Identification of Two Regulatory Elements within the High Mobility Group Box Transcription Factor XTCF-4*

Some members of the Wnt family of extracellular glycoproteins regulate target gene expression by inducing stabilization and nuclear accumulation of β-catenin, which functions as a transcriptional activator after binding to transcription factors of the T-cell factor/lymphoid enhancer factor (TCF/LEF) family. Three different members of this family have been identified in Xenopus laevis thus far that differ in their ability to influence mesodermal differentiation and to activate expression of the Wnt target gene fibronectin. Here we report on the isolation and characterization of additional variants of XTCF-4. We show that the differential ability of these proteins and other members of the TCF family to activate target genes is neither due to preferential interaction with transcriptional cofactors of the groucho family or SMAD4 nor to different DNA binding affinities. Expression of these proteins in an epithelial cell line reveals differences in their ability to form a ternary complex with DNA and β-catenin. Interestingly, formation of this ternary complex was not sufficient to activate target gene expression as previously thought. Our experiments identify two amino acid sequence motifs, LVPQ and SFLSS, in the central domain of XTCF-4 that regulate the formation of the DNA-TCF-β-catenin complex or activation of target genes, respectively. Biochemical studies reveal that the phosphorylation state of these XTCF-4 variants correlates with their ability to form a ternary complex with β-catenin and DNA but not to activate target gene expression. The described variants of XTFC-4 with their different properties in complex formation provide strong evidence that in addition to the regulation of β-catenin stability the isoforms of TCF/LEF transcription factors and their posttranslational modifications define the cellular response of a Wnt/wingless signal.

Some members of the Wnt family of extracellular glycoproteins regulate target gene expression by inducing stabilization and nuclear accumulation of ␤-catenin, which functions as a transcriptional activator after binding to transcription factors of the T-cell factor/ lymphoid enhancer factor (TCF/LEF) family. Three different members of this family have been identified in Xenopus laevis thus far that differ in their ability to influence mesodermal differentiation and to activate expression of the Wnt target gene fibronectin. Here we report on the isolation and characterization of additional variants of XTCF-4. We show that the differential ability of these proteins and other members of the TCF family to activate target genes is neither due to preferential interaction with transcriptional cofactors of the groucho family or SMAD4 nor to different DNA binding affinities. Expression of these proteins in an epithelial cell line reveals differences in their ability to form a ternary complex with DNA and ␤-catenin. Interestingly, formation of this ternary complex was not sufficient to activate target gene expression as previously thought. Our experiments identify two amino acid sequence motifs, LVPQ and SFLSS, in the central domain of XTCF-4 that regulate the formation of the DNA-TCF-␤-catenin complex or activation of target genes, respectively. Biochemical studies reveal that the phosphorylation state of these XTCF-4 variants correlates with their ability to form a ternary complex with ␤-catenin and DNA but not to activate target gene expression. The described variants of XTFC-4 with their different properties in complex formation provide strong evidence that in addition to the regulation of ␤-catenin stability the isoforms of TCF/LEF transcription factors and their posttranslational modifications define the cellular response of a Wnt/wingless signal.
Wnt-1 was originally identified as a proto-oncogene in mice (1) and was found to be a vertebrate homolog of the Drosophila segment polarity gene wingless. To date at least 15 additional vertebrate homologs have been identified (for review see Ref. 2). Members of the Wnt family of extracellular glycoproteins are able to activate at least two different signaling pathways in vertebrates (reviewed in Refs. 3 and 4). The canonical Wnt pathway is involved in different developmental processes such as cell fate specification and cell migration. The major cytoplasmic effector of this pathway, ␤-catenin, accumulates in the cytoplasm in response to a Wnt signal. Subsequently, ␤-catenin can enter the nucleus where it binds to the N terminus of TCF/LEF 1 transcription factors to function as a transcriptional coactivator. Deregulation of the Wnt/␤-catenin pathway leads to carcinogenesis. Mutations increasing the stability and thus the cytoplasmic/nuclear pool of ␤-catenin were found in colon carcinomas and malignant melanomas (4). The identification of c-myc as a target gene of TCF-4/␤-catenin links Wnt signaling with cell proliferation and thus with carcinogenesis (5). Another important target gene of TCF/LEF in this context leading to cell cycle deregulation was identified as cyclin D1 (6).
Members of the TCF/LEF family were originally identified as T-and B-lymphocyte-specific transcription factors. Sequencespecific DNA binding is mediated by the HMG box, and the target motif of these factors is given by the sequence (C/G)TT-TG(A/T)(A/T) (7). Members of this family are unable to activate transcription of reporter gene constructs that carry multiple copies of their minimal binding motifs (8) and thus lack activation properties on their own. This raises the question how these proteins work as transcriptional regulators. One well characterized target gene regulated by LEF-1 is the T-cell receptor ␣ gene, TCR␣. Activation of the TCR␣ promoter by LEF-1 is strictly context-dependent. Under these circumstances LEF-1 functions as an architectural transcription factor by bending DNA, thus facilitating stable interactions of other known transcription factors like CREB/ATF, PEBP2␣, and Ets-1 (9).
Despite binding to the same target site, members of the TCF/LEF family differ in their function. When overexpressed on the ventral side of Xenopus embryos, mLEF-1 induces the formation of a secondary body axis via activation of the Wnt target gene siamois (13,14), whereas XTCF-3 fails in this assay (15). In epithelial A6 cells LEF-1 activates the Wnt target gene fibronectin, whereas XTCF-3 does not (21). Two hybrid studies demonstrated that unlike LEF-1, XTCF-3 binds transcriptional corepressors of the groucho family (22) as well as CtBP (23). This led to the model that XTCF-3 functions as a transcriptional repressor in the absence of ␤-catenin by recruiting corepressors, whereas LEF-1 is unable to function as a repressor due to lack of interaction with grouchos or CtBP (23). However, Levanon et al. (24) showed physical and functional interaction of the human groucho homolog TLE-1 with hLEF-1. The contradictory results may be explained by different splice variants of LEF-1 that were used by Roose et al. (22) and Levanon et al. (24), respectively.
Analysis of the genomic structure of the human TCF-1 gene and analysis of the detected mRNA transcripts revealed extensive alternative splicing events. These result in a maximum of 96 theoretical splice variants, at least 5 of which have been shown to be expressed in vivo (25). Alternative splice variants also have been reported for TCF-3, TCF-4, and LEF-1. These observations further support the idea that the binding of corepressors or coactivators depends on the expressed splice variant.
In this study we report on the isolation of novel XTCF-4 variants. Furthermore, we analyzed the transactivation potential, DNA binding affinities, protein-protein interactions, and posttranslational modifications of these variants in comparison to mLEF-1/XLEF-1 and XTCF-3. Our studies revealed that the complex formation of XTCF-4 and ␤-catenin and the ability of this complex to activate target genes depends on the absence of two short amino acid motifs, LVPQ and SFLSS. Unexpectedly, these motifs neither influenced groucho nor SMAD4 binding to the transcription factors. Instead, we found that XTCF-4-␤catenin-DNA complex formation was accompanied by a dephosphorylation event observed in XTCF-4 isoforms lacking these motifs.

EXPERIMENTAL PROCEDURES
DNA Constructs Used in This Study-Additional TCF-4 variants were isolated exactly as described (26). In short, TCF-4 clones were isolated from an A6 -ZAP cDNA library. Complete cDNA clones were obtained by 5Ј-rapid amplification of cDNA ends and combining overlapping fragments. The XTCF-3 clone was isolated from a gastrula stage cDNA library and completed by 5Ј-rapid amplification of cDNA ends. The Xenopus LEF-1 and the XTCF-3⌬grg constructs were kind gifts of O. Destreé (22,27). Independently, an identical XLEF-1 clone has been isolated in our laboratory by means of reverse transcriptase-PCR. The C-terminal truncated form of XTCF-3 has been constructed by PCR methods using proofreading Pwo polymerase (Roche Molecular Biochemicals). Primer sequences used for this construct were 5Ј-GCGC-CTCGAGATGCCTCAGCTCAACAGCGGC-3Ј and 5Ј-GGGCCCCTC-GAGTCTCTCTCAACTAGTCACTGGATCTGGTCAC-3Ј. The correct sequence of this construct was verified by DNA sequencing. All of these constructs were subcloned into pCS2ϩ containing a cytomegalovirus promoter as well as an sp6 promoter.
For the isolation of a full-length Xenopus ESG-1 clone, a ZAP Xenopus gastrula cDNA library was screened with a 600-base pair fragment of ESG-1 (28) obtained by PCR amplification of a Xenopus gastrula library. Only partial cDNA clones were isolated, the longest encoding amino acids 34 -756. The N-terminal region of ESG-1 (756 base pairs) was amplified by PCR of a Xenopus oocyte cDNA library and subcloned into the BamHI site of ESG-1. The obtained full-length sequence was 95% homologous to the previously isolated ESG-1 and submitted to GenBank TM (accession number AF289027). For GST pulldown assays with GST-SMAD4, human SMAD4 cloned into pGex4T3 (Amersham Pharmacia Biotech). The truncated version of XTCF-3 lacking the HMG domain and C terminus (amino acids 1-322) which was used as a negative control was prepared by PCR and subcloned into the EcoRI-XhoI sites of pCS2ϩ. The primers used were 5Ј-CGGGAAT-TCATGCCTCAACTAAACAGCGGC-3Ј and 5ЈCGGCTCGAGACTGGG-CTTCTTCTCTTCTTCC-3Ј.
Cell Culture and Reporter Gene Assays-Transfection of epithelial A6 cells with different expression constructs and luciferase reporter gene assays were performed as described (21), except for the use of Effectene transfection reagent (Qiagen) instead of Lipofectin (Life Technologies, Inc.). Cell extracts used in bandshift studies were prepared in 50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM DTT, and 1 mM phenylmethylsulfonyl fluoride. The used phosphatase inhibitors were sodium vanadate (10 mM), sodium fluoride (10 mM), and sodium molybdate (10 mM).
Bacterial Expression of HMG Box Transcription Factors and in Vitro Translation-For expression of His-tagged fusion proteins, cDNAs were subcloned into pRSETA (Invitrogen) with PCR techniques using proofreading Pwo polymerase (Roche Molecular Biochemicals) according to the manufacturer's instructions. Transformed BL21(DE3) bacteria were induced by 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 2 h at room temperature. After centrifugation, bacterial pellets were lysed by sonification in lysis buffer (50 mM NaH 2 PO 4 , 300 -1000 mM NaCl, 10 mM imidazole, pH 8.0) including protease inhibitors (complete, Roche Molecular Biochemicals), and the cleared supernatant was loaded onto nickel-nitrilotriacetic acid resins for 1 h at 4°C in a batch procedure. After washing with 50 mM NaH 2 PO 4 , 300 -1000 mM NaCl, 10 -30 mM imidazole, pH 8.0, the protein was eluted 10ϫ with 1 ml of 50 mM NaH 2 PO 4 , 300 -1000 mM NaCl, 100 -300 mM imidazole, pH 8.0. Each fraction was analyzed by SDS-PAGE followed by Coomassie staining using standard procedures. For Western blot analyses of His-tagged fusion proteins, a commercially available antibody was used (RGS-His by Qiagen). For the electrophoretic mobility shift assays proteins were dialyzed against 1000-fold access of binding buffer (20 mM Tris/HCl, pH 8.0, 50 mM NaCl, 5% glycerol, 0.1 mg bovine serum albumin per ml, 1 mM dithiothreitol, 1 mM MgCl 2 ), and protein concentrations were determined by Bradford assay. In vitro translation was done using a coupled transcription/translation kit (Promega).
Far Western Analyses-Bacterially expressed fusion proteins were separated by SDS-PAGE and transferred to a nitrocellulose membrane. Blocking was done in 3% bovine serum albumin in TBST overnight at 4°C. The ESG protein was translated in vitro in the presence of [ 35 S]methionine and added to the blocked nitrocellulose filter in TBST for 8 h at 4°C. After 3 washes in TBST, 0.25% bovine serum albumin the filter was exposed to a PhosphorImager screen and further analyzed. The negative control was a His-tagged fusion protein comprising the cytoplasmic domain of Xenopus cadherin-11 (29).
GST Pull-down Assay-Interaction assays were performed by addition of bacterially expressed GST or GST-SMAD4 proteins, 2-4 l of 35 S-labeled TCF/LEF proteins, and 300 l of binding buffer (20 mM Tris/HCl, pH 8.0, 50 mM KCl, 2.5 mM MgCl 2 , 1 mM DTT, 10% glycerol, and 0.2% Nonidet P-40) to 20 l of glutathione-Sepharose beads. Binding reactions were rolled for 4 h at 4°C and washed four times with wash buffer (20 mM Tris/HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA, and 0.2% Nonidet P-40). The beads were boiled in sample buffer and subjected to SDS-PAGE. After staining with Coomassie to verify the integrity of GST and GST-SMAD4, the gel was dried and subjected to PhosphorImager analysis to visualize bound TCF/LEF proteins.
Quantitative Electrophoretic Mobility Shift Assay Studies-The oligonucleotide sequences used were as shown in Fig. 3B and as published (21). Before labeling, oligonucleotides were purified by denaturing PAGE. Both DNA strands were labeled separately and annealed to form the duplex DNA. Competition studies were performed with a 100-fold excess of either unlabeled probe or an unrelated oligonucleotide as previously published (21), 5Ј-CAATAAAAAAGGGATCTCGCCTGTTA-ATGA-3Ј. For quantitative gel shift analysis, the total concentration of the dimer in the binding reaction was 13 nM. Binding conditions using different concentrations of fusion proteins were done exactly as described (21). The protein concentrations used are given in the corresponding figures. Quantification of the band shift studies was performed using a PhosphorImager (Fuji). The results shown in Fig. 4, A and B, are the mean of 3-5 independent experiments. Error bars give standard error means. Data analysis was done using CricketGraph III software (Computer Associates International, Inc. Islandia, NY). For supershift analysis 1 l of either XTCF-4 antibody (Biomol) or ␤-catenin antibody (Sigma) was added.
Dephosphorylation Procedure-Phosphatase treatment lysates were prepared as described above in 50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM DTT, and 1 mM phenylmethylsulfonyl fluoride. Phosphatase treatment was done by adding 1 l of potato acidic phosphatase (Roche Molecular Biochemicals) and 0.5 l of calf intestine alkaline phosphatase (Roche Molecular Biochemicals) to 1 ml of lysis buffer. Controls were in the absence of phosphatase in presence of 10 mM NaF and 10 mM sodium molybdate. Optimal results were obtained after 10 min of treatment at 30°C as longer treatment (30 min) resulted in a nearly complete loss of DNA binding.
Phosphorylation of Fusion Proteins by Casein Kinase II-Phosphorylation of bacterially expressed fusion proteins by casein kinase II (CKII) was performed with commercially available CKII (New England Biolabs) according to the manufacturer's instructions within 10 min.

Screening for XTCF-4 Reveals Different Splice Variants-In
our attempt to isolate Xenopus members of the TCF/LEF family, we recently identified a Xenopus homolog of TCF-4 (26). In addition to the published XTCF-4 sequence, which we now refer to as XTCF-4B (GenBank TM accession number AF207708), we isolated in this screen two highly identical variants that differ in two short amino acid motifs (Fig. 1A). Whereas XTCF-4A (GenBank TM accession number AF287150) contains both of these stretches, XTCF-B lacks the sequence SFLSS at position 288 -292. XTCF-4C (GenBank TM accession number AF287151) additionally lacks 4 amino acids (LVPQ) at position 256 -259. In comparison to other known members of the TCF/LEF family, XTCF-4A resembles XTCF-3, whereas XTCF-4C is similar to murine LEF-1 (see Fig. 1A). As identical gaps are present in the mLEF-1 sequence, we exclude the possibility that these sequence variations might be an artifact. Based on the fact that the multiple TCF-1 isoforms are due to alternative splicing (25), we assume that the isolated XTCF-4 variants were also generated by alternative splicing.
We also isolated Xenopus homologs of TCF-3 (GenBank TM accession number AF287149) and mLEF-1 (GenBank TM accession numbers AF287147 and AF287148) that are almost identical to the published sequences (15,27). The isolated Xenopus homologs of mLEF-1 not only lack the two described short amino acid stretches but also the region in between (Fig. 1A). Comparison with the recently characterized genomic structure of the human TCF-1 gene (25) reveals that this region is encoded by the alternatively used exon IVA of hTCF-1 (underlined in Fig. 1A). So far we have not been able to detect a Xenopus homolog of XLEF-1 or TCF-1 encompassing this exon. For functional studies, these splice variants were subcloned into pCS2ϩ containing a cytomegalovirus promoter allowing expression of the constructs in culture cell lines.
Differential Ability of TCF/LEF Transcription Factors to Activate Target Gene Expression-To compare the transactivating activities of these HMG box transcription factors, we used the Xenopus fibronectin gene as a direct target for the classical Wnt/␤-catenin pathway in Xenopus. As shown previously, in epithelial A6 cells mLEF-1 activates a Ϫ499/ϩ20 fibronectin (FN) reporter gene construct significantly stronger than XTCF-3 (Ref. 21 and Fig. 1C). We first tested whether in this assay XLEF-1 behaves in a similar fashion to its mouse counterpart even though it lacks the sequences corresponding to the TCF-1 exon IVA. Indeed, under comparable conditions XLEF-1 was able to activate the FN promoter to the same extent as the mouse homolog (Fig. 1C). We conclude that the sequence differences between mLEF-1 and XLEF-1 are nonrelevant with respect to activation of the fibronectin promoter.
Fibronectin is highly expressed in XTC fibroblasts, and this is due to an activated Wnt signaling pathway (21). We therefore analyzed the different TCF/LEF factors also in this cell line. As the Wnt pathway is already active in these cells, transfection of XLEF-1 did not lead to a further activation of the FN promoter. However, transfection of XTCF-3, XTCF-4A, or XTCF-4B led to an inhibition of FN expression (Fig. 1E). In contrast, XTCF-4C behaved as XLEF-1. Thus, those TCF/LEF factors that failed to activate the FN promoter in epithelial A6 cells are working as transcriptional repressors in XTC fibroblast, whereas those TCF/LEF factors that activate the reporter in A6 cells do not show repressing activity in XTC cells.
In all cases, activation of the reporter gene construct was dependent on the presence of a functional TCF/LEF target site that we further refer to as the Wnt-responsive element, Wnt-RE. Introduction of point mutations into the WntRE of the Ϫ499/ϩ20 fragment (21), thereby interfering with binding of the transcription factors to their target site, completely eliminated the responsiveness of the promoter upon transfection of TCF/LEF members both in epithelial A6 cells or XTC fibroblasts ( Fig. 1, D and F).
In summary, despite the high percentage of amino acid sequence identity between different variants of XTCF-4 that differ in the LVPQ and SFLSS motifs, these factors show distinguishable transactivation behavior in two different cell lines.
Differential Ability of TCF/LEF Factors to Activate Target Genes Is Not Due to Complex Formation with groucho, CtBP, or SMAD4 -In previous reports it had been speculated that LEF-1 functions as a transcriptional activator, whereas XTCF-3 acts as a repressor of transcription (22,23). This assumption was based on the following observations: (i) mLEF-1 activates expression of the homeobox transcription factor siamois and induces axis duplication in early Xenopus embryos, whereas XTCF-3 does not, and (ii) XTCF-3 binds to transcriptional repressors of the groucho family, whereas for LEF-1 contradictory results have been obtained (22,24). In agreement with this groucho-repressor model was the observation that deletion of the groucho-binding site in XTCF-3 converts the repressor into an activator (22), and we were able to confirm this observation using fibronectin as a target for Wnt signaling (21).
We therefore asked whether the differential potential of TCF/LEF members to activate the fibronectin promoter might be due to differences in their ability to interact with transcriptional repressors of the groucho family. For this purpose, a far Western was performed (9) with purified His-tagged fusion proteins of mLEF-1, XLEF-1, XTCF-3, XTFC-4A, and XTCF-4C using SDS-PAGE. After electrophoresis, proteins were transferred onto a nitrocellulose membrane and overlaid with radioactively labeled Xenopus ESG-1, a member of the groucho family of transcriptional repressors. In contrast to Roose et al. (22), but supporting Levanon et al. (24), we found that all members of the TCF/LEF family can interact with ESG-1 in vitro (Fig. 2). The cytoplasmic domain of Xenopus cell-cell adhesion molecule cadherin-11 was used as negative control demonstrating the specificity of interaction between ESG-1 and TCF/LEF factors. Based on these results, we conclude that all members of the TCF/LEF family have the potential to interact with members of the groucho family and that there is no simple correlation between the ability of a TCF/LEF factor to bind groucho members and their role as transcriptional activators or repressors, respectively.
Another transcriptional corepressor that had been shown to interact with the C-terminal region of XTCF-3 is CtBP (23).
Binding of CtBP to XTCF-3 depends on two PLSL(T/V) motifs in the C terminus of XTCF-3 that are also present in hTCF-4 but not XTCF-4. Since this region is missing in mLEF-1, this opens the possibility that CtBP might be involved in repressing fibronectin expression by XTCF-3. We therefore constructed a C-terminal deletion construct of XTCF-3 (Fig. 1B) that has been shown to be unable to bind CtBP (23), and we tested the ability of this construct to activate the FN-promoter in A6 cells.
In several independent experiments XTCF-3⌬C was unable to activate transcription from the Ϫ499/ϩ20 construct (Fig. 1C). This failure in activation is not due to a nonfunctional protein given the following two observations. First, XTCF-3⌬C is able to bind the Wnt-RE (see Fig. 5A below) and thus is functional with respect to DNA binding. Second, in XTC cells this construct was still able to repress FN expression (Fig. 1E) showing that this construct is still functional with respect to target gene regulation. This experiment together with the ability of XTCF-3⌬grg to activate fibronectin expression (21) demonstrate that CtBP is not the major corepressor component of the XTCF-3mediated transcription factor complex on the FN promoter.
A complex of SMAD4, LEF-1, and ␤-catenin has recently been shown to regulate the transcriptional activation of the Xenopus homeobox transcription factor twin (11,12). Subsequent in vitro precipitation experiments demonstrated that SMAD4 binds directly to the HMG box of LEF-1 (11). In the context of this study we therefore aimed to analyze this interaction with respect to other members of the TCF/LEF family. To determine whether SMAD4 preferentially binds to members of this family, GST pull-down experiments were performed. In vitro translated 35 S-labeled TCF/LEF proteins were precipitated with GST-SMAD4 or with GST alone. As shown in Fig.  2C, a strong interaction with SMAD4 was observed for all TCF/LEF factors tested, none of them interacting with GST alone. As a negative control, XTCF-3 lacking the HMG domain and the C terminus was used (XTCF3 (amino acids 1-322)). We conclude from these experiments that all Xenopus members of the TCF/LEF family can interact with SMAD4 thereby excluding the possibility that differential complex formation with SMAD4 accounts for the observed differences in transactivation of HMG box transcription factors.
Different Members of the TCF/LEF Family Bind to Their Target Sequence with Different Affinity-We next reasoned that the differences in transactivation activity of TCF/LEF members might be due to differences in DNA binding affinity. Different TCF/LEF fusion proteins were purified as judged by Coomassie staining (Fig. 3A), and binding affinities toward two well defined WntRE in Xenopus were determined. The DNA duplexes were derived from the TCF/LEF-binding sites found in the Xenopus fibronectin and siamois promoters (21, 30) (Fig.   3B). We recently reported that mLEF-1 specifically binds to the WntRE of the fibronectin promoter (21). Here we demonstrate that XTCF-3, XTCF-4A, and-4C are also able to bind specifically to the FN-WntRE (Fig. 3C). Whereas addition in excess of unlabeled oligonucleotides containing the FN-WntRE prevented binding of the fusion proteins to the labeled fragment, addition of an unspecific competitor did not (Fig. 3C). Identical results were obtained with the sia-WntRE indicating the specific binding of TCF/LEF factors (not shown).
To determine the apparent dissociation constant to the used oligonucleotides derived from the siamois or the FN promoter, respectively, we performed quantitative bandshift studies with variable amounts of TCF/LEF fusion proteins (Fig. 4, A and B). The percentage of bound oligonucleotide was plotted as a function of the fusion protein concentration, and the protein concentration that resulted in a binding of 50% of the oligonucleotide was estimated as the apparent K D value. The K D values observed in our experiments were within a range of 0.3 and 4.6 M and thus within the same order of magnitude as recently determined for the murine LEF-1 HMG box only (31). In addition, to verify these constants by a different plotting procedure we performed classical Scatchard plot analysis. For this purpose, the ratio of bound protein to free protein concentration (bound/free) is plotted versus the concentration of bound protein. By using this procedure, the K D value can be estimated by using the slope of the obtained line. Additionally, the total number of binding sites represented by the used oligonucleotide can also be estimated by the abscissa intercept (for a more detailed description of Scatchard plots see Ref. 32).
The most obvious outcome of these experiments is the fact that all tested fusion proteins have a significant higher affinity toward the sia-WntRE than the FN-WntRE. However, the absolute difference in binding affinity was dependent on the fusion protein used. We found for mLEF-1 a 2-fold higher affinity toward the siamois-derived target site in comparison to the FN-derived sequence, whereas the fold difference in binding affinity for the others were as follows: XTCF-4A, 2-fold higher; XTCF-4C, 5-fold higher; and XTCF-3, 5-fold higher (Fig. 4, A  and B). One major difference between the two target oligonu-

FIG. 2. Interaction of TCF/LEF transcription factors with ESG-1 groucho.
A, fusion proteins were purified on nickel-agarose and separated on polyacrylamide gels. XLEF-1 and mLEF-1 were stained by Coomassie (left), whereas the other proteins (right) were detected by an anti-His-tag antibody (Qiagen). The asterisk labels the band for XLEF-1. Molecular weights are indicated. Xcad-11 is a His-tagged fusion protein of the intracellular domain of Xenopus cadherin-11 that served as a negative control (29). B, overlay blot with 35 S-labeled Xenopus ESG-1, a member of the groucho family. Lanes were as in A. Note that all proteins are interacting with ESG-1 the exception of Xenopus cadherin-11. C, SMAD4 binds directly to TCF/LEF transcription factors. In vitro translated 35 S-labeled TCF/LEF proteins (lanes 1-6) were precipitated with GST (lanes 7-12) or SMAD4-GST (lanes 13-18). All tested factors bound GST-SMAD4 except TCF-3 lacking the HMG domain and C terminus. cleotides became obvious by comparing the binding affinities for XTCF-3. Whereas in case of the FN-WntRE XTCF-3 displays sigmoidal binding with comparable low affinity, it exhibits a hyperbolic binding to the sia-WntRE with a significant higher affinity. A similar behavior was observed for XTCF-4C. XTCF-4A showed sigmoidal binding to both oligonucleotides. In these cases we were not able to plot linear Scatchard plots but obtained graphs with a maximum, which is an indication for positive cooperative binding behavior (32). This lower binding affinity toward the FN-WntRE is most likely not due to protein misfolding or degradation as this experiment was done with different protein preparations, and the same batches of proteins showed a higher binding affinity toward the sia-WntRE. These data raise the possibility that the two different bases in the core sequence as well as the flanking regions influence the binding behavior of a given HMG box transcription factor. In addition, the extended C termini of XTCF-3 and XTCF-4 might influence the binding characteristics on the FN-derived oligonucleotide. Indeed, we found that truncation of the C terminus of XTCF-3 significantly shifts the binding behavior toward the mLEF-1-derived one (Fig. 4C).
Most strikingly, although bacterially expressed XTCF-4C binds to the FN-WntRE in vitro with much lower affinity than mLEF-1, it is a more potent in vivo activator of the FN promoter than mLEF-1. We conclude that although members of the TCF/LEF family differ in their in vitro DNA binding affinities toward different promoter target sites, this difference does not correlate with their transactivation potentials.
Members of the TCF/LEF Family Differentially Bind ␤-Cate-nin When Bound to DNA-We next asked whether there might be differences in complex formation mediated by the analyzed TCF/LEF factors in vivo. We therefore transfected epithelial A6 cells with expression constructs encoding for the transcription factors. After 3 days we prepared cell lysates under mild conditions and analyzed the behavior of the overexpressed proteins in gel shift assays. Western blot analysis of these lysates revealed that all overexpressed proteins were expressed at comparable levels (Fig. 5C). Although XTCF-4A and XTCF-4C are of nearly identical size, the formed DNA-protein complexes show different migrating behavior (Fig. 5A). The XTCF-4C induced DNA-protein complex was of larger size compared with the XTCF-4A one. Whereas the mLEF-1-DNA complex behaved similar to the XTCF-4C-DNA complex, the XTCF-3-induced complex migrates as fast as the XTCF-4A one. Thus, the ability of these transcription factors to trigger target gene activation is paralleled by the migration behavior of the corresponding DNA-protein complexes. These differences are probably not due to differences in DNA bending as we never observed such differences when using bacterially expressed fusion proteins (see Fig. 3 as example). This raises the possibility that the observed differences in migration behavior reflect differential complex formation with transcriptional coactivators, namely ␤-catenin, rather than corepressors. We aimed to study this possibility by supershift analyses. We focused on the different TCF-4 variants as antibodies are commercially available that specifically recognize XTCF-4 and ␤-catenin. Our attempt to include mLEF-1 and XTCF-3 into this study was hampered by the lack of antibodies For the supershift experiments the buffer and gel conditions were slightly altered compared with the experiments shown in Fig. 5A to achieve a higher resolution for the slower migrating complexes (Fig. 5B). This led to the detection of a faster migrating complex in XTCF-4B-and XTCF-4C-transfected cells that runs at the level of the complex also found in parental and XTCF-4A-transfected cells (for reasons described below and labeled as TCF/DNA in Fig. 5B). Whereas upon XTCF-4A transfection and in untransfected cells only this faster migrating complex was seen, XTCF-4B and XTCF-4C transfection revealed an additional slower migrating complex that corresponds to the main signal seen in Fig. 5A. One additional faster migrating band was also observable in untransfected cells indicating that this signal is of endogenous origin and not due to the overexpressed TCFs (labeled as non TCF in Fig. 5B).
For all tested XTCF-4 variants, addition of an antibody against TCF-4 supershifted the faster migrating band, indicat- ing that this band represents an XTCF-4-DNA complex. However, the slower migrating upper band was not supershifted by addition of TCF-4 antibody indicating that most probably the binding epitope for the TCF-4 antibody may be blocked by an additional protein. On the other site, adding an antibody against ␤-catenin interfered with formation of the upper band and resulted in a strong increase in the TCF-4/DNA band (Fig.  5B) identifying ␤-catenin as the additional protein in the slower migrating band. Thus, the slower migrating band that we observed in XTCF-4B and XTCF-4C cells, but not in those transfected with XTCF-4A, indicates the presence of a XTCF-4-␤-catenin-DNA trimer. This difference in complex formation with ␤-catenin is not due to an up-regulation of the overall amount of ␤-catenin as judged by Western blot studies of transfected cells (Fig. 5D). As the TCF-4 variants used differ in the LVPQ and SFLSS motifs, this indicates that the complex formation of the analyzed XTCF-4 variants with ␤-catenin is directly or indirectly dependent on the two described amino acid motifs. Another striking observation obtained by these experiments using cell lysates was that in vivo expressed XTCF-4A apparently binds with lower affinity to the FN-Wn-tRE than XTCF-4B and -C, although all proteins were expressed at comparable levels (see Fig. 5, B and C).
TCF/LEF-␤-Catenin Complex Formation Is Dependent on the Phosphorylation of TCF-By comparing the expression level of TCF-4 variants in cell lysates, we made the observation that XTCF-4B and XTCF-4C appeared in Western blots as a double band (Fig. 5C). This raises the question whether there are differentially phosphorylated forms of XTCF-4. Indeed, treating the lysates with calf intestine alkaline phosphatase, as well as acidic potato phosphatase, resulted in an increased appearance of a faster migrating band in all XTCF-4 variants (Fig. 6A). We conclude from these data that members of the TCF family are the subject of phosphorylation in vivo. As we never observed an XTCF-4A-␤-catenin-DNA complex in our bandshift experiments (Fig. 5B) or the faster migrating form of XTCF-4A in the absence of phosphatase activities (Fig. 6A), we reasoned that the complex formation of XTCF-4 with ␤-catenin might be dependent on the existence of a dephosphorylated form. By having established the identity of different retarded FIG. 5. TCF-␤-catenin complex formation on the fibronectin oligonucleotide. A, epithelial A6 cells were transfected with expression constructs as indicated, and cell lysates prepared in the presence of phosphatase inhibitors were assayed for their ability to bind specifically the fibronectin-derived oligonucleotide. The unspecific competition was performed as described under "Experimental Procedures." The data for XTCF-3⌬C are from an independent experiment in comparison to the other shown results. B, bandshifts were done as in A in presence of phosphatase inhibitors, but antibodies against TCF-4 or ␤-catenin were added. Note that XTCF-4B and XTCF-4C display two distinct signals in the absence of antibodies. The faster migrating signal can be supershifted by adding TCF-4-specific antibodies. Formation of the slower migrating complex is prevented by adding ␤-catenin antibodies indicating the presence of ␤-catenin in the slower migrating complex. XTCF-4A transfectants never showed this slower migrating complex. C, immunoblot of A6 cells transfected as indicated and probed for expression of TCF-4. All constructs are expressed at similar levels. Note that XTCF-4B and XTCF-4C clearly show two distinct bands (arrowheads), whereas the second signal is barely visible for XTCF-4A. D, immunoblot for ␤-catenin protein (arrowhead) in transfected A6 cells showing no significant differences in expression level of ␤-catenin after transfection. bands in our mobility shift studies, we therefore repeated these studies with XTCF-4A transfectants in the presence or absence of exogenously added phosphatase activity. These experiments revealed two important observations. First, the overall binding affinity of XTCF-4A decreased after phosphatase treatment because we needed comparable longer exposure times to detect signals. This loss in binding is most probably not due to any protease contaminations since all experiments were done in the presence of protease inhibitors. Furthermore, we tested the added phosphatases negative for any protease contaminations by treating purified TCF/LEF fusion proteins under identical experimental conditions (inset, Fig. 6B). Therefore, this observation suggests that the binding behavior of TCF/LEF factors is dependent on pre-phosphoryla- tion. A similar effect has recently been described for Drosophila HMG1 proteins, non-sequence-specific, DNA-binding proteins containing several HMG boxes (33). Pre-phosphorylation has been shown to be casein kinase II (CKII)-dependent for HMG1 proteins. As all TCF/LEF transcription factors have multiple CKII target sites and CKII is a ubiquitously found enzyme, we tested whether these factors are phosphorylated by CKII and whether phosphorylation influences the binding behavior of TCF/LEF factors. In fact, all tested TCF/LEF factors were found to be phosphorylated by CKII (Fig. 6C), and this phosphorylation was accompanied by an increased binding affinity (Fig. 6D).
Second and more important, these experiments indicate that the formation of a TCF-4-␤-catenin-DNA complex is dependent on the presence of a partially dephosphorylated form of the transcription factor (Fig. 6B). Treatment of cell lysates with phosphatases did not only result in dephosphorylation of the proteins but also in formation of the slower migrating complex indicating entrance of ␤-catenin into the complex.
To compare the results shown in Fig. 6B and in Fig. 5B, and to test whether this dephosphorylation might also occur in vivo, we prepared lysates of TCF-4A transfectants in the presence or absence of added phosphatase inhibitors. As shown in Fig. 6B, addition of a phosphatase inhibitor mixture to the lysis buffer (as in Fig. 5B) prevented formation of TCF-4-␤-catenin-DNA complexes, whereas a lack of phosphatase inhibitors resulted in formation of the slower migrating band. This experiment clearly indicates that endogenous phosphatases are able to dephosphorylate TCF/LEF factors and that most probably this dephosphorylation results in the recruitment of ␤-catenin into the TCF-DNA complex. DISCUSSION We report herein that different, highly similar XTCF-4 variants show divergent behavior in target gene activation assays. By using different biochemical approaches, we provide evidence that these differences are neither due to preferential interaction with corepressors of the groucho family or CtBP nor to differential binding of the transcriptional activator SMAD4. We demonstrate that the formation of a TCF-␤-catenin-DNA complex and the activation of the target gene fibronectin depends on two short amino acid motifs. These motifs, LVPQ and SFLSS, are involved in regulating posttranslational modifications of XTCF-4. Since they are not present within all variants of TCF/LEF factors (e.g. mLEF-1 versus XTCF-3), and since cofactors of the TCF/LEF complex are expressed in a tissuespecific manner, a complex pattern of regulation is emerging that has to be considered when studying Wnt signaling.
Implications on TCF/LEF-mediated Complex Formation-In our attempts to elucidate the molecular mechanism responsible for the different transactivation properties of TCF/ LEF factors, we here present data that support multiple levels of regulation of TCF/LEF-mediated transcription factor complex formation.
First, there is no correlation between DNA binding affinities of TCF/LEF factors and target gene activation. The transcriptional activator XTCF-4 C has a significant lower DNA binding affinity than mLEF-1 but activates the expression from the FN promoter stronger than mLEF-1.
Second, although the oligonucleotides used for quantitative band shift analyses both contain the canonical TCF/LEF target site, (C/G)TTTG(A/T)(A/T), they show significant differences in complex formation with TCF/LEF. All studied transcription factors showed a higher affinity toward the sia-WntRE than toward the FN-WntRE. These differences in binding affinity and behavior of TCF/LEF transcription factors might indicate that the nucleotides flanking the conserved core region of the target site influence the binding affinity by providing additional contacts. These putative interactions do not only modify DNA binding strength but also the type of binding behavior. XTCF-3 displays hyperbolic binding behavior on the sia-Wnt-RE but sigmoidal binding behavior on the FN-WntRE indicating additional protein/DNA interactions. In accordance with this hypothesis is the observation that deleting the C-terminal domain of XTCF-3 shifts the DNA binding curve of this mutant toward the mLEF-1-derived one. Further experiments have to confirm whether these additional, promoter-specific protein/ DNA interactions mediated by the C-terminal part of XTCF-3 can be proven and, if so, whether this holds true for XTCF-4C as well. This is of general interest, since it has recently been shown that LEF-␤-catenin-SMAD4 complexes are involved in regulating the twin promoter (11,12), although the same factors are not involved in regulating the c-myc gene (11). Thus, with respect to activation of Wnt target genes for future promoter analysis, the fine structure of the analyzed promoter region has to be taken into account.
Third, we found, that at least in vitro, all TCF/LEF transcription factors are able to interact with ESG-1, a member of the groucho family. This is of general interest as it is in contrast to an earlier hypothesis suggesting that differences in TCF/LEF-mediated transactivation is a biased ability of these factors to interact with transcriptional corepressors. Although we acknowledge that there still might be differences in the ability to interact in vivo, our results clearly show that additional levels of complexity with respect to TCF/LEF-mediated complex formation have to be postulated. The only tested TCF/ LEF-groucho protein interaction using Xenopus proteins described so far showed that XTCF-3 can interact with Xgrg-5. However, this result was only based on two-hybrid analyses (21). The same study further provides indirect evidence for an interaction of XTCF-3 with Xgrg-5 and Xgrg-4, since both are translocated to the nucleus in XTCF-3-transfected COS-1 cells. Further experiments are required to analyze the in vivo interactions of all known TCF/LEF and groucho proteins from one species (e.g. Xenopus or mouse) under different cellular conditions. Fourth, another important and striking observation is the fact that complex formation of XTCF-4B with DNA and ␤-catenin is not sufficient for target gene activation. This is in agreement with a previous observation in a different cell culture system (34). In this case, ␤-catenin and LEF-1 can activate target gene expression in transformed Jurkat T-cells but not in normal T-lymphocytes despite the nuclear localization of both factors. Thus, additional tissue-specific components of TCF/ LEF-mediated transcription factor complexes or additional posttranslational modifications have to be postulated. Of the known components of TCF/LEF-mediated transcription factor complexes, neither p300/CBP nor TBP are expressed in a tissue-specific manner. However, in Xenopus we recently described that Xpontin and Xreptin are expressed in a tissuespecific manner (18). Also some groucho members show a distinct expression pattern (35). Beside the tissue-specific expression of cofactors, tissue-specific posttranslational modifications have to be taken into account.
Protein Phosphorylation as an Additional Level of Regulation in TCF/LEF-mediated Complex Formation-The experiments described here clearly identify phosphorylation as an additional level of TCF/LEF regulation. First, we found a strict correlation between the presence of a faster migrating dephosphorylated band of XTCF-4B and XTCF-4C and the formation of a ternary complex of DNA, TCF-4, and ␤-catenin. A dephosphorylated band in Western blots was never observed for XTCF-4A which was paralleled by the absence of a DNA su-pershift. However, treatment of cell lysates derived from XTCF-4A-transfected cells with different phosphatases resulted in the appearance of the dephosphorylated form and, in parallel, in the formation of a slower migrating band in electrophoretic mobility studies. In other words, dephosphorylation of XTCF-4 most probably at one or more specific sites allows complex formation with ␤-catenin, whereas phosphorylation of XTCF-4 prevents complex formation. Since complex formation between TCF/LEF transcription factors and ␤-catenin is a prerequisite for target gene activation, signal transduction pathways that lead to a phosphorylation of TCF/LEF factors thus might function as inhibitors of Wnt signaling. One candidate kinase that might be responsible for this effect is NLK (nemolike kinase) as it phosphorylates TCF-4 and TCF-3 and inhibits TCF-␤-catenin complex formation (36,37). With respect to these previous publications our results offer an attractive hypothesis. As deletion of SFLSS results in the appearance of the dephosphorylated form of XTCF-4, this sequence motif might either represent the target site of NLK phosphorylation or might be involved in TCF-4/NLK interaction. The observation that the phosphorylated form of XTCF-4 is also present in those variants that lack the SFLSS motif excludes the first possibility and strengthens the latter one. We also note that this motif is serine-rich and thus might also be subject to phosphorylation events that regulate DNA/protein interactions. Further supporting the idea that posttranslational modifications regulate complex formation of TCF/LEF factors with ␤-catenin in vivo, we found no differences in interaction when using bacterially expressed fusion proteins in coimmunoprecipitation or pulldown studies (not shown). Although the absence of the SFLSS motif allows dephosphorylation and complex formation with ␤-catenin, it is not sufficient to turn XTCF-4B into an efficient activator of transcription. However, this is achieved when both SFLSS and LVPQ motifs are deleted indicating another important regulatory role for the latter. Further experiments will reveal whether this short motif is involved in protein-protein interactions and how those might influence the transactivation behavior of TCFs.
Furthermore, we observed that the overall DNA affinity of TCF/LEF transcription factors decreased when they were dephosphorylated. In a reverse experiment we were able to increase the DNA binding affinity of unphosphorylated bacterially expressed fusion proteins by treatment with casein kinase II. All of the tested TCF/LEF factors have several CKII target site motifs of (S/T)XXE. Two of these are located within the HMG box and represent the only positionally conserved target sites for CKII. Although we have no evidence that this kind of phosphorylation occurs in vivo, these experiments clearly implicate an additional possibility to modify the outcome of Wnt signaling.
Implications for Our Understanding of Wnt Signaling-In summary, we show that complex formation of TCF/LEF factors with ␤-catenin correlates with posttranslational phosphorylation/dephosphorylation events. Furthermore, we provide additional evidence that binding of TCF/LEF factors with ␤-catenin is not necessarily linked with transcriptional activation. In addition to an active Wnt signal that leads to accumulation of ␤-catenin in the nucleus, additional signals are required to convert TCF/LEF into a transcriptional activator. Beside the SFLSS motif that is involved in regulation of TCF-␤-catenin complex formation, we provide evidence that the LVPQ motif is involved in turning TCF/LEFs into a transcriptional activator and might be involved in interactions with transcriptional repressors. Thus, the outcome of a Wnt signal is additionally regulated at the level of transcription factors of the TCF/LEF family. With respect to TCF/LEF-mediated complex formation, our data also highlight the differences between in vivo and in vitro experiments. Further studies regarding regulation of Wnt signaling at the level of transcription factors thus have to consider the following: (i) which splice variants of the transcription factors are expressed, (ii) their posttranslational modifications, (iii) what kind of cofactors are expressed in these cells, and (iv) the actual components of the finally assembled complex. This regulation might be by far more complex than previously thought.