B cells play a crucial role as antigen-presenting cells and collaborate with inflammatory cytokines in glucose-6-phosphate isomerase-induced arthritis

Abstract

Anti-glucose-6-phosphate isomerase (GPI) antibodies from K/BxN mice directly induce arthritis; however, the transfer of these antibodies from mice with GPI-induced arthritis does not induce arthritis. CD4⁺ T cells play an important role in the induction and effector phase in this model; however, the roles of B cells and immunoglobulins (Igs) have not been elucidated. We investigated the roles of B cells and Igs in GPI-induced arthritis by using adoptive transfer system into SCID mice. Transfer of splenocytes of male DBA/1 mice immunized with GPI into SCID mice induced arthritis on day 6 in the latter, in association with the production of anti-GPI antibodies. Co-localization of C3 and IgG on the articular surface was identified in arthritic SCID mice. Inoculation of IgG (or anti-GPI antibodies) and CD19⁺-depleted splenocytes from arthritic DBA/1 mice induced arthritis in SCID mice, but not CD19⁺-depleted or CD4⁺-depleted splenocytes from DBA/1 mice. In vitro analysis of cytokine production by splenocytes from DBA/1 arthritic mice demonstrated production of large amounts of tumour necrosis factor (TNF)-α and interleukin (IL)-6 in an antigen-specific manner (P < 0·01), and production was dominated by CD19⁺-depleted than CD4⁺-depleted splenocytes (P < 0·05). Addition of IgG from DBA/1 arthritic mice to the culture enhanced TNF-α but not IL-6 production, and this effect was blocked by anti-Fcγ receptor antibody. In vivo analysis of neutralization with TNF-α protected arthritis completely in SCID mice. Our results highlight the important role of B cells in GPI-induced arthritis as autoantibody producers, and these autoantibodies can trigger joint inflammation in orchestration with inflammatory cytokines, especially TNF-α.

Introduction

Rheumatoid arthritis (RA) is a common chronic autoimmune disease of unknown aetiology characterized by progressive inflammatory process and destruction of joints. Several autoantigens play a role in arthritis [1], and one of the candidate arthritogenic antigens, glucose-6-phosphate isomerase (GPI), was identified in the K/BxN model of arthritis [2]. GPI is a ubiquitous cytoplasmic enzyme, and anti-GPI antibodies in K/BxN mice induce arthritis directly. The effector mechanisms of anti-GPI antibodies have been confirmed by the requirement of innate immune system players, e.g. complement cascade, FcγR, especially FcγRIII, neutrophils and mast cells [3–6]. In addition, GPI accumulates on the synovium and joint articular surfaces, and the formation of a specific immunocomplex on the joint cavity leads ultimately to arthritis in the K/BxN serum transfer model [7]. These results indicate that ubiquitous antigens might be the targets of arthritogenic antibodies.

Recent studies have reported that immunization of DBA/1 mice with human GPI provoked arthritis, supporting the notion that autoimmunity to GPI plays a direct role in arthritis in genetically unaltered mice [8,9]. CD4⁺ T cells were necessary for both the induction and the effector phase of the disease because arthritis was ameliorated by depletion of CD4⁺ T cells with anti-CD4 monoclonal antibodies (mAbs). On the other hand, the role of B cells in this form of arthritis is still obscure. Immunoglobulin (Ig)G purified from arthritic DBA/1 mice did not induce arthritis in naive DBA/1 mice; however, FcγR^−/− mice developed mild arthritis following GPI immunization [8]. Moreover, both B cell-deficient C3H.Q and B10.Q mice are resistant to GPI-induced arthritis [9]. These results suggest that GPI-induced arthritis is B cell-dependent, although it is not clear that these cells are required as autoantibody-producing cells similar to antigen-presenting cells (APCs).

In the present study, we assessed the role of B cells and Igs in GPI-induced arthritis in DBA/1 mice using adoptive transfer into immunodeficient SCID mice. SCID mice were inoculated with splenocytes from GPI-immunized DBA/1 mice plus GPI. They developed arthritis with evident immune complex activation on the articular surface. Splenocytes lacking B and CD4⁺T cells from arthritic DBA/1 mice failed to induce arthritis in SCID mice. SCID mice recipients of both IgG (or purified anti-GPI antibodies) from GPI immunized DBA/1 mice and B cell-depleted splenocytes developed arthritis, whereas SCID mice recipients of IgG (or anti-GPI antibodies) only did not. Moreover, in vitro analysis of splenocytes of arthritic mice showed production of tumour necrosis factor (TNF)-α and interleukin (IL)-6 in an antigen-specific manner, driven mainly by B cell-depleted splenocytes. TNF-α, in particular, was produced mainly by CD11b⁺ cells. In vivo neutralization of TNF-α protected arthritis development of SCID mice completely. These results suggest that B cells play a crucial role as antibody producers, and that antigen-induced cytokine production, especially TNF-α, seems to enhance the development of GPI-induced arthritis.

Materials and methods

Induction of GPI-induced arthritis in DBA/1 mice

Male DBA/1 mice (6–8 weeks old) were obtained from Charles River Laboratories (Yokohama, Japan). Recombinant human GPI was prepared as described previously [10]. Mice (n = 10) were immunized by intradermal injection of 300 µg of recombinant human GPI-gluthathione S-transfererase (GST) (hGPI) in emulsified Freund's complete adjuvant (CFA) (Difco, Detroit, MI, USA). As a control, we immunized another group of DBA/1 mice (n = 10) with 300 µg of GST in CFA. The experimental protocol was approved by the Ethics Review Committee for Animal Experimentation of Tsukuba University.

Arthritic animals were assessed clinically and ankle thickness was recorded. We used the following arthritis scoring system to evaluate the disease state (clinical score): 0 = no evidence of inflammation, 1 = subtle inflammation or localized oedema, 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, yielding a maximum possible score of 12 per mouse.

Human recombinant GPI/GST fusion protein was produced by Escherichia coli with pGEX vector (GE Healthcare, Uppsala, Sweden), as described previously [2]. GPI/GST fusion protein was purified from lysate with gluthathione sepharose 4B (GE Healthcare). The volume of GPI/GST fusion proteins was determined at 280 nm and the purity of proteins checked using standard sodium dodecyl sulphate gels.

Induction of arthritis in SCID mice

CB17/ICR-Prkdc^scid (SCID) mice (8–10 weeks old) were purchased from Charles River Laboratories. The spleens were removed from arthritic DBA/1 mice on day 14 after immunization. The harvested splenocytes were suspended in phosphate-buffered saline (PBS) and erythrocytes were lysed. The remaining cells were washed in PBS, then separated by magnetic affinity cell sorting (MACS; Militenyi Biotech, Bergisch Gladbach, Germany) using anti-CD4⁺ (T cells) or anti-CD19⁺(B cells)-depleted splenocytes, estimated by fluorescence activated cell sorter (FACS) (> 99% cells were depleted). These cells were inoculated intraperitoneally with 100 µg GPI into SCID mice.

Enzyme-linked immunosorbent assay

The enzyme-linked immunosorbent assay (ELISA) microtitre plates were coated with 5 µg/ml rh-GPI in PBS (Sumitomo Bakelite, Tokyo, Japan) overnight at 4°C. The plates were then washed and saturated with 300 µl blocking solution (Dainippon Sumitomo Pharma, Tokyo, Japan) at room temperature. After 2 h, they were washed and 1/500 diluted serum with blocking solution was added. Incubation was carried out for 2 h at room temperature. The plates were washed and 150 µl alkaline phosphatase-conjugated Fc-specific anti-mouse IgG antibody (American Qualex, San Clemente, CA, USA) diluted at 1:5000 with blocking solution was added. After incubation at room temperature for 1 h, the plates were detected with 150 µl of substrate solution (9·6% 2-aminoethanol, 2·4 mM MgCl2 in distilled and deionized water, pH 9·8). Colour development was read by a microplate reader at 405 nm.

Antibody purification

Antibodies were purified from sera of DBA/1 mice immunized with 300 µg rh-GPI or GST. Serum samples were diluted 10-fold with binding buffer and then poured over a protein G column (GE Healthcare, Uppsala, Sweden) to purify IgG. Anti-GPI antibodies were also purified by affinity column (GE Healthcare), following the method described [2]. Purified antibodies were changed buffer to PBS by centricon YM-50 (Millipore, Billerica, MA, USA).

Histological examination

Mice were killed and hind-paw joints were fixed with 4% paraformaldehyde at 4°C for 6 h. The method used for decalcification was described previously [11]. The tissues were then embedded in optimal cutting temperature compound (Miles Scientific, Naperville, IL, USA) and frozen rapidly at −80°C. Frozen sections (5-µm thick) were cut on a cryostat and placed on magnesium aluminum silicate-coated glass microscope slides and allowed to air-dry. Joints were stained with haematoxylin and eosin (H&E) or fluorescent staining. Fluorescent antibodies were anti-C3 fluorescein isothiocyanate (FITC) (ICN Biomedicals, Solon, OH, USA) and anti-IgG Texas Red (EY Laboratories, San Meteo, CA, USA).

In vitro analysis of cytokine production by splenocytes from DBA/1 arthritic mice

Spleens were removed from arthritic GPI-induced mice on day 14. The spleens were harvested and haemolyzed with 0·83% NH₄Cl, 0·12% NaHCO₃ and 0·004% ethylenediamine tetraacetic acid 2Na in PBS. Single-cell suspensions were prepared in RPMI-1640 medium (Sigma-Aldrich, St Louis, MO, USA) supplemented with 10% FBS, 100 U/ml of penicillin, 100 µg/ml of streptomycin and 50 µM of 2-mercaptoethanol. CD4⁺ T cells, CD11b cells, CD11c cells or CD19⁺ cells were isolated and enriched by MACS (Miltenyi Biotech). The cell purity was confirmed by flow cytometry (>90%). Whole splenocytes or MACS-separated cells (1 × 10⁶ cells/ ml) were cultured with 5 µg/ml of GPI (or GST) at 37°C in 5% CO₂ for 12 h. Anti-FcγR II/III receptor antibody (BD Bioscience, San Jose, CA, USA) was used at 1 µg/ml as an Fc blocker. Supernatants were assayed for TNF-α, interferon (IFN)-γ, IL-17 and IL-6 by Quantikine ELISA kit (R&D Systems, Minneapolis, MN, USA) or ELISA Ready-SET-Go! (eBioscience, San Diego, CA, USA).

In vivo analysis using mAb for neutralizing cytokines

We used commercially available anti-TNF-α mAbs (eBioscience) and anti-IL-6 mAbs (R&D Systems) to neutralize the respective cytokines. As a control antibody, we used the same amount of Rat IgG1 isotype control (R&D Systems). In SCID-transferred arthritis, each mouse received a single injection of 100 µg of anti-TNF-α mAb, anti-IL-6 mAb or control Ig was injected on the day of splenocytes transferred (day 0).

Statistical analysis

All data were expressed as mean ± standard error of the mean. Differences between groups were examined for statistical significance using the Mann–Whitney U-test. A P-value less than 0·05 denoted a statistically significant difference.

Results

The GPI-induced arthritis in DBA/1 mice

Arthritis was induced in DBA/1 mice with 300 µg rh-GPI emulsified in CFA. Beginning on day 8 after immunization, the paws and ankles of mice were examined daily for clinical signs of arthritis. Joint swelling reached maximum around day 14, then resolved gradually (Fig. 1a). Arthritic changes were observed mainly in the paws (Fig. 1b, right) and ankles of immunized DBA/1 mice, but not in digits. Control (GST) immunization did not lead to apparent arthritis (Fig. 1b,c, left). Histopathological examination showed synovium proliferation (Fig. 1c, right), resulting in bone destruction (data not shown). Immunohistochemical analysis confirmed co-localization of IgG and C3 on the surface of cartilage on day 14 in arthritic DBA/1 mice (Fig. 1d right; control immunization on left). These findings suggest that immune complex activation in local joints is involved in the development of GPI-induced arthritis.

Fig 1.

Clinical and histological evaluation of glucose-6-phosphate isomerase (GPI)-induced arthritis in DBA/1 mice. Mean of clinical score (a) (±standard error of the mean, 10 mice) followed days after immunization. (b) Paw of control DBA/1 mouse treated with control antigens [gluthathione S-transfererase (GST)] 300 µg (left). DBA/1 mice were immunized with rh-GPI 300 µg in Freund's complete adjuvant (CFA) (right). (c) Histological examination of ankle joints of the control (left) and GPI-induced arthritis on day 14 showing severe synovium proliferation (haematoxylin and eosin staining, right). (d) Anti-C3 (green) and anti-immunoglobulin (Ig)G (red) staining in joints of control (left) and arthritic DBA/1 mice (right). Nuclei were counterstained with 4,6-diamino-2-phenylindole (blue). C3 and IgG were co-localized on the surface of cartilage of ankle joints (right). Magnification of original photographs: ×40 (c) or ×600 (d); spl: splenocytes.

Successful transfer of GPI-induced arthritis into immunodeficient mice

Splenocytes (1 × 10⁷ cells) from arthritic DBA/1 mice were inoculated into SCID mice on day 14 post-immunization with 100 µg of GPI. Spleens from control SCID mice (Fig. 2a left) or SCID mice inoculated with splenocytes (1 × 10⁷ cells) from arthritic DBA/1 mice (Fig. 2a right, on day 14) was shown. Arthritis developed in splenocyte-inoculated SCID mice (Fig. 2b right, c). However, arthritis was not observed in both SCID mice inoculated with the same number of splenocytes from arthritic DBA/1 mice without GPI, and SCID mice inoculated with splenocytes from naive DBA/1 mice with GPI (Fig. 2b left, c) These results indicate that splenocytes from arthritic mice plus GPI contain important factor(s) in the induction of arthritis.

Fig 2.

Transfer of arthritis in SCID mice. Glucose-6-phosphate isomerase (GPI)-immunized DBA/1 mice were killed on day 14, and 1 × 10⁷ splenocytes (spl) were isolated and transferred into SCID mice with 100 µg of GPI. (a) Spleen of control SCID mice (left) and SCID mice inoculated with splenocytes of arthritic DBA/1 mice 1 × 10⁷ cells (right on day 14). (b) Feet of SCID mice inoculated with splenocytes from naive DBA/1 mice (left) and SCID mice inoculated with splenocytes from GPI-induced arthritic DBA/1 mice (right). (c) Clinical score and development of arthritis in SCID mice. Swelling of paws was observed on day 6 in SCID mice inoculated with whole splenocytes plus 100 µg of GPI (♦). Control mice were inoculated with 1 × 10⁷ splenocytes only from immunized DBA/1 mice (▵) or 1 × 10⁷ splenocytes from control immunized DBA/1 mice plus 100 µg of GPI (○). Control mice did not develop arthritis. Data are mean ± standard error of the mean of five mice in each group. *P < 0·05 by Mann–Whitney U-test.

Histological analysis of arthritic SCID mice

Histopathological examination of the arthritic joints of SCID mice showed synovial hyperplasia in arthritic SCID mice inoculated with splenocytes from arthritic DBA/1 mice (H&E staining, Fig. 3b,c), but not in SCID mice inoculated with splenocytes from naive DBA/1 mice (Fig. 3a). Immunohistochemical study showed co-localization of IgG and C3 on the cartilage surface of arthritic SCID mice (Fig. 3d), but not in joints of SCID mice inoculated with splenocytes from naive DBA/1 mice. These findings suggest that Igs produced by inoculated splenocytes from DBA/1 mice attach to the articular surface of SCID mice and result in arthritis by complement activation.

Fig 3.

Histological evaluation of joints of SCID mice. Joints of control mice inoculated splenocyte (spl) from naive DBA/1 mice with 100 µg of glucose-6-phosphate isomerase (GPI). (b,c) Synovial hyperplasia in a representative SCID mouse inoculated with splenocytes from arthritic DBA/1 mice. (d, right) Co-localization of immunoglobulin (Ig)G (red) and C3 (green) on the articular surface of SCID mice on day 14 after transfer by fluorescent staining. Nuclei were counterstained with 4,6-diamino-2-phenylindole (blue). Magnification of original photographs: ×40 (a), ×100 (b, c) or ×600 (d).

Importance of T and B cells in arthritis of SCID mice

To evaluate the role of CD19⁺ or CD4⁺ cells in arthritis of SCID mice, we inoculated 1 × 10⁷ CD19⁺- or CD4⁺-depleted splenocytes of arthritic DBA/1 mice plus 100 µg of GPI into SCID mice. In these inoculi, the percentage of CD19⁺ and CD4⁺ cells in depleted splenocytes was less than 1%. Neither CD19⁺ nor CD4⁺ cell-depleted splenocytes induced arthritis in SCID mice (Fig. 4a), or produced anti-GPI antibodies (Fig. 4b), suggesting that both CD19⁺ and CD4⁺ cells play important roles in the induction of arthritis in SCID, and that production of anti-GPI antibodies may be indispensable for such induction.

Fig 4.

Importance of anti-GPI antibodies in transfer of arthritis. CD19-depleted or CD4⁺-depleted splenocytes (spl) from arthritic DBA/1 mice obtained on day 14 after immunization were inoculated with glucose-6-phosphate isomerase (GPI) into SCID mice. (a) Mean clinical score. (b) Anti-GPI antibodies detected by enzyme-linked immunosorbent assay (ELISA) at 405 nm (b). (♦) SCID mice that received 1 × 10⁷ of splenocytes from arthritic DBA/1 mice plus 100 µg GPI; (▪) SCID mice recipients of 1 × 10⁷ CD4⁺-depleted cells plus 100 µg GPI; (○) SCID mice recipients of 10⁷ CD19-depleted cells plus 100 µg GPI. Data are mean ± standard error of the mean of five mice in each group. *P < 0·05 by Mann–Whitney U-test.

Importance of B cells as producers of antibodies in arthritis of SCID mice

It has been reported previously that B cell-deficient mice are resistant to GPI-induced arthritis [9]. However, whether these cells act as autoantibody-producing cells as well as APCs is unknown at present. To investigate the role of autoantibodies, we inoculated SCID mice with IgGs from arthritic DBA/1 mice. IgGs were purified from the sera of DBA/1 mice on day 14 after immunization. Injection of 3 mg IgG alone from arthriticDBA/1 mice did not result in overt arthritis in SCID mice, even if we added 100 µg of GPI (Fig. 5a). However, injection of IgG with CD19⁺-depleted splenocytes and GPI resulted in the development of arthritis in SCID mice (Fig. 5a). To investigate further the arthritogenicity of anti-GPI antibodies, we used affinity purified anti-GPI antibodies from arthritic DBA/1 mice. Injection of 3 mg anti-GPI antibodies alone did not result in arthritis in SCID mice, even if we added 100 µg of GPI (Fig. 5b). However, with CD19⁺-depleted splenocytes, even if we used 1mg of affinity purified anti-GPI antibodies from GPI-induced mice instead of IgG, clear arthritis was developed in SCID mice (Fig. 5b). These findings suggest that CD19⁺ cells play an important role as producers of antibody (especially anti-GPI antibodies) in arthritis of SCID mice; however, anti-GPI antibodies alone from GPI-induced arthritis do not have arthritogenecity.

Fig 5.

Role of B cells in induction of arthritis in SCID mice. IgG from arthritic DBA/1 mice [alone or with glucose-6-phosphate isomerase (GPI)] or immunoglobulin (Ig)G plus CD19⁺-depleted cells were inoculated into SCID mice. Development of arthritis in SCID mice was monitored. (a) Mean clinical score is depicted. (♦) SCID mice that received 1 × 10⁷ splenocytes (spl) plus 100 µg GPI; (□) SCID mice recipients of 1 × 10⁷ CD19-depleted cells from arthritic DBA/1 mice with 3 mg of IgG plus 100 µg of GPI; (▵) SCID mice recipients of 3 mg of IgG from arthritic DBA/1 mice plus 100 µg GPI; (•) SCID mice recipients of 3 mg of IgG from arthritic DBA/1 mice alone. (b) Affinity purified anti-GPI antibodies from arthritic DBA/1 mice (alone or with GPI), or anti-GPI antibodies plus CD19⁺-depleted cells were inoculated into SCID mice and monitored. (♦) SCID mice that received 1 × 10⁷ spenocytes plus 100 µg GPI; (□) SCID mice recipients of 1 × 10⁷ CD19-depleted cells from arthritic DBA/1 mice with 1 mg of anti-GPI antibodies plus 100 µg of GPI; (▵) SCID mice recipients of 3 mg of anti-GPI antibodies from arthritic DBA/1 mice plus 100 µg GPI; (•) SCID mice recipients of 3 mg of anti-GPI antibodies from arthritic DBA/1 mice alone. Data are mean ± standard error of the mean of five mice in each group. *P < 0·05 by Mann–Whitney U-test.

Importance of TNF-α in the development of arthritis in SCID mice

To determine the humoral factors that were mediated by arthritis with splenocytes from GPI-induced arthritis plus GPI in SCID mice, we screened in vitro cytokine production from splenocytes plus GPI. We selected two proinflammatory cytokines in these experiments based on the preliminary results of cytometric beads array analysis, which revealed antigen-specific expression of TNF-α and IL-6 (data not shown); they have recently proved to be important in the induction of GPI-induced arthritis [12]. Indeed, the addition of GPI to the culture medium induced the production of large amounts of TNF-α and IL-6, while control antigen did not induce these cytokines (Fig. 6a). We also examined the production of these cytokines by CD19⁺- and CD4⁺-depleted cells. TNF-α and IL-6 levels were enhanced in the presence of CD19⁺-depleted cells compared with CD4⁺-depleted cells (Fig. 6a), and enriched slightly in CD19⁺-depleted cells compared with whole splenocytes. To examine the role of IgG from DBA/1 arthritic mice, CD19⁺-depleted splenocytes were stimulated with GPI and/or IgG in vitro. IgG triggered weak production of TNF-α and Fcγ blockade suppressed TNF-α production (Fig. 6b, P < 0·05). On the other hand, IL-6 production was regulated by neither IgG nor Fcγ blockade (Fig. 6b). To confirm the dependency of these inflammatory cytokines of arthritis in SCID mice, neutralizing mAbs were injected in vivo on the day of inoculation of splenocytes. Surprisingly, anti-TNF-α mAb protected arthritis completely in SCID mice, whereas anti-IL-6 mAb blocked arthritis partially (Fig. 6c). These findings suggest that TNF-α in particular (partially IL-6) induced by GPI may contribute to the development of arthritis, although IgG from arthritic mice contributed weakly to the production of TNF-α via Fcγ receptors.

Fig 6.

In vitro cytokine production by splenocytes from arthritic DBA/1 mice and in vivo neutralization of inflammatory cytokines in SCID mice. Cytokine concentrations in supernatant of cultured splenocytes (spl) from arthritic DBA/1 mice were assessed by enzyme-linked immunosorbent assay. (a) Whole splenocytes or separated splenocytes (10⁶ cell/ml) were cultured with 5 µg/ml of glucose-6-phosphate isomerase (GPI) or gluthathione S-transfererase (GST). (b) CD19⁺-depleted splenocytes were cultured with GPI and immunoglobulin (Ig)G with/without FcγR blocker. IgG was purified from arthritic DBA/1 mice on day 14 after GPI immunization. Naive IgG (nIgG) was purified from naive DBA/1 mice. Inhibition rate was calculated to be divided by amount of productions from CD19⁺-depleted splenocytes stimulated with GPI. Representative data of three independent experiments with three individual mice per experiment. (c) Neutralization of inflammatory cytokines was performed in vivo by monoclonal antibody (mAb), five mice in each group. SCID mice recipients of 1 × 10⁷ splenocytes from arthritic DBA/1 mice plus 100 µg GPI and 100 µg of control IgG (◊), anti-tumour necrosis factor (TNF)-α mAb (▪) or anti-interleukin (IL)-6 mAb (*). Data are mean ± standard error of the mean of five mice in each group. *P < 0·05, ^†P < 0·01, by Mann–Whitney U-test.

CD11b⁺ cells collaborating with CD4⁺ T cells produce predominantly TNF-α

To analyse further the dominant cell populations that can produce TNF-α and IL-6, MACS-separated cells were co-cultured with GPI or GST (Fig. 7a,b). TNF-α was produced by several cell populations, driven mainly by CD11b⁺ cells (Fig. 7a). It is possible that TNF-α production from CD11b cells was induced by the collaboration of activated T cells containing CD11b when cultured with GPI. On the other hand, IL-6 was produced predominantly by CD11c⁺ cells (Fig. 7b). This cytokine production was enhanced by adding CD4⁺ cells (P < 0·05), thus CD4⁺ T cells might also contribute for producing inflammatory cytokines.

Fig 7.

Exploration of dominant cell population of inflammatory cytokines, and B cell functions as an antigen-presenting cells (APCs) in vitro. Whole splenocytes (spl) or independent magnetic affinity cell sorting (MACS) separated (CD4⁺, CD19⁺, CD11b⁺ and CD11c⁺ cells) splenocytes (total 1 × 10⁶ cells/ml) were cultured with 5 µg/ml of glucose-6-phosphate isomerase (GPI) or gluthathione S-transfererase (GST). Inflammatory cytokines [tumour necrosis factor (TNF)-α (a) and interleukin (IL)-6 (b)] and T cell-secreted cytokines [IL-17 (c) and interferon (IFN)-γ (d)] were compared between CD19⁺ cells and other APCs (CD11b⁺ CD11c⁺ cells). Cytokine concentrations in supernatant of cultured splenocytes from arthritic DBA /1 mice were assessed by enzyme-linked immunosorbent assay. The purity of cells was estimated by fluorescence activated cell sorter flow cytometry (> 90%). Data are mean ± standard error of the mean of three mice in each group. *P < 0·05, ^†P < 0·001, by Mann–Whitney U-test.

Exploring antigen-presenting function of B cells

Finally, to evaluate B cell function as APCs, MACS-separated CD4⁺ T cell and CD19⁺ cells were co-cultured with GPI. IFN-γ and IL-17 production were used to indicate the barometer of the antigen presentation function of B cells. Both IFN-γ and IL-17 were up-regulated clearly by adding CD19⁺ splenocytes (P < 0·05), indicating that CD19⁺ cell may function as APCs (Fig. 7c,d).

Discussion

Anti-GPI antibodies from K/BxN mice are well known as arthritogenic autoantibodies, and their effector mechanisms have been identified in several elegant studies [2–7]. Briefly, the key players involved in the development of arthritis after the transfer of anti-GPI antibodies included Fcγ receptor (particularly FcγRIII), alternative complement pathways such as factors B, C3, C5 and C5aR [3], subsets of Fcγ receptor or C5a receptor-bearing cells [4–6] and some inflammatory cytokines such as IL-1 and TNF-α[3]. In particular, a dominant pathological action driven by anti-GPI antibodies is a local association between GPI and anti-GPI on the articular surface, which leads to complement activation in the joints [7,13].

However, anti-GPI antibodies from GPI-induced arthritis did not induce overt arthritis in naive mice [8]. A previous report showed that B cell-deficient C3H.Q and B 10.Q mice were resistant to GPI-induced arthritis [9]. Moreover, FcγR-deficient mice were protected from GPI-induced arthritis, whereas mice deficient in inhibitory FcγRIIB developed severe arthritis [8]. These results show that B cells play an essential role in arthritis by producing autoantibodies that result in FcγR activation in this model. In our immunohistological study, a clear complement activation by immune complex was observed in joints of mice with GPI-induced arthritis. This finding suggests that local immune complex (probably GPI–anti-GPI antibodies) activation in the joints also plays an important role in GPI-induced arthritis.

To determine the role of B cells more precisely, we set up a transfer system using SCID mice. SCID mice inoculated with splenocytes from GPI-immunized DBA/1 mice together with GPI developed arthritis, and the immune complex activation was also noted on the articular surface of SCID mice. In GPI-induced arthritis, the expression of TNF-α mRNA in inflammatory joints and serum was increased on day 7 when detectable amounts of anti-GPI antibodies were produced (reference [12] and our unpublished data). B cell-depleted splenocytes from arthritic DBA/1 mice could not induce arthritis in SCID mice. On the other hand, SCID mice that received IgG (or anti-GPI antibodies) with B cell-depleted splenocytes from GPI-immunized DBA/1 mice developed arthritis, whereas SCID mice that received only IgG did not. These results suggest that B cells play a crucial role as antibody producers, followed by immune complex deposition on the articular surfaces in arthritis.

Our scenario is similar to adoptive transfer of collagen-induced arthritis (CIA) to SCID mice [14–16]. However, the GPI-induced arthritis in SCID mice occurred earlier (5–6 days) than CIA (14–16 days). The other difference between these two types of arthritis is that IgG from GPI-induced arthritis did not have arthritogenic capacity, whereas CIA IgG alone exhibit such capacity. Thus, anti-GPI antibodies produced by inoculated splenocytes play an important role in the induction of arthritis. However, these antibodies could not induce arthritis when injected alone, and thus we need to know about other humoral factors that trigger arthritis.

Our in vitro assay with splenocytes from GPI-induced arthritis plus GPI indicated that TNF-α and IL-6 may be crucial for the induction of arthritis. An earlier study from our laboratories identified the therapeutic efficacy of both anti-TNF-α mAb and anti-IL-6 mAb in GPI-induced arthritis [12]. Moreover, we clearly confirmed a protective effect of anti-TNF-α mAb in SCID-transferred arthritis. These results indicate that arthritis in SCID recipients may be enhanced mainly not only by anti-GPI antibodies, but also humoral factors such as TNF-α and IL-6. In particular, the development of arthritis was associated with the production of anti-GPI antibodies in SCID mice, thus autoantibodies might play a crucial role especially in the local joints, collaborating with inflammatory cytokines.

Concerning the other role of B cells, our in vitro assay suggests that B cells had a weak capacity of producing TNF-α, as well as antigen-presenting function with GPI culture. A recent paper reported that subsets of dendritic cells (DC) could express CD19 [17], thus it is possible that these cells comprise such functions of B cells. However, we tested in vivo experimentally with CD19⁺-depleted cells, suggesting that the antoantibody produced indeed contributed to the development of arthritis.

What is the role of T cells in GPI-induced arthritis? Based on our experiments, splenocytes lacking CD4⁺ cells failed to induce arthritis in SCID mice. The lack of anti-GPI antibodies in the serum of SCID recipients of the CD4⁺T cell-depleted cell population suggests that production of autoantibodies is CD4⁺ T cell-dependent. Moreover, our in vitro assay identified CD19⁺-depleted cells (probably comprising APCs plus T cells) as the main source of inflammatory cytokines that can trigger arthritis. TNF-α and IL-6 production was enhanced by adding CD4⁺ cells, as confirmed by in vitro assay. In this regard, in GPI-induced arthritis, administration of anti-CD4 mAb on days 11 and 14 after immunization induced rapid remission of the arthritis [8]. We have reported previously that GPI-specific CD4⁺ T cells were differentiated to T helper type 1 (Th1) and Th17 [18]. The administration of anti-IL-17 mAb on day 7 ameliorated arthritis significantly, whereas that administered on day 14 did not affect the disease. Moreover, our vitro assay using splenocytes on day 14 could detect tiny amounts of IL-17 with GPI ([17], and our unpublished data). These findings show that CD4⁺ T cells (particularly Th17 cells) are necessary in the induction phase, and they function as supporters of production with autoantibodies and inflammatory cytokines in the effector phase of GPI-induced arthritis.

Are these scenarios relevant to human RA? High titres of anti-GPI antibodies are found in patients with severe forms of RA, but in only a few control individuals [10,19,20]. We reported recently that a FCGR3A-158V/F functional polymorphism was associated with RA in anti-GPI antibody-positive individuals, because 89% of healthy subjects positive for anti-GPI antibodies possessed homozygous low-affinity genotype FCGR3A-158F [21]. Moreover, among anti-GPI antibody-positive individuals, GPI-reactive CD4⁺ T cells, especially Th1 cells, are detected specifically in peripheral blood mononucleocytes of patients with RA who share either human leucocyte antigen (HLA)-DRB1 *0405 or *0901 haplotypes [22]. These findings suggest that arthritis in anti-GPI antibody-positive individuals depends on several important factors, such as GPI-reactive T cells, HLA-DR*0405/*0901 and FcγRIII.

What of the role of anti-GPI antibodies in GPI-induced arthritis? The H2^q haplotype confers severe form of arthritis [9]. High titres of anti-GPI antibodies were also found in arthritis-resistant C57BL/6(H2^b) mice, although their T cells had weak GPI responses ([8], and our observations) compared with arthritis-susceptible DBA/1 mice. In addition, FcγR^−/− mice are protected from GPI-induced arthritis, whereas FcγRIIB^−/− mice developed pronounced arthritis [8]. These findings indicate that anti-GPI antibodies do not induce arthritis per se; it is probable that unique activation of major histocompatibility complex class II and antigen-specific T cells might be indispensable. In this regard, GPI-induced arthritis appears to be akin to human RA.

In conclusion, we identified that B cells play a crucial role in GPI-induced arthritis as autoantibody producers. This finding might explain how autoantibodies orchestrate the induction of arthritis with inflammatory cytokines such as TNF-α in patients with RA.