Development of Selective Axonopathy in Adult Sensory Neurons Isolated From Diabetic Rats: Role of Glucose-Induced Oxidative Stress

Dorsal root ganglion sensory neurons from adult Sprague-Dawley male rats were isolated and dissociated using a previously described method (17,–19). Rats were age-matched control or 3- to 5-month STZ-diabetic rats that included cohorts receiving insulin implants (two Linplant implants placed subcutaneously; LinShin Canada, North Scarborough, ON, Canada). Rats were made diabetic with a single intraperitoneal injection of 75 mg/kg STZ (Sigma, St. Louis, MO). End points for body weight, plasma glucose, and hemoglobin A1C are presented in supplementary Table 1, which is available in an online appendix at http://diabetes.diabetesjournals.org/cgi/content/full/db09-0034/DC1. Cells were plated onto poly-dl-ornithine and laminin-coated 12-well plates (Nunclon Surface, Ottawa, ON, Canada) for the neuronal survival study, 25-mm glass cover slips (German glass no. 1; Electron Microscopy Sciences, Hatfield, PA) for neurite outgrowth and for immunocytochemistry, and 35-mm glass-bottom dishes for ROS measurement (catalog no. W20; Bioscience Tools, San Diego, CA). Neurons were cultured in Ham's F12 medium (Invitrogen, Carlsbad, CA) with 10% fetal bovine serum (FBS; HyClone, Logan, UT) for long-term cultures (>5 days). For short-term cultures, neurons were grown in defined medium with modified Bottenstein and Sato's N2 medium containing: 0.1 mg/ml transferrin, 20 nmol/l progesterone, 100 μmol/l putrescine, 30 nmol/l sodium selenite, and 10 mg/ml BSA (the modified N2 formulation used did not include insulin; all obtained from Sigma). Neuronal survival was quantified by established methods (20,21). Viable neurons were counted before experimental treatment and at time points after treatment. Neurons that died in the intervals between examination points were absent, and the viability of the remaining neurons was assessed by morphologic criteria. Neurons with membranes and soma with a smooth round appearance were viable, whereas neurons with fragmented or distended membranes and vacuolated soma were nonviable. Four images were collected from each well and from the same place using marked grids on each well bottom at 1, 4, 7, 21, and 28 days using a light microscope (phase contrast Nikon Diaphot). Survival was confirmed by trypan blue exclusion at the last time point. For neurite outgrowth measurements, images of cultures were collected from 10 random fields from each well. The longest axon and total number of neurons and number of intersects of their neurites with a vertical grid were counted using a morphometric approach and SigmaScan Microsoft software. The total number of intersects per neuron was taken as the parameter of total axon outgrowth as previously described (22).

ROS levels were detected using real-time fluorescence microscopy on a Carl Zeiss LSM510 inverted confocal microscope using two dye-based approaches:1) dihydrorhodamine 123 (DHR123) and 2) 5-(and-6)-chloromethyl-2′7′-dichlorodihydrofluorescein diacetate acetyl ester (CM-H₂DCFDA). Neurons were loaded with 5 μmol/l of DHR123 (in 100% EtOH; Sigma) or 1.2 μmol/l CM-H₂DCFDA (in 100% anhydrous DMSO; Molecular Probes) for 30 or 15 min, respectively, at 37°C, and then washed three times with F-12 and visualized using the argon laser (at 2% power). For DHR123, excitation (max) was 505 nm and emission (max) 534 nm, and for CM-H₂DCFDA, excitation (max) was 488 nm and emission collected above 510 nm. Images were collected and analyzed as pixel intensity at the level of the axon or cell body. For axonal signals, all axons in each field were assessed, and, using the Carl Zeiss software package, a line was placed along the length of the axon and pixel intensity collected per micrometer of axon. Control experiments were performed to confirm that the DHR123 and CM-H₂DCFDA signals were sensitive to antioxidant (N-acetyl-cysteine) or pro-oxidant (H₂O₂) treatment (not shown).

Dorsal root ganglion neurons were fixed with 4% paraformaldehyde in phosphate buffer at pH 7.4 for 15 min followed by permeabilizing with 0.3% Triton X-100 in PBS. Nonspecific binding was blocked by incubation with blocking reagent combined with FBS and 1.0 mol/l PBS in proportions of 3:1:1 (catalog no. 1 096 176; Roche, Indianapolis, IN) for 1 h at room temperature and washed with PBS three times. Cells were then incubated with antibodies to β-tubulin isotype III (1:1,000; Sigma, Oakville, ON, Canada), caspase 3 (1:100; Chemicon, Temecula, CA), (E)-4-hydroxy-2-nonenal (4-HNE) adducts (1:500 anti-4-HNE adducts Pab; Alexis Biochemicals, San Diego, CA), manganese superoxide dismutase (MnSOD; 1:300; StressGen, Ann Arbor, MI), and phosphorylated neurofilament H (NFH; 1:500 SMI-31; Covance, Berkeley, CA). For localization of mitochondria, live cultures of dorsal root ganglion neurons were treated for 15 min with 200 nmol/l MitoFluor Green dye (Molecular Probes, Eugene, OR) before being fixed for immunostaining for anti-NFH. Primary antibodies were incubated with slides overnight in a humidified chamber followed by fluorescein isothiocyanate– and CY3-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA; 1:250) for 3 h at room temperature. Fluorescence signals were examined and quantified using a Carl Zeiss Axioscop-2 mot microscope with AxioVision 3 software and equipped with fluorescein isothiocyanate, CY3, 4′,6-diamidino-2-phenylindole filters, and an AxioCam camera.

Because of the technically difficult nature of the cultures and the image acquisition, not all images were collected in a blinded fashion, although a random approach was taken for image capture from each well. Where appropriate, data were subjected to one-way ANOVA with post hoc comparison using Tukey's test (Prism 4; GraphPad Software, San Diego, CA). In all other cases, standard two-tailed unpaired Student's t test with Welch's correction was performed, which does not assume equal variances, with significance levels of P = 0.05, using GraphPad Prism 4.

Initial experiments were designed to test the influence of high concentrations of glucose (50 mmol/l) on survival of sensory neurons in culture, the induction of apoptosis, and markers of oxidative stress. Sensory neurons from adult control rats were grown up to 4 weeks in F12 medium with 10% FBS. Healthy neurons were phase bright, and by 1 week, cultures exhibited high numbers of nonneuronal cells, including Schwann cells, satellite cells, and fibroblasts (Fig. 1,A and B). High glucose concentration had no effect on survival of neurons at any time and did not affect axon outgrowth on day 1 (Fig. 1,C and insert). Surviving dorsal root ganglion neurons did decline over time and had diminished by 50% by the 28th day; however, there was no effect of 50 mmol/l glucose on this process. Oxidative stress and/or induction of apoptosis was assessed by staining for adducts of 4-HNE and caspase 3 activation. There was no effect of 50 mmol/l glucose on caspase 3 or 4-HNE adduct expression in neuronal perikarya at 2 or 4 weeks (Fig. 1 D and E). However, there was a small (10–20%), but statistically significant, increase of activation of caspase 3 and 4-HNE adduct expression in the 4- vs. 2-week group of dorsal root ganglion neurons; this effect was not perturbed by 50 mmol/l glucose.

A second set of experiments tested whether high glucose concentration induced oxidative stress in neuronal perikarya or axons in cultures of adult sensory neurons under defined conditions in F12 + modified N₂medium. To assess oxidative stress, a real-time video microscopy approach was taken using the fluorescent dye CM-H₂ DCFDA. This dye is believed to give a readout of cellular ROS levels, including peroxynitrite, hydrogen peroxide, and hydroxyl radicals. Neurons were grown 24 h under control (10 mmol/l glucose) or high (25 mmol/l) glucose. SupplementaryFig. 1 shows no effect of 25 mmol/l glucose on CM-H₂DCFDA intensity in neuronal perikarya (supplementaryFig. 1,A) or axons (supplementaryFig. 1,B). High glucose concentration did not elevate 4-HNE adducts in cell bodies (supplementaryFig. 1,C) or axons (supplementaryFig. 1 D).

We next determined whether cultured neurons from STZ-diabetic rats were also refractory to high glucose concentration. Neurons from age-matched control rats were grown in defined F12 + modified N₂medium under low (10 mmol/l) glucose with 10 nmol/l insulin, whereas neurons from 3- to 4-month STZ-diabetic rats were grown in F12 + modified N₂medium with 25 mmol/l glucose with no insulin support (to attempt to mimic diabetes in vivo). Cell survival of neurons from normal and diabetic rats over a 4-day period and under defined conditions were identical. Figure 2 shows the phase contrast images of sensory neurons from control (Fig. 2,A, upper and lower panels) and STZ-diabetic (Fig. 2,B) rats after 4 days in culture. Diabetic dorsal root ganglion neurons exhibited abnormal morphologic changes in neurite outgrowth, which were characterized by less axonal outgrowth (Fig. 2,C) and appearance of phase-dark axonal swellings and beading along neurites (Fig. 2,B, lower panel, inset). Separate cultures from control or STZ-diabetic rats were grown for 1 day and assessed for level of ROS in cell bodies and axons using CM-H₂ DCFDA (Fig. 3,A, B, and E) or DHR123 (Fig. 3,C, D, and F) fluorescence imaging. ROS levels in cell bodies did not differ between control and diabetic neurons (not shown). Axonal ROS levels in neurons from STZ-diabetic rats showed at least a twofold elevation compared with control using both dyes (Fig. 3,E and F). Cultures were then grown for 4 days and immunostained for 4-HNE adducts and β-tubulin III. Control cultures exhibited normal axons with very rare instances of axonal swellings (or varicosities); such swellings were negative for 4-HNE adduct staining (Fig. 4,A, C, and E). In comparison, cultures from STZ-diabetic rats demonstrated large numbers of axonal swellings that were characterized with positive staining for 4-HNE adducts (Fig. 4 B, D, and F). The cell bodies of control and diabetic cultures revealed similar levels of highly positive staining for 4-HNE adducts.

It was important to determine whether differences in axonal morphology, ROS levels, and 4-HNE staining between control and diabetic cultures was controlled by acute effects in the culture of glucose concentration and/or insulin. Dorsal root ganglion sensory neurons from STZ-diabetic rats were cultured at 10 mmol/l glucose, 25 mmol/l glucose, and 15 mmol/l mannitol (+ 10 mmol/l glucose) in defined F12 + modified N₂medium without insulin support. ROS level in axons was measured after 1 day of culture (Fig. 5,A and B). Quantification of ROS accumulation in axons using CM-H₂ DCFDA dye showed a statistical increase in ROS levels after high glucose administration compared with low glucose concentration (Fig. 5,E). High mannitol concentration revealed an intermediate level of ROS production in axons (Fig. 5,E). 4-HNE adduct accumulation in axons of sensory neurons from STZ-diabetic rats was measured after 3 days of culture (Fig. 5,C and D). High glucose concentration induced a significant increase in 4-HNE adduct accumulation in axons compared with those cultured at 10 mmol/l glucose and mannitol (Fig. 5,F). Neurons isolated from STZ-diabetic rats treated for the final month with insulin that significantly lowered hyperglycemia (supplementary Table 1) did not respond to high glucose with elevated ROS or 4-HNE accumulation in axons (Fig. 5 E and F). Studies were also performed in which STZ-diabetic neurons were treated with 10 nmol/l insulin in the presence of low or high glucose concentration, and under such conditions there was no impact on ROS or 4-HNE levels of the insulin treatment (data not shown). In addition, an alternative control was used for mannitol, 15 mmol/ll-glucose, and results were the same as those seen with mannitol.

Experiments now determined whether oxidative stress was causally linked to aberrant axonal structure in neurons derived from STZ-diabetic rats. The antioxidant N-acetyl cysteine (NAC) was tested for its ability to lower ROS. Cultures from control rats were assessed for ROS using DHR123 imaging and treated acutely with 1 mmol/l NAC (Fig. 6). Figure 6,A–D shows that NAC rapidly neutralized ROS levels. In a separate experiment, STZ-diabetic cultures were plated for 1 day in the presence or absence of 1 mmol/l NAC, and effect on axonal outgrowth was determined; NAC was clearly able to enhance levels of total axonal outgrowth (Fig. 6,E). Cultures of sensory neurons from control or STZ-diabetic rats were maintained for 3 days (at this time point, diabetic neurons exhibited 4-HNE staining) (Fig. 5,C and D). During the last 24 h of culture, control and STZ-diabetic neurons were treated with 1 mmol/l NAC for a further 24 h. After 4 days all groups of cultures were treated with MitoFluor Green dye to specifically stain mitochondria and then fixed and immunostained for phosphorylated NFH.Figure 7,A and B shows control neurons with normal axonal structure and uniform axonal localization of mitochondria. Cultures from STZ-diabetic rats demonstrated abnormal accumulations of mitochondria in axonsthat colocalized with phosphorylated NFH staining (Fig. 7,C and D). Treatment with 1 mmol/l NAC significantly lowered the number of axonal swellings in normal and STZ-diabetic cultures (Fig. 7 E–G).

Increased ROS under high glucose concentration in diabetic neurons may reflect an impaired capacity to scavenge free radicals. Therefore, the effect of diabetes and high glucose on the expression of MnSOD was assessed in isolated mitochondrial preparations from dorsal root ganglia and in cultured sensory neurons. In Fig. 8, we present Western blot data demonstrating that MnSOD expression was reduced by 32% (P < 0.005) in mitochondria isolated from dorsal root ganglia of STZ-diabetic rats, and this was prevented by insulin therapy.Figure 9,A and B shows axonal staining of MnSOD in sensory neurons from age-matched control rats cultured for 1 day under 10 mmol/l glucose. Results were similar under 25 mmol/l glucose (images not shown). Neurons from STZ-diabetic rats cultured under 10 mmol/l glucose for 1 day exhibited diminished immunostaining for MnSOD (Fig. 9,C and D); however, if cultured in the presence of 25 mmol/l glucose, then no alteration in MnSOD expression was observed (Fig. 9,E and F). When cultures were maintained for 3 days in vitro, the ability of 25 mmol/l glucose to maintain or raise MnSOD expression was significantly impaired (Fig. 9 G and H).

This work demonstrates for the first time that high glucose concentration can induce oxidative stress and axonal degeneration in adult sensory neurons, but only if these neurons exhibit an STZ-induced diabetic phenotype. Neurons from normal rats did not exhibit oxidative stress or cell death under high glucose concentrations. High glucose concentration in neurons from diabetic rats triggered elevated ROS levels and adducts of 4-HNE specifically in axons, with the perikarya being unaffected, and this distally directed process of oxidative stress resulted in axonal swelling and axon degeneration. This neurodegenerative process may have been triggered or exacerbated by impaired anti-oxidant defenses, as evidenced by the diminished expression of MnSOD in axons. However, these neurons did not undergo any process of cell death. The appearance and content of the axonal swellings closely resembled dystrophic axons observed in human disease.

Previous studies on cultures of embryonic sensory neurons have shown that high glucose concentration triggers oxidative stress, activation of apoptosis, and death of neurons (4,–6). In vitro studies by Russell and colleagues (3,–5) and Feldman and colleagues (4,6) used glucose concentrations of 45–50 mmol/l and higher and showed, using an array of markers of oxidative stress and apoptosis, that these indexes were enhanced in cultured embryonic neurons. Instead, we found that survival of adult sensory neurons from lumbar dorsal root ganglia of normal or diabetic rats did not change under high glucose concentrations. This supports previous in vitro studies on sensory neuron cultures from adult rats and mice (dissociated and explant), where exposure to a hyperglycemic environment failed to induce neuronal death over 1–8 days (23,24). In addition, embryonic sensory neurons given time to mature in vitro for several weeks also become resistant to the apoptosis-inducing effects of high glucose concentration (25). Differences in culture conditions between embryonic and adult preparations may play a role here, and in particular the presence of nonneurons in the mature and adult cultures may be protective against hyperglycemia.

However, this glucose insensitivity of adult rat sensory neurons with respect to cell survival is in good agreement with in vivo studies on experimentally diabetic rodents (8,–10,12,26). In the STZ and BB rat models of type 1 diabetes, there is no loss of dorsal root ganglion neurons until 10–12 months of disease, and by this time there is already extensive distal axon loss (9,–11,27). There are clearly a range of effects of STZ-induced diabetes on neuronal survival and fiber loss in rodent models, depending on the rodent and/or strain. Clearly, in rat models, either in vitro or in vivo, neuronal cell loss is not a feature of the pathogenesis that underlies distal axon loss. It should be noted that sensory neurons under diabetic conditions can express markers of apoptosis, e.g., activated caspase-3, but these indexes are dissociated from cell death (8). These varied effects of the diabetic state on neuronal survival in the animal models must be judged with caution given the findings that neuronal cell loss in the human form of the disease is not a major feature of the pathology. Studies on postmortem lumbar dorsal root ganglion and sympathetic ganglia derived from type 2 diabetic patients showed ultrastructure of the neuronal perikarya to be abnormal, with signs of shrinkage and axonal dystrophy; however, there was no evidence of any cell loss (14,15). This lack of clear neuronal cell loss occurred while there was overt distal axonal loss in the epidermis (28,–30) and sural nerve (1,2,31) of type 1 and type 2 diabetic patients.

The key finding of the current work is that high glucose concentration induced oxidative stress in the axons of sensory neurons isolated from STZ-diabetic rats. This effect was not caused by osmotic stress, and the presence or absence of insulin in the culture was irrelevant to the development of neurodegeneration. The failure of glucose to enhance such processes in normal neurons clearly reveals the importance of the “STZ-induced diabetic phenotype” of neurons, which we propose provides susceptibility of these neurons to glucose-induced metabolic stress. Whether such neurons have a metabolic memory related to previous high glucose concentration–related stress is impossible to ascertain at this juncture. A more profitable line of reasoning relates to growth factor starvation and alterations in the phenotype of the neuron. During 3–5 months of STZ-diabetes, the loss of neurotrophic support from insulin, glial cell line–derived neurotrophic factor, nerve growth factor, and neurotrophin-3 has been well described, and it is known that these growth factors combine to modulate sensory neuron gene expression (32,,–35). Of particular interest is the ability of insulin and neurotrophin-3 to correct abnormalities in mitochondrial function and calcium homeostasis (17,18,36). Therefore, in the cultures from diabetic rats studied herein, it is feasible that the ability to respond to glucose-induced oxidative stress was impaired as a consequence of suboptimal mitochondrial function and dyshomeostasis of Ca²⁺ (37). We believe abnormalities in Ca²⁺ homeostasis leading to raised intracellular Ca²⁺ concentration in neurons in diabetes is important because this key molecular marker of neurodegeneration appears more rapidly in the sensory neurons with the longest axons (36). We know of no other molecular marker of disease that is specifically associated with lumbar dorsal root ganglion sensory neurons with the longest axons. This abnormal Ca²⁺ homeostasis could be attributable to global gene expression changes, including affects on endoplasmic reticulum resident proteins (e.g., SERCA2 [sarcoplasmic reticulum Ca²⁺-ATPase 2]) (38), and in addition diabetes may impair expression of proteins associated with antioxidant defenses (e.g., MnSOD and thioredoxin-interacting protein) (39) and growth factor receptors (e.g., the p75^NTR receptor) (40). Furthermore, sensitivity of sensory neuron axon outgrowth to extracellular matrix proteins is known to alter under stress, e.g., neurons become more sensitive to fibronectin at the expense of laminin after axonal injury (41). These changes in cellular phenotype could combine to sensitize the cell to high glucose concentration and/or changes in growth factor support.

The axonal swellings observed in diabetic neurons were comprised of accumulation of phosphorylated NFH and mitochondria, as seen in dystrophic axonal swellings in autonomic ganglia and nerve, lumbar dorsal root ganglia, and epidermal nerve fibers of skin in humans with diabetic neuropathy (14,15,42). We hypothesize that axonal swellings are sites of mitochondrial and structural protein (e.g., NFH) accumulation caused by blockade of their axonal transport, and the outcome is a dearth of functioning mitochondria and structural proteins at distal axonal sites. The absence of structural proteins and ATP-generating organelles in the distal axon, where high ATP concentration is required for optimal axonal plasticity in the form of axon sprouting and regeneration (43), will trigger pathological conditions that lead to an absence of efficient axonal plasticity. This process may precede or predispose the axon to degeneration and dissolution, and distal axonal degeneration will follow. A key aspect is that the neurodegenerative process within the epidermis caused by the “STZ-induced diabetic state” will not be repaired through normal regenerative processes because these repair functions will also be targeted. The diabetic state as envisaged would include toxic effects of high glucose concentration in addition to abnormalities in neurotrophic support.

We believe thesedegenerative processes proceed more vigorously in the axon than the perikaryal region of sensory neurons, and this explains the distal dying back nature of diabetic sensory polyneuropathy. The axonal environment during axon sprouting and regeneration would have a very high ATP demand; in fact, 50% of all ATP synthesized is required for growth cone motility during axon sprouting and regeneration (43). Also, ROS scavenging systems, primarily MnSOD, may not be as efficient in the axon because of reduced expression.Figure 9 shows that sensory neurons with a STZ-diabetic phenotype can mount an MnSOD-based antioxidant defense in axons in response to high glucose; however, by 3 days in vitro, the upregulation of MnSOD expression shows signs of failing and may contribute to enhanced oxidative stress and associated aberrant axon structure (Fig. 9 H). These events may combine to place an additional load on mitochondrial bioenergetics not seen in the perikarya and sensitize the physiology of the organelle to elevations in Ca²⁺ and high glucose concentration.

E.Z. was supported by grants to P.F. from the Canadian Institutes of Health Research (ROP-72893 and MOP-84214) and the Juvenile Diabetes Research Foundation (1-2008-193). D.R.S. was supported by a grant to P.F. from the Manitoba Health Research Council. This work was also funded by the St Boniface General Hospital and Research Foundation and the Manitoba Medical Services Foundation.

No potential conflicts of interest relevant to this article were reported.

We thank Gordon Glazner, University of Manitoba and St Boniface Hospital Research Centre, for permitting access to the Carl Zeiss LSM 510 and Jason Schapansky and Randy Van der Ploeg for expert technical assistance. We appreciate the efforts of Douglas Zochodne, University of Calgary, and Alex Verkhratsky, University of Manchester, for critically reading the manuscript before submission.