Neurotrophin-3 Prevents the Proximal Accumulation of Neurofilament Proteins in Sensory Neurons of Streptozocin-Induced Diabetic Rats

Male Wistar rats (300 g) were made diabetic by a single injection of STZ (55 mg/kg i.p.; Sigma) and maintained in parallel with age-matched control animals. Tail blood glucose was assayed 3 days after injection using glucose test strips (BM-Accutest; Roche Diagnostics, Basel, Switzerland) to confirm diabetes. All diabetic animals had blood glucose values >30 mmol/l. Rats were maintained for 14 weeks with free access to water and laboratory diet. One group of diabetic rats received injections (5 mg/kg s.c.) of human recombinant NT-3 (a gift from Regeneron Pharmaceuticals, Tarrytown, NY) three times weekly for the final 10 weeks of the 14-week diabetes study. Final body weights were as follows: age-matched controls, 629 ± 18.5; STZ-induced diabetic rats, 347 ± 16.0; and STZ-diabetic + NT-3 rats, 389 ± 19.0 (means ± SE; n = 6–8; P < 0.001 for control vs. other groups by one-way ANOVA). Immediately after the death of the animals, a blood sample was taken and the plasma glucose concentration was determined using the GOD-PERID test kit (Boehringer Mannheim, Germany).

NCVs were measured by subtracting the distal latencies for M waves (MNCV) and H reflexes (SNCV) elicited by stimulation at the sciatic notch and Achilles tendon. The principles and practicalities of these methods have been described in detail elsewhere (26). In the present investigation, the measurements were performed on rats under isofluorane anesthesia. The core body temperature was monitored by a rectal probe and maintained at ∼37°C. The temperature adjacent to the sciatic nerve was monitored by a microthermocouple connected to an electronic thermometer (Comark Electronics, Rustington, U.K.). NCVs were assessed in all animals during the final week before being culled. The procedure was approved as part of the Home Office Project License awarded to D.R.T.

Sural nerves were processed for light microscopic examination by immersion fixation in 2.5% phosphate-buffered glutaraldehyde (37°C) for ∼3 h. The nerves were then fixed overnight in 2.5% phosphate-buffered glutaraldehyde at 4°C. Tissue was postfixed in 1% aqueous osmium tetroxide for 3–4 h before being dehydrated using a series of graded alcohols and propylene oxide. After being infiltrated with a 1:1 mixture of propylene oxide and araldite for 4 h, nerves were placed in 100% araldite overnight before being embedded in fresh araldite resin. Thick sections (1 μm) were cut with glass knives and stained with p-phenylenediamine before they were studied with a light microscope. Computer-assisted analyses of axonal size-frequency distributions of myelinated fibers were performed on nerves, as described earlier (25). Single sections of the sural nerve from each animal were sampled. A video image was obtained for each nerve with an Olympus BH-2 light microscope and Cohu 5000 series television camera interfaced with a Macintosh Quadra 850AV computer running the National Institutes of Health’s Image 1.55 software. Myelinated fibers with axons >1 μm in diameter were individually identified and selected before being sorted with an automated process into bins based on axonal diameter. In each case, systematic sampling of nonoverlapping fields resulted in sampling 50–80% of the total myelinated fibers in each nerve. Neurofilament density and number in the sural nerve were calculated as previously described (27). Thin sections (60 nm) were cut with a diamond knife and stained with uranyl acetate and bismuth subnitrate before being viewed with a Zeiss 10 electron microscope operated at 80 keV. Electron micrographs of cross-sections of axons from 3–15 myelinated fibers per nerve were taken at a magnification of ×10,000 and printed at a final magnification of ×58,000. An acetate with templates of evenly spaced circles was put over each print, and the number of neurofilaments within each circle was counted to determine density. The total area of each axon was then digitized and multiplied by neurofilament density to determine total neurofilament number.

The L4−L5 DRG, proximal nerve (5 mm of spinal nerve directly adjacent and distal to the DRG, containing mainly sensory and motor fibers), and sural nerve (mainly sensory, but with some autonomic content) were isolated from control, diabetic, and NT-3–treated animals and immediately frozen on dry ice. The tissues were homogenized in 450–500 μl homogenization buffer (50 mmol/l Tris-HCl [pH7.4], 1% NP-40, 0.25% sodium deoxycholate; 150 mmol/l NaCl; 1 mmol/l EGTA; 1 mmol/l phenylmethylsulfonyl fluoride; 1 μmol/ml aprotinin; leupeptin; pepstatin; 1 mmol/l Na₃VO₄; and 1 mmol/l NaF). Protein concentrations were determined using a bromophenol-blue–based protein assay. For SDS-PAGE, 5 μg of protein were loaded into each well and then transferred to nitrocellulose. The nitrocellulose paper was rinsed with 0.5% Tween-20 in PBS for 10 min. After being blocked with 5% nonfat dry milk in PBS for 1 h at room temperature, the blotted nitrocellulose was incubated with diluted primary antibodies against total NF-H (COOH-terminal antibody from Chemicon; diluted 1:10,000), total NF-M (RM044 from Zymed; diluted 1:10,000), phosphorylated NF-H and NF-M (SMI-31 from Sternberger; diluted 1:10,000), total NF-L (Chemicon; diluted 1:1,000), total JNK (Santa Cruz Biotechnology, Santa Cruz, CA; code FL; diluted 1:1,000), phosphorylated JNK (Santa Cruz; code G-7; diluted 1:200; phosphorylated on Thr-183 and Tyr-185), total p38 (New England Biolabs, Beverley, MA; code 9212; diluted 1:1,000), phosphorylated p38 (Biosource International, Camarillo, CA; code 44–684; diluted 1:2000; phosphorylated on Thr-180 and Tyr-182), and phosphorylated ERK (New England Biolabs; code 9101; diluted 1:4,000; phosphorylated on Thr-202 and Tyr-204). Incubations were followed by treatment with secondary antibody (horseradish peroxidase−conjugated anti-mouse/rabbit secondary antibody [Cell Signaling, 1:2,500]) at room temperature for 1.5 h. Enhanced chemiluminescence (LumiGlu; Cell Signaling) was used to detect the signal from the blot. The bands in the X-ray film were scanned and quantitated using the Molecular Analysis image analysis software package (Version 1.5, Biorad).

Lumbar DRGs from control, diabetic, and NT-3 treated animals (n = 3) were fixed in situ, as previously described (10). These animals were used at the 12-week time point of the 14-week study. Animals were anesthetized by injection of sagatal (1 ml/kg i.p.). Blood vessels were perfused with ice-cold saline via the left ventricle with a peristaltic pump (Watson Marlow, Falmouth, U.K.) at a rate of 25 ml/min for 8 min to replace blood, followed by perfusion with ice-cold 4% paraformaldehyde (in 0.1 mol/l phosphate buffer [pH 7.4]) for 15 min. L4−L5 DRG were dissected and postfixed in 4% paraformaldehyde for 4 h at 4°C. The fixed tissues were washed in PBS overnight at 4°C and soaked in 30% sucrose for 2 h at 4°C. The tissues were put in embedding medium and placed on dry ice until frozen. Sections (11 μm) were cut and placed on polylysine-coated slides. Nonspecific binding was blocked by incubation with 10% donkey serum for 2 h at room temperature. The slides were then washed with PBS three times. Fixed tissues were incubated with the primary antibodies described above overnight at 4°C followed by fluorescein isothiocyanate (FITC)-conjugated secondary antibody (1:200) for 1.5 h at room temperature. Fluorescence was examined using a Leica DMR fluorescence microscope with an FITC filter, and images were collected with a Hamamatsu C4742–95 digital camera.

For quantification of the SMI-31 signal intensity in DRG, the serial sections were individually analyzed using MetaFluor/MetaMorph software (Universal Imaging, Westchester, PA). Single neuron perikarya were identified and the signal for fluorescence intensity collected. At least three DRG sections for each animal were analyzed, and three animals were studied for each treatment. Every neuronal perikaryon in every section was analyzed; the cell numbers assessed for each treatment group ranged from 35 to 100.

Where appropriate, data were subjected to one-way ANOVA using the statistical package SPSS/PC+ (SPSS, Chicago, IL). Where the F ratio gave P < 0.05, comparisons between individual group means were made by Scheffe’s multiple-range test at significance levels of P = 0.05. Where nonhomogeneity of variance was apparent, single comparisons were performed using Student’s t test at significance levels of P < 0.01.

The STZ-induced diabetic rats exhibited characteristic reduced body weight and increased plasma glucose levels (Table 1). NT-3 treatment had no effect on these parameters. MNCV and SNCV were both significantly lower in STZ-induced diabetic rats, and NT-3 treatment prevented the deficit in SNCV but not MNCV (Table 1). No qualitative differences were evident in axons of myelinated fibers from the sural nerve (Fig. 1), suggesting that STZ-induced diabetes and NT-3 treatment had no effect. Likewise, quantitative measures of axonal diameter and neurofilament density and number in myelinated axons of the sural nerve were not significantly altered in STZ-induced diabetic rats (Table 1), although there was a nonsignificant 27% reduction in the number of neurofilaments. NT-3 treatment raised axonal diameter in STZ-induced diabetic rats by ∼9%, but this effect was not statistically significant (P < 0.18).

Levels of total NF-H and NF-M proteins (detected using antibodies specific for phosphorylation independent epitopes) were raised by 3.3- and 2.5-fold, respectively, in proximal nerve (5-mm nerve segment adjacent and distal to DRG) of STZ-induced diabetic rats, and these accumulations were completely prevented by NT-3 treatment (Fig. 2). In comparison, NF-H and NF-M phosphorylation status (detected using antibody SMI-31) was reduced by 50 and 80%, respectively; NT-3 normalized this abnormality. There was no diabetes-induced alteration in NF-L protein levels (Table 2).

Levels of total and phosphorylated NF-H and NF-M were lowered by ∼50% in sural nerve of STZ-induced diabetic rats (Fig. 3). NT-3 treatment did not significantly reverse this decline in neurofilament protein levels; however, there was an ∼25% increase in total and phosphorylated NF-M. There was no change in NF-L protein levels (Table 2).

In lumbar DRG, the levels of total NF-H were raised fourfold in STZ-induced diabetic rats, with NT-3 significantly ameliorating this effect (Fig. 4). Immunohistochemical analysis of DRG revealed elevated expression of total NF-H in individual sensory neurons that was prevented by NT-3 (Fig. 5D−F; indicated by arrows). The Western blot analysis showed that the phosphorylation status of NF-H was also raised in diabetes by ∼30% (P < 0.05), although NT-3 had no effect on this parameter. Total levels of NF-M were reduced by ∼30% in STZ-induced diabetic rats, and NT-3 treatment had no effect (Fig. 4). Phosphorylation levels of NF-M were raised by 20–30% in STZ-induced diabetic rats, with the highest levels seen in the NT-3-treated animals. In contrast to the Western blotting data, the immunohistochemical analysis, using the SMI-31 antibody that detects phosphorylated epitopes on NF-H and/or NF-M, showed that phosphorylation of NF-H and NF-M was clearly upregulated in sensory neuron cell bodies of STZ-induced diabetic animals (Figs. 5A–C; arrows show sensory neurons in DRG of STZ-induced diabetic animals with particularly high signal intensity). Quantification of the fluorescence intensity of the SMI-31 signal showed that large (>30 μm in diameter) and small (<30 μm) sensory neurons underwent an elevation in NF-H/NF-M phosphorylation in DRG of STZ-induced diabetic rats (Fig. 6). The effect was greatest in large neurons, and NT-3 significantly prevented this increase in phosphorylation (Fig. 6). Again, NF-L protein levels measured using Western blotting were not significantly affected by diabetes or NT-3 treatment (Table 2).

The neurofilament protein kinases, ERK-1 and -2, JNK (46-kDa and 54- to 56-kDa isoforms), and p38 all exhibited elevated activation (1.5- to 3.5-fold increase over control) in DRG of STZ-induced diabetic animals (Fig. 7). NT-3 failed to significantly affect the activation status of any of these protein kinases.

The results of the present study showed that in proximal nerve and perikarya of sensory neurons in STZ-induced diabetic rats maintained for 14 weeks, there is an accumulation of NF-H and NF-M, but not NF-L, proteins. This proximal buildup coexisted with a 50% decrease of NF-H and NF-M, but not NF-L, in sural nerve. Treatment with NT-3 corrected the abnormalities in DRG and proximal nerve but only had a partial effect in restoring sural nerve neurofilament levels to normal.

Although sensory and motor NCV were reduced, the axon caliber in sural (Table 1) and peroneal nerve (i.e., motor axon; data not shown) was not significantly affected by diabetes. This observation is controversial in diabetic neuropathy, but these data confirm that early alterations in NCV are not associated with structural abnormalities in nerve in experimental diabetes. Therefore, relating the NCV change in STZ-induced diabetic rats to those observed in humans should be approached with caution. NT-3 treatment prevented the sensory NCV deficit but had no effect on motor NCV, a finding that conflicts with those of our previous study (3). This disparity may relate to differences in rate and extent of maturation of rats. The previous study was performed in the Mizisin/Calcutt laboratory using younger animals that may be more sensitive to NT-3 treatment with regard to NCV physiology. The current results showed that abnormalities in neurofilament biology in STZ-induced diabetic rats lag far behind the decrease in NCV, thereby supporting the view that electrophysiological alterations in nerve function are not attributable to endoskeleton-based structural changes in the axon.

NF-H and NF-M protein levels were reduced by 50% in sural nerve (Fig. 3) but did not correlate with loss in axonal caliber, alterations in neurofilament density (Table 1), or changes in axon integrity. This indicates that nodal gaps were normal and that there was no hydropic axoplasm (Fig. 1). Although there was a 27% decrease in neurofilament number in sural nerve of STZ-induced diabetic rats, this was not statistically significant. These results match the control data and fall between the values for neurofilament numbers obtained by Scott et al. (1) at 2 and 6 months of STZ-induced diabetes. In our previous study at 3 months of STZ-induced diabetes, the mean axonal diameter of the largest myelinated fibers in sural nerve was significantly reduced by 7%, a reduction associated with reductions in NCV and loss of NF-H and NF-M protein. NT-3 treatment corrected the deficits in NCV and axonal caliber and partially normalized alterations in neurofilament phosphorylation but did not prevent the loss of NF-H and NF-M protein in sural nerve (3). These observations suggest that at 3 months of STZ-induced diabetes, the loss of NF-H and NF-M protein and the trend toward the reduction in neurofilament numbers (Table 1) in sural nerve are not critical determinants in the regulation of axonal caliber.

These findings confirm, in part, the results of neurofilament gene targeting studies in transgenic mice. Knockout of the NF-H gene or removal of the COOH-terminal of NF-H has no effect on axonal caliber in sciatic nerve (28–30); however, complete loss of NF-M expression does reduce axon caliber in sensory axons, possibly through impaired control of NF-L assembly (31,32). Studies in double-knockout mice for NF-H and NF-M show severe axonal atrophy associated with impaired expression and axonal transport of NF-L in sciatic nerve (32). The complete absence of NF-H and NF-M results in the failure of NF-L subunits to assemble, leading to loss of 10-nm filaments (32). Furthermore, the loss of NF-L expression via targeted disruption results in the failure of NF-H and NF-M to assemble, a scarcity of 10-nm filaments, and severe axonal atrophy (33). In the present study, NF-L protein levels were sustained at normal levels in the DRG and proximal and sural nerve of STZ-induced diabetic rats (Table 2). Our results suggest that more severe loss of NF-H and NF-M protein expression and delivery to the axon is required before an impact on NF-L assembly is produced and subsequent axonal atrophy develops. Therefore, if we had extended this study by 1–2 months, the axonal atrophy may have appeared and could have been associated with a significant loss in neurofilament number. With a longer duration of STZ-induced diabetes (≥6 months), axonal atrophy is more pronounced, is observed in both myelinated and unmyelinated fibers, and is associated with the loss of 10-nm filament numbers in axons (1,2).

The accumulation of NF-H within the perikarya of sensory neurons in the DRG was accompanied by a small but significant increase in phosphorylation (Figs. 4B, 5A–C, and 6). This accumulation of NF-H proteins was not associated with any ultrastructural changes within the DRG; for example, perikaryal sizes were normally distributed and intraganglionic nerve fibers appeared unaffected (data not shown). Although phosphorylation-dependent retardation of axonal transport of NF-H is feasible (34,35), the COOH-terminal of NF-H, which is the main site of phosphorylation, is not critical for the regulation of axonal transport of cargoes (30). Furthermore, the accumulation of NF-H and NF-M in the proximal nerve of diabetic rats was associated with a 50–80% decrease in the level of phosphorylation (Fig. 2B). Hypophosphorylated NF-H and NF-M undergo more rapid axonal transport (34) and, importantly, more readily form stable complexes with kinesin (36). Furthermore, postnatal deletion of KIF-5A, a neuronal-specific kinesin, results in impaired axonal transport of neurofilament and induces accumulation in sensory neuron cell bodies and axonal shrinkage in dorsal root (37). The energy dependency of kinesin-associated cargo motility, coupled with the mitochondrial dysfunction (38,39) and energy deficiency (40) characteristic of diabetes, may therefore explain impaired protein transport observed in STZ-induced diabetes (6).

NT-3 treatment normalized aberrant NF-H phosphorylation in small- and large-sized sensory neurons (Figs. 5 and 6). At a dosage of 5.0 mg/kg, NT-3 may influence neurofilament biology in a wide range of sensory neuron phenotypes via binding to trkC, trkA, and/or p75^NTR (41,42). Although NT-3 reduced NF-H phosphorylation in perikarya of DRG (Figs. 5A–C and 6), it raised NF-H and NF-M phosphorylation in the proximal nerve (Fig. 2). We believe all these results are independent of direct modulation of neurofilament function by phosphorylation and instead are indicative of NT-3–dependent translocation of NF-H and NF-M away from the site of accumulation in the proximal axon. The increased activation of SAPKs (neurofilament kinases) in DRG of STZ-diabetic rats may have been expected to contribute significantly to the raised NF-H phosphorylation observed (Figs. 4B and 7). However, NT-3 failed to alter the phosphorylation status of these neurofilament kinases and, hence, involvement of these proteins in the etiology of neurofilament accumulation is unlikely. In fact, in the proximal axon, the SAPKs were activated two- to threefold (43) in association with a decrease in phosphorylation of NF-H and NF-M (Fig. 2B). NT-3 may act at two major loci to normalize neurofilament biology. NT-3 could augment neurofilament synthesis at the transcriptional and/or posttranscriptional levels; in fact, we have unpublished data showing that NT-3 can reverse the STZ diabetes-induced decrease in NFH and NFM mRNA levels. However, we favor a role for NT-3 in optimizing energy reserves within the sensory axon, possibly via direct modulation of mitochondrial function. Preliminary data demonstrating that NT-3 increases the mitochondrial membrane potential in cultured adult sensory neurons and can reverse deficits in this parameter in STZ-induced diabetic rats supports this notion (T. Huang, N.M.S., A. Verkhratsky, and P.F., unpublished observations). The NT-3 pathway regulating mitochondrial function involves activation of phosphoinositide 3-kinase and downstream regulation of metabolic input into the electron transport chain.

Table 3 summarizes the changes in neurofilament biology occurring in the STZ-diabetic rat. Early reductions (i.e., at 1–2 months of STZ-induced diabetes) in axonal transport of neurofilament are most closely correlated with reduced synthesis of transcripts (4–6,44). Between 2 and 3 months of STZ-induced diabetes, the aberrant phosphorylation of neurofilament in the DRG is triggered, which may induce accumulation of NF-H and NF-M and exacerbate the inhibition of neurofilament transport. Alternatively, the increased phosphorylation in DRG is a reflection of a further cessation of neurofilament transport as a consequence of impaired energy reserves (as discussed above), associated with a hypophosphorylation of neurofilament in the proximal axon. In contrast, in the distal axon at a longer time point (6 months) of diabetes, there is an elevation of neurofilament phosphorylation that is associated with a gross loss of neurofilament and possibly related shrinkage of axon caliber. This increased distal phosphorylation of neurofilament may reflect an attempt to optimize caliber in response to the loss of neurofilament polymers.

The reduced expression and delivery of neurofilament subunits to the distal axon has been proposed as a critical factor in the etiology of the axonal degeneration observed in diabetic animal models and humans (1,10,11). We propose that the accumulation of NF-H and NF-M in the proximal axon represents a retardation of axonal transport of neurofilament subunits to the distal axon, resulting, in the longer term, in a diminution of distal axon caliber.

NT-3 expression is downregulated in muscle in STZ-diabetic rats (24); therefore, reduced occupancy of trk receptors and p75^NTR at nerve endings may contribute to a distally generated stress signal (43) that triggers mitochondrial dysfunction and abnormalities in neurofilament biology (11,24). Although NT-3 treatment corrected the abnormality in neurofilament accumulation, it did not succeed in completely preventing the loss of NF-H and NF-M in sural nerve. The partial 25% increase in NF-M levels in the treated animals may be a reflection of the duration of NT-3 treatment. Two months of treatment may not have been long enough to allow NT-3 to correct the diabetes-induced abnormality causing the accumulation of neurofilament subunits and then, given the slow rate of neurofilament axonal transport, permit the transport of the released subunits to the distal axon.

The present study showed that the impairment in axonal transport and delivery of neurofilament in sensory neurons in an animal model of type 1 diabetes involves an accumulation of subunits in the proximal area of the peripheral sensory neuraxis. The etiology of this defect may involve loss of neurotrophic support, as NT-3 treatment successfully overcame this proximal build-up of protein. Accumulation of neurofilaments may have an impact on the slow anterograde axonal transport of other proteins and explain a range of deficits in diabetes (45). We propose that the outcome of a proximal accumulation of these critical cytoskeletal proteins will, in the longer term, induce detrimental effects on NF-L expression and axonal transport, leading to a failure to maintain the distal axon. These abnormalities in neurofilament biology could explain the distal axonal atrophy seen in experimental and possibly human diabetic neuropathy.

This work was funded by the Juvenile Diabetes Research Foundation (to P.F.) and National Institutes of Health Grants NS/DK 38855 and DK57629 (to N.A.C.).