Lineage-committed osteoclast precursors circulate in blood and settle down into bone



Osteoclasts are derived from the monocyte/macrophage lineage, but little is known about osteoclast precursors in circulation. We previously showed that cell cycle–arrested quiescent osteoclast precursors (QOPs) were detected along bone surfaces as direct osteoclast precursors. Here we show that receptor activator of NF-κB (RANK)-positive cells isolated from bone marrow and peripheral blood possess characteristics of QOPs in mice. RANK-positive cells expressed c-Fms (receptors of macrophage colony-stimulating factor) at various levels, but scarcely expressed other monocyte/granulocyte markers. RANK-positive cells failed to exert phagocytic and proliferating activities, and differentiated into osteoclasts but not into dendritic cells. To identify circulating QOPs, collagen disks containing bone morphogenetic protein-2 (BMP disks) were implanted into mice, which were administered bromodeoxyuridine daily. Most nuclei of osteoclasts detected in BMP-2–induced ectopic bone were bromodeoxyuridine-negative. RANK-positive cells in peripheral blood proliferated more slowly and had a much longer lifespan than F4/80 (a macrophage marker)-positive macrophages. When BMP disks and control disks were implanted in RANK ligand-deficient mice, RANK-positive cells were observed in the BMP disks but not in the controls. F4/80-positive cells were distributed in both disks. Administration of FYT720, a sphingosine 1-phosphate agonist, promoted the egress of RANK-positive cells from hematopoietic tissues into bloodstream. These results suggest that lineage-determined QOPs circulate in the blood and settle in the bone. © 2011 American Society for Bone and Mineral Research


Osteoclasts, multinucleated cells responsible for bone resorption, differentiate from the monocyte/macrophage lineage under the tight regulation of osteoblasts, mononuclear cells responsible for bone formation.1–3 Osteoblasts express two cytokines essential for osteoclastic differentiation, macrophage-colony stimulating factor (M-CSF, also called colony-stimulating factor 1) and receptor activator of NF-κB ligand (RANKL).4–7 M-CSF is constitutively expressed by osteoblasts, while RANKL is inducibly expressed by osteoblasts in response to osteotropic hormones and factors such as 1α,25-dihydroxyvitamin D3 [1α,25(OH)2D3], parathyroid hormones (PTH), and prostaglandin E2 (PGE2).7, 8 Osteoblasts also produce osteoprotegerin (OPG), a soluble decoy receptor for RANKL, which inhibits osteoclastogenesis by blocking the RANKL-RANK interaction.9–11 Osteoclast precursors such as bone marrow-derived macrophages express RANK (RANKL receptors) and c-Fms (M-CSF receptors), and differentiate into osteoclasts in the presence of RANKL and M-CSF.12, 13

Osteoclasts are formed from monocytes and macrophages in cultures treated with RANKL and M-CSF. Therefore, researchers have believed that most cells of the monocyte/macrophage lineage existing in bone marrow or near bone can differentiate into osteoclasts at the correct site in vivo. However, our experimental results challenge this idea. We examined the characteristics and dynamics of osteoclast precursors in vivo, and found that the precursors are specific myeloid cells, not common monocytes or macrophages.14 We named those precursors “cell cycle–arrested quiescent osteoclast precursors (QOPs)” after their characteristics. QOPs had a long lifespan, and differentiated into osteoclasts in response to M-CSF and RANKL without cell cycle progression. Cell cycle withdrawal of osteoclast precursors was associated with downregulation of cyclins and cyclin-dependent kinases (Cdks) and upregulation of p27KIP1. Mononuclear cells expressing RANK and c-Fms but not Ki67, a marker of cell proliferation, were detected in the vicinity of osteoblasts in bone tissue in RANKL-deficient (RANKL−/−) mice,14 suggesting that osteoblasts play a role in the maintenance of QOPs.

Another line of experiments showed that the site where RANKL and M-CSF are expressed is not a determinant of the correct site for osteoclastogenesis.15 When collagen disks containing bone morphogenetic protein (BMP)-2 (BMP disks) or vehicle (control disks) were subcutaneously implanted into wild-type mice, tartrate-resistant acid phosphatase (TRAP; a marker of osteoclasts)-positive osteoclasts and alkaline phosphatase (ALP; a marker of osteoblasts)-positive osteoblasts simultaneously appeared in the BMP disks but not in the control disks. Using this system of ectopic bone formation and RANKL−/− mice, we examined how the site of osteoclastogenesis is determined. BMP disks or control-disks were implanted into RANKL−/− mice, which were intraperitoneally injected with RANKL or vehicle.15 Osteoclasts were totally absent in BMP-2–induced ectopic bones because of the RANKL deficiency. TRAP-positive osteoclasts appeared in the BMP disks, in response to RANKL injections. Osteoclasts were detected in close proximity to ALP-positive osteoblasts. In contrast, osteoclasts were not observed in the control disks, even those treated with RANKL. These results suggest that osteoblasts also play important roles in osteoclastogenesis by providing a suitable microenvironment for the action of RANKL, or for the homing of osteoclast precursors to bone. The distribution of QOPs in bone may determine the correct site for osteoclastogenesis.

Circulating osteoclast precursors were initially identified as biomarkers for erosive disease in psoriatic arthritis patients,16 and subsequently in rheumatoid arthritis, spondyloarthritis, osteoporosis, and tophaceous gout.17–19 Conversely, the loss of circulating osteoclast precursors has been correlated with nonerosive inflammatory arthritis.20, 21 Circulating osteoclast precursors are shown to be CD11b+Gr-1lowRANKlowc-Fmshigh cells.22, 23 They are largely quiescent, but their proliferation can be induced by M-CSF, tumor necrosis factor α (TNF-α) and systemic autoimmunity. Normal and TNF-α–induced generation of circulating osteoclast precursors are independent of RANK signaling, but this signal is required for lineage commitment and osteoclastogenesis.24, 25

In the present study, we further characterized osteoclast precursors in vivo. RANK-positive cells detected in bone marrow and also in the peripheral blood possessed characteristics of QOPs. Osteoclasts detected in BMP-induced ectopic bone were formed from circulating QOPs. FYT720, a sphingosine 1-phosphate agonist, was shown to stimulate migration of osteoclast precursor–containing monocytoid populations in vivo.26 Administration of FYT720 promoted the egress of RANK-positive cells from hematopoietic tissues into bloodstream. Our results support the current concept that osteoclast precursors circulate to find the correct site for osteoclastogenesis.

Materials and Methods


Three-, 6-, and 8-week-old male ddY mice were obtained from Japan SLC (Tokyo, Japan). RANKL−/− mice (C57BL/6) were generated in one of the authors' laboratories.27 All experiments were conducted in accordance with the guidelines for studies with laboratory animals of the Matsumoto Dental University Experimental Animal Committee.


The antibodies used for the immunohistochemical analysis were anti-F4/80 (A3-1) from BMA Biomedicals (Augst, Switzerland), biotin-conjugated anti-RANK and anti-c-Fms from R&D Systems (Minneapolis, MN, USA), and anti-Ki67 from Novocastra Laboratories (Newcastle, UK). The antibodies used for the flow cytometric analysis were phycoerythrin (PE)-conjugated anti-RANK (R12-31), biotin-conjugated anti-c-Fms (AFS98), PE-conjugated anti-F4/80 (BM8), allophycocyanin (APC)-conjugated anti-CD11b (M1/70), APC-conjugated anti-Ly-6G (Gr-1) (RB6-8C5), and APC-conjugated anti-F4/80 (BM8) from eBioscience (San Diego, CA, USA), and PE-conjugated B220 (RA36B2), fluorescein isothiocyanate (FITC)-conjugated CD3e (145-2C11), FITC-conjugated anti-CD11c (HL3), and anti-CD86 (GL-1) from BD Biosciences (San Jose, CA, USA).

Osteoclast preparations

Fluorescence-activated cell sorting (FACS)-sorted cells were cultured in α-modified essential medium (α-MEM; Sigma, St. Louis, MO, USA) containing 10% fetal bovine serum (FBS; JRH Bio-sciences, Lenexa, KS, USA) with 104 units/mL of M-CSF (Kyowa Hakko Kirin, Tokyo, Japan) and 5 nM RANKL (GST-RANKL; Oriental Yeast, Tokyo, Japan) in the presence or absence of 50 µM hydroxyurea (MP Biomedicals, Eschwege, Germany) in 96-well plates (1 × 104 cells/well). After being cultured for 5 days, osteoclasts were fixed and stained for TRAP.28 Cell growth was estimated by the Alamar Blue assay (Biosource, Camarillo, CA, USA).

Phagocytosis assay

The phagocytic activity of cells was determined as described previously.29 Cells were prepared on cell desks (13.5 mm in diameter; Sumitomo Bakelite, Tokyo, Japan) in 24-well plates. Cells were maintained in serum-free medium in the presence of M-CSF (104 units/mL) for 4 hours, incubated with latex beads (Polysciences, Warrington, PA, USA) for 40 minutes, fixed, treated with 0.2% Triton X-100 in PBS, and stained with rhodamine-conjugated phalloidin (Molecular Probes, Eugene, OR, USA) and 4,4′-diamidino-2-phenylindole (DAPI) (Vector Laboratories, Burlingame, CA, USA). The latex beads appeared as green dots under FITC-filtered light.

Dendritic cell differentiation

Dendritic cell differentiation was induced as reported previously.30 Cells were cultured for 3 days in RPMI 1640 medium (GIBCO/Invitrogen, Karlsruhe, Germany) supplemented with 5% FBS in the presence of granulocyte-macrophage colony-stimulating factor (GM-CSF) (R&D Systems) (10 ng/mL), and further treated with GM-CSF (10 ng/mL) and lipopolysaccharide (Sigma) (1 µg/mL) for 24 hours. Immunostaining for dendritic cells was performed as described below (Immunohistochemical analysis).

Ectopic bone formation and tissue preparation

Ectopic bone formation induced by BMP is generally thought to be a model for normal bone formation through endochondral ossification.31 The 5-mm2 collagen disks were prepared from MedGEL (P19) (MedGEL, Kyoto, Japan). BMP-2 (R&D Systems) (5 µg/disk) was added to the collagen disks. The disks were implanted into the left dorsal muscle pouches of wild-type mice or RANKL-deficient mice. Wild-type mice were given 1 mg/mL of bromodeoxyuridine (BrdU) (Sigma) in drinking water. Two weeks after the implantation, the disks were recovered, decalcified with EDTA, and embedded in paraffin (Sakura Finetek, Tokyo, Japan). Disk sections were processed for TRAP staining and for immunostaining using specific antibodies.

RNA extraction and GeneChip analysis

FACS-sorted cells were lysed with Trizol (Invitrogen, Carlsbad, CA, USA), and total RNA was extracted from the lysate using a PureLink RNA Mini Kit (Applied Biosystems/Ambion, Austin, TX, USA). Biotin-labeled complementary DNA (cDNA) was prepared using a WT-Ovation Pico RNA Amplification System (NuGen Technologies, San Carlos, CA, USA), WT-Ovation Exon Module (NuGen Technologies), and FL-Ovation cDNA Biotin Module V2 (NuGen Technologies). Each cDNA sample was hybridized to an Affymetrix GeneChip® mouse gene 1.0 ST array (Affymetrix, Santa Clara, CA, USA). Probe arrays were scanned using the Affymetrix GeneChip system confocal scanner 3000. Data were analyzed with the Affymetrix GeneChip operating software (GCOS 1.4). To enable a comparison between chips, data were normalized to a global intensity using the robust multi-array average (RMA) procedure.

Immunohistochemical analysis

The paraffin-embedded 4-µm-thick sections were double stained for TRAP and BrdU or F4/80 and BrdU. BrdU incorporated into cells was detected by using a BrdU immunohistochemistry kit (Exalpha Biologicals, Watertown, MA, USA) or 5-Bromo-2′-deoxy-uridine Labeling and Detection Kit II (Roche Applied Science, Mannheim, Germany), and counterstained with hematoxylin. BrdU-positive and BrdU-negative nuclei in TRAP-positive osteoclasts, osteocytes, and F4/80-positive cells were counted in 10 images of 0.03 mm2 (220 × 140 µm) in the ectopic bone surface region and central region. For immunofluorescent staining, tissues were frozen in hexane using a cooling apparatus (PSL-1800; Tokyo Rikakikai, Tokyo, Japan) and embedded in a 5% carboxymethyl cellulose (CMC) gel. Sections 5 µm thick were prepared using Kawamoto's film method (Cryofilm transfer kit; Finetec, Tokyo, Japan).32 The sections and cultured cells were fixed in ice-cold 5% acetic acid in ethanol. The sections and cultured cells were subjected to staining for RANK, c-Fms, F4/80, CD11c, CD86, and Ki67 using specific antibodies. NorthernLights (NL) 557-conjugated anti-sheep immunoglobulin G (IgG) (R&D Systems.), NL 557-conjugated anti-rabbit IgG (R&D Systems.), horseradish peroxidase (HRP)-conjugated anti-FITC (PerkinElmer, Boston, MA, USA), and Histofine Simple Stain Mouse MAX-PO (Rat) (Nichirei Biosciences, Tokyo, Japan) were used as the secondary antibodies. Immunoreactivity was visualized with Tyramide Signal Amplification kits for FITC and for Cy3 (PerkinElmer). Nuclei were detected by DAPI staining (Vector Laboratories). F4/80-positive cells and RANK-positive cells in ectopic bone were counted in 10 images of 0.03 mm2 (220 × 140 µm) in the bone surface region. Images were obtained using a microscope (Axioplan 2 imaging and Axiovert 200; Carl Zeiss, Göttingen, Germany) with AxioVision software.

Flow cytometric analysis

Bone marrow cells and peripheral blood cells were prepared from 6-week-old mice, and layered onto a lympholyte-M (Cedarlane Laboratories, Ontario, Canada) and lympholyte-Mammal (Cedarlane Laboratories) gradient, respectively. After centrifugation, mononuclear cells were collected and stained with the PE-conjugated anti-RANK and biotin-conjugated anti-c-Fms antibodies. Simultaneously, cells were also stained with APC-conjugated antibodies in some experiments. After two washes, they were incubated with Streptavidin FITC conjugate (BD Biosciences). Bone marrow cells were subdivided into three cell populations; RANKhighFmslow, RANKhighFmshigh, and RANKlowFmshigh (see Fig. 1B). We used “high and low” instead of “+ and −” in the naming of cell populations because the positive shifts of the RANKlow and Fmslow populations were still greater than those of IgG isotype controls. In addition, Fmslow and RANKlow cell populations responded to M-CSF and RANKL, respectively. The BrdU incorporated into cells was detected by using a FITC BrdU Flow kit (BD Biosciences). Cells were analyzed or sorted using a Cytomics FC 500 (Beckman Coulter, Fullerton, CA, USA) and FACS Vantage SE (BD Biosciences).

Figure 1.

Identification of QOPs in mouse bone marrow. (A) Tibial sections prepared from 6-week-old mice were stained for RANK (green) and DAPI (nuclei, blue) (middle panel). The lower panel shows an enlargement of the boxed area in the middle panel. Dotted lines indicate bone. Biotinylated goat IgG was used for the control of anti-RANK antibody. The upper panel shows the control for RANK staining. (B) Bone marrow cells obtained from tibiae were analyzed for the expression of RANK and c-Fms using FACS. Percentages of RANKhighFmslow cells, RANKhighFmshigh cells, and RANKlowFmshigh cells were provided in each square fraction. (C) RANKhighFmslow, RANKhighFmshigh, and RANKlowFmshigh cells were isolated from bone marrow and were cultured with M-CSF (104 units/mL) and RANKL (5 nM) in the presence or absence of hydroxyurea (50 µM) in 96-well plates (1 × 104 cells/well). After culturing for 5 days, cells were fixed and stained for TRAP (upper panels). TRAP-positive cells containing more than three nuclei were counted as osteoclasts (middle panel). The total number of nuclei in osteoclasts were counted (lower panel). Results are expressed as the mean ± SD for three cultures. Significantly different from control, ap < 0.05, bp < 0.01. (D) RANKhighFmslow cells (blue histogram) and RANKlowFmshigh cells (red histogram) were analyzed for expression of CD11b, Gr-1, and F4/80 by FACS. Black histograms correspond to IgG isotype controls. Data are representative of three independent experiments.

FYT720 treatment

Eight-week-old mice were injected intraperitoneally with FTY720 (3 mg/kg body weight; Cayman Chemical, Ann Arbor, MI, USA) in sterile water containing 5% dimethylsulfoxide and 30% fatty acid–free bovine serum albumin.26 Control animals received 5% dimethylsulfoxide and 30% fatty acid–free bovine serum albumin in sterile water. Five hours later, peripheral blood cells were collected and analyzed by FACS.

Statistical analysis

Stat View 5.0 software (SAS Institute, Cary, NC, USA) was used for all statistical analyses. Data were evaluated by a one-way ANOVA followed by Fisher's projected least significant difference (PLSD) test. The results of all experiments were expressed as the mean ± SD for three to eight cultures; p < 0.05 was considered statistically significant. Each experiment was repeated at least three times and similar results were obtained.


Characteristics of QOPs in bone marrow

Using immunohistochemical techniques, we identified RANK and c-Fms double-positive cells as QOPs that exist along bone surfaces in RANKL−/− mice.14 However, immunostainable RANK-positive cells were rarely observed in bone marrow away from bone surfaces (Fig. 1A), suggesting the RANK expression by osteoclast precursors in bone marrow to be weak. Then, we tried to isolate osteoclast precursors from bone marrow, using FACS with anti-RANK and c-Fms antibodies (Fig. 1B). Bone marrow cells were subdivided into three populations: RANKhighFmslow, RANKhighFmshigh, and RANKlowFmshigh. When the three populations were cultured with M-CSF and RANKL for 5 days, TRAP-positive osteoclasts were similarly formed (Fig. 1C). We previously reported that hydroxyurea, an inhibitor of DNA replication, inhibited RANKL-induced osteoclastic differentiation from bone marrow macrophages (BMMs) but not from QOPs.14 Hydroxyurea was added together with RANKL and M-CSF to the cultures of the three populations. A similar number of osteoclasts were formed in RANKhighFmslow cultures even in the presence of hydroxyurea. The number of osteoclasts formed in RANKhighFmshigh cultures was significantly lower in the presence of hydroxyurea. Hydroxyurea almost completely inhibited osteoclastogenesis in the RANKlowFmshigh population. Each osteoclast has various numbers of nuclei; we counted the number of nuclei in TRAP-positive osteoclasts (Fig. 1C, lower panel). The number of nuclei in osteoclasts was well correlated with the number of osteoclasts formed. The ability of cells to differentiate into osteoclasts is influenced by the number of cells in culture. RANKL-induced osteoclast formation in RANKlowFmshigh cell cultures at two different cell numbers (1 × 103 cells/well, 1.5 × 105 cells/well) was similarly inhibited by adding hydroxyurea (data not shown). These results suggest that most cells in the RANKhighFmslow population are cell cycle–arrested QOPs, and the RANKhighFmshigh population contains some cells that can differentiate into osteoclasts through cell cycle progression.

The expression of markers for monocytes, macrophages, and granulocytes of osteoclast precursors isolated by FACS have been shown in previous studies.23, 33, 34 We also examined the expression of those markers in the RANKhighFmslow population, in comparison with the RANKlowFmshigh population (Fig. 1D). CD11b (monocyte/macrophage lineage cells), Gr-1 (monocyte/granulocyte lineage cells), and F4/80 (macrophage lineage cells) were highly expressed in cells of the RANKlowFmshigh population. In contrast, these markers were expressed at low levels in the RANKhighFmslow population.

To further analyze expression markers of QOPs, GeneChip analysis was performed on RANKhighFmslow cells and RANKlowFmshigh cells. We confirmed that the expression levels of monocyte-macrophage markers such as Emr1 (F4/80), Itgam (CD11b), and Csf1r (c-Fms) were lower in the RANKhighFmslow than RANKlowFmshigh population (Table 1). In contrast, cells in the RANKhighFmslow population expressed higher levels of osteoclast markers such as Car II (carbonic anhydrase II), Mmp9 (matrix metalloproteinase 9), Acp5 (acid phosphatase 5), and Tfrc (transferrin receptor). These results suggest that RANKhighFmslow cells express few of the phenotypes of monocytes, and their differentiation into osteoclasts occurs at a slightly more advanced stage than that of the RANKlowFmshigh population.

Table 1. Relative Expression for Macrophage Markers and Osteoclast Markers in RANKhighFmslow Cells and RANKlowFmshigh Cells
Gene descriptionGene symbolFold change (normalized) RANKhighFmslow/RANKlowFmshigh
  1. RANKhighFmslow cells and RANKlowFmshigh cells were isolated from bone marrow in mice by FACS. Differential expression levels of macrophage and osteoclast markers were determined by GeneChip analysis. The numbers were calculated by dividing the fold changes of genes in RANKhighFmslow cells by the fold changes of genes in RANKlowFmshigh cells. EGF = epidermal growth factor; RANK = receptor activator of NF-κB.

Macrophage markers
 EGF-like module containing, mucin-like, hormone receptor-like sequence 1Emr1 (F4/80)0.3
 Integrin alpha MItgam (CD11b)0.6
 Colony stimulating factor 1 receptorCsf1r (c-Fms)0.2
Osteoclast markers
 Carbonic anhydrase IICar II4.0
 Matrix metallopeptidase 9Mmp93.7
 Acid phosphatase 5, tartrate-resistantAcp5 (TRAP5b)2.9
 Transferrin receptorTfrc2.6

Using a transmission electron microscope, we examined ultrastructural features of RANKhighFmslow cells in comparison with authentic osteoclasts and F4/80-positive cells isolated from the bone marrow (Supplemental Fig. S1). F4/80-positive cells showed characteristics of immature macrophages, such as many electron-dense granules and free ribosomes. In contrast, RANKhighFmslow cells possessed some fine structures often observed in osteoclasts, such as developed vacuoles and the bleb-like cell membrane structure. However, the numbers of mitochondria and Golgi complexes were much lower in RANKhighFmslow cells than osteoclasts. Thus, RANKhighFmslow cells appeared to be ultrastructurally quiescent and different from macrophages or osteoclasts.

The characteristics of RANKhighFmslow cells were further examined in comparison with RANKlowFmshigh cells (Fig. 2). RANKlowFmshigh cells proliferated in response to M-CSF (Fig. 2A). M-CSF supported the survival of RANKhighFmslow cells, but did not induce cell-cycle progression. Phagocytic activity of RANKhighFmslow cells was weaker than that of RANKlowFmshigh cells (Fig. 2B). When RANKlowFmshigh cells were cultured for 3 days in the presence of GM-CSF, and further treated with GM-CSF and lipopolysaccharide, dendritic cells double-positive for CD11c and CD86 appeared in the cultures (Fig. 2C). In contrast, RANKhighFmslow cells failed to differentiate into dendritic cells under the same conditions. Thus, RANKhighFmslow cells isolated from bone marrow possess characteristics of QOPs as reported previously.14

Figure 2.

Characterization of RANKhighFmslow cells and RANKlowFmshigh cells in bone marrow. RANKhighFmslow cells and RANKlowFmshigh cells were isolated from bone marrow using FACS. (A) Cell growth. RANKhighFmslow cells and RANKlowFmshigh cells were cultured for 3 and 6 days in the presence of M-CSF (104 units/mL). Cell growth was measured by the Alamar Blue assay and expressed as the fold-increase over RANKlowFmshigh cells at day 3. (B) Phagocytosis. RANKhighFmslow cells and RANKlowFmshigh cells were cultured in serum-free medium for 4 hours in the presence of M-CSF (104 units/mL), and then incubated for 40 minutes with latex beads (green). Cells were stained for rhodamine-conjugated phalloidin (red) and DAPI (blue) (upper panels). Arrows indicate bead-positive cells. Cells incorporating more than 30 beads were counted as bead-positive cells (lower panel), and percentages of bead-positive cells were determined. (C) Dendritic cell differentiation. RANKhighFmslow cells and RANKlowFmshigh cells were cultured for 3 days with GM-CSF (10 ng/mL), and further treated for 24 hours with GM-CSF (10 ng/mL) and lipopolysaccharide (1 µg/mL). Cells were stained for CD11c (green), CD86 (red), and DAPI (blue) (left panels). Arrows indicate CD11c and CD86 double-positive cells (yellow cells). Percentages of CD11c and CD86 double-positive (CD11c+CD86+) cells were determined (right panel). Results are expressed as the mean ± SD for three cultures. Significantly different from cultures of RANKlowFmshigh cells, ap < 0.05, bp < 0.01.

Identification and dynamics of circulating QOPs

We next examined whether QOPs circulate in blood. The RANKhighFmslow and RANKlowFmshigh populations were also detected in peripheral blood (Supplemental Fig. S2A). RANKhighFmslow cells isolated from peripheral blood could differentiate into TRAP-positive cells more effectively than RANKlowFmshigh cells in the presence of hydroxyurea (Supplemental Fig. S2B). Expressions of CD11b, Gr-1, and F4/80 in the peripheral blood-derived RANKhighFmslow and RANKlowFmshigh populations were similar to those in the bone marrow-derived RANKhighFmslow and RANKlowFmshigh populations (Supplemental Fig. S2C; Fig. 1D).

To further identify circulating QOPs in vivo, collagen disks containing BMP-2 (BMP disks) were implanted into dorsal muscle pouches in mice. BrdU in drinking water was given to those mice. After 2 weeks, ectopic bone had formed in the BMP disks. TRAP-positive osteoclasts were detected on the surfaces of BMP-induced ectopic bone (Fig. 3A). QOPs failed to strongly express F4/80 (Fig. 1D; Supplemental Fig. S2B; Table 1). Expression of F4/80 in osteoclast precursors was markedly decreased during their differentiation into osteoclasts (Supplemental Fig. S3). Then the distribution of F4/80-positive cells was examined in BMP-induced ectopic bone (Fig. 3A, lower panel). F4/80-positive cells were distributed around bone tissues. Most of the nuclei (86%) in osteoclasts were negative for BrdU (Fig. 3B). About 56% of F4/80-positive cells possessed BrdU-positive nuclei. In sharp contrast, most of the nuclei in osteocytes in ectopic bone were labeled with BrdU (87.5%), indicating that osteoblasts proliferated in the bone induced to form by BMP-2. These results suggest that the osteoclasts detected in ectopic bones were formed from circulating QOPs.

Figure 3.

Incorporation of BrdU into osteoclasts, osteocytes, and F4/80-positive cells in BMP-2–induced ectopic bone. BMP disks were implanted for 2 weeks into mice, which were given BrdU in drinking water (1 mg/mL). BMP-disks were then recovered and processed for tissue observation. (A) Sections of BMP disks were double-stained for TRAP (red) and BrdU (brown) (upper panels). Sections of BMP disks were double-stained for F4/80 (brown) and BrdU (blue) (lower panels). The surface region (sr) and central region (cr) of ectopic bones are indicated in squares in the left panels. The right panels show enlargement of the images of surface and central regions in the left panels. Arrows and asterisks indicate BrdU-negative nuclei in osteoclasts and BrdU-positive nuclei in osteocytes, respectively. Arrowheads indicate BrdU-positive nuclei in F4/80-positive cells. (B) BrdU-positive and -negative nuclei were counted and percentages of BrdU-positive nuclei were calculated. Results are expressed as the mean ± SD for three animals.

Our FACS analysis revealed that QOPs were included in the RANKhigh cell populations (Fig. 1C). Therefore, we then analyzed the behavior of RANK-positive cells and F4/80-positive cells in peripheral blood (Fig. 4A). Both were detected as different cell populations in peripheral blood (Fig. 4B). When BrdU was given to mice for 2 weeks, it labeled more than 80% of nuclei in F4/80-positive cells (Fig. 4C, D; left panel). In contrast, about 25% of the nuclei of RANK-positive cells had incorporated BrdU after 2 weeks. To determine lifespans of RANK-positive cells and F4/80-positive cells, mice were given BrdU for 2 weeks, and further maintained without BrdU for an additional 2 weeks (Fig. 4C, D; right panel). Most of the F4/80-positive cells that had incorporated BrdU disappeared in the peripheral blood within 2 weeks after the removal of BrdU. In contrast, about 10% of RANK-positive cells still possessed BrdU-containing nuclei. The proportion of BrdU-positive cells remaining was much higher among RANK-positive cells than F4/80-positive cells. Half-lives of RANK-positive cells and F4/80-positive cells in peripheral blood were calculated to be about 11 days and 3 days, respectively. These results suggest that RANK-positive cells proliferate slowly and possess long lifespans.

Figure 4.

Incorporation of BrdU into nuclei of F4/80-positive cells and RANK-positive cells in peripheral blood. (A) The experimental protocol. One group of mice was given BrdU in drinking water (1 mg/mL) for 2 weeks. The other group was given BrdU in drinking water for 2 weeks, and further maintained for 2 weeks without BrdU. Mice were euthanized to collect peripheral blood. (B) Peripheral blood mononuclear cells were analyzed for the expression of RANK and F4/80 using FACS. Percentages of F4/80highRANKlow cells, F4/80highRANKhigh cells, and F4/80lowRANKhigh cells were provided in each square fraction. (C) F4/80-positive cells and RANK-positive cells were gated (left panels). The incorporation of BrdU into F4/80-positive cells (upper panels) and RANK-positive cells (lower panels) was analyzed by FACS. Blue and black histograms indicate the results obtained using anti-BrdU antibodies and IgG isotype controls, respectively. Percentages of BrdU-positive cells are provided in each figure. Data are representative of three independent experiments. (D) Percentages of BrdU-positive cells among F4/80-positive cells and RANK-positive cells were determined in the first FACS analysis (left panel). The percentage of residual BrdU-positive cells was calculated by dividing the percentage of BrdU-positive cells in the second FACS analysis by that in the first FACS analysis (right panel). Results are expressed as the mean ± SD for three experiments. Significantly different from F4/80+, ap < 0.01.

We next tried to identify RANK-positive cells in ectopic bone induced to form by BMP-2 in RANKL−/− mice, because QOPs but not osteoclasts are present in RANKL−/− mice. BMP disks and control disks were implanted into RANKL−/− mice for 2 weeks, and subjected to staining for F4/80 and RANK. There was no significant difference in the distribution of F4/80-positive cells between BMP disks and control disks (Fig. 5A, B). On the other hand, the number of RANK-positive cells was much higher in BMP disks than control disks. Most of the RANK-positive cells were positive for c-Fms (Fig. 5C; middle panels). Furthermore, most nuclei of RANK-positive cells (>70%) were negative for Ki67, a marker of proliferating cells (Fig. 5C; right panels, D). When RANKL was injected into those mice, osteoclasts formed in the BMP disks but not the control disks.15 These results suggest that QOPs circulate in the bloodstream and settle in the right place for osteoclastogenesis.

Figure 5.

Identification of QOPs in BMP-disks in RANKL−/− mice. BMP disks and control disks were implanted into RANKL−/− mice, and recovered 2 weeks later. (A) Sections of the BMP disks and control disks were stained for F4/80 (red) or RANK (green). Biotinylated goat IgG and rat IgG were used for the control of anti-RANK and anti-Ki67 antibodies, respectively. Nuclei were detected by DAPI staining (blue). (B) Numbers of F4/80-positive cells (left) and RANK-positive cells (right) were counted. Results are expressed as the mean ± SD for three experiments. Significantly different from control-disks, ap < 0.01. (C) Sections of the BMP-disks were subjected to double staining of RANK (green) and c-Fms (red) (middle), or to double staining of RANK (green) and Ki67 (red) (right). Nuclei were detected by DAPI staining (blue). The arrows in the middle panel indicate mononuclear cells double-positive for RANK and c-Fms (yellow cells). The arrowheads in the right panel indicate RANK-positive and Ki67-negative cells. Biotinylated goat IgG, sheep IgG, and rabbit IgG were used for the control of anti-RANK, anti-Fms, and anti-Ki67 antibodies, respectively. (D) Ki67-positive and Ki67-negative nuclei in RANK-positive cells were counted and percentages of Ki67-positive nuclei were calculated. Results are expressed as the mean ± SD for three BMP disks.

Sphingosine 1-phosphate (S1P) signaling is now recognized to modulate the trafficking of T cells, B cells, and dendritic cells.35, 36 FYT720, an agonist of S1P, increased circulating F4/80 and CX3CR1 double-positive osteoclast precursors and regulated bone homeostasis.26 Finally, we examined the effects of FYT720 on the mobilization of circulating RANK-positive cells (Fig. 6A). Consistent with a previous report,26 treatment of mice with FYT720 for 5 hours led to a rapid decrease in the percentage of CD3 (T-cell marker)-positive cells and B220 (B-cell marker)-positive cells in the blood. In contrast, the percentages of circulating RANK-positive cells as well as F4/80-positive cells were increased by the injection of FTY720 (Fig. 6A). The increase in the percentage of RANK-positive cells in the blood cell population was not due to the decrease in the numbers of T cells and B cells. The number of RANK-positive cells in blood increased from 0.69 × 105 cells/mL to 1.05 × 105 cells/mL in response to FTY720 injection. These results indicated that FYT720 promoted the egress of QOPs from hematopoietic tissues.

Figure 6.

Effect of FYT720 on peripheral QOPs in mice. (A) Eight-week-old mice were injected with FYT720 (3 mg/kg body weight) or vehicle (control). Five hours later, mice were sacrificed. Peripheral blood cells were collected, and analyzed for the expression of CD3e, B220, F4/80, and RANK by using FACS. Data show the representative FACS profiles. Similar profiles of blood cells were obtained in the two additional independent experiments. (B) Schematic diagram of osteoclastogenesis in vivo. Lineage-committed QOPs are generated in hematopoietic organs. Some QOPs enter the bloodstream and home to the correct location for osteoclastogenesis. RANK expression of QOPs is upregulated there. QOPs fuse each other, and differentiate into osteoclasts in response to bone-resorbing stimuli without cell-cycle progression. Osteoblasts may play an essential role in the homing of QOPs to bone.


Characteristics of in vivo osteoclast precursors, QOPs

The concept of circulating osteoclast precursors was established by pioneering studies.16–25 We show here that lineage-committed osteoclast precursors circulate in the bloodstream to find the correct site for osteoclastogenesis (Fig. 6B). The characteristics of osteoclast precursors have been investigated extensively using culture systems.7 Osteoclasts are formed from macrophage lineage cells such as peripheral blood-derived monocytes and BMMs in culture.13, 23 Dendritic cells also can differentiate into osteoclasts under specific conditions in vitro and in vivo.37 In sharp contrast, osteoclasts were formed from QOPs but not from F4/80-positive cells in vivo. Niida and colleagues38 reported that when M-CSF was administered into osteopetrotic (op/op) mice, osteoclasts appeared in bone but F4/80-positive macrophages did not. These findings suggest that osteoclast differentiation is much more strictly regulated in vivo than in vitro. On the other hand, the suppression of phenotypes of macrophages such as F4/80 expression in osteoclast precursors preceded their osteoclastic differentiation (Supplemental Fig. S3). Therefore, it is likely that macrophages such as F4/80-positive cells may differentiate into QOPs in vivo.

It was reported that the suppression of negative regulators for osteoclastogenesis such as interferon regulatory factor-8 (IRF8) and v-maf musculoaponeurotic fibrosarcoma oncogene family, protein B (Mafb) in osteoclast precursors was necessary for their differentiation into osteoclasts.39–41 Our preliminary experiments showed that the expression of IRF8 is suppressed in RANKhighFmslow cells (data not shown). Thus, the suppression of phenotypes of monocytes/macrophages in osteoclast precursors is a crucial event at the commencement of osteoclastogenesis.

c-Fos-deficient (c-Fos−/−) mice exhibit severe osteopetrosis due to osteoclast deficiency, and bone marrow transplantation is the curative treatment.42 The adoptive transfer experiment was performed using QOPs and c-Fos−/− mice. Wild-type bone marrow cells and FACS-sorted RANK-positive cells were transplanted into c-Fos−/− mice that had been myelosuppressed with busulfan.43 Osteoclasts were detected in tibiae of c-Fos−/− mice transplanted with bone marrow cells but not with RANK-positive cells (data not shown). At present, the reason why transplanted RANK-positive cells were not able to differentiate into osteoclasts in vivo is uncertain. It took 8 hours to prepare enough amounts of RANK-positive cells for the in vivo experiment. RANK-positive cells may have received damage during the preparation. It was shown that the hematopoietic stem cell niche within bone marrow protect stem cells from environmental insults, including cytotoxic chemotherapy, and pathogenic immunity.44, 45 Such a defense mechanism may not operate for RANK-positive cells.

Cell fate decision of QOPs

In previous experiments,14 osteoclasts formed from QOPs in RANKL−/− mice and in op/op mice, in response to RANKL and M-CSF, respectively. This suggests that the cell fate of QOPs is decided in the absence of RANKL and M-CSF. However, our recent findings suggest c-Fms-mediated signals to be required for the cell-fate decision in QOPs. Interleukin-34 (IL-34) is a newly discovered cytokine which binds to c-Fms and exerts its action in the same way as M-CSF.46 On exploring the mechanism of osteoclastogenesis in op/op mice, we found IL-34 to be involved in the appearance of QOPs in spleen47 (Y Nakamichi, unpublished data, 2011). IL-34 was strongly expressed in spleen but not bone in op/op mice. When the spleen was removed, no osteoclasts formed in bone in response to M-CSF injections. These results suggest that QOPs develop in hematopoietic tissues in the presence of M-CSF and/or IL-34. RANKL-mediated signaling may be also necessary for the generation of QOPs. TNF-α simulates osteoclastic differentiation of BMMs in culture48, 49 and in RANKL−/− mice.50 These results suggest RANKL-mediated or TNF-α-mediated signals to be necessary for the development of QOPs. Consistent with these findings, the expression of osteoclast markers was slightly but significantly higher in RANKhighFmslow cells than RANKlowFmshigh cells. Taken together, these results suggest that the fate of osteoclast lineage cells is determined in hematopoietic tissues. M-CSF or IL-34 together with RANKL or TNF-α may be involved in the cell fate decision.

The level of RANK in QOPs detected on bone surfaces is much higher than that in QOPs detected in bone marrow or in peripheral blood. This suggests that some cytokines expressed by osteoblasts must be involved in the upregulation of RANK expression in QOPs. M-CSF is reported to enhance RANK expression in osteoclast precursors.51 We also confirmed that M-CSF stimulated RANK expression in BMMs even in the absence of RANKL (see Supplemental Fig. S3C). These results suggest the upregulation of RANK expression induced by osteoblast-derived cytokines such as M-CSF to be essential for the differentiation of hematopoietic precursors into QOPs.

In vivo dynamics of QOPs

BMP-induced ectopic bone creates bone marrow, in which hematopoietic stem cell niches are formed.52 Therefore, hematopoietic stem cells formed in the niche may differentiate into osteoclasts there. However, this idea seems to be unlikely for the following reasons. Cell-cycle progression in hematopoietic stem cells was necessary for their differentiation into QOPs in vivo.14 Cell-cycle–arrested RANK-positive cells were detected in both bloodstream and BMP-induced ectopic bone. The ratio of BrdU-positive nuclei in RANK-positive cells in the BMP-induced bone was similar to that in peripheral blood. We have also shown that TRAP-positive cells and osteoblasts appeared simultaneously in BMP-containing collagen disks within 7 days.15 Neither calcified bone nor bone marrow was formed in the BMP-disks at that time. These results suggest that QOPs are carried to the BMP-induced ectopic bone through the bloodstream.

Our study suggests that the lineage-committed QOPs must enter the bloodstream. This idea raises the question of whether all osteoclasts in vivo are derived from circulating QOPs. In developing bone rudiments, osteoclast precursors enter the bone through blood vessels.53 It was reported that matrix metalloproteinase 9 and vascular endothelial growth factor are essential for the recruitment of osteoclasts into developing long bones.54, 55 Maes and colleagues56 generated tamoxifen-inducible transgenic mice bred to Rosa26R-LacZ reporter mice to follow the fates of stage-selective osteoblast lineage cells, and showed that osteoblast precursors give rise to trabecular osteoblasts and osteocytes inside the developing bone. During this process, osteoclasts appeared just after the invasion of blood vessels. Fischman and Hay57 found that osteoclasts that appeared in regenerated salamander forelimbs were formed from postmitotic precursors circulating in blood. These findings may suggest that osteoclasts develop from QOPs circulating in blood. However, it is also possible that QOPs generated in bone marrow move to sites of bone resorption, where they directly differentiate into osteoclasts.

Increased QOPs in peripheral blood by FYT720 appear to be derived form hematopoietic organs. Characteristics of RANK-positive cells detected in peripheral blood were similar to those of RANK-positive cells in bone marrow. Using interval two-photon microscopy, Ishii and colleagues26 showed that FYT720 increased dynamic migration of osteoclast precursors between the blood and bone. These results suggest that FYT720 promoted the egress of QOPs from hematopoietic tissues.

The cellular mechanism of the egress of QOPs from hematopoietic organs into the bloodstream remains to be elucidated. Circulating QOPs were increased in response to FYT720 treatment in mice, suggesting lipid-mediated chemotaxis to be involved in the egress of QOPs from hematopoietic tissues. QOPs were detected in BMP disks but not in the control disks in RANKL−/− mice, suggesting that they recognize signals from bone tissues to settle down. Our data also indicate that neither RANKL nor M-CSF is responsible for the chemotaxis of QOPs. Omatsu and colleagues58 reported that CXC chemokine ligand (CXCL) 12-abundant reticular (CAR) cells play important roles in maintaining hematopoietic stem cells and B lymphocytes in bone marrow. CAR cells have been shown to play a critical role in the homing of hematopoietic stem cells to bone. CAR cells are specifically located in endochondral bone tissues, especially in the space between blood vessels and bone.59 Therefore, we speculate that reticular cells in bone tissues, like CAR cells, aid the homing of QOPs to bone through interaction between chemokines and chemokine receptors. Further experiments are required to elucidate the dynamics of circulating QOPs.


All the authors state that they have no conflicts of interest.


A Muto, T Mizoguchi, and N Udagawa contributed equally to this work.

This work was supported in part by Grants-in-Aid for Scientific Research [20791365 (TM), 20200019 (TM), 21890274 (AM), and 22390351 (NT)] from the Ministry of Education, Culture, Sports, Science and Technology of Japan and by a grant from Naito Foundation Natural Science (TM). GST-RANKL was kindly provided by Dr. Hisakata Yasuda (Oriental Yeast, Tokyo, Japan).

Authors' roles: AM, TM, SI, IK, YA, AA, and SH performed basic experiments; IK performed transmission electron microscope analysis; YA performed Gene Chip assay; JMP generated RANKL−/− mice; TM, NU, YK, YN, TN, and NT designed the study; and AM, TM, and NT wrote the manuscript.