Smad4/DPC4 and Smad3 Mediate Transforming Growth Factor-β (TGF-β) Signaling through Direct Binding to a Novel TGF-β-responsive Element in the Human Plasminogen Activator Inhibitor-1 Promoter*

The transforming growth factor-β (TGF-β) family of cytokines mediates multiple biological effects by regulating the expression of target genes. The Smad family proteins function as intracellular signal transducers downstream of the receptors to transmit the TGF-β signal from cell membrane to nucleus. The mechanisms by which Smads mediate transcriptional activation of target genes is largely unknown. Here we report the identification of a novel TGF-β-responsive element in the human type 1 plasminogen activator inhibitor promoter that is required for mediating strong transcriptional activation of this gene by TGF-β. Smad3 and Smad4 are incorporated into a TGF-β-inducible complex formed on this element following TGF-β stimulation of human hepatoma cells. Both Smad3 and Smad4 bind directly to this TGF-β-responsive element through their conserved MH1 domains. These results indicate that Smad3 and Smad4 mediate TGF-β signaling by directly interacting with a specific response element in a physiological target gene.

The transforming growth factor-␤ (TGF-␤) family of cytokines mediates multiple biological effects by regulating the expression of target genes. The Smad family proteins function as intracellular signal transducers downstream of the receptors to transmit the TGF-␤ signal from cell membrane to nucleus. The mechanisms by which Smads mediate transcriptional activation of target genes is largely unknown. Here we report the identification of a novel TGF-␤-responsive element in the human type 1 plasminogen activator inhibitor promoter that is required for mediating strong transcriptional activation of this gene by TGF-␤. Smad3 and Smad4 are incorporated into a TGF-␤-inducible complex formed on this element following TGF-␤ stimulation of human hepatoma cells. Both Smad3 and Smad4 bind directly to this TGF-␤-responsive element through their conserved MH1 domains. These results indicate that Smad3 and Smad4 mediate TGF-␤ signaling by directly interacting with a specific response element in a physiological target gene.
The transforming growth factor-␤ (TGF-␤) 1 family of growth factors regulates diverse biological processes including proliferation, differentiation, development, and extracellular matrix production (1). TGF-␤ signaling is mediated through two types of serine/threonine kinase receptors (2). The highly conserved Smad proteins have been identified as downstream signal transducers (3)(4)(5)(6). Smad2/3 interact directly with the TGF-␤activated receptor complexes and become phosphorylated by the activated type I receptor (7,8). The phosphorylation of Smad2/3 results in the formation of hetero-oligomeric complexes of Smad2/3 and Smad4/DPC4 and their nuclear translocation (3)(4)(5)(6). Studies on the transcriptional activation of Xenopus Mix2 and Drosophila vestigial provided the first clues on the molecular mechanisms of Smad function. In the former case, the association of Smad2 and Smad4 with the activinresponsive element in the Xenopus mix2 promoter is mediated by Xenopus FAST-1 (9). In the latter, Drosophila Mad protein can bind to an enhancer sequence of vestigial directly and mediates its activation by decapentaplegic (10). Recently, mammalian homologues of the Xenopus FAST-1 have been identified that bind to DNA elements adjacent to Smad binding sites and modulate the expression of target genes (11,12). Smad proteins have also been reported recently to be able to bind to specific DNA sequences (13,14).
PAI-1 is the major physiologic regulator of plasminogen activation (15), and its expression is tightly regulated by hormones and cytokines including TGF-␤ (6,16). The TGF-␤ regulation of PAI-1 and other molecules involved in extracellular matrix interactions may play an important role in the control of cell growth, differentiation, tissue remodeling, and wound healing (1). In this report, we show that Smad3 and Smad4, through their MH1 domains, bind directly to a novel TGF-␤responsive sequence (TRS) in the human PAI-1 promoter that is associated with the ability to mediate strong transcriptional activation of this gene by TGF-␤. These results demonstrate that Smad3 and Smad4 function as sequence-specific transcriptional activators in mediating transcriptional activation by TGF-␤.
Transfection and Luciferase Assays-Hep3B cells were transfected with luciferase reporter constructs using FuGENE 6 (Boehringer Mannheim). At 12 h after transfection, the cells were treated with 50 pM TGF-␤ for 24 h, and luciferase activity was measured using the Promega luciferase assay system.
Cell Extracts and Protein Purification-Hep3B whole cell extracts were prepared after stimulation with 100 pM TGF-␤ for 30 min as described (13). Whole cell extracts from COS-1 cells were prepared 48 h after transfection with expression plasmid using LipofectAMINE (Life Technologies, Inc.). GST-Smads were purified as described (19). The concentration and purity of the fusion proteins were determined by SDS-polyacrylamide gel electrophoresis and Coomassie Blue staining using bovine serum albumin as standard.

RESULTS AND DISCUSSION
Previous analyses of the human PAI-1 promoter have mapped TGF-␤-responsive elements to different regions and * This work was supported by Grants CA22729 and DK46010 from the National Institutes of Health. 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 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed: Dept. of Human Genetics, University of Michigan Medical School, Ann Arbor, MI 48109-0618. Tel.: 734-647-3159; Fax: 734-763-3784; E-mail: czsong@ umich.edu. 1 The abbreviations used are: TGF-␤, transforming growth factor-␤; PAI-1, plasminogen activator inhibitor-1; TRS, TGF-␤-responsive sequence; GST, glutathione S-transferase; EMSA, electrophoretic mobility shift assay; bp, base pair(s). implicated different sequence-specific transcription factors in mediating TGF-␤ response in human hepatoma cell lines (20 -22). To better understand how the TGF-␤ signal activates target genes, we first carried out mutational analyses of the PAI-1 promoter. Consistent with the observation of Keeton et al. (20), deletion of sequence between Ϫ740 and Ϫ700 from the PAI-1 promoter significantly reduced TGF-␤ induction from 30-fold to less than 4-fold, indicating that this region contains sequences that are required for optimal induction by TGF-␤ (data not shown). Therefore, this region was further characterized, and it was found that a 12-base pair (bp) sequence from Ϫ732 to Ϫ721 (AGACAAGGTTGT) is capable of conferring TGF-␤ responsiveness to a minimal promoter construct, pE1b-luc. A reporter construct containing four copies of this 12-bp TRS was induced more than 4-fold by TGF-␤, whereas a construct containing six copies of TRS was induced nearly 40-fold by TGF-␤ (Fig. 1). In contrast, the same vector containing the E1b TATA box alone or six copies of sequence from Ϫ720 to Ϫ708 (TGACACAA-GAGAG), containing a TRE-like sequence (20), is unresponsive to TGF-␤ (Fig. 1). Point mutations and double or triple base substitutions in the TRS completely abrogated TGF-␤ responsiveness (Fig. 1). These results demonstrate that the 12-bp TRS is both necessary and sufficient for mediating transcriptional responsiveness to TGF-␤, thus defining this minimal TRS as a critical target for transcriptional activation by TGF-␤.
EMSA revealed the formation of two complexes with TRS ( Fig. 2A, lanes 1-4). Complex II is strongly enhanced by TGF-␤, whereas complex I is relatively unchanged. Point mutations in TRS, which resulted in loss of TGF-␤ responsiveness (Fig. 1), eliminated the formation of both complexes ( Fig. 2A, lanes  5-8). These results indicate that cellular proteins bind specifically to the TRS and that the formation of these complexes correlates with transcriptional activation by TGF-␤.
Since Smad proteins have been identified as mediators of TGF-␤ signaling (3-6), we tested whether they are components of the TRS-binding complex. Supershift EMSA using anti-Smad3 and Smad4 antibodies revealed the presence of Smad3 and Smad4 in the TGF-␤-inducible complex (II) but not in the TGF-␤-independent complex (I) (Fig. 2, B and C). EMSA using cell extracts from COS cells transfected with expression vectors for Myc-Smad4 (8,17) revealed the formation of a distinct TRS-binding complex from extracts prepared from cells transfected with Myc-Smad4 expression vector but not with the empty vector (Fig. 3). The Smad4-dependent complex is supershifted by anti-Myc but not by the control anti-Flag antibody (Fig. 3), confirming the existence of Smad4 in the TRS-protein complex. Competition EMSA showed that the binding of Myc-Smad4 to TRS is specifically competed by wild type but not mutant TRS, demonstrating the specificity of the Smad4-TRS interaction (Fig. 3). Taken together, these results indicate that Smad3 and Smad4 are components of the TRS-binding complex.
We next determined whether Smad3/4 could directly interact with TRS. EMSA using purified GST fusion proteins containing different regions of Smad4 demonstrated that the DNA binding is mediated by the MH1 domain of Smad4 (Fig. 4A). A missense mutation (R100T) in the Smad4 MH1 domain, originally identified in a pancreatic carcinoma (23), has been reported to disrupt both growth inhibition and transcriptional activation by TGF-␤ (24). Arg-100 in Smad4 corresponds to Arg-133 in Smad2 that is also mutated in colon carcinoma (25); the corresponding mutation in Drosophila Mad eliminated its DNA binding activity (10). The R100T mutation completely eliminated TRS binding by Smad4 even when added at 100-fold greater concentration than the wild type protein (Fig. 4A and data not shown). The wild type and mutant GST-Smad proteins were purified to similar purity (data not shown), and the exact same amount of each protein was used in the EMSA. Competition EMSA was performed to further establish the specificity and functional relevance of the Smad4-TRS interaction. Wild type TRS effectively competed the binding of Smad4 (Fig. 4B); however, mutations in TRS that abolished its ability to mediate TGF-␤ responsiveness (Fig. 1) also abolished (m1, m2, m4) or substantially diminished (m3) its ability to compete for binding to Smad4. These results demonstrate that Smad4, either purified in vitro or expressed in vivo, binds specifically to the TRS and that the Smad4-TRS interaction requires the functional integrity of both Smad4 and TRS.
Full-length GST-Smad3 showed no detectable binding to TRS. However, The MH1 domain of Smad3 showed sequencespecific DNA binding activity comparable with that of the MH1 domain of Smad4 (Fig. 4C), suggesting an inhibitory effect of Smad3 MH2 domain on the DNA binding activity of its MH1 domain. In contrast, the MH1 domain of Smad2, which is also a substrate and mediator of TGF-␤ and activin receptors (7), showed no detectable binding to TRS (Fig. 4C). Competition EMSA with wild type and mutant TRS revealed that Smad3 had the same specificity of binding to TRS as Smad4 (data not shown). These observations together with previous structural and functional studies (24, 26 -28) support the notion that in the absence of TGF-␤ stimulation, Smad3 exists in an inactive configuration in which the MH1 and MH2 domains mutually inhibit their DNA binding and effector functions, respectively.
It has been reported that both Smad3 and Smad4 bind to a synthetic octamer sequence termed SBE (Smad binding ele-  (lanes 1, 2, 5, and 6) or 3 l (lanes 3, 4, 7, and 8) of extracts from Hep3B cells treated with (ϩ) or without (Ϫ) TGF-␤ for 30 min. Oligonucleotides containing four copies of the wild type or mutant TRS were used as probes. B, the TGF-␤-inducible complex (shift II) contains Smad3. The cell extracts (3 l) and the wild type probe are the same as those used in A. EMSA reactions contained 4 g (lanes 3 and 7) or 6 g (lanes 4 and 8) of anti-Smad3 antibody or 4 g (lane 5) or 6 g (lane 6) of anti-Flag antibody, respectively. C, the TGF-␤-inducible complex (shift II) contains Smad4. The EMSA reactions are the same as those in B except that 1 g (lanes 3 and 7) or 2 g (lanes 4 and 8) of anti-Smad4 antibody or 1 g (lane 5) or 2 g (lane 6) of anti-Flag antibody were used, respectively. ment), which mediates strong TGF-␤ responsiveness (14). Although the overall sequence of the TRS appears to be quite different from the SBE, the 5Ј-AGAC-3Ј sequence in the TRS is identical to the half-site of the 8-bp palindromic sequence (GTCTAGAC) in SBE. After our manuscript was submitted, Dennler et al. (29) reported the identification of Smad3 and Smad4 binding sequences, termed "CAGA boxes," in the human PAI-1 promoter. The most distal "CAGA box," which showed higher affinity for binding to Smad4 and mediated stronger TGF-␤ induction than the two more proximal elements, partially overlaps with the TRS identified in this report. However, the TRS described here contains no CAGA sequence, since it does not included the 5Ј-cytosine. Of note, both TRS and the CAGA boxes contain the 4-bp AGAC element that is the half-site of SBE and also exists in the Smad4 binding site in the collagenase promoter in the 3TP lux (13). The existence of AGAC sequence in all these Smad binding sites suggests that this element may play an important role in mediating Smad binding and TGF-␤ induction. Indeed, mutations in AGAC (TRS m1 and TRS m2) in TRS abrogated both Smad3 and Smad4 binding and TGF-␤ induction. Although the AGAC in the TRS element is absolutely required, TRS mutants in which the AGAC remains intact (TRS m3 and TRS m5) also failed both to mediate TGF-␤ induction ( Fig. 1) and to compete with wild type TRS for Smad3 and Smad4 binding ( Fig. 4B and data not shown). Therefore, sequences beside AGAC in TRS also play an important role in both Smad3/4 binding and in mediating transcriptional activation by TGF-␤. Further work will be required to determine the exact bases in these TGF-␤-responsive sequences with which Smad3 and Smad4 interact and the contribution of sequences outside AGAC to TGF-␤ induction.