Surface-bound anti–type II collagen–containing immune complexes induce production of tumor necrosis factor α, interleukin-1β, and interleukin-8 from peripheral blood monocytes via fcγ receptor IIA: A potential pathophysiologic mechanism for humoral anti–type II collagen immunity in arthritis




Type II collagen (CII) is a major component of hyaline cartilage, and antibodies against CII are found in a subgroup of patients with rheumatoid arthritis. We undertook this study to investigate whether and how antibodies directed against CII can form solid-phase immune complexes (ICs) with cytokine-inducing properties in a model theoretically resembling the situation in the inflamed joint, in which CII is exposed for interaction with anti-CII antibodies during periods of inflammation.


Sixty-five arthritis patients with varying levels of anti–native CII antibodies and 10 healthy controls were evaluated concerning anti-CII and cytokines induced in a solid-phase IC model. Monocytes were either depleted or enriched to define responder cells. Antibodies blocking Fcγ receptors (FcγR) were used to define the responsible T cell surface receptors.


ICs containing anti-CII from arthritis patients induced the production of tumor necrosis factor α (TNFα), interleukin-1β (IL-1β), and IL-8. We found a close correlation between enzyme-linked immunosorbent assay optical density values and induction of TNFα (r = 0.862, P < 0.0001), IL-1β (r = 0.839, P < 0.0001), and IL-8 (r = 0.547, P < 0.0001). The anti-CII–containing IC density threshold needed for cytokine induction differed among peripheral blood mononuclear cell donors. Anti-CII–containing IC–induced cytokine production was almost totally abolished (>99%) after monocyte depletion, and receptor blocking studies showed significant decreases in the production of TNFα, IL-1β, and IL-8 after blocking FcγRIIa, but not after blocking FcγRIII.


These findings represent a possible mechanism for perpetuation of joint inflammation in the subgroup of arthritis patients with high levels of anti-CII. Blockade of FcγRIIa and suppression of synovial macrophages are conceivable treatment options in such patients.

Collagen fibrils are structural tissue proteins. Among more than 15 different types of collagens that have been identified in different tissues of vertebrates, type I, II, and III collagens are most abundant in the body. Type II collagen (CII) is the predominant collagen in joint cartilage, constituting >50% of its dry weight (1).

Electron microscopy studies have shown that a thin layer of proteinaceous material covers the healthy cartilage surface facing the joint space (2, 3). Exposure of cartilage CII epitopes to the inflammatory and immune systems differs between healthy and inflamed joints. In healthy joints, the proteinaceous layer covering the intact articular surface inhibits anti-CII antibody binding. In joint inflammation, cartilage CII epitopes are exposed to anti-CII antibodies due to disruption of this proteinaceous layer. Prerequisites thereby exist for the formation of surface-bound immune complexes (ICs) containing anticartilage antibodies, as has been reported in joint inflammation in a rabbit arthritis model (4, 5) and in rheumatoid arthritis (RA) (6), but not in healthy joints (2).

Anti-CII antibodies are present in a subpopulation of RA patients. In different investigations, between 3% and 27% of RA patients have elevated anti–native CII levels (7–11). Anti-CII is spontaneously produced by cells in the synovial tissue (12) and synovial fluid (SF) (13, 14), but less so by peripheral blood B cells (13, 14). Anti–native CII levels are accordingly higher in SF than in the serum of RA patients (15, 16). Antibodies against native CII recognize epitopes dependent on the intact triple-helical conformation of native CII. When CII is denatured (e.g., by heat), the chains are separated and the epitopes specific for native CII are destroyed. RA patients have different antibody reactivities against native and denatured CII (11, 17). Whereas antibodies against denatured CII also occur in patients with other diseases (18), reactivity against the intact native CII conformation is more closely associated with RA and a few other diseases, such as relapsing polychondritis.

In mice susceptible to experimental collagen-induced arthritis (CIA), immunization with CII leads to development of an autoimmune polyarthritis and production of anti-CII antibodies (19, 20). Injection of anti-CII alone into experimental animals creates either a mild self-limiting synovitis (21–25) or severe joint-destroying arthritis (26, 27). In the latter study, a mixture of 3 monoclonal antibodies (mAb) directed against discrete CII epitopes was used. The investigators hypothesized that this circumstance created a situation with optimal spatial configuration of antibody epitopes. They also noted that the antibody-mediated arthritis seemed to affect the cartilage directly, in contrast to CIA induced by CII immunization, in which the development of a pannus-like structure preceded cartilage and bone destruction (27). Numerous studies of RA patients have also emphasized epitope differences between anti-CII in individual RA patients (10, 28–31). Other groups have stressed the importance of both T cells and B cells in CIA development (32, 33).

Recent studies by our group have shown that circulating systemic lupus erythematosus (SLE) ICs can stimulate the production of cytokines such as interleukin-10 (IL-10) and IL-6 (34) from peripheral blood mononuclear cells (PBMCs), and that cryoglobulins from patients with lymphoproliferative diseases stimulate the production of tumor necrosis factor α (TNFα) and IL-10 (35). Investigators in our group have previously reported that anti–native CII is produced by antibody-producing cells obtained from the SF of patients with active RA (13). We have now extended these studies to antigen-specific ICs with direct significance for immune reactions in arthritic joints, and we have investigated the potential of such anti-CII–containing ICs to induce proinflammatory cytokines of importance in RA pathogenesis.


Patients and controls.

Sera from 65 arthritis patients (73% women and 27% men, mean age 59 years [range 28–82 years]) and 10 healthy controls were selected for investigation. Eleven of the arthritis sera were obtained from patients fulfilling the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 revised criteria for the classification of RA (36). These sera had in earlier studies demonstrated antibody reactivity to chicken CII (data not shown). Ten of these sera were investigated together with control sera in the initial experiments of the present study, during which the technique used for cytokine induction was established. Samples from the other 54 arthritis patients were obtained at inclusion in the early arthritis (disease duration of <12 months) cohort at the Department of Rheumatology, Karolinska University Hospital. These 54 patients were chosen on technical grounds to represent varying levels of anti-CII antibodies based on an earlier investigation of anti–denatured CII antibodies (data not shown).

Control sera were obtained from healthy blood donors (3 men and 7 women, mean age 38 years [range 20–53 years]). The presence of inflammatory disease was ruled out in all controls before sampling. All sera were separated and stored at −70°C within 4 hours of sampling. As sources of responder cells for cell culture experiments, buffy coat preparations were obtained from healthy blood donors, and heparinized blood was obtained from laboratory personnel. All sera and cells were obtained after informed consent. The investigation was approved by the Ethics Committees of Akademiska Sjukhuset/Uppsala University Hospital, Uppsala, and the Karolinska University Hospital Stockholm.

Preparation of PBMCs.

Buffy coat preparations were diluted 1:4, and heparinized blood was diluted 1:2 in sterile phosphate buffered saline (PBS) at room temperature. After purification on Ficoll-Paque Plus density gradients (Amersham Pharmacia Biotech, Uppsala, Sweden) according to the manufacturer's instructions, PBMCs were diluted to 1 × 106/ml in RPMI 1640 medium with L-glutamine (Invitrogen, Paisley, UK) supplemented with 1% penicillin/streptomycin, 1% HEPES buffer, 12.5 μg/ml of polymyxin B sulfate (Sigma-Aldrich, Stockholm, Sweden), and 1% Ultroser G (Flow, Irvine, UK). Ultroser G is a serum supplement for in vitro cell cultures which, in previous studies in the laboratory, have been shown to sustain IC-induced cytokine production in otherwise serum-free systems (37).

Previous experience in our laboratory has demonstrated that different donors might differ considerably concerning their PBMC responses to artificial or patient-derived ICs. Since some PBMC donors might yield either universally low or high cytokine responses without any noticeable effect of the addition of ICs, PBMCs from at least 2 donors were always investigated in parallel in the present study. Only data from the PBMC donor exhibiting the strongest net IC reactivity are presented in the Results. To try to overcome this PBMC variability, 3 monocytoid cell lines were tested, but they were abandoned due to low or absent cytokine responses (U937, MonoMac6) or lack of reactivity to various IC preparations (THP-1).

CII preparations and blocking agents.

Initial experiments were performed using native rat CII (a kind gift from Helena Erlandsson Harris, Karolinska Institute, Stockholm, Sweden). After defining the IC stimulation technique, we changed to native human CII (enzyme-linked immunosorbent assay [ELISA] grade; Chondrex, Redmond, WA). Since the native collagen was not endotoxin tested, these experiments were performed with the addition of polymyxin B (see above).

In initial experiments, plates were blocked with bovine serum albumin (BSA) or with PBS alone. To obtain a fully human experimental setup, these blocking reagents were replaced with human serum albumin (HSA) intended for intravenous use (Albumin Immuno; Baxter, Kista, Sweden).

Solid-phase IC assay.

Different protocols for the manufacture of CII-containing ICs were investigated, utilizing either soluble or surface-bound ICs formed with either biotinylated or nonbiotinylated CII. The technique eventually selected for IC stimulation of PBMCs has similarities to an ELISA. However, we found that higher concentrations of both CII and serum were needed to obtain cytokine stimulation and reproducible differences between individuals compared with conventional ELISAs.

In our first experiments, we used native rat collagen and blocking with BSA instead of HSA. Whereas human CII generally yielded higher ELISA optical density (OD) values than rat CII, rat CII–containing solid-phase ICs induced higher TNFα production. BSA blocking induced a significant nonspecific TNFα production and was therefore exchanged for sterile endotoxin-free HSA without intrinsic cytokine-inducing properties (Figure 1b) in the final experimental setup. In the final protocol used, nonbiotinylated CII (10 μg/ml; 50 μl/well) was directly bound to MaxiSorp ELISA plates (Nunc, Roskilde, Denmark) overnight at 4°C. Plates were then blocked with 1% HSA in sterile PBS for 1 hour at room temperature on a shaker. After discarding the blocking solution, 50 μl of sera from patients and controls was added at a 10% dilution in sterile PBS for 2 hours at room temperature on a shaker. After further washing, 300 μl of responder cell suspensions was added and left to stand for 20 hours in a 37°C cell incubator with 5% CO2. Earlier experiments in our laboratory have defined 20 hours of cell culture to be optimal for measurement of IC-induced cytokine responses (34, 35, 37).

Figure 1.

Distribution of tumor necrosis factor α (TNFα) induction and enzyme-linked immunosorbent assay (ELISA) optical density (OD) values by immune complexes (ICs) containing antibodies from 10 rheumatoid arthritis sera and 10 healthy control sera. a, Cytokine induction from complete solid-phase ICs containing type II collagen (CII) and blocked with human serum albumin (HSA). b, TNFα induction from the corresponding wells only blocked with HSA. c, OD values from ELISA wells coated with CII and subsequently blocked with HSA. d, OD values from wells only blocked with HSA.

During method development, varying CII concentrations (0, 0.1, 1.0, and 10 μg/ml) and varying dilutions of anti-CII–containing sera (1:10, 1:100, and 1:1,000) were used. The use of plastics intended for sterile cell cultures was abandoned due to low binding of CII, low levels of cytokine production, and nonreproducible results. Previous experiments by our group using a Limulus amebocyte assay showed no measurable amount of endotoxin in MaxiSorp ELISA plates and no adverse effects on the outcome and intrawell variability in short-term cell cultures.


Half of the wells in high-binding ELISA plates (MaxiSorp) were coated with 50 μl of human native CII (ELISA grade), 10 μg/ml in PBS at 4°C overnight. Thereafter, the coating solution was discarded and the full plates were blocked. To facilitate comparison with the cytokine stimulation experiments, plates were blocked for 1 hour at room temperature with PBS with 1% HSA of the same quality as that used in cell culture experiments. Sera diluted 1:100 in PBS were then added to the plates in duplicate, both to the CII-coated half and on the half that had only been blocked, and left to stand for 2 hours. After washing with PBS–Tween, the detection antibody, an alkaline phosphatase–coupled goat F(ab′)2 antibody against the human IgGγ chain (diluted 1:10,000 in PBS/HSA; Jackson ImmunoResearch, Cambridgeshire, UK), was incubated for 1 hour at room temperature on a shaker. After further washing, the reaction was developed using 4-nitrophenyl phosphate disodium in diethanolamine buffer and read in a spectrophotometer at 405 nm after 45 minutes. This time point was chosen because it yielded the best discrimination between sera with high and low OD values. OD values were calculated as follows: (mean OD in the CII-coated wells) − (mean OD in wells only blocked).

While there were major variations of TNFα values induced in CII-coated plates in a set of 10 RA and 10 control sera (Figure 1a), there was almost no variation of TNFα values induced in parallel plates coated with HSA alone (Figure 1b). The correlation between anti-CII OD values and anti-CII–containing IC–induced TNFα levels did not change with subtraction of TNFα levels in blank wells (r = 0.86, P < 0.0001 and r = 0.89, P < 0.0001, respectively). We therefore decided not to subtract serum-specific blank wells in the cytokine induction experiments. In parallel ELISA investigations, we recorded a broad distribution of OD values in the wells coated with CII and subsequently blocked with HSA (Figure 1c), but we also recorded considerable variability in parallel control wells coated with HSA alone (Figure 1d and data not shown). We therefore decided that there was a need for the subtraction of serum-specific blank wells in the ELISA experiments accompanying the cytokine induction experiments. None of the 10 control sera gave higher OD values in CII-coated wells compared with control wells only blocked with HSA, and none of the control sera induced cytokine levels above background levels (HSA-blocked wells) when incorporated into solid-phase ICs containing CII and anti-CII.

Monocyte depletion and enrichment experiments.

Monocyte depletion and enrichment experiments were performed using RosetteSep reagents (StemCell Technologies, Vancouver, British Columbia, Canada) according to the manufacturer's instructions. This enrichment protocol yields totally unmanipulated monocytes for subsequent functional studies. Cell depletion and enrichment efficacy was monitored by staining with fluorescein isothiocyanate–conjugated anti-human CD14 antibodies (clone TYK; Dako, Glostrup, Denmark) and immediate analysis using a FACSCalibur flow cytometer (Becton Dickinson, Mountain View, CA). Cells depleted and enriched for monocytes were diluted in cell culture medium to the same final concentration as that used for untreated PBMCs.

Fcγ receptor (FcγR) blocking experiments.

For studies of the interaction between ICs and cell surface receptors, mouse mAb IV.3 against FcγRIIa (Fab fragment) and 3G8 against FcγRIII (F[ab′]2 fragment) were obtained from Medarex (Nutley, NY). PBMCs were incubated with FcγR-blocking antibodies (1.5 μg/ml; preliminary experiments had shown an equivalent blocking effect using either 1.5 or 4 μg/ml) for 30 minutes at 4°C before addition to CII- and anti-CII–coated plates. Cell suspensions were then incubated for 20 hours before supernatants were harvested for ELISA evaluation of cytokine levels. The mAb IV.3 has earlier been shown to react primarily with the activating human FcγRIIa, but not with the inhibitory FcγRIIb (38, 39). The mAb 3G8 is more effective in blocking FcγRIII-mediated functions compared with other antibodies (40).

Cytokine ELISAs.

Supernatants were harvested after 20 hours of incubation, and cytokine ELISAs for TNFα, IL-1β, and IL-8 were performed to measure cytokines released from stimulated PBMCs. All these methods have been established in our laboratory for the measurement of IC-induced cytokine responses in vitro (34, 35, 37, 41). The antibodies and recombinant cytokine standards used in the assay were purchased from R&D Systems (Abingdon, UK). For capture, mouse mAb MAB610 (2.0 μg/ml; for TNFα), MAB601 (2.0 μg/ml; for IL-1β), and MAB208 (4.0 μg/ml; for IL-8) were used, and for detection, biotinylated polyclonal goat antibodies BAF210 (0.1 μg/ml; for TNFα), BAF201 (2.5 μg/ml; for IL-1β), and BAF208 (0.02 μg/ml; for IL-8) were used.

MaxiSorp ELISA plates (96-well) were coated with 50 μl of capture antibody diluted in PBS and incubated at 4°C overnight. Thereafter, wells were blocked with 100 μl of PBS with 1% BSA (PBS/BSA) for 1 hour at room temperature on a shaker. Supernatants (50 μl) from stimulated PBMC cultures or recombinant cytokine standards diluted in RPMI were added in duplicate. The plates were then incubated for 2 hours at room temperature on a shaker. After washing with PBS–Tween, 50 μl detection antibody diluted in PBS/BSA was added for 2 hours. After further washing, 50 μl of streptavidin–horseradish peroxidase (R&D Systems) was added for 20 minutes on a shaker. Following additional washing, 50 μl of substrate solution (3,3′,5,5′-tetramethylbenzidine; Dako) was then added, and the plates were placed in the dark at room temperature until color development. The reaction was stopped with 50 μl of 1M H2SO4, and the plates were finally read at 450 nm in a spectrophotometer.

Statistical analysis.

For correlation analyses, Pearson's product-moment correlation with Fisher's r-to-z transformation was used. The Wilcoxon test was used to analyze the effect of FcγR-blocking antibodies. P values less than 0.05 were considered significant.


Correlation between cytokine induction and ELISA OD values.

When the 65 arthritis patients and 10 controls were investigated in parallel concerning the induction of TNFα, IL-1β, and IL-8, there was a good correlation between ELISA OD values and the production of TNFα (r = 0.862, P < 0.0001) (Figure 2a), IL-1β (r = 0.839, P < 0.0001) (Figure 2b), and IL-8 (r = 0.547, P < 0.0001) (Figure 2c). For the 65 arthritis patients investigated separately, the statistical results were almost the same (for TNFα, r = 0.862, P < 0.0001; for IL-1β, r = 0.836, P < 0.0001; and for IL-8, r = 0.568, P < 0.001). All controls yielded low levels of anti-CII and low levels of IC-induced TNFα, IL-1β, and IL-8. No significant correlations were found when the controls were analyzed separately (data not shown).

Figure 2.

Overall correlation between ELISA OD values and induction of a, TNFα, b, interleukin-1β (IL-1β), and c, IL-8 in a cross-sectional study when solid-phase ICs created with sera from 65 arthritis patients and 10 healthy controls were investigated in parallel. All normal sera and many arthritis sera without elevated anti-CII levels gave somewhat higher OD values in wells with HSA alone compared with CII-containing wells. These individuals therefore show net ELISA values below zero. See Figure 1 for other definitions.

When results were divided between arthritis patients and healthy controls, it became evident that only arthritis patients showed high anti-CII levels (Figure 3a) and high production of TNFα (Figure 3b), IL-1β (Figure 3c), and IL-8 (Figure 3d). We found no difference between patients with established RA and patients with early arthritis with high OD values in anti-CII ELISA, since sera from both groups of patients induced TNFα, IL-1β, and IL-8 after incorporation into solid-phase ICs. A number of patients initially chosen due to anti–denatured CII antibodies did not have antibodies against native CII, and the number of individuals displaying high levels of anti–native CII eventually became lower than expected.

Figure 3.

Distribution between arthritis patients and healthy controls of a, net anti-CII OD values and bd, the solid-phase IC-induced production of TNFα (b), interleukin-1β (IL-1) (c), and IL-8 (d) from 1 single experiment. See Figure 1 for other definitions.

Different threshold anti-CII densities for induction of TNFα from PBMCs obtained from different donors.

When the cytokine responses of PBMCs obtained from 2 different donors were compared concerning their reactivity to solid-phase anti-CII–containing ICs created with sera from the 65 arthritis patients and 10 controls, the correlation between the responses was strong and highly significant (r = 0.817, P < 0.0001). It was obvious, however, that responder cell populations from the 2 donors started to respond with TNFα production at different threshold densities of anti-CII in the solid-phase ICs (Figure 4a). Titration of anti-CII in solid-phase ICs showed that PBMCs obtained from 5 different donors differed in the extent of cytokine responses as well as in threshold anti-CII density (Figures 4b–d).

Figure 4.

Difference in the anti-CII–containing IC density threshold needed for cytokine induction from peripheral blood mononuclear cells (PBMCs) obtained from different donors. a, Effects of anti-CII–containing ICs on TNFα induction in a cross-sectional investigation. Sera from 65 arthritis patients and 10 healthy controls were investigated for anti-CII–induced TNFα production with 2 parallel PBMC responder populations. PBMCs from donor B were more easily stimulated than PBMCs from donor A. bd, Using 1 high-level anti-CII serum titrated into solid-phase ICs as described in Patients and Methods, PBMCs obtained from 5 donors were investigated in parallel concerning their dependency on anti-CII levels in solid-phase ICs for the production of b, TNFα, c, interleukin-1β (IL-1β), and d, IL-8. Anti-CII levels are given using a standardized ELISA with a reference range of <29 units/ml. See Figure 1 for other definitions.

Induction of proinflammatory cytokines from PBMCs by solid-phase ICs containing CII and anti-CII is dependent on monocytes.

When PBMCs were depleted of CD68-positive monocytes (from 8–15% to <0.2%) with mAb using a technique leaving all other PBMCs untouched, solid-phase CII- and anti-CII–containing IC induction of TNFα and IL-1β was almost totally abolished (Figures 5a and b) (<1% remaining cytokine production). Comparable data were obtained when solid-phase anti-CII–containing ICs were used to stimulate PBMCs from 3 different donors. Monocyte depletion also decreased anti-CII–containing IC–induced production of IL-8 to <1% of values obtained in untreated PBMCs (data not shown). These data are consistent with our previous findings of monocyte dependency of TNFα production induced by purified cryoglobulins (35). In control experiments, a normal human serum (NHS) was used instead of a high-titer anti-CII serum for coating the CII surface before addition of responder cells. Concordant with our earlier observations, this surface did not induce significant levels of TNFα, IL-1β, or IL-8, irrespective of whether monocytes were depleted or not (Figures 5a and b; data not shown for IL-8).

Figure 5.

Effects of monocyte depletion (a and b) and enrichment (c and d) on solid-phase ICs containing both CII and anti-CII (anti-CII IC)–induced production of TNFα (a and c) and interleukin-1β (IL-1β) (b and d) from 3 cultures of peripheral blood mononuclear cells (PBMCs) obtained from healthy controls (donors A–C). Values above bars indicate percentages of monocytes in the corresponding cell cultures. For each PBMC donor, the 2 left bars represent results from a solid-phase anti-CII–containing IC created with a high-titer anti-CII serum, whereas the 2 right bars depict the results when a normal human serum (NHS) was used instead of the anti-CII serum. The results for IL-8 (data not shown) were comparable with the results obtained for TNFα and IL-1β. Bars representing monocyte-depleted cultures and NHS controls are not visible in all panels due to unmeasurable cytokine concentrations. However, positions of the bars are the same in all panels. See Figure 1 for other definitions.

In other experiments, PBMCs were enriched for monocytes (from 5–11% to 57–61%) using a technique leaving monocytes unmanipulated before they were stimulated with solid-phase anti-CII–containing ICs. Again, results were uniformly consistent among parallel cultures of PBMCs obtained from 3 donors, demonstrating that monocyte enrichment increased the amount of TNFα (246–432% increase) (Figure 5c), IL-1β (209–620% increase) (Figure 5d), and IL-8 (up to a 492% increase) (data not shown). Control experiments with NHS instead of high-titer anti-CII serum yielded lower cytokine levels, which often increased somewhat with monocyte enrichment (Figures 5c and d; data not shown for IL-8).

Induction of proinflammatory cytokines from PBMCs by solid-phase ICs containing CII and anti-CII is dependent on FcγRIIa, but not FcγRIII.

When responder cells were incubated with Fab or F(ab′)2 fragments of antibodies blocking FcγRIIa and FcγRIII, we obtained uniform results for all 3 cytokines investigated (Figures 6a–c). There was a significant decrease in the production of TNFα (P = 0.015), IL-1β (P = 0.008), and IL-8 (P = 0.008) when FcγRIIa was blocked on PBMCs before addition to solid-phase ICs. Blockade of FcγRIIa decreased the solid-phase anti-CII–containing IC–induced production of TNFα between 30% and 81%, and the production of IL-1β between 48% and 92%. The effect of FcγRIIa blockade on IL-8 production was weaker (a decrease of 14–69%). There was no uniform effect of FcγRIII blockade: TNFα production changed between −23% and 23%, IL-1β production changed between −68% and 640%, and IL-8 production changed between −52% and 382%. As in previous studies (34, 35), we found a nonspecific stimulatory effect of the anti-FcγRIII antibody 3G8, as shown by significant stimulation of TNFα production by control NHS ICs without anti-CII antibodies (Figure 6a).

Figure 6.

Blockade of Fcγ receptor IIa (FcγRIIa) by monoclonal antibody (mAb) IV.3, but not FcγRIII by mAb 3G8, suppresses the production of a, TNFα, b, interleukin-1β (IL-1β), and c, IL-8 induced by solid-phase ICs containing human native CII and anti-CII from patients with rheumatoid arthritis (RA). Left graphs show median cytokine induction from 9 experiments with combinations of 3 different RA sera and 6 different healthy responder peripheral blood mononuclear cell populations. Right graphs show corresponding data from 3 healthy control sera without anti-CII. Monoclonal antibody 3G8 against FcγRIII sometimes induces nonspecific cytokine stimulation. NS = not significant (see Figure 1 for other definitions).


RA is a systemic autoimmune disease that primarily affects diarthrodial joints. CII, the most abundant collagen type in joint cartilage, has been implicated as a possible autoantigen candidate in RA. One source of evidence for this is the demonstration of antibodies to CII in sera and SF from a subpopulation of RA patients (7–11), as well as the spontaneous production of anti–native CII antibodies within RA joints (12–14). Susceptible rats and mice, when immunized with native CII in adjuvant, also develop polyarthritic and partly anti-CII–dependent CIA that has immunologic and pathologic similarities to RA.

In this study, we defined a method to investigate the possibility that surface-bound ICs containing anti-CII antibodies from RA patients may induce production of proinflammatory cytokines such as TNFα, IL-1β, and IL-8 from PBMCs. After defining an appropriate cellular readout technique, we performed a technical study in which we evaluated 65 arthritis patients and 10 healthy controls without anti-CII. We found a strong overall correlation between the IgG anti-CII OD values and IC-induced cytokine induction.

In initial experiments, we also investigated other means of producing artificial CII-containing ICs by combining pure or biotinylated antigen with human anti-CII in fluid phase and thereafter stimulating responder PBMCs with ICs either in soluble form or after binding to streptavidin-coated plates, respectively. Although another human IC system (tetanus toxoid combined with antitetanus hyperimmune serum from tetanus-vaccinated recipients) worked well using soluble antigen, with predictable production of TNFα, this was not the case for anti-CII–containing ICs. Another reason that we abandoned fluid-phase systems was the obvious risk of antibody excess that was evident as a lower TNFα production with high concentrations of antitetanus in the tetanus toxoid experiments (data not shown). We supposed that such effects would make it very difficult to make parallel comparisons of cytokine induction from ICs constructed with different sera.

Cytokine production in the chosen solid-phase IC system is not a general phenomenon working with any antigen–antibody combination. Wells treated with only HSA yielded considerably different OD values in ELISA using different sera, without parallel differences in cytokine induction. When we coated plates with the citrulline-containing peptide cfc9 (42) and used strongly anti–cyclic citrullinated peptide–positive RA sera, we also found very strong ELISA reactivity, but without any corresponding cytokine production in the cell culture system used for anti-CII–containing ICs (data not shown).

In joint inflammation, CII epitopes are exposed to anti-CII due to disruption of the proteinaceous covering that protects the intact cartilage surface (2, 4–6). Prerequisites therefore exist for the formation of cartilage surface–bound anti-CII–containing ICs. The mechanisms elucidated in this study might be responsible for anti-CII–dependent production of proinflammatory cytokines in inflamed joints. In this cross-sectional study, we demonstrate that surface-bound CII-containing ICs can induce the production of cytokines of pathogenetic importance in the rheumatoid joint. This hypothesis is corroborated by Kim et al (15, 16), who reported that RA patients with elevated levels of anti–native human CII have increased levels of TNFα and IL-6, together with higher erythrocyte sedimentation rates and C-reactive protein levels, compared with anti-CII–negative patients. It is evident that both RA and CIA are not purely antibody-mediated diseases. Nonetheless, anti-CII antibodies can transfer self-limiting joint symptoms (21–25) or cartilage destruction (26, 27). Since our hypothesis for the action of anti-CII according to our experimental model is one of a direct action on exposed cartilage epitopes, it is interesting that different investigators of anti-CII–transferred arthritis have also noted an accumulation of transferred anti-CII at the cartilage surface (22) as well as a rapidly evolving pathology directly involving cartilage and without pannus development (27).

In a recent study, we found that production of IL-10 induced by artificial ICs and ICs precipitated from SLE sera (34), as well as TNFα production induced by cryoglobulins associated with lymphoproliferative disease (35), was dependent on FcγRIIa, and that the FcγRII receptor density on monocytes in responder PBMC populations correlated with IC-induced production of IL-10 and IL-6 (34). In the present study, we have parallel findings concerning the effects of FcγRIIa blockade on the IC-induced production of TNFα, IL-1β, and IL-8, with even more unequivocal results (up to 92% blockade of anti-CII–containing IC–induced production of IL-1β). High TNFα levels seem to be induced by ICs from sera with a certain threshold OD value in ELISA, and this threshold differs between PBMC responder populations. We speculate that this might be due to divergent expression of FcγRIIa receptors on cells interacting with the solid-phase CII-containing ICs, and experiments to confirm this hypothesis are in progress. This hypothesis is also consistent with a recent study showing that RA patients have increased expression of FcγRII in the synovium, that RA macrophages have enhanced FcγRII expression and higher production of TNFα after stimulation with artificial ICs, and that matrix metalloproteinase 1 expression in synovium correlated with the expression of FcγRII (43).

Earlier studies have suggested that monocytes are the main human responder cells that produce cytokines following IC stimulation (44). By performing parallel monocyte depletion and enrichment experiments, we have unambiguously demonstrated that this is the case for the production of TNFα, IL-1β, and IL-8 induced by solid-phase ICs containing CII and anti-CII. Obviously, the effect of solid-phase anti-CII–containing ICs on the production of cytokines from RA monocytoid cells compared with healthy control monocytoid cells should be studied and correlated with FcγR density and polymorphisms, and such experiments are currently in progress.

The activating FcγRIIa receptor expressing an immunoreceptor tyrosine–based activation motif is exclusively expressed in higher primates, including humans, and is present on the cell surface of many inflammatory cell types (45). Most animal studies have therefore been performed in species not expressing the FcγRIIa receptor, and instead, they often pinpoint FcγRI, FcγRIII, and FcγRIIb on the surface of macrophages as the important mediators and suppressors of inflammation in IC-mediated rodent arthritis models (46–48). The full expression of rodent CIA is dependent on anti-CII (21, 22, 49) and can only be elicited in susceptible strains with a suitable genetic background. An otherwise nonsusceptible strain has been reported to be susceptible to CIA after transfer of the human FcγRIIa gene (45), further supporting the role of this otherwise primate-restricted receptor in arthritis development. In addition, passive transfer of anti-CII induced more rapid and severe arthritis in FcγRIIa-transgenic mice than in nontransgenic animals, and IC-induced production of TNFα was shown to be predominantly mediated via the transgenic FcγRIIa receptor (50).

Collectively, these data stress the importance of FcγRIIa in human IC–mediated arthritis as well as the importance of human FcγRIIa for the appearance of cytokine-mediated and anti-CII–dependent joint inflammation in humans. It is conceivable that mouse FcγRIII plays a proinflammatory role similar to the primate-specific FcγRIIa, and that the 2 receptors may act synergistically in humans, as has been suggested by others (45).

In the present study, serum samples were chosen on a technical basis with the aim of defining the mechanisms underlying solid-phase anti-CII–containing IC–induced production of proinflammatory cytokines. In a followup study, we have pursued our investigations by first defining a standardized anti-CII ELISA and then screening a defined cohort of patients with early RA for anti–native CII antibodies. RA patients with high-titer anti–native CII have been identified and followed up over time, and changes in levels of antibodies and IC-induced cytokines have correlated with changes in clinical parameters (Mullazchi M, et al: unpublished observations). We have thus presented clinical data supporting the importance of anti-CII for acute inflammation at the time of first appearance of RA.

High levels of antibodies against native CII are present in a minority of RA patients. These antibodies induce the proinflammatory cytokines TNFα, IL-1β, and IL-8 when incorporated into solid-phase ICs. In our model system, cytokine production can be discontinued either by depleting monocyte/macrophages or by blocking the primate-specific FcγRIIa receptor in the PBMC cultures. CII epitopes are exposed for interaction with anti-CII in acutely inflamed joints, and we hypothesize that analogous ICs also form in vivo at the cartilage surface in inflamed joints of RA patients with high anti-CII levels. Macrophage suppression and FcγRIIa blockade therefore represent conceivable treatment options in RA patients with high levels of anti-CII and possibly also in other forms of IC-mediated inflammation.


We are thankful to Professor Lars Klareskog for supplying RA sera. We also thank Associate Professor R. A. Harris for linguistic advice.