Competition between Ski and CREB-binding Protein for Binding to Smad Proteins in Transforming Growth Factor-β Signaling*

The family of Smad proteins mediates transforming growth factor-β (TGF-β) signaling in cell growth and differentiation. Smads repress or activate TGF-β signaling by interacting with corepressors (e.g. Ski) or coactivators (e.g. CREB-binding protein (CBP)), respectively. Specifically, Ski has been shown to interfere with the interaction between Smad3 and CBP. However, it is unclear whether Ski competes with CBP for binding to Smads and whether they can interact with Smad3 at the same binding surface on Smad3. We investigated the interactions among purified constructs of Smad, Ski, and CBP in vitro by size-exclusion chromatography, isothermal titration calorimetry, and mutational studies. Here, we show that Ski-(16-192) interacted directly with a homotrimer of receptor-regulated Smad protein (R-Smad), e.g. Smad2 or Smad3, to form a hexamer; Ski-(16-192) interacted with an R-Smad·Smad4 heterotrimer to form a pentamer. CBP-(1941-1992) was also found to interact directly with an R-Smad homotrimer to form a hexamer and with an R-Smad·Smad4 heterotrimer to form a pentamer. Moreover, these domains of Ski and CBP competed with each other for binding to Smad3. Our mutational studies revealed that domains of Ski and CBP interacted with Smad3 at a portion of the binding surface of the Smad anchor for receptor activation. Our results suggest that Ski negatively regulates TGF-β signaling by replacing CBP in R-Smad complexes. Our working model suggests that Smad protein activity is delicately balanced by Ski and CBP in the TGF-β pathway.

As positive regulators of Smad function, the transcriptional coactivators CREB-binding protein (CBP) and p300 are important components of TGF-␤ signaling and mediate the anti-oncogenic functions of Smad2 and Smad4 (18). These Smad proteins directly interact with the nuclear coactivator CBP/p300 to activate the transcription of TGF-␤-responsive genes (18 -21). The MH2 domain of R-Smads can interact with the C-terminal domain (amino acids 1891-2175) of CBP (18), and CBP directly acetylates the MH2 domain of Smad3 at Lys 378 to positively regulate its transactivation activity (22).
Ski has been shown to interfere with the interaction between Smad3 and CBP in vivo (12). However, the interactions among Smad, Ski, and CBP have not been completely characterized. It is unclear whether Ski- (16 -192) or CBP-  can interact with Smad3 at the SARA-binding surface and whether there is direct competition between CBP and Ski for binding to Smads.
To resolve the stoichiometric identity of Ski and CBP for binding to Smads and to provide insights into the regulation of TGF-␤ signaling, size-exclusion chromatography and isothermal titration calorimetry were used in in vitro interactive assays to measure the binding of purified Ski- (16 -192) and CBP-(1941CBP-( -1992 to Smads. Here, we present biochemical and calorimetric evidence that Ski and CBP directly interact with R-Smad and that Ski competes with CBP for binding to Smad.
The phosphorylated Smad2 and Smad3 constructs used to analyze Smad/Ski interactions in this study include their linker and MH2 domains. The phosphorylated Smad2 construct extends from residues 186 to 467 and is referred to as S2LC-2P. The phosphorylated Smad3 construct extends from residues 145 to 425 and is referred to as S3LC-2P. The Smad4 construct comprises its MH2 domain and part of its linker domain, extending from residues 273 to 552, and is referred to as S4AF (23). S2LC-2P and S3LC-2P were prepared using the intein-mediated phosphopeptide ligation method as described previously (16,24). S4AF was purified as described previously (23). The Smad heteromeric S4AF⅐S3LC-2P complexes and S4AF⅐ S2LC-2P were purified as described previously (8).
plexes tested were phosphorylated Smad2, Smad3, the heterotrimer of S2LC⅐S4AF, and the heterotrimer of S3LC⅐S4AF. The heats of dilution for Ski- (16 -192) or GST-CBP-  were measured by titrating Ski- (16 -192) or GST-CBP-  into ITC buffer; these values were subtracted for data analysis. Data were analyzed with Origin 7.0 software (Micro-Cal, LLC) using a single-site binding model. ⌬H, ⌬S, and K a values were determined experimentally, and ⌬G was calculated from the following equation: ⌬G ϭ ϪRT ln K a .
Size-exclusion Chromatography-To obtain the stoichiometry of Ski and CBP for binding to Smads, size-exclusion chromatography was performed on a Superdex 200 HR 10/30 column using the Á KTA Explorer 10 FPLC system (GE Healthcare). All runs were performed at room temperature in FPLC buffer (20 mM HEPES (pH 7.3), 0.1 mM EDTA, 100 mM NaCl, and 1 mM dithiothreitol). Prior to loading onto the column, protein samples were incubated in 1 mM tris(2-carboxyethyl)phosphine for 60 min at room temperature. FPLC operation and data analysis were done with UNICORN software. The column was calibrated with blue dextran (to determine the void volume) and molecular mass standards (ovalbumin (43 kDa), albumin (67 kDa), aldolase (158 kDa), catalase (232 kDa), and ferritin (440 kDa)). Elution fractions (0.5 ml) were collected at room temperature at a flow rate of 0.5 ml/min.
GST Pulldown Assay-The GST fusion form of a protein was immobilized onto glutathione-Sepharose beads, which were then cleaned with wash buffer (20 mM Tris (pH 7.4), 0.1 mM EDTA, 10 mM NaCl, and 1 mM dithiothreitol). A second protein purified without a GST tag was then added to the beads and incubated at 4°C for 1 h. The beads were washed again and analyzed by SDS-PAGE, and the gels were stained with Coomassie Blue. The standard conditions were as follows: 100 l of beads, 100 g of GST fusion protein, and 200 g of binding partner. Unless stated otherwise, the beads were washed six times (each time with 1 ml of wash buffer), rapidly mixed, and centrifuged, and the buffer was removed.
To further characterize the interactions between Smads and Ski- (16 -192), ITC was used to analyze the thermodynamics and stoichiometry of heteromeric Smad⅐Ski- (16 -192) complex formation. The results showed that Ski- (16 -192) titrated into both S2LC-2P and S3LC-2P at a relative molar ratio of 1.0 ( Fig.   2C and Table 1), meaning that a single molecule of Ski- (16 -192) can interact with one molecule of S2LC-2P or S3LC-2P. The ITC results are consistent with the stoichiometry observed in the size-exclusion analysis. Taken together, these results suggest that the heteromeric Smad⅐Ski- (16 -192) complex is a heterohexamer with a preferred stoichiometry of three subunits of S2LC-2P or S3LC-2P and three subunits of Ski- (16 -192) (Fig. 2D).
Homotrimeric S2LC-2P or S3LC-2P Interacts with CBP to Form a Heterohexamer-The analysis above showed that Ski- (16 -192) bound to homotrimeric S2LC-2P and S3LC-2P as well as to heterotrimeric S2LC-2P⅐S4AF and S3LC-2P⅐S4AF. To compare the binding of Ski and CBP to Smads, the same conditions used for Ski- (16 -192) were used for CBP-  to test its interaction with Smads.
A previous study showed that the MH2 domain of R-Smads can interact with the C-terminal domain (amino acids 1891-2175) of CBP (18). Furthermore, Wu et al. (13) identified a 26-residue sequence motif of CBP/p300 (amino acids 1955-1980) that is both necessary and sufficient to form a stable complex with R-Smads. To confirm these results, we carried out in vitro binding assays using CBP- , CBP-(1931CBP-( -1992, and CBP- . An in vitro GST-mediated pulldown assay was used to confirm that CBP-(1941-1992) is sufficient to format a stable complex with the S3LC-2P homotrimer and the S3LC-2P⅐S4AF heterotrimer (data not shown).
Because CBP-(1941CBP-( -1992 was unstable, GST was required to stabilize it. Before testing the interaction of GST-CBP-  with Smads, it was necessary to test whether GST alone interacts with Smad2. Our results show that GST did not interact with Smad2 and Smad3 (Fig. 4B, upper panels). To further investigate whether Smads and GST-CBP-  interact directly in the purified state, size-exclusion chromatography and ITC were performed. GST-CBP-  interacted directly with S2LC-2P or S3LC-2P as judged by coelution on a size-exclusion column (Fig. 4, A and B). When mixed at a 2:1 molar ratio, a significant portion of GST-CBP-  co-eluted in the same fractions as trimeric S2LC-2P or S3LC-2P. The remaining uncomplexed GST-CBP-  eluted in a dimeric form, presumably through GST dimerization. The elution peak of the S2LC-2P⅐GST-CBP-  or S3LC-2P⅐GST-CBP-(1941-1992) complex shifted 2.0 fractions forward relative to the peak of trimeric S2LC-2P or S3LC-2P, respectively. The ITC results for formation of the heteromeric Smad⅐GST-CBP-  complex are shown in Fig. 4C and Table 1. GST-CBP-(1941-1992 titrated into both S2LC-2P and S3LC-2P at a relative molar ratio of 1.0, meaning that a single molecule of GST-CBP-  can interact with one molecule of S2LC-2P or S3LC-2P. Thus, these results suggest that the heteromeric Smad⅐GST-CBP-  complex is a heterohexamer with a preferred stoichiometry of three subunits of S3LC-2P and three subunits of GST-CBP-(1941-1992) (Fig. 4D).

The Heterotrimeric S2LC-2P⅐S4AF or S3LC-2P⅐S4AF Complex Interacts with CBP to Form a Heteropentamer-GST-CBP-
1992). However, the binding determinants for Ski and CBP are not identical either, as mutations of Phe 303 and Tyr 323 , which abolished or reduced Ski- (16 -192) interactions, had no effect on CBP-  interactions.
In addition, hydrophobic residues in Ski were involved in contact with Smad3 (Fig. 7C). Mutations of Leu 21 or Phe 24 in Ski abol-ished its interaction with S3LC or S3LC-2P, and mutation of Leu 26 dramatically reduced its interaction with S3LC-2P because the K d was 3-fold greater than that of the wild-type protein (Table 1). Several point mutations of Lys 19 , Glu 22 , Leu 26 , Phe 38 , Ile 126 , Leu 127 , or Leu 131 in Ski had no significant effect on its interaction with S3LC or S3LC-2P.   -(1941-1992) and S2LC-2P⅐S4AF or S3LC-2P⅐S4AF were loaded as 2:1 molar mixtures onto a size-exclusion column, and their interactions are indicated by co-elution as a heteromeric complex with a peak around fraction 13. Excess uncomplexed GST-CBP-(1941-1992) eluted as a dimer with a peak around fraction 17. mAU, milli-absorbance units. B, GST-CBP co-eluted with trimeric S2LC-2P⅐S4AF (left panels) or S3LC-2P⅐S4AF (right panels). Constructs (indicated to the right of each panel) were injected into the size-exclusion column, and eluted fractions were analyzed by SDS-PAGE. Molecular mass standards are shown above and to the left; fraction numbers are shown below. Gels were stained with Coomassie Blue. C, CBP interacted with the heterotrimer of Smad2⅐Smad4 or Smad3⅐Smad4 to form a heteropentamer of Smad2⅐Smad4⅐CBP or Smad3⅐Smad4⅐CBP.

DISCUSSION
Here, we have shown that both CBP and Ski directly interact with R-Smad proteins with slightly different affinities. These results are the first direct stoichiometric evidence from ITC and size-exclusion chromatography for interactions among Smad, CBP, and Ski. Taken together with previous reports (2, 6 -10, 16, 18, 20, 26 -28), the present evidence suggests a model of Smad protein regulation by Ski and CBP in the TGF-␤ pathway (Fig. 8). Upon phosphorylation at its C-terminal serines by the activated type I TGF-␤ receptor kinase, Smad2 or Smad3 is induced to form a homotrimer, which interacts with Smad4 to form a functional signaling heterotrimer that translocates into the nucleus. CBP directly interacts with R-Smad⅐ Smad4 to activate the transcription of TGF-␤-responsive genes, whereas Ski directly interacts with R-Smad⅐ Smad4 to repress the transcription of TGF-␤-responsive genes. Moreover, Ski can repress TGF-␤ signaling activity by displacing CBP from R-Smads. Similarly, CBP can activate TGF-␤ signaling activity by displacing Ski from R-Smads.
Regulation of Smad proteins in the TGF-␤ signaling pathway can be controlled at many levels such as by coactivators/corepressors and by cell type. Because the affinities of Smad for Ski and CBP are almost equal (Table 1), the transcriptional activity of Smads in the nucleus is delicately balanced by steady-state levels of Ski and CBP. Moreover, the levels of Ski and CBP depend on cell-type specificity because Smad coactivators or corepressors may vary among different cell types (2). Ski is ubiquitously expressed in virtually all adult and embryonic tissues at low levels, but its expression is increased in many human tumor cells (29). Ski may play an important role in binding to Smads to depress TGF-␤ signaling activity in tumor cells. These observations are consistent with the finding that the recruitment of transcriptional coactivators or corepressors to target genes in TGF-␤ signals depends on cell type-specific partner proteins (30).   (16 -192) or S3LC-2P⅐Ski- (16 -192) and S4AF were loaded as a 1:4 molar mixture onto the size-exclusion column. The eluted fractions were analyzed by SDS-PAGE. Molecular mass standards are shown above and to the left; fraction numbers are shown below. Gels were stained with Coomassie Blue. B, Ski and CBP bind to a portion of the SARA-binding site. GST-Ski- (16 -192) was used to detect Ski- (16 -192) interactions with S3LC mutants (left panel). GST-CBP-(1941-1992 was used to detect CBP interactions with S3LC mutants (right panel). C, hydrophobic residues in Ski mediate direct interactions with S3LC-2P (left panel) or S3LC (right panel). GST-Ski- (16 -192) mutants were used to detect interactions with S3LC or S3LC-2P.
To interact with R-Smads, Ski and CBP compete for similar binding sites. The MH2 domain of Smad3 is a versatile protein/ protein interaction module. At the receptor complex, the MH2 domain of Smad3 interacts with SARA and the receptor kinase (17,(31)(32)(33)(34)(35)(36). In the nucleus, the MH2 domain of Smad3 further interacts with Ski and CBP (11,16,18,21,22,37). In this study, we therefore evaluated how the residues that bind SARA in the MH2 domain of Smad3 affect the interaction between Smad3 and Ski or CBP. Trp 405 was identified in the MH2 domain of Smad3 as a key residue for interactions with both Ski and CBP (Fig. 7B). Trp 405 resides in the conserved H5 helix of Smad3, which binds the ␤ structure of SBD in SARA (Fig. 1B) (16), and is fully conserved among most Smads (23). Phe 303 was also identified in the MH2 domain of Smad3 as a key residue for interaction with Ski, but not for interaction with CBP (Fig. 7B). This finding is consistent with those of Mizuide et al. (37) showing that amino acids 261-314 in the MH2 domain of Smad3 are important for interaction with c-Ski. Phe 303 resides in the conserved ␤6 sheet of Smad3, which binds the helical structure of SBD in SARA (Fig. 1B) (16), suggesting that Ski and CBP can interact with Smad3 at similar binding sites on Smad3. This finding explains why both activation and repression of gene expression use the same set of activated Smad proteins (2).
We have found that Smad, Ski, and CBP could exist as a stable complex of Smad⅐Ski⅐CBP regardless of whether Ski was incubated with the Smad3⅐CBP heteromer or whether CBP was incubated with the Smad3⅐Ski heteromer. This result also provides evidence that both activation and repression of gene expression use the same set of activated Smad proteins. The Smad⅐Ski⅐CBP complex may provide a large platform for the binding of multiple Smad⅐coactivator or Smad⅐corepressor molecules. Consistent with our result, another group reported finding c-Ski and CBP in the same complex after stimulating mouse embryonic cells with TGF-␤ (38), suggesting that the TGF-␤ signaling pathway cooperatively regulates genes with Smad-binding element-containing promoters.
Purified proteins were used in our experiments to test whether Ski disrupts the formation of the R-Smad⅐Smad4 heterotrimer. Our results demonstrate that the R-Smad⅐Smad4 complex was stable whether Ski was used to directly interact with the R-Smad⅐Smad4 complex (Fig. 3B) or whether Smad4 was used to directly interact with the R-Smad⅐Ski complex (Fig. 7A). These results provide evidence that the binding of Ski still permits formation of the R-Smad⅐Smad4 heterotrimer. This result is consistent not only with previous results from our laboratory (16), but also with coimmunoprecipitation results presented by Ueki and Hayman (15) showing that Ski does not disrupt Smad3⅐Smad4 heteromer formation. They also demonstrated that Ski can use different domains to interact with Smad3 and Smad4 and that recruitment of Ski to the Smad3⅐Smad4 complex through binding to either Smad3 or Smad4 is both necessary and sufficient for repression. Wu et al. (13) also demonstrated that Ski can use different domains to interact with Smad3 and Smad4. Although their results indicated that Ski disrupts Smad3⅐Smad4 heteromer formation, they still proposed that Smad4 and Smad2⅐Smad3 remain bound to the Ski protein in an inactive conformation that no longer stimulates gene expression. In contrast to full-length Ski, which can interact with R-Smad and Smad4, the Ski domain that was used in our study can interact only with R-Smad.
On the basis of the report by Wu et al. (13), full-length Ski likely binds to both R-Smads and Smad4 in the complex. To better understand the role of Ski in repressing Smad activity, including the level of oligomerization when interacting with Smads, it would be necessary to use full-length Ski or a Ski construct that contains the domains for binding to both R-Smad and Smad4. Unfortunately, our attempts to express these constructs of Ski were unsuccessful. Thus, whether fulllength Ski can interact with both R-Smad and Smad4 and whether CBP can compete with full-length Ski for binding to R-Smad remain unknown. Although Ski represses TGF-␤ signaling by multiple mechanisms, our results suggest that the R-Smad⅐Smad4 complex will still occupy the promoter of the target gene and that Smads will actively repress transcription when Ski is bound to Smads.
In summary, we have demonstrated that purified Ski and CBP directly interact with purified Smads and that Ski and CBP compete for binding to Smad with slightly different affinities. In addition, Ski and CBP interact with Smad at a portion of the  APRIL 13, 2007 • VOLUME 282 • NUMBER 15 binding sites on Smad. Finally, our model suggests that Smad activity is delicately balanced by Ski and CBP in the TGF-␤ signaling pathway. To determine whether this model is also applicable to bone morphogenetic protein signaling, additional studies are required to investigate the interactions of Ski with Smad1 and Smad5.