Heteromeric and Homomeric Interactions Correlate with Signaling ...

Download PDF
8 downloads 38 Views 1MB Size Report
Comment
Two homologs, the receptor-associated Smad3 and the tumor suppressor. Smad4/DPC4 ... ing protein named TRIP-1 (3) have been shown to associate with and to be ...... Gelbart. 1995. Genetic characterization and cloning of mothers against dpp, a gene required for decapentaplegic function in Drosophila melanogaster.
MOLECULAR AND CELLULAR BIOLOGY, May 1997, p. 2521–2528 0270-7306/97/$04.0010 Copyright q 1997, American Society for Microbiology

Vol. 17, No. 5

Heteromeric and Homomeric Interactions Correlate with Signaling Activity and Functional Cooperativity of Smad3 and Smad4/DPC4 RUI-YUN WU, YING ZHANG, XIN-HUA FENG,

AND

RIK DERYNCK*

Departments of Growth and Development and of Anatomy, Programs in Cell Biology and Developmental Biology, University of California at San Francisco, San Francisco, California 94143-0640 Received 1 November 1996/Returned for modification 10 December 1996/Accepted 6 February 1997

Homologs of Drosophila Mad function as downstream mediators of the receptors for transforming growth factor b (TGF-b)-related factors. Two homologs, the receptor-associated Smad3 and the tumor suppressor Smad4/DPC4, synergize to induce ligand-independent TGF-b activities and are essential mediators of the natural TGF-b response. We now show that Smad3 and Smad4 associate in homomeric and heteromeric interactions, as assessed by yeast two-hybrid and coimmunoprecipitation analyses. Heteromeric interactions are mediated through the conserved C-terminal domains of Smad3 and Smad4. In Smad3, the homomeric interaction is mediated by the same domain. In contrast, the homomeric association of Smad4 requires both the N-terminal domain and the C-terminal domain, which by itself does not homomerize. Mutations that have been associated with impaired Mad activity in Drosophila or decreased tumor suppressor activity of Smad4/ DPC4 in pancreas cancer, including a short C-terminal truncation and two point mutations in the conserved C-terminal domains, impair the ability of Smad3 and Smad4 to undergo homo- and heteromeric associations. Analyses of the biological activity of Smad3 and Smad4 and their mutants show that full signaling activity correlates with their ability to undergo efficient homo- and heteromeric interactions. Mutations that interfere with these interactions result in decreased signaling activity. Finally, we evaluated the ability of Smad3 or Smad4 to induce transcriptional activation in yeast. These results correlate the ability of individual Smads to homomerize with transcriptional activation and additionally with their biological activity in mammalian cells. Transforming growth factor b (TGF-b)-related factors play important regulatory roles in cell differentiation and proliferation and in tissue morphogenesis in animals ranging from flies and nematodes to mammals (12, 16). Among the many members of the TGF-b superfamily, which include activins and bone morphogenetic proteins (BMPs) (12, 16), TGF-b1 is considered the prototype factor to characterize the various cellular responses and the mechanism of receptor signaling. The growth-inhibitory effect of TGF-b in a variety of cell types, such as epithelial cells, and its ability to induce gene expression that leads to increased extracellular matrix deposition have been best characterized (5). Similarly to BMP-2/4 and activin, TGF-b exerts its activities through two types of cell surface serine/threonine kinase receptors, type II and type I receptors. The type II and type I receptors have the ability to form a heterotetrameric complex, consisting of two type II and two type I receptors, and this complex is thought to represent the fully functional receptor complex that mediates TGF-b signaling following ligand binding. In this tetrameric complex, the constitutively active type II receptors phosphorylate the type I receptor cytoplasmic domains, and this phosphorylation is required for activation of the signaling pathways leading to the TGF-b response (4, 18, 22). The downstream kinase targets of the serine/threonine kinase receptors are as yet poorly characterized. The a-subunit of farnesyl transferase (14, 20, 21) and a WD-repeat-contain-

ing protein named TRIP-1 (3) have been shown to associate with and to be phosphorylated by the type I and type II receptor cytoplasmic domains, respectively, but a role of these proteins as signaling effectors has not been documented. Recent genetic analyses of Drosophila melanogaster and Caenorhabditis elegans and developmental and biochemical analyses of Xenopus laevis embryos and mammalian cells have identified Drosophila Mad, C. elegans Sma, and the vertebrate homologs, Smads, as signaling mediators downstream from the receptors for TGF-b-related factors (6, 18). All Mad-related proteins show structural conservation of an N-terminal and a C-terminal domain, which are separated by a proline-rich spacer sequence of variable length and sequence (6). In vertebrates, Smad1 has been implicated as a signaling mediator for BMP-2 and BMP-4 (9, 13, 17) whereas the closely related Smad2 and Smad3 have been implicated as effectors of the activin and TGF-b responses, respectively (2, 9, 24). In contrast, Smad4, which is identical to the human tumor suppressor DPC4 (11), is involved in the response to not only TGF-b (24) but also activin and BMP-2/4 (16a, 25). We have previously shown that Smad3 and Smad4 strongly synergize to induce a potent TGF-b-like response in the absence of ligand stimulation. Furthermore, overexpression of dominant negative mutants of Smad3 or Smad4 blocks the natural TGF-b response. These results reveal a functional cooperation of Smad3 and Smad4 and show that both Smads are required effectors of TGF-b receptor signaling (24). Smad3 physically associates with the receptor complex and is phosphorylated. Its efficient phosphorylation by the type I receptor kinase in vitro suggests that Smad3 may be a direct kinase target for the TGF-b receptor complex. In contrast, Smad4 is not associated with the receptor and is not phosphorylated

* Corresponding author. Mailing address: Department of Growth and Development, University of California at San Francisco, San Francisco, CA 94143-0640. Phone: (415) 476-7322. Fax: (415) 476-1499. E-mail: [email protected]. 2521

2522

WU ET AL.

(24). Finally, Smad3 and Smad4 are not only present in the cytoplasm but are also concentrated in the nucleus of transfected cells (25), where they are thought to play a role in transcriptional activation of TGF-b responsive genes (6, 17, 18). The functional cooperation of Smad3 and Smad4 in TGF-b receptor signaling led us to examine whether they can physically associate. Using yeast two-hybrid assays and coimmunoprecipitation analyses of transfected cells, we show that Smad3 and Smad4 form homomeric and heteromeric associations. The domains required for these interactions were characterized, and our results emphasize an important role of the conserved C-terminal domains in these interactions. Naturally occurring mutations of Smads disrupt the ability of Smad3 and/or Smad4 to undergo homomeric and heteromeric interactions, which correlates with impaired signaling. Taken together, our findings show that the abilities of Smad3 and Smad4 to form homomeric and heteromeric associations correlate with efficient signaling activity. MATERIALS AND METHODS Yeast two-hybrid assays. The LexA-based yeast two-hybrid system was used to assay for protein interactions as described previously (10, 23). cDNA sequences encoding full-size Smad2, Smad3, or Smad4 or the cytoplasmic domain of the Tsk7L type I receptor (amino acids 147 to 509) (7) were isolated by incorporating 59 EcoRI and 39 XhoI sites during PCR amplification with Pfu polymerase (Stratagene) and subcloned in plasmids pEG202 and pJG4-5. The Smad1 coding sequence was released from pRK5-Smad1 (24) with EcoRI and SalI and was ligated into pEG202 and pJG4-5. All mutations of Smad3 and Smad4 were introduced by PCR-based techniques. All Smad3 and Smad4 plasmids were confirmed by DNA sequencing (U.S. Biochemical Corp.) following subcloning. Yeast EGY48 was transformed with the expression plasmids by using Alkali Cation (BIO 101, Inc.). The expression of the fusion proteins of the appropriate size was verified by Western blotting with anti-HA antibody, which recognizes the epitope-tagged fusion proteins (1). Protein interactions and resulting transcriptional activation were determined by scoring b-galactosidase activity on X-Gal (5-bromo-4-chloro-3-indolyl-b-D-galactoside) plates containing galactose or glucose (1). b-Galactosidase activity was measured in solution with o-nitrophenyl-b-D-galactoside as the substrate as described previously (1). Enzymatic activity was calculated from the following equation: enzyme units 5 1,000(A420/ A600tv) (t 5 time and v 5 volume). Transient transfections and immunoprecipitations. pRK5-based expression plasmids for Smad3 and Smad4 have been described previously (24). Mutations were introduced by PCR-based methods. COS-1 cells were transfected with Lipofectamine (Life Technologies, Inc.) as specified by the manufacturer. At 48 h after transfection, the cells were lysed in MLB lysis buffer consisting of 20 mM Tris-HCl (pH 8.0), 137 mM NaCl, 0.5% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 5 mM benzamidine, and 10 mg each of aprotinin and leupeptin per ml. Lysates were precleared with anti-mouse immunoglobulin G (Jackson Laboratories) and protein A-Sepharose CL4B (Pharmacia), and immunoprecipitations were performed with the epitope-specific monoclonal antibodies 9E10 (anti-myc) or M2 (anti-Flag) or with preimmune antiserum. Immunoprecipitated proteins were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis and probed in Western blots with anti-myc or anti-Flag antibodies. PAI-1 luciferase reporter assay. Expression plasmids for wild-type and mutant Smad3 and Smad4 were cotransfected with the reporter plasmid p800luc (15) to assay for expression from the TGF-b-inducible PAI-1 promoter. Plasmid pRKbgal (8), which expresses b-galactosidase under the control of the cytomegalovirus promoter, was included in all transfections to allow normalization of the luciferase activity for transfection efficiency. SW480.7 cells were transfected with Lipofectamine in six-well dishes. In each transfection, 2 mg of each expression plasmid, 0.5 mg of p800luc, and 0.5 mg of pRKbgal were used. At 12 h after transfection, the cells were treated with 5 to 10 ng of TGF-b1 per ml in medium containing 0.2% fetal bovine serum or were left untreated. After 24 h, the cells were lysed and luciferase assays were performed as described previously (8). GST fusion proteins and adsorption. Glutathione-S-transferase (GST) fusion proteins of Smad3 and Smad4 were expressed as described previously (24). 35 S-labelled Smad3 or Smad4 or defined protein segments were generated by in vitro transcription from pRK5 plasmids containing the corresponding cDNA sequences and translation in the presence of [35S]methionine. The translation products were mixed with purified GST-Smad3 or GST-Smad4 immobilized on glutathione-Sepharose in 50 mM Tris-HCl (pH 7.5)–120 mM NaCl–2 mM EDTA–0.1% Nonidet P-40 for 2 h. Bound proteins were analyzed by SDS-PAGE and exposed for autoradiography for 3 days. Yeast transcription assays. Wild-type and mutant Smad3 and Smad4 were expressed as fusion proteins with the LexA DNA binding domain in plasmid

MOL. CELL. BIOL. pEG202 and introduced into yeast EGY48 containing the b-galactosidase reporter gene plasmid pSH18-34 (1). Transformants were grown on glucose-containing X-Gal plates. Transcriptional activity was monitored by measuring b-galactosidase activity, similarly to the yeast two-hybrid assays.

RESULTS Direct homomeric and heteromeric associations of Smad proteins. We have previously shown that Smad3 and Smad4 synergize to induce TGF-b-like responses and that their functional interaction is required to mediate the natural TGF-b response (24). To explore the biochemical basis of this cooperativity, we investigated the ability of Smad3 and Smad4 to physically and functionally associate with each other. For this purpose, we used the yeast two-hybrid system, which lacks endogenous Smads and scores physical interactions between proteins, based on transcriptional activation of a reporter gene such as the b-galactosidase gene. While testing the heteromeric associations of these two Smads, we also evaluated their ability to engage in homomeric interactions, and we incorporated Smad1 and Smad2 in these assays. Two groups of expression plasmids were constructed by fusing full-length Smad1, Smad2, Smad3, or Smad4 to the LexA DNA binding domain (plasmid pEG202) as bait or to the nuclear localized B42 acidic activation domain (plasmid pJG4-5) as prey. Coexpression of any combination of Smads as bait and prey in the yeast strain pEGY48 resulted in a strong transcriptional activation of the b-galactosidase reporter gene, as detected by the intensely blue color following staining for enzymatic b-galactosidase activity (Fig. 1). In these assays, Smad3 and Smad4 associated in both homomeric and heteromeric interactions. Furthermore, Smad1 and Smad2 also showed homomeric interactions and associated with Smad4. The homomeric and heteromeric interactions were similar in strength, as assessed by the b-galactosidase activities, and were comparable to the association between the type I Tsk7L receptor cytoplasmic domain and FKBP12. The Tsk7L cytoplasmic domain did not associate with any of the Smads. To evaluate whether homomeric and heteromeric associations of Smad3 and Smad4 also occur in vivo, we performed coimmunoprecipitation analyses of transfected cells. Thus, COS-1 cells were cotransfected with expression plasmids for Smad3 and/or Smad4, differentially tagged with a C-terminal myc or Flag epitope. Lysates from unlabelled cells were subjected to immunoprecipitations with tag-specific antibodies followed by Western blot analysis with an antibody against the second tag. As shown in Fig. 2A, anti-Flag antibody not only precipitated Flag-tagged Smad3 or Smad4 (middle panel) but also coexpressed myc-tagged Smad3 or Smad4 (right panel). Thus, both Smad3 and Smad4 associate in homomeric interactions. To assess heteromeric interactions, we coexpressed differentially tagged Smad3 and Smad4. As shown in Fig. 2B, immunoprecipitation of Smad3 resulted in coprecipitation of Smad4 and vice versa, thus demonstrating heteromeric complex formation between the two Smads. To further assess the specificity of the immunoprecipitations, we coexpressed myctagged Smad4 with Flag-tagged Smad3 or Smad4. As shown in Fig. 2C, the control antibody did not precipitate Flag-tagged Smad3 or Smad4 (left panel) or myc-tagged Smad4. Mapping the interaction domains of Smad3 and Smad4. All Smads have two highly conserved domains, one in the Nterminal third and one in the C-terminal third of the proteins, which are separated by a proline-rich sequence of variable length and sequence (6, 18). To define the domains of Smad3 and Smad4 responsible for the homomeric and heteromeric associations, segments of Smad3 and Smad4 were cloned into the prey plasmid pJG4-5 and tested for their ability to associate

VOL. 17, 1997

INTERACTIONS OF Smad3 AND Smad4

2523

Smad4 did not undergo homomeric interactions with the Smad4 C-terminal domain but associated efficiently with the C-terminal domain of Smad3 (Fig. 3B). Finally, short C-terminal segments (amino acids 343 to 424 in Smad3 and amino acids 439 to 552 in Smad4) of the conserved C-terminal domains of both Smads did not undergo homomeric or heteromeric interactions with full-size Smad3 or Smad4, suggesting that this segment is not sufficient to mediate Smad interactions. Taken together, our results show that the conserved C-terminal domain of Smad3 is sufficient to mediate both homomeric and heteromeric interactions whereas the corresponding domain in Smad4 is fully responsible for the heteromeric interaction with Smad3 but cooperates with N-terminal sequences to mediate the strong homomeric interactions. Furthermore, the C-terminal domain of Smad4 does not show homomeric associations.

FIG. 1. Homomeric and heteromeric associations of Smads in yeast twohybrid assays. The full-length coding sequences of Smad1, Smad2, Smad3, and Smad4 and FKBP12 were fused in frame with the transcriptional activation domain B42 in plasmid pJG4-5. Individual plasmids were transformed into yeast containing the coding sequences for Smad1, Smad2, Smad3 or Smad4 or the type I Tsk7L receptor cytoplasmic domain, fused in frame with the LexA DNA binding domain in plasmid pEG202. Transformants were tested on galactosecontaining X-Gal plates, which allowed scoring of the association by the presence of a dark blue color. (A) Photograph of the dark blue yeast colonies, indicative of the associations between Smads. (B) The associations between Smads and control proteins are scored on a scale from 2 to 1111. White colonies after a 48-h incubation are scored as 2. The blue colonies are scored from dark blue (1111) to light (11) blue after 24 h of incubation.

with full-size Smad3 and Smad4 expressed as LexA fusion proteins. As shown in Fig. 3A, the N-terminal segment of Smad3, consisting of the conserved domain and the prolinerich sequence, did not associate with either Smad3 or Smad4. In contrast, the C-terminal conserved domain of Smad3 showed homomeric and heteromeric interactions with full size Smad3 and Smad4, respectively, with a b-galactosidase activity similar to that when full-size Smad3 was used as bait. Thus, this conserved domain of Smad3 mediates both homomeric and heteromeric interactions. In contrast to Smad3, both the N-terminal and C-terminal segments of Smad4 associated with full-size Smad4, although to a lesser extent than in the homomeric interactions of the full-size proteins. However, only the C-terminal conserved domain of Smad4, and not the N-terminal segment, associated with full-size Smad3, and this association was similar in strength to the heteromeric association between full-size Smad3 and Smad4 (Fig. 3A). The C-terminal domains of Smad3 and Smad4 were also expressed as LexA fusion proteins and tested for their ability to associate with the same C-terminal domains of Smad3 or Smad4. The C-terminal domain of Smad3 interacted efficiently with the C-terminal domains of both Smad3 and Smad4. In contrast, the C-terminal domain of

FIG. 2. Homomeric and heteromeric associations of Smad3 and Smad4 in transfected cells. COS-1 cells were transfected with expression vectors for mycor Flag-tagged Smad3 and/or Smad4. Cell extracts were immunoprecipitated with either anti-myc or anti-Flag antibody. Following gel electrophoresis, the Western blots were probed with anti-myc or anti-Flag, as indicated. (A) Homomeric interactions of Smad3 and Smad4. Lanes: 1, Smad3-myc and Smad3-Flag were coexpressed; 2, Smad4-myc and Smad4-Flag were coexpressed. (B) Heteromeric associations of Smad3 and Smad4. Lanes: 3, Smad3-myc and Smad4Flag were coexpressed; 4, Smad3-Flag and Smad4-myc were coexpressed. (C) Antibody control of the immunoprecipitations. Smad4-myc was coexpressed with Smad3-Flag or Smad4-Flag as indicated, and immunoprecipitations (I.P.) were performed with the anti-Flag antibody or a control antiserum (lanes C) as shown, followed by immunoblotting with anti-Flag or anti-myc antibodies as shown. The control antibody did not immunoprecipitate Flag-tagged or myc-tagged Smads. The immunoglobulin B band which is recognized by the secondary antibody in Western blots of immunoprecipitations is marked with an asterisk.

2524

WU ET AL.

FIG. 3. Homo- and heteromerization of Smad3 and Smad4 segments in yeast two-hybrid assays. (A) Full-length Smad3 and Smad4, and defined segments of Smad3 and Smad4, as schematically shown, were used to define the interacting domains. These segments and full-size forms were made as LexA fusion proteins, and their ability to associate with full-size Smad3 or Smad4 was scored based on the induction of b-galactosidase activity. 1111 indicates dark blue and 1 indicates light blue after 24 h of incubation on galactose-containing X-Gal plates. 2 denotes white colonies after 48 h of incubation. (B) Similar assays scored the ability of the C-terminal domains of Smad3 and Smad4 to homomerize or heteromerize. The C-terminal domain of Smad4 did not homomerize.

These observations in yeast two-hybrid assays were also confirmed by using GST fusion proteins. 35S-labelled Smad3 and Smad4 and their C-terminal domains, generated by in vitro transcription-translation, associated efficiently with the GSTSmad3 and GST-Smad4 fusion proteins (data not shown). Effects of mutations in Smad3 and Smad4 on association and biological activity. Smad3 and Smad4 strongly synergize to induce gene expression from the PAI-1 promoter with luciferase as the reporter gene (24). This activity correlated with the ability of Smad3 and Smad4 to undergo homomeric and heteromeric interactions (Fig. 1 and 3). Since the C-terminal domain of Smad3 associated efficiently with full-size Smad4 (Fig. 4A), we evaluated the biological activity of this Smad3 domain when coexpressed with full-size Smad4. As shown in Fig. 4B, coexpression of these two proteins induced a PAI-1 response that was comparable to the synergistic response of full-size Smad3 and Smad4. In contrast, the N-terminal domains of Smad3 or Smad4, when coexpressed with full-length Smad4 or Smad3, respectively, were unable to induce a potent PAI-1 response (Fig. 4B). We also tested the biological activity of the C-terminal segment of Smad4 in combination with fullsize Smad3. Even though these proteins heteromerized efficiently (Fig. 4A), they showed only a weak biological activity (Fig. 4B). This inactivity does not correlate with a lack of heteromerization but with an inability of the C-terminal domains of Smad4 to homomerize (Fig. 3B). Accordingly, coexpression of the C-terminal domains of Smad3 and Smad4 does not induce PAI-1 luciferase activity (25), which is consistent with the inability of the Smad4 C-terminal domain to homomerize, even though the heteromerization of both Smad3

MOL. CELL. BIOL.

and Smad4 domains is efficient (Fig. 3B). These results suggest that both homomeric and heteromeric interactions are required for the biological activity of Smad3 and Smad4. To further explore the correlation between the biological activity of Smad3 and Smad4 and their ability to form a heteromeric complex, we examined the effects of several mutations in Mad or Smads, which have previously been isolated as inactivating mutations. One of these was a truncation of the C-terminal 39 amino acids in Drosophila Mad, which contributed to the initial identification and characterization of Mad. The corresponding mutation has also been identified in Smad4/DPC4 in pancreas carcinomas (11). This mutation incorporated into Smad3 and Smad4 resulted in their inactivity, as assessed by the inability of Smad3DC to synergize with Smad4 or the inability of Smad4DC to synergize with Smad3 in the PAI-1–luciferase assay (24). Since the conserved C-terminal domain mediates association, we evaluated the ability of these truncated Smads, i.e., Smad3DC and Smad4DC, to undergo homomeric and heteromeric associations in yeast twohybrid assays. As shown in Fig. 4A, Smad3DC did not detectably interact with Smad3 or Smad4, and Smad4DC did not detectably interact with Smad3 and interacted only weakly with Smad4. In addition, the same truncation incorporated into the C-terminal domain of Smad3 (Smad3CDC) or Smad4 (Smad4CDC) abolished the ability of this domain to associate with full-size Smad3 and/or Smad4 (Fig. 4A). Thus, the lack of biological activity correlated with an inability of the truncated forms to confer homomeric or heteromeric interactions. In Drosophila, a missense mutation of glycine at position 409 to serine confers a phenotype similar to that of truncated Mad, suggesting that this G/S mutation inactivates Mad (19). This mutation, incorporated in the corresponding position in Smad3, decreased its ability to form homomers with wild-type Smad3 and to form heteromers with Smad4 about 10- and 20-fold, respectively (Fig. 4A). To evaluate the biological activity of Smad3(GS), we coexpressed this mutant with Smad4 and scored their ability to induce luciferase expression from the PAI-1 promoter. The G/S mutation conferred a 30% decrease in activity of Smad3 (Fig. 4B). Another mutation resulting in an aspartic acid-to-histidine replacement at position 512 in Smad4 was identified in the Smad4/DPC4 gene in human pancreatic carcinomas (11). This point mutation in Smad4 abolished its ability to interact with Smad3 and strongly decreased the homomerization efficiency of Smad4. This mutation incorporated into the C-terminal domain of Smad4, expressed by itself, Smad4C(DH), also abolished its ability to interact as a heteromer with Smad3 and as a homomer with Smad4 (Fig. 4A). In the PAI-1–luciferase assay, this mutation resulted in a decrease of Smad4 activity of about 80% (Fig. 4B). In parallel experiments, the wild-type and mutant forms of Smad3 and Smad4 were expressed at similar levels in COS-1 cells (Fig. 4C) and SW480 cells (data not shown). Taken together, our analyses showed that these mutations abolished or decreased the ability of Smad3 and Smad4 to engage in homomeric or heteromeric interactions. This decreased ability to interact correlated with decreased biological activity, which further supports our conclusion that both homomeric and heteromeric interactions may be required for biological activity. Transcriptional activity of Smad3 and Smad4 in yeast. Previous results have shown that the C-terminal half of Smad1 and Smad4/DPC4 induce transcriptional activation in mammalian cells when fused to a GAL4 binding domain which allows direct interaction with DNA (17). However, these assays were performed in the presence of the widely expressed, endogenous Smads, with which the assayed Smad could associate, an important caveat considering the strong synergism of Smad1,

VOL. 17, 1997

INTERACTIONS OF Smad3 AND Smad4

2525

FIG. 4. Effects of mutations in Smad3 and Smad4 on association and biological activity. (A) Analysis of mutants of Smad3 and Smad4, expressed as LexA fusion proteins with full-size Smad3 and Smad4. Interactions were scored based on the induction of b-galactosidase activity. 1111 indicates dark blue, and 1 indicates light blue after 24 h of incubation on galactose-containing X-Gal plates. 2 denotes white colonies after 48 h of incubation. The numbers indicate the units of b-galactosidase activity quantitated from enzymatic assays in solutions. The values shown are means of duplicate assay results. (B) Effects of Smad3 and Smad4 mutations on their ability to induce PAI-1 expression. The combination of Smad3 and Smad4 mutants is as shown. PAI-1 induction was scored in both the presence (■) and absence (h) of TGF-b. The results shown are means of triplicate assays, which differed less than 10%. (C) Expression of wild-type and mutant forms of Smad3 or Smad4. COS-1 cells were transfected with expression plasmids for the Flag-tagged Smads as shown, and the Smads were detected by immunoblotting of the lysates with anti-Flag antibody.

Smad2, or Smad3 with Smad4 (24, 25). We therefore set up a transcriptional activation assay in yeast which lacks endogenous Smads and evaluated the ability of Smad3 and Smad4 and their N- and C-terminal segments to induce transcriptional activation. Smad3 and Smad4 were individually fused to the LexA DNA binding domain and coexpressed in yeast with a reporter plasmid in which b-galactosidase expression is under the control of the LexA operator. This one-hybrid system stands in contrast to the two-hybrid system, in which a high level of b-galactosidase expression is produced as a result of the ability of two proteins to interact with each other and the resulting interaction of the B42 activation domain with transcriptional regulators upstream from the reporter gene. As shown in Fig. 5, full-length Smad3 and Smad4 had weak transcriptional activity. However, the C-terminal domain of Smad3 had a strong activity in this assay. This activity was abolished by the C-terminal truncation, which inactivates this interaction domain and abolishes the biological activity of Smad3. Fur-

thermore, the G/S point mutation also inactivated the transcriptional activity of this domain, consistent with its ability to interfere with the association. In contrast to Smad3, the Cterminal domain of Smad4 had no detectable activity, and the weak activity of full-size Smad4 was associated with the Nterminal segment. In addition, the DH mutation abolished the transcriptional activity of Smad4. Thus, the C-terminal domain of Smad4 behaves remarkably differently from Smad3 in this transactivation assay. In addition, mutations that decrease or abolish homomeric associations also abolish transcriptional activation. Thus, these data strongly suggest that homomerization is required for transcriptional activation. DISCUSSION Homologs of Drosophila Mad have recently been implicated as mediators of receptor signaling by TGF-b-related factors (6, 18). The vertebrate homologs Smad3 and Smad4/DPC4, a tu-

2526

WU ET AL.

FIG. 5. Transcriptional activation by Smad3 or Smad4 in yeast. Full-length and mutant forms of Smad3 or Smad4 were expressed as fusion proteins with the LexA DNA binding domain and transformed into yeast containing the b-galactosidase reporter gene for transcriptional activation. Transformants were grown on glucose-containing X-Gal indicator plates for 3 days. (A) Transcriptional activation is shown by the intensity of the blue color. The numbers correspond to the individual Smad3 or Smad4 mutants or segments shown in panel B. (B) Schematic presentation of individual Smad3 or Smad4 segments fused to the LexA domain. The intensity of the blue color is indicative of transcriptional activity and was scored on a scale from 111 to 2. The numbers represent b-galactosidase activity measured in solution, as in panel A.

mor suppressor, have been identified as effectors of the TGF-b response downstream from the receptors. Thus, both Smad3 and Smad4 are required for TGF-b receptor signaling, and their coexpression results in a potent TGF-b-like response in the absence of ligand stimulation (24). The use of quantitative

MOL. CELL. BIOL.

assays in characterizing the Smad3 and Smad4 activities revealed a remarkable ability to strongly synergize with each other. This cooperativity between the two Smads forms the basis of their role as signaling mediators, since inactivation of either Smad3 or Smad4 activity by dominant negative interference blocks receptor signaling (24). However, this cooperativity is not restricted to the functional interactions between Smad3 and Smad4, since Smad4 also synergizes with Smad2 and Smad1 (16a, 25). In contrast, Smad1, Smad2, and Smad3 do not synergize with each other to induce gene expression. Thus, Smad4 plays a central role in signaling by different TGFb-related factors and receptor systems, distinct from the other Smads. Consistent with this function of Smad4 are the observations that Smad4 does not associate with the receptor and is not phosphorylated whereas Smad1, Smad2, and Smad3 are phosphorylated, presumably as a result of their association with receptors, as has been shown for Smad3 (24). Finally, Smad4 is also structurally distinct from the other, more closely related Smads. Smad4 contains an insert in the C-terminal conserved domain and is considerably larger than the other Smads (6, 11). In this study, we have evaluated the structural basis of the functional cooperativity of Smad3 and Smad4, which is at the basis of Smad signaling, and have characterized the homomeric and heteromeric physical associations as being required for Smad activity. Heteromeric and homomeric interactions were evaluated by a combination of yeast two-hybrid assays, which score proteinprotein associations, and immunoprecipitations of Smads coexpressed in transfected cells. The observed heteromeric association of Smad4 with Smad3 may explain the functional cooperation of the two Smads in TGF-b receptor signaling and their ability to synergize in inducing a ligand-independent response (24). In addition, Smad4 associates similarly with Smad1 and Smad2, which is consistent with the ability of Smad4 to synergize with these Smads and with their functional cooperativity (16a, 25). Besides the heteromeric interactions, we also observed equally strong homomeric associations of the individual Smads. These homomeric interactions are, in the case of receptor-associated Smads, consistent with the presence of two type II and two type I receptors in the receptor complex, with which Smads are most probably associated as homomers. Finally, Smad1, Smad2, and Smad3 also form heteromeric interactions with each other. This may functionally correlate with an association with receptor complexes for heteromeric ligands, such as inhibin or BMP-2/6 and BMP-2/7. These receptor complexes most probably contain two different type II or type I receptors. A characterization of the domains which mediate homo- and heteromeric interactions reveals differences between Smad3 and Smad4 and shows a clear correlation between the biological activity of Smad3 and Smad4 and their ability to associate. In Smad3, the conserved C-terminal domain mediates both homomeric interactions and heteromeric association with Smad4 (Fig. 3). Accordingly, when coexpressed with full-size Smad4, the C-terminal domain of Smad3 induces full biological activity, comparable to full-size Smad3 and Smad4 (Fig. 4B), thus indicating that the conserved C-terminal domain of Smad3 is sufficient to synergize with Smad4. Similarly, the ability of the C-terminal domains of Smad1 and Smad2 to mimic the activity of their full-length counterparts (2, 17) is probably due to cooperation with endogenous Smad4. In contrast, the C-terminal domain of Smad4 mediates the heteromeric interactions with Smad3 but not the homomeric interactions. Furthermore, the conserved N-terminal domain of Smad4, but not of Smad3, mediates homomerization, although not as efficiently as full-size Smad4 does (Fig. 3). Accordingly,

VOL. 17, 1997

coexpression of the C-terminal domain of Smad4 with full-size Smad3 induces only a weak response (Fig. 4B), and the coexpressed C-terminal domains of both Smad3 and Smad4 are biologically inactive (25). These two C-terminal domains show efficient heteromerization and homomerization for Smad3 but no homomerization for Smad4 (Fig. 3). Taken together, our data strongly suggest that homomerization and heteromeric interactions are both required for full synergistic biological activity and emphasize the functionally distinct nature of Smad4. To further explore the correlation between the ability of Smad3 and Smad4 to interact and their biological activity, we characterized three mutations in the conserved C-terminal domain that have been associated with impaired activity of either Mad in Drosophila (19) or the tumor suppressor Smad4/DPC4 in pancreas cancers (11). One of these mutations is a short C-terminal truncation, which in both Smad3 and Smad4 strongly decreases their ability to homo- or heteromerize (Fig. 4A). Accordingly, this mutation inactivates Smad3 and Smad4 signaling (24) (Fig. 4B). Another mutation is a single Gly-toSer replacement, which was identified in Drosophila Mad and impaired development (19). This mutation strongly decreased the ability of Smad3 to homo- or heteromerize and decreased the activity of Smad3 when coexpressed with Smad4. However, Smad3(GS) was not inactive in the functional reporter assay. Another single point mutation results in an Asp-to-His replacement in Smad4/DPC4 and was originally detected in pancreas carcinoma (11). This mutation abolished the interaction of Smad4 with Smad3 and decreased the ability of mutant Smad4 to associate with wild-type Smad4, consistent with a role of the N-terminal domain in Smad4 homomerization. Accordingly, the ability of Smad4(DH) to cooperate with full-size Smad3 in signaling activity was strongly impaired (Fig. 4). These results further emphasize the role of homo- and heteromerization in Smad signaling and reveal the mechanistic basis for impaired Mad function in Drosophila and Smad4/DPC4 function in pancreas cancer. We also tested the ability of Smad3 and Smad4 to induce transcriptional activation in yeast. Full-size Smad3 or Smad4, or derivative segments or mutants, were fused to the LexA domain, which allows subsequent DNA binding and activation of a reporter gene. Since this LexA domain mediates dimerization (1), the mere dimerization of fusion proteins is insufficient to induce transcriptional activation. Conceptually similar experiments have revealed the ability of Smad1 and DPC4 to induce transcriptional activation in transfected mammalian cells (17). However, vertebrate cells contain endogenous Smads, and so the ability of a tested Smad to transactivate might have been biased by its interaction with endogenous full-size Smads. Our results show that the C-terminal domain of Smad3 has a strong ability to induce transcriptional activation. This activity was considerably higher than the activity of full-size Smad3, which is consistent with a potential inhibitory role of the N-terminal domain on Smad activity (17, 18). The ability of the C-terminal domain of Smad3 to activate transcription correlates well with its ability to homomerize, since the C-terminal truncation and the GS point mutation in this domain impaired both transcriptional activity (Fig. 5) and homomerization (Fig. 3A and 4A). The ability of Smad4 to induce transcription in yeast was remarkably different from that of Smad3. The C-terminal conserved domain of Smad4 was inactive (Fig. 5), which is again consistent with its inability to homomerize (Fig. 3B). This result contrasts with the transactivation by a similar Smad4 segment in mammalian cells (17), which presumably resulted from a heteromeric interaction with endogenous Smads. However, full-size Smad4 had a weak tran-

INTERACTIONS OF Smad3 AND Smad4

2527

TABLE 1. Correlations of association with transcriptional activation and the synergistic biological activity of Smad3 and Smad4a Smad

Homomer formation

Heteromer formation

Transcriptional activity

Synergistic activity

Smad3 Smad3N Smad3C Smad3CDC Smad3DC Smad3(GS) Smad3C(GS) Smad4 Smad4N Smad4C Smad4CDC Smad4DC Smad4(DH) Smad4C(DH)

1111 2 1111 2 2 1 1 1111 1 1 2 1 1 2

1111 2 1111 2 2 1 1 1111 2 1111 2 2 2 2

1 1 111 2 2 2 2 11 11 2 2 2 2 2

1111 2 1111 ND 2 111 ND 1111 1 11 ND 2 11 ND

a The ability to interact was scored against full-length Smad3 or Smad4, and the biological activity was evaluated when coexpressed with full-length Smad3 or Smad4. ND, not determined.

scriptional activity in yeast which correlates with the weak homomerization (Fig. 3A and 4A) and transactivation (Fig. 5) by the N-terminal domain of Smad4. Remarkably, the DH point mutation in the C-terminal domain of Smad4 inactivated the transcriptional activation compared to wild-type Smad4, even though the N-terminal domain that plays a role in the homomerization was not mutated. This interference may be associated with a functional interaction between the conserved N-terminal and C-terminal domains of Smads (17). Our results thus clearly correlate the ability of Smads to homomerize with transcriptional activity. Furthermore, the ability of Smads to induce transcription in yeast now enables us to map the structural requirements for transcriptional activation by Smads by using genetic screens with randomly mutagenized libraries of Smad fragments. In summary, our results have characterized the ability of Smad3 and Smad4 to form homomeric and heteromeric associations and emphasize the distinct structural and functional nature of Smad4 in mediating Smad and receptor signaling. The cooperativity of Smad4 with Smad3 and, by extension, with other Smads is clearly at the basis of receptor signaling by Smads. Our findings, summarized in Table 1, show a consistent correlation between the abilities of Smads to homo- and heteromerize, the transcriptional activation in yeast, and the biological activity in mammalian cells. Naturally occurring mutations that impair Smad signaling are also correlated with decreased or abolished associations. The yeast transactivation system will now allow us to precisely map the structural requirements for the biological activity of Smad3. ACKNOWLEDGMENTS This research was supported by NIH grant CA54826 to R.D., a postdoctoral NIH training grant of the Cardiovascular Research Institute at UCSF to Y.Z., and an American Cancer Society postdoctoral fellowship to X.-H.F. We thank D. J. Loskutoff for plasmid p800luc.

ADDENDUM Following submission of this paper, Lagna et al. (16a) showed a physical and functional interaction of Smad4/DPC4 with the other Smads. Their finding that Smad4 physically interacts with Smad1 or Smad2 is in agreement with our data,

2528

WU ET AL.

MOL. CELL. BIOL.

although they show that this interaction is increased following ligand stimulation. REFERENCES 1. Ausubel, F. M., R. Brent, R. E. Kingston, D. D. Moore, J. G. Seidman, J. A. Smith, and K. Struhl (ed.). 1994. Current protocols in molecular biology, p. 113.14.1. John Wiley & Sons, Inc., New York, N.Y. 2. Baker, J. C., and R. M. Harland. 1996. A novel mesoderm inducer, Madr2, functions in the activin signal transduction pathway. Genes Dev. 10:1880– 1889. 3. Chen, R. H., P. J. Miettinen, E. M. Maruoka, L. Choy, and R. Derynck. 1995. A WD-domain protein that is associated with and phosphorylated by the type II TGF-beta receptor. Nature 377:548–552. 4. Derynck, R. 1994. TGF-b receptor-mediated signaling. Trends Biochem. Sci. 19:548–553. 5. Derynck, R. 1994. Transforming growth factor-beta, p. 319–342. In A. W. Thomson (ed.), The cytokine handbook. Academic Press, Inc., San Diego, Calif. 6. Derynck, R., and Y. Zhang. 1996. The Mad way to do it. Curr. Biol. 6:1226– 1229. 7. Ebner, R., R.-H. Chen, L. Shum, S. Lawler, T. F. Zioncheck, A. Lee, A. R. Lopez, and R. Derynck. 1993. Cloning of a type I TGF-b receptor and its effect on TGF-b binding to the type II receptor. Science 260:1344–1348. 8. Feng, X. H., E. H. Filvaroff, and R. Derynck. 1995. Transforming growth factor-beta (TGF-beta)-induced down-regulation of cyclin A expression requires a functional TGF-beta receptor complex. Characterization of chimeric and truncated type I and type II receptors. J. Biol. Chem. 270:24237– 24245. 9. Graff, J. M., A. Bansal, and D. A. Melton. 1996. Xenopus Mad proteins transduce distinct subsets of signals for the TGFb superfamily. Cell 85:479– 487. 10. Gyuris, J., E. Golemis, H. Chertkov, and R. Brent. 1993. Cdi1, a human G1 and S phase protein phosphatase that associates with Cdk2. Cell 75:791–803. 11. Hahn, S. A., M. Schutte, A. T. Hoque, C. A. Moskaluk, L. T. da Costa, E. Rozenblum, C. L. Weinstein, A. Fischer, C. J. Yeo, R. H. Hruban, and S. E. Kern. 1996. DPC4, a candidate tumor suppressor gene at human chromosome 18q21.1. Science 271:350–353. 12. Hogan, B. 1996. Bone morphogenetic protein: multifunctional regulators of vertebrate development. Genes Dev. 10:1580–1594.

13. Hoodless, P. A., T. Haerry, S. Abdollah, M. Stapleton, M. B. O’Connor, L. Attisano, and J. L. Wrana. 1996. MADR1, a MAD-related protein that functions in BMP2 signalling pathways. Cell 85:489–500. 14. Kawabata, M., T. Imamura, K. Miyazono, M. E. Engel, and H. L. Moses. 1995. Interaction of the transforming growth factor-beta type I receptor with farnesyl-protein transferase-alpha. J. Biol. Chem. 270:29628–29631. 15. Keeton, M. R., S. A. Curriden, A. van Zonneveld, and D. J. Loskutoff. 1991. Identification of regulatory sequences in the type 1 plasminogen activator inhibitor gene responsive to transforming growth factor beta. J. Biol. Chem. 266:23048–23052. 16. Kingsley, D. M. 1994. The TGF-b superfamily: new members, new receptors, and new genetic tests of function in different organisms. Genes Dev. 8:133– 146. 16a.Lagna, G., A. Hata, A. Hemmati-Brivanlou, and J. Massague´. 1996. Partnership between DPC4 and SMAD proteins in TGF-b signalling pathways. Nature 383:832–836. 17. Liu, F., A. Hata, J. C. Baker, J. Doody, J. Ca ´rcamo, R. M. Harland, and J. Massague´. 1996. A human Mad protein acting as a BMP-regulated transcriptional activator. Nature 381:620–623. 18. Massague´, J. 1996. TGFb signaling: receptors, transducers, and Mad proteins. Cell 85:947–950. 19. Sekelsky, J. J., S. J. Newfeld, L. A. Raftery, E. H. Chartoff, and W. M. Gelbart. 1995. Genetic characterization and cloning of mothers against dpp, a gene required for decapentaplegic function in Drosophila melanogaster. Genetics 139:1347–1358. 20. Ventura, F., F. Liu, J. Doody, and J. Massague´. 1996. Interaction of transforming growth factor-b receptor I with farnesyl-protein transferase-a in yeast and mammalian cells. J. Biol. Chem. 271:13931–13934. 21. Wang, T., P. K. Donahoe, and A. S. Zervos. 1994. Specific interaction of type I receptors of the TGF-b family with the immunophilin FKBP-12. Science 265:674–676. 22. Wrana, J. L., L. Attisano, R. Wieser, F. Ventura, and J. Massague´. 1994. Mechanism of activation of the TGF-b receptor. Nature 370:341–347. 23. Zervos, A. S., J. Gyuris, and R. Brent. 1993. Mxi1, a protein that specifically interacts with Max to bind Myc-Max recognition sites. Cell 72:223–232. 24. Zhang, Y., X.-H. Feng, R.-Y. Wu, and R. Derynck. 1996. Receptor-associated Mad homologues synergize as effectors of the TGF-beta response. Nature 383:168–172. 25. Zhang, Y., T. Musci, and R. Derynck. The tumor suppressor Smad4/DPC4 as a central mediator of Smad function. Curr. Biol., in press.