Convergence of p53 and Transforming Growth Factor β (TGFβ) Signaling on Activating Expression of the Tumor Suppressor Gene maspin in Mammary Epithelial Cells*

Using two-dimensional difference gel electrophoresis, we identified the tumor suppressor gene maspin as a transforming growth factor β (TGFβ) target gene in human mammary epithelial cells. TGFβ up-regulatesMaspin expression both at the RNA and protein levels. This up-regulation required Smad2/3 function and intact p53-binding elements in the Maspin promoter. DNA affinity immunoblot and chromatin immunoprecipitation revealed the presence of both Smads and p53 at the Maspin promoter in TGFβ-treated cells, suggesting that both transcription factors cooperate to induce Maspin transcription. TGFβ did not activate Maspin-luciferase reporter in p53-mutant MDA-MB-231 breast cancer cells, which exhibit methylation of the endogenous Maspin promoter. Expression of ectopic p53, however, restored ligand-induced association of Smad2/3 with a transfected Maspin promoter. Stable transfection of Maspin inhibited basal and TGFβ-stimulated MDA-MB-231 cell motility. Finally, knockdown of endogenous Maspin in p53 wild-type MCF10A/HER2 cells enhanced basal and TGFβ-stimulated motility. Taken together, these data support cooperation between the p53 and TGFβ tumor suppressor pathways in the induction of Maspin expression, thus leading to inhibition of cell migration.

Members of the transforming growth factor ␤ (TGF␤) 2 family play an important role in the regulation of cellular fate both in embryonic development and adult tissue homeostasis (1). It is generally accepted that they have both tumor suppressor and tumor promoter functions (reviewed in Refs. 2 and 3). The TGF␤ ligands bind to cognate serine/threonine kinase transmembrane receptors, which, in turn, phosphorylate and activate the Smad family of signal transducers. Once activated, Smad2 and Smad3 translocate to the nucleus where they control gene expression in association with Smad4 and transcriptional coregulators (4). The intrinsic DNA binding activity of Smads is of relative low affinity and specificity. While this activity is sufficient to drive Smad-regulated transcription of artificial reporter constructs, tissue-specific regulation of transcription by Smads in vivo is thought to depend on interactions with additional site-specific DNA binding factors. The most studied Smad-dependent transcriptional responses are those involved in cell cycle arrest and apoptosis, essential for the tumor suppressor role of the TGF␤s. Mutational inactivation or loss of TGF␤ receptors and/or Smads is permissive for epithelial cell transformation and carcinogenesis (1,2). On the other hand, introduction of dominant negative TGF␤ receptors into metastatic cancer cells inhibits epithelial-to-mesenchymal transdifferentiation, motility, invasiveness, and survival, supporting the tumor promoter role in TGF␤ in fully transformed cells (reviewed in Ref. 5). In addition, excess production and/or activation of TGF␤ by cancer cells contributes to tumor progression by paracrine mechanisms that modulate the tumor microenvironment (6).
Most carcinomas attenuate or lose the Smad-dependent anti-mitogenic effect but gain pro-metastatic abilities in response to TGF␤. In most tumors, this change occurs without acquiring genetic defects involving Smads or TGF␤ receptors, suggesting that alterations in other regulatory molecules can have a profound influence of the cellular response to TGF␤. One possible explanation is the presence of different Smad partners in different cell types, which converge with Smads at the level of target gene expression, thus modifying the biological output of TGF␤ signaling.
In an effort to understand the molecular mediators of crosstalk between TGF␤ and ErbB receptor signaling, we performed two-dimensional multi-variable difference gel electrophoresis (DIGE) coupled with mass spectrometry in MCF10A/HER2 cells treated with TGF␤. In these cells, TGF␤ induces signaling programs associated with enhanced motility and survival (12,13). One of the proteins induced by TGF␤ was Maspin, a tumor suppressor related to the serpin (serine protein inhibitor) family of protease inhibitors (14). Maspin was equally induced in MCF10A cells not overexpressing HER2 (controls), suggesting it was not involved in the transformed phenotype induced by TGF␤ in cells overexpressing the oncogene. Indeed, knockdown of Maspin with RNA interference in MCF10A/HER2 cells was permissive for TGF␤-induced motility, suggesting Maspin had retained its tumor suppressor function in these cells. Because of the known regulation of maspin gene transcription by members of the p53 family (15,16), we examined the coregulation of Maspin expression by p53 and Smads. Using Maspin reporter constructs with mutations in p53 and/or Smad-binding elements (SBEs) as well as RNA interference in non-tumorigenic and breast cancer cell lines, we show herein that Maspin is transcriptionally co-regulated by p53 and Smads, thus representing a point of convergence of two tumor suppressor pathways.
Two-dimensional DIGE Mass Spectrometry (MS)-MCF10A/ HER2 cells were treated with TGF␤ for different times (0, 8, 24, and 40 h), washed with phosphate-buffered saline, and harvested with ice-cold Nonidet P-40 lysis buffer (20 mM Tris (pH 7.4), 150 mM NaCl, 1% Nonidet P-40, 0.1 mM EDTA, plus protease and phosphatase inhibitors). After sonication for 10 s and centrifugation (14,000 rpm), protein concentration in the supernatants was measured using the BCA protein assay reagent (Pierce). The treatment timecourse was repeated independently in triplicate, and the global protein expression profiles of over 1,000 resolved proteins (including charged isoforms and processed proteins) from the resulting 12 samples were quantitatively analyzed and identified using minimal labeling two-dimensional DIGE and MS as described (20,21). The resulting mass spectral data were used to interrogate sequences present in the Swiss-Prot and NCBInr data bases to generate statistically significant candidate identifications using GPS Explorer software (Applied Biosystems) running the MASCOT search algorithm. Maspin was identified from 16 peptide ions, 8 of which provided independent MS/MS matches to provide a combined MOWSE score of 591 and 49% sequence coverage (21).
Extraction of Total RNA and RT-PCR-Total RNA was extracted using the RNeasy mini-kit (Qiagen). RT-PCR was carried out using the titanium one-step RT-PCR kit (BD Biosciences). For each RT-PCR reaction, 100 ng RNA were added to a 50-l reaction system according to the manufacturer's protocol (50°C for 1 h followed by 30 cycles of PCR amplification) using the same primers for Maspin cDNA cloning. The PCR products were analyzed in 1.2% agarose gels.
Transfection and Luciferase Reporter Assays-Cells were seeded in 6-well plates and transfected with 0.5 g of luciferase reporter plasmid with or without 1 g of p53 expressing plasmid pCEP4-p53 or empty vector, along with 0.01 g of pCMV-Renilla using FuGENE 6 according to the manufacturer's protocol. Firefly and Renilla reniformis luciferase activities were measured using the dual luciferase assay system (Promega) as reported previously (22).
DNA Affinity Immunoblotting (DAI) and Chromatin Immunoprecipitation (ChIP)-DAI and ChIP assays were performed as described (12). For DAI assay, the wild-type Maspin promoter DNA and a promoter with a mutation in its p53 site I were released from pM-Luc(Ϫ297) and pM-Luc(Ϫ297mt1), respectively, by restriction enzyme digestion and subsequently labeled with biotin. Labeled probes were incubated with 200 g of nuclear extract prepared from cells that had been treated or not with TGF␤. Streptavidin magnetic beads (0.1 mg) were added for 1 h and then washed with DNA-binding buffer 5 times, boiled in 20 l of 2ϫ SDS-gel loading buffer before separation by SDS-PAGE followed by immunoblot. For ChIP assay, PCR (30 cycles) was performed using the Maspin promoter primers for pM-Luc(Ϫ154) cloning, i.e. 5Ј-ctaggtacgcatgtctgtacgtatg and 5Ј-gattgaaagcttagaagcagcggtggctcacc. Primer 5Ј-gagtggccctatcaaatgtta and the reverse primer for Maspin cDNA cloning, which amplify a 194-bp Maspin coding region, were used for control PCR.
Electrophoretic Mobility Shift Assay (EMSA)-Oligonucleotides 5Ј-ctaggtacgcatgtctgtacgtatgcatgtaact and 5Ј-ctagagttacatgcatacgtacagacatgcgtac, 5Ј-ctaggtacgcatCtAGAtacgtatgcatgtaact and 5Ј-ctagagttacatgcatacgtaTCTaGatgcgtac, and 5Ј-ctaggtaATATAgtctgtacgtatgcatgtaact and 5Ј-ctagagttacatg-catacgtacagacTATATtac were annealed to generate the DNA probes p53-II, mtp53-II and mtp53-IIS, respectively. The probes were radiolabeled with [␣-32 P]dCTP by incubation with Klenow DNA polymerase in the presence of dNTP (minus dCTP) as described elsewhere (23). Approximately 50,000 cpm of the 32 P-labeled probe were added to 20 g of nuclear extract prepared from MCF10A cells. The mixture was incubated for 30 min at room temperature in a binding system containing 10 mM HEPES (pH 7.5), 50 mM KCl, 1 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 1% Triton X-100, 5% glycerol, and 2 g of poly(dI-dC). For supershift experiments, 1 g of p53 antibody (Cell Signaling) or Smad4 antibody (Santa Cruz Biotechnology) was added to the mixture after 30 min and incubated for additional 30 min before gel loading. Samples were separated in a 4.5% polyacrylamide gel in a buffer containing 10 mM HEPES (pH 7.5), 1 mM EDTA, and 0.5 mM EGTA at 150 V and 4°C. The gel was subsequently dried and subjected to autoradiography.
Cell Motility and Invasion Assays-Cells growing in complete medium were allowed to reach confluence on a 6-well plate and then incubated in serum-free medium for 24 -40 h. The monolayers were wounded with a plastic pipette tip as described (22) and replenished with fresh serum-free medium Ϯ 2 ng/ml TGF␤1. Phase contrast images were photographed at different times following the addition of ligand. Invasion assay was performed in BD Bio-Coat growth factor-reduced Matrigel invasion chambers (BD Biosciences) according to the manufacturer's protocol.
Indirect Immunofluorescence Assay (IFA)-Cell IFA were performed as described previously (24). Fluorescent images were captured using a Princeton Instruments cooled CCD digital camera from a Zeiss Axiophot upright microscope.

RESULTS
Characterization of maspin as a TGF␤ Target Gene-We used two-dimensional DIGE to identify TGF␤-induced changes in protein abundance in MCF10A/HER2 human mammary epithelial cells (21). Two species of Maspin resolved by isoelectric focusing, i.e. the acidic and basic forms, were found as proteins regulated by TGF␤. Both forms of Maspin, which may indicate post-translational modifications, were up-regulated by TGF␤ treatment starting at 8 h and continuing to increase at 24 and 40 h. At 40 h, there was a 65% (1.65-fold average increase; n ϭ 3, p ϭ 0.003) and 67% (1.67-fold average increase; n ϭ 3, p ϭ 0.0069) increase in the acidic and basic forms of Maspin, respectively (Fig. 1A). This time-dependent up-regulation was confirmed in immunoblots of MCF10A cells  1-4) or biotin-labeled Maspin promoter with a mutation at the p53-binding site (p53-I; lanes 5-8) were incubated with nuclear extracts from MCF10A cells that had been treated or not with TGF␤ for 1 h. Where indicated, a 10-fold excess of unlabeled (CAGA) 12 competitor DNA (lanes 3, 4, 7, and 8) was added to the incubation mix. Input nuclear extracts were also loaded (lanes 9 and 10). The presence of Smads and p53 on Maspin promoter DNA was detected in Streptavidin pulldowns subjected to immunoblot analysis using specific antibodies. A c-Jun antibody was used as a negative control. C, MCF10A cells were treated with TGF␤ at the indicated times prior to ChIP. After precipitation with the antibodies indicated to the left, primers encompassing Maspin promoter region from Ϫ154 to ϩ87 were used for PCR amplification. The c-Jun antibody was used as a negative control. To control for equal chromatin input, cell lysates before immunoprecipitation were used as templates for PCR (top four rows). When primers encompassing a 194-bp Maspin coding region were used for PCR, an amplified PCR product was not detectable in any of the ChIP samples (bottom three rows).
TGF␤ Up-regulates Maspin Expression via Smad Pathway-To determine whether Smad signaling is required for TGF␤induced Maspin expression, we tested this induction over a time course in MCF10A cells infected with an adenovirus encoding the inhibitory Smad7 or ␤-galactosidase. Expression of Smad7 abolished TGF␤-induced Smad2 phosphorylation and attenuated the induction of Maspin protein ( Fig. 2A). At the mRNA level, induction of Maspin mRNA by TGF␤ occurred as early as 1 h, and this was also attenuated by forced expression of Smad7 (Fig. 2B).
p53 and Smads Cooperate at the Maspin Promoter to Activate maspin Transcription-It has been reported that the tumor suppressor p53 can physically interact with Smad2 and therefore plays a key role in the cellular response to TGF␤ (25). Since the Maspin promoter contains a p53-binding site that can be transactivated by p53 (16), we speculated that TGF␤ activates maspin transcription by inducing an interaction between Smads and p53 at the Maspin promoter. To test this, we tran-siently transfected 293 cells with luciferase reporter genes driven by various Maspin promoter regions (Fig. 3A). Reporters containing Maspin promoter regions starting from Ϫ759 and Ϫ297 were activated by TGF␤ ϳ8-fold. However, a reporter containing a mutation in the p53-binding site (designated as p53-I) exhibited a markedly impaired response to TGF␤ (Fig.  3A), suggesting that the p53-I site is required for TGF␤-induced maspin gene transcription.
We next confirmed the interaction between Maspin promoter regions and Smads by DAI using biotin-labeled Maspin promoters released from pM-Luc(Ϫ297) and pM-Luc(Ϫ297mtp53) containing a p53 wild-type (wt)-binding site or a mutation at this site (p53-I), respectively. After incubating the labeled probes with nuclear extracts from MCF10A cells, the wt Maspin promoter showed increased binding to Smads and p53 after treatment with TGF␤ (Fig. 3B). TGF␤ did not affect p53 levels in nuclear extracts (Fig.  3B, lanes 9 and 10). This suggests that TGF␤ increases p53 binding to the Maspin promoter by increasing the binding of Smads to this same promoter. Consistent with this speculation, addition of unlabeled competitor DNA, (CAGA) 12 , which specifically binds to Smads, decreased both Smads and p53 binding. Compared with the wt Maspin promoter, less p53 and Smads were bound to the probe mutated in the p53-I site, suggesting that binding of p53 to this site may also stabilize Smads binding. Interestingly, the mutation in the p53-I site decreased but did not completely abolish p53 binding to the Maspin promoter (Fig. 3B, lanes 5 and 6). This suggested that other p53-binding sites, besides p53-I, may be present in Maspin promoter. Indeed, a second p53-binding site (designated as p53-II) was found in subsequent experiments (Fig. 4).
Finally, to prove that TGF␤ induces binding of p53 to the Maspin promoter in vivo through Smads, ChIP over a time course was performed. TGF␤ induced binding of p53 to the Maspin promoter as early as 15 min reaching a maximum at 30 min and 1 h. Detectable maximal binding of Smads to this promoter, as indicated by recovery of the Maspin promoter PCR product in Smad2/3 antibody pulldowns, reached a maximum at 1 h after treatment with exogenous ligand (Fig. 3C, top two  rows). When primers encompassing a 194-bp Maspin coding region were used for PCR, none of the ChIP samples gave amplified bands (Fig. 3C, bottom panel).  1, 6, and 9). Antibody against p53 or Smad4 was used to generate the supershifted bands (lanes 4 and 5). S, specific band; SS, supershifted band. C, schematic representations of the Maspin promoter luciferase reporters and their response to TGF␤. All of the Maspin promoter reporters have the common 3Ј ends (ϩ87) and different truncated 5Ј ends. MCF10A cells were transiently transfected with Maspin promoter reporters and 8 h later treated or not with TGF␤. At 24 h, the cells were harvested and tested for luciferase reporter activity as indicated under "Experimental Procedures." Each bar represents the mean normalized luciferase activity Ϯ S.D. calculated from three wells. D, MCF10A cells were transiently co-transfected with maspin promoter reporters (Ϫ154, Ϫ154mtp53-II, or Ϫ154mtp53-IIS), a p53 plasmid, or a control vector. Eight h post-transfection, cells were treated or not with TGF␤; 24 h later, the cells were harvested and tested in a luciferase reporter assay. Fold activation by p53 was calculated based on the luciferase activity in vector-transfected cells which is set at 1. Each bar represents the mean normalized luciferase activity Ϯ S.D. calculated from three wells.
Mutations of p53 Sites in the maspin Promoter Impair TGF␤induced Transcription-By sequence analysis we identified four potential SBEs and two p53-binding sites in the Maspin promoter region from Ϫ297 to ϩ87 (Fig. 4A). To determine how these sites are involved in p53-mediated induction of Maspin by TGF␤, we constructed a series of Maspin promoterluciferase reporters with various truncated 5Ј ends and point mutations to eliminate the p53 and SBEs (Fig. 4C). The function of the new p53 site (p53-II) that overlaps with the SBE-II site was confirmed by EMSA and transient transfection assays. Incubation of a 32 P-labeled p53-II probe with nuclear extracts from TGF␤-treated MCF10A cells resulted in the formation of a shifted band which underwent supershift upon the addition of p53 or Smad4 antibodies, suggesting that both p53 and Smads directly bind to this promoter site (Fig. 4B, lanes 1-5). In contrast, no shifted bands were observed when probe mtp53-II, with mutations at both p53-II and SBE-II sites, or probe mtp53-IIS carrying mutant p53-II but wild-type SBE-II were used (Fig.  4B, lanes 6 -11).
In cells co-transfected with Maspin reporters and a plasmid encoding p53 cDNA, mutation of the p53-II/SBE-II site (Ϫ154mtp53-II) or p53-II site alone (Ϫ154mtp53-IIS) resulted in 60 -75% reduction of p53-induced reporter activity (Fig. 4D). Transfection of a Maspin promoter construct with a mutation in the p53-I site (Ϫ297mtp53-I compared with Ϫ297; shown in Fig. 4C) into MCF10A cells partially abrogated TGF␤-mediated induction (from 6.7-to 4.0-fold; Fig. 4C), consistent with the experiment shown in Fig.  3A in 293 cells. Mutation of p53-II/ SBE-II site (Ϫ154mtp53-II compared with Ϫ154) also reduced the response to TGF␤ from 6.9-to 3.2fold) (Fig. 4C). The mutant reporter pM-Luc(Ϫ154mtp53-IIS) with mutant p53-II but wt SBE-II also exhibited a lower response (3.1-fold) to TGF␤, suggesting that p53 binding to this site is required. Deletion of the two distal SBEs (Ϫ154 compared with Ϫ297) did not impair the promoter response to TGF␤, whereas the two proximal SBEs, one overlapping with p53-II and the other one adjacent to p53-I, were required for Maspin reporter transcription. Mutation or deletion of either SBE-I or SBE-II (Ϫ154mtSBE-I, Ϫ116, Ϫ116mtSBE-I) also impaired the response to TGF␤ (Fig. 4C).
TGF␤ Fails to Activate Maspin Promoter in p53-mutant Cells-MDA-MB-231 human breast cancer cells harbor a point mutation in codon 280 (G/A) of the p53 gene. We transiently transfected these cells with a Maspin promoter luciferase reporter. Treatment with TGF␤ did not induce Maspin promoter activity but still induced Smad reporter p(CAGA) 12 -Luc (Fig. 5A). Co-transfection with a wt p53 plasmid restored TGF␤-induced Maspin promoter activity (Fig. 5A), suggesting that p53 function is necessary for the induction of Maspin transcription by TGF␤. ChIP in cells transfected with a Maspin promoter and wt p53 showed that both p53 and Smads bound to the Maspin promoter upon treatment with TGF␤.
Due to the p53 mutation and the reported methylation of endogenous Maspin promoter in MDA-MB-231 (26), detectable Maspin protein levels are very low in these cells and are not up-regulated by TGF␤ (Fig. 5C, lanes 1 and 2). To determine the possible significance of low Maspin expression, we stably transfected a Maspin-encoding plasmid into MDA-MB-231 cells (Fig. 5C). Added TGF␤ induced MDA-MB-231 cell motility and invasion in vitro, but these responses were abolished in cells overexpressing Maspin (Fig. 5, D and E). This suggests that loss of Maspin expression is permissive for TGF␤-induced motility in cancer cells such as MDA-MB-231.  1-3). Cells were treated with TGF␤ for 0 min, 30 min, or 1 h prior to ChIP. DNA-protein complexes were precipitated with the antibodies indicated to the left of each panel followed by PCR amplification using primers specific to the maspin promoter as indicated under "Experimental Procedures." To control from chromatin input, whole cell lysates (before the antibody pulldown) were used as templates for the PCR (bottom row). C, MDA-MB-231 cells stably transfected with Maspin or control vector were treated or not with TGF␤ for 24 h. Cell lysates were separated by SDS-PAGE followed by immunoblot analysis with the indicated antibodies. D, close-to-confluent monolayers of MDA-MB-231 cells stably expressing Maspin or control vector were serum-starved for 24 h before wounding. Serum-free medium Ϯ TGF␤ was replenished and wound closure was followed 20 h later as indicated under "Experimental Procedures." E, MDA-MB-231 cells stably expressing maspin or control vector were seeded (2.5ϫ10 4 cells/well) on Matrigel-coated transwells and allowed to invade toward growth medium Ϯ TGF␤. At 24 h, the cells invading into the Matrigel were counted; each bar represents the mean Ϯ S.D. of three wells.
Knockdown of Maspin Facilitates TGF␤-induced MCF10A/ HER2 Motility-We have previously reported that TGF␤ induces the motility of MCF10A/HER2 cells via PI3K-dependent activation of the Rac GTPase and its substrate Pak1 (11,13). Using siRNA against human maspin gene, we knocked down both basal and TGF␤-induced Maspin expression in MCF10A/ HER2 cells (Fig. 6A). In wound closure assays, TGF␤-induced motility was accelerated in cells expressing Maspin siRNA compared with serum-starved cells expressing control siRNA (Fig. 6B). Meanwhile, the number of focal adhesions was significantly decreased in cells expressing Maspin siRNA as detected by immunofluorescence using an antibody against Vinculin (Fig. 6C). In addition, both basal and TGF␤-induced cell invasion in Matrigel were enhanced in cells expressing Maspin siRNA (Fig. 6D). This result further suggests that low levels or absence of the Maspin tumor suppressor protein enhance the tumor promoting effects of TGF␤ in transformed cells.

DISCUSSION
We performed two-dimensional DIGE in MCF10A mammary epithelial cells overexpressing HER2. Treatment with TGF␤ resulted in a time-dependent change in the abundance of over 60 protein species. Among these, two forms of Maspin protein with the same molecular weight but slight variations in their isoelectric points (pI), were characterized. The maspin gene encodes a protein related to the serpin (serine protease inhibitor) family of protease inhibitors. It was first identified as a tumor suppressor gene with a potential role in human breast cancer (14). Maspin protein is expressed in non-tumor mammary epithelial cells but is absent in most breast cancer cell lines (14). In MMTV/TGF␣ transgenic mice, Maspin expression inversely correlates with mammary tumor progression (27). Restoration of Maspin protein expression in MDA-MB-435 breast cancer cells reduces the abilities of these cells to form tumors and metastasize in nude mice (14). Other studies have shown that Maspin acts at the cell membrane to inhibit the invasion and motility of prostate and breast cancer cell lines (28). Treatment of breast cancer cells with exogenous Maspin increases adhesion to fibronectin and inhibits Rac GTPase activity (29). Maspin can be tyrosine-phosphorylated in both normal and tumor mammary cells; recombinant Maspin protein can be phosphorylated by the epidermal growth factor receptor tyrosine kinase in vitro (30), although the biological significance of this modification is unknown. Interestingly, another serpin family member, plasminogen-activator inhibitor (PAI)-1, is strongly induced by TGF␤. In contrast to Maspin, however, both PAI-1 and PAI-2 are overexpressed in malignant tumors (28,(31)(32)(33).
Two major mechanisms regulate Maspin expression: transcription by p53 family members and promoter methylation. Adenovirus-encoded wild-type p53 strongly up-regulates Maspin gene transcription and protein levels in breast and prostate cancer cells as a result of binding of p53 to a p53 consensus site (i.e. the p53-I site shown in Fig. 4A) in the Maspin promoter (16). Consistent with this finding, tissue microarray analyses have shown that Maspin expression is down-regulated in cancers with a high histological grade and low expression correlates with presence of p53 mutations in various tumor types (34). Interestingly, Maspin is overexpressed in lung cancer, suggesting it may have a role other than tumor suppressor (35). p63, a homolog of p53, transactivates the Maspin promoter via directly binding to the p53 site in cancer cells (15). In addition, aberrant cytosine methylation of the Maspin promoter resulting in gene silencing has been found in mammary cell lines including MDA-MB-231 cancer cells but not in the nontumorigenic MCF10A cells (26). Therefore, p53 mutation and promoter methylation can cooperate to silence Maspin expression in cancer cells. Finally, transactivation through an Ets and Ap1 sites as well as transcriptional repression through a negative hormonal responsive element recognized by the androgen receptor are also found in the Maspin promoter (36,  In this study we show that 1) TGF␤ induces Maspin expression through Smads, and 2) wt p53 binding to the Maspin promoter is required for TGF␤-induced transcription. p53 promotes the activation of other TGF␤ target genes during Xenopus embryonic development. p53-deficient mammalian cells display impaired responses to TGF␤ including the induction of p21 WAF1 , PAI-1, and MMP2 as well as growth arrest (25). Based on these findings, a model for cooperation between p53 and Smads has been proposed: p53 associates with Smad2/3 in vivo, while at the same time it contacts its own cognate DNA site on promoters of a subset of TGF␤ target genes to facilitate and/or enhance TGF␤-induced transcription of (25,38). In this model, Smad2/3 enter the nucleus where they associate with Smad4 and specific cofactors to bind target sequences at gene promoters. The presence of a nearby p53binding site and p53 binding to DNA are both necessary for the synergistic transcriptional activation by p53 and Smads. The convergence between p53 and TGF␤ signaling can also co-repress target gene expression such as ␣-fetoprotein. The ␣-fetoprotein promoter contains an overlapping Smad binding and p53 regulatory element (SBE/p53RE), where p53 and Smads cooperate to induce histone deacetylation and DNA methylation to inhibit TGF␤-induced promoter activity (39).
We identified two p53-binding sites in the human Maspin promoter, both required for full response to TGF␤ stimulation. The previously reported site p53-I is only 9 bp away from a downstream SBE, which when mutated will impair the response to TGF␤ (as in pM-Luc(Ϫ154mtSBE-I) and pM-Luc(Ϫ116mtSBE-I); Fig. 4C). A new p53 site identified in this study, designated as p53-II, also overlaps with an SBE and, therefore, allowed the p53:Smad2 interaction. ChIP in MCF10A cells indicated rapid recruitment of p53 and Smads to the Maspin promoter as early as 15 min after the addition of TGF␤, with p53 reaching maximal binding at 30 min and Smad2/3 at 1 h (Fig. 3C). Interestingly, the amount of p53 that binds to the Maspin promoter was almost undetectable 4 h after the addition of TGF␤ (Fig. 3); this was followed by a second increase at 8 h. This second increase in promoter binding was not observed for Smad2/3, suggesting the possibility of another TGF␤-induced signaling pathway contributing to this secondary, p53-dependent activation of Maspin transcription.
The results herein suggest that p53 is required for the cellular response to TGF␤. Data in support of the opposite have also been reported. For example, TGF␤ depletion results in decreased p53 phosphorylation in Ser-18 and an altered cellular response to DNA damage (40), implying that TGF␤ maybe essential for p53-mediated cell fate decisions in situ. Thus, we speculate that convergence between p53 and TGF␤ signaling on the co-regulation of gene expression, such as maspin in this report, may represent a common pattern of cooperation between two tumor suppressive pathways that affect a broader set of targets. These gene targets may not only be related to an antiproliferative response but also, as in the case of Maspin, to the suppression of cell migration and invasion, thus linking the tumor suppressor function of p53 with TGF␤-regulated extracellular cues at the cell surface during transformation.