BMP2, but not BMP4, is crucial for chondrocyte proliferation and maturation during endochondral bone development

The BMP signaling pathway has a crucial role in chondrocyte proliferation and maturation during endochondral bone development. To investigate the specific function of the Bmp2 and Bmp4 genes in growth plate chondrocytes during cartilage development, we generated chondrocyte-specific Bmp2 and Bmp4 conditional knockout (cKO) mice and Bmp2,Bmp4 double knockout (dKO) mice. We found that deletion of Bmp2 and Bmp4 genes or the Bmp2 gene alone results in a severe chondrodysplasia phenotype, whereas deletion of the Bmp4 gene alone produces a minor cartilage phenotype. Both dKO and Bmp2 cKO mice exhibit severe disorganization of chondrocytes within the growth plate region and display profound defects in chondrocyte proliferation, differentiation and apoptosis. To understand the mechanism by which BMP2 regulates these processes, we explored the specific relationship between BMP2 and Runx2, a key regulator of chondrocyte differentiation. We found that BMP2 induces Runx2 expression at both the transcriptional and post-transcriptional levels. BMP2 enhances Runx2 protein levels through inhibition of CDK4 and subsequent prevention of Runx2 ubiquitylation and proteasomal degradation. Our studies provide novel insights into the genetic control and molecular mechanism of BMP signaling during cartilage development.


Introduction
During skeletal development, the majority of the bones in the body are established by the endochondral bone formation process, which is initiated by mesenchymal cell condensation and subsequent mesenchymal cell differentiation into chondrocytes and surrounding perichondrial cells. The committed chondrocytes proliferate rapidly forming the cartilage growth plate where cells are arranged in columns of proliferating, differentiating and terminally hypertrophic chondrocytes. Chondrocytes near the center of the cartilage elements exit the cell cycle initiating the process of hypertrophic differentiation to generate a calcified cartilage matrix. Eventually, the local vasculature, perichondrial osteoblasts and various other types of cells invade the calcified cartilage, replacing the terminally mature chondrocytes with marrow components and trabecular bone matrix. Primary ossification occurs with osteoblast-mediated bone formation, which initially occurs on the calcified cartilage template. Chondrocyte maturation and the endochondral bone development process is tightly regulated by a series of growth factors and transcription factors, including bone morphogenetic proteins (BMPs), fibroblast growth factors (FGFs), indian hedgehog (Ihh), parathyroid hormone-related protein (PTHrP), Wnt signaling proteins and Runt-related transcription factor 2 (Runx2) (Yoon and Lyons, 2004;Ornitz, 2005;Kronenberg, 2003;Komori, 2003;Kolpakova and Olsen, 2005).
BMPs are multi-functional growth factors that belong to the transforming growth factor b (TGF-b) super family. In vivo evidence suggests that BMP signaling is primarily mediated through the canonical BMP-Smad pathway in chondrocytes (Yoon et al., 2005). BMPs bind the type II receptor and phosphorylate type I serine or threonine receptors, which subsequently phosphorylate Smad1, Smad5 and Smad8 (R-Smads). The activated R-Smads form a complex with Smad4 before entering the nucleus to regulate target gene transcription. Several lines of evidence suggest that the BMP-Smad pathway has a crucial role in endochondral bone development. Removal of Bmp2 and Bmp4 specifically from mesenchymal cells leads to defects in skeletal development (Bandyopadhyay, 2006). Deletion of the Smad1 and Smad5 genes or the Bmpr1a and Bmpr1b genes in cartilage results in chondrodysplasia (Yoon et al., 2005;Retting et al., 2009). In addition to the role of BMPs in early mesenchymal cell differentiation (Haas and Tuan, 1999;Hatakeyama et al., 2004), they also have crucial roles during later stages of chondrocyte proliferation and differentiation (Shukunami et al., 2000;Leboy et al., 2001;Valcourt et al., 2002). However, as a result of overlapping and redundant functions among different BMP genes, the regulatory role of these genes during chondrocyte proliferation and maturation in vivo remains undefined.
Bmp2 and Bmp4 mRNAs are highly expressed in prehypertrophic and hypertrophic chondrocytes of the growth plate (Feng et al., 2003;Nilsson et al., 2007) (supplementary material Fig. S1). Bmpr1a is highly expressed in pre-hypertrophic chondrocytes, and phosphorylated Smad1, Smad5 and Smad8 proteins are detected in the lower region of proliferating columnar zone and pre-hypertrophic chondrocytes (Sakou et al., 1999;Yoon et al., 2006). The specific expression patterns of these genes suggest an essential role for Bmp2 and/or Bmp4 in chondrocyte proliferation and maturation during endochondral bone development. In vitro studies have shown that BMP2 and BMP4 stimulate the progression of chondrocyte hypertrophy (Hatakeyama et al., 2004;Leboy et al., 1997;De Luca et al., 2001;Minina et al., 2001;Horiki et al., 2004;Clark et al., 2009). Similarly, expression of constitutively active Bmpr1a in chondrocytes induces the acceleration of chondrocyte differentiation into hypertrophic chondrocytes (Kobayashi et al., 2005). These findings suggest that Bmp2 and Bmp4 have a similar and redundant role in chondrocyte maturation. To determine which one of these BMP genes is required for chondrocyte development in vivo, we have generated chondrocyte-specific Bmp2 and Bmp4 cKO mice and Bmp2,Bmp4 dKO mice. Chondrocyte-specific deletion of these BMP genes is achieved by breeding Col2a1CreER T2 transgenic mice  with the Bmp2 or Bmp4 floxed mice (Bmp2 fx/fx and Bmp4 fx/fx ). Chondrocyte-specific gene deletion is achieved by intraperitoneal injection of a single dose of tamoxifen (TM) to the pregnant female carrying embryos at embryonic day 12.5 (E12.5). We then assessed changes in chondrocyte maturation in these mutant embryos at E14.5 and E18.5. Our studies demonstrate that deletion of only Bmp2 or both Bmp2 and Bmp4 genes led to severe defects in chondrocyte proliferation and maturation during endochondral bone development. By contrast, chondrocyte-specific deletion of only the Bmp4 gene caused minor changes in chondrocyte maturation. Our findings indicate that Bmp2 has a crucial and non-redundant role in chondrocyte proliferation and maturation during endochondral bone development.

Results
Deletion of Bmp2 and Bmp4 or Bmp2 alone impairs skeletal development To investigate the role of endogenous Bmp2 and Bmp4 genes in growth plate chondrocyte maturation and skeletal development, pregnant mice with embryos at E12.5 were injected with TM. E18.5 embryos were collected and whole skeletal Alizarin Red and Alcian Blue staining was performed. Whole skeletons and individual skeletal elements of Bmp2 and Bmp4 (Bmp2/4) dKO and Bmp2 cKO embryos were very small compared with their Cre-negative littermate controls, suggesting impaired skeletal development in Bmp2/4 dKO and Bmp2 cKO embryos (Fig. 1A). Calvaria of these mutant embryos were smaller than those of Crenegative littermates, and cartilaginous occipital bones were nearly absent, demonstrating that intramembranous bone formation was also impaired. Compared with Cre-negative littermates, the deformed thoracic cavities of Bmp2/4 dKO and Bmp2 cKO embryos were significantly smaller with minimal bone formation. Spines and hind limbs of Bmp2/4 dKO and Bmp2 cKO embryos were also markedly shorter than Cre-negative littermates (Fig. 1B). However, only minor differences were observed in all skeletal elements analyzed from the Bmp4 cKO embryos compared with Cre-negative littermates, suggesting that the Bmp4 gene has a minor role in normal embryonic skeletal growth and development or is complemented by the expression of other BMP genes (Fig. 1A,B).
Formation of primary ossification center is delayed in Bmp2/4 dKO and Bmp2 cKO embryos To further analyze changes in skeletal development in Bmp2/4 dKO and Bmp2 cKO embryos, histological staining was performed on tibias sections of E14.5 Bmp2/4 dKO, Bmp2 and Bmp4 cKO embryos and the Cre-negative littermates. In Crenegative embryos, chondrocytes in the middle of tibia began the differentiation process forming a hypertrophic zone, which stained weakly with Alcian Blue or Safranin O compared with the adjacent immature chondrocytes. In Bmp2/4 dKO embryos, the whole tibia was smaller than that of Cre-negative embryos and chondrocyte hypertrophy was absent, as was evidence of formation of the primary ossification center. Bmp2 cKO embryos showed a very similar delay in the formation of the hypertrophic zone of tibia compared with Bmp2/4 dKO embryos. By contrast, there were minimal changes in the tibiae of Bmp4 cKO embryos when compared with the Cre-negative littermates ( Fig. 2A). Bmp4 cKO embryos had evidence of chondrocyte hypertrophy and formation of the primary ossification center. These findings demonstrated that chondrocyte hypertrophy is severely delayed by deletion of the Bmp2 gene, but not the Bmp4 gene, in Col2a1positive chondrocytes during embryonic skeletal development.
Chondrocyte maturation is impaired in Bmp2/4 dKO and Bmp2 cKO embryos Further histological analysis was performed on E18.5 embryos. The results demonstrated that the lengths of proliferative and hypertrophic zones of Bmp2/4 dKO and Bmp2 cKO embryos were significantly reduced with disorganized columnar chondrocyte structure (Fig. 2B). Normal hypertrophic chondrocytes were replaced with smaller number of enlarged hypertrophic chondrocytes, with expansion of both the cytoplasm and nucleus in Bmp2/4 dKO and Bmp2 cKO embryos ( Fig. 2B-D). The reduced size of the growth plate was associated with less endochondral bone formation, although ectopic matrix deposition was observed at the perichondrial region surrounding the abnormal cartilage in Bmp2/4 dKO and Bmp2 cKO embryos (Fig. 2C,D). Because Col2a1CreER T2 mice do not target perichondrial cells , this ectopic matrix formation suggests a non-cell-autonomous effect in Bmp2/4 dKO and Bmp2 cKO embryos. To rule out the toxic effect of TM on embryonic skeletal development, we injected TM in pregnant WT mice with embryos at E12.5. E18.5 embryos were collected and histology staining was performed. No significant difference in skeletal development was found by injection of TM (supplementary material Fig. S2). In L4 vertebrae, hypertrophic chondrocyte area was reduced over 50% in Bmp2/4 dKO and Bmp2 cKO embryos compared with those in Cre-negative embryos. The reduced chondrocyte hypertrophy and decreased matrix deposition observed in the center of the vertebral body Bmp2 and cartilage development 3429 (Fig. 2E,upper panel) indicates that the chondrocyte maturation process is also delayed in vertebral bones in Bmp2/4 dKO and Bmp2 cKO embryos. By contrast, only minor changes in growth plate chondrocyte maturation were found in E18.5 Bmp4 cKO embryos ( Fig. 2B,C,E), suggesting that the expression of the Bmp4 gene is not absolutely required for chondrocyte maturation and cartilage development. To further determine changes in cellular function in growth plate chondrocytes, we performed proliferating cell nuclear antigen (PCNA) staining and terminal deoxynucleotidyl transferasemediated dUTP-biotin nick end labeling (TUNEL) staining using tibia sections of E18.5 embryos. The results of PCNA staining demonstrated that cell proliferation was dramatically reduced in Bmp2/4 dKO and Bmp2 cKO embryos. By contrast, no significant reduction in PCNA-positive proliferating chondrocytes was found in Bmp4 cKO embryos ( Fig. 3A and B). The TUNEL staining images of E18.5 embryos demonstrated that chondrocyte apoptosis was significantly increased in Bmp2/4 dKO and Bmp2 cKO embryos (Fig. 3C).

Defects in chondrocyte differentiation in Bmp2/4 dKO and Bmp2 cKO embryos
To examine chondrocyte differentiation, we performed in situ hybridization assays using Col2a1, Col10a1 and Mmp13 probes. Col2a1 is highly expressed in growth plate chondrocytes in the resting and proliferating chondrocytes in E18.5 Cre-negative embryos. Opposing Col2a1 expression, Col10a1 is highly expressed in pre-hypertrophic and hypertrophic chondrocytes. Mmp13 is expressed in terminal hypertrophic chondrocytes that are proceeding into the final apoptotic stage where cartilage matrix is degraded and replaced by bone matrix. In this study, we found that the expression of all of these chondrocyte marker genes was significantly reduced in E18.5 Bmp2/4 dKO and Bmp2 cKO embryos by in situ hybridization (Fig. 4A). To further determine changes in chondrocyte marker gene expression, we also performed real-time RT-PCR assays and found that expression of Sox9, Acan (aggrecan) and Col2a1 was significantly reduced in chondrocytes in which the Bmp2 or Bmp2/4 genes were deleted ( Fig. 4B-D). Compared with the changes in gene expression in Bmp2/4 dKO and Bmp2 cKO embryos, there are minor changes in the expression of these chondrocyte marker genes in Bmp4 cKO embryos ( Fig. 4B-D). In this assay, primary chondrocytes were isolated from E18.5 mutant and Cre-negative embryos. It has been reported that BMP-2 regulates itself and several other BMP genes, including Bmp4, Bmp5, Bmp6 and Bmp8a (Harris et al., 1994;Ghosh-Choudhury et al., 1995;Chen et al., 1997;Edgar et al., 2007). The expression of several BMP family members was examined in Bmp2-and Bmp4-deleted chondrocytes. The expression of Bmp5, Bmp7, Bmp8b and Bmp9 were significantly downregulated in the chondrocytes in which the Bmp2 gene was deleted ( Fig. 4E-I). By contrast, no significant change in the expression of these genes was found in Bmp4-deficient chondrocytes ( Fig. 4J-N). These results suggest that these BMP genes are regulated by endogenous BMP2. In addition, we found that Bmp4 expression was upregulated in Bmp2-deficient chondrocytes and Bmp2 expression was upregulated in Bmp4-deficient chondrocytes (supplementary material Fig. S3), suggesting that expression of Bmp4 and Bmp2 genes was regulated by endogenous BMP2 and BMP4. To determine the interaction of Wnt/b-catenin and BMP signaling pathways, we isolated primary sternal chondrocytes from Bmp2/4 fx/fx mice. The cells were infected with Ad-Cre or Ad-GFP (control) and treated with BIO (1 mM), a GSK-3b inhibitor, and Wnt3a (100 ng/ml). We found that BIO-and Wnt3a-induced Alp expression was significantly inhibited in Bmp2/4-deficient chondrocytes (Fig. 4O), suggesting that canonical Wnt/b-catenin signaling may stimulate chondrocyte differentiation partially through a Bmp2/4-dependent mechanism. Significant amounts of ectopic matrix deposition were found in perichondrial areas of Bmp2/4 dKO and Bmp2 cKO embryos. To determine whether bone-specific markers and key transcription factors are upregulated in these areas, we examined Runx2 and Osterix expression by immunocytochemistry. We found that the numbers of Runx2-and Osterix-positive cells, and staining intensity were significantly increased in perichondrial areas of Bmp2/4 dKO and Bmp2 cKO embryos (Fig. 5A,B). By contrast, Runx2 expression in the proliferating and pre-hypertrophic areas was significantly reduced in Bmp2/4 dKO embryos (Fig. 5A). Taken together, the findings suggest that chondrocyte functions are severely impaired when the Bmp2 gene, but not the Bmp4 gene, is deleted in E18.5 Bmp2/4 and Bmp2 mutant embryos.

BMP-2 upregulates Runx2 protein levels by downregulation of CDK4 expression
It has been well documented that BMP2 induces Runx2 mRNA expression (Chen et al., 1998;Hassan et al., 2006). In the present studies, we examined the effects of BMP2 on Runx2 mRNA and protein expression in chondrocytes. We found that BMP2 induced Runx2 mRNA expression up to 3.5-fold, but enhanced Runx2 protein levels up to 10-fold (Fig. 6A,D,E). These observations suggest that, in addition to its transcriptional regulation, BMP2 also regulates Runx2 expression at the posttranscriptional level. Our previous report demonstrated that Runx2 protein levels are regulated by the ubiquitin-proteasome pathway through a cyclin-D1-CDK4-induced phosphorylation of Runx2 (Shen et al., 2006). To determine whether BMP2 regulates CDK4 expression, we performed western blot analysis and found that BMP2 significantly inhibited CDK4 expression in chondrogenic RCS cells (Fig. 6A). Similarly to BMP2, BMP4 also inhibited CDK4 expression in a time-dependent manner (supplementary material Fig. S4). The BMP2-mediated enhancement of Runx2 protein levels could be partially inhibited by expression of CDK4 in these cells (Fig. 6A). The regulatory role of BMP2 in Runx2 protein degradation was further confirmed by a Runx2 ubiquitylation assay. BMP2 inhibited Runx2 ubiquitylation whereas overexpression of CDK4 partially reversed the inhibitory effect of BMP2 on Runx2 ubiquitylation in RCS cells (Fig. 6B). To further determine the role of CDK4 in regulation of Runx2 protein expression, we transfected Cdk4 siRNA into RCS cells and found that similar to the addition of BMP2, transfection of Cdk4 siRNA also enhanced Runx2 protein levels. Addition of noggin blocked BMP2-induced Runx2 expression. By contrast, noggin had no effect on Cdk4 siRNA-induced upregulation of Runx2 protein (Fig. 6C), suggesting that CDK4 works downstream of BMP2 in regulation of Runx2 protein levels. We further analyzed the doseresponse effect of BMP2 on Runx2 protein levels and demonstrated that BMP2 upregulated Runx2 protein levels in a dose-dependent manner. Overexpression of CDK4 significantly inhibited BMP2-induced upregulation of Runx2 protein (Fig. 6D). Interestingly, we also found that over-expression of CDK4 also inhibited BMP2-induced Runx2 mRNA expression in RCS cells (Fig. 6E). To determine if BMP2 affects cyclin-D1-CDK4 interaction, we performed immunoprecipitation assays in the absence or presence of BMP2. We found that BMP2 significantly inhibited the interaction between cyclin D1 and CDK4 in chondrocytes (Fig. 6F). Addition of noggin abolished

Journal of Cell Science 124 (20) 3434
the inhibitory effect of BMP2 on the cyclin-D1-CDK4 interaction (Fig. 6F). These results indicate that BMP2 might prevent Runx2 degradation through downregulation of CDK4 expression and inhibition of cyclin-D1-CDK4 interaction in chondrocytes. It has been reported that Sox9 interacts with Runx2 and inhibits Runx2 function (Akiyama et al., 2004). In this study, we examined the effect of Sox9 siRNA on Runx2 levels in chondrogenic RCS cells. We found that transfection of Sox9 siRNA upregulated basal and BMP2-induced Runx2 protein levels in RCS cells (supplementary material Fig. S5). These results suggest that BMP2-regulated Runx2 expression is not Sox9 dependent. Taken together, these results suggest that part of the effect of BMP2 on Runx2 upregulation is mediated through downregulation of CDK4 expression and subsequent inhibition of Runx2 ubiquitylation in chondrocytes.

Discussion
In vitro studies suggest that BMP2 and BMP4 have similar functions. For example, both BMP2 and BMP4 induce mouse embryonic stem cell and human mesenchymal stem cell differentiation into chondrocytes (Kramer et al., 2000;Steinert et al., 2009). BMP2 and BMP4 also stimulate chondrocyte proliferation and hypertrophy (De Luca et al., 2001;Minina et al., 2001;Hatakeyama et al., 2004;Leboy et al., 1997). The expression of BMP2 and BMP4 proteins is detected in chondrocytes during endochondral ossification in fracture callus with the strongest expression detected in hypertrophic chondrocytes (Yu et al., 2010). During ectopic bone formation induced by implantation of Saos-2 cells into nude mice, both BMP2 and BMP4 are upregulated in mature chondrocytes (McCullough et al., 2007). Overexpression of Bmp2 or Bmp4 induces ectopic bone formation through a mechanism that is similar to endochondral ossification (Alden et al., 1999;Kubota et al., 2002;Jane et al., 2002). Because in vivo environments are different from in vitro studies, the in vitro findings need to be confirmed through an in vivo approach. Homozygous Bmp2 mutant embryos (conventional deletion of the Bmp2 gene) die between E7.5 and E10.5 and have defects in cardiac development (Zhang and Bradley, 1996); whereas homozygous Bmp4 mutant embryos (conventional deletion of the Bmp4 gene) die between E6.5 and E9.5, and show little or no mesodermal differentiation (Winnier et al., 1995). Because skeletal development begins around E10.5-E11.5, Bmp2 and Bmp4 conventional KO mouse models cannot be used to study skeletal biology. Deletion of the Bmp2 or Bmp4 gene specifically in the limb bud mesenchyme leads to severe chondrodysplasia, suggesting crucial roles of both Bmp2 and Bmp4 in early mesenchymal cell differentiation (Bandyopadhyay et al., 2006).
To determine the specific functions of Bmp2 and Bmp4 in chondrocyte proliferation and maturation during endochondral bone development in vivo, we have generated chondrocytespecific Bmp2 and Bmp4 cKO mice and Bmp2/4 dKO mice using Col2a1CreER T2 transgenic mice in which the expression of the CreER transgene is induced by tamoxifen and is restricted to cartilage. In our studies, deletion of the Bmp2/4 or Bmp2 gene in Col2a1-expressing chondrocytes resulted in severe defects in endochondral bone development, which differs from the results obtained by deletion of the Bmp2 and Bmp4 genes in mesenchymal progenitor cells (mediated by Prx1Cre transgenic mice). These findings suggest that during early mesenchymal cell differentiation, functions of Bmp2 and Bmp4 might be at least partially compensated by each other. However, during late stage chondrocyte maturation, Bmp2 function cannot be compensated by Bmp4 or other BMP genes in chondrocytes. In postnatal Bmp2 cKO mice (mediated by Prx1Cre), the fracture healing process is delayed. However, the fracture healing process was not affected in Bmp4 cKO mice (mediated by Prx1Cre) (Tsuji et al., 2006).
In Bmp2/4 dKO mice and Bmp2 cKO embryos, both chondrocyte proliferation and maturation are impaired. Chondrocyte columns in the proliferating and hypertrophic zones are disorganized with a dramatic decrease in Col2a1 and Col10a1 expression. In Bmp2/Bmp4 dKO and Bmp2 cKO embryos, ectopic bone formation was observed in perichondrial areas with enhanced Runx2 and Osterix expression. Because Col2a1CreER T2 mice do not target perichondrial cells, it seems that this ectopic bone formation reflects the secondary effect of deletion of the Bmp2 gene. In contrast to the perichondrial area, Runx2 expression in the proliferating and pre-hypertrophic areas was significantly reduced in Bmp2/4 dKO embryos. To investigate the regulatory mechanism of BMP2 on Runx2 expression, we examined the effect of BMP2 on Runx2 mRNA and protein levels in chondrogenic RCS cells. In addition to its stimulatory effect on Runx2 mRNA expression, BMP2 had much greater effect on Runx2 protein levels than its effect on Runx2 mRNA expression. Our in vitro studies demonstrate that BMP2 prevents Runx2 protein ubiquitylation through downregulation of CDK4 expression and inhibition of cyclin-D1-CDK4 interaction. Sox9 is an important downstream mediator of the BMP2 and hedgehog signaling pathways in osteoblasts. A Smad responsive element responsible for BMP2 activation was identified in the Sox9 promoter (Pan et al., 2008). Sox9 expression is upregulated by BMP2 in mesenchymal progenitor cell line (Zehentner et al., 1999). It has also been reported that Sox9 interacts with Runx2 and inhibits Runx2 function (Akiyama et al., 2004). In the present studies, we examined the effect of Sox9 siRNA on Runx2 protein In situ hybridization demonstrates that Col2a1, Col10a1 and Mmp13 expression is significantly reduced in E18.5 Bmp2/4 dKO and Bmp2 cKO embryos. (B-D) Total RNA was extracted from primary chondrocytes isolated from E18.5 Cre-negative control, Bmp2 cKO, Bmp4 cKO and Bmp2/4 dKO embryos. The expression of Sox9, Acan and Col2a1 genes was analyzed by real-time PCR. Results demonstrate that the expression of these chondrocyte marker genes is significantly reduced in Bmp2/4 dKO and Bmp2 cKO chondrocytes. By contrast, the expression of Sox9 and Acan was slightly but significantly reduced in Bmp4 cKO embryos. (E-I) Total RNA was extracted from primary chondrocytes derived from E18.5 Bmp2 cKO and Cre-negative embryos. The expression of BMP genes was analyzed by real-time PCR. Results demonstrated that the expression of Bmp5, Bmp7, Bmp8b and Bmp9 genes was significantly reduced in Bmp2deficient chondrocytes. (J-N) Total RNA was extracted from primary chondrocytes derived from E18.5 Bmp4 cKO and Cre-negative embryos. The expression of BMP genes was analyzed by real-time PCR. Results demonstrated that the expression of Bmp5, Bmp7, Bmp8b and Bmp9 genes is not significantly changed in Bmp4-deficient chondrocytes. (O) Primary sternal chondrocytes were isolated from 3-day-old Bmp2/4 fx/fx mice and were infected with Ad-Cre or Ad-GFP (control). 48 hours after infection, cells were treated with BIO (1 mM) or Wnt3a (100 ng/ml). Cell cultures were stopped 24 hours later and total RNA was extracted and expression Alp was examined by real-time PCR. BIO or Wnt3a-induced Alp upregulation is significantly inhibited in the Bmp2/4-deficient chondrocytes (Ad-Cre infected cells). *P,0.05, unpaired Student's t-test, n53. Values are means + s.e.m. levels and found that silencing of Sox9 upregulated Runx2 protein levels, suggesting that Sox9 has an inhibitory effect on Runx2, and that BMP2-mediated Runx2 upregulation is Sox9 independent.
BMP2 induces the expression of molecular marker genes characteristic of hypertrophic chondrocytes, such as Col10a1 and Alp (Valcourt et al., 2002). When BMP signals are transduced through R-Smads, Smad1 can interact with the transcription factor Runx2 . It has been reported that BMP2 promotes Col10a1 and Smad6 gene transcription through the conserved Runx2 binding sites (Leboy et al., 2001;Zheng et al., 2003;Wang et al., 2007). Our studies suggest that BMP2 might regulate Col10a1 expression through Smad1-Runx2 interaction at the 59 promoter region of the Col10a1 gene. Fig. 6. BMP2 protects Runx2 protein degradation through inhibition of CDK4 expression. (A) CDK4 expression construct was transiently transfected into chondrogenic RCS cells. 24 hours after transfection, the cells were treated with BMP2 (100 ng/ml) for 24 hours. Runx2 and CDK4 protein expression was detected by western blotting. BMP2 inhibits CDK4 expression and enhances Runx2 protein level. Expression of CDK4 partially inhibits BMP2-induced Runx2 upregulation. (B) Runx2 ubiquitylation assay. CDK4 expression construct was transiently transfected into RCS cells. 24 hours after transfection, the cells were treated with BMP2 (100 ng/ml) for 24 hours. Proteasome inhibitor MG132 (10 mM) was added to the medium 4 hours before cell lysates were collected. Ubiquitylated proteins were pulled down using an UbiQapture-Q kit and polyubiquitylated Runx2 was detected using an anti-Runx2 antibody. BMP2 inhibits Runx2 ubiquitylation and expression of CDK4 partially reverses the inhibitory effect of BMP2 on Runx2 ubiquitylation. (C) Cdk4 siRNA was transfected into RCS cells. Cells were treated with BMP2 (100 ng/ml) with or without noggin (300 ng/ml) 24 hours after Cdk4 siRNA transfection. The expression of Runx2 and CDK4 protein was detected by western blotting. Silencing of CDK4 results in an upregulation of Runx2 protein levels. Noggin inhibits the effect of BMP2 on Runx2 upregulation but could not inhibit the effect of Cdk4 siRNA on Runx2 protein upregulation. (D) CDK4 expression construct was transiently transfected into RCS cells. Cells were treated with BMP2 for 24 hours at different concentrations (0, 20, 100, 200 ng/ml) 24 hours after CDK4 transfection. The expression of Runx2 and CDK4 protein was detected by western blotting. BMP2 upregulates Runx2 protein levels in a dose-dependent manner. Expression of CDK4 partially inhibits BMP2-induced Runx2 protein upregulation. (E) RCS cells were treated with different concentrations of BMP2 with or without transfection of CDK4. Runx2 mRNA expression was determined by real-time PCR. CDK4 partially inhibits BMP2-induced Runx2 mRNA expression. (F) RCS cells were transfected with cyclin D1 and CDK4 expression plasmids and treated with BMP2 (100 ng/ml). Cell lysates were collected and subjected to IP using an anti-cyclin-D1 or anti-CDK4 antibody followed by western blotting using the anti-CDK4 or anti-cyclin-D1 antibody. Treatment of BMP2 inhibits cyclin-D1-CDK4 interaction and addition of noggin blocks the inhibitory effect of BMP2 on cyclin-D1-CDK4 interaction.
Our previous observations and reports from other laboratories demonstrated that both Bmp2 and Bmp4 genes are expressed in chondrocytes during embryonic development and early postnatal stages at similar levels (Feng et al., 2003) (supplementary material Fig. S1). Thus, the phenotypic difference of skeletal development observed in Bmp2 and Bmp4 cKO embryos could not be explained by the expression patterns of these two genes in chondrocytes. Previous reports suggest that BMP2 regulates the expression of other BMP family members in mesenchymal progenitor cells, osteoblasts and chondrocytes (Harris et al., 1994;Ghosh-Choudhury et al., 1995;Chen et al., 1997;Ghosh-Choudhury et al., 2001;Edgar et al., 2007), suggesting that BMP2 serves as an upstream regulator of other BMP genes in chondrocytes. In terms of the mechanism by which BMP2 and BMP4 regulate chondrocyte maturation, the main difference between these two growth factors is that BMP2 controls Bmp2 and expression of other BMP genes through autocrine and paracrine regulatory mechanisms. This notion is supported by several lines of evidence. (1) Our previous studies demonstrate that BMP2 upregulates Bmp2 gene transcription through the Bmp2 proximal promoter element (Ghosh-Choudhury et al., 2001). (2) In a fracture healing study, it has been shown that Bmp2 expression reaches its maximal level at day 1 after fracture. Gdf5 showed maximal expression at day 7. The expression of Bmp4, Bmp7 and Bmp8 was detected from day 14 to day 21, whereas Bmp5, Bmp6 and Gdf10 were expressed from day 3 to day 21 (Cho et al., 2002). (3) It has been reported that in a bone marrow cell culture, addition of BMP2 neutralizing antibody reduced the expression of endogenous levels of BMP2, BMP3, BMP5 and BMP8a, whereas addition of BMP2 had the opposite effect (Edgar et al., 2007). (4) In the present studies, we demonstrated that expression of Bmp5, Bmp7, Bmp8b and Bmp9 was downregulated in Bmp2-deficient chondrocytes. By contrast, expression of these BMP genes was not significantly changed in the Bmp4-deficient chondrocytes.
The phenotype of Bmp2 cKO embryos is similar to that of Smad1/ 5 dKO embryos (Retting et al., 2009) and Bmpr1a/Bmpr1b dKO embryos (Yoon et al., 2005), including impaired skeletal development, disorganized growth plate formation, decreased cartilage matrix deposition, and decreased chondrocyte proliferation and increased chondrocyte apoptosis in the growth plate. Ectopic matrix deposition at the perichondrial area surrounding prehypertrophic and hypertrophic chondrocytes was also seen in Smad1/5 dKO embryos, which is consistent with what we observed in Bmp2 cKO mice. However, skeletal development was more severely impaired in Smad1/5 dKO embryos and Bmpr1a/ Bmpr1b dKO embryos compared with Bmp2 cKO embryos. One possibility is that Bmp2 cKO was induced at stage E12.5 in our studies. By contrast, the Smad1/5 and Bmpr1a/Bmpr1b genes were deleted earlier than E12.5, which could have led to more severe defects in chondrogenesis. In summary, our findings indicate that Bmp2 is required for chondrocyte maturation and endochondral bone formation during embryonic development.

Materials and Methods
Generation of Bmp2/4 dKO mice and Bmp2 and Bmp4 cKO mice Col2a1-CreER T2 mice were generated in our lab . Bmp2 fx/fx mice were generated in the lab of Stephen Harris at the University of Texas Health Science Center at San Antonio, TX (Singh et al., 2008) and Bmp4 fx/fx mice were a gift from Brigid Hogan (Duke University, town, state) (Kulessa and Hogan, 2005;Gluhak-Heinrich et al., 2010). To generate KO embryos, Bmp2 fx/fx ;Bmp4 fx/fx , Bmp2 fx/fx and Bmp4 fx/fx mice were crossed with Col2a1-CreER T2 ;Bmp2 fx/fx ;Bmp4 fx/fx , Col2a1-CreER T2 ;Bmp2 fx/fx and Col2a1-CreER T2 ;Bmp4 fx/fx mice, respectively. The pregnant mice were injected with tamoxifen (1 mg/10 g body weight, i.p.) at E12.5 and sacrificed at E14.5 and E18.5. The Cre-positive embryos were used as KO embryos and the Cre-negative littermates were used as controls.

Whole embryo Alizarin Red and Alcian Blue staining
Embryos at E18.5 were collected and the skin, viscera and adipose tissues were carefully removed. Whole skeletons were fixed in 95% ethanol for 2 days followed by fixation in acetone for an additional day, and stained with 0.015% Alcian Blue and 0.005% Alizarin Red for 3 days. Images of the skeletons were taken when most of the soft tissue was digested in 1% potassium chloride.

Histological analysis
Tibiae from E14.5 and E18.5 embryos and vertebrae from E18.5 embryos were fixed in 4% paraformaldehyde, decalcified, dehydrated and embedded in paraffin. Serial midsagittal sections (3 mm thick) of tibias and vertebrae were cut and stained with Alcian Blue and hematoxylin and eosin or Safranin O and Fast Green, respectively, for morphometric analysis.

PCNA staining
Paraffin sections (3 mm thick) of E18.5 tibias were rehydrated and blocked in 3% H 2 O 2 in methanol for 15 minutes and digested in Proteinase K (10 mg/ml) for 10 minutes at room temperature. PCNA staining was performed with a PCNA staining kit (Promega, WI).

TUNEL staining
Rehydrated paraffin sections (3 mm thick) of E18.5 tibias were fixed with 4% formaldehyde solution in PBS for 15 minutes followed by digestion in Proteinase K (10 mg/ml) for 10 minutes. TUNEL staining was performed using Fluorometric TUNEL System (Promega, WI). After mounting with DAPI reagent to stain nuclei, the samples were analyzed under a fluorescence microscope using a standard fluorescein filter set to view the green fluorescence of fluorescein at 520¡20 nm; and view blue DAPI fluorescence at 460 nm.

In situ hybridization
Radiolabeled probes for Col2a1, Col10a1 and Mmp13 were created by transcribing linearized antisense complementary deoxyribonucleic acid (DNA) in the presence of [ 35 S]UTP using T7 polymerases at 37˚C for 2 hours. DNA was removed with RNAse-free DNase. The labeled RNA was purified using a mini Quick Spin RNA Column. All probes have been previously characterized (Dong et al., 2010). Sections (5 mm thick) of E18.5 tibias were prepared. After dewaxing and rehydration, sections were pretreated with 10 mg/ml Proteinase K, 0.2N hydrogen chloride and 0.1 M triethanolamine at room temperature. Hybridization was performed at 55˚C for 18 hours. Non-specific binding was reduced by adding 10 mg/ml RNase A and several washes in SSC. After dipping in nuclear-type emulsion emulsion, the slides were exposed for 3-7 days at 4˚C followed by developing and fixation with Kodak developer and fixer. The slides were counterstained with Toluidine Blue, dehydrated and coverslipped.

Primary chondrocyte isolation
Three-day-old neonatal mice were euthanized and genotyped using tail tissues obtained at the time of death. The anterior rib cage and sternum were harvested, washed with phosphate buffered saline (PBS), and then digested with 2 mg/ml Pronase (Roche) in PBS in a 37˚C water bath with continuous shaking for 60 minutes. This was followed by incubation in a solution of 3 mg/ml collagenase D (Roche)/Dulbecco's modified Eagle's medium (DMEM) for 90 minutes at 37˚C. The soft tissue debris was thoroughly removed. The remaining sterna and costosternal junctions were further digested in a fresh collagenase D solution in Petri dishes in a 37˚C incubator for 5 hours with intermittent shaking. The digestion solution was filtered to remove all residual bone fragments, and centrifuged. The cells were washed and collected for RNA analysis.

Cell culture and transfection
Rat chondrosarcoma (RCS) cells were cultured in DMEM supplemented with 10% fetal calf serum at 37˚C under 5% CO 2 . DNA plasmids were transiently transfected into RCS cells in 6 cm culture dishes using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). Empty vector was used to keep the total amount of transfected DNA constant in each group in all experiments. FLAG-EGFP plasmid was cotransfected as an internal control for transfection efficiency. Western blot and immunoprecipitation (IP) assays were performed 24 hours after transfection.

Western blotting and ubiquitylation assay
Western blotting and in vivo ubiquitylation assay were performed as described previously (Shen et al., 2006;Zhang et al., 2010). For the Runx2 ubiquitylation assay, the proteasome inhibitor MG132 (10 mM) was added to the cell culture 4 hours before cells were harvested. The rat anti-Runx2 monoclonal antibody was purchased from Marine Biological Laboratory (MBL, town, MA). The rabbit anti-CDK4 (C-22) polyclonal antibody and the rabbit anti-ubiquitin (FL-76) polyclonal antibody were purchased from Santa Cruz Biotechnology (Santa Cruz, CA).

Real-time quantitative polymerase chain reaction (qPCR)
Total RNA was extracted from RCS cells and primary mouse chondrocytes using RNAzol B solution (Tel-Test, town, TX). DNAse-I-treated total RNA was reverse transcribed-using oligo-(dT) and cDNA was amplified by PCR in a total volume of 20 ml solution containing 10 ml SYBR Green Master Mix (Thermo Scientific), 1 ml of the diluted (1:5) cDNA, and 10 pM of forward and reverse primers specific for the genes listed in supplementary material Table S1.