Aldose Reductase–Deficient Mice Are Protected From Delayed Motor Nerve Conduction Velocity, Increased c-Jun NH₂-Terminal Kinase Activation, Depletion of Reduced Glutathione, Increased Superoxide Accumulation, and DNA Damage

AR^+/+ and AR^−/− mice generated previously (13) and backcrossed to C57BL/6N for more than five generations were used. Animal experiments were carried out under guideline set forth by the committee on the use of live animals in teaching and research at the University of Hong Kong.

Mice (6 weeks old, 20–25 g) were induced to become diabetic with streptozotocin (in 0.1 mol/l citrate buffer, pH 4.5; Sigma, St. Louis, MO) by injection (200 mg/kg body wt i.p.). Control animals were injected with citrate buffer. Blood glucose was monitored with a glucose meter (Bayer, Leverkusen, Germany) 2 days later. Mice with a blood glucose level >25 mmol/l were considered diabetic, and those with blood glucose level <8 mmol/l were considered nondiabetic.

The 4, 8, and 12 weeks’ diabetic AR^−/− mice and their appropriate control animals were anesthetized with Hypnorm (Janssen, Oxford, U.K.)/Hypnorval (Hoffmann-La Roche, Nutley, NJ) at a concentration of 0.1 ml per 10 g body wt (the resulting mixture of Hypnorm/Dormicum contained 1.25 mg/ml midazolam, 2.5 mg/ml fluanisone, and 0.079 mg/ml fentanyl). Body temperature was maintained automatically at a mean rectal temperature of 37.5–37.9°C by the use of a heat pad (Fine Science Tools, Vancouver, BC, Canada). The sciatic nerve was stimulated (5–10 V, 0.05 ms single square-wave pulses) proximally at the level of the sciatic notch and distally at the level of the ankle with platinum needle electrodes (Grass, Quincy, MA). Compound muscle action potentials were recorded from the ipsilateral foot between digits 1 and 2, amplified, stored, and displayed on a computer (Spike 2; CED, Cambridge, U.K.). The first compound action potential from individual stimulation was used for the measurement of motor latency, whereas the second one was used for the measurement of sensory latency. Averaged distal and proximal motor and sensory latencies from 10 separate recordings, together with the nerve length between the two stimulation sites, were used for calculation of MNCV and sensory nerve conduction velocity (SNCV).

The sciatic nerves of 4 weeks’ diabetic AR^−/− mice and their appropriate control animals were dissected and frozen in liquid nitrogen until processed. Dry weight of the sciatic nerves was determined after drying the nerves in a SpeedVac (AS160; Savant Instruments, Farmingdale, NY) until the weight was constant. Carbohydrates were extracted from sciatic nerves of mice as described previously (22). Carbohydrate extracts were separated on a DX500 high-performance liquid chromatography system (Dionex, Sunnyvale, CA) equipped with a Carbopac (MA-1) anion exchange column (Dionex) with isocratic elution by 420 mmol/l NaOH. Peak integration was performed, using Peaknet software (Dionex). The peak areas were normalized with the internal control, and carbohydrates were quantified against a known standard curve run in the same run. The concentrations of carbohydrates were expressed in nanomoles per milligram dry weight of sciatic nerves.

Sciatic nerves were stored in liquid nitrogen until homogenization in 1 ml ice-cold 6% perchloric acid. After centrifugation at 4,000g for 10 min, nerve perchloric extracts were neutralized by 5 mol/l potassium carbonate and centrifuged at 4,000g for 5 min. The supernatant was then mixed with 0.89 ml of 20 mmol/l EDTA in 1 mol/l Tris-Cl buffer (pH 8.1). The reaction was initiated by the addition of 0.01 ml of 0-pathaldialdehyde (10 mg in 1 ml methanol; Sigma). Reaction products were detected/quantified by fluorescence spectroscopy at λ-excitation 345 nm and λ-emission 425 nm (23).

Dihydroethidium (DHE; 2 μmol/l; Molecular Probes, Eugene, OR) was topically applied onto 6-μm cryosections of tissue-freezing medium embedded sciatic nerves, incubated in a humidified chamber at 37°C (30 min, in dark). Sections were counterstained with 4′,6-diamidino-2-phenylindole dihydrochloride hydrate (DAPI; Sigma). Signals were captured with an Olympus IX71 microscope equipped with a Spot RT digital camera (Diagnostic Instruments, Detroit, MI).

The 10-μm longitudinal paraffin-embedded sections of 4% paraformaldehyde-fixed sciatic nerves from 12 weeks’ diabetic AR^−/− mice and their appropriate control mice were deparaffinized in xylene and pretreated with 0.3% H₂O₂ to eliminate endogenous peroxidase activity. Mouse monoclonal anti–poly(ADP-ribose) (PAR) antibody (Alexis, San Diego, CA) was applied onto the sections at a concentration of 1:200 overnight at 4°C. PAR immunoreactivity was detected with a biotinylated anti-mouse secondary antibody and streptavidin-biotin-peroxidase complex according to the instructions from Vector Elite kit (Vector Laboratories, Burlingame, CA). Diaminobenzidine was used as substrate for peroxidase. Sections were then counterstained with hematoxylin, dehydrated, and mounted. Images were captured with an Olympus IX71 microscope equipped with a Spot RT digital camera (Diagnostic Instruments).

Sciatic nerves stored in liquid nitrogen were homogenized in ice-cold 0.1 mol/l NaCl, 0.05 mol/l Tris-Cl (pH 7.4), 0.001 mol/l EDTA, and a mixture of protease inhibitors (5 μl/ml Complete; Roche, Basel, Switzerland). SDS-PAGE (8, 10, and 15% acrylamide) was performed with 50 μg of total proteins. Separated proteins were transferred to a reinforced nitrocellulose membrane (Schleicher & Schuell, Dassel, Germany) using a “tank” buffer system (Bio-Rad, Hercules, CA). Primary antibodies used were antibodies against total p46 c-Jun NH₂-terminal kinase (JNK; 1:1,000; Santa Cruz Biotechnology, Santa Cruz, CA), antibodies against phosphorylated p54 JNK and phosphorylated p46 JNK (1:1,000, phosphorylated on Thr-183 and Tyr-185; Cell Signaling Technology, Danvers, MA), antibodies against total extracellular signal–regulated kinase-1 (ERK1) and ERK2 (1:1,000; Cell Signaling Technology), and antibodies against phosphorylated ERK1 and ERK2 (1:1,000, phosphorylated on Tyr-204; Santa Cruz Biotechnology). Detection was achieved using enhanced chemiluminescence Western blotting detection reagents (Amersham Biosciences, Buckinghamshire, U.K.). The immunoblots were scanned on a flatbed scanner (UMAX, Dallas, TX) and converted into numerical values by ImageQuant (version 5.1; Amersham Biosciences, Sunnyvale, CA).

AR^+/+ and AR^−/− mice (13) were mated with thy1-YFP mice (24,25) to introduce yellowish-green fluorescent protein (YFP) into sensory/motor neurons and their processes for analyzing the effect of AR deficiency on the organization of peripheral nerves in live animals. The images of fluorescent YFP nerve fibers in the ear, abdomen, diaphragm, and lower leg of these mice under normal conditions were captured with a digital camera (DC500; Leica, Bensheim, Germany) that was attached to a fluorescence stereomicroscope (MZ FLIII; Leica).

Sural nerves were fixed in 2.5% glutaraldehyde and postfixed in 1% osmium tetroxide. After dehydration, midportions of the fixed sural nerves were embedded in epon and polymerized, and 1-μm-thick transverse nerve sections were stained with toluidine blue. Fascicular area and myelinated fiber number and size were measured at a magnification of 1,600× using a computer-assisted image analyzing system (National Institutes of Health image, Agfa Arcusscanner connected to a Quadra 700; Macintosh, Cupertino, CA).

All data are the means ± SE. Either Student’s t test or one-way ANOVA together with Bonferroni’s post hoc test were performed. P < 0.05 was considered statistically significant.

The AR^−/− mice backcrossed to C57BL/6N for at least five generations had mild impairment in water reabsorption in the kidney, leading to slightly increased urine output and increased water consumption, as described previously (13). Under diabetic conditions, there was no significant difference in the urine output, water consumption, and level of blood glucose between AR^−/− and AR^+/+ mice (data not shown).

There was no significant difference in MNCV and SNCV between nondiabetic AR^+/+ and AR^−/− mice. However, after 4, 8, and 12 weeks of diabetes, AR^+/+ mice showed a significant reduction of MNCV compared with their nondiabetic counterparts (P < 0.001 for each stage). However, a significant reduction of MNCV in diabetic AR^−/− mice was not observed (Fig. 1). Similarly, the reduction of SNCV observed in diabetic AR^+/+ mice (P < 0.001 for each stage) was also not observed in diabetic AR^−/− mice (Fig. 1). Not surprisingly, the ARI fidarestat also protected diabetic AR^+/+ mice from a reduction of MNCV almost equivalent to that of AR-deficient mutants after 12 weeks of diabetes (Fig. 1).

Nerve glucose levels in nondiabetic AR^+/+ and AR^−/− mice were indistinguishable (Table 1). After 4 weeks of diabetes, both diabetic AR^+/+ and AR^−/− mice showed significant increases in glucose content compared with their nondiabetic counterparts (P < 0.001). Glucose content in diabetic AR^+/+ mice was not significantly different from that of diabetic AR^−/− mice (Table 1).

Under normal conditions, sorbitol and fructose contents in the sciatic nerves of AR^−/− mice were significantly reduced when compared with that of AR^+/+ mice (P < 0.001). After 4 weeks of diabetes, both AR^+/+ and AR^−/− mice showed significant increases in sorbitol and fructose in their sciatic nerves compared with those of the nondiabetic control animals (P < 0.001). Sorbitol and fructose levels in diabetic AR^−/− mice were significantly lower than that in diabetic AR^+/+ mice (P < 0.001). The myo-inositol levels between nondiabetic AR^+/+ and AR^−/− mice were similar. The myo-inositol levels in diabetic AR^+/+ and AR^−/− mice were not statistically different from their nondiabetic counterparts (Table 1).

Glutathione levels in nondiabetic AR^+/+ and AR^−/− mice were similar. A significant reduction of glutathione in the 4 weeks’ diabetic AR^+/+ mice was observed (P < 0.001), but no reduction in glutathione level was found in diabetic AR^−/− mice. The sciatic nerves of 12 weeks’ diabetic AR^−/− mice or ARI-treated diabetic mice also showed similar protection from the depletion of reduced glutathione in the sciatic nerves (Fig. 2). Also, we found a significant increase in the urine 8-OHdG (8-hydroxy-2′-deoxyguanosine) content in 8 weeks’ diabetic AR^+/+ mice, but the increase was not statistically significant in diabetic AR^−/− mice (data not shown).

The oxidative fluorescent dye DHE, represented as red fluorescence in the nucleus of the cells, was used to determine the level of O₂^· in the sciatic nerve. Under normal conditions, ethidium fluorescence was barely detectable in the nuclei of cells from AR^+/+ and AR^−/− mice (∼6% of nuclei were positive for DHE). Sciatic nerves of 12 weeks’ diabetic AR^+/+ mice showed a marked increase in DHE staining compared with that of nondiabetic AR^+/+ mice. The percentage of DHE-stained nuclei was also significantly increased (P < 0.001). However, the 12 weeks’ diabetic AR^+/+ mice treated with ARI or fidarestat or the diabetic AR^−/− animals showed much less accumulation of superoxide in the sciatic nerve (Fig. 3A). Quantitation results showed that the increase of DHE-positive nuclei as observed in diabetic AR^+/+ mice was ameliorated in ARI-treated diabetic AR^+/+ mice or diabetic AR^−/− mice (P < 0.001) (Fig. 3B).

To further determine other indicators of oxidative stress, PAR polymerase (PARP) reactivity, which is induced by increased free radicals and DNA damage (2), was determined by PAR immunoreactivity in nucleus. PAR-reactive nuclei were barely detectable in the sciatic nerves of nondiabetic AR^+/+ and AR^−/− mice. A marked increase in PAR immunoreactivity was observed in AR^+/+ mice after 12 weeks of diabetes. Moreover, 12 weeks’ diabetic AR^−/− mice were protected from PARP activation. In addition, oral treatment with the ARI fidarestat during 12 weeks of diabetes protected AR^+/+ mice from PARP activation, similar to that observed in 12 weeks’ diabetic AR^−/− mice (Fig. 4).

The expression level of p46 JNK in the sciatic nerve was similar in the nondiabetic AR^+/+ and AR^−/− mice, and it was not affected by 4 weeks of diabetes in both groups of mice. The expression levels of phosphorylated p54 JNK and phosphorylated p46 JNK were also similar in the nondiabetic AR^+/+ and AR^−/− mice. After 4 weeks of diabetes, there was a significant increase in the ratio of phosphorylated protein to total protein of p46 JNK in the diabetic AR^+/+ mice when compared with the nondiabetic AR^+/+ mice (P < 0.001). However, the phosphorylated protein–to–total protein ratio of p46 JNK in the AR^−/− mice was unchanged after 4 weeks of diabetes. Unlike p46 JNK, the phosphorylated protein–to–total protein ratio of p54 JNK was not affected after 4 weeks of diabetes in both AR^+/+ and AR^−/− mice (Fig. 5). Similar analysis was also made for ERK activation, but there was no significant difference between experimental groups (data not shown).

Under normal conditions, AR deficiency did not affect the structure of nerves, as indicated by the comparison of the total fascicular area, myelinated fiber size, and total myelinated fiber of the sural nerves of AR^+/+ and AR^−/− mice (Table 2). These findings were also confirmed by analyzing the peripheral nerves of live AR^−/− mice mated with thy1-YFP transgenic mice, which have YFP protein expression in all motor and sensory neurons, including the sural nerves, and allow direct viewing of nerves in the live animals. The gross innervation patterns of major organs at all the ages studied (4, 8, 12, and 24 weeks), such as abdomen, diaphragm, kidney, bladder, ear, and lower leg, in AR^−/−mice were not obviously different from those of AR^+/+ mice under normal conditions (Fig. 6), suggesting that AR deficiency does not affect the normal development of the nervous system.

In the AR^+/+ mice, 8 weeks of diabetes caused a significant reduction in total fascicular area, mean myelinated fiber size, and fiber number (Table 2). Similar changes in these morphologic parameters were also observed after 12 weeks of diabetes. In the AR^−/− mice, 8 or 12 weeks of diabetes did not cause a significant reduction in fascicular area and fiber number. Mean myelinated fiber size, however, was reduced after 8 or 12 weeks of diabetes. The effect of ARI on diabetes-induced structural changes of the nerves was also tested in wild-type mice. The results showed that mice with 12 weeks of diabetes treated with ARI also showed a reduction in fascicular area and mean myelinated fiber size (Table 2). However, the reduction in fascicular area was less severe than that found in diabetic mice not treated with ARI. Furthermore, ARI treatment also protected the nerve against diabetes-induced fiber loss. However, there was no significant difference in fascicular area or total fiber number in the sural nerves from 8 and 12 weeks’ diabetic AR^−/− and AR^+/+ mice or 12 weeks’ diabetic AR^+/+ mice treated with ARI.

As previously described, AR-deficient mice had a mild impairment in water reabsorption in the kidney, leading to slightly increased urine output and increased water consumption under normal conditions without affecting the levels of serum electrolytes (13). However, under diabetic conditions, there was no significant difference in urine volume, water consumption, and blood and nerve glucose levels between AR^−/− and AR^+/+ mice. In addition, morphometric analysis of sural nerve biopsies and gross comparisons of peripheral nerves of AR^+/+ or AR^−/− mice mated with thy1-YFP transgenic mice showed no obvious effect of AR deficiency on the peripheral nervous system under normal conditions. Taken together, these observations indicate that the AR^−/− mouse model would be an invaluable model for studying the role of AR deficiency in the pathogenesis of diabetic neuropathy.

Previous studies using ARIs in experimental animal models (4,8) and diabetic patients (26) showed that inhibition of AR activity could partially restore MNCV slowing, which is one of the major features of diabetic neuropathy. Transgenic mice that overexpress AR either ubiquitously (4) or specifically in Schwann cells (5) developed more severe MNCV deficits. Here, we demonstrated that diabetic AR^−/− mice were protected against MNCV and SNCV reduction. Furthermore, ARI (fidarestat) treatment protected diabetic AR^+/+ mice from MNCV reduction, similar to its protective effect on SNCV reduction, as demonstrated by others (27). Our current data suggest that exaggerated flux through the polyol pathway plays a key role in the pathogenesis of acute diabetic neuropathy, possibly acting directly on the peripheral nerves. Although the AR level in mouse sciatic nerve is low (28), the presence of AR mRNA was clearly demonstrated (5). Furthermore, in diabetic mice, AR activity (5) and sorbitol accumulation (21) were found to be increased in this tissue. The low level of sorbitol in mouse sciatic nerve does not truly reflect the level of polyol pathway activity. We have previously shown that the sorbitol level in the sciatic nerves of diabetic C57BL/10N, nondiabetic, and diabetic sorbitol dehydrogenase–deficient mice were increased 4.3, 16.6, and 38.1-fold, respectively, above that of nondiabetic C57BL/10N mice (21), indicating that sorbitol is quickly converted to fructose by sorbitol dehydrogenase and that a significant amount of glucose is fluxed through the polyol pathway, particularly in the diabetic mice. Interestingly, trace amounts of sorbitol and fructose were found in the diabetic AR/sorbitol dehydrogenase double-mutant mice (data not shown) and the AR^−/− mice. These may be contributed by other proteins with AR-like activity, such as aldehyde reductase (29), the fibroblast growth factor–regulated protein gene (Fgfrp) (30), and the androgen-regulated vas deferens protein gene (Avdp) (30).

We found that the glutathione level was decreased in the sciatic nerves of diabetic AR^+/+ mice, but not in diabetic AR^−/− or ARI-treated diabetic AR^+/+ mice, indicating that AR activity contributes to hyperglycemia-induced glutathione depletion. Potential mechanisms of the polyol pathway’s contribution to diabetes-induced oxidative stress have been discussed previously (2). The difference in the absolute level of glutathione reported here compared with that of the previous study (5) may be the result of differences in mouse substrains or the adaptation of a more sensitive detection method in the current study. Previous studies demonstrated that an increase of oxidative stress could lead to the activation of stress-activated protein kinases, such as MAPK (10). Our current data showed that diabetes caused a significant activation of p46 JNK in the sciatic nerves of wild-type mice, which was prevented by AR deficiency. It is likely that amelioration of oxidative stress by AR deficiency under diabetic conditions might contribute to attenuation of JNK activation in the sciatic nerve in AR^−/− mice because other signs of oxidative stress, such as superoxide accumulation (determined by DHE staining) or PARP activation (determined by PAR immunohistochemistry), in sciatic nerves observed in diabetic AR^+/+ mice were also ameliorated in diabetic AR^−/− mice or ARI-treated diabetic AR^+/+ mice. Taken together, our data suggest that AR is a key enzyme in the pathogenesis of diabetic neuropathy and that diabetes-induced oxidative stress might be primarily the consequence of increased flux of glucose through the polyol pathway in the nervous tissue, similar to nerve cells in the diabetic retina (14).

We observed small but significant loss of myelinated nerve fibers in our diabetic wild-type mice. Previous animal studies did not find such structural changes (4,31). This discrepancy may be due to the fact that the previous studies did not count all of the nerve fibers within the nerve fascicle. The diabetes-induced reduction of fascicular area and fiber number observed in the sural nerves of wild-type mice were largely attenuated in AR^−/− mice, indicating that AR activity contributes to these hyperglycemia-induced lesions. ARI treatment only partially alleviated these structural changes. This may be attributable to an insufficient amount of drug to completely inhibit AR in the target tissue. Interestingly, diabetes-induced reduction in mean myelinated fiber size was also not alleviated by an AR-deficiency mutation or ARI treatment, suggesting that other toxic effects of hyperglycemia, such as glycation or inappropriate activation of PKC, might contribute to this diabetes-induced structural change of the nerves. It is likely that diabetic neuropathy involves multiple pathogenic mechanisms.

Here, we demonstrate that AR or the polyol pathway plays a key role in the pathogenesis of diabetic neuropathy. Our AR^−/− mouse model is a useful model for determining its contribution to the cascade of molecular events, such as reduced number of cytoskeletons in distal axons (32–34), transport of cytoskeltons or nerve growth factors (34), or aberrant phosphorylation (35), to further explain the role of AR deficiency in functional, biochemical, and structural recovery of peripheral nerves. Of course, these AR-deficient mice can also serve to determine the interactions of AR with other hypothesized contributors to diabetic neuropathy, such as insulinopenia, impaired neurotrophic support, or glycation.

This project was partly supported by grants from the Hong Kong Research Grant Council (HKU7313/04M) and the Biotechnology Research Institute, Hong Kong University of Science and Technology (to S.K.C.).