A nuclear antagonistic mechanism of inhibitory Smads in transforming growth factor-beta signaling.

Inhibitory Smads (I-Smads), including Smad6 and Smad7, were initially characterized as cytoplasmic antagonists in the transforming growth factor-beta signaling pathway. However, I-Smads are also localized in the nucleus. Previously, we have shown that Smad6 can function as a transcriptional co-repressor. In this study, we found both Smad6 and Smad7 interact with histone deacetylases (HDACs). Acetylation state of core histones plays a critical role in gene transcription regulation. An HDAC inhibitor, trichostatin A, released Smad6-mediated transcription repression. Moreover, class I HDACs (HDAC-1 and -3), not class II HDACs (HDAC-4, -5, and -6), were co-immunoprecipitated with Smad6. Endogenous HDAC-1 was also shown to interact with both Smad6 and Hoxc-8. Mapping of the interaction domain indicates Smad6 MH2 domain is mainly involved in recruiting HDAC-1. Most interestingly, Smad6 also binds to DNA through its MH1 domain, and the MH2 domain of Smad6 masks this binding activity, indicating that Smad6 MH1 and MH2 domains associate reciprocally and inhibit each other's function. Hoxc-8 induces Smad6 binding to DNA as a transcriptional complex. Our findings revealed that I-Smads act as antagonists in the nucleus by recruiting HDACs.

Inhibitory Smads (I-Smads), including Smad6 and Smad7, were initially characterized as cytoplasmic antagonists in the transforming growth factor-␤ signaling pathway. However, I-Smads are also localized in the nucleus. Previously, we have shown that Smad6 can function as a transcriptional co-repressor. In this study, we found both Smad6 and Smad7 interact with histone deacetylases (HDACs). Acetylation state of core histones plays a critical role in gene transcription regulation. An HDAC inhibitor, trichostatin A, released Smad6-mediated transcription repression. Moreover, class I HDACs (HDAC-1 and -3), not class II HDACs (HDAC-4, -5, and -6), were co-immunoprecipitated with Smad6. Endogenous HDAC-1 was also shown to interact with both Smad6 and Hoxc-8. Mapping of the interaction domain indicates Smad6 MH2 domain is mainly involved in recruiting HDAC-1. Most interestingly, Smad6 also binds to DNA through its MH1 domain, and the MH2 domain of Smad6 masks this binding activity, indicating that Smad6 MH1 and MH2 domains associate reciprocally and inhibit each other's function. Hoxc-8 induces Smad6 binding to DNA as a transcriptional complex. Our findings revealed that I-Smads act as antagonists in the nucleus by recruiting HDACs.
Transforming growth factor ␤ (TGF-␤) 1 superfamily members, which include TGF-␤s, activins, and bone morphogenetic proteins (BMPs), play a very important role during embryonic development and maintaining adult tissue homeostasis (1). TGF-␤ signaling is mediated by two transmembrane serinethreonine kinase receptors, type II and type I receptors (2). Upon ligand binding, the constitutively active type II receptors phosphorylate and activate type I receptors, leading to the propagation of signaling by recruiting and phosphorylating a group of specific proteins, Smads (3). Smads are pivotal intracellular nuclear effectors of TGF-␤ family members, which transduce the signal from the cell membrane to the nucleus (3). Smads contain two highly conserved domains: MH1 and MH2 domains (1). Commonly, MH1 domain binds to DNA, whereas MH2 domain is the protein-protein interaction and transactivation domain (1). These two domains interact reciprocally and inhibit each other's function (1,4). Based on their function and sequence similarity, Smads are divided into three subgroups. 1) The receptor-regulated Smads (R-Smads) are the targets of the activated type I receptors. Smad1, Smad5, and Smad8 mediate BMP signaling (3), whereas Smad2 and Smad3 mediate TGF-␤ signaling (5). 2) The common partner Smads (Co-Smads), Smad4 being the only one identified in mammals thus far, are shared by all of the R-Smads (1). 3) Inhibitory Smads (I-Smads), including Smad6 and Smad7, stably bind to activated type I receptors and block phosphorylation of R-Smads (6,7). Both TGF-␤ and BMP induce I-Smad expression, indicating their negative feedback function in TGF-␤ signaling (8,9).
Smad6 preferably inhibits BMP signaling, whereas Smad7 is more a general inhibitor (10). Smad6 and Smad7 are expressed at the earliest stage during embryo development and highly expressed in the developing cardiovascular system, eyes, bones, and other tissues (11,12). The expression of Smad6 overlaps BMP-2, -4, and -7 expression, which orchestrates BMP-mediated cardiac development (11). Aortic ossification and elevated blood pressure were reported in viable Smad6 mutants (13). BMP induces the ventral mesoderm formation (14,15), and the overlapped expression of Smad6 and BMP indicates that Smad6 is a key protein in balancing the function of BMP during embryo development. Smad6 inhibits BMP-induced osteoblast differentiation (16). It is also reported that Smad6 inhibits adipogenesis (17). Smad7, not Smad6, was identified as inhibiting TGF-␤ function during embryonic vasculogenesis (18) and lung development, injury, and repair (19,20).
The biological activities of Smads are closely associated to their cellular localization. R-Smads are located in the cytoplasm in the absence of signaling. Upon ligand stimulation, they are phosphorylated at their extreme carboxyl end SSXS motif and recruit the common partner Smad4 into the nucleus, where they act as transcriptional activators (21). I-Smads are located in both the cytosol and the nucleus (22)(23)(24)(25). Smad6 cellular distribution is not affected by TGF-␤ or BMP treatment (24,25) whereas TGF-␤ induces nuclear export of Smad7 (22). The inhibitory mechanisms of I-Smads have been characterized in the cytoplasm. I-Smads interact with the activated type I receptors, which then block the phosphorylation of the R-Smads (6,7). Smad6 was also demonstrated to interact with the phosphorylated Smad1 in the cytoplasm, competing with Smad4 to form an inactive Smad1-Smad6 complex (26). We have shown that, in the nucleus, Smad6 acts as a transcriptional co-repressor on osteopontin promoter by interacting with Hoxc-8 (25), a transcriptional repressor in BMP signaling (27,28). The mechanism of Smad6 repressive function in the nucleus remains unclear.
Signaling to chromatin through histone modification is demonstrated as a major step for regulating target gene transcription (29). One of the modifications of histones is acetylation, where specific lysine residues are functional targets for histone acetyltransferases (HATs) and histone deacetylases (HDACs) (30 -32). The acetylation of chromosomal histones loosens the structure of the target gene promoter and results in increased accessibility of the chromatin for transcription factors (30). Conversely, hypoacetylation of a gene regulatory region is strongly related to silencing gene expression (30). Several transcription repressors and co-repressors have been demonstrated to recruit HDACs to specific genes for silencing gene expression (33). For example, TGIF, a Smad transcriptional co-repressor, was identified as interacting with HDAC-1 and inhibiting TGF-␤-induced gene transcription (34). In addition to HDAC-1, eight other HDACs have been identified thus far (35,36).
Given that Smad6 displays similar activity as TGIF and other transcriptional repressors, all of which appear to recruit different HDAC complexes, we considered whether histone deacetylase activity might extend to Smad6 and affect its transcriptional activity in the nucleus. To address this issue, we characterized the mechanism of Smad6-mediated gene transcription. We demonstrated that an HDAC activity inhibitor, trichostatin A (TSA), rescued the repressive function of Smad6 in BMP signaling, indicating that HDACs engaged in the inhibitory effect of Smad6 in the nucleus. Furthermore, Smad6 was shown to interact with HDAC-1 and -3 in an immunoprecipitation assay. HDAC activity was detected in HDAC assay with an immunoprecipitated Smad6 protein complex. These data suggest that Smad6 represses gene transcription by recruiting class I HDACs and modifying chromatin conformation.
Transient Transfection and Luciferase Assay-To identify the function of HDACs on osteopontin promoter, osteopontin promoter containing Hoxc-8 binding element (Hox-pGL3) was used as the reporter plasmid (27). Mv1Lu cells (5 ϫ 10 4 cells/12-well plate) were transiently co-transfected with Smad1B, Smad6, and Hoxc-8, respectively, in the absence or presence of the HDAC inhibitor TSA (100 nM) as indicated in Fig. 1 using Tfx-50 according to the manufacturer's instructions (Promega). Using the same reporter plasmid, HDAC-1 was co-transfected in Mv1Lu cells with Smad1B, Smad6, and Hoxc-8 (Fig. 3D). Cells were passively lysed 48 h after transfection, and luciferase activity was measured and normalized by Renilla luciferase activity expressed from pRL-SV40 reporter plasmid using the Dual-Luciferase assay kit (Promega) according to the manufacturer's directions (Promega). Luciferase values shown in the figures are representative of transfection experiments performed in triplicate in at least three independent experiments.
Immunoprecipitation Assay-HA-tagged Smad6 was co-transfected with FLAG-tagged HDAC-1, HDAC-3, HDAC-4, HDAC-5, and HDAC-6 into COS-1 cells using Tfx-50 according to the manufacturer's instructions (Promega). The cells were lysed 48 h after transfection. Anti-HA polyclonal antibody (poly-HA, Babco) was added into the same amount of protein samples and rotated at 4°C for 3 h. The protein complexes were immunoprecipitated by protein G-Sepharose beads (Amersham Biosciences, Inc.), and the samples were loaded and run in a 12% SDS-PAGE. The precipitates were transferred to nitrocellulose mem-brane and immunoblotted by anti-FLAG M2 monoclonal antibody (Sigma). FLAG-tagged Smad6, Smad6 truncated proteins, and Smad7 were co-transfected in COS-1 cells with HA-tagged HDAC-1 using Tfx-50 according to the manufacturer's instructions (Promega). Anti-FLAG M2 monoclonal antibody was added into the same amount of protein samples for immunoprecipitation, and the poly-HA was used for immunoblotting the precipitated protein complex. To determine the interaction of Hoxc-8 and HDAC-1, the HA-tagged Hoxc-8 and FLAG-tagged HDAC-1 were co-transfected in COS-1 cells. The cell lysates were immunoprecipitated by poly-HA, and the protein complexes were immunoblotted by anti-FLAG M2 monoclonal antibody. For the interaction of Hoxc-8 and Smad6 with endogenous HDAC-1, HA-tagged Hoxc-8 and Smad6 were transfected into COS-1 cells. The cells were lysed 48 h after transfection. Anti-HA monoclonal antibody (Babco) was used for immunoprecipitation, and anti-HDAC-1 polyclonal antibody (Upstate Biotechnology, Inc.) was used for immunoblotting the precipitated protein complex.
HDAC Assay-HDAC-1, Smad6, and Hoxc-8 were co-transfected in COS-1 cells, as indicated in Fig. 3B, using Tfx-50 according to the manufacturer's description (Promega). Anti-FLAG M2 monoclonal antibody or poly-HA was used to immunoprecipitate the protein complexes. The 3 H-acetylated histones (1.5 g/reaction) were mixed with these protein complexes and incubated at 30°C for 45 min (38,39). The reactions were stopped by adding 15 l of 1 N HCl, and the released 3 H was extracted by adding 1 ml of ethyl acetate. The whole reaction tubes were centrifuged, and the remaining [ 3 H]acetylated histones were maintained in the inorganic layer. Histone acetylation degree was measured using the Whatman filter assay described under "HAT Assay" (39).
Gel Shift Assay-Gel shift assays were performed as described previously (40). GST fusion constructs of Smad6, Smad7, and Hoxc-8 were generated in our previous study (25,27). Smad6 truncated fragments were generated by PCR and cloned into pGEX vectors. These constructs were transformed into BL21 Escherichia coli. The proteins were extracted as described previously following induction with isopropyl-␤-D thiogalactopyranoside (27). Osteopontin Hox binding element (OPN-5) (27) was used as the probe to test the interaction between Hoxc-8 and Smad6. This probe was also used for testing whether Smad6 binds to DNA. We randomly mutated some base pairs at the flanking regions of the Hox binding site for identifying the Smad6 binding site. The upstream strain of the wild type probe is 5Ј-GGGTA-GTTAATGACATCGTTCATCAG-3Ј, mutant probe 1 is 5Ј-GGACAAGT-AATGACATCGTTCATCAG-3Ј, and mutant probe 2 is 5Ј-GGGTAGTT-AATGTCAGCACTCATCAG-3Ј.

An HDAC Inhibitor Releases Both Smad6-and Hoxc-8mediated Transcriptional Repression-We have shown that
Hoxc-8 functions as a transcriptional repressor in the BMP signaling pathway (27,28). Smad6 interacts with Hoxc-8 as a transcriptional co-repressor in response to BMP stimulation (25). These findings revealed Smad6 antagonistic function in the nucleus. Therefore, it is important to understand the mechanism of Smad6-repressed gene transcription. Histone acetylation is one of the major chromatin modifications in gene transcription regulation, and hypoacetylation of histones leads to repression of gene expression, which is mediated by HDACs (30 -32). TSA, an HDAC inhibitor, was used to examine if HDACs affect Smad6 repressive activity in the nucleus. Smad1B, the interaction domain of Smad1 with Hoxc-8, dislodges Hoxc-8 binding from its binding sites and activates BMP downstream gene transcription (25,28,37). BMP-inducible reporter construct, Hox-pGL3 (27), was co-transfected with Smad1B and/or Smad6, Hoxc-8 in mink lung epithelial cells (Mv1Lu) treated with or without TSA (100 nM). As shown in Fig. 1, Smad1B stimulated the transcriptional activity of osteopontin promoter, whereas Smad6 and Hoxc-8 inhibited it. Importantly, TSA released the transcription repression by Smad6 and Hoxc-8, suggesting that HDACs are involved in transcription repression of the osteopontin promoter mediated by Smad6 and Hoxc-8.
Smad7 is the other inhibitory Smad located in the nucleus in the absence of TGF-␤ stimulation (22). Unlike Smad6, it does not interact with Hoxc-8 or other Hox proteins (25). Nevertheless, the MH1 and MH2 domains are highly conserved between Smad6 and Smad7 (26). Smad7 may also interact with HDACs. Smad6 interacts with class I HDACs; therefore, we examined the interaction between Smad7 and HDAC-1. FLAG-tagged Smad7 was co-transfected with HA-tagged HDAC-1 in COS-1 cells. As expected, Smad7 was co-immunoprecipitated with HDAC-1 (Fig. 2B), indicating interaction of Smad7 with HDAC-1. This finding suggests that I-Smads are transcriptional repressors or co-repressors in the nucleus.
We also mapped the interaction domain of Smad6 with HDAC-1. A series of truncated Smad6 expression plasmids linked to FLAG (25) were co-transfected with HA-tagged HDAC-1 in COS-1 cells. Immunoprecipitation assay results in Fig. 2D showed that the Smad6 MH2 domain was mainly involved in interaction with HDAC-1, whereas the NH 2 -terminal domain and linker region had a very weak interaction. The linker region alone has no interaction with HDAC-1.
HDAC Activity in the Smad6 and Hoxc-8 Complex-Transcription repressors and co-repressors have been demonstrated to recruit HDACs to specific target gene promoters to deacetylate chromosomal histones in silencing gene transcription. Therefore, we examined the HDAC enzymatic activity in the Smad6 transcriptional complex because I-Smads and Hoxc-8 have been shown to interact with HDACs. FLAG-tagged Smad6 or HA-tagged Hoxc-8 plasmids were co-transfected with HDAC-1 expression plasmids in COS-1 cells as indicated in Fig. 3B. Cell lysates were immunoprecipitated with anti-FLAG antibody or anti-HA antibody for measuring HDAC activity (Fig. 3B). The acetylated histones, the substrate for HDAC assays, were prepared with GST-CBP (aa 1099 -1758), which contains a HAT domain, as described previously (38) (Fig. 3A). The resulted acetylated histones were incubated with the immunoprecipitated complexes at 30°C for 45 min. The degree of histone acetylation was measured by liquid scintillation counting (Fig. 3A). Both Smad6-HDAC-1 and Hoxc-8-HDAC-1 complexes deacetylated about one third of 3 H-labeled histones in comparison with controls, indicating HDAC activity in the Smad6 or Hoxc-8 complex. Furthermore, we examined endogenous HDAC activity in the Smad6 and Hoxc-8 immunoprecipitated complex. In this case, FLAG-Smad6 or HA-Hoxc-8 plasmids were transfected alone into COS-1 cells, and the cell lysates were immunoprecipitated with anti-FLAG antibody or anti-HA antibody. As expected, the Smad6 and Hoxc-8 immunoprecipitated complex exhibited significant endogenous HDAC activities (Fig. 3B). These results clearly suggest that Smad6 and Hoxc-8 inhibit gene transcription by recruiting HDACs.
Smad6 MH1 Domain Binds to DNA-We have shown that Hoxc-8 binds to osteopontin promoter, Smad6 and Hoxc-8 form a complex on the DNA element, although Smad6 alone does not bind to this DNA element (25). Both Smad6 and Hoxc-8 interact with HDAC-1, which implies that Smad6 may also be a DNA-binding protein. Smad proteins contain two highly con- served domains: the MH1 and the MH2 domain. In general, the MH1 domain binds to DNA, and the MH2 domain is the protein-protein interaction and transactivation domain (1). Smad1 (41), Smad3, and Smad4 (42,43) have been identified as binding to Smad binding element, which is characterized as GTCT Smad box or CAGA Smad box (42,43). The MH2 domain of Smad6 is highly conserved with other Smads, but the MH1 domain is distinct from the R-Smads and the Co-Smad (44). To determine whether Smad6 binds to DNA, we performed gel shift assays with a series of GST-Smad6 truncated proteins (Fig. 4A) using the osteopontin promoter that contains the Hoxc-8 binding element (OPN-5) as the probe (25,27). The results in Fig. 4B showed that the first 101 amino acids of Smad6 strongly bind to DNA. More interestingly, as the length of the Smad6 DNA binding domain was extended to its MH2 domain, the DNA binding activities were reduced. The fulllength Smad6 did not bind to the probe. These results indicate that, like other Smads, Smad6 MH1 domain binds to DNA and its MH2 domain inhibits the MH1 domain DNA binding activity. To determine the binding site of Smad6, we mutated several bases on the probe at the flanking regions of the Hoxc-8 binding site. The gel-shift results showed that the MH1 domain of Smad6 bound to all of the mutated probes (data not shown), suggesting that specificity of binding of the heterodimer is mainly determined by Hoxc-8. The interaction affinity of fulllength Smad6 with HDAC-1 appeared much weaker than that of Smad6 MH2 domain in immunoprecipitation assay. Again, it suggests that Smad6 MH1 domain and MH2 domain inhibit each other's function.

Smad6 MH1 Domain Is Required for the Formation of Hoxc-8-Smad6
Complex-Having identified that the Smad6 MH1 domain binds to DNA and the MH2 domain interacts with Hoxc-8 and HDAC-1, we mapped the domain(s) of Hoxc-8 that interacts with Smad6. Hox hexapeptide motif is highly conserved and located just upstream of the homeodomain, which is known as interacting with their partners (45)(46)(47)(48). To examine if Hoxc-8 hexapeptide motif is responsible for the interaction, we performed gel shift assays with Smad6 and a series of truncated Hoxc-8 GST fusion proteins (Fig. 4C). Because fulllength Smad6 does not bind to DNA, the homeodomain of Hoxc-8 was maintained for all the truncated Hoxc-8 proteins. The results were shown in Fig. 4D. All the proteins containing the hexapeptide motif, such as Hoxc-8NH, Hoxc-8HH, and full-length Hoxc-8, form shifted bands with Smad6. However, Hoxc-8HC, which does not contain the hexapeptide motif, failed to form a protein complex with Smad6. These results suggest that the hexapeptide motif of Hoxc-8 is important in interacting with Smad6.
Thus far, we know the interaction domains of both Smad6 and Hoxc-8. Incubation of these two interacting domains together with the same probe (OPN-5) should yield a supershift band. Unexpectedly, Smad6 MH2 domain did not form a complex with Hoxc-8HH. However, the full length of Smad6 formed a shifted band with Hoxc-8HH (Fig. 4E). These results strongly   (8,9). I-Smad interacts with the activated type I receptors, blocking the phosphorylation of R-Smads (6,7). Particularly, Smad6 competes with Smad4 to form an inactive complex with phosphorylated Smad1 (26). However, antagonistic function of I-Smads is not well characterized in the nucleus. I-Smads are localized in both the cytosol and the nucleus, and TGF-␤ or BMP stimulation does not change the cellular distribution of Smad6. Initially, we reported that Smad6 functions as a transcriptional co-repressor through interacting with Hoxc-8 (25). In this report, we demonstrated that both Smad6 and Smad7 were co-immunoprecipitated with HDAC-1, which is normally recruited by transcriptional repressors or co-repressors in inhibition of gene transcription. Importantly, we also find that Smad6 is a DNAbinding protein as a partner of Smad6/Hoxc-8 complex. Like other Smads, Smad6 binds to DNA through its MH1 domain, whereas Smad6 MH2 domain recruits HDAC-1. These findings revealed an important functional aspect of I-Smads in the nucleus.
Chromatin modification plays a critical role in regulating gene transcription. Acetylation is one of the major modification processes. HATs acetylate lysine residues on histone N-tails, leading to loosening of the chromatin structure and increasing the accessibility of transcriptional factors to the chromosome, whereas HDACs mediate silencing of gene transcription. R-Smads translocate to the nucleus upon BMP or TGF-␤ stimu- FIG. 3. The Smad6 and Hoxc-8 immunoprecipitated complexes contain HDAC activity. A, histones were acetylated by CBP intrinsic HAT domain. GST protein and GST-CBP (aa 1099 -1758 (CBP HAT domain GST fusion protein)) were bacterially expressed and subsequently purified. Each protein (3 g) was tested for HAT activity using the standard HAT assay as described under "Experimental Procedures." 3 H-Labeled acetyl-CoA incorporation into histones was determined by liquid scintillation counting. The acetylated histones were used as substrates in panel B. B, Smad6 and Hoxc-8 protein complexes have HDAC activity. FLAG-tagged Smad6 or empty pcDNA3 vector was co-transfected in COS-1 cells with HA-tagged HDAC-1, and the protein complexes were immunoprecipitated by anti-FLAG M2 monoclonal antibody (lanes 2, 5, 7, and 8). HAtagged Hoxc-8 or empty pcDNA3 vector was co-transfected in COS-1 cells with FLAG-tagged HDAC-1, and the protein complexes were immunoprecipitated with poly-HA (lanes 1, 3, 4, and 6). Anti-HA monoclonal antibody was used for immunoprecipitation, and anti-HDAC-1 polyclonal antibody was used for immunoblotting. D, HDAC-1 mediates the transcriptional repression of Smad6 and Hoxc-8. HDAC-1 (100 ng) plasmid was co-transfected in Mv1Lu cells with Hox-pGL3 (500 ng), Smad1B (300 ng), Smad6 (100 ng) and Hoxc-8 (25 ng). Cell lysates were assayed for luciferase activity normalized to Renilla luciferase levels at 48 h after transfection. Experiments were repeated three times in triplicate. lation, where they function as transcriptional activators through interacting and recruiting p300/CBP to specific gene promoters (34,41). TGIF, a Smad co-repressor in the TGF-␤ signaling pathway, interacts with HDAC-1 and inhibits target gene expression (34). HDACs inhibit gene transcription by tightening the chromosome structure. TSA, an HDAC inhibitor, rescued Smad6-and Hoxc-8-mediated transcription repression. Interaction of Smad6 and Smad7 with HDAC-1 and recruitment of endogenous HDAC activity clearly indicate the nuclear function of I-Smads as transcription repressors or co-repressors. Smad7 does not interact with Hox proteins (25). Therefore, Smad7 was not shown as a transcriptional repressor or co-repressor. However, the interaction of Smad7 with HDAC-1 suggests its potential function as a transcriptional repressor or co-repressor in different transcription mechanisms because other cytokines such as TNF-␣ induce Smad7 expression (49). Thus, Smad7 could also mediate cross-talk between TGF-␤ and other signaling pathways (49,50). The detailed mechanism of Smad7 in the nucleus remains to be characterized. The finding of Smad6 binding to DNA suggests that all of the Smads are DNA binding proteins. I-Smads contain both MH1 and MH2 domains, which are necessary for full activities of Xenopus Smad6 and Smad7 (51). The MH2 domain is highly conserved with other Smads, but the MH1 domain is distinct (44). MH1 domains of R-Smads and Smad4 bind DNA Smad box (GTCT) (41)(42)(43). Like other Smads, Smad6 binds to DNA also through its MH1 domain. The full length of Smad6 does not bind to DNA, whereas, in the presence of Hoxc-8, Smad6 binds to DNA and forms a heterodimer. Importantly, the MH1 domain of Smad6 is required for the Hoxc-8-Smad6 complex formation although it is the MH2 domain of Smad6 that interacts with Hoxc-8. It appears that the MH2 domain of Smad6 masks its MH1 domain DNA binding activity, and that the MH1 and MH2 domains of Smad6 interact reciprocally and inhibit each other's function. Indeed, Smad6 MH1 DNA binding activity decreased when the length of Smad6 truncated MH1 domain extended to MH2 domain. This observation suggests that Hoxc-8 induces Smad6 conformation change, in which Smad6 MH1 domain becomes available to bind to DNA. Phosphorylation of Smad6 could be another way to change its conformation for DNA binding activity, which could also serve as a cross-talk with other signaling pathways.
Taken together, we demonstrated a specific interaction of I-Smads with HDACs. As a model, we describe a transcription repression mechanism of Smad6. In response to BMP or TGF-␤ stimulation, Smad6 interacts with Hoxc-8 and binds to DNA as a heterodimer, which inhibits Smad1-induced gene transcription by recruiting HDACs. Importantly, Smad6 is also a DNAbinding protein. Like other Smads, Smad6 binds to DNA through its MH1 domain, and Smad6 MH2 domain interacts with HDAC-1 and Hoxc-8. Consistent with the observation of I-Smads cellular distribution, our data indicate that I-Smads can function as transcription repressors or co-repressors in the nucleus as antagonistic feedback loop of the TGF-␤ signaling pathway.