ^{51 58} Nevertheless, it is conceivable that EPO takes a cytoprotective effect through an “on–off” mechanism by which the binding between EPO and EpoR(s) turns on cytoprotective signaling pathway(s). Once the pathway(s) is turned on, the sequential protective effect persists for a certain period irrespective of EPO availability in the vitreous. The present data indicate that a significant protective function of EPO can be provided through episodic delivery into the vitreous. This feature confers a therapeutic or a prophylactic function on EPO to treat DR in its earliest stages. It has been reported that in neonatal mice, early administration of EPO prevented initial retinal vessel loss and inhibited subsequent pathologic proliferation, whereas late injection of EPO did not. This phenomenon suggests that the efficacy of EPO intervention lies in the temporal dependency for treating retinopathy (Chen et al. IOVS 2007;48:ARVO E-Abstract 1963). 

The molecular mechanisms and signaling pathways by which EPO exerts its antiapoptotic function in the diabetic retina have not been elucidated. Several well-known pathways that are thought to promote cytoprotection, such as ERK, p38, and c-jun-NH2-kinase (JNK) kinases, Akt, and STATs are being tested to understand the ability of EPO for diabetic retinal cells in our laboratory. ^{59 60 61 62 63 64} In this article, exogenous EPO is clearly able to activate the ERK1/2 pathway. This fact (Fig. 2D)has verified that the protective effect of intravitreal injection of EPO acts through an EPO/EpoR-dependent pathway in the diabetic retina. In contrast, the activation of the STAT5 pathway has not been found in the present animal model. Several possibilities may be attributed to this finding. First, EPO at relatively low concentration (0.5 U/mL) in the vitreous may fail to stimulate the STAT5 pathway in retinal cells. A similar finding was reported by using VSMCs in an in vitro model. ⁶⁵ Second, EPO-induced phosphorylation of STAT5 in retinal cells may be transient but not a persistent fashion. This possibility was demonstrated in other cell lines. ⁶⁶ Third, in the present study, a mixture of retinal cells from the whole neurosensory retina was used to detect a relatively low-expression protein (STAT5) compared with a highly expressed ERK protein (data not shown). The subtle difference in STAT5 expression by individual retinal cell types may be diluted by a larger mixed-cell population in the present study. However, other possible mechanisms of EPO protection in the early diabetic retina, particularly in some individual cell types, will be studied further. 

Since EPO is a cytoprotectant as well as an angiogenic factor, the possible hazardous effect for retinal diseases that involve neovascularization is a genuine concern. ^{26 67} To clarify this possible adverse effect, the EPO/EpoR system including downstream signaling pathways of retinal cells in a neovascular condition merits further study. Particularly, whether EPO/EpoR is still a fully functional system in the neovascular stage—in other words, whether EpoR is dissociated with the EPO/EpoR system—should be the first question to be answered. ⁶⁸  

 Supported by the Ministry of Science and Technology (2004CB720300); Shanghai Science and Technology Committee Grants 04ZR14149, 054107058, and 200483; and the Unrestricted Research Fund of the Clear Vision Foundation, Philadelphia, Pennsylvania.

 Submitted for publication June 15, 2007; revised September 27, 2007; accepted December 11, 2007.

 Disclosure: J. Zhang, None; Y. Wu, None; Y. Jin, None; F. Ji, None; S.H. Sinclair, None; Y. Luo, None; G. Xu, None; L. Lu, None; W. Dai, None; M. Yanoff, None; W. Li, None; G.-T. Xu, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Blood glucose levels (A) and body weights (B) of diabetic rats with and without intravitreal EPO injection at 4 weeks after onset of diabetes. Data are expressed as the mean ± SE with n ≥ 7. (•) Nondiabetic control; (○) diabetic rats without EPO treatment; (▾) diabetic rats treated with a single intravitreal EPO injection (50 ng/eye) on the day of diabetes onset. Sham injection (normal saline) was performed in both control and diabetic rats.

Figure 2.

 

EPO/EpoR system and ERK pathway in the neurosensory retina of diabetic and nondiabetic rats. Samples from STZ-treated rats were collected at 1, 2, 4, and 9 weeks after diabetes onset. (A) Western blot of EPO and EpoR. N: normal control; Dxw: x-week diabetic rats (x = 1, 2, 4, and 9). (B) The mRNA levels of EPO and EpoR of D4w demonstrated by RT-PCR. (C) The IHC study for EpoR in retinas without nuclear staining. Left: EpoR immunostaining for three groups (200× for NC: negative control; N: normal retina; D: diabetic retina of D4w). Right: retinal layers in the diabetic rats showing EpoR-positive cells in the different layers. Magnification: left, ×200; right, ×1000. (D) The Western blot result of both phospho-ERK and total ERK. (Da) No: normal retina without intravitreal injection; N: normal retina; D: diabetic retina at day 7 and 14, respectively; E: EPO (50 ng/eye)-treated retina at 7 and 14 days, respectively. The subscript number shows the day the rats were killed after the onset of diabetes. (Db) The densitometric values of phospho-ERK/ERK in (Da).

Figure 3.

 

Effect of insulin on EBP in the retina in STZ-diabetic rats in the 2-week period after the onset of diabetes. C: nondiabetic control; E+I: diabetic rats treated with both insulin and EPO; E–I: diabetic rats treated with EPO but without insulin injection; D: diabetic rats. The EPO was injected into the vitreous (5 ng/eye) in both EPO-treated groups. Sham injection (normal saline) was performed to both control and diabetes groups. In the E+I group, insulin was given at a low dosage (4 IU per week) by SC injection, so that the blood glucose level could be maintained at a high level. There is no significant difference between the rats treated with insulin and those without insulin treatment (n = 6; P > 0.05).

Figure 4.

 

Time course of EBP of the retina of STZ-diabetic rats treated with and without different doses of EPO. Three dosages of EPO (5, 20, and 50 ng/eye) were tested. Sham injection (normal saline) was performed in both the control and diabetes groups. Each data point is the mean ± SE. n ≥ 7 in each group. *Significantly different from the diabetes group (P < 0.05); ^#not significantly different in comparison with diabetes group (P > 0.05).

Figure 5.

 

Dose-dependent protection by EPO of the BRB of STZ-diabetic rats. Different doses of EPO (0.05–200 ng/eye) were injected intravitreally at the onset of diabetes. Sham injection (normal saline) was performed in both the control and diabetic rats. The rats were killed after 2 weeks of treatment. Each point represents the mean ± SE of results in six rats.

Figure 6.

 

Protecting normal retinal thickness in STZ-diabetic rats by intravitreal injection of EPO. The dose of EPO was 50 ng/eye. Sham injection (normal saline) was performed to both nondiabetic control rats and diabetic rats. Three rats were killed at each week up to 4 weeks; the representative sections were C: nondiabetic control; D: diabetes; and E: EPO-treated diabetic rats. The thickness of the retina was measured on HE-stained samples. Magnification, ×200.

Figure 7.

 

Preventing loss of retinal neurons in STZ-diabetic rats by EPO. The dose of EPO was 50 ng/eye for the EPO group. The diabetes and nondiabetic control received a sham injection (normal saline). Three rats from each group were killed at 4 weeks. The number of retinal neurons in different layers was counted systematically at the same orientation from 12 sections of each rat. Twelve sections from each rat (six from each eye) were examined; two areas in each section were counted, at 1 mm on the nasal and temporal sides of the optic nerve; 25 μm in width was covered in each area for the final count. The data are expressed as the mean ± SE of 72 counts for each group. The cell nuclei were counted after HE staining. C: nondiabetic control; D: diabetes; E: EPO-treated diabetic rats. The cell count in the ONL of the diabetic rat retina was significantly reduced (n = 12, P < 0.001) and resumed completely by EPO treatment (n = 12, P < 0.001). In the INL, the cell count in EPO-treated rats was significantly higher than that in the INL of diabetic rats (n = 12, P < 0.001). Magnification, ×1000.

Figure 8.

 

Representative micrographs of EPO protection of retinal neurons from death detected by TUNEL staining. The diabetic rats were treated with a single intravitreal injection of EPO at the diabetic onset (E) with the dose of 50 ng/eye. Sham injection (normal saline) was performed in both nondiabetic control rats (C) and diabetic rats (D). The subscript numbers indicate the weeks after onset of diabetes, in (Cn) control rats; (Dn) diabetic rats; and (En) EPO-treated diabetic rats.

Figure 9.

 

Representative electron micrographs of different retinal cells of control and diabetic rats that were treated or untreated with intravitreal injection of EPO. C: nondiabetic control; D: diabetic rats; E: diabetic rats treated with EPO (50 ng/eye). The rats were killed 1 week after the onset of diabetes. The images show the changes in retinal vascular endothelial cells (en-C, en-D, and en-E; magnification ×9700), retinal pigment epithelium (rpe-C, rpe-D, and rpe-E; magnification ×5800), rod and cone layer (r/c-C, r/c-D, and r/c-E; magnification, ×9700), and the outer nuclear layer (ONL, onl-C, onl-D, and onl-E; magnification, ×5800). Arrows: The changes resulting from diabetes, including apoptotic vascular endothelial cells, edema/exudates around the vascular endothelial cells (en-D), rupture of BM, and collagen accumulation in the widened space between the RPE and BM (rpe-D), the destroyed structure and intercellular space left by the dissolved discs (r/c-D), and the evident pyknosis in ONL (onl-D). The changes were largely prevented in the EPO-treated rats.

The authors thank Jing Yi and Jie Yang of the Department of Cell Biology, Institute of Basic Medicine (Shanghai Jiao Tong University School of Medicine) for consulting and assisting in the EM related work.