Identification of a Promoter-specific Transcriptional Activation Domain at the C Terminus of the Wnt Effector Protein T-cell Factor 4*

Wnt growth factors control numerous cell fate decisions in development by altering specific gene expression patterns through the activity of heterodimeric transcriptional activators. These consist of β-catenin and one of the four members of the T-cell factor (TCF) family of DNA-binding proteins. How can the Wnt/β-catenin pathway control various sets of target genes in distinct cellular settings with such a limited number of nuclear effectors? Here we asked whether different TCF proteins could perform specific, nonredundant functions at natural β-catenin/TCF-regulated promoters. We found that TCF4E but not LEF1 supported β-catenin-dependent activation of the Cdx1promoter, whereas LEF1 specifically activated the Siamoispromoter. Deletion of a C-terminal domain of TCF4E preventedCdx1 promoter induction. A chimeric protein consisting of LEF1 and the C terminus of TCF4E was fully functional. Therefore, the TCF4E C terminus harbors a promoter-specific transactivation domain. This domain influences the DNA binding properties of TCF4 and additionally mediates an interaction with the transcriptional coactivator p300. Apparently, the C terminus of TCF4E cooperates with β-catenin and p300 to form a specialized transcription factor complex that specifically supports the activation of the Cdx1promoter.

Multicellular organisms characteristically employ a limited number of signaling systems in order to generate the panoply of different cell types found in the body. The repeated use of the same growth factor families and signaling pathways poses the question of how the same effector proteins can generate distinct, tissue-specific responses (1). The canonical Wnt/␤-catenin signal transduction pathway provides an interesting model system to address this problem. Wnt growth factors constitute a large family of secreted glycoproteins that control numerous developmental processes in a wide range of organisms (2,3). In order to evoke transcriptional responses, the Wnt/␤-catenin signaling cascade utilizes a small number of bipartite transcription factor complexes. These complexes are formed by ␤-catenin, which provides a transcriptional activation function, and by a member of the T-cell factor (TCF) 1 family of DNA-binding proteins, which guides ␤-catenin to the promoter regions of specific target genes (4,5). ␤-Catenin belongs to an evolutionarily conserved family of proteins, which, as a signature motif, carry multiple copies of a 42-amino acid module, the Armadillo repeat (6). In ␤-catenin, 12 copies of this repeat form the large, central domain of the protein and provide the interaction surface for most of the known binding partners of ␤-catenin (7)(8)(9)(10). The Armadillo repeat domain is flanked by short N-and C-terminal extensions, which harbor the transcriptional activation domains of ␤-catenin (11)(12)(13). In addition to its role in Wnt signaling, ␤-catenin also performs a function in cellcell adhesion, where it is crucial for the formation of cadherin-catenin complexes (2,3).
In mammals, four genes encode the TCF family members TCF1, LEF1, TCF3, and TCF4, which have some structural features in common (5). The extreme N terminus harbors the binding site for ␤-catenin. The recognition and occupation of specific DNA sequence motifs is mediated by nearly identical HMG-box domains, which are located near the C terminus or in the middle of TCFs. Interspersed between the ␤-catenin-binding domain and the HMG-box are sequences that interact with Groucho/TLE transcriptional corepressors (14 -17). Groucho/ TLE factors are histone-binding proteins and additionally interact with a histone deacetylase (18,19). This suggests that their function is to set up a specialized repressive chromatin structure that prevents inappropriate activation of ␤-catenin/ TCF target genes in the absence of a Wnt signal. Since TCF proteins are functionally neutral on their own, their role in gene regulation is likely to serve as chromosomal docking sites for various interaction partners, thus promoting the formation of different transcription factor complexes with distinct properties.
The participation of Wnt/␤-catenin signaling in multiple developmental programs necessitates that ␤-catenin-TCF complexes induce only subsets of all potential Wnt/␤-catenin target genes at any particular time and that different sets of genes are addressed depending on the particular cellular background. How this is achieved is not understood. Several different mechanisms have been described whereby the formation and activity of ␤-catenin-TCF complexes is controlled by covalent modification of TCFs or competitive binding to ␤-catenin (3)(4)(5). These mechanisms, however, appear to affect the expression of Wnt target genes indiscriminately. Alternatively, ␤-catenin could employ transcriptional coactivators in a promoter-specific manner. Among the cofactors of ␤-catenin are BCL9/Legless, Brg-1, p300, and the closely related CBP (20 -25). CBP and p300 are widely used transcriptional coactivators that can provide a link to the basal transcription machinery or can target chromatin structure through their intrinsic acetylase activity (26). Both p300 and CBP have been implicated in differential promoter activation by ␤-catenin (21,27). In addition, several observations closely link TCF family members to mechanisms that generate promoter-specific transcriptional responses. TCFs can interact with Smad proteins, and this interaction appears to be critically involved in the combinatorial regulation of the Xenopus laevis Twin gene promoter by the transforming growth factor ␤ and Wnt signaling pathways (28,29). Similarly, an interaction between LEF1 and the basic helix-loop-helix/leucine zipper protein microphtalmia-associated transcription factor has been implicated in the melanocyte-specific expression of Wnt target genes (30). In addition, multiple isoforms with different functionalities arise from TCF genes by way of alternative splicing and the use of dual promoters, and different TCF family members perform distinct tasks in developmental processes (31)(32)(33)(34)(35)(36). It thus appears that TCF family members and their isoforms are intrinsically different and can support the execution of different developmental programs.
To better understand the mechanisms whereby Wnt signals selectively activate target genes, we began to compare TCF family members with respect to their ability to support ␤-catenin-mediated activation of different ␤-catenin/TCF target gene promoters. Here we report that LEF1 and the TCF4E isoform have opposing capabilities to synergize with ␤-catenin at the Siamois and the Cdx1 promoters. A promoter-specific transcriptional activation domain was identified at the C terminus of TCF4E, which can interact with the p300 transcriptional adaptor. The interaction between TCF4E and p300 differs from the ␤-catenin/p300 interaction, because ␤-catenin induces a posttranslational modification of p300, whereas TCF4E does not. We propose that TCF4E supports the promoter-specific assembly of a dormant transcription factor complex, the transcriptional activity of which is triggered by ␤-catenin-induced phosphorylation of p300.

EXPERIMENTAL PROCEDURES
Plasmids-Expression vectors for murine ␤-catenin, human p300, human TCF4E, and mouse LEF1 were the same as in Hecht et al. (21). All other plasmids were generated by standard cloning techniques or polymerase chain reaction-based strategies (37). Details of the constructs are available upon request.
Electrophoretic Mobility Shift Assays (EMSA)-For EMSA, LEF1 and TCF4 proteins were transcribed and translated in vitro using the SP6 TNT system (Promega). Translation efficiency was analyzed by Western blotting with anti-HA antibodies (3F10; Roche Applied Science). DNA probes used were complementary pairs of synthetic oligonucleotides derived from the TCR␣ gene enhancer (TCR␣25, 5Ј-TCGA-CCTGCAGGTAGGGCACCCTTTGAAGCTCTCCC-3Ј and 5Ј-TCGAGG-GAGAGCTTCAAAGGGTGCCCTACCTGCAGG-3Ј; TCF consensus motifs underlined) or a BamHI/XbaI fragment with Cdx1 promoter sequences from positions Ϫ224 to Ϫ52 (38,39). DNA fragments were labeled by fill-in with Klenow enzyme and [␣-32 P]dCTP (3000 Ci/mmol). For DNA binding, 0.5, 1.0, and 2.0 l of the TNT samples containing similar amounts of LEF1 and TCF4 proteins and 5 fmol of radiolabeled probe were combined in 20 mM Hepes/NaOH, pH 7.9, 75 mM NaCl, 1 mM dithiothreitol, 100 g of bovine serum albumin and 0.25 g of poly(dI: dC). Total volume was 15 l. Control reactions received 2 l of unprogrammed reticulocyte lysate diluted 1:1 with phosphate-buffered saline. Binding reactions were incubated for 30 min on ice, supplemented with 2 l of a solution containing 0.25% (w/v) bromphenol blue in 10 mM Tris/HCl, pH 7.8, and loaded onto 5 or 6% polyacrylamide gels with 0.5ϫ Tris-borate-EDTA running buffer (37). Electrophoresis was carried out at 150 V constant voltage. Gels were transferred onto Whatman paper, dried, and exposed to Biomax MR film (Eastman Kodak Co.) at Ϫ70°C.
Cell Culture, Transient Transfections, and Reporter Gene Assays-Human 293 embryonic kidney cells (ATCC number CRL-1658) and U-2 OS osteosarcoma cells (ATCC number HTB-96) were cultured as described (21). To monitor TCF protein expression, 2.5 ϫ 10 5 293 cells, seeded into 35-mm dishes, were transfected with 2.5 g of expression vector using FuGENE6 reagent (Roche Applied Science). For immunoprecipitations, 293 cells were transfected as before (21). For reporter gene assays, cells were plated into 24-well plates (5 ϫ 10 4 cells/well) and transfected 4 h later with a FuGENE6/DNA mixture containing 10 ng of pRL-CMV or pCMV␤ (Promega) as internal standard, 100 ng of luciferase reporter plasmid, and expression vectors for ␤-catenin (50 ng), TCF (10 ng), and p300 (250 ng) together with increasing amounts of competitor as indicated. Total amounts of DNA were kept constant by adding empty expression vector where needed. Firefly luciferase reporters were p01234, pCdx1-4 Luc, and pCdx1 (Ϫ350/ϩ72) Luc (39,40). Firefly luciferase and ␤-galactosidase activities in cell lysates were determined as before (21) using 96-well plates and a Labsystems Luminoskan Ascent luminometer. To measure Renilla luciferase activity, 100 l of a solution with 0.5 M coelenterazine (Calbiochem) in 25 mM Tris/HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA was injected per well of a 96-well plate, and after a delay of 0.5 s, light emission was recorded for 5 s. Reporter gene activities shown are average values and their S.D. values, obtained from at least three independent experiments after normalization against ␤-galactosidase or Renilla luciferase activities.
Western Blotting and Immunoprecipitation-Protein extracts from cells transfected in 35-mm dishes were made by lysing cells in 200 l of SDS-PAGE sample buffer containing "Complete" protease inhibitor mix (Roche Applied Science). After boiling for 5 min and sonication, onetenth of each lysate was loaded onto an 8% SDS-polyacrylamide gel. Co-immunoprecipitations were performed as described (21) except that TCF4/p300 immunoprecipitates were washed for 10 min at 4°C once with 1 ml of immunoprecipitation buffer (50 mM Tris/HCl, pH 7.6, 5 mM MgCl 2 , 0.1% Nonidet P-40) containing 120 mM NaCl and twice with 1 ml of immunoprecipitation buffer with 75 mM NaCl. For Western blot analyses, immunoprecipitates were eluted from protein G-Sepharose beads in SDS-gel loading buffer and run on 6% polyacrylamide gels. After electrophoresis, proteins were transferred onto nitrocellulose membranes and probed with anti-HA (3F10; Roche Applied Science), anti-p300 (catalog no. 05-267; Upstate Biotechnology, Inc.) or anti-␤catenin (Transduction Laboratories) primary antibodies and the appropriate horseradish peroxidase-labeled secondary antibodies (Jackson ImmunoResearch Laboratories). Antibody-antigen complexes were visualized by chemiluminescence (ECL system; Amersham Biosciences).
GST Pull-down Assays-For GST pull-down assays, 2 g of GST or the various GST-TCF4 fusion proteins were immobilized on glutathione-Sepharose beads as described (21). After preincubation in binding buffer containing 0.5% bovine serum albumin for 30 min at 4°C, [ 35 S]methionine-labeled p300 was added, after which binding proceeded for 2 h at 4°C. Following extensive washing with binding buffer without bovine serum albumin, material retained on the glutathione-Sepharose matrix was eluted in SDS-PAGE loading buffer, separated by SDS-PAGE on 6% gels, and visualized by fluorography. Radiolabeled p300 for these experiments was transcribed and translated in vitro using the SP6-based TNT system (Promega). For each binding reaction, we used a 4-l aliquot of a 50-l TNT reaction.
Phosphatase Assays-Transfected 293 cells were lysed for 30 min on ice in 1 ml of IPK buffer (50 mM Tris/HCl, pH 7.6, 75 mM KCl, 1 mM dithiothreitol, 10 mM NaF, 0.1 mM sodium orthovanadate, and "Complete" protease inhibitor mix). After clearance by centrifugation at 20,000 ϫ g for 15 min, 1-2 l of cell lysate were treated with -phosphatase (New England Biolabs) for 20 min at 37°C in a total volume of 10 l in the buffer supplied with the enzyme. Untreated or phosphatase-treated samples were loaded onto 6% SDS-polyacrylamide gels and separated by electrophoresis at 75 V. Run times were extended for up to 2 h after the bromphenol blue tracking dye had left the gel. After transfer onto nitrocellulose membranes, p300 was visualized by Western blotting as described above.

LEF1 and TCF4E Exhibit Promoter-specific Transactivation
Properties-To test whether downstream components of the Wnt/␤-catenin signal transduction pathway contribute to the differential regulation of Wnt target genes, we asked whether LEF1 and TCF4E were equally capable of supporting activation of Wnt-regulated promoters by ␤-catenin and p300. LEF1 and the TCF4E splice variant are representatives of the short and long isoforms of TCF family members (5). Their main differences are the extended C terminus in TCF4E with bind-ing motifs for the transcriptional corepressor CtBP and a contextdependent transactivation domain at the N terminus of LEF1 (Fig. 1A). Transactivation properties of LEF1 and TCF4E were compared at the promoters of the mouse Cdx1 gene and the Siamois gene of X. laevis (39,40). Both of these genes are Wnt/␤-catenin targets, which contain multiple TCF-binding elements (TBEs) in their promoter regions, and in their natural contexts they are regulated in a highly cell type-specific manner. Combinations of expression vectors for epitope-tagged TCF4E and LEF1, an activated form of ␤-catenin with Ala substitutions in the N-terminal destruction box (␤-catS33A) (3,41), and p300 were transfected into the human embryonic kidney cell line 293 and the human osteosarcoma cell line U-2 OS together with the pCdx1-4 Luc and p01234 luciferase reporter constructs (Fig. 1, C and D). When ␤-catenin, TCF4E, LEF1, and p300 were expressed individually, they had little or no effect on promoter activity. Pairwise combinations of ␤-catenin and LEF1 or ␤-catenin and TCF4E only weakly activated the luciferase reporters. Low levels of reporter gene expression were also induced by ␤-catenin and p300. In the absence of transfected TCFs, ␤-catenin presumably employs a limited pool of endogenous TCFs (14). In contrast, high levels of reporter gene activation were obtained when ␤-catenin, p300, and one of the TCFs were expressed simultaneously. Importantly, however, TCF4E and LEF1 clearly differed in their ability to synergize with ␤-catenin and p300, although they are expressed at similar levels ( Fig. 1, B and D). In both 293 and U-2 OS cells, TCF4E supported ␤-catenin/p300-dependent activation of the Cdx1 promoter but not of the Siamois promoter. Conversely, LEF1 mediated activation of the Siamois promoter but not of the Cdx1 promoter (Fig. 1D). Even when we raised the amount of transfected plasmid, LEF1 did not support activation of the Cdx1 reporter (not shown). Thus, the two TCF family members are functionally different and act in a promoter-specific manner.
The C Terminus of TCF4E Harbors a Promoter-specific Activation Domain-To gain insight into the mechanisms underlying its promoter-specific activity, we determined which domain in TCF4E is required for activation of the Cdx1 promoter. A panel of deletion mutants was constructed ( Fig. 2A), and Western blotting experiments were performed to confirm that all mutants were expressed at similar levels ( Fig. 2B). Functional testing revealed that all mutants that lack the C1.1 domain (amino acids 436 -482 of the TCF4E C terminus) are unable to synergize with ␤-catenin and p300 in Cdx1 promoter activation (TCF4⌬C, TCF4⌬C1, and TCF4⌬C1.1) (Fig. 2D). Constructs that retain this domain but have deletions in other parts of the TCF4E C terminus (TCF4⌬C2 and TCF4⌬C1.2) activate the Cdx1 promoter. From these results, we conclude that the C1.1 domain contains at least essential parts, if not all, of a promoter-specific transactivation domain. In addition, because activation of the Cdx1 promoter by TCF4E was seen with the short reporter construct used in these experiments (Fig. 2C), it appears that the proximal promoter region with its TBEs contains all of the determinants that make the Cdx1 promoter specifically responsive to TCF4E.
LEF1 Forms a Nonproductive Transcription Factor Complex at the Cdx1 Promoter-To rule out the possibility that LEF1 and the inactive TCF4 mutants were unable to activate the Cdx1 promoter because they could not access it, we performed competition experiments. For this, activation of the Cdx1 promoter by ␤-catenin, p300, and TCF4E was challenged by coexpressing increasing amounts of LEF1 and TCF4⌬C. As shown in Fig. 3A, higher levels of both factors progressively inhibited Cdx1 reporter activation. Deletion of the DNA-binding domain of LEF1 (Fig. 3A, LEF1⌬HMG) and, to a lesser extent, removal of the ␤-catenin-binding domain (Fig. 3A, LEF1⌬N) impaired this effect. Accordingly, the dominant negative activity of LEF1, and presumably of the TCF4⌬C mutant, most likely results from a combination of promoter blockade and sequestration of ␤-catenin. Therefore, LEF1 is capable of binding to the Cdx1 promoter in living cells and of competing with TCF4E. However, LEF1 and TCF4 isoforms lacking the C1.1 domain do not form productive transcription factor complexes at the Cdx1 promoter.
Since LEF1 can occupy the Cdx1 promoter construct, we wondered whether the addition of the C-terminal activation domain of TCF4E would endow LEF1 with the ability to stimulate Cdx1 promoter activity. A LEF1-TCF4E chimera was generated (Fig. 3, C and D), and when coexpressed together with either ␤-catenin or p300 alone, the LEF1-T4C fusion mediated even higher levels of Cdx1 reporter activity than TCF4E (Fig. 3B). In the presence of both p300 and ␤-catenin, TCF4E and LEF1-T4C activated the Cdx1 promoter equally well. For its function, the chimeric protein required an intact DNAbinding domain. This shows that the LEF1-T4C chimera must be able to bind to the Cdx1 promoter in order to activate and rules out the possibility that LEF1-T4C stimulates the reporter only indirectly. Thus, the presence of the TCF4E C terminus turns LEF1 into an activator at the Cdx1 promoter. Together with the previous results, this demonstrates that the transactivation domain present at the C terminus of TCF4E is necessary and sufficient to confer promoter specificity upon TCF proteins. LEF1 and TCF4 Possess Different DNA Binding Properties in Vitro-Wnt target gene activation depends on sequencespecific promoter recognition by TCFs and on TCF-␤-catenin complex formation. However, no differences were seen with respect to the ability of LEF1, TCF4E, or TCF4⌬C to interact with ␤-catenin (not shown). Therefore, we compared the DNA binding properties of LEF1, wild-type and mutant TCF4E, and the LEF1-T4C chimera by EMSA with two different DNA probes (Fig. 4). The TCR␣25 probe harbors a single consensus TBE derived from the T-cell receptor ␣-chain enhancer (38). The second probe was a Cdx1 promoter fragment, which contains three TCF binding motifs: TBE3a, TBE3, and TBE4 (39). The TBE3a element is a newly discovered TBE that was not reported previously. The TCF proteins were transcribed and translated in vitro, and similar amounts of the various factors, as determined by Western blotting (Fig. 4A), were used. Although both probes were bound by LEF1 and TCF4E, LEF1 exhibited an affinity for the single binding site of the TCR␣25 probe that was at least 5 times higher than that of TCF4E (Fig.  4C, compare lanes 3-5 with lanes 6 -8). Interestingly, this difference was also seen when oligonucleotides with a single TBE (either TBE3a, -3, or -4) derived from the Cdx1 promoter were used as probes (not shown). The C terminus of TCF4E appears to be at least partly responsible for the reduced DNA binding capacity of TCF4E, as indicated by the complementary increases and decreases in the DNA-binding capabilities of TCF4⌬C1.1 and the LEF1-T4C chimera (Fig. 4C, compare  lanes 6 -8 with lanes 9 -11 and lanes 3-5 with lanes 12-14).
Whereas TCF4E possesses a markedly reduced affinity for a single TBE when compared with LEF1, this difference was not seen in EMSAs with a Cdx1 promoter fragment containing all three TBEs together. TCF4E, TCF4⌬C1.1, and the chimeric LEF1-T4C protein readily generated protein-DNA complexes representing single, double, and triple occupancy of the Cdx1 probe (Fig. 4C, lanes 20 -28). In contrast, under the conditions used, LEF1 generated primarily a single species of protein-DNA complexes (Fig. 4C, lanes 17-19). Still, LEF1 can recognize all three Cdx1 TBEs also in the context of the entire promoter fragment as shown by DNase I footprinting experiments. 2 Because the same protein preparations were used for EMSA with the Cdx1 promoter fragment and the single TBEs, the observed differences in DNA binding are inherent properties of the TCF proteins and not due to variable efficiencies of protein folding. In addition, these experiments also confirm the potential regulatory functions of the TCF4E C terminus. As seen in EMSAs with a single TBE, we observed that deletion of 2 A. Hecht, unpublished observations.  Fig. 1. All constructs contain the C-terminal HA epitope tag. B, expression of the TCF4E deletion mutants was analyzed by Western blotting with an anti-HA antibody as before (Fig.  1B). M w , molecular weight standard. Asterisks mark posttranslationally modified forms of TCF4E and its derivatives. C, schematic representation of the pCdx1(Ϫ350/ϩ72)Luc reporter. The promoter fragment in this reporter harbors only TBE3a, -3, and -4. D, 293 cells were transfected with combinations of expression vectors for the TCF4E deletion mutants, ␤-catS33A, p300, the Cdx1 reporter, and pCMV␤ as indicated. Luciferase reporter gene activities were determined as in Fig.  1. Reporter gene activity without ␤-catenin (␤-cat), TCF4E, and p300 was arbitrarily assigned the value of 1. WT, wild type. the C1.1 domain increased the ability of TCF4E to bind to the Cdx1 promoter fragment. This is evident from the greater abundance of DNA complexes with all three TBEs being simultaneously occupied by the TCF4⌬C1.1 mutant (Fig. 4C, compare lanes 20 -22 with lanes 23-25). In addition to this inhibitory activity, which seems to be linked to the C1.1 domain, the TCF4E C terminus also appears to harbor an opposing activity, which promotes the binding of TCFs to multiple TBEs. We conclude this from the observation that the LEF1-T4C chimera interacts with the Cdx1 promoter more efficiently than the parental LEF1 protein (Fig. 4C, compare lanes 17-19 with  lanes 26 -28). Aside from this, the pattern of Cdx1 promoter complexes formed by the LEF1-T4C chimera differed from all other patterns seen. The molecular weights of TCF4E and LEF1-T4C are very similar, and with the TCR␣25 probe, TCF4E and LEF1-T4C complexes migrated at nearly identical positions (Fig. 4C, lanes 6 -8 and 12-14). However, the LEF1-T4C-Cdx1 complexes are much more retarded than expected (Fig. 4C, compare lanes 20 -22 and 26 -28). Since TCF proteins are architectural transcription factors that are known to change DNA topology (5), it is possible that these migratory differences between TCF4E and LEF1-T4C arise from different conformations of the DNA complexes containing either TCF4E or LEF1-T4C. Taken together, the results of our comparative DNA-binding experiments show that the DNA binding properties of TCF4E differ significantly from those of LEF1 and that the affinity of TCF4E for DNA-binding elements changes depending on the number of motifs present and in response to regulatory functions present at its C terminus.
The Promoter-specific Activation Domain of TCF4E Interacts with the Transcriptional Coactivator p300 -High level induction of the Cdx1 reporter constructs required simultaneous expression of ␤-catenin, p300, and TCF4E. However, pairwise combinations of p300 and TCF4E and especially the LEF1-T4C chimera also elicited a significant response of the Cdx1 promoter (e.g. Fig. 3B). Therefore, we speculated that the C terminus of TCF4E might contribute to the formation of an active transcription factor complex at the Cdx1 promoter through protein-protein interactions with p300. Indeed, p300 co-immunoprecipitated with epitope-tagged TCF4E from lysates of transfected 293 cells (Fig. 5A, lane 3). To clarify whether the presence of p300 was simply due to coprecipitation with ␤-catenin, which was also detected in the immunoprecipitate and which can interact with p300 on its own (21), we performed control experiments with the TCF4⌬N mutant lacking the ␤-catenin-binding domain of TCF4E. As expected, ␤-catenin no longer coprecipitated with TCF4⌬N, whereas p300 was still present in the immunoprecipitate (Fig. 5A, lane 4). In contrast, deletion of the C1.1 domain significantly reduced the amount of p300 coprecipitating with TCF4⌬C1.1 even in the presence of ␤-catenin (Fig. 5A, lane 5). Apparently, the C terminus of TCF4E contributes to an interaction between TCF4E and p300. Complex formation between TCF4E and p300 was further characterized by GST pull-down experiments with bacterially expressed GST-TCF4 fusions and p300, which was transcribed and translated in vitro. All GST-TCF4C fusions containing the C1.1 domain interacted with p300 (Fig. 5, B and C, constructs A, D, and F), whereas all mutants lacking the C1.1 domain did not (Fig. 5, B and C, constructs B, C, and E). Thus, the C1.1 domain in TCF4E, which is required for promoter-specific transactivation, also mediates an interaction with p300.
␤-Catenin, but Not TCF4E, Induces Phosphorylation of p300 -The observation that the C1.1 domain can interact with p300 raises the possibility that p300 is recruited to the Cdx1 promoter by TCF4E, which would make ␤-catenin seemingly dispensable for promoter activation. However, when we analyzed p300 in total cell lysates using extended electrophoresis on denaturing gels and Western blotting, we noticed that p300 from 293 cells expressing both p300 and ␤-catenin migrated more slowly than p300 from control cells or cells expressing p300 and TCF4E (Fig. 6A). This indicates that ␤-catenin induces a covalent, posttranslational modification of p300, whereas TCF4E does not. Since p300 is a phosphoprotein and regulated phosphorylation contributes to its functionality as a transcriptional coactivator (26), we tested whether p300 becomes phosphorylated in the presence of ␤-catenin. Upon treating protein samples containing p300 with -phosphatase, the p300 bands sharpened, and material from control cells and ␤-catenin-expressing cells migrated at the same relative positions (Fig. 6B). Control experiments performed in the presence of the phosphatase inhibitors NaF and EDTA or in the absence of Mn 2ϩ ions (essential phosphatase cofactors) prove that the changes in electrophoretic mobility of p300 are due to dephosphorylation rather than protein degradation (Fig. 6C). Based on this, we propose that TCF4E recruits a dormant form of p300 to the Cdx1 promoter, whereas ␤-catenin might trigger promoter activation by inducing phosphorylation of p300. DISCUSSION TCFs are multifunctional transcription factors that control the expression status of Wnt target genes by forming multipro-tein promoter complexes that incorporate either repressing or activating cofactors (5). Due to a common overall structure and seemingly identical DNA binding specificities, TCF proteins were long considered to be largely interchangeable components of the Wnt signaling cascade. However, recent reports revealed functional differences between TCF family members and between isoforms derived from the same TCF gene. For example, LEF1 appears to mainly act as a ␤-catenin-dependent transcriptional activator, whereas the repressor activities of TCF3 prevail over its activating function (42)(43)(44). The molecular basis for these differences is unknown, but in the case of TCF proteins from X. laevis it was shown that activating properties require the presence of an alternatively spliced exon at their N termini (31,35). Here we report that also LEF1 and TCF4 have different transactivation capacities, although both factors contain the activating N-terminal exon. Additionally, whereas both LEF1 and TCF4 can act as activators, they appear to have distinct spectra of target genes, and a novel, promoter-specific transcriptional activation domain was identified in the TCF4E variant.
The domain of TCF4E responsible for its promoter-specific activity is located at the C terminus and contains the "CRARF" amino acid motif and a second block of mostly positively charged amino acids. The "CRARF" domain is evolutionarily conserved and specifically present in the "E" variants of TCF1 and TCF4 gene products (5,34). However, to date, no specific function has been assigned to this domain. Based on our results, it may be involved in at least two different processes, namely DNA binding and selective activation of certain promoters. Although LEF1 and TCF4E recognized the same DNA sequences, their affinities for single or multimerized recognition motifs varied considerably. Pukrop et al. (31) also observed 3-7-fold differences in affinity for single TBEs between LEF1 and other TCF family members. These differences are likely to be of importance, because TCF proteins are not only mediators of Wnt signaling but also contribute to Wnt/␤-catenin-independent gene regulation, for example at the TCR␣ enhancer or at the HIV-1 promoter (38,45). How can inappropriate activation of these regulatory elements by ␤-catenin-TCF complexes be prevented? Perhaps significantly, both the TCR␣ enhancer and the HIV-1 promoter contain single TCF binding motifs (38,45), whereas the known Wnt target genes typically contain multiple TCF recognition elements. Under competitive conditions, when TCF levels are low, or when only a particular TCF family member is expressed in a cell, the differential recognition and occupancy of single versus multiple binding sites could be one way to distinguish Wnt target genes from nontarget genes.
Although differences in DNA binding may contribute to the differential activation of the Cdx1 promoter by LEF1 and TCF4E, we believe that the C-terminal activation domain in TCF4 performs additional functions. LEF1 and TCF4 mutants lacking the C1.1 domain displayed differences in promoter FIG. 5. The C terminus of TCF4E mediates an interaction with p300. A, TCF4E and p300 co-immunoprecipitate from cell lysates. 293 cells were transfected with expression plasmids for Glu-Glu-tagged p300 (p300-EE; 15 g) and HA-tagged wild-type (WT) or mutant TCF4E (5 g) as indicated. Cell extracts were prepared 40 h after transfection and used for immunoprecipitation (IP) with anti-HA antibodies. Immunoprecipitated material and a fraction of each cell lysate were resolved by SDS-PAGE and analyzed by Western blotting (IB) with antibodies as shown. B, the promoter-specific activation domain at the TCF4E C terminus interacts with p300 in vitro. GST or GST-TCF4 fusion proteins shown schematically in C were bound to glutathione-Sepharose beads and incubated with radiolabeled p300. Proteins retained by GST or the GST-TCF4 fusions and 10% of the input material were analyzed by SDS-PAGE and fluorography. C, structure of the GST-TCF4 fusion proteins and, for comparison, of TCF4E. Amino acid end points of the TCF4 fragments are given. Functionally relevant domains and interaction sites for TCF4Ebinding factors are indicated as before. Hatched bar, GST.
FIG. 6. ␤-Catenin (␤-cat), but not TCF4E, induces phosphorylation of p300 in vivo. A, altered electrophoretic mobility of p300 in the presence of ␤-catenin. 293 cells were transfected with expression plasmids for p300-EE (15 g), ␤-catS33A, or HA-tagged TCF4E (5 g) as indicated. Cell extracts were prepared 40 h after transfection. A fraction of each cell lysate was resolved by SDS-PAGE and analyzed by Western blotting with antibodies against p300. The position of anti-p300 immunoreactive material in the presence or absence of ␤-catS33A or TCF4E is indicated at the right side of the blot. B, reduced electrophoretic mobility of p300 in the presence of ␤-catenin is due to phosphorylation. Cell extracts prepared as in A were treated with increasing amounts of -phosphatase ( PPase) for 20 min at 37°C. Samples were then resolved by SDS-PAGE on 6% gels and analyzed by Western blotting with anti-p300 antibodies. C, specificity of the phosphatase effect. Fractions of the same cell extracts as in B were subjected to phosphatase treatment either in the absence of the essential -phosphatase cofactor MnCl 2 or in the presence of the phosphatase inhibitors EDTA and sodium fluoride (NaF). Samples were analyzed by SDS-PAGE and Western blotting as in B.
recognition, but there was no clear correlation between the ability to activate the Cdx1 promoter and the mode of promoter binding in vitro. TCF4⌬C and TCF4⌬C1.1 did not activate the Cdx1 promoter and blocked Cdx1 induction by wild-type TCF4E in our competition experiments, yet they bound to the Cdx1 promoter in vitro even more efficiently than TCF4E or LEF1 (Figs. 3 and 4). Moreover, the pattern of DNA-protein complexes produced by the LEF1-T4C chimera differed from those of all other factors analyzed, although the chimera is functionally equivalent to TCF4E. Thus, an influence on promoter occupation or topology is unlikely to be the only mechanism by which the Cdx1-specific activation domain at the C terminus of TCF4E functions. An additional mode of action could be that the C1.1 domain mediates protein-protein interactions. Indeed, it interacts with p300 in vitro, and p300 coimmunoprecipitated together with TCF4 from cellular lysates. Although p300 may not be the only factor that interacts with the C1.1 domain, the match between the ability to form a complex with p300 and the ability to support activation of the Cdx1 promoter strongly suggests that the interaction between TCF4E and p300 is of physiological relevance.
Expression of the murine Cdx1 gene begins in ectodermal and mesodermal cells of gastrulating embryos around day 7.5 of gestation. It occurs in a graded fashion along the body axis with highest levels of expression in posterior parts. From day 14 of embryonic development onward, Cdx1 is also expressed in the endoderm of the developing intestine (46 -48). The latter tissue expresses both TCF4 and LEF1 (49,50). Based on our results, TCF4E may be responsible for Cdx1 regulation in the intestine. However, TCF4 is not expressed in the early embryonic Cdx1 expression domain (51). We suggest that there activation of the Cdx1 promoter is mediated by an isoform derived from the TCF1 gene. Preliminary results indicate that TCF1E, but not TCF1B, which resembles LEF1, can cooperate with ␤-catenin and p300 to activate the Cdx1 promoter. 2 Temporal and tissue-specific inducibility of the Cdx1 gene by Wnt signaling thus may in part be restricted through the availability of an appropriate TCF isoform. The presence of TCF4E or TCF1E may hence be a prerequisite, but it is not sufficient for Cdx1 activation, because the expression domains of TCF1 and TCF4 exceed the expression domains of the Cdx1 gene (51). Many factors have been identified that interfere with the gene regulatory activity of ␤-catenin and TCFs and could help to shape specific gene expression patterns. However, inhibitors like ICAT act upstream of target gene promoter activation and cause a global inhibition of Wnt signaling rather than a selective and differential regulation of Wnt target genes (3,5,52). More likely, mechanisms that govern tissue-specific Wnt target gene regulation act at the level of individual regulatory elements and their associated transcription factors. For example, a hierarchy of regulatory events permits Wnt effectors to maintain expression of the murine Brachyury gene but only after the initiation of transcription by a different regulatory system (53). Alternatively, the activation of Wnt target genes may depend on the simultaneous input from multiple signaling pathways such as a combination of Wnt and transforming growth factor ␤ signals (54,55). At the molecular level, the two pathways are integrated through physical interactions of the FIG. 7. Models for activation of the Cdx1 promoter by transcription factor complexes containing TCF4E, ␤-catenin (␤-cat), and p300. A, in this model, promoter specificity is generated, because the assembly of a transcriptionally active complex at the Cdx1 promoter requires a particular promoter topology that is induced specifically by TCF4E (I) but not LEF1 (II). B, the combinatorial input model is an alternative, but not mutually exclusive, possibility. Here, TCF4E and an additional promoter-specific DNA-binding protein (X) synergize to recruit p300 to the Cdx1 promoter by providing multiple contact points (I). In this case, LEF1 fails to support ␤-catenin and p300-dependent activation of the Cdx1 promoter, because, unlike TCF4E, it does not provide an essential interaction site for p300 (II). XRE, factor X-responsive promoter element. Transcriptionally active complexes may assemble at the promoter de novo, or, as indicated here, ␤-catenin may enter preformed complexes containing p300 and TCF4E. By inducing phosphorylation of p300, ␤-catenin would thereby turn the resident but dormant protein into an active cofactor (symbolized by the appearance of a star-shaped outline around p300). Note that the exact composition of the protein complexes and the stoichiometric relationships are unknown.
Smad proteins and members of the TCF family, necessitating the presence of DNA recognition motifs for both families of DNA-binding proteins (28,29). Thereby, only those genes with the proper combination of cis-active DNA elements in their promoter regions will respond to the combinatorial input from Wnt and transforming growth factor ␤-signaling pathways. This underscores that the DNA sequences of promoter regions and the interaction partners of TCF proteins are critical determinants for the differential regulation of Wnt target genes.
Although p300 and CBP have already been linked to differential gene regulation by ␤-catenin (21,27), it was unexpected to find an interaction between TCF4E and a transcriptional coactivator like p300. Intuitively, one would surmise that recruitment of p300 by TCF4E catalyzed transcription even in the absence of a Wnt signal. However, even the low level Cdx1 promoter activation upon expression of p300 and TCF4E was seen only if TCF4 contained the N-terminal ␤-catenin binding domain. Thus, Cdx1 transcriptional activation is strictly dependent on ␤-catenin and simple recruitment of p300 by TCF4 is insufficient for promoter activation. A similar separation between recruitment of p300 or the highly related CBP and transcriptional activation has been described by Soutoglou et al. (56). An additional step that could be required for transcriptional activation is the phosphorylation of p300, which we found can be induced by ␤-catenin. Phosphorylation of p300 by several different kinases has been observed, but both activating and inhibitory effects have been described (26,57,58). A possible explanation for these conflicting observations could be that the particular settings at a given promoter ultimately determine the molecular mechanisms by which p300 is used as a transcriptional coactivator and whether its activity needs to be additionally regulated. Such a context dependence would explain how p300 can be utilized by ␤-catenin at both the Siamois and the Cdx1 promoters but apparently in two different manners. At the Siamois promoter and in conjunction with LEF1, recruitment and activation of p300 by ␤-catenin appears to be sufficient, whereas at the Cdx1 promoter an additional contribution by TCF4E is required. What could be the promoter-specific determinants for these distinctions? The interaction of LEF1 and TCF4E with Cdx1 promoter fragment resulted in protein-DNA complexes with distinct topologies. Different three-dimensional structures of the resulting transcription factor complexes may preclude or even require the incorporation of specific coactivators in order to facilitate the formation of a productive transcription complex (59) (Fig. 7A). Indeed, the promoters of the Siamois, Twin, and cyclin D1 genes and an artificial reporter construct displayed a differential responsiveness to coexpression of ␤-catenin and p300 (21). An alternative but not necessarily mutually exclusive explanation is based on the ability of p300 to interact with many different transcription factors. Based on this property of p300, it is possible that its recruitment to the Cdx1 promoter depends on multiple interactions not only with ␤-catenin and TCF4E but in addition with an as yet unknown factor that specifically recognizes Cdx1 promoter sequences (Fig. 7B). According to this model, LEF1 fails to activate the Cdx1 promoter because it cannot provide a vital contact for p300. Such a mechanism represents a new variation of the combinatorial input and multiple contact model that is already used for the regulation of Wnt target gene expression. Whether formation of a transcriptionally active complex occurs in one step or whether ␤-catenin joins a preformed complex and triggers its activity by inducing phosphorylation of p300 remains to be determined. Other issues that need to be addressed are the mechanisms that prevent activation of the Siamois promoter by TCF4E and make this promoter dependent on LEF1 or TCF3 (40). In addition, the re-peated cooperation in differential gene regulation and the varying functional implications of their joint activities warrants further analysis of the interactions between ␤-catenin, TCFs, and p300/CBP. This will provide further insight into the mechanisms of tissue-specific and inducible gene regulation by Wnt growth factors during embryonic development.