Experimental Diabetic Neuropathy With Spontaneous Recovery: Is There Irreparable Damage?

Procedures were approved by the animal care committee of the University of Calgary in accordance with the Canadian Council on Animal Care. Adult male Swiss Wistar mice (20–30 g, n = 60; Department of Biosciences, University of Calgary) were housed two per plastic cage (8) on a 12-h light/dark cycle, with food and water available ad libitum. Mice were assigned randomly to either a diabetic or control group, and diabetes was initiated by three consecutive injections of streptozotocin (STZ; 50 mg/ml Zanosar; Upjohn, Don Mills, ON, Canada) in citrate buffer (pH 4.8) during the fasting state (one intraperitoneal injection per day: 85 mg/kg on day 1; 70 mg/kg on day 2; and 55 mg/kg on day 3). Control animals received equivalent volume doses of the citrate buffer solution.

Hyperglycemia was verified 1 week later by sampling from a tail vein. A fasting whole–blood glucose level ≥16 mmol/l (normal 5–8 mmol/l) was our criterion for experimental diabetes. Whole–blood glucose tests were carried out using an Accuchek IIm (Boehringer Mannheim, Dorval, PQ, Canada), whereas plasma glucose was checked with a glucose oxidase method (Ektachem DT-II analyzer; Eastman Kodak, Rochester, NY). The cohort of mice studied had diabetes of long duration (9 months) intermixed with a subset that recovered and were euglycemic between 6 and 9 months after diabetes initiation (all mice in the recovered group were originally diabetic until 6 months’ post–STZ injection).

Electrophysiological recordings were made under anesthesia (Viking I; Nicolet, Madison, WI), as reported elsewhere in rats (9). Motor conduction velocity in sciatic-tibial fibers was assessed by stimulating at the sciatic notch and knee while recording the M wave (compound muscle action potential) from the tibial-innervated dorsal interossei foot muscles. Caudal conduction (sensory conduction velocity and mixed nerve action potentials) was recorded in the tail of mice by stimulating distally and recording proximally at fixed distances. Temperature near nerves was kept constant at 37°±1°C using a subdermal thermistor and heating lamp.

At the end point, tibial and sural nerves were removed and processed for epon embedding (10). Animals were subsequently killed with an overdose of sodium pentobarbital. Briefly, samples were fixed in 2.5% glutaraldehyde in 0.025 mol/l cacodylate buffer overnight, serially washed in 0.15 mol/l cacodylate buffer, postfixed in 2% osmium tetroxide in 0.12 mol/l cacodylate, dehydrated using a series of graded alcohols and propylene oxide, and embedded in epon. Transverse sections (1.0 μm thick) were cut with an ultramicrotome (Reichert, Vienna, Austria) using glass knives and were stained with toluidine blue. Morphologic examination of specimens used a JAVA-based image analysis program (Jandel Scientific) (11) obtained with a light microscope (Axioskope; Zeiss, North York, ON, Canada) and attached video camera (AVC-D7; Sony) interfaced with a computer. All counting was performed with the microscopist masked to the identity of the animal group.

Dorsal root ganglia embedded in epon were serially cut (1 μm thick) on an ultramicrotome. The beginning of dorsal root ganglion tissue was considered to be the point at which at least 10 neuronal profiles were seen per transverse section (this became known as section A, or the reference section). A subsequent section was made 1 μm away from section A and served as a look-up section (section B). Neurons were counted as those with nuclear profiles that were visible in section A but not in section B. Neurons were differentiated from satellite and Schwann cells based on size and the presence of Nissl substance. The next pair of sections were processed in a similar manner at a predetermined distance interval, k (50 μm), away from the first set (n + k). Pairs of sections were then obtained for counting every 50 μm (n + 2k, n + 3k, + nk), until only 10 neurons were seen once again (indicating the end of the dorsal root ganglion). Total neuronal number was then calculated based on the formula: N (number of neurons counted in all Axioskope sections) × k (distance interval). Sections were visualized (500×) and quantitated under a light microscope (Axioskope; Zeiss). Images were captured using a digital camera (Axiocam; Zeiss) and neuronal diameters measured using an interactive measurement module contained in an image analysis program (Axiovision 2.05; Zeiss). Only neurons with nuclear profiles were used to measure neuron diameters. All counting was performed with the microscopist masked to the identity of the experimental group.

Footpads were removed from the right plantar hind foot and fixed in Zamboni’s fixative (2% paraformaldehyde in 0.1 mol/l phosphate buffer) overnight. The same anatomical footpads were used in each animal for comparison. Tissues were then washed 3 × 5 min in 0.1 mol/l PBS, followed by a 3 × 5-min DMSO wash, and finishing with a 3 × 5-min 0.1 mol/l PBS wash once again. Footpads were then cryoprotected overnight in 20% glycerol/0.1 mol/l PBS and subsequently frozen in optimum cutting temperature compound. Frozen sections were cut at 35 μm on a cryostat and thawed onto electrostatically charged slides (Superfrost Plus; VWR, Edmonton, Canada). For immunostaining, sections were subjected to antigen retrieval using a 20-min immersion in boiling sodium citrate (10 mmol/l) to expose the epitope. They were then lyophilized in 1% Triton X-100/PBS. After washes in 0.1 mol/l PBS (3 × 5 min), slides were incubated for 24 h in a humidified box at 4°C with rabbit anti–protein gene product 9.5 (1:800 dilution; Ultraclone, Cambridge, U.K.) diluted with 0.1 mol/l PBS, 1% NGS, and 0.3% Triton X-100. Slides were washed and incubated with Alexa 488 (1:500 dilution; Molecular Probes, Eugene, OR) anti-rabbit secondary antibody for 1 h at room temperature and mounted with Vectashield. Images were captured using an Olympus laser scanning confocal microscope equipped with epifluorescence. Z-stack images were obtained at 0.35-μm intervals through the section and combined (γ = 0.7) using Fluoview software (Olympus). Three fields per section and five sections per animal underwent quantitative analysis under a fluorescent microscope, and a mean value per mouse was calculated. Epidermal fibers were counted as those that extended from and above the dermal papillae. Fiber length was measured using an interactive measurement module (Axiovision 2.05; Zeiss) as the straight length the fiber extended into the epidermis irrespective of tortuosity. Sweat gland innervation was scored using a 5-point Likert scale (0 being no innervation, 5 being complete innervation). All counting and scoring was carried out with the microscopist blinded to the identity of each footpad.

All data are presented as the means ± SE. Data were analyzed by a one-way ANOVA with Bonferroni corrected post hoc Student t test comparisons. For comparing diameter histograms (group versus diameter), a two-way ANOVA was used, primarily to assess the interaction effect. In all tests, statistical significance was set at α = 0.05.

Long-term diabetic animals demonstrated features of hyperglycemia, including polydipsia, polyuria, and a decrease in activity. Whole–blood glucose measurements verified significant hyperglycemia beginning 1 week after STZ injection and persisting at 6 and 9 months postinjection. Recovered mice had elevated blood glucose levels at 1 week and 6 months, but by 9 months they had spontaneous recovery of islet function, with glycemia approaching the control range. Weights of diabetic and recovered mice were similar and significantly less than control mice by the end point (P < 0.01) (Table 1).

Diabetic mice had slowing of motor conduction velocity, reduced compound muscle action potential amplitudes (M waves), a trend (nonsignificant) toward sensory conduction velocity slowing, and reduced caudal nerve action potential amplitudes. In contrast, recovered mice did not demonstrate any abnormal electrophysiological measurements by the 9-month end point (Table 2). Indeed, there was a nonsignificant trend toward higher motor conduction velocity in recovered mice.

Myelinated fiber density and number in tibial fascicles, prominently populated by motor axons, were not altered by long-term diabetes. Mean fiber and axonal diameters and areas were smaller in diabetic mice (P < 0.05), but the recovered group had near-normal calibers (Fig. 1A). Myelin in tibial fascicles was thinner in diabetic mice but not in recovered mice (P < 0.05) (Table 3). Diabetic sural nerves populated by sensory (and some autonomic) axons also had preserved fiber density and numbers. Mean fiber, but not axonal, diameter was reduced in both the diabetic and recovered groups (P < 0.05) (Fig. 1B) and was attributed to thinner myelin in the former (P < 0.05) (Table 4).

Diabetic animals had significantly reduced numbers of epidermal fibers in plantar footpads (P < 0.001) (Fig. 2A). There was also dermal denervation with decreased sweat gland–associated axons as judged using a rating scale (diabetic: 1.3 ± 0.7; recovered: 3.8 ± 0.6; nondiabetic: 3.8 ± 0.5; P < 0.05; maximum score = 5). Recovered animals had epidermal fiber numbers similar to nondiabetic animals (Fig. 2B). Dermal and sweat gland innervation were also similar between recovered and nondiabetic groups. Recovered animals, however, did have longer epidermal fibers than either diabetic or nondiabetic animals (diabetic: 21.5 ± 1.5 μm; recovered: 25.0 ± 0.9 μm; nondiabetic: 21.9 ± 0.6 μm; P < 0.05).

Using the unbiased physical dissector stereological counting method, nondiabetic animals had a mean of 2,538 ± 200 neurons with a mean diameter of 23.6 ± 1.2 μm in the L₅ dorsal root ganglia. Diabetic animals had significantly fewer neurons compared with controls (P < 0.05). Mice with recovery also had fewer neurons (P < 0.05), with values similar to unrecovered diabetic animals (diabetic: 1,690 ± 143; recovered: 1,803 ± 145) (Fig. 3A). There was no difference in mean neuronal diameter among all three groups (diabetic: 23.4 ± 0.6 μm; recovered: 23.3 ± 0.6 μm) (Fig. 3B). Though not quantitated, neurons of diabetic and recovered mice appeared darker, with less nucleolar definition (Fig. 4).

The major findings of the present study were: 1) long-term diabetic mice developed indexes of hyperglycemia with significant weight loss, similar to that of untreated humans; 2) a significant number of previously hyperglycemic diabetic mice reverted to near-euglycemia, indicating spontaneous recovery from diabetes; 3) long-term diabetic mice demonstrated electrophysiological abnormalities observed in human disease (i.e., reduced motor conduction velocity, decreases in amplitudes of M waves and nerve action potentials; electrophysiological abnormalities disappeared in recovered mice); 4) other structural abnormalities of the long-term murine model of diabetes closely modeled human disease (i.e., axon atrophy, myelin thinning, loss of epidermal axons, loss of sweat gland innervation, and sensory neuron loss); and 5) recovered mice had recovery of axonal structural abnormalities but no recovery of lost neurons.

Our long-term diabetic mouse model helps with some difficulties that have arisen in studies of STZ-induced diabetic rat models. STZ-induced diabetic mice reportedly develop pancreatic lesions similar to patients with type 1 diabetes. Cytokines or reactive radicals released during T-cell–mediated pancreatic islet inflammation may involve activation of poly(ADP-ribose) polymerase or inducible nitric oxide synthase (12–14), with selective targeting of GLUT2 by STZ (15,16). Our mice survived long periods of diabetes without exogenous insulin administration. Because STZ-induced diabetic mice do not progress into ketoacidosis (unlike NOD mice), it is likely that the pancreas continues to produce small amounts of insulin. An unexpected finding, however, was the reversion to near-euglycemia of a subset of these chronically diabetic animals, as also reported by Li et al. (17), Hartman et al. (18), and George et al. (19). All of the animals in the recovered group had diabetes for at least 6 months, with a critical reversion point thereafter. Given the loss of sensory neurons, it is clear that neuropathy had developed before recovery. In other cohorts of mice, identically studied, electrophysiological abnormalities are well established by 3–4 months (well before the 6-month reversion time point in this study). It is also important to note that despite dramatically improved glycemia in the recovered mice, mean plasma glucose persisted somewhat above control values. Although STZ-induced diabetic rat models similarly develop significant slowing of motor and sensory conduction velocity (20,21), they do not later exhibit the loss of M waves or nerve action potentials or the myelin thinning that are observed in humans.

In diabetic neuropathy, a sequence of reversible, largely electrophysiological abnormalities likely involves metabolic derangements, such as polyol accumulation (22–26), decreased Na-K-ATPase activity, and intra-axonal sodium accumulation leading to slowed conduction (24,27). Sensory neurons may be more susceptible to irreversible structural damage. Degeneration of both myelinated and unmyelinated fibers particularly in sural sensory nerves are often noted in nerve biopsies from diabetic patients (28). The very distal loss of fibers in sural nerves also seen in the murine model (29) may support the concept that defects in growth factor synthesis, uptake, or signaling contribute to diabetic neuropathy (30–33). Growth factor deficiency may also be compounded by slowed retrograde axonal transport, limiting delivery of growth factors to proximal neurons (34–36). Akkina et al. (37) identified early involvement of nonpeptidergic unmyelinated afferents in STZ-induced diabetic mice that were rescued by trophic therapy. Our finding of epidermal nerve fiber loss is a striking parallel to human disease. Axonal degeneration in human diabetic neuropathy is often also associated with attempted regeneration (38,39). We did not detect evidence of regenerative clusters or sprouts in diabetic mice.

In our long-term mice, there was substantial loss in the number of sensory neurons in diabetic and recovered animals. Russell et al. (40) have argued for an aggressive circle of apoptosis in diabetic rat dorsal root ganglion neurons, secondary to mitochondrial dysfunction. This level of loss, however, has not been observed in actual rat diabetic models. Kishi et al. (41) and Zochodne et al. (5) failed to demonstrate sensory neuron loss in long-term diabetic rats. Later events in diabetic neuropathy may also center around impaired axonal transport of neurofilaments or other cytoskeletal structures (42). Limited cytoskeletal support by perikarya may result in flawed axons or axons that are incapable of dynamic restructuring (because of rapid glycosylation of proteins). Excessive production of free radicals in diabetes or oxidative stress may generate “all or none” irreversible damage.

Had recovery been associated with replenishment of sensory neuron populations, this would have been a surprising finding, given the long-term evidence that perikaryal loss is essentially irretrievable. The robust collateral reinnervation of the skin arising in recovered diabetic mice is of considerable interest. For example, it may be that preventing some perikaryal loss is impossible in diabetic patients. The knowledge that remaining neurons may be capable of reinnervating some denervated target tissues if β-function recovers may be important to exploit. Although such striking collateral reinnervation may simply arise because neurons are no longer exposed to hyperglycemia, restored insulin signaling may be more relevant. Insulin receptor expression appears to rise in diabetes, and direct administration of insulin to sensory neurons appears capable of restoring neuron function independently of glycemic levels (43–45). Insulin support may repair neuronal function by reversing deficits in mitochondrial depolarization (46). Insulin also appears to offer significant support to regenerative sprouting (47,48) and could support collateral sprouting through synergistic actions with neurotrophic growth factors that provide such support (49–51).

Our results demonstrate that long-term experimental murine diabetes models human diabetic neuropathy in a number of important aspects. It seems evident, from this model, that neuropathy is not a single continuous process, but rather it represents interchanging activities differentially involving in sensory and motor components. Spontaneous recovery from diabetes rendering near-euglycemia allows many features of neuropathy to recover. Importantly, however, it cannot resurrect sensory neurons once lost. Clinical approaches to reconstitute islet function may also need to protect the sensory neuron directly.