Insulin Prevents Depolarization of the Mitochondrial Inner Membrane in Sensory Neurons of Type 1 Diabetic Rats in the Presence of Sustained Hyperglycemia

Male Wistar rats (300 g) were made diabetic by a single intraperitoneal injection of STZ (55 mg/kg; Sigma). Age-matched control groups were also created. Tail blood glucose was assayed 3 days after injection using glucose test strips (BM-Accutest; Roche Diagnostics, Basel, Switzerland) to confirm diabetes, and the animals were tested weekly thereafter. All diabetic and insulin-treated diabetic animals had blood glucose values >30 mmol/l. Rats were maintained for 8 weeks with free access to water and food. One group of STZ-induced diabetic animals received one-half of an insulin implant (LinPlant; LinShin, Scarborough, Canada) in the first week.

Nerve conduction velocities (NCVs) were measured by subtraction of distal latencies for M-waves (for MNCV) and H-reflexes (for SNCV) elicited by stimulation at the sciatic notch and Achilles tendon. The principles and practicalities of these methods are described in detail elsewhere (29). In the present investigation, the measurements were performed on rats under isofluorane anesthesia and the measurement of NCV was accomplished within 5 min of the animal becoming unconscious; this is significantly faster than that described in our previously published study and may explain the higher NCVs produced in the present study (29). Core body temperature was monitored by a rectal probe and maintained at ∼37°C. The temperature adjacent to the sciatic nerve was monitored by microthermocouple connected to an electronic thermometer (Comark Electronics, Sussex, U.K.). The upper flank of the animal was warmed using an infrared lamp to bring the near-nerve temperature to 37°C at the time of recording action potentials. The procedure is approved as part of the U.K. Home Office Project License awarded to D.R.T.

Animals were exsanguinated postmortem in accordance with U.K. Home Office regulations and whole blood collected in sodium heparin-coated vacutainers (Becton Dickinson Vacutainer Systems, Oxford, U.K.). Glycated hemoglobin was separated from the unglycated form using an affinity column kit (441-B; Sigma-Aldrich, Poole, U.K.). Spectrophotometric analysis at 415 nm measured the percentage of glycated hemoglobin compared with total hemoglobin.

Sciatic nerves were frozen and stored at −70°C. Sugars and polyols were extracted from freeze-dried nerves by boiling for 15 min in a solution of distilled water and 30 μg α-methylmannoside to act as an internal standard. Trimethylsilyl derivatives of the extracted nerve sugars and internal standard were produced and assayed by gas chromatography using a Hewlett Packard (Avondale, PA) 5890A gas chromatograph fitted with an Ultra 1 capillary column and a flame ionization detector.

Sensory neurons from DRG of adult rats were isolated and dissociated using a previously described method (11). The cells were plated onto poly-l-ornithine-laminin-coated 1.1-cm glass coverslips in serum- and insulin-free F12 medium (Life Technologies, Paisley, U.K.) in the presence of modified N2 additives (containing no insulin) at 37°C in a 95% air/5% CO₂ humidified incubator. Lumbar DRG sensory neurons from treated animals were cultured for 3–4 h and then assessed for mitochondrial function. For in vitro experiments, sensory neurons were cultured for up to 24 h with or without insulin (1.0 or 10 nmol/l) or 50 mmol/l glucose, and then mitochondrial inner membrane potential (Δψ_m) was assessed.

This technique was based on previously described procedures (1,30,31). Cultured DRG sensory neurons were loaded with 10 μmol/l rhodamine 123 (R123) (for mitochondrial analysis) and/or fura-2/AM (for calcium) for 10 min at room temperature in standard physiological saline (in mmol/l): 140 NaCl, 3 KCl, 2 CaCl₂, 2 MgCl₂, 10 glucose, and 20 HEPES/NaOH, pH 7.4. After loading, the cells were washed in normal saline for an additional 30 min to ensure the de-esterification of R123. Glass coverslips with stained cells were transferred into the 500-μl perfusion chamber mounted on the stage of an upright microscope (BX50WI; Olympus, Tokyo, Japan) equipped with 20× water immersion objectives. Neurons were exposed to excitation light provided by a monochromator (Polychrom IV; TILL Photonics, Gräfelfing, Germany) at 490 nm (for R123) and 340 and 380 nm (for fura-2/AM). Emitted fluorescent light was collected at 530 ± 15 nm (for R123 and fura-2/AM) by a frame-transfer, cooled, intensified, charge-coupled device camera (Gene IV; Roper Scientific, Marlow, U.K.). The imaging data were acquired and analyzed using MetaFluor/MetaMorph software (Universal Imaging, West Chester, PA). The fura-2/AM signal was calibrated using an ionomycin-based in situ procedure (32); the parameters R_min, R_max, and K* were 0.2, 1.0, and 451 nmol/l, respectively.

Figure 1A shows the semiquantitative R123-based technique used to assess Δψ_m. Carbonyl cyanide m-chlorophenylhydrazone (CCCP), an uncoupler (or protonophore), was used to collapse the proton gradient and inner membrane potential across the mitochondrial inner membrane. Collapse of mitochondrial membrane potential by application of 10 μmol/l CCCP caused an increase in R123 fluorescence due to unquenching and the release of R123 from the mitochondrial matrix into the cytosol. This elevated fluorescence intensity was then normalized to the baseline level of fluorescence before CCCP treatment (F₀).

Real-time NAD(P)H autofluorescence [a relative indicator of NAD(P)H redox state] was measured at an excitation wavelength of 340 nm, and autofluorescence was collected at 510 ± 15 nm from single neurons. Fluorescence changes were expressed as F/F₀, where F is the autofluorescence at any given time and F₀ is the initial autofluorescence. ΔF/F₀ CCCP is the maximal response of fluorescence due to the addition of CCCP and is used to semiquantify the relative NAD(P)H level associated with the mitochondria of a single neuron. There are significant limitations to this technique. For example, while the major autofluorescence signal derives from NAD/NADH associated with the mitochondria (as much as 80–90% of the total fluorescence) (33), there is the possibility of alterations in NADP/NADPH levels in the cytoplasm impacting the fluorescence level.

Cultured DRG neurons were harvested with homogenization buffer (50 mmol/l Tris-HCl, pH 7.4, 1% NP-40, 0.25% sodium deoxycholate; 150 mmol/l NaCl, 1 mmol/l EGTA, 1 mmol/l phenylmethylsulfonyl fluoride, 1 μmol/ml each aprotinin, leupeptin, and pepstatin, 1 mmol/l Na₃VO₄, and 1 mmol/l NaF), and 30 μl Laemmli sample buffer was added. Total protein (5 μg) was subjected to SDS-PAGE and then transferred to nitrocellulose, which was then incubated with primary antibody (anti-insulin receptor-α, 1:500 [Santa Cruz Biotech] and anti-insulin receptor-β, 1:250 [BD Transduction]) in 3% milk PBS overnight at 4°C, followed by incubation of secondary antibody (horseradish peroxidase-conjugated anti-rabbit secondary antibody, 1:2,500 [Cell Signaling]) at room temperature for 1 h. Enhanced chemiluminescence (LumiGlu; Cell Signaling) was used to detect the signal from the blot.

For immunocytochemistry, cultured DRG neurons were fixed with ice-cold 4% paraformaldehyde (in 0.1 mol/l phosphate buffer, pH 7.4) for 30 min at 4°C followed by permeabilizing with 1% Triton X-100 in PBS for 30 min. Nonspecific binding was blocked by the incubation of 10% donkey serum for 1 h at room temperature and washed three times with PBS for 15 min. Fixed cells were then incubated with anti-insulin receptor-β antibody (1:100; BD Transduction) overnight in a humidified chamber followed by fluorescein isothiocyanate (FITC)-conjugated secondary antibody (1:200) for 1 h at room temperature. Fluorescence was examined using Leica fluorescence microscopy equipped with FITC and DAPI filters.

Where appropriate, data were subjected to one-way ANOVA using the Statistical Package for Social Scientists (SPSS/PC+; SPSS, Chicago, IL). Where the F ratio gave P < 0.05, comparisons between individual group means were made by Scheffe’s multiple range test at a significance level of P = 0.05. In all other situations, a standard Student’s t test was used for single comparisons.

We first confirmed that cultured adult sensory neurons expressed receptors for insulin. Figure 2A shows that cultured DRG neurons express insulin receptor (IR) subunits-α (IRα) and -β (IRβ). The molecular weight of the IRα subunit (∼135 kDa) confirms that this receptor is of the peripheral type, with a higher molecular weight than the corresponding receptor found in the adult brain. Immunocytochemical labeling for IRβ confirmed a neuronal localization with strong staining of the plasma membrane (Fig. 2B).

Dissociated adult rat DRG sensory neurons from normal animals were cultured for up to 24 h with insulin, and the effect on Δψ_m was analyzed using R123 fluorescence and the CCCP-dependent collapse of the mitochondrial membrane potential (Fig. 1A for method). Acute treatment with 10 nmol/l insulin (≤20 min) had no effect on R123 fluorescence intensity (normalized) (Fig. 1B). Treatment with 1.0 nmol/l insulin for 6 or 24 h significantly raised the Δψ_m (as determined using the normalized amplitude of CCCP-induced increase in R123 fluorescence) by ∼50 and 100%, respectively (P < 0.05 for both; Fig. 1C). In all insulin treatment studies there were no changes in neuron survival during the period of study (in the presence or absence of insulin, data not shown).

In an attempt to mimic hyperglycemia, the cultures were treated with a high concentration of glucose (50 mmol/l). Acute treatment with 50 mmol/l glucose added to standard physiological saline for ≤20 min had only a fast transient effect on R123 fluorescence intensity and NAD(P)H autofluorescence levels (Fig. 3A and B). Over a period of 2–3 min, the glucose treatment caused a transient depolarization of the mitochondrial membrane that coincided with a similarly transient lowering of the NAD(P)H autofluorescence signal. The latter was most likely due to oxidation of NADH within the mitochondrion, but oxidation of NADPH within the cytoplasm cannot be ruled out. When we treated neurons with 50 mmol/l glucose with adjustment for alterations in osmolarity, no effect on R123 fluorescence was observed for ≤6 h during the continuous application of high glucose-containing solution (data not shown). Therefore, we conclude that rapid mitochondrial depolarization caused by constant treatment with 50 mmol/l glucose within the normal bathing saline solution resulted in hyperosmotic stress. Cultures were then continually treated with 50 mmol/l glucose in the presence or absence of 1.0 nmol/l insulin for 1 day (Fig. 3D). Constant treatment with insulin caused a significant 100% increase in Δψ_m, while 50 mmol/l glucose was without effect.

The results in Figs. 1 and 3 suggested that insulin, but not glucose, was a key regulator of mitochondrial function in vitro. An in vivo experiment was therefore designed in which STZ-induced diabetic rats were treated with suboptimal insulin concentrations that provided background insulin, albeit at a dose that did not affect hyperglycemia (Fig. 4A describes experimental design). A subgroup of STZ-induced diabetic rats received one-half (∼1 unit/day) of an insulin implant (Linplant; Linshin) from week 1 onward. Figures 4B and C show that the insulin therapy raised the total body weight of STZ-induced diabetic animals by ∼50 g over the 8-week duration of the study. At study end, the insulin-treated group had body weights ∼150 g lower than those of the age-matched control animals. The insulin therapy regimen had no effect on blood glucose levels in STZ-induced diabetic rats, as measured using test strips (all STZ-induced diabetic and insulin-treated animals showed >25 mmol/l glucose; data not shown). Additionally, insulin treatment had no significant effect on the raised glycated hemoglobin levels in STZ-induced diabetic rats (Fig. 5A). Sciatic nerve polyol levels, including glucose, fructose, and sorbitol were all raised significantly in the STZ-induced diabetic group (Fig. 5C– E). Insulin therapy did not reduce the polyol accumulation; in fact, there was evidence that glucose and sorbitol levels were raised (Fig. 5C and D). For example, insulin treatment increased glucose levels in nerve to 41.6 ± 4.4 nmol/mg compared with 31.5 ± 9.6 in untreated STZ-induced diabetic animals (P < 0.05). Furthermore, insulin treatment approximately doubled sorbitol levels in nerve from 3.47 ± 1.33 nmol/mg in STZ-induced diabetic animals to 6.25 ± 2.33 nmol/mg in the treated group. Myoinositol levels were significantly reduced in the STZ-induced diabetic group, and insulin treatment had no effect (Fig. 5B). Finally, insulin treatment partially, although significantly, normalized the MNCV value (P < 0.02) and almost completely corrected the diabetes-induced deficit in the SNCV (P < 0.02; as measured using the H-reflex) (Fig. 4D and E).

Lumbar DRG from each animal group were acutely isolated and cultured, and Δψ_m was measured using R123 fluorescence as described above. Sensory neurons from STZ-induced diabetic animals had a 50% decrease in the level of the Δψ_m compared with age-matched control animals (Fig. 6A and B). Insulin treatment completely prevented the diabetes-induced deficit in Δψ_m. Calcium homeostasis in neurons is intimately linked to mitochondrial function, therefore resting intracellular calcium was also analyzed; STZ-induced diabetes raised this parameter from ∼100 to 175 nmol/l, and insulin treatment ameliorated this abnormality (Fig. 6C).

The results show that insulin can prevent diabetes-induced deficits in Δψ_m in adult DRG sensory neurons. These findings complement those of Srinivasan et al. (5), who first showed mitochondrial membrane depolarization in sensory neurons of STZ-induced diabetic rats. However, in the present study the effect of insulin in vivo was not mediated through normalization of hyperglycemia because the insulin concentration used was too low to significantly affect glucose metabolism. Measurement of glycated hemoglobin and sciatic nerve polyol levels confirmed that hyperglycemia in STZ-induced diabetic rats was not affected by the background insulin regimen. In vitro experiments on normal cultured DRG sensory neurons treated with 50 mmol/l glucose also showed no direct effect on Δψ_m, whereas treatment with insulin significantly raised the Δψ_m by 100% (Figs. 1 and 3).

In the present study, the failure of glucose to induce mitochondrial depolarization in cultured adult sensory neurons contradicts a recent investigation (34). There are a number of methodological discrepancies that may account for this. First, Russell et al. (34) used embryonic sensory neurons that have different neurotrophic requirements from the adult neurons used in the present study. Second, Russell et al. used a fluorimetric analysis of cell suspensions, which does not identify neuronal-specific signals, whereas our technique allows detection of neuron-specific signals. Their use of tetramethylrhodamine methylester may also impede interpretation, because when illuminated, tetramethylrhodamine methylester triggers opening of the permeability transition pore, generation of ROS, and release of apoptotic signals (35). In contrast, we used R123, which does not have this limitation under the conditions used in the present study. Finally, data obtained by the confocal imaging technique used by Russell et al. is also difficult to interpret because any alterations in plasma membrane potential, possibly induced by glucose, can impact the fluorescence signal emitted from the mitochondria (31). This is why we used single neuron imaging at the light microscope level. Any or all of the above might account for the discrepancies between the two studies and give us confidence in our methodology and findings.

The ability of insulin to prevent deficits in Δψ_m in sensory neurons in diabetes does suggest a role for this neurotrophic factor in the etiology of the mitochondrial depolarization observed; however, some caution must be exercised before discounting hyperglycemia completely. The in vitro studies only involved short-term exposure of neurons to high glucose concentrations. The study by Nishikawa et al. (3), which demonstrated high glucose-induced ROS production linked to excessive oxidative phosphorylation, involved the in vitro exposure of endothelial cells for at least 1 week. Unfortunately, such studies cannot be performed in adult sensory neurons because of the involvement of proliferating non-neurons at longer time points in culture. Therefore, a role for hyperglycemia in the etiology of mitochondrial dysfunction and cellular degeneration remains a possibility; however, what is clear is that insulin-dependent neurotrophic effects may also be involved in the pathogenesis and are certainly capable of inducing repair of mitochondrial abnormalities.

Circulating insulin is obviously downregulated in type 1 diabetes, and it has been determined that background levels fall significantly <1–4 nm in STZ-induced diabetic animals and in humans (16). DRG sensory neurons express receptors for insulin (14,15) (Fig. 2), and insulin does regulate neuronal phenotype (16). For example, insulin elevates axonal outgrowth in embryonic and adult sensory neurons (10,11), and in clonal cell lines, neurofilament and tubulin mRNA expression are enhanced in association with increased neurite elongation (12,13). Insulin mediates survival responses in neurons and non-neurons, in part, through phosphoinositide (PI) 3-kinase-dependent regulation of protein kinase B (PKB or Akt) with associated modulation of mitochondrial function (23–25). PKB promotes neuronal cell survival by growth factors against several apoptotic stimuli through modulation of Bcl-2 protein function (24–26). In STZ-induced diabetic rats, impaired mitochondrial function is associated with reduced Bcl-2 expression (5), and in Schwann cells (26) exposed to high glucose the resulting degeneration is linked to impaired functioning of Bcl-xL. In Schwann cells, overexpression of Bcl-xL afforded protection to high glucose, as did treatment with IGF-I operating through a PI 3-kinase-dependent pathway (26). These important regulators of mammalian apoptosis function, in part, by altering mitochondrial function, through changes in permeability transition pore function (1). Furthermore, loss of neurotrophic support in neurons induces translocation of Bim and Bax to the mitochondrial outer membrane and is a key pro-apoptotic event in embryonic and mature neurons involving mitochondrial membrane depolarization and cytochrome C release (27,28). Thus, we hypothesize that Bax is one possible target for insulin signaling, and that its PI 3-kinase/Akt-regulated translocation may regulate Δψ_m in the absence of an effect on cell survival, since significant levels of sensory neuron cell death are not observed in cell culture in the absence of insulin (11) and/or presence of high glucose concentrations (36). Furthermore, in acute STZ-induced diabetes (2–3 months) there is no evidence of neuron loss in DRG (37), and this corresponds with maintenance of normal DRG neuron numbers in postmortem samples from patients with diabetes (38).

Apart from putative changes in expression/localization of Bcl-2 family proteins, the insulin-mediated effect on mitochondrial membrane potential could be regulated by enhanced glucose transport, glucose metabolism, or alterations in electron transfer chain redox state. Insulin-dependent changes in glucose transport are unlikely because this process is not limiting in sensory neurons (39). This may be true for neurons, however the polyol data showing enhanced glucose, sorbitol, and fructose levels in the sciatic nerve (Fig. 5) suggest that the background insulin treatment may trigger enhanced glucose utilization in Schwann cells and other non-neurons located in the nerve fiber of STZ-induced diabetic rats. In addition, insulin is known to regulate glucose metabolism at both the posttranslational and transcriptional levels in liver, muscle, and fat tissues (40). At the posttranslational level, insulin directly controls the activities of a set of metabolic enzymes involved in gluconeogenesis and glycolysis (41). Insulin increases transcription of glycolytic enzymes such as glucokinase, pyruvate kinase, and acetyl-CoA carboxylase (42). These phenotype changes may have a significant impact on the levels of redox equivalents [e.g., NAD(P)H] that might directly affect the Δψ_m through an increase in electron donation. Also, modulation of mitochondrial function can also be mediated through alterations in the operation of the electron transfer chain. For example, the level of cytochrome C (a critical and possibly rate-limiting component within the electron transfer chain) can also be regulated through the cAMP response element binding protein (CREB) transcription factor upon serum stimulation (43,44). In fact, our preliminary data show that insulin activates CREB in cultured DRG neurons (T.-J.H., A.V., P.F., manuscript in preparation). Intriguingly, one-third of CREB-regulated genes (genes that contain CREB binding sites) are glucose metabolism related (including hexokinase and cytochrome C) (45).

Diabetes-induced alterations in calcium homeostasis may also underlie abnormalities in mitochondrial polarization status. Mitochondrial function is intimately associated with intracellular calcium homeostasis and calcium signaling. First, mitochondria provide ATP for numerous Ca²⁺ pumps responsible for regulation of resting [Ca²⁺]_i and Ca²⁺ removal from the cytoplasm (46). Second, mitochondria act as a dynamic calcium store, capable of buffering substantial Ca²⁺ loads and release accumulated Ca²⁺, thus shaping cytoplasmic Ca²⁺ signals (47). In fact, in peripheral nerve of 5-month-old STZ-induced diabetic rats, the levels of Ca²⁺ in axoplasm and in mitochondria were significantly raised, as measured using electron probe X-ray microanalysis (48). Both ATP production and mitochondrial Ca²⁺ accumulation are directly controlled by Δψ_m, and chronic mitochondrial depolarization may consequently affect Ca²⁺ homeostatic cascades. As we have demonstrated previously (30), STZ-induced diabetes considerably destabilizes Ca²⁺ homeostasis and Ca²⁺ signaling in sensory neurons, but most importantly, in vivo treatment with background insulin fully restores resting [Ca²⁺]_i (Fig. 6C). This effect of long-lasting insulin treatment could be extremely important in the context of diabetic neuropathies, as impaired cytosolic Ca²⁺ homeostasis has a well-documented role in various neurodegenerative disorders and neuronal death (49). Furthermore, limited ATP production in diabetes may affect Ca²⁺ homeostasis within the endoplasmic reticulum, which may in turn significantly contribute to the development of neuropathic changes (50).

In vitro sensory neurons are quiescent but do not die in the absence of insulin, while enhanced insulin signaling could optimize mitochondrial function and, presumably, increase ATP production, thereby maintaining synthetic processes. The depolarization of the neuronal mitochondrial membrane in diabetes will result in a corresponding decrease in the driving force for ATP production. Energy status in the sciatic nerve of STZ-induced diabetic rats is compromised (51). Obrosova et al. (51) have shown that although levels of ATP were not significantly altered (a 20% decrease was observed) in the sciatic nerve of STZ-induced diabetic rats, there was a significant reduction in the levels of stored high-energy phosphates, such as phosphocreatine (51). Diabetic neuropathy is characterized by a range of alterations in sensory neuron gene expression and an impaired regenerative response in the event of peripheral axon damage (9). Additionally, slow axonal transport of a range of axoplasmic materials, including neurofilaments, is impaired (52,53). All of these aspects of neuropathic pathology have a common requirement for energy, mainly in the form of ATP, and clearly, if mitochondrial function is impaired, considerable detriment to the neuron will result.

In summary, the results show that a lack of insulin-dependent neurotrophic support may underlie mitochondrial dysfunction in sensory neurons in diabetic neuropathy. While we cannot be certain of an involvement of loss of insulin in the etiology of mitochondrial dysfunction, the results show that therapy with insulin can correct diabetes-induced disturbances in mitochondrial function. It can be proposed that such a decrease in mitochondrial membrane polarization will impact a plethora of sensory neuron functions through a decrease in available energy (ATP and/or other stores) and may be a critical component of the etiology of diabetic sensory polyneuropathy.

This work was funded by grants from the Juvenile Diabetes Research Foundation (to P.F.), from Diabetes U.K. (to P.F. and A.V.), and from the Medical Research Council (to D.R.T.). T.-J.H. was a recipient of an Overseas Research Scholarship.