Smad5 and DPC4 Are Key Molecules in Mediating BMP-2-induced Osteoblastic Differentiation of the Pluripotent Mesenchymal Precursor Cell Line C2C12*

Since the bone morphogenetic proteins (BMPs) are members of the transforming growth factor-β (TGF-β) superfamily that induce the differentiation of mesenchymal precursor cells into the osteogenic cells, we identified the relevant signaling molecules responsible for mediating BMP-2 effects on mesenchymal precursor cells. BMP-2 induces osteoblastic differentiation of the pluripotent mesenchymal cell line C2C12 by increasing alkaline phosphatase activity and osteocalcin production. As recent studies have demonstrated that cytoplasmic Smad proteins are involved in TGF-β superfamily signaling, we plan to isolate the relevant Smad family members involved in osteoblastic differentiation. We identified human Smad5, which is highly homologous to Smad1. BMP-2 caused serine phosphorylation of Smad5 as well as Smad1. In contrast, TGF-β failed to cause serine phosphorylation of Smad1 and Smad5. We found Smad5 is directly activated by BMP type Ia or Ib receptors through physical association with these receptors. Following phosphorylation, Smad5 bound to DPC4, another Smad family member, and the complex was translocated to the nucleus. Overexpression of point-mutated Smad5 (G419S) or a C-terminal deletion mutant DPC4 (DPC4ΔC) blocked the induction of alkaline phosphatase activity, osteocalcin production, and Smad5-DPC4 signaling cascades upon BMP-2 treatment in C2C12 cells. These data suggest that activation of Smad5 and subsequent Smad5-DPC4 complex formation are key steps in the BMP signaling pathway, which mediates BMP-2-induced osteoblastic differentiation of the C2C12 mesenchymal cells.

The bone morphogenic proteins (BMPs) 1 are members of the transforming growth factor-␤ (TGF-␤) superfamily which have been implicated in embryogenesis, organogenesis, and morphogenesis (1). Among BMPs, BMP-2 and BMP-4 have been shown to promote the development of bone and cartilage by inducing the differentiation of undifferentiated mesenchymal cells into the osteoblastic cells or cartilage cells, respectively (2). BMP-2 has been shown to induce ectopic bone or cartilage formation when implanted in muscular tissue in vivo (2,3) and stimulate osteoblastic differentiation of the mesenchymal cells in vitro as assessed by the stimulation of calcification, alkaline phosphatase (ALP) expression, or osteocalcin production (4,5). However, the molecular mechanisms responsible for the effects of BMP on differentiation of these cells toward bone and cartilage are poorly understood.
BMPs exert their diverse biological effects through two types of transmembrane receptors, BMP type I (BMPIR) and type II (BMPRII) receptors (6,7), which possess intrinsic serine/threonine kinase activity (6,7). BMPIR is further subclassified into BMP type IaR (also called ALK3) and IbR (also called ALK6) (6,7), but their functional difference in BMP signaling is unknown at the present time. Upon binding to the type II receptors, BMPs induce heterodimerization between BMP type I and type II receptors, and transduce signals into the cytoplasm (6,7). Recent studies have shown that cytoplasmic signaling molecules, including Mad (mother against dpp), the Xenopus homologue of Mad, Xmad1, Xmad2, and several human homologues of Mad, Smads, play critical roles in TGF-␤ superfamily signaling (7)(8)(9). To date, seven Smad family members that possess ligand selectivity have been identified (7,8). For example, Smad1 has been implicated in BMP responses (10 -12), whereas Smad2 (13,14) and Smad3 (15,16) are activated upon treatment with TGF-␤. DPC4 (Smad4) was initially found to be a tumor suppressor in pancreatic cancers (17). DPC4 is not phosphorylated but forms a complex with Smad1, Smad2, and Smad3 upon BMP or TGF-␤ treatment (18,19). Thus, DPC4 may have unique roles in TGF-␤ superfamily signaling. Interestingly, Smad6 (20) and Smad7 (20,21) are found to possess unique structures compared with other Smad family members and both Smad6 and Smad7 inhibit TGF-␤ effects. Despite these data, it is not known whether these Smad family members are specific for these growth factors, or whether the growth factors may utilize different Smad family members for different biological effects. Furthermore, considering the multifunctional properties of the BMPs, it is possible that there are still unidentified Smad family members.
In the present study, we first sought Smad family members that might be involved in BMP-induced osteoblastic differentiation of pluripotent mesenchymal precursor cells. We isolated several human Smad family members including Smad5. Because the functional roles of Smad5 in TGF-␤ superfamily  signaling have not been fully characterized as yet, we focused  our efforts on Smad5 and found that Smad5 was directly activated by BMP type Ia or Ib receptors upon BMP-2 stimulation  and transduced BMP-2 signals to the nucleus by forming a  complex with DPC4. Moreover, we demonstrate that Smad5 and DPC4 play a critical role in the induction of the osteoblastic differentiation of the pluripotent mesenchymal cells C2C12 by BMP-2.

MATERIALS AND METHODS
Cells and Antibodies-293 cells, L6 cells and C2C12 cells were cultured in DMEM containing 10% fetal bovine serum (FBS; HyClone Laboratories, Logan, UT). NMuMg cells were purchased from ATCC and cultured in DMEM containing 10% FBS and 10 g/ml insulin (Sigma). Anti-HA polyclonal antibody, anti-phosphotyrosine monoclonal antibody, and anti-Flag monoclonal antibody were purchased from BAbCO, Transduction Laboratories, and IBI, respectively. Antiphosphoserine and anti-phosphothreonine polyclonal antibodies were purchased from Zymed Laboratories Inc.
cDNA Cloning-Human homologues of Mad cDNA were isolated using the PCR-nested cloning approach. Degenerate oligonucleotide primers for PCR were designed based on the conserved region of Drosophila Mad (22,23) and Caenorhabditis elegans Sma-2 (24) proteins (forward primer; 5Ј-CA(C/T)AT(A/C/T)GGNAA(A/G)GGNGT-3Ј encoding HIGKGV; reverse primer; 5Ј-TG(A/T/G)AT(C/T)TC(A/G/T)ATCCA-(A/G)CANGGNGT-3Ј, encoding TPCWIEIH). After PCR amplification, predicted size PCR products were subcloned into TA cloning vector (Invitrogen) and their DNA sequences were determined by dideoxy DNA sequencing kit (Upstate Biotechnology, Inc., Lake Placid, NY). The PCR products were released from TA-cloning vector with EcoRI, and radiolabeled with random primed labeling kit (Boehringer Mannheim) and [ 32 P]dCTP (NEN Life Science Products). Human 293 cell cDNA library (gifted by Joseph Schlessinger and Ivan Dikic) were screened with radiolabeled probes and positive clones were isolated by two additional round of screening. The isolated clones were subjected to DNA sequence analysis.
Immunoprecipitation, Western Blotting, and Metabolic Labeling-The cells were serum-starved with DMEM containing 0.2% FBS for 16 h, and treated with 100 ng/ml BMP-2 or 10 ng/ml TGF-␤ for 15 min. The cells were washed three times with ice-cold phosphatebuffered saline buffer (PBS), and solubilized in lysis buffer (20 mM Hepes (pH 7.4), 150 mM NaCl, 1 mM EGTA, 1.5 mM MgCl 2 , 10% glycerol, 1% Triton X-100, 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, 0.2 mM sodium orthovanadate) (25). The lysates were centrifuged for 15 min at 4°C, 16,000 ϫ g. The lysates were incubated with antibodies for 4 h at 4°C, followed by immunoprecipitation with protein A-Sepharose (Zymed Laboratories Inc.) or protein G-agarose (Boehringer Mannheim). Immunoprecipitates were washed five times with lysis buffer and boiled in SDS sample buffer containing 0.5 M ␤-mercaptoethanol, and supernatants were recovered as immunoprecipitate samples. These samples were separated by SDS-PAGE, transferred to nitrocellulose membranes, immunoblotted with anti-antibodies. The samples were visualized with horseradish peroxidase coupled to protein A (KPL) or horseradish peroxidase-coupled anti-mouse IgG antibodies (Cappel), and enhanced by ECL detection kits (Amersham). For 32 P metabolic labeling, serum-starved cells were incubated phosphate-free DMEM for 3 h, and then incubated with 0.5 mCi/ml 32 P i (NEN Life Science Products) for 3 h. The lysates of cells treated with or without 100 ng/ml BMP-2 or 10 ng/ml TGF-␤ for 15 min were immunoprecipitated with anti-Flag antibody and subjected to SDS-PAGE, followed by autoradiography.
Immunofluorescent Staining-The cells were serum-starved with DMEM containing 0.2% FBS for 16 h, and treated with 100 ng/ml BMP-2 for 40 min. The cells were washed three times with ice-cold PBS and fixed with 3.8% paraformaldehyde-PBS. After 15 min of incubation with 0.1% Triton-PBS, the cells were blocked with 1% bovine serum albumin-PBS, incubated with anti-Flag antibody (1:500) for 2 h, and washed six times with 0.1% Triton-PBS, followed by incubation with fluorescein isothiocyanate-conjugated anti-mouse IgG antibody (Jackson Immunoresearch Laboratories Inc.). The cells were extensively washed with PBS and visualized by fluorescence microscope (Zeiss). The number of cells whose nucleus was stained by anti-Flag antibody was counted per hundred cells at four independent fields.
GST Fusion Proteins and in Vitro Binding Assay-Smad5 and DPC4 cDNA were subcloned into pGEX-2T (Pharmacia Biotech Inc.) in frame. GST fusion proteins were expressed and purified by glutathione-agarose affinity column (25). Recombinant GST or GST fusion proteins (0.5 g) immobilized on agarose beads were incubated with the cell lysates for in vitro binding assay (25). After extensively washing the beads, the molecules associated with GST fusion proteins were determined by Western blotting using anti-HA polyclonal or anti-Flag monoclonal antibodies.
Specificity of Smad5 Responsiveness for BMP-2-Since Smad5 shows high homology to Smad1 that is phosphorylated  1 and 2) or BMPIbR (lanes 3 and 4) was co-transfected into 293 cells with BMPRII. Cells were stimulated with 100 ng/ml BMP-2 (lanes 2 and 4) for 15 min, lysed, and incubated with GST-Smad5, and the complexes were determined by immunoblotting with anti-HA antibody. The difference in the position of two bands is due to different molecular sizes for BMPIaR and BMPIbR. C, GST-Smad5, but not GST nor GST-DPC4, binds to activated BMPIbR. 293 cells were transfected with HA-tagged-BMPIbR and BMPRII, and stimulated with 100 ng/ml BMP-2 (lanes 2, 4, and 6) for 15 min. The cell lysates were incubated with GST (lanes 1 and 2), GST-Smad5 (lanes 3 and 4), or GST-DPC4 (lanes 5 and 6), and the complexes were visualized by immunoblotting with anti-HA antibody. D, MH2 domain of Smad5 is required for binding to BMPIbR. 293 cells were transfected with HA-tagged-BMPIbR and BMPRII, and stimulated with 100 ng/ml BMP-2 (lanes 2, 4, and 6) for 15 min. The cell lysates were incubated with GST fulllength (lanes 1 and 2), MH1 domain-deleted (⌬MH1; lanes 3 and 4) or MH2 domain-deleted (⌬MH2; lanes 5 and 6) Smad5, and the complexes were visualized by immunoblotting with anti-HA antibody.
Association of Smad5 with BMPIR-To further determine the interaction of Smad5 with BMPIR, we next explored whether Smad5 physically associates with BMPIRs. We performed co-immunoprecipitation experiments in 293 cells that were co-transfected with HA-BMPIbR and Flag-Smad5. As shown in Fig. 3A, Smad5 was co-immunoprecipitated with BMPIbR in a BMP-2-dependent manner (lanes 1 and 2). Consistent with this result obtained in living cells, GST-Smad5 associated with BMP-2-stimulated BMPIaR (Fig. 3B, lanes 1  and 2) or BMPIbR (Fig. 3, lanes 3 and 4 in B and C). The difference in the position of the bands of BMPIaR and BMPIbR observed in Fig. 3B is due to the difference of molecular size of BMPIaR and BMPIbR. GST alone (Fig. 3C, lanes 1 and 2) or GST-DPC4 (lanes 5 and 6) was unable to bind to BMPIR. We also found that GST-Smad5 was phosphorylated by BMPIRs in vitro (Fig. 2E). These findings suggest that activated BMPIRs . These cells were then treated for 3 days with BMP-2 (300 ng/ml), and ALP activity (B) and osteocalcin production (C) were determined as described in the text. Data are shown as mean Ϯ S.D. D, Mutant Smad5 (G419S) inhibits serine phosphorylation of wild-type Smad1 and Smad5 in C2C12 cells. Equal amounts (0.5 g) of wild-type Flag-Smad1 (lanes 1-4), Flag-Smad5 (lanes 5-8), or Flag-Smad3 (lanes 9 -12) were co-transfected with equal amounts (5 g) of pcDNA3 (Invitrogen) (lanes 1, 2, 5, 6, 9, and 10) or G419S mutant Flag- Smad5 (lanes 3, 4, 7, 8, 11, and 12) into C2C12 cells. Cells were stimulated with 300 ng/ml BMP-2 for 15 min, lysed, immunoprecipitated with anti-Flag antibody, and immunoblotted with anti-phosphoserine antibody. Expression levels of Flag-Smad1, Flag-Smad5, or Flag-Smad3 were determined by immunoblotting with anti-Flag antibody (bottom). E, mutant DPC4 (DPC4⌬C) inhibits heterocomplex formation between Smad5 and DPC4, but not serine phosphorylation of Smad5 in C2C12 cells. Wild-type Flag- Smad5 (1 g, lanes 1-4) and 0.5 g of wild-type HA-DPC4 (lanes 1-4) were co-transfected with pcDNA3 (Invitrogen) (5 g) (lanes 1 and 2) or mutant HA-DPC4(DPC4⌬C) (5 g) (lanes 3 and 4). Cells were stimulated with 300 ng/ml BMP-2 for 15 min, lysed, and immunoprecipitated with anti-Flag antibody, and the immunoprecipitates were determined for the heterocomplex formation of Smad5 with DPC4 and serine phosphorylation of Smad5 by immunoblotting with anti-HA (top panel) and anti-phosphoserine (second panel) antibody, respectively. Expression levels of Flag-Smad5 or HA-DPC4 were determined by immunoblotting with anti-Flag (third panel) antibody or anti-HA (bottom panel), respectively. directly phosphorylate Smad5 through physical association. Interestingly, the mutant Smad5 (G419S) that is not phosphorylated by BMP-2 was also able to bind to activated BMPIbR (Fig. 3A, lane 4). The result suggests that the mutant Smad5 (G419S) may compete with the intact Smad5 for the binding to BMPIRs regardless the state of phosphorylation. We also found that GST-Smad5 lacking the MH2 domain was unable to bind to activated BMPIbR (Fig. 3D, lanes 5 and 6), whereas MH1 domain-deleted GST-Smad5 still retained binding capacity (Fig. 3D, lanes 3 and 4). These data indicate that the MH2 domain is responsible for the physical association of Smad5 to BMPIbR.
Association of Activated Smad5 with DPC4 -Smad1 that has high homology with Smad5 is shown to form a heterocomplex with DPC4 (18,19). To unravel further downstream of Smad5, we determined the interaction of Smad5 with DPC4 by co-immunoprecipitation and an in vitro binding assay using GST-DPC4 (25). As shown in Fig. 4A, Smad5 associated with DPC4 in a BMP-2-dependent manner (lane 2). In contrast, the mutant Smad5 (G419S) that is not phosphorylated by BMP-2 failed to bind to DPC4 (Fig. 4A, lane 4). Of note, BMP-2activated Smad5 failed to associate with the C-terminal deletion mutant DPC4 (DPC4⌬C) (Fig. 4A, lane 6). Consistent with these results obtained in co-immunoprecipitation experiments, BMP-2-activated Smad5 also associated with GST-DPC4 (Fig.  4B, lane 2), but not with GST-DPC4⌬C (lane 6) in vitro. Furthermore, mutant Smad5 (G419S) was not able to bind to GST-DPC4 (lane 4). These results demonstrate that the phosphorylated Smad5 forms heterocomplex with DPC4 and suggest that the phosphorylation of Smad5 and the presence of the C terminus of DPC4 are required for heterocomplex formation between Smad5 and DPC4. Since DPC4 did not physically associate with BMPIRs (Fig. 3C, lane 6), DPC4 most likely serves as a downstream molecule of Smad5 in the BMP-2 signaling pathway.
Translocation of Smad5 and DPC4 into Nucleus by BMP-2-We next examined whether Smad5-DPC4 complex translocates into the nucleus, as shown in the case of Smad1 and Smad2 (11,31). Immunofluorescent staining demonstrated Flag-Smad5 and Flag-DPC4 clearly translocated and accumulated in the nucleus as early as 40 min after BMP-2 treatment (Fig. 5, A, B, E, and F). The mutant Smad5 (G419S) or mutant DPC4 (DPC4⌬C) did not show nuclear translocation by BMP-2 treatment (Fig. 5, C, D, G, and H). These results suggest that phosphorylated Smad5, following complex formation with DPC4, translocates to the nucleus and might function as a transcription-regulating factor, as is the case of Smad2 (31).
Role of Smad5 and DPC4 in Osteoblastic Differentiation of C2C12 Cells-Since BMP-2 plays an important role in osteogenesis by regulating the differentiation of the undifferentiated mesenchymal cells in vivo (2,3,32) and in vitro (4,5,27), we next explored the biological role of Smad5 in a pluripotent mesenchymal cell line C2C12, which shows osteoblastic differentiation in the presence of BMP-2 (27). BMP-2 induced serine phosphorylation of Flag-Smad5 that was transfected in C2C12 cells (Fig. 6A). In conjunction with this, BMP-2 also induced a marked increase in ALP activity (Fig. 6B) and production of osteocalcin (Fig. 6C) in C2C12 cells as reported previously (27). ALP and osteocalcin are widely recognized phenotypic markers of cells of osteoblast lineage (27). Importantly, overexpression of the mutant Smad5 (G419S) in C2C12 cells strongly inhibited BMP-2-induced ALP activity (Fig. 6B) and osteocalcin production (Fig. 6C). In addition, co-transfection of the mutant Flag-Smad5 (G419S) with wild-type Flag-Smad5 into C2C12 cells abolished BMP-2-induced serine phosphorylation of wild-type Flag-Smad5, showing dominant negative effects of the mutant Smad5 (G419S) on Smad5 activation (Fig. 6D). This dominantnegative effect is probably due to a competition for the binding to BMPIR between wild type and mutant Smad5 (Fig. 3A). The data demonstrate an inhibition of Smad5 phosphorylation by dominant-negative mutant Smad5 is associated with an inhibition of BMP-2-induced osteoblastic differentiation of C2C12 cells and suggest that the phosphorylation of Smad5 is necessary for the osteoblastic differentiation of C2C12 cells induced by BMP-2.
We also determined the biological role of DPC4 in the osteoblastic differentiation of C2C12 cells using Flag-DPC4⌬C, which is unable to associate with the phosphorylated Smad5 (Fig. 4, A and B). DPC4⌬C profoundly decreased BMP-2-induced ALP activity and osteocalcin production in C2C12 cells (Fig. 6, B and C). The dominant-negative effects of DPC4⌬C have been reported previously (15,18). We found that DPC4⌬C did not affect the phosphorylation of Smad5 but instead specifically blocked the association of Smad5 with intact DPC4 (Fig.  6E). Thus, the results suggest that the heterocomplex formation of phosphorylated Smad5 with DPC4 is also essential for the osteoblastic differentiation of C2C12 cells.

DISCUSSION
The bone-and cartilage-inducing effects of the BMPs have been studied extensively (2-5, 27, 32). Recent studies have markedly increased our understanding of BMP signal transduction pathways at molecular levels through identification of BMP-activated cytoplasmic signaling molecules including Smads and DPC4 (7,8). Nevertheless, the precise role of these signaling molecules in the bone-inducing effects of BMPs is not defined to date. In this study, we isolated human Smad5 and found that Smad5 was involved in BMP-2 signaling cascades, which mediate the bone-inducing effects of BMP-2. Smad5 was directly serine-phosphorylated by BMPIR through a physical interaction. The activated Smad5 subsequently formed a complex with DPC4, and this complex was then translocated to the nucleus. Overexpression of mutant Smad5 (G419S), which inhibits the phosphorylation of intact Smad5, blocked the BMP-2-induced osteoblastic differentiation of C2C12 cells. Furthermore, suppression of the complex formation of the Smad5 with DPC4 by overexpressing DPC4⌬C also blocked BMP-2-induced osteoblastic differentiation of C2C12 cells. Thus, interruption of BMP-2 signaling cascades by inhibiting Smad5 activation or interfering with the association between Smad5 and DPC4 abolished the osteogenic effects of BMP-2 on C2C12 cells. These results strongly suggest that both activation of Smad5 and following heterocomplex formation with DPC4 are critical to the BMP-2 signaling, which mediates the induction of the osteoblastic differentiation of the pluripotent mesenchymal cells C2C12. In contrast, TGF-␤ did not induce the phosphorylation of Smad5 in living cells and in vitro, constitutively active TGF-␤RI failed to phosphorylate Smad5, Smad5 was unable to bind to the activated TGF-␤RI, and a previous study has shown that TGF-␤ fails to promote osteoblastic differentiation of C2C12 cells (27). Collectively, these results suggest that Smad5, in addition to Smad1, is an intracellular molecule specifically involved in the BMP signaling. Thus, Smad5 is a new cytoplasmic signaling molecule of human Smad family members that mediates the osteogenic effects of BMP-2.
In conflict with our data, an earlier study has reported that mouse homologues for Smad5 and Smad1, dwarfin-C and dwarfin-A, respectively, are phosphorylated by TGF-␤, but not by BMP-2 (28). To examine whether this apparent discrepancy between that study and ours is due to a difference in experimental models, we performed identical experiments to those described here in NMuMg cells that were used in the previous study. In our hands, Smad5 was selectively activated by BMP-2 and TGF-␤ did not activate Smad5 in NMuMg cells. It is possible that Smad1 and Smad5 antibodies used in the previous study might recognize other Smads including Smad2 and Smad3 that are responsive to TGF-␤ (13)(14)(15)(16), since homology between these Smad members is high (7,8). Indeed, the authors raised the same possibility in the report. Furthermore, we experienced that the antibody we generated against GST-Smad5 recognized other Smad molecules including Smad1, Smad3, and DPC4. 2 However, it still remains possible that endogenous Smad5 behaves in different manners from that of transfected exogenous Smad5 or Smad1. Resolution of this issue awaits generation of specific antibodies for Smad5. Finally, our results suggest that there are no distinctive functional differences between Smad1 and Smad5 in BMP-2 signal transduction and BMP-2-induced osteoblastic differentiation in C2C12 cells. Dominant-negative Smad1, like dominant-negative Smad5, inhibited BMP-2-induced C2C12 differentiation into osteoblasts. Dominant-negative Smad5 interfered with Smad1 phosphorylation by BMP-2 stimulation. Whether Smad5 characterized herein has specific roles that are distinguishable from those of Smad1 in BMP-2 signaling that mediates the biological effects of BMP-2 is an important issue. However this is beyond the scope of the present study. Antibodies that specifically recognize Smad5 and development of additional experimental models to C2C12 cells may clarify this point.
In conclusion, we demonstrate an important role of Smad5 and DPC4 in the BMP-2-induced osteoblastic differentiation of the pluripotent mesenchymal stem cell C2C12.