Diabetic polyneuropathy is the most common acquired diffuse disorder of the peripheral nervous system. It is generally assumed that insulin benefits human and experimental diabetic neuropathy indirectly by lowering glucose levels. Insulin also provides potent direct support of neurons and axons, and there is a possibility that abnormalities in direct insulin signaling on peripheral neurons relate to the development of this disorder. Here we report that direct neuronal (intrathecal) delivery of low doses of insulin (0.1–0.2 IU daily), insufficient to reduce glycemia or equimolar IGF-I but not intrathecal saline or subcutaneous insulin, improved and reversed slowing of motor and sensory conduction velocity in rats rendered diabetic using streptozotocin. Moreover, insulin and IGF-I similarly reversed atrophy in myelinated sensory axons in the sural nerve. That intrathecal insulin had the capability of signaling sensory neurons was confirmed by observing that fluorescein isothiocyanate-labeled insulin given intrathecally accessed and labeled individual lumbar dorsal root ganglion neurons. Moreover, we confirmed that such neurons express the insulin receptor, as previously suggested by Sugimoto et al. Finally, we sequestered intrathecal insulin in nondiabetic rats using an anti-insulin antibody. Conduction slowing and axonal atrophy resembling the changes in diabetes were generated by anti-insulin but not by an anti-rat albumin antibody infusion. Defective direct signaling of insulin on peripheral neurons through routes that include the cerebrospinal fluid may relate to the development of diabetic peripheral neuropathy.
Diabetic polyneuropathy is a common and debilitating acquired disorder of the peripheral nerve (1). Work in experimental models and in humans has demonstrated correction or prevention of polyneuropathy in diabetic subjects treated with insulin infusions to render euglycemia (2–4). Implicit in these studies, however, has been the assumption that prevention of neuropathy is exclusively an indirect benefit of insulin’s actions on glycemia. This, however, may not be the case. There is a possibility that lack of insulin itself, resistance to its actions, or decreased availability of its cousin molecules, the IGFs, contributes to neuropathy. Insulin and IGFs act as potent neuronal growth factors (5–7). Of particular interest are actions that promote neuronal survival, prompt neurons to synthesize neurofilament axon lattice proteins, and reverse diabetes-induced changes in neuronal mitochondrial function (8–10). In experimental models, low doses of insulin that do not influence hyperglycemia can correct conduction abnormalities in diabetic axons (10,11). Through forms of tyrosine kinase signaling similar to those used by classic neurotrophic molecules, insulin and IGFs signal through a number of common downstream second messengers. Indeed, impaired support of neurons by neurotrophic molecules and related growth factors, such as NGF, NT-3, and CNTF, has been considered relevant in the pathogenesis of diabetic neuropathy (12–14). However, a distinct feature of insulin and IGF-I may be a more widespread, or “pantrophic,” action on several classes of neurons.
In this article, we address the possibility that insulin, applied directly to perikarya, rather than peripherally or systemically, may be capable of supporting sensory neurons and altering properties of their peripheral axons in experimental diabetes. The impact of intrathecal infusions of insulin, its relative IGF-I, or carrier was examined on downstream motor and sensory conduction velocity and axon caliber. In additional work, we address whether sequestration of insulin, using an anti-insulin antibody in the cerebrospinal fluid, might generate distal axonal abnormalities that resemble those of diabetes. Overall, the findings support the hypothesis that deficient direct insulin signaling may play an important role in the development of diabetic neuropathy.
RESEARCH DESIGN AND METHODS
Sprague-Dawley rats weighing 300–350 g were made diabetic by a single intraperitoneal injection of freshly prepared streptozotocin (65 mg/kg; Sigma, St. Louis, MO) in vehicle (100 mmol/l citrate-phosphate buffer [pH 4.8]) during the fasting state. Control animals received equivalent volume doses of the citrate buffer solution. Fasting whole-blood glucose ≥16 mmol/l (normal 5–8 mmol/l) was our criterion for experimental diabetes. All procedures were carried out under pentobarbital anesthesia (65 mg/kg). The protocols were reviewed and approved by the University of Calgary Animal Care Committee using the Canadian Council of Animal Care guidelines.
Eight-week diabetic (n = 10) and nondiabetic control rats (n = 10) were anesthetized with sodium pentobarbital (65 mg/kg). We placed a silicone catheter (0.012 in × 0.025 in) connected to a second silicone catheter (0.025 in × 0.047 in) into the lumbar intrathecal space between the L6 and S1 vertebrae connected to a 2-week Alzet infusion pump in the dorsal back subcutaneous space. To confirm intrathecal delivery and catheter placement, we added India ink to some pumps to verify the presence of a subarachnoid infusate. The infusate had access to an extensive longitudinal proportion of the intrathecal space. Diabetic and nondiabetic rats received either 0.1–0.2 IU/day of regular humulin insulin and IGF-I equimolar to 0.1 IU insulin or its saline carrier daily for 4 weeks. In addition, nondiabetic rats (n = 6) were intrathecally administered either insulin-fluorescein isothiocyanate (FITC) (I2383; Sigma) equimolar to 0.1 IU insulin or its saline carrier for 2 weeks. Nondiabetic rats (n = 8) were also intrathecally infused with either anti-insulin (rabbit polyclonal antibody to rat insulin [H-86], 1:1,000; Santa Cruz Biotechnology, Santa Cruz, CA) or anti-albumin (rabbit polyclonal to BSA; Novus Biologicals, Littleton, CO) for 4 weeks.
Electrophysiology.
Recordings were carried out as described in previous work (15). Motor conduction in sciatic-tibial fibers was performed by stimulating at the sciatic notch and knee while recording the M wave (compound muscle action potential) from the tibial-innervated dorsal interossei foot muscles. Caudal sensory conduction was obtained in the tail by stimulating at distances of 40 and 60 mm from the recording electrodes. All stimulating and recording used 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.
Morphometry.
After 4 weeks of intrathecal infusion, left sciatic and sural nerves and left L4, L5, and L6 dorsal root ganglia (DRG) were harvested and processed as described in previous work (16). Samples were fixed in 2.5% glutaraldehyde in 0.025 mol/l cacodylate buffer overnight. Semithin (1 μm) sections of sural nerve, sciatic nerve, and lumbar L4–L6 DRG were cut on an ultramicrotome (Reichert, Vienna, Austria) and were stained with toluidine blue. Morphometric analysis was carried out using a Zeiss Axioskop at 1,000× magnification. Computer-assisted image analysis allowed for the determination of number and caliber of intact myelinated fibers. All morphological measurements were performed by a single microscopist blinded to the origin of the samples. The axonal area of 100 myelinated fibers within the sural or sciatic nerve and 200 perikarya containing nucleoli within the L4–L6 DRG of each rat were measured using Scion Image v.4.0.2.
FITC-insulin labeling and immunohistochemistry.
Sciatic nerve and L4, L5, and L6 DRG were harvested from rats after 2 weeks of intrathecal infusions of insulin-FITC (equimolar to 0.1 IU daily insulin infusions as above) or saline. Tissues were fresh frozen and sectioned at 14 μm. Slides were visualized on a fluorescent microscope (Axioplan 2; Carl Zeiss Canada, Toronto, ON, Canada). For immunohistochemistry, L4–L6 DRG from nondiabetic rats were harvested and immediately mounted in optimal cutting temperature. Cryostat sections (14 μm) were immunostained with a rabbit polyclonal antibody directed against the insulin receptor β subunit as primary antibody (anti-IRβ [C-19], 1:100; Santa Cruz Biotechnology). Incubation with FITC-conjugated goat anti-rabbit (1:50; Sigma, Mississauga, ON, Canada) as the secondary antibody was used to visualize insulin receptors.
Statistical analysis.
Data were calculated as means ± SE. In the electrophysiological work, data were analyzed by one-way ANOVA with appropriate Bonferroni corrected post hoc Student’s t test comparisons. In the morphometric work, mean axonal area was determined for each treatment group and compared using the Student’s t test (nonpaired two tailed). In all tests, statistical significance was set at α = 0.05.
RESULTS
Intrathecal delivery of insulin accesses and can signal sensory neurons.
Insulin labeled with FITC, delivered over a 2-week period by intrathecal catheter and miniosmotic pump, labeled L4–L6 DRG sensory neurons. There was discrete labeling of neuronal perikarya by insulin-FITC that had an intense and striking association with cell membranes and was occasionally perinuclear (Fig. 1A–D). No labeling of sciatic nerves was identified. That insulin receptors were present on ganglia sensory neurons to bind infused insulin was also confirmed. A large majority of L4–L6 DRG neurons exhibited cytoplasmic labeling by immunohistochemistry with an antibody directed to the β subunit of the insulin receptor, as described by Sugimoto et al. (17) (Fig. 1E).
Direct delivery of insulin or IGF-I improves conduction abnormalities in diabetic rats.
By 3 months of diabetes, hyperglycemia was present in diabetic rats (whole blood glucose >24 mmol/l). Insulin (or IGF-I) intrathecal and subcutaneous infusions did not alter glycemia. Sciatic-tibial motor and caudal sensory conduction velocities were significantly lower than in nondiabetic rats, as expected from the model (Fig. 2A and B). One month of subsequent saline carrier infusion had no impact on conduction slowing. In contrast, insulin infusion at 0.1 IU daily improved motor conduction slowing in diabetic rats (Fig. 2A). Infusion of a higher dose (0.2 IU insulin or equimolar [to 0.1 IU of insulin] IGF-I [5 × 10−6 mol/l] also improved motor conduction slowing and completely reversed slowing of sensory conduction velocity in diabetic rats (Fig. 2B). One month of 0.1 IU insulin daily subcutaneous infusion over the dorsal trunk of the rat did not alter slowing of motor or sensory conduction velocity (Fig. 2C and D). There was a trend, but not statistically significant, for higher conduction velocities in nondiabetic rats receiving insulin.
Direct delivery of insulin or IGF-I prevents axonal atrophy in diabetic rats.
The caliber of myelinated sensory fibers in sural nerve sections from nondiabetic (Fig. 3A) and diabetic (Fig. 3B) rats were measured after 3 months of diabetes and 1 month of infusion with an operator blinded to the origin of the nerves. Diabetic rats infused with a saline carrier developed evidence of myelinated fiber atrophy, with the mean axonal area significantly lower than that of nondiabetic rats. Infusion of 0.1 IU/day insulin, 0.2 IU/day insulin, or equimolar (to 0.1 IU insulin) IGF-I prevented myelinated fiber axon atrophy, such that the mean sural myelinated axonal area at the end point was similar to that of nondiabetic rats (Fig. 3C). In contrast, the mean surface area of 200 L5 dorsal root ganglia sensory neurons with identified nucleoli and their size distributions were not altered among the groups. Neuron area size distributions were also comparable between diabetic and nondiabetic rats (Fig. 3D).
Sequestering endogenous intrathecal insulin generates axonal abnormalities.
To address the possibility that endogenous intrathecal insulin contributes normally to the support of sensory neurons and axons, we sequestered native insulin in nondiabetic rats by intrathecally infusing an anti-insulin antibody (1/1,000) over a 4-week period. As a control, an additional group of rats were infused with anti-rat albumin antibody at an identical antibody dilution (1/1,000). Anti-insulin infusion was associated with the development of slowing of motor conduction velocity (Fig. 4) and atrophy of sciatic myelinated fibers (anti-albumin: 32.6 ± 0.8 μm2; anti-insulin: 24.8 ± 0.7 μm2, data not shown). Anti-rat albumin had no impact on conduction velocity.
DISCUSSION
The major findings of this work were as follows: 1) intrathecal administration of insulin has access to dorsal root ganglia sensory neurons where receptors for it are expressed, likely on perikaryal membranes; 2) local delivery of insulin, or IGF-I, to sensory neurons by intrathecal delivery influenced downstream physiological and structural alterations in axons rendered by diabetes: motor and sensory conduction velocity slowing and atrophy of sural myelinated axons; and 3) endogenous insulin in cerebrospinal fluid may offer normal maintenance support for peripheral nerve axon function and caliber.
Models addressing the pathogenesis of human diabetic neuropathy almost exclusively rely on hyperglycemia generated by loss of circulating insulin or resistance to its action. Hyperglycemia itself is highly likely to alter neurons, axons, and Schwann cells through one of several mechanisms that include excessive polyol flux (18), myo-inositol depletion (19), alterations in protein kinase C isoforms (20–22), glycosylation of structural nerve proteins such as tubulin or neurofilament (23,24), oxidative stress triggered by glucose auto-oxidation, or nerve and ganglion microvascular ischemia (25). Nonetheless, these insults arise in the setting of insulin loss or altered signaling, potentially a requisite source of ongoing support for peripheral neurons. Insulin receptors indeed exist both on perikarya and axons of peripheral neurons, also confirmed in this work (17,26). Its uptake could occur at one of multiple sites where neurons are relatively exposed to the circulation, for example, at the level of the less robust blood-ganglion barrier (27).
Commonalities in downstream tyrosine kinase activation may allow insulin, IGF-I and classic neurotrophin, and nonneurotrophin growth factors to support neurons in similar ways. As in this work, such growth factors ameliorate features of experimental diabetes separately from an action on hyperglycemia (13,28–30). In the case of insulin and IGF-I, this relationship is very close among these “cousin” molecules, since they can cross-occupy each other’s membrane receptors (5,31). Both signal through insulin receptor substrate (IRS)-1 and IRS-2 scaffold docking proteins to influence neuron survival and neurite outgrowth (32–34). Second messenger pathways affected by the IRS proteins include phosphatidylinositol 3-kinase, GRB2, NCK, CRK, fyn, SHB, SOCS1, SOCS3, and SHP2 (7). Of particular interest to the current work is the capability of insulin to influence neurofilament synthesis in neuron-like neuroblastoma cells (9).
IGF-I appears to protect sensory neurons from hyperglycemia-induced apoptosis (8), an important potential benefit in situations of incipient diabetes-related neuron loss. In our work, however, we suggest its impact was unrelated to neuron survival, since loss of sensory perikarya and axons is not a feature of early streptozotocin-induced diabetes in rats. Indeed, in much longer duration models than studied here, rigorous counting of neurons and axons has not identified loss out to 12 months of diabetes (16,35). This relative resistance to neuron loss in rat streptozotocin diabetes appears to differ from overt loss in diabetes induced by streptozotocin in mice (36; J.M. Kennedy, D.W.Z., unpublished data). The relationship between IGF-I and nonlethal alterations in diabetic neurons may be more challenging to dissect. Free levels of IGF-I are reduced in diabetic subjects, but such changes are also complicated by alterations in its binding proteins (6).
The benefits of insulin and IGF-I in our model appear more subtle than those of simple neuron protection, but they are not easily explained. Distal axon atrophy in diabetic models may be a function of reduced investment with neurofilament, either from impaired axonal transport or declines in perikaryal synthesis (37–39). The present study confirmed, as suggested by previous studies, that loss of neuronal axon caliber truly occurs distally first, since perikaryal atrophy did not accompany it. Such distal changes, however, may not be directly related to loss of neurofilaments, since mice that completely lack neurofilaments in their axons exhibit accelerated nerve atrophy when they become diabetic (D.W.Z., Sun, Cheng, and Eyer, unpublished data). Instead, a generalized decline in the synthesis of critical structural and other proteins, including neurofilament, appears to reflect the development of a degenerative disorder of sensory neurons (16). Diabetic neurofilaments also exhibit hyperphosphorylation, a change linked to heightened mitogen-activated protein kinase signaling in sensory neurons (40,41). Hyperphosphorylation of KSP domains on the heavy and medium subunits of the neurofilament triplet protein may alter their spacing, transport, or lifespan. Its apparent reversal by NT-3 also indicates how such support pathways may converge to reverse several diabetic alterations (13).
An early and potentially critical derangement recently described in diabetic sensory neurons is an inappropriate inner membrane depolarization of mitochondria (10). This change was reversible with subhypoglycemic doses of insulin. Failure of mitochondria to elaborate sufficient quantities of high-energy phosphates to maintain the function of neurons could explain a range of abnormalities that ensue. Direct delivery of insulin or IGF-I to neurons in our model might in turn correct a range of diabetic abnormalities by restoring optimal mitochondrial function both at the level of the perikarya and also downstream after normal mitochondrial transport. That such an influence was directed only to perikarya and secondarily to axons was indicated by the absence of insulin-FITC in sciatic nerves after intrathecal infusion.
Conduction velocity is directly related to axon caliber. In diabetic models, however, atrophy or neurofilament loss appears much later than changes in conduction velocity. Instead, diabetic conduction slowing may be better linked to metabolic changes including excessive sorbitol flux, reduced sodium-potassium ATPase pumping, intra-axonal accumulation of sodium, axonal depolarization, and impaired action potential ion flux (42). These changes in axon membrane excitability in turn might be associated with impaired mitochondrial generation of high-energy metabolites. Alternatively, changes in nodal protein investment or properties, including specific ion channels, might alter axon properties (43). Moreover, such slowing is not confined to myelinated axons (4).
Finally, we provide early evidence that there may be a role for endogenous insulin in normally supporting peripheral axons. An antibody directed against rat insulin, but not albumin, administered for 4 weeks generated slowing of motor conduction velocity and evidence of sciatic axon atrophy. Whereas these findings should be confirmed elsewhere using other kinds of approaches, they do provide evidence for a potent ongoing role of insulin in maintaining both physiological and structural properties of downstream axons. Our antibody may have sequestered insulin synthesized by the pancreas that penetrates cerebrospinal fluid from the bloodstream, indicating a major route of insulin access to peripheral neurons. Others have argued, however, that there may only be a limited relationship between blood insulin levels and brain insulin signaling (44). It is also possible that the nervous system synthesizes its own insulin and circulates it through cerebrospinal fluid, as has been suggested (45).
Atrophic axons likely withdraw terminals from tissue targets and eventually may be associated with death of parent perikarya. In human type 1 diabetes, insulin deficiency is severe and even intensive replacement regimens likely fail to mimic direct ongoing stable insulin actions. Nonetheless, intensive insulin therapy partly prevents the development of diabetic neuropathy. In type 2 human diabetes, many patients develop insulin resistance such that intermittent elevated circulating insulin levels fail to transduce intracellular signals. It is not inconceivable that neurons might share such resistance to the ongoing trophic support of insulin. Lack of insulin access to peripheral neurons may represent one additional failed support mechanism that contributes to neuropathy in the face of metabolic, structural, and ischemic insults.
Insulin-labeled FITC and the β insulin receptors in DRG. A and B: Absence of staining in L4–L6 DRG from rats injected intrathecally with saline for 2 weeks. Immunofluorescence (A) and light microscopy (B) of the same sections are shown. C: Several neurons in DRG labeled with membrane-bound insulin-FITC, viewed by immunofluorescence. D: Same section viewed by light microscopy. E: Insulin receptor β subunits expressed on the perikarya of sensory neurons in intact rat DRG. Scale bar = 20 μm.
Insulin-labeled FITC and the β insulin receptors in DRG. A and B: Absence of staining in L4–L6 DRG from rats injected intrathecally with saline for 2 weeks. Immunofluorescence (A) and light microscopy (B) of the same sections are shown. C: Several neurons in DRG labeled with membrane-bound insulin-FITC, viewed by immunofluorescence. D: Same section viewed by light microscopy. E: Insulin receptor β subunits expressed on the perikarya of sensory neurons in intact rat DRG. Scale bar = 20 μm.
Insulin and IGF-I reversed slowing of nerve conduction velocity. The motor (A, C) and sensory (B, D) nerve velocity (m/s) studies were performed after 4 weeks of treatment with saline, insulin, and IGF-I. ▪, control rats; 
, diabetic rats; 0, baseline; s, treated with saline; 0.1, insulin-treated rats (0.1 IU/day); 0.2, insulin-treated rats (0.2 IU/day); IGF-I, IGF-I-treated rats. C and D: , intrathecal insulin (0.1 IU/day) injection in diabetic rats for 1 month; ▪, subcutaneous insulin (0.1 IU/day) injection for 1 month. * and **, significant versus saline-treated diabetic group, P < 0.05 (A and B). *P < 0.05 (C and D).
Insulin and IGF-I reversed slowing of nerve conduction velocity. The motor (A, C) and sensory (B, D) nerve velocity (m/s) studies were performed after 4 weeks of treatment with saline, insulin, and IGF-I. ▪, control rats; , diabetic rats; 0, baseline; s, treated with saline; 0.1, insulin-treated rats (0.1 IU/day); 0.2, insulin-treated rats (0.2 IU/day); IGF-I, IGF-I-treated rats. C and D: , intrathecal insulin (0.1 IU/day) injection in diabetic rats for 1 month; ▪, subcutaneous insulin (0.1 IU/day) injection for 1 month. * and **, significant versus saline-treated diabetic group, P < 0.05 (A and B). *P < 0.05 (C and D).
Insulin and IGF-I reversed distal sensory axonal atrophy. A: Sural nerve from control rat (mean axonal area 22.7 ± 0.6 μm2) (A) and sural nerve from diabetic rat (mean axonal area 18.4 ± 0.5 μm2) (B). C: No significant difference in axonal area between insulin or IGF-I infused diabetic and saline control rats. Cs, saline control rats; Ds, saline diabetic rats; D0.1, insulin diabetic rats (0.1 IU/day); D0.2, 0.2 IU/day insulin; DIGF-I, IGF-I diabetic rats. D: In the frequency size distribution of neuronal surface area of DRG neurons, neither diabetes nor insulin changed the mean area of L5 neuronal perikarya. *P < 0.05.
Insulin and IGF-I reversed distal sensory axonal atrophy. A: Sural nerve from control rat (mean axonal area 22.7 ± 0.6 μm2) (A) and sural nerve from diabetic rat (mean axonal area 18.4 ± 0.5 μm2) (B). C: No significant difference in axonal area between insulin or IGF-I infused diabetic and saline control rats. Cs, saline control rats; Ds, saline diabetic rats; D0.1, insulin diabetic rats (0.1 IU/day); D0.2, 0.2 IU/day insulin; DIGF-I, IGF-I diabetic rats. D: In the frequency size distribution of neuronal surface area of DRG neurons, neither diabetes nor insulin changed the mean area of L5 neuronal perikarya. *P < 0.05.
Intrathecal anti-insulin antibody generated axonal abnormalities. Intrathecally injected anti-insulin antibody induced slowing of peripheral motor conduction velocity. Nerve conduction studies were performed on nondiabetic rats before and after 2 or 4 weeks of treatment with anti-insulin or anti-albumin antibody (1/1,000 dilution). All data are presented as means ± SE. *Significant versus the baseline group (P < 0.05); **significant versus the anti-albumin-treated group (1/1,000) (P < 0.05).
Intrathecal anti-insulin antibody generated axonal abnormalities. Intrathecally injected anti-insulin antibody induced slowing of peripheral motor conduction velocity. Nerve conduction studies were performed on nondiabetic rats before and after 2 or 4 weeks of treatment with anti-insulin or anti-albumin antibody (1/1,000 dilution). All data are presented as means ± SE. *Significant versus the baseline group (P < 0.05); **significant versus the anti-albumin-treated group (1/1,000) (P < 0.05).
Article Information
This study was supported by an operating grant from the Canadian Institutes of Health Research and the Canadian Diabetes Association. D.W.Z. is a Senior Medical Scholar of Alberta Heritage Foundation for Medical Research.
We thank Brenda Boake for providing expert secretarial assistance.
