Smad3 and Smad4 mediate transcriptional activation of the human Smad7 promoter by transforming growth factor beta.

Smad7 is an inducible intracellular inhibitor of transforming growth factor-beta (TGF-beta) signaling that is regulated by diverse stimuli including members of the TGF-beta superfamily. To define the molecular mechanisms of negative control of TGF-beta signaling, we have isolated the human SMAD7 gene and characterized its promoter region. A -303 to +672 SMAD7 region contained a palindromic GTCTAGAC Smad binding element (SBE) between nucleotides -179 and -172 that was necessary for the induction of a Smad7 promoter luciferase reporter gene by TGF-beta. Electrophoretic mobility shift assays using oligonucleotide probes demonstrated that TGF-beta rapidly induced the binding of an endogenous SBE-binding complex (SBC) containing Smad2, Smad3, and Smad4. Transfection assays in mouse embryonic fibroblasts (MEFs), with targeted deletions of either Smad2 or Smad3, and the Smad4-deficient cell line MD-MBA-468 revealed that both Smad3 and Smad4, but not Smad2, were absolutely required for induction of the Smad7 promoter reporter gene by TGF-beta. Furthermore, the TGF-beta-inducible SBE-binding complex was diminished in Smad2-deficient MEFs when compared with wild type MEFs and not detectable in Smad3-deficient MEFs and MD-MBA-468 cells. Taken together, our data demonstrate that TGF-beta induces transcription of the human SMAD7 gene through activation of Smad3 and Smad4 transcription factor binding to its proximal promoter.

Transforming growth factor-␤ (TGF-␤) 1 is the prototype of a cytokine superfamily with important roles in cell cycle control, differentiation, and apoptosis. TGF-␤ initiates signaling through the ligand-dependent activation of a complex of heteromeric transmembrane serine/threonine kinases, consisting of type I and type II receptors (1,2). Upon activation, type I receptor associates with and activates Smad2 and/or Smad3, two signaling mediators of the SMAD protein family (3)(4)(5)(6). Activated Smad2 and/or Smad3 associate with the shared partner Smad4 and translocate to the nucleus, where Smad protein complexes participate in transcriptional activation of target genes (7)(8)(9).
The TGF-␤/Smad signaling system is notable for an autoinhibitory feedback loop which involves Smad7, a structurally and functionally divergent Smad protein of the subfamily of "inhibitory Smads" (10 -12). Smad7 interacts stable with ligand-activated type 1 receptor and interferes with receptor binding and phosphorylation of substrate Smads (10). Thus, Smad7 may have an essential role in the regulation of the TGF-␤/Smad signaling system by controlling the accessibility of ligand-activated type 1 receptor for substrate Smad2 and/or Smad3. Several reports indicate that Smad7 expression is strongly and rapidly induced by TGF-␤ itself (12,13) by the Jak1/Stat1 pathway following stimulation with IFN-␥ (14), by activated NF-B, 2 and by fluid shear stress acting on endothelial cells (11). Together, these observations point to a broad role for Smad7 in trans-modulation of signaling pathways.
Because of the potentially central roles of Smad7 as an effector in an autoregulatory feedback loop in TGF-␤/Smad signaling and as a mediator of inhibitory signaling cross-talk between opposing pathways and the TGF-␤/Smad pathway, we reasoned that knowledge of the molecular mechanisms that control the expression of Smad7 would advance the understanding of the regulation of the TGF-␤/Smad pathway. Here we report a molecular mechanism by which TGF-␤ induces transcription of the human Smad7 promoter. We have identified a palindromic Smad binding element that binds a protein complex containing Smad2, Smad3, and Smad4 and shown that is necessary for the transcriptional activation of the Smad7 promoter by TGF-␤. In cells that lack Smad3 or Smad4, TGF-␤ is unable to induce Smad7 promoter activity.

MATERIALS AND METHODS
Cell Culture and RNA Analysis-A spontaneously immortalized human keratinocyte cell line (HaCaT), and SV40-transformed mouse mesangial cells were obtained from Dr. Norbert Fusenig and Dr. Fuad Ziyadeh, respectively. NIH3T3 murine fibroblasts and the Smad4-deficient human mammary adenocarcinoma cell line MDA-MB 468 were obtained from the American Type Culture Collection (ATCC) (Manassas, VA). Wild type and Smad2-deficient (Smad2 dex2/dex2 ) mouse embryonic fibroblasts were derived from d10.5 embryos as described (15). Wild type and Smad3-deficient MEFs were derived from day 12.5 embryos (16). All cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and antibiotics. Recombinant human TGF-␤1, TGF-␤2, and TGF-␤3 were purchased from R & D Systems, and interferon-␥ (IFN-␥) was obtained from Genzyme. Recombinant murine TNF-␣ was from Roche Molecular Biochemicals, and human epidermal growth factor (EGF) was obtained from Promega. Actinomy-* This work was supported by American Heart Association Grant-inaid 9950349N and by National Institutes of Health Grant DK-56077-01 (to E. P. B.). 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  cin D and cycloheximide were purchased from Sigma and used in concentrations recommended by the supplier. RNA was isolated using Trizol Reagent (Life Technologies, Inc.) following the manufacturer's protocol. For Northern blot analysis, RNA was electrophoresed on 1% agarose gels and transferred to a filter. Filters were then hybridized in QuickHyb solution (Stratagene) with 32 P-labeled cDNA probes for murine Smad7 and analyzed by phosphor imagery.
Isolation of the Human Smad7 Promoter-A down-to-the-well human genomic PAC library screening system (Genome Systems) was screened with a polymerase chain reaction probe generated by a primer pair (primers E5 (5Ј-GCCTCCGGGAGACTGG) and E6 (5Ј-GAGAAAA-GTCGTTGGCCTG)) located in the 5Ј-untranslated region of human Smad7 cDNA to give clones 806N1 and 529P14. DNA prepared from both PAC clones was digested with restriction endonucleases (BamHI, EcoRI) and electrophoresed on 1% agarose gel in 1ϫ Tris/acetic acid/ ethylenediaminetetraacetate (TAE). DNA fragments were transferred to Hybond N ϩ membrane (Amersham Pharmacia Biotech) and hybridized with the 170-bp polymerase chain reaction probe. A 4.6-kilobase pair EcoRI fragment was identified, gel-purified, and ligated with pBluescript KSϩ/Ϫ vector DNA (Stratagene). This fragment contained mostly 5Ј-flanking sequence of the SMAD7 gene and was used for promoter analyses.
Deletion Constructs, Transfections and Transcriptional Reporter Assays-5Ј-and 3Ј-deletions were generated by endonuclease digestions from the isolated EcoRI fragment. Seven distinct fragments were ligated into the promoterless luciferase reporter vector pGL3-basic (Promega) (see Fig. 2A). For transcriptional reporter assays, cells (2.5-6 ϫ 10 4 /well) were seeded in 24-well or six-well dishes and transfected with the indicated luciferase reporter constructs and pRSV-Gal (Promega), using the Superfect Reagent (Qiagen) according to the manufacturer's protocol. Transfected cells were incubated in 0.2% fetal bovine serum starvation medium for 20 h and then either left untreated or treated with TGF-␤1 (1 ng/ml) for 4 h. Luciferase and galactosidase activities in transfected cells were determined using assay kits from Promega. Luciferase activities was measured using an AutoLumat LB953 (EG & G Berthold) luminometer. Galactosidase activities were measured with a Labsystems Multiscan MCC/340 plate reader at 405 nm. To correct for differences in transfection efficiencies, luciferase units were normalized for galactosidase activities in the same cell lysate. Corrected luciferase units were then expressed as ratio (-fold induction) compared with the luciferase readings mediated by the empty vector pGL3 basic in the same experiment. Experiments were performed in triplicate.
Primer Extension Analysis-An oligonucleotide ( ϩ98 CACGCGGCTC-GTCGTTCGCTCACAC ϩ64 ) complementary to the human SMAD7 DNA was labeled with [␥-32 P]ATP (Amersham Pharmacia Biotech), hybridized to human kidney mRNA (CLONTECH) and reverse transcribed into cDNA using the avian myeloblastosis virus Reverse Transcriptase System (Promega) following the manufacturer's protocol. Sequencing of genomic SMAD7 DNA contained in the pS7-5 plasmid was performed using a Sequenase version 2.0 sequencing kit (U.S. Biochemical Corp.) following the manufacturer's protocol. The sequencing primer was GT-GCGCCGAGCAGCAAGCGAG. The radiolabeled cDNA primer extension products were analyzed in parallel with the sequencing reactions using an 8 M urea denaturing polyacrylamide gel.
Site-directed Mutagenesis-Site-directed mutagenesis was carried out in the pS7-5 construct using a QuickChange kit (Stratagene) following the manufacturer's instructions. Thymidine at position Ϫ176 and adenine at position Ϫ175 in the center of the Smad7 Smad binding element were replaced with adenine and thymidine, respectively (lowercase italic type), using complementary oligonucleotides (mSBEfw, Ϫ185 CAGGGTGTCatGACGGCCAC Ϫ166 ; mSBErv, Ϫ166 GTGGCCGTCat-GACACCCTG Ϫ185 ) to generate the mutant construct pS7-5mSBE. Sequence fidelity was confirmed by sequencing.
Preparation of Nuclear Protein Extracts and Electrophoretic Mobility Shift Assays-Nuclear protein extracts were prepared from subconfluent cell cultures on 100-mm dishes. Cells were washed twice in cold phosphate-buffered saline and lysed in 1 ml of ice-cold hypotonic lysis buffer (10 mM Hepes, pH 7.9, 10 mM KCl, 0.1 mM EDTA, pH 8.0, 0.1 mM EGTA, 1 mM dithiothreitol, 0.6% Nonidet P-40 containing AEBSF, leupeptin, aprotinin, pepstatin A, antipain, sodium vanadate, sodium fluoride, and okadaic acid at concentrations recommended by the manufacturers). The cells were allowed to swell for 15 min and then scraped, collected, and washed with hypotonic lysis buffer without detergent. Nuclei were pelleted by centrifugation at 13,000 rpm for 20 s in a microcentrifuge and resuspended in 20 l of nuclear extraction buffer (lysis buffer with 20 mM Hepes, pH 7.9, and 420 mM NaCl). Nuclear lysates were incubated for 20 min on a shaker and cleared of debris by centrifugation.
Electrophoretic mobility shift assays were performed as described previously (17), using nuclear extracts prepared from either untreated cells or cells treated with TGF-␤1 (1 ng/ml) for 1 h. Complementary oligonucleotides "S7SBE" containing the Smad binding element (SBE) sequence were ␥-32 P-end-labeled by T4 polynucleotide kinase reaction and annealed. S7SBE probe (50,000 cpm) was incubated with 1 g of nuclear extract in binding buffer (16% glycerol, 20 mM Hepes, pH 7.9, 0.1 mM EDTA, 30 mM KCl, 3 g of poly(dI-dC), 0.8 mM NaP i , pH 7.8, 4 mM spermidine, 4 mM MgCl 2 ) with or without preincubation for 10 min with a 50-or 100-fold molar excess of cold annealed competitor at 4°C for 30 min. For antibody interference studies (supershift analysis), nuclear extracts were incubated overnight at 4°C with 2 g of the following antibodies prior to or following the addition of radiolabeled probe as indicated: mouse monoclonal anti-Smad2 (S 66220; Transduction Laboratories) or goat polyclonal anti-Smad2 (sc-6200 X), anti-Smad3 (sc-6202 X), and anti-Smad4 (sc-1909 X) (all from Santa Cruz Biotechnology, Inc., Santa Cruz, CA). DNA-binding protein complexes were separated by nondenaturing 4% polyacrylamide gel electrophoresis at 4°C and visualized by autoradiography.

TGF-␤ Regulates SMAD7 by Transcriptional Activation-
Smad7 is a member of the Smad protein family that has been shown to antagonize TGF-␤ receptor signaling. Several TGF-␤ family members including activin and BMP7 were found to increase Smad7 expression and induce its interaction with activated TGF-␤ type I receptor (10,12,13). Thus, it has been proposed that Smad7 mediates an autoregulatory negative feedback loop in TGF-␤ signaling (18). To determine whether the regulation of Smad7 by TGF-␤ is mediated at the level of gene transcription, we examined Smad7 mRNA levels in response to TGF-␤ in the absence or presence of actinomycin D, an inhibitor of transcription (Fig. 1). TGF-␤-mediated up-regulation of Smad7 mRNA was completely blocked by pretreatment of cells with actinomycin D (Fig. 1, lanes 3 and 4), indi- cating that TGF-␤ activates transcription of the SMAD7 gene. Pretreatment with cycloheximide had no effect on the transcriptional activation of SMAD7 by TGF-␤, suggesting that de novo protein synthesis was not required for this activity (data not shown).
Characterization of the Human Smad7 Promoter-We isolated the human SMAD7 gene by screening a human P1-artificial chromosome library with a 170-bp polymerase chain reaction probe corresponding to 5Ј-untranslated sequence of the human Smad7 cDNA. 3 An EcoRI-SphI genomic fragment spanning 4.6 kilobase pairs of 5Ј-flanking sequence of SMAD7 was isolated and subcloned into the promoterless luciferase reporter vector pGL3basic. In order to define the basal and TGF-␤-inducible Smad7 promoter elements in this region, we generated a total of seven 5Ј-and 3Ј deletion constructs (pS7-1 to pS7-7) using convenient restriction sites ( Fig. 2A). Transfections of these constructs into NIH3T3 fibroblasts revealed that the constructs pS7-1 to pS7-5 mediated both TGF-␤-inducible and basal promoter activity (Fig. 2B). Further 5Ј deletion of the Smad7 promoter (pS7-6) between a KpnI (Ϫ303) and a BssHII (Ϫ146) site resulted in a complete loss of TGF-␤-inducibility of the Smad7 promoter without affecting basal promoter activity (Fig. 2B). In contrast, the region between a HindIII and a SphI site (pS7-7) mediated luciferase activities that were not different from empty control vector, indicating that the basal Smad7 promoter region was located upstream of this fragment (Fig. 2B).
To examine whether the Ϫ303 to ϩ672 SMAD7 fragment was TGF-␤-inducible in different cell types, we transfected the pS7-5 plasmid into murine mesangial cells and HaCaT cells in addition to NIH3T3 fibroblasts. The pS7-5 plasmid gave rise to comparable basal and TGF-␤-inducible luciferase activities in all three cell lines, indicating that the TGF-␤-responsive element is activated in a cell type-independent manner (Fig. 2D).
These experiments identified a 975-bp fragment (KpnI to HindIII) of the SMAD7 gene that contained both the TGF-␤responsive and basal promoter elements. This DNA fragment was sequenced in its entirety. Sequence analysis using the MatInspector version 2.2 program (19) did reveal several putative binding sites for transcription factors (Fig. 3A). The absence of a TATA-box and the presence of multiple Sp1 sites in this region suggested that the SMAD7 gene has a TATA-less promoter (20). To identify putative transcription initiation sites in the Smad7 promoter, we performed a series of primer extension analyses with poly(A) RNA from human kidney (CLON-TECH). A major extension product of 49 bp was obtained in multiple experiments and compared with sequence analysis of genomic SMAD7 DNA, using the same primer (Fig. 3B). This major initiation site was designated as the ϩ1-position (Fig.  3A).
Smad  binding sequence (the SBE) at positions Ϫ179 to Ϫ172 (Fig.  3A). This sequence has been shown to interact with recombinant Smad3 and Smad4 and was sufficient to confer transcriptional activation by TGF-␤ upon a heterologous promoter reporter construct (21). To determine whether the SBE in the Smad7 promoter was necessary to confer TGF-␤-inducibility, we used site-directed mutagenesis to change thymidine Ϫ176 to adenine and adenine Ϫ175 to thymidine in pS7-5, resulting in pS7-5mSBE (Fig. 4A). These point mutations were expected to abolish binding of Smad3 and/or Smad4 to the SBE completely (21). When transfected into NIH3T3 cells, wild type pS7-5 conferred 2.7-fold induction of luciferase activity by TGF-␤ (Fig. 4B). In contrast, pS7-5mSBE was able to mediate basal promoter activity but did not confer induction by TGF-␤ (Fig. 4B), demonstrating that the SBE at Ϫ179 to Ϫ172 was necessary for induction of the Smad7 promoter by TGF-␤.
Next, we used radiolabeled oligonucleotide probes spanning positions Ϫ185 to Ϫ166 in electrophoretic mobility shift assays (EMSAs) to examine whether the Smad7 SBE (S7SBE) was able to interact with nuclear protein complexes. Nuclear protein extracts were prepared from untreated and TGF-␤-treated NIH3T3 fibroblasts. DNA binding of a protein complex labeled the SBE-binding complex (SBC) was specific and strongly increased in nuclear extracts from TGF-␤ treated NIH3T3 when compared with untreated NIH3T3 cells (Fig. 4C, lanes 3 and 2,  respectively). Time course experiments indicated that the induction of SBC was detectable as early as 10 min and strongest after 40 min of TGF-␤ treatment (data not shown). Preincubation of nuclear extracts from TGF-␤-treated NIH3T3 cells with anti-SMAD2, anti-SMAD3, and anti-SMAD4 antibodies revealed significantly reduced SBC binding in the presence of anti-SMAD2 antibodies (Fig. 4C, lane 6) or supershifted SBC complexes in the presence of anti-SMAD3 (lane 7) or anti-SMAD4 (lane 8), suggesting that SBC contained Smad2, Smad3, and Smad4 antigens. We obtained similar results when nuclear extracts from untreated NIH3T3 were used, albeit the intensity of the SBC signal was much weaker throughout the experiment (Fig. 4C, lanes 9 -14). These results indicated that a nuclear protein complexes containing Smad2, Smad3, and Smad4 formed at the SBE in the Smad7 promoter at base line and that TGF-␤ treatment strongly increased the amount of bound Smad protein complexes, resulting in transcriptional activation of the Smad7 promoter. The SBE probe specifically interacted with a higher molecular weight complex in most experiments (labeled with an asterisk, Fig. 4C). The binding characteristics of this complex were not significantly altered by TGF-␤ or anti-SMAD antibodies.
Since it has been reported that the GTCTAGAC sequence does not interact with recombinant Smad2 (21), we further investigated whether Smad2 antigens participated in SBC on the Smad7 promoter. We used, in addition to the polyclonal anti-SMAD2 antibody, a monoclonal anti-SMAD2 antibody that specifically detected Smad2 but not Smad3 or Smad4 (see Fig. 5A). This antibody supershifted the SBC irrespective of whether it was added to the binding reaction before or after the addition of the SBE probe (Fig. 4D), suggesting that Smad2 participates in the SBC.
Smad3 and Smad4 Are Required for Induction of the Smad7 Promoter by TGF-␤-In order to provide conclusive evidence for a role of Smad proteins in the induction of the SMAD7 gene by TGF-␤, we obtained MEFs with either targeted deletions of Smad2 (15) or Smad3 (16) and the SMAD4-deficient breast cancer cell line MD-MBA468. Western blot analysis using anti-SMAD2 and anti-SMAD3 antibodies confirmed that both Smad2 Ϫ/Ϫ MEFs and Smad3 Ϫ/Ϫ MEFs lacked Smad2 or Smad3 protein, respectively (Fig. 5A). EMSAs using nuclear extracts prepared from untreated and TGF-␤-treated wild type control MEFs confirmed both the basal and inducible binding of SBC to the SBE probe (Fig. 5B, lanes 1 and 2). SBC formation was also observed in Smad2 Ϫ/Ϫ nuclear protein extracts, albeit the signal intensity of basal and induced SBCs appeared weaker compared with wild type nuclear protein extracts (Fig. 5B,  lanes 4 and 5). In contrast, the binding of TGF-␤-inducible SBC was dramatically reduced in Smad3-deficient nuclear protein FIG. 3. Promoter sequence and primer extension analysis of the Smad7 promoter. A, the Smad7 promoter sequence contained in the functional pS7-5 construct is presented between positions Ϫ303 and ϩ672 relative to a major transcription start site (designated as ϩ1 and denoted with an arrow). Putative binding sites for transcription factors were identified using the MatInspector version 2.2 program. Several of those putative binding sites are underlined. A putative Smad binding element (SBE (21)) is boxed. The sequence of the primer extension primer is shown. B, primer extension analysis using human kidney mRNA as template showing a major extension product (lane 2). Lane 1 shows radiolabeled DNA markers (40, 48, and 66 bp, respectively). Lane 3, result of the control reaction in which RNA was omitted. All reactions and markers were loaded together and run next to a DNA sequencing reaction (lanes 4 -7) to determine the relative positions of the transcription start site. extracts (Fig. 5B, lane 8), and neither basal nor TGF-␤-inducible SBC were detectable in nuclear extracts derived from Smad4-deficient MD-MBA468 (Fig. 5B, lanes 10 and 11). The high molecular weight DNA-protein complex (*) was detectable irrespective of the presence or absence of Smad2 and Smad3 but was not observed in Smad4-deficient cells (Fig. 5B). Our data suggest that Smad4 is absolutely required for the formation of basal and TGF-␤-inducible SBC and that Smad3 is a major component of the TGF-␤-inducible SBC. In contrast, Smad2 is not required for binding of SBC to the SBE probe. FIG. 4. An inducible complex consisting of Smad2, Smad3, and Smad4 interacts with a Smad binding element that is required for activation of the Smad7 promoter by TGF-␤. A, oligonucleotide sequence containing wild type (S7SBE) and mutant (S7mSBE) Smad binding elements. B, corrected luciferase activities (relative luciferase units (RLU)) in untreated (black bars) and TGF-␤-treated (gray bars) NIH3T3 fibroblasts after transfection with wild type pS7-5 (pS7-5wt) or FIG. 5. Smad3 and Smad4 but not Smad2 are necessary for SBC binding. A, Western blot analysis of cell lysates from untreated (Ϫ) or TGF-␤-treated (ϩ) wild type (wt), Smad2-deficient (Smad2 Ϫ/Ϫ ) and Smad3-deficient (Smad3 Ϫ/Ϫ ) MEFs. The membrane was probed with a monoclonal anti-Smad2 antibody (Transduction Laboratories) and a polyclonal anti-Smad3 antibody (Zymed Laboratories Inc.). The same blot was probed for GDP dissociation inhibitor (GDI) to control for equal protein loading. B, EMSA using the S7SBE oligonucleotide probe and nuclear protein extracts from wild type (wt) (lanes 1-3), Smad2-deficient (Smad2 Ϫ/Ϫ ) (lanes 4 -6), Smad3-deficient (Smad3 Ϫ/Ϫ ) (lanes 7-9) MEFs and MDA-MB-468 cells (lanes 10 -12). SBC and an asterisk denote TGF-␤-inducible and constitutive S7SBE-binding protein complexes, respectively. mutant pS7-5 (pS7-5mSBE) Smad7 promoter luciferase reporter gene constructs. C, EMSA demonstrating interaction of a double-stranded oligonucleotide probe (S7SBE) with a TGF-␤-inducible SBC in nuclear protein extracts from untreated (Ϫ) or TGF-␤-treated (ϩ) NIH3T3 fibroblasts (2.5 ng/ml TGF-␤1 for 40 min). Open and filled arrowheads denote supershifted complexes when nuclear extracts were incubated with anti-Smad3 (␣-Smad3) or anti-Smad4 (␣-Smad4) antibodies before the addition of the probe, respectively. The asterisk denotes a complex of constitutive SBE-binding proteins (see "Results"). A 50-fold molar excess of unlabeled, annealed SBE oligonucleotides ablates SBC binding (lane 4). Nonimmune goat IgG and anti-Smad2 (␣-Smad2) antibody are shown. D, EMSA comparing the effect on the SBC of a polyclonal goat anti-Smad2 (␣-Smad2 SC) or a monoclonal anti-Smad2 (␣-Smad2 TL) added to the binding reaction either before or after the S7SBE probe, respectively. Next, we transfected the TGF-␤-responsive Smad7 promoter reporter construct pS7-5 into wild type control MEFs, Smad2deficient MEFs, Smad3-deficient MEFs, and Smad4-deficient MD-MBA468 cells to examine the functional roles of these Smad proteins in the transcriptional regulation of the SMAD7 gene. TGF-␤ treatment resulted in significant 2.4-and 2.1-fold increases of luciferase activities in transfected wild type and Smad2-deficient MEFs, respectively (Fig. 6A). In contrast, TGF-␤ had no significant effect on luciferase activities in pS7-5-transfected Smad3-deficient MEFs (1.2-fold induction) and Smad4-deficient MD-MBA468 cells (0.9-fold induction). To confirm the essential roles for both Smad3 and Smad4 in the induction of the Smad7 promoter by TGF-␤, we reconstituted the deficient Smad proteins by cotransfection of cytomegalovirus promoter/enhancer expression vectors for Smad2 (pFS-mad2), Smad3 (pFSmad3), or Smad4 (pHASmad4), respectively, together with the pS7-5 reporter plasmid (see Fig. 6B). Reconstitution of Smad2 in Smad2-deficient MEFs had no significant effect on either base-line or TGF-␤-inducible pS7-5mediated luciferase activities (Fig. 6B). In contrast, reconstitution of Smad3 in Smad3-deficient MEFs resulted in a strong increase of both base-line and TGF-␤-inducible luciferase activities. Reconstitution of Smad4 in MD-MBA468 cells rescued the inducibility of the Smad7 promoter luciferase activity by TGF-␤ and had no effect on basal promoter activity (Fig. 6B). These results provide conclusive evidence that both Smad3 and Smad4 are essential for the induction of the Smad7 promoter by TGF-␤. DISCUSSION We report a molecular mechanism that may have a central role in negative autoregulation of TGF-␤/Smad signaling. Our results demonstrate that ligand-dependent activation of TGF-␤ receptor complexes induces the interaction of Smad2, Smad3, and Smad4 transcription factor complexes with a palindromic consensus Smad binding element in the human Smad7 promoter. Mutations in this cis-acting element or deletion of either Smad3 or Smad4, but not Smad2, ablate the ability of TGF-␤ to induce the human Smad7 promoter. Thus, the transcriptional regulation of Smad7, an intracellular inhibitor of the TGF-␤ type I receptor (10,12), by TGF-␤ itself is mediated through a rapid and direct Smad3-and Smad4-dependent signaling mechanism.
Detailed molecular studies of a number of TGF-␤-responsive promoters suggest that the TGF-␤/Smad pathway regulates transcription by at least two distinct mechanisms. The first mechanism involves the interaction of Smad proteins with other transcription factors at their specific binding sequences in TGF-␤-responsive promoters. For example, the TGF-␤ or activin response of the Mix.2 promoter is mediated by a Fast-2-dependent transcriptional activator complex consisting of Fast-2 and Smad2-Smad4 complexes (9). In contrast, a number of examples support a distinct mechanism of regulation in which Smad3 and Smad4 activate transcription through direct interaction with specific DNA sequences (i.e. CAGA) or socalled Smad binding elements (22)(23)(24). Our observations that the Smad7 promoter contains a palindromic GTCTAGAC sequence that mediates binding of and transcriptional activation by Smad3 and Smad4 now provide an additional important example for direct Smad3/Smad4-dependent transcriptional regulation by TGF-␤.
However, our observation that Smad2 is associated with Smad3 and Smad4 in the TGF-␤-inducible SBE-binding complex that interacts with the GTCTAGAC element in the Smad7 promoter has not previously been reported and raises additional questions. First, we show that Smad2 is associated with the basal and TGF-␤-inducible SBE-binding complex, but in contrast with Smad3 and Smad4, Smad2 is not required for induction of the Smad7 promoter by TGF-␤. A functional role for Smad2 in the SBC therefore remains to be established. In addition, the GTCTAGAC sequence has been identified in a oligonucleotide-based screening by its binding to the major homology-1 domain of recombinant Smad3 and Smad4 (21). The major homology-1 domain of Smad2 was unable to bind to this artificial binding sequence directly. However, the human Smad7 promoter represents the first naturally occurring gene that contains the GTCTAGAC sequence as a functional SBE. Since our binding studies were performed using whole nuclear protein extracts, instead of recombinant proteins, it is likely that Smad2 does not bind DNA directly but participates in a heterotrimeric Smad-binding complex containing at least Smad2, Smad3, and Smad4 (25). Additional studies will be needed to clarify this issue.
Several reports indicate that the expression of Smad7 is induced by independent pathways including TGF-␤, activin, BMP-7, IFN-␥, shear stress, and NF-B pathways (11,13,14). 2 Our results provide evidence that signaling by the TGF-␤ superfamily members TGF-␤, activin, and BMP-7 may converge on the SBE in the Smad7 promoter. Whereas both activin and BMP-7 stimulation resulted in a small but significant increase of the Smad7 promoter reporter gene activity, induction mediated by the three TGF-␤ isoforms ␤1, ␤2, and ␤3 was considerably stronger (see Fig. 2C). These results are consistent with a previous report demonstrating strong induction of Smad7 mRNA expression by TGF-␤1, compared with moderate induc-FIG. 6. Smad3 and Smad4 are necessary and sufficient to mediate activation of the Smad7 promoter by TGF-␤. A, corrected luciferase activities in untreated (black bars) or TGF-␤-treated (gray bars) cells with deletions of individual Smads as indicated. -Fold induction of TGF-␤-inducible pS7-5 reporter activity by TGF-␤ in cells cotransfected with the reporter construct and pRSV-␤Gal is indicated. B, rescue of the induction of the Smad7 promoter reporter construct pS7-5 by TGF-␤ following reconstitution of Smad2, Smad3, and Smad4 in Smad-deficient cells, respectively. Smads were reconstituted by cotransfection with expression plasmids for Smad2 (pFSmad2), Smad3 (pFSmad3), and Smad4 (pHASmad4), as indicated. tion by activin and BMP-7 (13). In contrast, our results indicate that TNF-␣, IFN-␥, and EGF may regulate the Smad7 gene through cis-and trans-acting elements independent of the SBE and its binding complex. These data support a model in which the overall degree of Smad7 gene expression under physiological or pathophysiological conditions may be determined through combinatorial activation of distinct regulatory elements at the level of the Smad7 promoter.
Our findings provide new insights into the regulation of the major TGF-␤/Smad signaling pathway that support an oscillating rather than a static mode of feedback regulation involving inhibitory Smad7. In such a model, activation of TGF-␤ receptor complexes by its ligands results in the activation of Smad3 and Smad4 and leads to the rapid and transient transcriptional activation and/or repression of a set of target genes that can be characterized as immediate early gene responses and include the SMAD7 gene. Smad7 may then directly inhibit further activity of the receptor complex by associating with it at the intracellular domain and blocking activation of receptor-regulated Smads. Thus, Smad7 may turn off its own Smad3/Smad4dependent transcription. Depending on the rate of dissociation and turnover of TGF-␤ receptor and Smad7 complexes, cells would regain responsiveness to new TGF-␤ stimulation and reactivation of the cycle. Such a model is consistent with observations that Smad7 mRNA expression shows an early peak at 30 -90 min and additional peaks between 4 and 24 h of exposure to TGF-␤ (13). 2 The molecular mechanisms of regulation (this report) and inhibitor function (10,12) of Smad7 are consistent with an emerging theme in feedback during signaling. For example, several negative regulators that are positively transcriptionally regulated by the pathway that they inhibit have been identified in epidermal growth factor receptor signaling in Drosophila (26 -28). We anticipate that the continued characterization of Smad7 and related molecules will provide novel approaches to the design of inhibitors of the TGF-␤/Smad signaling family for therapeutic use in oncogenesis and fibrogenesis.