Interferon-γ regulates susceptibility to collagen-induced arthritis through suppression of interleukin-17




The enhanced expression of experimental arthritis in the absence of interferon-γ (IFNγ) suggests that IFNγ suppresses arthritis. Interleukin-17 (IL-17) is a pivotal T cell cytokine in arthritis, and in vitro studies have indicated that IFNγ suppresses IL-17 production. We undertook this study to test the hypothesis that resistance to collagen-induced arthritis (CIA) in C57BL/6 (B6) mice is regulated by IFNγ-mediated suppression of IL-17.


Wild-type (WT) B6 mice, IFNγ-knockout (KO) B6 mice, and DBA/1 mice were immunized with type II collagen (CII) in Freund's complete adjuvant (CFA). Lymphocytes from immunized mice were analyzed for cytokine production ex vivo by intracellular staining or restimulation with CII and enzyme-linked immunosorbent assays. In vivo blockade of IL-17 was achieved with an anti–IL-17 monoclonal antibody (mAb).


CII restimulation of T cells from CII/CFA-immunized mice resulted in an ∼5-fold increase in IL-17 production in IFNγ-KO B6 mice compared with WT B6 mice. Neutralization of IFNγ increased IL-17 production in WT B6 mice, and neutralization of IL-4 had a synergistic effect. Interestingly, the prototypical CIA-susceptible strain DBA/1 also demonstrated a high IL-17 and a low IFNγ cytokine profile compared with WT B6 mice. Administration of the anti–IL-17 mAb attenuated arthritis in DBA/1 mice and almost completely prevented expression of arthritis in IFNγ-KO B6 mice.


These results indicate that sensitivity of IFNγ-deficient B6 mice to CIA is associated with high IL-17 production and that this cytokine is required for expression of arthritis in this strain.

Despite the powerful inflammatory effect of cytokines such as tumor necrosis factor α (TNFα) produced by the innate immune system, abundant evidence indicates that T cells are required for initiation and/or chronicity both in human rheumatoid arthritis (RA) and in mouse models of RA (1–3). Interleukin-17 (IL-17) has emerged as a critical T cell cytokine in the pathogenesis of human RA and several mouse models of this disease (for review, see ref.4). IL-17 promotes inflammation through enhancing the production of other proinflammatory cytokines, IL-1 and TNFα, by monocytes (5) and has a synergistic effect in TNFα-induced IL-1, IL-6, and IL-8 synthesis by synovial fibroblasts (6, 7). In addition, IL-17 is implicated in cartilage destruction and bone resorption (8, 9). A necessary role for this cytokine has been demonstrated in several experimental models of arthritis by attenuation of murine arthritis with IL-17 neutralization by anti–IL-17 antibodies or with mice genetically deficient in IL-17 or IL-17 receptor (IL-17R) (10–12).

IL-17 is produced by activated memory/effector T cells. Recent studies suggest that CD4+ T cells producing IL-17 represent a distinct subpopulation of T helper cells that have been designated Th17 (13, 14). At least in vitro, IL-17 production is negatively regulated by both interferon-γ (IFNγ) and IL-4 (13, 14). One of us and other investigators previously reported that a lack of IFNγ paradoxically enhanced autoimmunity in diseases thought to be mediated by Th1, such as experimental autoimmune encephalomyelitis (EAE) and collagen-induced arthritis (CIA) (15–18). In CIA, IFNγ deficiency renders the normally resistant C57BL/6 (B6) strain susceptible to disease (16–18), and lack of IFNγ or signaling through the IFNγ receptor enhances the severity of arthritis in susceptible strains such as DBA/1 (19, 20). Lack of IFNγ was thought to promote disease through lack of activation-induced cell death and up-regulation of IL-1 (15, 17). In the present study, we tested a different hypothesis, that increased susceptibility to arthritis is explained by loss of IFNγ-mediated suppression of IL-17.



Wild-type (WT) B6 mice, IFNγ-knockout (KO) B6 mice, and DBA/1 mice were obtained from The Jackson Laboratory (Bar Harbor, ME) and Taconic (Germantown, NY) and kept in a modified specific pathogen–free facility. Mice were used at age 8–10 weeks under a protocol approved by the Institutional Animal Care and Use Committee.

Cytokines and antibodies to cytokines.

Rat anti-mouse IL-17 monoclonal antibody (mAb) (clone M210, IgG2a) was generated at Amgen (Seattle, WA) by immunization with recombinant mouse IL-17. M210 was shown to bind to IL-17 specifically, but not to TNFα, IL-1β, IL-6, or IFNγ, in enzyme-linked immunosorbent assays (ELISAs) using M210 as a capture antibody (data not shown). Furthermore, M210 did not block TNFα cytotoxicity in L929 cells (data not shown). Fine specificity of M210 was tested by its ability to neutralize IL-17 family members in a bioassay that measures IL-17 stimulation of keratinocyte-derived chemokine (KC) secretion by fibroblasts (21). We used the following cytokines and antibodies to cytokines for ELISA and cell culture: recombinant IL-17, purified rat anti-mouse IL-17, and biotinylated goat anti-mouse IL-17 (R&D Systems, Minneapolis, MN); biotinylated anti-mouse IL-4, IL-6, and IL-12/IL-23 (p40) and purified anti-mouse IL-12 (p35) and TNFα (BioLegend, San Diego, CA); purified anti-mouse IL-6 and recombinant IFNγ (BD PharMingen, San Diego, CA); biotinylated anti-mouse IFNγ (Caltag, South San Francisco, CA); anti-mouse IL-23 (p19), biotinylated rabbit anti-mouse TNFα, functional grade purified anti-mouse IFNγ, and recombinant IL-4, IL-23, IL-6, IL-12, and IFNγ (eBioscience, San Diego, CA); and anti-mouse IL-4 (National Cancer Institute Biological Resources Branch Preclinical Repository, Rockville, MD).

Ex vivo cell culture and ELISAs for cytokines.

Inguinal lymph nodes and spleen were harvested 10–14 days postimmunization, and single-cell suspensions were prepared. Cells were cultured in RPMI 1640 containing 10% fetal calf serum (FCS) at 5 × 106/ml in round-bottomed 96-well plates in the presence of type II collagen (CII). Cytokines or mAb to cytokines were added to the culture for 72 hours. Cytokines in supernatants were quantified by ELISA according to the manufacturers' instructions.

IL-17 bioassay.

NIH3T3 cells derived from mouse embryonic fibroblasts (American Type Culture Collection, Manassas, VA) were plated at 1 × 105/well in round-bottomed 96-well plates in Dulbecco's modified Eagle's medium containing 10% FCS. Mouse IL-17 (10 ng/ml; R&D Systems) or IL-17F (300 ng/ml; Amgen) in combination with human TNFα (0.5 ng/ml; R&D Systems) was added to the culture. In some experiments, IL-17 or IL-17F and TNFα were preincubated for 45 minutes with M210 (30 μg/ml) or mouse IL-17R-Fc fusion protein (30 μg/ml; Amgen) prior to adding to the culture. Cells were cultured for 24 hours, and supernatant was assayed for KC by ELISA (R&D Systems) according to the manufacturer's instructions.

Induction of CIA and anti–IL-17 mAb treatment.

Native bovine CII was provided by Dr. Marie Griffiths (University of Utah, Salt Lake City, UT), and native chicken CII was obtained from Sigma-Aldrich (St. Louis, MO). CII was dissolved in 0.1M acetic acid and emulsified in Freund's complete adjuvant (CFA; Difco, Detroit, MI) containing 5 mg/ml of killed Mycobacterium tuberculosis H37Ra. Mice were injected intradermally at the base of the tail with 100 μg CII in CFA and boosted with an intraperitoneal (IP) injection of 100 μg CII on day 21. For in vivo anti–IL-17 mAb treatment in DBA/1 mice, arthritic mice were randomized into 3 groups and treated with phosphate buffered saline, normal rat IgG, or anti–IL-17 mAb (M210) at 100 μg IP every other day. For IFNγ-KO B6 mice, on the day of initial immunization, mice were injected IP with 200 μg anti–IL-17 mAb (M210) or normal rat IgG2a weekly for a total of 4 doses. Arthritis was assessed using a scoring system as described (16). Mice were killed on day 56, and the limbs were removed. Joint pathology was evaluated on decalcified hematoxylin and eosin–stained sections (16).

Flow cytometry.

Intracellular cytokine staining was performed using an intracellular staining kit (BD PharMingen). Lymphocytes from CII-immunized mice were stimulated with phorbol myristate acetate (0.05 μg/ml) and ionomycin (0.5 μg/ml) in the presence of GolgiStop solution (BD PharMingen) for 6 hours and stained with fluorescein isothiocyanate–conjugated anti-CD4, allophycocyanin-conjugated anti-IFNγ (clone XMG1.2; BioLegend), and phycoerythrin-conjugated anti–IL-17 (clone TC11-18H10.1; BD PharMingen) and analyzed on a FACSCanto (BD Biosciences, San Jose, CA).

Statistical analysis.

Results were analyzed with Student's t-test.


Antigen-specific and dose-dependent exaggerated T cell IL-17 response to CII in IFNγ-KO B6 mice.

Since recent studies have shown that IFNγ suppresses IL-17 production in vitro (13, 14) and IL-17 is intimately associated with inflammatory responses in several experimental models of arthritis (for review, see ref.4), we tested the hypothesis that susceptibility to CIA in IFNγ-KO B6 mice was due to lack of repression of IL-17 production. We therefore induced CIA in WT B6 and IFNγ-KO B6 mice and compared T cell IL-17 production immediately ex vivo and following restimulation with CII in vitro.

CII immunization of IFNγ-KO B6 mice induced a 2–3-fold higher proportion of IL-17–producing CD4+ T cells compared with WT B6 mice when analyzed immediately ex vivo (Figure 1a). The prototypical CIA-susceptible mouse strain DBA/1 also showed a higher proportion of IL-17–producing T cells and a lower proportion of IFNγ-producing T cells compared with WT B6 mice. Ex vivo restimulation of T cells from IFNγ-KO B6 mice primed with CII in vivo revealed an ∼5-fold increase in IL-17 production compared with WT B6 mice (Figure 1b). Since IL-17 production was only detected in IFNγ-KO B6 mice that were immunized with CFA/CII and restimulated with CII, these responses were antigen specific and were also dose dependent (Figure 1c). The increased IL-17 cytokine response was not accompanied by enhanced production of all inflammatory cytokines in IFNγ-KO B6 mice: no differences were observed in IL-12 or TNFα responses (data not shown), whereas IL-6 production was significantly higher in IFNγ-KO B6 mice (Figure 1d).

Figure 1.

Increased interleukin-17 (IL-17) produced by CD4+ T cells from interferon-γ–knockout C57BL/6 mice (IFNγ-KO B6 mice) in response to type II collagen (CII) immunization. Wild-type (WT) B6 mice (B6), IFNγ-KO B6 mice (B6-IFN-γ KO), and DBA/1 mice were injected with Freund's complete adjuvant (CFA) containing buffer alone or bovine CII, as described in Materials and Methods. Ten days postimmunization, inguinal lymph nodes and spleen were harvested and examined for inflammatory cytokine production. a, Cells obtained immediately ex vivo were stimulated with phorbol myristate acetate/ionomycin for 6 hours, and intracellular production of IL-17 and IFNγ by CD4+ T cells was detected by flow cytometry. Results are representative of 4–6 mice analyzed in each group. b, Mononuclear cells were cultured with medium (Med) alone, ovalbumin (OVA), or CII for 72 hours. IL-17 production in the supernatant was measured by enzyme-linked immunosorbent assay (ELISA). c, CII was added at concentrations ranging from 0 μg/ml to 100 μg/ml. Results shown in b and c are presented as the mean and SD and are pooled from 6–9 mice in each group. d, Mononuclear cells were restimulated with CII as above. IL-6 in the supernatant was quantified by ELISA. Results are presented as the mean and SD and are pooled from 4–6 mice in each group. ∗ = P = 0.04 versus WT B6 mice; ∗∗ = P = 0.013 versus WT B6 mice. PBS = phosphate buffered saline.

Synergistic action of IFNγ and IL-4 to suppress the IL-17 response to CII.

Since we proposed that enhanced IL-17 production in IFNγ-KO B6 mice was caused by loss of IFNγ suppression, we added IFNγ back to T cell restimulation cultures and quantified IL-17 in the supernatants. As shown in Figure 2a, IL-17 production was almost completely suppressed by IFNγ. To determine the effect of suppression of IFNγ on IL-17 production by WT B6 mice in the restimulation assay, we added a neutralizing anti-IFNγ mAb during the 3-day culture period. Neutralization of IFNγ in vitro resulted in a statistically significant increase in IL-17 production, although the IL-17 concentrations were lower than in IFNγ-KO B6 mice (Figure 2b). Since IL-4 has also been implicated in IL-17 suppression, we compared the effects of neutralization of IFNγ, neutralization of IL-4, and neutralization of both cytokines together. Despite a lack of effect of neutralizing IL-4 alone and detection of very low levels of IL-4 in T cell restimulation cultures (not shown), neutralization of both cytokines had a striking synergistic effect (Figure 2c).

Figure 2.

IFNγ suppression of IL-17 production by T cells from IFNγ-KO B6 mice in response to CII. a, Spleen and lymph node mononuclear cells from immunized IFNγ-KO B6 mice were cultured with CII (100 μg/ml) as described in Figure 1, except that IFNγ was added at the indicated concentrations (in ng/ml) at the beginning of culture, and IL-17 was quantified on day 3. b, To determine the effect of neutralization of IFNγ on IL-17 production in vitro, mononuclear cells from immunized WT B6 mice were cultured as described in Figure 1, except that anti-IFNγ monoclonal antibody (mAb) was added to the culture at the indicated concentrations (in μg/ml). c, In vivo–primed lymphocytes from WT B6 mice were stimulated with CII in the presence of 10 μg/ml anti-IFNγ, 10 μg/ml anti–IL-4, or a combination of mAb. Results shown in ac are presented as the mean and SD and are pooled from 3 experiments (n = 6 mice in each group). See Figure 1 for other definitions.

Suppression of arthritis by in vivo treatment with anti–IL-17 mAb.

We first examined the fine specificity of the anti–IL-17 mAb, M210, using a bioassay that quantifies IL-17 stimulation of KC secretion by fibroblasts (21). As shown in Figure 3a, M210 completely blocked IL-17–mediated KC production, but not IL-17F–mediated KC production, by the murine fibroblast cell line NIH3T3. The IL-17–neutralizing capacity of the anti–IL-17 mAb M210 was equivalent to that of the IL-17R-Fc fusion protein. This demonstrates a high degree of specificity of M210 and its ability to neutralize IL-17 bioactivity. Since IL-17–neutralizing antibodies have previously been shown to be effective in the treatment of CIA in DBA/1 mice (10), we further tested M210 for functional activity in this arthritis model in vivo. As shown in Figure 3b, M210 significantly attenuated progression of arthritis even when administered after disease onset, consistent with previous studies using polyclonal anti–IL-17 antibodies (10).

Figure 3.

Suppression of arthritis by blockade of IL-17 in vivo. a, NIH3T3 cells were cultured in tumor necrosis factor α (0.5 ng/ml)–supplemented medium in the presence of IL-17 (10 ng/ml) or IL-17F (300 ng/ml) for 24 hours. Anti–IL-17 monoclonal antibody (M210; 30 μg/ml) or IL-17 receptor-Fc fusion protein (IL-17R-Fc; 30 μg/ml) was preincubated with IL-17 or IL-17F prior to adding to the culture. Keratinocyte-derived chemokine (KC) was measured by ELISA. bd, DBA/1 and IFNγ-KO B6 mice were immunized with CII in CFA and boosted with an intraperitoneal (IP) injection of CII on day 21. b, DBA/1 mice were treated with M210 (100 μg/mouse IP, n = 20), control rat IgG (100 μg/mouse IP, n = 6), or PBS (n = 15) every other day after the onset of arthritis. Severity of arthritis was monitored using a scoring system as described elsewhere (16). c, Clinical score and cumulative incidence of arthritis in IFNγ-KO B6 mice. On the day of immunization, IFNγ-KO B6 mice were treated with M210 (200 μg/mouse IP; n = 10) or with normal rat IgG2a (n = 10) followed by weekly injections for a total of 4 doses. Mice were followed up for 56 days for development of arthritis. Results shown in ac are presented as the mean and SD. d, Hematoxylin and eosin–stained sections from the hind paws of IFNγ-KO B6 mice obtained 56 days postimmunization. A and B, Sections from rat IgG2a–treated mice showing massive synovitis, erosion, and joint destruction. C and D, Sections from anti–IL-17–treated mice, representative of paws with no (C) or minimal (D) inflammation. See Figure 1 for other definitions.

IFNγ-KO B6 mice produced higher levels of IL-17 in response to immunization with CII in CFA as compared with WT B6 mice. To determine whether IL-17 was necessary for the development of CIA, we determined whether neutralization of IL-17 with the anti–IL-17 mAb M210 could prevent CIA in IFNγ-KO B6 mice. These mice were injected with 200 μg of anti–IL-17 mAb or control rat IgG2a on the day of CII immunization, followed by weekly injections of mAb for a total of 4 doses. Development of arthritis was followed for 56 days after initial immunization and scored using a system described previously (16). Eighty percent of mice treated with normal rat IgG2a developed arthritis with a typical clinical course with a mean onset on day 29 after initial immunization (Figure 3c), as observed previously (16). In contrast, only 20% of the mice in the anti–IL-17–treated group developed arthritis, and this was mild and had a delayed onset (Figure 3c). Histologic examination of control IgG2a-injected joints showed extensive synovitis, pannus formation, marginal erosion, cartilage destruction, and joint architectural changes (Figure 3d, parts A and B) which are typically seen in CIA (16). In contrast, most joints from anti–IL-17–treated animals had minimal changes (Figure 3d, part C) or a very mild synovitis in joints showing clinical inflammation (Figure 3d, part D). These results indicated that IL-17 is necessary for the expression of arthritis in IFNγ-deficient B6 mice.

Association of heightened IL-17 production in CIA-susceptible DBA/1 mice with low IFNγ responses to CII.

DBA/1 mice have previously been shown to produce high levels of IL-17 in response to CII (22). We observed that, like IFNγ-KO B6 mice, DBA/1 mice had a higher proportion of CD4+ T cells producing IL-17 but markedly fewer IFNγ-producing T cells compared with WT B6 animals when studied ex vivo. Furthermore, highly significant differences in IL-17 and IFNγ production were observed between DBA/1 and WT B6 mice upon restimulation in vitro (Figures 1a and c).


In this study, we demonstrated that lack of IFNγ in a normally resistant strain of mouse leads to heightened IL-17 responses to CII immunization, and that neutralization of IL-17 almost entirely prevents expression of disease. This strongly suggests that conversion from CIA resistance to susceptibility is explained by the release of IL-17 from IFNγ-mediated suppression. Our findings are consistent with those of Murphy et al (23), who reported that IL-12 (p35)–deficient mice on a B6 or (B6 × 129)F2 background developed CIA and had increased numbers of IL-17–producing T cells. It is of interest that exaggerated IL-17 responses were associated with increased IL-6, but not TNFα, production in vitro. Since IL-6 may be both downstream and, together with transforming growth factor β, upstream of IL-17 (24–26), and since IL-17–producing cells also secrete IL-6 (27), a positive feedback loop may be in operation in arthritis.

It was reported that both IFNγ and IL-4 can suppress IL-17 production in vitro (13, 14). Our ex vivo data indicate that in WT B6 mice, IFNγ was the dominant T cell cytokine response to CII in CFA, whereas IL-4 was only weakly detected. Nevertheless, IFNγ and IL-4 together had a synergistic effect on suppression of CII-specific IL-17 production during CII restimulation in vitro. The suppressive effect of IFNγ in vivo also appears to be dominant in CIA, as mentioned above, and in myelin oligodendrocyte glycoprotein peptide–induced EAE, in which IFNγ-KO mice developed a fatal disease, with 50% mortality as compared with no mortality in IFNγ-intact mice (15). Since IFNγ may actually be required for IL-4 production by Th2 cells (28, 29), a possible reduction of both IL-17–suppressing cytokines may explain the striking effect on disease severity when IFNγ is deficient. In contrast, B6 mice genetically deficient in IL-4 are resistant to CIA (18), and IFNγ-KO B6 mice that were treated with anti–IL-4 mAb appeared to be resistant to CIA, leading to the suggestion that IL-4 may actually promote CIA (18). Based on the results reported here and the known suppression of IFNγ by IL-4, perhaps a lack of IL-4 leads to increased IFNγ, resulting in suppression of IL-17. However, it should be noted that high concentrations of exogenous IL-4 or systems overexpressing IL-4 were reported to suppress arthritis (30, 31), and this effect was associated with decreased IL-17 in arthritic joints (31).

A particularly interesting observation in the present study was that high IL-17 production by T cells in response to CII restimulation in DBA/1 mice was associated with low IFNγ secretion. It was recently reported that DBA/1-derived natural killer (NK) T cells have defective production of IFNγ and IL-4 in response to stimulation with a natural agonist, α-galactosylceramide, as compared with NK T cells from B6 mice (32). This defect occurred despite equivalent numbers of NK T cells. These findings are consistent with the idea that a relative lack of IFNγ, possibly in conjunction with IL-4, may contribute to disease susceptibility to CIA in DBA/1 mice.

Previous studies have demonstrated the relative deficiency of IFNγ (33, 34), but abundant IL-17 (for review, see ref.4), in RA patients. CD4+ T cells from RA patients produce low levels of IFNγ in response to IL-12 or IL-12/IL-18 stimulation (35), although it remains to be determined whether this defect is primary or secondary to suppression by other proinflammatory cytokines, since the IFNγ response to IL-12 or IL-18 was partially restored by TNFα blockade (35). As with the low IFNγ–producing mouse strains we studied in RA patients, reduced IFNγ responses to CII or other joint-derived antigens may allow excessive IL-17 production that, in turn, drives chronic inflammation. Further studies are required to better understand how IFNγ regulates CD4+ T cell production of IL-17 during arthritis development. Therapeutic intervention designed to redirect T cell cytokine production from IL-17 to IFNγ may prove to be beneficial in RA patients.


Dr. Elkon had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Chu, Tocker, Elkon.

Acquisition of data. Chu, Swart, Alcorn.

Analysis and interpretation of data. Chu, Tocker, Elkon.

Manuscript preparation. Chu, Tocker, Elkon.

Statistical analysis. Chu.


We would like to thank Drs. Chris Wilson and Jacques Peschon for discussion and Daniel Kim for technical assistance.