Epidermal Nerve Fiber Length Density Estimation Using Global Spatial Sampling in Healthy Subjects and Neuropathy Patients

Transcription

1 J Neuropathol Exp Neurol Copyright Ó 2013 by the American Association of Neuropathologists, Inc. Vol. 72, No. 3 March 2013 pp. 186Y193 ORIGINAL ARTICLE Epidermal Nerve Fiber Length Density Estimation Using Global Spatial Sampling in Healthy Subjects and Neuropathy Patients Páll Karlsson, MSc, Anette Torvin MLller, MD, PhD, Troels Staehelin Jensen, MD, DMSc, and Jens Randel Nyengaard, MD, DMSc Abstract Assessment of intraepidermal nerve fiber density (IENFD) has become a useful tool for the investigation of patients with suspected smallfiber neuropathy (SFN). Here, we estimate epidermal nerve fiber lengths in 12 patients with SFN and 36 healthy controls using global spatial sampling and compare the lengths with IENFD and axonal swelling ratios. Skin biopsies were analyzed on 50-Km-thick free-floating sections immunostained for the neuronal cytoplasmic marker PGP 9.5. Mean IENFD in SFN patients was 2.22 T 1.63 mm j1 versus 7.51 T 2.17 mm j1 in controls; mean length density was 112 T 82.6 mm j2 in SFN patients versus 565 T 240 mm j2 in controls (p G for both). The correlation between the nerve fiber length and the IENFD was r = 0.16 for healthy subjects and r = 0.39 for patients, suggesting that these variables provide different quantitative information. There were significant differences in axonal swelling ratios between healthy subjects and patients, that is, per IENFD and per nerve fiber length. Together, these results suggest that, although length estimation requires more time and additional equipment, it is as effective as IENFD in differentiating SFN patients from healthy subjects. Estimating nerve fiber length may increase mechanistic understanding beyond IENFD estimation and improve efficiency in diagnosing SFN. Key Words: Global spatial sampling, Intraepidermal nerve fiber density, Nerve fiber length, Skin biopsy, Stereology. INTRODUCTION Small nerve fibers include myelinated A-delta and unmyelinated C fibers, which mediate pain, thermal sensation, and autonomic function. Damage to these fibers occurs in patients with small-fiber neuropathy (SFN) (1, 2). Patients with SFN may have sensory symptoms, including sensations of burning, tingling, numbness, allodynia, that is, pain caused by normally not painful stimulus, hyperalgesia, and hypersensitivity to heat and/or cold. Symptoms typically arise distally in a glove and From the Danish Pain Research Center, Aarhus University Hospital, Aarhus, Denmark (PK, TSJ); Department of Neurology, Aarhus University Hospital, Aarhus, Denmark (ATM, TSJ); and Stereology and Electron Microscopy Laboratory, Centre for Stochastic Geometry and Advanced Bioimaging, Aarhus University Hospital, Aarhus, Denmark (PK, JRN). Send correspondence and reprint requests to: Páll Karlsson, MSc, Danish Pain Research Center, Aarhus University Hospital, Norrebrogade 44, Bldg 1A, DK-8000 Aarhus C, Denmark; pkar@svf.au.dk The authors would like to acknowledge the Villum Foundation for supporting the Centre for Stochastic Geometry and Advanced Bioimaging, Aarhus University. 186 stocking pattern. Patients may acquire an SFN in association with diabetes as the most frequently known cause or after injury, infection, alcoholism, cancer, or any of many other diseases (2). However, in many cases, no specific cause of the neuropathy can be found, and it is classified as idiopathic neuropathy. Verifying the clinical suspicion of SFN can be difficult because electroneuronography is often normal. Quantitative sensory tests are a psychophysic measurement of small-nerve function, whereas skin punch biopsies can provide an objective specific measurement of C fiber density. A combination of quantitative sensory testing, examination of the damage to the small and large nerve fibers, and skin punch biopsy, where epidermal nerve C-fiber density is calculated, is often used for diagnosis. Determining intraepidermal nerve fiber density (IENFD) or nerve fiber intersections from the dermis to epidermis by immunostaining the C fibers that penetrate the epidermal layer is becoming an increasingly used tool for investigation of patients with suspected SFN (3). Intraepidermal nerve fibers are unmyelinated nociceptors that arise from a subepidermal neural plexus: axon bundles just beneath the dermal-epidermal basement membrane in the superficial dermis (4). Skin biopsy has proven to be both a specific and sensitive way for diagnosing peripheral neuropathy, with a diagnosing efficiency up to 90% (5). Skin biopsies from patients with peripheral neuropathy show a significant reduction of small nerve fibers in the epidermis in comparison with age- and sex-matched healthy subjects (6). The European Federation of Neurological Societies and the Peripheral Nerve Society have developed guidelines on the use of skin biopsies for the diagnosis of SFN (7). They recommend the use of light microscopy and preferably the use of the free-floating immunohistochemistry protocol described by McCarthy et al (8). The free-floating protocol is most frequentlyusedwhenvisualizing small nerve fibers from skin biopsies. Morphologic changes of the nerve fibers have also been investigated. Increased axonal swellings may predict degeneration of epidermal fibers, although swellings are also observed in healthy subjects (4, 9Y11). Despite high diagnostic efficiency, at least 12% of SFN patients show a normal IENFD (5). When investigating the morphology of the nerve fibers, their length within the epidermis varies substantially, possibly hiding further extractable information from the skin biopsy about the neuropathy. There is one previous investigation of nerve fiber length density J Neuropathol Exp Neurol Volume 72, Number 3, March 2013

2 J Neuropathol Exp Neurol Volume 72, Number 3, March 2013 within the epidermis (12). However, in that study, the epidermal nerve fiber length density was estimated in only 2 healthy subjects, leaving the parameter largely uninvestigated both in healthy subjects and in patients with SFN. Obtaining an accurate estimation of nerve fiber length is technically challenging. The 3-dimensional stereologic sampling technique of global spatial sampling generates isotropic focal planes, thereby making it possible to obtain an accurate estimation of nerve fiber length density within the epidermis (13). When a feature has a random orientation, it is said to be isotropic. This is seldom the case for biologic features and does not apply to skin nerve fibers (14). The aim of this study was to estimate IENFD using global spatial sampling in healthy subjects and SFN patients and to compare nerve fiber lengths with IENFD and axonal swellings. MATERIALS AND METHODS Subjects Skin biopsies from 36 healthy subjects aged 26 to 80 years (5 women; mean age, 53.4 years) and 12 patients with SFN aged 36 to 77 years (5 women; mean age, 60.5 years) were studied. We also included an additional 6 healthy subjects to investigate possible quantitative differences between 2 immunostaining protocols. The diagnosis of SFN was made on the basis of clinical examination and results from quantitative sensory testing. All participants gave their written informed consent for the study, which was approved by the local ethics committee (No ) in accordance with the Declaration of Helsinki. Skin Biopsies All skin biopsies were taken under sterile conditions from the distal leg 10 cm above the lateral malleolus with a 3-mm disposable biopsy punch (Miltex, York, PA). The participants were offered topical anesthesia with lidocaine before the procedure. After 4-hour fixation in 4% phosphate-buffered paraformaldehyde, the biopsies were washed in 0.1 mol/l PBS and stored in 10% sucrose with 0.1 mol/l PBS for 2 days. Immunostaining Protocols To investigate whether there were any quantitative differences using the immunostaining protocol used here and the one recommended by the European Federation of Neurological Societies and the Peripheral Nerve Society (7), we obtained additional biopsies from 6 healthy individuals (10 cm above the lateral malleolus on both legs). One biopsy was analyzed using our protocol (Protocol I) and the other using the recommended protocol (Protocol II). The biopsies used for analysis with Protocol II were fixed in 2% paraformaldehydelysine-periodate overnight, followed by cryoprotection in a solution containing 20% glycerol and 0.08 mol/l Sorenson PBS overnight. All biopsies were frozen to j55-cintissue-tek(sakura, Alphen aan den Rijn, Netherlands). Cryostat sections were cut into 50-Km thicknesses vertically in relation to the epidermis direction (Microm Cryostat HM 500 OM, Zeiss, Germany). Each biopsy was divided into 3 sets using systematic, uniformly Epidermal Nerve Fiber Length Estimation random sampling, with each set containing 10 to 12 tissue sections. Until further treatment, the biopsies were stored in a j20-c freezer in De Olmers antifreeze liquid. ForimmunostainingProtocolI,1setofsectionsfrom each biopsy was randomly selected and washed in PBS before blocking endogenous peroxidase and target retrieval with TEG buffer (Tris, EGTA, water). The sections were immunoreacted free-floating with rabbit anti-human PGP 9.5 (1:1000; AbD Serotec, Dusseldorf, Germany) as a primary antibody and horseradish peroxidaseymarked goat anti-rabbit as secondary antibody (1:200; DakoCytomation, Glostrup, Denmark). The antibodies were detected using diaminobenzidine, mounted on slides, rehydrated, counterstained with hematoxylin, and dehydrated. For immunostaining Protocol II, 1 set of sections from each biopsy was randomly selected and washed in PBS before a blocking step using 0.5% powdered milk, 1% Triton X-100, and 4% normal serum in Tris-buffered saline for 2 hours. Sections were incubated free-floating with a mixture of anti-pgp 9.5 (1:1000) as a primary antibody, 0.5% Triton X-100, 2% normal serum, and 0.5% powdered milk overnight. On Day 2, the sections were immunoreacted with biotinylated goat antirabbit secondary antibody (1:100; Vector Laboratories, Burlingame, CA) using the same solution as with primary antibody for 1 hour. After endogenous peroxidase removal with 33% methanol/pbs and 3.3% H 2 O 2 for 30 minutes and 1 hour in Vectastain (Vector Laboratories), the antibodies were detected using a peroxidase substrate SG kit (Vector Laboratories) before mounting, rehydration, counterstaining, and dehydration. Analysis Sections were analyzed using an Olympus BX51 microscope connected to an Olympus DP70 camera (Olympus, Tokyo, Japan), a Heidenhain ND 281 encoder (Heidenhain, Schaumburg, IL), and a Prior Proscan II motorized stage (Prior Scientific Inc., Rockland, MA). All hardware was connected to a computer and controlled via newcast software (Visiopharm, Hoersholm, Denmark). A minimum of 3 sections per individual was analyzed. The length of each section was measured at 10 magnification by drawing freehand along the epidermis of each section. Small fibers originating in the dermis and crossing the basal membrane to the epidermis were counted under a light microscopy to achieve the number of crossing nerve fibers per length (in millimeters) of epidermis (IENFD). Only a single fiber crossing the membrane was counted, excluding secondary branches. The morphology of the fibers was evaluated with a 60 oil immersion lens (Olympus UPlanSApo; NA = 1.35). All sections were counted by the same investigator (Páll Karlsson) blinded to the origin of the biopsies. To ensure the validity of results, 30 random sections were recounted at the end of the study. Global Spatial Sampling Global spatial sampling generates isotropic section planes by introducing virtual isotropic test planes (Fig. 1). Randomly oriented isotropic virtual planes are superimposed on the computer screen over the region of interest, thereby creating isotropy of the test planes, so they have an equal probability of Ó 2013 American Association of Neuropathologists, Inc. 187

3 Karlsson et al J Neuropathol Exp Neurol Volume 72, Number 3, March 2013 intersecting the nerve fibers. The virtual planes are randomly oriented in virtual 3-dimensional sampling boxes containing parallel planes separated by a fixed distance (d). The software superimposes sampling boxes over the sample until a predefined percentage of the area of interest has been estimated. For each field of view, the orientation of the virtual plane is changed at random. Global spatial sampling can be used on thick sections with an arbitrary section orientation, making it possible to estimate the length of tubular objects such as nerve fibers (10). The skin sections were investigated with randomized isotropic virtual planes in systematically sampled fields of view. A minimum of 3 skin sections in which IENFD had already been estimated was selected using systematic sampling. Global nerve fiber length density in epidermis, L V (nerve fiber/epidermis) is estimated by: L v ðnervefiber=epiþ ¼ ~QðnervefiberÞ ~að planeþ ¼ 2pðboxÞ avg aðplaneþ : ~QðnervefiberÞ ~PðepiÞ where ~QðnervefiberÞ is the sum of epidermal nerve fibers counted, p(box) is the number of test points in 1 sampling box, avg a(plane) is the average of the sum of areas of isotropic oriented planes in 1 sampling box, and ~PðepiÞ is the number of sampling box corners (equal to test points) hitting the region of interest. The sampling box volume divided by the plane distance is a(plane). Before the sampling, the height of the sampling box, excluding the upper and lower heights of the thick section to avoid artifacts, was set to 15 Km becauseaz analysis had shown a constant number of ~QðnerverfiberÞ in this interval. The size of the sampling box area was 4,800 Km 2, the sampling steps were Km (dx, dy), and the plane separation distance (d )was25km. After delineating the area of interest, the sampling sites were chosen systematically until a predefined fraction of the area was measured. At each sampling site, there was 1 sampling box containing randomly oriented virtual planes. Nerve fibers that intersect with the virtual planes at each field of view were counted before moving on to the next field of view until all fields of view had been examined. Each nerve fiber that crossed the isotropically oriented virtual plane was marked (Fig. 2), and the length density was then estimated using the equation above. FIGURE 1. Schematic representation of the intersection between the probe (isotropic virtual plane) and the focal plane (physical section). The virtual plane is represented on the computer screen as a line at the place where the virtual plane crosses the focal plane. When focusing along the z axis, the line moves accordingly. N p = normal plane. Modified with permission from Larsen et al (13). 188 Quantification of Axonal Swellings Axonal swellings with a diameter equal to or greater than 1.5 Km were counted on all nerve fibers evaluated during IENFD measurement by using the same microscope setup as previously described. An example of an axonal swelling is shown in Figure 3. Swellings below this value and swellings in the terminal part of the nerve fibers were considered having normal morphology and were not counted (9). Two different swelling ratios were calculated for healthy subjects and neuropathy patients, namely, number of swellings/ienfd and number of swellings/nerve fiber length. Statistics Pearson correlation was used to estimate the correlation between the nerve fiber intersections and the nerve fiber length density and Student t-test for comparison between healthy subjects and patients. The ratio of means was used to get the average nerve fiber length density per individual. To measure the observed relative variability for nerve fiber intersections, nerve fiber length density, swelling ratios, and tissue shrinkage, the coefficient of variance (CV = SD/mean) was calculated. We also measured how much of the total variance, CV(tot) in nerve fiber length could be explained by biologic variance and error variance: CV 2 ðtotþ ¼CV 2 ðbioþþce 2 ðsteþ: The coefficient of error CE(ste) is the error variance and can be estimated for nerve fiber length density being a ratio estimator (11): vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi n ~PðepiÞ 2 CEðsteÞ¼ n 1 2 þ ~IðnervefiberÞ 2 ub B 2 2~PðepiÞ :IðnervefiberÞCC AA ~PðepiÞ:~IðnervefiberÞ ~PðepiÞ ~IðnervefiberÞ where n is the number of sections counted, P(epi) is the Ó 2013 American Association of Neuropathologists, Inc.

4 J Neuropathol Exp Neurol Volume 72, Number 3, March 2013 Epidermal Nerve Fiber Length Estimation FIGURE 2. Estimation of nerve fiber length density using light microscopy connected to newcast in a section immunostained for the neuronal cytoplasmic marker PGP 9.5. Left: Overview of a 3-mm skin biopsy (4 objective). Right: The green line is a moving line and represents the place where the focal plane and the virtual plane intersect. When the plane cross, the nerve fibers in focus get a mark, and the software estimates their length. The two black lines delineate the area of interest. The sampling box is Km. number of sampling box corners hitting the region of interest, andi(nerve fiber) is the sum of epidermal nerve fibers marked. The equation requires that CE(P(epi)) is less than 0.1. RESULTS Epidermal Nerve Fiber Intersection Density and Length Density Mean nerve fiber intersections in healthy subjects was 7.51 T 2.17 mm j1, and mean fiber length density was 565 T 240 mm j2 (Figs. 4, 5). Mean nerve fiber intersections in SFN patients was 2.22 T 1.63 mm j1, and mean fiber length density was 112 T 82.6 mm j2 (Figs. 4, 5). Differences are significant for both parameters ( p G 0.001). The number of intersections declined with age in healthy subjects, whereas length density did not (Table). The correlation between the nerve fiber length density and the nerve fiber intersections was analyzed by Pearson correlation, giving r = 0.16 for healthy subjects, and r =0.39for patients. The mean coefficient of error for healthy subjects was 0.20 (ranging between 0.04 and 0.43); for SFN patients, it was 0.39 (ranging from 0.16 to 0.64). The coefficients of variance for IENFD and the length estimation were calculated, giving a CV(tot) of 0.29 and 0.37 for healthy subjects and 0.55 and 0.84 for patients. When calculating the impact of the error variance on the total variation for nerve fiber length density estimation, we used the ratio: CE2 ðsteþ CV 2, which should be in the ðtotþ range of 0.1 to 0.5. This was the case. Therefore, we can conclude that the main contributor to the relatively large total variance for nerve fiber length was the biologic variation and not the error variance from stereologic sampling. Other Findings Mean nerve fiber intersections using Protocol I was 5.76 T 1.48 mm j1 ; mean nerve fiber intersections using Protocol FIGURE 3. Example of an axonal swelling. The section is immunostained for the neuronal cytoplasmic marker PGP 9.5. The image was taken with a 60 objective of a section from a healthy male stained with protocol I. Blue scale bar = 1.6 Km; red scale bar = 20 Km. Ó 2013 American Association of Neuropathologists, Inc. 189

5 Karlsson et al J Neuropathol Exp Neurol Volume 72, Number 3, March 2013 FIGURE 4. (A) Epidermal nerve fiber length density plotted against age. (B) Epidermal nerve fiber length density plotted against nerve fiber intersections. (C) Nerve fiber intersections plotted against age. Healthy subjects are represented by closed dots and neuropathy patients with open dots. II was 5.98 T 1.73 mm j1 (p ). The correlation between the 2 methods was analyzed by Pearson correlation, giving r = Coefficient of variance for Protocol I was 0.26, and for Protocol II, CV was We calculated 2 axonal swelling ratios: the number of swellings per IENFD and the number of swellings per epidermal nerve fiber length density. Swelling ratios for healthy subjects were (number of swellings/ienfd, CV = 0.97) and (number of swellings/lv, CV = 1.10). Swelling ratios for SFN patients were 0.22 (number of swellings/ IENFD, CV = 1.86) and 0.05 (number of swellings/lv, CV = 2.31; p = for number of swellings/ienfd ratio; p = for number of swellings/lv ratio). To investigate tissue thickness shrinkage, we selected 5 sections at random from 8 controls (aged 26Y80 years; mean, 54.1 years) and measured the mounted section thickness at 3 different sites in the epidermis and dermis. Mean tissue thickness in the epidermis was 29.6 Km (CV = 0.11); in the dermis, the mean thickness was 22.8 Km (CV = 0.16). To investigate shrinkage on the xy axis, we measured the longest FIGURE 5. (A) Nerve fiber intersections for healthy subjects (closed dots) and neuropathy patients (open dots). (B) Epidermal nerve fiber length density for healthy subjects (closed dots) and neuropathy patients (open dots). Horizontal lines indicate means. 190 Ó 2013 American Association of Neuropathologists, Inc.

6 J Neuropathol Exp Neurol Volume 72, Number 3, March 2013 TABLE. Intraepidermal Nerve Fiber Density and Epidermal Nerve Fiber Length Density in Healthy Subjects Grouped by Age Age, years Mean Intraepidermal Nerve Fiber Density, no./mm Mean Epidermal Nerve Fiber Length Density, mm j2 20Y29 (n = 2) Y39 (n = 2) Y49 (n = 12) Y59 (n = 8) Y69 (n = 7) Y80 (n = 5) section for each participant. All participants had sections equal to or longer than 3 mm, ranging from 3.00 to 3.53 mm (CV = 0.05). DISCUSSION Epidermal Nerve Fiber Intersection Density Here we show that nerve fiber intersections decline with age in healthy subjects, and that there are significant differences between controls and SFN patients, as has been shown in other studies (9, 15Y18). Estimates of nerve fiber intersections in biopsies from the distal leg can be used in clinical practice as long as reliable normative reference values are known (18) and standardization of quantitative data from PGP 9.5Yimmunostained sections is possible. Fiber Length Density Up to 12% of SFN patients show normal IENFD (5). However, how far inside the epidermis the nerve fibers penetrate varies greatly, leading to questions of whether there is any difference in nerve fiber length in the epidermis between healthy subjects and neuropathy patients and whether this parameter increases the efficacy of the diagnosis. Few studies have dealt with length estimation of nerve fibers in human skin and none in the epidermis. Studies have tried to estimate nerve fiber length in the dermis (or a surrogate parameter of nerve fiber length) but with different methods and each with its limitation (19Y21). At least 2 studies found a correlation between IENFD and dermal nerve fiber length. In one of those studies, a combination of IENFD and dermal nerve fiber length gave higher diagnostic specificity and sensitivity (19, 21). To date, however, no method has been published where nerve fiber length can be estimated in a design-based manner by the use of 50-Km-thick skin sections analyzed by light microscopy, as recommended by the European Federation of Neurological Societies (6). The global spatial sampling method excludes the need for isotropic or vertical sections when estimating length (13). Although this method requires more complex hardware and software and takes a longer time to analyze than IENFD, nerve fiber length estimation provides a foundation for mechanistic understanding of peripheral neuropathy and adds new possibilities for assessing treatment effects. Estimating IENFD is a fast and effective method to differentiate patients from healthy Epidermal Nerve Fiber Length Estimation subjects, and it is a very simple counting method that does not take into consideration the biologic complexity of the skin and the nerve fibers. Estimating nerve fiber lengths takes into account the nerve fibers form and twists and turns in space. By adding epidermal nerve fiber length measurement to the IENFD, along with results of quantitative sensory testing, a more complete profile of the patients can be achieved, which may improve the diagnostic yield for SFN. To obtain unbiased nerve fiber length estimation, either the nerve fibers or the stereologic probe, that is, a test plane, needs to be isotropic. It cannot be assumed that nerve fibers in skin are isotropic. There are several ways to generate isotropic test planes: 1) thin isotropic sections and a 2-dimensional counting frame; 2) thick vertical sections using total projection; and 3) thick arbitrary sections and global spatial sampling or space balls (virtual planes) (22). Thin isotropic sections would provide very few counts per section, and it would probably also be difficult to recognize all the small positive nerve profiles. Total projection of skin nerve fibers in thick sections may not be feasible because of their complex organization. The space ball method uses spheres to create isotropy; global spatial sampling uses randomly rotated parallel virtual planes (22). It may seem ideal to use spheres to create isotropy; however, the problem is that the curved lines in a sphere may suffer from a bias. The bias is determined by the ratio of the diameter of the probe and the diameter of the structure that is being measured (23). Because the sphere is asymmetric because of its curvature, it may either get false-positive marks or false-negative marks that do not balance each other out, thereby yielding a net positive bias. This bias does not occur in global spatial sampling where straight lines are used. We have shown that it is possible to achieve a quantitative estimation of the nerve fiber length in human skin biopsies. Furthermore, we report a clear significant difference in epidermal nerve fiber length between healthy subjects and neuropathy patients. The estimation can serve as another parameter when investigating suspected neuropathy and add further information to the patient s neurologic profile. The mean coefficient of error, which measures the precision of the length density estimation, was calculated to be The coefficient of variance for the length estimation for healthy subjects was 0.37; for patients, 0.74, indicating a high biologic variance. A poor correlation was observed between nerve fiber density and length density for patients (Pearson, r =0.39), and the correlation was even lower for healthy subjects (Pearson, r = 0.16), yielding a coefficient of determination (r 2 )of0.15forpatientsand0.03for healthy subjects. These results suggest that these variables are quite different. The reason for the difference in correlation between healthy subjects and patients may be because nerve fiber length density has an unknown maximal value, which is not related to the density of nerve fibers and/or because patients typically have degenerated nerve fibers in combination with abnormal morphology, resulting in a low nerve fiber length density per squared area, which would correlate with a low IENFD. Approximately 15% of the change in nerve fiber length density can be explained by IENFD in patients and only 3% in healthy subjects. Furthermore, whereas IENFD Ó 2013 American Association of Neuropathologists, Inc. 191

7 Karlsson et al J Neuropathol Exp Neurol Volume 72, Number 3, March 2013 declines with age, epidermal nerve fiber length does not correlate with age. A clear correlation between low IENFD, loss of sensory function, and presence of pain in patients has not yet been shown. MLller et al (24) suggested that the lack of correlation might be explained by some present nerve fibers having an increased function in patients who have allodynia and/or hyperalgesia in the painful area. The present analysis of 36 healthy subjects (plus 6 additional healthy subjects to compare 2 immunohistochemical staining methods) and 12 SFN patients has a relatively small sample size for statistical analysis. Although we would not expect mean values to change significantly, the sample size should be increased further to verify the results reported here. It is of interest to investigate further the significance of the difference of epidermal nerve fiber length observed between healthy subjects and neuropathy patients, including investigating the correlation between the nerve fiber length with other tests available such as quantitative sensory testing to investigate the correlation between shorter epidermal nerve fibers and sensory symptoms. Normative reference values should be obtained. Estimating fiber length density is more time consuming than estimating nerve fiber intersections. Scanning through 1 section takes about 20 minutes for a trained investigator. Axonal Swellings Axonal swellings are observed even when the IENFD is in the reference range and might, therefore, be a predegenerative marker of nerve fibers. We found a statistically significant difference in swelling ratio (number of swellings/ienfd) between healthy subjects and patients, which fits well with previous findings (8Y10, 25). We further report a statistically significant difference in swelling ratio versus nerve fiber length (no swellings/lv) between healthy subjects and patients. There was high variability in both swelling ratios (data not shown). However, there is no clear definition of axonal swelling in nerve fibers, and different methods have been used to analyze data, which makes comparison between studies difficult. For example, Herrmann et al (10) define swellings as enlargements more than twice the diameter of the parent fiber, whereas Lauria et al (9) define swellings as enlargements more than 1.5 Km. In a third study, swellings were divided into large swellings that are up to 5-fold the diameter of the parent fiber and small swellings that are 2- to 3-fold the diameter of the parent fiber (25). Lauria et al (9) calculated the swelling ratio (number of swellings/ienfd) and compared the findings between healthy subjects and patients, whereas other studies have counted the number of swellings according to their definition with no regard to the number of epidermal fibers (8Y10, 25). 192 Other Methodological Considerations The cut section thickness was set at 50 Km at the cryostat; however, during the histologic preparation, frozen sections shrink in the z axis. Although the section shrinks, the nerve fibers may or may not shrink to the same degree. This may introduce a bias in the estimation of the nerve fiber length that can be difficult to correct. We selected 5 random sections from 8 participants and measured the mounted section thickness at different sites in the epidermis and dermis. The measurements showed shrinkage of approximately 40% of the epidermis and approximately 50% of the dermis. In contrast to what we expected, younger subjects did not show more shrinkage compared with older subjects. To see if there was any xy shrinkage, we measured the longest section for each participant. All participants had sections equal to or longer than 3 mm, suggesting little or no xy shrinkage, corresponding to data in the literature (26). Estimates of nerve fiber length density and IENFD (as well as of axonal swellings) provide results as ratios, for example, fiber length density is length of nerve fibers per volume epidermis. These ratios may not necessarily allow conclusions about total length (27). Therefore, densities should be multiplied with the reference volume or area to avoid dubious conclusions (27). It is difficult to estimate the total volume of epidermis or the surface of the epidermal basal membrane in living subjects. The conclusions drawn from these ratios do, therefore, require that the total volume of epidermis or the surfaces of the epidermal basal membrane are unchanged between the patients and healthy subjects. We did not identify significant quantitative differences between results using Protocol I and Protocol II, that is, the one used in a large reference study (6). Protocol I takes less time and has fewer steps, whereas Protocol II sometimes shows nerve fibers clearer, making morphologic changes in the nerve fibers easier to detect. CONCLUSIONS We report the first study in which fiber length density is estimated in epidermis from skin biopsies using global spatial sampling. Small-fiber neuropathy patients had significantly shorter nerve fibers in the epidermis versus those in healthy subjects. Although estimating IENFD is a quick, yet effective, method to differentiate patients from healthy subjects, estimating nerve fiber length may increase mechanistic understanding. Length estimation is more time consuming and requires additional equipment than IENFD but is equally effective in differentiating patients from healthy subjects. REFERENCES 1. Pan CL, Tseng TJ, Lin YH, et al. Cutaneous innervation in Guillain-Barré syndrome: Pathology and clinical correlations. Brain 2003;126:386Y97 2. Tavee J, Zhou L. Small fiber neuropathy: A burning problem. Cleveland Clinic J Med 2009;76:297Y Lauria G, Lombardi R, Camozzi F, et al. Skin biopsy for the diagnosis of peripheral neuropathy. Histopathology 2009;54:273Y85 4. Wendelschafer-Crabb G, Kennedy WR, Walk D, et al. Morphological features of nerves in skin biopsies. J Neurol Sci 2006;242:15Y21 5. Devigili G, Tugnaoli V, Penza P, et al. The diagnostic criteria for small fibre neuropathy: From symptoms to neuropathology. Brain 2008;131: 1912Y25 6. Lauria G, Bakkers M, Schmitz C, et al. Intraepidermal nerve fiber density at the distal leg: A worldwide normative reference study. J Peripher Nerv Syst 2010;15:202Y7 7. Lauria G, Hsieh ST, Johansson O, et al. European Federation of Neurological Societies/Peripheral Nerve Society Guideline on the use of skin biopsy in the diagnosis of small fiber neuropathy. Report of a joint task force of the European Federation of Neurological Societies and the Peripheral Nerve Society. Eur J Neurol 2010;17:903Y12 8. McCarthy BG, Hsieh ST, Stocks A, et al. Cutaneous innervation in sensory neuropathies: Evaluation by skin biopsy. Neurology 1995;45: 1848Y55 Ó 2013 American Association of Neuropathologists, Inc.

8 J Neuropathol Exp Neurol Volume 72, Number 3, March Lauria G, Morbin M, Lombardi R, et al. Axonal swellings predict the degeneration of epidermal nerve fibers in painful neuropathies. Neurology 2003;61:631Y Herrmann DN, McDermott MP, Henderson D, et al. Epidermal nerve fiber density, axonal swellings and QST as predictors of HIV distal sensory neuropathy. Muscle Nerve 2004;29:420Y Gibbons CH, Griffin JW, Polydefkis M, et al. The utility of skin biopsy for prediction of progression in suspected small fiber neuropathy. Neurology 2006;66:256Y Stocks EA, McArthur JC, Griffin JW, et al. An unbiased method for estimation of total epidermal nerve fiber length. J Neurocytol 1996;25:637Y Larsen JO, Gundersen HJ, Nielsen J. Global spatial sampling with isotropic virtual planes: Estimators of length density and total length in thick, arbitrarily orientated sections. J Microsc 1998;191:238Y Nyengaard JR. Stereologic methods and their application in kidney research. J Am Soc Nephrol 1999;10:1100Y Pan CL, Lin YH, Lin WM, et al. Degeneration of nociceptive nerve terminals in human peripheral neuropathy. Neuroreport 2001;12:787Y Goransson LG, Mellgren SI, Lindal S, et al. The effect of age and gender on epidermal nerve fiber density. Neurology 2004;62:774Y Pittenger GL, Ray M, Burcus NI, et al. Intraepidermal nerve fibers are indicators of small fiber neuropathy in both diabetic and nondiabetic patients. Diabetes Care 2004;27:1974Y Bakkers M, Merkies ISJ, Lauria G, et al. Intraepidermal nerve fiber density and its application in sarcoidosis. Neurology 2009;73:1142Y48 Epidermal Nerve Fiber Length Estimation 19. Vlckova-Moravcova E, Bednarik J, Dusek L, et al. Diagnostic validity of epidermal nerve fiber densities in painful sensory neuropathies. Muscle Nerve 2008;37:50Y Tschachler E, Reinisch CM, Mayer C, et al. Sheet preparations expose the dermal nerve plexus of human skin and render the dermal nerve end organ accessible to extensive analysis. J Invest Dermatol 2004; 122:177Y Lauria G, Cazzato D, Poretta-Serapiglia C, et al. Morphometry of dermal nerve fibers in human skin. Neurology 2011;77:242Y Calhoun ME, Mouton PR. Length measurement: New developments in neurostereology and 3D imagery. J Chem Neuroanat 2000;20:61Y Gundersen HJ. Stereological estimation of tubular length. J Microsc 2002;207:155Y MLller AT, Bach FW, Feldt-Rasmussen U, et al. Autonomic skin responses in females with Fabry disease. J Peripher Nerv Syst 2009;14: 159Y Ebenezer GJ, McArthur JC, Thomas D, et al. Denervation of skin neuropathies: The sequence of axonal and Schwann cell changes in skin biopsies. Brain 2007;130:2703Y Dorph-Petersen KA, Nyengaard JR, Gundersen HJ. Tissue shrinkage and unbiased stereological estimation of particle number and size. J Microsc 2001;204:232Y Braendgaard H, Gundersen HJ. The impact of recent stereological advances on quantitative studies of nervous system. J Neurosci Meth 1986; 18:39Y78 Ó 2013 American Association of Neuropathologists, Inc. 193