Uncoupling Proteins Prevent Glucose-Induced Neuronal Oxidative Stress and Programmed Cell Death

  1. Andrea M. Vincent1,
  2. James A. Olzmann1,
  3. Michael Brownlee2,
  4. W.I. Sivitz3 and
  5. James W. Russell14
  1. 1Department of Neurology, University of Michigan, Ann Arbor, Michigan
  2. 2Department of Medicine, Diabetes Research Center, Albert Einstein College of Medicine, Bronx, New York
  3. 3Department of Internal Medicine, University of Iowa Hospitals and Clinics, Iowa City, Iowa
  4. 4Geriatric Research, Education and Clinical Center (GRECC), Ann Arbor Veterans Administration Medical Center, Ann Arbor, Michigan
  1. Address correspondence and reprint requests to James W. Russell, MD, MS, University of Michigan, Department of Neurology, 200 Zina Pitcher Pl., 4410 Kresge III, Ann Arbor, MI 48109-0588. E-mail: jruss{at}umich.edu


The central role of mitochondria in most pathways leading to programmed cell death (PCD) has focused our investigations into the mechanisms of glucose-induced neuronal degeneration. It has been postulated that hyperglycemic neuronal injury results from mitochondria membrane hyperpolarization and reactive oxygen species formation. The present study not only provides further evidence to support our model of glucose-induced PCD but also demonstrates a potent ability for uncoupling proteins (UCPs) to prevent this process. Dorsal root ganglion (DRG) neurons were screened for UCP expression by Western blotting and immunocytochemistry. The abilities of individual UCPs to prevent hyperglycemic PCD were assessed by adenovirus-mediated overexpression of UCP1 and UCP3. Interestingly, UCP3 is expressed not only in muscle, but also in DRG neurons under control conditions. UCP3 expression is rapidly downregulated by hyperglycemia in diabetic rats and by high glucose in cultured neurons. Overexpression of UCPs prevents glucose-induced transient mitochondrial membrane hyperpolarization, reactive oxygen species formation, and induction of PCD. The loss of UCP3 in DRG neurons may represent a significant contributing factor in glucose-induced injury. Furthermore, the ability to prevent UCP3 downregulation or to reproduce the uncoupling response in DRG neurons constitutes promising novel approaches to avert diabetic complications such as neuropathy.

Diabetic neuropathy is the most common neuropathy in the western world. Although several etiologies have been proposed for this disease, an association has been found between hyperglycemia and dorsal root ganglion (DRG) injury in both cell culture and animal models of diabetes (16).

One unifying model of injury lies in the ability of high glucose to both enhance production as well as decrease scavenging of reactive oxygen species, leading to cellular oxidative stress and impaired mitochondrial function (1,710). Consistent with this view, blocking oxidative stress in the diabetic animal prevents the development of neuropathy (11) and restores sciatic and saphenous nerve conduction velocities in streptozotocin (STZ)-induced diabetic rats (7,10,12). Excess glucose leads to an oversupply of electrons in the mitochondrial transfer chain that results in mitochondrial membrane hyperpolarization and the formation of reactive oxygen species (ROS) (13,14). These free radicals damage proteins and lipids and lead to dysfunction of mitochondria, the central mediators of programmed cell death (PCD) (2,15,16). Damaged mitochondria release proapoptotic factors that activate the cysteine protease family of caspases, which in turn propagate a death cascade (15,17,18).

Cellular energy and mitochondrial membrane function in response to glucose are regulated in part by a group of uncoupling proteins (UCPs). Members of this family of inner mitochondrial membrane proteins all function as proton carriers and can disperse the mitochondrial proton gradient, bypassing the production of ATP by oxidative phosphorylation (19,20). Recent investigations (13,21) have demonstrated that UCPs can prevent mitochondrial ROS formation. The different members of the UCP family have distinct tissue distributions (19,22,23). Tissue localization as well as regulation confer different roles for the family members. For example, UCP1 is strictly restricted to brown adipose tissue, and its principal function is believed to be the generation of body heat (19,24). UCP3 tissue distribution is predominantly limited to skeletal muscle, whereas UCP2 expression is widespread in spleen, lung, stomach, white adipose tissue, and brain (23,25). Both UCPs regulate energy metabolism and weight regulation (19,20,26).

The current study investigates the expression of UCPs in DRG neurons as well as the function of UCPs in glucose-mediated PCD. In a novel finding, DRG neurons express UCP3, which is rapidly lost in both animal and tissue culture models of diabetes. Overexpression of UCPs in cultured neurons blocks glucose-induced PCD by preventing mitochondrial hyperpolarization and formation of ROS. These results define a central role for UCP3 regulation of mitochondrial membrane hyperpolarization and ROS formation in glucose-mediated neuronal injury.


DRG culture and STZ animal models of hyperglycemia.

DRG neurons from embryonic day 15 Sprague-Dawley rats were cultured as previously described (1,2). Under these culture conditions almost all Schwann cells, satellite cells, and fibroblasts are removed and pure neuronal cultures are obtained. Basal glucose in culture medium is 25 mmol/l, which is necessary for DRG neuron survival (1,2). To produce a hyperglycemic insult, 20 mmol/l additional glucose (total 45 mmol/l glucose) was added to the media for the period specified in individual experiments. Sprague-Dawley rats were made diabetic with STZ and maintained under standard conditions for diet and environment as previously described (1).

Western blotting for UCP expression.

Equal amounts of protein (40 μg, unless stated otherwise) were loaded onto a 12.5% polyacrylamide gel and incubated with 5 μg/ml primary antibody overnight at 4°C and then secondary antibody (goat anti-rabbit horseradish peroxidase) (1:3,000; Santa Cruz Biotech, Santa Cruz, CA) for 1 h at room temperature. Blots were subsequently stripped and reprobed using an antibody against the mitochondrial protein cytochrome c oxidase IV (COX-IV; Molecular Probes, Eugene, OR) to correct any differences in gel loading (at least three experiments per condition). Specific rabbit polyclonal antibodies against UCPs 1–3 (RDI, Flanders, NJ) were tested for specificity against 1 μg/μl control peptide.

Localization of UCPs using confocal microscopy.

DRG neurons were fixed, stained, and viewed by confocal microscopy using dual excitation and emission in the red and green channels as described (2). Primary antibodies were UCPs and COX-I (RDI). Secondary antibodies were donkey anti-rabbit Cy3 (1:500) for UCPs and donkey anti-goat fluorescein isothiocyanate (1:500) for COX-I. Experiments were repeated in quadruplicate.

Gene transfer using adenoviral constructs.

Adenovirus (Ad) constructs containing cDNA for UCP1, UCP3, oxoglutarate malate carrier (OMC), or green fluorescent protein (GFP) were prepared as previously described (27,28) and purified at the University of Iowa Gene Transfer Vector Core. Ad.OMC was kindly provided by Dr. Thomas Scholz, University of Iowa (28). At 24 h, 95–100% of DRG neurons was infected using a multiplicity of infection of 1,000, determined by counting GFP-positive neurons.

Caspase-3 and TdT-mediated dUTP-biotin nick-end labeling staining.

Staining for caspase-3 activation was performed as previously described with 0.1 μg/ml anti-cleaved caspase-3 (Pharmingen, San Diego, CA) and 7.5 μg/ml Texas Red conjugated goat anti-rabbit antibody (Molecular Probes, Eugene, OR) for 1 h as previously described (1,2) and was performed on duplicate samples on three distinct occasions. The nuclear chromatin was counterstained with 1 μg/ml bisbenzamide in PBS. The specificity of the antibody was determined by immunoabsorbing against a specific recombinant peptide of cleaved caspase-3. TdT-mediated dUTP-biotin nick-end labeling (TUNEL) staining was performed using specific positive and negative controls as previously described (1).

Measurement of mitochondrial membrane potential and ROS formation.

The degree of polarization of the mitochondria was determined in real time in DRGs by loading with tetramethylrhodamine (TMRM; Molecular Probes, Eugene, OR). Accumulation of TMRM in mitochondria is directly dependent on the polarization state of the mitochondrial membrane (29). Each culture dish was incubated with 50 nmol/l TMRM for 15 min at 37°C and then rinsed with Hank’s balanced salt solution (HBSS) (10 mmol/l HEPES, pH 7.4, 150 mmol/l NaCl, 5 mmol/l KCl, 1 mmol/l MgCl2, 1.8 mmol/l CaCl2). The concentration and loading time of TMRM was selected following experiments using the uncoupling agent FCCP [carbonylcyanide p-(trifluoromethoxy)phenylhydrazone] (2.5 μmol/l) or the hyperpolarizing agent oligomycin (1 μmol/l). FCCP induced depolarization of mitochondria within 2 min, and oligomycin induced rapid and prolonged hyperpolarization under the selected loading conditions. Dishes were immediately placed in a fluorimeter (Fluroskan Ascent II; Labsystems, Helsinki, Finland), and the absorbance of TMRM was determined with 485-nm excitation and 590-nm emission. Results from five different experiments with each experimental paradigm performed in duplicate were pooled for statistical analysis.

Oxidative stress was measured with cell-permeant CM-H2DCFDA (chloromethyl-dichlorodihydrofluorescein diacetate) (Molecular Probes) in DMSO at a final dilution of <0.1% to prevent toxic or oxidative effects. Neuronal cultures were incubated in a final concentration of 2 μmol/l CM-H2DCFDA at the start of glucose administration to measure any transient increases in ROS production. Measurement of ROS was performed using confocal microscopy as previously described (2). The mitochondria pixel intensity for 100 individual neurons was averaged in each experimental paradigm and the background fluorescence subtracted.

The cell permeant dihydroethidium (DHE) (Molecular Probes) was used to assess real-time formation of superoxide in DRG neurons. DHE was dissolved at a concentration of 10 mg/ml in DMSO, then diluted to 3 μmol/l in HBSS immediately before use. The diluted DHE solution was applied to DRG neurons for 30 min, and then neurons were rinsed once rapidly in HBSS and immediately placed in a fluorimeter (Fluroskan Ascent II). For each sample, the ratio of red (518 nm excitation and 605 nm emission) over blue (485 excitation and 520 emission) fluorescence was determined with a 1-s integration to assess real-time formation of superoxide in DRG neurons as previously described (2). Results were obtained from duplicate samples in five different experiments.

Statistical analysis.

Assumptions about the Gaussian distribution of data and rules for transformation of nonnormative data were made as previously described (1,2). Comparison of dependent variables was performed using factorial ANOVA with 95% CIs. An observer blinded to the experimental condition made measurements. Bar graphs illustrate the means ± SE.


Rat DRG neurons express UCP3.

Specific antibodies against UCP1, UCP2, and UCP3 were used to assess UCP expression in E15 DRGs by Western blotting (Fig. 1A). There was significant expression of UCP3 protein in DRG neurons, which is of interest because UCP3 is reported to be more prevalent in skeletal muscle and brown adipose tissue (30,31). UCP3 was at least as abundant in DRG neurons as in skeletal muscle preparations. Consistent with previous work that demonstrates that UCP1 is restricted to brown adipose tissue (19), UCP1 was not detected in DRG neurons. The expression of UCP2 was low and not consistently detected in DRG neurons, which is in contrast to high UCP expression in brain, as previously reported (22) and shown in Fig. 1, lane 5.

The specificity of the UCP3 antibody used was confirmed by the preincubation of the antibody with the immunogenic peptides or with a random peptide (Fig. 1B). Binding of the antibody to its target was specifically prevented following preabsorption by the immunogenic peptide but not by another random peptide.

UCP3 in DRG neurons is reduced in diabetic animals.

Since UCP3 is abundantly expressed in DRG neurons, we postulated that a reduction of UCP3 may result in loss of regulation of the mitochondrial membrane potential (ΔψM) and prevent diabetes-induced neuronal PCD in vivo. To determine whether UCP3 expression is altered by diabetes, DRGs were isolated from eight control rats and eight 4-week STZ-induced diabetic rats (Fig. 2A). UCP expression was analyzed by Western blotting. Five whole DRGs from each animal were solubilized and loaded in one lane of the polyacrylamide gel. Mean serum glucose concentrations were 4.2 ± 0.15 in control and 18.0 ± 0.59 mmol/l in diabetic rats. Similar to the expression of UCP3 in cultured E15 DRG neurons, UCP3 was expressed in the DRG neurons from adult control rats. In each diabetic animal, the UCP3 band was almost completely lost (Fig. 2A). Densitometric measurement of the bands from all of the animals demonstrates that UCP3 is significantly downregulated in the 4-week diabetic animals compared with the weight-matched controls. Similar to E15 DRGs, UCP1 was not detected and UCP2 was difficult to detect in adult control rat DRGs (data not shown). Figure 2B illustrates a Western blot for UCP3 in DRG neurons that were exposed to 20 mmol/l added glucose for increasing periods of time. The expression of UCP3 was decreased within 3 h following the addition of 20 mmol/l glucose (P < 0.01). UCP3 continued to decrease and was not re-expressed at any time during the 24-h course of the experiment. Densitometric measurements are expressed as a ratio against the mitochondrial protein, COX-IV (Fig. 2B, bar graph). To confirm that the effect of UCP3 downregulation is specific for hyperglycemic injury and not a hyperosmotic shock, the nonglycolytic control o-methyl glucopyranose (20 mmol/l) (OMG) was also applied to DRG neuron cultures for 12 h and then immunoblotted for UCP3 (Fig. 2C). Again UCP3 expression was normalized against COX-IV expression, and there was no significant difference between untreated control cultures and those exposed to 20 mmol/l added OMG (Fig. 2C, bar graph).

Native and exogenous UCPs are expressed in the mitochondria.

To determine whether native and overexpressed UCPs function only at the level of the mitochondria and not at a cytosolic level, the subcellular localization of the UCPs in DRGs was examined using double staining for native or adenoviral UCP with the mitochondrial protein COX-I. Figure 3 demonstrates that native UCP3 (Fig. 3A), adenoviral UCP1 (Fig. 3B), or adenoviral UCP3 (Fig. 3C) are all strictly coexpressed with COX-I in the DRG neuronal mitochondria, suggesting that the cellular activities of these proteins are affected at the mitochondrial level. Interestingly, neither UCP2 nor UCP3 was convincingly expressed in pure cultured Schwann cells.

Overexpression of UCPs prevents glucose-mediated neuronal PCD.

The loss of UCP3 in DRG neurons after exposure to glucose at concentrations known to mediate PCD suggests a pivotal role for UCP3 in DRG survival. Thus, overexpression of UCP3 may in turn prevent glucose-induced PCD. To test this hypothesis, cultured DRG neurons were infected with adenoviral constructs containing the full-length sequence of rat UCP1 or UCP3 or GFP as a control. Development of PCD was then assessed in DRG neurons overexpressing UCP1, UCP3, or GFP in the presence of 20 mmol/l added glucose (Fig. 4). Figure 4A illustrates typical images of DRG neurons stained for caspase-3 cleavage. At the 6-h period following the addition of 20 mmol/l extra glucose, caspase-3 cleavage is evident in >60% of neurons. UCP3 expression blocks glucose-mediated caspase-3 cleavage. Caspase-3 cleavage was quantitated by counting the proportion of stained neurons at 6 h (Fig. 4B). In control conditions with GFP alone, caspase-3 activation increased from 14 ± 3% in basal glucose to 69 ± 5% 24 h after adding 20 mmol/l glucose. The application of the nonglycolytic osmotic control, OMG, did not increase the percentage of neurons with cleaved caspase-3 compared with that of untreated control cells. The overexpression of UCP1 or UCP3 slightly, but not significantly, decreased caspase cleavage in basal glucose from 14 ± 3% to ∼11%. The degree of caspase-3 cleavage was significantly decreased in UCP1- and UCP3-overexpressing neurons compared with neurons in high glucose alone (Fig. 4B).

The neuroprotective effects of UCPs were also examined using TUNEL staining (Fig. 4C). Similar to previous studies (1), the addition of 20 mmol/l extra glucose increased evidence of DNA fragmentation in control (GFP-expressing) cultures from 17 ± 3 to 63 ± 12% over 24 h, but 20 mmol/l OMG had no effect. The overexpression of either UCP1 or UCP3 decreased glucose-induced DNA fragmentation compared with neurons in high glucose alone (Fig. 4C).

UCPs prevent hyperglycemia-induced hyperpolarization of the mitochondrial membrane.

Having demonstrated that UCPs are expressed in DRG mitochondria and can prevent glucose-mediated PCD, we then asked whether UCPs prevent mitochondrial damage by decreasing glucose-induced mitochondrial membrane hyperpolarization. Changes in the ΔψM were measured using fluorescent TMRM (29) (Fig. 5). The validity of this method of determining a change in the ΔψM at the concentrations used has been previously established using specific controls (2). The changes in ΔψM in response to 20 mmol/l added glucose or the nonglycolytic and osmotic control OMG were determined over a 24-h period (Fig. 5A). There was a rapid increase in ΔψM that peaked ∼3 h following exposure to 20 mmol/l added glucose (P < 0.01 compared with basal fluorescence). The ΔψM then decreased and by 24 h was lower than that of the untreated control neurons. This increase in ΔψM was specific for the metabolism of glucose, and there was no significant change in ΔψM with 20 mmol/l OMG.

Next, the abilities of UCPs to prevent mitochondrial hyperpolarization were assessed by comparing ΔψM in control DRG neurons or neurons infected with Ad.GFP, Ad.UCP1, Ad.UCP3, or Ad.OMC in basal glucose conditions or at 3 or 6 h following the addition of 20 mmol/l glucose (Fig. 5B). Overexpression of UCP1 or UCP3 completely prevented the peak in ΔψM at the 3-h time point observed in control or mitochondria-targeted Ad.OMC-infected DRG neurons. Although TMRM levels were slightly lower with UCP3 overexpression at 6 h compared with control at 0 h, this was not significantly different.

UCPs prevent the production of glucose-induced ROS formation.

One consequence of an increase in ΔψM is the production of ROS. In particular, stasis of the electron transport chain prolongs the lifetime of superoxide-generating electron-transport intermediates such as ubisemiquinone (13,14). When the ΔψM exceeds a threshold value that can be produced in hyperglycemia (32), superoxide production is markedly increased. This would lead to mitochondrial damage and the development of PCD. To further support our assertion that glucose-induced PCD is generated through this mechanism and that UCPs may prevent this process, the production of ROS in DRGs was measured during high-glucose injury.

Initial investigations employed CM-H2DCFDA, which produces a green fluorescent derivative (2′,7′-dichlorofluorescein, [DCF]) following oxidation by free radicals. Preliminary experiments using this probe demonstrate that the application of 20 mmol/l extra glucose causes a rapid burst of ROS formation in the mitochondria within 1 h, followed by a more sustained peak between 4 and 6 h (Fig. 6A). The application of an agent that is known to cause ROS production (antimycin A, 2.5 μmol/l) caused a threefold increase in DCF fluorescence over 1 h (data not shown). Next, the formation of ROS was determined in basal glucose conditions as well as 5 h following the application of 20 mmol/l extra glucose in control DRGs overexpressing the mitochondrial carrier OMC or overexpressing UCP1 or UCP3. Production of ROS in basal glucose was slightly decreased in UCP1-overexpressing DRG neurons compared with that of control OMC-infected cells (Fig. 6B) (P < 0.05). The peak of ROS formation following a 20-mmol/l increase in glucose concentration was observed in OMC-infected neurons but was completely inhibited in the UCP-overexpressing neurons. The ROS production at the 5-h period in UCP1- and UCP3-expressing neurons was significantly lower than that in the basal levels in control cells (P < 0.05), demonstrating that UCPs provide powerful protection against ROS formation under high-glucose conditions.

One of the most significant ROS generated at the hyperpolarized mitochondrial membrane is likely to be superoxide (13). Therefore, the probe DHE, which more specifically detects superoxide production (Figs. 6C and D), was also used. The utility of this probe was established in a series of control experiments (2). In particular, the inducer of ROS, antimycin A (2.5 μmol/l), significantly increased DHE oxidation over a 1-h period (Fig. 6C). The nonmetabolizable sugar OMG did not increase ROS above the level of untreated control neurons. Application of 20 mmol/l extra glucose increased superoxide to a similar level as antimycin A over the course of 1 h. A time course of glucose-induced superoxide formation was then determined (Fig. 6D). DRG neurons were infected with either the control virus containing OMC or UCP3. Neurons containing OMC produced the same results as uninfected control neurons, and only the OMC data are shown (Fig. 6D). In the presence of 20 mmol/l added glucose, superoxide increased compared with basal glucose in the Ad.OMC-infected cells, reaching significance within 3 h (P < 0.05), the same time point as the peak ΔψM (Fig. 5A). Superoxide formation continued to increase up to a peak at the 4-h time point before decreasing toward basal levels, corresponding temporally with mitochondrial membrane depolarization. In neurons infected with UCP3, superoxide production was significantly lower than in OMC control neurons up to 6 h. At this point, glucose-mediated superoxide declined to baseline levels (Fig. 6D) (P < 0.05). It is interesting to note that basal superoxide formation is decreased in UCP3-overexpressing cells compared with that of controls, but this level increases moderately following the application of 20 mmol/l glucose. Overexpression of UCP3 may increase the efficiency of electron transfer by carrying excess protons across the membrane when the electron acceptors are working at an optimum rate in the presence of adequate glucose. The process becomes less efficient as the glucose supply increases and electron transport is accelerated. More electrons are likely to escape from the transfer chain as the rate of flow increases, producing more superoxide. Superoxide formation in UCP1-overexpressing DRG neurons was assessed in a similar manner, and UCP1 also prevented glucose-induced superoxide formation, similar to UCP3 (data not shown).


Both animal and cell culture models demonstrate some evidence of neuronal PCD in primary sensory neurons (1,2,46) and may play a role in the development of diabetic neuropathy. In both cell culture and diabetic rats, high glucose induces oxidative stress and PCD. This likely occurs as a result of changes in the ΔψM. In cell culture, there is an initial hyperpolarization followed by depolarization of the inner ΔψM, which is associated with increased generation of ROS and activation of both upstream and downstream caspases (2). Recent evidence indicates that animal models of chronic diabetes show evidence of depolarization of the inner mitochondrial membrane (33), consistent with terminal changes in vitro. These findings of eventual loss of the ΔψM both in the STZ-induced diabetic rat DRGs (4,33) and after exposure to glucose in vitro (2) are consistent with activation of PCD pathways. In contrast, this study shows that UCPs are able to regulate the ΔψM and reduce mitochondrial hyperpolarization and the subsequent depolarization wave associated with PCD.

The basal concentration of glucose used in vitro reflects the increased energy requirements and metabolic rate of DRGs compared with nonneuronal cells (2). With a total in vitro concentration of 25–30 mmol/l glucose, there is optimal DRG neuronal survival, whereas glucose concentrations <25 mmol/l rapidly induce ATP depletion and an increase in caspase-3 cleavage (1,2). Thus, the culture model is a highly glucose-sensitive system that closely mimics the in vivo findings where both hyperglycemia and hypoglycemia are associated with neuronal and axonal injury. Interestingly, cultured embryonic DRG neurons depleted of Schwann cells used in this study are more susceptible to high-glucose-associated caspase activation than DRG neurons cultured with Schwann cells (34) or mature cultured rat neurons that are surrounded by Schwann-like satellite cells (33). These findings are consistent with protective neurotrophic support of DRG neurons by Schwann cells (34,35). These findings may in part explain differences in susceptibility to neuronal injury between isolated neurons and neuron/Schwann cell cocultures.

Most studies (2,46) to date indicate that there is activation of caspases in DRG neurons both in vitro and in vivo. There is also evidence of neuronal nuclear DNA fragmentation using in vivo studies (1,4,5) with rigorously applied controls showing convincing positive TUNEL staining with DNase. Most studies indicate some loss of DRG neurons (1,36), and in particular there is a statistically significant loss of large DRG neurons (36). In one study (37), using rigorous counting techniques of DRG nuclei in 6–12 pairs of sections from the whole DRG, it was concluded that there was no loss of neurons in the DRG from diabetic animals. In fact, this study showed a 14.5% decrease in the mean number of neurons (determined by nuclei) per ganglion in 12-month diabetic animals (31,629 ± 2,427 neurons) compared with that of control animals (36,986 ± 3,035). However, there was a large variance that may have resulted in inadequate power to detect a statistical difference. Thus, all diabetic animal studies to date show evidence of DRG neuronal loss in addition to any DRG neuronal atrophy that may be present (1,36,37). However, the number of DRG neurons showing evidence of caspase-3 cleavage or TUNEL staining may be greater than the measured loss of neurons, suggesting either that activation of caspases does not invariably result in neuronal death or that there is an intrinsic capacity for repair within the neuron resulting either from DNA repair or by activation of neurotrophic protective signaling pathways (35). DNA repair by poly(ADP-ribose) polymerase-1 is itself a double-edged regulator of cellular survival. When the DNA damage is moderate, poly(ADP-ribose) polymerase-1 participates in the DNA repair process. Conversely, in the case of massive DNA injury, elevated poly(ADP-ribose) polymerase-1 activation leads to rapid and massive NAD(+)/ATP consumption and cell death by necrosis (38). Thus, ultimate neuronal death is a balance between finely regulated but often opposing pathways that likely accounts for less neuronal loss than would be predicted if all caspase-3-positive neurons invariably underwent PCD.

Our current data support the concept that the development of hyperglycemic injury occurs in part through mitochondrial membrane hyperpolarization followed by depolarization. While results using fluorescent membrane-permeable cations have to be interpreted in relation to the specific properties of each probe, several different laboratories have independently shown using different probes that the net effect of hyperglycemia in DRG neurons both in vitro and in vivo is mitochondrial depolarization (2,4,33). With TMRM used in this study, there is rapid equilibration of the fluorescent cation across both the plasma and mitochondrial membranes (29) compared with the less cell-permeant rhodamine123. The mitochondrial matrix concentration will remain above the quench limit and will be unaffected by plasma membrane depolarization, whereas when the ΔψM collapses the subsequent large disequilibrium across the plasma membrane depolarization results in a rapid decay of whole-cell fluorescence (29). Thus, whole-cell recordings with TMRM are a sensitive measure of changes in the ΔψM and unlike isolated mitochondrial recordings, provide evidence of changes in physiological function of mitochondria within the whole-cell environment. Furthermore, there was no evidence of a significant change in control neuronal ΔψM using TMRM over a 24-h period using the methodology in this study.

The increase in the ΔψM will further drive formation of ROS, leading to neuronal injury. The formation of superoxide alone more closely paralleled the development of mitochondrial membrane hyperpolarization and the onset of PCD than that using DCF. In addition, the increase in superoxide was fivefold over baseline compared with a twofold increase in DCF levels. Thus, early superoxide formation may be the major ROS responsible for mitochondrial damage in hyperglycemia. UCPs may provide an important check to generation of ROS in DRG neurons. Since mitochondria are probably the central convergence point for the initiation of PCD (39), the induction of PCD in hyperglycemia may be entirely regulated at the level of mitochondrial function. UCPs also operate at this same level in the mitochondria and prevent PCD upstream of mitochondrial membrane hyperpolarization, ROS formation, and caspase activation. Thus, the ability to regulate mitochondrial uncoupling in the presence of hyperglycemia may be an effective therapeutic strategy against diabetic neuropathy. The novel finding that DRG neurons express UCP3 was determined by using a commercially available specific antibody for Western blotting and cytochemistry. The utility of this antibody was demonstrated in control tissues with known UCP family distributions. The specificity of the UCP antibodies was confirmed using blocking peptides (Fig. 1B) (27) and also by comparison with assays of mRNA expression (27,40). The data obtained using the UCP2 antibody were less convincing. We cannot rule out a potential role for UCP2 in DRG neurons, although clearly expression of UCP2 is greater in cerebral than DRG neurons. This study, therefore, focused on the important role of UCP3 in DRG neurons.

Despite the expression of UCP3 in normal DRG neurons, these cells remain sensitive to hyperglycemic injury through loss of ΔψM regulation and formation of ROS. This process appears to occur through the rapid downregulation of UCP3 protein in DRG neurons during hyperglycemia both in vivo and in vitro. In skeletal muscle and brown adipose tissue, which have previously been shown to express UCP3, the regulation of expression has also been examined. A recent brief report (41) demonstrated that UCP3 is decreased in skeletal muscle in type 2 diabetic patients. Interestingly, the expression of UCP3 protein does not correlate well with mRNA levels (41). One role for UCP3 in muscle may be to shift energy production to the oxidation of fatty acids (26,27). Therefore, it may be reasonable to predict that muscle cells would decrease UCP3 expression during hyperglycemia to transfer energy metabolism from fatty acids to glucose. Yet, neurons do not metabolize fatty acids and depend entirely on glucose to produce ATP, suggesting that the mechanism leading to UCP3 downregulation in DRGs is detrimental to the neurons. This mechanism is unclear at present, and future investigations will examine the release of UCP3 from the mitochondria following mitochondrial swelling (1) as well as the potential for degradation of UCP3 following oxidative damage.

Because UCPs are associated with the regulation of energy metabolism, their role in diabetes has received considerable attention. Mapping of the UCP2 gene to loci associated with obesity and hyperinsulinemia led to investigations into the role of this UCP in weight regulation and energy balance (42,43). It has been shown that UCP2 may be increased in pancreatic β-cells in the pre-diabetic state and that this relates to impaired glucose-induced insulin secretion (44). One mechanism for increased UCP2 in pre-diabetes is the presence of a polymorphism in the UCP2 promoter, which leads to increased expression of the gene (45). In concert with increased UCP2 in β-cells, the levels of UCP3 are decreased in muscle in type 2 diabetes (41). Thus, while high levels of UCP2 may be detrimental because of resultant hyperglycemia, it appears that it is the tissue location of the UCPs that is most important. UCP3 only appeared to be normally expressed in DRG neurons, and overexpression of UCP1 or UCP3 can prevent hyperglycemic injury to these cells. Therapeutic regimens designed to prevent mitochondrial membrane hyperpolarization to decrease hyperglycemic neuronal degeneration may require specific targeting to the neuronal tissues.

One concern with the overexpression studies is that in this system artifactual uncoupling through protein overexpression in the mitochondrial membrane cannot be distinguished from genuine uncoupling activity, as was recently reported (46). The adenoviral construct containing another mitochondrial carrier, OMC, did not prevent hyperglycemia-induced mitochondrial membrane hyperpolarization or ROS generation, supporting the idea that UCPs play a specific role in these processes and that these observations are not simply an artifact of abnormal protein folding or expression in the mitochondrial membrane. The present study demonstrates that the overexpression of UCPs does stabilize the ΔψM in the presence of hyperglycemia and decreases ROS production and PCD.

In summary, we provide further evidence that oxidative stress is associated with initial inner mitochondrial membrane hyperpolarization and that eventual depolarization is associated with neuronal injury. UCPs are able to prevent the sequence of events that leads to glucose-induced neuronal degeneration and do this through stabilization of the ΔψM. In a novel finding, DRG neurons normally express UCP3, and this is rapidly lost in hyperglycemia. While the function of UCP3 in DRG neurons requires further study, we propose that preventing the loss of UCP3 may avert glucose-induced neuronal injury. Furthermore, the potent ability for UCPs to prevent hyperglycemic neuronal injury suggests that pharmacological approaches to regulating the UCPs may provide therapeutic benefit in diabetic neuropathy.

FIG. 1.

Expression of UCPs in DRG neurons. A: Western blots of DRG neurons as well as control tissues were probed with specific antibodies against UCPs. Each lane was loaded with 40 μg protein except in lane 2 (100 μg protein). DRG neurons after 5 days in culture (1), DRG immediately after dissection from E15 rat embryos (2), skeletal muscle (3), brown adipose tissue (4), and brain (5). B: Antibodies were preincubated at room temperature for 3 h with control immunogenic peptides (control), a random peptide (random) of similar size, or no peptide (none) to establish specificity. Immunoabsorbed UCP2 was used to probe blots from rat brain, and immunoabsorbed UCP3 was used to probe samples from DRG neurons. The Western blots for UCP3 were stripped and reprobed for COX-IV to demonstrate equal gel loading.

FIG. 2.

UCP3 is downregulated in diabetic rats and in vitro hyperglycemia. A: DRGs from 4-week STZ-induced diabetic rats. DRG from eight diabetic and eight weight-matched control animals were used, and protein concentration was equalized before loading. Panel A illustrates four control and four diabetic rats on one blot. The bar graph shows the mean UCP densitometry for all 16 animals (diabetic < control animals, P < 0.01). B: DRG neurons were cultured for 3 days in basal glucose and then exposed to 20 mmol/l added glucose for 3, 6, or 12 h. The bar graph indicates that the ratio of the mean pixel densities of the bands decreases with increasing time of exposure to high glucose. *P < 0.01. C: Cultured DRG neurons were exposed to 20 mmol/l added glucose or 20 mmol/l OMG for 12 h and then blotted for UCP3 and COX-IV. †Glucose exposure significantly decreased UCP3 levels (P < 0.001) compared with the nonglycolytic control OMG.

FIG. 3.

Both native and adenoviral UCPs are expressed in the mitochondria. Cultured DRG neurons were double stained with UCP antibody and an antibody against the mitochondrial protein COX-I. Each staining paradigm is illustrated with the green channel for COX-I staining, the red channel for the UCP staining, and an overlay of the red and green with colocalization shown in yellow. A: Non-adenovirus-infected neurons stained for native UCP3. B: Ad.UCP1-infected neurons stained for UCP1. C: Ad.UCP3-infected neurons stained for UCP3.

FIG. 4.

Overexpression of UCP1 or UCP3 prevents glucose-induced PCD. Cultured DRG neurons were infected with adenovirus containing UCP1, UCP3, or GFP for 24 h before the application of 20 mmol/l extra glucose or 20 mmol/l OMG. Neurons were then fixed at 6 h and stained for caspase-3 cleavage or fixed at 24 h and stained for TUNEL. A and B: Caspase-3 staining in control GFP-expressing neurons or in neurons overexpressing UCP3 showing an increase in the number of neurons stained for cleaved caspase-3 in high glucose compared with UCP1- or UCP3-expressing neurons. Using 20 mmol/l OMG as an osmotic control did not alter caspase-3 staining in control or UCP-overexpressing neurons. C: DNA fragmentation using TUNEL. *Neurons with caspase-3 cleavage or TUNEL staining were increased with high glucose in GFP but not UCP1 or UCP3 neurons or in neurons exposed to OMG (P < 0.01).

FIG. 5.

UCP1 and UCP3 prevent glucose-induced mitochondrial membrane hyperpolarization. ΔψM was assessed by loading cultured DRG neurons with TMRM (50 nmol/l) for 15 min. The uptake of TMRM increases with increased ΔψM. A: The time course of changes in ΔψM following the application of 20 mmol/l glucose or 20 mmol/l OMG is shown. In control neurons exposed to high glucose, the ΔψM increased significantly between 1 and 2 h (+P < 0.05), continued to increase up to 3 h, and then decreased to below baseline at the 24-h period. In control neurons exposed to OMG, there were no changes in ΔψM over the time course of the experiment. B: Control neurons or neurons overexpressing UCP1, UCP3, or OMC were exposed to 20 mmol/l added glucose for 0, 3, or 6 h. *The ΔψM was significantly higher in noninfected or OMC-expressing neurons at the 3-h period compared with other time points and also to both UCP1- and UCP3-infected neurons at 3 h (P < 0.01).

FIG. 6.

UCP overexpression prevents ROS formation. A and B: ROS were measured with DCF. ROS production increased significantly within 1 h following the application of 20 mmol/l glucose and remained elevated (*P < 0.01, †P < 0.05). ROS formation was decreased in UCP1-overexpressing neurons in basal glucose (†P < 0.05) and UCP1- and UCP3-overexpressing neurons in high glucose at 5 h (*P < 0.005) compared with OMC-expressing neurons. C: Control cultures received antimycin A (2.5 μmol/l), OMG (20 mmol/l), or glucose (20 mmol/l) for 1 h, and superoxide generation was measured using the ethidium-to-DHE ratio. D: DRG neurons were infected with either Ad.OMC or Ad.UCP3 and exposed to high glucose. The ethidium-to-DHE ratio was significantly decreased in UCP compared with OMC-overexpressing controls at the time points shown (*P < 0.01).


This work was supported in part by National Institutes of Health Grant NS42056, by the Juvenile Diabetes Research Foundation Center for the Study of Complications in Diabetes, Office of Research Development (Medical Research Service), by the Department of Veterans Affairs (to J.W.R.), by National Institute of Diabetes and Digestive and Kidney Diseases Grant #5P60DK-20572, and by the Michigan Diabetes Research and Training Center. The Ad.OMC was kindly provided by Dr. Thomas Scholz, University of Iowa.

The authors thank Denice Janus for her help with the manuscript preparation.


    • Accepted December 1, 2003.
    • Received February 2, 2003.


| Table of Contents