Intraepidermal Nerve Fibers Are Indicators of Small-Fiber Neuropathy in Both Diabetic and Nondiabetic Patients

Skin-punch biopsy and quantitative sensory testing were performed on 48 patients (3 type 1 and 22 type 2 diabetic and 23 nondiabetic) who presented at The Strelitz Diabetes Institutes with positive symptoms of neuropathy (Table 1). Pathologies of nondiabetic neuropathy patients included eight autoimmune (e.g., chronic inflammatory demyelinating polyneuropathy); two glucose intolerance; one each of ethanol induced, paraneoplastic, mononeuritis multiplex, porphyria, arsenic poisoning, and systemic lupus erythematosus; and seven idiopathic. In addition, biopsies were performed on 20 healthy control subjects for similar evaluation. None of the patients had symptomatic nephropathy or evidence of other autoimmune diseases. These studies were performed with the approval of the Eastern Virginia Medical School institutional review board, and all subjects gave informed consent.

All patients had a complete history and physical examination, with special attention given to neurological evaluation. Neurological symptom scores and neurological disability scores were generated by completion of a modified questionnaire based on those developed by Dyck (15). Neuropathy was established by the criteria suggested by the American Academy of Neurology and the American Diabetes Association (16), including two of the following characteristics: presence of symptoms or signs, abnormal quantitative electrophysiology, abnormal electromyography, and autonomic dysfunction. Sensory testing included measures of vibration, temperature, and current perception (touch, prickling, and pain) thresholds performed as previously described (17). Autonomic function was assessed by three tests: the heart rate variability during deep breathing at six breaths per minute (E:I ratio) and the R:R variation in response to the Valsalva maneuver and postural change.

Skin-punch biopsy (3 mm) was performed under local anesthesia. Biopsies were collected from each patient from dorsal forearm, lateral thigh (10 cm proximal to patella), lateral proximal leg (10 cm distal to fibular head), and lateral distal leg (10 cm proximal to lateral malleolus). Tissue samples were immediately fixed in 2% buffered paraformaldehyde/lysine/periodate solution for 12–24 h at 4°C. They were subsequently cryoprotected in phosphate buffer with 20% glycerol overnight and frozen for later cryosectioning.

The procedures for identifying neurons in skin biopsies used immunofluorescence techniques in a modification of the protocol described by McCarthy et al. (11). Biopsy specimens were cut into 50-μm sections on a Reichert cryostat. Melanin was bleached with a 0.25% KMnO₄ solution followed by a 5% oxalic acid solution. The tissues were permeabilized and blocked with a solution of Tris buffer with 0.5% powdered milk, 1.0% Triton X-100, and 4% normal goat serum. Sections were stained in the same solution with rabbit anti–PGP 9.5 antibody (Chemicon, Temecula, CA) overnight at 4°C and incubated with fluorescein isothiocyanate–labeled goat anti-rabbit IgG secondary antibody for 2 h. After washing, coverslips were mounted and sealed with nail polish.

The prepared slides were viewed and images captured by a blinded observer using an Olympus Fluoview Scanning Laser Confocal Microscope. Six random fields from at least three different sections from each specimen were collected for analysis. Confocal images were captured at 1.6-μm intervals, and the image stack was superimposed on each other to produce the image for quantitation. Quantification was performed only on the area of the photomicrographs showing epidermis. Images were analyzed for mean dendritic length (MDL) (in micrometers), mean dendrite number, and IENFs per millimeter of epidermis using Image Pro image analysis software. Results were expressed as means ± SE, and statistics were evaluated using ANOVA, with P < 0.05 considered significant. Post hoc Tukey-Kramer analysis was performed to compare individual groups.

Neuropathy symptom and disability scores as seen in Table 1 indicate that this patient population had neuropathy that was moderate to severe and predominantly sensory (15,16). There were no significant differences in neurological symptom and disability scores between diabetic and nondiabetic neuropathy groups, either as total scores or as the subscales of motor, sensory, and autonomic scores.

Quantitative sensory testing (warm and cold thermal perception threshold, heat and cold pain, and pressure sense) was performed at all four sites that were biopsied. There was a greater loss of sensory fiber function in the distal leg compared to the forearm. Cold (P < 0.03) and warm (P < 0.05) perception thresholds, as well as heat pain perception (P < 0.02), were significantly elevated in neuropathy patients at the distal leg compared with the control subjects, but there was no significant difference for any of the quantitative sensory measures among the groups at the forearm. This indicates that the small-fiber function was more impaired at the distal leg than the forearm in these groups of patients.

There was a significant correlation between age and distal leg MDL (r = −0.39, P < 0.005) and age and distal leg IENF (r = −0.39, P < 0.005). Data from other sites showed no relationship with age. There was a clear negative IENF gradient from proximal to distal sites in patients with neuropathy (ANOVA, P < 0.001) (Fig. 1A). In contrast, biopsies from control subjects showed only a small, nonsignificant decrease (Fig. 1A), even at the distal leg, where the density was lowest in neuropathy patients. Comparing results at each site, IENF in the forearm was not significantly different between patients (55.5 ± 4.5 fibers/mm tissue) and control subjects (37.4 ± 9.0, P > 0.05), neither was there a difference in the thigh or proximal leg. Only in the distal leg was there a significant decrease in IENF (P < 0.05) of patients (17.5 ± 3.3) compared with control subjects (32.3 ± 5.9).

MDL was similar across all sites for control subjects. MDL was significantly decreased in skin from neuropathy patients with a progressive decrease from proximal to distal sites (P < 0.002) (Fig. 1B). The differences in MDL between samples from control subjects and patients at the thigh (4.4 ± 0.6 vs. 2.6 ± 0.1 μm, respectively), the proximal leg (3.7 ± 0.3 vs. 2.5 ± 0.2, respectively), and the distal leg (3.6 ± 0.4 vs. 1.9 ± 0.2, respectively) were significantly different, but not in skin from the forearm (3.2 ± 0.3 vs. 2.8 ± 0.1, respectively). There was a direct, positive correlation between IENF and MDL in patients only at the distal leg. In control subjects there was no significant correlation between IENF and MDL.

Of particular interest was the dichotomy between IENF and MDL in the distal leg and the effect of duration of diabetes. IENF was significantly greater in patients with <5 years’ duration of diabetes (37.4 ± 7.1 fibers/mm tissue) than in those with >5 years ’ duration (7.8 ± 7.1, P < 0.01) and was not different from control subjects (34.0 ± 6.0). It should be noted that five of six patients >5 years after diagnosis of diabetes had IENF values of 0. In contrast to IENF, MDL was reduced in a linear fashion in the distal leg from the time of diagnosis. When IENF at the distal leg site was compared with clinical measures of neuropathy taken at the same site, a series of negative correlations were found. Warm (r = −0.42, P < 0.005) and cold (r = −0.31, P < 0.05) thermal sense and heat pain thresholds (r = −0.54, P < 0.0001) all correlated negatively with IENF density. In contrast, there was no correlation between pressure sense and IENF. These results are similar to those found with MDL, in which tests of warm sense (r = −0.31, P < 0.05), cold sense (r = −0.36, P < 0.02), and heat pain threshold (r = −0.46, P < 0.002) were significantly related and pressure sense not significantly related. Finally, there was a significant negative correlation between distal leg IENF, but not MDL, and neuropathy disability scores for total neuropathy (P < 0.03) and total sensory (P < 0.03) evaluation.

These data demonstrated both a shortening of IENFs from proximal to distal sites and a significant loss of fibers at the distal site of the lower limb in patients with neuropathy. There were significant changes in IENF in diabetic neuropathy related to duration of diabetes, with a significant fiber loss in patients who were >5 years from diagnosis. In contrast, MDL showed a linear relationship with time since diagnosis of diabetes and with shortening of nerve fibers, irrespective of duration. Furthermore, MDL also showed a shortening of nerve fibers at all sites compared with control subjects. Aging reduced IENF as well as MDL. Finally, in our data there was a correlation between IENF and MDL only at the distal leg and only in neuropathy patients, but not in control subjects, indicating that these patients were at a stage of neuropathy involving only the longest epidermal fibers.

Many studies have demonstrated the loss of peripheral nerve fibers in neuropathy, especially in diabetic patients or even in the pre-diabetic, glucose-intolerant stage. The earliest studies (18,19) used nerve biopsy and both light and electron microscopy techniques to demonstrate the reduction and pathology of distal nerve fibers, especially the sural nerve. Early studies (20,21) with skin biopsy described the pathology of nerve fibers in the lower limb in diabetes, particularly in association with neuropathy. In addition, the shortening of nerve fibers in diabetic neuropathy has previously been reported (12), as well as the correlation of mean fiber length and fiber density. These studies sought to extend the observations to multiple sites and to relate the findings to quantitative sensory tests at the sites of biopsy.

In these patients there was no demonstrable loss of fiber density in the distal upper limb. Quantitative sensory tests performed at the sites of biopsy also showed no deficit in the upper limb, whereas the distal lower limb showed a mild deficit in C-fiber–mediated modalities (warm and cold sense and heat pain) but not in pressure sense, an A-δ fiber–mediated test. Interestingly, the distal fiber loss is seen not just in diabetic neuropathy, but also in a group of patients with other peripheral small-fiber neuropathies. Thus, there is a correlation between gross presentation of neuropathy and the microanatomical evaluation of nerve fibers. This is in concordance with previous reports (4,12,22).

However, our data show that in patients within 5 years of diagnosis there is not significant nerve fiber loss at the distal leg compared with control subjects. This contrasts with reports from other investigators (2,4) that there is a loss of epidermal fibers, even in the pre-diabetic, glucose-intolerant period. These differences may be due to the small numbers of patients analyzed in these studies. Another possibility is that there are separate subgroups of patients who respond in different ways to the onset of metabolic disturbances. Our data in patients <5 years from diagnosis are consistent with a report from Properzi et al. (23) in 1993, who also used skin biopsy, which demonstrated no decrease in PGP 9.5–positive nerve fibers in skin of patients with diabetic neuropathy within 3 years of diagnosis, but an increase in vasoactive intestinal peptide–positive fibers in five of six of the same patients. In addition, Lauria et al. (24) demonstrated a depletion of fibers at the site of deficit, with a normal but disordered nerve fiber distribution in the unaffected area, indicating a more general response to nerve fiber deficits.

The presence of fibers does not address the functional capability of those fibers. The length of the epidermal fibers appears to be linearly related to duration of disease, indicating a slow but steady decline in fiber length in the face of diabetes. This suggests that the mechanisms for reducing fiber length and fiber density are different over time with diabetes, regardless of the overall correlation between the two measures across all groups. MDL may directly reflect the pathological mechanism for reduction of fibers, resulting in shorter and shorter fiber extension, whereas IENF reflects the functional state of epidermal fibers, in which fewer fibers reflects loss of sensation.

One implication of these observations is that they predict difficulty correlating symptom scores and disability scores with epidermal nerve fiber density in the early period after diagnosis. It appears from our data that correlations of IENF with specific quantitative sensory testing measures remain valid. Staining for neurotransmitters such as vasoactive intestinal peptide, calcitonin gene–related peptide, or substance P in combination with PGP 9.5 may provide more detailed functional information.

It has been suggested (22,25) that the use of nerve fiber quantitation in skin biopsies as an early end point in diabetic neuropathy studies might prove useful. Our results would argue against the utility of skin biopsy and measuring IENF alone as a means for detecting and evaluating neuropathic changes in early diabetes, although MDL alone seems promising. There is not a significant difference in PGP 9.5–positive distal leg nerve fibers between control subjects and diabetic neuropathy patients <5 years from diagnosis. Some combination of fiber density and fiber length with neurotransmitter analysis, or even multisite analysis, might prove to be the most revealing method for staging the progression of neuropathy, using skin biopsy measures as a possible end point in clinical trials.

We have previously demonstrated that the microvascular effects of diabetes mimic premature aging (17), and this may be further confirmation that diabetes complications arise similarly to physiological changes observed with aging. Early studies (25) with PGP 9.5 staining showed that epidermal fiber density loosely correlated with the number of abnormal neurophysiological tests. However, these investigators (26) were subsequently unable to detect consistent, specific neurophysiological deficits related to epidermal fiber loss. In our studies, IENF and MDL correlated with clinical measures of small-fiber neuropathy that reflect small-fiber dysfunction, i.e., thermal perception, but not with clinical tests that reflect large-fiber functions. This is in contrast to a previous report (4) that IENF did not correlate with clinical measures of neuropathy. This may have been due to the smaller number of patients in this study, to the different methodology for evaluating IENF utilized by the authors, or to failure to distinguish effectively between the different nerve fiber types. It appears critical to understand the time course for development of diabetic neuropathy in order to evaluate skin biopsy results. A mix of early and late disease patients might statistically mask the relationship with measurable nerve deficits. If a similar evaluation for the time course of development of nondiabetic neuropathies could be developed, it might be that those patients too would show an initial neurotrophic response in an attempt to overcome developing deficits. This possibility remains to be studied.

It is also important to note that there is a difference in evaluation of IENF dependent on whether immunohistochemistry or immunofluorescence is used. The normal range for IENF in the distal leg in reports from various groups utilizing immunohistochemistry is between 7 and 20 fibers/mm tissue (4,11,14), whereas those groups using confocal microscopy, including our data, report IENF ranges from ∼25 to 40 fibers/mm tissue (12,13). Whether this reflects the higher resolution of confocal microscopy, the greater sensitivity of photomultiplier tubes compared to the human eye, or some other methodological issue remains to be resolved.

In summary, we found that IENF and MDL reflect the proximal to distal loss of epidermal nerve fibers described previously by nerve biopsy. These fiber changes correlate negatively with results of quantitative sensory testing, suggesting a direct relationship. However, there was an increase in PGP 9.5–positive IENFs early after the diagnosis of diabetes that was subsequently reduced to nearly zero, whereas the relationship of fiber length to time since diagnosis was linear. In addition, there have been recent reports (9,27) indicating that PGP 9.5 may not stain all epidermal nerve fibers, particularly in neuropathic conditions. Thus, it may be that it will be important to not only measure characteristics of PGP 9.5–positive fibers but also to measure other sensory neurotransmitters in the skin in order to make skin biopsy a useful technique for the early detection of or the efficacy of treatment for diabetic neuropathy.

This work was supported by the Diabetes Institutes Foundation, Norfolk, Virginia.