Intranasal Insulin Ameliorates Experimental Diabetic Neuropathy

We studied a total of 484 male CD1 wild-type mice with initial weight of 20 to 30 g housed in plastic sawdust covered cages with a normal light-dark cycle and free access to mouse chow and water. In all cases, mice were raised and studied in strict pathogen-free environments. All protocols were reviewed and approved by the institutional animal care and use committee at Regions Hospital (21 mice, experiment 1) and the University of Calgary Animal Care Committee using the Canadian Council of Animal Care guidelines (463 mice, experiment 2). Mice were anesthetized with pentobarbital (60 mg/kg) before all procedures. At the age of 1 month, 304 mice were injected with streptozotocin (STZ) (Sigma, St. Louis, MO) intraperitoneally once daily for each of 3 consecutive days with doses of 60 mg/kg, 50 mg/kg, and then 40 mg/kg with the remaining 180 mice injected with carrier (sodium citrate) for 3 consecutive days. Studies using harvested tissues occurred after 1 month (36 diabetic mice, 30 nondiabetic mice), 3 months (48 diabetic mice, 30 nondiabetic mice), 5 months (60 diabetic mice, 30 nondiabetic mice), and 8 months (160 diabetic mice, 90 nondiabetic mice).

Whole blood glucose measurements were performed monthly with puncture of the tail vein and a blood glucometer (OneTouch Ultra Meter; LifeScan Canada, Burnaby, BC, Canada). Hyperglycemia was verified 1 week after STZ injections with a fasting whole blood glucose level of 16 mmol/l or greater (normal 5–8 mmol/l), our definition for experimental diabetes. All animals were weighed monthly. Mice were followed and harvested at 1, 3, 5, or 8 months of diabetes (>9 months of life). Mice that did not develop diabetes were excluded from further assessment.

Animals were inspected twice daily and examined for signs of depressed level of consciousness, ataxia, or general malaise. When such signs were identified, whole blood glucose testing was performed with a measurement of less than 3.5 mmol/l defined to represent hypoglycemia. No intervention was performed at any time with regard to additional insulin, glucose, or fluid delivery. In situations in which the mouse was obviously ill, euthanasia was performed. In circumstances in which severe hyperglycemia was found (>33 mmol/l) in an ill mouse, euthanasia was performed.

We studied cohorts with a maximum of eight mice in each group initially as a result of resource limitations. After the initial cohorts containing eight mice each were studied, a second cohort was used to obtain additional mouse data for mouse cohorts with higher levels of mortality. For any animal that experienced mortality after the 20-week point of the sensorimotor studies, the data were carried through using the last obtainable data point.

In this work, delivery of subcutaneous saline is indicated as “S-S,” subcutaneous insulin as “S-I,” intranasal saline as “I-S,” and intranasal insulin as “I-I.”

¹²⁵I-labeled I-I administration was performed to determine distribution of intranasally delivered insulin in 21 nondiabetic mice. Before experimentation, animals were acclimated for handling during awake intranasal delivery for 2 weeks. ¹²⁵I-labeled I-I was provided to 12 CD1 mice (male, 6–8 weeks; Charles River) and ¹²⁵I-labeled S-I was provided to nine mice under pentobarbital anesthesia (60 mg/kg). Insulin (Humulin R; Eli Lilly, Toronto, Canada) with an initial concentration of 100 units/ml or 4033.98 mg/ml was dissolved in PBS and custom-labeled with ¹²⁵I (GE Healthcare, Piscataway, NJ). Radiolabeled insulin solution contained 344.3 uCi/μg at synthesis. ¹²⁵I-labeled I-I delivery was performed in a fume hood behind a lead-impregnated shield with anesthetized mice placed supine. A mixture of ¹²⁵I insulin (15.8 μCi) and unlabeled insulin (3.3 μg) were administered as I-I or S-I. ¹²⁵I I-I was delivered as eight 3-μl drops with an Eppendorf pipette over alternating nares every 2 min for a total volume of 24 μl. For subcutaneous delivery, ¹²⁵I S-I was delivered with a single subcutaneous injection of 24 μl in a fume hood behind a lead-impregnated shield.

At each of 1, 2, and 6 h after ¹²⁵I I-I or S-I delivery, cardiocentesis was performed to extract blood followed by euthanasia using transcardial perfusion using 120 ml of 4% paraformaldehyde under anesthesia. To quantify ¹²⁵I distribution, portions of the nervous systems were harvested along with blood, urine, lymphatic, and visceral organ structures. Olfactory epithelium and trigeminal nerve were examined as a result of their role in intranasal delivery into the nervous system and cerebrospinal fluid. Gamma signal was quantified in each tissue using a Packard Cobra II auto-gamma counter (PerkinElmer Life and Analytical Sciences, Waltham, MA). Concentrations of ¹²⁵I insulin were calculated based on the gamma counting data, tissue weight, specific activity of the insulin administered, and measured standards.

Daily I-I (Humulin R; Eli Lilly) and intranasal saline (I-S) was administered to either diabetic or nondiabetic male CD1 mice after a 1-week training period immediately after STZ injection using only intranasal saline for accustoming mice before diabetes verification. Although each mouse was held in a supine position while in neck extension, a total of 24 μl containing either a total of 0.87 IU of insulin or 0.9% saline only was provided as four drops of 6 μl each through Eppendorf pipette over alternating nares every 1 min. Daily S-I (0.87 IU) (Humulin R) and S-S were also administered daily to either diabetic or nondiabetic male CD1 mice at the same dose. All therapies began immediately after confirmation of presence of diabetes for each cohort. In the first week, daily glucometer testing was performed for all mice followed by once-monthly testing and during times of illness.

We attempted to use other control groups but found their usefulness to be limited in each case. The delivery of S-I by a sliding scale approach requiring daily whole blood glucose sampling through repeated punctures or the use of a protected venous catheter led to intolerable rates of infection and/or tail amputation, so the morbidity accrued with this procedure was deemed unacceptable and a potential confounder with any behavioral studies. Delivery of reduced subcutaneous doses of insulin failed to modify glycemic levels. Therefore, we selected the S-I dose to be equivalent to I-I dosing (0.87 IU) for the cohorts studied.

Electrophysiological assessment of sciatic nerve function was performed as previously described (13) under halothane anesthesia. Initial baseline studies were carrier out before STZ or carrier injections; no significant difference between groups was identified. Of all cohorts of mice receiving I-I, S-I, I-S, or S-S, both diabetic and nondiabetic, at least five mice in each group underwent monthly electrophysiological testing beginning before induction of diabetes and after 1, 2, 4, 6, and 8 months of diabetes. For orthodromic sensory conduction studies, the sural nerve was used with a fixed distance of 30 mm from platinum subdermal stimulation needle electrodes (Grass Instruments, Astro-Med, West Warwick, RI) to the sciatic notch where recording electrodes were placed to measure the sensory nerve action potential (SNAP) amplitude and sensory nerve conduction velocity (SNCV). Near-nerve temperature was kept constant during testing at 37 ± 0.5°C using a heating lamp.

A total of 10 mice in each cohort had behavioral testing performed twice monthly to evaluate mechanical and thermal sensation. A 2-week training period was performed to acclimatize mice to the procedure immediately after STZ injections and during diabetes verification. Mice were placed in a Plexiglas cage on a glass plate (for thermal testing) or on a plastic mesh floor (for tactile testing) and were allowed to acclimate for at least 5 min before recording in all cases. Thermal sensation was tested using the Hargreaves apparatus (24). In brief, a radiant heat source was applied individually to the middle of either hind paw for up to 60 s with the latency (seconds) to withdrawal measured. Heating rate ramped from 30 to 58°C over 60 s in consistent fashion on each occasion. Paws were inspected before and after thermal testing to ensure that no evidence of thermal damage was present. There were 5-min intervals provided between a total of three trials performed during the same day. To quantify mechanical sensitivity of the foot, withdrawal in response to a stimulus consisted of sequentially smaller von Frey filaments (25). Five trials were performed at each filament size with a total of three or more withdrawals to up-and-down movements indicating sensitivity at the smallest positive filament size. Mechanical and thermal testing was performed on identical days with an interval of at least 1 h between the two tests.

Harvesting of tissues and their subsequent morphometric processing and analysis along with quantification of epidermal fibers and all other molecular testing (Western immunoblotting, polymerase chain reactions, and electrophoretic mobility shift assays) have been described in detail in an online appendix (available at http://diabetes.diabetesjournals.org/cgi/content/full/db08-1287/DC1).

All statistical comparisons were intended between the following groups: diabetic I-I and S-I; diabetic I-I and I-S; diabetic I-I and control I-I; diabetic S-I and S-S; diabetic S-I and control S-I; control I-I and S-I; control I-I and I-S; and control S-I and S-S. Comparison testing was not performed between other grouped cohorts with Bonferroni corrections applied as appropriate for these group comparisons.

Data collected in the groups were expressed as mean ± SE in all cases. Data from each individual mouse were used to obtain means in each case. One-way matched/unmatched ANOVA and Student's t tests were performed to compare means between diabetic and control groups. For immunohistochemistry comparisons demonstrated as low/medium/high intensity, the individual values were compared using unmatched ANOVA testing. Also, AUC statistical testing was performed for behavioral testing performed and was calculated using the trapezoidal method. Again, only the groups intended to have statistical comparisons were analyzed as such. Correlational relationships for AUC were tested using multiple linear regression analysis. For the purposes of molecular studies and comparisons, in some cases, only one control (nondiabetic) group was used as a control value with subsequent comparisons to other diabetic groups for the molecular test studied; Bonferroni corrections were made as appropriate depending on the number of relevant comparisons.

At 1 h after I-I or S-I delivery, insulin concentrations were higher within cervical spinal cord, DRGs, and spinal dura with I-I delivery (Fig. 1). Insulin concentrations in blood were substantially higher after S-I delivery (850× greater), but insulin concentrations within kidney, liver, and lung were higher after I-I delivery.

At 6 h after I-I or S-I delivery, insulin concentrations were higher in DRGs and spinal dura after S-I delivery compared with I-I delivery (Fig. 1). After 6 h, blood concentrations of insulin were higher after I-I delivery with this peak detected at 6 h. Insulin concentrations in systemic organs were now higher after S-I delivery. When 2-h data are considered, I-I delivery led to peaks in insulin concentration within DRGs and systemic organs after 1 h and peaks in blood concentrations of insulin after 6 h. S-I delivery, in contrast, led to peaks in insulin concentration at DRG and systemic organs after 6 h, whereas blood concentrations peaked after 1 h. Blood concentrations of insulin after S-I delivery peaked at a value nearly 1,000× the peak value obtained with I-I delivery.

Mice receiving I-I treatment maintained good health throughout the 1-, 2-, and 6-h monitoring periods before being euthanized, whereas S-I delivery led to more frequent development of hypoglycemia-induced illness, including death, and also to reduced consciousness levels in many mice.

After STZ injection, mice developed diabetes within 2 weeks in greater than 85% of animals, and in each case, diabetes was maintained over the length of the study. Diabetic mice were smaller than nondiabetic mice within 1 month after STZ injection, and diabetic mice had smaller body weights throughout life (Table 1); diabetic mice receiving I-I maintained weight better than the cohort receiving I-S. Hyperglycemia was identical in mice receiving I-I or I-S, but S-I mice had more documented hyperglycemia and more episodes of illness or death associated with confirmed hypoglycemia (Table 1). Mouse glycated hemoglobin was increased in all diabetic mice after more than 9 months of life and was identical between I-I and I-S mice but was reduced in surviving S-I mice (Table 1). The mortality rate in diabetic mice was significantly higher than in nondiabetic mice, although diabetic I-I mice had improved mortality relative to diabetic I-S, diabetic S-S, and diabetic S-I mice (Table 1).

Before induction of diabetes, there were no baseline differences in sensory behavior testing between any of the mouse cohorts identified. All diabetic mice developed tactile allodynia after 5 to 7 weeks of diabetes when compared with nondiabetic mice (Fig. 2). Diabetic I-I mice developed less allodynia in the first 3 months of diabetes (Fig. 2) as compared with the other diabetic cohorts. In the later stages of diabetes, I-I mice had less tactile sensory loss than other diabetic mouse cohorts, although diabetic S-I mice had greater maintenance of sensation for tactile stimuli than diabetic mice receiving S-S or I-S (Fig. 2). During serial thermal testing, each cohort of diabetic mice also developed thermal hypersensitivity when compared with nondiabetic mice after several weeks, earlier than identified in other species or mouse strains (Fig. 2). However, diabetic I-I mice demonstrated less hypersensitivity in the early stages of diabetes and maintained thermal sensation better in the later stages of diabetes. Diabetic S-I mice had less thermal sensory loss in the later stages of diabetes (Fig. 2) when compared with diabetic mice receiving S-S and S-I. Within the first 20 weeks, AUC measurements were statistically different for diabetic I-I mice as compared with other diabetic mouse cohorts for both tactile and thermal testing, and all diabetic cohorts had thermal hypersensitivity and tactile allodynia when compared with nondiabetic mice (Fig. 2).

Before induction of diabetes, there were no baseline electrophysiological differences between any of the mouse cohorts identified. Sensory amplitudes and sensory conduction velocities demonstrated age-related declines over time in both diabetic and nondiabetic cohorts (Fig. 3, with more rapid declines in diabetes. Overall reductions in SNAP amplitudes and in SNCV occurred in diabetic mice relative to nondiabetic mice beginning after 2 to 3 months, similar to previous results (13). Protection against declining SNAPs and SNCV occurred beginning at 3 to 4 months in diabetic I-I mice as compared with all other cohort groups (Fig. 3) and slightly later in diabetic S-I mice (Fig. 3). Overall, in comparison to S-S, I-S, and S-I diabetic mice, diabetic I-I mice displayed significant protection against electrophysiological deterioration in sensory function.

In diabetic mice, sciatic nerves had axonal atrophy without axonal loss identified (supplemental Table 1), and sural nerves from diabetic mice developed a loss of fiber density and axonal area (atrophy) relative to nondiabetic controls (Table 2 and supplemental Fig. 1). Diabetic I-I mice were protected from declines in sural axon density and from axonal atrophy in both the sural and sciatic nerves after 8 months of diabetes (Tables 2 and 3) when compared with diabetic S-I, S-S, and I-S mice (supplemental Tables 1 and 2 and supplemental Fig. 1). Diabetic sural and sciatic nerve myelin thickness was also reduced after 8 months of diabetes (supplemental Tables 1 and 2) with some protection offered by I-I delivery.

DRG neurons developed mild neuronal atrophy and loss of density with exposure to long-term diabetes (supplemental Fig. 2 and Table 3), except in the case of diabetic I-I mice, in which no significant loss in neuronal density was identified, indicating neuronal protection (Table 3). Similarly, diabetic I-I mice again had preservation of DRG neuronal size and density when compared with diabetic I-S and diabetic S-S mice (supplemental Fig. 2 and Table 3). There was evidence of partial protection among diabetic S-I mice (supplemental Fig. 2 and Table 3). DRG neurons from diabetic I-I mice had elevated levels of PI3K, Akt, and pAkt mRNA and protein (Fig. 4).

The hind footpad epidermal nerve fiber density of all mouse cohorts with diabetes was reduced compared with nondiabetic mice (Fig. 5). Overall, diabetic I-I mice were protected from epidermal fiber loss (Fig. 5), whereas diabetic S-I mice had better maintained epidermal nerve fiber density than diabetic S-S or diabetic I-S mice by the 8-month end point. The separate methods of analyses for epidermal nerve fiber densities yielded similar results.

Quantification of mRNA and protein for Akt and PI3K demonstrated general downregulation in diabetic tissues with at least partial reversal of both PI3K and Akt mRNA and protein levels occurring in diabetic I-I mice (Fig. 4) in DRG, but not in sciatic nerve. Diabetic I-I mice also had partial protection from downregulation of pGSK3β, GSK3β, and pCREB (Fig. 4). Finally, CREB protein binding to DNA was identified to be depressed with diabetes (supplemental Fig. 3) with partial resolution in diabetic I-I mice.

I-I protected diabetic mice from behavioral, structural, and molecular changes associated with DPN. We propose that insulin's neuroprotective effects on the PNS are the result of restoration of the PI3K/Akt pathway components (Fig. 4). Additionally, I-I led to less mortality than S-I delivery and provided greater protection against the effects of long-term diabetes on the PNS.

Replacement of insulin in a type I model of diabetes through either I-I or S-I delivery led to improvements in behavioral, electrophysiological, morphological, and molecular status (Figs. 2,FIG. 3,FIG. 4–5 and supplemental Figs. 1–3) related to diabetes. Paradoxically, S-I delivery in diabetic mice led to greater mortality (Table 1), relating in part to episodes of hypoglycemia, a complication avoided by I-I. Although S-I led to improved glycated hemoglobin levels at final end point, this effect was not seen in diabetic I-I mice. These results suggest that the beneficial effects of I-I in diabetes are not primarily related to corrections in hyperglycemia (Table 1).

Insulin, a highly conserved peptide, has now emerged as a key neurotrophic factor in the nervous system, a role that is lost in type 1 diabetes. The major site of insulin's activity, the insulin receptor, is found in high concentrations among DRG neurons and myelinated sensory root fibers and in lesser concentrations on myelinated anterior root fibers and in the ventral horn of the spinal cord (16,26). Intrathecal insulin prevents degeneration and promotes regeneration in injured peripheral nerve (16). Meanwhile, systemic or intrathecal insulin delivery prevents diabetes-mediated electrophysiological changes (27), whereas intrathecal insulin restores distal skin epidermal innervation (17). In vitro, insulin exerts a direct neuritic outgrowth effect through insulin receptors or perhaps through crossactivation of IGF-1 receptors (28). Insulin's greatest impact appears to be at the level of the DRG, where insulin may prevent a “dying-back” that begins in the most distal epidermal fibers (4,29).

Insulin stimulation upregulates protein-tyrosine phosphorylation (30) through downstream activation of IRS-2 (18). Insulin also modulates the inner mitochondrial membrane potential through activation of the PI3K pathway (31), stimulating phosphorylation of Akt and Akt substrates such as CREB (32,,–35). PI3K promotes translocation of voltage-dependent calcium channel currents to the neurolemma in an Akt-dependent manner (36). Activated Akt is important for sensory neurite extension and branching (37), and the PI3K–Akt pathway has a positive regulatory effect on myelin-associated glycoprotein (MAG) expression in Schwann cells, Schwann cell differentiation (38), and promotion of myelination (39) through Schwann cell biosynthesis of fatty acids (40). Similar to insulin, IGF-I also activates the PI3K/Akt pathway (41), leading to phosphorylation of Akt effectors, including CREB and GSK-3β (19). IGF-I also leads to accumulation of pAkt within DRG neuronal nuclei and increases CREB-mediated transcription (19). In our studies, prevention of downregulation of PI3K/Akt (Fig. 4) in murine DRG was associated with amelioration of behavioral, electrophysiological, and morphological changes pertaining to diabetes.

CREB is also a neuroprotective molecule; CREB phosphorylation inhibits apoptosis in embryonic neurons (42), whereas the loss of CREB results in impaired axonal growth (43). Meanwhile, GSK-3β, downstream of PI3K/Akt, is a neuron-specific promoter of apoptosis in DRG neurons when it is active (nonphosphorylated state) (44). Phosphorylation of GSK-3β by Akt renders it inactive, leading to antiapoptotic properties (44,45). GSK-3β also regulates the transcriptional activities of CREB (46,47) and may regulate gene expression and activity of transcriptional factor binding to the MAG promoter region (38). GSK-3β also promotes Schwann cell differentiation, suggesting that the PI3K/Akt/GSK-3β pathway is crucial for initiation and possibly maintenance of myelination through promotion of MAG expression (38). Thus, insulin may be important in maintaining conduction velocities (Fig. 3) by direct effects on Schwann cells. In our studies, I-I delivery was associated with elevation of pCREB and pGSK-3β levels and reversal of diabetes-associated suppression of CREB-DNA binding within diabetic mouse DRGs (Fig. 4 and supplemental Fig. 3).

Intranasal administration allows insulin to bypass the blood-brain barrier and enter the brain and spinal cord parenchyma, as well as CSF, within 1 h. Its entry into the nervous system and CSF likely occurs through extracellular bulk flow transport along both olfactory and trigeminal neural pathways and may use perivascular channels of blood vessels entering the CNS (21). This method of insulin delivery permitted us to study the impact of insulin without affecting glycemia levels such as occurs with systemic insulin delivery, easing difficulty in dissecting the relative contributions of hyperglycemia and insulin's trophic properties (48). Prior studies using intranasal delivery of insulin-like molecules such as IGF-1 have demonstrated safety and efficacy in experimental stroke (22). I-I delivery in humans has led to improvements in memory (23) within minimal impact on plasma glucose levels, which remain in the euglycemic range (20,23). Although the use of I-I for the management of systemic diabetes has been limited, in our mouse cohorts, I-I was also associated with better maintenance of body weight and improved mortality (Table 1).

Limitations of our results using I-I delivery in diabetic mice must be acknowledged. Our results must be considered under the limitations of working with a murine model, and the inability to achieve a long-term model of murine type 1 diabetes with optimal glycemic management as a suitable control group. The mouse cohorts were subjected to intensive testing throughout their lifetime, which may have led to stress impacting on behavioral testing results. Diabetic CD1 mice developed sensory behavioral changes earlier than has been observed in other rodent models of diabetes, which may limit portability of these findings to other models. It is also possible that hypoglycemia may have impacted on sensorimotor testing; the impact of hypoglycemia on the diabetic I-S and control I-S cohort groups was anticipated but unavoidable. Distribution of insulin within the diabetic nervous system may differ from results obtained in nondiabetic mice examined in our radiolabeling studies. In addition, although I-I–obtained concentrations were higher in nervous system tissues, tissue concentrations at later time points were not different between I-I and S-I delivery, suggesting that differences in systemic exposure and possibly different metabolic rates for insulin may also play a role in our results. Based on our studies, it is difficult to develop a more appropriate control group of diabetic mice with long-term glycemic control based on the STZ-induced diabetic model. However, our results also provide evidence for potentially robust benefits of insulin independent of its actions on hyperglycemia. These results support its role as an important neurotrophic factor in the management of diabetic neuropathy. Our results support the development of human I-I clinical trials for the prevention and slowing of the development of DPN.

This study was supported by an operating grant from the Alberta Heritage Foundation for Medical Research (AHFMR) 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).

W.F. is the inventor of a patent for intranasal insulin (Neurologic agents for nasal administration the brain. World Intellectual Property Organization. PCT priority date 5.12.89, WO 91/07947; 1991), a patent and technique that has no goal of commercialization and therefore no measurable financial conflict of interest. Although W.F. is an inventor on this patent about intranasal insulin, this patent is wholly owned by Chiron/Novartis and, to the best of our knowledge, the company has no intent to commercialize intranasal insulin. None of the other co-authors have any relationship with Chiron/Novartis. No other potential conflicts of interest relevant to this article were reported.