Endoneurial Capillary Abnormalities Presage Deterioration of Glucose Tolerance and Accompany Peripheral Neuropathy in Man

Between 1975 and 1979, 69 men with type 2 diabetes, 51 men with IGT, and 62 men with NGT were classified in the Malmö area and matched for age, height, and BMI. Between 1989 and 1991, they each underwent neurophysiological evaluation of peripheral nerve function (20). For the baseline examination in 1992–1993, after a neurological and neurophysiological examination, 10 subjects with diabetes, 10 with IGT, and 10 with NGT underwent a whole-sural nerve biopsy (10). At a follow-up examination between 1998 and 1999, 5–6 years after the biopsy, all subjects were invited to receive an OGTT and a detailed neurological and neurophysiological examination, but no second biopsy was performed. Among the 30 individuals, 22 accepted the complete follow-up investigation, 1 with IGT did not participate in the clinical examination, 3 died (2 with diabetes and 1 with NGT), 2 refused to participate in the clinical and neurophysiological examination (1 with diabetes and 1 with NGT), and 1 with diabetes had moved from the Malmö area. Accordingly, 7 of 10 subjects originally classified as having diabetes, 9 of 10 originally classified as having IGT, and 9 of 10 originally classified as having NGT at baseline were included at the follow-up examination.

At baseline, the subjects were examined by a neurologist, and patients with absent ankle reflex and/or altered/reduced sensory perception and/or neuropathic symptoms in toes or feet were considered to have clinical peripheral neuropathy (10). At the follow-up examination, a modified version of Dyck’s original Neuropathy Symptom Score and Neuropathy Disability Score (NDS) was used to define the incidence and severity of peripheral neuropathy (21). In our protocol, the subject was asked if the following symptoms were present in the feet: numbness, abnormal sensation of heat or cold, sensation of pins and needles, contact dysaesthesia to bed clothes, burning pain, lancinating pain, and dull pain. If positive, the subject was asked whether the symptoms occurred sometimes, often, or regularly at night. Each of the seven symptoms was scored; lack of symptoms was scored as 0, sometimes as 1, often as 2, and during most nights as 3. The scores were added into the Neuropathy Symptom Score, giving a range of 0–21. Sensory perception was assessed on the great toe, anterior foot, lateral malleolus, pretibia, and knee bilaterally with regard to the modalities of light touch (cotton wool), pin prick (needle), vibration (vibration fork), and cold (cold metal item) together with an evaluation of the patellar and ankle reflexes. Lack of sensation for the individual modality was given a score of 1. The sensory modalities were added into an NDS subscore A (NDS A) (the sum was divided by 2), giving a range of 0–20. Reflex findings were added into an NDS subscore B (NDS B) in which normal reflex was given a score of 0, reduced reflex a score of 1, and absent reflex a score of 2, giving a range of 0–8. NDS A and NDS B were also added together to create a composite NDS ranging from 0 to 28.

Whole blood glucose was measured fasting (0 min) and 120 min after 75 g glucose was ingested. In accordance with World Health Organization criteria (22), a fasting blood glucose concentration >6.1 mmol/l and/or a 120-min blood glucose >11.0 mmol/l defined diabetes. A fasting value of <6.1 mmol/l combined with a 120-min value of 7.8–11.0 mmol/l defined IGT, and a fasting blood glucose <6.1 mmol/l combined with a 120-min value of <7.8 mmol/l defined NGT. OGTTs were conducted in all subjects except one with previously classified IGT. This subject was considered to have IGT in all analyses except those involving glucose tolerance status at follow-up, from which he was excluded.

Plasma insulin was measured by radioimmunoassay, according to the method of Heding (23), at fasting (0 min), 40 min, and 120 min after the 75-g dose of glucose was ingested.

Stimuli were applied to the sural nerve at the calf ∼15 cm proximal to the lateral malleolus. Antidromic sensory neurography of the sural nerve was performed with subcutaneous needle electrodes placed at the lateral malleolus with a reference electrode placed 3 cm distally.

The surgical procedure for the whole-sural nerve biopsy performed at baseline has been previously described in detail (24). Briefly, the sural nerve was exposed posterior to the lateral malleolus during local 1% lignocaine anesthesia and a whole, 5- to 6-cm length of sural nerve was obtained and divided into five 1-cm segments. The first, fourth, and fifth (proximal to distal) segments were immediately frozen in liquid nitrogen and stored at −70°C. The second and third segments were fixed in 0.1 mol/l cacodylate buffered (pH 7.3) 2.5% gluteraldehyde. The fixed 1-cm segments were rinsed and further divided into three equal segments, postfixed in 1% osmium [4% sucrose, 1.5% K₃Fe(CN)₆ in cacodylate buffer], dehydrated in ascending concentrations of ethanol (50–100%), and infiltrated with Epon 812 resin using propylene oxide as an intermediary before setting in resin blocks.

We have previously reported on the morphologic findings of the myelinated fibers conducted at the University of Michigan Nerve Biopsy Laboratory, Michigan (10). In the current study, we report on the morphometric evaluation of endoneurial capillaries conducted in Manchester, U.K., by one of us (R.M.). The biopsies were recorded with random identification numbers to conceal their identities from the analysis. The study was approved by the Ethics Committee of Lund University.

The endoneurial capillary density was determined directly by light microscopy from semithin sections stained with Thionin and Acradine Orange. Sampling was not used because all endoneurial capillaries were counted directly in all fascicles of the whole nerve and related to the fascicular area to calculate a density. Electron micrographs (magnification ×6,000) were prepared on an average of 10 randomly selected endoneurial capillaries per biopsy (range 8–12; no difference between the three groups of different glucose tolerances), and the luminal, endothelial cellular, and basement membrane area were derived by tracing the image analysis cursor around each capillary profile. The endothelial cell profile nuclear number and pericyte cell nuclear number per capillary were counted directly from each micrograph. The methodology for quantifying microangiopathy has been described in detail elsewhere (25–28). The following variables were analyzed: endoneurial capillary density (in number per millimeters squared), luminal area (in micrometers squared), basement membrane area (in micrometers squared), and endothelial/pericyte nuclear ratio. For technical reasons, capillary density could not be assessed in three subjects: one with diabetes, one with IGT, and one with NGT at baseline. Similarly, endothelial/pericyte ratio, luminal area, and basement membrane area could not be assessed in three subjects: two with IGT and one with NGT at baseline. In addition, MNFD could not be assessed in four subjects, one with diabetes, one with IGT, and two with NGT at baseline.

Data are presented as median (interquartile range). Differences were compared using the Kruskal-Wallis and Mann-Whitney U tests. Correlations were analyzed by Spearman rank-correlation test.

All seven subjects with diabetes still had diabetes, whereas five of the nine IGT subjects had developed diabetes, and three of the nine NGT subjects had developed IGT. Accordingly, among the 25 subjects at follow-up, 12 had diabetes, 7 had IGT, and 6 had NGT. At follow-up, plasma insulin concentrations were significantly higher in diabetic subjects compared with NGT subjects at fasting (22.0 [14.0] vs. 9.5 [10.0] μU/ml; P = 0.0050) and 120 min (68.0 [65.0] vs. 34.0 [29.0] μU/ml; P = 0.0342), and in IGT subjects compared with NGT subjects at 120 min (76.0 [23.0] vs. 34.0 [29.0] μU/ml; P = 0.0282).

Subjects with diabetes were slightly younger than other subjects (Table 1), but there were no significant differences in height or BMI. As previously reported in the baseline study (10), subjects with diabetes had significantly lower sural nerve action potential (SNAP) amplitudes compared with those of the subjects with IGT (P = 0.0409) and NGT (P = 0.0142). In addition, subjects with diabetes had significantly lower sural nerve conduction velocities (SNCVs) compared with those of subjects with IGT at baseline (P = 0.0140) (Table 1). However, MNFD did not differ significantly between the three groups with different glucose tolerances. At follow-up, there were no significant differences in sural nerve electrophysiology or MNFD between the three groups with different glucose tolerances (Table 2).

The prevalence of clinical peripheral neuropathy was 60% (6 of 10) in the diabetic group, 44% (4 of 9) in the IGT group, and 0% in the NGT group at baseline (Table 1). There were no significant differences in age, height, or BMI between groups with or without peripheral neuropathy at baseline (Table 3). Patients with clinical peripheral neuropathy (all with diabetes or IGT) showed significantly lower SNAP (P = 0.0029) and MNFD (P = 0.0005) compared with subjects with diabetes or IGT and without neuropathy (Table 3). Hence, axonal loss and low MNFD were associated with clinical peripheral neuropathy. Subjects with clinical peripheral neuropathy at baseline showed a significantly higher symptom score (Neuropathy Symptom Score) at follow-up than those without (1.5 [3.0] vs. 0.0 [1.0]; P = 0.0078), indicating that the symptom score used may be relevant in setting criteria for peripheral neuropathy.

When classified by metabolic status at follow-up 6 years after nerve biopsy, patients with diabetes (n = 10) had a significantly higher capillary density compared with subjects with NGT (n = 5) (86.0 [24.3] vs. 54.9 [17.1] number/mm²; P = 0.0200) (Fig. 1A and Table 2). However, Fig. 1A also shows that subjects with IGT at baseline who progressed to diabetes (n = 4) 6 years later exhibited a significantly higher capillary density in their baseline biopsies compared with those with persistent IGT (n = 4; 86.7 [25.2] vs. 54.1 [14.6] number/mm²; P = 0.0433). One baseline subject with NGT, who progressed to IGT, had a very high capillary density (128.1 number/mm²) at baseline. If this subject was excluded from the analysis, capillary density was significantly increased among patients with diabetes (n = 9) compared with subjects with NGT at baseline (n = 8; 75.3 [22.3] vs. 55.1 [15.8] number/mm²; P = 0.0343). Hence, high capillary density was positively associated with future and current diabetes.

Tables 1 and 2 show that luminal area did not differ between the three groups with different glucose tolerances. However, Fig. 1B shows that subjects whose glucose tolerance deteriorated at follow-up (three from IGT to diabetes and two from NGT to IGT) had significantly lower luminal area than subjects whose glucose tolerance did not deteriorate (constant NGT [n = 6] and constant IGT [n = 4; 11.9 {2.4} vs. 20.8 {7.8}; P = 0.0200]). Hence, reduction in luminal area was associated with deterioration in glucose tolerance in IGT and NGT.

The endothelial/pericyte nuclear ratio was significantly increased in both subjects with diabetes (P = 0.0141) and IGT (P = 0.0299) compared with subjects with NGT at baseline but not at follow-up (Tables 1 and 2). The basement membrane area did not differ among the three glucose tolerance groups (Tables 1 and 2).

In subjects with diabetes at baseline, serum insulin levels at fasting, 40 min, and 120 min correlated negatively with the basement membrane area (r_s = −0.748, P = 0.0249; r_s = −0.729, P = 0.0286; and r_s = −0.602, P = 0.0710, respectively), whereas there were no significant correlations with capillary density or MNFD.

In subjects with diabetes or IGT at baseline, high capillary density was associated with disturbed nerve conduction velocity at follow-up (low SNCV; r_s = −0.691, P = 0.0166) (Fig. 2). Among subjects with diabetes or IGT, those with clinical peripheral neuropathy at baseline (n = 10) had a significantly increased basement membrane area compared with that of those without peripheral neuropathy (n = 7; 114.6 [68.8] vs. 75.3 [28.7] μm²; P = 0.0084) (Figs. 3 and 4). Furthermore, there was a significant and inverse correlation between the basement membrane area and MNFD in those with diabetes or IGT at baseline (r_s = −0.559; P = 0.0304) (Fig. 5A). Moreover, among subjects with diabetes at baseline, there was a close correlation between basement membrane area and axonal loss (low SNAP) at baseline (r_s = −0.894; P = 0.0073) (Fig. 5B).

This prospective study has quantified the key alterations in the endoneurial vasculature in relation to the development and progression of glucose intolerance and neuropathy. Firstly, we have demonstrated that sural nerve endoneurial capillary density is increased in subjects with IGT who develop diabetes and in patients with manifest diabetes compared with subjects with NGT. Secondly, we have demonstrated a reduction in capillary luminal area in subjects progressing from NGT to IGT and from IGT to diabetes. If it is indicative of reduced tissue blood flow, this finding links tissue hypoperfusion to the deterioration of glucose tolerance and suggests that increased endoneurial capillary density may be compensatory for endoneurial hypoperfusion. This is consistent with our previous observation (19) that poor aerobic capacity correlates with increased skeletal muscle capillary density. With regard to nerve, sural nerve oxygen tension is reduced in diabetic patients with established neuropathy (29) but without a significant alteration in either luminal area or endoneurial capillary density (26–28,30,31). However, our recent study of patients with subclinical neuropathy (32) demonstrates that increased epineurial nerve blood flow may reflect arterio-venous shunting. This supports the concept that increased endoneurial capillary density may be considered a compensatory reaction to hypoperfusion and hypoxia.

The changes described thus far are of relevance to the pathogenesis of diabetic neuropathy. Endoneurial vessels maintain the endoneurial microenvironment and are therefore important for nerve fiber function. A reduction in endoneurial capillary density and hence increased intercapillary distance is considered one of the key alterations regulating endoneurial perfusion (33). A recent detailed study (34) in a variety of advanced neuropathies, including diabetic neuropathy, has demonstrated a significant increase in epineurial but not endoneurial capillaries. In the current study, we demonstrated a negative correlation between endoneurial capillary density and disturbed nerve function, which is suggestive of a compensatory increase in vascular density in early neuropathy. However, in diabetic patients with established neuropathy, no alteration in endoneurial capillary density has been demonstrated (26–28,30,31). Hence, there is a reduction in endoneurial capillary density with progression of neuropathy.

A variety of other pathological alterations have been associated with the severity of human diabetic neuropathy. The hallmark of diabetic microangiopathy is basement membrane thickening. Many studies (26–28,30,31) have shown a strong correlation between basement membrane thickening versus neuropathic severity in diabetic and nondiabetic human neuropathies. The current study has confirmed that this is an early process and that it relates specifically to neuropathy. Our patients with neuropathy had a significant increase in basement membrane area compared with those without neuropathy. Furthermore, we demonstrated a significant correlation with axonal loss and increased basement membrane area at this early stage of nerve damage. This not only confirms the findings in patients with established (27,28,30,31) and mild (26) neuropathy but also shows that basement membrane thickening may precede more severe nerve damage (16).

The lack of correlation between the basement membrane area and the degree of glucose tolerance in our study suggests that increased basement membrane thickness is not simply an epiphenomenon related to hyperglycemia but is related specifically to the development and progression of peripheral neuropathy. In addition, increased basement membrane area correlated with decreased plasma insulin concentration in our diabetic subjects. Hence, as suggested, hypoinsulinemia may promote the development of polyneuropathy, regardless of the degree of hyperglycemia (35). The lack of correlation between plasma insulin concentration and MNFD supports our hypothesis that hypoinsulinemia primarily affects the basement membrane area, thereby promoting the development of polyneuropathy. Luminal occlusion has been demonstrated in some (36) but not all (26–28,30,31) studies. In the current study, there was no difference in luminal area between those with and without neuropathy, and there was no relation between luminal area with nerve function or axonal loss, which supports the majority of these studies (26–28,30,31).

Because insulin resistance and hyperinsulinemia with associated alteration in capillary function are hallmarks of IGT and early type 2 diabetes, it is tempting to speculate that early endoneurial capillary microangiopathy in these groups could be attributed to insulin resistance and/or hyperinsulinemia. While there is an established relationship between insulin resistance and alterations in the function and structure of blood vessels in classically insulin-responsive tissues such as muscle and fat, little is known of this relationship in other tissues (37). Thus, among its many actions, insulin is a vasoactive hormone, which at physiological concentrations increases skeletal muscle tissue perfusion by recruiting microvascular beds (38). By definition, insulin-resistant states exhibit diminished insulin-mediated glucose uptake into peripheral tissues but also display impaired insulin-mediated vasodilatation (39). We have previously shown (19) that increased skeletal muscle capillary density correlates with insulin levels and that it precedes the development of diabetes in subjects with IGT. Our observations of increased capillary density in both nerve and muscle suggest a compensatory response to insulin resistance. However, the complex effects of insulin are exemplified in rats with insulinoma, in which hyperinsulinemia without obvious insulin resistance is associated with increased endoneurial capillary density (40). Moreover, hyperinsulinemia in diabetic animals (41) and in diabetic patients with insulin neuritis (32) leads to arterio-venous shunting and a reduction in endoneurial oxygen tension (41). Although insulin concentrations were increased in subjects with diabetes or IGT in the current study, no correlation with insulin concentrations and capillary density was noted. This suggests that insulin resistance rather than hyperinsulinemia per se is associated with increased capillary density.

Although we are limited by a small sample size, we still cautiously conclude that increased capillary density in the sural nerve precedes the development of diabetes in subjects with IGT. Furthermore, decreased capillary luminal area in the sural nerve precedes deterioration in glucose tolerance in both IGT and NGT. In addition, endothelial cell hyperplasia and/or pericyte degeneration in sural nerve occur in IGT and diabetes. Moreover, endoneurial microangiopathy, and in particular basement membrane thickening, is related to clinical, neurophysiological, and morphologic measures of neuropathy in subjects with diabetes and IGT. These data are consistent with the hypothesis that endoneurial capillary microangiopathy presages deterioration in glucose tolerance and is an early and persistent feature in the processes underlying diabetic peripheral neuropathy. Additionally, whereas insulin resistance/hyperinsulinemia is generally invoked as a cause of increased macrovascular risk in subjects with IGT or type 2 diabetes, these results would suggest that a microvascular complication, i.e., diabetic neuropathy, might also be associated with these elements of what is generally referred to as “the metabolic syndrome.”

This study was supported by grants from the Albert Påhlsson Foundation, the Crafoord Foundation, the Swedish Heart Lung Fund, the Lundström Foundation, Research Funds Malmö University Hospital, the Swedish Diabetes Association, the Thorsten and Elsa Segerfalks Foundation, and the Swedish Medical Research Council (VR 084; 72KX-14531).

We thank Christina Rosborn and Ann Radelius for excellent technical assistance and Dr. Jan-Åke Nilsson, Department of Statistics and Information Processing, Malmö University Hospital, Malmö, Sweden, for statistical advice.