Male DBA/1 mice (aged 6 to 8 weeks) were obtained from Charles River (Yokohama, Japan). Recombinant human GPI was prepared as described previously 14. Mice were immunized by intradermal injection of 300 μg recombinant human GPI-glutathione S-transferase (GST) fusion protein (hGPI) in emulsified complete Freund's adjuvant (Difco, Detroit, MI, USA). Control mice were immunized with 300 μg GST in complete Freund's adjuvant. The experimental protocol was approved by the Ethics Review Committee for Animal Experimentation of Tsukuba University School of Medicine. Arthritic animals were clinically assessed and ankle thickness recorded. We used the following arthritis scoring system to evaluate the disease state (clinical index): 0 = no evidence of inflammation, 1 = subtle inflammation or localized edema, 2 = easily identified swelling but localized to either dorsal or ventral surface of paws, and score 3 = swelling on all aspects of paws. All four limbs were evaluated using a constant tension caliper and graded, yielding a maximum possible score of 12 per mouse.

At the indicated time points, the ankles of the mice were removed, fixed, decalcified and paraffin-embedded. Sections (5 μm) were stained with hematoxylin and eosin, and evaluated for histologic changes indicating inflammation, pannus formation, and cartilage and bone damage.

Spleens were dissected from immunized DBA/1 or B6 mice (on day 8 after immunization) and immediately immersed in phosphate-buffered saline (PBS; Gibco, Grand Island, NY, USA). Single-cell suspensions were prepared. Red blood cells were lysed by incubation of the suspension in NH₄Cl (0.83% in 0.01 mol/l Tris-HCl [pH 7.2]). The number of splenocytes was then counted, centrifuged again, and resuspended in RPMI (Gibco, Grand Island, NY, USA). For culture, we used RPMI supplemented with 100 μg/ml streptomycin, 100 U/ml penicillin, 10% fetal bovine serum, and 50 μM 2-mercaptoethanol. After counting the cells, the medium was added to make the final concentration 2.5 × 10⁶/ml. Next, CD4⁺ T cells were isolated by positive selection with anti-mouse CD4⁺ antibody (T cell isolation kit; Miltenyi Biotec, Bergisch Gladbach, Germany). The labeled cells were then passed through separation columns (MiniMACS columns; Miltenyi Biotec). The cells contained more than 97% CD4⁺ T cells. T-depleted spleen cells were treated with 50 μg/ml mitomycin C (Kyowa Hakko Kogyo, Tokyo, Japan) for 30 minutes at 37°C and were used as antigen-presenting cells (APCs).

CD4⁺ T cells (1 × 10⁶ cells/ml) were stimulated with 5 μg/ml GPI (or GST) and APCs (2 × 10⁵ cells/ml) in 1 ml volume in 48-well culture plates (Nunc) for 12 hours. The culture supernatants were collected and cell-free samples were stored at -30°C until the cytokine assay. The concentrations of TNF-α, IFN-γ, IL-2, IL-4, IL-5, IL-6, IL-10, and IL-12p70 were measured using cytometric bead array (CBA) with a series of anti-cytokine mAb-coated beads and PE-conjugated anti-cytokine mAbs, followed by Epics XL flow cytometric analysis (Beckman-Coulter Electronics, Fullerton, CA, USA), using the CBA kit (BD Bioscience, San Jose, CA, USA) and software (BD).

We used commercially available anti-TNF-α mAb (eBioscience, San Diego, CA, USA; 10 μg/ml), anti-IFN-γ mAb (BD Biosciences; 1 μg/ml), and anti-IL-12 mAb (BD; 0.3 μg/ml) to neutralize the respective cytokines. These concentrations were selected based on more than 80% blockade of the respective cytokine. CTLA-4Ig (BD; 1 μg/ml), anti-inducible co-stimulator (ICOS) mAb (BD; 0.5 μg/ml), and anti-CD40L mAb (BD; 1 μg/ml) were used to block co-stimulatory pathways. As a control antibody, we used the same amount of rat IgG₁ isotype control (R&D Systems, Minneapolis, MN, USA). Inhibition study was conducted by adding the above concentration at the start of culture. Three independent experiments were performed.

On day 8 after the onset of arthritis, each mouse received a single injection of 100 μg of anti-TNF-α mAb, anti-IL-12 mAb, anti-IFN-γ mAb or anti-IL-6 mAb. A single injection of anti-IL-6 mAb on day 14 was also administered. On the other hand, two injections of 100 μg CTLA-4Ig were administered on days 8 and 12, or on days 12 and 16 after the onset of arthritis.

Serum samples were collected at the indicated time points. The serum levels of TNF-α, IL-6, IL-1β and IFN-γ were determined with the respective enzyme-linked immunosorbent assay kits (BD). To detect the levels of anti-GPI antibodies, we used hGPI and GST at 5 μg/ml (diluted in PBS) to coat microtiter plates (12 hours, 4°C). After washing twice with washing buffer (0.05% Tween 20 in PBS), Block Ace (diluted 1/4 in 1 × PBS; Dainippon Pharmaceuticals, Osaka, Japan) was used for saturation (2 hours at room temperature). After two washes, sera (diluted 1/500) were added and the plates incubated for 2 hours at room temperature. After washing, alkaline phosphatase (AP)-conjugated anti-mouse IgG (Fc-fragment specific; Jackson Immunoresearch Laboratories, West Grove, PA, USA) was added to the plate (dilution 1/5,000, 1 hour, room temperature). After three washes, color was developed with AP reaction solution (containing 9.6% diethanolamine and 0.25 mmol/l MgCl₂ [pH 9.8]) with AP substrate tablets (Sigma Chemical Co., St. Louis, MO, USA; one AP tablet per 5 ml AP reaction solution). Plates were incubated for 30 minutes at room temperature and the optical density was measured by plate spectrophotometry at 405 nm. Determinations were conducted in triplicate, and standardized between experiments by reference to a highly positive mouse anti-GPI serum. The primary reading was processed by subtracting optical density readings of control wells (coated with GST for hGPI).

All data were expressed as mean ± standard error of the mean. Differences between groups were examined for statistical significance by using Mann-Whitney's U test. P < 0.05 denoted the presence of a statistically significant difference.

To investigate whether our own GPI immunization procedure can induce arthritis, we immunized DBA/1 mice using human recombinant GPI prepared in our laboratories. As reported previously 91015, all mice developed arthritis after immunization with 300 μg recombinant GPI. Arthritis appeared at day 8, and severe arthritis was noted at day 14, with maximum ankle swelling on day 14 (data not shown). GST immunization did not induce apparent arthritis (data not shown).

To identify the dominant cytokines at the onset of antigen-induced arthritis (day 8), we established the CBA array system using spleen CD4⁺ T cells plus mitomycin-treated APCs cultured in GPI. In this system, treatment of APCs with mitomycin is designed to kill autoreactive APCs. The results demonstrated the production of large amounts of TNF-α and IFN-γ by the spleen of arthritic mice (Figure 1). In contrast, cells cultured with control antigen (GST) instead of GPI did not produce these cytokines (Figure 1). APC plus antigen alone produced such amounts of cytokines. Very low but detectable levels of IL-2 and IL-6 were produced, but almost no T-helper-2 type cytokines (such as IL-4, IL-5, and IL-10) were detected (Figure 1). These results indicate that exposure to the GPI antigen results in induction of TNF-α and IFN-γ by immunocytes, and suggest that these cytokines could play a crucial role in the induction of arthritis in GPI-induced mice.

To delineate the separate contributions of TNF-α and IFN-γ, we performed blocking experiments using neutralizing mAbs for anti-TNF-α, IFN-γ, and IL-12 using the CBA array system. TNF-α production was inhibited by anti-TNF mAb (64.7 ± 2.7%; Figure 2a), but not by anti-IL-12 mAb (0%; Figure 2a). On the other hand, IFN-γ production was inhibited by anti-IFN-γ mAb (82.5 ± 1.2%; Figure 2b) as well as by anti IL-12 mAb (67.5 ± 2.5%; Figure 2b), and weakly by anti-TNF-α mAb (17.2 ± 9.2%; Figure 2b). These results suggest that TNF-α production is not regulated by IFN-γ, although IFN-γ is partially regulated by TNF-α.

To determine the effect of co-stimulatory molecules in established arthritis, we conducted the same in vitro experiments by using CTLA-4Ig, anti-ICOS, and anti-CD40L mAbs. CTLA-4Ig suppressed TNF-α (18 ± 2.1%; Figure 2c), and IFN-γ (42.9 ± 2.1%; Figure 2d) production, but not anti-ICOS or anti-CD40L mAb. These findings suggest that the antigen-induced cytokines are mainly driven by CD28/B7-1,2 co-stimulator.

To identify the pathogenic cytokine that can provoke the onset of arthritis, we conducted in vivo experiments using neutralizing mAbs. A single injection of 100 μg of anti-TNF-α mAb at day 8 ameliorated the disease (Figure 3a). In contrast, injection of the same dose of anti-IFN-γ or anti-IL-12 mAb had no such effect on the course of the disease, but rather tended to exacerbate the arthritis (Figure 3b). Histopathological examination of the joints of treated mice showed a clear therapeutic effect for anti-TNF-α mAb (Figure 3f, on day 21) as compared with that of control antibody (Figure 3g, on day 21). These results suggest that TNF-α blockade has clear therapeutic effect in GPI-induced model, irrespective of the minor role of 'conventional' T-helper-1 autoimmunity.

To investigate the effect of CTLA-4Ig in vivo, we treated arthritic mice with CTLA-4Ig on days 8 and 12, or on days 12 and 14. A marked improvement was seen after the second treatment (on days 8 and 12), probably because of a reduction in effector T cells at that stage (Figure 4c, and hematoxylin and eosin staining on day 21 in Figure 4h). Moreover, if we admininstered treatment on days 12 and 16, clear therapeutic efficacy was observed after the first treatment. This finding suggests that CTLA-4Ig is also therapeutically potent, especially on day 12, in mice with GPI-induced arthritis.

IL-6 is also an important cytokine in arthritis, and it is considered a promising target for the treatment of RA 78. Serum IL-6 concentrations were elevated in arthritic mice, especially during the disease effector phase (Figure 4). In the next step, we assessed the effect of IL-6 blockade in mice with GPI-induced arthritis. Surprisingly, anti-IL-6 treatment on day 8 resulted in improvement in the clinical index (Figure 3e), although treatment on day 14 had no effect on the course of the disease (data not shown), suggesting that IL-6 is also pathologically crucial in the early effector phase in arthritis.

To determine the effects of inflammatory cytokines during the effector phase of arthritis, we measured the serum concentrations of TNF-α, IL-6, IL-1β, and IFN-γ at days 0, 7, 14, and 28 in DBA/1 mice after GPI immunization. Serum TNF-α concentration was upregulated at disease onset (day 7), but gradually decreased to the basal level by day 28 (Figure 4). On the other hand, serum IL-6 concentration was upregulated gradually, especially during the disease effector phase (days 7 and 14; Figure 4). In contrast, serum IL-1β and IFN-γ concentrations were persistently low and below the detection limit (4 pg/ml) in GPI-induced mice throughout the study (Figure 4). These findings suggest a systemic TNF-α/IL-6 imbalance in arthritic mice.

Anti-GPI antibodies have potent arthritogenic capacity in K/B × N mice. However, anti-GPI antibodies from mice with GPI-induced arthritis do not solely cause arthritis (Schubert and coworkers 11 and our preliminary observations). In GPI-induced arthritis, IgG and C3 are co-localized on the articular surface of arthritic joints (Tanaka and coworkers, unpublished data). Accordingly, we compared the effects of anti-cytokine mAbs, immunomodulatory molecule CTLA-4Ig, and control immunoglobulin on the production of anti-GPI antibodies in mice with GPI-induced arthritis. The antigen was injected on day 8, and then sera were collected on day 14. As shown in Figure 5, anti-TNF-α, anti-IL-6, and CTLA-4Ig tended to suppress the production of anti-GPI antibodies, whereas IL-12 mAb slightly enhanced the production of the antibodies. These findings suggest that effective treatments might also alter autoantibody production during this phase of GPI-induced arthritis.