Abstract
Objective: To determine whether the cytokine tumor necrosis factor α (TNF-α) acts as a pain mediator in neuropathic pain in humans.
Background: In animal models, inflammatory cytokines such as TNF-α have been shown to facilitate neuropathic pain.
Methods: The expression of TNF-α was analyzed immunohistochemically in 20 human nerve biopsy specimens of patients with painful (n = 10) and nonpainful (n = 10) neuropathies. Additionally, serum soluble TNF-α receptor I (sTNF-RI) levels were determined in 24 patients with neuropathies, 16 of which were painful and 8 that were painless.
Results: Colocalization studies by confocal fluorescence microscopy for S-100 and TNF-α showed expression of TNF-α in human Schwann cells. Patients with painful neuropathies showed a stronger TNF-α immunoreactivity in myelinating Schwann cells relative to the epineurial background staining compared with patients with nonpainful neuropathy (0.949 ± 0.047 vs 1.010 ± 0.053, p < 0.05). Although there was no difference in sTNF-RI levels between painful (n = 16) and nonpainful (n = 8) neuropathies (sTNF-RI: 1412 ± 545 pg/mL vs 1,318 ± 175 pg/mL), patients with a mechanical allodynia (n = 9) had elevated serum sTNF-RI (1627 ± 645 pg/mL vs 1233 ± 192 pg/mL, p < 0.05) compared with patients without allodynia (n = 15).
Conclusions: TNF-α expression of human Schwann cells may be up-regulated in painful neuropathies. The elevation of sTNF-RI in patients with centrally mediated mechanical allodynia suggests that systemic sTNF-RI levels may influence central pain processing mechanisms.
Regardless of their cause, neuropathies can be divided into painful and nonpainful forms. Despite the remarkable progress that has been made in identifying pain mechanisms in animal models, it is still unclear why some neuropathies are painful in certain patients but painless in others.
Because clinical experience shows that inflammatory neuropathies are generally painful, and nonsteroidal anti-inflammatory drugs (NSAID) are only partly effective in treating neuropathic pain,1 attention has been focused on inflammatory pain-mediating mediators other than prostaglandins. Cytokines such as tumor necrosis factor α (TNF-α) and interleukin-6 have been recognized to be important factors in the mediation of neuropathic pain. TNF-α, for example, induces ectopic activity in nociceptive primary afferent neurons2 and elicits hyperalgesia, peripherally3,4⇓ or centrally.5 TNF-receptor I (TNF-RI) antibodies6 and antibodies against the interleukin-6 receptor7 are able to reduce neuropathic pain. Furthermore, the expression of TNF-α is up-regulated in the spinal cord in a painful mononeuropathy model.8
These findings suggest that TNF-α may be involved in peripheral and central pain-mediating mechanisms. Hence, immunohistochemical analysis of peripheral nerve biopsy specimens could determine whether TNF-α also contributes to peripheral neuropathic pain generation. Altered serum cytokine or their soluble receptor levels also could indicate whether systemic immune changes influence neuropathic pain perception.
Methods.
Patients with sural nerve biopsies.
An immunohistochemical analysis was performed to determine TNF-α expression in peripheral nerve biopsy specimens of 20 patients collected between 1993 and 1999 at the Department of Neurology, University Hospital, Basel, Switzerland. Ten patients had painful and 10 nonpainful neuropathy ( table 1). Nine of 10 patients in the painful group had an inflammatory origin of the neuropathy, whereas this was the case in only four of 10 patients of the nonpainful group. Patients for whom the pain state at the time of the biopsy could not be determined retrospectively were not included. Clinical data of the patients are given in table 1. Biopsy specimens were taken of either the sural nerve or a superficial sensory branch of the peroneal nerve.
Clinical data of patients undergoing a nerve biopsy
Prospectively examined patients with serum soluble TNF-RI (sTNF-RI) analysis.
Twenty-four patients (nine women, 15 men; mean age, 59.1 ± 13.4 years) with neuropathies were studied prospectively ( table 2), of whom five also had undergone a biopsy. Patients were recruited consecutively at the Department of Neurology at the University Hospital Basel, were examined clinically by the same observer (M.E.), and were also examined for pathologic pain conditions such as allodynia, and scored with the Dyck Neurological Disability Score.9 Patients were considered to have an allodynia if the observer’s gentle tactile stimulation of the extremities caused pain. Patients were asked whether they had actual neuropathic pain, and those patients who had painful neuropathy also answered the German version of the McGill pain questionnaire.10,11⇓ The difference between pain and only unpleasant, but not painful, paresthesia was explained to every patient. Exclusion criteria included ongoing infection (documented with a blood cell count and determination of the C-reactive protein) and pain of other origin. Sixteen patients had painful neuropathy (mean age, 54.6 ± 13.4 years); the neuropathy of six (38%) had an inflammatory origin. Eight patients had nonpainful neuropathy (mean age, 68.1 ± 8.1 years); the neuropathy of one (13%) had an inflammatory origin. Nine patients with painful neuropathy also showed allodynia, whereas none of the patients without pain did.
Clinical data of patients with serum soluble tumor necrosis factor receptor I analysis
Immunohistochemistry.
Frozen transversal nerve sections (8 μm) were mounted on gelatin-coated slides and fixed for 1 hour in 10% formalin. Unspecific binding was blocked by incubation with serum according to the protocol for peroxidase immunohistochemistry (ABC Elite Kit, Vectastain, Burlingame, CA) or with diluted (1:20) normal goat serum for immunofluorescence for 15 minutes. For peroxidase staining, endogenous peroxidase was inhibited before blocking unspecific binding. The monoclonal mouse anti–human TNF-α antibody (clone JID9, donated by Dr. Jez McLaughlin, Liverpool, UK) was diluted 1:200 in phosphate buffer, for fluorescence staining 1:50 in phosphate buffer, and incubated overnight at 4 °C. Other primary antibodies (S-100, HAM-56, and CD68: all DAKO, Glostrup, Denmark; NF 160: Affinity Research Products Ltd., UK) were incubated 1 hour at room temperature for both staining methods. The avidin–biotin–peroxidase complex method (ABC Elite Kit, Vectastain, Burlingame, CA) was used as second antibody system for peroxidase staining. As an enzyme substrate, AEC (3-amino-9-ethyl carbazole) was used, which results in a red reaction product. For immunofluorescence staining, second antibodies with either fluorochrome Cy-3 (Jackson Labs, West Grove, PA) or Cy-2 (Jackson Labs, West Grove) or fluorescein isothiocyanate (FITC; Sigma, St. Louis, MO) were used. Slides were mounted in Kaisers glycerin gelatin (Merck, Darmstadt, Germany) or Fluorosave (Calbiochem, Bad Soden, Germany) for fluorescence staining.
Competition assay.
To ensure that the antibody recognizes specifically human TNF-α, a competition immunohistochemistry assay was performed. An approximately 10-fold molar excess amount of TNF-protein (donated by Dr. Jez McLaughlin, Liverpool, UK) was bound to solid-phase Sepharose (NHS-activated Sepharose 4 fast flow, Pharmacia Biotech, Uppsala, Sweden), and the diluted antibody in the usual concentration was incubated overnight with the Sepharose-bound protein. The supernatant used for the competition immunohistochemistry resulted in no staining ( figure 1A).
Figure 1. Peroxidase immunohistochemistry for tumor necrosis factor α (TNF-α) in human peripheral nerve biopsies showed TNF-α expression in the perineurium, blood vessels, and Schwann cells. (A) A competition assay with TNF-α antibody and TNF-α protein resulted in no staining. (B) Overview with stained perineurium (arrowhead), stained vessel, and endoneurially stained cells. (C and D) Typical sickle-shaped morphology of stained myelinating Schwann cells (arrow). The arrowheads in A, B, and C mark the perineurium; the asterisks in C and D mark the place of an axon located in the center of a myelin sheet.
Image analysis and quantification of the peroxidase staining.
All image analyses were performed by one blinded observer (M.E.) with a qualitative and quantitative evaluation. For the quantitative evaluation, we used a “density of red” measurement, in the following simply called “intensity.” The intensity of the different stained structures (i.e., perineurium, vessels) was measured relative to the respective background intensity of the epineurium. Thus, the results of the intensity measurements represent a ratio of two intensities. This analysis was performed in the awareness that an intensity measurement by peroxidase immunohistochemistry does not fulfill the criteria of constant measurement conditions that are required for linear quantification.
Color images were taken with a Leitz Dialux microscope equipped with a digital camera (Panasonic Switzerland, WV-CD130L/G, Littau-Luzern, Switzerland) and KS-100 image analysis software (Kontron Electronics, Eching, Germany) under constant conditions with a 1:40 lens. Three images were made of three fascicles with three intensity measurements for each structural compartment of interest. If there were more than three fascicles, those to be measured were chosen in a fixed manner in the form of an isosceles triangle pointing to the bottom. For endoneurial measurements, three images were chosen in the same way. With the help of the image analysis software vessels, perineurium, epineurium, or Schwann cells were marked interactively. Only cells that had the typical sickle-shaped form of myelinating Schwann cells around a myelinated axon were marked as Schwann cells (see figure 1, C and D). In three biopsy specimens of the painful group and three of the nonpainful group, the staining of the Schwann cells was too little to be quantified. The mean intensity in red (i.e., mean density in red), ranging from 0 (highest intensity) to 255 (least intensity), was measured. For each biopsy, the mean of the intensity measurements of perineurium, epineurium, vessels, and Schwann cells was calculated, and the ratio of the mean intensity of the structure of interest and of the mean intensity of the epineurium was used for statistical analysis, as mentioned previously.
All biopsy specimens had been subjected to routine diagnostic procedures. The routine evaluation for the degree of infiltrating epineurial and endoneurial lymphocytes (marked with anti-CD3 antibody) or macrophages (marked with anti-CD68 antibody) was performed by the same observer (B.E.) and rated in categories of weak/strong. A Zeiss Axiovert 100 M LSM510 confocal microscope (Carl Zeiss AG Switzerland, Feldback, Switzerland) was used for fluorescence staining analysis.
ELISA.
Blood for cytokine analysis was drawn in the morning, stored at 4 °C for 30 minutes, centrifuged, and the serum stored at −20 °C until assayed. The level of the sTNF-RI (a more stable serum parameter than TNF-α itself) was determined with a commercially available ELISA Kit (R&D Systems, Wiesbaden, Germany).
Student’s t-test was used for statistic analysis of density measurements or cytokine levels, Mann–Whitney U test for non-normally distributed values, and the Fisher’s exact χ2 test for dichotomic values.
Results.
Qualitative evaluation of the TNF-α immunohistochemistry.
TNF-α staining was observed in different intensities in the perineurium, (epineurial and endoneurial) blood vessels, in single cells that morphologically suggested lymphocytes or macrophages, and endoneurially in cells that appeared as myelinating Schwann cells (see figure 1B for an overview; stained myelinating Schwann cells are shown in figure 1, C and D). Approximately 5% of the Schwann cells showed TNF-α immunoreactivity. With double-immunofluorescent analysis, a colocalization of TNF-α and S-100 was demonstrated ( figure 2). Colocalization of macrophages and TNF-α also was observed, but the immunoreactivity was weaker and restricted to a few cells (data not shown).
Figure 2. Double fluorescence staining showed a clear colocalization (yellow) for tumor necrosis factor α (TNF-α) (red) and S-100 (green) in a human peripheral nerve biopsy. The green immunoreactivity for S-100 shows the typical sickle-shaped morphology of a myelinating Schwann cell (arrow) with a central accumulation of TNF-α immunoreactivity, which appears yellow as it is colocalized. The asterisk marks the place where the axon is presumably located in the center of the nonstained myelin sheet.
Quantification of the TNF-α-immunoreactivity.
Because the staining intensity of corresponding structures differed between patients, a quantification of the staining intensity was performed to evaluate a possible correlation to the pain state. A stronger TNF staining was found in myelinating Schwann cells relative to the epineurial background staining of patients with painful neuropathy than in those with nonpainful neuropathy (mean intensity in red of Schwann cells/mean intensity in red of epineurium: 0.949 ± 0.047 vs 1.010 ± 0.053, p < 0.05; figure 3).
Figure 3. The mean density in red of myelinated Schwann cells relative to the background mean density in red of the epineurium of biopsy specimens of the painful group is lower (p < 0.05), and thus more intense than that of the nonpainful group. The upper and the lower bars indicate the group SD; the middle bar represents the mean of the respective group. The low variance is explained by using a ratio of two values. Data for Schwann cells were available for seven patients from each group.
The other structures measured showed no difference in staining intensity between painful and nonpainful neuropathies (for perineurial staining: 0.980 ± 0.165 vs 0.969 ± 0.185; NS; for vessel staining: 0.934 ± 0.056 vs 0.953 ± 0.060; NS) The low variance is explained by using a ratio of two values.
Other parameters.
Furthermore, we assessed whether infiltrating immune cells or the duration of neuropathy had an influence on the pain state. There was no relation between the degree of infiltration endoneurially or epineurially of lymphocytes or macrophages determined in routine diagnostics and the pain state (Fisher’s exact χ for endoneurial lymphocytes: p = 0.24; for epineurial lymphocytes: p = 0.53; for endoneurial macrophages: p = 0.30; for epineurial macrophages: p = 0.62). However, patients with painful neuropathy had a shorter duration of illness (8.2 ± 11.9 months, n = 9; vs 65.6 ± 61.9 months, n = 9; Mann–Whitney U test: p < 0.05).
Prospectively examined patients with serum sTNF-RI analysis.
In the group of prospectively examined patients with blood sample collection, those with painful neuropathy were younger (54.6 ± 13.4 vs 68.1 ± 8.1 years, p < 0.01) and had a shorter duration of illness (21.4 ± 25.7 months vs 61.6 ± 68.2 months; p < 0.05). Furthermore, the patients in the painful group were less disabled by their neuropathy than the patients in the nonpainful group, although the difference in the Dyck score failed to reach significance (8.9 ± 7.8 vs 14.8 ± 8.8; p = 0.08 for the right and 8.6 ± 6.1 vs 14.6 ± 9.4; p = 0.06 for the left side). The mean McGill score for patients with neuropathic pain was 25.5 ± 7.9.
There was no significant difference in sTNF-RI levels (1,412 ± 545 pg/mL vs 1,318 ± 175 pg/mL) between patients with painful (n = 16) and nonpainful (n = 8) neuropathies, although patients with painful neuropathy tended to have higher values. However, patients with mechanical allodynia (n = 9) had elevated serum sTNF-RI (1,627 ± 645 pg/mL vs 1,233 ± 192 pg/mL, p < 0.05; figure 4) compared with patients without allodynia (n = 15).
Figure 4. The mean soluble tumor necrosis factor receptor I levels of patients with allodynia are higher (p < 0.05) than those of the patients without allodynia. The upper and the lower bars indicate the group SD; the middle bar represents the mean of the respective group.
Discussion.
The immunohistochemical analysis in our study showed two major findings. First, we showed that in human nerve biopsies from patients with peripheral neuropathy, TNF-α is expressed in the perineurium, in epineurial and endoneurial blood vessels, in macrophages, and in Schwann cells. These results are in agreement with a previous study of cytokine expression in human nerve biopsy specimens,12 which described a strong messenger RNA expression of TNF-α in the perineurium, blood vessels, and endoneurial cells, suggestive of Schwann cells. In contrast, another immunohistochemical analysis of human nerve biopsy specimens could not detect TNF-α immunoreactivity in endoneurial cells appearing as Schwann cells, but localized the most intense staining in phagocytosing macrophages.13
Although some of the macrophages adhering to myelin may have been considered as Schwann cells by the peroxidase staining procedure, we verified the specificity of TNF-α expression in human Schwann cells by confocal microscopy, in which TNF-α and S-100 expression were clearly colocalized. Furthermore, neither in the peroxidase staining nor in the fluorescence staining did macrophage markers such as HAM-56 or CD68 show a typical sickle-shaped morphology. Thus, it seems unlikely that Schwann cells were confounded with macrophages.
In our study, the expression of TNF-α in macrophages was generally weaker than that reported in other studies.12,13⇓ This could be attributable to the anti-HAM 56 antibody, which was used for double-fluorescence staining, although the staining pattern was similar to that with anti-CD68 antibody (not shown).
The second major finding was that the intensity of TNF-α staining in Schwann cells was enhanced in patients with painful neuropathy. These data are consistent with the possibility that TNF-α contributes to neuropathic pain generation as it is described in animal experiments, in which TNF-α was shown to play a role in the development of neuropathic pain and hyperalgesia.4,14,15⇓⇓ Interestingly, the occurrence of circulation-derived immune cells, which also release TNF-α, did not influence the pain state. Nevertheless, it is possible that the predominance of inflammatory causes of the neuropathies in the painful group (90 vs 40%) might contribute to the stronger expression of TNF-α rather than the latter is related to the pain state itself. But because inflammation and painful neuropathies are associated in clinical terms, a clear-cut distinction between the proinflammatory and the potential algesic qualities of TNF-α cannot be assessed in a clinical retrospective study. In view of the recent report that myelinated fibers display ectopic activity in nerve injury,16 it seems at least possible that myelinating Schwann cells contribute to the generation of neuropathic pain (i.e., due to release of potential algesic mediators such as TNF-α). The TNF-α expression of other structures was not related to the occurrence of pain in our study.
Because demyelinating and axonal neuropathies had an approximately equal distribution between the two groups, it is not likely that they had a confounding influence. Although no specific morphologic correlate of neuropathic pain has been found yet,17,18⇓ other factors such as the presence of axonal degeneration or regeneration in the biopsy have been suggested to be associated with neuropathic pain.18 It seems improbable that these two parameters caused a bias in our study (see table 1).
Patients with painful neuropathy had a shorter duration of illness, probably because they seek medical attention earlier. The fact that they have a biopsy earlier in the course of the disease could also contribute to an enhanced TNF-α expression, because a stronger TNF-α expression also has been reported to occur in animal experiments in the initial stages of nerve injury.19 Because these initial stages constitute a period of hours, and baseline levels are returned to on day 15 in the animal model, it cannot be excluded that the stronger expression is due to biopsy earlier in the course of the disease in terms of months; however, this seems of minor importance.
With regard to systemic changes, sTNF-RI serum levels did not differ between patients with painful or nonpainful neuropathies. Thus, neuropathic pain and soluble receptor serum levels (i.e., a systemic immune activation) do not seem to be directly related to each other. However, we observed a systemic elevation of sTNF-RI in patients with mechanical allodynia. Mechanical allodynia is a centrally mediated neuropathic pain that occurs when the central perception of a normally innocuous A beta-fiber input (i.e., touch) is altered.20 Because circulation-derived cytokines are known to enter the brain,21 and central administration of TNF-α facilitates hyperalgesia,5 it seems possible that elevated systemic sTNF-RI levels might influence the facilitation of allodynia. This hypothesis is based on the premise that sTNF-RI approximately reflect TNF-α serum activity, because sTNF-RI and TNF-α are released by the same stimuli.22 In view of the small differences and the methodologic limitations for the quantifying of the immunohistochemistry, these results have to be considered preliminary until corroborated by further studies.
Acknowledgments
M.E. was financed by an ENS-fellowship.
Acknowledgment
The authors thank Dr. A. Probst, Institute for Neuropathology, University of Basel, Switzerland, for the routine histologic examinations, and Mrs. J. Benson for carefully reading the manuscript.
Footnotes
-
M.E. is currently affiliated with the Department of Neurology, Ludwig-Maximilians University, Munich, Germany.
- Received September 12, 2000.
- Accepted January 30, 2001.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.