Sensory Neurons With Activated Caspase-3 Survive Long-Term Experimental Diabetes

Male Sprague-Dawley rats (weight 200–300 g at onset) were used for the experiment. Diabetes was induced by a single intraperitoneal injection of streptozotocin (STZ) (65 mg/kg). Controls received the STZ citrate buffer carrier alone. Additional control rats were noninjected male Sprague-Dawley rats weighing 200–300 g. The cohort of rats used in this study formed the basis of other previously reported investigations (5,20). The protocol was reviewed and approved by the University of Calgary Animal Care Committee based on the guidelines of the Canadian Council on Animal Care.

Rats were maintained on sawdust-covered plastic-bottomed cages and had access to rat water ad libitum. The whole blood glucose level was measured with a glucometer (AccuCheck; Boehringer Mannheim Canada, Dorval, Canada). Rats with a whole blood glucose level of ≥16.0 mmol/l were considered diabetic. Hyperglycemia was confirmed using fresh plasma and a glucose oxidase method (Ektachem DT6011 Analyzer; Kodak, Rochester, NY). In the same cohort of rats previously reported by our laboratory (5), there was slowing of motor and sensory conduction velocity, and diabetic rats weighed less than nondiabetic controls. One year later, diabetic and nondiabetic control rats were killed and tissue was harvested. L4 and L5 dorsal root ganglia (DRG) were rapidly removed during anesthesia (within ∼30 min) for immunohistochemical and light and electron microscopic studies.

Tissue samples were fixed in modified Zamboni’s fixative (2% paraformaldehyde, 0.5% picric acid, and 0.1% phosphate buffer) overnight at 4°C. Tissues were then washed in PBS five times, cryo-protected in 20% sucrose/PBS, and left at 4°C overnight. After embedding in optimum cutting temperature (OCT) compound (Miles Laboratories, Elkart, IN), 16-μm thick sections were placed on poly-l-lysine–coated slides. For Bcl-2 immunohistochemical staining, samples were fast frozen, embedded in OCT, and stored at −80°C. Then 14-μm sections were placed on slides and postfixed with acetone at −20°C for 10 min.

Neurons and their process were labeled by monoclonal antineurofilament 200 antibody (1:800, Sigma). Macrophages were recognized by mouse monoclonal anti-ED-1 (1:1,000, Chemicon). The following apoptosis-related primary antibodies were used: rabbit antiprocaspase-3 and the cleaved p20 fragment caspase-3 polyclonal antibody (1:100, Chemicon), anticleaved caspase-3 rabbit polyclonal antibody (1:100, Trevigen), monoclonal mouse antibody recognizing poly(ADP-ribose) polymerase (PARP) (1:100, Trevigen), which recognizes both the full-length and cleaved molecule, polyclonal rabbit anti-rat/mouse Bcl-2 antibody (BD PharMingen), and mouse monoclonal anticytochrome c (1:50, Zymed Laboratories, San Francisco). Secondary antibodies were sheep anti-mouse IgG CY3 conjugate (1:100, Sigma) or Alexa Fluor 488 goat anti-rabbit IgG (H + L) conjugate (1:500, Cedarlane Laboratories, Hornby, ON, Canada).

For indirect immunofluorescence, slides were incubated for 48 h at 4°C with primary antibodies, washed with PBS, and incubated with secondary antibodies for 1 h at room temperature. After further PBS washing, coverslips were mounted with bicarbonate-buffered glycerol (pH 8.6), and slides were viewed with a fluorescent microscope (Axioskop; Zeiss).

Controls included omission of primary or secondary antibodies on parallel sections. DRG sections from young nondiabetic rats were also used as controls for detecting the normal pattern of some antibody labeling. A marker of nuclei, 4′,6-diamidino-2-phenylindole (DAPI), was used to counterstain DNA. After immunohistochemical labeling, Vectashield mounting medium with DAPI (Vector Laboratories, Burlington, ON) was used for mounting slides.

After immunofluorescent staining, digital sectioning (digital “confocal” microscopy) was performed for selected studies by using a Leica DM-RXA2 fluorescent microscope. Series images were taken by Princeton Instruments ST-138 with KAF 1660 cooled to −40°C, collected at 0.5- to 1.0-μm intervals. Stacks of images with slices at the same Z position (±0.05 μ) were collected for each fluorochrome. The deconvoluted images were obtained using Vaytek Microtome version 6.1 (Fairfield, IA). Images were converted to 8-bit red-green-blue (RGB) on false color. The deconvoluted images were rescaled to cover the entire 255-value gray range and further processed in Adobe Photoshop (25).

Apoptotic cell death was detected by using an in situ cell death detection fluorescein kit (Roche Molecular Biochemicals) based on labeling of DNA strand breaks (transferase-mediated dUTP nick-end labeling [TUNEL] technology). L4 and L5 DRG from 1-year diabetic, 1-year control, and normal rats (n = 3 for each group) were removed, fast frozen, and immediately embedded in OCT compound in precooled 2-methylbutane in dry ice. Sections 14 μm thick were mounted on slides and postfixed in freshly made 4% paraformaldehyde in PBS for 20 min at room temperature. After washing for 30 min in PBS, slides were permeabilized in 0.1% Triton X-100/0.1% sodium citrate for 2 min on ice and then incubated in TUNEL reaction mixture at 37°C for 1 h. After TUNEL staining, slides were washed in PBS three times and covered. Each experiment included a negative and a positive control slide. For negative control, slides were incubated in the label solution only (without the terminal deoxynucleotidyl transferase). The positive control slides were treated with 3,000 units/ml DNase I for 10 min before incubating in the TUNEL reaction mixture. All slides were viewed with the fluorescent microscope.

L4 and L5 DRG were taken from 1-year control and diabetic rats and fixed in 2.5% glutaraldehyde in 0.025 mol/l cacodylate buffer overnight. After washing in 0.15 mol/l cacodylate, samples were postfixed with 2% osmium tetroxide in 0.12 mol/l cacodylate, dehydrated in graded alcohols, and embedded in epon. Sections 1 μm thick were cut using an ultramicrotome, stained with toludine blue, and viewed by a light microscope (Axioskop; Zeiss). Light microscopy with blinded grading has been reported by our lab in this cohort before (5). For electron microscopy, 70-nm thick sections were placed on copper grids and stained with 10% uranyl and 0.1% lead citrate. The grids were examined with an H7000 transmission electron microscope (X2500-15000). Neurons were examined from the entirely randomly selected sections.

For comparing the immunohistochemical staining of caspase-3 (procaspase-3 and the cleaved p20 fragment caspase-3), sections of L4 DRG from diabetic and control groups were postfixed and stained with this antibody. Six digital DRG images were taken randomly from each rat at the same magnification (400×). Caspase-3 immunoreactive profiles were counted and the mean calculated from six sections of each of three rats per group.

For counting the Bcl-2–positive neurons between diabetic and control groups, sections from L5 DRG of each rat were mounted on slides at 80-μm intervals. After immunohistochemical staining, Bcl-2 immunoreactive neurons were counted in sequential sections until 200 Bcl-2 immunoreactive neurons were reached. The percentage of Bcl-2 neurons out of total neuron profiles was then calculated (n = 3 for each group).

Results were compared as means ± SE using two-tailed Student’s t tests.

As previously reported (5), diabetic rats weighed less than controls (415 ± 15 vs. 760 ± 16 g), had whole blood hyperglycemia (18.1 ± 0.6 vs. 4.9 ± 0.2 mmol/l), and had slowing motor (43.9 ± 0.9 vs. 52.5 ± 1.1 m/s) and sensory (54.9 ± 0.9 vs. 56.8 ± 0.7) conduction velocities (n = 26 diabetic and 27 nondiabetic rats in the cohort).

We applied two anti–caspase-3 antibody markers. The first was directed toward procaspase-3 (32 kDa) and the cleaved p20 fragment, and the second primary antibody exclusively recognized cleaved caspase-3 only. Discrete profiles labeled with the procaspase/p20 antibody were detected in proximal axon segments of neurons (Fig. 1A and B), which was confirmed by double labeling in transverse and longitudinal sections with NF200 antibody directed against the heavy neurofilament subunit [Fig. 1C, C(a), and C(b)]. This caspase-3 antibody generally did not label perikarya. There was a significant increase in labeled profiles observed in diabetic animals compared with the nondiabetic control group (Fig. 1).

Labeling with an antibody to cleaved caspase-3 antibody also confirmed that there were rises in diabetic neurons with immunoreactivity in proximal axon segments, as above, but also within perikarya. The latter labeling, only very rarely observed in nondiabetic control neurons, was diffuse cytoplasmic, multifocal cytoplasmic, and occasionally nuclear (Fig. 1D–H). Occasional satellite cells in both nondiabetic (in both young control rats and citrate-injected nondiabetic rats at 12 months) and diabetic DRG were labeled with the antibody to cleaved caspase-3. Sural nerves had no evidence of caspase-3 immunoreactivity using either antibody approach.

In nondiabetic DRG, PARP immunohistochemical labeling was largely confined to neuron nuclei. Similar findings were encountered whether young nondiabetic rat DRG were examined or the 12-month nondiabetic control group. Small PARP profiles outlined most neuron nuclei, which was confirmed by double labeling with the nuclear marker DAPI (Fig. 2A–B). PARP expression differed in diabetic neurons, (Fig. 2C–D) with somewhat more prominent labeling of neuronal nuclei and cytoplasm, and also in proximal axons, seen using serial optical imaging (Fig. 2E). Some PARP immunoreactivity was also observed in the nuclei of the satellite cells (Fig. 2E), a pattern not observed in nondiabetic controls. PARP immunoreactivity did not generally colocalize with that of caspase-3.

Assuming that the activation of caspase-3 was inexorably linked to genomic DNA fragmentation, we applied TUNEL to sections of diabetic and nondiabetic DRG. Each series of labeled slides included a positive control using DNase I to fragment DNA and a negative control slide to eliminate terminal deoxynucleotidyl transferase. In positive control slides, numerous TUNEL-positive cells included satellite cells, Schwann cells, DRG capsule cells, and a few neurons (Fig. 3A and B). Slides were counter stained with toludine blue to confirm the identity of the labeled cells (Fig. 3B). Negative control slides did not have TUNEL, as expected.

In ganglia of long-term diabetic and nondiabetic rats, only occasional satellite cells were TUNEL positive (Fig. 3D). Neurons were not labeled. In addition to TUNEL staining, we examined the pattern of DAPI staining of ganglia cell populations. No characteristic nuclear clumping or fragmentation expected of apoptosis was observed in neurons (Fig. 3E–H).

Random sampling of thin electron microscopy sections was made and included a minimum of 60–80 neurons in each of three rats from both the nondiabetic and diabetic groups. None of the neuronal nuclei had chromatin condensation or fragmentation to suggest apoptosis, and cytoplasmic blebbing associated with neurons or in other fields was not observed (Fig. 4). There was an apparent increase in lipofuscin pigment in diabetic neurons.

To evaluate an antiapoptosis signaling factor that could contribute to neuron survival in diabetes, we labeled ganglia with an antibody directed against Bcl-2, a member of the antiapoptotic cascade. Bcl-2 immunohistochemical staining was largely confined to the cytoplasm of small-diameter sensory DRG neurons in both diabetic and nondiabetic control neurons (Fig. 5). There was no difference in the proportion of Bcl-2–labeled neurons between nondiabetic control and diabetic ganglia (Fig. 5G). Discrete Bcl-2 labeling was also observed in axons and blood vessels without detectable differences between diabetic and control subjects. Cytochrome c neuronal labeling appeared confined to discrete profiles most likely representing cytoplasmic mitochondria (data not shown). There was no apparent difference between diabetic and control subjects and otherwise no evidence of cytochrome c presence in the cytosol.

In DRG harvested from young rats, rare nitrotyrosine-labeled profiles were observed in the cytoplasm of Schwann cells (data not shown). In 1-year diabetic and nondiabetic control groups, nitrotyrosine antibody also labeled the cytoplasm of occasional small neurons and occasional proximal axon segments (Fig. 6). Nitrotyrosine immunoreactivity appeared more prevalent in diabetic neurons, but labeling was judged not sufficiently discrete to permit specific counting. Rare ED-1–labeled macrophage profiles were observed in both diabetic and nondiabetic DRG without a difference between the groups.

The major findings of this study are as follows. First, DRG sensory neurons in a model of long-term experimental diabetes (12 months) failed to exhibit any features of apoptosis, such as nuclear fragmentation, clumping on DAPI staining or electron microscopy, or TUNEL. Second, while apoptosis did not occur in diabetic sensory neurons, there was heightened expression of activated caspase-3 in nuclei, cytoplasm, and proximal axon segments. Third, PARP labeling was enhanced in diabetic neurons and was expressed in nuclei, cytoplasm, and proximal axon segments. Fourth, Bcl-2 expression was neither upregulated nor downregulated in experimental diabetes. Finally, nitrotyrosine is expressed in DRG neurons of diabetic and nondiabetic rats of older age with a trend toward greater labeling in diabetic subjects. Macrophage infiltration was not increased in diabetes.

The above evidence needs to be coupled with our findings reported previously in the same model (5), indicating that: 1) unbiased physical counts of the entire L5 DRG did not identify neuron dropout; 2) downstream sural axons were preserved in number (chronic cumulative neuron dropout over 12 months should have been easily detectable by either method); and 3) blinded assessment of 100–150 neurons under light microscopy in each of five diabetic and five nondiabetic rats did not identify significant evidence of degeneration, loss (nests of Nageotte), or apoptosis accompanying diabetes.

Several approaches are available to evaluate possible activation of the death-inducing signaling cascades in diabetes. Upstream activators include Fas/tumor necrosis factor-α, which recruits procaspases through the death-inducing signaling complex, eventually activating effector caspases 3, 6, and 7 (26). As postulated in diabetes, activation of death pathways may instead be more likely to occur through mitochondrial pore formation (27). The latter step in turn releases cytochrome c and, with Apaf-1 and caspase-9, forms the apoptosome, which then generates a downstream cascade of caspase activation. Other putative players in this complex pathway not addressed in the present study are upstream members of the caspase family, Jun NH₂-terminal kinase, nuclear factor κ-B, phosphatidylinositol 3-kinase/Akt, second mitochondria-derived activator of caspase/DIABLO, and others (27,28). In this study, we selected several key steps putatively linked to diabetic neurons by previous studies: caspase-3 proenzyme and its activated product as critical downstream activators of cell death; Bcl-2, an inhibitor of mitochondrial pore formation; and PARP, a target of caspase cleavage and also a repair enzyme.

We applied particular rigor to addressing the thorny question of whether apoptosis was actually ongoing by combining unbiased physical dissector counting (not done before in the models), morphological changes under the light and electron microscope, and three potential assays of apoptosis. An immunohistochemical approach was required for all of the players examined in this study because immunoblots or other approaches would not provide information about selective protein expression in neurons alone. Abnormal vacuolation of sensory neurons has been considered an index of apoptosis-related mitochondrial dysfunction in diabetes in some studies (29). While mitochondrial dysfunction may be a key element of cellular damage in diabetes, microvacuolar change as observed under light microscopy has been nonselective and can occur as a function of postmortem harvesting (X.-Q. Li and D.W.Z., unpublished observations).

We identified procaspase-3 and the p20 cleaved fragment labeling to be largely in axons, whereas activated caspase-3 was also observed in neuronal soma and nuclei. The presence of limited staining in nondiabetic control ganglia axons and cytoplasm may reflect low-level constitutive expression of the proenzyme. In diabetes there was more widespread expression, which was also in nuclei, unlike controls. Mechanisms postulated to account for heightened activation of caspase-3 in diabetes include oxidative and nitrative stress, growth factor deficiency, ischemia, and perhaps excessive polyol flux (8,16,19,21,22,30–34). In our model, activated caspase-3 was not detectable in sural axons. The overall pattern is different from an apparent retrograde activation of caspase-3 after target removal in olfactory neurons reported by Cowan et al. (35).

The findings indicate that sensory neurons are directly vulnerable to diabetic damage. Despite heightened caspase-3 activation, we provide convincing evidence against ongoing apoptosis. Cell shrinkage, aggregation of chromatin at the nuclear membrane, and formation of membrane blebbing was not observed under light and electron microscopy. Furthermore, unbiased physical dissection counts of the entire L5 DRG did not identify significant neuron loss in the same cohort of rats. Is it possible that apoptosis occurs too rapidly to be detected in neurons at a given time point, particularly if only a smaller proportion of neurons are dying at any given time? Any significant form of ongoing loss, however, should have led to a reduction of counted neurons, but it did not.

It has been widely accepted that the activation of effector caspases, such as caspase-3, will eventually lead to apoptosis of a cell. This view, however, has recently been challenged (36). In some instances, the activation of caspase-3 does not commit a cell to apoptosis. The balance of pro- and antiapoptotic cascades and their interaction with adjacent supporting cells is complex. Reddien et al. (37) reported that in C. elegans, cells expressing CED-3 caspase—the homologue of mammalian caspase-3—survive and even differentiate, passing through a stage called “mostly dead,” which resembles the early phase of apoptosis (38). Hinck et al. (39) noted that hyperglycemia induced caspase-mediated nuclear chromatin degradation but that high glucose–induced nuclear fragmentation was not linked to caspases. Here we provide new evidence that caspase-3 activation is not always fatal in vertebrates.

TUNEL may reflect chronic mild DNA injury in some models that is repaired by enzymes such as PARP. A more recent report (40) suggested the presence of caspase activation and extensive apoptosis in a model of experimental diabetes lasting 12 months. In this study the number of DRG neurons expressing caspase-3 was also increased at earlier time points. TUNEL cells, which are increased in diabetes, were used to identify apoptosis. In our study we argue, as have others, that morphological criteria of neurons undergoing apoptosis, unequivocal cell and axon loss, and replacement of neurons by nests of Nageatte are critical in proving that apoptosis exists. Regardless of these differences, our studies do support the concept that apoptotic actors are activated by diabetic stress. A further report examining intact neurons in STZ-induced diabetes only identified TUNEL-positive DRG sensory neurons after axon injury (41). Apoptotic cell death was not observed in sensory neurons of mouse spinal ganglia (42).

PARP is a substrate of both caspase-3 activation in apoptosis and an enzyme activated by peroxynitrite. Its nuclear localization in nondiabetic ganglia is an expected feature of its function in repairing DNA injury. In diabetes, however, PARP had heightened not only cytoplasmic localization but also movement out into proximal axon segments. We also identified novel PARP expression in diabetic perineuronal satellite cells. There is evidence that this population may provide critical support for neurons, and it is also possible that satellite cells retain their supportive role for neurons by proliferating when they are directly targeted. The axonal localization of PARP may reflect activation of mechanisms to repair damaged mitochondrial DNA, a postulated mechanism of cell death in diabetic neurons (24,43,44). Our antibody did not distinguish whether increased PARP in diabetes was from the parent molecule or cleaved product.

Nagayama et al. (45) reported that after sublethal transient global ischemia, activation of PARP in rat hippocampus might contribute to cellular recovery. The cleavage of PARP by caspases (such as caspase-3) facilitates cellular disassembly and serves as a marker of cells undergoing apoptosis. In other conditions, such as exposure to peroxynitrite and saturating levels of DNA damage, excessive PARP activity has been thought to lead to cell death through unregulated consumption of high-energy metabolites by the enzyme (46,47).

Chronic low-grade rises in peroxynitrite elaboration, and perhaps toxicity, may occur in diabetic ganglia. In a previous study, we found evidence that long-term (12 months) diabetic ganglia had increased activity of nitric oxide synthase, thus likely generating local rises in nitric oxide content (20). The reaction of nitric oxide with superoxide forms the toxic metabolite peroxynitrite, whose footprints in turn appear as nitrotyrosine immunoreactivity. Endocardial biopsies from diabetic patients identified both nitrotyrosine labeling and apoptosis (48).

Another unexpected finding in our study was the absence of alterations in Bcl-2 intensity in diabetes. When cells are exposed to stimuli that trigger apoptosis, cytochrome c is released from mitochondria into cytoplasm, where it activates caspases. By preventing the release of cytochrome c, Bcl-2 protein interferes with this activation of caspases and prolongs cell survival. Srinivasan et al. (24) reported decreased Bcl-2 in 3- to 6-week diabetic DRG, suggesting increased risk for apoptosis. We found no such change, which further supports arguments against active apoptosis occurring in our population and evidence that other “repair” cascades may be involved in protecting diabetic sensory neurons. Also, since Bcl-2 appeared to be mainly concentrated in small-diameter neurons, whereas caspase-3 distribution was more widespread, other factors are also likely to protect neurons from apoptosis. One of the candidates might be HSP-27 (heat shock protein 27, a survival protein), which was found to have elevated mRNA levels in DRG neurons in our previous report of this model (5,49).

In summary, diabetic stress triggers molecular actors in the apoptotic pathway, but unique defense mechanisms likely protect long-term diabetic sensory neurons from apoptosis. Understanding and exploiting such mechanisms may offer new avenues in the treatment of human diabetic polyneuropathy.

This study was supported by an operating grant from the Canadian Institutes of Health Research. D.W.Z. is a Senior Medical Scholar of the Alberta Heritage Foundation for Medical Research.

Brenda Boake provided expert secretarial assistance. Dr. W.M. Schoel, W-X. Dong, and R.W. Humphrey of the University of Calgary Microscopy and Imaging Facility kindly assisted with deconvolution and electron microscopy.