Smad4 is required for maintaining normal murine postnatal bone homeostasis

Bone remodeling is central to maintaining the integrity of the skeletal system, wherein the developed bone is constantly renewed by the balanced action of osteoblastic bone formation and osteoclastic bone resorption. The cellular growth and differentiation of osteoblasts and osteoclasts are precisely coordinated to maintain the bone mass homeostasis. Multiple signaling pathways, such as insulin-like growth factor, parathyroid hormone and transforming growth factor β (TGFβ) have been shown to couple bone formation to bone resorption (Harada and Rodan, 2003). However, much less is known about the intracellular control of osteoblastic function.

TGFβ is an important physiological regulator of osteoblast differentiation and function as coupling factors during bone remodeling (Martin and Sims, 2005). TGFβ is stored in large amounts in the bone matrix and subsequently activated by the acidic microenvironment created by bone-resorbing osteoclasts (Oreffo et al., 1989; Oursler, 1994). Once activated, TGFβ stimulates the proliferation of osteoblast precursor cells as well as osteoblast matrix production and induces bone formation in vivo (Noda and Camilliere, 1989; Yamada et al., 2000). Increasing evidence suggests that allelic variations at the TGFB1 gene contribute to the development of osteoporosis (Yamada et al., 1998; Yamada et al., 2000; Keen et al., 2001). It has also been reported that bone loss in old male mice results from diminished activity and availability of TGFβ (Gazit et al., 1998). Consistently, decreased bone mass and bone elasticity have been shown in mice lacking the Tgfb1 gene (Geiser et al., 1998). Osteoblastic overexpression of TGFβ2 results in a high-turnover osteoporosis owing to the increase in both bone resorption and bone formation (Erlebacher and Derynck, 1996), whereas transgenic mice expressing a truncated dominant negative TGFβ receptor II in osteoblasts show an age-dependent increase in trabecular bone owing to imbalanced bone formation and bone resorption (Filvaroff et al., 1999). These results suggest that TGFβ is a key mediator of the coupling of osteoblast differentiation to osteoclastic bone resorption. However, the molecular mechanism underlying the function of TGFβ during bone remodeling is still largely unknown.

Bone morphogenetic proteins (BMPs) were originally identified by their ability to induce ectopic bone formation (Wozney et al., 1988). BMPs are involved in embryonic and postnatal osteogenesis by stimulating the commitment of mesenchymal cells into osteoblastic lineage and promoting osteoblast differentiation (Canalis et al., 2003). However, the mechanism of BMP signaling on bone formation and postnatal bone metabolism remains unclear owing to the embryonic lethality resulting from mutations of Bmp2 and Bmp4, and their receptors (Mishina et al., 1995; Winnier et al., 1995). Recent studies have begun to reveal the necessary roles of BMP receptors in bone remodeling (Mishina et al., 2004). Postnatal osteoblast specific disruption of Bmpr1a results in low bone mass in relatively young mutants and an opposite phenotype in aged mice because of impaired bone formation and bone resorption (Mishina et al., 2004). The transgenic mice expressing a truncated dominant-negative Bmpr1b in osteoblasts exhibit osteopenia owing to impairment of bone growth and reduction of bone mineral density (Zhao et al., 2002). Transgenic mice overexpressing BMP antagonists, such as noggin and sclerostin, all exhibit low bone mass, which is due to decreased BMP signaling (Wu et al., 2003). It is noteworthy that BMP2 has been linked with osteoporosis in humans (Styrkarsdottir et al., 2003).

Smads are primary cytoplasmic signal transducers of TGFβ and BMPs. TGFβ activates Smad2 and Smad3, whereas BMP signaling is conveyed by Smad1, Smad5 and Smad8. Smad4 is the common mediator Smad shared by signaling pathways for BMPs and TGFβ/activin (Derynck and Zhang, 2003). It has been shown that the loss of Smad3 results in a lower rate of bone formation and osteopenia in addition to defects in chondrocyte differentiation (Borton et al., 2001; Yang et al., 2001). By contrast, recent studies revealed that a reduction in TGFβ signaling via Smad3 deletion enhanced the mineral concentration of the bone matrix and the bone mass (Balooch et al., 2005). In addition, Smad3 interacts with class IIa histone deacetylases to mediate TGFβ signaling, which inhibits osteoblast differentiation (Kang et al., 2005). As a common mediator Smad of TGFβ signaling, Smad4 is expressed in osteoblasts and mediates effects of both BMP and TGFβ signals in cultured human osteoblasts (Lai and Cheng, 2002). Several in vitro studies suggest that Smad4 plays an important role in osteoblasts by interacting with other molecules, such as Runx2 and Fos (Lee et al., 2000; Zhang et al., 2000; Lai and Cheng, 2002). However, it is difficult to access the in vivo function of Smad4 during bone remodeling owing to the early embryonic lethality of the Smad4-knockout mice (Sirard et al., 1998; Yang et al., 1998). To understand the role of Smad4-mediated TGFβ signals in postnatal bone formation and metabolism, we disrupted the Smad4 gene specifically in differentiated osteoblasts using the Cre-loxP system. Our results suggest that Smad4-mediated TGFβ signaling is important for regulating postnatal bone homeostasis.

To investigate the function of Smad4 in osteoblasts, we generated a transgenic mouse strain (OC-Cre) that expressed the Cre recombinase under the control of the mouse osteocalcin (OC) promoter (Dacquin et al., 2002). Osteoblast-specific Cre activity was examined using OC-Cre;ROSA26R double transgenic mice. Although the Cre activity revealed by LacZ staining was barely detected in mandible bones of 1-day-old double transgenic mice (supplementary material Fig. S1A,B), it was observed in osteoblasts on trabecular bone (Fig. 1A), cortical bone (Fig. 1B), mandible bone (supplementary material Fig. S1C,D) and calvarial bone (data not shown) surfaces in 10-day-old and 6-week-old double transgenic mice. Most osteocytes derived from osteoblasts were stained positively (Fig. 1C). To examine the temporal expression of the Cre transgene under control of the mouse osteocalcin promoter, primary osteoblast cultures derived from bone marrow cells of double heterozygous and control mice were established and assayed at multiple time points for LacZ staining. Positive staining was observed in osteoblasts when they started to mineralize (Fig. 1D-I).

Inactivation of Smad4 in mature osteoblasts was achieved by breeding a mouse strain containing the Smad4 conditional alleles (Smad4^Co/Co) (Yang et al., 2002) with the OC-Cre transgenic mice. Cre-mediated excision of exon 8 in different tissues isolated from a Smad4^Co/Co;OC-Cre mouse was detected by Southern blot. The results confirmed that Cre-mediated recombination occurred exclusively in calvaria, femurs and vertebrae, all of which are bone tissues containing osteoblasts (Fig. 1J). To follow the kinetics of Smad4 deletion during the differentiation of the cultured primary osteoblasts, genomic PCR was used to detect the Cre-mediated excision. The results showed that the deletion was barely detectable in the early days of culture and gradually increased in later stages when mineralizing osteoblasts appeared (see supplementary material Fig. S1E). To examine when deletion of Smad4 occurs in vivo, we performed a Southern blot on DNA extracted from calvarial cells at different ages. Cre-mediated DNA excision was easily detected in 2-week-old mice (Fig. 1K). All the data demonstrated that the OC-Cre transgenic mice could be used to achieve Cre-mediated recombination specifically in differentiated osteoblasts.

To determine the effect of Smad4 deficiency on TGFβ and BMP signaling, reporter assays were performed in primary osteoblasts derived from calvaria of Smad4^Co/Co mice. The Smad4^Co/Co osteoblasts were transfected with a 12×SBE-OC-Luc reporter construct with or without the Cre gene under the control of the CMV promoter. Addition of BMP2 stimulated luciferase activity of this reporter in Smad4^Co/Co osteoblasts without the CMV-Cre. By contrast, co-transfection with the CMV-Cre in osteoblasts significantly inhibited the reporter activity, suggesting that BMP signaling is dramatically reduced in Smad4^Co/Co osteoblasts after Cre-mediated recombination (Fig. 1L). TGFβ1-stimulated luciferase activity of p3TP-Lux was also inhibited by the Cre-mediated deletion of the Smad4 gene in osteoblasts (Fig. 1M). These results indicated that ablation of Smad4 in osteoblasts diminished the responsiveness of osteoblasts to both BMP-2 and TGFβ1.

The Smad4^Co/Co;OC-Cre mice were born normally and had similar body size and body weight to littermate controls (data not shown). They exhibited smaller body size as early as 3 weeks and more pronounced dwarfism at 6 weeks of age (Fig. 2A). The Smad4^Co/Co;OC-Cre mice were 50-70% smaller than wild-type controls at 6 weeks, and remained significantly smaller than their littermates throughout their lifetime. A quantitative measurement of femoral length indicated that the femurs of Smad4 homozygous mutant mice were significantly shorter than wild-type controls at 3 (0.918±0.015 cm in male mutant vs 1.097±0.031 cm in controls, n=6, P<0.01), 6 (1.079±0.084 cm in male mutant vs 1.545±0.044 cm in controls, n=6, P<0.01) and 12 weeks (1.217±0.074 cm in male mutant vs 1.616±0.050 cm in controls, n=6, P<0.01). X-ray analysis of 6-week-old knockout mice showed shorter and more lucent femurs and tibias and suggested severe osteopenia (Fig. 2B). A dramatic decrease in bone mineral density (BMD) was evident at 6 and 12 weeks of age in both male and female Smad4^Co/Co;OC-Cre mice compared with the littermate controls (Fig. 2C).

To characterize the skeletal abnormalities in the Smad4^Co/Co;OC-Cre mice in more detail, the proximal tibias and calvarial bones at different developmental stages were sectioned for histological analysis. At day 16, slightly decreased levels of trabecular bone were observed in knockout mice (data not shown). Histological sections of 7-week-old mutant tibias revealed that the number of trabecular bones was reduced significantly compared with the littermate controls (Fig. 2D,E). There was a marked decline in the amount of cortical bone in knockout mice, as assessed by cortical width of femoral midshaft at 8 weeks (103.9±15.5 μm in mutants vs 170.5±34.4 μm in controls, n=9, P<0.01). Calvarial thickness was decreased significantly in mutant mice compared with controls at this stage (57.1±10.0 μm in Smad4 mutants vs 85.5±3.8 μm in controls, n=6, P<0.01). A loose connection of interparietal bones in the sagittal and coronal sutures was observed in Smad4 mutants (Fig. 2F,G and data not shown). The growth plate of Smad4 mutant mice was shorter than control littermates (Fig. 2E and see supplementary material Fig. S1), suggesting that endochondral ossification could be affected in the absence of Smad4 in osteoblasts. Taken together, these observations reveal that targeted disruption of Smad4 in osteoblasts causes osteopenia in young mutant mice.

To investigate the cellular basis underlying the bone mass loss in Smad4 mutants, we performed static and dynamic bone histomorphometric analysis in sections of undecalcified distal femurs and tibias at 4 and 8 weeks. Von Kossa staining revealed significantly reduced bone volume in 7-week-old mutant mice (Fig. 3A,B). Calcein labeling analysis, a histomorphometric measurement of osteoblast activity in vivo, confirmed a remarkably decreased bone formation rate (BFR) associated with Smad4 deficiency in 4- and 8-week-old animals (Fig. 3C,D and data not shown). Goldner trichrome staining revealed a 51% reduction in bone volume in 8-week-old mutant mice (Fig. 3G). Eight-week-old knockout mice also demonstrated a striking decrease in trabecular number (Fig. 3H), and an in trabecular separation (Fig. 3I) compared with normal siblings. Among bone formation parameters, the percentage of bone surface occupied by osteoblasts (Ob.S/BS) was significantly decreased by 27% in Smad4 mutant mice compared with wild-type controls (Fig. 3J). Mineral apposition rate (MAR, an index of individual osteoblast activity) and bone formation rate/bone surface (BFR/BS is determined from the number and the function of osteoblasts) were markedly reduced in Smad4 mutants (Fig. 3K,L). Consistently, formation of mineralized extracellular matrix (ECM) in neonatal calvarial osteoblast cultures from Smad4 mutant mice was significantly delayed compared with that from control mice at day 18 of culture (Fig. 3E,F). All these results indicate that the decreased bone density in Smad4 mutant mice possibly results from a decrease in the bone-forming activity of osteoblasts.

We further examined the expression of bone ECM genes in Smad4 mutants by in situ hybridization. The expression of osteopontin (Spp1, secreted phosphoprotein 1), an early osteoblast marker gene, was decreased in Smad4 mutants at 4 weeks compared with controls (Fig. 4A,B). The expression of collagen I (Col1a2) and osteocalcin, late markers of osteoblasts, were also reduced in mutants (Fig. 4C,D and data not shown). Real-time PCR analyses confirmed that alkaline phosphatase (Akp1) (Fig. 4E), an early differentiation marker of committed osteoblasts, was also downregulated in bone extracts of Smad4-deficient mice. The mRNA levels of Col1a2 and osteocalcin were all decreased (Fig. 4F,G). These data were consistent with the reduced number of osteoblasts described in the above histomorphometric measurements. We also found that the expression of Runx2 was decreased (Fig. 4H). The mRNA levels of Akp1, Col1a2, osteocalcin and Runx2 were all decreased in mutant calvarial cells that had been isolated under standard cultural condition for osteoblasts (supplementary material Fig. S2A,B and data not shown). The molecules in the Wnt/β-catenin signaling pathway were also changed. The mRNA levels of low-density lipoprotein-receptor-related protein 5 (Lrp5) and β-catenin were decreased, whereas lymphoid enhancer binding factor 1 (Lef1) was not changed. Levels of Dickkopf 2 (Dkk2), which blocks Wnt1-induced transcription of Lef1/Tcf target genes, were increased (see supplementary material Fig. S2C-F). To determine whether osteoblast proliferation was affected in Smad4 mutant mice, we measured the BrdU-positive cells on calvarial periosteal surfaces. Significantly fewer labeled nuclei were detected in sections from Smad4 mutant mice compared with those derived from control littermates (Fig. 4I,J). These results indicate that the proliferation of osteoblasts is decreased owing to the loss of Smad4.

Although the Smad4 mutant mice remained smaller than controls throughout their lifetime, the differences were reduced gradually in aged mice. The trabecular bone volume in knockout mice was increased when animals were older than 7 months, whereas the calvarial and cortical bone remained thinner than that of controls. This phenomenon was especially significant in 1-year-old mutant mice (Fig. 5A-D and data not shown). Histomorphometric measurements revealed increased trabecular bone volume (Fig. 5E), trabecular number (Fig. 5F) and a significantly decreased trabecular separation (data not shown) in 11-month-old mutant mice compared with normal siblings. Calcein labeling analysis detected markedly decreased mineral apposition rate (Fig. 5G) and bone formation rate (Fig. 5H) in aged Smad4 mutants. All these data suggest that increased bone mass in aged Smad4 mutants is possibly due to downregulation of osteoclast function.

A consequence of reduced osteoblast activity would be the reduction of osteoclast function. TRAP (tartrate-resistant acid phosphatase) staining on femur sections at 8 weeks revealed that this indeed was the case. A marked decrease in the number of TRAP-positive multinucleated osteoclasts demonstrated a decrease in bone resorption in Smad4 mutant mice (Fig. 6A,B). This was confirmed by histomorphometric analyses. Bone resorption parameters, the percentage of bone surface covered by mature osteoclasts (OcS/BS) and the number of mature osteoclasts (OcN/BPm) were significantly decreased in Smad4 mutant mice (Fig. 6C,D). Real-time PCR analysis revealed that the expression of TRAP and cathepsin K (Ctsk), lysosomal enzymes essential for osteoclastic bone resorption, were significantly reduced in calvaria of Smad4 mutant mice (Fig. 6E,F). When a common osteoclast progenitor population derived from the wild-type spleens was co-cultured with primary cultured osteoblasts derived from Smad4 mutant or control bone marrows, TRAP⁺ cells were significantly reduced in co-cultures with Smad4 mutant osteoblasts compared with controls (264.4±12.6 per well with mutants vs 169.25±7.8 per well with controls, n=5, P<0.001), indicating that the osteoclastogenic activity of the Smad4 mutant osteoblasts was significantly decreased (Fig. 6G-I). These data suggest that increased trabecular bone volume in aged Smad4 mutant mice is probably due to the decreased bone resorption.

The observation that the number of osteoclasts was reduced in Smad4 mutant mice prompted us to check the expression of components of the RANKL-OPG axis by RT-PCR. The transcripts of both RANKL and OPG genes were significantly increased in 7-week-old Smad4 mutant calvaria (Fig. 6J). Real-time PCR showed that the expression of the RANKL gene was elevated twofold in the calvaria of Smad4 mutants (Fig. 6K), whereas the expression of OPG was increased by a factor of seven (Fig. 6L). We also examined the expression of Bmp2, Bmp4 and Tgfb1, which regulate coupling between osteoblasts and osteoclasts by RT-PCR and real-time PCR. The expression of Tgfb1 was decreased (Fig. 6J,M), whereas changes in the expression levels of Bmp2 and Bmp4 were not significant (Fig. 6J,N and data not shown).

Skeletal homeostasis is controlled by the intricate coordination of constituent cells including osteoblasts and osteoclasts (Karsenty and Wagner, 2002). The number and the function of cells present in the bone microenvironment determine skeletal homeostasis and are regulated by systemic hormones and local bone growth factors (Canalis et al., 2003). Ablation of Smad4 in differentiated osteoblasts resulted in a growth retardation and decreased bone mass in young knockout mice. The Smad4 mutants showed a decreased number and impaired function of osteoblasts, suggesting that Smad4 is required for control of osteoblast function and regulation of bone mass. Interestingly, the differentiated osteoblast specific disruption of Smad4 resulted in an alteration of the RANKL/OPG axis and eventually led to decreased number and reduced activity of osteoclasts that caused higher trabecular bone volume in aged mutant mice. This indicates that responsiveness of osteoblasts to TGFβ signaling mediated by Smad4 plays an important role in the coupling of bone formation to bone resorption. Thus, the current study is the first direct evidence of an important function for Smad4 in maintaining bone homeostasis.

Previous studies have shown that the disruption of BMP signaling during embryo genesis affects skeletogenesis (Kingsley et al., 1992; Storm et al., 1994; Dudley et al., 1995; Luo et al., 1995; Thomas et al., 1997; Thomas et al., 1996; Katagiri et al., 1998; Solloway et al., 1998). Recently, tissue specific downregulation of BMP signals in mice has revealed an important function of BMP signaling in postnatal bone formation (Mishina et al., 2004). It is conceivable that the osteoblast-specific Smad4-knockout mutant mice largely copied the phenotypes observed in the osteoblast-specific Bmpr1a-knockout mutant mice and transgenic mice overexpressing a dominant-negative form of Bmpr1b in osteoblasts (Zhao et al., 2002; Mishina et al., 2004), given the fact that ablation of Smad4 in osteoblasts impaired the responsiveness of osteoblasts to BMP2. Previous studies have shown that BMPs might stimulate osteoblast differentiation by inducing expression and activation of Runx2, and inhibition of BMP signaling disrupts the ability of Runx2 to stimulate osteoblast differentiation and gene expression (Chen et al., 1998; Zhao et al., 2002).

In this study, we showed that the Smad4 mutant mice suffered from a decreased bone formation rate and a defect in bone mineralization, correlated with the downregulation of Runx2, providing in vivo evidence that Smad4 regulates osteoblast differentiation and maturation through Runx2. It has been shown recently that reduction of TGFβ enhanced the bone mass and the mechanical properties (Balooch et al., 2005). Although Smad4 deletion also impaired the responsiveness of osteoblasts to TGFβ1, the phenotypes of the Smad4 mutant mice were more similar to the mutants with reduced BMP signaling (Zhao et al., 2002; Wu et al., 2003; Mishina et al., 2004; Balooch et al., 2005) than those with diminished TGFβ signals (Balooch et al., 2005). This suggests that the differentiated osteoblasts at later stages might be less sensitive to TGFβ. Notably, in vitro reporter assays with TGFβ and BMP reporter-luciferase constructs revealed ∼70% reduction in TGFβ and BMP responsiveness, suggesting that some Smad4-independent compensatory mechanisms might exist both in vitro and in vivo. Indeed, previous studies show that TGFβ signals regulate osteoblast differentiation and apoptosis through Smad-independent signaling pathways (Hay et al., 2001; Celil and Campbell, 2005).

We noted that ablation of Smad4 in differentiated osteoblasts resulted in decreased proliferation and a reduced number of osteoblasts, which is different from the osteoblast-specific Bmpr1a-knockout mice which exhibit a normal number of osteoblasts (Mishina et al., 2004). This could be largely due to the downregulation of Tgfb1 in Smad4 mutant mice (Fig. 6J). Indeed, previous studies have shown that TGFβ1 increases the cell population that will differentiate into osteoblasts by inducing chemotaxis and proliferation (Janssens et al., 2005). The downregulation of TGFβ1 in Smad4 mutants raised the possibility that Smad4 might mediate a positive feedback loop regulating TGFβ1 expression in osteoblasts.

It has been shown that TGFβ signals can function as coupling factors to maintain the bone formation and bone resorption and thus skeletal homeostasis (Harada and Rodan, 2003). A striking phenotype in osteoblast-specific Smad4-knockout mice is that the trabecular bone volume was significantly higher than that of controls at 7 months of age, because of reduced osteoclast bone resorption. This suggests that Smad4-mediated TGFβ signals in osteoblasts play an important role in the control of bone resorption. Previous studies have shown that TGFβ has both direct enabling effects on osteoclast formation (Fox et al., 2000; Fuller et al., 2000; Kaneda et al., 2000), and indirect effects through the regulation of osteoblastic RANKL and OPG expression (Murakami et al., 1998; Takai et al., 1998; Quinn et al., 2001; Thirunavukkarasu et al., 2001). Consistently, we found that deletion of Smad4 in osteoblasts downregulated the expression of TGFβ1 and altered the RANKL/OPG ratio, and eventually caused a reduced rate of osteoclast differentiation. It is quite unexpected that the expression of OPG is greatly upregulated in Smad4 mutant mice, because both BMP and TGFβ have been shown to be able to increase OPG production in vitro (Thirunavukkarasu et al., 2001). A recent study has suggested that in vitro effects of TGFβ1 on the RANKL-OPG axis and differentiation of osteoclasts depend greatly on TGFβ1 concentration (Karst et al., 2004). Our results indicated that the in vivo effects of TGFβ on RANKL-OPG axis expression are much more complicated than we appreciated. Nevertheless, the downregulation of TGFβ1 and the shift from the RANKL-OPG axis in favor of OPG suppressed osteoclast formation and prevented excessive bone loss.

Mice that were homozygous for the floxed Smad4 allele (Smad4^Co/Co) (Yang et al., 2002) were bred with transgenic mice in which the 1.3 kb mouse OC promoter (Dacquin et al., 2002) controls Cre recombinase expression (OC-Cre) to generate Smad4^Co/+;OC-Cre mice. Smad4^Co/Co;OC-Cre mice were obtained by breeding the Smad4^Co/+;OC-Cre mice with Smad4^Co/Co mice. For routine genotyping, the Smad4 locus and the Cre transgene were detected by PCR using primers described previously (Yang et al., 2005; Yang et al., 2002).

To determine the specificity of Cre-mediated recombination, OC-Cre transgenic mice were bred with ROSA26 reporter mice (Soriano, 1999). The bones were stained as previously described (Yang et al., 2005). Osteoblasts derived from the bone marrow cells of double heterozygous or control mice were also stained with X-gal as previously described (Yang et al., 2005).

Radiographic analyses were carried out using a soft X-ray system (Contour Plus). BMD was measured from femurs by dual energy X-ray absorptiometry with a Piximus Mouse Densitometer (GE Lunar Medical System).

Tissues were fixed in 4% paraformaldehyde at 4°C overnight, decalcified in 5% EDTA-PBS and embedded in paraffin. The sections were stained with hematoxylineosin and TRAP by standard methods. For in vivo fluorescent labeling, 4-week-old animals were injected with calcein (20 mg/kg body weight) intraperitoneally at day 6 and day 3, whereas 8-week-old mice were injected at day 12 and day 2 before sacrifice. 5 μm plastic sections were stained using the modified Goldner trichrome technique (Gruber, 1992) and sections serial to the stained sections were left unstained for fluorescent analyses. An image analysis system (Osteometrics) was used for all histomorphometric analysis. Parameters for the trabecular bone were measured in an area 1.8 mm in length from 0.3 mm below the growth plate of the distal femur. For 11-month-old female mice, bone histomorphometric analysis was performed in an area 1.5 mm in length under the growth plate of the proximal tibias. All histomorphometric parameters are reported in accordance with the ASBMR nomenclature (Parfitt et al., 1987).

Calvarial cells from 3-day-old neonatal mice were established as previously described (Zhao et al., 2002). Bone marrow cells were inoculated at a density of 1.5×10⁶ cells/well in 24-well plates, 50 μg/ml L-ascorbic acid and 10 mM β-glycerophosphate were added at day 3 of culture. Osteoblast-osteoclast co-culture experiments were performed as described (Cao et al., 2003). Briefly, bone marrow cells were incubated for 7 days in differentiation medium and the osteoblast-like cells were plated at 2×10⁴ per well in 24-well plates. Upon confluency, non-adherent monocyte/macrophage progenitor cells derived from wild-type spleen (5×10⁵ per well) were then plated, and the cultures were complemented with 10^–8 mol/L 1,25-dihydroxy vitamin D₃. TRAP staining was performed after 5 days of culture.

Mice at 3 weeks of age were injected intraperitoneally with 10 μl of 10 mM BrdU in PBS per gram of body weight, and sacrificed 24 hours later (Zhang et al., 2005). Calvarial sections were detected with an anti-BrdU antibody (Sigma). All BrdU-positive (brown) nuclei in an area of 0.3 mm in length from the sagittal suture were counted using Osteomeasure software. Five animals per genotype were used and statistical differences between groups were assessed using the Student's t-test.

The reporter construct, 12×SBE-OC-Luc (Zhao et al., 2004) or p3TP-Lux (Chen et al., 1996) together with pCMV-Cre and Renilla control plasmids were co-transfected in osteoblasts isolated from calvaria of Smad4^Co/Co mice using lipofectamine 2000 (GIBCO BRL). 24 hours later, the medium was replaced by fresh medium containing 1% FCS in the presence or absence of 50 ng/ml BMP2 or 5 ng/ml TGFβ1 (R&D Systems). Cells were lysed 48 hours after transfection, and firefly luciferase activity was assayed using a dual luciferase reporter assay system (Promega) and normalized by Renilla luciferase activity. Each experiment was performed using triplicate data points and performed three times.

In situ hybridization was performed using standard procedures. Probes were labeled with [³⁵S]UTP using the MAXIscript in vitro transcription kit (Ambion). Slides were dipped in emulsion (Amersham Pharmacia) and exposed for 2-10 days before developing.

Total RNA was isolated from calvarial bone of 7-week-old mice using the TRizol reagent (Invitrogen) according to the manufacturer's instructions. 5 μg total RNA was reverse transcribed to cDNA with the use of the first-strand cDNA synthesis kit (Invitrogen). Real-time PCR was performed using the LightCycler system (Roche) with the FastStart DNA Master SYBR Green. The standard curve method of quantification was used to calculate the expression of target genes relative to the housekeeping gene Gapdh. The wild-type expression level was set to 1 as described previously (Glass, 2nd et al., 2005). Experiments were repeated at least three times. The following primers were used: OC, 5′-ACCCTGGCTGCGCTCTGTCTCT-3′ and 5′-GATGCGTTTGTAGGCGGTCTTCA-3′; Col1a2, 5′-CAGCGAAGAACTCATACAGCC-3′ and 5′-TTGGAGCAGCCATCGACTA-3′; Akp1, 5′-AGGGCAATGAGG TCACATCC-3′ and 5′-GCATCTCGTTATCCGAGTACCAG-3′; TRAP, 5′-AGACCCAGACCCTGAACACC-3′ and 5′-CGCCCAAGAAAGCTCTACCTAA-3′; Ctsk, 5′-CACGGCAAAGGCAGCTAAAT-3′ and 5′-CCATAGCCCACCACCAACAC-3′; OPG, 5′-ACGGACAGCTGGCACACCAG-3′ and 5′-CTCACACACTCGGTTGTGGG-3′; RANKL, 5′-CATGACGTTAAGCAACGG-3′ and 5′-AGGGAAGGGTTGGACA-3′; Lrp5, 5′-ATTGAAAGGGTCCACAAGGTC-3′ and 5′-GATAGCCACATCGTTGTTGTTAG T-3′; β-catenin, 5′-GCCATCTGTGCTCTTCGTC-3′ and 5′-ACACCCTTCTACTATCTCCTCC-3′; Lef1, 5′-TTCAGGTACAGGTCCCAGAATG-3′ and 5′-AGTCGGCGCTTGCAGTAGA-3′; Dkk2, 5′-CAGGGTAAACAATCAGTAGTCC-3′ and 5′-CAATGCCATTCCTTCACAA3′; Gapdh, 5′-TGCCCAGAACATCATCCCT-3′ and 5′-GGTCCTCAGTGTAGCCCAAG-3′. The primer pairs for Tgfb1, Bmp4 and Runx2 have been described previously (De Ranieri et al., 2005).

All results were expressed as mean ± s.d. All statistical analyses were performed using the SPSS software package for Windows release 11.0. P<0.05 was considered significant.

We thank Gerard Karsenty for providing the OC promoter, Hua Gu for the plasmid carrying the Cre gene, Yeguang Chen for 3TP-Lux, Bin Zhao for undecalcified sections, Yebin Jiang and Jing Ma for helpful discussion, Michael Zuscik for proofreading the manuscript. This work was supported by the National Natural Science Foundation of China (30430350), National Basic Research Program of China (2005CB522506; 2006CB943501), Key Technologies R&D Program (2006AA02Z168), National Science Supporting Program (2006BAI23B01-3) and Beijing Science Projects (Z0006303041231).