Receptor for Advanced Glycation End Products (RAGEs) and Experimental Diabetic Neuropathy

We studied 128 male Swiss Webster wild-type (WT) mice with initial weight of 20–30 g, housed in plastic sawdust-covered cages with a normal light-dark cycle and free access to mouse food and water. All protocols were reviewed and approved by the University of Calgary Animal Care Committee using the Canadian Council of Animal Care guidelines. Mice were anesthetized with pentobarbital (60 mg/kg) before all procedures. At the age of 1 month, 82 mice were injected with streptozotocin (STZ) (Sigma, St. Louis, MO) intraperitoneally for each of 3 consecutive days, with once-daily doses of 60, 50, and then 40 mg/kg, with the remaining 46 mice injected with carrier (sodium citrate) for three consecutive days. Whole blood glucose measurements were performed monthly following electrophysiological testing using the tail vein and a blood glucose meter (OneTouch Ultra Meter; LifeScan Canada, Burnaby, BC, Canada). Hyperglycemia was verified 1 week later by sampling from a tail vein. A fasting whole blood glucose level of ≥16 mmol/l (normal 5–8 mmol/l) was our criterion for experimental diabetes. All animals had whole blood glucose sampling monthly and animals were weighed monthly as well. Animals were followed for an additional 1, 3, 5, or 8 months of diabetes (up to 9 months of life). Male and female RAGE^−/− mice were constructed on an SVE129×C57BL/6 background (129/B6) as previously described (16) and were backcrossed >10 generations into C57BL/6 before enrollment in studies. WT C57BL/6 mice were used as control littermates in these experiments. RAGE^−/− mice were also injected with STZ as above (n = 15) or carrier (n = 11) and, along with control RAGE^−/− nondiabetic littermates and WT diabetic and nondiabetic littermates, were followed for 1, 3, or 5 months of diabetes. In all cases, those mice that did not develop diabetes as defined above after STZ injections were excluded from further assessment.

The main cohort of diabetic mice was injected, followed, and harvested at the University of Calgary. RAGE^−/− mice were injected, followed, and harvested at Columbia University. In all cases, mice were raised and studied in strict pathogen-free environments. Tissues from all mice, irrespective of their origin, were amassed, identically processed, and appraised at one site (Calgary), with the measurements carried out by the investigator blinded to their origin (C.T.).

Electrophysiological assessment of sciatic nerve conduction was performed as previously described (17) under halothane anesthesia. Initial baseline studies were carrier out before STZ or carrier injection and identified no significant differences between groups studied. For sensory conduction studies, the tibial nerve was used with a fixed distance of 30 mm from stimulation electrodes to the sciatic notch, where recording electrodes were placed to measure the sensory nerve action potential amplitude and sensory nerve conduction velocity in orthodromic fashion. All stimulating and recording electrodes were platinum subdermal needle electrodes (Grass Instruments, Astro-Med, West Warwick, RI), with near-nerve temperature kept constant at 37 ± 0.5°C using a heating lamp. The cohort of WT and RAGE^−/− mice, both diabetic and nondiabetic, underwent electrophysiological testing after 5 months of diabetes, whereas WT mice were assessed monthly until 8 months of diabetes.

After 1, 3, 5, or 9 months of diabetes, mice were killed and the following tissues were harvested: bilateral L3–L6 dorsal root ganglia (DRG), sciatic nerves, sural nerves, and hind and fore footpads. Blood for glycated hemoglobin measurements was taken before death. One-half of all tissues were placed either in Zamboni's fixative (left-sided tissues) for later immunohistochemistry or were fixed in cacodylate-buffered glutaraldehyde and then cacodylate buffer for later epon embedding for morphometric studies. The remaining tissues (right-sided tissues) were immediately fresh-frozen at 80°C or placed in Trizol fixative (Life Technologies, Rockville, MD) and stored at −80°C. Additional tissues to be used as control samples including brain, spinal cord, liver, and pancreas were removed and fixed in similar manners.

RAGE^−/− mouse tissues and corresponding littermate tissues were harvested after 1, 3, and 5 months of diabetes and were placed in Zamboni's fixative. Bilateral L3–L6 DRG, sciatic nerves, sural nerves, and hind and fore footpads were harvested as above.

For peripheral nerve and DRG specimens, samples were fixed in 2.5% glutaraldehyde in 0.025 mol/l cacodylate buffer overnight, before washing with repeated 0.015 M cacodylate and glucose-buffered cacodylate, progressive dehydration with alcohols and propylenoxide, staining with osmium tetroxide, and placement in epon. Semi-thin (1-μm) sections of peripheral nerve and L4–L6 DRGs were cut on an ultramicrotome (Reichert, Vienna, Austria) and were stained with 0.5% toluidine blue. Image analysis was performed by a single examiner blinded to the origin of the sections (Zeiss Axioskope at 400× and 1,000× magnification using Scion Image v.4.0.2 [Scion, Fredrick, MD]) with measurements of the number, axonal area, and myelin thickness of all myelinated fibers within 25 nonadjacent transverse nerve sections. Secondary measurements included macrophage number, degenerating profile number, and regenerating fiber cluster number. For DRGs, neurons with visible nuclei were used for counting within a sized area for 25 nonadjacent sections separated by ∼300 μm for each L4–L6 DRG for neuronal density and to provide calculations for total estimated neuronal counts.

For tissues designated for immunohistochemistry, specimens were fixed in 2% Zamboni's fixative overnight at 4°C, washed in PBS, kept overnight in 25% glucose PBS solution, embedded in optimal cutting temperature embedding solution, and stored at −80°C until sectioning. Cryostat transverse and longitudinal nerve sections (10 μm) were placed onto poly-l-lysine–and acetone-coated slides. Sections were stained with anti-human RAGE IgG primary antibody (supplied by Dr. A.M. Schmidt, Columbia University, New York, NY). Antigen retrieval was performed with slides placed in sodium citrate in an 80°C water bath, a PBS wash for 5 min, blocking with 10% goat serum for 1 h, and further PBS washing. Slides were incubated with primary antibody (anti-human RAGE 1:100) overnight at 4°C. After PBS washing, secondary fluorescent antibody (anti-rabbit IgG fluorescein isothiocyanate, 1:100; Zymed, San Francisco, CA) was applied with incubation for 1 h at room temperature.

Additional immunohistochemistry was performed using antibodies to CML [NH₂-(carboxymethyl)lysine] (anti-CML monoclonal antibody, 1:200; CosmoBio, Tokyo, Japan), nuclear factor (NF)-κB (anti–NF-κB p50, 1:200; Santa Cruz Technology, Santa Cruz, CA), PKCβII (anti-PKCβII, 1:200; GeneTex, San Antonio, TX), HNE (4-hydroxy-29-nonenal) Adduct (anti-HNE Adduct, 1:200; Cedarlane Laboratories, Hornby, ON), and C-Rel (anti–C-Rel, 1:200; Cell Signaling Technology, Danvers, MA) to examine RAGE-associated pathways. S100 was used to label Schwann cells (anti-S100, 1:100; Chemicon International, Temecula, CA), NF-200 and β-tubulin were used to label peripheral nerve axons (anti-neurofilament, NF-200, 1:100; Chemicon; anti–β-tubulin, 1:100; Abcam, Cambridge, MA), and MAP-2 and NeuN were used to label neurons (anti–MAP-2, 1:100; Sigma-Aldrich, Oakville, Ontario; anti-NeuN, 1:100; Chemicon International).

For these immunohistochemistry slides, secondary antibody incubation was performed as above (anti–fluorescein isothiocyanate or bovine anti-mouse IgG Cy3, 1:100; Zymed), with the exception of the C-Rel immunostaining, where a peroxidase-labeled streptavidin biotin antibody (LSAB; Dako, Copenhagen, Denmark) was used for secondary detection. To perform triple localization, consecutive samples were collected on alternating slides. Analysis used Image Pro Plus software (Image Pro Plus 5.0; MediaCybernetics, Silver Spring, MD) and Adobe Photoshop software (Adobe Photoshop 7.0, Adobe, San Jose, CA, 2002). Positive profiles were tallied on every 10-μm-thick section on 20 random sections for each DRG or nerve, and the luminosity of individual neurons, glia, or nerve fibers was measured. The numbers of neurons, axons, or glia with luminosity was classified as none-low (luminosity value of 0–150), moderate (150–250), or high (>250) (scale of 0–255 with arbitrary units). Neurons and glia were assessed for nuclear labeling for NF-κB recorded as well using a predetermined luminosity measurement threshold of 150 (no units). Measurement of the activation of PKC, believed to be related to translocation from cytosol to neurolemma, and the extent of translocation was assessed using a predetermined luminosity measurement threshold of 150 (no units).

Epidermal foot pads from both hind feet and forefeet were fixed in 2% Zamboni's fixative overnight at 4°C, washed in PBS, kept overnight in 20% glucose PBS solution, embedded in optimal cutting temperature embedding solution, and stored at −80°C until sectioning. The 30-μm sections were prepared using a cryostat, and samples were placed onto poly-l-lysine–and acetone-coated slides. Immunohistochemistry within foot pads was performed in identical fashion as described above, using anti-Human RAGE IgG primary antibody and the secondary antibody anti-rabbit IgG fluorescein isothiocyanate and using the panaxonal marker PGP 9.5 to identify all epidermal axons (anti-mouse PGP 9.5 antibody, 1:500; Jackson ImmunoResearch Laboratories, West Grove, PA; and secondary antibody, anti-goat IgG Cy3, 1:100; Chemicon International).

Analysis of epidermal fibers within foot pads was performed as described previously (18). In brief, epidermal nerve fibers extending above the dermal papillae were quantified per square millimeter of epidermal area. We chose to use this three-dimensional reconstruction to clearly capture all fibers present within the microsection. In addition to this, we also captured linear densities by calculating the number of fibers within each microsection expressed as a function of epidermal length, which has the disadvantage of calculation of epidermal length over a nonlinear surface, as well as of a possible overcounting of undulating fibers identified in multiple sections within three-dimensional sectioning. A single Zeiss fluorescent microscope was used in all cases. Digital photography (Zeiss) of all specimen portions was performed, and Adobe Photoshop 7.0 was used to visualize each specimen for counting purposes. For each animal, the number of PGP 9.5-immunoreactive profiles was counted by a single observer blinded to the source of each specimen in each of the 30-μm-thick microsections, which were performed through the entire footpad. A minimum of 200 microsections was examined in each case.

In situ hybridization for IκB using digoxigenin-labeled RNA probes was performed as described previously (19). DRG cryosections (16 μm) were fixed in 4% paraformaldehyde and washed twice in PBT (PBS with 0.1% Tween-20). After bleaching with 6% H₂O₂/PBT and washing, sections were treated with 1 mg/ml proteinase K/PBT. Then sections were re-fixed and washed in PBT before hybridization overnight at 70°C. Sections were washed three times in 50% formamid/5 × SSC/1% SDS at 70°C, followed by two washes in 50% formamid/2 × SSC at 65°C. Staining was visualized using an Nitro-Blue Tetrazolium Chloride (NBT)/5-Bromo-4-Chloro-3′-Indolyphosphate p-Toluidine Salt (BCIP) (Roche, Laval, PQ) substrate.

Peripheral nerve, DRG, liver, and pancreas tissues had protein quantified from fresh frozen samples that were maintained at −80°C for a maximum of 1 month before protein quantification studies. Tissue portions for protein studies were homogenized using a RotorStator Homogenizer in ice-cold lysis buffer (10% glycerol, 2% SDS, 25 mmol/l Tris-HCl, pH 7.4, Roche Mini-Complete Protease Inhibitors). Samples were then centrifuged at 10,000g for 15 min. Supernatant was stored at −20°C before SDS-PAGE and immunoblotting analysis. Equal amounts (15 μg) of protein were loaded, and samples were separated by SDS-PAGE using 10% polyacrylamide gels with 800 V hours of current applied. Separated proteins were transferred onto nitrocellulose paper (Bio-Rad) over 16 h at 200 mA in Towbin transfer buffer (25 mmol/l Tris, 192 mmol/l glycine, 20% vol/vol methanol, 0.1% vol/vol SDS). Separate blots were blocked overnight in 7.5% (wt/vol) milk (Nestle, Carnation) in TBS (50 mmol/l Tris, 137 mmol/l NaCl, 51 mmol/l KCl, 0.05% [vol/vol] Tween-20).

Anti-human RAGE IgG primary antibody (1:500) anti–NF-κB p50 (1:1,000), anti-PKCβII (1:1,000), and anti–β-actin (1:100, Biogenesis, Poole, U.K.) were applied to separate blots. Secondary anti-rabbit or anti-mouse IgG horseradish peroxidase–linked antibody (Cell Signaling) was applied at 1:5,000 in each case as appropriate. Signal detection was performed by exposing of the blot to enhanced chemiluminescent reagents ECL (Amersham) for 2 min. The blots were subsequently exposed and captured on Kodac X-OMAT K film.

Total RNA was extracted from peripheral nerve, DRG, brain, spinal cord, liver, and pancreas tissue using Trizol reagent (Life Technologies). Total RNA (1 μg) was processed directly to cDNA synthesis using the Superscript II Reverse Transcriptase system (Invitrogen). RAGE primers and probe sequences were as follows: forward, 5′-GGACCCTTAGCTGGCACTTAGA-3′; backward, 5′-GAGTCCCGTCTCAGGGTGTCT-3′; and probe, 5′-ATTCCCGATGGCAAAGAAACACTCGTG-3′. β-Actin primers and probe sequences were as follows: forward, 5′-CCTGAGCGCAAGTACTCTGTGT-3′; backward, 5′-GCTGATCCACATCTGCTGGAA-3′; and probe, 5′-CGGTGGCTCCATCTTGGCCTCAC-3′. Cyclophyllin primer sequences were as follows: forward, 5′-TGTGCCAGGGTGGTGACTT-3′, and backward, 5′- TCAAATTTCTCTCCGTAGATGGACTT-3′. PKCβII primers sequences were as follows: forward, 5′-GGTGGCATGTAGAAAGTGCTGC-3′, and backward, 5′-CAAGCATTTTCTCTCCCGTGG-3′. NF-κBp65 primers sequences were as follows: forward, 5′-TGTGCGACAAGGTGCAGAAA-3′, and backward, 5′-ACAATGGCCACTTGCCGAT-3′.

RT-PCR was done using SYBR Green dye. All reactions were performed in triplicate in an ABI PRISM 7000 Sequence Detection System. Data were calculated by the 2-^ΔΔCT method and are presented as the fold induction of mRNA for RAGE in diabetic tissues normalized to cyclophyllin compared with nondiabetic tissues (defined as 1.0-fold).

All data were represented as mean ± SE. The t testing or ANOVA testing with multiple comparisons of independently assessed samples and groups were performed as appropriate in all cases with Bonferroni corrections as needed.

Mice injected with STZ developed diabetes within 1–4 weeks after injection in >80% of animals, and in each case, diabetes was maintained over the length of the study. The time at which diabetes was detected was taken as the onset of diabetes and was within 2 weeks of the last STZ injection in all cases (or otherwise mice were not included in further studies) and within 1 week of STZ injection in over 90% of mice studied. WT diabetic mice were smaller than WT nondiabetic mice at 1 month after STZ injection, and diabetic mice had smaller body weights throughout life (Table 1). Although the diabetic mice were smaller, they nevertheless continued to gain weight over the course of the study. Mouse glycated hemoglobin was increased in WT diabetic mice at 9 months of life. The mortality rate in WT diabetic mice was significantly higher than in controls, with only 40% of all diabetics surviving the entire 9-month duration of diabetes. Kaplan-Meier survival statistics identified an increased mortality rate in diabetic mice over the duration of the experiment, with significant mortality present after 5 months of diabetes. RAGE^−/− diabetic mice were viable and were not physically different than their nondiabetic littermates. RAGE^−/− mice displayed levels of hyperglycemia similar to those of WT diabetic mice.

There were no differences in motor or sensory nerve electrophysiological parameters between WT and RAGE^−/− nondiabetic or diabetic mice present at baseline. Electrophysiological recordings identified dysfunction in diabetic WT mice within 5 months of diabetes for all sensory and motor measurements (Fig. 1), as previously described in diabetic rodent models (15). For WT diabetic mice, declines in sensory nerve conduction velocities and sensory nerve action potential amplitudes were present after 2 months of diabetes, similar to previously reported results (20–23). In marked contrast, RAGE^−/− mice with diabetes had electrophysiological results similar to RAGE^−/− mice without diabetes and WT mice without diabetes after 5 months, without slowing of nerve conduction velocities or compound motor action potential and sensory nerve action potential amplitudes (Fig. 1). Overall, in comparison to WT diabetic mice, diabetic RAGE^−/− mice displayed significant protection against dysfunction in both motor and sensory function.

There were no differences in morphological parameters of sural or sciatic nerves between WT and RAGE^−/− nondiabetic or diabetic mice present at 1 month after study initiation. The sural nerves from WT diabetic mice developed a loss of fiber density and axonal area (atrophy) relative to nondiabetic controls (Fig. 2) and sciatic nerve axonal atrophy was also detected in WT diabetic mice (Fig. 2) at later time points. However, sciatic fiber density was not changed over the period of exposure to diabetes. Diabetic sural and sciatic nerve myelin thickness was reduced after 5 months of diabetes (Tables 2 and 3). There were no differences in the number of macrophages or degenerating fiber clusters seen between WT mice with or without diabetes, but an increased number of degenerating axonal profiles were identified in both sciatic and sural diabetic nerves compared with control nerve (Tables 2 and 3). In contrast to WT diabetic mice, RAGE^−/− mice with diabetes were protected from declines in sural axon density and from axonal atrophy in the sural and sciatic nerves after 5 months of diabetes (Fig. 2 and Table 3).

CML deposition was verified in DRG neurons and within peripheral nerves, and there was heightened expression with diabetes, irrespective of RAGE genetic status of the mouse (supplementary Fig. 1, found in an online-only appendix at http://doi.org/10.2337/db07–0339). RAGE expression was of greater intensity within axons and Schwann cells of both the sciatic nerve and especially the sural nerve of WT mice with diabetes (Fig. 3 and Table 4). Expression of RAGE was also noted within endothelial cells of vasa nervorum. Identification of Schwann cells using the S-100 antibody revealed extensive co-localization with RAGE within diabetic peripheral nerve and within DRGs of WT diabetic mice (Fig. 4 and Table 4). RAGE was also expressed in peripheral nerves of WT mice without diabetes but at lower levels (Fig. 3 and Table 4). Despite the long-term diabetes, no apoptosis was detected within DRG neurons or Schwann cells. Evidence for apoptosis could be detected within satellite cells of DRGs exposed to diabetes, with greater predominance found in diabetic WT mice compared with diabetic RAGE null mice (supplementary Fig. 2).

DRG neurons of WT diabetic mice were smaller in area (neuronal atrophy) than DRG neurons in WT mice without diabetes beginning after 5 months of diabetes (Table 5). In addition, at later time points available for study in the Swiss Webster WT mice (8 months), diabetic DRGs had a decline in profile neuronal density compared with nondiabetic control mice, suggesting neuronal dropout (Fig. 4). In contrast to the WT diabetic mice, RAGE^−/− diabetic mice, available for analysis and studied at 5 months of diabetes, were protected from neuronal atrophy. RAGE expression was greater within neurons and Schwann cells of diabetic WT mouse DRGs compared with nondiabetic controls (Fig. 3 and Table 4).

The footpad epidermal nerve fiber density of WT mice with diabetes was reduced compared with control foot pads at and after 3 months of diabetes (Fig. 5 and Table 6). RAGE^−/− mice with diabetes were protected from epidermal fiber loss (Fig. 5 and Table 6), with no difference identified between epidermal fiber densities in RAGE^−/− diabetic mice compared with those without diabetes after 5 months (Fig. 5 and Table 6). Results of epidermal nerve fiber densities were similar using two separate methods of quantification, either with epidermal area or epidermal length.

RAGE expression was identified in both control and diabetic epidermal axons, but there was greater expression within residual diabetic epidermal axons (Fig. 5). RAGE labeling was also increased within both sebaceous and sweat glands, as well as within dermal blood vessels of diabetic mice.

Diabetes was associated with a relative increase in RAGE mRNA within DRG and peripheral nerve (Fig. 6) compared with the housekeeping gene cyclophyllin. Western blotting similarly confirmed the presence of increased RAGE protein within diabetic WT mouse DRG and peripheral nerve relative to nondiabetic WT mouse tissues (Fig. 6), when RAGE protein levels were normalized for tissue β-actin content.

WT mice with diabetes also demonstrated heightened expression of both protein and mRNA for NF-κB and PKCβII when compared with WT mice without diabetes (Fig. 6). In contrast, RAGE^−/− mice had very low levels of NF-κB and PKCβII for both protein and mRNA in PNS tissues, independent of their diabetic status (Fig. 6).

Consistent with these findings, immunohistochemistry of peripheral nerve and DRGs demonstrated upregulation of NF-κB and its increased nuclear translocation (activation) in DRG sensory neurons in WT mice with diabetes when compared with WT mice without diabetes and RAGE^−/− mice with or without diabetes (Fig. 7). Both cytoplasmic and nuclear expression of NF-κB was attenuated in DRG neurons in RAGE^−/− mice with diabetes when compared with WT mice with diabetes (Fig. 7). However, the more striking finding in both DRG and peripheral nerve was the overexpression of NF-κB in Schwann cells, already exhibiting heightened expression of RAGE (above), from the WT mice with diabetes (Fig. 7). The transcription factor C-Rel was upregulated in WT diabetic DRG neurons, but not in RAGE null diabetic neurons (supplementary Fig. 3). HNE Adduct, a marker of lipid peroxidation and oxidative stress, was upregulated in diabetic DRG neurons independent of RAGE genetic status (supplementary Fig. 4). In situ hybridization for the inhibitor of NF-κB activation, IκB, identified its significant upregulation in WT diabetic mice (supplementary Fig. 5) compared with other groups. The prominent expression of NF-κB in Schwann cells was attenuated in RAGE^−/− diabetic mice. Furthermore, the intensity of NF-κB expression was associated with RAGE expression in both WT nondiabetic and diabetic DRG neurons, especially in diabetic mice (Fig. 8). Although peripheral nerve and DRGs also exhibited upregulation of PKCβII in DRG sensory neurons in WT mice with diabetes when compared with nondiabetic WT mice and RAGE^−/− mice with or without diabetes, these results were much more modest than observed changes with NF-κB. Taken together, our results demonstrate that NF-κB, a potential downstream signal of RAGE overactivation, dramatically upregulates in experimental diabetic neuropathy dependent on the presence of RAGE.

AGEs are generated by aging and hyperglycemia, but with an accelerated rate in chronic diabetes (7). On their own, AGEs induce permanent abnormalities in extracellular matrix component function and have also been shown to be mutagenic (24). AGEs such as CML derived from glyceraldehyde and glycolaldehyde induce apoptosis and decrease viability in cultured Schwann cells, where they also increase the release of tumor necrosis factor-α and interleukin-6, as well as enhance activation of NF-κB (25). In the murine PNS, diabetes was associated with CML deposition, and its presence was unrelated to RAGE genetic status. This is an expected result, based on similar levels of hyperglycemia in WT and RAGE null diabetic mice driving the formation of AGEs such as CML. Although AGEs bind to other scavenger receptors, the best-characterized receptor for AGE is RAGE, which is a multi-ligand member of the immunoglobulin superfamily of cell surface molecules. AGE-RAGE ligation leads to cell activation and increased expression of extracellular matrix proteins, vascular adhesion molecules, cytokines, growth factors, and the generation of reactive oxygen intermediates. In vitro studies have identified RAGE to be important for mediation of neurite outgrowth and regulation of gene expression through NF-κB, a transcription factor, through the Ras–mitogen-activated kinase (MAP) kinase pathway (26). Activation of these signal transduction pathways may contribute to development of reactive oxygen species, such as with NADPH oxidase activation (27). One of these pathways includes activation of p21^ras, followed by activation of p44/p42 MAP kinases and nuclear translocation of NF-κB (28,29). NF-κB–regulated gene expression is detected in pathological samples where both RAGE and AGEs are present at high levels. RAGE and NF-κB had an associated upregulation in nondiabetic DRG neurons (Fig. 8) and a prominent association in diabetic DRG neurons. In this study, diabetic WT mouse DRG neurons and Schwann cells also had upregulated C-Rel nuclear translocation. Together with data revealing upregulated IκB, these findings link RAGE to upregulation and expression of NF-κB.

In vitro studies have also shown that administration of AGEs in a high-glucose environment potentiates PKC activation (30,31) in both macrophages and renal tubular cells. AGE-induced modification of signal transduction pathways occur through the induction of PKC activation, but may also occur independent of PKC activation (32). In mesangial cells, the independent activation of NF-κB and PKC pathways due to AGE presence in vitro is thought to be an early event contributing to oxidative stress (33). The interaction of AGEs with RAGE also leads to perturbation of vascular and inflammatory cells (9,11,12,24,26,34–36). RAGE and AGEs act to diminish vascular barrier function as demonstrated by the tissue-blood isotope ratio (37); vascular permeability may be reduced because of an enhanced expression of vascular cell adhesion molecule-1 (VCAM-1) (37,38). In our model, AGE- and RAGE-mediated changes of signaling pathways for NF-κB and PKC also target the diabetic PNS in our model.

Mice with long-term experimental diabetes developed changes in both electrophysiology and structure of their nerves and sensory ganglia relative to nondiabetic control littermates. Beginning at the DRG, sensory neurons demonstrate atrophy (loss of neuronal area) and mild neuronal loss after prolonged diabetes (15,39). Within the mixed motor-sensory sciatic nerve, diabetic axons are smaller in caliber and have thinner myelin relative to control fibers. These fibers had parallel electrophysiological abnormalities with progressive loss of motor potentials (compound motor action potentials) and declines in conduction velocity. In the sensory-dominant sural nerve, diabetic mice had frank loss of axons with fibers undergoing degeneration, in addition to atrophy and myelin thinning. Sciatic-tibial sensory nerve conduction studies demonstrated impairment of electrophysiological parameters in diabetic nerves relative to controls. At the most distal site of the PNS, diabetic mice had a loss of epidermal nerve fibers. Thus, this model replicated most of the features of progressive human diabetic polyneuropathy.

In this context, we identified robust RAGE expression at all levels of the PNS in experimental diabetes, including DRG, peripheral nerve, and epidermal fibers. In this model and in human peripheral nerves, the deposition of AGEs (10) likely contributes in turn to RAGE upregulation (40). In many pathological lesions, the abundance of RAGE-expressing cells has a strict correlation with accumulation of RAGE ligands, such as in diabetic vasculature, where highly expressing RAGE cells can be found proximal to AGE-abundant regions (41,42). Although RAGE expression in the peripheral nerve is clearly upregulated within neurons and axons, Schwann cells within diabetic peripheral nerve and DRGs have an even greater apparent overexpression of RAGE in this long-term model of diabetes. The presence of RAGE in epidermal axons of skin footpads is of considerable interest. If, as suggested by our findings, its overexpression in abnormal and residual diabetic epidermal axons reflects ongoing injury, its presence may be relevant to the targeting of small axons by diabetes, leading to small fiber neuropathy. Epidermal nerve fiber densities are indeed decreased in diabetic patients, findings that are associated with loss of sensory function and the development of neuropathic pain (43). Both diabetic and control epidermal axons expressed RAGE, but diabetics had loss of axons with higher expression in residual axons.

RAGE^−/− mice permitted a unique assessment of the potential role of the RAGE pathway in the development of neuropathy. Although AGE actions independent of RAGE ligation may contribute to ongoing toxicity within the extracellular matrix, AGE-RAGE interactions were significantly mitigated in RAGE null mice. The absence of RAGE attenuated both structural and electrophysiological changes within the peripheral nerves and the DRG after 5 months of diabetes. Previous studies have not evaluated the impact of RAGE deletion during the full repertoire of changes from neuropathy in a longer term model. Preliminary work by Bierhaus et al. (13) demonstrated that RAGE^−/− mice did not develop expected hypalgesia of diabetes. In addition, NF-κB activation was heightened in diabetic WT mice and was diminished in RAGE^−/− mice or with the intervention of soluble RAGE in WT mice (13). In our studies, RAGE^−/− mice peripheral neurons and supporting cells were protected from diabetes-induced upregulation of both NF-κB and PKCβII, even in the presence of long-term diabetes. Long-term upregulation of NF-κB, and possibly PKCβII, have both been previously linked to the development of neuropathy in diabetes (44). The upregulation of PKC in our mouse model was statistically significant but modest compared with that observed in the case of NF-κB. Previous studies in other species, or in experiments with shorter duration of diabetes, have demonstrated either mild PKC upregulation associated with microvasculature or absence of definite changes in PKC levels (44–46). It is important to note that RAGE^−/− diabetic mice did demonstrate evidence of mild morphological and electrophysiological deterioration nonetheless. It appears unlikely that RAGE- and AGE-mediated processes provide an exclusive explanation for the abnormalities identified in diabetic neuropathy. It is of interest that RAGE^−/− nondiabetic mice also had mild attenuation of age-related changes in morphology and electrophysiology, suggesting a possible role of RAGE in age-related decline.

In addition to its expression in neurons, NF-κB expression was also identified in glial cells. The role of NF-κB is complex and depends on age and on injury type and timing (47,48). When diabetic peripheral nerve is exposed to ischemia followed by reperfusion, NF-κB expression in Schwann cells of diabetic peripheral nerve rises (49). In human Schwann cell cultures, tumor necrosis factor-α contributes to the transient activation of NF-κB in the absence of apoptosis (50). In human patients with chronic inflammatory demyelinating polyneuropathy, Schwann cell nuclei demonstrate increased levels of NF-κB (50). For Schwann cells exposed to hyperglycemia in vitro, NF-κB expression is similarly elevated (51). In development, activation of the transcription factor NF-κB is an essential differentiation signal for axon-associated peripheral myelin formation, with its expression progressively declining during adulthood, presumably to prevent excessive myelination (52). Additionally, NF-κB activation within Schwann cells after angiotensin II's interaction with its receptor may be important in peripheral nerve regeneration (53). The present data demonstrate a consistent relationship with age- and diabetes-dependent upregulated expression of NF-κB in peripheral nerves, attenuated in RAGE^−/− mice with diabetes.

In conclusion, our data identify an important role for RAGE in a long-term model of experimental diabetes. Not only is heightened RAGE expression prominent and progressive at multiple levels (and in multiple cells) of the PNS, but its expression strictly parallels structural and electrophysiological alterations. Deletion of RAGE prevented a number of electrophysiological and morphological changes in the diabetic nervous system. Moreover, we provide evidence that a prominent downstream target of RAGE (NF-κB) is increased in the diabetic PNS, but not with RAGE deletion. Whereas overexpression of RAGE, and the consequent aberrant signaling of neurons and glial cells, is not the only mechanism of neurological damage in diabetes, our data suggest that it has an important contribution to neuropathy in this animal model of diabetes.

This study was supported by an operating grant from the Canadian Institutes of Health Research (CIHR) and the Canadian Diabetes Association (CDA). C.T. is a Clinical Investigator of the Alberta Heritage Foundation for Medical Research and D.W.Z. is a Scientist of the Alberta Heritage Foundation for Medical Research (AHFMR). A.M.S., L.L.R., and F.S. were supported by grants from the United State Public Health Service.