Smad3 Inhibits Transforming Growth Factor-β and Activin Signaling by Competing with Smad4 for FAST-2 Binding*

Transcriptional regulation by transforming growth factor-β and activin is mediated by interaction of Smad2 and Smad3 with specific transcription factors and/or DNA elements. However, Smad3 behaves differently from Smad2 in regulating transcription by a winged-helix transcription factor, FAST-2, on an activin-responsive element (ARE) in the Xenopus Mix.2 promoter. Smad3 alone was able to stimulate the ARE through FAST-2, but inhibited the ARE transactivation mediated by Smad2/Smad4 following receptor activation. We characterized the functional domains that are involved in these two activities of Smad3. Deletion of the MH1 domain as well as mutations of four lysine residues in the MH1 domain abrogated the inhibitory activity of Smad3, but did not compromise the self-stimulatory function. In contrast, deletion of the MH2 domain or a point mutation of glycine 379 within this domain obliterated the self-stimulatory activity of Smad3, but not the inhibitory activity. In an electrophoretic mobility shift assay, we found that Smad3 was able to associate with the FAST-2·ARE complex and that this association was dependent on FAST-2. In addition, Smad3 was not able to directly bind the ARE in a DNase I protection assay, in which FAST-2 binds the ARE around a motif (TGTGTATT) previously characterized to associate with the human FAST-1 protein. Interestingly, Smad4 was also able to directly associate with the FAST-2·ARE complex through binding with FAST-2. In a gel shift assay, the association of FAST-2 with Smad4 was mutually exclusive from the association with Smad3. Taken together, these data indicate that Smad3 exerts the inhibitory activity by competitive association with FAST-2.

Transcriptional regulation by transforming growth factor-␤ and activin is mediated by interaction of Smad2 and Smad3 with specific transcription factors and/or DNA elements. However, Smad3 behaves differently from Smad2 in regulating transcription by a wingedhelix transcription factor, FAST-2, on an activin-responsive element (ARE) in the Xenopus Mix.2 promoter. Smad3 alone was able to stimulate the ARE through FAST-2, but inhibited the ARE transactivation mediated by Smad2/Smad4 following receptor activation. We characterized the functional domains that are involved in these two activities of Smad3. Deletion of the MH1 domain as well as mutations of four lysine residues in the MH1 domain abrogated the inhibitory activity of Smad3, but did not compromise the self-stimulatory function. In contrast, deletion of the MH2 domain or a point mutation of glycine 379 within this domain obliterated the self-stimulatory activity of Smad3, but not the inhibitory activity. In an electrophoretic mobility shift assay, we found that Smad3 was able to associate with the FAST-2⅐ARE complex and that this association was dependent on FAST-2. In addition, Smad3 was not able to directly bind the ARE in a DNase I protection assay, in which FAST-2 binds the ARE around a motif (TGTG-TATT) previously characterized to associate with the human FAST-1 protein. Interestingly, Smad4 was also able to directly associate with the FAST-2⅐ARE complex through binding with FAST-2. In a gel shift assay, the association of FAST-2 with Smad4 was mutually exclusive from the association with Smad3. Taken together, these data indicate that Smad3 exerts the inhibitory activity by competitive association with FAST-2.
Smad protein, when translocated into the nucleus, functions as a transcriptional regulator that controls the expression of target genes (3,19,20). Smad exerts its transcriptional regulatory activity by interacting with either a specific transcription factor or a specific DNA element. In Xenopus, Smad2 and Smad4 participate in the activin-mediated transcriptional induction of the Mix.2 promoter through an interaction with a specific DNA-binding transcription factor, FAST-1 (forkhead activin signal transducer-1), a member of the winged-helix forkhead transcription factor family (21,22). FAST-1 has two functional domains that mediate DNA binding and Smad association, respectively. The DNA-binding motif of FAST-1 mediates the interaction of FAST-1 with an activin-responsive element (ARE) on the Xenopus Mix.2 promoter, and the Smad interaction domain is involved in the association with the Smad2⅐Smad4 complex (22). Smad proteins have been found to functionally interact with other transcription factors, including the mammalian FAST-1 homologues human FAST-1 and mouse FAST-2 (23-25), AP1 (26), Sp1 (27), and TFE3 (28). In addition, recent studies have also suggested that Smad can directly bind DNA. In Drosophila, Mad protein is able to directly bind a GC-rich region of various enhancers (29). A polymerase chain reaction-based screening with random sequences has led to the discovery of specific binding of Smad3 and Smad4 with a palindromic DNA sequence (30). Smad3 and Smad4 were also found to bind a CAGA motif in the promoter of plasminogen activator inhibitor-1 (31).
Once Smad protein associates with either a transcription factor or a DNA element, its C-terminal Mad homology domain-2 (MH2) domain may exert a transactivation function. The C-terminal domain of Smad proteins fused to a heterologous DNA-binding domain was found to induce a transcriptional response (5). This function of the MH2 domain was supported by the recent discovery that Smad may interact with the general transcription coactivators CBP and p300 (32)(33)(34)(35). For example, the adenoviral protein E1A is able to inhibit a TGF-␤-initiated transcriptional response by its interaction with CBP/p300 (33,35). CBP/p300 may bridge the general transcription machinery and Smad proteins as well as the Smad-associated transcription factors, e.g. FAST-1 or FAST-2, and enable the transcriptional regulation of target genes.
In theory, regulation of TGF-␤ and activin signaling may occur at the transcriptional level by modulating the interaction of Smad with a transcription factor and/or a Smad-binding DNA element. Such regulation through DNA binding was recently exemplified by studies with the promoter of goosecoid, which is a homeobox gene required for dorsal-ventral patterning in the frog (36). A mouse FAST-1 homologue, FAST-2, was found to mediate the TGF-␤-initiated transcriptional regulation of the goosecoid promoter (24). It was also found that Smad3 was able to bind a DNA element in the goosecoid promoter and compete with Smad4 for binding. This competition has been proposed to cause Smad3 inhibition of the FAST-2mediated goosecoid promoter transactivation through Smad2/ Smad4 hetero-oligomers following TGF-␤ signaling. We have independently isolated mouse FAST-2 and found that it is able to mediate the transcriptional response downstream of TGF-␤ and activin signaling when using the Xenopus Mix.2 ARE as a transcriptional reporter. Similar to the findings with the goosecoid promoter, we observed an inhibition of the receptormediated (through Smad2 and Smad4) transactivation of the Mix.2 ARE by Smad3. However, we found that Smad3 itself was able to stimulate the promoter in the absence of receptor activation. To further delineate the molecular mechanisms underlying the dual (self-stimulatory and inhibitory) activities of Smad3 in FAST-2-mediated ARE regulation, we studied the structural determinants that mediate these two activities of Smad3. As a result, we discovered a novel means by which Smad3 inhibits the transcriptional regulation induced by Smad2/Smad4 following activation of TGF-␤/activin receptors.

MATERIALS AND METHODS
Cell Culture and Cell Transfection-Human embryonic kidney (HEK) 293 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum supplemented with penicillin and streptomycin. Transient cell transfection was performed by the calcium phosphate method (11).
Plasmid Construction and GST Fusion Proteins-All of the Smad and FAST-2 plasmids used in cell-transfected assays were subcloned into pcDNA3 (Invitrogen) under the control of the cytomegalovirus promoter. Wild-type Smad2, Smad3, and Smad4 and the constitutively active activin type I receptor have been described before (11). The Nand C-terminal truncations of Smad3 were generated by restriction enzyme digestion that resulted in deletion of the first 114 and last 148 amino acid residues, respectively. The Smad3(G/S) mutant was generated by a polymerase chain reaction strategy that specifically changed glycine 379 to serine. The Smad3(K/Q) mutant was generated by polymerase chain reaction that substituted all four lysine residues at positions 40, 41, 43, and 44 with glutamine residues. All GST fusion proteins were generated by in-frame fusion of the full-length protein with pGEX-4T2 (Amersham Pharmacia Biotech). The constructs were transformed into the Escherichia coli BL21 strain (Amersham Pharmacia Biotech), and the GST fusion proteins were purified according to the protocol of Amersham Pharmacia Biotech.
Promoter Assay-For the pAR3-lux promoter assay, ϳ5 ϫ 10 4 cells/ well in a six-well plate were transfected with different combinations of plasmid DNA by the calcium phosphate method. A pCMV-␤-gal plasmid was used as an internal control for the transfection efficiency. The pcDNA3 plasmid was used to make up a total of 5 g of DNA for the transfection in each well. The cells were harvested at 40 h after transfection by lysis with 400 l of Nonidet P-40 lysis buffer (37). Twenty microliters of the lysate was used in the ␤-galactosidase assay as reported before (11), and 10 l of the lysate was used in the luciferase assay with a Promega luciferase assay kit. The luciferase activity was recorded as the averaged relative light unit/s for a duration of 10 s.
Electrophoretic Mobility Shift Assay (EMSA)-The Mix.2 ARE probe was generated by polymerase chain reaction with two primers corresponding to both ends of one ARE (21) and labeled with 32 P by T4 polynucleotide kinase. The probes of ϳ5 ϫ 10 4 cpm were incubated with GST fusion proteins in a buffer containing 4% glycerol, 10 mM Tris (pH 7.5), 1 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 50 mM NaCl, and 0.1 g/l poly(dI-dC). The reaction was incubated at room temperature for 1 h, separated by 4% nondenaturing polyacrylamide gel electrophoresis in 0.5ϫ Tris borate/EDTA, and detected by autoradiography.
DNase I Protection Assay-A DNA fragment containing three ARE repeats and a TATA box was generated by XhoI and KpnI digestion of the pAR3-lux plasmid. This fragment was labeled at one end with [ 32 P]dCTP by Klenow DNA polymerase. The probe was incubated with GST fusion proteins in a buffer containing 10 mM HEPES (pH 7.9), 0.2 mM EDTA, 7% glycerol, 1 mM dithiothreitol, and 1 g of poly(dI-dC). The reaction was then treated with DNase I (0.02-0.2 unit) for 2 min at room temperature. After phenol extraction and ethanol precipitation, the DNA pellet was resuspended in formamide-containing loading buffer, separated by 8% denaturing polyacrylamide gel electrophoresis, and detected by autoradiography. The relative positions of the protected bands on the gel were determined by a G/A reaction using the protocol of Maxam and Gilbert (38).

Characterization of Functional Domains of Smad3
-We recently isolated and characterized mouse FAST-2 that is homologous to Xenopus FAST-1. 2 Using the ARE of the Xenopus Mix.2 promoter as a reporter, we found that FAST-2 was involved in the transcriptional regulation by TGF-␤ and activin signaling pathways. Activation of the TGF-␤ type I receptor (ALK-5) or the activin type I receptor (ALK-4) was capable of inducing a FAST-2-dependent ARE stimulation through Smad2 and Smad4. However, we noticed that Smad3 exerted a different role in regulating the FAST-2-mediated transcriptional regulation of the Mix.2 ARE. Expression of Smad3 alone, but not Smad2 alone, was able to stimulate the promoter at a moderate level in a dose-dependent manner (as the self-stimulatory activity of Smad3). Meanwhile, Smad3 was able to inhibit the Smad2/Smad4-mediated promoter transactivation following activation of the activin type I receptor (as the inhibitory activity of Smad3).
To characterize the structural determinants that mediate these differential activities of Smad3 in the regulation of ARE transactivation, we constructed different mutants of Smad3 and expressed them in HEK293 cells. This cell line was chosen because its high transfectability allows simultaneous expression of multiple proteins. We then determined the properties of these proteins in regulating pAR3-lux, which is composed of a luciferase gene driven by three tandem repeats of the Xenopus Mix.2 ARE and a TATA box (39). Functional studies using both Xenopus FAST-1 and mouse FAST-2 have indicated that this promoter is specifically responsive to both TGF-␤ and activin signaling (39,40). As shown in Fig. 1, expression of wild-type Smad3, but not wild-type Smad2, in these cells moderately stimulated pAR3-lux (ϳ15-fold) in the presence of FAST-2. Coexpression of Smad2 and Smad4 with a constitutively active activin type I receptor (CA-ALK-4) bearing an activating mutation of threonine 206 to aspartic acid led to a strong stimulation of the promoter (ϳ76-fold). Smad3 could inhibit this receptor/Smad2/Smad4-mediated transactivation and reduce the -fold induction to a level that was comparable to that induced by Smad3 alone. We then determined the effect of either the C-terminal (MH2 domain) deletion of Smad3 or a point mutation of Smad3 with a mutation of glycine 379 to serine (Smad3(G/S) mutation) on ARE transcription. The mutation of glycine 379 to serine was originally discovered in Drosophila Mad, in which it led to a compromised development characteristic of a defective decapentaplegic pathway (41). This glycine residue is conserved in all Smad proteins identified so far and is located in the Smad L3 loop, which has been proposed to be involved in protein-protein interaction (42). Interestingly, we found that both the C-terminal deletion of Smad3 and the Smad3(G/S) mutant lost their ability to stimulate the promoter by themselves, indicating that the MH2 domain and the conserved glycine residue are required for the self-stimulatory activity of Smad3. However, both of the mutants appeared to reserve the ability to inhibit the receptor/Smad2/Smad4-mediated ARE stimulation. Both of these mutants were able to significantly reduce the promoter induction by the receptor/ Smad2/Smad4 (Fig. 1), suggesting that the MH2 domain and the conserved glycine residue are not involved in the inhibitory activity of Smad3.
We then determined the ability of two other mutants of Smad3 to transcriptionally regulate the ARE. When the Nterminal (MH1) deletion mutant of Smad3 was expressed in the cells, it led to a strong stimulation of pAR3-lux by itself, almost reaching the level achieved by expression of CA-ALK-4 with Smad2 and Smad4. This super-stimulatory effect of the MH1 deletion mutant of Smad3 could be explained by the hypothesis that the MH1 domain of Smad proteins may serve as an inhibitory motif that blocks the transactivating activity of the MH2 domain (43). We also investigated another Smad3 mutant that bears substitution of four lysine residues at positions 40, 41, 43, and 44 with glutamine within the MH1 domain (Smad3(K/Q) mutant). These four lysines are located at the helix H2 region in the MH1 domain (44). This mutant was originally generated to test the hypothesis that the four lysine residues may be involved in nuclear translocation of Smad3. However, we did not detect a change in nuclear translocation of this mutant (data not shown). Interestingly, the Smad3(K/Q) mutant was still able to stimulate the ARE by itself (Fig. 1), similar to wild-type Smad3. These data therefore indicate that neither the MH1 deletion nor the Smad3(K/Q) mutant was defective in its self-stimulatory activities. In contrast, both of these mutants appeared to have a loss of inhibitory activity. Expression of either the N-terminal deletion of Smad3 or the Smad3(K/Q) mutant did not significantly inhibit the promoter transactivation induced by the receptor/Smad2/Smad4, indicating that the MH1 domain as well as the lysine residues in this domain are involved in the inhibitory activity of Smad3.
Association of Smad3 with the FAST-2⅐ARE Complex-Recent studies on the transcriptional regulation by Smad protein have indicated two possible mechanisms by which Smad affects the promoter activity of a target gene: interaction of Smad with a transcription factor or direct binding to a specific DNA motif. Smad2 and Smad4 have been shown to associate with Xenopus FAST-1 upon activin or TGF-␤ stimulation (21,22). In addition, Smad3 and Smad4 have been shown to bind a consensus palindromic DNA sequence (30). The direct binding of Smad3 with particular DNA sequences has also been identified during the transcriptional regulation of plasminogen-activator inhibitor-1 and JunB promoters (31,45). To determine if Smad3 exerts its unique activities in Mix.2 ARE regulation by a direct interaction with FAST-2 and/or a DNA element, we used an EMSA to determine the interaction of the purified GST fusion proteins with the Mix.2 ARE. We have previously detected a direct association of FAST-2 with the ARE by the same assay. 2 To determine if Smad3 was able to associate with the FAST-2⅐ARE complex, we included GST-Smad3 in the gel shift assay together with FAST-2 and the ARE probe. As shown in Fig. 2A, FAST-2 was able to associate with the ARE as indicated by the shift of the probe. Smad3 led to a further shift of the FAST-2⅐ARE complex, indicating an association of Smad3 with the complex. This association was further confirmed by the dosedependent supershift using an anti-Smad3 antibody.
Smad3 Does Not Directly Bind the ARE-To determine if Smad3 directly bound the ARE or needed FAST-2 for the association, we studied the ability of Smad3 to shift the ARE probe in the presence or absence of FAST-2. As shown in Fig.  2B, Smad3 was not able to associate with the probe by itself. However, inclusion of FAST-2 could lead to a dose-dependent association of Smad3 with the FAST-2⅐ARE complex. These data indicate that the Smad3 association with the ARE is dependent on FAST-2 and that Smad3 itself is not able to directly associate with the ARE. This is consistent with the observation that the consensus Smad3-binding motif is absent in the Mix.2 ARE. In addition, we also found that the MH2 domain of Smad3 is required for the association of Smad3 with FAST-2. As shown in Fig. 3, both GST protein alone and GST-Smad3 with C-the terminal deletion did not associate with the FAST-2⅐ARE complex in a gel shift assay. This is consistent with the report demonstrating that the MH2 domain of Smad is involved in the interaction with Xenopus FAST-1 (22).
To further determine the DNA sequence in the Mix.2 ARE that confers the binding of FAST-2 as well as the FAST-2⅐Smad3 complex, we performed a DNase I protection assay with the probe generated from the three tandem ARE repeats of the pAR3-lux plasmid that was used in the cell transfection assay. As shown in Fig. 4, Smad3 alone could not protect any region of the probe, consistent with our previous experiment in Fig. 2B revealing the inability of Smad3 to directly bind the ARE. As expected, FAST-2 was able to protect one region in each of the ARE repeats. FAST-2 was able to protect a region of 16 base pairs centered by a consensus FAST-binding motif (TGTGTATT) as determined by human FAST-1 (23). In addition, Smad3 was not able to introduce further protection beyond those sequences protected by FAST-2. Therefore, these data not only characterize the FAST-2-binding motif in the Mix.2 ARE, but also confirm that Smad3 is not able to directly associate with the ARE, a phenomenon that is different from the finding with the goosecoid promoter (24).
Smad3 Does Not Inhibit Receptor Signaling by Sequestering Smad4 or CBP Pools-One of the possibilities that may underlie the inhibitory activity of Smad3 is that Smad3, when overexpressed, may sequester Smad4 and make the Smad4 pool insufficient for binding receptor-activated Smad2, resulting in a reduced transcriptional response. To investigate this hypoth-esis, we asked if overexpression of Smad4 could abrogate the inhibitory activity of Smad3 or the Smad3(G/S) mutant in the receptor/Smad2-mediated ARE transactivation. We reasoned that if the Smad3-mediated inhibition is caused by the sequestration of Smad4, overexpression of Smad4 should provide enough partners for Smad2 and detour the Smad3 inhibition. As shown in Fig. 5, expression of CA-ALK-4 with Smad2 increased ARE stimulation by ϳ40-fold, possibly mediated by endogenous Smad4. Cotransfection of Smad4 with the receptor and Smad2 stimulated the promoter by ϳ115-fold. As expected, expression of either Smad3 or the Smad3(G/S) mutant could significantly inhibit the transcriptional response induced by the receptor/Smad2/Smad4 cotransfection. However, expression of Smad4 at different amounts was not able to abrogate the inhibitory activity of both of those Smad3 proteins, even when Smad4 was expressed 7-fold more than Smad3 or the Smad3(G/S) mutant. These data do not support the hypothesis that Smad3 inhibits the TGF-␤/activin-triggered transcriptional response by sequestering Smad4.
An alternative possibility underlying the inhibitory activity of Smad3 is that overexpression of Smad3 may take up or sequester CBP/p300, which is required for the transcriptional regulation by Smad proteins (33,35). To address this issue, we determined if overexpression of CBP is able to rescue the Smad3 inhibition of the ALK-4/Smad2/Smad4-mediated transcriptional response. As shown in Fig. 6, expression of Smad3 was able to stimulate AR3-lux activity, and this effect was enhanced by coexpression of CBP, consistent with the observation that CBP/p300 is involved in the transcriptional response  . 4. Determination of FAST-2-binding sequence by DNase I footprinting assay. The Mix.2 ARE was protected by FAST-2, but not Smad3. The whole promoter region of the pAR3-lux plasmid was labeled with 32 P. This probe was incubated with GST, GST-Smad3, or GST-FAST-2 and treated with DNase I. The reaction was separated by denaturing polyacrylamide gel electrophoresis and detected by autoradiography. The relative positions of the bands on the gel were determined by a G/A reaction (not shown) using the protocol of Maxam and Gilbert (38). The protected regions are labeled by three hatched boxes on the right. The full sequence of one ARE is shown here, and the consensus sequence that was identified to bind human FAST-1 is underlined.
by Smad proteins. In the same experiment, both Smad3 and the Smad3(G/S) mutant were able to inhibit the CA-ALK-4/ Smad2/Smad4-induced AR3-lux activation. However, overexpression of CBP did not seem to be able to abrogate this inhibitory effect imposed by either wild-type Smad3 or the Smad3(G/S) mutant. These data indicate that the inhibitory activity of Smad3 in FAST-2-mediated Mix.2 ARE transactivation by the activin receptor is unlikely a result of sequestration of CBP by Smad3.
Association of Smad4 with FAST-2-After excluding the possibility that Smad3 inhibits receptor signaling by taking up the available Smad4 pool, we turned to another hypothesis: Smad3 may compete with receptor-activated Smad2/Smad4 for the association with the FAST-2⅐ARE complex. To address this hypothesis, we generated a GST-Smad4 fusion protein and determined if Smad4 was able to directly associate with FAST-2 in the gel shift assay. As shown in Fig. 7A, Smad4 itself was not able to associate with the ARE probe. Similar to Smad3, Smad4 did associate with the FAST-2⅐ARE complex in the presence of FAST-2. Interestingly, we observed three retarded protein complexes when Smad4 was used in this assay, suggesting that Smad4 might bind FAST-2 as a monomer, dimer, or trimer. It is noteworthy that Smad3 led to only one shifted complex (Figs. 2 and 7A), indicating that Smad3 may not be able to self-assemble to form multi-oligomers under our experimental conditions.
Association of Smad4 and Smad3 with the FAST-2⅐ARE Complex Is Mutually Exclusive-We next addressed the question of whether Smad4 and Smad3 could bind the FAST-2⅐ARE complex concomitantly or mutually exclusively. The concomitant association of Smad3 and Smad4 with FAST-2 should lead to a supershift of the three retarded Smad4⅐FAST-2⅐ARE complexes in the gel shift assay because inclusion of Smad3 would add molecular weight to the complexes. As shown in Fig. 7A, we did not observe any change in the mobility shift with the three Smad4-retarded complexes, even though Smad3 itself was able to associate with the FAST-2⅐ARE complex. These data indicate that Smad3 is unlikely to coexist with Smad4 through binding at different sites of FAST-2. However, this experiment could not exclude the possibility that Smad3 may form a dimer or trimer with Smad4, and the Smad3/Smad4 heterodimers or heterotrimers may be retarded at similar positions on the gel as the Smad4 homodimers or homotrimers. To address this possibility, we used the anti-Smad3 antibody in the assay. This antibody should supershift the heterodimers or heterotrimers formed between Smad3 and Smad4, but not the Smad4 homodimers or homotrimers. As shown in Fig. 7B, the antibody was able to supershift the Smad3⅐FAST-2⅐ARE complex (compare lane 5 with lane 4). However, it had no effect on the behavior of the putative dimeric or trimeric complexes (compare lane 7 with lane 6). Taken together, our data suggest a mutually exclusive nature of the association of Smad3 and Smad4 with FAST-2. Furthermore, if our hypothesis were right, the inhibitory activity of Smad3 would be correlated with its affinity for FAST-2. To test this idea, we determined the FAST-2 binding ability of the Smad3(K/Q) and Smad3(G/S) mutants. In the gel shift assay, the GST-Smad3(K/Q) fusion protein had a largely reduced ability to associate with the FAST-2⅐ARE complex as compared with wild-type Smad3 (Fig. 8), consistent with its loss of inhibitory activity in the activin receptor-mediated transcriptional response. On the other hand, the Smad3(G/S) mutant was able to weakly bind the FAST-2⅐ARE complex, partly correlating with its activity to inhibit the Smad2/Smad4-mediated activin signaling. In addition, we also found that fulllength Smad2 was not able to associate with the FAST-2⅐ARE complex. Taken together, these data provide further evidence supporting the competition theory for the transcriptional inhibitory function of Smad3.

DISCUSSION
Regulation of the TGF-␤ and activin signaling pathways may occur at different levels. First, they can be regulated by extracellular factors. For example, follistatin is able to directly bind and inactivate activin (46). Inhibin may block activin signaling by competing for the same type II receptor on the cell surface (47). In addition, the local activity of TGF-␤ can be regulated by a latency-associated protein that is recognized by specific integrin molecules on the plasma membrane (48). Modulation of the TGF-␤ family receptors through the inhibitory Smad proteins constitutes another level of intracellular regulation. Both Smad6 and Smad7 have been found to antagonize BMP and TGF-␤ signaling by directly associating with the receptor and/or sequestering pathway-specific Smad protein (13)(14)(15)(16)(17)(18). Likewise, the regulation may occur at the transcriptional level. Smad proteins regulate gene transcription either by association with a specific transcription factor or by binding with a specific DNA element. The regulation of TGF-␤/activin signaling at the transcriptional level may happen at these two steps, i.e. by regulating Smad binding to the transcription factor or to the DNA element. Regulation though DNA binding has been found with the goosecoid promoter (24). It has been proposed that Smad3 inhibits the transactivation of this promoter by competing with receptor-activated Smad4/Smad2 for binding to the same DNA motif that is located in a region distinct from the FAST-2-binding element (24). Our experiments indicated another means of regulation at the transcriptional level by a competition for binding to the same transcription factor, FAST-2, in the Mix.2 ARE. In contrast to the observation in the goosecoid promoter, Smad3 associated with the Mix.2 ARE only through its interaction with FAST-2, not directly binding the promoter DNA. Smad4 was also able to associate with the ARE through FAST-2. It appeared that Smad3 and Smad4 could not concomitantly associate with FAST-2, but rather mutually exclusively. It is likely that this mutually exclusive or competitive binding of Smad3 and Smad4 with FAST-2 may partly explain the ability of Smad3 to inhibit the receptor-activated transactivation through Smad2/Smad4. Alternatively, the inhibitory activity of Smad3 may be due to its relatively higher affinity for FAST-2 than the affinity of Smad4 for this transcription factor, as Smad4 overexpression could not abrogate the inhibitory effect of Smad3 (Fig. 5). However, these hypotheses remain to be elucidated in the future.
Our experiments also indicated that the dual activities of Smad3 in regulating the FAST-2-mediated Mix.2 ARE transcription are contributed by the MH1 and MH2 domains, respectively. Deletion of the MH1 domain or point mutations of the four lysine residues inside this domain (Smad3(K/Q) mutant) led to an abrogation of the inhibitory activity, but not the self-stimulatory activity, of Smad3. Meanwhile, deletion of the MH2 domain as well as the point mutation of glycine 379 abrogated the ability of Smad3 to stimulate the Mix.2 ARE by itself, but did not compromise the inhibitory activity of Smad3. These data indicate that the MH1 domain of Smad3 as well as the lysine residues are crucial for the inhibitory activity of Smad3. Because the inhibitory activity of Smad3 may be caused by competition with Smad4 for binding to FAST-2, the inability of the MH1 deletion mutant or the lysine mutant of Smad3 to inhibit Smad2/Smad4 signaling is likely a result of a compromised affinity for FAST-2. This hypothesis was supported by our finding that the Smad3(K/Q) mutant had a largely reduced association with FAST-2 by the gel shift assay. In addition, the differential contributions of the MH1 and MH2 domains to the dual regulatory roles of Smad3 may be caused by the possibility that the two domains of Smad3 might interact differently with FAST-2 in the "quiescent" (unphosphorylated) or receptor-activated (phosphorylated) state.
Our data suggest that the MH2 region of Smad3 and the conserved glycine residue are required for the self-stimulatory activity of Smad3. This stimulatory activity is correlated with the transactivating activity of the Smad MH2 domains that have been shown to stimulate gene transcription when fused with the GAL4 DNA-binding domain (5). Deletion of the MH2 domain of Smad3 would lead to a loss of this transactivating activity. The phenotype of the Smad3 mutant that changes the conserved glycine residue is very perplexing. This glycine residue is located at the L3 loop, which has been proposed to be involved in protein-protein interaction (42). A recent study has indicated that this loop is involved in the interaction with the type I receptors of the TGF-␤ superfamily and that the specificity of the interaction is determined by the amino acid residues inside the loop (49). If this glycine is involved only in receptor interaction, the Smad3(G/S) mutant should not be compromised in its self-stimulatory activity. However, our data suggest that this mutation may also affect the transactivating activity of Smad3. In addition, the transactivating activity of the Smad MH2 domain is related to its association with the general transcription coactivators CBP and p300, which have endogenous histone acetyltransferase activities. The MH2 domain of Smad proteins has been shown to associate with the C-terminal domain of CBP/p300, and this association is blocked by adenoviral protein E1A (32)(33)(34)(35). Therefore, it is likely that the self-stimulatory activity of Smad3 is related to the interaction at its MH2 domain with these transcription coactivators. It is also of importance to determine in the future how this glycine 379 to serine mutation is involved in the interaction with CBP/p300.
Based on the studies of Smad3 in both goosecoid and Mix.2 promoters as reported here, Smad3 appears to have a function distinct from Smad2. Smad2 functions as a direct signaling mediator that receives the signals from the activated receptor through association with Smad4. Smad3 may perform multiple functions. First, Smad3 itself may mediate the signal from the receptor activation that leads to Smad3 phosphorylation and association with Smad4, identical to the activity of Smad2. Second, Smad3 may stimulate the promoter independently of receptor activation by a direct interaction with a transcription factor, e.g. FAST-2, through its MH1 domain. Once Smad3 binds the transcription factor, it stimulates gene transcription through its transactivating activity at its MH2 domain. Third, Smad3 may inhibit the receptor-activated Smad2/Smad4 signaling by competing with Smad4 for binding to either a transcription factor, e.g. FAST-2, or a DNA motif, e.g. the goosecoid promoter. This complicated behavior of Smad3 may be used by a cell to determine the basal tone of the promoter activity as well as the magnitude of the transcriptional response that the receptor activation could trigger. For example, if a cell has a relatively high concentration of Smad3 in comparison with Smad2, the basal transcriptional activity of the TGF-␤-and activin-regulated promoters may be higher than in a cell with a lower Smad3 concentration. Likewise, TGF-␤ or activin may be able to initiate a stronger transcriptional response in cells with a lower concentration of Smad3 than in those with a higher Smad3 level. However, because our finding on the selfstimulatory activity of Smad3 is based on an ectopic expression system, future experiments are required to establish if the basal transcriptional activity and the TGF-␤/activin response are affected by the relative protein expression level of endogenous Smad2 and Smad3.
Recent studies using mouse knockout technology have provided a hint as to the functional difference between Smad2 and Smad3. Whereas the Smad2-deleted mouse dies at a very early developmental stage with defective egg cylinder elongation and mesoderm induction (50 -52), the Smad3 knockout mouse grows to adulthood, but develops multiple metastasizing colon cancers (53). These studies indicate that Smad2, but not Smad3, is indispensable for the early development of the mouse. They also suggest that the tumor suppressor activity of Smad3 could not be replaced by Smad2 during the colon cancer formation in the mouse (53). With the lack of detailed information on the temporal and spatial distribution of Smad2 and Smad3, it is still hard to judge at present if the functional difference between Smad2 and Smad3 reflects only the differential distribution of these two proteins or the actual functional diversity of them. However, these mouse studies, together with the work at the molecular level, definitely justify further experiments to dissect the functions of these two very similar Smad proteins in mediating and regulating the signals downstream of TGF-␤ and activin that exert a large variety of biological activities in the body.