Tumor Necrosis Factor-α Increases Circulating Osteoclast Precursor Numbers by Promoting Their Proliferation and Differentiation in the Bone Marrow through Up-regulation of c-Fms Expression*

Osteoclasts are essential cells for bone erosion in inflammatory arthritis and are derived from cells in the myeloid lineage. Recently, we reported that tumor necrosis factor-α (TNFα) increases the blood osteoclast precursor (OCP) numbers in arthritic patients and animals, which are reduced by anti-TNF therapy, implying that circulating OCPs may have an important role in the pathogenesis of erosive arthritis. The aim of this study is to investigate the mechanism by which TNFα induces this increase in OCP frequency. We found that TNFα stimulated cell division and conversion of CD11b+/Gr-1-/lo/c-Fms- to CD11b+/Gr-1-/lo/c-Fms+ cells, which was not blocked by neutralizing macrophage colony-stimulating factor (M-CSF) antibody. Ex vivo analysis of monocytes demonstrated the following: (i) blood CD11b+/Gr-1-/lo but not CD11b-/Gr-1- cells give rise to osteoclasts when they were cultured with receptor activator NF-κB ligand and M-CSF; and (ii) TNF-transgenic mice have a significant increase in blood CD11b+/Gr-1-/lo cells and bone marrow proliferating CD11b+/Gr-1-/lo cells. Administration of TNFα to wild type mice induced bone marrow CD11b+/Gr-1-/lo cell proliferation, which was associated with an increase in CD11b+/Gr-1-/lo OCPs in the circulation. Thus, TNFα directly stimulates bone marrow OCP genesis by enhancing c-Fms expression. This results in progenitor cell proliferation and differentiation in response to M-CSF, leading to an enlargement of the marrow OCP pool. Increased marrow OCPs subsequently egress to the circulation, forming a basis for elevated OCP frequency. Therefore, the first step of TNF-induced osteoclastogenesis is at the level of OCP genesis in the bone marrow, which represents another layer of regulation to control erosive disease.

Mature osteoclasts are essential effector cells for normal bone remodeling and pathologic bone loss seen in many forms of erosive diseases, such as rheumatoid arthritis (RA). 3 The importance of osteoclastic resorption in this process has been proven in studies with various knock-out mice deficient in genes essential for osteoclastogenesis, which develop osteopetrosis because of the accumulation of un-resorbed bone matrix within the bone marrow cavity (1,2). Similarly, these mice are completely resistant to bone destruction of affected joints when they are induced to develop erosive arthritis (3)(4)(5). Osteoclasts are derived from common osteoclast/monocyte precursors that are generated in bone marrow and travel to peripheral tissues through the bloodstream (6). In patients or animals with RA, these precursor cells constantly migrate to inflamed joints perhaps from the following two directions: "outside in," from blood to the pannus-bone interface, and "inside out," from epiphysial bone marrow to the subchondral bone. They then differentiate to mature osteoclasts in response to high levels of osteoclastogenic cytokines, including receptor activator NF-B ligand (RANKL), macrophage-colony-stimulating factor (M-CSF), and tumor necrosis factor-␣ (TNF␣), produced by inflammatory cells in the synovium (7,8). Although the teams of mature osteoclasts mediate focal erosions via resorption of the periarticular and subchondral bone over long periods of time (months to years), the life span of individual osteoclasts is only a few weeks. Thus, mature osteoclasts must be constantly replaced by a perpetual supply of osteoclast precursors (OCPs). Currently, the molecular mechanism by which joint inflammation sustains the perpetual supply of OCPs is poorly understood.
OCPs are derived from c-Kit ϩ multipotent hematopoietic stem cells in the bone marrow through a series of differentiation processes (9). The first step of OCP ontogeny is the commitment of hematopoietic stem cells to the myeloid lineage under the control of the Ets transcription factor PU.1 (10). Then the myeloid progenitors survive, proliferate, and differentiate into various "downstream" lineages, including OCPs. These events have been characterized by changes in the surface expression of distinct markers. For instance, early myeloid progenitors do not express detectable levels of c-Fms, the receptor for M-CSF, an essential survival factor for OCPs. As such, they exhibit lower proliferation and differentiation potency. However, under the influence of hematopoietic factors such as stem cell factor and perhaps M-CSF itself, c-Fms Ϫ cells differentiate to c-Fms ϩ cells (11). The conversion of c-Fms Ϫ cells to c-Fms ϩ cells is an important landmark for the progression of myelopoiesis because the M-CSF signal at this stage is critical for cell survival (anti-apoptosis), proliferation, and differentiation. However, the regulation of this conversion in chronic inflammatory bone diseases has not been well studied.
TNF␣ is one of the most potent pro-inflammatory cytokines, and its role in RA has been formally established by the development of anti-TNF therapy. Additional proof comes from studies of TNF transgenic (TNF-Tg) mice that develop erosive arthritis featured with intense synovial inflammation and destruction of cartilage and bone (12). These mice have increased numbers of OCPs in spleen and blood and increased numbers of mature osteoclasts in the affected joints (13). The role of TNF␣ in osteoclastogenesis has been studied extensively in the last decade. Administration of TNF␣ into wild-type (WT) mice greatly increases the number of tartrateresistant acid phosphatase (TRAP)-positive osteoclasts locally and systemically (13,14). In vitro, TNF␣ directly stimulates mature osteoclast formation from OCPs in the absence of osteoblasts/stromal cells through activation of NF-B and nuclear factor of activated T cells pathways (15). It also promotes the production of M-CSF by T lymphocytes and RANKL by osteoblasts, thereby indirectly stimulating osteoclast formation (16,17). However, whether TNF␣ promotes OCP generation from early myeloid precursors and the mechanisms involved in this process have not been investigated.
In this study, we test the hypothesis that TNF␣ increases OCP numbers through regulation of c-Fms expression. We demonstrate that TNF␣ increases the proliferation of OCPs in vivo and in vitro and promotes the differentiation of c-Fms Ϫ cells to c-Fms ϩ cells. TNF␣-induced OCP genesis is directly related to increased blood OCP frequency. Thus, our results reveal a new mechanism for TNF␣ in the control of peripheral osteoclast numbers and provide an additional regulatory step to control osteoclast formation and bone resorption in inflammatory erosive diseases.
Animals-TNF-Tg mice in a CBA ϫ C57Bl/6 background (3647 TNF-Tg line) were obtained from Dr. G. Kollias (12). Four-month-old TNF-Tg mice were used because this is the age when they typically develop severe joint disease with elevated serum human TNF␣ concentrations (13). Acute TNF␣ in vivo treatment was performed as we described previously (13). In brief, 2-month-old C57/B6 male mice were randomly divided into TNF␣ and PBS groups. Murine TNF␣ (0.5 g in 25 l of PBS) or the same volume of PBS was injected four times/day for 3 days into the subcutaneous layer overlying the calvariae of WT mice. Bone marrow and PBMCs were collected for cell cycle analysis, as described below. The Institutional Animal Care and Use Committee approved all animal studies.
Generation of Osteoclasts-Cells from several sources were used to generate osteoclasts: 1) freshly isolated bone marrow cells; 2) M-CSF or TNF␣ pretreated nonadherent bone marrow cells; and 3) flow-sorted bone marrow or blood cells. Cells were cultured in ␣-modified essential medium (Invitrogen) with 10% fetal calf serum (Hyclone Laboratories, Logan, UT), RANKL (5 ng/ml), and M-CSF (10 ng/ml) for 3-5 days when multinucleated cells typically were observed under an inverted microscope. Cells were then fixed and stained for TRAP activity, and TRAP ϩ cells containing Ն3 nuclei were counted as mature osteoclasts as we described previously (13).
Fluorescence-activated Cell Sorting (FACS) Analysis and Cell Sorting-For FACS analysis, freshly isolated bone marrow, PBMCs or cultured OCPs were incubated with anti-murine CD16/32 to block Fc receptormediated antibody binding. Cells were then stained with various fluorescent-labeled antibodies. Data were acquired using a FACSCalibur flow cytometer (BD Biosciences) and analyzed using the Cellquest soft-ware (version 3.1), as described previously (13). For cell sorting, pooled bone marrow cells or peripheral blood cells from either TNF-Tg or WT mice were double-stained with FITC-anti-CD11b and PE-anti-Gr-1 antibodies, and then sorted by a FACSVantage SE Turbo sorter. CD11b ϩ /Gr-1 Ϫ/lo , CD11b Ϫ /Gr-1 Ϫ , and CD11b ϩ /Gr-1 hi cells were collected separately, reanalyzed to ensure their purity (Ն98%), and used for osteoclastogenesis assays, as described above.
Cell Cycle Analysis of Blood and Bone Marrow OCPs-PBMCs or bone marrow cells were stained with FITC anti-CD11b and PE anti-Gr-1 antibodies first, and the labeled cells were then incubated with Hoechst 33342 (5 g/ml) for 45 min at 37°C in a reaction buffer (19) prior to FACS analysis. Cell cycle distributions of CD11b ϩ /Gr-1 Ϫ/lo , CD11b ϩ /Gr-1 hi , and CD11b Ϫ /Gr-1 Ϫ cells were analyzed using a FACS-Vantage flow cytometer (BD Biosciences) equipped with UV light and 488-nm laser excitations. The percentage of the cells in G 1 , S, and G 2 /M phases of the cell cycle was determined using the ModFit program (Verity Software, Topsham, ME).
In Vitro Carboxyfluorescein Diacetate Succinimidyl Ester (CFSE) Labeling-Fresh bone marrow cells were isolated from femur and tibia of WT mice. Red blood cells were removed using ACK lysis buffer (0.15 M NH 4 Cl, 1.0 mM KHCO 3 , 0.1 mM Na 2 EDTA). The cells were then resuspended in PBS at a density of 5 ϫ 10 7 /ml, and an equal volume of CFSE (10 M) was added by drop while vortexing. The reaction was incubated at 37°C for 10 min and mixed once during the incubation. The labeled cells were washed with ␣-minimum Eagle's medium plus 10% fetal bovine serum three times and seeded in a 6-well plate at a density of 4 -5 ϫ 10 6 /well.
Quantitative Real Time RT-PCR-RNA from TNF␣-treated and nontreated bone marrow cells or sorted CD11b ϩ /Gr-1 Ϫ/lo cells was extracted using the RNeasy kit and the QiaShredder from Qiagen (Valencia, CA). cDNA was synthesized with the use of 20 l of RT reaction solution containing 1 g of total RNA, 10 mM Tris-HCl buffer (pH 8.3), 50 mM KCl, 5 mM MgCl 2 , 1 mM deoxynucleoside triphosphates, 2.5 M random hexamers, 20 units of RNase inhibitor, and 50 units of Moloney murine leukemia virus reverse transcriptase (all from Roche Applied Science). Quantitative PCR amplification was performed with gene-specific primers using a Rotor-Gene 2000 real time amplification operator (Corbett Research, Mortlake, Australia). The primer sequences included the following: 1) c-Fms primers, 5Ј-GTCA-GAGGCCCCGTTTGTT-3Ј and 5Ј-AGTAAATATAGAGGCTAG-CACTGTGAGAAC-3Ј; and 2) actin primers, 5Ј-AGATGTGGATCAG-CAAGCAG-3Ј and 5Ј-GCGCAAGTTAGGTTTTGTCA-3Ј. The relative standard curve method was used to calculate the amplification difference for each primer set (20). The standard curve was made from four points corresponding to 10-fold cDNA dilution series for each gene. For each sample, the relative amount was calculated from their respective relative standard curves. The relative c-Fms expression value was then obtained by dividing each value by the actin value. Standards were run in triplicate, and samples were run three times in triplicate.
Statistics-All results are given as means Ϯ S.E. Comparisons were made by analysis of variance and Student's t test for unpaired data. p values less than 0.05 were considered statistically significant.

CD11b and Gr-1 Are Cell Surface Markers for Blood and Bone
Marrow OCPs-Previously, we have used CD11b as a single cell surface marker for OCPs in the spleen and demonstrated that TNF-Tg arthritic mice have increased CD11b ϩ OCP frequency (13). However, CD11b protein is broadly expressed on the surface of various myeloid lineage cells, especially on granulocytes that consist of more than 70% of the total blood (72.4 Ϯ 3.4%) or bone marrow (77.4 Ϯ 2%) CD11b ϩ cells in adult C57/B6 mice. Because of this, OCP numbers may be overestimated if the frequency of CD11b ϩ cells is used to assess the number of OCPs in these compartments. In this study, we first characterized the osteoclastogenic potential of blood and bone marrow cells using CD11b and Gr-1 to eliminate CD11b ϩ /Gr-1 hi granulocytes (Fig. 1). Peripheral blood mononuclear cells (PBMC) and bone marrow cells from WT adult mice were stained with anti-CD11b and Gr-1 antibodies, and antibodies and subjected to FACS analysis. The living cells were gated using forward and side scatter (R1). The CD11b ϩ /Gr-1 Ϫ/lo (R2), CD11b ϩ /Gr-1 hi (R3), and CD11b Ϫ /Gr-1 Ϫ (R4) populations were sorted and cultured with M-CSF and RANKL to generate osteoclasts. Photography (ϫ4) of the TRAP ϩ osteoclasts formed from purified PBMCs and from bone marrow are shown in the upper and lower panels, respectively. C, the number of TRAP ϩ osteoclasts formed from above cultures was quantified and presented as means plus S.E. of three wells. Experiments were repeated once with similar results. three populations were fractionated by FACS on the basis of their surface expression: CD11b ϩ /Gr-1 Ϫ/lo (R2), CD11b ϩ /Gr-1 hi (R3), and CD11b Ϫ /Gr-1 Ϫ (R4). The sorted cells were cultured with M-CSF and RANKL to generate osteoclasts that were assessed by TRAP staining. In blood, only CD11b ϩ /Gr-1 Ϫ/lo cells (R2) formed osteoclasts. Although both CD11b ϩ /Gr-1 Ϫ/lo (R2) and CD11b Ϫ /Gr-1 Ϫ (R4) marrow cells gave rise to osteoclasts (Fig. 1), the number of TRAP ϩ osteoclasts was much higher in the CD11b ϩ /Gr-1 Ϫ/lo (R2) fraction than that in the CD11b Ϫ /Gr-1 Ϫ (R4) fraction (Fig. 1, B and C). The CD11b ϩ /Gr-1 hi cells did not form TRAP ϩ osteoclasts and died off under our culture conditions. Because it is likely that marrow CD11b Ϫ cells differentiate to TRAP ϩ osteoclasts through the CD11b ϩ stage and only blood CD11b ϩ / Gr-1 Ϫ/lo cells give rise to osteoclasts, we reasoned that CD11b ϩ /Gr-1 Ϫ/lo cells include the majority of OCPs and can be used as OCP surface markers.
TNF␣ Promotes the Proliferation of Bone Marrow OCPs-To examine whether increased bone marrow CD11b ϩ /Gr-1 Ϫ/lo OCPs in TNF-Tg mice results from an alteration in proliferation, cell cycle analysis was performed using a combination of anti-CD11b and Gr-1 antibodies and Hoechst 33342 DNA dye staining. This approach allows us to assess cell cycle status in different fractions of CD11b-and Gr-1stained cells simultaneously and avoid cell sorting and in vitro labeling. Compared with CD11b Ϫ /Gr-1 Ϫ and CD11b ϩ /Gr-1 hi cells, the CD11b ϩ / Gr-1 Ϫ/lo population contained cells in the S/G 2 /M phase of the cell cycle in both TNF-Tg and WT mice (Fig. 3A). Furthermore, more TNF-Tg CD11b ϩ /Gr-1 Ϫ/lo cells were in the S/G 2 /M phase than those of WT littermates, and there was no difference in the frequency of apoptosis (Fig. 3B, upper panel). In contrast to marrow cells, PBMCs were not in  the cell cycle, and the majority of them were at the G 0 /G 1 phase (Fig. 3B,  lower panel).
To investigate whether increased cycling of marrow OCPs leads to increased total numbers of OCPs in the bone marrow and subsequently in the peripheral blood, we injected TNF␣ (0.5 g/injection, four times/ day for 3 days) into WT mice and examined the cell cycle status of marrow OCPs, as well as the number of OCPs in blood and bone marrow by FACS analysis. A 3-day regimen was used because our preliminary experiments showed that it took 3 days for TNF␣ to increase blood OCPs (data not shown). TNF␣ significantly increased the percentage of cycling marrow OCPs, which started after the 2nd day and peaked after the 3rd day of injection. Correspondingly, total OCP numbers in bone marrow increased at 3 days. Blood OCPs were slightly elevated after 2 days and increased significantly after 3 days of TNF␣ treatment, in parallel with bone marrow cells (Fig. 3, C and D). Together, these findings suggest that TNF␣ may increase bone marrow OCP numbers by promoting their proliferation.
To determine the direct effects of TNF and M-CSF on OCP proliferation in vitro, we labeled WT bone marrow cells with 5-(and 6-)-carboxyfluorescein diacetate succinimidyl ester (CFSE) to monitor cell division. The labeled cells were then treated with PBS, M-CSF (10 ng/ml), TNF␣ (10 ng/ml), or M-CSF/TNF␣ for 24, 48, and 72 h, stained with anti-CD11b, and subjected to FACS analysis. The percentage of dividing cells in the CD11b ϩ population was measured as an indicator of OCP division. Because our initial results demonstrated that 100% of CD11b ϩ /Gr-1 hi cells die off, leaving only CD11b ϩ Gr-1 Ϫ/lo cells 24 h after culture (data not shown), Gr-1 staining was omitted in these experiments. TNF␣ alone had no effect on cell division at any time point, and M-CSF stimulated OCP division after treatment for 48 h (Fig. 4). In contrast, TNF␣ synergized with M-CSF to stimulate OCP division from 33% in M-CSF-treated cells alone up to 70% in TNF␣ plus M-CSFtreated cells (Fig. 4, A and B).
TNF␣ Induces c-Fms Expression of Bone Marrow OCPs-Increased M-CSF-dependent proliferation after TNF␣ treatment suggests that  TNF␣ may affect the factors essential for M-CSF signaling. Because the M-CSF receptor, c-Fms, is expressed on the surface of late stage of myeloid precursors, including OCPs (21), we hypothesized that TNF-Tg CD11b ϩ / Gr-1 Ϫ/lo OCPs may contain more c-Fms ϩ cells to account for the TNF␣induced M-CSF-dependent proliferation. To test this, we compared the percentage of c-Fms ϩ cells in the gated CD11b ϩ /Gr-1 Ϫ/lo population of TNF-Tg with that of WT mice, and we found that the TNF-Tg mice had significantly more c-Fms ϩ cells in CD11b ϩ /Gr-1 Ϫ/lo OCPs in both blood and bone marrow (Fig. 5A).
To determine whether TNF␣ promotes the differentiation of c-Fms Ϫ cells to c-Fms ϩ cells, we examined the induction of CD11b ϩ /c-Fms ϩ cells in vitro in response to TNF␣. WT marrow cells were treated with PBS, M-CSF (10 ng/ml), TNF␣ (10 ng/ml), or M-CSF/TNF␣, and the percentage of CD11b ϩ /c-Fms ϩ cells was examined after 24, 48, and 72 h in nonadherent and adherent fractions (Fig. 5). TNF␣ increased the percentage of CD11b ϩ /c-Fms ϩ cells at 24 h, and this peaked at 72 h. M-CSF decreased c-Fms expression over the first 12-24 h, which may reflect ligand-induced receptor internalization (22). However, after 48 h, M-CSF significantly stimulated c-Fms expression. TNF␣ alone increased the percentage of CD11b ϩ /c-Fms ϩ cells in nonadherent cells by 9 -10-fold and synergized with M-CSF to stimulate c-Fms expression in those cells (Fig. 5B). TNF␣ also stimulated the percentage of CD11b ϩ /c-Fms ϩ cells in adherent cells, but to a lesser extent (by 3-4-fold, data not shown).
To confirm the functional consequence of increased c-Fms expression by TNF␣ in osteoclastogenesis, nonadherent TNF␣-or PBStreated cells were cultured to assess their osteoclast-forming potency. Cells primed by TNF␣ formed significantly more osteoclasts when they were subsequently cultured with M-CSF and RANKL (Fig. 5B). To examine if TNF␣ stimulates c-Fms gene transcription, mRNA expression levels of c-Fms were determined in TNF␣-treated nonadherent cells or purified CD11b ϩ /Gr-1 Ϫ/lo OCPs. TNF␣ significantly increased the c-Fms expression levels in both cell preparations (Fig. 5C). Finally, we treated cells with TNF␣ in the presence of M-CSF neutralizing antibody to determine whether TNF-induced c-Fms expression is mediated by M-CSF. This treatment completely prevented M-CSF-induced c-Fms expression in a dose-dependent manner but had little effect on TNF␣-induced c-Fms up-regulation (Fig. 5D). These findings indicate that TNF␣ can directly stimulate c-Fms expression by OCPs.

DISCUSSION
TNF␣ is a pleiotropic cytokine that induces focal erosions in inflamed arthritic joints by several mechanisms. Recently, we have demonstrated that one of these mechanisms involves an increase in the frequency of peripheral blood CD11b ϩ OCP in arthritic patients and in TNF-Tg arthritic mice. We also found that this increased OCP frequency is reduced by and associated with a successful outcome of anti-TNF therapy (4,13), implying that circulating OCP numbers may have an important role in the pathogenesis of inflammatory erosive arthritis. However, the mechanism by which systemic TNF␣ induces this increase in OCP frequency remains unknown.
Although it has been well established that osteoclasts are derived from myeloid precursors (23,24), the identity, function, and regulation of OCPs in vivo remain poorly understood, partly because of the absence of unique cell surface markers to identify them. We and other have previously used CD11b ϩ as a single surface marker for spleen OCPs (13,25). Recently, we established that many of these cells are blood granulocytes that express CD11b, and the number of OCP may be significantly overestimated by this method. Apart from CD11b, c-Fms and RANK expression also has been used to estimate OCPs, based on the essential role of M-CSF and RANKL for osteoclastogenesis (11). However, CD11b ϩ /c-Fms Ϫ and CD11b ϩ /c-Fms ϩ cells as well as CD11b ϩ /RANK Ϫ and CD11b ϩ /RANK ϩ cells all can give rise to osteoclasts. Thus, they also are suboptimal OCP markers (13,26). To try to improve on this, here we assessed CD11b ϩ /Gr-1 Ϫ/lo cells, excluding CD11b ϩ /Gr-1 hi granulocytes, which include both early (c-Fms Ϫ or/and RANK Ϫ ) and late OCPs (c-Fms ϩ or/and RANK ϩ ). We found that 60 -70% of CD11b ϩ /Gr-1 Ϫ/lo cells can differentiate to TRAP ϩ mononuclear osteoclasts when cultured with M-CSF and RANKL, indicating that most of these cells have osteoclast-forming potential. However, it is important to realize that CD11b/Gr-1 Ϫ/lo cells are not composed of only OCPs because they can differentiate into macrophages and dendritic cells under the appropriate conditions. At present, it is not known if there are committed mono-potential OCPs that can be separated from other precursor cells. Recently, Geissmann et al. (27) proposed that blood monocytes can be divided into two functional subsets according to expression patterns of an array of surface proteins: CD11b ϩ /Gr-1 ϩ /CX3CR1 lo /CCR2 ϩ /CD62L ϩ cells are the cells that are actively recruited to inflamed tissues, and CD11b ϩ /Gr-1 Ϫ /CX3CR1 ϩ /CCR2 Ϫ /CD62L Ϫ cells are resident monocytes that give rise to specialized cell types, including osteoclasts. According to this classification, our CD11b ϩ /Gr-1 Ϫ/lo cells appear to belong to resident monocytes. Although the significance of this kind of classification in vivo is controversial (28), it may be possible to define further OCPs by using multiple surface markers.
Using CD11b ϩ /Gr-1 Ϫ/lo as broad OCP markers, we demonstrated that TNF␣ promotes the differentiation of bone marrow CD11b ϩ / Gr-1 Ϫ/lo /c-Fms Ϫ cells to CD11b ϩ /Gr-1 Ϫ/lo /c-Fms ϩ cells. Thus, the first step of TNF␣-induced osteoclastogenesis is the conversion of c-Fms Ϫ OCPs to c-Fms ϩ OCPs. This results in enhanced sensitization of precursor cells to M-CSF and provides more M-CSF-responding cells. TNF␣-induced c-Fms mRNA expression has been reported in murine tissue macrophages (29). However, tissue macrophages are terminally differentiated cells that do not have osteoclastforming potential (30), and the findings from macrophages may differ from those of osteoclasts or their precursors. Our findings provide the first experimental evidence of TNF increasing c-Fms expression by OCPs, and the functional consequence of this increased c-Fms expression on osteoclastogenesis (Fig. 5). In addi- tion, the failure of neutralizing M-CSF antibodies to block TNF␣induced c-Fms expression indicates that this is a direct effect and is not secondary to M-CSF autoregulation (Fig. 5). Recently, Kitaura et al. (31) used chimerical WT and TNF receptor knock-out mice to dissect the contribution of cells in the osteoclast lineage from that of stromal cells in the response to TNF␣-mediated osteoclastogenesis in vivo. From the findings of this elegant study, they conclude that in conditions where TNF␣ concentrations are high, TNF␣ can fully exert its bone erosive effects by directly promoting the differentiation of osteoclast precursors independent of cytokine-responsive stromal cells and T lymphocytes (31). Our results are consistent with their findings and further point out that one of the mechanisms by which TNF promotes OCP differentiation is though up-regulation of c-Fms expression.
We found that M-CSF alone increases c-Fms ϩ cell numbers (Fig. 5), indicating that M-CSF can induce its own receptor expression and thereby forms an autocrine loop to amplify M-CSF-mediated signals. This is not a surprising because c-Fms expression is controlled by PU.1, one of the Ets transcription factors that is stimulated by M-CSF treatment (32). Because M-CSF is constitutively expressed in the bone microenvironment in vivo, and TNF␣ stimulates osteoblasts and T lymphocytes to up-regulate their expression of M-CSF (16,17), TNF␣induced c-Fms expression must work through both M-CSF-dependent and -independent mechanisms. We have reported previously that TNF␣ promotes mature osteoclast survival (33). However, we did not observe clear changes in OCP apoptosis. Thus, TNF␣ likely has a different effect on OCPs versus mature osteoclasts; it stimulates proliferation and differentiation of the former to provide more OCPs and increases survival of the latter to prolong the duration of bone resorption (Fig. 6).
TNF␣ administration stimulates bone marrow OCP proliferation 1 day before the increase in total OCPs numbers are observed in both bone marrow and blood (Fig. 3). This finding, along with similar findings in TNF-Tg mice, suggests that TNF␣-induced OCP proliferation and the subsequent increase in the bone marrow OCP pool may lead to the increased OCPs numbers in the peripheral blood. Based on these observations, we propose that in chronic inflammatory arthritis, where TNF␣ levels are elevated, TNF␣ increases marrow OCP numbers, which may directly result in a "push" of cells out of the bone marrow (marrow cell release) into the bloodstream because of the limited space available in the bone marrow cavity. Thus, chronic exposure to TNF␣ could increase the total turnover rate of OCPs. Unfortunately, we do not yet have experimental evidence to prove that the increase in blood OCP numbers in TNF-Tg mice is a direct consequence of an enlarged bone marrow pool. However, it is known that proliferation is not a pre-requisite for cell release from the bone marrow. In fact, the majority of common mobilizing factors stimulates marrow cell release without influencing their proliferation (18). Thus, TNF␣-induced elevation in blood OCPs might result from different mechanisms than common mobilization and represent a unique response to chronic inflammation.
One important question regarding circulating OCPs is the role of these cells in normal bone remodeling, because this process occurs exclusively within the bone marrow cavity. Do OCPs need to travel in the blood to initiate physiological bone remodeling or do they simply receive signals from osteoblast/lining cells adjacent to them on or near bone surface identified for remodeling? We do not know the answers to these questions at this time. However, given that focal erosions develop from chronic inflammation over a number of years and the active osteoclasts that mediate these erosions have a half-life of only a few weeks, these cells must be replaced continuously. Because these osteoclasts are known to enter the joint through synovium as OCPs, our conclusion is that circulating OCPs serve as the pre-osteoclast reservoir. Identification of the mechanisms whereby OCP generation and trafficking are regulated should help answer these important questions.
In summary, our studies reveal a new mechanism by which TNF␣ stimulates osteoclast-mediated bone resorption in chronic inflammatory arthritis. We propose that the first influence of TNF␣ on osteoclastogenesis is to increase bone marrow OCP genesis through the following two mechanisms (Fig. 6): 1) TNF␣ directly stimulates the conversion of CD11b ϩ /Gr-1 Ϫ/lo /c-Fms Ϫ cells to c-Fms ϩ cells; and 2) it stimulates osteoblasts/stromal cells to produce M-CSF, which subsequently increases c-Fms ϩ cell proliferation. These CD11b ϩ /Gr-1 Ϫ/lo /c-Fms ϩ cells are then released to the bloodstream, through a yet to be defined mechanism, and then home to inflamed joints where they differentiate in response to RANKL and other osteoclastogenic signals.