The MSG1 Non-DNA-binding Transactivator Binds to the p300/CBP Coactivators, Enhancing Their Functional Link to the Smad Transcription Factors*

The MSG1 nuclear protein has a strong transcriptional activating activity but does not bind directly to DNA. When cotransfected, MSG1 enhances transcription mediated by the Smad transcription factors in mammalian cells in a manner dependent on ligand-in-duced Smad hetero-oligomerization. However, the mechanism of this MSG1 effect has been unknown. We now show that MSG1 directly binds to the p300/cAMP-response element-binding protein-binding protein (CBP) transcriptional coactivators, which in turn bind to the Smads, and enhances Smad-mediated transcription in a manner dependent on p300/CBP. The C-termi-nal transactivating domain of MSG1 is required for binding to p300/CBP and enhancement of Smad-medi-ated transcription; the viral VP16 transactivating domain could not substitute for it. In the N-terminal region of MSG1, we identified a domain that is necessary and sufficient to direct the specific interaction of MSG1 with Smads. We also found that the Hsc70 heat-shock cognate protein also forms complex with MSG1 in vivo, suppressing both binding of MSG1 to p300/CBP and enhancement of Smad-mediated transcription by MSG1. These results indicate that MSG1 and B. and p3TP-Lux J. Transfection and Transcriptional Activation Assay— Cells were transfected using LipofectAMINE PLUS reagent (Life Technologies, Inc.) as described (11). Transfection efficiency was evaluated by co-transfecting the b -galactosidase expression plasmid pSV b Gal (Pro- mega) and by enzyme activity assay. Transcriptional activation in mammalian cells was assayed as described previously (11, 32). Human recombinant TGF- b 1 was purchased from R&D Systems. All assays were repeated at least three times and are represented as mean 6 S.E. In all cell lines used in the present study, background transcription mediated by the GAL4 DNA-binding domain without fusion was negli- gible, and MSG1 cotransfection did not affect it in the presence or absence of TGF- b (data not shown). B16-F10 cell clones stably expressing HA-tagged mouse MSG1 or controls were generated by transfecting cells with an expression plas- mid (1) or pRc/CMV vector (Invitrogen), respectively, followed by selec-tion with 2 mg/ml G418 for 2 weeks and clonal isolation. Antibodies, Western Blotting, and Immunoprecipitation— Our anti-MSG1

MSG1 is a 27-kDa nuclear protein that has strong transcriptional activating activity in mammalian cells but does not bind directly to DNA (1,2). Although MSG1 had been isolated as a melanocyte-specific protein (1,3,4), it was subsequently found that the msg1 mRNA transcript is expressed also in several non-melanocytic tissues of mouse embryos (5). Although possible roles of MSG1 in cell differentiation and/or development have been predicted based on its inducible expression (4) and tissue distribution (5), its precise physiological roles remain to be elucidated. Members of the MSG1 family nuclear proteins share a well conserved C-terminal acidic domain (the CR2 1 domain) that accounts for their strong transcriptional activating activity (2), whereas none of them has significant DNA binding motifs or activity (1,2,5). Thus, the MSG1 family represents a novel family of non-DNA-binding transcriptional activators.
Several non-DNA-binding transactivators have been demonstrated to interact with sequence-specific DNA-binding proteins and function as transactivating subunits of multisubunit transcription factors (6 -10). Consistent with this concept, we have elucidated previously that MSG1 specifically interacts with the Smad DNA-binding transcription factors in a manner dependent on the transforming growth factor ␤ (TGF-␤) signaling, enhancing Smad-mediated transcription (11). Receptorphosphorylated Smads (Smads 1, 2, 3, 5, and 8) form heterooligomers with Smad4, a "common" Smad that is involved in all active Smad complexes, and translocate to the nucleus, where they bind to promoters of target genes and activate transcription (12)(13)(14)(15). Although specific interaction of MSG1 with Smad4 was demonstrated by yeast two-hybrid assay (11), the mechanism of the enhancement of Smad-mediated transcription by MSG1 has been unknown.
Recent studies have elucidated that the mechanism of Smadmediated transcriptional activation involves the p300/cAMPresponse element-binding protein-binding protein (CBP) transcriptional coactivators. Both the receptor-phosphorylated Smads (16 -20) and Smad4 (21) bind directly to p300/CBP, and Smad-mediated transcription is dependent on the coactivator function of p300/CBP. To understand the mechanism of the MSG1 action, it is, therefore, important to determine whether MSG1 functions as an additional Smad-interacting transcriptional coactivator that functions in a manner independent of p300/CBP (14) or MSG1 enhances Smad-mediated transcription in a p300/CBP-dependent fashion.
In the present study, we show biochemical evidence that MSG1 is a p300/CBP-binding protein and augments the functional link between Smads and p300/CBP, resulting in enhancement of Smad-mediated transcription.

EXPERIMENTAL PROCEDURES
Cells-NMuMg, NIH3T3, and COS-1 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and antibiotics. Culture conditions and MSG1 expression in * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: CR, conserved region; ARE, activin responsive element; CAT, chloramphenicol acetyltransferase; CBP, cAMP-response element-binding protein-binding protein; GAL4DB, GAL4 DNA binding domain; GST, glutathione S-transferase; HA, hemagglutinin; SAD, Smad4 activation domain; SBE, Smad binding element; SID, Smad interaction domain; TGF-␤, transforming growth factor ␤. murine melanoma cells (B16-F1, B16-F10) were described previously (1,3,4).
A baculovirus expression plasmid for His 6 -tagged human MSG1 was constructed on the pBlueBacHis2 transfer vector (Invitrogen), and the recombinant baculovirus was generated in Sf9 insect cells using a kit (Invitrogen). His 6 -MSG1protein was purified using a metal-chelating column (Amersham Pharmacia Biotech) to homogeneity.
Bacterial expression plasmids for glutathione S-transferase (GST) fusion fragments of p300 (27) were gifts from S. R. Grossman and D. M. Livingston. A plasmid for GST fusion human TAF II 32 (28) was from R. Tjian. GST fusion proteins were purified using glutathione beads (Amersham Pharmacia Biotech) following the manufacturer's instructions.
Transfection and Transcriptional Activation Assay-Cells were transfected using LipofectAMINE PLUS reagent (Life Technologies, Inc.) as described (11). Transfection efficiency was evaluated by cotransfecting the ␤-galactosidase expression plasmid pSV␤Gal (Promega) and by enzyme activity assay. Transcriptional activation in mammalian cells was assayed as described previously (11,32). Human recombinant TGF-␤1 was purchased from R&D Systems. All assays were repeated at least three times and are represented as mean Ϯ S.E. In all cell lines used in the present study, background transcription mediated by the GAL4 DNA-binding domain without fusion was negligible, and MSG1 cotransfection did not affect it in the presence or absence of TGF-␤ (data not shown).
B16-F10 cell clones stably expressing HA-tagged mouse MSG1 or controls were generated by transfecting cells with an expression plasmid (1) or pRc/CMV vector (Invitrogen), respectively, followed by selection with 2 mg/ml G418 for 2 weeks and clonal isolation.
For transfection-based coimmunoprecipitation analyses, COS-1 cells (1 ϫ 10 6 per 6-cm dish) were transfected with expression plasmids and cultured for 48 h in the presence of 10% serum before cell lysis. Binding of endogenous MSG1 to CBP or Hsc70 was demonstrated using B16-F1 cells without transfection. Cells were lysed in 0.5 ml per dish of IP lysis buffer (50 mM Hepes/NaOH (pH 7.5), 150 mM NaCl, 2.5 mM EGTA, 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, 1% Triton X-100) supplemented with phosphatase inhibitors (50 mM NaF, 25 mM sodium glycerophosphate, 1 mM boiled sodium orthovanadate) and protease inhibitors (10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride). Protein expression from plasmids was evaluated by Western blotting as described (4). For immunoprecipitation cell lysates were precleared by non-immunized mouse or rabbit IgG with protein G-agarose beads (Pierce), and target proteins were precipitated using specific antibodies and protein G-agarose beads for 2 h at 4°C. Beads were then washed with the IP lysis buffer five times at 4°C. Precipitated and coprecipitated proteins were extracted by boiling the beads in 2ϫ Laemmli buffer for 5 min and subjected to Western blotting for detection. Direct binding of purified GST fusion proteins and baculovirus recombinant proteins was also evaluated by immunoprecipitation in the IP lysis buffer at 4°C.

Specificity of MSG1 on Activation of Smad Binding
Promoters-Interactions of non-DNA-binding transactivators with sequence-specific DNA-binding proteins often depend on specific features of their target promoters (6,10,(33)(34)(35). Transcriptional activation by the DNA-binding Smad complexes is also affected strongly by the location, direction, and the numbers of the Smad binding elements in the promoters (36). To determine whether the enhancement of Smad-mediated transcription by MSG1 is affected by specific features of Smad-binding promoters, we evaluated the effects of MSG1 on well characterized synthetic promoters (Fig. 1A) (13, 14). The AREx3 promoter contains the activin responsive element (ARE) and requires Smad2 and Smad4, as well as the FAST-1 ARE-binding protein, for its activation (37,38). The SBEx4 promoter contains the Smad binding element (SBE), to which Smad3 and Smad4 directly bind (30). Cotransfection of MSG1 enhanced TGF-␤induced transcriptional activation from these promoters by about 3-fold, whereas MSG1 did not affect the basal level transcription in the absence of TGF-␤. When Smad-mediated transcription was evaluated using a heterologous system consisting of a GAL4-dependent promoter and full-length Smad4 fused to the GAL4 DNA-binding domain, cotransfection of MSG1 enhanced TGF-␤-induced transcription by more than 10-fold. In contrast, TGF-␤-induced transcription from the 3TP promoter, which also requires Smad3 and Smad4 for activation (31,39,40), was rather suppressed by MSG1 by about 60%. These results indicate that the enhancement of Smad-mediated transcription by MSG1 is affected by specific features of Smadbinding promoters, implying that MSG1 may interact with limited conformations of the Smad complexes.
We previously demonstrated interaction of Smad4 with MSG1 by yeast two-hybrid assay (11). Recently, we also demonstrated that Smad4 directly binds to the N-terminal regions of the p300/CBP transcriptional coactivators through its Smad4 activation domain (SAD) (21), which is required for transcriptional activation by Smad4 (32). To determine whether the enhancement of Smad-mediated transcription by MSG1 requires the SAD, we evaluated effects of MSG1 on transcription mediated by wild type or SAD-defective mutant Smad4 fused to the GAL4 DNA-binding domain. As shown in Fig. 1A, Smad4 lacking the SAD did not activate transcription significantly, and MSG1 had no effect on it. Protein expression of the wild type and the SAD-defective Smad4 was comparable when evaluated by anti-GAL4 Western blotting (data not shown). These results indicate an involvement of the SAD, and hence suggest an involvement of the SAD-binding p300/CBP coactivators, in the mechanism of the transcriptional enhancement by MSG1.
Physiological Levels of MSG1 Can Enhance Smad-mediated Transcription-MSG1 enhanced Smad-mediated transcription in mammalian cells when it was introduced by transient transfection ( Fig. 1A and Ref. 11), under which conditions MSG1 protein could have been overexpressed far beyond its physiological levels and therefore not biologically relevant. To determine whether expression of MSG1 at amounts comparable to endogenous levels in expressing cells is also capable of enhancing Smad-mediated transcription, we evaluated effects of stable expression of mouse MSG1 on TGF-␤-induced, Smad-mediated transcription in mouse B16-F10 melanoma cells (Fig. 1B). Three stable clones that expressed MSG1 protein at lower amounts than the endogenous levels of MSG1 in B16-F1 melanoma cells have been isolated; the MSG1 expression in the highest-expression clone (clone 1) was 85% of that in B16-F1 cells when evaluated by Western blotting and densitometry (Fig. 1B, right). Expression of MSG1 in normal human epider-mal melanocyte culture was comparable to that in B16-F1 cells (1). TGF-␤-induced, Smad-mediated transcription in these MSG1-expressing stable clones was stronger than the neo control clones by about 2.5-5-fold, showing a rough correlation between the MSG1 expression levels and the strength of the TGF-␤-induced transcription (Fig. 1B, left). Stable expression of MSG1 did not affect the basal Smad-mediated transcription in the absence of TGF-␤ stimulation. These results indicate that physiological levels of MSG1 are sufficient to significantly enhance Smad-mediated transcription in vivo.
MSG1 Binds to the p300/CBP Transcriptional Coactivators-Enhancement of Smad-mediated transcription by MSG1 required the SAD (Fig. 1A), which is a p300/CBP binding do-main of Smad4 and necessary for Smad4-mediated transcriptional activation (21,32). Therefore, we speculated that the mechanisms of MSG1 action may involve the p300/CBP transcriptional coactivators. To determine whether MSG1 binds to p300/CBP in vivo, MSG1 was transfected into COS-1 cells, and its binding to endogenous p300/CBP was evaluated by immunoprecipitation. MSG1 coprecipitated with both p300 ( Fig. 2A) and CBP (Fig. 2B). A blocking peptide that disrupted binding of p300 to the anti-p300 antibody abolished the coprecipitation of MSG1 with p300 ( Fig. 2A), demonstrating that the MSG1 coprecipitation was specific.
To demonstrate the existence of endogenous protein complexes involving MSG1 and p300/CBP, mouse CBP was immu-

FIG. 1. Enhancement of Smad-mediated transcription by MSG1 transient cotransfection or stable expression.
A, transient transfection assays. NMuMg cells were transiently transfected with luciferase reporter plasmids for Smad-binding promoters (the AREs, the SBEs, the 3TP promoter, or GAL4 binding elements (GAL4BE)), together with expression plasmids for DNA-binding proteins (FAST-1 or Smad4 fusion protein with GAL4DB) and human MSG1 as indicated. Smad4⌬SAD is a Smad4 mutant lacking the Smad4 activation domain. Cells were then cultured in a low (0.5%) serum condition in the presence (filled columns) or absence (open columns) of 10 ng/ml TGF-␤ for 24 h, followed by luciferase activity assay. B, stable expression analyses. B16-F10 cells were stably transfected with an expression plasmid for HA-tagged mouse MSG1 (HA-MSG1) or vector (Neo). Expression of MSG1 protein in stable clones and control mouse melanoma cells was evaluated by Western blotting using anti-MSG1 and anti-HA antibodies (top three panels). To evaluate the strength of Smad-mediated transcription, stable clones were transfected with a GAL4-dependent CAT reporter plasmid together with an expression plasmid for a GAL4 DNA-binding domain fusion protein of the Smad4 C-terminal domain (amino acids 302-552). Cells were then cultured in a low (0.5%) serum condition in the presence or absence of 10 ng/ml TGF-␤ for 24 h, followed by CAT activity assay (bottom panel).
noprecipitated from lysates of non-transfected B16-F1 mouse melanoma cells, which strongly expressed endogenous MSG1 (Fig. 1B) and CBP (data not shown), and coprecipitated endogenous MSG1 was detected by anti-MSG1 Western blotting. As shown in Fig. 2B, MSG1 coprecipitated specifically with CBP; the amount of the CBP-associated MSG1 was estimated to be less than 20% of total cellular MSG1 (quantitative data is not shown).
We attempted to identify domains of MSG1 required for binding to p300/CBP by evaluating the in vivo CBP-binding activity of a series of MSG1 deletion mutants. As shown in Fig.  2C, the C-terminal region of MSG1 was necessary for binding to CBP. When the C-terminal transactivating domain (the CR2 domain, amino acids 145-193 (11); see also Fig. 3A) was deleted, MSG1 did not bind to CBP and, concomitantly, completely lost its activity to enhance Smad-mediated transcription. A deletion of a region immediately before the CR2 domain (amino acids 61-144) also resulted in loss of the CBP binding activity and a concomitant strong reduction in the Smad-enhancing activity. Whereas the ⌬61-144 mutant MSG1 showed weak but detectable Smad-enhancing activity (up to 18% of the wild type MSG1) when strongly overexpressed, the ⌬CR2 mutant MSG1 never showed any detectable effect on Smad-mediated transcription (data not shown; see Ref. 11), suggesting that the CR2 domain is more directly involved in the Smad enhancing activity. On the other hand, an MSG1 mutant lacking the near N-terminal amino acids 30 -60, a region that we previously had demonstrated to be required to enhance Smadmediated transcription (11), coprecipitated with CBP as strongly as the wild type MSG1, indicating that this region is not involved in interaction with p300/CBP. Interestingly, an MSG1 mutant lacking the N-terminal amino acids 1-29 enhanced Smad-mediated transcription more efficiently than the wild type MSG1. We speculated that this region may be in-

FIG. 2. MSG1 binds to the p300/CBP coactivators in vivo.
A, transfected MSG1 binds to endogenous p300. COS-1 cells were transfected with HA-tagged human MSG1, and their cell lysate was subjected to immunoprecipitation using an anti-p300 antibody in the presence or absence of excess amounts of a blocking peptide against this antibody. Coprecipitated MSG1 (asterisk) was detected by anti-HA Western blotting. B, endogenous MSG1 and CBP form a stable protein complex. Cell lysate of non-transfected B16-F1 cells was subjected to immunoprecipitation using a rabbit anti-CBP polyclonal antibody or non-immunized rabbit IgG (control). Coprecipitated mouse MSG1 (asterisk) was detected by anti-MSG1 Western blotting. C, requirement of the C-terminal region of MSG1 for the binding to p300/CBP. HA-tagged human MSG1 and its deletion mutants were transfected to COS-1 cells, and their in vivo binding to the endogenous CBP was evaluated by immunoprecipitation followed by Western blotting (top). Comparable amounts of protein expression of the MSG1 mutants were confirmed by anti-HA Western blotting of cell lysates (middle). Transcriptional enhancement of Smad-mediated transcription by the MSG1 mutants was evaluated as described for Fig. 1B (bottom). volved in negative regulation of the MSG1 activity; this issue is addressed in detail below. The MSG1 effect showed saturation; the wild type and the ⌬1-29 mutant MSG1 showed quantitatively similar effects on Smad-mediated transcription when they were strongly overexpressed (11). Similar results were obtained when binding of MSG1 mutants with p300 was evaluated (data not shown).
The Smad Interaction Domain (SID) of MSG1 Directs Specific Functional Interaction with the Smad Complexes-An MSG1 mutant lacking amino acids 30 -60 did not enhance Smad-mediated transcription (Fig. 2C and Ref. 11), whereas it bound to p300/CBP in vivo (Fig. 2C). Because MSG1 interacted with Smad4 in yeast (11), we speculated that the region consisting of amino acids 30 -60 (which we had tentatively designated SID in our previous publication because of its requirement for MSG1 to enhance Smad-mediated transcription (11)) might interact, at least functionally, with Smads. Supporting this hypothesis indirectly, MRG1, another member of the MSG1 family that shares the CR2 domain but does not have this region, neither interacted with Smad4 in yeast nor enhanced Smad-mediated transcription (11). To address this hypothesis, we utilized a series of domain-swapping mutants of MSG1 and MRG1 and characterized their ability to enhance Smad-mediated transcription. As shown in Fig. 3, MSG1[SIDϾMRG1], an MSG1 mutant the SID of which was replaced by the corresponding sequence of MRG1, did not enhance Smad-mediated transcription. The wild type MRG1 did not affect Smad-mediated transcription either (11). However, an MRG1 mutant that harbored the SID of MSG1 (MRG1ϩSID) enhanced Smad-mediated transcription even more strongly than the wild type MSG1: the strength of the effect of MRG1ϩSID was comparable to that of the MSG1 mutant lacking the N-terminal 29 amino acids (see Fig. 2C). These results provide strong evidence that the SID of MSG1 is necessary and sufficient to direct the specific functional interaction of MSG1 with Smads.
Direct Binding of MSG1 to p300 in Vitro-To determine whether MSG1 binds directly to p300, in vitro binding of purified fusion proteins was evaluated by immunoprecipitation. As shown in Fig. 4, MSG1 strongly bound to the middle region of p300, whereas much weaker but reproducible binding of MSG1 to the N-terminal region of p300 was also observed. MSG1 bound neither to a C-terminal region of p300 nor to the control non-fusion GST, indicating that the observed in vitro MSG1-p300 binding was specific. Thus, MSG1 bound directly to p300 through multiple MSG1 binding sites on p300. Several other transcriptional regulators, such as c-Jun, have also been reported to interact with multiple sites on the p300/CBP sequence (Fig. 4B) (18,41).

MSG1 Enhances Smad-mediated Transcription in a Manner
Dependent on the Coactivator Function of p300/CBP-We next attempted to determine whether p300/CBP are functionally required for the enhancement of Smad-mediated transcription by MSG1. Because the C-terminal acidic domain of MSG1 (the CR2 domain; amino acids 145-193) has strong transcriptional activating activity (2) and is essential for the enhancement of Smad4-mediated transcription by MSG1 (11), we first determined whether p300/CBP are involved in transactivation by this domain. To address this, we evaluated the effects of the adenovirus E1A oncoprotein on transcription mediated by fulllength MSG1, the CR2 domain, or MSG1 mutants that were fused to the GAL4 DNA-binding domain (Fig. 5A). Wild type E1A binds to both p300/CBP and the RB tumor suppressor and inactivates their transcriptional activities. The E1A⌬2-36 mutant binds to RB only, whereas the E1Am928 mutant binds to p300/CBP only (22,42). Transcriptional activation mediated by full-length MSG1 or the CR2 domain alone was strongly suppressed by wild type E1A as well as by the m928 mutant, whereas it was not affected by the ⌬2-36 mutant. A CR2defective MSG1 mutant, which did not bind to p300/CBP (Fig.  2C), had no detectable transactivating activity. The VP16 transactivating domain, which activates transcription by interacting with the basal transcription factors such as TAF II 32 (28,43), activated transcription as strongly as the CR2 domain when fused to the GAL4 DNA-binding domain but was not affected by E1A. An MSG1 mutant the CR2 domain of which was replaced by the VP16 transactivating domain also strongly activated transcription but was not affected by E1A. These results strongly suggest that the mechanism of transcriptional activation by the CR2 domain is dependent on the coactivator function of p300/CBP. We also confirmed that the GAL4 DNA binding domain (GAL4DB) fusion VP16 does not coimmunoprecipitate with p300 when coexpressed in COS-1 cells, whereas it coprecipitated with the human TAF II 32 TATA-binding proteinassociated factor as reported by others (28); on the other hand, TAF II 32 did not coprecipitate with GAL4DB fusion MSG1 (data not shown).
To determine whether p300/CBP are involved in the enhancement of Smad-mediated transcription by MSG1, we first evaluated whether an MSG1 mutant the CR2 domain of which was replaced by the VP16 transactivating domain, which activates transcription as strongly as the CR2 domain but through a mechanism not involving the p300/CBP as shown above, can enhance Smad-mediated transcription. As shown in Fig. 5B, such a mutant MSG1 enhanced Smad-mediated transcription only poorly, whereas its protein expression was comparable to that of the wild type MSG1. This result supported the hypothesis that p300/CBP are involved in the efficient enhancement of Smad-mediated transcription by MSG1.
To obtain further evidence, we evaluated effects of cotrans- , fusion to an MSG1 mutant the CR2 domain of which was replaced by the VP16 transactivating domain) and E1A oncoproteins (wt, wild type; ⌬2-36 and m928, mutants of E1A (see text for explanations)) as indicated. Cells were then cultured for 24 h in the presence of 10% serum, followed by CAT activity assay. B, an MSG1 domain-swapping mutant the CR2 domain of which was replaced by the VP16 viral transactivating domain does not effectively enhance Smad-mediated transcription. Cells were transfected with pG5CAT reporter plasmid and GAL4DB-Smad4(302-552) together with HA-tagged MSG1 or its domain-swapping mutant the CR2 domain of which was replaced by the VP16 transactivating domain (HA-MSG1.VP16), followed by CAT activity assay. Protein expression of HA-MSG1 and HA-MSG1.VP16 was evaluated by anti-HA Western blotting of cell lysates. C, enhancement of Smad-mediated transcription by MSG1 is further augmented by exogenous p300 and suppressed by E1A. NIH3T3 cells were transfected with pG5CAT reporter plasmid and GAL4DB-Smad4(302-552), together with HA-tagged MSG1, p300, or E1A as indicated. Left, effectors (p300 or E1As) were transfected separately; right, effectors were cotransfected. Cells were cultured for 24 h after transfection in the presence of 10% serum, followed by CAT activity assay. Protein expression of HA-MSG1 and HA-p300 was evaluated by anti-HA Western blotting of cell lysates. D, absence of MSG1 effect on p300 directly fused to GAL4 DNA-binding domain. NIH3T3 cells were transfected with pG5CAT reporter plasmid and GAL4 DB alone or GAL4DB fusion full-length p300 together with HA-MSG1 as indicated. The amount of HA-MSG1 cotransfected with GAL4DB was equal to the largest amount cotransfected with GAL4DB-p300. Cells were cultured for 24 h after transfection in the presence of 10% serum, followed by CAT activity assay. fection of p300 and/or E1A on the enhancement of Smad-mediated transcription by MSG1. Cotransfection of p300 augmented the MSG1 effect (Fig. 5C), whereas it had no effect on Smad-mediated transcription in the absence of MSG1 cotransfection (data not shown). The wild type and m928 mutant (p300-suppressing) E1As strongly suppressed the MSG1 effect, whereas the ⌬2-36 mutant (RB-suppressing) E1A had no effect (Fig. 5C, left). Moreover, the suppression of the MSG1 effect by E1Am928 was overcome by p300, and overexpression of p300 resulted in more than 7-fold enhancement of the MSG1 effect, even in the presence of E1Am928 (Fig. 5C, right panel; the smallest amount of p300 cotransfection in the right panel corresponds to the p300 amount in the left panel). Cotransfection of p300 or E1As did not affect protein expression of MSG1 (Fig.  5C, Western blotting). When we considered these data together, we concluded that p300/CBP are functionally involved in the mechanism of the MSG1 effect that enhances Smadmediated transcription.
Our results strongly suggest that MSG1 strengthens the functional link between the Smad complex and the p300/CBP coactivators. To eliminate the possibility that MSG1 functions as a non-specific enhancer of the p300/CBP-dependent transcription, we determined whether MSG1 affects transcription mediated by p300 directly fused to the GAL4 DNA-binding domain (Fig. 5D). GAL4-p300 activated transcription in mammalian cells as reported previously (26) and, as expected, cotransfection of MSG1 (at amounts that strongly activated Smad4-mediated transcription) had no effect on it. Thus, MSG1 does not enhance p300-mediated transcription when p300 is directly recruited to the promoter by physical linking to a DNA-binding protein.
The Hsc70 Heat-shock Cognate Protein Binds to MSG1 and Suppresses Its Physical and Functional Interaction with p300/ CBP-Our results have suggested that the activity of the wild type MSG1 is somehow suppressed in vivo through a mechanism that requires the N-terminal 29 amino acids of MSG1 (Figs. 2C and 3B). Based on a hypothesis that the MSG1 activity may be suppressed in vivo by another MSG1-binding protein, we extended our yeast two-hybrid screening for MSG1binding proteins as described previously (11) and isolated one additional cDNA clone of which the product interacted with MSG1, but not with MRG1, in yeast (data not shown). The new MSG1 interactor clone encoded a part of the human Hsc70 heat-shock cognate protein, a multifunctional molecular chaperone.
We characterized binding of MSG1 to Hsc70 in mammalian cells by transfection-based immunoprecipitation assay. When the full-length human Hsc70 was cotransfected with human MSG1 in COS-1 cells, MSG1 coprecipitated with Hsc70 (Fig.  6A). Several deletion mutants of MSG1, including the one lacking the N-terminal 29 amino acids, did not bind to Hsc70, indicating that the MSG1-Hsc70 binding is specific. By this approach, we found that a large portion of MSG1 N-terminal region is required for the binding to Hsc70, whereas the Cterminal CR2 domain is not required (Fig. 6, A and B).
To determine whether Hsc70 and MSG1 form an endogenous protein complex, mouse MSG1 was immunoprecipitated from lysates of non-transfected B16-F1 mouse melanoma cells, which strongly expressed both MSG1 and Hsc70 proteins (Fig.  6C). Hsc70 coprecipitated with MSG1 by an anti-MSG1 antibody but not by non-immunized control rabbit IgG, indicating that endogenous MSG1 forms stable protein complex with Hsc70. Treatment of B16-F1 cells with TGF-␤ did not affect the MSG1-Hsc70 coprecipitation, indicating that the MSG1-Hsc70 binding is not affected by Smad-mediated signaling.
To determine whether the binding of Hsc70 to MSG1 affects the binding of MSG1 to p300/CBP, effects of Hsc70 cotransfection on binding of MSG1 to endogenous p300 were evaluated by immunoprecipitation (Fig. 6D). An increase in amounts of cotransfected Hsc70 resulted in a concomitant decrease in amounts of MSG1 that coprecipitated with p300, whereas Hsc70 transfection did not affect expression of the MSG1 protein. These results strongly suggest that Hsc70 blocks the binding of MSG1 to p300 in vivo. Finally, we determined whether cotransfection of Hsc70 inhibits the enhancement of Smad-mediated transcription by MSG1. As shown in Fig. 6, E and F, when increasing amounts of Hsc70 were expressed in cells, increased suppression of the transcriptional enhancement by the wild type MSG1was observed, whereas expression of MSG1 protein was not affected by Hsc70 transfection. However, enhancement of Smad-mediated transcription by the MSG1 mutant lacking the N-terminal 29 amino acids (hence not binding to Hsc70) was greater than the wild type MSG1 by about 3.5-fold and not at all suppressed by Hsc70 cotransfection. Taken together, we concluded that the Hsc70 heat-shock cognate protein binds to MSG1 and inhibits binding of MSG1 to p300/CBP in vivo, thus suppressing the MSG1 activity to enhance Smad-mediated transcription.

DISCUSSION
In the present study, we have shown that MSG1 is a p300/ CBP-binding protein and enhances Smad-mediated transcription in a manner dependent on the coactivator function of p300/CBP. The C-terminal region of MSG1 is involved in physical and functional interaction with p300/CBP, whereas the SID located in the N-terminal region of MSG1 directs functional interaction of MSG1 with Smads. Although physiological levels of total cellular MSG1 are sufficient to enhance Smadmediated transcription in vivo, a significant portion of the MSG1 activity is inhibited by the Hsc70 heat-shock cognate protein, which binds directly to MSG1 and suppresses physical and functional interaction of MSG1 with p300/CBP. When p300 was directly fused to a DNA-binding domain, such a fusion protein strongly activated transcription, and MSG1 did not enhance it. These results indicate that, as a non-DNA-binding transcriptional coactivator, MSG1 strengthens the functional link between the DNA-binding Smad transcription factors and the p300/CBP coactivators, thus enhancing Smad-mediated transcription. The functional domain structure of MSG1 described above is summarized in Fig. 7A.
We previously demonstrated that MSG1 interacted with the Smad4 C-terminal domain (amino acids 302-552) in yeast (11), in which no endogenous p300/CBP homologues are expressed. In the present study, we have demonstrated that the SID of MSG1 is necessary and sufficient to direct the specific functional interaction of MSG1 with Smads (Fig. 3) and that the SID is not involved in binding to p300/CBP (Fig. 2C). Thus, it is clear that, at least functionally, MSG1 specifically interacts with Smads through the SID in a manner independent of p300/CBP. However, although it is likely that the MSG1-interacting counterpart resides in the C-terminal region of Smad4, we have been unable to characterize the physical binding of MSG1 to Smads because of the experimental difficulty to demonstrate stable protein complexes containing MSG1 and Smads. This might suggest that the binding of MSG1 to Smads is so weak or transient that the conventional biochemical approaches (such as immunoprecipitation or electromobility gel shift assay) cannot demonstrate it. Alternatively, the MSG1-Smad interaction can be indirect in mammalian cells; the direct target of the SID might not be Smads themselves but Smadassociated proteins. We plan to address these possibilities with future studies.
MSG1 is a founding member of the MSG1 family non-DNA-binding transcriptional activators that are characterized by their conserved C-terminal acidic transactivating domain (the CR2 domain) (Fig. 7B) (1,2,5). Because the CR2 domain of MSG1 is involved in binding to p300/CBP, it is plausible that this family may represent a novel group of p300/CBP-binding transcriptional regulators. Consistent with this notion, a splicing variant of MRG1, another member of the MSG1 family, has been recently isolated as a p300/CBP-binding protein, although the authors proposed it as a competitive inhibitor of HIF-1␣mediated transcription (44). While this report was in preparation, another report appeared indicating that MRG1 directly binds to the LIM homeodomain transcription factor, Lhx2, as well as to p300 and TATA-binding protein and enhances transcription of the glycoprotein hormone ␣-subunit gene (45). Moreover, we have recently isolated a novel member of this family, Mrg2/Cited4, from mouse mammary gland; Mrg2 copre-FIG. 6. The Hsc70 heat-shock cognate protein binds to MSG1 and suppresses the physical and functional interaction of MSG1 with p300/CBP. A, transfected Hsc70 and MSG1 bind in vivo. COS-1 cells were transfected with Myc-tagged Hsc70 and HA-tagged MSG1 (or its deletion mutants) as indicated. Cell lysates were subjected to immunoprecipitation using an anti-HA antibody, and coprecipitated Hsc70 was detected by anti-Myc Western blotting (top panel). Expression of the transfected proteins was evaluated by Western blotting of cell lysates (middle and bottom panels). B, summary of results from the top panel of A. C, endogenous MSG1 and Hsc70 form a stable protein complex. Cell lysate of non-transfected B16-F1 cells was subjected to immunoprecipitation using a rabbit anti-MSG1 antibody or non-immunized rabbit IgG, and coprecipitated Hsc70 was detected by anti-Hsc70 Western blotting (left). Expression of endogenous MSG1 and Hsc70 proteins was evaluated by Western blotting of the cell lysate (right). D, Hsc70 cotransfection suppresses binding of MSG1 to endogenous p300 in vivo. COS-1 cells were transfected with a fixed amount of HA-MSG1 and varying amounts of Myc-Hsc70. Cell lysates were then subjected to immunoprecipitation using an anti-p300 antibody, and coprecipitated MSG1 was detected by anti-HA Western blotting (top panel). Expression of the transfected proteins was evaluated by Western blotting of cell lysates (middle and bottom panels). E, Hsc70 cotransfection suppresses enhancement of Smad-mediated transcription by MSG1. NIH3T3 cells were transfected with a GAL4-dependent CAT reporter plasmid and GAL4DB-Smad4(302-552) fusion protein, together with a fixed amount of HA-MSG1 and varying amounts of Myc-Hsc70. Cells were then cultured for 24 h in the presence of 10% serum, followed by CAT activity assay. A representative CAT assay is shown (top panel). Rel.CAT, relative CAT activity. Expression of HA- cipitated with endogenous p300 when transfected into COS-1 cells. 2 Because the amino acid sequences of MRG1 or Mrg2/ Cited4, other than the CR2 domain, are very different from that of MSG1 (Fig. 7B), it is likely that the CR2 domain, by itself, is directly involved in binding to p300/CBP. Whereas a deletion of a region immediately before the CR2 domain of MSG1 (⌬61-144) resulted in loss of binding to p300/CBP and reduction in the Smad-enhancing activity (Fig. 2C), the suppression of the latter was incomplete (11). A reasonable interpretation of the behavior of the ⌬61-144 would be that this deletion might somehow have affected the accessibility of the CR2 domain to p300/CBP, reducing the affinity between them so that their interaction was undetectable by the stringent immunoprecipitation assays but detectable by more sensitive reporter assays. On the other hand, the N-terminal regions of the MSG1 family proteins show minimal similarity, and no identifiable DNA-binding motifs are found in any of them (Fig.  7B). The SID is unique to MSG1; this is consistent with the fact that three other members of the MSG1 family did not affect Smad-mediated transcription. 2 Based on these observations, it is tempting to speculate that a common feature of the MSG1 family proteins is their ability to bind to p300/CBP through their conserved C-terminal regions (the CR2 domain) while interacting at least functionally with their target DNA-binding proteins through the unique N-terminal regions, thus regulating p300/CBP-dependent transcriptional activation. Justification of this hypothesis would require elucidation of the target DNA-binding proteins for each of the MSG1 family proteins, as well as characterization of their physiological roles.
Our results indicate that the effects of MSG1 on Smadmediated transcription are dependent on the specific features of Smad-binding promoters (Fig. 1A). This may reflect variations of the conformation of the Smad complexes formed on different promoters (13,14). Consistent with this possibility, several non-DNA-binding transactivators have been reported to interact with their target DNA-binding proteins only in the context of specific subtypes of promoters. Thus, the B-cell specific Bob1 non-DNA-binding transactivator interacts with the Oct-1 DNA-binding protein only when Oct-1 is bound to the immunoglobulin promoter, whereas Oct-1 also binds to many other promoters (33). The DNA-binding protein LEF-1 interacts with ␤-catenin and ALY non-DNA-binding transactivators; the former is involved in LEF-1-mediated transactivation of the Wnt target genes but the latter is not, whereas the latter 2 T. Yahata and T. Shioda, unpublished results. is exclusively involved in LEF-1-mediated transcriptional activation of the TCR␣ gene (10,34,35). Demonstration of the possible promoter context-dependent regulation of Smad-mediated transcription by MSG1 awaits isolation of examples of natural Smad-binding promoters that are strongly affected by MSG1.
We have demonstrated that the Hsc70 heat-shock cognate protein binds to MSG1 in vivo and efficiently suppresses its physical and functional interaction with p300/CBP (Fig. 6). However, we presently do not know whether or not Hsc70 regulates the MSG1 activity in physiological contexts. Hsc70 is a multifunctional molecular chaperone (46), and its roles in regulating DNA-binding transcription factors have been reported (47,48). The function of Hsc70 is, in turn, regulated by complex protein-protein interactions with several Hsc70-associating proteins (49,50). It remains to be determined by future studies whether the binding of MSG1 to Hsc70 is regulated or not.
In summary, we have demonstrated that MSG1 is a p300/ CBP-binding protein and enhances Smad-mediated transcription in a manner dependent on the coactivator functions of p300/CBP. MSG1 interacts with Smads and p300/CBP through its N-terminal SID sequence and the C-terminal CR2 domain, respectively, augmenting the functional link between the Smads and p300/CBP. Hsc70 binds to MSG1 and inhibits interaction of MSG1 with p300/CBP as well as the enhancement of Smad-mediated transcription by MSG1. These findings provide important insights into our understanding of the molecular mechanisms of MSG1 activity and Smad-mediated transcriptional activation.