Reduced Vascular Endothelial Growth Factor Expression and Intra-Epidermal Nerve Fiber Loss in Human Diabetic Neuropathy

RESEARCH DESIGN AND METHODS—

This study was approved by the local research ethics committee, and all patients gave informed consent to participate. Patients and nondiabetic volunteers were invited unless they had absent pedal pulses (to exclude peripheral vascular disease that may have an impact on both neuropathy and VEGF expression but also wound healing after the biopsy). All subjects were screened for other causes for neuropathy including vitamin B₁₂, thyroid-stimulating hormone, antinuclear antibody, and serum protein electrophoresis, and those with a family history of neuropathy or a disease known to cause neuropathy were excluded.

Neuropathy evaluation

Neuropathic symptoms were assessed using the diabetic neuropathy symptom score (10) and the visual analog score (VAS) for pain (11). Neurological deficits were assessed using the neuropathy disability score (NDS) (12) which ranges between 0 and 10; neuropathy was diagnosed if NDS was ≥3/10. We also performed quantitative sensory tests including cooling detection threshold, minimal (HP-VAS 0.5) intermediate (HP-VAS 5.0), and differential (HP-VAS 0.5–5.0) heat as pain thresholds (13,14) and deep breathing heart rate variability (15) with the CASE IV (WR Medical Electronics, Stillwater, MN). Patients also underwent electrophysiological assessment with a Dantec Keypoint system (Dantec Dynamics, Bristol, U.K.) to evaluate amplitudes (microvolts), conduction velocities (meters per second), and latency (milliseconds) of responses.

Laser Doppler flowmetry

A subset of the patients comprising 11 control subjects and 46 diabetic patients who underwent biopsy also had skin blood flow laser Doppler flowmetry with a large area laser scanner (Moor Instruments, Devon, U.K.) on the dorsum of the foot at the site of the biopsy. A MIC iontophoresis controller (Moor Instruments, Devon, U.K.) delivered 20 μA for 5.5 min to the skin through MIC-ION plastic chambers (Moor Instruments) containing 1% acetylcholine and 1% sodium nitroprusside. The results were measured in arbitrary units of flux for baseline, maximal responses, and differential increases in blood flow (Δ).

Skin biopsy

Two 3-mm punch skin biopsy samples were taken from the dorsum of the foot, ∼2 cm above the second metatarsal head, under 1% plain lidocaine local anesthesia.

IENFD

The first specimen was immediately fixed in 4% paraformaldehyde and after 18–24 h was rinsed in Tris-buffered saline and soaked in 33% sucrose (2–4 h) before cryoprotection in OCT and rapid freezing in liquid nitrogen. Sections (50 μm) were cut using a cryostat (model OTF; Bright Instruments, Huntington, U.K.), and floating sections were transferred onto a 96-well plate. Melanin bleaching (0.25% KMnO₄ for 15 min and then 5% oxalate for 3 min) was performed on four sections per case followed by a 4-h protein block with a Tris-buffered saline solution of 5% normal swine serum, 0.5% powdered milk, and 1% Triton X-100; sections then were incubated overnight with 1:1,200 Biogenesis polyclonal rabbit anti-human PGP9.5 antibody (Serotec, Oxford, U.K.). Swine anti-rabbit secondary antibody 1:300 (1 h) was then applied; sections were quenched with 1% H₂O₂ in 30% MeOH-PBS (30 min) before a 1-h incubation with 1:500 horseradish peroxidase–streptavidin (Vector Laboratories, Peterborough, U.K.). Nerve fibers were demonstrated using 3,3′-diaminobenzidine chromogen (Sigma-Aldrich, Manchester, U.K.). Sections were mildly counterstained with eosin to better localize the basement membrane. Negative controls comprised sections that underwent the same run except that the primary antibody was omitted and the developing time was exactly the same for all sections that were processed synchronously.

Vascular factor immunostaining

The second biopsy sample was fixed in 4% paraformaldehyde for 18–24 h, embedded in paraffin wax, and cut into 5-μm sections. Thin sections were mounted on positively charged slides (three per slide), dewaxed in xylene, and gradually rehydrated through decreasing ethanol dilutions. Trypsinization was used to disclose the antigen for sections used to immunostain for blood vessels. Optimal visualization was obtained by microwaving for VEGF-A and VEGFR-2 and by adding a tyramide amplification reagent (CSA I; Dako) for HIF-1α. In all cases, epidermal melanin was bleached with 0.25% KMnO₄ and 5% oxalate before serum protein block. Anti-human primary antibodies were applied overnight at 5°C: mouse monoclonal antibodies to CD31 and von Willebrand factor (diluted 1:100; Dako) and VEGFR-2 (1:50; Santa Cruz); and rabbit polyclonal antibodies to VEGF-A (1:300; Santa Cruz) and HIF-1α (1:300; Abcam). Negative controls comprised sections that underwent the same runs except that the primary antibody was omitted. Developing time was exactly the same for all sections in each separate run and in each run the sections were processed synchronously.

Image analysis

Patterns of immunostaining were examined by light microscopy (Leitz DM RB microscope). Digital images were captured at ×400 magnification with a Nikon digital camera and analyzed with Leica QWin Standard V2.4 (Leica Microsystem Imaging, Cambridge, U.K.) set to detect color intensities in a fixed and constant range (16).

IENFD was defined as the number of fibers per millimeter of basement membrane length and expressed as numbers per millimeter (nerve density per length). If two nerve endings appeared closer than five fiber diameters they were counted as one single nerve (17).

The protocol for assessment of VEGF-A, VEGFR-2, and HIF-1α included a fixed light intensity, a condenser set between 2 and 3, and default Nikon color settings. Every image was evaluated using a standardized Leica program to quantify the amount of stained and total areas (Leica QWin Standard V2.4). The epidermis was assessed including the basal, granular, and spinous layers, but the keratin layer was excluded. The positively stained area was divided by the total area considered to quantify the amount of staining as a percentage. Blood vessel cross-sections were counted manually and divided by the dermal area to obtain a density (number per square millimeter). All observations were performed on coded slides to prevent observer bias.

Statistical analysis

Statistical analysis was performed using SPSS 14.0 for Windows (SPSS, Chicago, IL). The data are presented as means ± SEM. Gaussian data were subjected to parametric ANOVA and non-Gaussian data nonparametric ANOVA (Kruskal-Wallis). After Levene's test for homogeneity of variances, Tukey's or Dunnett's T3 test was used for multiple comparisons as appropriate. Comparison between groups was made by an unpaired t test for Gaussian data and a Mann-Whitney test for non-Gaussian data. Logarithmic transformation was used to normalize the datasets. Correlations were studied with the Spearman coefficient (r_s), and P < 0.05 was considered statistically significant.

RESULTS—

Demographics of the 53 diabetic patients (17 with type 1 diabetes) and 12 age-matched control subjects are listed in Table 1. Diabetic patients had no (n = 12), mild (n = 18), moderate (n = 12), and severe (n = 11) neuropathy on the basis of the NDS. Electrophysiology, quantitative sensory, and autonomic function testing showed a progressive worsening with increasing neuropathic severity (Table 1). Laser Doppler flowmetry (11 control subjects, 11 patients with no neuropathy, 18 with mild neuropathy, 9 with moderate neuropathy, and 8 with severe neuropathy) showed no differences for baseline values (Table 1).

IENFD

IENFD showed a significant and progressive reduction with increasing neuropathic severity (ANOVA, P < 0.001) (Fig. 1D). Comparison with control subjects (10.6 ± 0.7) demonstrated a progressive reduction in diabetic patients without (8.3 ± 1.1, NS) and with mild (6.6 ± 0.9, P < 0.05), moderate (4.8 ± 1.0, P < 0.01), and severe (1.8 ± 0.4, P < 0.001) neuropathy (Fig. 2A and B).

Epidermal VEGF-A

VEGF-A staining was prominent in the basal and spinous layers of the epidermis and was significantly reduced in diabetic patients (22 ± 3%) compared with control subjects (53 ± 8%) (P = 0.001) . VEGF-A expression correlated positively with IENFD (r_s = 0.361, P < 0.01) (Table 2, Fig. 3A and D).

Epidermal VEGFR-2

Epidermal VEGFR-2 expression did not differ between control subjects (14 ± 6%) and diabetic patients (9 ± 2%) (Fig. 3B and E) and was independent of neuropathic severity (Fig. 1B); however, it correlated significantly with epidermal VEGF-A staining (r_s = 0.397, P < 0.01) (Fig. 3B and E).

Epidermal HIF-1α.

Epidermal HIF-1α expression was low and did not differ between groups (Fig. 1C). The VAS correlated with epidermal HIF-1α expression (r_s = 0.361, P < 0.05) (Fig. 3C and F).

Immunolocalization of vascular factors in the dermis

In the upper dermis, VEGF-A, VEGFR-2, and HIF-1α were immunolocalized in fibroblasts, pericytes, endothelium, and smooth muscle of the vascular wall. HIF-1α staining was present both in the nucleus and in the cytoplasm of each cell type and the highest intensity was observed in the pericytes. The amount of staining of VEGF-A, VEGFR-2, and HIF-1α in the dermis was lower than that in the epidermis (Table 2), and there was no significant difference between groups. VEGF-A staining was significantly reduced in diabetic patients compared with control subjects in the upper dermis (P < 0.05) and on blood vessels (P < 0.05) and was the lowest in those with severe neuropathy compared with control subjects (P < 0.05). VEGFR-2 staining did not differ in the dermis but was significantly reduced on blood vessels in diabetic patients with severe neuropathy compared with control subjects (P < 0.05) and diabetic patients without neuropathy (P < 0.01) (Table 2). It correlated inversely to blood vessel density (r_s = −0.412, P < 0.01). HIF-1α expression on blood vessels differed between groups (P < 0.05) and dermal HIF-1α expression correlated with the VAS (r_s = 0.288, P < 0.05).

Dermal microvessel density and vasodilator responses

Dermal microvessel density (number per square millimeter) differed significantly between groups (P < 0.01). It was increased in patients with severe neuropathy (145 ± 18), reaching significance in those with moderate neuropathy (180 ± 17) compared with control subjects (129 ± 12) and in patients without (109 ± 10, P < 0.05) and with mild (105 ± 12) neuropathy (P < 0.01) (Fig. 1E). Baseline blood flow showed a weak negative correlation (r_s = −0.316, P = 0.04) with epidermal expression of VEGF. The maximal blood flow response to acetylcholine was significantly lower in patients with neuropathy (P < 0.05) (Table 1). There was no difference in blood flow response to sodium nitroprusside. Blood flow data showed no correlation to dermal microvessel density.

CONCLUSIONS—

Vascular factors are thought to be central in the development of diabetic neuropathy (18,19), and VEGF has neuroprotective effects (20,21). The present study combined detailed clinical, neurological, and immunohistological evaluation in skin biopsies from the dorsum of the foot. The central finding of our study is a reduction in epidermal VEGF expression and IENFs in diabetic patients with increasing neuropathic severity. Maximal expression of VEGF was observed in the epidermis consistent with previous studies (22), particularly by keratinocytes (23,24). Although these results do not directly imply cause and effect, they do provide the first clinical data in support of a link demonstrated in experimental diabetic neuropathy (3,4).

Tissue hypoxia has been proposed to upregulate HIF-1α expression (25), which is considered to be the main stimulator of VEGF-A synthesis. A previous study in skin wounds has shown increased VEGF-A after topical nerve growth factor administration but did not assess HIF-1α expression (26). In the present study, HIF-1α expression was not reduced in intact skin of diabetic patients without peripheral vascular disease, suggesting adequate tissue oxygenation, which is supported by the findings of a baseline skin blood flow not significantly different from that in control subjects even in patients with advanced neuropathy. It is possible that hyperglycemia per se induces a reduction of VEGF expression in the skin; clearly, further mechanistic studies are needed to clarify this relationship.

Dermal blood vessel density was increased in diabetic patients with more severe neuropathy, particularly those with moderate neuropathy, despite a reduction in VEGF and VEGFR-2 expression. Although no significant changes were found in baseline blood flow and HIF-1α did not increase in neuropathic patients, it may reflect an angiogenic response to the impaired endothelium-dependent response observed in these patients. In nerves of diabetic rats there has been some evidence of transient HIF-1α upregulation (6) and a blunted postinjury rise in nerve blood vessel numbers (27). Finally, despite the overall low expression of HIF-1α, it was inversely related to the intensity of pain assessed using the VAS. This finding requires confirmation but emphasizes that vascular factors may play an important role in the development of painful symptoms (28) and also provides a rationale for the use of vasodilating drugs in painful diabetic neuropathy (29).

In summary, this is the first translational study to suggest that there may be a link between skin VEGF expression, loss of IENFs, and the severity of human diabetic neuropathy. These data demand further quantitative studies assessing VEGF protein and/or mRNA levels to establish a more direct causal link between VEGF and diabetic neuropathy to provide a basis for the beneficial effects reported to date (3–5,30).