To address the controversy of whether TNFα can compensate for RANKL in osteoclastogenesis in vivo, we used a TNFα-induced animal model of inflammatory arthritis and blocked RANKL/RANK signaling. TNFα increased osteoclast precursors available for RANK-dependent osteoclastogenesis. RANK signaling is not required for the TNFα-stimulated increase in CD11bhi osteoclast precursors but is essential for mature osteoclast formation.
Introduction: Although critical roles of TNFα in inflammatory arthritis and RANKL in bone resorption have been firmly established, a central controversy remains about the extent to which TNFα can compensate for RANKL during osteoclastogenesis and the stage at which RANK signaling is required for osteoclastogenesis. Here, we used the human TNFα transgenic mouse model (TNF-Tg) of erosive arthritis to determine if there are both RANK-dependent and -independent stages of osteoclastogenesis in TNFα-induced erosive arthritis.
Materials and Methods: Osteoclastogenesis and osteoclast precursor (OCP) frequency were analyzed using histology, fluorescence-activated cell sorting (FACS), and cell culture from (1) TNF-Tg mice treated with the RANKL antagonist, RANK:Fc, or (2) TNF-Tg × RANK−/− mice generated by crossing TNF-Tg mice with RANK−/− mice.
Results: Treatment of TNF-Tg mice, which have increased OCPs in their spleens, with RANK:Fc dramatically reduced osteoclast numbers on the surface of their arthritic joints and within their bones, but did not decrease CD11bhi OCP numbers in their spleens. Long-term RANK:Fc administration alleviated joint erosion. Furthermore, TNF-Tg × RANK−/− mice had severe osteopetrosis, no osteoclasts, and no joint erosion, but increased CD11bhi precursor numbers that failed to form mature osteoclasts in vitro.
Conclusion: RANK signaling is essential for mature osteoclast formation in TNFα-mediated inflammatory arthritis but not for the TNFα-induced increase in CD11bhi OCP that subsequently can differentiate into osteoclasts in inflamed joints.
OSTEOCLASTS ARE THE principal bone-resorbing cells and are formed by fusion of mononuclear precursors, which are in the monocyte/macrophage lineage.(1) Considerable progress has been made in our understanding of osteoclastogenesis by applying cell culture techniques(2) and generating transgenic and knockout mice.(3) From these studies, two distinct signals required for osteoclastogenesis have been identified. The first is delivered by macrophage-colony-stimulating factor (M-CSF), which signals through its receptor c-Fms. The second is mediated by RANKL through RANK. Cell culture conditions have been established in which M-CSF and RANKL are the only requirements for osteoclastogenesis,(4, 5) and mice genetically deficient in M-CSF or c-Fms(6, 7) or RANKL or RANK(8, 9) are incapable of osteoclastogenesis and suffer from osteopetrosis.
TNFα, a potent osteoclastogenic cytokine, promotes severe focal bone loss at sites of chronic inflammation. A relationship between TNFα and osteoclastic resorption has been firmly established in diseases that are associated with erosive bone loss, such as rheumatoid arthritis.(10, 11) Previous studies have demonstrated that TNFα stimulates RANKL production by stromal cells,(12) lymphocytes,(13, 14) and endothelial cells.(15) TNFα also stimulates M-CSF production by murine or human stromal cells.(16) Thus, it is well established that TNFα can induce osteoclast formation through this indirect mechanism. Furthermore, TNFα may directly promote osteoclastogenesis by affecting cells in the osteoclast lineage(17) and promote the bone resorptive activity of mature osteoclasts derived from spleen cells in a stromal cell/osteoblast-free environment.(18) Two groups have proposed that TNFα can compensate for RANKL during osteoclastogenesis in vitro under conditions of RANKL blockade.(19, 20) However, other investigators have argued that “permissive” levels of RANKL are required for TNFα-induced osteoclastogenesis in vitro.(21) Importantly, no study has demonstrated that osteoclastogenesis can occur in the absence of RANK signaling in vivo under physiological conditions. Adding fuel to this controversy are the various degrees of osteoclast inhibition observed from RANK blockade in the adjuvant-induced,(22) serum-induced,(23) collagen-induced,(24) and TNFα-induced(25) models of erosive arthritis.
Other areas under active investigation are definition of the various stages of osteoclastogenesis and assessment of their functional significance in erosive joint disease. Recently, we demonstrated that patients with psoriatic arthritis (PsA) have a remarkable increase in the number of circulating osteoclast precursors (OCP) in their peripheral blood mononuclear cell (PBMC) population, which correlated with erosive disease.(26) Furthermore, this increase in OCP was reversible with anti-TNF therapy and correlated with amelioration of clinical signs and symptoms. In a preclinical study, we have shown that transgenic and exogenous TNFα also markedly increased the OCP frequency in spleens and that all of these cells are contained within the CD11bhi population.(27)
In this study, we used the human TNFα transgenic mouse model (TNF-Tg mice) of erosive arthritis(28) to determine if there are both RANK-dependent and −independent stages of osteoclastogenesis in TNFα-induced erosive arthritis. Two in vivo models of RANKL blockade were used: (1) TNF-Tg mice treated with a RANKL antagonist, RANK:Fc, and (2) TNF-Tg mice in a RANK null background (TNF-Tg × RANK−/−). In these studies, we found that RANKL/RANK signaling is not required for the TNFα-stimulated increase in CD11bhi OCP. However, elevated levels of TNFα could not compensate for the absence of RANK signaling for mature osteoclast formation in arthritic joints and within bones. Thus, in chronic inflammatory bone loss, TNFα may first affect osteoclastogenesis by increasing CD11bhi osteoclast precursors in the periphery through a RANKL/RANK-independent mechanism. These osteoclast precursors then respond to RANKL and become mature osteoclasts at sites of bone resorption.
MATERIALS AND METHODS
Murine RANK:Fc was provided by Dr W Dougall (Amgen Inc., Seattle, WA, USA). Recombinant human M-CSF was purchased from R&D Systems Inc. (Minneapolis, MN, USA). Anti-murine CD11b (M1/70) and related isotype controls were purchased from eBioscience Inc. (San Diego, CA, USA). Anti-murine CD16/32 (FcγIII/II) was obtained from Pharmingen (San Diego, CA, USA).
TNF-Tg mice (3647 TNF-Tg line) in a CBA × C57Bl/6 background were obtained from Dr G Kollias, and RANK−/− mice in a C57Bl/6 background were obtained from Dr W Dougall (Amgen Inc.).(9) TNF-Tg × RANK−/− mice were generated by inter-crossing TNF-Tg and RANK+/− mice to generate the TNF-Tg × RANK+/− F1 generation; these F1 mice were crossed with RANK+/−, and the progeny were genotyped by tail polymerase chain reaction (PCR). To accommodate the absence of teeth, adult TNF-Tg × RANK−/− and RANK−/− mice were fed with powdered mouse chow. The Institutional Animal Care and Use Committee approved all studies.
Preparation and histomorphometry of bone sections
The limbs were removed from mice after death, fixed in 10% buffered formalin, decalcified in 10% EDTA, and embedded in paraffin. Sections (5 μm thick) were then stained for TRACP activity and counter-stained with H&E. Histomorphometric analysis was performed in sections of TNF-Tg mice treated with RANK:Fc or PBS, as described previously,(29) using an Osteomeasure image analysis software (Osteometrics, Atlanta, GA, USA). Osteoclast numbers were expressed per tibia in longitudinal sections and per millimeter of eroded knee joint surface.
ELISA for human TNFα in mouse serum
Blood was drawn from TNF-Tg mice by cardio-puncture, and the serum was collected by centrifugation. The levels of human TNFα were detected according to the manufacturer's instruction (R & D Systems, Minneapolis, MN, USA). The whole procedure was carried at room temperature. Briefly, 96-well plates were coated with 4 μg/ml of capture antibody (MAB610) overnight and blocked with PBS containing 1% bovine serum albumin (BSA), 5% sucrose, and 0.05% NaN3 for 2 h. Serum samples and standards were added and incubated for 2 h. The plates were incubated with 200 ng/ml of biotinylated detection antibody (BAF210) for 1 h and streptavidin HRP (DY 998) for 20 minutes. The color reaction was developed by adding substrate solutions to the plates, and the optical density (OD) was read at 450 nm.
In vivo blockade with RANK:Fc
For short-term blockade of RANKL signaling, TNF-Tg mice (five per group) at 4 months of age (established arthritis) were given intraperitoneal injections of either RANK:Fc (10 mg/kg/day) or PBS for 2 weeks. For long-term blockade, TNF-Tg mice (five per group) at 3 months of age (onset of erosive arthritis) were given RANK:Fc (1 mg/kg) by intraperitoneal injections twice a week for 8 weeks. The mice were killed 1 day after the last injection. Spleen cells were harvested for fluorescence-activated cell sorting (FACS), and legs were processed for histological analysis.
In vitro osteoclastogenesis assay
Splenocytes were incubated with ammonium chloride solution for 10 minutes to lyse red blood cells and then cultured (1.75 × 105 cells/well in 96-well plates) in α-modified essential medium (α-MEM; GIBCO BRL, Grand Island, NY, USA) supplemented with 10% fetal calf serum (FCS; Hyclone Laboratories, Logan, UT, USA) in the presence of RANKL (100 ng/ml) and M-CSF (10 ng/ml). Cultures were maintained for 5 days at 37°C in an atmosphere of 5% CO2/air, and media were changed every 2 days by replacing one-half of the spent media with fresh media supplemented with RANKL/M-CSF. Cells were fixed and stained for TRACP activity using the Diagnostics Acid Phosphatase Kit (Sigma, St Louis, MO, USA) to identify osteoclasts, as described previously.(30)
Surface protein staining was performed on freshly isolated mouse splenocytes. After red blood cell lysis, a single cell suspension was incubated with anti-murine CD16/32 to block Fc receptor-mediated antibody binding. Cells were then labeled with fluorescent probes, as described previously.(31) Data were acquired using a FACScalibur instrument (Beckton Dickenson, Bedford, MA, USA) and analyzed by Cellquest software (version 3.1).
All results are given as means ± SE. Comparisons were made by ANOVA and Student's t-test for unpaired data. p values <0.05 were considered statistically significant.
To elucidate the RANKL dependence of TNFα-enhanced osteoclast formation in vivo, we used TNF-Tg mice that overexpress a human TNFα transgene. Human TNFα is detectable in the serum of these mice around 2 months of age (119.6 ± 47.8 pg/ml) when the first macroscopic sign of inflammatory arthritis, swollen ankle joints, can be observed. In contrast, murine TNFα concentrations are undetectable in these mice.(32) TNF-Tg mice with established arthritis (4 months old) were treated with PBS or RANK:Fc (10 mg/kg, daily, IP, for 2 weeks) and were killed 1 day after the last RANK:Fc injection. The knees of these mice were processed for histology. RANK:Fc treatment effectively reduced osteoclast number in both wildtype (PBS 112.7 ± 10.7 per tibia versus RANK:Fc 2.3 ± 0.3 per tibia, p < 0.01) and TNF-Tg mice (Fig. 1). Most striking was the depletion of osteoclasts at the leading edge of the subarticular erosions and below the growth plate (Figs. 1Ae and 1Af).
To determine if RANK:Fc could block the TNFα-mediated increase in OCPs, we collected splenocytes from the mice described in Fig. 1 and assessed the frequency of CD11bhi OCP and their osteoclastogenic potential. Splenocytes were divided into two aliquots. One was stained with antibodies for CD11b and analyzed by FACS to determine the frequency of CD11bhi OCP cells; the other was cultured in the presence of M-CSF and RANKL for 5 days to generate osteoclasts and assessed by TRACP assay. RANK:Fc treatment did not decrease the percentage of CD11bhi splenocytes (Fig. 2A) or the osteoclastogenic potential of these OCP (Fig. 2B).
We next examined if longer-term blockade of RANKL signaling with RANK:Fc could inhibit the bone erosion in inflammatory arthritis by reducing osteoclast numbers. TNF-Tg mice were treated with 1 mg/kg of RANK:Fc twice a week for 8 weeks at the time when erosive arthritis begins to develop (3 months old). At this dose, RANK:Fc also significantly reduced osteoclast numbers at the site of inflammation (Figs. 3A and 3B), but did not reduce the increased frequency of CD11bhi OCPs in the periphery (Fig. 3C). Serum human TNFα concentrations remained elevated in these RANK:Fc-treated TNF-Tg mice: 47 ± 19.6 pg/ml versus 85 ± 42 pg/ml (p > 0.05) in untreated TNF-Tg mice and undetectable values in wildtype mice. This low dose of RANK:Fc efficiently ameliorated bone erosion in TNF-Tg mice compared with placebo-treated control mice (Figs. 3A and 3B).
To further define the role of RANK signaling in TNFα-induced osteoclastogenesis in vivo, we crossed the TNF-Tg mice with RANK knockout mice (RANK−/−). These mice (TNF-Tg × RANK−/−) were smaller in size than their wildtype and TNF-Tg littermates, had no tooth eruption, and were severely osteopetrotic by radiographic analysis (Fig. 4A). Gross analysis failed to identify any differences in phenotype between the TNF-Tg × RANK−/− mice and their RANK−/− littermates. Histological analysis of 4-month-old TNF-Tg × RANK−/− mice showed a complete absence of osteoclasts in their bones (Fig. 4B). Consistent with the RANK:Fc treatment results in Figs. 1 and 2, spleens from TNF-Tg × RANK−/− mice contained a larger CD11bhi population than either of their wildtype or RANK−/− littermates (Fig. 5A). Using our standard culture conditions, we did not detect any osteoclastogenic potential in TNF-Tg × RANK−/− splenocytes (Fig. 5B). The TNF-Tg × RANK−/− mice had an increase in CD11bhi precursors that were unable to differentiate into mature osteoclasts in vivo.
Several studies have revealed a novel RANK-independent mechanism by which TNFα mediates osteoclastogenesis using in vitro cell culture models.(18–20) Administration of high doses of exogenous TNFα also leads to the formation of occasional osteoclast-like cells in RANK−/− mice at the site of calvarial injection, suggesting that TNFα may substitute for RANKL and induce osteoclastogenesis in vivo.(33) However, RANK-independent osteoclastogenesis has not been documented in a pathological setting in vivo. Previously, we investigated RANK-independent osteoclastogenesis in animal models of wear debris-induced osteolysis(34) and fracture healing,(35) where TNFα is present in large quantities. However, we did not observe any RANK-independent osteoclastogenesis. Here, we again show that, in a model of chronic TNFα overexpression and aggressive bone erosion, osteoclastogenesis does not occur in the absence of RANKL/RANK signaling.
Osteoclasts are derived from multipotent stem cells in the bone marrow. The stem cells that are committed to the osteoclast lineage undergo proliferation and differentiation to become TRACP+ osteoclast precursors. Arai et al.(36) have used a comprehensive panel of cell surface markers, cell sorting, CFU-M colony assay, and bone marrow cultures to separate TRACP− osteoclast precursors into two groups: early stage cells characterized as “c-Fms+/CD11blo/RANK−,” and late stage precursors characterized as “c-Fms+/CD11bhi/RANK+.” Using a similar strategy, we recently characterized osteoclast precursors in mouse spleen as “c-Fms+/−/CD11bhi/RANK+/−.”(27) Our findings demonstrated that CD11bhi alone can be used as a marker to identify TRACP− osteoclast precursors in the periphery. Using CD11bhi as a marker for peripheral OCPs, we found that TNFα expressed in transgenic mice increases peripheral CD11bhi OCP frequency in vivo and that etanercept (a TNF blocker) treatment reduced the number of CD11bhi splenocytes and their osteoclastogenic potential to wildtype levels. More importantly, we demonstrated that patients with psoriatic arthritis have a marked increase in the number of CD11bhi OCPs in their peripheral blood compared with normal controls and patients with osteoarthritis.(26) This increase also seems to be reversible with anti-TNF therapy, indicating that the increase in CD11bhi OCP frequency is mediated by TNFα.(27)
To examine the requirement for RANKL/RANK in the TNFα-mediated increase in CD11bhi OCPs as well as in mature osteoclast formation at the site of inflammation in diseased joints, we used two animal models where TNFα levels are high, but RANK signaling has been blocked. In the first, we injected the RANKL antagonist, RANK:Fc, into TNF-Tg mice that have elevated levels of TNFα in their circulation.(27) RANK:Fc is a soluble fusion protein consisting of the extracellular domain of RANK fused to the Fc domain of IgG1.(37) In the second animal model, we generated TNF-Tg/RANK−/− mice. In both models, we found that RANK blockade abolishes mature osteoclast formation but has no effect on the increased CD11bhi OCP frequency, indicating that the TNFα-mediated increase in CD11bhi OCPs in vivo is RANK independent. However, we found that RANK signaling is indispensable for mature osteoclast formation.
One caveat about these conclusions is that inhibition of RANK signaling could affect TNF levels in vivo. It is not known if RANKL can stimulate TNFα production. However, given the restricted expression pattern of RANK protein, which is only expressed on the surface of osteoclast precursors and dendritic cells, and the fact that synovial cells are a major source of TNFα in these TNF-Tg mice,(32) it is unlikely that blockade of RANKL/RANK signaling in our animal models influenced TNFα levels. Indeed, our data show that the serum TNFα levels in RANK:Fc-treated TNF-Tg mice remained elevated with an associated increase in OCP frequency. Although we did not measure TNFα levels in the TNF-Tg × RANK−/− mice, these animals had increased numbers of CD11bhi/Gr-1− cells, the values being similar to those in the TNF-Tg mice and significantly higher than RANK−/− mice. Furthermore, injection of murine TNFα increased the CD11bhi OCP frequency in the peripheral blood of RANK−/− mice (Li P, Schwarz EM, Boyce BF, and Xing L, unpublished data, 2003).
The activation of NF-κB and AP-1 signaling pathways by TNFα have been well documented in bone marrow-derived macrophages, which become enriched for CD11b+ cells after M-CSF treatment.(21, 38) These pathways regulate proliferation, differentiation, and survival in many cell types. However, we have found that TNFα does not affect proliferation or apoptosis of CD11bhi OCPs in vivo or differentiation of the precursors in vitro. Because osteoclast precursors originate from hematopoietic stem cells in the bone marrow, it is possible that TNFα may affect the release of precursors into the periphery, leading to an increase in circulating CD11bhi OCPs. Indeed, administration of TNFα into wildtype mice rapidly increased the number of CD11bhi cells in the periphery, suggesting that expedited mobilization of this late stage OCP from bone marrow may be the underlining mechanism of this TNFα-mediated process.
Together with our knowledge of the pathology associated with focal bone loss in RA(39, 40) and the presence of abundant amounts of RANKL at these sites,(41, 42) our data suggest that RANKL/RANK signaling plays distinct roles in two phases of systemic TNFα-mediated osteoclastogenesis. In the first phase, RANK signaling is dispensable for a TNFα-mediated increase in the number of CD11bhi/c-Fms+/−/RANK+/− OCPs(27) in the periphery. In the second phase, an obligatory RANKL/RANK signal is required for mature osteoclast formation at the sites of bone erosion where TNFα and/or other mediators of osteoclastogenesis are increased. In this phase, RANKL/RANK signaling is the “check point” for osteoclastogenesis in vivo. Thus, a rational clinical approach to prevent erosive inflammatory arthritis would be a combination of anti-TNF therapy and OPG or RANK:Fc in RA patients who do not respond to or develop resistance to anti-TNF therapy.
The RANK:Fc used in this study were provided by Amgen Inc. The authors thank W Dougall for critical advice and J Harvey for technical assistance with the histology. This work was supported by research grants from the National Institutes of Health (PHS AR45791, AR43510, AR44220, and AR48697).