An essential role for Mad homology domain 1 in the association of Smad3 with histone deacetylase activity*.

The Smads are a family of sequence-specific DNA-binding proteins that modulate transcription in response to transforming growth factor beta (TGFbeta) by recruiting transcriptional activators like the histone acetyltransferase, p300/CBP, or repressors like the histone deacetylase, HDAC1, to TGFbeta target genes. The association of Smads and HDAC1 is mediated in part by direct binding of Smads to the HDAC1-associated proteins, TG-interacting factor, c-ski, and SnoN. Although ectopic expression of these proteins inhibits Smad-activated transcription, the contribution of histone deacetylase enzymatic activity to transcriptional repression by TGFbeta is unknown. Here, the biological requirements for the interaction between Smads and endogenous histone deacetylase activity are investigated. We identify residues in Mad homology domain 1 of Smad3 that are required for association with histone deacetylase activity. An amino acid change at one of these critical residues does not disrupt the association of Smad3 with c-ski, SnoN, and transforming growth-interacting factor but does abrogate the ability of Smad3 to repress transcription. These findings indicate that the association of Smad3 and histone deacetylase activity relies on additional protein mediators that make contact with Smad3 at its amino terminus. Moreover, these data suggest that the suppressive effect of Smad3 on transcription is dependent upon its association with histone deacetylase enzymatic activity.

kine that elicits a diverse array of cellular responses in a variety of cell types (1). The combined actions of these cellular responses are likely to mediate the more global effects of TGF␤, including its role in development, wound healing, immune responses, and the pathogenesis of cancer (2). TGF␤ initiates downstream signaling inside the cell through a heteromeric receptor complex containing the type I and type II TGF␤ receptors at the cell surface (3,4). In its activated state, the type I receptor directly binds and phosphorylates specific members of the Smad protein family (5)(6)(7)(8). Each member of this protein family contains two highly conserved globular domains, an amino-terminal Mad homology domain 1 (MH1) and a carboxyl-terminal Mad homology domain 2 (MH2), connected by a less conserved proline-rich linker sequence. Phosphorylated Smad2 and Smad3 associate with Smad4 and each other (9 -13). These Smad oligomers enter the nucleus and modulate the transcription of various TGF␤-responsive genes (2,14).
Studies of the Smad proteins have revealed much about the mechanism of TGF␤-mediated transcriptional modulation. By binding directly to Smad DNA consensus sites and to other promoter-bound transcription factors, the Smads activate the transcription of several genes in response to TGF␤. For instance, Smads directly bind to sites in the plasminogen activator inhibitor-1 (PAI-1) promoter itself and to the transcription factors, transcription factor meE3 and Sp1, on the promoter to activate PAI-1 transcription (15)(16)(17)(18)(19)(20).
Far less is known, however, about the mechanism through which TGF␤ represses the expression of target genes that maximally produce transcript in the absence of ligand. This set of genes includes osteocalcin (21)(22)(23), transin/stromelysin (24), and the proto-oncogene, c-myc (25). TGF␤-mediated transcriptional repression of these genes is important for many cellular effects of TGF␤s. For example, TGF␤-mediated inactivation of c-myc transcription is essential for the anti-proliferative capacity of TGF␤ in epithelial cells (25).
Given the established role of Smads as effectors of TGF␤mediated transcriptional regulation, we postulate that the Smads may mediate transcriptional repression of these genes. Although the molecular mechanism underlying this transcriptional repression remains largely unknown, the recent findings showing that Smads associate with several components of the histone deacetylase 1 (HDAC1) transcriptional repressor complex may provide a clue. It has been shown that recruitment of this large multi-nucleoprotein complex to DNA by site-specific DNA binding transcription factors induces localized regions of deacetylated chromatin, resulting in the tight packing of nucleosomes and the effective repression of transcription (for review see Refs. 26 and 27)). Several studies describe direct interactions between Smads and several HDAC1-associated proteins, TGIF, c-ski, and SnoN (28 -33). Ectopic expression of TGIF, c-ski, or SnoN blocks the ability of TGF␤ to modulate transcription. Furthermore, c-ski and TGIF were shown to compete with the co-activator, p300, for binding at the MH2 domain of Smad3. Therefore, the reported inhibitory effects of these HDAC1-associated proteins on TGF␤-activated transcription may result from competition between transcriptional co-activators and co-repressors for Smad binding.
Although these studies establish a physical link between the Smads and HDAC1, the contribution of Smad-mediated recruitment of histone deacetylase enzymatic activity to specific promoters has not been evaluated. To address this issue, we show that the ability of Smad3 to associate with endogenous HDAC enzymatic activity correlates with its ability to repress transcription. Using random mutagenesis and "reverse" twohybrid analysis, a cluster of residues in the MH1 domain of Smad3 that are critical for the ability of Smad3 to associate with HDAC enzymatic activity are identified. A mutation at one of these residues does not affect the ability of Smad3 to activate transcription or to bind TGIF, c-ski, or SnoN in vitro. Surprisingly, this same mutation does abrogate the ability of Smad3 to repress transcription. Together, these results indicate that the association of Smad3 with histone deacetylase activity may require interactions with an unidentified HDACassociated factor(s) at its amino terminus and that this association is critical for Smad3 to repress constitutive transcription. Moreover, these findings suggest that Smad3 may have distinct functional motifs associated with either the activation or repression of transcription.

EXPERIMENTAL PROCEDURES
Cell Culture-293T and Mv1Lu cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, penicillin, and streptomycin.
Yeast Two Hybrid Screen-A yeast two-hybrid screen of a HaCaT cell cDNA library (CLONTECH) was performed using Smad3 as bait according to the CLONTECH Matchmaker protocol and as described (35). Briefly, an Hf7c yeast line carrying Smad3 in the Gal4 DNA binding (DBD) domain fusion vector, pGBT9 (CLONTECH) and expressing DBD-Smad3 was transformed with the HaCaT cDNA library fused to the Gal4 activation (AD) domain in pACT2 (CLONTECH). 5 ϫ 10 6 transformants were screened for their ability to grow in the absence of histidine. 5 mM 3-aminotriazole was included in the media to reduce background HIS3 reporter activity. 242 transformants grew under these conditions and were positive for lacZ reporter gene activity. Bait dependence of the interaction was established by reintroducing fusion vectors for AD clone 349 with DBD alone, with DBD-Smad3, or with the DBD fused to an unrelated protein, G 12 , into Hf7c parental yeast. Co-transformants were re-screened for the HIS3 and lacZ reporter gene activities according to the CLONTECH Matchmaker protocol.
Reverse Two-hybrid--Performed according to the protocol developed by Vidal et al. (see Ref. 36). The MaV103 strain of Saccharomyces cerevisiae carries the SPAL10::URA3 marker. Yeast two-hybrid protein-protein interactions confer sensitivity to 5-fluoorotic acid (FOA) in yeast carrying this marker. A clonal line of MaV103 that expressed AD-349 isolated from the yeast two hybrid screen was transformed with 300 ng of the mutagenized Smad3 library and 150 ng of BamHI-linearized pGBT9. Several thousand transformants were first selected for growth in the presence of 0.1% FOA. These FOA-resistant transformants were tested for ␤-galactosidase (␤-gal) activity using the 5-bromo-4-chloro-3-indolyl-␤-D-galactoside (X-gal) filter lift assay (CLONTECH MatchMaker protocol). FOA-resistant lines with lacZ reporter activities less than wild type Smad3 were tested for Smad3 protein expression by Western analysis with anti-Gal4 DBD (Santa Cruz) monoclonal antibody. Clonal lines that expressed full-length, DBD-Smad3 fusion proteins at levels approximately equal to the wild type DBD-Smad3-expressing control line were considered positives. The Smad3 cDNAs from these lines were rescued and reintroduced with AD 349 into parental MaV103. Cotransformants were selected on synthetic dropout without tryptophan or leucine. Smad3 alleles with ␤-gal levels less than wild type Smad3 were sequenced.
Reporter Assays-Luciferase assays were performed as previously described (38). LipofectAMINE (Life Technologies, Inc.) transfections were performed using the manufacturer's protocol. All transfections were normalized to ␤-galactosidase activity by co-transfection of 0.5 g of a cytomegalovirus ␤-galactosidase (CMV-␤-gal) expression vector. Twelve hours after transfection, 10 ng/ml TGF␤-1 in Dulbecco's modified Eagle's medium containing 0.1% fetal bovine serum was added. Luciferase activity was measured 20 -24 h later. Chloramphenicol acetyltransferase assays were performed according to a standard protocol (39,40). 293T cells were transfected for 7-12 h using calcium phosphate. 24 h later cells were assayed for chloramphenicol acetyltransferase activity.

RESULTS
The Smads Associate with HDAC1 in Yeast-To elucidate the mechanism of Smad-mediated transcriptional modulation of TGF␤-responsive genes, we set out to isolate transcription factors that bind to the Smads. A yeast two-hybrid screen of a human kerationcyte (HaCaT) cDNA library was performed using Smad3 as a bait. A cDNA clone (AD-349) encoding amino acids 211-482 of the histone deacetylase HDAC1 was isolated (Fig. 1A). Western analysis of yeast lysates confirmed that the expected protein product was produced (data not shown). To verify bait dependence of the interaction, AD-349 was reintroduced into the parental yeast line Hf7c in combination with expression vectors encoding the DBD alone or protein fusions of DBD-Smad3 or DBD-Smad4. As a positive control for interaction, DBD-p53 and AD-T antigen were introduced into the parental yeast strain. As negative controls for the interaction, either DBD-Smad3, DBD-Smad4, or the AD empty expression vector were also introduced into Hf7c. Transformants were selected on synthetic dropout media lacking tryptophan and leucine.
The Hf7c parental line carries two reporter genes for protein interaction, HIS3 and lacZ, each containing multiple consensus Gal4 DNA binding sequences upstream of their respective coding regions. Thus, growth on media lacking histidine and elevated ␤-gal activity are indicative of fusion protein interaction. As shown in Fig. 1B, both DBD-Smad3-and DBD-Smad4expressing cells survive in the absence of histidine only if AD-349 is present, suggesting that Smad3 and Smad4 associate with amino acids 211-482 of HDAC1. The inability of cells expressing DBD and AD-349 to survive confirms the bait dependence of these interactions. Similar results were found by ␤-gal activity assays (data not shown).
In vitro binding experiments show that the Smad/HDAC1 interaction was not direct ( Fig. 5C and data not shown) but may have been mediated by highly conserved yeast homologs to components of the HDAC repressor complex. Conserved proteins in this complex include yeast homologs to Sin3A, HDAC1, HDAC2, RbAP46, RbAP48, SAP18, and SAP30 (41)(42)(43)(44)(45). Indeed, yeast proteins have been attributed with bridging between bait and prey in other yeast two-hybrid screens (46). Consistent with this possibility, Smad2, Smad3, and Smad4 bound directly to Sin3A in vitro (data not shown and Fig. 5C).
Smads Associate with Histone Deacetylase Enzymatic Activity in Mammalian Cells-These data support the findings of several studies that describe interactions between the Smads and other HDAC1-binding proteins, TGIF, c-ski, and SnoN. Taken together, these results indicate that the Smads may make direct contact with several members of the repressor complex, with the functional end point being an association with histone deacetylase enzymatic activity.
To determine whether the Smads actually complex with an active HDAC enzyme, we measured the amount of endogenous histone deacetylase activity associated with the Smads. The method used to show that the transcriptional repressor Mad1 and the Sin3A and N-CoR/SMRT co-repressors associate with endogenous HDAC enzymatic activity was employed ( Fig. 2A) (47)(48)(49)(50). Carboxyl-terminal FLAG-tagged Smads were overexpressed in 293T cells and immunoprecipitated with an antibody directed against the FLAG epitope. These immunoprecipitated complexes were incubated with [ 3 H]acetate-labeled histones, and the amount of histone deacetylase activity was recorded as the number of dpm released from the reactions. Smad1, Smad2, Smad3, and Smad4 co-immunoprecipitated with significant levels of histone deacetylase activity that could be inhibited upon treatment with trichostatin A, a reversible inhibitor of histone deacetylases (51). Smad1 and Smad2 consistently associated with less histone deacetylase activity than either Smad3 or Smad4. The levels of HDAC activity bound to Smad3 and Smad4 were ϳ4-fold less than that associated with the HDAC1 binding partner and transcriptional repressor, Sin3A. A FLAG immunoblot of the immunoprecipitations confirmed that trichostatin A had no FIG. 1. Smad3 and Smad4 associate with HDAC1 in yeast. A, cDNA clone (349) encoding the carboxyl-terminal portion of the HDAC1 cDNA was isolated in a yeast two-hybrid screen using Smad3 as bait. A schematic depiction of clone 349 coding sequence compared with full-length HDAC1 is shown. B, Smad3 and Smad4 associated with clone 349 in S. cerevisiae. Protein interaction of DBD fusion proteins with AD fusion proteins activate the transcription of the HIS3 and lacZ marker genes in the S. cerevisiae strain, Hf7c. Here we show that the interaction of DBD-Smad3 or DBD-Smad4 with AD-349 activates the HIS3 reporter. Yeast strain Hf7c was transformed with the yeast expression vectors encoding DBD-Smad3 and either AD or AD-349, DBD-Smad4, and either AD or AD-349 and DBD with AD-349 according to the CLONTECH Matchmaker protocol. As a positive control for interaction, DBD fused to amino acids 72-390 of p53 (DBD-p53) and AD fused to amino acids 84 -708 of SV40 large T antigen (AD-T antigen) were also included. Transformants were isolated on synthetic dropout media that lacks both tryptophan and leucine to select for cells carrying both a DBD fusion vector and the AD fusion vector, respectively. These cells were streaked onto synthetic dropout media that lacked tryptophan and leucine and either contained (ϩhistidine) or lacked histidine (Ϫhistidine). 5 mM 3-aminotriazole was included in the Ϫhistidine plate (see "Experimental Procedures") WT, wild type. effect on the amount of Smad protein immunoprecipitated from each sample (Fig. 2B).
Screen for Loss-of-function Mutations in Smad3-We reasoned that the identification of the amino acids in Smad3 required for its association with histone deacetylase activity would provide us with tools to dissect the role of histone deacetylases in Smad-directed transcription. To identify specific residues in Smad3 involved in the association between Smad3 and HDAC1, a reverse two-hybrid screen of a randomly mutagenized Smad3 library was performed. The entire Smad3 cDNA was mutated using random PCR mutagenesis. These PCR products were then introduced into a MaV103 yeast line expressing the AD-349 clone isolated in the screen for Smad3 binding partners (Fig. 1). The MaV103 yeast strain carries the SPAL10::URA3 marker in addition to the Gal4::lacZ and Gal4::HIS3 markers previously described (36). Yeast two-hybrid protein-protein interactions confer sensitivity to FOA in lines carrying this reporter. Several thousand transformants were screened for their ability to grow in the presence of 0.1% FOA and their inability to produce a blue color in the presence of X-gal. From this set of positives, clones that expressed full-length DBD-Smad3 proteins to a level similar to the wild type Smad3 line were selected. In all, nine clones met the selection criteria. A map of these mutants is depicted in Fig. 3. As a measure of lacZ reporter gene activity, the level of ␤-gal activity produced by yeast expressing AD-349 and each Smad3 mutant is shown.
Confirming the validity of the screen, ␤-gal activity for each Smad3 mutant allele was significantly reduced compared with that of wild type. Although reporter gene activity for some of the mutations is approximately half that of wild type Smad3, other mutant alleles, particularly L38P and N123D/I298M, give ␤-gal activities comparable with negative controls for the interaction. Notably, all of the clones contain a missense mutation in the MH1 domain of Smad3, suggesting that the MH1 domain may make contact with HDAC1-binding proteins. One particular amino acid, Tyr-127, was mutated in two different clones, highlighting its importance for the interaction. Two other mutants, Smad3F111S/W394R and Smad3N123D/ I298M, contain two missense mutations, one in the MH1 domain and one in the MH2 domain. A subset of these mutations displaying the lowest ␤-gal activities was subcloned into mammalian expression vectors to be studied further.
Characterization of Smad3 HDAC Activity Mutants-Given that the yeast two-hybrid interaction between Smad3 and HDAC1 may be dependent upon yeast proteins, the HDAC1containing complexes that associate with Smad3 in our yeast two-hybrid interaction may be very different from those that associate with Smad3 in mammalian cells. To determine if these mutations have any effect on the association of Smad3 with endogenous histone deacetylases in human cells, we measured the level of histone deacetylase activity associated with individual mutants in 293T cells (Fig. 4A). Approximately equal amounts of wild type and mutant Smad3-FLAG fusion proteins were immunoprecipitated, suggesting that each mutant is stable in mammalian cells. In this assay, each mutant was significantly deficient in its ability to associate with HDAC activity, demonstrating the importance of these particular MH1 domain residues in the association of active histone deacetylases with Smad3.
Similar assays using Smad3 deletion mutants confirmed the importance of the MH1 domain for the coupling of HDAC activity with Smad3. Smad3NL, Smad3C, and Smad3⌬c were overexpressed, immunoprecipitated, and analyzed for HDAC activity (Fig. 4, B and C). Strikingly, Smad3NL associated with 3-fold more HDAC activity than full-length Smad3. The level of HDAC activity associated with Smad3NL was comparable with Sin3A-bound HDAC activity, confirming that Smad3 binds biologically relevant levels of HDAC activity through its MH1 domain. Smad3⌬c co-immunoprecipitated with levels of associated HDAC activity similar to full-length Smad3. In contrast, Smad3C associated with little or no HDAC enzymatic activity.
Examination of the crystal structure of Smad3 MH1 domain revealed that many of these mutations alter amino acids in the hydrophobic core of the protein (52). Thus, some of these amino acid changes may lead to gross conformational changes in the MH1 domain that block several Smad3 functions such as DNA binding, nuclear entry, Smad4 binding, and transcriptional activation. To address these issues, the ability of Smad3 and the Smad3 mutants to activate the 4xSBE reporter was assessed (Fig. 5A). This reporter contains Smad consensus DNA binding sites (SBEs) to which Smad3 and Smad4 directly bind

Smad3 Binds Histone Deacetylase Activity to Repress Transcription
and activate transcription. In this assay, the ability of mutant L38P to activate transcription was severely inhibited. In contrast, F111S/W394R, Y125C, and Y127C activated transcription to levels similar to those of wild type Smad3. Activation of transcription by P95S and N123D/I298M was partially hindered. Mutated amino acids that are accessible to the surface of the wild type protein are noted (Fig. 5B). Taken together, these data suggest that unlike other mutations isolated from the reverse two-hybrid screen, the F111S and Y125C mutations might disrupt a discreet binding surface for HDAC1-binding proteins without inhibiting the DNA binding, nuclear entry, Smad4 cooperation, and transcriptional activation functions of Smad3 mutants were screened for their inability to associate with AD-349. The location of each mutation in the Smad3 sequence is noted by the wild type residue, codon number, and the mutant residue. Note that two alleles contain double mutations, F111S/W394R and N123D/I298M indicated by * and #, respectively. In addition, residue Tyr-127 was mutated in two different clones. The interaction of each mutant Smad3 allele with AD-349 was quantified as a function of GAL1::lacZ reporter gene expression as determined by ␤-gal activity assays using chlorophenol red-␤-D-galactopyranoside as substrate (CLONTECH MatchMaker protocol). Values for the AD-349/DBD-Smad3 wild type interaction were standardized to 100, and reporter activity for each interaction is shown as a percentage of this standard. Reporter activities for AD-349ϩDBD alone and AD aloneϩDBD-Smad3 are shown as negative controls for the interaction. Reporter activity for AD-T antigen ϩ DBD-p53 is shown as a positive control, and reporter activity for ADϩDBD is shown as a negative control for the assay. Error indicates the S.D. of ␤-gal activities from two different transformant colonies.

FIG. 4. The association of Smad3 with endogenous histone deacetylase activity requires specific residues in the MH1 domain.
Smad3 mutant alleles were subcloned into the mammalian expression vector, pRK5FLAG, and assayed for their ability to co-immunoprecipitate with histone deacetylase activity as described under "Experimental Procedures" and in Fig. 3. A, a graph

Smad3 Binds Histone Deacetylase Activity to Repress Transcription
Smad3. Supporting this notion, the MH1 domains of bacterially produced GST-Smad3NL-F111S and GST-Smad3NL-Y125C retain the ability to directly associate with the SBE by electrophoretic mobility shift assay (data not shown).
c-ski, SnoN, and TGIF have been shown to mediate the association of specific Smads with HDAC1. The inability of the mutant clones derived from the reverse two-hybrid screen to associate with HDAC activity in mammalian cells may be the result of deficient association with these co-repressor proteins. To address this issue, the ability of in vitro TNT c-ski, SnoN, and TGIF to bind GST fusion proteins of Smad3, Smad3F111S/ W394R, and Smad3Y125C was assessed (Fig. 5C). The inability of GST-Smad3 to bind TNT-HDAC1 was mentioned previously and is shown here as a negative control for the binding assays. As a positive control, the results of in vitro binding reactions between each Smad3 mutant and TNT-c-Jun is shown (35,53). c-Jun forms critical contacts with lysine 41 in the MH1 domain of Smad3 (54). Although Smad3F111S/W394R bound TNT-c-Jun with similar efficiency as wild type Smad3, consistently the association of Smad3Y125C and TNT-c-Jun was slightly diminished. Therefore, at least under in vitro conditions, the Y125C mutation may cause minor alterations to overall protein structure rather than solely affecting a discreet surface. Consistent with this observation, the ability of Smad3Y125C to associate with other TNT products was slightly diminished. In contrast, no consistent difference between the affinities of Smad3 and Smad3YF111S/W394R for TNT-c-ski, TNT-SnoN, and TNT-TGIF was observed, indicating that direct binding to c-ski, SnoN, and TGIF is unaffected by this mutation.
Transcriptional Repression by Smad3 HDAC Activity Mutants Is Deficient-Based on evidence showing a connection between HDAC activity and the repression of transcription, we postulated that the ability of the Smads to associate with HDAC activity would allow them to repress constitutive transcription. To quantify Smad-mediated repression, a reporter construct containing GAL4 DNA binding sites upstream of the constitutively active SV40 promoter was employed (Fig. 6A). Previous reports show that Gal4 fusions of the transcriptional repressors, Mad1 and Rb, inhibit the expression of this same reporter (55,56). In this assay, Smad3 and Smad4 significantly repressed transcription when tethered to DNA (Fig. 6, B and C, and data not shown). Reporter activity was repressed ϳ8-fold further by Gal4-Smad3NL but was unaffected by Gal4-Smad3C expression.
If the ability of Smad3NL to repress transcription was indeed dependent upon its association with HDAC enzymatic activity, we predicted that repression of reporter gene expression by Smad3 HDAC activity mutants would be deficient. To address this, the effect of the NL domains of the Smad3Y125C and Smad3F111S HDAC activity mutants on reporter gene activity was assessed (Fig. 6D). Consistent with a role for the Smad3-HDAC interaction in the repression of active transcription, the ability of each mutant to repress reporter gene expression was significantly inhibited compared with wild type Smad3NL. In contrast, the expression level of each mutant was similar to wild type Smad3 NL (Fig. 6E). DISCUSSION The ability of TGF␤ to repress the expression of critical cell cycle regulators like c-Myc indicates that transcriptional repression by TGF␤ may contribute to TGF␤-mediated inhibition of cellular proliferation. Thus, understanding the mechanisms through which TGF␤ represses transcription may provide insight into the anti-proliferative capacity of TGF␤ and the role of Smads in tumor suppression. To this end, we show that Smad1, Smad2, Smad3, and Smad4 associate with endogenous histone deacetylase activity. Thus, Smad⅐HDAC complexes are enzymatically active and may couple histone deacetylase activity with nuclear Smad activities like DNA binding. A mutant lacking the MH2 domain, Smad3NL, associated with 3-fold more histone deacetylase enzymatic activity than Smad3. The amount of Smad3NL-associated activity is comparable with the level of HDAC activity that co-immunoprecipitated with the transcriptional repressor and direct HDAC1-binding protein, Sin3A. Thus, Smad3 associates with levels of HDAC activity that are biologically significant in the repression of transcription.
The difference in HDAC activity isolated from Smad3 fulllength and Smad3NL immunoprecipitations suggests that the MH2 domain may inhibit the association of histone deacetylases with full-length Smad3. A previously characterized autoinhibitory interaction between the MH1 and MH2 domains may be responsible for this effect. Activation of Smad3 by TGF␤ is thought to interrupt the binding of the MH1 and MH2 FIG. 5. Effects of the mutations on Smad3 function. A, quantification of the ability of mutant Smad3 alleles to activate transcription from Smad DNA binding sites. Either Smad3/pRK5F (0.5 g), mutant Smad3/pRK5F (0.5 g), or empty CMV vector, pCMV5 (0.5 g) was co-transfected with Smad4/pRK5F (0.5 g) into Mv1Lu cells along with the 4xSBE-lux (1.0 g) and CMV-␤-gal (0.5 g) reporters described under "Experimental Procedures." Total transfected DNA was kept constant with pCMV5. Luciferase values were normalized to ␤-galactosidase activity to control for transfection efficiency, giving relative luciferase units domains to each other and may facilitate the association of Smads with HDAC-binding proteins (57). Alternatively, histone acetyltransferase activity of p300/CBP bound at the MH2 domain of Smad3 (58 -62) may effectively reverse the effects of co-immunoprecipitated HDACs, resulting in reduced Smadassociated HDAC activity levels in the presence of the MH2 domain of Smad3 and Smad3⌬c.
We have further dissected this endogenous HDAC⅐Smad3 complex by identifying particular residues in Smad3 that are required for Smad3 to either associate with the endogenous HDAC complex or affect the enzymatic activity of the bound HDAC complex. Consistent with the Smad3 deletion analysis, all of the isolated mutants contain sequence alterations in the MH1 domain of Smad3, revealing the importance of this domain in the coupling of HDAC enzymatic activity with Smad3. Amino acid residues Leu-38, Tyr-88, and Val-122 interact with each other and constitute a portion of the MH1 domain hydrophobic core (52). Asn-123 and Tyr-127 are also buried within the MH1 domain and are conserved among Smad members. Mutations at these residues most likely alter the shape of the entire MH1 globular domain and in this way alter the association of Smad3 with the endogenous HDAC activity. Three mutations in the MH1 domain, P95S, F111S, and Y125C, however, change conserved surface residues. They are located on the side of the MH1 domain opposite the DNA binding surface and may define an area of co-repressor protein contact. Pro-95 and Tyr-125 are a part of the "double loop" region defined by crystallographic analysis (52). This region, characterized by an extensive surface area that includes the L2 and L4 loops, has been proposed to contribute to macromolecular interactions. Phe-111 is located on an adjacent surface that is of particular interest because it contains the hydrophobic Phe-111 residue surrounded by several hydrophilic residues, Glu-110, Lys-117, Asp-118, and Glu-119. The structural distinctiveness of this surface suggests it may, indeed, contribute toward a proteinprotein interface.
In vitro binding of TGIF, c-ski, or SnoN to the HDAC activity mutant identified in these studies, Smad3F111S/W394R, was essentially unaffected. It is possible that, TGIF, c-ski, and SnoN are not present in these complexes, but they do make up and may be essential for the formation of other Smad⅐HDAC complexes not isolated or characterized in these assays. Supporting this possibility, our Smad3 HDAC activity mutants were based on an interaction between Smad3 and HDAC1 in S. cerevisiae, an organism with no homologues to TGIF, c-ski, or SnoN. Furthermore, previous studies show that the Smad MH2 domain rather than the MH1 domain makes contact with c-ski and SnoN (29 -32, 63, 64). We demonstrate here, however, that Smad3NL, a Smad3 fragment lacking the MH2 domain, associates with biologically relevant levels of HDAC enzymatic activity.
Alternatively, TGIF, c-ski, and SnoN do participate in the HDAC complexes isolated in the studies presented here, but their association with Smad3 is not sufficient to mediate the interaction between Smad3 and HDAC activity when HDAC binding at the MH1 domain is blocked. In this scenario, TGIF, c-ski, and SnoN association at the MH2 domain may serve to stabilize the essential MH1 interaction. Competition with p300 for binding at the MH2 domain may have masked this role of TGIF, c-ski, and SnoN in the HDAC activity assays reported here. Indeed, previous reports show that p300/CBP competes with TGIF and c-ski for Smad binding (32,65). If p300/CBP does compete with TGIF, c-ski, and SnoN in our studies, we predict that inhibition of p300 binding in these assays may result in the co-immunoprecipitation of even more HDAC activity with Smad3 and Smad3⌬c. In contrast, a block of Smad3C/p300 association is not expected to elevate HDAC activity levels associated with Smad3C since the MH1 domain residues essential for the interaction between Smads and HDAC activity are not present in this particular mutant.
In the transcriptional assay used here, the Smads are tethered to a viral promoter that is constitutively transcribed. We show that Smad3 represses constitutive transcription from this promoter. Inconsistent with previous reports that Smad-activated transcription is blocked by Smad binding partners, transcriptional repression by Smad3 in this assay was independent of TGF␤-activated or Smad-activated transcription (28, 30 -32, 64, 66). In the context of this assay, transcriptional repression by Smad3F111S and Smad3Y125C was significantly abrogated compared with wild type Smad3. Consistent with these findings, Smad3C, which did not associate with HDAC activity, did not repress transcription. Together, these data suggest that, in the context of this assay, the HDAC enzymatic activity of the Smad3⅐HDAC complex is required for the repression of constitutive transcription by Smad3.
The ability of the Smad3 HDAC-activity mutants to repress FIG. 6. The ability of Smad3 mutants to repress transcription is inhibited. A, map of the pSVEC reporter used in B and D with functional domains noted. B, reporter activity in the presence of Smad3 or Smad3 deletion mutants. Approximately 2 ϫ 10 6 293T cells were co-transfected with pSVEC (1.0 g) and CMV-␤-gal (0.5 g) reporter constructs and 5 g of either pRK5Gal4, Smad3/pRK5Gal4, Smad3NL/ pRK5Gal4, or Smad3C/pRK5Gal4 or as a positive control, Gal4-Mad1, into 239T cells. 36 h post-transfection the cells were lysed and normalized for transfection efficiency. Normalized lysates were assayed for chloramphenicol acetyltransferase activity and analyzed by Western blot (C). C, Western analysis of normalized lysates in B using anti-Gal4DBD monoclonal antibody (Santa Cruz, sc-510). D, reporter activity in the presence of Smad3NL or Smad3NL HDAC1 binding mutants. 293T cells were transfected as in B with pSVEC (1.0 g) and CMV-␤-gal (0.5 g) reporter constructs and 5 g of pRK5Gal4, Smad3NL/ pRK5Gal4, or Smad3NL/pRK5Gal4 mutants. Lysates were processed as in B. E, Western analysis of lysates normalized for transfection efficiency using the sc-510 antibody as in C.
transcription was, however, partially retained. This may simply reflect the nature of the SV40 promoter that is sensitive to HDAC-independent modes of repression (56). Thus, a portion of the inhibitory effect of Smad3 on reporter gene activity may be independent of HDACs, suggesting that in addition to the recruitment of histone deacetylase activity to a particular promoter, the Smads may repress constitutive transcription through HDAC-independent mechanisms. Nevertheless, a significant portion of Smad3-mediated transcriptional repression found here is dependent upon the association of Smad3 with histone deacetylase activity. The finding that the repressiondeficient mutants, Smad3F111S/W394R and Smad3Y125C, retain the ability to activate 4xSBE transcription indicates that the activation and repressive effects of Smad3 on transcription may exist as two separable functions of the Smad3 protein. We envision a model where, under specific promoter contexts, association of Smad3 and p300/CBP is favored and transcription is activated. The competition between p300/CBP and HDACassociated repressors would determine the extent of transcriptional activation. On other promoters, HDAC complexes associate with Smad3 at the MH1 domain and repress transcription. This mode of Smad3-mediated transcriptional repression may not involve the competition between co-activators and co-repressors. Supporting this model, loss of association with HDAC enzymatic activity was not found to result in a hyperactivation of 4xSBE-mediated transcription by Smad3F111S or Smad3Y125C. Instead, 4xSBE-mediated transcription is activated similarly by wild type Smad3, and each repression-deficient Smad3 mutant examined in this study.
Our data suggest that an interaction between the MH1 domain of Smad3 and an unknown but conserved HDAC-binding factor is required for the coupling of Smad3 to histone deacetylase enzymatic activity. A plausible candidate for this factor is the co-repressor Sin3A, which has a highly conserved homologue in yeast and directly binds to HDAC1. In vitro binding experiments with TNT-Sin3A, however, demonstrate that Smad3F111S/W394R and Smad3Y125C associate with Sin3A as efficiently as wild type Smad3. Thus, the identity of this factor remains unknown. Furthermore, whether Smad-associated histone deacetylase activity is the effect of one specific HDAC complex or the net effect of many distinct complexes, each containing a specific combination of HDACs and co-repressor proteins, also remains unclear (68). Recruitment of distinct HDAC complexes by the Smads may facilitate promoter-specific transcriptional repression. Future studies are required to further characterize the components of Smad⅐HDAC complexes and any distinct functions they may have.
Taken together, these findings suggest that by targeting HDAC1 activity to TGF␤-repressed genes, the Smads may directly mediate the repression of transcription of target genes, like c-myc, in response to TGF␤. Future in vivo studies of such TGF␤ target genes using the Smad3 HDAC activity mutants identified here will elucidate the contribution of the Smad-HDAC interaction to the transcriptional modulation of a particular target gene by TGF␤.