c-Ski acts as a transcriptional co-repressor in transforming growth factor-beta signaling through interaction with smads.

Smads are intracellular signaling mediators of the transforming growth factor-beta (TGF-beta) superfamily that regulates a wide variety of biological processes. Among them, Smads 2 and 3 are activated specifically by TGF-beta. We identified c-Ski as a Smad2 interacting protein. c-Ski is the cellular homologue of the v-ski oncogene product and has been shown to repress transcription by recruiting histone deacetylase (HDAC). Smad2/3 interacts with c-Ski through its C-terminal MH2 domain in a TGF-beta-dependent manner. c-Ski contains two distinct Smad-binding sites with different binding properties. c-Ski strongly inhibits transactivation of various reporter genes by TGF-beta. c-Ski is incorporated in the Smad DNA binding complex, interferes with the interaction of Smad3 with a transcriptional co-activator, p300, and in turn recruits HDAC. c-Ski is thus a transcriptional co-repressor that links Smads to HDAC in TGF-beta signaling.

development of Xenopus. FAST-1 was isolated as a factor that binds to the activin-responsive element in Mix.2 (11). Smad2 interacts with FAST-1 in a ligand-dependent manner, and the FAST-1-Smad2-Smad4 complex is required for the expression of Mix.2. DNA binding motifs for several Smads have been revealed. Drosophila Mad was shown to bind to GC-rich sequences in genes induced by Decapentaplegic (Dpp), a BMP homologue in Drosophila (12). A palindromic sequence, GTCTAGAC, was identified as a consensus Smad-binding site through polymerase chain reaction-based screening of random sequences (13). Another sequence containing CAGAC has been found as a Smad binding motif in a variety of TGF-␤-responsive genes (14,15). The MH2 domain possesses intrinsic transactivation activity (16). Smad3 was shown to act as a transcriptional co-activator for vitamin D receptor through its direct association with vitamin D receptor (17).
Recently Smads are shown to interact with transcriptional co-activators such as p300 and CBP (8). p300 and CBP are structurally related proteins that possess histone acetyltransferase (HAT) activity. p300/CBP neutralizes positive charges of histones in chromatin by acetylation, thereby loosening the chromatin structure and enhancing the accessibility of the transcription factors to the target DNA. p300/CBP also recruits the basal transcription machinery. E1A, an adenoviral oncoprotein, inhibits TGF-␤ signaling. E1A binds to and antagonizes p300/CBP and, in addition, interacts with R-Smads and interferes with the p300/CBP-Smad interaction. 2 p300/CBP interacts with a wide variety of transcription factors through several distinct domains. p300 has been shown to integrate Smad and STAT signaling pathways through concomitant interaction with the two proteins (18). In contrast, histone deacetylases (HDs) with activity opposing that of HATs repress transcription. At least six HDs (HDAC1-6) have been identified in humans (19). For example, Mad (different from the Drosophila Smad described above), an antagonist of c-Myc, represses transcription by recruitment of HDAC through its interaction with mSin3 (20). TGF-␤ down-regulates the expression of cdc25A, which is likely to be one of the mechanisms of growth regulation by TGF-␤. HDAC1 has been implicated in the transcriptional repression of cdc25A (21).
We conducted a yeast two-hybrid screen to identify proteins that interact with Smad2. One of the positive clones encoded c-Ski. c-ski is the cellular counterpart of the v-ski oncogene isolated from the Sloan-Kettering retroviruses (22). c-Ski has been shown to induce the expression of myogenic markers (23,24). c-Ski also represses the activity of a reporter gene containing c-Ski-binding sites (25) and retinoic acid receptor-mediated transactivation (26). More recently, c-Ski was shown to recruit a co-repressor mSin3 associated with HDAC (27). We demonstrate that c-Ski interacts with Smad2 and Smad3 in a TGF-␤-dependent manner but not with BMP-specific R-Smads. Smad4 is also incorporated in the complex. c-Ski strongly inhibited the transcription of various reporter genes activated by TGF-␤ but not transactivation by BMP stimulation. c-Ski is tethered to Smad-binding sites, recruiting HDAC1. c-Ski and p300 bind to Smad3 in a competitive manner. Finally, we show that c-Ski suppresses Smad-associated HAT activity. c-Ski thus counteracts with p300 and acts as a transcriptional corepressor for Smads in the TGF-␤ signaling pathway.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The constructions of plasmids containing Smads, receptors, and p300 have been described (7,28). The mammalian expression plasmids for the deletion mutant B (del B) (amino acids 491-728) of c-Ski was obtained by digesting the original two-hybrid clone with EcoRI and XhoI and following subcloning into FLAG-pcDNA3 (7). The deletion mutant A (del A) (amino acids 338 -728) of c-Ski was constructed by subcloning the EcoRI-EcoRI fragment (amino acids 338 -490) from the two-hybrid clone into the EcoRI site of del B. The full-length c-Ski for mammalian expression was constructed as follows. The NcoI-SalI fragment from the original c-Ski cDNA (gift of N. Nomura) was subcloned between NcoI and XhoI sites in pSKiMOD13 (28) to add an EcoRI site to the 5Ј-end of the coding region. The resulting EcoRI-EcoRI fragment (amino acids 1-490) was subcloned into the EcoRI site of del B. In order to make the deletion mutant C (del C) (amino acids 338 -490), the EcoRI-EcoRI fragment of the two-hybrid clone was subcloned into pSKiMOD13 to add a stop codon and an XhoI site at the C-terminal end. Then the NotI-XhoI fragment was inserted between the corresponding sites of del A. The deletion mutant D (del D) (amino acids 1-490) was obtained by digesting the full-length c-Ski with EcoRI and subcloning the resulting fragment into del C digested with EcoRI. The deletion mutant E (del E) (amino acids 1-309) was obtained by digesting the full-length c-Ski with EcoRI and ApoI followed by subcloning the resulting fragment into del C digested with EcoRI. c-Ski expression plasmids for the two-hybrid assay were constructed by subcloning each deletion mutant between EcoRI and XhoI sites of pJG4-5 (29) or into the EcoRI site of pJG4-53 (30). Various Smads in pEG202 (29) used as baits were constructed as Smad3 in pEG202 (28).
Yeast Two-hybrid Screen and Assay-Two-hybrid screen with fulllength Smad2 was conducted essentially as described (29). Briefly, EGY48 was transformed with Smad2 in pEG202 and a cDNA expression library made from serum-starved WI-38 cells (gift of R. Brent) in pJG4-5. Smad2 as a bait has moderate background activity, and thus the selection was done only by ␤-galactosidase assay. The inserts of the positive clones were isolated and sequenced.
Cell Culture and Plasmid Transfection-COS-7 cells and mink R mutant lung epithelial cells were cultured in Dulbecco's modified Eagle's medium (Sigma) containing 10% fetal bovine serum and 100 units/ml penicillin. P19 cells (gift of T. Momoi) were cultured in ␣-minimal essential medium containing 10% fetal bovine serum and 100 units/ml penicillin. The cells were maintained in humidified atmosphere with 5% CO 2 at 37°C. Transfection was performed using Fu-GENE 6 (Roche Molecular Biochemicals) according to the manufacturer's protocol.
Immunoprecipitation and Immunoblotting-COS-7 cells were used for the detection of protein-protein interaction in vivo. Cells were transfected with an appropriate combination of expression plasmids, washed, scraped, and solubilized in a buffer containing 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 1% aprotinin, and 1 mM phenylmethylsulfonyl fluoride, or IPH buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, 1% aprotinin, and 0.1 mM phenylmethylsulfonyl fluoride). Lysates were cleared and incubated with anti-FLAG M2 antibody (Sigma), followed by incubation with protein G-Sepharose beads (Amersham Pharmacia Biotech). The beads were washed with solubilization buffer, and the immunoprecipitates were eluted by boiling for 3 min in SDS sample buffer (100 mM Tris-HCl, pH 8.8, 0.01% bromphenol blue, 36% glycerol, 4% SDS) containing 10 mM dithiothreitol and subjected to SDS-gel electrophoresis. Proteins were electrotransferred to nitrocellulose filters, immunoblotted with anti-Myc 9E10 antibody, anti-HA antibody, anti-mSin3A antibody (Santa Cruz Biotechnology), anti-human HDAC1 antibody (Santa Cruz Biotechnology), or anti-FLAG M2 antibody. The bands were detected using the enhanced chemiluminescence detection system (Amersham Pharmacia Biotech). Some of the lysates were directly subjected to Western blotting without immunoprecipitation.
Luciferase Reporter Assay-Luciferase assays were carried out using various reporters and mink R mutant lung epithelial cells or P19 cells. Cells were transiently transfected with an appropriate combination of the reporter, expression plasmids, and pcDNA3 (Invitrogen). Total amounts of the transfected DNAs were the same throughout the experiments, and luciferase activities were normalized using the sea-pansy luciferase activity under the control of the thymidine kinase promoter.
HAT Assay-HAT assay was performed according to the method previously described (31) with minor modification. COS-7 cells were transfected with an appropriate combination of expression plasmids, washed with PBS, and resuspended in IPH buffer. The lysates were incubated on ice for 30 min and then cleared by centrifugation. The proteins were immunoprecipitated with anti-FLAG antibody as described above and suspended in 30 l of IPH buffer. The samples were incubated with 1.25 l of 20 mg/ml core histones (Sigma) and 1 l of 5 Ci/ml [1-14 C]acetyl-CoA (Amersham Pharmacia Biotech) at 30°C for 1 h. The reaction products were resolved by SDS-PAGE followed by detection using a Fuji BAS 2500 bio-imaging analyzer (Fuji Photo Film). The histones used in the assays were stained using Gel Code Blue stain reagent (Pierce).

c-Ski Interacts with Smad2/3 in a TGF-␤-dependent
Manner-To identify proteins that play a role in TGF-␤ signaling, we conducted a two-hybrid screen of a human cDNA library using full-length Smad2 as a bait. Among the positive clones, we identified several nuclear proteins such as JunB that are likely to act as a nuclear partner of Smad2 (data not shown). One of such clones encoded c-Ski, the cellular counterpart of the retroviral v-ski oncogene product. The clone lacked the N-terminal 337 amino acids, which is shown as del A in Fig. 1A. We first tested whether full-length c-Ski interacts with Smad2 in vivo. FLAG-tagged c-Ski and 6Myc-tagged Smad2 were cotransfected into COS-7 cells in the absence or presence of a constitutively active form of T␤R-I, T␤R-I(TD). The interaction was detected by immunoprecipitation with anti-FLAG antibody followed by immunoblotting with anti-Myc antibody. c-Ski interacted with Smad2 in a TGF-␤-dependent manner (Fig.  1B). Interaction between c-Ski and Smad3 was also tested. A similar result was obtained, although a weak interaction between Smad3 and c-Ski was detected in the absence of T␤R-I(TD). Thus c-Ski interacts with Smad2/3 in a ligand-dependent manner in vivo.
We next determined the c-Ski interaction domain in Smad3. Four deletion mutants of Smad3 were tested (Fig. 1C). Fulllength Smad3 interacted with c-Ski in the presence of T␤R-I(TD) as in Fig. 1B, whereas the MH1 domain or the MH1ϩlinker region did not interact with c-Ski. In contrast, the linkerϩMH2 and MH2 regions interacted with c-Ski in a ligand-independent manner. The result indicates that Smad3 interacts with c-Ski through its MH2 domain, and the MH1 domain inhibits the interaction, which is released by the phosphorylation of Smad3 by T␤R-I. We then tested which Smads interact with c-Ski (Fig. 1D). Smad2 and Smad3 interacted with c-Ski, whereas Smad1 and Smad5 did not. Smad4 interacted with c-Ski, but Smad6 and Smad7 failed to co-precipitate with c-Ski. Thus c-Ski associates with TGF-␤/activin-specific R-Smads and Co-Smad but not with BMP-regulated R-Smads or I-Smads.
To gain insights into the components of the c-Ski-Smad complex, we tested whether the association of Smad3 and Smad4 with c-Ski is mutually exclusive. As shown in Fig. 1E, coexpression of Smad3 and Smad4 did not diminish the interaction of each Smad with c-Ski, suggesting that Smad heterooligomers can bind to c-Ski. c-Ski has been shown to form oligomers (32,33). We examined whether the expression of Smad3 affects the c-Ski oligomerization. As shown in Fig. 1F, Smad3 did not decrease the association of FLAG-c-Ski and 6Myc-c-Ski, suggesting that c-Ski oligomers bind to Smad3. Taken together, the c-Ski-Smad complex consists of oligomers of both Smad and c-Ski.
c-Ski Has Two Distinct Smad Interacting Domains-To de-termine the Smad interacting domain in c-Ski, we generated an array of deletion mutants shown in Fig. 1A. del A is the original clone obtained from the two-hybrid screen. del B contains the most C-terminal region between the internal EcoRI site and the stop codon. del C is the N-terminal part of del A. del D contains the region between the start codon and the internal EcoRI site. del E contains the N-terminal region of c-Ski between the start codon and an internal ApoI site. We first used the two-hybrid assay. As shown in Fig. 2A, del A interacted with Smad2/3 but not with Smad4. However, an N-terminal deletion of del A caused loss of interaction. Consistent with this result, del C interacted with Smad2/3 but not with Smad4. Thus the region containing amino acids 338 -490 is an interaction domain for Smad2/3 but not for Smad4. This result suggested that another region is responsible for the interaction with Smad4. del E interacted with both Smad2/3 and Smad4, suggesting that the region spanning amino acids 1-309 is a second interaction domain for Smads. Thus c-Ski has two distinct Smad interaction domains with different binding properties. Consistent results were obtained in COS-7 cells (Fig. 2B). c-Ski Represses Transactivation by TGF-␤-We studied the effect of c-Ski on TGF-␤-induced transcription in mink R mutant lung epithelial cells lacking functional T␤R-I (28). p3TP-Lux is one of the standard reporters for TGF-␤. Expression of Smad3 activated the reporter, and c-Ski strongly inhibited the activation (Fig. 3A). Co-expression of T␤R-I(TD) with Smad3 further enhanced the reporter activity, which was again suppressed by c-Ski almost completely. c-Ski suppressed the activity of p3TP-Lux in a dose-dependent manner (Fig. 3B). As c-Ski slightly inhibited the basal activity of the reporter, we calculated the fold induction upon TGF-␤ stimulation. The inset in Fig. 3B clearly shows that c-Ski inhibited the TGF-␤-dependent induction of the reporter activity. We next tested another reporter, pAR3-Lux, containing FAST-1-binding sites. c-Ski efficiently suppressed the activation of the reporter by TGF-␤ (Fig.  3C). (CAGA) 9 -MLP-Luc contains 9 repeats of the CAGA Smad binding motif. c-Ski suppressed the activation of the reporter as well (Fig. 3D).
We then examined whether the repression is specific to TGF-␤. pTlx-Lux is a BMP-responsive reporter (34). We have used mink lung cells in the experiments above, but pTlx-Lux does not respond to BMPs in these cells (data not shown). We thus used P19 cells for pTlx-Lux assay. c-Ski rather enhanced the activity of the reporter in a dose-dependent manner, whereas it repressed the activity of p3TP-Lux in the same assay (Fig. 3E). Thus the repression by c-Ski is specific to TGF-␤, which is consistent with the result that c-Ski interacts with Smad2/3 but not with Smad1/5.
Finally we studied whether c-Ski antagonizes p300 in the activation of p3TP-Lux. p300 enhanced the activation of p3TP-Lux by TGF-␤, and c-Ski canceled the enhancement by p300 in a dose-dependent manner (Fig. 3F). Notably, comparison of the presence and absence of p300 shows that the effect of p300 is completely abolished in the presence of a higher dose of c-Ski.
c-Ski Is Incorporated in the DNA Binding Complex of Smads-The transcriptional repression by c-Ski could be due to the inhibition of DNA binding of Smads. We tested this possibility using gel mobility shift assays. First we used the 3xCAGA motif identified in the TGF-␤-responsive plasminogen activator inhibitor-1 promoter (14) (Fig. 4A). TGF-␤ induced efficient binding of Smad3 and Smad4 (complex 1) to the probe, whereas c-Ski alone did not bind to the probe. Co-expression of c-Ski caused a new shifted band (complex 2) with a slower mobility, and the Smad3-Smad4 complex decreased in parallel, suggesting that c-Ski associates with Smads on DNA. Addition of antibodies against each of the proteins supershifted the shown with a triangle) of Smad3 were co-expressed with FLAG-c-Ski and complex 2 (data not shown), confirming that the complex contains Smads and c-Ski. c-Ski was shown to bind to the palindromic GTCTAGAC sequence (25) that is also identified as a Smad binding motif (13). We used this sequence (2xGTCT) as a probe (Fig. 4B). Under the condition employed, we could not detect the binding of c-Ski alone to the probe. Co-expression of c-Ski with Smads induced the appearance of a new band with a slower mobility in a dose-dependent manner, reproducing the result obtained with the 3xCAGA probe. These results argue against the idea that c-Ski interferes with the DNA binding of Smads.
c-Ski Competes with p300 in Binding to Smad3-Another possible mechanism for the transcriptional repression by c-Ski is that c-Ski competes with p300 in binding to Smad2/3. We tested this possibility using immunoprecipitation coupled with immunoblotting in COS-7 cells. p300 interacted with Smad3 in the presence of T␤R-I(TD) (Fig. 5). Increasing amounts of c-Ski abrogated the p300-Smad3 interaction. This result supports the model that c-Ski competes with p300/CBP in binding to Smad2/3. This mutually exclusive binding of c-Ski and p300 to Smads at least partially accounts for the transcriptional repression by c-Ski.
c-Ski Recruits HDAC to Smad-c-Ski has been shown to interact with transcriptional co-repressor mSin3 that binds HDAC (27). We investigated whether Smad can recruit HDAC1 via c-Ski (Fig. 6). Whereas Smad3 interacted with c-Ski in the presence of T␤R-I(TD), Smad3 did not directly interact with mSin3A or HDAC1. When Smad3 associated with c-Ski, mSin3A and HDAC1 were recruited. This result indicates that Smad forms a complex with HD activity that antagonizes the HAT activity of p300/CBP. c-Ski Inhibits Smad-associated HAT Activity-Smad2 and Smad3 interact with p300/CBP; however, whether Smad complexes possess HAT activity has not been reported. We thus studied whether Smads exert HAT activity. COS-7 cells were transfected with the expression plasmids for Smad3 and p300 in the absence or presence of T␤R-I(TD). The immunoprecipitates from the cell lysates were subjected to in vitro HAT assay using histones and [1-14 C]acetyl-CoA (Fig. 7A). In another experiment, Smad3 in the presence of T␤R-I(TD) showed a low level of HAT activity that is presumably due to the involvement of endogenous p300/CBP (data not shown). When p300 was co-expressed in the presence of T␤R-I(TD), the immunoprecipitates showed strong HAT activity. A similar result was obtained with Smad2, but Smad4 did not show significant HAT activity (data not shown). We then studied the effect of c-Ski on the Smad-associated HAT activity (Fig. 7B). c-Ski suppressed the HAT activity of the Smad3-p300 complex in a dose-dependent manner, which accounts for the transcriptional repression by c-Ski. DISCUSSION c-ski was identified as the cellular counterpart of the v-ski oncogene isolated from Sloan-Kettering retroviruses (22,35,36). c-ski is related to another gene, sno (ski-related novel gene), expressing several splice variants such as snoN, snoN2, snoA, and snoI (35,(37)(38)(39). c-ski and sno are expressed in a variety of cell lines and tissues (27,35,38). c-Ski is a nuclear protein with DNA binding ability (40). c-Ski and SnoN form homo-and hetero-oligomers (32,33). The DNA binding domain of c-Ski is mapped to the N-terminal cysteine-rich region (40,41). The C-terminal coiled-coil region with a leucine zipper-like motif is responsible for the oligomerization of c-Ski and Sno (32,33,42). v-Ski induces muscle differentiation of quail embryo cells, whereas it also stimulates proliferation, transformation, and anchorage-independent growth in the same cells (43). The mechanism of these paradoxical effects has not been elucidated. Screening of oligonucleotides resulted in the identification of GTCTAGAC as the consensus v-Ski/c-Ski/SnoN-binding site (25,44). Intriguingly, the sequence is identical with the Smad-binding site revealed by Zawel et al. (13). Efficient DNA binding of c-Ski, however, requires other cellular proteins (25,27,40,44). Ski represses transcription of the reporters with its consensus binding sites (25). v-Ski associates with the retinoic acid receptor complex and represses transcription from a retinoic acid response element (26). c-Ski is one of the components of the N-CoR-mSin3-HDAC transcriptional co-repressor complex (27). This large complex is required for the transcriptional repression by Mad (antagonist of Myc), thyroid hormone receptor, the retinoblastoma gene product (Rb) (27,45), and presumably other transcriptional repressors. v-Ski abrogates the Rbmediated repression, which may eventually contribute to tumorigenesis.
In the present report, we showed that Smad2 and Smad3 interact with c-Ski in a TGF-␤-dependent manner. Smad4 is also incorporated in the complex. Smad3 bound to c-Ski through its MH2 domain. c-Ski has two distinct Smad binding 6Myc-c-Ski. 0.6 g of Smad3 (shown with ϩ) was used in the control. The cell lysates of transfected cells were subjected to immunoprecipitation with anti-FLAG antibody followed by immunoblotting with anti-Myc antibody, which detects the homo-oligomerization of c-Ski. The top panel shows the interaction and the lower three panels show the expression of each protein as indicated. A, the effect of c-Ski on transactivation by TGF-␤ was examined in mink R mutant cells using the p3TP-Lux luciferase reporter. Luciferase activity was normalized against the cotransfected sea-pansy luciferase activity. Experiments were done in duplicate, and the standard deviations are shown in vertical lines. The values represent fold induction as determined by comparing with the basal p3TP-Lux activity. B, increasing amounts of c-Ski were cotransfected into mink R mutant cells with p3TP-Lux and T␤R-I(TD), and luciferase activity was assayed. The inset shows fold induction upon TGF-␤ stimulation calculated by dividing the luciferase activity in the presence of T␤R-I(TD) by the corresponding activity in the absence of T␤R-I(TD). C, the effect of c-Ski on transactivation by TGF-␤ was examined using the pAR3-Lux luciferase reporter in mink R mutant cells. D, the effect of c-Ski on transactivation by TGF-␤ was examined using the (CAGA) 9 -MLP-Luc luciferase reporter in mink R mutant cells. E, the effect of c-Ski on transactivation by BMP was examined using the pTlx-Lux luciferase reporter in P19 cells. BMPR-II represents BMP type II receptor. The effect of c-Ski on transactivation of p3TP-Lux in P19 cells was examined in the same experiment. F, the effect of c-Ski on the enhancement of the p3TP-Lux activity by p300 was examined in mink R mutant cells.
has not been examined in this study. c-Ski potently inhibits transactivation of various reporters by TGF-␤, whereas it does not suppress the activation of pTlx-Lux by BMP stimulation. c-Ski rather moderately enhances the activity of pTlx-Lux. The effect of c-Ski on transcription is highly dependent on the cell type as well as on the growth/differentiation status of the cells.
c-Ski induced the transactivation of myosin light chain 1/3 and muscle creatine kinase in myoblasts but not in fibroblasts (23). c-Ski suppressed the promoter/enhancer activity of myogenin in proliferating C2C12 myogenic cells, whereas it potentiated the activity in differentiated myotubes (24). Furthermore, c-Ski suppressed the myogenin transcription in non-myogenic cells. In our study, the repression of TGF-␤ signaling by c-Ski was observed at least two different cell lines including mink lung epithelial cells and murine P19 embryonal carcinoma cells. In addition, the enhancement of the pTlx-Lux activity by c-Ski is FIG. 4. Effect of c-Ski on DNA binding of Smads. A, gel mobility shift assay using the 3xCAGA probe was conducted. Complexes 1 and 2 represent those of Smad3-Smad4 and Smad3-Smad4-c-Ski, respectively. B, gel mobility shift assay using the 2xGTCT probe was conducted. Complexes 1 and 2 represent those of Smad3-Smad4 and Smad3-Smad4-c-Ski, respectively. 0.1 (ϩ) or 1.2 (ϩϩ) g of the c-Ski plasmid was used. We investigated possible mechanisms of the repression of TGF-␤-induced transactivation by c-Ski. c-Ski does not inhibit DNA binding of Smads. In fact c-Ski can be incorporated in the Smad DNA binding complex. The GTCTAGAC sequence is of particular interest, since it has been identified as a binding motif both for c-Ski and Smads (13,25). c-Ski and Smads may thus compete with each other in DNA binding. However, we were unable to detect c-Ski binding to the sequence under our experimental condition. Nicol and Stavnezer (25) used 10 times more of the labeled probe (2 ϫ 10 5 cpm). The GTCTAGAC probe used in our assay is based upon the report on Smad4 by Le Dai et al. (46), and the sequence contains ATA between two GTCTAGAC sequences, whereas the sequence used by Nicol and Stavnezer (25) does not have any intervening base between the two motifs. These differences may account for the difference of the results. As mentioned above, efficient DNA binding of c-Ski requires other cellular proteins (25,27,40,44). Perhaps Smads may serve as such proteins. The results of gel mobility shift assays also suggest that c-Ski does not antagonize TGF-␤ signaling prior to DNA binding of Smads, which is in contrast to the actions of I-Smads and dominant-negative Smad3 that antagonize TGF-␤ at the receptor level (47)(48)(49)(50).
Smads interact with transcriptional co-activators such as p300 and CBP (28,(51)(52)(53)(54)(55). c-Ski may compete with p300/CBP in Smad binding, since both proteins bind to the MH2 domain in Smad3 (28). Our result shown in Fig. 5 supports this model. We next tested the possibility that Smad may recruit HDAC in association with c-Ski. The result shown in Fig. 6 demonstrated that this is the case. Thus Smad can form a complex associated with HD activity that represses transcription. Both of these two mechanisms can concurrently contribute to the repression of TGF-␤-induced transcription by c-Ski. Finally, we investigated the effect of c-Ski on Smad-associated HAT activity. Immunoprecipitation of Smad3 revealed that HAT activity is associated with Smad3 when p300 is present. c-Ski inhibited the Smad3-associated HAT activity in a dose-dependent manner (Fig. 7B). The decrease of HAT activity is not due to the variation of the expression levels of Smad3 or p300 as shown in the figure. The result is consistent with that of Fig. 3F in which c-Ski abolished the enhancing effect of p300 in the transactivation of p3TP-Lux by TGF-␤. In conclusion, c-Ski displaces p300/CBP from Smad2/3 and, in turn, recruits the mSin3/ HDAC transcriptional co-repressor complex that plays a critical role in a variety of transcriptional regulatory pathways involving tumor suppressor genes and nuclear hormone receptors. Interestingly, Rb, Mad, and Smad2/3 are all negative growth regulators, and these proteins could be targets of oncogenic v-Ski.
Several nuclear proteins have been implicated in transcriptional repression in Smad signaling. Hoxc-8, a homeodomain protein, was isolated as a Smad1-interacting protein (56). Hoxc-8 binds to the BMP-responsive osteopontin promoter and represses its transactivation. BMP induces the interaction of Smad1 with Hoxc-8 and transactivates osteopontin by displacing Hoxc-8 from its binding site. SIP1 is a member of the ␦EF1/Zfh-1 family of zinc finger/homeodomain proteins. SIP1 binds to the promoter of brachyury in Xenopus and inhibits its expression (57). Activin-induced interaction of SIP1 with Smad2/3 may prevent the DNA binding of SIP1, leading to the expression of brachyury. R-Smads can thus act as a derepression factor in TGF-␤ superfamily signaling. In Drosophila, Brinker was isolated as a protein responsible for the repression of Dpp-inducible genes (58 -60). The molecular mechanism of the repression, however, remains to be elucidated. Recently, a homeodomain protein of TALE class, TGIF, was identified as a transcriptional co-repressor for Smad2 and Smad3 (61). TGIF was originally isolated as a protein that binds to retinoid X receptor-responsive element (62). TGIF interacts with Smad2 in a TGF-␤-dependent manner. TGIF represses TGF-␤-induced transcription by recruiting HDAC. TGIF also competes with p300 in binding to Smad2. Thus TGIF acts in a similar fashion to c-Ski, although the structures are completely divergent. Whether mSin3 is involved in the Smad-TGIF complex is not known. Co-repressors may not only actively repress transcription but also set a threshold for genes that are induced by TGF-␤. Our results indicate that Smads can act as both transactivators and transcriptional repressors depending on the nuclear partners. There have been known a number of other transcription factors that can act as either activators or repressors depending on protein-protein interactions dictated by the promoter and physiological context (63,64). Our results argue that Smads are such bi-modal transcriptional regulators.
Whereas TGF-␤ activates the transcription of a variety of genes, it down-regulates cell cycle regulators such as c-Myc and cdc25A (21), which contribute to growth arrest by TGF-␤. Al- though Rb has been implicated in the down-regulation of c-Myc (65), the precise molecular mechanism remains unclear. TGF-␤ also represses the expression of genes such as urokinase (66), transin/stromelysin (67), elastase (68), collagenase (69), and tissue glutaminase (70), some of which are up-regulated depending on the cell type. In addition, TGF-␤ represses the expression of cytokines such as interferon-␥ and various interleukins in splenocytes, and the repression is abrogated in Smad3-null mice (71). It would be a future focus to investigate whether transcriptional co-repressors such as c-Ski play a role in the transcriptional regulation of these genes. It also remains to be elucidated how the mutually exclusive interaction of Smads with co-activators or co-repressors is regulated in the physiological context.