Smad1 Interacts with Homeobox DNA-binding Proteins in Bone Morphogenetic Protein Signaling*

Bone morphogenetic proteins (BMP) transduce their signals into the cell through a family of mediator proteins known as Smads. Upon phosphorylation by the BMP receptors, Smad1 interacts with Smad4 and translocates into the nucleus where the complex recruits DNA-binding protein(s) to activate specific gene transcription. However, the DNA-binding protein(s) involved in BMP signaling has not been identified. Using a yeast two-hybrid approach, we found that Smad1 interacts with Hoxc-8, a homeodomain transcription factor. The interaction between Smad1 and Hoxc-8 was confirmed by a “pull-down” assay and a co-immunoprecipitation experiment in COS-1 cells. Interestingly, purified Smad1 inhibited Hoxc-8 binding to the osteopontin Hoxc-8 site in a concentration-dependent manner. Transient transfection studies showed that native osteopontin promoter activity was elevated upon BMP stimulation. Consistent with the gel shift assay, overexpression of Hoxc-8 abolished the BMP stimulation. When a wild type or mutant Hoxc-8 binding element was linked to an SV40 promoter-driven reporter gene, the wild type but not the mutant Hoxc-8 binding site responded to BMP stimulation. Again, overexpression of Hoxc-8 suppressed the BMP-induced activity of the wild type reporter construct. Our findings suggest that Smad1 interaction with Hoxc-8 dislodges Hoxc-8 from its DNA binding element, resulting in the induction of gene expression.

Transforming growth factor-␤ (TGF-␤) 1 -related molecules, or BMPs, regulate embryonic development, vertebral patterning, and mesenchymal cell differentiation (1,2). BMP-2 and -4 have been identified as bone inductive growth factors and are important signaling molecules during the development of the skeleton in vertebrates (1,3,4). Signal transduction in the TGF-␤ superfamily requires the interaction of two types of serine/threonine transmembrane receptor kinases (5). The signaling is mediated by direct phosphorylation of Smad proteins. Specifically, Smad2 and Smad3 are phosphorylated by TGF-␤ and activin receptors (5,6), whereas phosphorylation of Smad1 is induced by BMPs (7,8). Upon phosphorylation, these Smads interact with a common partner, Smad4, which then translocates to the nucleus where the complexes recruit DNA-binding protein(s) to activate specific gene transcription (5,7,9). The downstream DNA-binding proteins in the TGF-␤ signaling pathway, such as Fast-1, Fast-2, and TFE3, have been reported (10 -12). However, little is known about the downstream DNAbinding protein(s) beyond Smad1 in the BMP signal transduction machinery.
It has been suggested that homeobox genes play a role in downstream events in BMP signaling (13,14). In vertebrates, there are 39 Hox transcription factor genes organized into four separated chromosome clusters, which play critical roles in the patterning of vertebrate embryonic development (15). These 39 genes are subdivided into 13 paralogous groups on the basis of duplication of an ancestral homeobox cluster during evolution, sequence similarity, and position within the cluster (16). Each paralog group has been demonstrated to be responsible for the morphogenesis of a particular embryonic domain or structure (15).
Hoxc-8, as one of the three members of paralog VIII, is predominantly expressed at a high level in the limbs, backbone, and spinal cord in early mouse embryos (17,18). Null mutant mice showed that Hoxc-8 is expressed in the neuron, chondrocyte, fetal liver, and adult bone marrow (19,20). Bending and fusion of the ribs, anterior transformation of the vertebrae, and abnormal patterns of ossification in the sternum were observed in adult Hoxc-8 null mice (20). Studies published recently demonstrated that tissue-specific overexpression of a Hoxc-8 transgene inhibits chondrocyte maturation and stimulates chondrocyte proliferation (21). The other two members in the Hox VIII group are Hoxb-8 and Hoxd-8. Hoxb-8 has been shown to activate the Sonic hedgehog gene, an essential mediator in forelimb development (22,23), whereas generalized expression of Hoxd-8 modifies Drosophila anterior head segments (24).
Despite the fact that homeobox genes are DNA-binding proteins, little has been learned about their natural DNA response elements and role in transcription (25). In the current study, we report that Smad1 interacts with Hoxc-8, and this interaction specifically activates the osteopontin gene transcription in response to BMP stimulation. Our data suggest that Hoxc-8 functions as a transcription repressor and that the interaction of Smad1 with Hoxc-8 dislodges Hoxc-8 binding from its element resulting in initiation of gene transcription.

EXPERIMENTAL PROCEDURES
Two-hybrid Library Screening-A full-length Smad1 coding sequence from pBluescript-Smad1(9) was cloned into SalI/PstI sites of pGBT9 (CLONTECH) to generate the pGBT9/Smad1 bait plasmid. The human U2 OS osteoblast-like pACT2 cDNA library was screened according to the manufacturer's instruction (CLONTECH). To confirm the interaction between Hoxc-8 and Smad1, a full-length mouse Hoxc-8 cDNA (18) was subcloned into pACT2 vector (CLONTECH) between the XhoI and EcoRI sites. The pACT/Hoxc-8 was co-transformed with pBGT9/Smad1 into Y190, and the colonies were assayed for the production of ␤-galactosidase using both filter lift and liquid assays.
GST Pull-down Assay-[ 35 S]Methionine-labeled Hoxc-8 protein was synthesized using the TNT-coupled transcription and translation system (Promega) with linearized pBluescript-Hoxc-8 plasmid according to the manufacturer's instruction. The production of labeled protein was confirmed by SDS-polyacrylamide gel electrophoresis. An equivalent amount (1 g) of purified GST or GST-Smad1 fusion protein was preincubated with 35 S-labeled Hoxc-8 protein (5 l) for 30 min on ice. Following the addition of GST-agarose, the samples were incubated for another 30 min at 4°C. The agarose beads were washed four times in a phosphate-buffered saline, 0.1% Triton X-100 solution, and bound proteins were eluted by boiling in 2ϫ SDS buffer for 5 min before loading onto 10% SDS-polyacrylamide gel electrophoresis.
Immunoprecipitation and Western Blot-HA-tagged Hoxc-8 was subcloned from pACT2/Hoxc-8 into a mammalian expression vector pcDNA3 (Invitrogen) at BglII/BamHI and XhoI. Expression vectors for FLAG-tagged Smad1 and Smad4 were provided by Dr. Rik Derynck (University of California, San Francisco, CA). Expression plasmids for constitutively active BMP type IA (ALK3) and IB receptors (ALK6) (7) were provided by Dr. Jeffrey L. Wrana (Hospital for Sick Children, Toronto, Canada). COS-1 cells were transfected with expression constructs as indicated in Fig. 2B using Tfx-50 according to the manufacturer's description (Promega). Cells were lysed 48 h post-transfection, and lysates were immunoprecipitated with anti-HA antiserum (Babco) and immunoblotted with anti-FLAG M2 (Eastman Kodak) as described (7).
Transfection-The osteopontin promoter from region Ϫ266 to Ϫ1 relative to the transcription start site was amplified by polymerase chain reaction from CH10T1/2 cell genomic DNA and cloned into SmaI and XhoI sites of the pGL3-basic vector (Promega) to generate a luciferase reporter construct (OPN-266). Hox-pGL3 reporter bearing the Hoxc-8 binding site (Ϫ290 to Ϫ166) was constructed using the same strategy but was put into the pGL3-control vector (Promega). The Hox recognition core TAAT was replaced with GCCG in Hox-pGL3 by polymerase chain reaction to create mutant Hox-pGL3 (mHox-pGL3). C3H10T1/2 cells (2.5 ϫ 10 5 cells/60-mm dish) were transfected using Tfx-50 with 0.5 g of luciferase reporter plasmid (OPN-266, Hox-pGL3, or mHox-pGL3) and different expression plasmids as indicated. Total DNA was kept constant by the addition of pSV-␤-galactosidase plasmid. Luciferase activities were assayed 48 h post-transfection using the dual luciferase assay kit (Promega) according to the manufacturer's direction. Values were normalized with the Renilla luciferase activity expressed from pRL-SV40 reporter plasmid. Luciferase values shown in the figures are representative of transfection experiments performed in triplicate in at least three independent experiments.

RESULTS AND DISCUSSION
Yeast Two-hybrid Library Screening-To investigate the transcription mechanism in BMP-induced gene activation, we used a yeast two-hybrid system to identify transcription factors that interact with Smad1. An intact Smad1 cDNA fused with the Gal4 DNA-binding domain was used as a bait plasmid to screen a human U-2 OS osteoblast-like cell cDNA library constructed in the pACT2 plasmid. After two rounds of screening, we obtained 25 positive clones. DNA sequence analysis identified one clone as Hoxc-8 and two clones as Smad4. Because our objective is to identify downstream transcription factors in the BMP signaling pathway and Hoxc-8 is a homeodomain DNAbinding protein, we chose the Hoxc-8 cDNA clone for further study. Cloning of Smad4 provided a positive control for the two-hybrid library screening because the interaction between Smad1 and Smad4 is known. The other 22 clones were not characterized.
The initial Hoxc-8 cDNA clone (Fig. 1B, clone 19) encodes amino acids 68 -237 of a 242-amino acid Hoxc-8 protein. Fig. 1A shows the growth properties of the two-hybrid clones, suggesting that there is an interaction between Smad1 and Hoxc-8 in vivo. The yeast bearing both Smad1 and Hoxc-8 plasmids grew on medium deficient in Trp, Leu, and His. The interaction between Hoxc-8 and Smad1 was further confirmed with a ␤-galactosidase filter lift assay (data not shown) and quantified by a liquid ␤-galactosidase assay (Fig. 1B). When the full length of Hoxc-8 fused with the Gal4 transcriptional activation domain was tested in the two-hybrid system, it showed an interaction similar to clone 19. The assays of both the empty prey vector (pACT2) with Smad1 in the bait plasmid and the empty bait vector (pGBT9) with full-length Hoxc-8 in the prey vector showed very little activity. Compared with the interaction between Smad1 and Smad4, the interaction of Smad1 with Hoxc-8 is weaker in the yeast two-hybrid ␤-galactosidase assay (Fig. 1B).
Smad1 Interacts with Hoxc-8 in Vitro and in COS-1 Cells-The interaction between Smad1 and Hoxc-8 was examined in an in vitro pull-down experiment using [ 35 S]methionine-labeled Hoxc-8 and purified GST-Smad1 or GST alone. As shown in Fig. 2A, Hoxc-8 was precipitated with the purified GST-Smad1 fusion protein (lane 3) but not with GST alone (lane2), demonstrating a direct interaction between the two proteins in vitro.
BMP-2 stimulates phosphorylation of Smad1, and phosphorylated Smad1 in turn binds to Smad4 and takes the complex into the nucleus. It is of interest whether Smad1, Smad4, or the complex of Smad1 and Smad4 also interacts with Hoxc-8 in cells. COS-1 cells were transiently co-transfected with expression plasmids for FLAG-Smad1, FLAG-Smad4, HA-Hoxc-8, and/or constitutively active BMP type IA receptor ALK3 (Q233D). The cell lysates were immunoprecipitated with anti-HA antibody and immunoblotted with anti-FLAG antibody. Osteopontin Promoter Contains a Hoxc-8 Binding Element-To examine the effect of the interaction between Hoxc-8 and Smad1 on Hoxc-8 DNA binding activity, we turned our attention to BMP-2 inducible genes. Putative Hox binding sites that have served as markers for osteogenic differentiation were found in four BMP-2 inducible genes, including bone sialoprotein, osteopontin, osteonectin, and osteocalcin (3, 31). These genes have served as markers for osteoblast differentiation. The osteopontin promoter was examined for this purpose be-cause its mRNA expression is rapidly induced in response to BMP-2 treatment in C3H10T1/2 mesenchymal cell (3). Five putative Hox binding sites with a core sequence of Tt/aAT (32) were identified within the first 382 base pairs of the 5Ј-flanking region in the osteopontin gene (Fig. 3A). When a 212-base pair DNA fragment from Ϫ382 to Ϫ170 (OPN-1 in Fig. 3A) containing all five putative Hox sites was used for a gel shift assay with purified GST-Hoxc-8 protein, one shifted band (Fig. 3B,  lane 3) was observed. This band was not present in lane 1, containing probe only, or in lane 2, containing probe with GST (Fig. 3B). This result indicates that there is only one Hoxc-8

FIG. 1. Specific interaction of Smad1 with Hoxc-8 in a yeast two-hybrid system.
A, growth properties of two-hybrid clones. The interaction was assayed in a yeast strain, which requires His, Leu, and Trp to grow. pGBT9-Smad1 and pACT2-Hoxc-8 plasmids carry Trp and Leu as their selective markers, respectively. The interaction between Smad1 and Hoxc-8 enables the yeast to synthesize His. Only clones bearing both pGBT9-Smad1 and pACT2-Hoxc-8 plasmids grew in medium lacking His, Leu, and Trp. All assays were done in medium containing 45 mM 3-aminotriazole (3-AT), which inhibits growth because of nonspecific interaction. B, ␤-galactosidase liquid assay for two-hybrid interaction. ␤-Galactosidase activities for yeast bearing plasmids as indicated were plotted. binding site in this osteopontin promoter fragment. Further gel shift assays with shorter probes (OPN-2 and OPN-3 in Fig. 3A) indicated that OPN-2 contains this Hoxc-8 binding element (Fig. 3B, lane 6). When three single putative Hox binding probes (OPN-4, -5, and -6, Fig. 3A) were used, Hoxc-8 only bound to OPN-5, located at Ϫ206 to Ϫ180 (Fig. 3C, lane 8).
The specificity of the Hoxc-8 binding to the DNA was determined by a gel shift competition assay. Unlabeled Hoxc-8 DNA binding element inhibited the shifted band in a concentration dependent manner (Fig. 3D, lanes 3-5) in which a 100-fold excess of the specific cold probe eliminated the Hoxc-8 binding, whereas a 100-fold excess of the Msx-2 DNA binding element (33) did not (Fig. 3D). Msx-2 is a homeodomain-containing protein, but it does not belong to the HOX family. The Msx-2 DNA binding element was identified from the osteocalcin promoter, and its flanking regions of the core sequence is different from Hoxc-8 binding site.
There are three TAAT and two TTAT putative Hox sites identified from the osteopontin promoter. Hoxc-8 binds to only one of the TAAT core sequences (Ϫ206 to Ϫ180), suggesting that the flanking regions are also important for Hoxc-8 binding. The Hoxc-8 binding site, including its flanking regions, is highly conserved in chicken, mouse, pig, and human. The other four putative Hox sites may be involved in other homeodomain protein binding or may not be authentic Hox binding sites.
Smad1 Inhibits Binding of Hox Proteins to DNA-Purified GST-Smad1 was examined for the effect of its interaction with Hoxc-8 on Hoxc-8 DNA binding activity. When GST-Hoxc-8 protein and its DNA binding element (OPN-5) were incubated with increasing amounts of GST-Smad1 protein, the binding of Hoxc-8 to the DNA probe was inhibited in a concentrationdependent manner (Fig. 4A, lanes 5-7). The same amount of GST protein did not have an effect on Hoxc-8 binding activity (Fig. 4B, lane 4). These results suggest that the interaction of Smad1 with Hoxc-8 dislodges Hoxc-8 from its response element.
Because the signaling networks of the TGF-␤ superfamily are very complex, it is important to understand the specificity of the interaction between Hox and Smad proteins. Hoxa-9 was chosen as a well characterized homeobox DNA-binding protein (34,35) to examine its interaction with different Smad proteins. Two other homeodomain proteins, Msx-1 and Msx-2, were also used for gel shift assays for the same purpose. Msx-1 and Msx-2, found at different loci than the Hox gene clusters, are involved in development of teeth. The expression of both genes is coordinately regulated by BMP-2 and BMP-4 (36 -38).
To estimate the relative strength of the interactions between the Smads and homeodomain proteins, the same amounts of Hoxc-8 and Hoxa-9 or Msx-1 and Msx-2 proteins were used in each of the gel shift assays with a fixed amount of different Smad proteins (Fig. 4, B and C). Smad1 and Smad4 inhibited both Hoxc-8 and Hoxa-9 binding, and the inhibition was enhanced when both Smad proteins were added together (Fig. 4B,  lanes 7 and 14). In contrast, neither Smad2 nor Smad3 interacted with these two Hox proteins. Fig. 4C showed that neither of the Msx proteins interacted with any of the four Smad proteins. GST did not affect Hox or Msx protein binding (Fig.  4B, lanes 4 and 11, and Fig. 4C, lanes 4 and 10). The homeodomain, a well conserved DNA binding motif, is the region highly conserved between Hoxc-8 and Hoxa-9, suggesting that Smad1 interacts with other Hox proteins involved in BMP signaling.
moter fragment containing the Hoxc-8 binding site into the pGL3-basic luciferase reporter vector to generate an OPN-266 reporter plasmid (Fig. 5A). Transfection of the OPN-266 construct in C3H10T1/2 mesenchymal cells showed that the reporter activity was stimulated moderately when Smad1 or Smad4 expression plasmids were co-transfected. The luciferase activity was significantly enhanced when the OPN-266 reporter construct was co-transfected with ALK3 (Q233D), Smad1, and Smad4 expression plasmids. Furthermore, the ALK3 (Q233D)-induced transcriptional activity was completely abolished when Hoxc-8 was overexpressed (Fig. 5B).
To further define the transcription activity of the Hoxc-8 binding site, we linked a shorter osteopontin promoter fragment containing the Hoxc-8 binding site to a luciferase reporter vector under the control of the SV40 promoter (Hox-pGL3, Fig.  5A). When the Hox-pGL3 construct was co-transfected in C3H10T1/2 cells with ALK3 (Q233D) or ALK6 (Q203D), luciferase reporter activity was induced more than 13-and 11-fold, respectively. Most importantly, overexpression of Hoxc-8 suppressed the ALK3 (Q233D)-induced or ALK6 (Q203D)-induced reporter activity (Fig. 5C). These results suggest that the Hox binding site mediates BMP signaling and that Hoxc-8 functions as a transcription repressor. In comparison with the osteopontin native promoter, the Hox-pGL3 construct does not require overexpression of Smad1 and -4 in responding to BMP stimulation. This is an SV40 promoter-driven construct with a much shorter osteopontin promoter fragment, which does not contain many other transcription elements like the native osteopontin promoter construct.
Several Smad downstream transcription factors have been characterized in the TGF-␤ pathway. Here, we first show that Hoxc-8 interacts with Smad1 as a downstream DNA-binding protein in the BMP pathway. Our data demonstrate that Hoxc-8 binds to the osteopontin promoter and represses the gene transcription. BMP stimulation activates gene transcription by derepressing the Hoxc-8 protein through the interaction promoter; mHox-pGL3 contains the mutated osteopontin Hox binding site. B, BMP activates the osteopontin promoter. The OPN-266 plasmid was co-transfected in C3H10T1/2 mesenchymal cells with Hoxc-8, Smad1, or Smad4 plasmids alone or in a combination of all three in the presence or absence of ALK3 plasmid. C, the osteopontin Hox binding site mediates BMP-induced transcription. Hox-pGL3 construct was co-transfected with ALK6 or ALK3 in C3H10T1/2 mesenchymal cells. D, mutation of Hox binding site abolishes BMP stimulation. Hox-pGL3 construct or mHox-pGL3 pGL3 control plasmid was co-transfected with ALK6, ALK3, or Hoxc-8 plasmids in C3H10T1/2 mesenchymal cells. Cell lysates in B, C, and D were assayed for luciferase activity normalized to Renilla luciferase levels 48 h after transfection. Experiments were repeated twice in triplicates.
of Smad1 with the Hoxc-8 protein. The direct interaction between Smad1 and Hox protein(s) suggests their functional relationship and the mechanisms in BMP-induced skeleton development.