The Murine Gastrin Promoter Is Synergistically Activated by Transforming Growth Factor-β/Smad and Wnt Signaling Pathways*

The transforming growth factor-β (TGF-β) and Wnt/wingless pathways play critical roles in the specification of cell fate during development and also contribute to cancer formation and progression. Whereas Wnt signaling is clearly pro-oncogenic, TGF-β signaling is cell- and context-dependent, manifesting both inhibitory and proliferative effects. The growth factor, gastrin, has previously been shown to be a downstream target of the Wnt pathway and a promoter of gastrointestinal cancer. In this study, we show that the mouse gastrin promoter is regulated synergistically by TGF-β/Smads and β-catenin/T-cell factor (TCF). Co-transfection of Smad3/Smad4 and β-catenin expression constructs synergistically activated mouse gastrin promoter activity 30–60-fold in AGS cells with minimal effect seen with either construct alone. This activation was further potentiated by TGF-β1 treatment. Mutating either the TCF binding site or the Smad-binding element (SBE) diminished the activation of gastrin expression by Smad3/Smad4 and β-catenin and led to a loss of gastrin promoter responsiveness to TGF-β1 treatment. Wnt and TGF-β regulated endogenous gastrin mRNA levels in AGS cells in a similar fashion, as revealed by small interference RNA studies or overexpression of Smads and TCF4/β-catenin. Electrophoretic mobility shift assays and DNA affinity precipitation assays showed that the putative SBE and T-cell factor (TCF) sites were able to bind a complex containing Smads and β-catenin/TCF4. In addition, the synergy between Smads and β-catenin/TCF4 was dependent on CREB-binding protein (CBP)/P300, as demonstrated by overexpression of CBP or E1A. Moreover, by using a heterogeneous promoter reporter system, we showed that this complex containing Smads/TCF4/β-catenin complex was able to up-regulate transcription at isolated SBE or TCF sites. Thus, the Wnt signaling pathway is able to activate some target genes through its actions as a co-activator at non-TCF sites and has the potential to profoundly alter transcriptional responses to TGF-β signaling.

APC tumor suppressor gene, which forms a complex with axin and glycogen synthase kinase 3␤ (GSK-3␤) 1 to form a ␤-catenin degradation complex, which phosphorylates ␤-catenin, thus targeting the protein for subsequent ubiquitination and degradation (1). In the absence of phosphorylation and degradation, ␤-catenin accumulates in cytoplasm and moves into the nucleus, where it functions as a cofactor for the T-cell factor/ lymphoid-enhancing factor (TCF/LEF) transcription factor family. Whereas it does not bind DNA itself, ␤-catenin is essential for the transcriptional activity of TCF/LEF, which binds to a consensus DNA sequence of 5Ј-CCTTTGTCT-3Ј within promoters of target genes to activate transcription. Wnt signaling is important in embryonic development, maintenance of stem cell populations, and the progression of gastrointestinal and other cancers (2). A number of TCF/LEF target genes have been identified, some of which (including c-myc, cyclin D1, MMP7, EPHB2/3) have important implications in understanding the role of Wnt signaling in cancer, and many additional potential Wnt targets have been identified by large scale DNA microarrays (3).
Gastrin is a gastrointestinal hormone and growth factor that has recently been shown to be a functionally relevant downstream target of the Wnt signaling pathway (4). Incompletely processed gastrins (progastrin and glycine-extended gastrin) are capable of inducing colonic proliferation both in vitro and in vivo (5)(6)(7)(8)(9). Deletion of the gastrin gene results in a decrease in colonic proliferation, and overexpression of incompletely processed gastrins results in increased colonic proliferation, accelerated colonic neoplasia in response to azoxymethane or Apc/ Min-dependent polyposis (10 -12). Incompletely processed gastrins are also up-regulated in colon carcinomas, and several studies have also found that the plasma level of progastrin is significantly elevated in colorectal cancer patients (13). In initial studies, we demonstrated that induced Apc expression was capable of suppressing gastrin gene expression and that overexpression of activated ␤-catenin could stimulate the human gastrin promoter 2-3-fold through a putative TCF-binding site (4). The mechanisms by which ␤-catenin promotes gene transcription are not well understood but are thought to involve interactions with other cofactors.
One pathway that appears to be able to interact with the Wnt signaling pathway and that also plays an important role in cancer, is the TGF-␤ signaling pathway. TGF-␤ was originally named as a transforming growth factor based on its ability to induce malignant behavior by normal fibroblasts (14). It was later recognized that TGF-␤ had profound growth-suppressive effects on epithelial and lymphoid cells and for a time was considered primarily as a tumor suppressor. However, it is now accepted that TGF-␤ signaling has a dual role in cancer; whereas TGF-␤ signaling may act as a tumor suppressor at early stages of carcinoma development, in later stages TGF-␤ signaling contributes to invasion, metastases, and progression of advanced disease (15,16). Thus, whereas the loss of TGF-␤ responsiveness may provide a distinct advantage for developing tumors, only a minority of tumors have genetic alterations in this pathway. Instead, TGF-␤ expression is up-regulated during late stage cancer progression and in most tumors is associated with a loss of TGF-␤ growth inhibition through uncertain mechanisms. TGF-␤ signaling is clearly cell typeand context-dependent, and in cancer TGF-␤ responses appear to be realigned in favor of growth promotion (17).
Members of the TGF-␤ superfamily exert their effects through cell surface receptors that have been well studied (18). The signaling cascade is activated by the dimerization of type II and type I receptors upon the binding of ligands such as TGF-␤. The type I receptor is subsequently phosphorylated and then phosphorylates the receptor-regulated Smads (R-Smads, Smad2, and Smad3) (18). These proteins in turn form a heteromeric complex with Smad4, translocate to the nucleus, and regulate expression of target genes in close association with a variety of transcriptional co-activators and co-repressors (19). Smads typically activate gene transcription, but they are also involved in TGF-␤-mediated repression of transcription. Furthermore, Smads function to regulate transcriptional responses in several distinct ways. They can bind to DNA at a consensus-binding site (GTCT) and directly regulate transcription by recruiting co-activators or co-repressors to the promoter. However, Smads appear to bind to DNA with rather low sequence specificity and affinity, and DNA binding alone by Smads is often insufficient to mediate sequence-specific transcriptional responses. Interactions with other DNA-binding partners are typically required for Smad-dependent transcriptional responses. Thus, Smads have been reported to interact with other sequence-specific DNA transcription factors, and in this context they may function as co-modulators of transcription (18,19).
One link that has received some attention is the interaction between Smads and the Wnt signaling pathway. Previous observations indicated cooperation between Wnt and TGF-␤ pathways in the transcription regulation of a number of target genes. Wnt cooperates with TGF-␤ during embryonic development and tissue maintenance, and both are required for the establishment of a dorsal signaling center, Spemann's organizer (20). Activation of both Wnt and TGF-␤ signaling leads to translocation of DNA-binding proteins to the nucleus, and di-rect interactions between the downstream components has been demonstrated. Importantly, LEF1/TCF and ␤-catenin have been shown to form a specific complex with Smad3 or Smad4 (20 -22). Studies have reported that Smad3 and Smad4 can associate with ␤-catenin in the renal proximal tubule cell line, HK-2 (23). In Xenopus, the interaction between LEF1/␤catenin and Smad3/Smad4 synergistically activates expression of the twin (Xtwn) gene during Spemann's organizer formation (20). Labbe et al. (21) showed that this activation results from a physical interaction between Smad3 and the HMG box domain of LEF1. In this model promoter, Smad binding to DNA at an SBE site stabilized the transcriptional activation complex of ␤-catenin/LEF1 attached to an adjacent TCF site. Importantly, activation of TCF sites by TGF-␤ signaling was completely dependent on the presence of adjacent Smad-binding sites (21,24). However, the requirement for adjacent sites has been challenged by reports that ectopic expression of Smad3 and Smad4 could activate LEF/TCF-specific reporter genes and that ectopic LEF1 expression could weakly induce SBE-specific reporter activity (25).
Given the recognized role of gastrin as a growth factor for colon cancer and its reported regulation by the Wnt pathway, we explored the possible role of TGF-␤ signaling in the modulation of gastrin gene expression. Our results indicate that the murine gastrin gene is synergistically transactivated by TGF-␤/Smads and Wnt/␤-catenin at the transcription level. Furthermore, the results support the notion that Wnt signaling may function as a co-activator, transactivating genes through Smad-binding sites.
On the day before transfection, 1 ϫ 10 5 cells were seeded into 12-well plates. 0.3 g of reporter construct and 0.3 g of each expression construct were transfected using Superfect transfection reagent (Qiagen, Valencia, CA) according to manufacturer's instructions. 0.005 g of the promoter TK-Renilla luciferase construct was transfected as internal control. The total DNA amount was kept at 1.5 g/well. Corresponding empty vectors were used to make up the difference between treatments. Luciferase activity was measured 48 h later using a Monolight TM 3010 Luminometer (PharMingen, San Diego, CA). The dual luciferase assay kit was purchased from Promega (Madison, WI). All experiments were done by triplicates and performed at least three independent times. All results were shown as relative luciferase activity and represented the mean Ϯ S.E. For TGF-␤ treatment, cells were starved in Dulbecco's modified Eagle's medium without fetal bovine serum for 24 h and then cultured for another 24 h after the addition of 3 ng/ml TGF-␤1 (Sigma).
Plasmids-The promoter luciferase reporter constructs of murine gastrin were generated by PCR (see Table I for primer sequences) and digested with SmaI and BamHI and then ligated into pXP2 (26) predigested with the same enzymes. The mutation of putative LEF/TCF or  (26). The integrity of all constructs was verified by DNA sequencing. Cytomegalovirus early promoter-driven 5Ј FLAG-tagged human Smad3 and Smad4 mammalian expression vectors, pCF-S3 and pCD-D43F(R), were gifts from Dr. Shioda (27,28). Constitutively active human ␤-catenin (S37A) and wild type human TCF4 expression vectors were from Upstate Biotechnology, Inc. (Lake Placid, NY). The reporter construct p3TP-luc (29) and mammalian expression vectors for p300 (30) and E1A (31) have been described previously.
Coimmunoprecipitation and Western Blotting-AGS cells were cotransfected with FLAG-tagged Smad3 or Smad4 or both along with ␤-catenin and TCF4 expression constructs. Cells were harvested 48 h later and lysed in RIPA buffer (150 mM NaCl, 1% Nonidet P-40 (Nonidet P-40), 0.5% deoxycholate, 0.1% SDS, and 50 mM Tris-HCl, pH 8.0). 2000 g of proteins from each preparation was incubated with 0.5 g of appropriate immunoprecipitation antibody in 500 l of immunoprecipitation buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 1ϫ proteinase inhibitor mixture) for 1 h at 4°C with gentle rotation. Incubation was continued overnight after the addition of 30 l of protein-A Sepharose flurry (Amersham Biosciences). Protein A-precipitated protein complexes were recovered by brief centrifugation and washed five times with immunoprecipitation buffer and then disassociated by adding 40 l of 2ϫ SDS sample buffer and boiling for 5 min. The proteins were detected by Western blot using the indicated antibodies. Anti-␤-catenin and anti-FLAG antibodies were purchased from Cell Signaling (Beverly, MA). Anti-Smad2/3 and anti-TCF4 antibodies were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA).
Preparation of Nuclear Extract and Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts from TGF-␤1 or vehicle-treated AGS cells were prepared as previously described (32,33). For EMSA, 32 Plabeled double-stranded DNA oligomers corresponding to the murine gastrin SBE2 (5Ј-TGG GGA GTC TGG CCT CAC CTG GAA) and the TCF binding site (5Ј-TTC CAG GCC CAT TTC TCT TGC TGT GGG) were incubated with 6 g of nuclear extracts on ice for 20 min. For competition assays, 100ϫ cold wild type or mutant oligonucleotides (mutant SBE2: 5Ј-TGG GGA tTC TGt CCT CAC CTG GAA; mutant TCF site: 5Ј-TTC CAG GCC taa TTC TCT TGC TGT GGG, with mutated nucleotides in lowercase) were mixed with 32 P-labeled probes and then incubated with nuclear extracts from TGF-␤1-treated AGS cells. For supershift assays, the antibody against Smad3, Smad4, ␤-catenin, or TCF4 or control antibody against cyclin D1 was added to the mixture individually and incubated on ice for 30 min before resolving in 8% nondenaturing polyacrylamide gel.
DNA Affinity Precipitation Assays (DAPA)-DNA Affinity Precipitation Assays were performed essentially as previously described (34,35). Briefly, 5Ј-biotin end-labeled sense and antisense oligonucleotides corresponding to the wild type or mutant Smad-or TCF/LEF-binding site (see above) of murine gastrin promoter were custom synthesized by Invitrogen. The oligomers were annealed and purified as previously described (35). 50 g of nuclear extracts from TGF-␤1-treated AGS cells were incubated with 0.25 g of wild type or mutant oligonucleotide probe in 400 l of binding buffer (60 mM KCl, 12 mM HEPES, pH 7.9, 4 mM Tris-HCl, 5% glycerol, 0.5 mM EDTA, 1 mM dithiothreitol) on ice for 1 h. The DNA-protein complexes were then incubated overnight at 4°C with gentle rotation with 40 l of Tetralink TM Avidin Resin (Promega, Madison, WI), which was washed and then pre-equilibrated with binding buffer for 1 h. Avidin-precipitated DNA-protein complexes were washed five times with binding buffer and then disassociated by adding 40 l of 2ϫ SDS sample buffer and boiling for 5 min. The proteins were resolved on polyacrylamide gel and followed by immunoblotting using specified antibodies.
Small Interference RNA (siRNA) and Real Time PCR-Doublestranded RNA oligonucleotides for RNA interference of control (enhanced green fluorescence protein), Smad3, and TCF4 were custom-synthesized by Qiagen (Valencia, CA) and transfected into AGS cells using Oligofectamine (Invitrogen) according to the manufacturer's instruction. Total RNA was extracted from transfected cells by the Trizol method, and cDNA was synthesized using the Superscript first-strand reverse transcriptase-PCR kit (Invitrogen). mRNA levels of gastrin and the RNAi targeting genes were quantified by real time PCR using the SYBR Green kit (Qiagen) and the ⌬⌬ TM method. The oligonucleotide sequences for real time PCR are as follows: human gastrin, 5Ј-GCG CTG GCC GCC TTC TCT GA (sense) and 5Ј-CAT CCA TAG GCT TCT TCT TCT TCC (antisense); GAPDH, 5Ј-GAC ATC AAG AAG GTG GTG AAG C (sense) and 5Ј-GTC CAC CAC CCT GTT GCT GTA G (antisense); Smad3, 5Ј-GGC CAC CGT CTG CAA CAT C (sense) and 5Ј-TAC TGG TCA CAG TCT GTC (antisense); TCF4, 5Ј-TGA AGG CAG CTG CCT CAG C (sense) and 5Ј-GTG GGT GGC CTC AGC GAG C (antisense).

The Murine Gastrin Promoter Is Synergistically Activated by
Overexpression of Smad3/Smad4 and ␤-Catenin-Previous studies by our group (4) as well as others (36) have demonstrated that gastrin is a downstream target of the Wnt signaling pathway. However, TGF-␤ is also up-regulated during colon cancer progression and has been shown to synergize with the Wnt pathway (20 -22, 24, 25). Recent reports suggest that there are important differences between genes that are acti-FIG. 1. The murine gastrin promoter is synergistically activated by Smad3/Smad4 and ␤-catenin. The murine gastrin promoter (1 kb to ϩ50 bp) construct pmGAS1kb was transfected alone or in combination with expression constructs for Smad3/Smad4 and/or ␤-catenin (see "Materials and Methods"). For TGF-␤ stimulation, cells were serum-starved for 24 h and then treated with or without 3 ng/ml TGF-␤1 for another 24 h. Reporter firefly luciferase activity was normalized to internal control Renilla luciferase activity and expressed as relative luciferase activity. All assays were performed in triplicate and repeated at least three times. Mean Ϯ S.E. is shown. vated by Wnt signaling alone and those that are activated by both TGF-␤ and Wnt pathway (37). In order to understand the regulation of gastrin transcription by oncogenic pathways, we examined the role of these pathways in regulation of the murine gastrin promoter. A 1-kb fragment of murine gastrin promoter containing 50 base pairs downstream of the transcription initiation site was joined to the luciferase reporter gene (pmGAS1kb) and used in transient transfection reporter assays. The reporter construct was transfected into AGS cells alone or co-transfected in different combinations with overexpression constructs for Smad3, Smad4, and constitutively active ␤-catenin at a ratio of 1:1. Whereas co-transfection of ␤-catenin alone had minimal effects on the murine gastrin promoter, and Smad3/Smad4 resulted in a modest 3-5-fold promoter activation, the combination of all three (␤-catenin/ Smad3/Smad4) resulted in a 30 -60-fold activation of the murine gastrin promoter (or Ͼ10-fold synergy) (Fig. 1A). The transfected cells were also treated with 3 ng/ml TGF-␤1 or left untreated to evaluate the responsiveness of murine gastrin to TGF-␤. TGF-␤ treatment of the transfected cells resulted in a 2-fold stimulation of the 1-kb murine gastrin promoter, and the activation was potentiated 4-fold by cotransfection of Smad3/ Smad4 but unaffected by co-transfection of ␤-catenin (Fig. 1B). Overall, these initial studies suggested a strong synergistic interaction between TGF-␤/Smad and Wnt pathways at the murine gastrin promoter.
Regulation of Gastrin Gene Expression in Vivo by TGF-␤/ Smads and Wnt/␤-Catenin Signaling-In order to further validate our results with reporter genes, we examined the effects of TGF-␤ and Wnt signaling on endogenous gastrin gene expression in AGS cells. We employed two separate approaches. First, we evaluated endogenous gastrin mRNA level after transient transfection of overexpression constructs for Smad3/ Smad4 and TCF4/␤-catenin in various combinations. Quanti-

FIG. 2. Endogenous gastrin expression is regulated by Smad3/4 and TCF4/␤-catenin.
A, semiquantitative reverse transcriptase-PCR (top) of gastrin mRNA relative to GAPDH in transfected AGS cells. AGS cells were transfected with expression constructs for Smad3/Smad4, ␤-catenin, and/or TCF4 and then treated with or without TGF-␤1 for 24 h. The bottom shows the gastrin mRNA levels, relative to GAPDH and the untreated control, measured by quantitative real time PCR. The asterisks indicate that overexpression of the indicated constructs resulted in a significant increase (p Ͻ 0.05) over the control, which was transfected corresponding empty vectors. B, Smad3, TCF4, and control siRNA oligonucleotides were transfected into AGS cells using oligofectamine. Gastrin, Smad3, and TCF4 mRNA level was measured by real time PCR (see "Materials and Methods") and shown in percentage of control.
tative real time PCR showed that each of the combinations increased gastrin mRNA level relative to the GAPDH control, with the greatest increase observed with transfection of all four constructs (3-fold without and 5-fold with TGF-␤1 treatment) ( Fig. 2A). Simultaneous TGF-␤ stimulation resulted in a significant increase in gastrin mRNA level ( Fig. 2A). Second, we introduced gene-specific double-stranded siRNA oligonucleotides by transient transfection into AGS cells in order to knock FIG. 3. cis-Elements in the murine gastrin promoter mediate TGF-␤/Smad-and ␤-catenin-dependent activation. A, deletion analysis of the murine gastrin promoter. 5Ј deletion gastrin promoter-reporter gene constructs (Ϫ1 kb, Ϫ500, Ϫ140, Ϫ80, Ϫ55, and Ϫ30) were individually cotransfected into AGS cells with or without expression constructs for Smad3/Smad4 and ␤-catenin. Luciferase activity was normalized to that in cells transfected with vector alone and is shown as relative luciferase activity. On the right is shown the response to 3 ng/ml of TGF-␤1. B, sequence of the murine gastrin promoter. The location of the putative murine gastrin TCF/LEF-binding site, the Smad-binding elements SBE-1 and SBE-2, and the TATA-box are indicated. C, mutation of TCF or SBE2 sites inhibits gastrin promoter activation. Wild type (pmGAS140) or mutant reporter constructs containing block mutations in the TCF or SBE2 site (see "Materials and Methods") were cotransfected into AGS cells along with expression constructs for Smad3/Smad4 and ␤-catenin.
down the endogenous level of gene expression for Smad3, TCF4, or both (Fig. 2B). An enhanced green fluorescence protein siRNA served as a control. We initially tried to measure the endogenous Smad3 protein level by Western blot. However, commercially available antibodies were not sufficiently sensitive to detect endogenous Smad3 protein in AGS cells (data not shown). Therefore, we measured the abundance of the corresponding mRNA by quantitative reverse transcriptase-PCR. The mRNA level of Smad3 was decreased by more than 90%, whereas the TCF4 mRNA level was inhibited by more than 60% by their specific siRNAs, respectively (Fig. 2B, left and  middle panels). Accordingly, the mRNA level of gastrin was decreased more than 45% by Smad3 siRNA and about 25% by TCF4 siRNA. A decrease more than 55% of gastrin mRNA level was observed when the cells were exposed to both Smad3 and TCF4 siRNAs (Fig. 2B, right panel). Taken together, these results support the notion that Wnt and TGF-␤ pathways regulate the endogenous gastrin gene in the AGS gastrointestinal cancer cell line.
Both the TCF-binding Site and Smad-binding Element in the Murine Gastrin Promoter Can Mediate TGF-␤/Wnt-dependent Activation-In order to identify the cis-acting DNA elements responsible for Wnt-and TGF-␤-dependent activation of the murine gastrin promoter, a series of 5Ј deletion luciferase reporter constructs (containing 500, 140, 80, 55, and 30 bp of 5Ј-flanking sequence) were generated by PCR and cloned into the BamHI and the SmaI sites of pXP2 (26). Co-transfection of these reporter constructs with Smad3/Smad4 and ␤-catenin expression constructs into AGS cells showed that the cis-elements mediating response to ␤-catenin and Smad3/Smad4 included elements that were localized between Ϫ140 and Ϫ80 and between Ϫ80 and Ϫ55 (Fig. 3A). Deletion of these elements resulted in a significant loss of promoter activity and responsiveness to TGF-␤ under conditions of stimulation with ␤-catenin and Smad3/Smad4 (Fig. 3A). Interestingly, a putative TCFbinding site (5Ј-CCATTTCTCT) was identified between Ϫ140 and Ϫ80 region, an SBE (5Ј-GTCTG) was found between Ϫ80 and Ϫ55 (Fig. 3B), and another SBE was located downstream of the TATA-box. The putative TCF/LEF binding site in the murine gastrin promoter showed a high degree of homology to the consensus TCF-binding site (CC(A/T)TTG(A/C)TCT) (38).
To test whether those cis-elements were important in conferring TGF-␤/Wnt activation of murine gastrin expression, mutational analysis was carried out in the Ϫ140 bp mGAS-luc construct, pmGAS140. Mutating either the TCF-binding site or the distal Smad binding element (SBE2) was found to significantly diminish the activation of the murine gastrin promoter by the TGF-␤/Wnt pathways (Fig. 3C). Mutating both the TCF site and SBE2 did not result in any additional loss of responsiveness (data not shown). Surprisingly, block mutation of the TCF-binding site also greatly reduced the responsiveness of the gastrin promoter to TGF-␤, quite similar to the effect of mutating the distal Smad-binding element (Fig. 3C).
Smad Proteins and TCF4/␤-Catenin Bind to cis-Elements to Activate Murine Gastrin Expression-To further investigate the cis-elements and the nuclear factors involved in the activation of murine gastrin by TGF-␤/Smad and Wnt pathway, the murine gastrin Smad-binding element (SBE2) and the TCF-binding site were individually cloned into the enhancerless thymidine kinase-luciferase reporter vector pT81 to generate the heterologous promoter constructs pT81-SBE2 and pT81-TCF. These constructs were transfected into AGS cells in combination with overexpression constructs for ␤-catenin, Smad3/Smad4, or both. Both the SBE2 and the TCF binding site showed the ability to respond to TGF-␤/Smads and ␤-catenin but with somewhat different patterns. The Smad-binding element (SBE2) was activated about 30-fold by Smad3/Smad4 complex alone and more than 50-fold by the combination of Smad3/Smad4 plus ␤-catenin (Fig. 4A). In contrast, the putative TCF-binding site activated very weekly (3-4-fold) by Smad3/Smad4 alone but again was strongly activated (30-fold) when overexpression constructs for Smad3/Smad4 and ␤-catenin were co-transfected (Fig. 4B). Taken together, these results indicate that both the SBE and TCF binding site in the gastrin promoter were responsive to TGF-␤ and Wnt pathway signals in a context-dependent fashion.
We next performed EMSAs to evaluate the ability of these cis-elements from the murine gastrin promoter to recruit nuclear proteins involved in the TGF-␤ and Wnt pathways. Both SBE2 and TCF/LEF sites recruited nuclear protein complexes that were strongly enhanced by TGF-␤ treatment (Fig. 5, A and  B, lanes 2 and 3). Antibody-specific supershifts (ss) revealed that Smad3 (and Smad4), TCF4, and ␤-catenin were present in those complexes (Fig. 5, A and B). The control cyclin D1 antibody did not result in any detectable supershifted band (Fig. 5,  A and B). Furthermore, EMSAs performed with either the SBE2 or TCF/LEF element as a probe resulted in closely re- FIG. 4. The SBE2 and TCF/LEF-binding sites mediate gastrin promoter activation. A, co-transfection of expression constructs for ␤-catenin and Smad3/4 with a heterologous construct containing a single copy of the murine gastrin SBE2 binding site (pT81-SBE2). B, co-transfection of expression constructs for ␤-catenin and Smad3/4 with a heterologous construct containing a single copy of the murine gastrin TCF binding site (pT81-TCF). lated complexes, since competition with either wild type (but not mutant) element was able to compete away the binding (Fig. 5C).
In order to further verify the binding of Smads, TCF4, and ␤-catenin to the SBE2 and TCF binding site of murine gastrin promoter, DAPA were performed. After incubating either the 5Ј-biotin-labeled SBE2 or TCF binding site with TGF-␤1treated AGS cell nuclear extracts, avidin beads were used to precipitate the DNA-protein complexes. In these DAPA assays, both the SBE2 and TCF elements from the murine gastrin promoter were able to pull down Smad proteins, ␤-catenin, and TCF4 (Fig. 5D).
TCF4 Is the Mediator of the Cross-talk between TGF-␤/Smad and Wnt Pathways-It is generally appreciated that ␤-catenin alone is unable to transactivate eukaryotic promoters and typically requires interactions with the TCF/LEF family transcription factors in order to act on its target genes (3). In order to investigate the role of TCF4 in the activation of murine gastrin expression by TGF-␤ and Wnt pathways, we co-transfected a human TCF4 mammalian expression construct with Smad3/ Smad4 and/or ␤-catenin in combination with the Ϫ140 bp murine gastrin-luciferase reporter construct pmGAS140. Overexpression of TCF4 dramatically enhanced the activation of the murine gastrin promoter by Smad3/Smad4 and/or ␤-catenin (Fig. 6A, last lane). Interestingly, the combination of TCF4 and Smad3/Smad4 resulted in a very similar level of activation of the murine gastrin promoter as the combination of ␤-catenin plus Smad3/Smad4 (Fig. 6A). However, the combination of FIG. 5. Smads, ␤-catenin, and TCF4 bind to cis-acting elements in mouse gastrin promoter. EMSA using 32 P-labeled gastrin (A) SBE2 probe or (B) TCF probe, and nuclear extracts (NE) prepared from AGS cells with (ϩ) or without (Ϫ) TGF-␤ treatment. The specific shifted band (s) is indicated with the lower arrow, as well as the specific supershifted band (ss) generated with the specified antibodies against Smad3, Smad4, TCF4, and ␤-catenin. The control antibody against cyclin D1 resulted in no supershifted band. C, cross-competition between murine gastrin SBE2 and TCF probes. On the left is an EMSA study with SBE2 as a probe, and on the right is an EMSA study with TCF as a labeled probe. The cold competitors (100ϫ) included wild-type SBE2 (S), mutant SBE2 (mS), wild-type TCF (T), and mutant TCF (mT). D, DAPA reveal that both SBE and TCF/LEF site are able to bind Smad proteins and the TCF4/␤-catenin complex. Shown are immunoblots of avidin-precipitated DNA-protein complexes using antibodies specific for ␤-catenin, Smad3, or TCF4. The biotin-labeled DNA probes used corresponded to the wild-type or mutant SBE2 and TCF sites in the murine gastrin promoter. The input represented 10% of nuclear extracts used in each DAPA assay.
The heterologous promoter constructs containing an isolated copy of the SBE2 or TCF binding site provided further insight into the manner of transactivation by TGF-␤ and Wnt signaling. Co-transfection of TCF4, while on its own showing little or no effect, in combination with Smad3/Smad4 and ␤-catenin drastically enhanced the activation of both the SBE2 and TCF binding site (Fig. 6, B and C). However, TCF4 acted somewhat differently on those two sites. The heterologous construct containing an isolated TCF binding site was able to respond to Smad3/Smad4 only in the presence of co-transfected TCF4, whereas SBE2 showed the same level of activation by Smad3/ Smad4 with or without TCF4. In both cases, however, TCF4 was able to significantly augment the activation observed in the presence of Smad3/Smad4 and ␤-catenin (Fig. 6, B and C).
TCF4/␤-Catenin Complex Acts as a Co-activator of Smads-A number of studies have reported that the Wnt pathway can cooperate with the TGF-␤ pathway through the TCF/LEF proteins, which can interact with both ␤-catenin and Smad proteins (21,24). However, in this model, Smad3/Smad4 act as co-activators through binding to an SBE immediately adjacent to a TCF binding site. Thus, we examined the possibility that TCF4/␤-catenin complex might act as a co-activator for Smad proteins in the absence of its cognate binding site. When pT81-SBE2 was used in transient co-transfection studies, the TCF4/ ␤-catenin complex synergized with the Smad3/Smad4 complex to up-regulate its promoter activity (Fig. 6C). Another promoter reporter construct p3TP-luc (29), which contains three copies of the SBE and has previously been described as a TGF-␤-responsive element, was used to verify the generality of this observation. The promoter activity of p3TP-luc was activated 20-fold by co-transfection of Smad3/Smad4, whereas the addition of TCF4/␤-catenin (which showed no effect on their own) resulted in a 130-fold increase over basal activity, indicating again strong synergy between Smad3/Smad4 and TCF4/␤-catenin complexes (Fig. 7A). This activation was further enhanced by stimulation with exogenous TGF-␤ with 400-fold stimulation observed in combination with co-transfection of Smad3/Smad4, TCF4, and ␤-catenin (Fig. 7A).
Our promoter studies in transiently transfected cells strongly suggested the likelihood of a protein complex that contained Smad3/Smad4, ␤-catenin, and TCF4. Consequently, a co-immunoprecipitation assay was carried out to detect the interaction between these proteins. After overexpression of FLAG-tagged Smad3 and Smad4 (separately or together) along with TCF4 and ␤-catenin, anti-FLAG or anti-␤-catenin antibody was used to precipitate the interacting proteins. Both Smads and ␤-catenin were able to pull down TCF4 (Fig. 7B). These pull-down studies strongly supported the idea that the heteromeric TCF4/␤-catenin complex could act as co-factor for Smad proteins in their regulation of gene expression.
Since Smad proteins and ␤-catenin both have been shown to interact with co-activator p300/CBP (39,40), we investigated whether p300 played a role in murine gastrin promoter activation by Smads and ␤-catenin. Co-transfection of expression constructs for Smad3/Smad4 resulted in a 4 -5-fold increase of murine gastrin promoter over basal activity. The addition of expression constructs for p300 or ␤-catenin resulted in a more than 60-fold increase of murine gastrin promoter activity. This activation was abrogated by co-transfection of an expression construct for E1A, which is known to physically interact with p300/CBP and block its activation of specific target genes (Fig. 7C).

DISCUSSION
Previous studies have suggested the possibility of synergistic interactions between Wnt and TGF-␤ signaling (20 -22, 25), but the relevance of this interaction to cancer has not been clear. Furthermore, whereas gastrin is known to be expressed in a variety of epithelial cancers (9), the mechanism for the regulation of gastrin gene expression in cancer has been poorly understood. In this study, we show strong synergy between FIG. 7. The TCF4/␤-catenin complex serves as a co-activator for Smad3/4 at SBE sites. A, synergistic activation of the p3TP-luc constructed by co-expression of TCF4/␤-catenin. Co-transfection of TCF4 and ␤-catenin constructs enhanced the activation of Smad3/Smad4 on p3TPluc reporter construct. Treatment with TGF-␤1 further enhanced the promoter response. B, immunoprecipitation of Smads with TCF4. FLAG-tagged Smad3 and/or Smad4 and ␤-catenin/TCF4 were expressed in AGS cells. Cell lysates were immunoprecipitated with anti-FLAG antibody or with anti-␤-catenin antibody. The symbol IgG represents immunoprecipitation with unrelated mouse monoclonal anti-cyclin D1 (left) or rabbit polyclonal anti-integrin ␤1 (right) antibodies. Proteins were resolved on SDS-polyacrylamide gels and immunoblotted with the specified antibody. The input represented 10% of the proteins used in each immunoprecipitation assay. C, gastrin promoter activation is dependent on p300. The murine gastrin promoter-luciferase construct pmGAS140 was transfected into AGS cells along with various combinations of Smad3/Smad4, ␤-catenin, p300, and E1A. Wnt and TGF-␤ signaling in the regulation of the murine gastrin promoter in AGS cells. This cooperative regulation by Smad3/Smad4 and ␤-catenin/TCF4 could be shown both for transfected promoter constructs and for the endogenous gastrin gene. An additional 8-fold regulation of the gastrin promoter by treatment with TGF-␤1 could be shown to be mediated mainly through Smad proteins. The cis-acting DNA elements mediating the response to Smads and ␤-catenin/TCF4 could be mapped to a putative upstream SBE and TCF binding site. However, each of these sites showed binding to a Smad/ ␤-catenin/TCF4 complex, supporting the notion of direct interactions between the ␤-catenin/TCF4 and Smad complexes. Furthermore, when cloned upstream of a heterologous promoter, both the SBE and TCF sites showed synergistic activation by transfection of Smads plus ␤-catenin/TCF4. Finally, the synergistic interaction between Smads and ␤-catenin/TCF was strongly dependent on CBP/p300.
We have previously reported that gastrin is a downstream target gene of the TCF4/␤-catenin complex (4), but the possible modulation by other oncogenic pathways was not investigated. Furthermore, the activation of the human gastrin promoter by ␤-catenin alone was somewhat weak, with only a 2-fold activation in this prior study. However, whereas activation of the Wnt pathway may be one of the earliest changes in colorectal cancer pathogenesis, many additional pathways are activated during progression of colon cancer. Oncogenic Ras can also induce gastrin gene expression in colon cancer (41,42), and preliminary studies from our group indicate that K-Ras can also synergize with Wnt signaling to stimulate gastrin gene expression (43). TGF-␤ expression is typically up-regulated in most epithelial cancers and is thought to have pro-oncogenic effects, contributing to cancer invasion and metastases (17). Patients with HNPCC that have lost TGF-␤ receptor expression actually show a more favorable prognosis than patients where receptor expression is intact. The current findings suggest that the combination of TGF-␤ and Wnt signaling results in a strong synergistic promotion of gastrin gene expression. The ability of both amidated and incompletely processed gastrins to stimulate cell proliferation is well documented (10), and increased serum levels of progastrin-derived peptides have been noted in patients with colorectal cancer (13,44,45). The role of gastrin as an epithelial growth factor and promoter of malignancy has been demonstrated not only for colon cancer but also stomach, pancreatic, and lung cancer (46). The link between Wnt and TGF-␤ signaling to gastrin up-regulation would suggest that this synergistic interaction may be very relevant for cancer pathogenesis. Whereas many studies have investigated genes that are targets of specific oncogenic signaling pathways, insufficient attention has been paid to possible convergence or synergy between these pathways. In our study, we demonstrated the existence in gastric adenocarcinoma AGS cells of a complex composed of Smad proteins, TCF4, and ␤-catenin. Several published studies have recently reported on interactions between Wnt and TGF-␤ signaling (21,22,25,47,48). The TCF/LEF members have emerged as the integral players in coordinating Wnt and TGF-␤ signaling pathways. Whereas ␤-catenin cannot bind DNA itself but instead relies on interactions with TCF/ LEF factors to regulate target genes (2,49), Smad3 or Smad4 has also been shown to interact with LEF1 or TCF4 (20,21,25). A role for Smad3/Smad4 in the Wnt signaling pathway was first demonstrated for the Xenopus Xtwn gene, and in these studies it was suggested that a specific interaction between Smad3/Smad4 and TCF/LEF was responsible for the synergistic activation of the Xtwn promoter (20, 21). Hussein et al. (22) showed that the promoter of the homeobox-containing gene msx2 could also be cooperatively activated by Wnt/␤-catenin and Smad4-dependent pathways. Activation of the msx2 promoter by Wnt and Smads was dependent on p300/CBP, a finding similar to our own results with the murine gastrin promoter. The recruitment of CBP to the promoter would theoretically allow it to acetylate nearby histones and thereby promote transactivation. We also show in this study that cotransfection of CBP with Smads and ␤-catenin/TCF4 resulted in strong synergistic activation of the murine gastrin promoter. These findings support the hypothesis that interactions between Smads and ␤-catenin/TCF facilitate the recruitment of the co-activator of p300/CBP, which is usually presented in limited amounts and is required for promoter activation.
However, one of the remarkable findings of our studies was that overexpression of ␤-catenin and TCF4 could transactivate promoters that contained SBE but in the absence of TCF binding site. Whereas the activation of SBE might be due to the presence of a nearby cryptic TCF site, several lines of evidence strongly argue against such a presumption. First, using standard homology comparisons (including both computer-based searches and manual checks), we could not identify any TCFlike site within the DNA fragment of our murine SBE oligonucleotide sequence. Second, our functional studies confirmed our sequence analysis; we did not observe any activation of the gastrin SBE2-containing reporter after co-transfection and overexpression of TCF4/␤-catenin factors (Fig. 6C, fifth bar). This contrasts with the strong activation observed of the Top-Flash reporter that is known to contain the consensus TCF4 binding site (data not shown). Thus, this extremely sensitive reporter assay does not give any evidence for the existence of a functional TCF site within the murine gastrin SBE2 site. Third, a large excess of the mutated SBE2 oligonucleotide does not compete away complex formation with the radiolabeled wild-type TCF oligonucleotides (Fig. 5C, right). These data would suggest that the mutated SBE2 oligonucleotide cannot bind TCF4/␤-catenin complexes, thus excluding the presence of The interaction between the Smad3/4 and ␤-catenin/TCF complexes is mediated by TCF4, but this interaction could occur at adjacent TCF and SBE sites (shown in A), at isolated TCF sites (shown in B), or at isolated SBE sites (shown in C). In B, the Smad3/4 complex would be functioning as a co-activator, whereas in C, the ␤-catenin/TCF4 complex would be acting as a co-activator. Additional recruited cofactor p300/CBP, which is known to interact with both Smad3/4 and ␤-catenin, is also shown. adjacent TCF site on SBE oligonucleotide. Finally, we observed a similar transactivation by Wnt signaling for another SBE site located in the promoter of the plasminogen activator inhibitor gene I (p3TP-luc reporter; Fig. 7A). The p3TP-luc reporter sequence also does not contain an adjacent TCF-like site. Thus, it is highly improbable that two unrelated SBE sites from distinct genes that respond to Wnt signaling would contain adjacent "cryptic" TCF-like sites.
Previous studies by Labbe et al. indicated that activation of promoters by Wnt and TGF-␤ required the presence of adjacent TCF binding site and SBE, with no response observed in constructs lacking TCF sites (21). However, Eger et al. (25) reported that ectopic LEF-1 expression was able to weakly activate TGF-␤-specific reporter genes. The reason for these discrepancies is not clear but may relate to either the cell type or the specific promoter constructs used. Nevertheless, our results clearly show that transfection of ␤-catenin/TCF4 is able to transactivate SBE-dependent promoters in the absence of a TCF site. This suggests a model in which a complex containing ␤-catenin/TCF4/Smad3/Smad4 could bind to both SBE and TCF sites and interact to attract additional co-factors such as CBP to activate transcription (Fig. 8). In the case of the murine gastrin gene, disruption of any one of the sites would greatly diminish but not absolutely abrogate the effects of orchestrated action of TGF-␤/Smad and Wnt/␤-catenin. Thus, this is the first study to suggest that the heteromeric ␤-catenin/TCF4 complex is able to transactivate genes through its actions as a coactivator for genes that lack TCF sites.
It was previously reported that there exist both positive and negative effects of CBP/p300 on TCF mediated transcription (40,50). In our reporter assay, p300 strongly up-regulated the gastrin promoter in AGS cells, apparently via previously described interactions with ␤-catenin/TCF and Smad3/4 partners (39,40). The presence of these protein-protein interactions makes it reasonable to suggest a stabilizing role of p300 in formation of the activated ␤-catenin/TCF4-Smad3/4 complex on DNA, especially in the absence of a DNA binding site for one of the partners (Fig. 8, B and C).
Eger et al. (25) noted the importance of cooperation between ␤-catenin and TGF-␤ signaling in carcinogenesis using a model of mouse mammary epithelial cells (EpH4) stably expressing a fusion protein of the transcription factor c-Fos and the hormone-binding domain of the estrogen receptor (FosER). In this model, induction of FosER expression leads to increased ␤-catenin/TCF/LEF signaling, which cooperates with TGF-␤ signaling to maintain an undifferentiated mesenchymal phenotype (25). The ability of Wnt signaling to modulate TGF-␤ signaling suggests a number of possibilities with regard to the "dual effects" that have been described for TGF-␤. Whereas TGF-␤ has growth-suppressive effects on most epithelial cell types, these growth-inhibitory effects are lost during cancer progression. The up-regulation of TGF-␤ expression and the development of resistance to TGF-␤-dependent growth inhibition is a hallmark of invasive cancer (15)(16)(17), but the phenomenon is not well understood. Whereas some of this TGF-␤ "resistance" has been attributed to activating Ras mutations (51), the observation that Wnt signaling can synergize with TGF-␤ to stimulate growth factor genes such as gastrin raises the possibility that resistance could be attributed in part to activation of the Wnt pathway, which might change the type of genes activated by TGF-␤ or the manner of their regulation.