Involvement of Brain-Derived Neurotrophic Factor in Early Retinal Neuropathy of Streptozotocin-Induced Diabetes in Rats: Therapeutic Potential of Brain-Derived Neurotrophic Factor for Dopaminergic Amacrine Cells

All experimental procedures using animals were performed in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and our institutional guidelines for the care and use of laboratory animals. Adult male Wistar rats (Japan SLC, Hamamatsu, Japan), 9 weeks of age (250–300 g), were housed in standard lighting conditions (12-h light and 12-h dark cycle) with food and water ad libitum for at least 7 days before experimentation. STZ (70 mg/kg in 50 mmol/l sodium citrate buffer, pH 4.5; Sigma, St. Louis, MO) was administered intraperitoneally under general anesthesia with diethylether inhalation (16). Nondiabetic animals were injected with an equal volume of citrate buffer. Body weight and blood glucose concentrations were measured before the injection and weekly thereafter. Diabetes was confirmed by assaying the glucose concentration in blood obtained from the tail vein using FreeStyle (Nipro, Osaka, Japan). Rats with glucose levels >250 mg/dl were classified as diabetic. Animals were anesthetized by intraperitoneal injection of an overdose of chloral hydrate and killed by decapitation 4 weeks after the STZ injection, near the end of the light phase. The retinas were processed for further analyses.

Multiple intraocular injections, for a total of five times, were given every 3 days, beginning 2 weeks after the intraperitoneal injection of STZ or citrate buffer, under general anesthesia with diethylether and topical anesthesia with a drop of 2% lidocaine applied to the eyes. BDNF (5 μg) in 5 μl of 0.1% BSA (Fraction V; Sigma) in balanced salt solution (BSS) (Alcon, Fort Worth, TX) was injected into the vitreous space of one eye chosen at random according to a previous report (17). An equal volume of vehicle (0.1% BSA in BSS) was injected into the other eye as a control. Retinas were dissected 2 days after the last intraocular injection. Any animal with lens damage or visible vitreous hemorrhage was excluded from the analyses.

To determine tyrosine hydroxylase (TH) protein levels, retinal tissues were lysed, and their protein samples were subjected to SDS-PAGE followed by Western blotting as previously reported (18). Mouse anti-TH monoclonal antibody (19) (1:10,000 dilution; a gift from Dr. Hatanaka) and peroxidase-conjugated goat anti-mouse IgG antibody (1:10,000 dilution; Jackson ImmunoResearch Laboratories, West Grove, PA) were used as primary and secondary antibodies, respectively. Protein levels of TrkB, a high-affinity receptor for BDNF, in the retina were determined by Western blotting using mouse anti-TrkB monoclonal antibody (20) (1:3,000 dilution). Anti–β-actin (1:5,000 dilution; Chemicon, Temecula, CA) was used as a loading control for Western blotting. The intensity of the bands was quantified by densitometry using NIH Image version 1.60 (National Institutes of Health, Bethesda, MD). Protein concentration was assessed by a BioRad Protein Assay Kit (BioRad Laboratories, Hercules, CA).

Protein levels for BDNF (21), neurotrophin-3 (NT-3) (22), and Thy-1 (23) in the retina were quantified by enzyme-linked immunosorbent assay (ELISA) as previously described. ELISA samples (in duplicate) and standards (in triplicate) were applied to titer plates. The anti-BDNF antibody used in ELISA has no cross-reactivity against other neurotrophins or growth factors and did not detect BDNF protein in samples obtained from BDNF knockout mice (21). A standard curve of recombinant BDNF (1–100 pg/well), NT-3 (1–100 pg/well), or Thy-1 (300–6,000 pg/well) was plotted for each plate. The average value of the sample was normalized against the total protein concentration.

mRNA levels for BDNF, NT-3, and Thy-1 in the retinas were quantified by real-time PCR. Total RNA was extracted from retinas as previously described (11). The quality of the mRNA was evaluated by the ratio of 28S ribosomal RNA (rRNA) to 18S rRNA; samples with a ratio >1.6 were used for real-time PCR (n = 4 animals in each group) (24). Total RNA (50 ng) was amplified using an RT-PCR High Plus kit (Toyobo, Osaka, Japan), with the following sets of specific primers: for BDNF, 5′-TCCCTGGCTGACACTTTTGAG-3′ and 5′-ATTGGGTAGTTCGGCATTGCG-3′ (11); for NT-3, 5′-GATCCAGGCGGATATCTTGA-3′ and 5′-AGCGTCTCTGTTGCCGTAGT-3′; for Thy-1, 5′-CGCTTTATCAAGGTCCTTACTC-3′ and 5′-GCGTTTTGAGATATTTGAAGGT-3′ (25); and for β-actin, 5′-GGCATCCTGACCCTGAAGTA-3′ and 5′-GGGGTGTTGAAGGTCTCAAA-3′ (each experiment performed twice). These combinations of PCR primers gave single PCR products of 465, 182, 344, and 203 bp, respectively. The Thy-1 primers, which span an intron and thus produce different-sized products from genomic DNA (745 bp), were used to check for possible DNA contamination of the samples. Quantitative PCR amplification was carried out on a real-time PCR machine (LightCycler; Roche Diagnostics, Basel, Switzerland), using an initial incubation for 10 min at 60°C and 20 cycles of 1 s at 94°C and 15 s at 60°C, with fluorescence acquisition at 80°C in each cycle. Following the PCR cycles, a melting-curve analysis performed by a standard program showed a single melting peak for each pair of primers. Data were analyzed as previously described (24), and mRNA levels for BDNF, NT-3, and Thy-1 were standardized against β-actin mRNA levels in the same RNA sample. The mean value for the nondiabetic animals was set at 100%.

Tissue preparation, cryosectioning, and immunohistochemistry were performed as previously described (11). Rabbit anti-BDNF antibody (SC-546, 1:200 dilution; Santa Cruz Biotechnology, Santa Cruz, CA) or mouse anti-TH antibody (19) (1:1,000 dilution) was used as primary antibody. The specificities of both antibodies were confirmed by preabsorption with the antigens in the previous studies (11,19). For double immunostaining, rabbit anti-pan Trk antibody (C-14, 1:200 dilution; Santa Cruz Biotechnology) was used. The Alexa Fluor 488 goat anti-rabbit IgG antibody (1:200 dilution; Molecular Probes, Eugene, OR) and Alexa Fluor 546 goat anti-mouse IgG antibody (1:500 dilution; Molecular Probes) were used as secondary antibody. To detect individual apoptotic cells, dual-color staining for transferase-mediated dUTP nick-end labeling (TUNEL) and TH protein was carried out using a kit (In Situ Cell Death Detection Kit, Fluorescein; Roche Applied Science, Mannheim, Germany) according to the manufacturer’s instructions. Cell nuclei were counterstained with Hoechst 33342 (200 nmol/l; Molecular Probes) to visualize the layer structure of the retina. Specimens were observed under an AxioPhoto 2 microscope (Carl Zeiss, Jena, Germany), and photomicrographs were captured at 1,300 × 1,030-pixel resolution using a digital camera (AxioCam; Carl Zeiss) and image processing software (AxioVision; Carl Zeiss). The strength of the excitation light and length of exposure time were kept constant for precise comparisons. Immunohistochemistry was performed using four individual animals and was repeated at least twice.

Whole-mount immunohistochemistry for TH was performed on retinas of both nondiabetic and diabetic rats (n = 4 animals in each group), as previously reported (15), with some modifications. Briefly, fixed retinas were defattened and rehydrated in 0.1 mol/l PBS (26). Specimens were incubated with mouse anti-TH antibody (19) (1:1,000 dilution) in 5% skim milk in PBS containing 0.3% Triton X-100 for 3 days at 4°C with continuous agitation, followed by extensive washes in PBS. After a 1-day incubation with the Alexa Fluor 546 goat anti-mouse IgG antibody (1:500 dilution) at 4°C, retinas were washed in PBS and mounted flat on glass slides with the ganglion cell layer (GCL) oriented upward in 80% glycerol in PBS for observation. Image acquisition was performed as described above. Morphometric analyses were performed according to the methods described by Cellerino et al. (15): type I and type II TH-positive dopaminergic amacrine cells were identified by observing the intensity of their immunoreactivity and the size of their soma. Labeled cells in the sample consisting of five fields of 2.33 mm² for each retina were counted to determine the density of TH-positive cells.

All values are expressed as means ± SD. Statistical significance between groups was analyzed either by Welch’s t test or one-way ANOVA followed by Fisher’s projected least significant difference (PLSD) multiple comparison tests, as indicated. Values of P < 0.05 were taken to be significant.

TH protein level, which has been used as a marker for retinal dopaminergic amacrine cells (15), and actin level, as a control, were quantified by Western blotting (Fig. 1A). Band intensity was determined by densitometry, and the mean value obtained for nondiabetic animals was set as 100% for each protein species. TH protein levels were standardized to actin levels in the same lane because actin levels did not differ between nondiabetic (100 ± 4%) and diabetic animals (103 ± 11%, P = 0.73, Welch’s t test). TH levels obtained in diabetic animals were significantly lower than those of nondiabetic animals (P < 0.01, Welch’s t test, n = 4 animals in each group) (Fig. 1B).

We also quantified Thy-1 levels as an index of the number of RGCs (27) because Thy-1 protein levels had a strong positive correlation with the RGC number (23). Neither Thy-1 protein level nor mRNA level was significantly different between nondiabetic and diabetic animals (247 ± 13 vs. 227 ± 21 ng/mg Thy-1 protein level, P = 0.12; and 100 ± 7 vs. 92 ± 8% relative Thy-1 mRNA level, P = 0.30, in nondiabetic and diabetic animals, respectively, Welch’s t test, n = 4 animals in each group). β-Actin mRNA levels were also indistinguishable (nondiabetic, 100 ± 5%; diabetic animals, 101 ± 6%; P = 0.78, Welch’s t test).

To examine the effects of diabetes on the morphology of dopaminergic amacrine cells and their processes, we performed immunohistochemistry with an anti-TH antibody on radial sections of the rat retina. In both nondiabetic and diabetic rat retinas, TH-positive amacrine neurons were observed in the innermost row of the inner nuclear layer (INL) (Fig. 2A), as reported (15). TH-positive fibers innervating both stratum 1 (s1, arrows) and stratum 3 (s3, arrowheads) of the inner plexiform layer (IPL) could be detected in the nondiabetic rat retina (Fig. 2A, left panel). In the diabetic animals, TH-positive fibers in stratum 1 were thinning down, and those in stratum 3 were hardly visible (Fig. 2A, right panel). No positive signal could be observed in neighboring sections stained without primary antibody (data not shown). When specimens were analyzed by dual-color staining for TUNEL and TH protein, some TH-positive amacrine neurons undergoing apoptosis were identified in diabetic animals (Fig. 2B). In the nondiabetic animals, TUNEL-positive dopaminergic neurons could not be observed (data not shown).

Because dopaminergic amacrine neurons are very sparse, making it difficult to quantify their numbers in radial sections, we determined the density of TH-positive cells in the whole-mounted retina. Cells with a large soma that display intense TH immunoreactivity (type I) (Fig. 3A, arrowheads) and smaller cells with lighter TH immunoreactivity (type II, arrows) were distinguished in the rat retina, as reported (15). The distribution of TH-positive cells in the diabetic rat retinas (Fig. 3B, lower panel) was sparser than in the nondiabetic animals (Fig. 3B, upper panel). Furthermore, signal intensity of TH immunoreactivity in both the soma and cell processes of individual cells was much weaker in the diabetic rat retinas. Statistical analysis revealed a significant difference in the cell density of TH-positive dopaminergic amacrine cells in whole-mounted retinas. The densities of both type I and type II TH-positive neurons were decreased in diabetic animals (P < 0.05, Welch’s t test, n = 4 animals in each group) (Table 1). These results imply that degeneration of dopaminergic amacrine cells or TH downregulation in these cells occurs in the early stage of STZ-induced diabetes.

As shown in Fig. 4, BDNF protein levels in the retinas of diabetic animals were significantly decreased (P < 0.005, Welch’s t test, n = 4 animals in each group). In contrast, rather higher NT-3 protein levels were observed in diabetic animals compared with those in nondiabetic animals (P < 0.05, Welch’s t test, n = 4 animals in each group). BDNF mRNA levels were significantly lower in diabetic compared with nondiabetic animals (P < 0.05, Welch’s t test), whereas NT-3 mRNA levels did not differ significantly (Fig. 4C and D).

BDNF immunoreactivity was seen in the cytoplasm of cells in the GCL and INL and in the process-like structures of the GCL running through the IPL (Fig. 5A), which correspond to RGCs, amacrine cells, and Müller glia, respectively (11). In the diabetic rat retinas, all of these cell types exhibited reduced BDNF immunoreactivity (Fig. 5A, right panel), confirming the results obtained by ELISA and RT-PCR. Protein levels of TrkB, a high-affinity receptor for BDNF, in the retina were not altered by diabetic condition, as determined by Western blotting (P = 0.59, Welch’s t test, n = 4 animals in each group) (Fig. 5B). Trk immunoreactivity (Fig. 5C, green) was seen in the GCL, IPL, and INL and colocalized with TH immunoreactivity (red), even in the diabetic rat retinas (Fig. 5C, arrows indicate double-labeled cells), indicating that TH-positive amacrine neurons express Trk receptor.

To assess the therapeutic efficacy of BDNF against retinal neuropathy, multiple injections of BDNF were given to either eye of both nondiabetic and diabetic rats (n = 5 animals in each group). Overall TH protein levels in the retinas were determined by Western blotting (Fig. 6A) and standardized to actin protein levels. The mean TH protein level for the vehicle-treated nondiabetic rat retinas was set at 100%, and statistical analyses were performed by one-way ANOVA, followed by Fisher’s PLSD multiple comparison tests (Fig. 6B). The intraocular injection procedure itself did not alter retinal TH protein levels in nondiabetic animals when compared with noninjected nondiabetic ones (108 ± 9%, n = 4, P = 0.49, one-way ANOVA). Retinal TH levels in the vehicle-treated diabetic eyes were markedly reduced (P < 0.001 in comparison with the vehicle-treated nondiabetic rat retinas). BDNF treatment efficiently maintained TH levels in the diabetic rat retinas (P < 0.05 in comparison with the vehicle-treated diabetic rat retinas). Furthermore, BDNF application upregulated TH protein levels even in nondiabetic animals. To assess the therapeutic potential of BDNF morphologically, we conducted immunohistochemistry on whole-mounted retinas (n = 4 animals in each group). In the diabetic animals, BDNF treatment significantly increased the cell density of both type I and type II dopaminergic amacrine neurons and TH immunoreactivity in individual cells (Table 2 and Fig. 6C, lower panels). Even in nondiabetic animals, BDNF application increased the cell density of type II dopaminergic neurons and the intensity of TH immunoreactivity in cell processes.

Dysfunction of dopaminergic amacrine cells in diabetic retinas has been suggested by studies examining retinal dopamine content (2), TH activity (3,4), and electroretinogram (9). Using TH immunoreactivity as a marker for these cells (15), we demonstrated, for the first time, the morphological or functional degeneration of dopaminergic amacrine cells during the early stage of STZ-induced diabetes, an animal model of type 1 diabetes. In the retinas of diabetic rats, overall TH protein levels were decreased (Fig. 1), reflecting reductions in the cell density of dopaminergic amacrine cells (Table 1). Alternatively, downregulation of TH expression in individual cells might occur in response to diabetes. Either condition may lead to dopaminergic dysfunction in these cells. Because the intensity of TH immunoreactivity in individual cells was much weaker in the diabetic rat retinas, it cannot be ruled out that the numbers of TH-positive cells were underestimated due to immunoreactivity in some cells that was below the detection threshold of the immunohistochemical assay. We quantified retinal Thy-1 levels, which reflect RGC numbers (23,27), and did not find significant reduction in Thy-1 levels in diabetic rat retinas in the present study. In some studies, degeneration of RGCs in STZ-induced diabetic rats has been reported (5,6). The discrepancy between our results and these (5,6) might be due to the length of the induced diabetic condition, which was 4 weeks in our study but >15 weeks in others. As BDNF is a critical factor for RGC survival, long-term BDNF deficiency under diabetic conditions may lead to subsequent RGC death at a later time point.

There are several possible mechanisms underlying the degeneration of dopaminergic amacrine cells in the diabetic animals. First is severe insulin deprivation, which occurs in STZ-induced diabetes. Insulin is an essential factor for the survival of amacrine cells in vitro (28). In addition, systemically administered insulin accumulated in the vitreous (29) and augmented insulin receptor phosphorylation (30). A second possible underlying mechanism is hyperglycemia. Although a study (31) using the sand rat Psammomys obesus, which develops spontaneous hyperglycemia, did not demonstrate a statistically significant reduction in the number of dopaminergic amacrine cells, an in vitro study (32) has shown that excess glucose impairs insulin-stimulated cell survival and insulin receptor signaling in retinal neurons. A third possibility is dysfunction of Müller cells in the diabetic retina. Diabetic Müller cells showed reduced glutamate-aspartate transporter (GLAST) function (8) and impaired glutamine synthesis (7). These dysfunctions resulted in elevated glutamate levels in the diabetic retinas (7), which might induce excitotoxicity in amacrine cells.

In this study, we demonstrated reduced levels of BDNF in the diabetic retina, which may underlie the degeneration of dopaminergic amacrine cells. A null mutation of the bdnf gene in mice resulted in atrophy of dopaminergic amacrine cells (15). In addition, BDNF was reported to prevent the death of dopaminergic neurons in the substantia nigra (33) and retinal amacrine cells in vitro (13) and in vivo (14). Immunohistochemical analysis showed that BDNF protein levels in both RGCs and Müller cells were reduced in diabetic retinas (Fig. 5). This suggests that diabetes affects multiple cell types simultaneously. Diabetes-related changes in the expression levels of neurotrophins have been reported in peripheral nerves (34) and the brain (35). However, the endogenous neurotrophin levels in diabetic retinas had not been reported until this study, in which both protein and mRNA levels of BDNF were shown to be decreased in STZ-induced diabetic rats (Fig. 4A and C). This animal model of diabetes is associated with hypoinsulinemia. When under dietary restriction, another metabolic state associated with mild hypoinsulinemia, BDNF levels were reported to be upregulated in the brain (36); hence, it is controversial whether hypoinsulinemia itself reduces or increases BDNF levels in the central nervous system. In contradistinction to this reduction in BDNF levels, NT-3 did not show such a reduction. BDNF and NT-3 are subject to different transcriptional regulation (37), but how this process is altered under diabetic conditions is not known.

To confirm the neuropathological implications of the BDNF reduction, we administered BDNF protein into the vitreous cavities of diabetic rats. Our results demonstrate that BDNF injection efficiently rescues dopaminergic amacrine cells from neurodegeneration and counteracts the downregulation of TH level seen under diabetic conditions (Fig. 6 and Table 2). BDNF protects neurons from cell death through TrkB, activating downstream Ras/mitogen-activated protein kinase and phosphatidylinositol 3-kinase/Akt pathways (38), the latter of which is more relevant to neuronal survival (39). Neurotrophins have biological roles other than those involving the survival of dopaminergic neurons; BDNF elicited hypertrophy of dopaminergic amacrine cells in developing animals (15) and upregulated dopaminergic function of cultured ventral mesencephalic neurons (40). Although BDNF and insulin share the same intracellular signaling pathways (18), differences (18,41) and interplay (42,43) in the signaling between BDNF and insulin have been reported. Clearly, elucidation of the precise signaling mechanisms by which BDNF protects dopaminergic amacrine cells in the diabetic retina await further investigation. A previous report (44) demonstrated that TrkB localized to dopaminergic amacrine cells by means of double immunostaining using retinal specimens of various species of animals. Thus, TrkB is a strong candidate for having a role in this process. Another possibility is that BDNF, which has been shown to reduce systemic glucose levels (45), might also act in this fashion, locally in the retina.

Neurotrophins have been tested as candidate therapeutic agents for diabetic neuropathy. However, randomized, double-blind, placebo-controlled studies on the effect of subcutaneous injection of either nerve growth factor (46) or BDNF (47) in patients with diabetes failed to demonstrate a significant beneficial effect on diabetic polyneuropathy. The negative results of these clinical trials suggested that the manner and site of neurotrophin administration are critical. Furthermore, because neurotrophins do not cross the blood-brain barrier (48), it is difficult to ensure their local supply to retinal neurons. Although we injected BDNF directly to the eyeball in the present study, frequent injections of neurotrophic factors into the vitreous space of diabetic patients are not practical. An intravitreous implant that releases BDNF continuously or subretinal transplantation of BDNF-secreting cells might be more viable options. Furthermore, intraocular BDNF gene transfer using a viral vector could be attempted. However, the most practical approach is the induction of endogenous BDNF by external administration of agents with a smaller molecular weight, such as catechol compounds (49). In addition, nighttime illumination might be somewhat beneficial to the diabetic retina, by reducing oxygen consumption of the neural retina (50) and upregulating BDNF expression by light stimulation of the retina (11).

This work was supported by a research grant from the Ministry of Health, Labor and Welfare (15110101), the Japanese Society for the Promotion of Science (a Grant-in-Aid for Creative Scientific Research and Scientific Research [B]).

We are grateful to Dr. Shinichi Koizumi, Novartis Pharmaceuticals, for providing us with anti-TrkB antibody. We would like to thank Drs. Naoko Inamura, Ken Mizuno, Yasuhiro Sakai, and Noburo Ando for helpful discussion.