Commitment and Differentiation of Osteoclast Precursor Cells by the Sequential Expression of C-Fms and Receptor Activator of Nuclear Factor κb (Rank) Receptors

8–10-wk-old female C57BL/6 mice were purchased from Japan SLC. Bone marrow (BM) cells were freshly prepared from femur and tibia and used as a source of hematopoietic precursors.

1,25-dihydroxyvitamin D₃ (1,25-[OH]₂D₃) was provided by Dr. Ishizuka (Teijin Institute for Biomedical Research, Tokyo, Japan). Recombinant M-CSF and anti–mouse M-CSF neutralizing antibody were purchased from R & D Systems, Inc. Recombinant IL-7 and recombinant IL-3 were provided by Toray Industries Inc. Recombinant stem cell factor (SCF) was a gift from Chemo-Sero-Therapeutic Co., Ltd. Erythropoietin (Epo) was a gift from Snow-Brand Milk Product Co.

Mice were killed by cervical dislocation, and the tibiae and femora were removed and dissected free from adhering soft tissues. The bone ends were cut off with a scalpel, and the marrow was flushed with α-modified (α-)MEM (GIBCO BRL) containing 10% FCS (JRH Biosciences). Mononuclear cells were isolated by centrifugation of total BM cells on Lymphep™ (Nycomed Pharmaceuticals) according to the manufacturer's instructions.

A DNA fragment encoding the extracellular domain (Asp₇₆–Asp₃₁₆) of RANKL was prepared using reverse transcriptase (RT)-PCR from total RNA of ST2 stromal cells that were cultured in the presence of 1,25-(OH)₂D₃ (10⁻⁸ M) for 4 d. PCR primers were as follows: 5′-CCGCTCGAGCGTCTATGTCCTGAACTTTGA-3′ (sense) and 5′-CCCAAGCTTGATCCTAACAGAATATCAGAAGACA-3′ (antisense).

The PCR product was digested with HindIII and XhoI and ligated into the HindIII and XhoI sites of the pSecTag2 vector (Invitrogen Corp.) to yield pSecTag2–soluble (s)RANKL containing His₆ and myc tags. pSecTag2–sRANKL was transfected into COS7 cells cultured in DMEM (Life Technologies, Inc.) containing 10% FCS, and the supernatant was collected every 4 d for 12 d. sRANKL was purified from the supernatant using TALON and TALON Superflow Metal Affinity Resin (Clontech). The supernatant was applied to a chromatography column (Bio-Rad Labs.) filled with affinity resin, and the trapped proteins were eluted in 125 nM imidazol and fractionated. The high concentration fraction was collected and applied to a PD-10 column (Amersham Pharmacia Biotech, Inc.), eluted in PBS to eliminate the imidazol, and fractionated. Fractions containing a large amount of sRANKL were concentrated by a centricon concentrator (Amicon, Inc.).

Mouse BM–derived stromal cell line ST2 (a gift from Dr. Hayashi, Tottori University, Yonago, Japan) was maintained in RPMI 1640 (GIBCO BRL) supplemented with 10% FCS with 10⁻⁵ M of 2-ME (GIBCO BRL) at 37°C in humidified 5% CO₂ air. Cells were harvested every 3 d using 0.05% trypsin–EDTA (GIBCO BRL), and 10⁵ cells were passaged on 100-mm culture dishes (Falcon 3003; Becton Dickinson Labware).

The newborn calvaria-derived mouse stromal cell line OP9 (a gift from Dr. Kodama, Bayer Yakuhin Ltd., Kyoto, Japan) was maintained in α-MEM supplemented with 20% FCS at 37°C in humidified 5% CO₂ air. Cells were harvested every 3 d using 0.05% trypsin–EDTA, and 1.5 × 10⁵ cells were resuspended on 100-mm cell culture dishes.

The cell staining procedure for flow cytometry was performed as previously described 33. The mAbs used in immunofluorescence staining were anti–c-Kit antibody (ACK2; a gift from Dr. S.-I. Nishikawa, Kyoto University, Kyoto, Japan), anti–Mac-1 (CD-11b) antibody (M1/70; PharMingen), anti–c-Fms antibody (AFS98; a gift from Toray Industries Inc., anti-RANK antibody (muRANK-M395; a gift from Immunex Corp.), anti-CD51 antibody (H9.2B8; PharMingen), and anti-B220 antibody (RA3-6B2; PharMingen). The mAbs used were either biotinylated or fluoresceinated. Biotinylated mAbs were detected with streptavidin-conjugated allophycocyanin (Caltag Labs.) or streptavidin-conjugated Red613 (GIBCO BRL). The following rat Igs were used as isotype controls: biotinylated IgG2a and IgG2b and fluoresceinated IgG2a and IgG2b (PharMingen). Cells were incubated for 15 min on ice with CD16/32 (FcγIII/II receptor; 1:100; FcBlock™; PharMingen) before staining with the first antibody. 10⁶ cells/100 μl were suspended in 5% FCS/PBS (washing buffer). Cells were stained with the first antibody, incubated for 30 min on ice, and washed twice with washing buffer. The secondary antibody was added, and the cells were incubated for 30 min on ice. After incubation, cells were washed twice with washing buffer and suspended in washing buffer for FACS^® analysis. The stained cells were analyzed and sorted by FACSVantage™ (Becton Dickinson).

Sorted cells were centrifuged onto microscope slides using a cytospin centrifuge (Shandon Southern Products) and stained with May-Grünwald-Giemsa staining solution (Merck KGaA).

5 × 10³ ST2 cells were plated in 96-well plates (PRIMARIA; Becton Dickinson Labware) 1 d before coculture with sorted BM cells. Cultures were maintained in α-MEM/10% FCS with 10⁻⁸ M 1,25-(OH)₂D₃ and 10⁻⁷ M dexamethasone (Dex). Tartrate-resistant acid phosphatase (TRAP) staining was performed on day 4, and the TRAP solution assay (TRAP activity) was performed on day 7 or 10. Cocultures were scaled down to 100, 50, 40, 30, 20, 10, 8, 6, 5, 4, 2, and 1 cell in 100 μl for a limiting dilution assay.

To assay osteoclast formation without a stromal layer, sorted BM cells were cultured in α-MEM/10% FCS in the presence of recombinant M-CSF (100 ng/ml) and sRANKL (25 ng/ml).

After aspiration of medium, cells were fixed with 1% glutaraldehyde (Wako Pure Chemical Industries, Ltd.) in PBS for 15 min at room temperature. Cells were stained for TRAP using a histochemical kit (no. 387; Sigma Chemical Co.) according to the manufacturer's instructions. TRAP⁺ cells were scored as osteoclasts microscopically.

In the TRAP solution assay, enzyme activity was examined by the conversion of α-naphthyl phosphate (4 mmol/liter; Sigma Chemical Co.) to α-naphthol in the presence of 2 mol/liter l-tartrate solution (Sigma Chemical Co.) in each well. Absorbance was measured at 405 nm using a microplate reader (model 550; Bio-Rad Labs.).

c-Kit⁺Mac-1^dullc-Fms⁻ cells were cultured for 2 d in α-MEM/10% FCS in the presence of SCF (100 U/ml), and c-Fms expression was examined. Sorted c-Kit⁺Mac-1^dullc-Fms⁺ cells derived from BM were cultured for 3 d in α-MEM/10% FCS plus either M-CSF (30 ng/ml) or IL-3 (100 U/ml), and the expression of RANK was analyzed. To investigate whether c-Kit⁺Mac-1^dullc-Fms⁻ cells and c-Kit⁺Mac-1^dull c-Fms⁺ cells differentiate into B cells, these cells were cocultured with OP9 stromal cells for 10 d in RPMI 1640/10% FCS in the presence of IL-7 (20 U/ml). After 10 d in culture, the expression of B220 was analyzed using FACSVantage™.

An RNeasy mini Kit (QIAGEN GmbH) was used for isolation of total RNA from total BM cells or fractionated BM cells. Total RNA was reverse transcribed using an RT for PCR kit (Clontech). The cDNAs were amplified using an Advantage polymerase mix (Clontech) in a GeneAmp PCR system (model 9700; Perkin-Elmer Corp.) for 25–30 cycles. Sequences of gene-specific primers for RT-PCR were as follows: 5′-mRANK, CCAGGGGACAACGGAATCAG; 3′-mRANK, GGCCGGTCCGTGTACTCATC; 5′-mβ-actin, TCGTGCGTGACATCAAAGAG; and 3′-mβ-actin, TGGACAGTGAGGCCAGGATG. Each cycle consisted of 30 s denaturation at 94°C and 4 min annealing/extension at 70°C.

To compare CFU-C (culture) activity between c-Kit⁺Mac-1^dullc-Fms^± cells, 10³ cells were cultured in 1 ml of culture medium containing α-MEM, 1.2% methylcellulose (1,500 centipoise; Aldrich Chemical Co.), 30% FCS, 1% deionized BSA, 50 mM 2-ME, 100 U/ml IL-3, 2 U/ml Epo, and 100 ng/ml SCF. After 7 d in culture, aggregates of 50 or more cells were counted as a single colony.

For in vitro osteoclast colony formation assay, 10³ R3 cells were plated in methylcellulose medium. In brief, cells were embedded in 1 ml of 1.2% methylcellulose, 30% FCS, 1% deionized BSA, 50 mmol/liter 2-ME, and 100 ng/ml M-CSF in the presence or absence of sRANKL (25 ng/ml) in α-MEM. The culture dishes were incubated in humidified atmosphere at 37°C with 5% CO₂. After 7 d, the colonies were counted, picked up, and stained for nonspecific esterase, May-Grünwald-Giemsa, and TRAP.

To characterize and isolate osteoclast precursor cells, expression of cell surface markers was analyzed (Fig. 1 A). Of the c-Kit⁺ BM mononuclear cells, 37.5% were c-Fms⁺ and 56.1% were Mac-1⁺. Using contour blot analyses, four populations of c-Kit⁺ cells were detected: Mac-1^dull c-Fms⁺ (R3), Mac-1^highc-Fms⁺ (R4), Mac-1^dull c-Fms⁻ (R5), and Mac-1^highc-Fms⁻ (R6). These cells were fractionated, and further analyses were performed. The percentages of these populations in both BM mononuclear cells and c-Kit⁺ cells is shown in Table. The morphology of cells of each fraction was examined by May-Grünwald Giemsa staining (Fig. 1 B). R3 and R5 cells were immature cells showing a large N/C ratio, whereas most of the R4 and R6 cells were mature cells, including macrophages or neutrophils.

TRAP is an enzyme highly expressed in both immature and mature osteoclasts. To determine which fractionated cells differentiate into TRAP⁺ cells, 10³ cells of each fraction were cocultured with an ST2 stromal cell line in the presence of 1,25-(OH)₂D₃ and Dex (Fig. 1C and Fig. D). To detect TRAP⁺ mononuclear or multinuclear osteoclasts, TRAP staining was performed on day 4 (Fig. 1 C) and TRAP activity (TRAP solution assay) was measured on days 7 and 10 (Fig. 1 D). The number of TRAP⁺ cells in the R3 fraction was 10-fold higher than observed in unfractionated cells (Fig. 1 C). Relative TRAP activity (TRAP activity in each cell fraction/TRAP activity in unfractionated BM mononuclear cells) was highest in fraction R3 on day 7. In contrast, on day 10, the highest TRAP activity was detected in fraction R5 (1.55 ± 0.20) compared with R3 (1.17 ± 0.01). The R4 and R6 fractions showed low TRAP activities on both days 7 and 10.

These data suggest that the R3 fraction contains a higher proportion of osteoclast precursor cells than do other populations (R4, R5, and R6) and that R5 cells also contain osteoclast precursor cells that are less mature than R3 cells.

To investigate whether R5 cells (c-Kit⁺Mac-1^dullc-Fms⁻) differentiate to R3 cells (c-Kit⁺Mac-1^dullc-Fms⁺), the expression of c-Fms was analyzed after cultivation (Fig. 2 A). c-Fms⁻ cells in R5 or R6 were sorted and cultured in SCF (100 ng/ml), as they expressed c-Kit receptors. After 2 d in culture, 42.2% of cultured R5 cells (R5′) expressed c-Fms and 7.1% of c-Fms⁺ cells were also c-Kit⁺. In contrast, of cultured R6 cells (R6′), 9.3% were c-Fms⁺ cells and 0.4% were c-Kit⁺c-Fms⁺ cells. Moreover, of R5′ cells, c-Kit⁺ c-Fms⁺ cells were mainly Mac-1^dull (80.3%; Fig. 2 B). To determine if c-Fms⁺ cells in R5′ or R6′ could undergo osteoclastic differentiation, both cell fractions were cocultured with ST2 stromal cells for 4 d in the presence of both 1,25-(OH)₂D₃ and Dex and assayed for TRAP activity (Fig. 2 C). Although c-Fms⁺ cells of R5′ differentiated to osteoclasts, c-Fms⁺ cells of R6′ did not differentiate into TRAP⁺ cells. R6′ cells were mature granulocytes and macrophages.

To analyze the proportion of osteoclast precursor cells in unfractionated BM cells, R3, R5, or c-Kit⁺c-Fms⁺ R5′ cells, each fraction was cultured for 4 d on ST2 stromal cells and scored for TRAP⁺ cells by limiting dilution analysis (Fig. 2 D). Limiting dilution of BM cells, R3, R5, and c-Kit⁺ c-Fms⁺ R5′ cells revealed 2.0% (1/49.3; unfractionated BM), 20.4% (1/4.9; R3), 3.4% (1/29.8; R5), or 16.4% (1/6.1; c-Kit⁺c-Fms⁺ R5′) TRAP⁺ cells. R3 or c-Kit⁺c-Fms⁺ R5′ cells contained 10-fold more osteoclast precursor cells than did unfractionated BM cells. These data suggest that c-Fms⁻ cells differentiate to osteoclasts through the induction of c-Fms.

c-Fms expression increased during cultivation. As c-Kit⁺c-Fms⁺ R5′ cells contained the same proportion of osteoclast precursor cells as did R3 cells, we speculated that R5 cells are less mature than R3 cells. To examine the capacity of those osteoclast precursor cells to differentiate to other lineages, a colony assay using R3 or R5 cells was performed (Table). 10³ cells from each fraction were cultured with methylcellulose semisolid medium in the presence of SCF (100 ng/ml), IL-3 (100 U/ml), and Epo (2 U/ml) for 7 d. The numbers of colonies (CFU-GEMM [granulocyte, erythrocyte, macrophage, megakaryocyte] burst forming unit-erythrocyte, CFU-GM [granulocyte, macrophage], CFU-G, and CFU-M [macrophage]) were counted under an inverted microscope. Most colonies derived from R3 cells were macrophage colonies. In contrast, R5 cells contained more CFU-GM than CFU-M. Moreover, CFU-GEMM or burst forming unit-erythrocyte–derived colonies were observed in R5 cells.

To analyze the ability of R3 or R5 to differentiate into a B cell lineage, 10⁴ cells of each fraction were cocultured with the OP9 stromal cell line and IL-7 (20 U/ml; Table). After 10 d in culture, the number of expanded nonadherent cells was 73-fold more in R5 (7.9 ± 0.6 × 10⁶ cells) than in R3 (1.0 ± 0.2 × 10⁵ cells), and 47.0% of R3 or 93.9% of R5 cells were B220⁺ cells. These data suggest that R5 cells are more immature than R3 cells and that R5 cells can differentiate not only into osteoclast but also into myeloid, erythroid, or B cell lineages.

Expression of RANK in fractionated cells was examined by RT-PCR and FACS^®. To examine whether M-CSF induces RANK mRNA expression, unfractionated BM mononuclear cells were cultured for 72 h in the presence of IL-3 (100 U/ml) or M-CSF (30 ng/ml). Both IL-3 and M-CSF are able to support the differentiation of macrophagic differentiation. The expression of RANK mRNA was detected in a 24-h incubation with M-CSF and a 72-h incubation with IL-3 (Fig. 3 A). Subsequently, fractionated cells were cultured with M-CSF for 48 h. Before induction with M-CSF, very low levels of RANK mRNA were detected in all fractionated cells except R6 cells (Fig. 3 B). After incubation for 48 h, the expression of RANK mRNA was obvious in the R3 fraction. FACS^® analyses demonstrated that 5.4% of unfractionated BM cells were RANK⁺ (data not shown). Of R3 cells, 15.7% were RANK⁺, and 1.5% of R5 cells were RANK⁺ (Fig. 3 C). The expression of RANK protein in R3 or R5 cells was analyzed by FACS^® after incubation with M-CSF for 24 or 72 h (Fig. 3 C). The percentage of RANK⁺ cells in R3 increased with longer incubations with M-CSF (41.3% for a 24-h incubation and 58.4% for a 72-h period). Although a similar increase was also observed in R5 cells, the overall percentage of RANK⁺ R5 cells was lower than that of R3 cells (2.6% after a 24-h incubation and 11.5% after 72 h). To characterize RANK⁺ or RANK⁻ cells in the R3 fraction cultured with M-CSF, cells were sorted with a RANK mAb, and the expression of c-Kit, Mac-1, and c-Fms was analyzed. After 24 h in culture with M-CSF, sorted RANK⁺ cells were c-Kit⁻Mac-1^highc-Fms⁺, and sorted RANK⁻ cells were c-Kit⁻Mac-1^dullc-Fms⁺ (data not shown). RANK expression in R3 cells was induced by stimulation of M-CSF; however, R3 cells originally contained RANK⁺ cells. To examine whether M-CSF induces the expression of RANK on RANK⁻ R3 cells, R3 cells were subdivided into RANK⁺ and RANK⁻ fractions. RANK⁻ R3 cells were cultured in the presence of M-CSF with or without anti–mouse M-CSF neutralizing antibody, and the expression of RANK was analyzed by RT-PCR. The expression of RANK was detected in freshly isolated RANK⁺ R3 cells and cultured RANK⁻ R3 cells. However, RANK expression was significantly reduced by the addition of mouse M-CSF neutralizing antibody to the culture (Fig. 3 D). These results indicate that M-CSF stimulates the expression of RANK on osteoclast precursor cells.

Next we analyzed how osteoclast precursor cells differentiate into TRAP⁺ cells in the presence of M-CSF and sRANKL rather than ST2 stromal cells. Fractionated R3 cells were cultured for 72 h in IL-3 (100 U/ml) or M-CSF (30 ng/ml). As shown in Fig. 4 A, M-CSF induced RANK expression more efficiently than did IL-3. RANK⁺ or RANK⁻ cells precultured with IL-3 or M-CSF for 72 h were sorted and cultured in the presence of sRANKL and either IL-3 or M-CSF. The percentage of TRAP⁺ cells is shown in Fig. 4 B. Both IL-3– and M-CSF–precultured cells differentiated into TRAP⁺ cells in the presence of sRANKL and M-CSF. Moreover, RANK⁻ cells in each preculture condition showed a higher percentage of TRAP⁺ cells than RANK⁺ cells in the presence of sRANKL and M-CSF. M-CSF–precultured cells showed a higher percentage of TRAP⁺ cells than did IL-3–precultured cells.

To examine whether RANK⁺ cells differentiate to TRAP⁺ cells in the presence of sRANKL alone, primary R3 cells or R3 cells precultured with M-CSF were cultured for 2, 4, or 6 d and analyzed for TRAP staining. RANK⁺ cells precultured for 24 h differentiated into TRAP⁺ cells, and the percentage of TRAP⁺ cells decreased in the presence of sRANKL alone (Fig. 4 C). However, primary R3 cells (RANK⁺ or RANK⁻) or RANK⁻ cells precultured for 24 h did not differentiate into TRAP⁺ cells with sRANKL alone. No TRAP⁺ cells were observed when cells were cultured for 6 d with M-CSF alone (data not shown). These data suggest that M-CSF affects not only the survival factor but also the competence factor for osteoclast precursor cells during their differentiation.

To clarify the synergistic effect of M-CSF and sRANKL, the relationship between the onset of RANK expression and osteoclast differentiation was examined. Fig. 5 shows the flow chart of cell sorting and conditions of further cultivation. Sorted RANK⁺ or RANK⁻ cells from the primary R3 fraction were cultured for 2, 4, and 6 d with both sRANKL and M-CSF (Fig. 6A and Fig. B). On days 2 and 4, the percentage of TRAP⁺ cells was higher in RANK⁻ cells (11.8 ± 1.9% on day 2 and 97.5 ± 1.4% on day 4) than in RANK⁺ cells (7.0 ± 1.7% on day 2 and 42.5 ± 6.1% on day 4). On day 6, however, the percentage of TRAP⁺ cells was similar in RANK⁺ and RANK⁻ cells. Interestingly, the percentage of TRAP⁺ cells in R3 cells precultured for 24 h with M-CSF was similar between RANK⁺ and RANK⁻ cells (Fig. 6C and Fig. D), whereas in the case of R3 cells that were not precultured, RANK⁺ cells did not efficiently differentiate into TRAP⁺ cells (Fig. 6A and Fig. B). RANK⁻ cells showed a higher percentage of MNCs, which are fully matured osteoclasts, than did RANK⁺ cells. These MNCs from RANK⁻ cells were extremely large and contained a large number of nuclei.

Growth of RANK⁺ and RANK⁻ cells after 72 h of preculture with M-CSF was less than that of primary or 24-h–precultured R3 cells (Fig. 6 E). Although the percentage of TRAP⁺ cells was higher in RANK⁻ cells than RANK⁺ cells on day 4 (9.16 ± 4.3% in RANK⁺ and 95.6 ± 1.8% in RANK⁻ cells), the percentage was similar on day 6 (97.1 ± 0.5% for RANK⁺ and 99.5 ± 0.1% for RANK⁻ cells; Fig. 6 F). A delay in differentiation of TRAP⁺ cells from RANK⁺ cells was observed. Also RANK⁻ cells showed a higher percentage of TRAP⁺ MNCs than did RANK⁺ cells, but this number was lower than that observed in RANK⁻ cells precultured for 24 h. Moreover, the size of MNCs was reduced and the number of nuclei contained in MNCs was smaller in RANK⁻ cells precultured for 72 h than in cells precultured for 24 h.

In all culture conditions, especially 24-h preculture with M-CSF, the potential of cell growth of RANK⁻ cells was greater than that of RANK⁺ cells. Also, RANK⁻ cells formed a large number of MNCs, which contained high numbers of nuclei because of the increment of cell density induced by cell proliferation. In addition, when RANK⁺ and RANK⁻ cells were precultured with M-CSF for 24 h, a higher number of TRAP⁺ cells was observed in the latter, suggesting that RANK expression is necessary but not sufficient for osteoclast differentiation and that the timing of RANKL binding to RANK⁺ cells may be critical for commitment of osteoclast cells.

Fractionated cells (R3 or R5) can differentiate to lineages other than osteoclast (Table and Table); however, it is not clear whether a single osteoclast precursor cell is committed to differentiate into a TRAP⁺ cell or whether it can differentiate into other lineages. To understand the mechanism of osteoclast differentiation, R3 cells were cultured in methylcellulose instead of liquid culture (Fig. 7). 10³ R3 cells were cultured with methylcellulose medium containing M-CSF (100 ng/ml) in the presence or absence of sRANKL (25 ng/ml) for 7 d. In the absence of both cytokines, colony formation was not detected, whereas 139 ± 14.5 and 131 ± 1.7 colonies were observed in M-CSF alone and in M-CSF and sRANKL, respectively (Table). Whereas colonies formed in the presence of M-CSF alone were tightly compacted (Fig. 7 A), colonies observed in M-CSF and RANKL were of mixed type (Fig. 7 B). Individual colonies in the presence of M-CSF and sRANKL contained not only nonspecific esterase–positive macrophages but also TRAP⁺ cells (Fig. 7C–E). The percentage of TRAP⁺ cells in 20 colonies was 0 ± 0% (M-CSF alone) and 43.8 ± 30.4% (13.3–94.2%; M-CSF and sRANKL; Table). Any homogeneous colonies consisting of all TRAP⁺ cells were not observed. These data strongly suggest that single precursor cells can differentiate into TRAP⁺ cells and macrophages at a late stage of osteoclast differentiation.

Here we identify early and late stages of osteoclast precursor cells and describe the differentiation pathway of osteoclasts from hematopoietic precursor cells by coculture of those cells with the ST2 stromal cell line and 1,25-(OD)₂D₃. We also analyze the osteoclast commitment process by substituting M-CSF and sRANKL for the stromal cells.

Hematopoietic precursor cells exist in the c-Kit⁺ fraction 34,35. This fraction was clearly subdivided by the expression of c-Fms and Mac-1. Mac-1⁺ cells contain mainly mature granulocytes and macrophages, whereas Mac-1^dull cells are multipotential progenitor cells 36,37. We demonstrate that c-Kit⁺Mac-1^dullc-Fms⁺ cells in murine BM are early stage precursors of osteoclasts using a limiting dilution method. These cells are shown to be derived from c-Kit⁺Mac-1^dullc-Fms⁻ cells. We show directly that c-Fms⁻ cells differentiate into c-Fms⁺ cells after 2 d in culture with SCF. It has been previously reported that c-Fms is a key determinant in the differentiation of monocyte–macrophage lineage cells 38. c-Fms expression is regulated by a tissue-specific promoter. Although the precise mechanism of c-Fms expression is not known, transcription factors c-ets-1, c-ets-2, and PU.1 mediate induction of c-Fms 39,40,41. Thus, it is reasonable that c-Kit⁺ Mac-1^dullc-Fms⁺ cells differentiate into TRAP⁺ osteoclasts after 4 d, whereas c-Kit⁺Mac-1^dullc-Fms⁻ cells require >7 d. In addition, c-Kit⁺Mac-1^dullc-Fms⁺ cells express the RANK mRNA and protein in 24 h in the presence of M-CSF. Lacey et al. 21 demonstrated that osteoclast precursor cells were identified by sRANKL–FITC. They sorted sRANKL–FITC⁺ cells after cultivation of mouse BM cells in the presence of M-CSF and sRANKL for 1 d, and these precursor cells were therefore regarded as RANK⁺ cells. These sRANKL–FITC⁺ cells formed a larger number of multinucleated TRAP⁺ cells than sRANKL–FITC⁻ cells. In contrast to this finding, we found that a large number of multinucleated TRAP⁺ cells were derived from RANK⁻ rather than RANK⁺ cells, regardless of precultivation with M-CSF (Fig. 6). By cultivation with M-CSF and sRANKL for 1 d before cell sorting, it is speculated that sRANKL–FITC⁺ cells have already been committed to osteoclast lineage. We show a sequential change of phenotype in the differentiation pathway of osteoclasts (Fig. 8 A).

Although an mAb against c-Fms suppresses osteoclastogenesis 42, most osteoclast precursors in BM are c-Kit⁺ c-Fms⁻, and their numbers decrease along with expression of c-Fms 43. This observation is based on an assay of TRAP⁺ cells on day 6 of culture, which is consistent with our observation. We show that a later stage of osteoclast precursor cells (c-Fms⁺RANK⁺) differentiates into TRAP⁺ osteoclasts in 2 d. It was reported that osteoclast precursor cells were significantly higher in the c-Kit^low fraction, whereas myeloid cells of other lineages were higher in the c-Kit^high fraction 44. However, the c-Kit^high fraction contained a similar number of osteoclast precursor cells 45.

We showed that c-Kit⁺Mac-1^dullc-Fms⁻ cells can differentiate not only into the myeloid lineage but also into B220⁺ B cells in the presence of IL-7. On the other hand, c-Kit⁺Mac-1^dullc-Fms⁺ cells differentiate into osteoclasts at a frequency of 1:5 on stromal cells. Stromal cells such as ST2 express RANKL, and its expression can be upregulated by the bone-resorbing factors 1,25-(OH)₂D₃, IL-11, prostaglandin E2, and parathyroid hormone 46. However, the ST2 coculture system with 1,25-(OH)₂D₃ does not maximally stimulate osteoclastogenesis, as ST2 also produces the inhibitory molecule OPG/OCIF, which belongs to the TNFR family 47,48.

We investigated the commitment process of c-Kit⁺Mac-1^dullc-Fms⁺ cells to osteoclasts using M-CSF and sRANKL instead of stromal cells. Our data suggest that M-CSF stimulates c-Kit⁺Mac-1^dull c-Fms⁺ cells to induce RANK mRNA and protein in 24 h more efficiently than does IL-3. These cells can differentiate into osteoclasts in the presence of both M-CSF and sRANKL. With M-CSF only, they differentiate into macrophages but not into osteoclasts. RANKL is a differentiation factor for osteoclasts but not an exclusive osteoclast commitment factor, as RANK is expressed not only in osteoclasts but in T cells and dendritic cells 24. Mice with a disrupted RANKL (opgl) gene show severe osteopetrosis and lack all lymph nodes 32, suggesting that RANKL–RANK signaling plays several roles in organogenesis.

Preculture of c-Kit⁺Mac-1^dullc-Fms⁺ cells with M-CSF for 24 or 72 h results in different fates for osteoclast precursors. RANK⁺ cells after 24-h preculture differentiate into osteoclasts more efficiently than after 72-h preculture. RANK⁺ cells may autonomously differentiate into the macrophage lineage in the absence of sRANKL for 72 h. By contrast, in the presence of both sRANKL and M-CSF after 24-h preculture with M-CSF, RANK⁺ cells efficiently differentiate into osteoclasts. RANK⁻ cells, after preculture with M-CSF, may differentiate into RANK⁺ cells if they are continuously exposed to M-CSF. Once the cells express RANK, existing RANKL binds to the receptors. This is the most productive system for osteoclast differentiation, which may take place in vivo on bone surfaces. Thus, RANK⁻ cells differentiate into osteoclasts more efficiently than RANK⁺ cells. Moreover, as their proliferative activity is higher, multinuclear osteoclast formation from RANK⁻ cells is greater than that of RANK⁺ cells. This is dependent upon the cell density (data not shown).

c-Kit⁺Mac-1^dullc-Fms⁺ cells differentiate exclusively into osteoclasts (∼100%) in the presence of both M-CSF and sRANKL. As even RANK⁺ cells can differentiate into macrophages, we conclude that the stage of commitment of osteoclasts is late in the differentiation process. A clonal assay using colony formation in semisolid methylcellulose medium provided us with precise identification of constituent cells and revealed no pure osteoclast colonies in the presence of both M-CSF and sRANKL. A single colony contained 13.3–94.2% TRAP⁺ osteoclasts. This observation suggests the very low incidence of osteoclast colony forming cells, which are exclusively committed to osteoclasts. The growth of osteoclasts in methylcellulose is poorer than that observed in a liquid culture system, which enables osteoclasts to attach to the culture dish. Cell anchoring or polarity is critical for osteoclast differentiation. A study of cell adhesion molecules expressed by osteoclasts is underway.

Our findings suggest that M-CSF plays three roles in osteoclastogenesis: (a) it induces RANK, (b) it is a competence factor for differentiation, and (c) it stimulates cell survival and proliferation (Fig. 8 B). Although upregulation of RANK by M-CSF is higher than that by IL-3 (Fig. 3 A and 4 A) and GM-CSF (data not shown), RANK induction is not M-CSF specific. Even though expression of RANK on RANK⁻ cells was upregulated in the presence of M-CSF and its expression was significantly inhibited by addition of neutralizing M-CSF antibody (Fig. 3 D), further analysis in association with the signaling pathway would be necessary to elucidate the mechanism of c-Fms–RANK interaction.

As shown in Fig. 4 C, TRAP⁺ cells differentiated from M-CSF–precultured RANK⁺ cells but not from primary RANK⁺ cells in the presence of sRANKL alone, suggesting that unknown events induced by M-CSF are competent for differentiation to osteoclasts. Without M-CSF, RANKL–RANK could not induce osteoclasts even in the presence of IL-3. It will be interesting to define the molecular event induced by M-CSF in precursors. As RANK belongs to the TNFR family and does not stimulate cell proliferation, M-CSF–precultured RANK⁺ cells could not survive and proliferate without the continuous presence of M-CSF. It has been shown that RANKL stimulates mature osteoclasts to activate bone resorption 49. In the absence of RANKL, precursor cells autonomously differentiate into macrophages in the presence of M-CSF, a situation regarded as the default pathway of macrophagic differentiation. The determination of osteoclastic differentiation is achieved to avoid the default pathway of macrophagic differentiation from the signal of the RANK–RANKL system (Fig. 8 B).

Spontaneous recovery of adult op/op mice suggests the presence of unknown molecules 50,51. The osteoclast defect in M-CSF mutant osteopetrotic mice can be rescued by overexpression of the antiapoptotic protein Bcl-2 in the monocyte lineage 14. Although M-CSF is an indispensable factor for osteoclastogenesis, other molecules may induce RANK less efficiently than M-CSF. It is noteworthy that the intracellular domain of RANK directly binds TRAF2, TRAF5, and TRAF6 26,28, and TRAF6 deficiency results in osteopetrosis 29.

In conclusion, we have identified osteoclast precursor cells and clarified the function of M-CSF and RANKL. The cooperation of these two factors is critical for osteoclast differentiation. Osteoclastogenesis is thus a unique model system in the lineage determination of blood cell differentiation.

This work was supported by Grants-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Science, and Culture of Japan.