|Home | About | Journals | Submit | Contact Us | Français|
Osteoclasts are bone-resorbing cells essential for skeletal development, homeostasis, and regeneration. They derive from hematopoietic progenitors in the monocyte/macrophage lineage and differentiate in response to RANKL. However, the precise nature of osteoclast progenitors is a longstanding and important question. Using inducible peroxisome proliferator-activated receptor γ (PPARγ)-tTA TRE-GFP (green fluorescent protein) reporter mice, we show that osteoclast progenitors reside specifically in the PPARγ-expressing hematopoietic bone marrow population and identify the quiescent PPARγ+ cells as osteoclast progenitors. Importantly, two PPARγ-tTA TRE-Cre-controlled genetic models provide compelling functional evidence. First, Notch activation in PPARγ+ cells causes high bone mass due to impaired osteoclast precursor proliferation. Second, selective ablation of PPARγ+ cells by diphtheria toxin also causes high bone mass due to decreased osteoclast numbers. Furthermore, PPARγ+ cells respond to both pathological and pharmacological resorption-enhancing stimuli. Mechanistically, PPARγ promotes osteoclast progenitors by activating GATA2 transcription. These findings not only identify the long-sought-after osteoclast progenitors but also establish unprecedented tools for their visualization, isolation, characterization, and genetic manipulation.
Bone is a dynamic tissue that constantly remodels itself by balancing osteoclast-mediated bone resorption and osteoblast-mediated bone formation. Osteoclasts derive from hematopoietic progenitors (5) in the monocyte/macrophage lineage (41, 47); in contrast, osteoblasts are of mesenchymal lineage (38). Physiological osteoclast functions are essential for skeletal development, homeostasis, and regeneration in response to injury. However, pathological increases in osteoclast activities are associated with several diseases, including osteoporosis, arthritis, and bone metastasis of cancers (35).
Osteoclast lineage specification is a multistep process that requires osteoclast progenitor commitment (41, 47), macrophage colony-stimulating factor (M-CSF)-mediated osteoclast precursor proliferation (57), and RANKL (receptor activator of NF-κB ligand)-mediated osteoclast differentiation (8, 29, 56). Although the discovery of RANKL has revolutionized research in osteoclast biology, RANKL mainly acts at later stages of osteoclastogenesis. The cellular identity and the precise nature of the bona fide osteoclast progenitors are underexplored. Previous studies have elegantly characterized the cell surface markers that enrich osteoclast progenitors using flow cytometry (25); however, tools are lacking to label osteoclast progenitors in vivo for visualization, isolation, and lineage tracing, as well as to genetically manipulate osteoclast progenitors for functional characterization.
Peroxisome proliferator-activated receptor γ (PPARγ) is a member of the nuclear receptor family of transcription factors that can be activated by lipophilic ligands, including the diabetic drug rosiglitazone (BRL, or Avandia) (18, 49). Previous studies showed that PPARγ is highly expressed in both monocyte/macrophage precursors and mature osteoclasts (39, 48, 52). Loss of PPARγ function in mouse hematopoietic lineages causes osteoclast defects manifested as osteopetrosis (52). Gain of PPARγ function by pharmacological activation enhances osteoclastogenesis and bone resorption in mice and humans (52, 53, 59). These findings provide important mechanistic understanding of the clinically reported bone loss and higher fracture rates in diabetic patients treated with rosiglitazone. Here, we hypothesize that osteoclast progenitors reside in the PPARγ-expressing hematopoietic bone marrow population and that PPARγ regulation goes beyond osteoclast differentiation by also defining the osteoclast progenitors.
PPARγ-tTA TRE-H2BGFP mice (46), flox-DTA mice (30), and NICD-flox mice (55) have been described previously. PPARγ-tTA TRE-cre mice were bred with flox-DTA mice to generate PTDTA mice. PPARγ-tTA TRE-cre mice were bred with NICD-flox mice to generate PTNICD mice. All experiments were performed using littermate cohorts. All protocols for mouse experiments were approved by the Institutional Animal Care and Use Committee of the University of Texas Southwestern Medical Center.
To evaluate bone volume and architecture by micro-computed tomography (μCT), mouse tibiae were fixed in 70% ethanol and scanned using a Scanco μCT-35 instrument (Scanco Medical) at several resolutions for both overall tibial assessment (14-μm resolution) and structural analysis of trabecular and cortical bone (7-μm resolution). Trabecular bone parameters were calculated using the Scanco software to analyze the bone scans from the trabecular region directly distal to the proximal tibial growth plate. Histomorphometric analyses were conducted using Bioquant Image Analysis software (Bioquant). TRAP (tartrate-resistant acid phosphatase) staining of osteoclasts was performed using a leukocyte acid phosphatase staining kit (Sigma). ALP staining of osteoblasts was performed using an alkaline phosphatase staining kit (Sigma). As a bone resorption marker, urinary C-terminal telopeptide fragments of the type I collagen (CTX-1) was measured with the RatLaps enzyme immunoassay (EIA) kit (Immunodiagnostic Systems) and normalized by urinary creatinine measured with the Infinity Creatinine Reagent (Thermo Scientific). As a bone formation marker, serum osteocalcin was measured with the mouse osteocalcin EIA kit (Biomedical Technologies Inc.).
Osteoclasts were differentiated from mouse bone marrow cells as described previously (52, 53). Briefly, bone marrow cells were purified with a 40-μm cell strainer to remove mesenchymal cells, differentiated with 40 ng/ml of M-CSF (R&D Systems) in α minimal essential medium (α-MEM) containing 10% fetal bovine serum (FBS) for 3 days and then with 40 ng/ml of M-CSF and 100 ng/ml of RANKL (R&D Systems) for 3 days (unless otherwise stated), with or without BRL (1 μM) throughout the time course. Mature osteoclasts were identified as multinucleated (>3 nuclei) TRAP+ cells. Osteoclast differentiation was quantified by the RNA expression of RANKL-induced transcription factors and osteoclast function genes using reverse transcription-quantitative PCR (RT-QPCR) analysis.
Osteoclast precursor proliferation was quantified using a bromodeoxyuridine (BrdU) cell proliferation assay kit (GE Healthcare Life Sciences) (6). Mouse bone marrow cells were treated with M-CSF (40 ng/ml) for 3 days to stimulate osteoclast precursor expansion. On day 4, the cells were M-CSF starved for 6 h to synchronize the cell cycle. The cells were then restimulated with M-CSF for 4 h to induce S phase, during which BrdU was provided in the culture medium and integrated into the DNA of the proliferating cells. Osteoclast precursor proliferation was quantified as BrdU incorporation using the BrdU enzyme-linked immunosorbent assay (ELISA) in the kit.
RNA was reverse transcribed into cDNA using an ABI High Capacity cDNA RT Kit and analyzed using real-time quantitative PCR (SYBR green) in triplicate. All RNA expression was normalized by ribosomal protein L19.
Promoter sequence alignment was performed using Vector NTI Advanced 11 AlignX software (Invitrogen). Chromatin immunoprecipitation (ChIP) assays were performed using fluorescence-activated cell sorter (FACS)-sorted green fluorescent protein-positive (GFP+) and GFP− mouse bone marrow cells as previously described (52). The antibodies used were PPARγ, PU.1 (Santa Cruz), acetyl-histone H3 (Upstate/Millipore), and IgG negative control (BD Biosciences). ChIP output was quantified by real-time PCR in triplicate and normalized by 10% input.
Bone marrow cells were purified with a 40-μm cell strainer to remove mesenchymal cells, sorted by FACS into GFP+ and GFP− populations, and cultured with 40 ng/ml M-CSF overnight. Cells in suspension were removed, and adherent cells were transfected in the presence of 40 ng/ml M-CSF, using FuGene HD (Roche) for 3 days before RANKL-induced osteoclast differentiation. The transfection efficiency was >50%. For GATA2 gain of function, a plasmid encoding full-length GATA2 (Open Biosystems) or vector control was transfected. For GATA2 loss of function, small interfering RNA (siRNA) for GATA2 (siGATA2) or control siRNA (siCtrl) (Santa Cruz) was transfected.
The FACS analyses of bone marrow cells were performed using a BD FACScan flow cytometer and phycoerythrin (PE)-conjugated antibodies (all from BD Pharmingen). The FACS sorting of GFP+ and GFP− bone marrow cell populations was performed using a MoFlo Cell Sorter (Beckman Coulter).
All statistical analyses were performed with Student's t test, and the results are represented as means and standard deviations (SD).
To test this hypothesis, we employed our PPARγ-GFP reporter mice (PPARγ-tTA TRE-H2BGFP) (46). In these mice, a tet transactivator cassette (tTA) was inserted into the endogenous PPARγ locus. When combined with the tTA-responsive H2BGFP transgenic allele (TRE-H2BGFP) (26, 51), the bigenic mice marked PPARγ-expressing cells with GFP (Fig. 1A). We isolated GFP+ cells from the hematopoietic bone marrow population by FACS (http://www4.utsouthwestern.edu/wanlab/publications.htm) and compared their osteoclastogenic potential with that of GFP− cells using an ex vivo osteoclast differentiation assay. Expression of the osteoclast markers c-fos and TRAP (Fig. 1B), as well as Ctsk, Calcr, and CAR2 (not shown), indicated that GFP+ cells differentiated into osteoclasts, while such activity appeared to be lacking in GFP− cells.
To quantify osteoclast colony formation, we sorted GFP+ and GFP− bone marrow cells, plated them at 100 cells/well in 96-well plates, and cultured them with M-CSF until 50% confluence before RANKL stimulation. After 11 days, we stained the cells for TRAP and quantified mature osteoclasts (TRAP+ and >3 nuclei) in each well. The majority of the GFP+ cells in each well had already formed osteoclasts, and the remaining mononuclear cells were TRAP+ preosteoclasts in the process of maturation (Fig. 1C). For GFP+ wells, both macrophage and osteoclast colonies developed at 100% (96/96); in contrast, 55% (53/96) of GFP− wells developed TRAP− macrophage colonies, and only 7% (4/96) formed osteoclasts. The number of osteoclasts per well was also significantly higher for GFP+ than for GFP− wells; as a result, the frequency of osteoclast formation was 140-fold higher for GFP+ cells (Fig. 1D), indicating that >99% of osteoclast progenitors/precursors were PPARγ+. These results show that the osteoclast lineage resides in the GFP+ bone marrow population ex vivo.
Imaging analyses further supported the notion that the osteoclast lineage was derived from GFP+ cells ex vivo and in vivo. First, we performed ex vivo osteoclast differentiation using the unsorted bone marrow cells of PPARγ-GFP mice. The multinucleated osteoclasts developed in culture were all GFP+ (Fig. 1E). Second, we performed immunostaining of the femoral sections for CD115 (M-CSF receptor [M-CSFR]), a marker for the monocyte/macrophage lineage. The CD115+ multinucleated osteoclasts at the bone/marrow junction were GFP+, and notably, the CD115+ monocyte/macrophage precursors in the bone marrow also expressed GFP (Fig. 1F). Third, we performed TRAP staining as an independent method to identify osteoclasts. The TRAP+ osteoclasts also colocalized with GFP+ cells (Fig. 1G). We also observed GFP+ bone marrow cells that were TRAP−, which represent putative osteoclast progenitors and precursors (Fig. 1G). Thus, the osteoclast lineage resides in the GFP+ bone marrow population in situ.
To compare the gene expression profiles of GFP+ and GFP− cells, we performed microarray analysis. Approximately 800 genes were up- or downregulated in the GFP+ population by ≥1.8-fold (P ≤ 0.05) (Fig. 1H). The GFP+ cells displayed increased expression of stem cell/granulocyte-monocyte (GM) progenitor (26 genes) and macrophage/osteoclast (61 genes) markers and decreased expression of lymphocyte (20 genes), mast cell (37 genes), and megakaryocyte/erythrocyte (128 genes) markers (Fig. 1I) (http://www4.utsouthwestern.edu/wanlab/publications.htm). This suggests that (i) the PPARγ+ population is enriched for stem/progenitor cells and (ii) PPARγ expression specifically directs hematopoiesis toward the monocyte/macrophage lineage but away from lymphoid or other myeloid lineages, including megakaryocytes, erythrocytes, and mast cells. Consistently, methylcellulose colony-forming assays showed that GFP+ cells generated more macrophage colonies but fewer granulocyte, erythrocyte, and lymphocyte colonies than GFP− cells (http://www4.utsouthwestern.edu/wanlab/publications.htm). Together, these data indicate that we have identified the osteoclast progenitors in the PPARγ-expressing hematopoietic bone marrow cell population, and we can prospectively visualize, isolate, and characterize the osteoclast progenitors using the PPARγ-GFP reporter mice.
The inducible PPARγ-tTA TRE-H2BGFP reporter mice afforded a tool for osteoclast stem/progenitor cell marking by H2BGFP label retention, a quality that is often indicative of quiescent stem/progenitor cell populations (51). Histone H2B-GFP has been well characterized as a stable protein and a sensitive marker for chromosome dynamics in living cells (26, 51). In the absence of doxycycline (Dox), all PPARγ+ cells are labeled with GFP. When tTA activity is suppressed by Dox, only postmitotic and quiescent stem/progenitor cells retain the label, as H2BGFP is diluted in proliferating cells (26; http://www4.utsouthwestern.edu/wanlab/publications.htm) (Fig. 2A). Thus, we employed a pulse-chase strategy to distinguish slow-cycling (quiescent progenitor) from fast-cycling (proliferating precursor) cells (Fig. 2B). We first pulsed the reporter mice by allowing GFP expression in all PPARγ+ cells (−Dox) and then chased with Dox for 8 weeks to block new H2BGFP expression and thus mark only the label-retaining cells (LRCs). Terminally differentiated osteoclasts are also GFP+ and no longer proliferate. However, osteoclasts undergo constant turnover, with a half-life of 1.3 days in mice (32). Thus, the initial GFP+ osteoclasts were eliminated by apoptosis and replaced by osteoclasts differentiated from GFP− precursors by the end of the 8-week Dox chase (Fig. 2B). Previous studies have shown that 4 weeks to 4 months of Dox chase is sufficient to label progenitors as LRCs (26). In our study, we consistently Dox chased the mice for 8 weeks to ensure both proliferating precursors and mature osteoclasts became GFP−. Therefore, this pulse-chase strategy labels osteoclast progenitors as LRCs in the bone marrow.
To determine whether mature osteoclasts can be lineage traced to PPARγ+ progenitors, we employed both a proliferation-sensitive chaseable marker (H2BGFP) and a proliferation-insensitive indelible marker (LacZ) for PPARγ+ cells (http://www4.utsouthwestern.edu/wanlab/publications.htm). In the PPARγ-tTA TRE-H2BGFP reporter, only quiescent PPARγ+ progenitors remain GFP+ after Dox suppression of tTA activity, whereas in the PPARγ-tTA TRE-Cre ROSA-LacZ reporter, all cells that originate from PPARγ+ progenitors before Dox suppression at postnatal day 2 (P2) remain LacZ+, since the cre-flox recombination event is irreversible (http://www4.utsouthwestern.edu/wanlab/publications.htm). Thus, the combination of these two markers allows lineage tracing of PPARγ+ progenitors. The results showed that mature osteoclasts indeed originated from PPARγ+ progenitors, because they became GFP− but still remained LacZ+ after the Dox chase (Fig. 2C). To quantify the purity of osteoclast progenitors in the hematopoietic LRC population, we FACS sorted LRCs and plated them at a single cell per well in 96-well plates to assess their ability to differentiate into osteoclasts. The results showed that 94% of LRCs (+Dox) formed osteoclasts (90/96) whereas only 1% of GFP− cells (−Dox) formed osteoclasts (1/96) (Fig. 2D), indicating that the LRCs specifically labeled osteoclast progenitors.
We next further characterized the LRC (+Dox), GFP+ (−Dox), and GFP− (−Dox) populations by FACS analysis. First, LRCs represented ~1/3 of the total GFP+ cells (Fig. 2E). Second, stem/progenitor cell markers, including c-Kit (2, 24, 60), Sca-1 (44), CD135 (33), Siglec-F (3), and Notch-1 (6), were further enriched in the LRCs compared with either total GFP+ or GFP− (Fig. 2F) cells, whereas the macrophage marker CD11b was reduced in LRCs (Fig. 2F), indicating an increase in non-lineage-committed progenitors (4). Third, the erythrocyte marker Ter119 was excluded from both total GFP+ cells and LRCs (Fig. 2F). Fourth, >95% of LRCs expressed the leukocyte common antigen CD45, a marker found on all cells of hematopoietic origin except mature erythrocytes and platelets (23) but not on cells of mesenchymal origin (9), demonstrating that the LRCs purified by our method did not contain significant mesenchymal cell types (Fig. 2G). Fifth, 93% of total GFP+ cells and 85% of LRCs expressed CD115 (M-CSFR; c-fms), and 68% of total GFP+ cells and 12% of LRCs expressed CD265 (RANK), a receptor required for osteoclast precursors but not for myeloid progenitors or macrophages (15) (Fig. 2G). These results not only confirmed the microarray analysis (Fig. 1I) showing that the GFP+ population was highly enriched for the stem/progenitor cells of the monocyte/macrophage lineage, but also identified the LRC subpopulation as the osteoclast progenitors.
Notch signaling is a key regulator of osteoblastogenesis (17); however, the cell-autonomous function of Notch in osteoclastogenesis is incompletely understood. We found that Notch1 expression was 15-fold higher in LRCs than in total GFP+ cells (Fig. 2F), indicating that Notch signaling may regulate the quiescence-to-proliferation switch of the osteoclast progenitors. A previous study showed that loss of Notch function by Notch1 to -3 deletion enhances osteoclastogenesis by promoting osteoclast precursor proliferation (6). Nonetheless, the effect of gain of Notch function in the osteoclast lineage is unknown. Thus, if PPARγ+ cells are bona fide osteoclast progenitors, then Notch activation in these cells should impair osteoclastogenesis by restraining the quiescence-to-proliferation switch. To test this hypothesis, we exploited the PPARγ-tTA system, which enables not only osteoclast progenitor marking but also genetic manipulation therein. Specifically, PPARγ-tTA TRE-Cre mice permit rapid translation to in vivo models harboring flox-mediated inducible gene deletion or activation in osteoclast progenitors. To express a constitutively active Notch intracellular domain (NICD) in the osteoclast lineage, we bred PPARγ-tTA TRE-Cre mice with Stopflox/flox-NICD mice (55) to generate PTNICD mice (Fig. 3A).
Skeletal examinations indicated that the PTNICD mice developed high bone mass due to osteoclast defects. First, μCT imaging revealed a significant increase in trabecular bone in the PTNICD mice (Fig. 3B and C). Second, ELISA analyses showed that the bone resorption marker CTX-1 was markedly decreased by 56% (Fig. 3D) while the bone formation marker osteocalcin was unaltered (Fig. 3E). Third, histomorphometry showed that osteoclast surface and number (Oc.S/B.S and Oc.N/B.Ar) were significantly reduced (Fig. 3F and G), while osteoblast surface and number (Ob.S/B.S and Ob.N/B.Ar) were unaltered (http://www4.utsouthwestern.edu/wanlab/publications.htm). As often observed in osteopetrotic mice, PTNICD mice also exhibited extramedullary hematopoiesis in the spleen (Fig. 3H). These data suggested that the increased bone mass resulted mainly from decreased osteoclast numbers and bone resorption.
To assess the stage at which Notch activation blocks osteoclastogenesis and the cell-autonomous nature of the effects, we analyzed the osteoclast progenitors ex vivo. M-CSF-mediated osteoclast precursor proliferation was markedly reduced in the PTNICD cultures (Fig. 3I). Consistently, the bone marrow cells from PTNICD mice exhibited lower expression of RANK, PPARγ1, and c-fms than controls (http://www4.utsouthwestern.edu/wanlab/publications.htm). Moreover, RANKL-mediated and BRL-stimulated osteoclast differentiation was also blunted (Fig. 3J), and induction of osteoclast marker genes was severely decreased (Fig. 3K). NICD expression was significantly increased (Fig. 3L). In contrast, the tibial RANKL/OPG mRNA ratio was unaltered (Fig. 3M). These results all indicate that the impaired bone resorption in PTNICD mice was due to an osteoclast-autonomous defect. The simultaneous reduction in precursor proliferation and osteoclast differentiation suggested that Notch activation in PPARγ+ cells prevented the quiescence-to-proliferation switch of the osteoclast progenitors. This was consistent with the previous loss-of-function study showing that Notch is required to maintain the osteoclast stem cell fate (6). Importantly, these results demonstrated that osteoclast progenitors indeed reside in the PPARγ+ bone marrow population in vivo and can be marked and genetically manipulated by the PPARγ-tTA system.
To assess whether selective partial ablation of PPARγ+ cells by “diphtheria toxin attenuated” (DTA) (30) prevents osteoclastogenesis in vivo, we bred PPARγ-tTA TRE-Cre mice with Stopflox/flox- DTA mice to generate PTDTA mice (Fig. 4A). We found that the PTDTA mice also exhibited high bone mass. μCT revealed higher trabecular bone mass (Fig. 4B and C). CTX-1 was 72% lower (Fig. 4D), while osteocalcin was unaltered (Fig. 4E). Osteoclast surface and numbers were decreased (Fig. 4F and G), while osteoblast surface and numbers were unaltered (http://www4.utsouthwestern.edu/wanlab/publications.htm). Consistently, the PTDTA mice also exhibited extramedullary hematopoiesis in the spleen (Fig. 4H). These data indicated that the increased bone mass in the PTDTA mice was mainly caused by decreased osteoclast numbers and bone resorption.
Because this bone phenotype may be contributed by other PPARγ-expressing cell types, such as adipocytes, we next investigated the cell-autonomous nature of the resorption defects by examining the osteoclastogenic potential of the bone marrow from PTDTA mice ex vivo. While many osteoclasts developed in control cultures, few formed in PTDTA cultures (Fig. 4I and J). This was due to cell ablation and, consequently, decreased BrdU incorporation (Fig. 4K). Thus, ablation of PPARγ+ cells severely blunted osteoclastogenesis and bone resorption. Together, the PTNICD and PTDTA genetic models provide compelling in vivo evidence that the osteoclast lineage resides in the PPARγ+ bone marrow population under physiological conditions.
Estrogen deficiency, from menopause or ovariectomy (OVX), is an important cause of osteoporosis and debilitating fractures. Current notions indicate that estrogen deficiency enhances osteoclast survival (27, 34), but its specific effects on osteoclast progenitors remain unknown. To track the response of the osteoclast lineage to estrogen loss, we performed sham operations or ovariectomies on PPARγ-GFP reporter mice in the setting of placebo or Dox-induced reporter suppression, coupled with BrdU labeling 24 h before FACS analysis (Fig. 5A). We found that OVX significantly increased the GFP+ bone marrow population by 3.8-fold (Fig. 5B). This increase appeared secondary to enhanced cell proliferation, as both the percentage of BrdU+ cells in the GFP+ population and the percentage of BrdU+/GFP+ cells in the entire bone marrow population were elevated by 1.8- and 6.4-fold, respectively (Fig. 5C and D).
Next, we examined the effects of OVX on osteoclast progenitors (LRCs; +Dox) (Fig. 5E) and osteoclast precursors (total GFP+; −Dox) (Fig. 5F). We observed that the percentage of LRCs in total GFP+ cells (LRC/GFP+) was 15% in OVX mice compared to 66% in sham controls (66%), resulting in a 4.2-fold reduction (Fig. 5B). The percentage of stem cells (Notch-1+) was decreased, while the percentage of progenitor cells (Sca1+ or c-Kit+) was increased (Fig. 5E and F). These results suggested that OVX triggered the quiescence-to-proliferation switch of the osteoclast progenitors. Furthermore, Siglec-F is predominantly expressed in immature myelomonocytic precursors, and its expression is reduced upon macrophage/osteoclast differentiation (3). We found that the percentage of Siglec-F+ cells was increased in the LRCs but decreased in the GFP+ population (Fig. 5E and F), suggesting that OVX triggered not only the proliferation of the LRC progenitors, but also the differentiation of the GFP+ precursors. This notion was further illustrated by the elevated percentage of monocyte/macrophage lineage-committed cells (CD11b+) (Fig. 5E and F). Moreover, osteoclast differentiation assays showed that when equal numbers of GFP+ cells were seeded, more osteoclasts formed for the OVX mice than for the sham-treated controls (Fig. 5G), indicating that OVX increased both the number of GFP+ cells in the bone marrow and their differentiation potential. Consequently, OVX led to increased resorption (Fig. 5H). Together, these results indicate that OVX activated both the quiescence-to-proliferation switch in osteoclast progenitors and the proliferation-to-differentiation switch in osteoclast precursors (Fig. 5I), revealing previously unrecognized effects of estrogen deficiency on early osteoclast lineage specification.
The tracking system also allowed us to examine the response of the osteoclast lineage to drugs. As a model, we chose BRL, a PPARγ agonist and a diabetic drug that stimulates osteoclast differentiation and bone resorption (52, 53, 59). In ex vivo cultures, BRL attenuated M-CSF-mediated osteoclast precursor proliferation (Fig. 6A) yet exacerbated RANKL induction of osteoclast markers (c-fos and TRAP) (Fig. 6B), suggesting that PPARγ activation triggers a proliferation-to-differentiation switch toward osteoclasts. Interestingly, both BRL and RANKL suppressed the mature macrophage marker (MCP-1) (Fig. 6C), indicating that BRL promotes an osteoclast fate in part by shifting the progenitors away from a terminal macrophage fate.
To determine the effects of PPARγ activation on the osteoclast lineage in vivo, we next administered BRL or vehicle control to the reporter mice. By 2 weeks, BRL increased the GFP+ population within the marrow by 2.9-fold (Fig. 6D) yet reduced the percentage of stem/progenitor cells (Notch1+, Sca-1+, c-Kit+, or CD135+) in the GFP+ population (Fig. 6E). We did not observe changes in these markers in the GFP− (PPARγ−) population, indicating that the BRL effects were PPARγ dependent (Fig. 6E). Consistently, BRL increased the macrophage lineage-committed cells (CD11b+) (Fig. 6F) (45). Intriguingly, BRL significantly diminished the percentage of mature macrophages in the GFP+ population, as both CD14 (a lipopolysaccharide [LPS] receptor) (21) and MCP-1 (58) were downregulated (Fig. 6G). In contrast, BRL increased the osteoclast surface (Fig. 6H) and the bone resorption marker CTX-1 (Fig. 6I), leading to a decreased bone volume/tissue volume (BV/TV) ratio (Fig. 6H) but unaltered bone mineral density (BMD) (not shown). These in vivo data were consistent with the ex vivo results (Fig. 6B and C), indicating that ligand activation of PPARγ triggered the osteoclast progenitors to undergo differentiation, but toward osteoclasts and away from mature macrophages (Fig. 6J). Thus, PPARγ+ cells are osteoclast progenitors, yet PPARγ is also a molecular switch that translates an increased local concentration of PPARγ agonists into enhanced osteoclast differentiation. Importantly, both OVX and BRL, representing pathological and pharmacological resorption-enhancing stimuli, triggered the PPARγ+ cells to proliferate and differentiate, further supporting the notion that osteoclast progenitors reside in the PPARγ+ bone marrow population in vivo.
The GATA family of zinc finger transcription factors is an important regulator of hematopoiesis. GATA2 is required to generate osteoclast progenitors (50, 54), while GATA1 is dispensable for osteoclastogenesis but essential for erythropoiesis and megakaryocyte maturation (20, 37, 43). Therefore, the GATA2/GATA1 ratio in hematopoietic progenitors controls lineage divergence between osteoclasts and erythrocytes/megakaryocytes. In our microarray analysis, we found that this key GATA2/GATA1 ratio was 12.9-fold higher in the PPARγ+ (GFP+) cells than in the PPARγ− (GFP−) cells (Fig. 7A), owing to elevated GATA2 expression and diminished GATA1 expression (Fig. 7B). Since PPARγ is also critical for osteoclastogenesis (52), it may promote osteoclast progenitor commitment by activating GATA2 transcription.
To test this hypothesis, we examined the GATA2 promoter and identified three highly conserved PPAR response elements (PPREs) (Fig. 7C). To determine whether PPARγ directly binds to the mouse GATA2 promoter and induces its transcription, we performed a ChIP assay with antibodies for PPARγ or acetylated histone H3, a chromatin marker for activated transcription. In GFP+ cells, but not GFP− control cells, PPARγ bound to all three PPREs in the mGATA2 promoter, accompanied by elevated levels of acetylated histone H3 (Fig. 7D); in contrast, PPARγ did not bind to the GATA1 promoter (not shown), suggesting that PPARγ inhibits GATA1 expression via an indirect mechanism. PU.1 binding was also detected in these GATA2 regions, suggesting that PPARγ colocalization with PU.1 in the GFP+ cells specified GATA2 expression and osteoclast progenitors (Fig. 7D). We next assessed the functional requirement for GATA2 by both gain- and loss-of-function analyses. Ectopic GATA2 expression in PPARγ− cells to a level comparable to that in PPARγ+ cells partially rescued the osteoclast differentiation blockade (Fig. 7E). Conversely, GATA2 knockdown severely diminished both RANKL-mediated and BRL-stimulated osteoclast differentiation in the PPARγ+ cells (Fig. 7F). Together, these results indicate that PPARγ promotes osteoclast progenitor commitment, at least in part, by directly binding to the GATA2 promoter and activating its transcription (Fig. 7G).
The cellular identity and precise nature of osteoclast progenitors are longstanding and important biological questions. Based on our cellular, molecular, genetic, pathological, and pharmacological evidence, in vivo and ex vivo, we conclude that the osteoclast lineage resides in the PPARγ-expressing hematopoietic bone marrow cell population, and we have identified the quiescent PPARγ+ bone marrow cells as the osteoclast progenitors. Importantly, we have established PPARγ-tTA TRE-H2BGFP reporter mice as an unprecedented tool to visualize, isolate, quantify, and trace the lineage of osteoclast progenitors. As a complement, we have also established PPARγ-tTA TRE-Cre mice as a genetic tool to interrogate the function and regulation of osteoclast progenitors in vivo by inducing flox-mediated gene deletion or activation. Using these tools, we have uncovered previously unrecognized effects of ovariectomy and rosiglitazone, two resorption-enhancing stimuli, on the early osteoclast lineage. Mechanistically, we have identified GATA2 as a novel yet critical PPARγ target gene in osteoclast progenitors. Therefore, both conceptually and technically, this study opens an exciting new path to the fundamental understanding of both osteoclast lineage specification and PPARγ function.
In the PPARγ-tTA TRE-H2BGFP reporter mice, GFP+ cells also label adipocyte progenitors and mature adipocytes (46). Intriguingly, several reports show that mammalian cells of the adipocyte lineage and the macrophage lineage share numerous functional and antigenic properties. Gene expression profiling revealed that preadipocytes share a surprisingly closer signature with macrophages than with adipocytes, and preadipocytes can be effectively converted to macrophages in a macrophage environment (10). This appears to be an evolutionarily conserved phenomenon, because in invertebrates, such as Drosophila, hemocytes (blood cells) and fat bodies also share the expression of fate-determining genes (14). Our findings that PPARγ+ cells label both adipocyte progenitors and macrophage/osteoclast progenitors provided mechanistic evidence for convergence and/or plasticity in the adipocyte and macrophage lineage specification. Together with the dual roles of PPARγ ligand in stimulating both adipogenesis and osteoclastogenesis, our findings illuminate a potential molecular basis for the close correlation between insulin-sensitizing effects and bone loss effects, as well as the emerging connections between fat and bone.
Stem/progenitor cells are defined as multipotent; hence, in order to target osteoclast progenitors experimentally, it is impossible to completely rule out other differentiation outcomes, and specificity is only relative. For example, Tie2 labels not only osteoclast progenitors, but also all other hematopoietic progenitors, as well as endothelial cells (13, 52), whereas PPARγ labels osteoclast and adipocyte progenitors but not other hematopoietic lineages (Fig. 1 and and2).2). The advantages of the PPARγ-tTA-based mouse models include the following: (i) they distinguish macrophages/osteoclasts from other hematopoietic lineages; (ii) they target the entire osteoclast lineage, including osteoclast progenitors and mature osteoclasts; and (iii) they permit temporal control of inducible cell labeling and genetic manipulations in the osteoclast lineage. Our results show that osteoclast progenitors are 140-fold enriched in the PPARγ+ bone marrow cell population, and thus, only <1% of osteoclast progenitors may be derived from PPARγlow or PPARγ− cells. Moreover, previous studies have documented that Notch activation in lymphoid progenitors causes T-cell lymphoblastic leukemia in humans and mice (16, 36). In our study, PTNICD mice did not develop lymphoma, which further supports the notion that PPARγ specifically directs hematopoiesis toward the monocyte/macrophage lineage and that the PPARγ-expressing bone marrow population does not contain lymphoid progenitors. Since PPARγ also labels adipocyte progenitors (46), it is possible that Notch constitutive activation in PTNICD mice may also affect other PPARγ+ cells, such as adipocytes, in addition to osteoclast progenitors.
Several drivers targeting macrophage precursors or mature osteoclasts have been elegantly described; nonetheless, because they do not target osteoclast progenitors, they are not suitable for in vivo study of early osteoclast lineage specification. For example, CD11b or lysozyme drivers are useful to target macrophage precursors because they are upregulated only upon macrophage differentiation (12, 19, 25). In addition, Ctsk or TRAP drivers are useful to target preosteoclasts and mature osteoclasts because they are upregulated only upon osteoclast differentiation (11, 34). Therefore, the PPARγ-tTA-based models represent a novel osteoclast progenitor-targeting strategy that is complementary to other existing models for the comprehensive investigation of osteoclast lineage specification and differentiation.
Indeed, the PTNICD and PTDTA genetic models provide compelling in vivo evidence that the osteoclast lineage resides in the PPARγ+ bone marrow population under physiological conditions; in particular, the PTNICD model supports the notion that PPARγ+ cells represent osteoclast progenitors. In the PTDTA model, the DTA was an “attenuated” version of diphtheria toxin, thus explaining the relatively mild bone phenotype, which was supported by the survival of the PTDTA mice in contrast to the embryonic lethality in the global PPARγ knockout (KO) mice (7, 28, 40). Furthermore, both OVX and BRL, representing pathological and pharmacological resorption-enhancing stimuli, triggered the PPARγ+ cells to proliferate and differentiate, further supporting the notion that osteoclast progenitors reside in the PPARγ+ bone marrow population in vivo. The rapid increase (1 to 2 weeks) in the percentage of GFP+ bone marrow cells in response to BRL treatment in vivo indicates that this effect was not likely secondary to any BRL alteration of adipocytes followed by changes in hematopoiesis, which takes at least 4 weeks (1).
The Ets family transcription factor PU.1 is essential for the development of both myeloid and B-lymphoid cells (42). This suggests that additional transcription factors are required to function in combination with PU.1 and confer lineage specificity (22). Since osteoclasts are of myeloid lineage, PU.1 is also essential for the generation of osteoclast progenitors. Indeed, PU.1 deletion in mice precludes osteoclast development, leading to arrested bone resorption and osteopetrosis (47). A recent study revealed that, in macrophages, PPARγ colocalizes with PU.1 in areas of open chromatin and histone acetylation near a distinct set of hematopoietic genes (31). Our results suggest that in the PPARγ+ cells, PPARγ cooperates with PU.1 to activate the transcription of a subset of genes, including GATA2, thereby directing macrophage/osteoclast lineage commitment. In contrast, in the PPARγ− cells, the absence of PPARγ prevents GATA2 transcription and alters the subset of genes regulated by PU.1, thereby directing B-lymphoid lineage commitment. Therefore, our identification of PPARγ+ bone marrow cells as osteoclast progenitors provides in vivo evidence for the notion that the collaborative interaction between PPARγ and PU.1 on a subset of promoters is essential to activate the transcriptional program required for macrophage/osteoclast lineage commitment. Furthermore, our results suggest that the expression of PPARγ, rather than the ligand activation of PPARγ, promotes osteoclast progenitor specification by enhancing GATA2 expression, which is downregulated during the quiescence-to-proliferation switch and thus is absent in osteoclast precursors (53a) (Fig. 7G). Together, our current and previous studies reveal that PPARγ plays dual roles in osteoclastogenesis that involve multiple mechanisms and target genes (Fig. 7G): PPARγ expression promotes osteoclast progenitors by inducing GATA2, and PPARγ ligand activation stimulates osteoclast differentiation by inducing c-fos.
We thank J. Zerwekh for assistance with bone histomorphometry, L. Smith for assistance with μCT, and D. Mangelsdorf and S. Kliewer for helpful discussion.
This work was supported by the University of Texas Southwestern Medical Center Endowed Scholar Startup Fund (Y.W.), a BD Biosciences Research Grant Award (Y.W.), CPRIT (RP100841 [Y.W.]), the March of Dimes (5-FY10-1 [Y.W.]), The Welch Foundation (I-1751 [Y.W.]), NIH (R01 DK089113 [Y.W.] and R01 DK066556, R01 DK064261, and R01 DK088220 [J.M.G.]), and a postdoctoral fellowship (W.T.) and a predoctoral fellowship (D.Z.) from the American Heart Association South Central Affiliate. Y.W. is a Virginia Murchison Linthicum Scholar in Medical Research.
J.M.G. is a founder of Reata Pharmaceuticals. We declare that we have no financial conflict of interest.
Published ahead of print on 26 September 2011.