Inhibition of Transforming Growth Factor β-enhanced Serum Response Factor-dependent Transcription by SMAD7*

Transforming growth factor (TGF)-β is present in large amounts in the airways of patients with asthma and with other diseases of the lung. We show here that TGFβ treatment increased transcriptional activation of SM22α, a smooth muscle-specific promoter, in airway smooth muscle cells, and we demonstrate that this effect stems in part from TGFβ-induced enhancement of serum response factor (SRF) DNA binding and transcription promoting activity. Overexpression of Smad7 inhibited TGFβ-induced stimulation of SRF-dependent promoter function, and chromatin immunoprecipitation as well as co-immunoprecipitation assays established that endogenous or recombinant SRF interacts with Smad7 within the nucleus. The SRF binding domain of Smad7 mapped to the C-terminal half of the Smad7 molecule. TGFβ treatment weakened Smad7 association with SRF, and conversely the Smad7-SRF interaction was increased by inhibition of the TGFβ pathway through overexpression of a dominant negative mutant of TGFβ receptor I or of Smad3 phosphorylation-deficient mutant. Our findings thus reveal that SRF-Smad7 interactions in part mediate TGFβ regulation of gene transcription in airway smooth muscle. This offers potential targets for interventions in treating lung inflammation and asthma.

TGF␤ 2 is overexpressed in the airways of asthmatics, where it is thought to play a key role in the airway remodeling characteristic of their disease, by regulating an array of cellular functions (1) that includes gene expression, cell proliferation, differ-entiation (2), apoptosis (3), and migration (4) and remodeling of the extracellular matrix (5). To accomplish these diverse effects, TGF␤ binds to two unique cell surface receptors, designated TGF␤ type I (T␤RI) and type II (T␤RII), which are present in almost every cell type and are directly involved in signal transduction through their serine/threonine kinase activities (6). Binding of TGF␤ to T␤RII induces phosphorylation and activation of T␤RI, which in turn phosphorylates Smad proteins, the major transducers of TGF␤ signals (7). Once phosphorylated at their C terminus (the MH2 domain), these receptor-activated Smads or "R-Smads" (Smad2 and Smad3) associate with the common mediator Smad4 ("Co-Smad"); R-Smad-Co-Smad complexes then translocate into the nucleus, where they bind to regulatory regions of target genes (at the Smad-binding element (SBE)) and interact with other transcription factors, co-activators, and co-repressors to regulate expression of TGF␤-responsive genes.
In several cell types, the inhibitory Smads, Smad6 and Smad7 ("I-Smad"), negatively regulate TGF␤ signaling (8). After TGF␤ stimulation, Smad7 translocates to the plasma membrane where it binds to T␤RI and inhibits further signaling. Two mechanisms of suppression of TGF␤ signaling involving Smad7 are presently known, and both occur outside the nucleus. (i) Smad7 is able to interfere with the association between the R-Smads and activated receptors (9,10). (ii) A complex containing Smad7 and the E3 ubiquitin ligase Smurf1 or Smurf2 translocates to the plasma membrane and induces ubiquitination and consequent degradation of TGF␤ receptors (10).
Several genes with expression restricted to the smooth muscle (SM) lineages have been identified (11), and many of them are responsive to TGF␤ (12). In vascular smooth muscle cells, TGF␤ induces expression of ␣-sm-actin, an abundant SM protein whose transcriptional regulation is under at least partial control of a TGF␤ control element (TCE) located within the promoter region (13). A similar TCE, to which Kruppel-like transcription factors bind (14), is required for in vivo expression of SM22␣, one of the earliest and most widely expressed smooth muscle cell markers identified (11). Recently, TGF␤ was reported to up-regulate SM22␣ expression in fibroblasts through binding of Smad3 to two Smad-binding motifs located within the first exon (15,16). Furthermore, signaling through Smad2 and Smad3 plays an important role in the development of SM cells from totipotent embryonic stem cells (17). Analysis of the regulatory regions of SM-specific genes reveals that the vast majority of them (including SM22␣) contain CArG sequences, the binding site for the MADS-box transcription factor serum response factor (SRF) (18). Besides its essential role driving SM-and non-SM muscle-specific gene expression, SRF is also critical for regulation of cell proliferation and differentiation (19). The transcription promoting activity of SRF is modulated by its subcellular localization (20 -22) and through its association with a variety of other transcription factors, activators, and co-repressors (mostly via its MADS domain). Indeed, the competition for binding of various cofactors (e.g. myocardin versus Elk-1) to a common docking site on SRF may underlie the molecular mechanism of SM plasticity, its ability to switch between differentiated and proliferative phenotypes in response to extracellular cues (23).
Beyond its influence at the TCE, TGF␤ can also promote smooth muscle-specific gene transcription through effects on SRF. First, immunoprecipitation studies show that Smad3 can form complexes with SRF within the nucleus, and such complexes formed on the SM22␣ promoter (where Smad3 also binds to an SBE in the first exon) appear to enhance SM22␣ transcription (15). Second, artificial overexpression of inhibitory Smads 6 and 7 partially blocks SRF-VP16-mediated activation of the SM22␣ promoter (15), through an undetermined mechanism. Third, TGF␤ can stimulate the accumulation of the SRF protein (24). Together, these findings point to SRF as an important mediator of TGF␤ effects on smooth muscle-specific gene transcription.
Given the importance of TGF␤ in asthmatic airway remodeling, and in airway smooth muscle hypertrophy in particular (25), we have further explored how TGF␤ modulates SRF-dependent gene transcription, by testing the hypothesis that inhibitory Smad7 associates with SRF, thereby reducing its transcription-promoting activity. Our experiments demonstrate for the first time that this association does indeed occur in both myogenic and nonmyogenic cells and that TGF␤ stimulation reduces this interaction, thereby restoring the transcription-promoting potential of SRF.
Plasmids-We used several SRF-dependent reporter plasmids in our studies. In pSM22luc, luciferase expression is driven by bp Ϫ445 to ϩ41 of the mouse SM22␣ gene (26). SME plasmids containing CArG sequence mutations that prevent SRF binding to the 5Ј-CArG box, the 3Ј-CArG box, or both sites were described previously (27). The artificial p5xCArGluc (Stratagene, La Jolla, CA) contains five identical sites (CCTA-ATATGG) separated by 4 bases and a minimal TATA box. pSRE.Lluc (gift from Dr. N. Dulin, University of Chicago) contains two identical CArG sites (CCATATTAGG) separated by 23 bases. pSM-MHCluc contains 3.3 kb of 5Ј-flanking DNA upstream of the human SM-myosin heavy chain gene (28) driving luciferase expression. pMSVluc contains a strong viral pro-moter that is constitutively active and has been described previously (29). pSBEx2tkluc, which contains the thymidine kinase promoter plus two copies of the GTCT sequence (Smad/ski-binding site) which confer responsiveness to activated Smad pathway, was provided by Dr. E. Medrano, Baylor College of Medicine. pGREluc is an artificial promoter that harbors a glucocorticoid-responsive element and TATA box (Clontech). In pMSV␤gal, the murine sarcoma virus promoter drives ␤-galactosidase. pCMV5 and the expression plasmids pCMVSmad7-HA and pSmad3(3A)FLAG, which contain serine to alanine mutations in the Smad3 phosphorylation sites, were gifts from Dr. L. Attisano (University of Toronto). Plasmids encoding FLAG tag fused to amino acid residues 2-259 of human Smad7 (FN-FLAG-Smad7) or 206 -426 (FC-Smad7-FLAG) were provided by Dr. C-H. Heldin (Uppsala University, Sweden). FLAG-T␤RI-DN plasmid, which expresses a kinase domain mutant of T␤RI, was a gift from Dr. P. ten Dijke (The Netherlands Cancer Institute). EGFP-SRF fusion proteins were generated by cloning cDNAs encoding full-length or corresponding mutants SRF into pEGFP-C1 (Clontech) that encodes enhanced green fluorescent protein. In plasmid EGFP-mNLS-EGFP, amino acids 95 and 96 within the nuclear localization signal 95 RRGLKR 100 were mutated to EE. In plasmid pEGFP-mDM-SRF amino acid residues 183 VLLLV 187 within the dimerization domain were replaced with AAAAA.
Transfection-To measure the activity of SRE.L, 5xCArG, GRE, and SBE promoters, passage 2 CTSMC of 70 -80% confluence in 24-well plates was transiently transfected the day after plating with 0.1 g of promoter luciferase reporter plasmid, 0.1 g of pMSV␤gal, and 25 ng of expression plasmid DNA mixed with 1.5 l of Plus Reagent (Invitrogen). Liposomes were formed by adding 2 g of Lipofectamine (Invitrogen). After 4 h of transfection, the liposome suspension was removed, and cells were fed with DMEM/F-12 plus 10% fetal bovine serum for 5 h. After that, cells were serum-deprived overnight in DMEM/F-12 supplemented with 0.1% bovine serum albumin (Sigma) plus antibiotics. The following day, cells were treated with or without TGF␤1 (150 pM, R & D Systems, Minneapolis, MN) for 8 h in serum-free medium and then harvested. In some cases, canine airway myocytes were long term serum-deprived (4 -7 days) and, where indicated, transfected as described previously (20) prior to TGF␤ treatment. For transfection using SM22␣ and SME promoter constructs, cells were seeded in 6-well plates and co-transfected the following day with 1.8 g of indicated luciferase reporter, 600 ng of MSV␤gal plasmid, and 100 ng of empty or Smad7 expression vector with 12 g of Lipofectamine. After 6 h, the medium was changed to DMEM/F-12 plus insulin, transferring selenium mixture as described previously (20) with or without 150 pM TGF␤. Cells were harvested 48 h after transfection. In all cases, cells were lysed with M-PER detergent (Pierce) and frozen once, and luciferase activity was measured using a commercially available kit (Promega, Madison, WI). ␤-Galactosidase assays were performed as internal control to correct for differences in transfection efficiency. Promoter function was determined as normalized luciferase activity over ␤-galactosidase activity; TGF␤ effect was expressed as a ratio of normalized luciferase activity in TGF␤-treated cells relative to control untreated cells. Results are expressed as average Ϯ S.D. from at least three independent experiments, each performed in triplicate wells, and were analyzed by analysis of variance and Student's t test. A p value of 0.05 or less was considered to be significant.
Electrophoretic Mobility Shift Assay (EMSA)-EMSA analyses were performed as described (20). Probes were end-labeled oligonucleotides corresponding to either the 5Ј-CArG of SM22␣ promoter (CTGCCCATAAAAGGTTTTTC, SRF-binding site underlined), the SM22␣-TCE (TGGAGTGAGTGGGGCGGC-CCGG, TGF␤-responsive site underlined), or a PAI-1 probe (TCGAGAGCCAGACAAGGAGCCAGACAAGGAGCCAG-ACAC, TGF␤-responsive sites underlined). Between 5 and 20 g of nuclear protein was used per reaction, in a total volume of 15 l. For supershift studies, 1 l (1-2 g) of antibody was added after the 15-min incubation period, and the reactions were incubated for an additional 20 min.
Chromatin Immunoprecipitation (ChIP) Assay-Eight million HEK293 cells were treated with and without TGF␤ for 24 h and cross-linked with formaldehyde. Total lysates were prepared, sonicated, and divided into five tubes for ChIP experiments, using a commercially available kit (Upstate, Charlottesville, VA). One percent lysate was saved to isolate control input DNA. The manufacturer's protocol was followed with two modifications. To reduce nonspecific background, after sonicated lysates were precleaned with salmon sperm DNA/protein G-agarose and incubated with antibody, protein A bound-agarose beads that were previously blocked with 1% bovine serum albumin were added. In addition, immunoprecipitated chromatin was washed twice (with half of the recommended volume each time) with high salt buffer. One g of total antibody was used per ChIP reaction. Anti-SRF (G-20) and anti-Smad7 (N-19 and H-79 used in equal amounts) antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA); anti-RNA polymerase II antibody and normal IgG were from the ChIP kit (Upstate). Two percent of purified DNA (1 l), including saved input control DNA, was used as template for 32 cycles in PCR. Primers pairs employed to amplify the human SM22␣ promoter that harbors the CArG and TCE were as follows: forward, TCCATCTCCAAAGCATGCAG, and reverse, CCCTCCCT-GCTAGAGGAAGC, which map to bp Ϫ224 and ϩ18, respectively. For ChIP-positive control reaction, primers that amplify the human GAPDH promoter provided by the Upstate kit were used.
Knockdown of Smad7 by siRNA Interference-HEK293 cells were seeded in 24-well plates with antibiotic-free media the day before transfection and used at 80 -90% confluence the next day. Four wells were transiently transfected with a mixture that contains 100 ng of pSM22luc, 100 ng of pMSVbgal, 1.5 l of Plus Reagent, and 1 l of Lipofectamine in Opti-MEM medium along with siRNA duplex targeting either Smad7 (purchased from Santa Cruz Biotechnology) or Foxm1 (as control, a gift from Dr. V. Kalinichenko, University of Chicago). siRNAs were mixed with RNAiFect reagent (Qiagen, Valencia, CA) in a 1:6 ratio. In two experiments, a nonsilencing RNA duplex from Qiagen was used as an additional control. After 6 h, cells were fed with complete medium and harvested 54 h later. Luciferase and ␤-galactosidase activities were measured in three replicate wells, and total RNA was isolated from the fourth replicate well. The average of normalized luciferase activity in siRNA FIGURE 1. An SRF-dependent mechanism mediates TGF␤1 enhancement of the SM22␣ promoter activity. A, TGF␤ stimulates SM22␣ promoter function in a dose-dependent manner. Long term serum-deprived CTSMC were incubated with the indicated amounts of TGF␤1 (black bars) or left untreated (white bar). TGF␤1 effect was determined as normalized promoter activity relative to untreated cells. B, TGF isoforms ␤-1, ␤-2, and ␤-3 but not TGF␣ increased SM22␣ promoter function. Subconfluent serum-fed CTSMC were transfected with wild type SM22␣ promoter and treated with 100 pM of indicated TGF isoform (black bars) or left untreated (white bar). The effect of TGF isoforms on promoter SM22␣ activity was determined as in A. C, SRF-binding sites are necessary for stimulation of SM22␣ promoter activity by TFG-␤1. CTSMC were transfected with wild type (white bars) or the indicated SM22␣ constructs harboring point mutations in either the 5Ј-(SME1, gray bars) or the 3Ј-CArG sites (SME4, hatched bars) or both (dmSME, black bars). Cells were treated with 300 pM TGF␤1 (right bars) or left untreated (left bars). Mutation of CArG sequences that abolish SRF binding reduced but did not ablate TGF␤ induction of SM22␣ promoter activity (p Ͻ 0.01 between none and TGF␤treated cells, except for SME4 (p ϭ 0.06)).
Samd7-transfected cells was expressed relative to that obtained in control siRNA-transfected cells. Smad7 expression was analyzed by reverse transcription-PCR using cDNA prepared from 1 g of total RNA and a set of commercially available Smad7 primers from Santa Cruz Biotechnology. A 521-bp amplicon was seen following the manufacturer's instructions. Expression of Smad7 protein in the same lysates used for luciferase and ␤-galactosidase measures was analyzed by Western blot using the anti-Smad7 antibody N-19 from Santa Cruz Biotechnology, and was visualized by ECL from Pierce. siRNA experiments were performed four times with similar results.
Co-immunoprecipitation (IP) Assays-Unless indicated otherwise, CTSMC, COS, or HEK293 cells were transiently co-transfected with Smad7-HA and Myc-SRF expression vectors or their respective empty vector controls along with other members of the TGF␤ signaling pathway as indicated, in the presence of Lipofectamine as described above. In studies on regulation of SRF-Smad7 association, cells were treated with TGF␤ (4 ng/ml), IL-6 (50 ng/ml), or IL-13 (50 ng/ml) after transfection, as indicated. Nuclear and cytosolic extracts were prepared in the presence of protease inhibitors 24 h later using a Pierce kit. Equal amounts of protein from each sample (40 -200 g in 100-l final volume) were preincubated first with normal IgG-conjugated agarose beads (Santa Cruz Biotechnology) for 30 min on ice with occasional gentle mixing. Supernatants were recovered after centrifugation (30 s at 500 ϫ g) and incubated, unless indicated otherwise, with anti-Myc antibody-conjugated beads (Santa Cruz Biotechnology) for 3 h at 4°C in an orbital platform. Beads were precipitated as above, washed twice, and boiled in SDS loading buffer. Interacting proteins bound to the beads were resolved on SDS-PAGE, transferred to nitrocellulose, and probed with polyclonal anti-HA antibody unless indicated otherwise. Signal was detected by ECL (Pierce). For experiments using Smad7 mutants, membranes were incubated with anti-FLAG M2 antibody (Sigma). When SRF mutants were employed, extracts were incubated with anti-green fluorescent protein antibody-conjugated beads (Santa Cruz Biotechnology).

TGF␤1 Stimulates Transcriptional Activation of Smooth
Muscle-specific Genes in Myocytes from the Airway through an FIGURE 2. TGF␤ increases SRF-DNA binding capacity. A, NE from long term serum-deprived CTSM cells exposed to 150 pM TGF␤1 for the indicated times were incubated with a probe containing the SM22␣-5Ј-CArG site. TGF␤1 administration elevated SRF DNA binding activity (arrow) with increasing exposure times. B, crosslinked chromatin-containing lysates were prepared from untreated (Ϫ) or TGF␤-treated (ϩ) HEK293 cells. ChIP was performed with anti-SRF, anti-Smad7, or anti-RNA polymerase II (RNAP) antibodies as indicated, along with no added antibody as negative control (none). Chromatin that undergoes no IP serves as positive control (input). Amplification of SM22␣ promoter sequences that contain the CArG plus TCE sites (top) or of GAPDH promoter sequences (bottom) was performed by PCR from purified DNA immunoprecipitated from chromatin. Binding of SRF to the SM22␣ promoter increased in cells treated with TGF␤ (compare intensities of lane 1 versus lane 2). IP with anti-Smad7 antibodies was able to retrieve DNA containing the SRF-binding site (lane 7), although a fainter signal was obtained from TGF␤-treated cells (lane 8). Similar levels of binding to the GAPDH promoter, a gene reported to be SRF-independent, were achieved in chromatin treated with anti-RNAP II (lanes 13 and 14) but not with anti-SRF antibody (lanes 17 and 18). Negligible background signals were observed when no antibody (lanes 3, 4, 11, and 12) or DNA (lanes 9 and 19) was added. C, total cell lysates were prepared from control CTSMC or cells treated with 150 pM TGF␤ for 24 h. The membrane was probed first with anti-SRF antibody and then reprobed with anti-␤ actin antibody. D, NE were prepared from serum-fed CTSMC exposed to 150 pM TGF␤ for 24 h and incubated with either the SM22␣-5ЈCArG or the PAI-I probe that contains a highly TGF␤-responsive element. E, formation of DNA-protein complexes on the SM22␣-TCE. NE were prepared after exposure of serum-fed CTSMC to 100 pM TGF␤ for the indicated times. Probe contains the TCE from the SM22␣ promoter. TGF␤ induced appearance of two TCE binding activities (arrows) in a time-dependent fashion.

Smad7-SRF Interactions
SRF-dependent Pathway-We first tested whether TGF␤ affects promoter function of SM-specific genes in myocytes from the airway. Fig. 1A shows that TGF␤ treatment enhanced transcription from the SM22␣ promoter after transient transfection into long term serum-deprived myocytes by 3-9-fold in a dose-dependent manner. Comparable stimulation was observed for the human smooth muscle myosin heavy chain promoter, but not with the viral murine sarcoma virus long terminal repeat promoter under the same conditions (data not shown). This indicates that TGF␤ action on SM-specific promoters was selective. Transcriptional activation of SM22␣ promoter was also inducible by TGF␤ in subconfluent serum-fed cells, although to a lesser extent, and all three TGF␤ isoforms (but not TGF␣) increased SM22␣ promoter activity to a similar degree (Fig. 1B).
To delineate the sequence(s) responsible for TGF␤ responsiveness, we investigated the potential role of the CArG elements present in the SM22␣ promoter by inactivating them through site-directed mutagenesis. Point mutations that abolished SRF binding at either the 5Ј-CArG site at Ϫ275 bp or the 3Ј CArG site at Ϫ150 bp reduced basal SM22␣ promoter activ-ity (Fig. 1C, left bars) and also decreased the TGF␤-dependent enhancing effect (Fig. 1C, right  bars). Moreover, mutation of both SRF-binding sites inhibited even further the TGF␤-induced augmentation of SM22␣ promoter function. Interestingly, the TGF␤ effect was not completely abolished, supporting the notion that an SRF-independent mechanism also operate in airway myocytes.
TGF␤ Increases SRF-DNA Binding Activity-The mutagenesis studies described above prompted us to investigate whether TGF␤ influenced SRF DNA binding activity. To test this, we performed EMSA using nuclear extracts prepared from subconfluent airway smooth muscle cell cultures and a probe containing the SM22␣-5Ј-CArG site ( Fig. 2A). Compared with untreated controls, TGF␤ exposure increased the intensity of the DNA-protein complex in a time-dependent manner, up to 3 h of treatment. By competition and supershift analyses with anti-SRF antibody (not shown here but see Fig. 5D), we confirmed the presence of SRF within this complex.
We performed ChIP experiments to examine whether there is elevation of binding of endogenous SRF to the CArG sites within the SM22␣ promoter upon TGF␤ stimulation in vivo. Fig. 2B shows that anti-SRF antibody was able to immunoprecipitate cross-linked chromatin that contains DNA sequences harboring the CArG site from the SM22␣ promoter (lane 1). A stronger binding was observed from cells stimulated with TGF␤ (Fig. 2B, lane 2). These results indicate that SRF occupies the CArG site within the chromatin, and SRF binding to DNA is increased upon TGF␤ stimulation. Furthermore, Fig. 2B also demonstrates that anti-Smad7 antibody was capable of retrieving chromatin harboring SM22␣ promoter sequences that possess SRFbinding sites (lane 7). If cells were treated with TGF␤, the efficiency of recovering CArG sites-containing DNA was less evident (Fig. 2B, lane 8). As controls, TGF␤ induced no change in the signal for GAPDH promoter sequences that are precipitated using anti-RNA polymerase II antibody (Fig.  2B, lanes 13 and 14). Retrieval of sequences from the SM22␣ (Fig. 2B, lanes 3 and 4) or GAPDH (lanes 11 and 12) promoters were negligible in chromatin treated with no antibody. No GAPDH promoter sequences could be amplified from chromatin precipitated using anti-SRF antibody (Fig. 2B,  lanes 17 and 18). Altogether, our data indicate that Smad7 associates in vivo with SRF over the CArG element of the FIGURE 3. TGF␤ increases SRF-dependent transcriptional activity. CTSMC were co-transfected with two different SRF-dependent luciferase reporter plasmids 5xCArG (A) and SRE.L (B) plus empty CMV (left) or Smad7 (right) expression vectors. TGF␤ (100 pM) was added to selected wells (black bars), and controls were left untreated (white bars). Activity of both promoters increased more than 2-fold with TGF␤ treatment relative to untreated cells, and overexpression of Smad7 diminished basal as well as TGF␤-induced promoter function (#, p Ͻ 0.03; *, p Ͻ 0.01). C, cells were transfected with the artificial Smad-sensitive pSBEx2.tk promoter as above. TGF␤ stimulation enhanced more than 3-fold SBE promoter function (black bars) compared with control (white bars). This increase was abolished by expression of Smad7. *, p Ͻ 0.02. D, cells were co-transfected with GRE.luc, a glucocorticoid-dependent but SRFand Smad-insensitive promoter plasmid plus empty CMV or Smad7 expression vectors as above. No significant changes in promoter function were observed with TGF␤ stimulation (black bars) either with or without Smad7 overexpression relative to control (white bars). JULY 21, 2006 • VOLUME 281 • NUMBER 29 JOURNAL OF BIOLOGICAL CHEMISTRY 20387 SM22␣ promoter, and TGF␤ stimulates SRF-DNA binding activity to the SM22␣ promoter.

Smad7-SRF Interactions
Hautmann et al. (24) reported that increased SRF-DNA binding capacity was associated with augmented SRF content in TGF␤-treated vascular myocytes, but no such increase in SRF abundance was present in aorta endothelial cells. Western analysis of total cell lysates demonstrated that enhanced SRF-DNA binding activity in our system occurs with only minimal increase in SRF abundance (Fig. 2C). Fig. 2D shows that the level of stimulation of SRF SM22␣-5ЈCArG complex formation by TGF␤ administration was comparable with that observed by using a probe from the plasminogen activator inhibitor (PAI)-1 promoter, which is highly inducible by TGF␤ and contains Smad3/Smad4-binding sites. Finally, we demonstrated DNAprotein interactions using nuclear extracts from airway smooth muscle cells incubated with a probe containing the TCE from the SM22␣ promoter region (Fig. 2E). This sequence, which spans bp Ϫ132 to Ϫ111, is devoid of CArG sites. Formation of TCE binding activities from serum-fed airway smooth muscle cells was evident as early as 15 min of TGF␤ treatment.
Smad7 Is a Negative Regulator of SRF-dependent Transcription Activity-Previous studies have shown that Smad sites within the first exon of SM22␣ gene mediate TGF␤ responsiveness of SM22␣ and have identified a putative TCE in the 5Ј-flanking region. Thus, to establish whether TGF␤ induces smooth muscle gene transcription through activation of SRF transcription-promoting function independently of these sites, we performed transient transfection studies using two promoter reporter constructs solely responsive to SRF. In p5xCArGluc, five tandem copies of the CArG site and a TATA box drive expression of luciferase reporter; in pSRE.Lluc, the Ets-binding site in the c-fos promoter serum-response element is inactivated leaving only the SRF-binding sequences intact. Because TGF␤ action is mediated by Smad members in many cell types, we further hypothesized that overexpression of the inhibitor Smad7 would decrease SRF-dependent transcriptional activity induced by TGF␤ in airway myocytes. Fig. 3, A and B, shows that TGF␤ increased transcription from 5xCArG and SRE.L promoters by 2-fold, relative to untreated control cells. Moreover, overexpression of Smad7 markedly reduced transcription from these promoters, in the presence of TGF␤ and even in non-TGF␤-treated cells. The latter finding probably reflects autocrine secretion of TGF␤ by our airway myocytes, as addition of anti-TGF␤ antibodies to culture medium also reduces transcription from the SM22␣ promoter in non-TGF␤-treated cells. 3 We found parallel results for transcription from pSBEx2.tk.luc, which harbors a promoter sensitive to Smad activation. Relative to untreated cells, TGF␤ increased reporter transcription more than 3-fold (Fig. 3C, left bars), and co-expression of Smad7 reduced this enhancement to basal levels (Fig. 3C, black bars). Smad7 also diminished basal SBE promoter function in the absence of added TGF␤, again suggesting the presence of a TGF␤ autocrine pathway. As a negative control, TGF␤ treatment did not increase transcription from a glucocorticoid-sensitive (GRE) promoter, which contains neither Smad sites nor CArG boxes (Fig. 3D, left bars). Likewise, overexpression of Smad7 did not decrease GRE promoter function in treated or nontreated cells. Altogether, these results indicate that Smad7 specifically inhibits TGF␤-stimulated activation of SRF-dependent transcription in airway myocytes.
Silencing of Smad7 Stimulates SRF-dependent Promoter Function-To test the participation of endogenous Smad7 in regulating SRF-dependent promoter transcription, we performed promoter reporter assays in transiently transfected cells in which endogenous Smad7 was depleted via siRNA interference. Fig. 4 shows that knockdown of endogenous Smad7 in transiently transfected HEK293 cells is sufficient to more than double the promoter function of the SM22␣ promoter compared with cells treated with control siRNA. Thus, although overexpression of Smad7 is capable of inhibiting basal function of SRF-dependent promoters (Fig. 3, A and B and Fig. 7), blockade of Smad7 expression, conversely, results in exaggerated SRFdependent transcriptional activity.
SRF Interacts with Smad7-The ChIP studies described above strongly indicate that the inhibitory effect of Smad7 on SRF-dependent transcription stems from an association of Smad7 with SRF. To confirm this observation, we performed co-immunoprecipitation assays using nuclear (NE) and cytosolic (CE) extracts from COS or HEK293 cells that were co-transfected with plasmids encoding myc-SRF or Myc alone along with Smad7-HA expression plasmids or empty vector. Fig. 5A shows that Smad7 was pulled down from NE only when SRF was co-expressed (lane 8), indicating that SRF interacts with

Smad7-SRF Interactions
Smad7 within the nucleus. We were also able to demonstrate interaction of Smad7-HA or 6xmyc-Smad7 with EGFP-SRF or GST-SRF in COS, CTSM, and HEK293 cells as well. Interestingly, association of Smad7 to SRF in the cytosolic fraction was undetectable in all three cell types (see Fig. 6D, and data not shown). To corroborate this interaction between the endogenous molecules as well, we prepared NE and CE from nontransfected HEK293 and COS cells and immunoprecipitated Smad7 with a polyclonal anti-Smad7 antibody in the presence of conjugated IgA/IgG-agarose beads. Pulled down proteins were size-fractionated by PAGE, blotted onto nitrocellulose membrane, and probed with anti-SRF antibody. Fig.  5B shows that endogenous Smad7 associates with endogenous SRF inside the nuclei of both cells but not in the cytosol of HEK293 cells (note that in this preparation, SRF expression in COS cells was exclusively nuclear ( lane 6 versus lane 8)). These results rule out the possibility that interaction of Smad7 and SRF occurs only with overexpressed and chimeric recombinant proteins.
Smad members participate in protein-protein interactions via their MH2 domain localized within the C terminus of the molecule. To determine the domain of Smad7 that associates with SRF, we performed pulldown assays with extracts from cells co-transfected with myc-SRF and FN-Smad7-FLAG or FC-Smad7-FLAG, in which the first 259 residues or amino acids 206 -426 of Smad7, respectively, are fused to the FLAG tag. Fig. 5C shows that in HEK293 cells, SRF interacts with the C-terminal portion of Smad7 in the nucleus (lane 4), but binding is negligible in the cytoplasm (lane 8). On the other hand, virtually no FN-Smad7-FLAG could be pulled down by SRF in the nucleus (Fig. 5C, lane 2) under identical conditions. This indicates that structural features that enable Smad7 to associate with SRF reside in the second half of the Smad7 molecule. Moreover, the Smad7-SRF interaction appears to be maintained even when SRF is bound to DNA, which is consistent with our ChIP results. Indeed, Fig. 5D shows that complexes formed by airway smooth muscle nuclear proteins on the 3Ј-CArG-SM22␣ probe contains SRF (as anti-SRF antibody supershifts the bands; lane 2) and Smad7 (as anti-C Smad7 antibody prevents complex formation).
SRF contains a functional nuclear localization signal (NLS) within its N terminus, and the MADS domain of SRF participates in a variety of functional protein-protein interactions. We tested whether these two features were involved in the ability of SRF to associate with Smad7. Co-immunoprecipitation assay shown in Fig. 5E demonstrates that the EGFP-SRF fusion protein harboring mutations within either the nuclear localization signal (mNLS-SRF) or the dimerization domain (mDM-SRF)  1, 2, 5, and 6). NEs were prepared after 24 h, and equal amounts of protein were incubated with beads conjugated with anti-Myc antibody. Proteins bound to beads were resolved on SDS-PAGE, probed with anti-HA antibody. Arrow shows the expected position for Smad7. I, input; P, pulldown. B, NE and CE were prepared from HEK293 or COS cells. Endogenous Smad7 was immunoprecipitated with anti-Smad7 antibody and IgG/IgA-conjugated beads. Bound endogenous SRF was detected by anti-SRF Western blot. Interaction of Smad7 to SRF occurs in the nuclei of both cells (lanes 1 and 5) but not in the cytosol (lanes 3 and 7). In this experiment, SRF expression in COS cells is mostly nuclear. I, input; P, pulldown. C, NE and CE were prepared from HEK293 cells co-transfected with expression vectors encoding Myc-SRF plus either FN-Smad7-FLAG or FC-Smad7-FLAG, as indicated. IPs were performed with anti-Myc antibody beads. Blot was probe with anti-FLAG antibody. SRF was able to pull down FC-Smad7 (lower horizontal arrow) but not FN-Smad7 (upper oblique arrow), indicating that the Smad7 C terminus is involved in binding to SRF. Association of SRF with FC-Smad7 was observed only in the nuclear fraction. Asterisk indicates nonspecific bands. D, EMSA was performed with NE prepared from cells treated with TGF␤1 (lanes [2][3][4][5] or left untreated (lane 1) and SM22␣ 5Ј-CArG probe. Arrows indicate bands that contain SRF, as demonstrated by supershift of these bands with an anti-SRF antibody (lane 2). Anti-Smad7 antibody raised against the C terminus (lane 4) but not the N terminus (lane 3) of Smad7 interferes with SRF-DNA binding. E, SRF association with Smad7 in the nucleus requires the NLS and dimerization domain (DM) of SRF. NE were prepared from HEK293 cells that overexpress 6xmyc-Smad7 plus the indicated EGFP-SRF mutants (see text). SRF fusion proteins were pulled down with anti-GFP green fluorescent protein antibody-conjugated beads. Interacting Smad7 was detected by Western blot using anti-Myc antibody. Mutations within NLS or the dimerization domain (DM) interfere with SRF-Smad7 association. I, input; P, pulldown. JULY 21, 2006 • VOLUME 281 • NUMBER 29 exhibited a decreased ability to associate with Smad7 compared with wild type SRF.

Smad7-SRF Interactions
Modulation of Smad7-SRF Association by TGF␤-Because Smad7 has a pivotal role in controlling TGF␤ actions, and TGF␤ itself regulates Smad7 function, we hypothesized that SRF-Smad7 interaction might be modulated by activation of the TGF␤ cascade. To test this possibility, we performed pulldown assays using NE from cells treated with and without TGF␤. Fig. 6A shows that TGF␤-stimulated cells exhibited a weaker SRF-Smad7 interaction than untreated cells. Conversely, when TGF␤ action was interrupted by interfering with the propagation of the signal via expression of a defective Smad3 mutant (Fig. 6B) or at the receptor level by co-expression of a dominant negative T␤RI mutant (T␤RI-DN) deficient in its kinase activity (Fig. 6C), a stronger interaction of Smad7 and SRF was elicited.
These results prompted us to investigate whether other interventions also modulated Smad7-SRF association. Fig.  6D shows that interleukin (IL)-13 or IL-6 treatment weak-ened the binding of Smad7 to SRF in the nuclei of HEK293 cells. Surprisingly and in contrast to TGF␤ stimulation, SRF and Smad7 became capable of associating in the cytosol of these IL-treated cells. This indicates that additional pathways activated (or inhibited) by these inflammatory cytokines independently of TGF␤ influenced Smad7-SRF interaction.
Smad7 Inhibits SM22␣ Promoter Function-Finally, we tested the effect of TGF␤ treatment and Smad7 expression in the context of the native SM22␣ promoter, which harbors the SRF-binding sites, the TCE and the SBE. Fig. 7 shows that TGF␤ up-regulates SM22␣ promoter function, and expression of Smad7 is capable of decreasing both basal as well as enhanced SM22␣ promoter activity mediated by TGF␤ in transiently transfected airway SM cells. Therefore, Smad7 also interferes with activation of SMspecific genes. Taken together, these studies indicate that in smooth muscle as well as in non-smooth muscle cells, the positive action of TGF␤ on SRF function is regulated by association with Smad7, and thus modulation of the cross-talk between SRF and Smad7 may occur in cells under (patho)physiological conditions. This could have important biological consequences given the paramount relevance of TGF␤ signaling and SRF-sensitive pathways in many aspects of cell homeostasis.

DISCUSSION
In this study, we demonstrate that modulation of SRF-Smad7 interactions in part mediates TGF␤-stimulated transcription of smooth muscle-specific genes in airway myocytes. We found that TGF␤ stimulates transcriptional activity of SM-specific promoters through an SRF-dependent pathway and that overexpression of Smad7 reduces TGF␤-induced stimulation of purely SRF-dependent promoters. We demonstrated that SRF associates with Smad7 in cell nuclei and that this association is reduced by TGF␤ treatment. Thus, we propose a novel mechanism by which TGF␤ influences SRF-dependent transcription. 1) Smad7 interacts with SRF, thereby reducing its transcription promoting activity. 2) TGF␤ treatment reduces SRF-Smad7 interaction, thereby restoring transcription promoting activity of SRF. Fig. 8 illustrates possible ways by which Smad7 could inhibit SRF-dependent transcriptional activity. First, as is the case in other cell types such as lung epithelial cells (8), Smad7 may FIGURE 6. The TGF␤ signaling pathway modulates SRF-Smad7 interaction. A, COS cells were co-transfected with Myc-SRF and Smad7-HA expression vectors. After transfection, cells were fed with serumcontaining medium for 18 h and the next day treated with TGF␤ or left untreated for 3 h. Nuclear extracts were prepared and IP performed after extracts were preincubated with normal IgG-coupled beads, using anti-Myc antibody-bound beads. Western blot was done with anti-HA antibody to detect Smad7. P, pellet; S, supernatant after IgG preincubation. TGF␤ stimulation decreased the interaction of Smad7 with SRF. B, COS cells were co-transfected with Myc-SRF and Smad7-HA expression plasmids plus either a vector encoding the phosphorylation-deficient Smad3(3A) mutant (lanes 1 and 2), wild type Smad3 (lanes 3 and  4), or empty vector CMV5 (lanes 5 and 6). IP and Western analysis were performed with NE pretreated with normal IgG beads, as above. Overexpression of signaling-defective Smad3 mutant increased Smad7 binding to SRF (lane 2 versus 6). C, IP was performed with COS cells NE co-transfected with Myc-SRF and Smad7-HA vectors along with the empty (lanes 1-3) or kinase-defective T␤RI (T␤RI-DN, lanes 4 -6) expression plasmids. Smad7 bound to SRF was detected by using anti-Myc antibody-conjugated beads followed by anti-HA Western of inputs (I), pulldown (P), and supernatants (S), as described above. The results demonstrated that blockade of TGF␤ signaling by overexpression of a dominant negative T␤RI mutant augmented Smad7 interaction to SRF (lanes 2 versus 5). D, NE and CE were prepared from HEK293 cells co-transfected with Myc-SRF and Smad7-HA. After transfection, cells were treated with IL-6 (50 ng/ml) or IL-13 (50 ng/ml) as indicated, for 18 h. IP was performed with anti-Myc antibody beads. Bound Smad7 was detected by anti-HA Western blot. Binding of Smad7 to SRF in the nuclei is weakened in the presence of these cytokines, compared with untreated cells (lanes 2 and 3 versus 6). Interestingly, SRF and Smad7 became competent to associate in the cytosol of IL-treated cells (lanes 8 and 10 versus 12).
translocate into the cytoplasm upon binding of TGF␤ to T␤RII and prevent phosphorylation of R-Smads, propagation of TGF␤ signaling, and subsequent stimulation of SRF-dependent genes. Second, Smad7 may interfere with formation of transcriptionally competent active R-Smad-SRF complexes. In this regard, Qiu et al. (15) demonstrated that SRF could bind Smad3 in fibroblasts. In a third scenario elucidated here, Smad7 in the nucleus interacts with SRF, thereby reducing SRF function. The latter notion is consistent with previously known features of Smad7 as follows: (i) Smad7 localizes within the nucleus of some cells in the absence of TGF␤ (30); (ii) Smad7 possesses a repressing function in transcription assays, even independently of TGF␤ (Ref. 31 and our observations); (iii) Smad7 can interact with co-activators and transcription factors in the nucleus (32); and (iv) physiologically relevant regulation of Smad7 export from the nucleus has been reported (33). To our knowledge, we are the first to demonstrate an association of Smad7 with SRF within the nucleus and to postulate that Smad7 may perturb the capability of SRF to enhance transcriptional functions. Based on our IP results, we believe that this inhibitory activity does not likely involve gross changes in the subcellular localization of Smad7 or SRF.
Hautmann et al. (24) reported that a marked augmentation in SRF-DNA binding in aortic SM cells correlated with substantially higher levels of SRF expression in TGF␤-treated cells compared with control cells. Interestingly, we observed in airway smooth muscle that the SRF binding to SM22␣-CArG sites increased after TGF␤ treatment, but this occurred with only minor increase in SRF abundance. Whether this difference from the study of Hautmann et al. (24) stems from distinct experimental conditions or diverse properties of airway and vascular myocytes is not yet known.
Our results reveal that the C terminus of Smad7 is necessary for binding to SRF, which is consistent with previous reports that established that this region is involved in Smad7-protein interactions, including binding to T␤RI. As expected, import of SRF to the nucleus was important for SRF association with Smad7; in addition, the MADS domain of SRF was required for nuclear SRF-Smad7 interaction.
Recent studies reveal a prominent role of TGF␤and SRFdependent pathways in lung homeostasis. Smad7 adds to the list of proteins that regulate SRF function. That Smad7 associates with and inhibits SRF and that this functionally significant association is modulated by the external cell milieu may have important biological consequences. TGF␤ is abundant in asthmatic airways, in which airway smooth muscle hypertrophy is also a prominent feature. Expression of TGF␤ is also up-regulated in patients with chronic obstructive pulmonary disease (34) or with lymphangioleiomyomatosis, a disease of abnormal proliferation of smooth muscle cells within the lung (35). In our studies we demonstrated that TGF␤, IL-13, and IL-6 were able to disrupt SRF-Smad7 interaction, which may lead to alleviation of the inhibitory effect of Smad7 on SRF function. It is reasonable to speculate that overactivation of SRF by persistent TGF␤ stimulation might contribute to smooth muscle hypertrophy in inflammatory lung diseases.