Aldose Reductase Deficiency Prevents Diabetes-Induced Blood-Retinal Barrier Breakdown, Apoptosis, and Glial Reactivation in the Retina of db/db Mice

The AR null mutant mice (17) in C57BL/6J genetic background were mated with C57BL/KsJ db/m mice (The Jackson Laboratories, Bar Harbor, ME). The AR^−/− db/db offspring were backcrossed with C57BL/KsJ db/m for five generations. Then, sibling matings generated AR^+/+ db/m, AR^+/+ db/db, AR^−/− db/m, and AR^−/− db/db mice. The mice were killed by cervical dislocation at 15 months old according the approved protocol of The University of Hong Kong Committee on the use of animals for teaching and research. The eyes were immediately enucleated and fixed in 4% paraformaldehyde overnight at 4°C. They were then dehydrated with a graded series of ethanol and xylene and embedded in paraffin wax. Serial sections 7 μm thick were prepared for immunohistochemistry.

Paraffin sections of the retinas were deparaffinized in xylene and rehydrated with a graded series of ethanol. After washing, sections were blocked with 0.3% hydrogen peroxide for 15 min. The sections were then blocked for 1 h with 1.5% normal goat serum—for subsequent application of PECAM-1, GFAP, AR, VEGF, nitrotyrosine, S-100, proliferating cell nuclear antigen (PCNA), and cleaved caspase-3 antibodies—or Mouse on Mouse mouse Ig blocking reagent (Vector Laboratories, Burlingame, CA)—for subsequent application of α-smooth muscle actin (SMA) and poly(ADP-ribose) (PAR) antibodies. They were then incubated with diluted primary antibodies overnight at 4°C. The primary antibodies and their concentrations were rat anti-PECAM (1:400; PharMingen, San Diego, CA), rabbit anti-GFAP (1:2000; Dako, Carpinteria, CA), rabbit anti-AR (1:1,500), rabbit anti-VEGF (1:200; Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti-nitrotyrosine (1:200 Santa Cruz Biotechnology), rabbit anti–S-100 (1:500; Dako), rabbit anti-PCNA (1:300; Santa Cruz Biotechnology), rabbit anti–cleaved caspase-3 (1:200; Cell Signaling, St. Louis, MO), mouse anti–α-SMA (1:200; Sigma Aldrich, St. Louis, MO), and mouse anti-PAR (1:200; Alexis, Lausen, Switzerland). Immunoreactivity was detected with biotinated goat anti-rabbit, goat anti-rat, or Mouse On Mouse biotinated anti-mouse IgG secondary antibodies and the avidin-biotin-peroxidase complex (Vector Laboratories). An immunoreactive signal was developed, using diaminobenzidine as a substrate (Zymed Laboratories, San Francisco, CA) for 2 min. Photomicrographs were taken with an Olympus IX71 microscope system. All histological and immunohistochemical samples were coded and examined and graded in a blinded fashion.

Sections were prepared as described above and blocked with 2% normal goat serum for 1 h. They were then incubated with Mouse On Mouse biotinated anti-mouse IgG secondary antibody (Vector Laboratories) and avidin-biotin-peroxidase complex. The positive immunoreactivity of IgG was detected by exposing the sections to the substrate, diaminobenzidine, for 2 min.

Retinal sections stained with mouse IgG were used for blood vessel counting. Two digital photographs at ×20 magnification were taken on each side of the retina, starting from the middle. The number of blood vessels in the inner nuclear layer, where most of the capillaries were found, was counted, and the vessel density was calculated by the number of blood vessels per square pixel (25). One capillary was defined by an independent IgG staining of either a tube or dot shape. For counting pericyte density, retinal sections stained with α-SMA were used (26). Two digital photographs at ×20 magnification were taken on each side of the retina, starting from the middle. The number of pericytes in the inner nuclear layer was counted. The pericyte density was calculated by the number of pericytes per square pixel (27). One pericyte was defined by an independent α-SMA staining of a single cell. α-SMA staining of an incomplete single cell, because of the difference on the plane of focus, was also counted as one cell.

Data are means ± SE. Statistical analyses were performed using one-way ANOVA (Tukey’s multiple comparison tests) using GraphPad Prism software (San Diego, CA). In all cases, P < 0.05 was considered statistically significant.

Mice of different genotypes were generated as described in the research design and methods section. Before killing the animals at 15 months old, the body weight and blood glucose level for each animal were measured. The body weights of the AR^+/+ db/db (39.28 ± 0.54 g) and AR^−/− db/db (31.15 ± 1.5 g) mice were significantly heavier than that of the AR^+/+ db/m (27.52 ± 0.54 g) mice (n = 4 in all three groups, P < 0.01 for both comparisons). Blood glucose levels were also elevated significantly in both AR^+/+ db/db (28.08 ± 1.25 mmol/l) and AR^−/− db/db (26.52 ± 2.25 mmol/l) mice compared with that of AR^+/+ db/m (9.48 ± 1.09 mmol/l) mice (n = 4 in all three groups, P < 0.001 for both comparisons).

To understand the role of AR in diabetic retinopathy in mice, the locations of its gene expression in the retinas of the 15-month-old mice were determined (Fig. 1). In the retinas of the AR^+/+ db/m mice, AR immunoreactivity was present in astrocytes, Müller cells, retinal ganglion cells, and the neurons in the inner nuclear layer (Fig. 1A). In general, the intensity of AR immunoreactivity per cell and the number of AR-immunoreactive cells in all areas of the retina were increased in the AR^+/+ db/db mice compared with the AR^+/+ db/m mice (Fig. 1B). As expected, the retinas from the AR^−/− db/db mice did not show any AR immunoreactivity (Fig. 1C and F). In the blood vessels of the retina, AR was mainly present in the pericytes but not in the endothelial cells (Fig. 1D and E), and the expression level of AR in these cells was higher in the db/db mice than the db/m mice.

To assess pericyte loss, retinal sections were stained with antibody against α-SMA, which specifically marks pericytes in the capillaries. The number of α-SMA–positive cells in the inner nuclear layer of the retinas was counted and normalized against the area of the inner nuclear layer (Fig. 2A). The retinas of AR^+/+ db/db mice showed a >25% decrease in pericyte density compared with that of the AR^+/+ db/m mice (Fig. 2B). There was no decrease in pericyte density in the retinas of AR^−/− db/db mice.

In the retinas of the AR^+/+ db/db mice, formation of nitrotyrosine, an oxidative-nitrosative stress marker (28), was increased in retinal ganglion cells and in the neurons in inner and outer nuclear layers as well as in Müller cells (Fig. 3A2, indicated by arrows) compared with that of the AR^+/+ db/m mice (Fig. 3A1). The Müller cells and their processes seem to accumulate much less nitrotyrosine. In the retinas of the AR^−/− db/db mice, there was even less nitrotyrosine accumulation in these cell types (Fig. 3A3). Perhaps as a consequence of increased nitrosative stress that might lead to DNA damage, the level of PAR, the product of PAR polymerase (PARP) that is activated by DNA strand breaks (29), was increased in the retinal ganglion cells and neurons in the inner nuclear layer in the retinas of the AR^+/+ db/db mice (indicated by arrows) (Fig. 3B1 and B2). The retinas of the AR^−/− db/db mice showed no increased level of PAR (Fig. 3B3).

Immunostaining of retinal sections showed that the expression of PECAM-1 was localized in the vascular endothelial cells (Fig. 4A1–A3) and not in the pericytes surrounding the endothelial cells. The immunoreactivity of PECAM-1 was decreased in the retinas of the AR^+/+ db/db mice (Fig. 4A2) compared with those of the AR^+/+ db/m mice (Fig. 4A1). No decrease was observed in the retinas of AR^−/− db/db mice (Fig. 4A3).

Immunostaining of VEGF was found in the astrocytes, Müller cells, retinal ganglion cells, and neurons in the inner nuclear layer (Fig. 4B1–3). Only the endfeet of Müller cells around the capillaries in inner nuclear layer, but not the radial processes in both plexiform layers of Müller cells, were stained with VEGF (Fig. 4B2, indicated by arrows). The expression of VEGF was increased in the retinas of the AR^+/+ db/db mice (Fig. 4B2) but not in the AR^−/− db/db mice (Fig. 4B3), indicating that AR deficiency suppressed hyperglycemia-induced induction of this gene.

To determine the diabetes-induced glial response in the retinas, expression of GFAP, a marker for astrocytes (14) and reactive Müller cells after eye injury (30), was determined. In the retinas of AR^+/+ db/m mice, GFAP expression was observed in astrocytes in the inner limiting membrane and Müller cells with their cell bodies residing in the inner nuclear layer (Fig. 5A1). In the AR^+/+ db/db mice, more processes of the GFAP-stained Müller cells were observed in both inner and outer plexiform layers (indicated by arrows). Their locations confirmed that the processes are from Müller cells (Fig. 5A2). In addition, more GFAP immunoreactivity in the retinas of AR^+/+ db/db mice was observed in the inner limiting membrane (indicated by arrows), where the astrocytes reside, compared with that of AR^+/+ db/m mice (Fig. 5A1). The induction of GFAP expression in Müller cells and astrocytes in the retinas of AR^−/− db/db mice was not as high as in those of AR^+/+ db/db mice (Fig. 5A3).

To identify cells that are induced to proliferate by hyperglycemia, retinal sections were stained with antibody against PCNA, a marker for proliferating cells. The PCNA-positive cells were found mainly in the inner nuclear layer (Fig. 5B1), and they were shown to be Müller cells by staining the adjacent sections with S-100 antibody, a marker for Müller cells (Fig. 5B2). The number of PCNA-positive cells and the intensity of PCNA immunoreactivity was increased in the retinal ganglion cells and neurons in the inner nuclear layer, but not in the outer nuclear layer in the AR^+/+ db/db mice (Fig. 5C2), compared with those of the AR^+/+ db/m mice (Fig. 5C1). There was less PCNA immunoreactivity in the Müller cells of the AR^−/− db/db mice (Fig. 5C3).

To detect blood-retinal barrier breakdown, mouse IgG extravasation in the retinal sections of the mice was examined, and the results are shown in Fig. 6. The majority of IgG-positive blood vessels were present in the inner nuclear layer. Staining of IgG detected outside the lumen of vessels suggests leakage as a consequence of blood-retinal barrier breakdown (31). In the AR^+/+ db/m mice, IgG staining was located within the lumen of blood vessels and showed no signs of leakage out of the blood vessels (Fig. 6A). In the AR^+/+ db/db mice, a significant amount of IgG was found outside the lumen of the capillaries (Fig. 6B), whereas no extravasation of IgG was observed in the retinas of the AR^−/− db/db mice (Fig. 6C), suggesting that AR deficiency protects against hyperglycemia-induced blood-retinal barrier breakdown.

Apoptotic stress, as indicated by immunohistochemical staining of cleaved caspase-3, was found in the retinal ganglion cells, inner plexiform layer, neurons in the inner nuclear layer, and Müller cells (Fig. 7A, indicated by arrows). Both the immunoreactivity of cleaved caspase-3 per cell and the number of cleaved caspase-3–positive cells in retinal ganglion cells, inner plexiform layer, neurons in the inner nuclear layer, and Müller cells were increased in the retina of the AR^+/+ db/db mice (Fig. 7B) compared with AR^+/+ db/m mice (Fig. 7A). There were fewer cleaved caspase-3–positive cells in the AR^−/− db/db mice compared with AR^+/+ db/db mice (Fig. 7C).

Retinal capillaries, as identified by anti-IgG staining, were primarily located in the inner nuclear layer of AR^+/+ db/m retinas (Fig. 8A). Blood vessel density in this region of the retina was determined by counting the number of IgG-positive capillaries divided by the area of inner nuclear layer (Fig. 8B). There was a 1.5-fold increase of capillary density in the retinas of the AR^+/+ db/db mice compared with the AR^+/+ db/m mice (P < 0.01) (Fig. 8B). No significant increase of capillary density was observed in the retinas of the AR^−/− db/db mice.

A recent report showed that mice diabetic for 14 weeks from streptozotocin injection exhibited neuronal cell death in the retinal ganglion cell layer and shrinkage of the retinas (32). No abnormal vascular change was observed in these mice, presumably because of the short duration of the experiment. It is difficult to maintain the streptozotocin-treated mice for prolonged period. The 6-month-old db/db mice, which develop type 2 diabetes when they are ∼8 weeks old, have been shown to exhibit early features of diabetic retinopathy, such as pericyte and endothelial cell loss (23), basement membrane thickening (22), and increased blood flow (24). In this report, we demonstrate that the retinas of the 15-month-old db/db mice showed pericyte loss, breakdown of blood-retinal barrier, apoptosis of neuronal cells, glial reactivation, and even a feature of more advanced diabetic retinopathy: the proliferation of retinal capillaries. All of these hyperglycemia-induced changes were attenuated by AR null mutation, indicating that AR plays an important role in the pathogenesis of this disease. However, the mechanism is still unclear.

Increased flux of glucose through the polyol pathway is thought to increase osmotic and oxidative stress, activate protein kinase C, and increase advanced glycation end products (33). Depending on the cell type, one or more of these mechanisms may contribute to hyperglycemia-induced lesions. In the mouse retina, AR is expressed in the astrocytes, Müller cells, retinal ganglion cells, neurons in the inner nuclear layer, and capillary pericytes. In the diabetic mice, AR expression is increased in all of these cells, further exacerbating the toxic effect of hyperglycemia.

Pericyte loss is a hallmark of early diabetic retinopathy in human and animal models. Pericytes cultured in high glucose medium were shown to have decreased glutathione level and activation of caspase-3, leading to apoptosis (34). Treatment with an AR inhibitor or caspase-3 inhibitor was able to prevent such changes, suggesting that AR activity causes oxidative stress that activates the caspase-dependent apoptotic pathway. The pericytes wrap themselves around the vascular wall, making multiple contacts with the endothelial cells. These two types of cells communicate with each other through gap junctions. It is likely that pericyte loss would lead to endothelial cell death. This is supported by the observation that there was endothelial cell loss in mice with a null mutation in the platelet-derived growth factor gene, which is expressed in pericytes but not in endothelial cells, suggesting a close communication between these two types of cells (27). Therefore, although the endothelial cells have very low levels of AR, they are still sensitive to polyol pathway–mediated glucose toxicity.

The increased number of retinal vessels leaking IgG in the AR^+/+ db/db mice indicates breakdown of the blood-retinal barrier, a common sign of diabetic retinopathy in humans (35) and rats (36). AR deficiency prevented this blood-retinal barrier breakdown, indicating that AR contributes to the maintenance of the blood-retinal barrier. One possible mechanism causing blood-retinal barrier breakdown in diabetic mice is AR-mediated loss of pericytes and endothelial cells. We found that the expression of PECAM-1, which is primarily located in the endothelial cells, was decreased in the AR^+/+ db/db mice and normal in the AR^−/− db/db mice, suggesting that AR may inhibit its expression via its toxic effect on the pericytes. Deficiency in PECAM-1 has been shown to delay the establishment of tight junctions in the endothelial cells to restore the blood-brain barrier after inflammatory insult (11), suggesting that AR may also cause blood-retinal barrier breakdown via the downregulation of PECAM-1.

Increased oxidative and nitrosative stress has been suggested to cause blood-retinal barrier breakdown in diabetic rats (37). Treatment with a peroxynitrite scavenger reduced both oxidative and nitrosative stress and attenuated blood-retinal barrier breakdown. This could be mediated by VEGF because increased VEGF has been shown to increase blood-retinal barrier permeability (38), and reducing oxidative stress normalized VEGF level and prevented blood-retinal barrier breakdown (37). In the db/db mice, increased nitrosative stress in their retinas, as indicated by an increased level of nitrotyrosine immunoreactivity, was reduced by AR deficiency, suggesting that AR activity could also contribute to blood-retinal barrier breakdown by increasing oxidative and nitrosative stress.

Another indication of increased oxidative or nitrosative stress is an increased level of PAR. Increased reactive free radicals would cause DNA damage, leading to the activation of PARP to synthesize more PAR. AR inhibitor has been shown to attenuate the diabetes-induced increase in nitrotyrosine and PAR (28). This is confirmed by our observation that AR^−/− db/db mice accumulated less nitrotyrosine and PAR than AR^+/+ db/db mice. Increased oxidative stress and activation of PARP may have led to retinal degeneration in AR^+/+ db/db mice. We found that there were also more cleaved caspase-3–positive cells in the retinal ganglion cell and inner nuclear layer of their retinas, indicating that they underwent apoptosis (39). These were correlated with increased PCNA expression, which was suggested to play a role in selective postischemic survival, in these cells (40). Photoreceptor cells in the outer nuclear layer of the retinas with very little expression of nitrotyrosine and PAR were also negative in cleaved caspase-3 immunostaining.

There were indications that the astrocytes and Müller cells were experiencing stress because the expression of GFAP, which is induced after retinal damage (30), was increased. Signs of increased apoptotic cells in the Müller cells in 15-month-old AR^+/+ db/db mice was also shown by increased cleaved caspase-3 immunoreactivity. Furthermore, the level of PCNA, a marker for cell proliferation (41), was increased in the Müller cells. The expression of PCNA in the Müller cells was confirmed by the colocalization of PCNA and S-100, a Müller cell marker (42). The induction of GFAP expression and proliferation of glial cells, signs that the retina was undergoing degeneration (43), were not observed in the AR^−/− db/db mice, indicating that these hyperglycemia-induced pathological changes were mediated by the polyol pathway.

The retinas of the 15-month-old db/db mice had significantly more IgG-stained blood vessels than the nondiabetic db/m mice, suggesting the formation of new blood vessels. This represents a more advanced stage of diabetic retinopathy than previously reported for 14-week-old diabetic mice (32) or 26-month-old galactosemic mice (44). Because of the ease of genetic manipulation and availability of a large number of mutants, mice serve as a convenient animal model to study the pathogenesis of various diseases. Demonstration that they do develop early and more advanced features of diabetic retinopathy should make them an attractive model to study this disease. The key features of diabetic retinopathy in db/db mice, such as pericyte and endothelial cell loss, blood-retinal barrier breakdown, neuronal degeneration, glial reactivation, and neovascularization, were all attenuated by AR deficiency, indicating that aldose reductase plays an important role in the pathogenesis of this disease.

The work was supported by Research Grants Council of Hong Kong Grant HKU 7294/00M and HKU 7313/04M (to S.K.C.).