Chemoattraction of Human T Cells by IL-18¹ Free

Interleukin 18, a member of the IL-1 cytokine family, mediates important activities during both acquired and innate immune responses. Initially characterized by its capacity to promote Th1 responses in synergy with IL-12 (1, 2), IL-18 has recently been shown to drive either Th1 or Th2 responses, dependent on the cytokine microenvironment, suggesting a broader role in functional T cell differentiation than that originally recognized (3, 4). Sustained IL-18Rα expression on Th1 cells however suggests preferential activation of the latter in chronic lesions, e.g., in rheumatoid arthritis synovitis (5). A role in innate responses is also proposed. IL-18 enhances NK cell cytotoxicity and directly induces IFN-γ production by NK cells (6). Moreover, we recently demonstrated an important role for IL-18 in activating neutrophils and in promoting their recruitment to the peritoneal cavity in vivo (7). IL-18 can also directly induce monokine production by macrophages that constitutively express IL-18Rα (5).

IL-18 has been implicated as a critical factor in host responses in a broad range of infectious and autoimmune diseases. Protective Th1 responses during various bacterial and viral infections may be abrogated or enhanced by manipulation of IL-18 expression in IL-18-intact mice (8, 9). IL-18-deficient mice exhibit altered responsiveness to Mycobacterium bovis, Propionibacterium acnes, and Leishmania major, associated with modified T cell/NK cell function (6, 10). IL-18 mRNA is up-regulated in nonobese diabetic mice and the murine IL-18 gene maps to the idd2 susceptibility locus (11). Similarly, IL-18-deficiency is associated with altered myelin oligodendrocytic glycoprotein peptide-specific autoreactive T cell responses and amelioration of autoimmune encephalomyelitis (12). Several compelling data now implicate IL-18 in the pathogenesis of human inflammatory disease states including rheumatoid arthritis (RA),³ psoriasis, sarcoidosis, adult onset Still’s disease, and inflammatory bowel disease (13, 14, 15, 16). In particular, we have recently proposed that IL-18 exerts important proinflammatory activity in RA synovial tissues (17). IL-18-deficient mice develop significantly reduced incidence and severity of collagen-induced arthritis compared with wild-type mice (18) and Ab-, or IL-18-binding protein, mediated IL-18 neutralization suppresses streptococcal cell wall-induced arthritis and collagen-induced arthritis (19, 20). The mechanisms whereby IL-18 mediates effects in synovium remain unclear, but likely include modulating articular Th1 cell responses, directly promoting macrophage TNF-α production (5) and enhancing endothelial activation and angiogenesis (21).

We have explored the hypothesis that IL-18 can contribute to acquired immune response development and to the maintenance of chronic immune stimulation in diseases such as RA by promoting chemotaxis of activated T cells. The present report documents for the first time that IL-18 induces human CD4⁺ lymphocyte chemotaxis. IL-18 induced T cell polarization and migration into collagen gel matrices. Moreover, we demonstrated that such activity resides primarily in human Th1 cells that retain high levels of IL-18Rα expression. RA synovial CD4⁺, but not CD8⁺ T cells also migrate to IL-18, suggesting that our observations have physiologic relevance. Finally, we have shown that IL-18 can induce local mononuclear cell recruitment in vivo. Together these data show that IL-18 is chemotactic for T lymphocytes and as such provide a novel mechanism whereby IL-18 can promote and sustain inflammatory responses.

PE-conjugated anti-human IL-18R Ab and anti-human IL-18 Ab were obtained from R & D Systems (Abingdon, U.K.). Recombinant human IL-18 was cloned, expressed, and purified as described previously (5). Murine recombinant IL-18 was obtained from PeproTech (London, U.K.). Cytokines used were free of LPS as assessed by the Limulus amebocyte assay (Sigma-Aldrich, Poole, U.K.). FITC-conjugated anti-human CD4, CD8, CD19, CD14, CD56, and PE- and FITC-conjugated isotype control mouse IgG1 were obtained from Sigma-Aldrich. The medium routinely used was RPMI 1640 supplemented with 25 μM 2-ME (Sigma-Aldrich), penicillin (100 IU/ml), streptomycin (100 μg/ml), l-glutamine (2 mM), and FCS (10%) (all obtained from Life Technologies, Paisley, U.K.).

Heparinized venous blood was withdrawn gently from the forearm of healthy donors. PBMCs were purified by standard procedures using density gradient separation medium (Lymphoprep; Nycom Pharma, Oslo, Norway). The cells were washed extensively and either resuspended in RPMI 1640–10% FCS and used immediately for polarization assay or cultured overnight at 37°C in 5% CO₂ in 1 ml of medium (2 × 10⁶ cells/ml) in 24-well dishes in the presence or absence of staphylococcal enterotoxin B (SEB, 1 μg/ml; Sigma-Aldrich) as described previously (22). After washing, most of the cells reverted from the locomotive to spherical morphology. Human umbilical cord blood was obtained during full-term cesarean deliveries under appropriate ethical consent (North Glasgow Trust Ethical Committee, Glasgow, U.K.). CD4⁺ T cell populations were negatively purified from human umbilical cord blood by high-gradient magnetic sorting using the Viro MACS magnetic columns (Miltenyi Biotec, Auburn, CA) according to the manufacturer’s recommendations. The depletion mixture consisted of hapten-conjugated Abs (CD8, CD11b, CD16, CD19, CD36, and CD56). Purity of CD4⁺ T cells was >97% as judged by flow cytometry analysis using anti-CD14, CD3, CD19, CD4, CD56, and CD8 mAbs. To prepare Th1 cell populations, freshly purified CD4⁺ T cells were cultured in 25-ml tissue culture flasks (Nunc, Roskilde, Denmark) at 0.5 × 10⁶ cells/ml in the presence of PHA (1 μg/ml; Sigma-Aldrich), rIL-2 (100 ng/ml; R & D Systems), rIL-12 (2 ng/ml; BD PharMingen, San Diego, CA), and neutralizing anti-IL-4 Ab (200 ng/ml; BD PharMingen). Three days later, IL-2 (10 ng/ml) was added and the cells were harvested on day 5 for analysis. Synovial fluid was obtained from six patients with RA satisfying the American College of Rheumatology criteria (23). Mononuclear cells were separated as described above. The cells were used immediately after separation.

Surface markers of the cells were analyzed by FACSCalibur flow cytometry using CellQuest software (BD Biosciences, Mountain View, CA). To prevent nonspecific binding, all samples were preincubated for 30 min with human IgG (Sigma-Aldrich) to block the FcR. Aliquots of fresh cells or cultured cells (3 × 10⁵ cells/tube) were suspended in buffer containing 2% FCS and 0.02% sodium azide in PBS (staining buffer) and stained for 20–30 min with specific FITC- or PE-conjugated mAbs. Cells labeled with FITC- or PE-conjugated mouse isotype-matched Abs were used as controls.

This allows a direct phenotype visualization of polarized cells. The cells were fixed with 1% paraformaldehyde (Sigma-Aldrich) for 10–15 min to retain locomotor morphology, then attached to poly-l-lysine-coated coverslips (2 mg/ml; Sigma-Aldrich), and stained with appropriate FITC- or PE-labeled markers. Marker-positive and -negative, polarized and spherical cells were counted using a ×40 oil phase-contrast objective of a Zeiss Axioskop fluorescence microscope (Zeiss, Oberkochen, Germany).

Cell morphology using Giemsa-stained preparations is a useful preliminary to phenotype synovial cells (n = 6 patients). The morphology of these cells was observed under the optical microscope: neutrophils (77.50 ± 1.50%), lymphocytes (17.33 ± 1.23%), and monocytes (5.17 ± 0.6%).

This assay measures the change from a spherical shape to the shape change characterized by head-tail polarity, typical of locomotive cells (24). Freshly isolated or overnight-cultured PBMCs were resuspended in RPMI 1640 medium supplemented with 10% FCS and mixed with human IL-18 or IL-15 at concentrations ranging from 0.1 to 1 μg/ml. Experiments were conducted using polystyrene round-bottom tubes (110 × 16 mm; Sterilin, Stone, U.K.). The tubes, containing 200 μl of cell suspension at a concentration of 2 × 10⁶/ml, were incubated for 30 min at 37°C. A basal control value was established using RPMI 1640-FCS alone. The cells were fixed using 200 μl of 2.5% glutaraldehyde (Sigma-Aldrich) and the proportions of cells showing head-tail polarization typical of locomotive cells were determined. The proportion of cells scored as either spherical (nonmotile) or polarized (motile) was counted directly using a ×40 phase-contrast objective. Cells 250–300(250–300) were counted blind, and polarized cells were expressed as a percentage of viable cells. Data presented are values after subtracting the background values (without cytokine), which ranged from 5 ± 0.8% for freshly isolated cells, 18 ± 1.7% for cells cultured with medium alone overnight, and 23 ± 2.3% for cell cultured with SEB.

Rat tail collagen (type I) was prepared in solution from freshly obtained rat tail tendons by established methods (25). Gels were formed by bringing soluble collagen (1.5–2 mg/ml) in dilute acetic acid solution back to physiologic pH and osmolarity. IL-18 or FCS in culture medium was added before gelatinization. The mixture was then transferred into 24-well dishes (Sterilin) and allowed to gel. Cells were prepared and activated with appropriate activators as described above. They were washed and overlaid on the gel in 24-well dishes and incubated at 37°C for 20 h (unless otherwise stated) to allow cells to invade the gel. The proportion of invading cells was determined with an inverted microscope by scoring the number of cells remaining on top of the gel and the number of cells that had penetrated the gel, counting a minimum of 200 cells. The proportion of locomotor cells was calculated from the ratio of the numbers of invasive and noninvasive cells. To distinguish between chemotaxis and chemokinesis, IL-18 was also added on top of the collagen gel and also in the gel and incubated for 20 h.

IL-5 and IFN-γ concentrations in culture supernatants were determined by ELISA using paired Abs (BD PharMingen) and developed with tetramethylbenzidine peroxidase substrate (Insight Biotechnology, Middlesex, U.K.) according to the manufacturer’s recommendations. Sensitivity of the assay was 10 pg/ml. IL-18 in synovial fluid was similarly determined using paired Abs (Diaclone Research, Besancon, France). Sensitivity of the assay was 45 pg/ml. IL-18 levels in RA synovial fluids were up to 155.4 ± 28.7 pg/ml (mean ± SEM, n = 6).

This was conducted as previously described (26, 27). Briefly, male DBA/1 mice (Harlan Olac, Oxon, U.K.) and IL-18-deficient mice of the DBA/1 background (10, 18), age 10 wk, were injected i.p. with 500 μg of P. acnes and 7 days later received 500 ng of murine rIL-18 or PBS in the hind footpad. Footpad swelling was measured at 0, 8, 15, and 24 h after injection using a constant pressure dial caliper (Kroeplin, Munich, Germany). Following the measurement, the mice were sacrificed and the footpads were removed, Formalin fixed, and analyzed histologically by H&E staining. All mice were kept in the Biological Service facilities at the University of Glasgow according to the U.K. Home Office guidelines.

Synovial membranes (arthoplasty specimens) were obtained from RA patients satisfying the American College of Rheumatology diagnostic criteria (n = 12). Formaldehyde-fixed paraffin sections (3 μm) were stained by a standard streptavidin-HRP protocol using an anti-IL-18 mAb (a kind gift from Dr. A. Jackson, Cancer Research U.K., Leeds, U.K.) (5). Product was visualized with 3,3′-diaminobenzidine tetrachloride (Sigma-Aldrich) and counterstained with hematoxylin (Sigma-Aldrich). IL-18 expression was quantified by two independent observers using light microscopy. For verification of specificity, the anti-IL-18 Ab was replaced by an isotype-matched control (Sigma-Aldrich). Neutralization was performed by preincubation of primary anti-IL-18 Ab with either 1 μg of rIL-18 or IL-1β (R & D Systems) before tissue staining.

Values are expressed as mean ± SEM. Statistical analysis was performed using Student’s t test. A p < 0.05 was considered to be significant.

Lymphocyte shape change (polarization) and migration into collagen gels represent useful indicators of chemokinetic activity. We therefore determined the proportion of PBMCs exhibiting polarization in response to increasing concentrations of IL-18. IL-18 induced no reproducible effect on freshly isolated lymphocyte populations (Fig. 1,a) reflecting low levels of spontaneous IL-18Rα expression (data not shown). It has been shown that SEB promoted cytokine-mediated lymphocyte polarization (22). We therefore stimulated PBMCs with SEB to enhance cell polarization. After overnight culture in SEB, PBMCs polarized to IL-18 in a dose-dependent manner that was similar to the response induced by IL-15, a recognized T cell chemoattractant (23). Maximal polarization was seen with 100 ng/ml IL-18; at this dose, ∼20% of total PBMCs were polarized (Fig. 1,a). Flow cytometric analysis of the PBMC fraction contained 65 ± 5.6% CD3⁺ cells (n = 5), 6.8 ± 0.7% CD19⁺ cells (n = 5), 8.5 ± 0.5% CD14⁺ cells (n = 2), and ∼96% of the cells within the lymphocyte gate expressed CD45 (data not shown). Immunofluorescence microscopy analysis showed that the polarized cells were predominantly CD3⁺CD4⁺ T cells, with few CD8⁺ T cells (<5%) assuming polarized morphology (data not shown). This is not unexpected since CD4⁺ but not CD8⁺ cells are responsive to SEB (28) Commensurate with this, cells cultured for 16 h in SEB, but not in medium alone, increased the expression of IL-18Rα on CD4⁺ cells (up to 40% of total CD4⁺ T cell population in PBMCs measured by flow cytometry, data not shown). The polarized and nonpolarized cells are shown in Fig. 1,b. We also investigated the ability of the mononuclear cells from human cord blood (which contained principally naive lymphocytes) to polarize in response to IL-18. Cord blood cells polarized in a similar manner as PBMCs in response to IL-18 following activation by SEB, but did not respond to IL-18 when they were not activated with SEB (Fig. 1,c). A similar percentage of polarized cells was obtained whether purified CD4⁺ T cells or PBMCs were examined (data not shown). As specificity control, we further showed that neutralizing monoclonal anti-human IL-18 Ab abolished the chemotactic effects of IL-18 (Fig. 1,d). SEB-activated lymphocytes polarized within 5 min of IL-18 stimulation and reached maximal levels after 30 min (Fig. 1,e). Although measurement of shape change is a reliable correlate of chemoattraction/locomotion, the assay does not directly measure locomotion. To formally demonstrate this, we used the collagen gel invasion assay. PBMCs were cultured overnight in SEB, washed to remove the supernatant, and overlaid on a collagen gel previously impregnated with IL-18 (or diluent alone) in medium/10% FCS. The percentage of cells migrating into collagen gels in response to IL-18 after 20 h was significantly higher than that in the control gels (Fig. 1 f). Moreover, IL-12 (at a range of concentrations) did not have any activity in both polarization and gel invasion assays (data not shown), suggesting that this activity is not a general property of innate response monokines. These data therefore demonstrate clearly that IL-18 can rapidly induce not only polarization but also invasive locomotion in PBMC subsets.

Since we have previously shown that IL-18Rα is preferentially expressed on committed Th1 cells (29), we next explored the hypothesis that IL-18 induced locomotion in Th1 cells. CD4⁺ cells purified from human cord blood were cultured under Th1 differentiating conditions. Flow cytometry showed that whereas naive cord blood CD4⁺ T cells expressed a low level of IL-18Rα, 90% of committed Th1 cells expressed the receptor (data not shown). Predominant IFN-γ (1.1 ± 0.1 ng/ml), but not IL-5 (<30 pg/ml), production confirmed functional Th1 differentiation (data not shown). IL-18 induced human Th1 cell migration into collagen gels in a dose-dependent manner (Fig. 2,a). However, only ∼30% of the Th1 cells migrated into the gel in response to 200 ng/ml IL-18. This is consistent with earlier reports that lymphocytes must enter G₁ phase for full expression of a locomotor phenotype (30). Cell cycle analysis by flow cytometry showed that 60% of the committed Th1 cells were in the G₀-G₁ phase at the time of the migration assay (data not shown). Time course analysis comparing the proportion and migration distance of cells into collagen gels containing 200 ng/ml IL-18 or medium alone showed increasing responses up to 18 h (Fig. 2, b and c). To confirm that the migration was not due to random chemokinesis, the gel invasion assay was performed with or without an IL-18 concentration gradient (Fig. 3 d). Maximal Th1 cell migration occurred in a positive IL-18 concentration gradient between the top of the gel and inside the gel (chemotactic condition). In contrast, no significant migration was observed in the absence of an IL-18 gradient (chemokinetic condition) or in the absence of IL-18 (spontaneous migration). These results therefore demonstrate that IL-18 induced primarily a chemotactic response. We also compared the relative responsiveness of Th1 and Th2 cells to IL-18-induced polarization/cell migration. CD4⁺ T cells from human cord blood were driven to Th1 or Th2 lineages as described in Materials and Methods. Although relatively pure Th1 lines were easily obtained, a pure Th2 line proved difficult to establish. Viability was low after two rounds of cultivation. The Th2 lines retained a substantial and variable percentage of IL-18R⁺ IFN-γ producing cells following a single round of driving under the Th2 condition. These cells polarized and migrated in response to IL-18. However, there was a marked decrease in the percentage of IL-18R⁺ cells after the second round of driving and this decrease was clearly reflected in the substantial decrease in the percentage of cells migrating in response to IL-18 (data not shown). It should be noted that although committed Th2 cells are generally IL-18R⁻, Th0 cells and differentiating Th2 cells express a low level of IL-18R (29). Therefore, as much as IL-18R is preferentially expressed on committed Th1 but not Th2 cells, IL-18 polarizes Th1 rather than Th2 cells. It should also be noted that we have reported earlier that IL-18 attracts the migration of neutrophils, which also express IL-18R (7).

SEB represents only a surrogate for physiologic activation of lymphocytes. We therefore performed locomotion assays on lymphocytes that were activated in vivo. IL-18 expression is up-regulated in RA synovial tissues (5), particularly in lymphocyte-rich aggregates and adjacent to endothelia (Fig. 3). Mononuclear cells were prepared from synovial fluids freshly aspirated from RA patients and tested for their ability to polarize and migrate in the presence of graded concentrations of IL-18. Lymphocytes polarized in response to IL-18 in a dose-dependent manner (Fig. 4,a). Similar results were obtained when lymphocytes were tested for their ability to migrate through collagen gels (Fig. 4,b). Synovial lymphocyte fractions differ functionally and with respect to IL-18Rα expression. The expression of IL-18Rα on CD4⁺ and CD8⁺ synovial T cells was measured by FACS and revealed that whereas >50% of CD4⁺ synovial T cells expressed IL-18Rα, <10% of CD8⁺ T cells were IL-18Rα⁺ (Fig. 4, c and d). To investigate which cell subset(s) was the targets of the IL-18-mediated chemotaxis ex vivo, immunofluorescence microscopy was used to identify the responder cells. CD4⁺, but not CD8⁺, synovial T cells polarize significantly to IL-18 compared with control medium (Fig. 4 e). Moreover, CD19⁺ synovial B cells did not express IL-18Rα and did not respond to IL-18 in a polarization assay (data not shown).

The above data suggested that IL-18 was a potent chemoattractant for CD4⁺ T cells in vitro, raising the possibility that IL-18 may recruit T cells into local tissues during inflammatory responses. To investigate this, female DBA/1 mice were primed with P. acnes and then injected s.c. into a hind footpad with 500 ng of recombinant murine IL-18 or PBS alone. Mice injected with IL-18 developed significant footpad swelling compared with PBS-injected controls (Fig. 5). Moreover, IL-18 injection into footpads of IL-18-deficient mice, which express higher levels of IL-18R a priori (6, 10), led to significantly greater footpad swelling than IL-18-injected wild-type mice. Subsequent histologic analysis showed a substantial mononuclear infiltrate at the site of injection, whereas PBS-injected footpads were similar to noninjected controls (Fig. 6).

IL-18 has recently been attributed numerous properties commensurate with a vital role in innate and acquired immune responses. It is now clearly established that IL-18 can promote and regulate the evolution of functionally distinct T cell subsets, driving toward either type 1 or type 2 functional status dependent on ambient cytokine expression and costimulator expression on APC (3, 4). We now add to this the capacity to contribute to T cell recruitment to inflammatory sites via promotion of chemotactic responses. This may be manifest particularly upon Th1 cells that retain IL-18Rα expression as a phenotypic feature, although we cannot rule out similar effects manifest upon differentiating Th2 cells that transiently retain IL-18Rα expression through earlier stages of maturation. “Resting” IL-18 expression at the mRNA and protein (at least as pro-IL-18) levels has been observed in several tissues and cell types. By virtue of its early presence following tissue insult, the chemotactic properties demonstrated herein support a dual role for IL-18 in T cell activation. It may function as an important mediator in determining the nature of memory T cells recruited (preferentially IL-18Rα⁺CD4⁺ Th1 cells) at the same time as influencing, in synergy with IL-12 or IL-4, the functional destination of naive T cells present within a lesion. Importantly, we found no evidence of IL-18Rα expression on naive human T cells derived from cord blood nor any motile response by such naive cells to IL-18, and it is therefore unlikely that IL-18 has any significant role in recruitment of such T cells.

It will be important to determine the relationship that IL-18-induced T cell recruitment has to that mediated by other established chemokine receptor-mediated pathways, in particular those operating via CCR5, a putative Th1 cell marker (31). That IL-18-deficient mice generate inflammatory responses has been clearly demonstrated in several models (6, 10), indicating that IL-18 expression is not obligatory for host defense. Nevertheless, several reports suggest that IL-18 deficiency does prejudice host responses to some microbial species. Moreover, excess expression of IL-18 is sufficient to initiate or considerably enhance autoimmune responses. Transgenic cutaneous overexpression of caspase 1 induces high levels of IL-18 expression in the skin and leads to development of a marked local inflammatory infiltrate that is dependent in large part upon the presence of IL-18 (32). Similarly, IL-18 injection into male MRL/lpr mice considerably accelerates the development of autoimmune glomerulonephritis and intriguingly induces a pronounced cutaneous inflammatory infiltrate similar to that observed in caspase 1-transgenic mice (33). IL-18 administration is sufficient to induce inflammatory arthritis in collagen/IFA-primed DBA/1 mice (34). We now show that local introduction of recombinant mature IL-18 in vivo into a footpad is sufficient to induce a marked mononuclear cell infiltrate. The precise mechanisms whereby IL-18 mediates such effects in vivo require further analysis. Although the in vitro time course results (IL-18-induced T cell polarization within 5 min) presented herein render it unlikely, our data do not exclude the possibility that IL-18 could operate through inducing release of intermediate factors, including chemokines. This possibility is currently being investigated.

IL-18 may exert chemotactic responses in other cell lineages. Chemokinetic responses to IL-18 for endothelial cells may in part underpin the proangiogenic effects of IL-18 in several neovascularization models (21). We have also shown that IL-18 may promote neutrophil expression of adhesion molecule expression and recruitment to the peritoneal cavity in vivo (7). IL-18 has also been shown to up-regulate ICAM-1 expression on monocytes and on T cells, providing further mechanisms whereby IL-18 can promote T cell recruitment (35, 36).

Our data are of particular significance in determining the potential activities of IL-18 in chronic inflammatory diseases such as RA. IL-18 detection in inflammatory sites is characterized by high levels of expression in mononuclear cells adjacent to endothelial cells, providing circumstantial evidence that IL-18 is released in a geographically cogent area of an inflammatory lesion to facilitate further T cell recruitment (5). However, IL-18 could also operate within an inflamed tissue irrespective of trans-endothelial recruitment. We, and others, have recently proposed that recruitment of memory T cells and their subsequent cytokine-mediated activation within inflamed synovial membrane can drive downstream TNF-α release and consequent pathology (5, 37). Critical to this are cell contact-dependent interactions between activated T cells and macrophages that in turn promote proinflammatory cytokine release (38). In unpublished studies we have shown that IL-18 can considerably enhance T cell/macrophage cell contact-mediated cytokine release (J. A. Gracie and I. B. McInnes, unpublished data). It is attractive to propose that IL-18 released by macrophages generates microgradients within synovial tissue that then optimizes further T cell recruitment and activation adjacent to the macrophage, thereby perpetuating the chronic inflammatory process.

In conclusion, we provide here evidence that IL-18 has chemotactic activities in vitro and in vivo operating in particular upon human Th1 cells. Thus, in addition to its activities in functional regulation of naive T cell differentiation, IL-18 may have an important role in promoting recruitment of memory Th1 cells to an inflammatory lesion. This adds further to the portfolio of activities that render IL-18 an attractive therapeutic target in a variety of important autoimmune inflammatory disease states.

We thank Roderick Ferrier (Department of Pathology, Western Infirmary, Glasgow, U.K.) for technical assistance with histopathology. We also thank Dr. J. Hunter (Gartnavel General Hospital, Glasgow, U.K.) for providing additional synovial samples and Dr. Ian Mackay (Division of Immunology, Infection and Inflammation, University of Glasgow, Glasgow, U.K.) for assistance in statistical analyses.