An unexplained multisymptom illness, Gulf War veterans' illness (GWVI), has been described among allied force veterans of the first Gulf War (1990–1991). It has been proposed that some of its symptoms reflect an immune dysfunction, and rheumatologic symptoms including joint pain and stiffness are reported frequently. However, it is unknown whether synovial inflammation causes the articular symptoms. We examined synovial tissue from individuals with GWVI and joint pain for evidence of inflammation.
We compared synovial biopsy samples from 6 individuals with GWVI and joint pain with samples from 9 clinically asymptomatic controls (hematoxylin and eosin [H&E] stains only) and biopsy samples or surgically obtained specimens from 10 patients with rheumatoid arthritis (RA) and 12 with osteoarthritis (OA). Inflammatory changes were quantified in H&E stained sections with a modified synovitis score by immunostaining for CD3, CD20, CD38, CD68, Ki-67, and von Willebrand factor, and with a composite inflammation score based on these markers.
Normal histology was seen in the GWVI specimens, except for mild focal lining hyperplasia and rare low-grade perivascular infiltrates in 1 specimen each. Mean ± SEM synovitis scores were lowest and nearly identical in control (1.38 ± 0.30) and GWVI specimens (1.41 ± 0.29), intermediate in OA specimens (2.64 ± 0.39), and highest in RA specimens (6.0 ± 0.19). Likewise, inflammatory cells, cell division, vascular density, and composite inflammation score were lowest in the GWVI specimens.
Despite significant joint pain, the GWVI synovia did not differ from normal controls. These results agree with other studies that have failed to document inflammatory or immunologic etiologies in GWVI.
A complex of unexplained symptoms including fatigue, memory impairment, headache, respiratory symptoms, skin changes, and musculoskeletal pain and stiffness has been described among personnel returning from the first Persian Gulf War (August 1990 to June 1991) and has been termed Gulf War syndrome or Gulf War veterans' illness (GWVI) (1). Musculoskeletal symptoms have been among the most frequent symptoms, and in one study rheumatologic evaluation was the most frequent subspecialty consultation among 250 patients attending a clinical evaluation program for Gulf War veterans (2). That study failed to identify any single rheumatologic diagnosis specifically associated with GWVI (2). Another study evaluated rheumatologic findings among more than 900 Gulf War veterans evaluated in a screening clinic for unexplained symptoms (3). Musculoskeletal symptoms occurred in 16% of subjects, nearly all of whom experienced pain, most often in the knees, followed by the lower back and shoulders (3). Some studies have called into question whether a GWVI per se exists (4) and have suggested that the rheumatologic symptoms reported by Gulf War veterans resemble those of veterans of other wars (2) and that some symptom constellations occur more frequently among deployed Gulf War veterans but are also found in nondeployed personnel (5, 6).
Synovial inflammation may be a cause of joint pain. However, it is not known whether inflammation or other potential pathologic changes occur in the synovia of individuals with GWVI and joint pain. We therefore conducted a detailed histologic and immunohistochemical analysis of synovial biopsy samples obtained from individuals with joint pain who were referred for rheumatology consultation from a clinic for Gulf War veterans with unexplained symptoms. We found no evidence of significant inflammation or other obvious histologic abnormalities, suggesting that synovitis is not a frequent cause of the articular symptoms in individuals with GWVI.
PATIENTS AND METHODS
Patients and tissues.
This study was approved by the local institutional review board. The patients were deployed US veterans of the first Gulf War and originated from the referral area to the Philadelphia VA Medical Center. They were initially seen at this institution in a screening clinic for individuals with suspected GWVI and met entry criteria for a study of treatment for GWVI with doxycycline (7). All patients in that study had been deployed to the Persian Gulf area between August 1990 and August 1991, developed manifestations of GWVI after deployment, and had symptoms for at least 6 months (7). The patients included in the present study were referred for rheumatology consultation because of unusually severe musculoskeletal symptoms, including joint pain. Nine patients, in whom a knee was among the joints most affected by pain, consented to closed-needle biopsy of a symptomatic knee. None of these patients had knee effusions. Other identifiable causes of chronic musculoskeletal pain, including hepatitis virus infection, were excluded with the appropriate tests. Biopsies were performed in the years 1996–1999 (5–8 years after the Gulf War), essentially as described previously (8), prior to randomization for the doxycycline study. Specimens from 6 patients had adequate overall tissue integrity, consisted of synovial membrane as evidenced by the presence of synovial lining and vascularized subintima, and were therefore chosen for further analysis. Synovial tissues, which had been obtained by needle biopsy or during arthroplasty, from patients with rheumatoid arthritis (RA) with active disease despite disease-modifying antirheumatic drug treatment (n = 10) and from patients with osteoarthritis (OA; n = 12) were studied as disease controls.
Histology and immunohistochemistry.
Paraffin blocks were stored at room temperature and cut into 6–μm thick sections within 2 weeks before histologic analysis. To confirm the general preservation of tissue antigens, all specimens were immunostained for the cytoskeletal protein vimentin, which revealed strong staining in all cases (data not shown). Specimens were then stained with hematoxylin and eosin (H&E) for standard histologic assessments, or with standard 3-step immunohistochemistry for expression of CD68 (clone KP-1, macrophages), CD3 (T cells), CD20 (clone L26, B cells), CD38 (plasma cells), Ki-67 (clone K-2, cell proliferation), and von Willebrand factor (vWF; blood vessels). An automated immunostainer (BenchMark; Ventana Medical Systems, Tucson, AZ) was used, with nonspecific isotype or IgG as negative control and diaminobenzidine as chromogen. With the exception of the CD38 antibody, which was obtained from Novocastra Laboratories (Newcastle-upon-Tyne, UK), commercial antibody preparations (Ventana Medical Systems) were used according to standard staining protocols. Absence of staining due to technical failure was excluded by including appropriate positive control tissues in each staining run.
Quantification of histologic and immunohistochemical differences.
Histologic changes in the H&E stained sections were graded with the synovitis score according to Krenn et al (9, 10), with the modification that the average of all fields containing synovial lining was recorded per specimen. H&E stained sections from synovial biopsy samples from clinically asymptomatic normal knee joints were available from an unrelated study of needle biopsy samples from healthy volunteers (8) and were used as normal controls for the H&E-based evaluations. Tissue blocks or unstained sections for immunostaining were not available from these specimens. Immunohistochemical expression differences were determined by manual counting of positive staining cells in the lining or sublining using a Nikon OPTIPHOT-POL microscope (Optical Apparatus, Ardmore, PA) at 400× magnification (high-power field [hpf]). All histologically adequate fields, as defined by the presence of synovial lining with underlying vasculature, were evaluated and the averages of all fields were computed for each specimen. First, the microscopic field was adjusted such that the lining traversed the diameter of the field, and the number and percentage of positive staining cells were determined in the lining. For the evaluation of the subintima, the field was then moved so that it covered the immediately adjacent subintima. Vascular density was determined in the same subintimal fields by counting the number of vWF-positive vessels and was expressed as the number of vessels per hpf. A composite immunohistochemical inflammation score was calculated as follows. Because the absolute numbers of positive staining cells differed greatly among the cell markers assayed, all values were transformed into percentiles. For each cell marker, the specimen with the highest value (i.e., number of positive staining cells or vessels per hpf) was assigned the value of 100%. The relative percentage value was then computed for each of the other specimens stained for the same marker. The mean percentage value for each arthropathy (i.e., RA, OA, or GWVI) and cell marker was then computed by calculating the sum of the percentiles for one cell marker for each arthropathy and dividing this value by the number of specimens contained in this arthropathy. The composite inflammation score for a given arthropathy was the sum of the mean values for all cell markers. Statistical significance was determined by analysis of variance.
Demographic and clinical characteristics of the patients are summarized in Table 1, and musculoskeletal symptoms and findings are described in Table 2. Selected fields from H&E stained sections of all GWVI specimens are depicted in Figure 1. Normal histology was seen almost exclusively (Figure 1, panels A–F). The synovial lining was not hyperplastic, usually consisting of 1–2 cell layers. The lining cells had the typical elongate shape of noninflamed synovial lining. The subintimal fat was well preserved, and stromal fibroblasts predominantly displayed the elongate or spindle-shaped morphology typically seen in noninflamed synovium. There was no evidence of chronic or acute inflammatory changes in the blood vessels, such as fibrinoid deposits. Occasional inflammatory cells were seen, but there were no prominent perivascular infiltrates or lymphoid aggregates. The greatest abnormalities seen were mild and consisted of focal lining hyperplasia (Figure 1G) and mild perivascular lymphocyte infiltration (Figure 1H; compare also the immunostain depicted in Figure 2B) in 1 specimen each. There were no apparent associations between these abnormalities and the clinical features of the corresponding patients.
Recorded at the most recent prebiopsy evaluation. MCP = metacarpophalangeal joint; PIP = proximal interphalangeal joint; MRI = magnetic resonance imaging.
Multiple joint and muscle pain. Knee pain.
16/18 fibromyalgia tender points. No knee effusion. No limitation of range of motion.
Normal radiographs of knee, wrists, hands, and elbows.
Pain and stiffness in knees, ankles, shoulder, hands.
Extensive MCP and PIP tenderness; no swelling. Knees: full range of motion, no effusion.
Knee: joint space narrowing, no osteophytes.
Stiffness right shoulder, pain in lower back and both knees.
4/18 fibromyalgia points. Knees: full range of motion, no effusion; slight crepitus.
Normal radiographs of hips, knees, pelvis, and lumbar spine.
Generalized body pain, knee pain.
18/18 fibromyalgia tender points. Slight limitation to shoulder internal rotation. Hips: decreased internal rotation. Both knees: positive McMurray test, crepitus, and flexion limited to 100°.
Normal radiographs of hips and knees. MRI: lumbar disk disease.
Pain in multiple joints.
Multiple soft tissue tender points. Knees: full range of motion, no effusion.
Normal radiographs of hips, knees, and spine.
Generalized musculoskeletal pain, knee pain. Chronic left shoulder bursitis.
Left shoulder impingement. Biceps groove tenderness. 10/18 fibromyalgia points. Knees: full range of motion, no effusion.
Normal radiographs both knees.
To compare the histologic features in the GWVI synovia more precisely with those found in well-characterized arthropathies, a histologic grading system (9, 10) was used. Consistent with results from a previous validation study of this scoring system (10), the RA specimens displayed the highest scores (mean ± SEM 6.0 ± 0.19, corresponding to moderate synovitis), indicating the most severe inflammatory alterations, whereas the OA specimens had intermediate scores (mean ± SEM 2.6 ± 0.39, mild synovitis) (Figure 1I). Significantly, the normal and GWVI synovia had the lowest mean scores of 1.38 ± 0.30 and 1.41 ± 0.29, respectively, both corresponding to no synovitis (Figure 1I). These results agreed well with the predominantly normal histology seen in the H&E stains (Figure 1).
Assessment of inflammation-associated changes by immunohistochemistry
The specimens were then examined for the presence of inflammatory cells, cell cycle activity, and degree of vascularization by immunostaining for the appropriate markers. Figure 2 shows microscopic fields from the GWVI specimens with the highest numbers of CD68+, CD3+, CD20+, and Ki-67+ cells and vWF-positive blood vessels, and also illustrates the absence of CD38+ plasma cells. The results of the quantitative comparisons are shown in Figure 3. In the surface layer of the lining, the GWVI specimens possessed the lowest number of CD68+ macrophage-like cells (Figures 2A and 3A), a cell type that is typically increased in inflammatory arthropathies due to recruitment into the lining (11). Likewise, infiltrating subintimal inflammatory cells with positive staining for CD68, CD3, CD20, or CD38 were lowest in the GWVI biopsy specimens, whereas the OA specimens displayed intermediate numbers of these cells, and the RA specimens the highest numbers (Figures 2A–-E and 3B–E). Notably, only 1 GWVI specimen contained a few scattered CD20-expressing cells (Figure 2D), and CD38-expressing cells were absent in all GWVI specimens (Figures 2E and 3E). Cell division is increased in inflamed synovia and can be detected by staining for the nuclear marker Ki-67 (12). Ki-67 expression was least frequent in the GWVI specimens (Figure 3F) and, interestingly, localized predominantly to adipocytes (Figure 2F). Vascular density can increase in inflammatory arthropathies due to neovascularization (13). Subintimal blood vessel density, as assessed by enumerating the number of vWF-positive vessels per unit area, was lowest in the GWVI specimens, intermediate in OA specimens, and highest in RA specimens (Figure 3G).
Using the immunohistochemical inflammation score, we then compared the overall severity of inflammation-associated changes among the arthropathies. In keeping with all previous results, the GWVI specimens had the lowest scores, the OA specimens had intermediate scores, and the RA specimens had the highest scores (Figure 3H).
Using qualitative and quantitative light microscopic evaluations and a detailed immunohistochemical analysis, we failed to detect evidence of clinically significant inflammation in synovial tissue from patients with GWVI and joint pain. We chose OA as a presumably mildly or minimally inflamed disease control. There are no published reference values for the relative numbers of inflammatory cells in patients with OA, but the results of the cell marker quantifications clearly classified the OA specimens as substantially less inflamed than the RA group and, indeed, their mean synovitis score fell into the range of mild synovitis. Likewise, all parameters assayed were remarkably lower in the GWVI specimens than in the OA specimens. For instance, depending on the marker, the OA specimens contained 4.6–76 times more inflammatory cells than the GWVI specimens (Figure 3). In addition, the synovitis score classified the GWVI specimens as no synovitis. One may therefore conclude that the GWVI specimens form a group characterized by the absence of significant inflammation. Isolated histologic findings that might be interpreted as signs of a mild synovitis included focal lining hyperplasia in patient 6 (Figure 1G), mild perivascular infiltrates in patient 5 (Figures 1H and 2C), and a solitary cluster of B lymphocytes in patient 2 (Figure 2D). However, as we (8) and others (14, 15) have demonstrated in prior studies, such features are commonly found in clinically asymptomatic and otherwise histologically normal knee joints. Taken together, these findings suggest 1) that synovium in patients with GWVI does not exhibit clinically significant histologic evidence of inflammation and 2) that arthralgia in veterans with symptoms of GWVI is not usually due to an inflammatory process. These findings agree well with our clinical observation that joint effusions were not found in the patients in this study and that radiographic studies did not demonstrate erosive changes. All of the patients in our study had widespread arthralgia, 3 patients (50%) also had diffuse soft tissue pain and tenderness and 2 (33%) had significant fatigue, thus displaying characteristics of the fibromyalgia syndrome complex. This finding is consistent with the reported high frequency of fibromyalgia among Gulf War veterans with unexplained musculoskeletal pain (3, 16).
Exposure to toxic (17) and/or immunogenic substances (18) has been implicated etiologically in the GWVI symptom complex. Among the latter, squalene is an adjuvant that causes inflammatory arthritis in a rat model (19). Squalene is also an approved adjuvant in immunizations for humans, and exposure to it through vaccinations against biologic warfare agents has been implicated as a cause of GWVI symptoms (20). However, our results did not reveal any evidence of significant synovial inflammation and therefore do not support this model.
Our findings agree well with those of Amato et al (21), who evaluated muscle biopsy samples from Gulf War veterans with severe muscle fatigue, weakness, or myalgias and found only nonspecific histologic abnormalities, which were not believed to be clinically significant. In keeping with these findings, a 10-year followup health assessment of more than 1,000 Gulf War veterans revealed a higher than expected prevalence of nonspecific conditions such as fibromyalgia and chronic fatigue, although no specific inflammatory conditions were identified (5). Likewise, no abnormalities were detected by allergy testing of a series of 20 symptomatic Gulf War veterans (22) or by in vitro immune function testing of 52 symptomatic Gulf War veterans (23). These results contrast with a recent large-scale study in which mild activation of cellular immunity was detected in symptomatic Gulf War veterans (24), which correlated positively with the number of Gulf War–related immunizations received. However, activation was relatively small, as reflected by mild elevations in Th1 (interferon-γ) and Th2 (interleukin-10) cytokine synthesis, and may not be large enough to result in histologically detectable changes in an end organ, e.g., the synovial membrane.
Having excluded inflammation as a major cause of the joint pain in these patients, our results do not allow us to infer potential noninflammatory etiologies. Given the relatively small number of patients with biopsy samples, one also has to be aware of the possibility that these patients developed musculoskeletal symptoms coincidentally and that there was no causal relationship to their prior deployment in the Persian Gulf area. For instance, in patient 4 we cannot exclude a mechanical cause of the knee pain, such as the patellofemoral syndrome, although this would not explain his other widespread musculoskeletal symptoms.
A selection bias existed in this study in that only patients with severe symptoms underwent biopsy. However, this makes the absence of histologic abnormalities even more significant. Another potential caveat is that analysis of a larger number of specimens might reveal some specimens with more prominent evidence of inflammation. However, the absence of significant inflammation from all 6 specimens suggests that the percentage of inflamed specimens would be small. We did not assay the specimens for proinflammatory cytokines, which could potentially be synthesized by synovial cells such as lining or stromal fibroblasts. However, given the predominantly normal histology of the specimens, including normal fibroblast morphology, potentially present inflammatory mediators did not reach levels high enough to result in significant histologic alterations.
In summary, we failed to detect evidence of clinically significant inflammation in synovial biopsy samples from veterans with musculoskeletal pain and GWVI. These results support the conclusions from other studies, which have argued that inflammation is not a major cause of GWVI symptoms (25).
Dr. Pessler had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Funds from a collaborative research grant by Merck Research Laboratories (Rahway, NJ) were used to purchase some of the reagents used in this study. Merck Research Laboratories did not participate in study design, data collection, data analysis, or writing of the manuscript. Merck Research Laboratories agreed to submit the manuscript for publication and approved the content of the submitted manuscript without alterations.
We thank John Romano (Philadelphia VA Medical Center) for help with immunostaining and Catherine McKenna and Jan Dinnella for administrative support.