Quantitation of epidermal nerves in diabetic neuropathy

Treatment of diabetes by pancreas transplantation (PTx) results in euglycemia and improvement of patient well-being and general health. ^[1] We showed by sequential clinical and neurophysiologic testing of autonomic and somatic function that PTx also halts progression of diabetic neuropathy but we could not prove that reversal of neuropathy occurred. ^[2] Introduction of other potentially therapeutic agents for diabetic neuropathy creates an urgency to establish methods for collection of objective standards to evaluate the results of therapy.

We redirected our search for methods to assess post-PTx improvement to the investigation of nerve morphology. Study of nerves contained in skin biopsy specimens appeared to be a novel and direct approach because the most common sequelae of diabetic neuropathy are reduced sensitivity to light touch, temperature, and painful stimuli. Correlation of nerve structure with deficiencies of skin sensitivity have previously been impractical, mainly because of limitations in staining procedures. Advances in immunohistochemical staining and imaging of nerves now make it possible to visualize clearly all nerves in the skin. ^[3,4] The sensory modalities affected by diabetes are conveyed by sensory nerves whose endings were previously considered to reside in the papillary dermis and the basal layer of the epidermis. ^[5-8] However, recent recognition that large numbers of nerves occupy all layers of the epidermis indicates that depletion of these nerves must be seriously considered as the potential cause of insensitivity in diabetic neuropathy.

Over a century ago Langerhans ^[9] published drawings of nerve fibers that extend into superficial layers of the epidermis. Arthur and Shelley ^[10] reviewed the subject of cutaneous innervation and added their own illustrations of the vertical and horizontal extent of epidermal nerves. Still the idea persisted that epidermal innervation was sparse or absent in human skin. ^[11] The advent of immunohistochemical methods, particularly description of antibody to the general neuronal marker protein gene product 9.5 (PGP 9.5), ^[12,13] has led to conclusive evidence that many nerve fibers advance beyond the basal keratinocyte layer to penetrate deep into the epidermis. ^[3,4]

Epidermal nerve fibers (ENFs) project from a subepidermal plexus, penetrate the dermato-epidermal basement membrane, then rise between layers of keratinocytes toward the skin's surface; some appear to project into the stratum corneum. ^[3,4] ENFs are presumed extensions of sensory neurons, some of which express calcitonin gene related peptide (CGRP) or substance P (SP). We speculated that human ENFs continuously elongate and accompany keratinocytes during their upward migration to the stratum corneum, ^[4] perhaps in relationship to muscarinic and cholinergic receptors on keratinocytes. ^[14] Although assignment of ENFs to specific sensory modalities is speculative, it is logical to suspect a relationship between these nerves and the reduced sensitivity to light touch, thermal, and painful stimuli of our diabetic patients.

Reduced immunohistochemically demonstrated cutaneous innervation is described in diabetes ^[15,16] and other clinical disorders. ^[16-19] We report the results of using immunohistochemistry, laser scanning confocal microscopy (LSCM), and computer image analysis to quantify the number, branching patterns, length, and volume of ENFs in skin biopsy specimens from control and diabetic subjects.

Materials and methods.

Human skin was obtained with informed consent from 18 type I diabetic candidates for PTx (average age 37 years, average duration of diabetes 26 +/- 8 years) and from 18 sex-matched normal volunteers (average age 33 years). Peripheral and autonomic nerve function of the diabetic subjects was evaluated from graded results of a history and examination, motor and sensory nerve conduction determinations, quantitative testing of cardiorespiratory reflexes, warm and cold sensitivity, and sweating by methods previously reported. ^[2,20] Control subjects answered a health status questionnaire and underwent a neurologic examination, nerve conduction, and quantitative temperature and light touch sensitivity testing. Punch biopsy specimens (3 mm) were removed from the anesthetized calf skin of diabetic and control subjects and immediately placed in chilled Zamboni's paraformaldehyde/picric acid fixative for 18 hours at 4 degrees C, then cryoprotected with 20% sucrose in 0.1 M phosphate buffered saline (PBS). The project was approved by the University of Minnesota committee for use of humans in research.

Frozen, 100-micro meter thick sections were cut with a sliding microtome (American Optical, Buffalo, NY). Floating sections were stained using rabbit polyclonal antibody to PGP 9.5 (Ultraclone, Wellow, UK) with donkey anti-rabbit IgG labeled with cyanine 3.18 fluorescent probe (Jackson ImmunoResearch, West Grove, PA), and also with mouse monoclonal antibody to type IV collagen (Chemicon, Temecula, CA) reacted with donkey anti-mouse IgG labeled with cyanine 5.18 fluorescent probe (Jackson ImmunoResearch, West Grove, PA). A solution of 0.2 M PBS with 0.3% Triton X-100 (Sigma, St. Louis, MO) and 1% normal donkey serum (Jackson ImmunoResearch, West Grove, PA) was used as a dilutent and a wash solution. Nonimmune serum controls were run simultaneously. Sections were adhered to cover slips with agar, dehydrated with alcohol, cleared with methyl salicylate, and mounted in DPX (Fluka, Ronkonkoma, NY).

Samples were previewed with a Nikon Microphot-SA microscope equipped for epifluorescence (Lake Success, NY). Selected samples were imaged with a MRC-1000 Confocal Imaging System (BioRad, Boston, MA) mounted on a Nikon Optiphot2 microscope equipped with a times 20 plan apochromat objective (NA 0.75) and appropriate filters. Digitized images were collected in successive frames of 2-micro meter serial optical sections (z series) and projected into a single in-focus image of the 100-micro meter sections, or computerprocessed as three-dimensional objects. Measurements were made from z series collected at 4 to 6 sites across the entire length of epidermis in one section of each skin sample.

Measurement of ENFs was performed with the Eutectic NTS Neuron Tracing System (Sun Technologies, Inc., Raleigh, NC). Volume of epidermis was measured with the Image Volumes (IV) software package (Minnesota Datametrics Corporation, St. Paul, MN). The Eutectic system was run on a Gateway 2000 P5-90 Pentium personal computer, while IV was used with a Silicon Graphics Indigo2 workstation.

The Eutectic system and IV packages used the BioRad confocal image files directly, and analyzed the z series on an image-by-image basis before rendering three-dimensional reconstructions of the nerve structures and epidermal volumes; length and volume measurements were derived from these reconstructions. The LSCM images used in this analysis were 768 by 512 pixels in size (a pixel, or picture element, is the smallest coherent "dot" on a computer screen). At magnification times 20, used during collection of the epidermal LSCM images, one pixel corresponds to 0.83 micro meter, yielding an image size of 634.71 micro meter by 423.14 micro meter.

The Eutectic system was used to trace all ENFs, then measure their length, number of branch points, and summed length in three dimensions. Since nerve volume was not germane to our analysis, it was assumed that nerve diameter remained constant. Measurements were made on the first twenty-nine images in each series (approximately 60 micro meter) because diminishing contrast between the measured structures and background staining of the epidermis beyond this level made identification of the nerve structures and basement membrane difficult.

The tree-like pattern of each nerve fiber in the epidermis was traced from its origin, through successive branch points to the endings of each branch. ENF origins were taken to be the point where the trunk of the nerve penetrated the basement membrane to enter epidermis; the nerve count was therefore the number of nerve origins at the basement membrane. The Eutectic system provided the summed length of traced nerves, the aggregate number of branch points in the z series, and a break-down length of individual nerve fibers.

Because the thickness of the epidermis (basement membrane to stratum corneum) varied from subject to subject, IV was used to measure the epidermal volume represented in each LSCM z series. In addition, since images were collected in a manner that standardized the length of epidermis in each data set (same z series depth and magnification), variations in epidermal thickness from basement membrane to corneum were taken into account by measurement of the volume. The perimeter of the epidermal region on each image of the z series was drawn manually to identify the area of epidermis in the image. IV then rendered a three-dimensional structure from these areas, and calculated the volume of epidermis. To maintain compatibility with the Eutectic system nerve analysis, only the first twenty-nine images of the series were used in the volumetric rendering of the epidermis.

Results.

PGP 9.5 immunoreactivity (-ir) was distinctly localized in the nerve endings of epidermis. The intense labeling of nerves and faint background allowed clear image sets to be obtained for quantitation, but did not sharply demarcate the dermal-epidermal junction. Use of the second fluorophore to visualize type IV collagen-ir in basement membrane facilitated the determination of the point at which nerve fibers passed from the dermis through basement membrane into the epidermis. We did not observe any obvious thickening of the basement membrane.

The amount of nerve visible in biopsy specimens of normal calf skin varied from subject to subject Figure 1 B, D, F. In general, many nerves were present in the epidermis of normal subjects. The number of ENFs in biopsy specimens from diabetic subjects was generally reduced; some samples were without any ENFs while others were within the lower range of normal Figure 1 A, C, E. ENFs were unevenly distributed in normal and diabetic skin, but with larger distances between ENFs in diabetic skin. In biopsies from diabetic, but not normal, subjects, some nerves ended abruptly just below the basement membrane Figure 2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Laser scanning confocal microscope images of nerves in superficial skin from calf of diabetic subjects (A, C, E) and normal subjects (B, D, F). Images were selected to demonstrate the range of innervation from least (top) to most (bottom). Reactivity to antibody to PGP 9.5, a pan-neuronal marker, visualized using the indirect immunofluorescence method, reveals numerous nerve fibers in the epidermis of control subjects and a reduced number of nerve fibers in epidermis of diabetic subjects. Epidermal nerve fibers (ENFs) arise from a subepidermal neural plexus (SNP), which extends horizontally immediately proximal to the basement membrane of the epidermis. In severely neuropathic subjects nerves from the SNP truncate at the basement membrane and do not extend into the epidermis (A). In less severely affected diabetic subjects, a few nerve fibers penetrate the basement membrane and extend to the stratum corneum (C and E). Many nerve fibers in control subjects arise from the SNP to innervate the epidermis (B, D, F). They often form a candelabra-like pattern within the epidermis and extend the entire depth of the epidermis. All images were projected from 29 optical sections acquired at 2.0-micro meter intervals using a times 20 Nikon plan apochromat lens. Scale bars = 100 micro meter.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Dermal-epidermal junction region of diabetic subjects. This view shows a nerve fiber that approaches but does not cross the basement membrane (arrows). This phenomenon is common in skin from subjects with neuropathy. Scale bar = 50 micro meter.

The number of ENFs entering the epidermis per length of epidermis analyzed was decreased in diabetic subjects compared with skin from normal subjects Figure 3. Epidermal volume varied from subject to subject in both diabetics and controls; although the difference was not statistically significant, the epidermal volume of diabetic samples was generally less than that of controls. Total length of nerve per epidermal volume in normal samples exceeded that found in most diabetic samples Figure 4, with some crossover occurring in the lowest normal and the highest diabetic samples. In control subjects there was a decrease in the summed nerve length/volume with age, whereas no correlation was apparent in the diabetic subjects. This was presumably because most PTx patients were in the same two decades and the variation in degree of neuropathy of the PTx patients overshadowed any variation due to age.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Histogram comparing the number of epidermal nerve fibers (ENFs) per mm of epidermis in diabetic (n = 18) and control (n = 18) subjects. The number of ENFs is reduced in diabetes. Number of ENFs was derived from Eutectic nerve tracing analysis of confocal images as intersections of fibers with basement membrane. Length of epidermis was determined from measurements of confocal images.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Histogram comparing the length of epidermal nerve fibers (ENFs) per volume of epidermis in diabetic and control subjects. The total length of epidermal nerve in diabetic subjects is less than that seen in control subjects. Length of ENFs was derived from Eutectic nerve tracing analysis of confocal images. Volume of epidermis was determined from analysis with Image Volume software.

The aggregate nerve length and total number of branch points were proportional to the number of ENFs in samples from both diabetic and control subjects Figure 5. Nonparametric statistical analysis of these quantities yielded similar relationships for diabetic and control samples. Higher aggregate nerve lengths and a greater number of branch points in control samples were a result of the greater overall number of ENFs in control subjects compared with diabetic samples (see Figure 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. Length and branching patterns in normal and diabetic epidermal nerve fibers (ENFs). (A) Scatter plot comparing ENF length with number of ENFs, each standardized for epidermal length. Control and diabetic subjects have the same average length of individual ENFs, as evidenced by the comparable slopes for both normal and diabetic samples. Length data was derived from Eutectic nerve tracing analysis. (B) Scatter plot comparing number of ENFs with the average number of branch points, each standardized for epidermal length. Control and diabetic subjects have a similar number of branch points per ENF.

All diabetic subjects had polyneuropathy. The severity of the neuropathy, based upon the graded results of the history, examination, quantitative sensory, and neurophysiologic data ranged from mild to severe. Correlation of skin biopsy results with the clinical data revealed a linear relationship in diabetics with mild to moderate neuropathy, but ENFs were absent in most patients with severe neuropathy Figure 6.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6. Comparison of clinical neurologic evaluation scores and epidermal nerve fiber length per volume of epidermis. The amount of nerve present in calf biopsy specimens correlates well with the clinical evaluation of diabetic subjects when score values were less than 80. Subjects with the most nerve have the best (lowest) clinical scores, but when clinical scores were greater than 50, epidermal nerve was usually absent. The peripheral and autonomic systems of the diabetic subjects were evaluated from graded results of history and examination, nerve conduction, cardiorespiratory reflexes, warm and cold sensitivity, and sweat testing.

Discussion.

This paper describes methods for processing skin biopsy specimens to obtain clear z series images of ENFs that can be used for three-dimensional measurements. The investigators traced and counted ENFs and measured their summed length relative to the surface length and volume of epidermis imaged. Epidermis from diabetic patients exhibited a reduction in nerve counts and summed length of ENFs. Epidermis from control and diabetic subjects had similar individual nerve lengths and number of branch points per fiber. Several nerve fibers of diabetic subjects ended bluntly at the dermal surface of the basement membrane. Nerves of the subepidermal plexus appeared to be reduced but were not quantified.

Although ENFs were not separately identified and studied, overall cutaneous innervation is known to be reduced in diabetic neuropathy. Lindberger et al. ^[15] described a reduction of CGRP-ir and SP-ir in dermal nerves in calf skin of diabetic patients aged 23 to 64 that was more severe in those with clinical neuropathy. Levy et al. ^[16] correlated clinical neurophysiologic measurements of small fiber abnormalities with histologic findings and found wide variation of nerve content in skin from control subjects and nerve loss in the epidermal-papillary dermis area of skin from diabetic patients, greater in those with more advanced clinical neuropathy. PGP 9.5-ir and VIP-ir nerve content of sweat glands was the same in control and diabetic subjects. Properzi et al. ^[21] found reduced CGRP-ir and PGP 9.5-ir nerves in the epidermal-papillary dermis and reduced PGP 9.5-ir and VIP-ir nerves in sweat glands of patients with more severe neuropathy.

Efforts to perform quantitative measurements of cutaneous nerves include counting nerve segments more than 10 micro meter in length, ^[15] double-blind grading on an arbitrary scale, ^[16,17] and computer-based methods. Computer analysis of enhanced, low-intensity camera images from the epifluorescent microscope enabled Terenghi et al. ^[22] to analyze total field fluorescence, nerve fibers per field, and nerve intercepts per mm of immunoreactive nerve fibers; they reported decreased innervation in the epidermis-subepidermis region of skin from patients with Raynaud's phenomenon and in the dermis and sweat glands of patients with systemic sclerosis. Several investigators used the same computer analysis system including the above referenced reports by Levy et al. ^[16] and Properzi et al., ^[21] and also by Molina et al., ^[18] who showed increased innervation of lesioned skin of patients with nodular prurigo. McCarthy et al. ^[19] separately identified and counted ENFs per length of skin surface in thick immunostained sections and found marked reduction in patients with predominately sensory neuropathy and with neuropathy associated with autoimmune deficiency syndrome but less reduction in most HIV-positive, neurologically normal patients.

We encountered several problems during attempts to quantify the ENFs with traditional epifluorescence microscopy. Segmented epidermal nerves present in conventional frozen sections (8 to 10 micro meter thick) make accurate length analysis impossible. Use of thicker sections resulted in optical restrictions: only a defined narrow depth of tissue could be brought into sharp focus; out-of-focus light blurred nerve images; spatial relationships (such as distinguishing between two overlying nerve fibers) were poorly perceived; and the amount of fluorophore-emitted illumination appeared to be bright at the top of the thick section but diminished deeper in the specimen. Consequently, counting techniques gave inconsistent results. Use of an LSCM overcame most of these problems. The instrument uses a pinpoint of laser light to scan the specimen and excite fluorophores at narrowly defined elevations in a thick section. The resulting accumulation of images at sequential preselected focal intervals results in a z series of sharply focused digitized images that are optical slices of 100-micro meter thick sections.

Visualization of epidermal nerves is further enhanced when the epidermis is prevented from curling during the mounting procedure. Curling compromises accurate counting of nerve fibers and prevents precise nerve reconstruction. The curling tendency is overcome by adhering the stained sections to microscope slides with agar, rather than allowing them to air dry. The agar-adhered sections remain in place during subsequent dehydration, clearing, and mounting in DPX. Agar also prevents air-bubble formation and hinders compression of the section by supporting the coverslip. The cyanine dyes are stable in DPX mountant during long-term storage and no reduction in fluorophore intensity has been detected in samples stored over 2 years.

We used the Eutectic system and IV to analyze selected three-dimensional structures from LSCM image files. The Eutectic system enables researchers to make three-dimensional reconstructions of entire neurons, including axon and dendrites, from video projected microscope images; a recent enhancement accepts confocal z series. We reconstructed ENFs from the site where they penetrated the basement membrane to their endings. The decision to begin measurement at the basement membrane (instead of in the dermis, where single nerve fibers are often identified as branches from the SNP) provided a constant reference point, but increased the nerve fiber count as some nerves bifurcate or trifurcate below basement membrane. Accurate nerve fiber counts, measurements of fiber length, and branch point identification were possible from the traced images. Further details of the innervation pattern, such as clustering and surface projection of endings, will be possible with software development.

Decreased nerve counts of ENFs in diabetic subjects indicated nerve loss occurring at a site proximal to the dermal-epidermal basement membrane. Nerve counts should allow comparison with results of other authors. ^[19] The length of individual nerve fibers and the summed nerve length related to epidermal volume and surface length provides a more accurate representation of epidermal innervation than counts of the number of ENFs. In some of our patients, epidermal volume was noticeably reduced (not statistically significant). Also, we found that ENF length is not static. The ENFs can accommodate their length to an increased or decreased epidermal volume, having exaggerated lengths in psoriasis and shorter length in laser-treated skin. ^[23]

The observation that ENF length and degree of branching of surviving ENFs was the same in diabetic and control skin was unexpected considering the overall loss of ENFs in diabetic skin. We had anticipated increased intraepidermal branching akin to the common observation of nodal and endplate branching of alpha motor axons in diabetes. The absence of increased terminal branching suggests that ENF degeneration was proximal to basement membrane without partial retraction or "dying back" within the epidermis. Nerve regeneration or collateral branching, if present, is probably located in the dermis or more centrally.

Many nerve fibers in diabetic, but not control, subjects ended abruptly at or near the basement membrane of the dermal-epidermal junction. Similar blunt nerve endings were not present elsewhere in the papillary or reticular dermis as might be expected if the nerves were undergoing "dying back" degeneration. The nerves appeared to be attempting regeneration through basement membrane. If so, we question whether they were prevented from re-entering epidermis by the normal or an altered basement membrane. Changes of pore size or charge of the basement membrane mesh or of its glycoprotein constituents can be considered. Passage of solutes through kidney basement membrane, for instance, is influenced by the negative charge of heparin sulfate proteoglycans. ^[24] This proteoglycan is diminished in nephrotic diabetic mice. ^[25] Basement membrane thickness of the kidney, eye, and perhaps other organs is increased in diabetes. This may restrict pore size, which normally is about 9 to 32 nm, ^[26] considerably smaller than the diameter of ENFs. Additionally, other structural constituents of basement membrane or stored constituents, such as epidermal growth factor, basic fibroblast growth factor, and Ca sup ++, may be relevant for inhibition or stimulation of intraepidermal nerve growth at the dermal-epidermal junction. Little information is available about the effect of diabetes on basement membrane in skin. Of possible significance is the observation that some regenerating nerves to the mouse hind paw do penetrate the basement membrane, but in reduced numbers that advance only part way to the stratum corneum. ^[27] We did not observe obvious change of basement membrane thickness in skin biopsies from diabetic subjects.

The clinical studies performed on our diabetic patients are those routinely administered to all patients who present for a PTx at this institution. ^[2,20] We are encouraged that a correlation between clinical scores and ENFs is linear in diabetic subjects with mild to moderate neuropathy; assessment of more severely affected diabetics may require analysis of other nerves in skin, such as the subepidermal neuroplexus or sweat gland innervation, because ENFs are usually absent. This initial study suggests that clinical evaluation of patients, including tests for sensory and autonomic motor function, will be useful for determining the most advantageous site for skin biopsy. Selection of a more proximal biopsy site, as determined by the clinical evaluation and inclusion of patients with less severe neuropathy, will result in a greater number of ENFs for quantitation and improve the correlation with clinical findings.

A great advantage of performing skin biopsy to study peripheral nerves is the innocuous nature of the procedure. Scar formation is minimal and there is reduced potential for the sensory deficit, dysesthesia, and infection that can follow sural nerve biopsy. ^[28,29] The simplicity of the procedure increases the opportunity for biopsy of additional sites with consequent reduction of sampling error and also to perform follow-up biopsies to evaluate treatment. Sensory and sympathetic nerves can be identified at their terminals, allowing possible correlation with results of quantitative tests of touch and thermal sensation, sweating, and localized vasomotor reaction. This reduces the necessity of using electron microscopy to visualize and quantify unmyelinated nerves. ^[30] Because epidermal nerve content varies with subject age and site of biopsy, age-matched control values are necessary for each site studied.

We chose to study diabetic candidates for PTx because of our familiarity with the nature of their neuropathy, ^[2,20] and because of the potential opportunity to obtain future post-PTx skin biopsy specimens from patients who undergo successful transplants and who achieve a euglycemic state. The extreme neuropathic changes in these patients can be detected by examination of PGP 9.5 immunostained skin with the epifluorescent microscope; subtle changes in degree of neuropathy that may result from therapeutic intervention, however, require more exacting evaluation. Quantitation of the number and morphologic characteristics of surviving nerves in the epidermis and in other areas of skin will be necessary for successful evaluation of the efficacy of therapeutic attempts. If stabilization of blood glucose after successful PTx facilitates nerve regeneration, detection of improvement presumes an ability to differentiate between the initial abnormal innervation pattern and an improved but still abnormal innervation after treatment. This will require a degree of quantification that we believe can be accomplished by measuring and describing ENFs. This degree of quantitation may also be necessary for early diagnosis and longitudinal monitoring of polyneuropathy.

Acknowledgments

We thank Michele Illies, Sally Haugen, and Frederick Sahinen for their technical assistance. We are grateful to Dr. Will Town of Eutectic Electronics, Inc. (Division of Sun Technologies Group Inc.) for his generosity in facilitating use of the Eutectic system.