Follistatin-like protein-1 (FSTL-1) is a poorly characterized protein that is up-regulated in the early stage of collagen-induced arthritis and that exacerbates arthritis when delivered by gene transfer. The current study was designed to determine the mechanism by which FSTL-1 promotes arthritis. FSTL-1 was injected into mouse paws, resulting in severe paw swelling associated with up-regulation of IFN-γ transcript and the IFN-γ-induced chemokine, CXCL10. Mice depleted of T cells were protected. A central role for IFN-γ was confirmed by the finding that mice deficient in IFN-γ failed to exhibit paw swelling in response to injection of FSTL-1. Furthermore, IFN-γ secretion from mouse spleen cells exposed to a weak TCR signal was increased 5-fold in the presence of FSTL-1. FSTL-1 could be induced by innate immune signals, including TLR4 agonists and the arthritogenic cytokine, IL-1β, via an NFκB pathway. Finally, FSTL-1 was found to be overexpressed in human arthritis and its neutralization inhibited murine collagen-induced arthritis and suppressed IFN-γ and CXCL10 production in arthritic joints. These findings demonstrate that FSTL-1 plays a critical role in arthritis by enhancing IFN-γ signaling pathways and suggest a mechanism by which FSTL-1 bridges innate and adaptive immune responses.
While analyzing gene expression in collagen-induced arthritis (CIA),3 we discovered that a poorly characterized gene, follistatin-like protein 1 (FSTL-1), is highly overexpressed in mouse paws during early arthritis, especially at the interface of synovial pannus and eroding bone (1), suggesting that it plays a role in arthritis. FSTL-1, also known as FRP and TSC-36, is an extracellular glycoprotein belonging to the osteonectin (BM-40/SPARC) family of proteins containing both extracellular calcium-binding and follistatin-like domains. FSTL-1 was originally cloned from an osteoblast cell line as a TGF-β inducible gene (2). FSTL-1 was detected in the medium of all osteosarcoma and chondrosarcoma cell lines tested, and in some cells of the fibroblast lineage. FSTL-1 is highly conserved across mammalian species. Human and mouse FSTL-1 share 92% identity in their amino acid sequence.
In 1998, Tanaka et al. (3) cloned FSTL-1 from rheumatoid arthritis (RA) synovial tissue and demonstrated anti-FSTL-1 Abs in the serum and synovial fluid of RA patients, and suggested that FSTL-1 was an autoantigen. This group further reported that administration of human FSTL-1 to BALB/c mice with Ab-induced arthritis ameliorated disease (4), possibly by reducing synovial production of matrix metalloproteinases (5). The effect was modest and our own group subsequently demonstrated that FSTL-1 is a novel proinflammatory molecule with a previously unrecognized role in inflammation (6). Transfection of FSTL-1 into macrophages and fibroblasts leads to up-regulation of proinflammatory cytokines, including IL-1β, TNF-α, and IL-6. Overexpression of FSTL-1 in CIA results in significant exacerbation of arthritis. In addition, expression of FSTL-1 in normal mouse paws by gene transfer results in severe paw swelling and arthritis.
The present study was designed to further characterize the mechanism of action of this novel inflammatory molecule and to determine whether endogenous FSTL-1, produced during the inflammatory phase of CIA, contributes to arthritis.
Materials and Methods
A recombinant, E1a-E3-deleted replication defective adenovirus type 5 vector encoding the mouse FSTL-1 gene (NCBI Nucleotide database accession number BC028921) was generated through Cre-lox recombination as described by Hardy et al. (7). The control vector, Ad-BglII, is an E1a/E3-deleted replication-defective adenovirus type 5 lacking an insert. The vectors were grown in 293 cells and purified by CsCl gradient ultracentrifugation, dialyzed at 4°C against sterile virus buffer, aliquoted, and stored at −80°C. FSTL-1 expression was verified by both RT-PCR and Western blot from infected COS-7 cells.
Male IFN-γ-deficient C.129S7(B6)-Ifngtm1Ts/J and wild-type BALB/c mice were purchased from The Jackson Laboratory. Male DBA/1 mice, 6–10 wk of age, were purchased from Harlan Sprague Dawley. Mice were housed in the animal resource facility at the Children’s Hospital of Pittsburgh Rangos Research Center (Pittsburgh, PA). The study was approved by the Children’s Hospital of Pittsburgh’s Animal Research and Care Committee.
Induction and assessment of arthritis
CIA was induced by intradermal immunization of DBA/1 mice with bovine collagen type II (Elastin Products) and a booster given i.p. 21 days later, as previously described (8). Mice were evaluated for arthritis several times weekly by a blinded observer using a macroscopic scoring system ranging from 0 to 4 (0 = no detectable arthritis; 1 = swelling and/or redness of paw or one digit; 2 = two joints involved; 3 = three to four joints involved; and 4 = severe arthritis of entire paw and digits). The arthritic index for each mouse was calculated by adding the score of the four individual paws. The statistical significance was determined using the exact Wilcoxon test. Paw swelling was measured using calipers. p values <0.05 were considered significant. Some mice were injected in the paws with 1 × 109
Induction of FSTL-1
To inhibit NF-κB, the mouse osteoblast cell line MC3T3 (9) was infected with a retrovirus encoding a super IκB repressor and containing a puromycin resistance gene (10). Following 1 wk of selection in 2 μg/ml puromycin, confirmation of NF-κB inhibition was determined by transfecting cells with a plasmid encoding the luciferase reporter gene under control of an NF-κB responsive element (provided by Dr. J. Kolls, Children’s Hospital of Pittsburgh, Pittsburgh, PA) using the Fugene 6 transfection reagent as described by the manufacturer (Roche Diagnostics). Cells were assayed after 48 h for luciferase activity using the Promega Bright-Glo Luciferase Assay System according to the manufacturer’s instructions (Promega). Luciferase units were measured using a Packard microplate scintillation and luminescence counter. Control and super IκB-expressing MC3T3 cells (1 × 104 cells/well) were cultured for 3 days in triplicate in medium, TGF-β (2 ng/ml), LPS (100 ng/ml), or IL-1β (10 ng/ml) and supernatants were assayed by ELISA for FSTL-1.
Mouse paws were fixed in 10% neutral buffered formalin, decalcified, dehydrated in a gradient of alcohols, paraffin embedded, sectioned, mounted on glass slides, and stained with H&E as previously described (6).
FSTL-1 enhances T cell IFN-γ production
We previously found that injection of adenovirus encoding FSTL-1 into mouse paws results in severe paw swelling and synovitis (6). To understand how FSTL-1 mediates this effect, we incubated mouse spleen cells with FSTL-1 for 24 h and performed DNA microarray analysis on mRNA from these cells. Up-regulation of a number of IFN-γ-related genes was observed. Because T cells are a major source of IFN-γ we assessed the ability of FSTL-1 to induce inflammation in mice depleted of mature αβ+ T cells by treatment with the anti-αβ mAb, H57 (11). Hind paws were injected with Ad(FSTL-1) or a control adenovirus, Ad(BglII). Depletion of αβ+ T cells abrogated FSTL-1-induced paw swelling (Fig. 1⇓a). In a separate experiment, hind paws were injected with Ad(FSTL-1) or Ad(BglII) and mice were killed on days 1, 3, 6, and 8 and mRNA from the hind paws was assayed by real time PCR. Message for IFN-γ (Fig. 1⇓b) as well as the IFN-γ-induced chemokine, CXCL10 (Fig. 1⇓c) increased substantially with the appearance of paw swelling, suggesting a central role for IFN-γ. Other proinflammatory cytokines, including IL-1β (Fig. 1⇓d) and TNF-α (data not shown) were also increased. To confirm that FSTL-1 mediates its inflammatory effect through induction of IFN-γ, IFN-γ-null mice, and wild type controls were administered Ad(FSTL-1). FSTL-1-induced paw swelling was abrogated in mice deficient in IFN-γ (Fig. 1⇓d).
To further explore the ability of FSTL-1 to enhance T cell responses, we incubated mouse spleen cells with purified FSTL-1 protein. FSTL-1 alone induced a small, but statistically significant, amount of IFN-γ (Fig. 1⇑e). A strong synergistic effect was observed in the presence of a weak TCR signal delivered by low-titer anti-CD3 mAb, 2C11. At a titer of 8 ng/ml 2C11, coculture with FSTL-1 increased IFN-γ production by 5-fold. This effect was not observed with purified T cells, indicating that the activity of FSTL-1 requires an accessory cell population.
FSTL-1 is induced in response to mediators of innate immunity
FSTL-1 was originally described as a TGF-β inducible gene derived from the osteoblast cell line, MC3T3 (2). TGF-β signals through the SMAD pathway (12, 13). However, the ability of FSTL-1 to enhance T cell responses suggested a similarity to NF-κB-dependent cytokines, such as TNF-α, that are induced by innate immune signals and can enhance T cell activation (14, 15). To determine whether FSTL-1 is induced by innate signaling, we incubated MC3T3 cells with the TLR4 agonist, LPS, or with IL-1β, both of which signal through the NF-κB pathway. Both LPS and IL-1 β induced FSTL-1 (Fig. 2⇓a). A second group of MC3T3 cells were infected with a retrovirus encoding a super IκB inhibitor, in which the IκBα molecule is mutated. The mutation prevents the molecule from being phosphorylated by IκB kinase and subsequently degraded, which is normally required for the translocation of NF-κB to the nucleus (16). To confirm that NF-κB was inhibited, the cells were transfected with a reporter plasmid containing an NF-κB responsive element and encoding luciferase. Two days after transfection, luciferase activity was measured. The cells containing the inhibitor had ∼10-fold lower luciferase activity (data not shown), indicating that NF-κB activation was impaired. These super-IκB-expressing MC3T3 cells lost the ability to produce FSTL-1 in response to LPS and IL-1β, while still responding to TGF-β, demonstrating that FSTL-1 can be induced through NFκB in addition to SMAD. We observed a similar induction of FSTL-1 in vivo following injection of LPS (Fig. 2⇓b) or IL-1β (Fig. 2⇓d) into mouse paws, as well as injection of CFA (Fig. 2⇓c).
FSTL-1 is overexpressed in RA synovium
We have previously found FSTL-1 mRNA in RA synovial tissues (1). To determine whether this expression was constituitive or reflected active synovitis, RA synovial tissues were assayed by real-time PCR. RA synovium had a 2-fold increase in FSTL-1 mRNA, compared with control synovium obtained from patients undergoing knee arthroscopic anterior cruciate ligament repair (Fig. 3⇓a). The magnitude of this increase was similar to the induction of FSTL-1 mRNA in the paws of mice with CIA (Fig. 3⇓b). This finding is also consistent with a previous report suggesting that FSTL-1 is overexpressed by 2–3-fold in human RA synovial tissue, as compared with tissue from osteoarthritis (17). Together, these findings provide strong evidence for a role of FSTL-1 in human RA.
Endogenously produced FSTL-1 plays a role in CIA
We have previously demonstrated that administration of FSTL-1 by gene transfer exacerbates CIA (6), demonstrating that exogenously administered FSTL-1 could exacerbate arthritis. To determine whether endogenous FSTL-1 plays a role in CIA, we generated and affinity-purified rabbit anti-mouse FSTL-1 IgG and used it to neutralize FSTL-1 activity in vivo. DBA/1 mice were immunized with type II collagen (CII) on days 0 and 21 to induce CIA. On days 20, 22, 24, 26, and 28, mice were injected i.p. with 200 μg of either rabbit anti-FSTL-1 Ab or with rabbit IgG as a control. A third group of mice received no Ab. A substantial reduction of arthritis was observed in mice treated with anti-FSTL-1 Ab (Fig. 4⇓, a–c), indicating that endogenous FSTL-1 does indeed play a proinflammatory role in CIA, as its neutralization ameliorates arthritis. The incidence of arthritis was not reduced.
Analysis of paws demonstrated increased mRNA for IFN-γ CXCL10, and IL-1β on day 35 of CIA (Fig. 4⇑d). All of these were significantly reduced, down to baseline expression, in mice treated with anti-FSTL-1. These results support the conclusion from Fig. 1⇑ that FSTL-1 up-regulates IFN-γ and related genes. This finding is especially relevant to arthritis, because it has recently been demonstrated that CXCL10 is a central mediator of bone erosion in CIA (18). Bone and cartilage erosion was found to be substantially reduced in CIA following treatment of mice with anti-FSTL1 (Fig. 4⇑, e and f).
We have previously reported that over-expression of FSTL-1 by gene transfer exacerbates mouse CIA (6), demonstrating that exogenously administered FSTL-1 could exacerbate arthritis. This finding suggested, but did not conclusively demonstrate that endogenously produced FSTL-1 plays a role in arthritis. The current study indeed shows conclusively that endogenous FSTL-1 plays a proinflammatory role in CIA, as its neutralization ameliorates arthritis. Although we cannot be certain that the same is true in human RA, our finding that FSTL-1 is up-regulated in RA synovial tissues suggests that this might be the case. Furthermore, human and mouse FSTL-1 share 92% identity in their amino acid sequences.
These results shed further light on the possible mechanism of action of FSTL-1. FSTL-1 by itself has limited ability to induce IFN-γ. However, in the context of T cell activating signals, it functions to up-regulate IFN-γ secretion. In vitro, we observed that FSTL-1 dramatically increases IFN-γ secretion when T cells are stimulated with a weak signal provided by low-dose anti-CD3. We therefore hypothesize that a similar mechanism of action occurs in vivo. Paw swelling in response to Ad(FSTL-1) is only observed a week after injection. This would be explained if sufficient activated T cells need to reach the paws in response to adenovirus infection before FSTL-1 can act. This would also explain why FSTL-1, which is constituitively expressed at low levels in the joint (1), does not induce inflammation in the normal situation. The proinflammatory effect of FSTL-1 would require the presence of activated T cells. Further studies, now underway, will allow us to test this hypothesis.
These results indicate that FSTL-1 is induced by innate immune mediators central to arthritis. IL-1β is an especially important arthritis-promoting cytokine and its ability to induce FSTL-1 might explain the up-regulation of FSTL-1 observed in both mouse and human arthritic joints (1). The increased FSTL-1 would then be available to act upon activated T cells, leading to further inflammation. Thus, FSTL-1 may function as a bridge between innate and adaptive immune responses by being produced in response to innate signals and then amplifying T cell responses. In this fashion, it might function analogously to TNF which can promote T cell responses (14, 15). It has been suggested that signal strength drives T cells through hierarchical thresholds associated with proliferation preceding the acquisition of fitness and effector functions (19, 20). At the two extremes are anergy after a very weak signal strength and activation-induced cell death after excessive stimulation. The stimulation strength is determined by at least three independent parameters (Ag dose, costimulation, and duration). FSTL-1 may function to enhance stimulation strength following T cell activation by an as yet unknown mechanism, thereby helping to drive T cells to a mature effector state.
These findings also support previous studies demonstrating a role for IFN-γ in arthritis. We have previously shown that IFN-γ protects against CIA when given before disease onset but exacerbates CIA when given after disease onset (21, 22) consistent with the observation that IFN- γ-deficient mice on the C57BL/6 background are more susceptible than wild-type C57BL/6 (23). Thus, induction of IFN-γ by FSTL-1 would be a reasonable mechanism to explain its arthritogenic effect.
In summary, we provide evidence that FSTL-1 is a novel arthritogenic protein that plays a central role in arthritis through promotion of T cell activation and induction of IFN-γ. Studies now underway seek to understand the precise molecular mechanisms by which FSTL-1 acts.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 This work was supported in part by National Institutes of Health Grants AR052282 and AR48929, and the Children’s Hospital of Pittsburgh.
↵2 Address correspondence and reprint requests to Dr. Raphael Hirsch, Division of Rheumatology, Children’s Hospital of Pittsburgh, 3705 Fifth Avenue, Pittsburgh, PA 15213. E-mail address:
↵3 Abbreviations used in this paper: CIA, collagen-induced arthritis; FSTL-1, follistatin-like protein 1; RA, rheumatoid arthritis; CII, type II collagen.
- Received July 9, 2008.
- Accepted October 21, 2008.
- Copyright © 2009 by The American Association of Immunologists, Inc.