IL-1β and TNFα-initiated IL-6–STAT3 pathway is critical in mediating inflammatory cytokines and RANKL expression in inflammatory arthritis

Abstract

Rheumatoid arthritis (RA) is a chronic inflammatory disease that causes irreversible joint damage and significant disability. However, the fundamental mechanisms underlying how inflammation and joint destruction in RA develop and are sustained chronically remain largely unknown. Here, we show that signal transducer and activator of transcription 3 (STAT3) is the key mediator of both chronic inflammation and joint destruction in RA. We found that inflammatory cytokines highly expressed in RA patients, such as IL-1β, tumor necrosis factor alpha and IL-6, activated STAT3 either directly or indirectly and in turn induced expression of IL-6 family cytokines, further activating STAT3 in murine osteoblastic and fibroblastic cells. STAT3 activation also induced expression of receptor activator of nuclear factor kappa B ligand (RANKL), a cytokine essential for osteoclastogenesis, and STAT3 deficiency or pharmacological inhibition promoted significant reduction in expression of both IL-6 family cytokines and RANKL in vitro. STAT3 inhibition was also effective in treating an RA model, collagen-induced arthritis, in vivo through significant reduction in expression of IL-6 family cytokines and RANKL, inhibiting both inflammation and joint destruction. Leukemia inhibitory factor expression and STAT3 activation by IL-1β were mainly promoted by IL-6 but still induced in IL-6-deficient cells. Thus, our data provide new insight into RA pathogenesis and provide evidence that inflammatory cytokines trigger a cytokine amplification loop via IL-6–STAT3 that promotes sustained inflammation and joint destruction.

chronic inflammation, collagen-induced arthritis, IL-6, joint destruction, rheumatoid arthritis

Introduction

Rheumatoid arthritis (RA) is a disease characterized by chronic inflammation and joint damage causing joint destruction, which limits patients’ daily living activities due to impaired physical function and joint pain. RA is a complex disease resulting from known and unknown factors; however, it is primarily characterized by the presence of major pro-inflammatory cytokines, such as IL-1β, tumor necrosis factor alpha (TNFα) and IL-6. Indeed, the therapeutic antibodies or inhibitors of these inflammatory cytokines provide dramatic therapeutic effects in inhibiting disease activity. Nonetheless, some patients remain resistant to therapy, suggesting that targeting these cytokines individually is not sufficient to treat the pathogenesis of RA.

To date, various RA animal models such as transgenic mice over-expressing IL-1α (IL-1α Tg) or TNFα Tg, mice homozygous for a gp130 F759 mutation, collagen-induced arthritis (CIA) model and antibody against type2 collagen injection model reportedly exhibit RA-like phenotypes, including swelling and destruction in multiple joints (1–5). Analysis of these models indicates that increased serum pro-inflammatory cytokines, activation of inflammatory cells such as macrophages, neutrophils or CD4⁺ T cells or elevated levels of auto-antibodies underlie RA pathogenesis. Hirano’s group reported that an IL-17A-triggered positive feedback loop of IL-6 signaling, which activates both the nuclear factor kappa B (NFκB) and the signal transducer and activator of transcription 3 (STAT3), promotes autoimmune disease such as encephalomyelitis (6). Recently, they reported that local microbleeding induces an IL-17A-dependent IL-6 amplification loop by inducing local T_h17 accumulation followed by T_h17-mediated IL-6 expression and STAT3 activation in type I collagen-expressing cells (7). However, how inflammation and joint damage are sustained in a chronic manner or why a biological agent inhibiting a single pro-inflammatory cytokine occasionally fails to antagonize disease activity remains unknown.

STAT proteins are a family of transcriptional factors consisting of six members, STATs 1–6, all of which are activated by cytokine stimulation via Janus kinase (Jak) family proteins (8). In mice, STAT3 is implicated in development of visceral endoderm, and STAT3-null mice exhibit early embryonic lethality due to impaired mesoderm formation (9). Jak family proteins are non-receptor type tyrosine kinases consisting of Jak1, Jak2 and Jak3, which transduce various cytokine signals, including IL-6 family cytokines. Phenotypes seen in null mutants of each Jak gene differ, suggesting that each Jak has a unique function. CP690,550 is a small molecule, which was originally developed as a specific inhibitor of Jak3 (10). CP690,550 was utilized as an immunosuppressive agent for organ transplantation (10) and is now in clinical trials as treatment for RA patients (11). CP690,550 has also been utilized in an animal RA model and shown to reduce disease activity in vivo (12, 13).

RA causes induction of osteoclast formation, and osteoclasts are implicated in joint destruction due to their bone resorption activity at sub-chondral bone sites (14). Osteoclastogenesis is induced by the cytokine receptor activator of nuclear factor kappa B ligand (RANKL) (15), while osteoclast differentiation is driven by pro-inflammatory cytokines TNFα or IL-1β. Given their unique role in bone resorption, controlling osteoclast activity is considered crucial to prevent joint destruction in RA.

In the current study, we demonstrate that the major inflammatory cytokines elevated in RA, IL-1β, TNFα and IL-6, all function in an amplification circuit for IL-6 family cytokines and RANKL via direct or indirect activation of STAT3. STAT3 activation further induced IL-6 family cytokines as well as RANKL, and lack of STAT3 abrogated both IL-6 family cytokine and RANKL expression. Pharmacological inhibition of STAT3 also inhibited expression of IL-6 family cytokines and RANKL in osteoblastic cells induced by IL-1β, TNFα and IL-6 in vitro as well as in the joints of a CIA model in vivo. Furthermore, CIA-induced joint swelling and osteoclastogenesis were significantly inhibited by STAT3 abrogation, promoting significant improvements in arthritis scores in the CIA model. These findings indicate that STAT3 is a potential therapeutic target to prevent chronic inflammation and joint destruction in RA.

Methods

Mice

C57B6 mice and DBA/1J mice were purchased from Nihon Jikken Doubutsu (Tokyo, Japan). Animals were maintained under specific pathogen-free conditions in animal facilities certified by the Keio University School of Medicine animal care committee. Animal protocols were approved by the Keio University School of Medicine animal care committee.

CIA model

The experimental CIA model was generated in 6-week-old male mice by injecting 100 μl of emulsion containing 100 μg of type II collagen (CII) (Collagen Research Center, Tokyo, Japan) intradermally at the base of the tail. The basic emulsion was composed of 1 mg ml^–1 bovine CII dissolved in PBS and an equal volume of complete Freund’s adjuvant (Difco, Detroit, MI, USA). Three weeks later, a second immunization containing CII and Incomplete Freund’s adjuvant (Difco) was administered. Then, clinical symptoms of arthritis were evaluated visually in each limb and graded on a scale of 0–4; 0, no erythema or swelling; 0.5, swelling of one or more digits; 1, erythema and mild swelling of the ankle joint; 2, mild erythema and mild swelling involving the entire paw; 3, erythema and moderate swelling involving the entire paw and 4, erythema and severe swelling involving the entire paw. The clinical score for each mouse was the sum of the scores for all four limbs (maximum score 16).

Chemicals and reagents

CP-690550 was purchased from Selleck Chemicals (Houston, TX, USA). Cycloheximide was purchased from Sigma–Aldrich (St Louis, MO, USA); mouse IL-1β, mouse TNFα and mouse IL-6 were purchased from PeproTech Ltd (London, UK). Mouse oncostatin M (OSM) and sIL-6Rα were purchased from R&D systems (Minneapolis, MN, USA).

Cell culture

Primary osteoblasts (POBs) derived from calvaria of newborn mice were prepared as described (16, 17). POBs and MC3T3-E1 cells were maintained in αMEM (Sigma–Aldrich) containing 10% fetal bovine serum (FBS) (JRH Biosciences, KS, USA) with penicillin G and streptomycin. Splenocytes and lymphocytes were cultured in RPMI 1640 (Wako Pure Chemicals Industries, Osaka, Japan) supplemented with 10% FBS, 2-mercaptoethanol (Nacalai Tesque, Kyoto, Japan), penicillin G and streptomycin.

In vitro osteoclast formation

BM cells isolated from long bones (femurs and tibias) were cultured for 72 h in αMEM containing 10% heat-inactivated FBS and GlutaMax supplemented with macrophage colony stimulating factor (M-CSF) (50 ng ml^–1; Kyowa Hakko Kirin Co.). Adherent cells were then collected for analysis. Cells (10⁵) were cultured in 96-well plates with M-CSF and recombinant soluble RANKL (25 ng ml^–1; PeproTech Inc.) for 6 days. Osteoclastogenesis was evaluated by tartrate resistance acid phosphatase (TRAP) and May-Grünwald Giemsa staining (18, 19). For co-cultivation, M-CSF-dependent osteoclast progenitor cells and POBs were seeded at a density of 5 × 10⁵ cells and 2 × 10⁴ cells per 48-well plate, respectively, and co-cultured for 7 days in the presence of 50 ng ml^–1 OSM (R&D) or 10^-8 M 1,25(OH)₂D₃ (Wako) plus 1 μM prostaglandin E2 (PGE2) (Wako).

Enzyme-linked immunosorbent assay

ELISA assays for IL-6, IL-17 and IFNγ were undertaken following the manufacturer’s instructions (eBioscience, San Diego, CA, USA). A pSTAT3 ELISA kit (R&D) was utilized to measure pSTAT3 levels according to manufacturer’s instructions. The optical density at 450 nm was read on a Labsystems Multiscan MS (Analytical Instruments, LLC, MN, USA).

Real-time PCR analysis

Total RNAs were isolated from cultures or tissues using TRIzol reagent (Invitrogen, Tokyo, Japan). After denaturation of total RNAs at 70°C for 5 min, cDNAs were synthesized by reverse transcription from total RNA using oligo (dT) primer (Wako). Real-time PCR was performed using SYBR Premix ExTaq II (Takara Bio Inc., Shiga, Japan) with the DICE Thermal cycler (Takara Bio Inc.), according to the manufacturer’s instructions. Samples were matched to a standard curve generated by amplifying serially diluted products using the same PCR reactions. β-actin expression served as an internal control. Primer sequences were as follows:

β-actin forward: 5′-TGAGAGGGAAATCGTGCGTGAC-3′; β-actin reverse: 5′-AAGAAGGAAGGCTGGAAAAGAG-3′; IL-6 forward: 5′-CAAAGCCAGAGTCCTTCAGAG-3′; IL-6 reverse: 5′-GTCCTTAGCCACTCCTTCTG-3′; OSM forward: 5′-AACTCTTCCTCTCAGCTCCT-3′; OSM reverse: 5′-TGTGTTCAGGTTTTGGAGGC-3′; IL-11 forward: 5′-TGGGACATTGGGATCTTTGC-3′; IL-11 reverse: 5′-CATTGTACATGCCGGA-GGTAG-3′; Leukemia inhibitory factor (LIF) forward: 5′-TTCCCATCACCCCTGTAAATG-3′; LIF reverse: 5′-GAAACGGCTCCCCTTGAG-3′; RANKL forward: 5′-CAATGGCTGGCTTGGTTTCATAG-3′; RANKL reverse: 5′-CTGAACCAGACATGACAGCTGGA-3′; Ctsk forward: 5′-ACGGAGGCATTGACTCTGAAGATG-3′; Ctsk rev-erse: 5′-GGAAGCACCAACGAGAGGAGAAAT-3′; c-Fos forward: 5′-ATCGGCAGAAGGGGCAAAGTAG-3′; c-Fos reverse: 5′-GCAACGCAGACTTCTCATCTTCAAG-3′; NFATc1 forward: 5′-CAAGTCTCACCACAGGGCTCACTA-3′; NFATc1 reverse: 5′-GCGTGAGAGGTTCATTCTCCAAGT-3′

Immunoblotting analysis

Whole-cell lysates were prepared from cultures using radioimmunoprecipitation assay buffer (1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 150 mM NaCl, 10 mM Tris–HCl, pH 7.5, 5 mM EDTA and a protease inhibitor cocktail; Sigma–Aldrich). Equivalent amounts of protein were separated by SDS–PAGE and transferred to a PVDF membrane (Millipore, Billerica, MA, USA). Proteins were detected using the following antibodies: anti-pSTAT3 (#9131; Cell signaling technology, Inc., Beverly, MA, USA), anti-STAT3 (sc483; Santa Cruz Biotechnology, Inc., California, USA) and anti-actin (A2066; Sigma–Aldrich).

Histopathology and fluorescent immunohistochemistry

The lower ankle of CIA mice was fixed in 10% neutral-buffered formalin, decalcified in 10% EDTA, pH7.4, embedded in paraffin and then cut into 4-μm sections. Hematoxylin and eosin (H&E) and safranin-O stain were performed according to standard procedures. For each fluorescent immunohistochemistry assay, 4-μm sections were cut and subjected to microwave treatment for 5 min in 1 mM EDTA (pH 8.0) for antigen retrieval. After blocking with 0.1% BSA in 100 mM Tris–HCl (pH 7.6), 150 mM NaCl, 0.01% Tween-20 (TBST) for20 min, sections were stained for 1 h with mouse anti-rabbit pSTAT3 (1:100 dilution; Cell Signaling Techniques, Inc.), goat anti-mouse RANKL (1:150 dilution; R&D Systems, Inc.) or rabbit anti-mouse cathepsin K (1:100 dilution) at room temperature. After washing in TBST, sections were stained with Alexa Fluor 488-conjugated goat anti-mouse IgG (1:200 dilution; Invitrogen), Alexa Fluor 488-conjugated donkey anti-goat IgG (1:200 dilution; Invitrogen) or Alexa Fluor 546-conjugated rabbit anti-mouse IgG (1:200 dilution; Invitrogen) for 1 h at room temperature. Then, sections were mounted using Dako fluorescence mounting medium (Dako, Glostrup, Denmark). TOTO3 (1:750; Invitrogen) was used for a nuclear stain. Images were acquired with a laser confocal microscope (FV1000-D; Olympus, Tokyo, Japan).

Surface and intracellular cytokine staining

For intracellular cytokine staining, cultured cells were restimulated for 8 h with 50 nM phorbol myristate acetate (Sigma–Aldrich), 1 μg ml^–1 ionomycin (Sigma–Aldrich) and 1 μM brefeldin A (eBioscience). Surface staining was performed for 15 min with the corresponding mixture of fluorescently labeled antibodies as previously described (20). After surface staining, cells were suspended in fixation buffer (eBioscience), and intracellular cytokine staining was performed using the manufacturer’s protocol using anti-IL-17-PE (BD Pharmingen). Stained cells were analyzed with BD FACSAria2 (BD Biosciences).

Naive T-cell preparation and differentiation

For naive T-cell preparation and differentiation, CD4⁺CD25^– T cells were isolated from wild-type mouse spleens and lymph nodes by negative selection using biotinylated anti-CD25 (clone, BC96; eBioscience), CD8 (clone, 53–6.7; eBioscience), CD11b (clone, M1/70; eBioscience), B220 (clone, RA3-6B2; eBioscience), anti-NK1.1 (PK136; eBioscience), CD11c (N418; eBioscience) followed by streptoavidin-conjugated magnetic beads (Miltenyi Biotech, Bergisch Gladbach, Germany). For T_h17 differentiation, 1 × 10⁶ CD4⁺CD25^– T cells were cultured with 1 μg ml^–1 plate-bound anti-CD3 antibody (clone, 145-2C11; eBioscience), 0.5 μg ml^–1 soluble anti-CD28 antibody (clone, 37.51; eBioscience), 10 μg ml^–1 anti-IFN-γ antibody (clone, R4-6A2; PeproTech Inc.), 10 μg ml^–1 anti-IL-4 antibody (clone, 11B11; PeproTech Inc.), 0.2 ng ml^–1 recombinant murine IL-2 (PeproTech Inc.), 2 ng ml^–1 recombinant human transforming growth factor (TGF)-β1 (R&D Systems, Inc.) and 20 ng ml^–1 recombinant human IL-6 (R&D Systems, Inc.) for 3 days.

Results

IL-1β and TNFα activate STAT3 indirectly

In RA, high levels of inflammatory cytokines such as IL-1β, TNFα and IL-6 are detected and have been targeted by therapeutic antibodies or receptor antagonists (21–23). We found that treatment of osteoblastic cells with IL-1β further stimulated expression of various cytokines, particularly IL-6 family members such as IL-6, OSM, IL-11 and LIF, as well as RANKL, an essential cytokine for differentiation of osteoclasts (Fig. 1A and data not shown). IL-6 family cytokines and RANKL play a crucial role in inflammation and joint destruction, respectively. TNFα and IL-6 treatment also induced expression of these cytokines in osteoblastic cells (Supplementary Figure 1A and B is available at International Immunology Online), suggesting that inflammatory cytokines detected in RA trigger further expression of IL-6 family cytokines and RANKL in an autocrine/paracrine amplification manner. Interestingly, although STAT3 is not known to be a direct target of IL-1β and TNFα, STAT3 phosphorylation was induced in osteoblastic cells by IL-1β or TNFα treatment (Fig. 1B). However, STAT3 phosphorylation was strongly inhibited by pre-treatment of cells with cycloheximide, a protein synthesis inhibitor, suggesting that IL-1β and TNFα induce STAT3 indirectly (Fig. 1B). Indeed, we observed direct activation of IL-1β targets, such as JNK and NF-κB, prior to activation of STAT3 (data not shown), while IL-6 induced STAT3 phosphorylation at an early period (data not shown). These data suggest that major inflammatory cytokines up-regulated in RA, namely IL-1β, TNFα and IL-6, induce STAT3 phosphorylation but that IL-1β and TNFα likely induce STAT3 phosphorylation via IL-6.

Fig. 1.

IL-1β and TNFα stimulation induces IL-6 cytokine family and RANKL expression via STAT3 phosphorylation. (A) Total RNA was prepared from primary mouse osteoblasts treated with (+) or without (−) IL-1β (10 ng ml^–1), and expression of the IL-6 cytokine family genes IL-6, LIF, and RANKL relative to β-actin was analyzed by a quantitative real-time PCR. Data are means ± SD of the indicated genes/β-actin (*P< 0.001; n = 3). (B) Whole-cell lysates from MC3T3-E1 cells stimulated for the indicated times (minute) with IL-1β (10 ng ml^–1) or TNFα (10 ng ml^–1) in the presence or absence of cycloheximide (CHX, 20 μg ml^–1) were analyzed by immunoblotting to detect pSTAT3 and STAT3. Actin served as internal control. (C, E and G) Total RNA was prepared from wild-type (WT), STAT3 KO (KO) or IL-6 KO (KO) MEFs treated with (+) or without (−) IL-1β (10 ng ml^–1) or TNFα (10 ng ml^–1) for 24 h, and IL-6, LIF, IL-11 or RANKL expression relative to β-actin was analyzed by a quantitative real-time PCR. Data are means ± SD of the indicated genes/β-actin (*P < 0.001; n = 3). (D and F) IL-6 levels in supernatants of WT and STAT3 KO MEFs treated with (+) or without (−) IL-1β (10 ng ml^–1) or TNFα (10 ng ml^–1) for 24 h were assessed by ELISA. (H) Whole-cell lysates from WT or KO MEFs stimulated for 60 min with IL-1β (10 ng ml^–1) were analyzed by immunoblotting to detect pSTAT3 and STAT3. Actin served as internal control.

We then asked whether STAT3 is required for IL-6-related cytokine induction by IL-1β or TNFα. Expression of IL-6 family cytokines following IL-1β, TNFα or IL-6 stimulation was dramatically inhibited in STAT3-deficient cells (Fig. 1C and E and data not shown). IL-6 protein expression induced by IL-1β and TNFα was also significantly inhibited in STAT3-deficient cells (Fig. 1D and F). IL-6 was implicated in the pathogenesis of RA as an amplifier of cytokines (6), and indeed, the expression of IL-6 family cytokines such as IL-11 and LIF as well as RANKL by IL-1β or TNFα was significantly inhibited in IL-6 KO MEFs compared with wild-type MEFs (Fig. 1G). However, STAT3 activation by IL-1β was reduced but still detected in IL-6 KO MEFs (Fig. 1H). These results suggest that STAT3 activity stimulates a cytokine amplification loop that promotes sustained inflammation and RANKL expression in RA and that targeting STAT3 could antagonize chronic inflammation and joint destruction.

STAT3 is a therapeutic target in RA

To determine whether inhibition of STAT3 could have a potential benefit in a mouse model of RA, we undertook experiments using the drug CP690,550, a Jak3 inhibitor also known as Tofacitinib. Osteoblastic MC3T3-E1 cells were treated with IL-1β in the presence of indicated concentrations of CP690,550, and STAT3 phosphorylation was analyzed by western blot (Fig. 2A). CP690,550 treatment effectively decreased STAT3 phosphorylation (Fig. 2A). In addition, CP690,550 was further assessed for effects in a CIA model in vivo (Fig. 2B) and found effective in treating arthritis based on arthritis score.

Fig. 2.

The STAT3 inhibitor CP690,550 inhibits arthritis in vivo and the expression of IL-6 cytokine family in vitro. (A) Whole-cell lysates from MC3T3-E1 cells stimulated with IL-1β (10 ng ml^–1) plus CP690,550 at the indicated concentrations were analyzed by immunoblotting to detect pSTAT3 and STAT3. Actin served as an internal control. (B) 6-week-old DBA/1 male mice were given an initial injection of type 2 collagen on day −21, and arthritis was induced with a second injection on day 0. Vehicle or CP690,550 (15 mg kg^–1 day^–1) was administered intraperitoneally once daily for 2 weeks from day 0 (n = 4 per group). Arthritis scores were measured three times a week. (C and D) Total RNA was prepared from POBs treated with IL-1β (10 ng ml^–1), TNFα (10 ng ml^–1) or OSM (50 ng ml^–1) with (+) or without (−) CP690,550 (100 nM) for 24 h, and IL-6 expression relative to β-actin was analyzed by quantitative real-time PCR. Data are means ± SD of IL-6/β-actin (*P < 0.001; n= 3). (E) IL-6 protein levels in the supernatant of osteoblasts treated with IL-1β (left panel) or TNFα (right panel) plus indicated concentrations of CP690,550 for 24 h were assessed by ELISA. Data are means ± SD of IL-6 (picograms per milliliter).

CP690,550 treatment of osteoblastic cells significantly inhibited IL-1β-induced IL-6 family cytokine expression dose-dependently manner in vitro (Supplementary Figure 2A is available at International Immunology Online). IL-6 mRNA expression in osteoblastic cells induced by TNFα, IL-6 or OSM was also significantly inhibited by CP690,550 dose dependently (Fig. 2D and Supplementary Figure 2B is available at International Immunology Online). Furthermore, IL-6 protein expression induced by IL-1β or TNFα was significantly inhibited by CP690,550 dose dependently (Fig. 2E). These results suggest that STAT3 is the therapeutic target of CP690,550 in blocking a cytokine amplification loop and thereby inhibits sustained inflammation.

Inhibiting STAT3 reduces inflammation seen in a mouse model of RA

To determine whether STAT3 inhibition could ameliorate conditions associated with RA, CP690,550 or vehicle was administered to CIA mice by injection after inflammation had occurred. The arthritis score of CIA mice significantly improved following CP690,550 treatment compared with vehicle treatment (Fig. 3A). Treated mice showed serum IL-6 levels that were significantly down-regulated to a basal level (Fig. 3B), and expression of IL-6 family cytokines in joints was also significantly reduced compared with vehicle-treated control mice (Fig. 3C). Immunohistochemical analysis demonstrated improved fibrosis paralleling reduced STAT3 phosphorylation in joints of CP690,550-treated mice (Fig. 3D). Reduced STAT3 phosphorylation in the joints of CP690,550-treated mice compared with that of control mice was confirmed by western blot and ELISA to detect phosphorylated STAT3 (Fig. 3E). In contrast to STAT3, STAT1 and STAT5 activation was not detected in the joints of CIA mice (Supplementary Figure S3A is available at International Immunology Online), suggesting that STAT3 is specifically activated under inflammatory conditions in the joints. STAT3 is also reportedly activated by IL-23 in T cells (24), and the IL-23/STAT3 pathway is known to play a role in the pathogenesis of the CIA model (25). STAT3 activation by IL-23 in T cells is also inhibited by CP690,550 (supplementary Figure S3B is available at International Immunology Online), further indicating that CP690,550 inhibits STAT3.

Fig. 3.

STAT3 inhibition antagonizes arthritis effects in vivo. (A) 6-week-old DBA/J1 male mice were given initial injection of type 2 collagen on day −21, and arthritis was induced with a second injection on day 0. Vehicle or CP690,550 (15 mg kg^–1 day^–1) was administered intraperitoneally once daily for 2 weeks from day 7 (n = 4 per group). Arthritis scores were measured three times a week. (B) Total RNA was prepared from the tissue of hind paws of CIA-induced mice after 2 weeks of treatment by vehicle or CP690,550, and expression of IL-6 cytokine families (IL-6, OSM, IL-11 and LIF) relative to β-actin was analyzed by a quantitative real-time PCR. (C) IL-6 serum levels in sera of CIA-induced mice after 2 weeks of treatment by vehicle or CP690,550 were assessed by ELISA. (D) Specimens of ankle joints from CIA mice treated with vehicle or CP690,550 for 2 weeks were subjected to immunofluorescence staining for pSTAT3. Nuclei were visualized by TOTO3. Bar, 100 μm. (E) Whole-cell lysates were made from ankle joint tissues of CIA mice treated with or without CP690,550 for 2 weeks. Phosphorylated STAT3 was then analyzed by western blot (upper panel) and ELISA (lower panel). Results are representative of at least three independent experiments.

Blocking STAT3 inhibits RANKL expression, osteoclastogenesis and joint destruction in RA

Osteoclastogenesis, which is regulated by the cytokine RANKL, is a crucial event in joint destruction in RA. To examine the effect of STAT3 on RANKL expression, we undertook real-time PCR and immunohistochemistry. RANKL expression induced by IL-1β and TNFα was significantly inhibited in STAT3-deficient cells than wild-type cells (Fig. 4A and Supplementary Figure 4A is available at International Immunology Online). Induction of RANKL expression in osteoblastic cells by the inflammatory cytokines IL-1β and TNFα and by IL-6 family cytokines, such as IL-6 and OSM, was also significantly inhibited by CP690,550 dose dependently (Fig. 4B and Supplementary Figure 4B is available at International Immunology Online). Osteoclastogenesis induced in co-cultures of bone marrow-derived osteoclast progenitor cells and osteoblastic cells in the presence of OSM, which induces RANKL expression, was significantly inhibited by CP690,550 dose dependently (Fig. 4C). Vitamin D3 (VitD3) plus PGE2 also reportedly induces RANKL expression in osteoblastic cells and can thus induce osteoclastogeneis in a co-culture system of osteoblastic cells with osteoclast progenitors (26, 27). However, unlike OSM-induced osteoclastogenesis, CP690,550 did not inhibit VitD3 + PGE2-induced osteoclastogenesis (Fig. 4C). IL-1β and TNFα are known to activate the NFκB pathway (28, 29), but NFκB activation by IL-1β was not inhibited by CP690,550 (Supplementary Figure 4C is available at International Immunology Online). These results suggest that CP690,550 specifically inhibits osteoclastogenesis induced by STAT3-dependent RANKL expression in osteoblastic cells. Indeed, CP690,550 had little ability to inhibit osteoclastogenesis from bone marrow macrophages in the presence of M-CSF plus RANKL without osteoblastic cells and multinuclear TRAP-positive osteoclasts formed even in the presence of CP690,550 (Fig. 4D). c-Fos and NFATc1, transcription factors essential for osteoclast differentiation, were induced normally even in the presence of CP690,550 (Fig. 4E). These results suggest that the STAT3 plays a limited role in osteoclast differentiation of progenitor cells.

Fig. 4.

STAT3 inhibition blocks RANKL expression and osteoclastogenesis in vivo and in vitro. (A) Total RNA was prepared from wild-type (+/+) or STAT3 KO (−/−) MEFs treated with (+) or without (−) IL-1β (10 ng ml^–1) for 24 h, and expression of RANKL relative to β-actin was analyzed by quantitative real-time PCR. Data are means ± SD of the RANKL/β-actin (*P < 0.001; n = 3). (B) Total RNA was prepared from wild-type POBs treated with or without IL-1β (10 ng ml^–1) in the presence or absence of CP690,550 (100 nM) for 24 h, and expression of RANKL relative to β-actin was analyzed by quantitative real-time PCR. Data are means ± SD of RANKL/β-actin (*P < 0.001; n = 3). (C) Wild-type POBs and M-CSF-dependent osteoclast progenitor cells were co-cultured in the presence of OSM (50 ng ml^–1) or 1,25(OH)₂VitD₃ (10⁻⁷ M) + PGE₂ (1 μM) for 7 days, stained with TRAP (left panels) and the number of TRAP-positive cells containing more than three nuclei was scored (right panel) (*P < 0.001; n = 3). (D and E) M-CSF-dependent osteoclast progenitor cells were cultured in the presence or absence of RANKL (25 ng ml^–1) with or without indicated concentrations of CP690,550 for 5 days, stained with TRAP (left) and then the number of multinuclear TRAP-positive cells was scored (right). Data are means ± SD of multinuclear TRAP-positive cells (n = 6). Bar, 100 μm. (E) Total RNA was prepared from the above culture, and expression of c-Fos and NFATc1 relative to β-actin was analyzed by quantitative real-time PCR. Data are means ± SD of the indicated genes/β-actin (*P < 0.001; n = 3). (F) Total RNA was prepared from hind paw tissues of CIA-induced mice after 2 weeks of treatment by vehicle or CP690,550 (15 mg kg^–1 day^–1), and expression of RANKL and CTSK relative to β-actin was analyzed by quantitative real-time PCR. (G) Specimens of ankle joints from CIA-induced mice treated with vehicle or CP690,550 for 2 weeks were subjected to immunofluorescence staining for RANKL and CTSK. Nuclei were visualized by TOTO3. Bar, 100 μm. Representatives of at least two independent experiments are shown.

The mRNA of RANKL as well as Cathepsin K, an osteoclast differentiation marker, was highly induced in the joints of CIA model but was significantly inhibited by CP690,550 treatment compared with vehicle administration in vivo (Fig. 4F). Down-regulated RANKL and cathepsin K expression were also observed in CP690,550-treated CIA mice by immunofluorescence (Fig. 4G). Joint destruction seen in the CIA model was rescued by CP690,550 treatment (Fig. 5). Thus, inhibiting STAT3 in RA can inhibit the inflammatory cytokine loop triggering RANKL expression, thereby inhibiting joint destruction.

Fig. 5.

CP690,550 is effective in treating CIA in vivo. 6-week-old DBA/J1 male mice were given an initial injection of type 2 collagen on day −21 and arthritis was induced with a second injection on day 0. Vehicle or CP690,550 (15 mg kg^–1 day^–1) was administered interperitoneally once daily for 2 weeks from day 0 (n = 4 per group). Tissue specimens from the ankle of CIA mice administered vehicle or CP690,550 were stained with safranin O and methyl green. Bar, 100 μm. Representatives of at least two independent experiments are shown.

Antagonizing STAT3 inhibits differentiation of T_h17 cells

IL-17 producing helper T (T_h17) cells function in the pathogenesis of autoimmune diseases including RA (6, 7). Indeed, T_h17 cells accumulated in regional lymph nodes and spleen of CIA mice, and the appearance of T_h17 cells was slightly but significantly inhibited by CP690,550 treatment in vivo (Fig. 6A). In an ex vivo assay, IL-17, IL-6 and IFNγ production, which characterizes activation of T cells, induced by T-cell receptor oligomerization by anti-CD3 antibody or type 2 collage treatment was also significantly inhibited in the cells from CP690,550-treated CIA mice (Fig. 6B and data not shown). Interestingly, unlike the in vivo and ex vivo findings, differentiation of T_h17 cells in vitro induced by IL-6 and TGF-β was significantly stimulated by CP690,550 (Fig. 6C). A similar effect was reported by using another JAK inhibitor, pyridone6 (30). This effect was considered due to a more potent effect of the JAK inhibitor on IL-2, IL-4 and IFNγ than on IL-6, resulting in T_h17 differentiation. These results suggest that inhibition of T_h17 development by CP690,550 in vivo is likely due to IL-6 inhibition. Regulatory T cells (Tregs) are critical to inhibit autoimmune diseases (31), and CP690,550 likely activates Tregs. However, Treg development was inhibited in vitro and unchanged in vivo following CP690,550 treatment (SupplementaryFigure S5 is available at International Immunology Online). Taken together, our data suggests that STAT3 is a critical factor promoting sustained inflammation and osteoclastogenesis leading to chronic inflammation and joint destruction (Fig. 7) and that inhibition of STAT3 activation could constitute a novel anti-RA therapy.

Fig. 6.

CP690,550 suppresses T_h17 cells in vivo and ex vivo. (A) CD4 and intracellular staining of IL-17 in inguinal lymph node cells from CIA-induced mice harvested on day 21 after immunization were examined by flow cytometry (left), and the frequency of CD4+IL-17+ cells (%) was determined (right). (B) Inguinal lymph node cells from CIA-induced mice harvested on day 21 after immunization were cultured with or without anti-CD3 or type 2 collagen (100 μg ml^–1) for 5 days, and IL-6, IL-17 and IFNγ levels in the supernatant were assessed by ELISA (*P < 0.001; n = 3). (C) Naive T cells were cultured with 1 μg ml^–1 plate-bound anti-CD3 antibody (Ab), 0.5 μg ml^–1 soluble anti-CD28 Ab, 0.2 ng ml^–1 IL-2, 2 ng ml^–1 TGF-β1 and IL-6 (20 ng ml^–1) in the presence or absence of CP690,550 (50 nM) for 3 days. Foxp3 and IL-17A expression was then analyzed by flow cytometry (*P < 0.001; n = 3). Representatives of at least two independent experiments are shown.

Fig. 7.

Scheme of sustained inflammation and joint destruction via STAT3 in RA. (A) Major pro-inflammatory cytokines, namely, IL-6, TNFα and IL-1β, stimulate STAT3 either directly or indirectly. (B) STAT3 activation is followed by expression of IL-6 family cytokines, which in turn induce STAT3 activation in an autocrine/paracrine manner. STAT3 activation further induces RANKL expression, which promotes osteoclastogenesis and joint destruction. Thus, sustained inflammation and joint destruction are developed via STAT3, indicating that STAT3 could be a key target in treating RA.

Discussion

RA is a chronic inflammatory disease, which causes continuous joint swelling, pain and damage that limit patients’ quality of life (32). To date, various biological agents targeting major pro-inflammatory cytokines up-regulated in RA, such as TNFα, IL-6 and IL-1, have been established, and each of these therapies promotes significant reduction of disease activity (22, 33, 34). However, some populations of RA patients still suffer from continuous inflammation and joint problems (35). Thus, crucial targets to inhibit disease activity in these cases have been sought. Mechanisms underlying sustained inflammation and joint damage in RA remain largely unknown. In this study, we demonstrated that the major pro-inflammatory cytokines in RA, TNFα, IL-6 and IL-1, induced STAT3 activation either directly or indirectly and stimulated expression of IL-6 family cytokines and RANKL in an autocrine/paracrine manner in vivo and in vitro. Induced IL-6 family cytokines further activated STAT3 and then in turn induced IL-6 family cytokines and RANKL expression in a cytokine amplification circuit. RANKL induces osteoclastogenesis causing joint destruction by bone resorption. Thus, a loop including inflammatory cytokines, IL-6 family cytokines, RANKL and STAT3 loop likely mediates sustained inflammation and joint destruction in RA. Targeting STAT3 is likely crucial to block that loop.

High expression of inflammatory cytokines or the presence of auto-antibodies and autoreactive lymphocytes is observed and implicated in disease activity in RA and in RA models such as CIA, antigen-induced arthritis. Both were also observed in several reports of spontaneous autoimmune arthritis and in genetically engineered mouse models, including IL-1 Tg, TNF Tg, gp130/F759, ZAP70 SKG and K/BxN (36). In fibroblasts, an IL-17A-triggered positive-feedback loop of IL-6 expression through NFκB and STAT3 activation reportedly promotes arthritis (6), and such IL-17A- and IL-6-dependent arthritis is induced by local events such as microbleeding (7). Indeed, we found that IL-6 expression was significantly down-regulated by STAT3 inhibition caused by gene deletion or CP690,550 treatment in osteoblastic cells, fibroblasts and sera in vivo. IL-6 was reported as an amplifier of inflammatory cytokines (6, 7), and we found that STAT3 activation by IL-1 was reduced in IL-6 KO cells, suggesting that STAT3 was mainly activated by IL-6. However, we also found that STAT3 activation was reduced but still induced in IL-6 KO cells, suggesting that STAT3 was also stimulated likely by other IL-6 family cytokines induced by IL-1β, such as LIF, which was up-regulated in IL-6 KO cells by IL-1β or TNFα stimulation (Fig. 1G). In addition, NFκB activation was not inhibited and T_h17 cell development was rather stimulated by CP690,550, suggesting that STAT3 activation described here is, at least in part, IL-17A and IL-6 independent. Thus, we propose that IL-6–STAT3 is a key modulator of RA-related symptoms by inducing cytokine amplification.

STAT3 is reportedly a substrate of spleen tyrosine kinase (syk) in B-lineage leukemia/lymphoma cells, and syk is required for oxidative stress-mediated STAT3 activation in tumors (37). A Syk inhibitor was shown to be effective in inhibiting arthritis in the CIA model (38), suggesting that the Syk–STAT3 pathway is likely activated in RA. Inhibitors of Jak1 and Jak2, both upstream of STAT3, effectively improved CIA (39). Furthermore, IL-6 stimulation was shown to induce STAT3 activation as well as complex formation between STAT3 and cyclin-dependent kinase (CDK) 9 (40). Local gene transfer of CDK-inhibitors such as p16^INK4a or p21^Cip1 was also effective in treating CIA (41), suggesting that CDK potentially regulates STAT3 activation.

CP690,550 was originally developed as a Jak3 inhibitor and to a much lesser extent as an inhibitor of Jak1 and Jak2 to mediate immunosuppression (10). The CP690,550 IC50 indicated that it was more effective in inhibiting Jak3 than Jak1 and Jak2 (10, 42). Recently, CP690,550 was also shown to inhibit STAT1 activity (13). In our study, CP690,550 treatment inhibited a cytokine loop via STAT3 inhibition, indicating that STAT3 could be a therapeutic target in chronic inflammatory diseases.

Since IL-6 overproduction has been seen in RA patients and RA models, antagonizing IL-6 activity may be beneficial in treating RA. Indeed, a mAb to IL-6R has proven to be effective against RA. Previously, we reported that over-expression of suppressor of cytokine signaling 3 (SOCS3) or dominant negative STAT3 in joints effectively ameliorated CIA models (43). At that time, we proposed that blocking an intracellular cytokine signaling, particularly JAK1, could be effective against RA treatment. Our study here extends the potential therapeutic effects of JAK–STAT3 inhibitors and shows that they block an inflammatory cytokine-IL-6 circuit that amplifies STAT3 activation.

How does STAT3 contribute to RA? STAT3 has been shown to be necessary for synovial fibroblast proliferation (44), RANKL expression (45), osteoclastegenesis (46) and T_h17 induction (47). We have shown that CP690,550 inhibits many of these effects. Since expression of IL-6, which is required for T_h17 cell development, was significantly inhibited by CP690,550 in vivo (Fig. 6A), the reduced frequency of T_h17 cell fraction seen following treatment with CP690,550 in vivo is likely due to reduced IL-6 levels caused by the drug. However, the T_h17 cell fraction was enhanced by CP690,550 in vitro (Fig. 6C). The precise mechanisms underlying this discrepancy are not clear, but it is likely that CP-690550 inhibits other factors. Indeed, CP690,550 reportedly inhibits other Jaks and STATs (13), and a pan-Jak inhibitor, pyridone6, also stimulates T_h17 development in vitro by inhibiting STAT5 (30). Similarly, Treg development in vitro was inhibited by CP690,550, and such inhibition is also induced by pyridone6 (30), further suggesting that the in vitro effects of CP690,550 on T_h17 and Treg are due, at least in part, to inhibition of Jaks and STATs. Therefore, more specific STAT3 inhibitors are likely required for maximal therapeutic effect.

In RA, osteoclastic bone resorption is frequently elevated and is implicated in joint destruction (14). Some RA patients are treated with steroids, and steroid treatment potently induces osteoporosis, a disease marked by increased osteoclast activity. Thus, control of osteoclastogenesis is crucial for RA patients to avoid joint destruction or drug-induced osteoporosis. RANKL is an essential cytokine for osteoclastogenesis, and RANKL deficiency results in a complete lack of osteoclast formation (48). Thus, RANKL is a target to inhibit osteoclastogenesis in osteoporosis (49) and likely in joint destruction. RANKL expression is reportedly induced by various stimuli such as vitamin D3, parathyroid hormone or IL-6 family cytokines via vitamin D receptor, cyclic adenosine monophosphate or STAT3. In our study, RANKL expression was significantly inhibited in STAT3-deficient cells or in CIA joints of CP690,550-treated mice, suggesting that RANKL expression in RA is regulated by a STAT3-dependent mechanism. Thus, inhibiting STAT3 could benefit to inhibit both inflammation and osteoclastogenesis mediated by an IL-6–STAT3-dependent cytokine loop.

Supplementary data

Supplementary data are available at International Immunology Online.

Funding

Special grants-in-aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan, the Japan Society for the Promotion of Science, the Takeda Science Foundation, the Program for the Promotion of Fundamental Studies in Health Science of the National Institute of Biomedical Innovation, the Mochida Memorial Foundation, and the Uehara Memorial Foundation (A.Y.); grant-in-aid for Young Scientists (A) and by Precursory Research for Embryonic Science and Technology, the Takeda Science Foundation and the Keio Kanrinmaru project, Japan (T.M.); Global Center of Excellence Program at Keio University (T.M.).

We thank M. Asakawa, K. Fukuse, N. Shiino, Y. Sato and T. Kobayashi for technical assistance.