Bilateral deficiency of Meissner corpuscles and papillary microvessels in patients with acute complex regional pain syndrome

Immunofluorescence analysis of skin revealed disturbances in vascular microvessel and Meissner corpuscles but not in claudin-1, claudin-5, claudin-19, and nerve fiber density in cutaneous nerves.

Abstract

Complex regional pain syndrome (CRPS) presents postinjury with disproportionate pain and neuropathic, autonomic, motor symptoms, and skin texture affection. However, the origin of these multiplex changes is unclear. Skin biopsies offer a window to analyze the somatosensory and vascular system as well as skin trophicity with their protecting barriers. In previous studies, barrier-protective exosomal microRNAs were altered in CRPS. We here postulated that tissue architecture and barrier proteins are already altered at the beginning of CRPS. We analyzed ipsilateral and contralateral skin biopsies of 20 fully phenotyped early CRPS patients compared with 20 age- and sex-matched healthy controls. We established several automated unbiased methods to comprehensively analyze microvessels and somatosensory receptors as well as barrier proteins, including claudin-1, claudin-5, and claudin-19. Meissner corpuscles in the skin were bilaterally reduced in acute CRPS patients with some of them lacking these completely. The number of Merkel cells and the intraepidermal nerve fiber density were not different between the groups. Dermal papillary microvessels were bilaterally less abundant in CRPS, especially in patients with allodynia. Barrier proteins in keratinocytes, perineurium of dermal nerves, Schwann cells, and papillary microvessels were not affected in early CRPS. Bilateral changes in the tissue architecture in early CRPS might indicate a predisposition for CRPS that manifests after injury. Further studies should evaluate whether these changes might be used to identify risk patients for CRPS after trauma and as biomarkers for outcome.

1. Introduction

Complex regional pain syndrome (CRPS) develops after limb fracture or trauma. Based on the Budapest criteria, symptoms include pain disproportionate to the inciting event and signs of neurogenic inflammation with barrier disruption clinically evident as edema, autonomic changes, motor dysfunction, and sensory abnormalities.²⁷

In our previous study, quantitative sensory testing confirmed hyperalgesia and allodynia in patients with acute CRPS—both symptoms can originate in the peripheral nervous system.¹⁸ Corresponding receptors are rapidly adapting mechanoreceptors: Meissner corpuscles detect pressure and vibration, whereas Merkel cells are essential for light touch sensation. In a recent preclinical study, miswiring of Meissner corpuscles caused neuropathic pain.²² Similarly, an amputee study of 2 CRPS patients described Meissner corpuscles with disrupted Aβ-innervation in skin samples.² These results have not been confirmed in a larger cohort.

Barriers shield the nervous system through the blood–nerve barrier, including the perineurium and endoneurial vessels and the myelin barrier.^5,47 In preclinical neuropathy models, the blood–nerve barrier and the myelin barrier are leaky.^13,48,49 Complex regional pain syndrome patients showed decreased exosomal microRNA-183 in blood. MicroRNA-183 restores the microvascular barrier formed by the tight junction protein claudin-5 through inhibiting FoxO1 in preclinical models and cell cultures.⁴⁸ Similarly, tight junctions in the skin control edema formation.^38,58 Therefore, breakdown of barriers could be a common pathway for several symptoms in CRPS.

In skin, mainly 2 tight junction proteins have been studied before: epidermal claudin-1 shields against external harmful agents (eg, microbes) and protects against loss of water.^5,21 It is dysregulated in atopic dermatitis.^7,53 In peripheral nerves, claudin-1 seals the perineurium.^26,49 Endothelial claudin-5 is expressed in blood vessels and lymphatic vessels of the dermis and regulates their permeability.^1,32,38 It is altered in angiogenesis in psoriasis and protects against edema after UV injury. Another tight junction protein—claudin-19, not studied in skin yet—seals the myelin barrier in peripheral nerves at the site of the paranode.³⁹

Skin is an easily accessible tissue providing insight into not only the somatosensory system but also the immune, autonomic, and vascular skin system. Skin biopsies from CRPS patients have helped to elucidate pathophysiological mechanisms³: increased macrophage inflammatory protein-1, interleukin-6, tumor necrosis factor-α, and mast cell accumulation mediate inflammation; disbalance of endothelin-1 and nitrous oxide synthase-1 leads to vasomotor disturbances; increased adrenoceptor expression results in autonomic dysfunction; and activation of protease networks is responsible for neuropeptide accumulation.^{8,20–22,25,36,41} Morphologically, trophic changes like epidermal thickening were detected.⁸ Skin innervation to analyze neuropathy was performed only in chronic CRPS.^31,43,46 Furthermore, many of these studies lack control tissues from the same location from healthy individuals.

We hypothesized, that trophicity, vascular supply, and somatosensory system are altered in skin of patients with early CRPS. We further postulated that these alterations would be reflected by changes of barrier proteins in the skin cells and affect the morphology. Analysis of these structures might identify potential diagnostic markers or targets for individualized treatment.

2. Materials and methods

2.1. Patient cohort

This is an interim analysis within the study ResolvePAIN CRPS registered in the German clinical trial register (DRKS00016790) and approved by local authorities. Patients between 18 and 85 years presenting at the Center for Interdisciplinary Pain Medicine at the University Hospital Würzburg were included if they fulfilled the diagnostic Budapest criteria.²⁷ Healthy controls completely free of pain or sensory deficits served as control group. All participants gave written informed consent.

2.2. Clinical assessment, electrophysiology, and quantitative sensory testing

Study subjects underwent standardized clinical examination together with quantitative sensory testing and questionnaires to evaluate pain and psychological conditions (graded chronic pain scale,⁵⁵ neuropathic pain symptom inventory,¹⁰ Beck depression index II,⁶ state-trait anxiety inventory⁵²). In addition, basic blood tests were used to screen all participants for yet undiagnosed exclusion criteria: neurological, endocrinological, metabolic, or infectious disorders, especially inflammation, diabetes, vitamin deficiency, thyroid dysfunction, and kidney or liver failure. Routine electrophysiology was performed for the potentially affected nerves based on distribution of the symptoms. Patients were diagnosed with CRPS type II when the electrophysiology was abnormal and clinical data indicated nerve damage.

Characteristic symptoms like edema, allodynia, and skin temperature were clinically assessed following standardized protocols.^18,48 The disease duration was defined as time between diagnosis and inclusion. In addition, we assessed the duration since the inciting event. One patient had spontaneous CPRS, so the start of pain was taken for the calculation. Quantitative sensory testing followed the protocol established by the German Research Network on Neuropathic Pain and Z-scores of cold detection threshold (CDT), warm detection threshold (WDT), thermal sensory limen (TSL), cold pain threshold (CPT), heat pain threshold (HPT), mechanical detection threshold (MDT), mechanical pain threshold (MPT), mechanical pain sensitivity (MPS), wind up ratio (WUR), vibration detection threshold (VDT), and pressure pain threshold (PPT) were calculated as described using the provided reference values.⁵⁰

2.3. Standardized skin biopsies

Skin punch biopsies (4 mm) were taken from the radial proximal phalanx of an affected finger: digitus II (n = 17) or digitus III-V (n = 3). Care was taken that this covered a symptomatic area. Bilateral biopsies were obtained from patients and unilateral ones from controls. The biopsy site was not stitched. Patients did not report significant adverse events resulting from the biopsy. Each biopsy was divided into 2 pieces. For immunohistochemistry, one half was fixed in 4% phosphate-buffered paraformaldehyde for 30 minutes, washed with phosphate-buffered saline (PBS), incubated in 10% sucrose overnight for cryoprotection, and embedded in Tissue-Tek O.C.T. Compound (Sakura Finetek Europe B.V., Staufen, Germany). Sections measuring 40 µm were cut on a cryostat (Leica CM3050 S; Leica Biosystems, Wetzlar, Germany) and mounted in groups of 3 on SuperFrost-plus microscope slides (ThermoScientific, Waltham, MA). Sections on each slide had at least 80-µm distance to avoid double counting of structures. Slides were stored at −40°C until further processing. The second half was snap frozen in liquid nitrogen for RNA extraction for other studies.

2.4. Immunohistochemistry

Sections were blocked with 10% bovine serum albumin (BSA, Sigma Aldrich, St. Louis, MO, A7906) in PBS at room temperature for 30 minutes. Slides were incubated with primary antibodies diluted 1:100 in PBS with 1% BSA and 0.3% Triton X-100 (Sigma LifeScience, St. Gallen, Switzerland) at 4°C overnight. Details of antibodies are provided in Table 1. The following primary antibodies were used in 3 different stainings: (1) guinea pig-anti-cytokeratin 20 (K20), mouse-anti-PGP9.5, rabbit-anti-claudin-1, and goat-anti-collagen IV; (2) Alexa Fluor 488-conjugated anti-claudin-5, mouse-anti-CD31, and goat-anti-collagen IV; (3) guinea pig-anti-cytokeratin 20, chicken-anti-myelin basic protein, mouse-anti-sodium channel pan, and rabbit-anti-claudin-19.

Table 1.

Antibodies.

Antibody	Target	Company	Antibody ID
Guinea pig anti-cytokeratin 20	Merkel cells	ARP American Research Products Cat# 03-GP-K20	AB_1541036
Chicken anti-myelin basic protein	Myelinated fibres	Millipore Cat# AB9348	AB_11213157
Mouse anti-sodium channel Pan	Sodium channels	Sigma-Aldrich Cat# S8809	AB_477552
Rabbit anti-CLDN19	CLDN19	BiCell, 00219, rabbit
Mouse anti-PGP9.5	Axons	Bio-Rad Cat# 7863-1004	AB_620256
Goat anti-collagen IV	Collagen IV	SouthernBiotech Cat# 1340-01	AB_2721907
Rabbit anti-CLDN1	CLDN1	Thermo Fisher Scientific Cat# 51-9000	AB_2533916
Mouse anti-CD31	Endothelial cells	BD Biosciences Cat# 550389	AB_2252087
Anti-CLDN5 Alexa Fluor 488	CLDN5	Thermo Fisher Scientific Cat# 352588	AB_2532189
Anti-mouse Alexa Fluor 555 nm	Mouse IgG	Thermo Fisher Scientific Cat# A-31570	AB_2536180
Anti-guinea pig Alexa Fluor 405 nm	Guinea pig IgG	Abcam Cat# ab175678	AB_2827755
Goat-anti-chicken Alexa Fluor plus 488 nm	Chicken IgY	Thermo Fisher Scientific Cat# A32931	AB_2762843
Donkey-anti-rabbit Alexa Fluor plus 594 nm	Rrabbit IgG	Thermo Fisher Scientific Cat# A32754	AB_2762827
Donkey-anti-mouse Alexa Fluor plus 647 nm	Mouse IgG	Thermo Fisher Scientific Cat# A32787	AB_2762830
Anti-mouse Alexa Fluor 488 nm	Mouse IgG	Thermo Fisher Scientific Cat# A-21202	AB_141607
Anti-rabbit Alexa Fluor 555 nm	Rabbit IgG	Thermo Fisher Scientific Cat# A-31572	AB_162543
Anti-goat Alexa Fluor 647 nm	Goat IgG	Thermo Fisher Scientific Cat# A-21447	AB_2535864

Suitable secondary antibodies were diluted in 1:800 ratio in PBS with 1% BSA (Table 1) and incubated for 1 hour at room temperature. Exact procedures differed slightly. In staining 1, secondary antibodies were applied in 2 steps to avoid interferences of antibodies obtained from goat and antibodies against goat: slides were first incubated with donkey-anti-mouse Alexa Fluor 488, donkey-anti-rabbit Alexa Fluor 555, and donkey-anti-goat Alexa Fluor 647 for 1 hour; after washing in PBS, slides were incubated with goat-anti-guinea pig Alexa Fluor 405 for 1 hour. In staining 2, slides were incubated with donkey-anti-mouse Alexa Fluor 555 and donkey-anti-goat Alexa Fluor 647 for 1 hour. After washing in PBS, nuclei were stained with Hoechst33342 (Thermo Fischer) diluted 1:1000 in PBS for 10 minutes. In staining 3, slides were incubated with goat-anti-guinea pig Alexa Fluor 405, goat-anti-chicken Alexa Fluor plus 488, donkey-anti-rabbit Alexa Fluor plus 594, and donkey-anti-mouse Alexa Fluor plus 647 for 1 hour. After washing, sections were mounted with Aqua-Poly/Mount (Polysciences, Inc, Warrington, PA).

2.5. Microscopy and analysis

The same investigator blinded to sample allocation conducted tissue staining, microscopy, and analysis; 16-bit images were taken using the same microscope settings for each antibody. For each staining, 15 to 20 samples per group were analyzed. Per sample, 2 to 3 sections were examined. ImageJ macros that were used for semiautomated analysis can be found at https://github.com/katharinamehling/Paper-Skin.git.

To quantify claudin-1 in the epidermis, large-field (tile) microscopy was used to obtain images with a resolution of 4.4 pixels/micrometer of complete skin sections with a 20x/0.8 M27 objective (Axio Imager 2, Zeiss, Jena, Germany). Mean claudin-1 intensity of the epidermal area was measured in keratinocytes—corneal layers and margins were excluded. Average epidermal thickness was calculated by dividing the epidermal area by its length.

To analyze claudin-1 distribution in the epidermis, we developed a semiautomated method: claudin-1 intensity was plotted over a perpendicular line from the basal membrane to the translucent layer. All lines were drawn outside of epidermal papillae. To detect minima and maxima of the plot, the “find maxima …” function in ImageJ was used with its threshold set to 20% of the maximal intensity value or manually adjusted when resulting plots were not accurate. Peak size was measured between a minimum and the adjacent maximum. The first minimum (at the origin of the plot) and the first maximum were excluded from the analysis because they did not provide valid peaks. Three plots were analyzed in each section, and up to 9 plots were averaged for each sample.

For analysis of claudin-1 in the perineurium of dermal nerves, 2 images per sections were obtained with a resolution of 13.9 pixels/micrometer with a 63x/1.40 Oil M27 objective (Axio Imager 2, Zeiss). Within the region of interest, which was determined by the claudin-1 and collagen IV signals, the mean claudin-1 intensity was measured. Per image, 1 to 4 transversely cut nerves were captured, and per sample, 4 to 11 nerves were analyzed.

To assess the number of axons within the nerves, PGP9.5-positive maxima were counted using the “find maxima …” function of ImageJ in the area enclosed by perineurial region of interest. Quantification of axon densities was only performed for nerves with the whole circumference depicted in the image (77% of captured nerves).

K20-positive Merkel cells, Meissner corpuscles with PGP9.5-positive axons arranged as horizontal lamellae, MBP-positive myelinated fibers, and PGP9.5-positive intraepidermal nerve fibers (IENF) were counted using Olympus BX51 microscope or ZEISS Axio Imager 2 with 20x to 63x objectives. We identified PGP9.5-positive axons arranged as horizontal lamellae as Meissner corpuscles. Intraepidermal nerve fibers were counted following the Guidelines of European Federation of Neurological Societies.³⁵ Per sample, we counted 0 to 83 fibers in 1.6- to 8.7-mm epidermis, 0 to 48 Merkel cells in 1.8- to 18.2-mm epidermis of 2 to 6 sections; 0 to 13 Meissner corpuscles in 12 to 111 epidermal papillae; and 19 to 349 myelinated fibers in 2.3 to 37.2 mm² dermis, as described before.²⁸ To assess nodes of Ranvier and paranodal claudin-19, image stacks with a step size of 0.3 µm were acquired using confocal microscopy (Leica TCS SP8). Up to 2 regions with myelinated fibers were imaged in each section. 3D ImageJ Suite was used to measure intensity and volume of 0 to 32 (in average 9.5) nodes and 0 to 64 (in average 19) paranodes with claudin-19 immunoreactivity in 1 to 6 image stacks per sample.⁴⁴

Images of papillary vessels were obtained from 2 to 3 regions of the epidermal–dermal border with a resolution of 13.9 pixels/micrometer using a 63x/1.40 Oil M27 objective (Axio Imager 2, Zeiss). Vessels formed loops in the epidermal papillae and were counted as illustrated in Figure 1: Fully depicted vessels with an ascending and descending limb were counted as 2. Papillary vessels with only 1 limb were counted as 1. Per sample, 7 to 48 blood vessels were counted in 19 to 135 papillae of the epidermis. Mean claudin-5 intensity was measured in regions of interest, which were preselected using a manually adjusted threshold on merged claudin-5 and CD31 signals.

Figure 1.

Quantification of papillary blood microvessels. CD31-immunolabeled vessels ended as loops within the papillae of the epidermis. We created a novel method to quantify papillary vessels. Indentations of the epidermis were counted as papillae (red arrows). If a papilla had several subindentations, each subindentation was counted. The area within the papillae was delineated by a line (dotted line) connecting the base of the regions between the papillae. All vessels were considered for counting regardless of their size or configuration. Vessels with a change of directionality of more than 90°—resulting in 2 limbs—were counted as 2.

2.6. Statistics

Statistical analyses were performed using GraphPad Prism version 9.3.0 for Windows (GraphPad Software, San Diego, CA). Shapiro–Wilk normality test was used to evaluate distribution of data. As most data were not normally distributed, comparison between the 3 groups was performed using nonparametric Kruskal–Wallis test. Pairwise comparisons were P value corrected using Dunn test for multiple comparisons. All specified numbers in the following text indicate means with standard error of mean. Spearman correlation test was used to assess correlation. Differences were considered statistically significant if P < 0.05. QST data were analyzed by calculating z-scores as previously described by the German Research Network on Neuropathic Pain.⁵⁰

3. Results

3.1. Patient cohort

Demographic and basic clinical data are summarized in Table 2. Complex regional pain syndrome patients with affection of the upper extremity were mostly middle-aged women. Most of the patients had CRPS type I with an “initially warm” phenotype. The CRPS severity score was moderate—similar to pain ratings: patients experienced moderate mean pain and severe maximum pain. The pain had a neuropathic character, and half the patients experienced allodynia or mechanical hypersensitivity. In the affected limb, quantitative sensory testing documented loss of function in nonpainful stimuli (CDT, WDT, TSL, MDT, VDT) and gain of function in painful stimuli (MPT, PPT) compared with our healthy control group. The contralateral limb was also affected by the loss of function for WDT and MDT and the gain of function in MPT (Fig. 2).

Table 2.

Patients' characteristics.

	CRPS (n = 20)	Healthy controls (n = 20)
Age [y] (mean ± SD (range))	49 ± 12 (26-70)	51 ± 12 (22-71)
Gender (female %)	16 (80%)	17 (85%)
BMI [kg/m2] (mean ± SD)	27.5 ± 5.6	25.0 ± 4.8
CRPS type
Type I	11 (55%)
Type II	9 (45%)
CRPS initial phenotype
Warm	10 (50%)
Cold	5 (25%)
Unclear	5 (25%)
Disease duration [mo] (mean ± SD (range))
Since diagnosis	3.4 ± 4.7
Since trauma	6.9 ± 5.9
Pain (NRS: numerical rating scale (1-10))
Mean (mean ± SD)	5.4 ± 1.2
Max (mean ± SD)	7.8 ± 1.3
CSS* (mean ± SD)	11.1 ± 2.4
Symptoms
Edema	15 (75%)
Allodynia	10 (50%)
Questionnaires
GCPS† (mean ± SD)	1.5 ± 1.1	0.1 ± 0.2
NPSI‡ (mean ± SD)	38.5 ± 16.1
BDI II§ (mean ± SD)	16 ± 9.7	4.2 ± 4.0
STAI-T‖ (mean ± SD)	44.2 ± 5.9	42.1 ± 3.2

^*CRPS severity score, range 0 to 17.

^†Graded chronic pain scale—disability scores, range 0 to 100.

^‡Neuropathic pain symptom inventory, range 0 to 100.

^§Beck depression inventory II, range 0 to 63.

^‖State-trait anxiety inventory trait anxiety subscale, range 20 to 80.

BMI, body mass index; CRPS, complex regional pain syndrome.

Figure 2.

Some but not all patients with the expected changes, such as loss in nonpainful stimuli, gain in painful stimuli. Quantitative sensory testing profiles from ipsilateral and contralateral hands of CRPS patients and healthy controls. HC (healthy controls) n = 13, IL (ipsilateral) n = 20, CL (contralateral) n = 20. Black lines represent medians. Kruskal–Wallis test with Dunn test for multiple comparisons was used; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. CDT, cold detection threshold; CPT, cold pain threshold; CRPS, complex regional pain syndrome; HPT, heat pain threshold; MDT, mechanical detection threshold; MPS, mechanical pain sensitivity; MPT, mechanical pain threshold; PPT, pressure pain threshold; TSL, thermal sensory limen; VDT, vibration detection threshold; WDT, warm detection threshold; WUR, windup ratio.

3.2. Normal epidermal trophicity and claudin-1 expression in early complex regional pain syndrome

To assess epidermal trophicity and barrier integrity, we analyzed epidermal thickness and claudin-1 immunoreactivity in keratinocytes. Epidermal thickness was not significantly changed in ipsilateral CRPS skin compared with healthy controls (Figs. 3A–C). Mean intensities of claudin-1 immunoreactivity over epidermal cell layers (basal membrane to translucent layer) were similar in patients and healthy controls (Fig. 3D). We assessed the pattern of claudin-1 signal minima and maxima across the epidermis to detect claudin-1 location within each epidermal layer and within keratinocytes (Figs. 3E and F). A relatively flat curve with reduced peak sizes represents claudin-1 internalization; maxima represent cell–cell contacts connected by tight junctions expressing claudin-1, whereas minima reflect the cytoplasm that is normally devoid of barrier proteins. Within the epidermal layer, claudin-1 was expressed predominantly in the granular layer: In average, 8.3 ± 0.34 cell layers contained claudin-1 intensity maxima in healthy controls and 7.3 ± 0.43 in the affected limb of patients. Subcellular claudin-1 distribution in keratinocytes was normal in CRPS (Fig. 3G). In summary, our results did not confirm skin hypo- or hypertrophy in early CRPS and suggest a maintained epidermal barrier.

Figure 3.

Normotrophic epidermis with maintained skin barrier in early CRPS vs healthy controls. Skin biopsies (ipsilateral, contralateral second digit) were collected within 1 year after the diagnosis from patients with CRPS and healthy controls (HC). (A) Representative image of a skin section immunolabeled with collagen IV and claudin-1. (B) Region of interest (ROI) over the vivid epidermal layers and measurement of its length. (C) Average epidermal thickness calculated by dividing the epidermal area by its length. (D) Mean intensity of claudin-1 over the vivid cell layers of the epidermis (corneal layer and margins excluded). (E) Representative image with lines over which intensity was plotted and close-up. (F) Example of an intensity plot with selected minima (blue circles) and maxima (red crosses), and peak size (grey arrow). (G) Results of claudin-1 distribution analysis: data points represent the average mean peak size per sample. HC n = 15, IL n = 16, CL n= 17. Black lines represent medians. Kruskal–Wallis test with Dunn test for multiple comparisons was used. CRPS, complex regional pain syndrome; HC, healthy controls.

3.3. Sensory fibers and receptors: bilateral reduction of Meissner corpuscles in early complex regional pain syndrome

Densities of Meissner corpuscles, Merkel cells, and IENF were analyzed to evaluate alterations of the sensory system. The number of Meissner corpuscles per epidermal papilla was significantly decreased in the affected and contralateral unaffected limbs of CRPS patients compared with healthy controls (Figs. 4A–C). The number of papillae per millimeter of skin tissue was equal in controls and patients (8.5 ± 0.5 papillae/millimeter in healthy controls; 8.4 ± 0.6 papillae/millimeter in ipsilateral CRPS). Six of the CRPS patients (37.5%) did not have any Meissner corpuscles in all analyzed sections of the ipsilateral sample. Four of them (25.0%) were lacking corpuscles bilaterally, whereas only 1 healthy control (6.7%) had no Meissner corpuscles. Meissner corpuscle densities were independent of disease severity (CSS), disease duration, CRPS type, maximal pain, and allodynia in the clinical examination or vibration detection thresholds, mechanical detection thresholds, mechanical pain thresholds, and pressure pain threshold measured by QST (Supplementary Table 1, available at http://links.lww.com/PAIN/B989).

Figure 4.

Bilateral reduction of Meissner corpuscles in the skin of CRPS patients. Skin biopsies were immunolabeled with K20 for Merkel cells, PGP9.5 for axons, and collagen IV as basal membrane marker. (A) Meissner corpuscles (arrows) in healthy control compared with ipsilateral and contralateral CRPS skin. (B) Meissner corpuscles of each group. (C) Quantification of Meissner corpuscles relative to epidermal papillae. (D). Representative image of Merkel cells (arrows) in the epidermal basal cell layer. (E) Quantification of Merkel cells density. (F) Representative image of an intraepidermal nerve fiber passing the dermal–epidermal border. (G) Quantification of intraepidermal nerve fiber density (IENFD). (A–C, F, G): HC (healthy controls) n = 15, IL (ipsilateral) n = 16, CL (contralateral) n = 17; (D and E): HC n = 18, IL n = 19, CL n = 19. Black lines represent medians. Kruskal–Wallis test with Dunn test for multiple comparisons was used; *P < 0.05. CRPS, complex regional pain syndrome.

Merkel cells were observed as PGP9.5- and K20-positive oval-shaped structures in the basal layer of the epidermis. Densities were not significantly different between patients and healthy controls and did not correlate with pain or touch sensation measured in QST (Figs. 4D and E; Supplementary Table 1, available at http://links.lww.com/PAIN/B989). Our cohort of early CPRS patients had normal intraepidermal nerve fiber densities (Figs. 4F and G). Fiber densities did not correlate with QST measurements (Supplementary Table 1, available at http://links.lww.com/PAIN/B989). In summary, patients with early CRPS showed normal innervation apart from the bilaterally reduced Meissner corpuscles.

3.4. Unchanged dermal myelinated fiber densities and intact perineurial and myelin barrier

We investigated dermal nerves including their perineurial and myelin barrier to evaluate impairment of the somatosensory system in more detail. We did not observe a loss of PGP9.5-positive axons in dermal nerves in patients compared with healthy controls (Figs. 5A–C). Axon counts per nerve of healthy controls correlated with IENFD. Dermal nerves were immunoreactive for claudin-1 between layers of collagen IV–positive tissue forming a perineurium-like structure (Figs. 5D and E). Mean claudin-1 intensity over the perineurial area did not differ between healthy controls and CRPS patients (Fig. 5F).

Figure 5.

Intact perineurial and myelin barriers in CRPS skin. (A and B) Representative images of PGP9.5-positive axons (white crosses) within perineurium-like structure stained with collagen IV and claudin-1. The yellow line indicates the analyzed region of interest (ROI). (C) Axon density of all nerves within the analyzed image was averaged for each sample. (D and E) Representative ROI, determined by collagen IV staining, for intensity measurement of perineurial claudin-1. (F) Mean claudin-1 intensity was measured in the perineurium and averaged for each sample. (G) Representative image of myelinated fibers (MBP) with a close-up of 1 node of Ranvier (pan Na_V). Claudin-19 was located in the paranodal regions and in the Schmidt–Lanterman incisures (SLI). (H) Quantification of myelinated fiber relative to analyzed dermal area. (I) Mean paranodal claudin-19 intensities of patients and controls measured in a 3-dimensional ROI. (A–F): HC (healthy controls) n = 15, IL (ipsilateral) n = 16, CL (contralateral) n = 17; (G–I): HC n = 19, IL n = 20, CL n = 20. Black lines represent medians. Kruskal–Wallis test with Dunn test for multiple comparisons was used.

The densities of dermal myelinated fibers were similar in healthy controls and CPRS (Fig. 5H). Complex regional pain syndrome type II patients had slightly but not significantly lower myelinated fiber densities than CRPS type I patients (data not shown).

Claudin-19 seals the myelin barrier in the paranodal regions and Schmidt–Lanterman incisures. We observed claudin-19 immunoreactivity predominantly in paranodal regions adjacent to panNa_V-positive nodes of Ranvier and in Schmidt–Lanterman incisures (Fig. 5G). Nodes of patients and controls had similar intensity and volume of panNa_V immunoreactivity (data not shown). The intensity of paranodal claudin-19 immunoreactivity was also unaltered in patients (Fig. 5I). Overall, dermal nerves of CRPS patients showed intact neural barriers as assessed by tight junction proteins in the perineurium and the myelin barrier.

3.5. Bilateral loss of papillary microvessels in complex regional pain syndrome

We quantified blood vessels in the papillae of the epidermis and claudin-5 expression to evaluate vascular barrier integrity in CRPS. CD31-positive blood vessels forming loops in the epidermal papillae were counted in a standardized way (Fig. 1). Early CRPS patients showed a bilateral loss of microvessels compared with controls (Figs. 6A and B). The density of microvessels did neither correlate with the temperature of the affected extremity nor with the difference to the contralateral one (Supplementary Table 1, available at http://links.lww.com/PAIN/B989). In a preclinical model for chronic post ischemic pain, ischemia correlated with allodynia.³⁴ In CRPS patients with allodynia, the loss of microvessels was more pronounced than in patients without allodynia, but the difference was no longer significant after correction for multiple comparisons (Fig. 6C).

Figure 6.

Bilaterally decreased papillary blood vessels in the skin of CRPS patients. (A) Representative images of CD31-stained papillary vessels in skin of healthy controls and from the ipsilateral and contralateral side of patients. (B) Quantification of papillary vessel densities in skin. (C) Subgroup analysis of papillary vessel densities in ipsilateral biopsies of CRPS patients with allodynia, without allodynia, and healthy controls. (D) Examples of region of interest (ROI) for claudin-5 intensity measurement of a healthy control and ipsilateral CRPS sample. (E) Mean claudin-5 intensity over all ROIs within the analyzed image was averaged for each sample. (F) Subgroup analysis of claudin-5 intensity in ipsilateral biopsies of patients with edema, without edema, and healthy controls (HC). HC n = 17, IL (ipsilateral) n = 20, CL (contralateral) n = 19. Black lines represent medians. The arrows mark blood vessels. The dotted line marks the basal membrane. Kruskal–Wallis test with Dunn test for multiple comparisons; *P < 0.05. CRPS, complex regional pain syndrome.

Immunoreactivity of claudin-5—as the main tight junction protein of endothelial cells in papillary vessels—was not significantly altered in CRPS patients (Figs. 6D and E). To see if claudin-5 immunoreactivity is altered in patients specifically with edema, we performed subgroup analysis. But we could only observe a trend of increased claudin-5 immunoreactivity in patients with edema (Fig. 6F) or warm CRPS (data not shown). We also quantified the immunoreactivity of the endothelial marker CD31 as reference but found no difference (data not shown). Claudin-5 intensity did not correlate with vascular densities (Supplementary Table 1, available at http://links.lww.com/PAIN/B989). Although microvessel densities were bilaterally decreased, claudin-5 appeared unaltered in CRPS patients.

4. Discussion

In this study, we investigated the skin, including sensory receptors and barriers formed by tight junction proteins in bilateral biopsies, from patients with early CRPS and healthy controls. Meissner corpuscles and papillary blood vessels located in dermal papillae were bilaterally reduced in CRPS patients (Fig. 7).

Figure 7.

Schematic illustration of structural changes in bilateral skin of acute CRPS patients: Meissner corpuscles and papillary vessels are bilaterally decreased in CRPS patients. Other parameters assessed, such as epidermal thickness, epidermal barrier, or perineurial and myelin barrier, were not affected. Created with BioRender.com. CRPS, complex regional pain syndrome.

Merkel cells and Meissner corpuscles are responsible for the detection of low-frequency vibration and touch sensation, respectively. Beside their tactile qualities, Meissner corpuscles express classical nociceptive molecules like substance P or calcitonin gene–related peptide. Miswiring of Meissner corpuscles after nerve injury in preclinical models by sprouting nociceptors could explain tactile allodynia caused by light touch.²² Multiple studies described reduced Meissner corpuscles in other pain conditions, including diabetic, HIV, and chemotherapy-induced neuropathy,^{11,16,23,42,45} but only one study investigated Meissner corpuscles in CRPS.² The authors showed disorganized Meissner corpuscles with disruption of Aβ-innervation in skin samples from amputated extremities of 2 patients. Reduction of Meissner corpuscles could be a sign of neuropathy and result in the loss of fine tactile function. However, Meissner corpuscle density did not correlate with tactile function evaluated by QST in our cohort. This might be explained by the slightly different location of the investigated region because QST was measured on the back of the hand and at the base of the thumb, depending on the QST parameter. In other studies, a loss of Meissner corpuscles correlated with mechanical, vibration, or thermal thresholds in QST,^16,45 but there is no clear assignment of Meissner corpuscles to a specific tactile quality.

In our cohort, we could not confirm the loss of intraepidermal nerve fibers that was described in chronic CRPS.^43,46 This might be explained by the rather short disease duration. Small-fiber neuropathy might be a feature of chronic CRPS.

Papillary blood vessels were also bilaterally reduced. Literature on vascular abnormalities in CRPS is vast. A common hypothesis is that a combination of impaired sympathetic reactions and endothelial dysfunction causes vasomotor dysfunction³³: In the acute phase, a functional reduction of the sympathetic reaction with low catecholamine levels leads to relative vasodilatation and thus “warm” CRPS. In the long run, compensational upregulation of adrenoreceptors results in overactive vessels, vasoconstriction, and, finally, “cold” CRPS.^20,56 Acetylcholine-induced vasodilatation is reduced in CRPS, but an impairment of microvascular vasodilatation was not detected using laser Doppler imaging.^24,51 Dysfunctional endothelial cells produce an imbalance of endothelin-1 and nitric oxide, which is required for vasodilatation.²⁵ In skin obtained from amputated CRPS limbs, researchers observed hypertrophic blood vessels with thickened basal membranes and disrupted vascular innervation.^2,54 To our knowledge, quantification of blood vessels has not been done in CRPS skin before. Our patients with allodynia had a more pronounced loss of papillary blood vessels. In the chronic postischemia pain animal model, induction of ischemia and reperfusion produces a CRPS-like phenotype with hyperemia, edema, mechanical hyperalgesia, and allodynia. Here, microvascular dysfunction leads to a persistent inflammatory state and increased lactate levels in correlation with allodynia.^15,34 This indicates that microvascular features in the skin might contribute to allodynia. Reduced blood vessels cause hypoxia and thereby acidosis in skin,⁹ which might result in allodynia and pain if the tissue is damaged.

All in all, we did not find any differences in barrier proteins. The epidermal barrier (claudin-1), vascular (claudin-5), and the myelin barrier (claudin-19) were intact. It is possible that these changes occur rather in chronic CRPS, where trophic changes might be more pronounced. Alternatively, only subgroups could be affected. In our previous study, exosomal microRNA-183 was decreased in the serum of CRPS patients. In vitro, microRNA-183 restored microvascular barrier through recovery of claudin-5.⁴⁸ Thus, we expected that the lack of micro-RNA183 leads to decreased claudin-5 levels in CRPS clinically evident as edema. In the literature, most data also support this hypothesis: downregulated claudin-5 leads to leakiness of blood–nerve barrier in human sural nerve biopsies of chronic inflammatory demyelinating polyneuropathy and in various pain animal models.^12,30,40 However, our data did not reveal significant changes in claudin-5 immunoreactivity, which might be due to the small sample size.

Interestingly, we found bilateral losses of Meissner corpuscles and papillary vessels in CRPS affected and unaffected skin. But there were no correlations with disease duration. In the literature, contralateral hyperalgesia or symptoms have been reported and contralateral morphological or immunological changes are more common than expected.^{19–22,36,41,46} Most studies that found side differences, however, lacked a healthy control group for comparison.^8,25,29,57 Only some studies found differences only in the affected but not in the unaffected limb.^43,51 In preclinical nerve injury models, mirror-image allodynia has also been described by some groups: Mechanisms involve immune reactions in the central nervous system, persistent activation of spinal astrocytes and microglia, and neuronal changes in spinal or supraspinal structures.^4,14,37 Del Valle et al.¹⁷ could confirm such activation of microglia and astrocytes in the spinal cord of a patient with long-standing CRPS postmortem. Activation was predominant on the affected side, but it also reached the contralateral spinal cord. Central mechanisms might lead to contralateral alterations, such as the decrease of Meissner corpuscles and blood vessels, but seem premature in acute CRPS.

Considering the disease duration of our cohort, a preexisting status of low vascular supply or Meissner corpuscle densities might contribute to CRPS pathology. Under physiological conditions, fewer microvessels might still provide a sufficient cutaneous supply under normal condition. However, a “second hit” by trauma might overburden the delicate balance of the vascular system because of higher requirements during wound healing. Lower vessel counts might reflect insufficiency in angiogenesis and vasculogenesis—central mechanisms of wound healing. Finally, insufficient oxygen supply results in hypoxia and acidosis. Both might induce pain. Pathologic histological conditions as predisposition would also explain the lack of correlation between morphological findings and clinical symptoms.

Our study is limited by the number of samples. We only included patients with newly diagnosed CRPS. Thus, subgroup analyses were difficult. Complex regional pain syndrome has a great diversity of signs and symptoms and probably includes multiple different disease entities, which makes it difficult to identify pathophysiological pathways and mechanism. In addition, skin is a very heterogenous biomaterial. We took meticulous care to use exactly the same location of transition from glabrous to hairy skin. Still, sections contain multiple different structures cut in various directions. Our approach to address these difficulties was to focus our analysis on the dermal–epidermal junction, which provides consistent directionality. Therefore, further studies with larger patient cohorts are needed to confirm our results and put them in context. Conversely, we found significant differences even in our small cohort, which implicates a high statistical power of our results, and thus, clinical relevance is likely.

In conclusion, our study is the first to show a bilateral loss in papillary vessels and Meissner corpuscles in acute CRPS skin. We hypothesized that our findings might display preexisting tissue alterations and reflect a predisposition for CRPS development after trauma.

Conflict of interest statement

The University Hospital Wurzburg and H.L.R. received patient fees from Algiax for clinical trials and consultation fess from Gruenthal and Oreon. The authors declare no conflicts of interest.

Appendix A. Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/B989.