Concerted Action of Smad and CREB-binding Protein Regulates Bone Morphogenetic Protein-2-stimulated Osteoblastic Colony-stimulating Factor-1 Expression*

Bone remodeling depends upon proper osteoblast and osteoclast function. Bone morphogenetic protein-2 (BMP-2) stimulates differentiation of osteoblasts from pluripotent precursors. Osteoclast formation depends on the concerted action of osteoblast-derived receptor activator of NF-κB ligand and colony-stimulating factor-1 (CSF-1). BMP-2 stimulates receptor activator of NF-κB ligand expression. However, the effect of BMP-2 on CSF-1 expression has not been studied. We investigated the role of BMP-2 in CSF-1 expression in osteogenic C2C12 cells. Incubation of C2C12 cells with BMP-2 supported osteoclastogenesis of spleen cells with a concomitant increase in expression of CSF-1 mRNA and protein. To determine the mechanism, we identified a BMP-responsive element between –627 bp and –509 bp in the CSF-1 promoter. DNase I footprint analysis revealed the presence of consensus Smad binding motif in this region. Electrophoretic mobility shift assay showed BMP-2-stimulated binding of proteins to this motif. Mutation of core sequence as well as its 5′- and 3′-flanking sequences abolished the DNA-protein interaction resulting in inhibition of CSF-1 transcription. Supershift analysis detects the presence of Smads 1, 5, and 4 and the transcriptional coactivator CREB-binding protein in the BMP-responsive element-protein complex. In addition, Smads 1 and 5 alone or in combination with Smad 4 increased CSF-1 transcription. Furthermore, CREB-binding protein markedly increased transcription of CSF-1. These data represent the first evidence that BMP-2 increases the osteoclastogenic CSF-1 expression by a transcriptional mechanism using the canonical Smad pathway and provide a mechanism for BMP-2-induced osteoclast differentiation.

Bone morphogenetic proteins (BMPs) 3 have been shown to induce osteogenic activities in animals (1). BMPs influence osteoblast differentiation from uncommitted progenitors (2)(3)(4)(5). Bone formation is a result of the orchestrated action of osteoclast and osteoblast cells. Recombinant BMP-2, BMP-4, BMP-7, and BMP-9 have potential to stimulate the osteoinductive cascade (6). Endochondral bone formation, stimulated by BMP-2, involves differentiation of mesenchymal cells into chondrocytes followed by cartilage formation, resorption of the cartilaginous structure by chondroclasts or osteoclasts, and finally replacement with osteoblasts to form mature bone (7)(8)(9). This suggests that BMP-2 is involved in osteoclastic bone resorption as much as it is involved in cartilage and bone formation.
The role of BMP-2 in cartilage and bone formation is well studied (9 -11). Osteoblasts and stromal cells are essentially involved in the differentiation of hematopoietic progenitors into osteoclasts (12)(13)(14)(15). CSF-1 is produced by osteoblasts and stromal cells and is essential for osteoclast development (16 -19). Another critical factor for osteoclast development and differentiation is RANKL, also expressed in osteoblast and stromal cells (20). CSF-1 receptors and receptors for RANKL, namely, RANK, are present in hematopoietic osteoclast progenitors (21,22). Together, soluble RANKL and CSF-1 can induce osteoclast development and differentiation starting from mouse hematopoietic cells and human peripheral blood mononuclear cells even in the absence of osteoblasts and stromal cells (23). These data indicate that CSF-1 and RANKL are the two critical factors in osteoclast development and differentiation that are produced by the osteoblasts.
C2C12 is a subclone of the C2 mouse myoblastic line (24,25). Under normal conditions these cells differentiate to form myotubes and produce differentiated muscle proteins. Treatment with recombinant BMP-2 can cause inhibition of muscle protein production in these cells with subsequent formation of differentiated osteoblasts (10). This property of C2C12 cells gives this cell line a unique pluripotent phenotype. We report that BMP-2-treated C2C12 cells stimulate TRAP-positive multinucleated osteoclast formation by inducing CSF-1 gene expression and secretion.
BMPs induce intracellular signal transduction by binding to two types of transmembrane receptors, BMP receptor type I (BMPR I) and type II (BMPR II) (26). Three type I and three type II BMP receptors have been described (26 -28). Binding of BMP-2 to the BMPR II induces recruitment of BMPR I and its phosphorylation in the GS domain (29). The activated BMPR I then phosphorylates downstream substrates (30,31). Smads are downstream molecules that are activated by BMP-like ligands. They exist as monomers in the absence of ligand stimulation. Receptor-regulated Smads (R-Smads) are directly phosphorylated by BMPR I and subsequently associate with common Smad (Co-Smad). Heteromeric complexes then translocate into the nucleus and regulate transcription of target genes in association with other nuclear proteins (30,31). BMPR I activates Smads 1, 5, and 8 (R-Smads), which in turn heterodimerize with Smad 4 (Co-Smad) to transduce BMP-mediated intracellular signals (30,31). Smad binding elements have been identified to contain a consensus sequence CAGACA (CAGA sequences) in the promoters of the genes that are activated by Smad signaling (32).
BMPR and their downstream signaling molecules are expressed in osteoclasts, isolated from rabbit long bones, and BMP-2 stimulates the bone-resorbing activity of these osteoclasts (33). BMPR I mRNA was found to be expressed in granulocyte macrophage-colony stimulating factor-supported hematopoietic blast cells (34). It was recently reported that bone marrow macrophages and purified mature osteoclasts express BMPRIA (35). BMP-2 can stimulate parathyroid hormone-induced osteoclast formation, which was inhibited by BMP-2 antagonist noggin (36). BMP-2 was found to increase RANKL expression in murine osteoclast cultures and enhance osteoclast formation by interleukin-1␣ (37). BMP-2 was also reported to increase osteoclast survival and differentiation supported by RANKL but did not stimulate the survival of osteoclasts by itself (35). In addition, BMP-2 did not increase the survival of osteoclasts supported by CSF-1 (35). This suggests that BMP-2 signaling might converge to activate CSF-1 that in turn increases osteoclast differentiation. Thus in the presence of abundant exogenous CSF-1, BMP-2 does not show any additional stimulatory effects on osteoclastogenesis. Our goal for this study was to analyze the role of BMP-2 in CSF-1 gene expression and subsequent osteoclast differentiation.
In this report, we show that BMP-2 induces expression of CSF-1 mRNA and protein in the pluripotent C2C12 cells that differentiate into osteoblasts upon BMP-2 stimulation. We identified a BMP-2-responsive element (BRE) in the CSF-1 promoter. We show that upon BMP-2 stimulation CBP associates with BMP-specific Smads. We provide evidence that BRE interacts with BMP-specific Smads and the transcription cofactor CBP to regulate CSF-1 expression. In addition, we show that BMP-2-stimulated C2C12 cells can support multinucleated TRAP-positive osteoclast formation in a coculture system. Taken together, our data provide a molecular mechanism of CSF-1 expression in response to BMP-2 to support osteoclast survival and differentiation. Furthermore, this is the first report of the regulation of CSF-1production in osteoblasts by BMP-2 during osteoclastogenesis. Also this report unravels the mechanism by which BMP-2 induces CSF-1 transcription.

EXPERIMENTAL PROCEDURES
Materials-Recombinant BMP-2 was a gift from Anthony Celeste (Wyeth Research, Cambridge, MA). Tissue culture reagents and Lipofectamine Plus were obtained from Invitrogen. pGL3 luciferase plasmids (pGL3 basic and pGL3 promoter) and dual luciferase assay kits were purchased from Promega Inc. (Madison, WI). A nuclear fraction extraction kit was purchased from Pierce. Antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Plasmids expressing Smads 1, 4, and the constitutively active BMP receptors were kindly provided by Dr. Miyazono (University of Tokyo, Japan). Smad 5 expression plasmid was kindly provided by Dr. Nishimura (Osaka University, Japan). CBP expression plasmid was a gift from Dr. Christopher Glass (University of California at San Diego). RNAzol B was purchased from Biotex Laboratories (Houston, TX). Nylon filters were bought from Schleicher and Schuell (Keene, NH). A CSF-1 ELISA assay kit was purchased from R & D systems (Minneapolis, MN). A nuclear extract preparation kit and poly(dI⅐dC) were purchased from Pierce, and DNase I was purchased from Sigma-Aldrich.
Cell Culture-C2C12 cells were purchased from American Type Culture Collection (Manassas, VA). C2C12 cells undergo osteoblastic differentiation in the presence of BMP-2. The cells were grown in DMEM medium containing 10% fetal calf serum and 1% penicillin and streptomycin. For driving osteoblastic differentiation the cells are routinely grown up to 70% confluency and treated with 300 ng/ml recombinant BMP-2 in serumfree DMEM for the indicated periods of time.
RNA Preparation and Northern Analysis-RNA was isolated from C2C12 cells cultured in 10-cm tissue culture plates in serum-free DMEM containing 300 ng/ml BMP-2 for 6, 12, 24, and 48 h. Control cells were cultured for similar time periods in serum-free DMEM without BMP-2 in the culture medium. At the indicated time points the cells were harvested, and RNA was isolated using 5 ml of RNAzol B followed by chloroform extraction and precipitation of RNA with isopropanol (38). 20 g of RNA was separated by electrophoresis according to size in denaturing agarose gels and transferred to Nylon filters. The filters were hybridized with CSF-1 and glyceraldehyde-3-phosphate dehydrogenase cDNA probes (38). Northern analysis was repeated three times with different RNA isolations.
DNaseI Footprint Assay-Nuclear extract was prepared from C2C12 and fetal rat calvarial osteoblast cells as described earlier following the protocol specified by the vendor (40). The DNA fragment spanning a region, Ϫ627 to Ϫ509 bp, of CSF-1 promoter was cloned into plasmid vector. Purified DNA fragment was end-labeled using [␣-32 P]ATP. The DNaseI footprinting was carried out as follows: C2C12 or fetal rat calvarial nuclear proteins were allowed to bind the labeled DNA fragment at 25°C for 15 min in the presence of the binding buffer (50 mM NaCl, 0.1 mM EDTA, 20 mM Hepes/potassium hydroxide (KOH), pH 7.5, 0.5 mM dithiothreitol, and 10% glycerol) and 2 g of poly(dI⅐dC). The DNaseI treatment was then carried out for 45 s using 5 l of 1:10,000-fold diluted DNaseI at 25°C in the presence of 5 l of 50 mM MgCl 2 and 5 l of 10 mM CaCl 2 . The DNaseI was immediately inactivated by using 100 l of stop solution (0.375% SDS, 15 mM EDTA, 100 mM NaCl, and 100 mM Tris⅐Cl, pH 7.6) followed by proteinase K digestion and phenolchloroform extraction, ethanol precipitation, and electrophoresis on a 7% sequencing gel. A DNaseI protection assay (footprint assay) was repeated four times.
Electrophoretic Mobility Shift Assay-C2C12 cells were serum-starved for 24 h and then treated with recombinant BMP-2 for an additional 24 h in serum-free DMEM. Nuclear extracts were prepared as described above. The oligonucleotide probe spanning the BMP-2 response element in CSF-1 promoter was prepared by annealing the oligonucleotide 5Ј-GGA GGG AGC AAG CCA ATC TGC AAA CCC CGG GTT AAG GGC G-3Ј with its complementary strand. The core consensus Smad binding sequence is underlined. The mutant oligonucleotides were designed by (a) mutating the core binding sequence ATC TGC to gga TtC (MT), (b) mutating DNA sequences 3Ј to the core binding site CGG GTT to att cGg (3Ј-MT), and (c) mutating 5Ј sequence to the core binding site GGG AGC to ata Atg (5Ј-MT). The double-stranded oligonucleotides were labeled using [␥-32 P]ATP and T4 polynucleotide kinase (40). The EMSA was performed using 5 g of nuclear extract as described previously (40 -42). For determining the specificity of DNA-protein interaction the nuclear extract was incubated with increasing amounts of cold oligonucleotide before the addition of the radioactive probe. Specific antibodies were added to the nuclear extract to perform supershift analysis followed by incubation with radioactive probe as described previously (40). EMSA performed with unlabeled oligonucleotides, antibodies, and mutant oligonucleotides were repeated at least three to five times each with different preparations of nuclear extracts.
Transfection and Luciferase Assay-Sequential 5Ј deletion constructs of CSF-1 promoter driving the firefly luciferase reporter gene were transfected in C2C12 cells using Lipofectamine Plus reagent as described previously (40). To correct transfection efficiency a cytomegalovirus promoter-driven Renilla luciferase plasmid was cotransfected in the transient transfection assays. Luciferase activities were assayed using a dual luciferase assay kit.
Site-directed Mutagenesis-The BRE sequences were mutated using a QuikChange II site-directed mutagenesis kit from Stratagene (La Jolla, CA). The oligonucleotide used for mutating the core CAGA sequences was 5Ј-GGA GGG AGC AAG CCA gga TtC AAA CCC CGG GTT AAG GGC G-3Ј. Lowercase letters indicate mutated sequences. The reactions were carried out exactly as directed by the vendor's instruction manual.
Coculture Assay for TRAP-positive Multinucleated Osteoclast Formation-Coculture of C2C12 and mouse spleen cells was done as described by Otsuka et al. (43). In brief, 10 4 C2C12 cells were cultured in 24-well tissue culture plates in ␣-minimal essential medium. After 24 h, 10 6 mouse spleen cells were added per well to C2C12 cells in the presence of 10 Ϫ8 M 1,25dihydroxyvitamin D 3 (BIOMOL, Plymouth Meeting, PA), 10 Ϫ7 M dexamethasone, and 300 ng/ml recombinant human BMP-2 or bovine serum albumin (for vehicle control). Fresh medium was replenished every 2 days. After 6 days adherent cells were fixed in 10% formalin for 5 min and then treated with 1:1 mixture of ethanol and acetone for 1 min. Cultures were then dried and stained for TRAP activity using acid phosphatase kit and Fast Garnet dye (Sigma). TRAP-positive multinucleated cells (with three or more nuclei) were photographed using a light microscope (Nikon, Japan). C2C12 cells were cultured with mouse spleen cells in the absence (Ϫ) or presence (ϩ) of 300 ng/ml BMP-2. After 6 days the cells were fixed and stained for TRAP-positive multinucleated osteoclasts as described under "Experimental Procedures." B, effect of BMP-2 on CSF-1 mRNA expression in C2C12 cells. Cells were incubated with 300 ng/ml BMP-2 in the absence of serum for 6, 12, 24, and 48 h. Total RNA was isolated and analyzed by Northern blotting using 32 P-labeled CSF-1 cDNA as probe as described under "Experimental Procedures." The same blot was stripped and probed with 32 P-labeled glyceraldehyde-3-phosphate dehydrogenase cDNA to demonstrate equal loading. C, effect of BMP-2 on CSF-1 protein secretion. C2C12 cells were incubated with BMP-2 as described above, and the conditioned medium was examined for the presence of CSF-1 protein using an ELISA as described under "Experimental Procedures." Mean Ϯ S.E. of triplicate determinations is shown. *, p Ͻ 0.05 versus control.

BMP-2-induced CSF-1 Expression Leads to Osteoclastogenesis
Coimmunoprecipitation Assay-C2C12 cells were stimulated with 300 ng/ml BMP-2 for 24 h or left untreated for control. Cell lysates were prepared using radioimmune precipitation buffer as described earlier (44). Supernatants of cell lysates were immunoprecipitated with antibody against CBP, and the proteins were separated by SDS gel electrophoresis, as described earlier (40). The immunoprecipitated proteins were then immunoblotted with antibody against Smad 1/Smad5 (Smad 1/5) (40,44).
Statistical Analysis of Data-To determine the significance of the data we have used analysis of variance following Student-Newman-Keuls comparison. Significance level was considered at a p value of Ͻ0.05.

BMP-2 Stimulates Osteoclast Differentiation, CSF-1 mRNA
Expression, and Protein Secretion-In vitro osteoclast differentiation requires coculturing hematopoietic cells and mouse osteoblast or stromal cells. BMP-2 has been shown to induce osteoblast phenotype in the pluripotent mesenchymal precursor cell line C2C12 (10). We tested whether C2C12 cells can support osteoclast formation in the presence of BMP-2 by coculturing C2C12 cells and mouse spleen cells in the presence of 1,25-dihydroxyvitamin D 3 and dexamethasone with or without BMP-2. C2C12 cells treated with BMP-2 supported TRAPpositive multinucleated osteoclast formation, whereas only huge myotube formation was detected in the absence of BMP-2 ( Fig. 1A). Osteoclast formation essentially depends on production of RANKL and CSF-1 by cocultured osteoblast cells. Recently BMP-2 has been shown to induce RANKL expression in C2C12 cells (43). In this study we tested the effect of BMP-2 on CSF-1 gene expression using C2C12 cells as our model system. We examined the effect of BMP-2 on CSF-1 expression. Incubation of C2C12 cells with BMP-2 increased expression of CSF-1 mRNA in a time-dependent manner (Fig. 1B). To confirm this observation, we tested CSF-1 protein abundance in the conditioned medium of C2C12 cells using an ELISA. Incubation of C2C12 cells with BMP-2 increased secretion of CSF-1 protein in the medium in a time-dependent manner (Fig. 1C). In complete accordance with the mRNA data, CSF-1 protein secretion was optimally increased by 24 h of BMP-2 treatment (3-fold) and remained elevated at 36 and 48 h following BMP-2 treatment. These data provide the first direct evidence that BMP-2 stimulates CSF-1 protein expression in osteoblasts.
Identification of a BMP-responsive Region in CSF-1 Promoter-To understand the mechanism underlying CSF-1 gene regulation, we first used a CSF-1 promoter (Ϫ627/ϩ183 bp), containing the transcription start site (ϩ1), driving luciferase reporter gene (referred to as Ϫ627-Luc), in transfection assays using C2C12 cells. This construct showed significant basal promoter activity ( Fig. 2A). To systematically study the effect of BMP-2 on CSF-1 transcription, we examined the effect of BMP-2 on sequential 5Ј deletion constructs of CSF-1 promoter-driven luciferase reporter gene expression (Fig. 2B). C2C12 cells were transfected with the constructs containing the progressive deletions (Fig. 2B) and incubated with BMP-2. BMP-2 increased the luciferase activity in Ϫ627/ϩ183 promoter (Fig.  2C). BMP-2 did not have significant effect on the reporter con-structs containing the progressive deletions of CSF-1 promoter, i.e. Ϫ509, Ϫ329, Ϫ152, Ϫ88, and Ϫ43 bp, respectively (Fig. 2C). These data indicate the presence of BRE within Ϫ627 and Ϫ509 bp in the CSF-1 promoter.
Identification of a BRE in the CSF-1 Promoter-To identify the DNA elements involved in BMP-2-induced transcriptional increase in CSF-1 promoter, we generated an oligonucleotide probe spanning Ϫ627 and Ϫ509 bp by PCR. Using this probe, DNase I protection analysis (also known as footprint analysis) was performed with nuclear extracts prepared from control and BMP-2-stimulated C2C12 cells. In this assay DNA-protein interactions are denoted by clear areas indicating that binding of specific proteins to this region have protected these DNA regions from being degraded by DNaseI. Two distinct areas of DNA-protein interactions were identified within the CSF-1 promoter fragment spanning Ϫ627 and Ϫ509 bp (Fig. 3A, lane 5, indicated by square brackets). The protected region contains the DNA sequence TCTG (reverse of CAGA, a consensus sequence for binding BMP-specific Smad-proteins). Similar results were obtained when nuclear extracts were examined from BMP-2-stimulated primary cultures of fetal rat calvarial osteoblasts (Fig. 3A, lane 4). Nuclear extracts isolated from cells in the absence of BMP-2 did not show any protected region (Fig. 3A, lanes 2 and 3). Furthermore, the control lane in absence of any nuclear protein did not show any DNAprotein interaction (DNaseI-protected region) as expected (Fig.  3A, lane 1).
In an attempt to characterize the BRE that was identified in the CSF-1 promoter, we performed EMSA using different oli- gonucleotides designed to encompass the BRE. First we tested an oligonucleotide containing the putative Smad-binding element (TCTG) spanning Ϫ534 to Ϫ509 bp (BRE) in the EMSA for specific DNA-protein interaction. Two protein-DNA complexes were identified in the EMSA using nuclear extracts isolated from BMP-2-stimulated C2C12 cells and the ␥-32 P-labeled BRE oligonucleotide (Fig. 3B, lane 1). Increasing concentrations of the unlabeled BRE oligonucleotide competed with the labeled probe for binding the nuclear proteins (Fig. 3B,   lanes 2-7 ). An excess of oligonucleotide containing DNA sequences for binding an unrelated transcription factor AP2, however, did not compete with the radiolabeled probe for nuclear protein binding (Fig. 3B, lane 8). These data indicate that DNA elements present between Ϫ534 and Ϫ509 bp specifically bind to proteins in nuclear extracts of BMP-2-stimulated cells.
The DNA sequence spanning Ϫ534 to Ϫ509 bp contains the consensus BMP-responsive Smad binding element, CAGA, in reverse direction (TCTG). When used in EMSA with nuclear extracts prepared from C2C12 cells stimulated with recombinant BMP-2, this oligonucleotide showed significant increase in binding of proteins compared with the nuclear extract prepared from the untreated control cells (Fig. 3C, compare lane 1 to lane  2), indicating that BMP-2 stimulates binding of these specific nuclear proteins to the BRE present in the CSF-1 promoter. To test the requirement of the core sequence in nuclear protein binding, we mutated the TCTG sequence in the oligonucleotide as described under "Experimental Procedures." This mutated oligonucleotide (MT) was used as a probe in EMSA. The mutated DNA probe showed drastically reduced level of protein binding (Fig. 3C, compare lane 3 with lane 2, indicated by the arrow). For many transcription factors, DNA sequences flanking the core element play an important role in complex formation. Therefore, we mutated the sequence 3Ј (3Ј-MT) and 5Ј (5Ј-MT) to the TCTG core sequence separately and used them as probes in EMSA. Both mutants were unable to produce the similar binding pattern as compared with the wild-type (WT) oligonucleotide probe (Fig. 3C, compare lanes  4 and 5 with lane 2, indicated by the arrow). These data indicate that, in addition to the core sequence, the flanking sequences of the CSF-1 promoter BRE are important for nuclear factor binding.
Identification of BMP-specific Smads in BRE-protein Complex-We established above that BRE present in CSF-1 promoter binds specifically to nuclear proteins in response to BMP-2 (Fig. 3C, indicated by arrow). BMP-2 uses the interme-  2 and 3) of BMP-2. 10 g of nuclear extracts was incubated with the probe followed by DNase I digestion as described under "Experimental Procedures." Lane 1 represents products with no nuclear extracts. DNase I protected (footprint) area is indicated by brackets on the right, and the TCTG sequence was identified from the sequence of the protected region. B, characterization of nuclear protein-BRE complex. A double-stranded oligonucleotide representing Ϫ534 to Ϫ509 bp, which contains the TCTG sequence (at Ϫ518), was end-labeled with [␥-32 P]ATP and used in EMSA in the presence of nuclear extracts isolated from BMP-2-treated C2C12 cells as described under "Experimental Procedures." For competition analysis, the nuclear extracts were incubated with increasing concentrations of cold BRE oligonucleotide (5-, 10-, 20-, 50-, 100-, and 200-fold in excess to the labeled probe) prior to incubation with the radiolabeled probe (lanes 2-7). For specificity, an oligonucleotide recognizing AP2 transcription factor was incubated with the nuclear extracts before addition of BRE probe (lane 8). The arrows on the left indicate the DNA-protein complexes. C, specificity of BRE-protein interaction. EMSA was performed with nuclear extracts isolated from C2C12 cells incubated in the presence and absence of BMP-2 as described under "Experimental Procedures." Wild-type and three mutant probes were used in EMSA as described under "Experimental Procedures." MT, the core sequence is mutated; 3Ј MT and 5Ј MT contain mutations in the 5Ј-and 3Ј-flanking sequence of the core TCTG. The arrow indicates BMP-stimulated BREprotein complex. D, identification of BMP-specific Smads in BRE-protein complex. Nuclear extracts isolated from control and BMP-2-treated C2C12 cells were used in EMSA as described under "Experimental Procedures." Nuclear extracts were incubated with specific antibodies against Smad 1, Smad 5, and Smad 4, as indicated, before incubation with the BRE probe. An antibody against Erk1/2 protein was used as nonspecific control.

BMP-2-induced CSF-1 Expression Leads to Osteoclastogenesis
diate signaling proteins Smad 1 and Smad 5, both of which form dimers with common Smad, Smad 4, to stimulate gene transcription (45). Therefore, we tested the involvement of these Smads in BRE binding using EMSA. Specific antibodies against Smad 1, Smad 5, and Smad 4 were used in EMSA in the presence of the BRE probe with nuclear extracts isolated from BMP-2-stimulated C2C12 cells. Antibody against Smad 1 inhibited both the BRE-protein complex formation (Fig. 3D, compare lane 2 with lane 1). Similarly, both Smad 5 and Smad 4 antibodies inhibited the BRE-protein complex formation in the presence of BMP-2 (Fig. 3D, compare lanes 3 and 4 with lane 1). A nonspecific antibody against Erk1/2 did not affect BRE binding (Fig. 3D, lane 5). These data for the first time demonstrate that both Smad 1 and Smad 5 are present in the DNA-protein complexes formed in response to BMP-2. Thus these transcription factors may induce CSF-1 expression in osteoblasts.
To test the effect of BMP-2 on the CSF-1 promoter activity when the core Smad binding element is mutated, we introduced the same mutations described above in the Ϫ627-LUC reporter plasmid. This mutant plasmid (MT) was transfected in C2C12 cells, and the luciferase activity was compared with that transfected with the WT Ϫ627-Luc plasmid, in the presence and absence of recombinant BMP-2. The mutations at the core Smad binding sequence of the BRE totally abolished transcriptional activation of CSF-1 promoter by BMP-2 (Fig. 4A). These data proved that the core Smad binding sequence is an absolute requirement for BMP-2-induced activation of CSF-1 promoter.

BMP-specific Smad 1 and Smad 5 Stimulate CSF-1 Promoter Activity-
We established that BMP-2 stimulates transcription of CSF-1 from the Ϫ627-bp promoter (Fig. 2C). To investigate the involvement of BMP-2-specific Smads in this transcriptional activation, we transfected C2C12 cells with the Ϫ627-Luc construct along with expression vectors encoding the BMP receptor-specific Smad 1 or Smad 5 alone or in combination with the Co-Smad, Smad 4. Smad 1 increased CSF-1 transcriptional activity by 2.5-fold as compared with the vector-transfected cells (Fig.  4B). Similarly, Smad 5 and Smad 4 significantly increased CSF-1 transcriptional activity by 3.5-fold (Fig. 4B). Cotransfection of Smad 4 with Smad 1 had a modest effect as compared with the Smad 4 alone (Fig. 4B). However, Smad 4 exerted more significant effect on Smad 5-induced transcriptional response (6-fold) (Fig.  4B). These data provide the functional evidence demonstrating that BMP-specific Smad 1 and Smad 5 regulate CSF-1 gene expression.
For signaling, BMP-2 utilizes two type I receptors, IA and IB (45). In an attempt to examine the cooperation of BMP-specific Smads with the BMP receptors type I, we cotransfected Ϫ627-Luc plasmid with constitutively active BMPR IA or BMPR IB. Both of these receptors increased transcriptional activity of the CSF-1 promoter (Fig. 4C). Cotransfection of Smad 1 or Smad 4 increased both BMPR IA-and BMPR IB-induced transcriptional activity to the comparable extent (Fig. 4C). Similarly, cotransfection of Smad 5 also increased both BMPR IA-and IB-mediated CSF-1 gene transcription (data not shown). These data indicate that BMPspecific Smads cooperate with BMP receptor to induce transcription of CSF-1 gene.

BMP-2-induced CSF-1 Expression Leads to Osteoclastogenesis
mid in the C2C12 cells along with Ϫ627-Luc reporter plasmid. Smad 6 significantly inhibited basal and BMP-2-induced CSF-1 promoter activity (Fig. 4D). These data confirmed that CSF-1 gene expression is controlled by BMP-2 through Smads 1, 5, and 4. (46). To examine the involvement of CBP in CSF-1 transcription, we first performed EMSA using the CSF-1 promoter-derived BRE as probe in the presence of a CBP-specific antibody. We have shown above that BMP-2 stimulates formation of the Smad-BRE complex (Fig. 3D). Incubation of nuclear extracts with CBP-specific antibody inhibited this Smad-BRE complex formation (Fig. 5A, compare lane 3 with lane 2). These results conclusively indicate that CBP is recruited in the BMP-2-induced Smad-BRE complex. These data also suggest that CBP may induce transcription of the CSF-1 gene. To examine this notion, we transfected an expression vector containing CBP and Ϫ627-Luc into C2C12 cells. CBP significantly increased the transcriptional activity of CSF-1 promoter as compared with the vector-transfected cells (Fig.  5B). To confirm the role of CBP in BMP-2-induced CSF-1 gene transcription, we cotransfected constitutively active BMPR type IA or type IB along with CBP. BMPR IA did not have any additive effect on CBP-induced CSF-1 transcriptional activation (Fig. 5B). However, constitutively active BMPR IB significantly increased CBP-dependent transcription (Fig. 5B). These data indicate that the transcriptional coactivator CBP regulates CSF-1 gene transcription in C2C12 cells.

CBP Acts as a Transcriptional Coactivator in BMP-2-induced CSF-1 Expression-Transcriptional coactivators cooperate with the transcription factors to increase transcription
To confirm that the transcriptional induction of CSF-1 promoter is indeed dependent on recruitment of CBP we cotransfected an expression plasmid encoding adenovirus E1A promoter along with Ϫ627-Luc and CBP expression plasmid into C2C12 cells. Adenovirus E1A completely abrogated basal and CBP-induced CSF-1 promoter activity (Fig. 5C).

BMP-2 Induces CBP and Smads Interaction in C2C12 Cells-We
have established that the BRE present in CSF-1 promoter interacts with Smads 1, 5, and 4 and CBP. To test whether Smads 1 and 5 associate with CBP upon BMP-2 stimulation, we immunoprecipitated CBP from C2C12 cells stimulated with BMP-2. The immunoprecipitated proteins were then immunoblotted with an antibody that recognizes both Smads 1 and 5 to test whether Smads 1 and 5 associate with CBP upon BMP-2 stimulation. BMP-2 stimulated association of Smads 1/5 with CBP in C2C12 cells (Fig. 5D, upper left panel, compare lanes 1 and 2). The same immunoblot was probed with CBP antibody to demonstrate that an equal amount of CBP protein was immunoprecipitated from both the control and BMP-2-treated C2C12 cells (Fig. 5D, lower left panel, lanes 1 and 2). These data suggest that BMP-2 can induce Smads 1 and 5 to interact with CBP and thus stimulate recruitment of this complex to the BRE of CSF-1 promoter. BMP-2 signaling involves phosphorylation of Smads 1 and 5, which subsequently translocate into the nucleus to carry out transcriptional regulation by associating with other transcription factors. We tested whether Smads 1 and 5 are phosphorylated when associated with CBP in BMP-2-stimulated C2C12 cells. CBP immunoprecipitates from lysates of C2C12 cells were immunoblotted with phospho-Smad1/5 antibody. BMP-2-stimulated association of phosphorylated Smads 1 and 5 with CBP (Fig. 5D,  right upper panel, compare lane 2 with lane 1). These data indicate that BMP-2-stimulated phosphorylated Smad1/5 recruits CBP in the nucleus to induce CSF-1 gene transcription. A, identification of CBP in the BRE-protein complex. Nuclear extracts isolated from C2C12 cells were used in EMSA in the presence of 32 P-labeled BRE oligonucleotide described under "Experimental Procedures." Nuclear extracts prepared from BMP-2-stimulated C2C12 cells were incubated with an antibody specific for CBP or Erk1 before incubation with the radiolabeled probe. Arrows indicate protein-BRE complexes that contain CBP. B, BMP receptor cooperates with CBP in CSF-1 transcription. C2C12 cells were transfected with Ϫ627-Luc reporter construct along with expression vectors expressing CBP or in combination with constitutively active BMP receptors IA and IB. The luciferase activity was determined as described under "Experimental Procedures." Mean Ϯ S.E. of triplicate determination is shown. *, p Ͻ 0.05 versus control; **, p Ͻ 0.005 versus CBP alone. C, adenovirus E1A inhibits CSF-1 transcriptional activation by CBP. C2C12 cells were transfected with Ϫ627-Luc plasmid along with CBP expression plasmid and vector or E1A expression plasmid. The luciferase activities were determined as described under "Experimental Procedures." Mean Ϯ S.E. of triplicate determinations is shown. *, p Ͻ 0.05 versus control; **, p Ͻ 0.005 versus CBP alone. D, BMP-2 stimulates association of CBP with phosphorylated-Smads 1/5. C2C12 cells were treated with BMP-2 (ϩ) or were left untreated (Ϫ) for 24 h before preparing cleared cell lysates for immunoprecipitation with an antibody against CBP. The immunoprecipitated proteins were separated by SDS-polyacrylamide gel electrophoresis and immunoblotted with antibodies against Smads 1 and 5 (left upper panel ) or phosphorylated Smads 1/5 (right upper panel ). The same gels were immunoblotted with the anti-CBP antibody after stripping to show equal loading (lower panels).

BMP-2-induced CSF-1 Expression Leads to Osteoclastogenesis
BMP-2 Response Element from CSF-1 Gene Can Confer BMP-2 Sensitivity to Heterologous Promoters-Finally, to test whether the BRE identified and characterized in this study works universally for any promoters, we cloned 3ϫ BRE sequences at 5Ј of the SV40 promoter driving firefly luciferase reporter enzyme (pGL3 promoter). BMP-2 treatment of the C2C12 cells transfected with this plasmid showed 2-fold increase in luciferase activity (Fig. 6) indicating that these BRE sequences are capable of inducing BMP-2-mediated transcriptional activation from a heterologous promoter.

DISCUSSION
These studies represent the first direct demonstration that BMP-2 stimulates CSF-1 expression in osteoblasts by a transcriptional mechanism. Our data also provide evidence for the presence of BRE in the CSF-1 promoter. We show that BMPspecific Smad 1 and Smad 5 together with a common Smad, Smad 4, and the transcriptional coactivator CBP, interact with the BRE. Additionally, for the first time, we demonstrate that BMP-specific Smads regulate CSF-1 gene transcription. These results emphasize a mechanism in osteoblasts for CSF-1 protein expression, which in turn stimulates osteoclast progenitors to differentiate into mature osteoclasts (Fig. 7).
CSF-1 is essential for differentiating bone marrow precursor cells into bone-resorbing osteoclasts (16). Functional deletion of CSF-1 in a mutant mouse model (op/op) results in an osteopetrotic phenotype due to lack of osteoclast formation and impaired bone resorption (16). Alternative splicing of CSF-1 mRNA leads to synthesis of a soluble and a membrane-bound form of CSF-1 by osteoblasts and stromal cells (47,48). By targeting soluble CSF-1 expression to bone in op/op transgenic FIGURE 6. CSF-1-derived BRE can confer BMP-2 responsiveness to SV40 promoter. An oligonucleotide was designed to contain three copies of BRE sequences (3X BRE ) that have been used for EMSA experiments. This oligonucleotide was cloned 5Ј of the pGL3 promoter containing the SV40 basal promoter (3X BRE/pGL3 Promoter). Both pGL3 promoter and 3ϫ BRE/pGL3 promoter were transfected into C2C12 cells. Transiently transfected cells were treated with BMP-2 for 48 h. Luciferase activity was determined as described under "Experimental Procedures." Mean Ϯ S.E. of triplicate measurements is shown. *, p Ͻ 0.05 versus pGL3 Promoter in the presence of BMP-2. FIGURE 7. Schematic representation of BMP-2-stimulated osteoclast formation. The mechanism of BMP-2-induced CSF-1 gene expression as described in this report is diagrammatically represented. The RANKL expression by osteoblast cells is also essential for osteoclast formation and BMP-2 has been reported to induce its expression in the C2C12 cells (43).

BMP-2-induced CSF-1 Expression Leads to Osteoclastogenesis
mice we have recently demonstrated that the soluble form of CSF-1 is sufficient to drive osteoclast development in these CSF-1-deficient mice (39). Our study is aimed at understanding the mechanism of CSF-1 gene regulation in osteoblast cells. Previous studies have shown that interleukin-1 and tumor necrosis factor are potent stimulators of CSF-1 gene expression (49). CSF-1 is released from osteoblasts constitutively and in response to parathyroid hormone and parathyroid hormonerelated protein (50). Another interesting report shows the involvement of hydrogen peroxide in transforming growth factor-␤-induced CSF-1 expression (51). Very little has been reported on the transcriptional regulation of CSF-1 gene except for the recent reports elucidating the involvement of NFB and the NF1/CTF family of transcription factors in CSF-1 expression (52)(53)(54).
The importance of the BMP-2-signaling pathway has been implicated for osteoblastic differentiation from immature mesenchymal cells and myogenic cells (10,22). Recently the role of BMP-2 in osteoclast differentiation has been reported by several groups of investigators (35)(36)(37)43). These studies emphasized the role of BMP-2 in RANKL-induced osteoclastogenesis and did not show any effect of BMP-2 on CSF-1 gene expression. We showed that C2C12 myogenic cells, which could not support osteoclast formation in a coculture assay in the absence of BMP-2, generated TRAP-positive multinucleated osteoclasts when induced with recombinant BMP-2 (Fig. 1A). We provide the first evidence that BMP-2 stimulates CSF-1 mRNA expression in osteoblast cells (Fig. 1, B and C). Using CSF-1specific ELISA, we confirmed this observation that BMP-2 increased CSF-1 protein secretion from C2C12 cells in the medium (Fig. 1D). We have delineated the BMP-2-responsive region of CSF-1 promoter within Ϫ627 and Ϫ509 bp by analyzing CSF-1 promoter 5Ј deletion constructs (Fig. 2C). BMP-2 binding activates its cell-surface receptor to recruit and activate Smad proteins. Activated Smads in turn localize to the nucleus where they join the multisubunit complex of transcription activators and interact with DNA regions in the promoter for inducing gene transcription (55). The Smad binding elements reported so far are usually GC-rich and contain a 5-bp consensus of CAGAC that interacts with the MH1 domain of R-Smads and Co-Smads (56 -58). DNaseI footprint analysis of DNA spanning Ϫ627 bp and Ϫ509 bp of CSF-1 promoter identified a BRE containing a reverse CAGA sequence, TCTG (Fig. 3A), that has been implicated for Smad-mediated transcriptional activity (32,59). EMSAs revealed formation of two protein-DNA complexes in response to BMP-2 (Fig. 3C). Use of Smadspecific antibodies showed the presence of the BMP-specific Smads, Smad 1 and Smad 5, and the common Smad, Smad 4, in these DNA-protein complexes (Fig. 3D). Thus, the difference in mobility containing the same Smads in the two DNAprotein complexes may result from recruitment of other proteins necessary for expression of CSF-1 in response to BMP-2 in C2C12 cells. In fact, comparison of the Smad binding elements of relatively few BMP-responsive promoters characterized in vertebrates points to the fact that the activation of the target genes by BMP-2 are not solely dependent on Smads binding to a short Smad binding element sequence. EMSA using individual oligonucleotides mutated either in the core CAGA sequence or at the flanking 5Ј and 3Ј sequences demonstrated that the flanking DNA sequences are also required along with the core CAGA sequence for effectively binding Smad proteins in BMP-2-stimulated C2C12 cells (Fig. 3C). However, mutation of the core TCTG sequence in the CSF-1-promoter-reporter construct totally abolished BMP-2 response of CSF-1 promoter (Fig. 3C). These data indicate that the core Smad binding sequence is necessary for BMP-2-induced CSF-1 gene expression. Expression of BMP-specific R-Smads 1 and 5 as well as co-Smad 4 induced transcriptional activation of CSF-1 promoter (Fig. 4, B and C). These data provide a functional action of Smads that lead to CSF-1 mRNA and protein expression in osteoblast cells in response to BMP-2 treatment. The involvement of Smad 1 and Smad 5 in CSF-1 gene expression is further confirmed when Smad 6 expression totally inhibited BMP-2induced CSF-1 gene expression (Fig. 4D).
Efficient gene transcription is dependent on transcription initiation by RNA polymerase II and cooperative interactions of specific transcription factors with promoter and enhancer elements (60). CBP was initially identified as a protein interacting with the protein kinase A-phosphorylated form of cAMP response element-binding protein (CREB) (61). In addition to being transcriptional coactivator for CREB, it also interacts with many other transcription factors like c-Jun, c-Fos, c-Myb, nuclear hormone receptors, and MyoD (62,63). CBP itself has histone acetyl transferase activity, which suggests that it can disrupt the chromatin structure and thus can activate transcription (64,65). Smads, as GAL-4 fusion proteins, cooperate with CBP in increasing reporter transcription (66,67). In this study we show that CBP binds to the BREs in CSF-1 promoter (Fig. 5A) and can stimulate its transcriptional activation (Fig.  5B). The involvement of CBP in BMP-2-induced CSF-1 gene transcription was further supported by our observation that activation of CSF-1 gene transcription by CBP was augmented by cotransfection of the cells with constitutively active BMP receptor I (Fig. 5B). The early region 1A (E1A) protein coded by adenovirus 5 interacts with CBP, inhibiting its histone acetyl transferase activity, and prevents its interaction with transcription factors, thus inhibiting transcriptional activation involving CBP (68,69). Expression of E1A protein in C2C12 cells transfected with Ϫ627-Luc and CBP expression plasmids totally abolished CSF-1 promoter activity confirming the involvement of CBP in CSF-1 gene expression (Fig. 5C) suggesting that CBP is an essential transcriptional coactivator for CSF-1 gene expression. This is the first report to demonstrate that Smads 1 and 5, in association with the transcriptional coactivator CBP, regulate CSF-1 gene expression through a BRE. Next, by coimmunoprecipitation assay using antibodies against CBP, Smads1/5, and phospho-Smads1/5, we showed that BMP-2 induces physical association of these transcriptional activators that are then recruited on the BRE to induce CSF-1 gene expression (Fig. 5D). Finally, we confirmed that the BRE from CSF-1 described in this report can function independently and can confer BMP-2 responsiveness to otherwise unresponsive SV40 minimal promoter (Fig. 6).
In summary, our data demonstrate that CSF-1 gene expression is positively regulated by BMP-2 in osteoblast cells. Our data for the first time show that BMP-2 modulates transcrip-tion of the CSF-1 gene by signaling through BMPRI that in turn activate Smads 1 and 5 resulting in their binding to the CSF-1 promoter BRE along with the transcriptional coactivator CBP to facilitate CSF-1 gene expression and protein production. The stimulation of BRE by BMP-2 is not restricted to the CSF-1 promoter. We showed that BMP-2 stimulates expression of firefly luciferase gene driven by the SV40 promoter containing three tandem repeats of BRE (3ϫ BRE). We provide the functional significance of stimulation of CSF-1 by BMP-2 by demonstrating that C2C12 cells can support osteoclast formation from progenitor cells only in the presence of BMP-2. This study provides the first evidence for a mechanism of regulation of an osteoclast growth and differentiation factor CSF-1 by an essential bone-remodeling growth factor BMP-2 in osteoblast cells. A schematic diagram of our findings together with the known facts of osteoclast differentiation is provided in Fig. 7. CSF-1 is produced in osteoblast cells, and upon secretion it exerts its effect on osteoclast progenitor cells expressing receptors for CSF-1. This understanding of the regulatory mechanism for CSF-1 gene expression can be applied for future design of therapeutic molecules for controlling osteoclast function as is often desired in diseases such as osteoporosis and other bone abnormalities.