Transcriptional mechanisms of bone morphogenetic protein-induced osteoprotegrin gene expression.

Osteoprotegerin (OPG), an osteoblast-secreted decoy receptor, specifically binds to osteoclast differentiation factor and inhibits osteoclast maturation. Members of the transforming growth factor-beta superfamily including bone morphogenetic proteins (BMPs) stimulate OPG mRNA expression. In this study, we have characterized the transcription mechanism of BMP-induced OPG gene expression. Transfection of Smad1 and a constitutively active BMP type IA receptor ALK3 (Q233) stimulated the OPG promoter. Deletion analysis of the OPG promoter identified two Hoxc-8 binding sites that respond to BMP stimulation. Glutathione S-transferase-Hoxc-8 protein binds to these two Hox sites specifically. Consistent with the transfection results of the native promoter, ALK3 or Smad1 linker region, which interacts with Hoxc-8, stimulated the activation of the reporter construct with the two Hox sites. Overexpression of Hoxc-8 inhibited the induced promoter activity. When the two Hox binding sites were mutated, ALK3 or Smad1 linker region no longer activated the transcription. Importantly, Smad1 linker region induced both OPG promoter activity and endogenous OPG protein expression in 2T3 osteoblastic cells. The medium from cells transfected with Smad1 linker region expression plasmid effectively inhibited osteoclastogenesis. Collectively, our data indicate that Hox sites mediate both OPG promoter construct activity and endogenous OPG gene expression in response to BMP stimulation.

Bone remodeling mainly depends on a balance between bone resorption by osteoclasts and bone formation by osteoblasts (1,2). Cell to cell contact between osteoblasts and osteoclasts is critical for osteoclast differentiation and maturation (3,4). It is clear that osteoprotegrin (OPG) 1 and osteoprotegrin ligand (OPGL), both produced by osteoblasts, determine the destiny of osteoclast progenitors with the involvement of the osteoclast surface receptor, which is called receptor activator of NF-B. Thus, OPGL and its receptors mediate the communication between osteoblasts and osteoclasts and maintain the balanced bone metabolism (5,6). OPGL, also known as osteoclast differ-entiation factor, activates osteoclast differentiation. OPG is a secreted receptor of the tumor necrosis factor receptor family without transmembrane domain. It binds directly to OPGL on the osteoblasts/stromal cell surface, inhibiting OPGL-mediated osteoclastogenesis (7,8). OPG has also been shown to inhibit bone resorption both in vitro and in vivo (9,10). OPG-deficient mice exhibited severe osteopenia due to the accelerated bone resorption (11,12). In addition, OPG prevents bone and cartilage loss in a T-cell-dependent arthritis model (13). OPG mRNA was localized mainly to the cartilaginous aspects of developing bone, arteries, skin, and gastrointestinal tract (7), although OPG is also expressed in a variety of tissues, including the kidney, lung, liver, brain, placenta, heart, and hematopoietic cell lines (14,15).
OPG mRNA expression is regulated by many bone metabolic regulators, such as IL-1 (16), tumor necrosis factors (17), TGF-␤ (18,19), vitamin D (20), and prostaglandin E 2 (21) in osteoblastic cells. In particular, bone morphogenetic proteins (BMPs), osteotropic growth factors, significantly induce OPG expression both at the mRNA and the protein level in human fetal osteoblast cells (20). BMPs induce de novo bone formation in postfetal life through the process of intramembranous and endochondrial ossification. BMP-2, BMP-4, and BMP-7 are the most potent growth factors that promote new cartilage and bone formation (22)(23)(24)(25). The BMPs initiate the cascades of transcription program for mesenchymal stem cell differentiation into osteoblasts and chondrocytes. Studies of the mechanisms of BMP-induced OPG gene transcription should provide us with a better understanding of the roles of BMPs and OPG in skeletal formation and bone remodeling.
BMPs form the largest group in the TGF-␤ superfamily. The TGF-␤ signal transduction pathway has been elucidated in recent years (26,27). Like other members of the superfamily, BMPs bind to two types of transmembrane receptors, known as type I and type II, with serine/threonine kinase activity (28). Upon binding to BMP ligands, type I receptors phosphorylate a family of conserved downstream Smad proteins. Smad1 is the downstream effector of BMP signaling (29,30). The phosphorylated Smad1 interacts with Smad4, which then translocates into the nucleus and regulates gene transcription by associating with a nuclear transcription factor (31) or by binding directly to DNA (32). Previously, we reported that Smad1 interacts with Hoxc-8, a homeodomain transcription factor, and dislodges Hoxc-8 from its DNA binding element, resulting in the induction of gene expression (33). The interaction domains of Smad1 with Hoxc-8 have been characterized. Overexpression of these domains in osteoblast precursors stimulated osteoblast differentiation-related gene expression and led to mineralized bone matrix formation (34). Hox binding elements widely exist in promoters of osteoblast differentiation marker genes, especially those that rapidly respond to BMP stimulation, such as OPG (20), BMP-4 (35), and osteonectin (36). It appears that the interaction between Smad1 and Hox transcription factors may represent the major initiation mechanism of osteoblast differentiation in the BMP signaling pathway.
In the present study, we reported that constitutively active BMP type IA receptor ALK3 (Q233) stimulates OPG promoter activity. By characterization of the OPG promoter, we found that the two Hox binding sites are essential for OPG transcription activation induced by BMP. Most importantly, overexpression of the Smad1 linker region, which interacts with Hoxc-8, induced both OPG promoter activity and endogenous OPG protein expression in 2T3 osteoblastic cells. Finally, the medium from the cells transfected with the Smad1 linker region expression plasmids effectively inhibited OPGL-induced osteoclastogenesis.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-The OPG promoter from Ϫ1104 to ϩ59 relative to the transcription start site was amplified by polymerase chain reaction from human genomic DNA and cloned into MluI and BglII sites of the pGL3-basic vector (Promega) to generate a luciferase reporter construct (OPG-1). A deletion series (OPG-2, Ϫ999 to ϩ59; OPG-3, Ϫ937 to ϩ59; OPG-4, Ϫ700 to ϩ59; OPG-5, Ϫ564 to ϩ59) was also cloned into MluI and BglII sites of the pGL3-basic vector. OPG-SV40 (Ϫ1104 to Ϫ624) reporter bearing the two Hoxc-8 binding sites was constructed using the same strategy but was put into the pGL3-control vector (Promega). The mutants, mOPG-SV40, m1OPG-SV40, and m2OPG-SV40, were generated by replacing both, the first, or the second Hox recognition core sequences TAAT with TCCT (mutated nucleotides are in boldface type), respectively, in the OPG promoter region (Ϫ1104 to Ϫ624) using polymerase chain reaction. A Smad1C mutant construct (m-Smad1C) was generated by deletion of the Hoxc-8 interaction domain (amino acids 148 -191) of Smad1C (amino acids 148 -278).
Transfection and Luciferase Assay-C3H10T1/2 mesenchymal cells (5 ϫ 10 4 cells/12-well plate) were transiently transfected using Tfx-50 reagent (33) with a total of 0.3 g of luciferase reporter plasmid (OPG1, OPG-SV40, mOPG-SV40, m1OPG-SV40, or m2OPG-SV40) and different expression plasmids as indicated. Total DNA was kept constant by the addition of PCDNA3 plasmid. Luciferase activities were assayed 42 h post-transfection using the Dual Luciferase assay kit (Promega) according to the manufacturer's directions. Luciferase values shown in the figures are representative of transfection experiments performed in triplicate in at least three independent experiments.
Purification of GST Fusion Proteins-A GST fusion construct of GST-Hoxc-8 was generated as described in our previous study (33). The expression and purification of the fusion proteins were performed as described elsewhere (37).
Immunoblotting-2T3 cells were transiently transfected with 5 g of Smad 1C-NLS, the linker region of full-length Smad1, which is identified as the Hoxc-8 interaction domain (34). Cell lysate was collected at different times (0, 24, 33, 40, and 48 h), fractionated by 12% SDSpolyacrylamide gel electrophoresis, and electrophoretically transferred to nitrocellulose filters. Filters were blocked at room temperature for 1 h in 3% nonfat milk in phosphate-buffered saline containing 0.05% Tween 20 and then incubated for Ͼ1 h at room temperature with 1:100 anti-OPG polyclonal antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). After washing, the blots were incubated with a horseradish peroxidase-conjugated 1:1000 anti-goat antibody (Santa Cruz Biotechnology). The blots were washed five times and visualized by enhanced chemiluminescence kit (Amersham Pharmacia Biotech) and exposed to x-ray film. The specificity of the blot was examined by incubating 1 g/ml excess OPG blocking peptide (Santa Cruz Biotechnology) with primary antibody.
In Vitro Generation of Osteoclasts-Mouse bone marrow macrophages (BMMs) were isolated from 8-week mice (C57/B1) (38). To generate osteoclasts, BMMs were cultured in ␣-minimal essential medium containing 10% fetal bovine serum in 24-well plates (1 ϫ 10 5 cells/well) in the presence of murine recombinant M-CSF (10 ng/ml) from R & D Systems (Minneapolis, MN) and GST-OPGL (50 ng/ml) (39). Culturing medium was changed every 3 days, and M-CSF and GST-OPGL were freshly added to the culture after the medium change. Osteoclasts started to form at day 4, and the cultures were stained for tartrateresistant phosphatase (TRAP) activity using a commercial kit from Sigma at day 5. In selected wells, 25% of the culturing medium was the conditioned medium from 2T3 cells transfected with plasmid Smad1C or empty vector (PCDNA3). (20). To investigate the mechanisms of the BMP-activated OPG gene transcription, we have cloned an approximately 1-kilobase pair DNA fragment of the OPG promoter (Ϫ1104/ϩ59). The OPG promoter was ligated into a luciferase reporter expression vector to examine its responsiveness to BMP stimulation. Transient transfection of this construct in C3H10T1/2 cells with Smad1 and a constitutively active type IA BMP receptor (ALK3) expression plasmid resulted in a significant increase in luciferase activity. We then attempted to localize the BMP response elements by promoter deletion analysis. As shown in Fig. 1, a series of OPG 5Ј-flanking region deletion constructs were generated. These constructs were transfected in C3H10T1/2 cells and challenged with BMP stimulation. Deletion of the DNA fragment from Ϫ1104 to Ϫ938 abrogated BMP-induced transcription activity (Fig. 1). Further deletion did not significantly change luciferase activity. Analysis of the DNA sequence of this promoter fragment (Ϫ1104 to Ϫ938) revealed two putative Hox binding sites, located at Ϫ1003 to Ϫ1000 or Ϫ943 to Ϫ940, respectively. We have shown previously that Smad1 interaction with Hoxc-8 dislodges Hoxc-8 from its DNA binding element to activate osteopontin gene expression (33). We therefore examined whether the interaction between Hoxc-8 and Smad1 is also the transcription mechanism in response to BMP stimulation. Hoxc-8 was cotransfected with Smad1 and ALK3 in C3H10T1/2 cells. As expected, Hoxc-8 significantly inhibited BMP-induced OPG transcription (Fig. 1), suggesting that Hoxc-8 mediates BMP stimulation. This result indicates that the promoter fragment bearing two Hox binding sites may be responsible for the BMP-induced transcription.

Identification of BMP Response Elements in the OPG Promoter-BMPs induce OPG gene expression
To verify that the promoter region with two Hox sites function as BMP response elements, we cloned the fragment (Ϫ1104/Ϫ624) containing the two Hox binding sites into the pGL3-control luciferase reporter vector under the control of the SV40 promoter (OPG-SV40) ( Fig. 2A). When this construct was cotransfected in C3H10T1/2 cells with full-length Smad1 and ALK3 (Q233D), the transcription activity was significantly stimulated (Fig. 2B). Furthermore, the core nucleotides TAAT of these two Hox binding sites were mutated to TCCT, designated as mOPG-SV40. As expected, transfection of the mutant construct completely abolished the Smad1-and ALK3-induced reporter activity and eliminated Hoxc-8 inhibition in C3H10T1/2 cells. Effects of Hoxc-8 alone on both wild type and mutant promoter constructs were also examined. Hoxc-8 sig-nificantly inhibited basal activity in the wild type promoter (Fig. 2B). Two Hox sites were also mutated individually to assess the relative contribution to the BMP stimulation ( Fig.  2A). Transfection analysis of these two Hox mutant constructs suggests that mutation of either of the Hox sites abolish Hoxc-8-mediated inhibition (Fig. 2C). However, the OPG-Hox1 (Ϫ1003/Ϫ1000) appears to contribute more in response to BMP stimulation. These results indicate that the two Hox binding sites are essential to BMP-induced OPG promoter transcription.
Hoxc-8 Binds to OPG Promoter Specifically-To examine whether Hoxc-8 binds to the two consensus Hox binding sites, we performed EMSA. Isotope 32 P-labeled DNA fragments, containing either one or both of the Hox binding sites, were used in incubation with purified GST-Hoxc-8 (Fig. 3A). When a longer fragment comprising both Hox binding sites was used, Hoxc-8 protein formed a retarded band (Fig. 3B). Thus, Hoxc-8 protein binds to the DNA fragment. Then, two shorter fragments (Ϫ1013/Ϫ994 and Ϫ956/Ϫ932), each containing a Hox binding site, were used, respectively. As seen in Fig. 3C, Hoxc-8 binds   FIG. 1. Smad1 and ALK3 (Q233D) activate the OPG promoter. Left, schematic description of the constructs used in the transfection assay on the right. Nucleotides are numbered relative to the transcription start site. Right, BMP stimulates OPG promoter expression. C3H10T1/2 cells plated at a density of 5 ϫ 10 4 cells/well in 12-well plates for transfections. A deletion series of OPG promoter constructs (0.3 g) were cotransfected in C3H10T1/2 cells with or without Smad1 (0.1 g) and ALK3 (0.025 g) plasmids or in combination with Hoxc-8 plasmid. Transfected cells were harvested and lysed for 42 h. The luciferase activity was measured and normalized to Renilla luciferase level as internal controls. Data are shown with the mean Ϯ S.D. of triplicates.

FIG. 2. Hox binding sites mediate BMP stimulation.
A, schematic description of the constructs used in the transfection assay in B. Filled boxes indicate putative Hox binding sites (TAAT). Boxes marked ϫ indicate mutated Hox binding sites (TCCT). All of the constructs were generated by ligating DNA fragments from the OPG promoter to the SV40 promoter. B, OPG promoter fragment containing two Hox sites responds to Smad1 and ALK3 treatments. C3H10T1/2 cells plated at a density of 5 ϫ 10 4 cells/well in 12-well plates for transfections. OPG promoter constructs (0.3 g) with two Hox sites or with two mutated Hox sites were cotransfected in C3H10T1/2 cells, respectively, with or without Smad1 (0.1 g) and ALK3 (0.025 g) plasmids, Hoxc-8 plasmid, or a combination of these two. C, the first Hox site of the OPG promoter is more important for responding to Smad1 and ALK3 treatments. C3H10T1/2 cells plated at a density of 5 ϫ 10 4 cells/well in 12-well plates for transfections. OPG promoter constructs (0.3 g) with two Hox sites, the first Hox site mutation, or the second Hox site mutation were cotransfected in C3H10T1/2 cells, respectively, with or without Smad1 (0.1 g) and ALK3 (0.025 g) plasmids or in combination with Hoxc-8 plasmid. Transfected cells were harvested and lysed in 42 h. The luciferase activity was measured and normalized to the Renilla luciferase level as internal controls. Data are shown with the mean Ϯ S.D. of triplicates. efficiently to both of the individual Hox binding sites (Fig. 3C,  lanes 3 and 8). We have previously shown that Smad1 inhibits Hoxc-8 binding in activating gene transcription (33,34). The binding specificity of Hoxc-8 to these two Hox sites was examined by both DNA competition studies and Hox site mutation experiments. Each 50-fold unlabeled Hox site displaced the binding of Hoxc-8 to its corresponding Hox site, respectively (Fig. 3C, lanes 4 and 9). As expected, Hoxc-8 binding was abolished when mutated Hox sites were used as probes (Fig.  3C, lanes 5 and 10), in which the core sequence TAAT was mutated into TCCT for both Hox sites. Thus, these results confirm that Hoxc-8 binds to two Hox binding sites of OPG promoter specifically.
Hoxc-8 Interaction Domain of Smad1 Are Sufficient to Induce OPG Promoter Activation-We have previously established the BMP-induced gene transcription model (27,40), in which Smad1 removes Hoxc-8 binding via direct interaction between these two proteins (33). We also mapped the interaction domains for both proteins. The Smad1 linker region is one of the domains (Smad1C, Fig. 4A) that interacts with Hoxc-8 and blocks Hoxc-8 binding to its DNA element (34). The binding of Hoxc-8 to OPG promoter predicts that overexpression of the Smad1 linker region activates OPG promoter transcription activity. To determine whether such was the case, we cotransfected Smad1C expression plasmids in C3H10T1/2 cells with the OPG promoter construct (OPG1) or the SV40 promoter linked to the two OPG Hox sites (OPG-SV40). Smad1C dramatically stimulates transcription activity for both constructs, and Hoxc-8 inhibits Smad1C-induced transcription activity. Notably, Smad1C with deletion of Smad1 interactin domain did not have such an effect (Fig. 4B). The construct with both mutated Hox sites (mOPG-SV40) or a single Hox site mutation (m1OPG-SV40 and m2OPG-SV40) was transfected in C3H10T1/2 cells. Fig. 4C shows that mutation of one or both Hox sites significantly decreased Smad1C-induced OPG promoter transcription. These results establish that the interaction between Smad1 and Hoxc-8 mediates BMP signals and support our proposed model for BMP-induced gene transcription.
Smad1-Hoxc-8 Interaction Induces Endogenous OPG Gene Expression-Having characterized the BMP-induced transcription mechanism for the OPG promoter, we then examined if the OPG promoter analysis-derived transcription mechanism reflects Smad1-mediated endogenous OPG gene expression. C3H10T1/2 cells were transfected with Smad1C expression plasmids. The OPG protein expression levels were measured by Western blot from the Smad1C-transfected 2T3 cells. As seen in Fig. 5A, a 55-kDa band was significantly induced. Another band above 79 kDa appeared later at 32 h after Smad1C transfection, indicating that a large amount of monomer and a small amount of dimer exist in cells. Simonet et al. (7) demonstrated that OPG is synthesized as an ϳ55-kDa monomer within the cell, and the monomers are then converted to disulfide-linked dimers of ϳ110 kDa. The dimers are then secreted into the medium. The specificity of the OPG band recognized in Western blot is confirmed by completely blocking the band with excess of OPG antigen (Fig. 5B).
Finally, we examined the biological activity of secreted OPG in the medium from Smad1C-transfected cells. Osteoclasts normally differentiate from members of the monocyte-macrophage lineage when exposed to OPGL and M-CSF, but OPG inhibits FIG. 3. Characterize hox binding sites from OPG promoter. A, OPG promoter fragments used for gel shift assays in B and C. Nucleotides are numbered relative to the transcription start site. Filled boxes indicate putative Hox binding sites (TAAT). Boxes marked ϫ indicate mutated Hox binding sites (TCCT). B, Hoxc-8 binds to OPG-2Hox. 32 P-Labeled OPG-2Hox was used in gel shift assays in incubation with GST-Hoxc-8 (0.2 or 0.6 g). C, Hoxc-8 specifically binds to OPG-Hox1 and OPG-Hox2. A series of 32 P-labeled OPG promoter fragments, indicated in A, were used in gel shift assays in incubation with GST-Hoxc-8 (0.6 g), fragments alone (lanes 1 and 6), fragments with 0.6 g of GST (lanes 2 and 7), or 0.6 g of GST-Hoxc-8 (lanes 3 and 8). Lanes 4 and 8 contained a 50-fold molar excess of unlabeled indicated fragments. OPG-mHox1 and OPG-mHox2 were incubated with 0.6 g of GST-Hoxc-8 as shown in lanes 5 and 10, respectively. osteoclast differentiation. Primary mouse BMMs were cultured (38) in the presence or absence of M-CSF and OPGL in combination with one-fourth of the medium from Smad1C-or vectortransfected 2T3 cells. A significant quantity of TRAP ϩ multinucleated osteoclasts was formed in cells treated with M-CSF and OPGL (Fig. 6A), and the addition of medium from cells transfected with the empty vector did not change TRAP ϩ multinucleated osteoclast formation (Fig. 6C). The addition of Smad1C-transfected medium significantly blocked TRAP ϩ multinucleated osteoclast differentiation (Fig. 6D), similar to macrophages cultured without M-CSF and OPGL (Fig. 6B). These results demonstrate that Smad1C induces secretion of functional OPG protein, which inhibits osteoclastogenesis. DISCUSSION The aim of this study is to elucidate the transcription mechanism of BMP-induced OPG gene expression and further explore the interaction between Smad1 and Hoxc-8 as an initial cascade in BMP-induced osteoblast differentiation. We characterized BMP response elements in the OPG promoter. Promoter deletion analysis demonstrates that the OPG promoter fragment bearing two Hox binding sites responds to BMP stim-ulation. Deletion of the two Hox binding sites abolished BMPinduced OPG promoter activity, and further deletion of other putative Hox binding sites did not change the luciferase activity in response to BMP stimulation. This observation was validated by transfection studies with the construct containing mutations of the two Hox binding sites (Fig. 2B). Furthermore, gel shift assays demonstrate that Hoxc-8 binds to the two Hox sites specifically. There are other putative Hox binding sites identified by the core sequence TAAT in the OPG promoter. However, these Hox sites did not appear to mediate BMP signals in the osteogenic cells. The flanking regions of the "core" are important in determining the specificity of Hox binding, since there are 39 members in the Hox family. It is possible that these Hox binding sites play an essential role in regulating OPG gene expression in different tissues during development.
We then examined whether Smad1 interaction with Hoxc-8 mediates the BMP signals via the two Hox binding sites. We previously showed that the Smad1 linker region interacts with Hoxc-8 to induce osteopontin gene transcription. Overexpression of this interaction domain is sufficient to induce osteoblast differentiation and bone cell formation (34). Similar to osteopontin promoter, transfection of the Smad1 linker region significantly activated OPG promoter transcription in C3H10T1/2 cells, and the induced promoter activity is inhibited by Hoxc-8. Thus, these data suggest the interaction between Smad1 and Hoxc-8 is a general transcription mechanism in the BMP signaling pathway.
The most important functional finding is that the overexpression of Smad1 linker region stimulated endogenous OPG protein expression in C3H10T1/2 cells. The induced OPG protein was secreted into medium, which effectively inhibited osteoclastogenesis. It is known that BMPs play a critical role in bone remodeling and in maintaining the structural integrity of the skeletal system, due to their ability to induce osteoblast differentiation (22)(23)(24)(25). OPG, on the other hand, mediates the communication between osteoblast and osteoclast to sustain bone density. The BMP-induced OPG transcription mechanism revealed in the current study involves Smad1 interaction with Hoxc-8, which is a quick response to BMP stimulation, since it only requires Smad1 phosphorylation. This logically explains the importance of the balance between osteoblast activity and osteoclast activity. The regulation of bone cell differentiation is a very complicated process. OPG is a potent inhibitor of osteoclast-mediated bone resorption produced by osteoblast. It was recently shown that OPG mRNA expression is stimulated by many bone metabolic growth factors, such as tumor necrosis factor-␣ (20), TGF-␤ (18,19), and BMP-2 (20), while downregulated by osteoclast formation agents, such as 1,25-(OH) 2 D 3 , prostaglandin E 2 , and parathyroid hormone (19). To our knowledge, this is the first study to characterize the OPG transcription mechanism in osteoblastic cells, which may provide a potential anti-resorption therapeutic opportunity.
BMPs initiate the entire transcription program in osteoblast differentiation. Their signals are mediated by phosphorylation of BMP receptor-regulated Smads, including Smad1, Smad5, and Smad8. Hoxc-8 has been characterized as the downstream transcription factor of Smad1 in BMP-induced osteoblast differentiation (33), and our data also indicate that Hoxc-8 is specific for the BMP signaling pathway. Data presented in this paper provide additional evidence supporting the hypothesis that the interaction of Smad1 with Hoxc-8 represents the initiation mechanism for BMP-induced osteoblast differentiation. Smad1-Hoxc-8 interaction is the mechanism in activating OPG gene expression, another early marker gene, in response to BMP stimulation. A wide existence of putative Hox binding sites in many other early BMP-responsive genes (35, 36) also suggests the role of Smad1-Hoxc-8 interaction in their BMPinduced gene transcription. However, since Smad1 itself is a sequence-specific transcription factor and there are other BMP-regulated Smads, such as Smad5 and Smad8, an intricate BMP-induced transcription network is anticipated. The BMPinduced transcription cascades will become clear when more BMP downstream transcription factors are identified and more transcription mechanisms are characterized from BMP-regulated genes.