Smad4 is required to inhibit osteoclastogenesis and maintain bone mass

Bone homeostasis is maintained as a delicate balance between bone-resorption and bone-formation, which are coupled to maintain appropriate bone mass. A critical question is how bone-resorption is terminated to allow bone-formation to occur. Here, we show that TGFβs inhibit osteoclastogenesis and maintain bone-mass through Smad4 activity in osteoclasts. We found that latent-TGFβ1 was activated by osteoclasts to inhibit osteoclastogenesis. Osteoclast-specific Smad4 conditional knockout mice (Smad4-cKO) exhibited significantly reduced bone-mass and elevated osteoclast formation relative to controls. TGFβ1-activation induced expression of Irf8 and Bcl6, both of which encode factors inhibiting osteoclastogenesis, by blocking their negative regulator, Prdm1, in osteoclasts in a Smad4-dependent manner. Reduced bone-mass and accelerated osteoclastogenesis seen in Smad4-cKO were abrogated by Prdm1 deletion. Administration of latent-TGFβ1-Fc to wild-type mice antagonized LPS-induced bone destruction in a model of activated osteoclast-mediated bone destruction. Thus, latent-TGFβ1-Fc could serve as a promising new therapeutic agent in bone diseases marked by excessive resorption.

TGFβ 1 and insulin like growth factor 1 (IGF1) reportedly accumulate in bone matrix and are released following osteoclastic bone resorption, stimulating osteoblastic bone formation 23,24 . TGFβ 1 is initially produced in an inactive form (latent-TGFβ 1), which is then activated extracellularly under highly acidic conditions 25 . Activated TGFβ 1 reportedly promotes bone marrow stromal cell migration to enable bone formation 25 . Nonetheless, TGFβ 1 function in osteoclasts is controversial: various investigators report that TGFβ 1 exerts both stimulatory and inhibitory effects on osteoclast differentiation in vitro, the former of which was reported by majority as TGFβ 1 action on osteoclasts [26][27][28] . Recently, osteoclast-specific ablation of TGFβ receptor 2 in mice was reportedly resulted in no significant impact on osteoclast numbers or activity in vivo 29 . BMP2 reportedly stimulates osteoclastogenesis 30,31 ; however, osteoclast-specific conditional ablation of the gene encoding its receptor BMP receptor type 1A (BMPR1a) increases osteoclast differentiation in vivo 32 .
Here, we show that Smad4 is expressed in osteoclasts and report that osteoclast-specific Smad4 conditional knockout mice (Smad4-cKO: Cathepsin K (Ctsk) Cre/+ /Smad4 flox/flox ) exhibit significantly reduced bone mass due to accelerated osteoclast formation. High TGFβ 1 concentrations inhibited osteoclast differentiation of wild-type cells in vitro, and such inhibition was blocked in Smad4 cKO cells. We also show that TGFβ 1 inhibits Prdm1 expression, which in turn upregulates Irf8 and Bcl6 expression, inhibiting osteoclast differentiation. Reduced bone mass and elevated osteoclastogenesis in Smad4-cKO were abrogated in Smad4/Blimp1 doubly mutant mice. Latent-TGFβ 1 was converted to an active form by osteoclastic activity in cultured cells, and administration of latent-TGFβ 1-Fc to wild-type mice blocked LPS-induced bone destruction. We conclude that following bone resorption, inhibition of osteoclastogenesis by activated TGFβ 1 via Smad4 expressed in osteoclasts is crucial to maintain bone mass.

Results
Osteoclastogenesis is differentially regulated in osteoclast progenitor cells by high concentrations of TGFβ1 or TGFβ3 in vitro. To assess expression of TGFβ factors in bone, we undertook analysis of transcripts encoding these factors in bone tissues of wild-type mice and identified TGFβ 1, 2, 3 and BMP2 mRNAs (Fig. 1a). We also found that osteoclastogenesis, as assessed by expression of osteoclastic genes such as Cathepsin K (Ctsk) and NFATc1 following RANKL treatment of cultured Raw264.7 cells, was enhanced by co-incubation of cells with RANKL plus either TGFβ 1 or β 3 ( Supplementary Fig. 1). Osteoclast differentiation in bone marrow macrophages (BMMs) was also significantly inhibited by SB431542, a TGFβ inhibitor, in vitro ( Supplementary  Fig. 2a-c). Interestingly, in vitro osteoclastogenesis in wild-type BMMs was stimulated at a lower concentration (0.016 ng/ml) of either TGFβ 1 or TGFβ 3, while differentiation was significantly inhibited at higher concentrations (0.4, 2 or 10 ng/ml) of either TGFβ 1 or TGFβ 3 dose-dependently ( Fig. 1b-d, Supplementary Fig. 3 a nd Supplementary Fig. 4). In contrast, osteoclast differentiation from wild-type BMMs was stimulated by high concentrations of either TGFβ 2 (10 ng/ml) or BMP2 (200 ng/ml) (Fig. 1b,c). Osteoclastogenesis, as evidenced by appearance of multi-nuclear TRAP-positive cells, was stimulated by either 40 or 200 ng/ml BMP2 but inhibited by 1,000 ng/ml of BMP2 ( Supplementary Fig. 5). These results suggest that osteoclastogenesis is regulated in a complex manner by TGFβ superfamily members in the bone microenvironment.
Smad4 is required to inhibit osteoclastogenesis and maintain bone mass. As noted, lack of Smad4 results in abrogation of both TGFβ and BMP signaling. We detected Smad4 expression in osteoclasts (Fig. 2a). To assess roles of Smad4 and downstream signaling in regulating osteoclastogenesis and bone mass in vivo, we generated osteoclast-specific Smad4 conditional knockout mice (Smad4 cKO) using Ctsk-Cre mice (Fig. 2b). Based on DEXA analysis, Smad4 cKO mice exhibited significantly reduced bone mass with accelerated osteoclastogenesis as analyzed by TRAP staining and bone morphometric analysis compared with controls in vivo ( Fig. 2c-f). Osteoblastogenesis was normal in Smad4 cKO mice, while osteoclast formation was activated ( Fig. 2d-f). Thus, reduced bone mass seen in Smad4 cKO mice is likely due to elevated osteoclastogenesis in vivo.
TGFβ1 and β3 inhibit osteoclast differentiation via Smad4. We next focused on identifying osteoclast-inhibiting signals mediated by Smad4 in vitro (Fig. 3). Osteoclastogenesis in wild-type BMMs as analyzed by formation of multi-nuclear TRAP-positive cells in vitro, was significantly inhibited in the presence of high concentrations of TGFβ 1 or TGFβ 3, and inhibition was significantly reversed in Smad4 cKO cells (Fig. 3a,b and Supplementary Fig. 6a,b). Expression of the osteoclast differentiation markers Ctsk and NFATc1  Smad4 regulates Bcl6 and Irf8 expression. To define molecular mechanisms underlying TGFβ inhibition of osteoclastogenesis through Smad4, we analyzed expression of potential inhibitory factors following treatment of wild-type osteoclasts with TGFβ s. Candidates included Bcl6 and Irf8, both transcriptional repressors and reported inhibitors of osteoclastogenesis 39,40 . Bcl6 and Irf8 mRNA expression was upregulated following stimulation of wild-type osteoclasts with TGFβ 1 (Fig. 4a). Interestingly, Bcl6 and Irf8 upregulation was significantly blocked in Smad4 cKO cells (Fig. 4b), suggesting that such upregulation is dependent on Smad4. Thus, next we treated Bcl6-deficient BMMs with TGFβ 1 or β 3 and found that their inhibition of osteoclast formation was abrogated relative to wild-type cells (Fig. 4c,d, and Supplementary Fig. 7a,b). Likewise, decreased expression of the osteoclastic genes Ctsk and NFATc1 seen following TGFβ 1 or β 3 treatment in wild-type osteoclasts was significantly rescued in Bcl6-deficient osteoclasts (Fig. 4e,  Supplementary Fig. 7c). Similarly, Irf8-deficient cells were resistant to inhibition of osteoclastogenesis and suppression of osteoclastic gene expression by either TGFβ 1 or β 3 ( Fig. 4f-h, Supplementary Fig. 7d-f).

Blimp1 is a direct target of Smad4 in osteoclasts.
Both Bcl6 and Irf8 expression in osteoclasts is reportedly negatively regulated by Blimp1, a transcriptional repressor encoded by Prdm1 39,40 . Thus, we asked whether  (a) Osteoclast progenitors were isolated from wild-type mice and cultured in the presence or absence of M-CSF (M) and RANKL (R) with or without indicated concentrations of TGFβ 1. Irf8 and Bcl6 expression was then determined by realtime PCR. Data represent mean Bcl6 or Irf8 expression relative to β-actin ± SD (n = 3). (b) Osteoclast progenitors were isolated from control (Smad4 flox/flox or flox/+ ) or Smad4 cKO (cKO) mice and cultured in the presence or absence of RANKL with or without 10 ng/ml of TGFβ 1. Irf8 and Bcl6 expression was determined by realtime PCR. Data represent mean Bcl6 or Irf8 expression relative to β-actin ± SD (n = 3). (c-h) Osteoclast progenitors were isolated from wild-type, Bcl6-deficient (c-e) or Irf8-deficient (f-h) mice and cultured in the presence or absence of M-CSF (M) and RANKL (R) with or without 10 ng/ml TGFβ 1. Osteoclast formation was evaluated by TRAP staining (c,f), by the number of multi-nuclear TRAP-positive cells (d,g) and by Ctsk and NFATc1 expression as analyzed by realtime PCR (e,h). Irf8 and Bcl6 expression was determined in Bcl6 and Irf8-deficient mice, respectively, by realtime PCR (i). Data represent mean Ctsk or NFATc1 or expression relative to β-actin ± SD (n = 3). Bar = 100 μ m. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant. Representative data of at least two independent experiments are shown. elevated Bcl6 and Irf8 expression seen following TGFβ 1 or TGFβ 3 treatment was accompanied by decreased Prdm1 expression. In accordance, Prdm1 mRNA expression was significantly inhibited by either TGFβ 1 or TGFβ 3 treatment of wild-type osteoclasts (Fig. 5a, Supplementary Fig. 8), but such Prdm1 inhibition was abrogated in Smad4 cKO cells (Fig. 5a, Supplementary Fig. 8). To assess whether Prdm1 is a direct target of Smad in osteoclasts, we employed chromatin immune precipitation sequencing (ChIP seq) analysis using anti-Smad2/3 antibodies, and observed that Smad2/3 bound to an upstream region of the Prdm1 gene in osteoclasts under TGFβ 1 stimulation (Fig. 5b). When we generated osteoclast-specific Smad4/Prdm1 double knockout (DcKO: Ctsk Cre/+ /Smad4 f/f Prdm1 f/f ) mice, in which Prdm1 is deleted from Smad4 cKO mice, we found that the significantly decreased bone mass seen in Smad4 cKO mice was reversed and rather increased in DcKO mice (Fig. 5c-e). These observations suggest that Smad4 is required for Prdm1 inhibition in osteoclasts and to maintain bone mass following stimulation with either TGFβ 1 or TGFβ 3.

Latent-TGFβ1 inhibits LPS-induced osteoclast formation and bone destruction. As reported,
TGFβ 1 is converted from a non-active, latent-TGFβ 1 form to an activated form 25 . First, we established primary cultures of wild-type osteoclasts with or without latent-TGFβ 1, and found that osteoclastogenesis in vitro was inhibited when wild-type BMMs were treated with active TGFβ 1 but not by latent-TGFβ 1 (Fig. 6a). Then, we treated wild-type BMMs with supernatants from primary cultures in the presence of M-CSF and RANKL (Fig. 6b). Latent-TGFβ 1 is reportedly activated by osteoclastic bone-resorption 25 . In accordance, we found that osteoclastogenesis was inhibited in secondary cultures treated with supernatants from osteoclasts cultured with latent-TGFβ 1 (Fig. 6b). Based on these results, we concluded that administered latent-TGFβ 1 is converted to an active form by osteoclast to inhibit osteoclast formation. To test this hypothesis, we administered latent-TGFβ 1-Fc or control CD4-Fc protein by injection in vivo in a mouse model of LPS-induced bone destruction, in which LPS was injected on wild-type mouse calvariae. We found that LPS-induced bone-resorption and osteoclast formation as analyzed by micro CT (μ CT), and anti-Ctsk with anti-NFATc1 staining, respectively, were significantly inhibited by latent-TGFβ 1-Fc compared with CD4-Fc administration (Fig. 6c-f). TGFβ 1 signaling is known to promote differentiation of TH17 cells, a type of osteoclastogenic T cells implicated in bone destruction 41,42 . Indeed, in an LPS-induced model of bone destruction, we found that TH17 cell frequency significantly increased in mice treated with LPS together with latent-TGFβ 1-Fc compared with control mice treated with PBS plus latent-TGFβ 1-Fc (Fig. 6g). The fact that bone destruction was inhibited by latent-TGFβ 1-Fc, even under elevated TH17 cell conditions, suggests that latent-TGFβ 1-Fc could antagonize bone destruction in osteoclast-activating conditions.

Discussion
Numerous bone-regulating factors maintain bone homeostasis 1,43 . Among them, factors activating signals via Smad4, including TGFβ and BMP, reportedly support osteoblastic cell migration, proliferation, differentiation and bone formation in vivo and in vitro ( Supplementary Fig. 12a) 19,44 . This study demonstrates that Smad4 mediates osteoprotective signals that are coupled with osteoclastic bone resorption and acts as part of a negative feedback mechanism ( Supplementary Fig. 12b,c). Our findings suggest overall that Smad4 plays a role in both inhibiting bone resorption and activating bone formation ( Supplementary Fig. 12). Here, we show that latent-TGFβ is activated by osteoclasts, which inhibits their activity ( Supplementary Fig. 12b).
The activity of TGFβ superfamily members in osteoclasts reportedly varies [26][27][28][29]32 , and we show that TGFβ 1/β 3 inhibits osteoclastogenesis, while TGFβ 2/BMP2 stimulates it. However, the significant reduction in bone mass and elevated osteoclast formation we report here in Smad4 cKO mice suggests that in this system inhibitory signals via Smad4 are dominant over stimulators. Since Smad4 null mice exhibit embryonic lethality 45 , Smad4 function in osteoclasts and bones has not previously been characterized. The Cre/loxP system employed here did not completely abrogate Smad4 activity in osteoclasts, and some Smad4 function may remain. Nonetheless, it allowed us evaluate Smad4 function in osteoclastogenesis and bone at adult stages. Those signals via TGFβ result from conversion of latent-TGFβ to TGFβ 1, which in turn blocks expression of Prdm1, a repressor of osteoclastogenesis. Loss of the repressor encoded by Prdm1 upregulates Bcl6 and Irf8, both of which repress osteoclast differentiation ( Supplementary Fig. 12c). Although, at present, molecular mechanisms underlying are not clear, we found that Irf8 or Bcl6 expression was significantly inhibited in Bcl6-or Irf8-deficient osteoclasts, respectively (Fig. 4i), suggesting that these factors regulate each other in osteoclasts.
TGFβ and BMP signaling is regulated in a complex manner in osteoblasts 44 . Indeed, TGFβ 1 is reportedly required for osteoblastogenesis 19,44 , while it is also reported to inhibit osteoblastogenesis induced by BMP2 46 . However, there is net decrease in bone mass seen in osteoblast-specific Smad4-deficient mice 19 , suggesting that Smad4 signals in osteoblasts positively regulate bone formation. Thus overall, although why high concentration of BMP2 (1,000 ng/ml) inhibited osteoclast formation was not clear, Smad4 signaling in both osteoclasts and osteoblasts results in increases in bone mass.
Recent advances in developing anti-osteoporosis drugs have resulted in both anti-resorptive agents such as bisphosphonate or anti-RANKL antibodies, and bone-forming drugs, such as teriparatide [47][48][49][50] . Both types have significant therapeutic effects in increasing bone mass and preventing fractures in osteoporosis patients 51 . However, the broad effects of anti-resorptive or bone-forming agents in inhibiting or promoting osteoclast differentiation/function, respectively, can cause adverse side effects such as jaw osteonecrosis, super suppressive bone turnover or osteosarcoma formation 4,52 . As alternatives, investigators are currently seeking novel reagents targeting specific sites where bone formation is required following resorption. Our data strongly suggests that the Scientific RepoRts | 6:35221 | DOI: 10.1038/srep35221 Figure 6. Latent TGFβ1 is converted to an active form by osteoclastic activity. (a) Osteoclast progenitors from wild-type mice were cultured in the presence of M-CSF and RANKL with or without either active-or latent-TGFβ 1 (10 ng/ml each) for primary culture. Quantitative real-time PCR analysis of osteoclastic mRNAs was then undertaken. Data represent mean Ctsk or NFATc1 expression relative to β-actin ± SD (n = 3). (b) Primary culture supernatants were collected from wild-type cells and transferred to secondary cultures of wild-type osteoclast progenitors, which were then treated with M-CSF and RANKL, and Ctsk and NFATc1 expression was analyzed by realtime PCR. Data represent mean Ctsk or NFATc1 expression relative to β-actin ± SD (n = 3). (c-g) LPS (50 mg/kg) was administered subcutaneously onto the skull of living 8-week-old female wild-type mice with or without 16 mg of latent-TGFβ 1. Five days later, osteolysis in calvariae was analyzed by μ CT (c, low magnification; d, high magnification). PBS injection served as a negative control. The number of resorption pits per calvariae was scored. (e). Data represent mean resorption pit number per calvariae ± SD (n = 5). Sections were stained with mouse anti-Ctsk and rabbit anti-NFATc1 antibodies, followed by Alexa488-conjugated goat anti-mouse Ig' antibody, Alexa488-conjugated goat anti-rabbit Ig' antibody and DAPI. Sections were then observed by fluorescence microscopy (f). Spleen cells were stained with anti-CD4 and anti-IL-17 antibodies, and the frequency of TH17 cells (CD4 + IL-17 + cells) was analyzed by flow cytometry (g). Data represent mean TH17 cell frequency ± SD (n = 5). Bar = 100 μ m. *P < 0.05; NS, not significant. Representative data of at least two independent experiments are shown.
Scientific RepoRts | 6:35221 | DOI: 10.1038/srep35221 TGFβ /Smad4 system is specifically activated at such sites. Our observations therefore provide a molecular basis for developing agents that both inhibit bone-resorption and activate bone-formation.

Methods
Mice. Wild-type mice were purchased from Sankyo Labo Service (Tokyo, Japan). Ctsk cre/+ , Smad4 f/f , Prdm1 f/f , Bcl6-deficient and Irf8-deficient mice were prepared as previously described 39,40,53,54 . Animals were maintained under specific pathogen-free conditions in animal facilities certified by the Keio University Institutional Animal Care and Use Committee, and animal protocols were approved by that committee. All animal studies were performed in accordance with the Guidelines of the Keio University animal care committee.
Analysis of skeletal morphology. Ctsk cre/+ Smad4 f/f , Ctsk cre/+ Smad4 f/f Prdm1 f/f and control littermates were necropsied, and their hind limbs were removed, fixed in 70% ethanol, and subjected to DEXA analysis to measure bone mineral density, and analysis of bone histomorphometric parameters. Bones were collected from 8-week-old female mice. . Medium was changed every 2 days. Osteoclastogenesis was evaluated by TRAP staining, and TRAP-positive multi-nuclear cells containing more than three nuclei were scored as osteoclasts.
For some experiments, supernatants from osteoclast culture for five days with or without latent-TGFβ 1-Fc (10 μ g/ml, R & D Systems) were added to secondary cultures, and osteoclastogenesis was evaluated by TRAP staining or expression of osteoclastic genes.
Quantitative PCR analysis. Total RNAs were isolated from bone marrow cultures using TRIzol reagent (Invitrogen Corp.), and cDNA synthesis was performed using oligo(dT) primers and reverse transcriptase (Wako Pure Chemicals Industries). Quantitative PCR was performed using SYBR Premix ExTaq II reagent and a DICE Thermal cycler (Takara Bio Inc.), according to the manufacturer's instructions. β-actin (Actb) expression served as an internal control. Primers used for realtime PCR analysis were as follows.

Statistical analysis.
Results are expressed as the mean ± s.d. Differences between groups were examined for statistical significance using Student t test.