c-Fms-mediated differentiation and priming of monocyte lineage cells play a central role in autoimmune arthritis

In the in vitro studies, we used imatinib mesylate that was chemically synthesized and confirmed to be more than 98% pure by the Organic Synthesis Core Facility at Memorial Sloan-Kettering Cancer Center (New York, NY, USA). In the in vivo studies, we used imatinib mesylate tablets (Stanford Inpatient Pharmacy Services, Palo Alto, CA, USA), which were ground and resuspended in the vehicle. GW2580 provided by GlaxoSmithKline (Uxbridge, Middlesex, UK) was used in the studies on prevention of arthritis (Figures 1 and 2). GW2580 purchased from Calbiochem (San Diego, CA, USA) and GW2580 chemically synthesized and confirmed to be more than 99% pure by SRI International (Menlo Park, CA, USA) were used in the studies on the treatment of arthritis (Figures 1 and 2), the experiments shown in Figures 3, 4, 5 and 6, and the experiments shown in Additional file 1. Anti-c-Fms, anti-phospho-c-Fms, and isotype control antibodies were from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA).

c-Kit and Abl kinase activity in the presence or absence of small-molecule inhibitors was determined by using HTScan kinase assay kits (Cell Signaling Technology, Inc., Danvers, MA, USA) coupled with europium-labeled DELFIA assays (PerkinElmer, Waltham, MA, USA), and counts were measured by time-resolved fluorescence (PerkinElmer) in accordance with the protocols of the manufacturer. To assess c-Fms activity, we incubated human peripheral blood mononuclear cells with 20 ng/mL M-CSF in the presence or absence of small-molecule inhibitors and determined the percentage of macrophages, as described below. To assess PDGFR activity, we isolated human FLSs as previously described [12], stimulated them for 72 hours with 20 ng/mL PDGF-bb in the presence of small-molecule inhibitors, pulsed them with 1 μCi [³H] thymidine (ICN Pharmaceuticals, Costa Mesa, CA, USA) for the final 18 hours of the stimulation, and used a Betaplate scintillation counter (PerkinElmer) to quantify the radioactivity incorporated. Scintillation counts were used to generate nonlinear regression dose-response curves for each small-molecule inhibitor, and IC₅₀s (half inhibitory concentrations) were determined by using Prism software (GraphPad Software, Inc., San Diego, CA, USA).

Human synovial fluid and synovial tissue samples were collected from RA, osteoarthritis (OA), and psoriatic arthritis (PsA) patients who met the American College of Rheumatology criteria. Samples were collected in accordance with protocols approved by the Stanford University Institutional Review Board after procurement of informed consent.

Six- to eight-week-old male DBA/1 mice and female BALB/c mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA) and housed at Stanford University under protocols approved by the Stanford University Committee of Animal Research and in accordance with National Institutes of Health guidelines. Collagen-induced arthritis (CIA) in DBA/1 mice was induced and scored as previously described [23]. Briefly, DBA/1 mice were immunized by intradermal injection of 100 μg/mouse bovine collagen type II (CII) (Chondrex, Inc., Redmond, WA, USA) emulsified in complete Freund's adjuvant (CFA) containing 250 μg/mouse heat-killed Mycobacterium tuberculosis H37Ra (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). Twenty-one days after immunization, mice were given a subcutaneous boost injection (at the base of the tail) of 100 μg/mouse bovine CII emulsified in incomplete Freund's adjuvant (IFA). In BALB/c mice, anti-collagen antibody-induced arthritis (CAIA) was induced by intravenous injection of 1 mg of Arthrogen monoclonal antibody blend (Chondrex, Inc.) followed by 25 μg of lipopolysaccharide (LPS) (Chondrex, Inc.) 3 days later. K/BxN arthritis was induced in BALB/c mice by intraperitoneal (i.p.) injection of 1 μL of K/BxN serum per 1 g of mouse weight, followed 48 hours later by i.p. injection of 0.5 μL of K/BxN serum per 1 g of mouse weight. Arthritis severity was evaluated according to the following visual scoring system: 0 = no swelling or erythema; 1 = mild swelling and erythema of digits or paw; 2 = moderate swelling and erythema confined to the area distal to the mid-paw; 3 = more-pronounced swelling and erythema extending to the ankle; 4 = severe swelling, erythema, and joint rigidity of the ankle, foot, and digits. Each limb was assigned a score of 0 to 4, with a maximum possible score of 16 for each mouse. Paw thickness was determined by measuring the thickness of both hind paws with 0- to 10-mm calipers and calculating the mean of the two measurements.

Hind limbs from mice with autoimmune arthritis were fixed and decalcified in CalEx II (Fischer Scientific, Pittsburgh, PA, USA) for 3 days before being paraffin-embedded. Histological assessment of arthritis severity was made by blinded evaluation of Toluidine blue-stained joint sections in accordance with a previously described scoring system: 0 = normal; 1 = mild inflammation, mild hyperplasia of the synovial lining layer, and mild cartilage destruction without bone erosion; 2 to 4 = increasing degrees of inflammatory cell infiltrates, synovial lining hyperplasia and pannus formation, and cartilage and bone destruction [24].

Sections of paraffin-embedded synovium from RA patients and decalcified joint tissue from mice with autoimmune arthritis were deparaffinized, rehydrated, and subjected to antigen retrieval as described previously [25, 26].

Bone marrow cells were harvested from BALB/c mice and monocyte lineage cells were generated according to standard procedures [27]. After 4 to 5 days of culture, bone marrow-derived monocytes were incubated for 48 hours with 20 ng/mL M-CSF (PeproTech, Rocky Hill, NJ, USA) in the presence of 0 to 10 μM GW2580 or imatinib. To distinguish between monocytes and macrophages, we performed an α-napthyl acetate esterase assay, coupled with fluoride inhibition, in accordance with the protocol of the manufacturer (Sigma-Aldrich). At least 100 monocytes and macrophages were counted in triplicate for each experimental condition, and data are expressed as a percentage of macrophages in culture.

Twenty-four hours after their isolation from BALB/c mice, undifferentiated bone marrow cells were transferred to dentine disks (Immunodiagnostic Systems, Scottsdale, AZ, USA) or osteologic disks (BD Biosciences, San Jose, CA, USA) and cultured for 6 days in the presence of 50 ng/mL M-CSF and 50 ng/mL receptor activator of nuclear factor-kappa-B ligand (RANKL) (PeproTech) together with 0 to 5 μM small-molecule inhibitor. To identify multinucleated, tartrate-resistant acid phosphatase-positive (TRAP⁺) osteoclasts, we stained cells cultured on dentine disks with the acid phosphatase leukocyte kit (Sigma-Aldrich). ImageJ software was used to determine the dentine disk area that stained positive for TRAP⁺ multinucleated cells. Pit formation was assessed by measuring the removal of surface film on osteologic disks with the Bioquant Osteo II image quantification system (Bioquant Image Analysis Corporation, Nashville, TN, USA).

Bone marrow cells were harvested from BALB/c mice and macrophages were generated as previously described [27]. Macrophages were cultured overnight in complete RPMI media in the absence of M-CSF and then incubated for 3 hours in the presence of 0 to 50 ng/mL M-CSF and 0 to 5 μM small-molecule inhibitor, as described above. After 3 hours, cells were stimulated with 1 ng/mL LPS (Sigma-Aldrich) or 20 μg/mL plate-bound rat anti-mouse 2.4G2 (BD Biosciences) for 24 hours, as previously described [28], and supernatants were harvested for cytokine analysis by enzyme-linked immunosorbent assay (ELISA).

Splenocytes from CIA mice treated chronically with 80 mg/mL GW2580, 80 mg/mL imatinib, or vehicle were stimulated for 72 hours with 20 μg/mL whole, denatured bovine CII (Chondrex, Inc.). One microcurie of [³H] thymidine (ICN Pharmaceuticals) was added for the final 18 hours of culture, and radioactivity incorporation was quantified by using a Betaplate scintillation counter. Supernatants after 72 hours were harvested for cytokine analysis by ELISA.

Visual arthritis scores, paw thicknesses, and histology scores were compared by the Mann-Whitney U test with GraphPad InStat Version 3.0 (GraphPad Software, Inc.). Differences in arthritis scores were determined by the Fisher test with Analyse-it plug-in software (Analyse-it Software, Ltd., Leeds, UK) for Excel (Microsoft Corporation, Redmond, WA, USA). Macrophage differentiation, osteoclast differentiation, macrophage priming, and cytokine level were compared by unpaired t tests with GraphPad InStat Version 3.0 (GraphPad Software, Inc.).

To determine whether specific inhibition of c-Fms provides benefit in autoimmune arthritis, we explored the effects of GW2580 in several distinct models of RA and compared them with the effects of imatinib. Imatinib inhibits c-Kit, Abl, PDGFR, and c-Fms with IC₅₀s of 0.1, 0.25, 0.1, and 1.4 μM, respectively. On the basis of published pharmacokinetic profiles [12], imatinib was administered to mice orally, twice daily at a dose of 80 mg/kg. GW2580 was administered to mice orally, twice daily at doses of 30 or 80 mg/kg. Previous pharmacokinetic studies in mice have determined that oral administration of 80 mg/kg GW2580 yields a maximal plasma concentration of 5.6 μM [29]. To determine the IC₅₀ of GW2580 for the kinases c-Kit and Abl, we used cell-free kinase assays with time-resolved fluorescence. The IC₅₀s were 73.5 μM for Abl (Additional file 1a) and greater than 100 μM for c-Kit (Additional file 1b) and concentrations significantly above the maximal plasma concentrations of GW2580 achieved in mice receiving 80 mg/kg GW2580. Using cell-based assays, we showed that GW2580 potently inhibits c-Fms (IC₅₀ = 0.01 μM; Additional file 1c) and can inhibit PDGFR only at supraphysiological concentrations (IC₅₀ = 12.1 μM; Additional file 1d). Thus, dosing of mice with GW2580 at a concentration of 80 mg/kg or less should inhibit c-Fms but not Abl, c-Kit, or PDGFR. Indeed, in a cell-free assay that measures the specificity of small-molecule inhibitors, GW2580 at 10 μM abolished c-Fms activity and did not cross-react with nearly 200 other kinases [30].

CIA was induced by injection of DBA/1 mice with bovine CII emulsified in CFA, followed 21 days later by a boost injection of CII emulsified in IFA. When imatinib dosing was initiated 1 day before the induction of CIA, it significantly reduced the severity of arthritis (Figure 1a, b), in agreement with our previous findings [12]. Likewise, mean arthritis scores and paw thickness measurements were significantly lower in mice dosed prophylactically with 30 or 80 mg/kg GW2580 compared with mice dosed with vehicle. GW2580 was as efficacious as imatinib in preventing the development of arthritis. Furthermore, when the kinase inhibitors were administered after the induction of arthritis, both GW2580 and imatinib significantly inhibited the progression of arthritis (Figure 1c, d). Mice were sacrificed between days 48 and 50 as this represents the peak of synovitis and inflammation. In the CIA experiments presented, all mice developed arthritis by the time the experiment was terminated (100% incidence).

Imatinib has been shown to ameliorate CAIA [10]. We performed experiments to determine whether specific inhibition of c-Fms would yield a similar benefit in CAIA. We induced CAIA by injecting BALB/c mice with 1 mg of anti-collagen antibodies, followed by 25 μg of LPS 3 days later. Administration of GW2580 or imatinib was started 1 day before the transfer of antibodies. All CAIA mice developed arthritis by day 6 after antibody transfer (100% incidence). Arthritis was significantly less severe in CAIA mice treated with the c-Fms-specific inhibitor GW2580 compared with vehicle-treated CAIA mice (Figure 1e, f). The course of arthritis in GW2580-treated CAIA mice mirrored that in imatinib-treated CAIA mice.

We induced K/BxN arthritis in BALB/c mice by transferring 1 μL of serum/g of mouse weight, followed by 0.5 μL of serum/g of mouse weight 48 hours later. Administration of GW2580 or imatinib was initiated 1 day before the transfer of serum. All K/BxN mice developed arthritis by day 4 after serum transfer (100% incidence). Arthritis was significantly less severe in K/BxN mice treated with GW2580 or imatinib compared with vehicle-treated K/BxN mice (Figure 1g, h).

Histological analysis was performed on hind paws harvested from mice treated with 80 mg/kg GW2580, 80 mg/kg imatinib, or vehicle in the studies described above. Representative images of Toluidine blue-stained joint tissue sections from GW2580-, imatinib-, and vehicle-treated mice in the CIA prevention studies are presented (Figure 2a). Histopathologic evaluation by an investigator blinded to treatment groups for synovitis, formation of pannus, and erosion of cartilage and bone in paws derived from mice in CIA prevention (Figure 2b, n = 10 mice per group, with both hind paws from each mouse scored), CIA treatment (Figure 2c, n = 8 mice per group, with both hind paws from each mouse scored), CAIA (Figure 2d, n = 5 mice per group, with both hind paws for each mouse scored), and K/BxN (Figure 2e, n = 5 mice per group, with both hind paws of each mouse scored) studies. In contrast, these histological indices of arthritis were significantly reduced in paws from GW2580- or imatinib-treated mice in all four models of autoimmune arthritis.

Because imatinib has been shown to modulate T-cell function, we investigated whether specific inhibition of c-Fms with GW2580 affects T-cell function. Splenocytes harvested from CIA mice treated with 80 mg/kg GW2580, 80 mg/kg imatinib, or vehicle in the prevention studies were stimulated ex vivo with heat-denatured whole CII, and [³H] thymidine incorporation was used as a surrogate marker of T-cell proliferation. Cells harvested from vehicle- and GW2580-treated CIA mice proliferated extensively in response to CII, whereas cells harvested from imatinib-treated CIA mice exhibited a significantly reduced response (Additional file 2). In addition, splenocytes derived from imatinib-treated CIA mice stimulated with CII demonstrated significantly reduced production of the proinflammatory cytokines TNF and interferon-gamma compared with splenocytes derived from vehicle- or GW2580-treated CIA mice. There were no differences in IL-10 production in these same cell populations stimulated with CII. Thus, imatinib modulates T-cell function in vivo, whereas GW2580 does not.

To determine the effects of imatinib and GW2580 on macrophage infiltration of mouse joints, we assessed levels of total c-Fms and the macrophage-specific marker F4/80 in joint sections derived from CIA mice used in the prevention studies. We found that joints from CIA mice treated with vehicle exhibited marked expression of c-Fms protein, which colocalized with macrophages (Figure 3a). In contrast, in joints from CIA mice treated with 80 mg/kg GW2580 or 80 mg/kg imatinib, both c-Fms protein expression and macrophage infiltration were reduced.

To determine whether c-Fms inhibition affects the formation of macrophages, we isolated bone marrow cells from naïve BALB/c mice, treated them with M-CSF for 5 days to promote the maturation of monocytes, and cultured them for an additional 48 hours with M-CSF to promote their differentiation into macrophages, in the presence or absence of small-molecule inhibitors. Monocytes treated with M-CSF alone displayed a characteristic macrophage phenotype, including extension of multipolar processes and presence of heterogeneous cytoplasmic vacuoles and inclusion bodies (Figure 3b). In contrast, monocytes incubated with media alone and M-CSF-stimulated monocytes treated with 1 μM GW2580 morphologically resembled undifferentiated monocytes. Treatment of monocytes with M-CSF and 1 μM imatinib resulted in a heterogeneous mix of cells, of which some morphologically resembled monocytes and others resembled macrophages.

To confirm the effects of the c-Fms inhibitors on macrophage formation, we applied an assay for the detection of α-naphthyl acetate esterase activity, coupled with fluoride inhibition, to the cell culture system described above. In this assay, the granules of differentiated macrophages stain, whereas monocytes remain unstained [31]. Monocytes were generated as described above, cultured for 48 hours with M-CSF in the presence of 0 to 10 μM GW2580 or imatinib, and then fixed and stained for α-naphthyl acetate esterase activity. Approximately 20% of monocytes cultured in media alone formed macrophages; the addition of M-CSF increased the percentage of macrophages to nearly 90% of the total cell population (Figure 3c). Imatinib reduced macrophage formation in a dose-dependent manner. GW2580 suppressed macrophage formation at lower concentrations than imatinib, in keeping with GW2580 being the more potent c-Fms inhibitor. There were no differences in apoptosis between untreated cells and cells treated with imatinib or GW2580 (data not shown).

We performed experiments to determine whether GW2580 could suppress the formation of osteoclasts. Bone marrow cells were isolated from naïve BALB/c mice, and after 24 hours in culture with M-CSF, the suspension cells were transferred to plates containing dentine or osteologic discs and were co-cultured with M-CSF and RANKL in the presence or absence of GW2580 or imatinib. Treatment with M-CSF and RANKL led to marked formation of large TRAP⁺ multinucleated cells, which are indicative of osteoclasts. For each treatment condition, the dentine disk area that stained positive for TRAP⁺ multinucleated cells was calculated by ImageJ software and expressed as a percentage of the area stained positive for TRAP⁺ following treatment with M-CSF and RANKL alone. Both GW2580 and imatinib significantly reduced the formation of multinucleated TRAP⁺ osteoclasts, albeit with different potencies (inhibition achieved with 0.2 μM GW2580 was comparable to that achieved with 1 μM imatinib) (Figure 4a, b). Because osteoclasts erode bone in RA joints, we next examined the formation of pits in osteologic disks, an indication of the ability of osteoclasts to resorb bone. We found that treatment of cells with imatinib (0.2 or 1 μM) resulted in a dose-dependent reduction in pit formation and that treatment with 0.2 μM GW2580 blocked pit formation (Figure 4c).

Whether c-Fms activation promotes anti-inflammatory or proinflammatory activity in fully differentiated macrophages is controversial [32, 33]. To explore the role of c-Fms in macrophage production of the proinflammatory cytokine TNF, we performed priming studies with fully differentiated macrophages derived from bone marrow cells of BALB/c mice. Macrophages primed with M-CSF for 3 hours in the presence or absence of 5 μM GW2580 or imatinib were stimulated with 1 ng/mL LPS for 24 hours, and TNF protein in the cell culture supernatants was then measured by ELISA. Cells cultured in media alone or M-CSF alone produced very little TNF (Figure 5a). Low-dose LPS stimulated TNF production, which was significantly increased in macrophages primed with M-CSF for 3 hours before the addition of LPS. GW2580 or imatinib at 5 μM (a clinically relevant concentration) reduced TNF production to levels detected in cells treated with LPS alone.

Although LPS is routinely used in experimental settings for the stimulation of proinflammatory cytokine production by macrophages, it is not a true trigger for the production of TNF in RA. Important physiological triggers of inflammation in RA are immune complexes that bind and activate Fc receptors [34]. To examine the role of c-Fms in TNF production elicited by immune complexes, we used the FcR-specific antibody 2.4G2 as a surrogate for immune complex-mediated Fc activation. Stimulation of macrophages with 2.4G2 alone for 24 hours elicited a small increase in TNF production (Figure 5b). Macrophages that were first primed with M-CSF for 3 hours and then stimulated with 2.4G2 produced substantial amounts of TNF. Co-culturing of cells with 5 μM GW2580 or imatinib blocked the M-CSF- and 2.4G2-induced increase in TNF production.

To demonstrate involvement of the M-CSF/c-Fms axis in human RA, we used Luminex bead arrays to determine M-CSF protein levels in synovial fluid derived from patients with RA, OA, or PsA. M-CSF levels were significantly higher in human RA synovial fluid than in OA or PsA synovial fluid (Figure 6a). We also assessed levels of total c-Fms and phosphorylated c-Fms in human RA synovium and found that both expression and activation of c-Fms are high (Figure 6b).