Wnt controls the transcriptional activity of Kaiso through CK1ε-dependent phosphorylation of p120-catenin

The transmembrane E-cadherin protein constitutes the core of the adherens junction through its homophilic interaction with identical proteins from neighboring cells. E-cadherin function is controlled by a family of proteins called catenins, which bind to its cytosolic domain. Among these, β-catenin is required for recruiting the actin cytoskeleton. In addition to its function in cell adhesion, β-catenin is a central player in the Wnt pathway (Reya and Clevers, 2005). When released from the junctional complex, β-catenin translocates to the nucleus, where it interacts with the Tcf family of transcriptional factors and regulates the expression of a variety of genes involved in embryonic development and tumorigenesis. The translocation of β-catenin to the nucleus is stimulated by Wnt ligands that prevent the degradation of cytosolic β-catenin.

Another E-cadherin-associated protein, p120-catenin (also known as catenin δ-1), is also involved in the regulation of transcription. p120-catenin binds to a distinct site of E-cadherin to that bound by β-catenin and is necessary for the stabilization of E-cadherin at the cell membrane (Davis et al., 2003). Moreover, p120-catenin interacts with the transcriptional factor Kaiso (Daniel and Reynolds, 1999). Studies performed in Xenopus laevis have demonstrated that p120-catenin relieves the repression caused by Kaiso on Wnt target genes (Park et al., 2005). The inhibition by Kaiso has been attributed to its direct interaction with 5′-CTGCNA-3′ motifs present in Wnt target promoters, such as that encoding matrilysin (Spring et al., 2005). However, recent results have questioned this model because, according to Ruzov et al. (Ruzov et al., 2009a; Ruzov et al., 2009b), Kaiso binds to these motifs with low affinity. Alternatively, these authors proposed that inhibition of the Wnt pathway by Kaiso is dependent on its association with Tcf-3 or Tcf-4, which precludes the binding of this factor to DNA. Moreover, Kaiso is also capable of repressing the expression of other genes not modulated by β-catenin, through direct interaction with methylated CpG sequences in their promoters (Ruzov et al., 2009b).

p120-catenin was initially described as a v-Src (viral Src) substrate related to transformation (Reynolds et al., 1989). p120-catenin phosphorylation occurs not only on tyrosine but also on serine and threonine residues, which are mainly present in the N-terminal domain of the protein (Mariner et al., 2001; Xia et al., 2003). p120-catenin has several isoforms, which differ in this N-terminal regulatory domain. p120-catenin 1 has a 347-amino-acid N-terminal domain (the full-length protein, see Fig. 1A); isoforms 2 and 3 start at amino acids 55 and 102, respectively. p120-catenin 3 is the most abundant isoform in epithelial cells, which also express a shorter form, p120-catenin 4. p120-catenin 4 is devoid of the N-terminal domain, it consists of only the central armadillo domain, with the ten tandem 42-amino-acid repeats, and the C-terminal domain (Reynolds and Roczniak-Ferguson, 2004). Association of proteins to p120-catenin takes place through different elements; whereas Kaiso and E-cadherin interact with the central armadillo domain, other cofactors, such as the Fer or Fyn tyrosine kinases (Kim and Wong, 1995) and RhoA GTPase (Castaño et al., 2007; Yanigisawa et al., 2008), bind sequences in the N-terminal domain.

Although the modification of many different tyrosine, serine and threonine residues has been described, it is only in a few cases that the relevance of these modifications in the interaction of p120-catenin with the different cofactors is known. We have recently described that p120-catenin is phosphorylated on Ser268 and Ser269 by a mechanism depending on CK1ε activation (Casagolda et al., 2010). This protein kinase is stimulated by Wnt ligands that induce p120-catenin phosphorylation at this residue. Moreover, this modification has a functional significance, as it inhibits the interaction between p120-catenin and E-cadherin (Casagolda et al., 2010). We report here that Ser268 phosphorylation also increases binding of p120-catenin to Kaiso and modulates the effects of this transcriptional repressor.

Phosphorylation of p120-catenin by CK1 decreases its interaction with E-cadherin. Accordingly, upon phosphorylation, the interaction of p120-catenin with a GST-fusion protein comprising the cytosolic domain of E-cadherin (cytoE-cadh) is downregulated (Casagolda et al., 2010). As the p120-catenin armadillo domain is also the binding site for the Kaiso transcriptional factor, we analyzed the interaction of p120-catenin with Kaiso. We found that p120-catenin did not associate with Kaiso in the presence of E-cadherin (i.e. cytoE-cadh could compete with Kaiso to bind p120-catenin; supplementary material Fig. S1).

In contrast to the effect detected with E-cadherin, phosphorylation of p120-catenin by CK1 upregulated the association of p120-catenin with Kaiso (Fig. 1B). We analyzed the upregulation of Kaiso binding using different p120-catenin isoforms. As shown in Fig. 1B,C, p120-catenin 1 (amino acids 1–911) and 3 (amino acids 102–911) bound more Kaiso after phosphorylation by the catalytic fragment of CK1. Compared with p120-catenin 1, which is more abundant in mesenchymal cells, the epithelial isoform p120-catenin 3 (amino acids 102–911) presented a lower level of association with Kaiso in basal conditions, but bound more Kaiso than isoform 1 after CK1 phosphorylation (Fig. 1C). By contrast, a p120-catenin fragment similar to isoform 4 (amino acids 350–911) showed a lower level of interaction than the other isoforms and the binding was insensitive to CK1 phosphorylation.

Ser268 and Ser269 have been shown to be relevant in the regulation of the p120-catenin–E-cadherin interaction (Casagolda et al., 2010). To test whether these two residues were also involved in the association of p120-catenin with Kaiso, both residues were replaced by alanine residues (mutant S→A), thereby preventing their phosphorylation in a GST–p120-catenin (isoform 3) fusion protein. As shown in Fig. 1D, p120-catenin S→A was still sensitive to phosphorylation by CK1, although to a lesser extent than was the wild-type protein. Thus, the increase in Kaiso binding induced by CK1 was smaller in the S→A mutant than in the wild-type form. Moreover, the S→D mutant associated with more Kaiso than the wild-type form, although not as much as p120-catenin phosphorylated by CK1. These results indicate that, although phosphorylation of Ser268 and Ser269 contributes to the increased binding of Kaiso observed upon its phosphorylation with CK1, the modification of other, still unidentified, residues is also relevant.

Phosphorylation-specific monoclonal antibodies have been generated against several phosphorylated residues of p120-catenin (Xia et al., 2004). We used a monoclonal antibody specific for Ser268-P as a molecular tool to analyze the phosphorylation of Ser268 in cells. Eukaryotic expression plasmids encoding GST–p120-catenin (isoform 3) fusion proteins were expressed in SW-480 cells. Ser268-P was detected when GST–p120-catenin was transfected, but not upon transfection with the S→A or the S→D mutant constructs (Fig. 1E). Binding of Kaiso to p120-catenin was slightly decreased in the S→A mutant with respect to that in the control, whereas the S→D mutant showed a higher level of association (Fig. 1E).

Because CK1ε is required for phosphorylation of Ser268, we analyzed the effects of depletion of this isoform on Kaiso–p120-catenin binding. A short hairpin RNA (shRNA), specific for this isoform, markedly repressed the level of CK1ε (supplementary material Fig. S2A), by 75–90% in different experiments, as estimated by quantifying CK1ε levels. Transfection of a plasmid encoding this shRNA, but not that encoding a control scrambled sequence, downregulated the phosphorylation of p120-catenin Ser268 and the binding of p120-catenin to Kaiso, whereas the association with E-cadherin was increased (supplementary material Fig. S2A). Similar effects on the phosphorylation of p120-catenin Ser268 were detected after addition of a specific inhibitor of CK1, IC261 (Mashoon et al., 2000). Addition of this compound decreased the level of p120-catenin Ser268-P, increased the p120-catenin–E-cadherin association and decreased the p120-catenin–Kaiso association (supplementary material Fig. S2B).

Wnt3a is known to stimulate Ser268 phosphorylation (Casagolda et al., 2010) (see also Fig. 2C). Accordingly, the interaction of p120-catenin with E-cadherin was diminished by this factor in all the cell lines examined, such as in HT-29 M6 cells (Fig. 2A) that mostly express p120-catenin 1, and in SW-480 ADH cells, a subpopulation of SW-480 cells that express higher levels of E-cadherin and p120-catenin 3 (Fig. 2C) (see Materials and Methods). By contrast, Wnt3a upregulated the p120-catenin interaction with Kaiso in these two cell lines, given that Kaiso was enriched in p120-catenin immunoprecipitates. Similar results were obtained in HEK-293 fibroblasts (Fig. 2B), which express both p120-catenin isoforms. Consistent with the higher level of stimulation of Kaiso binding by CK1, and despite the predominant expression of p120-catenin 1, p120-catenin 3 was preferentially detected in the Kaiso immunocomplexes upon Wnt stimulation in these cells (Fig. 2B). Wnt-dependent upregulation of the p120-catenin–Kaiso interaction was dependent on the activity of CK1ε, given that the depletion of this enzyme by a specific shRNA prevented this increase (Fig. 2C).

Kaiso is known to be a bimodal transcriptional regulator because it is capable of directly binding DNA (Prokhortchouk et al., 2001; Daniel et al., 2002) and also associates with Tcf-4, thereby precluding the interaction of Tcf-4 with specific promoters (Ruzov et al., 2009a). Different Kaiso amino acid sequences are required for the interaction with Tcf-4 and DNA (Ruzov et al., 2009a).

Two different type of DNA-binding elements have been described for Kaiso: 5′-CTGCNA-3′ motifs, such as those present in the MMP7 (encoding matrilysin) promoter (Daniel et al., 2002), called KBS elements, and methylated CpG sequences, such as those present in the S100A4 (encoding metastasin) promoter (Prokhortchouk et al., 2001; Daniel et al., 2002). First, we determined the relative binding of Kaiso to these two elements. In these assays, we used equal amounts of the biotin-labeled oligonucleotides containing the KBS sequence from the MMP7 promoter (KBS) and the CpG element from S100A4 promoter, either methylated (^MeCpG) or non-methylated (CpG). DNA-bound proteins were purified by chromatography on streptavidin–agarose and analyzed by western blot [biotinylated-oligonucleotide-binding assays (BOPA), see Materials and Methods]. As shown in Fig. 3A, Kaiso bound to the methylated oligonucleotide better than to the KBS sequence; interaction with the non-methylated oligonucleotide was not detected. Therefore, we focused our attention on the effect of p120-catenin on the interaction of Kaiso with ^MeCpG elements.

We next determined whether Kaiso binding to methylated DNA was modified by its association with p120-catenin. Preincubation with p120-catenin did not substantially modify the association of recombinant Kaiso with the methylated oligonucleotide, unless a high excess of phosphorylated p120-catenin was used (Fig. 3B). Similar results were obtained with p120-catenin 1 (1–911) and 3 (102–911); therefore, p120-catenin does not substantially prevent the interaction of Kaiso with ^MeCpG. By contrast, p120-catenin interfered with the association of Kaiso and Tcf-4. Pull-down assays indicated that binding of Kaiso to Tcf-4 and p120-catenin was mutually incompatible. As shown in Fig. 3C, preincubation with p120-catenin, either isoform 1 or 3, decreased the amount of Tcf-4 associated with GST–Kaiso.

We also examined the effect of Kaiso on the Tcf-4 association with DNA, using an oligonucleotide containing the binding sequence for this factor that had been previously characterized in the Myc promoter. In accordance with previous reports (Ruzov et al., 2009a), binding of Tcf-4 to this oligonucleotide was prevented by the addition of recombinant Kaiso (Fig. 3D). Preincubation with p120-catenin reversed the inhibitory effect of Kaiso on Tcf-4 binding to the oligonucleotide. Next, we determined whether the Kaiso interaction with Tcf-4 and methylated DNA was mutually incompatible. Binding of Kaiso to methylated DNA was analyzed using BOPAs. As shown in Fig. 3E, the amount of Kaiso associated with the DNA was lower when the binding assay was supplemented with an extract from cells overexpressing Tcf-4 compared with when using a control extract from pcDNA3-transfected cells. Moreover, Tcf-4 was not detected in the DNA-bound fraction, demonstrating further that the Kaiso–Tcf-4 complex does not interact with methylated DNA.

Finally, we determined the effect of p120-catenin on the interaction of Kaiso with ^MeCpG in the presence of Tcf-4. Recombinant p120-catenin remarkably upregulated Kaiso–^MeCpG binding when Tcf-4 was also present in the reaction (Fig. 3F). Therefore, these results indicate that the Tcf-4–Kaiso complex prevents the interaction of both proteins with their respective DNA target elements, and disruption of this interaction by the association of p120-catenin with Kaiso relieves this blockage.

Because the binding of Kaiso to Tcf-4 was affected by p120-catenin, we also determined whether Wnt3a also regulated this interaction, as would be expected. When endogenous (Fig. 4A) or ectopic Tcf-4 (Fig. 4B) was immunoprecipitated with specific monoclonal antibodies, Kaiso was detected in both immunocomplexes. The reverse co-immunoprecipitation gave compatible results, as Tcf-4 was also present in Kaiso immunocomplexes (Fig. 4C), further demonstrating that these two proteins associated in SW-480 cells. The presence of Wnt3a downregulated this interaction, as it decreased the levels of Tcf-4 immunoprecipitated with Kaiso (Fig. 4C) and Kaiso with Tcf-4 (Fig. 4A,B). Interference of p120-catenin (Fig. 4A,C) or CK1ε (Fig. 4B) prevented the Wnt-induced downregulation of the Kaiso–Tcf-4 interaction. These results were validated in another cell line; as shown in Fig. 4D, in HT-29 M6 cells, Wnt3a also decreased the level of the Kaiso–Tcf-4 interaction and upregulated that of p120-catenin–Kaiso.

In accordance with these results, Kaiso overexpression blocked the Wnt-induced stimulation of β-catenin transcriptional activity, determined using the TOP reporter plasmid (Fig. 5). This Kaiso-mediated repression was relieved by p120-catenin overexpression, in agreement with the role of p120-catenin in preventing the interaction of Kaiso with Tcf-4. A similar effect on Kaiso-mediated repression was observed upon transfecting Tcf-4, suggesting that the effects of Kaiso on β-catenin transcriptional activity are due to inactivation of Tcf-4 (Fig. 5). Finally, and surprisingly, β-catenin overexpression also partially overcame the Kaiso-mediated repression of TOP activity, suggesting that β-catenin also affects the Tcf-4–Kaiso interaction.

The regulation by β-catenin on the Kaiso–Tcf-4 interaction was further explored. Because β-catenin contains an armadillo domain, which is similar to the p120-catenin domain involved in its binding to Kaiso, we considered the possibility that β-catenin also interacted with Kaiso. Pull-down assays performed with recombinant proteins confirmed that this interaction occurred. As shown in Fig. 6A–C and supplementary material Fig. S3, β-catenin was co-immunoprecipitated with a GST–Kaiso fusion protein and not with GST. The β-catenin interaction with Kaiso takes place through the first six repeats in the β-catenin armadillo domain, given that a fusion protein containing this β-catenin fragment was efficiently bound by GST–Kaiso (Fig. 6A). In addition, p120-catenin competed for the association of Kaiso with β-catenin (supplementary material Fig. S3), suggesting that both catenins interact with the same Kaiso domain. However, the association of β-catenin was less efficient than that of p120-catenin; when similar amounts of both recombinant catenins were used in the assay, more p120-catenin was found associated with Kaiso (Fig. 6B). Quantification experiments indicated that β-catenin binding to Kaiso is only 10% of the p120-catenin binding to Kaiso. β-catenin also prevented the association of Kaiso with Tcf-4. As shown in Fig. 6C, the amount of Tcf-4 retained by GST–Kaiso was decreased by a similar extent to that with p120-catenin when the binding assays were performed in the presence of recombinant β-catenin. Finally, and as expected, because the Kaiso-binding site in β-catenin partially overlapped with that of Tcf-4 (armadillo repeats 3–10) (Graham et al., 2000), addition of recombinant Kaiso decreased the amount of β-catenin associated with a fragment of Tcf-4 containing the first 53 amino acids of the protein (supplementary material Fig. S4). This Tcf-4 element contains the β-catenin-binding site (Miravet et al., 2002) but not the Kaiso-binding sequence, which is located in the HMG domain (Ruzov et al., 2009a); consequently Kaiso is not retained with a GST–Tcf4 (1–53) fusion protein.

The interaction between Kaiso and β-catenin was also detected by co-immunoprecipitation experiments. β-catenin was detected in Kaiso immunocomplexes obtained from SW-480 cells (Fig. 6D); the amount of β-catenin associated to Kaiso was decreased upon Wnt3a stimulation further confirming the specificity of this interaction. As controls for this experiment, Wnt3a was found to stimulate the Kaiso interaction with p120-catenin, whereas it decreased its binding to Tcf-4. In a similar manner, the downregulation of the Kaiso–β-catenin association was dependent on p120-catenin, as depletion of p120-catenin using a specific shRNA prevented it (Fig. 6D). The Kaiso–β-catenin interaction was confirmed by the reverse co-immunoprecipitation; thus, Kaiso was detected in β-catenin immunoprecipitates (Fig. 6E). As above, Wnt3a decreased this association and stimulated β-catenin binding to Tcf-4 (Fig. 6E). Similar results were obtained in HT-29 M6 cells (Fig. 6F), where Wnt3a induced the disruption of Kaiso–Tcf-4 and Kaiso–β-catenin interactions.

We also analyzed the subcellular compartment where the p120-catenin–Kaiso interaction takes place. Cellular extracts were separated into a nuclear fraction and a cytosolic fraction (which also contained membrane proteins). Expression of markers specific for the nucleus (lamin), cytosol (pyruvate kinase) or membrane (E-cadherin) demonstrate that the fractionation worked correctly. Wnt induced a partial translocation of p120-catenin from the cytosolic and membrane fraction to the nucleus (Fig. 7A). Quantification of different experiments indicated that the percentage of p120-catenin in the nucleus increased from 17 to 55% after 20 hours of incubation with Wnt3a (Fig. 7D). By contrast, Kaiso, which was almost totally nuclear in non-stimulated cells, was partially exported from the nucleus upon Wnt addition, when ~50% of Kaiso protein was detected in the cytosol (Fig. 7A,B,D). A timecourse analysis indicated that this translocation was detected after 6 hours of incubation with Wnt3a in SW-480 cells (Fig. 7A). Tcf-4 subcellular distribution was not altered by this factor.

The interactions established by Kaiso were analyzed in the two cellular compartments by immunoprecipitation and western blot, either in the absence or presence of Wnt3a. Association with p120-catenin was dependent on Wnt3a and was detected in both nuclear and cytosolic fractions (Fig. 7B). By contrast, the Kaiso–Tcf-4 complex was observed in the nucleus and decreased upon Wnt stimulation.

In order to determine whether the p120-catenin–Kaiso interaction took place initially in the nucleus or in the cytosol, nuclear export was inhibited using leptomycin B, an antifungal molecule widely used to block CRM1-dependent nuclear export (Kudo et al., 1998). Addition of this compound increased the levels of nuclear p120-catenin in Wnt-treated cells and prevented the nuclear export of Kaiso (Fig. 7C,D). Leptomycin did not prevent the Kaiso interaction with p120-catenin in the nucleus, suggesting that binding of these proteins occurs in the nucleus and that the protein complex is subsequently exported.

We also analyzed whether Wnt3a modulates the expression of genes sensitive to Kaiso-mediated repression. As shown previously, Kaiso associates preferentially to promoters containing methylated CpG sequences, such as that of CDKN2A (Lopes et al., 2008). First, we analyzed the extent of methylation of this promoter in different cell lines. As shown in Fig. 8A, in HT-29 M6 cells the CDKN2A promoter was not digested by the HpaII restriction enzyme, which is sensitive to methylation, whereas it was digested by the MspI isoschizomer, which is not sensitive to this modification, suggesting that this promoter is extensively methylated in this cell line. By contrast, in HEK-293T cells only 15% of the CDKN2A promoter was resistant to HpaII, whereas in SW-480 cells the percentage of digestion was 50% (Fig. 8A). Therefore, these results suggest that methylation in the CpG Kaiso-binding element of the CDKN2A promoter is complete in HT-29 M6, intermediate in SW-480 and practically nonexistent in HEK-293T cells. Incubation with Wnt did not affect the percentage of promoter methylation in any of these cells.

Wnt3a increased the amount of Kaiso binding to CDKN2 promoter (Fig. 8B). As shown by chromatin immunoprecipitation (ChIP) assays, Wnt3a-upregulated binding was observed in HT-29 M6 cells, where the promoter is methylated, but not in HEK-293T cells, where it is not. Treatment of HT-29 M6 cells with the demethylating agent azacytidine totally prevented Kaiso binding to the CDKN2A promoter (Fig. 8B). Wnt3a-enhanced Kaiso association to this promoter was accompanied by a lower expression of both CDKN2A mRNA (Fig. 8C) and its protein product p16 (Fig. 8D).

Besides its role in the control of E-cadherin stability, other functions for p120-catenin have been described in the past few years. For instance, it modulates the activity of the Rho and Rac small GTPases and controls the transcriptional activity of the Kaiso repressor (Anastasiadis, 2007). p120-catenin is found in two different cellular pools, bound or unbound to E-cadherin, which have different properties (Anastasiadis, 2007). Here, we described how the p120-catenin association with Kaiso is modulated by a Wnt3a-induced CK1ε-dependent phosphorylation. This stimulus induces the phosphorylation of Ser268 and Ser269, releasing p120-catenin from E-cadherin (Casagolda et al., 2010) and enhancing its binding to Kaiso. Although phosphorylation per se also increases the interaction with Kaiso, the upregulated p120-catenin–Kaiso association is probably a consequence of both this higher affinity and the downregulated binding to E-cadherin, given that both Kaiso and E-cadherin compete for p120-catenin association. In vitro experiments suggested that the phosphorylation of both Ser268 and Ser269 in p120-catenin are not the only modifications responsible for the increased binding to Kaiso: mutation of both serine residues to alanine residues, although affecting the association, did not totally abolish the increased Kaiso binding observed after phosphorylation by CK1. Notably, these two serine residues are located in the N-terminal regulatory domain of p120-catenin, not in the armadillo domain, the binding site for both E-cadherin and Kaiso. This result suggests that, as previously mentioned (Castaño et al., 2007; Yanigisawa et al., 2008), the N-terminal tail is spatially close to the armadillo repeats and controls the binding of factors to the armadillo domain. A similar model has been proposed for another armadillo-repeat-containing protein, β-catenin; in this case, the N- and C-tails regulate the association of several factors with the central armadillo domain (Solanas et al., 2004).

This effect of CK1 on the Kaiso–p120-catenin interaction has been detected for the two most abundant forms of this catenin: isoform 1 (1–911) and 3 (102–911). Both p120-catenins responded to Wnt3a similarly, increasing their interaction with Kaiso. However, they presented substantial differences. For instance, p120-catenin 1 showed a higher affinity for Kaiso in unstimulated cells, whereas upon Wnt addition, p120-catenin 3 bound slightly better. These results have interesting implications. For instance, because isoform 3 responds better to Wnt3a than isoform 1 (see Fig. 2B), the differential expression of p120-catenin isoforms would cause cells to become more or less sensitive to Wnt3a regarding Kaiso mobilization. Therefore, cells expressing p120-catenin 3, mostly epithelial cell lines, should respond better than more mesenchymal cells, which are enriched in p120-catenin 1. The second implication concerns the lower affinity for Kaiso detected for p120-catenin 3 compared with isoform 1. This result might mean that export from the nucleus in resting conditions is favored in cells expressing isoform 1, and might help to explain the different subcellular distributions observed for Kaiso in human tumors (van Roy and McCrea, 2005; Daniel, 2007).

Our model also indicates that Wnt3a-induced p120-catenin release from the adherens junctions would be a consequence of Ser268 and Ser269 phosphorylation, which decreases its interaction with E-cadherin. Although this modification should contribute to this release, because CK1 also phosphorylates other residues (Casagolda et al., 2010), we cannot formally discard the possibility that these other amino acids might be relevant for controlling the putative association of p120-catenin with other membrane-bound proteins. This might explain why phosphorylation of p120-catenin Ser268 can be catalyzed by a different protein kinase (i.e. not CK1), as has been detected in the membrane in some unstimulated cell lines (Xia et al., 2004).

We also discovered an interaction involving β-catenin and Kaiso. p120-catenin was able to compete with β-catenin for the association with Kaiso, suggesting that both catenins interact with the same element in Kaiso, the zinc-finger flanking sequences (Daniel and Reynolds, 1999). Similar to in p120-catenin, the first armadillo repeats in β-catenin are involved in Kaiso association. When the binding of Kaiso to both proteins was compared, p120-catenin presented a higher affinity in in vitro assays; moreover the difference in vivo might be even higher owing to the presence of additional factors in the nucleus that preferentially bind β-catenin. In any case, our results suggest a role for Kaiso as a ‘double-brake’ in β-catenin-dependent transcription, interacting and inactivating both Tcf-4 and β-catenin. Therefore, Kaiso not only prevents the formation of the β-catenin–Tcf-4 complex but also the Tcf-4 association with DNA and, possibly, interactions of β-catenin with other cofactors involved in gene activation. p120-catenin releases these two components from their interaction with Kaiso, facilitating β-catenin–Tcf4 transcriptional activity.

These results are presented in a model depicted in Fig. 9. In unstimulated cells, E-cadherin is bound to the p120-catenin–CK1ε complex and Kaiso is present in the nucleus, interacting with Tcf-4, precluding Tcf-4 binding to DNA, and with the residual β-catenin that might have escaped from APC-mediated degradation. Kaiso binding prevents the β-catenin association with Tcf-4 (Fig. 9A). Upon binding of Wnt ligands to the LRP5–LRP6–frizzled receptor complex, CK1ε is activated and p120-catenin becomes phosphorylated, downregulating its association with E-cadherin (Fig. 9B). p120-catenin released from the membrane complex translocates to the nucleus and associates with Kaiso, precluding the interaction of this factor with Tcf-4 and β-catenin, and therefore releasing the inhibition of Tcf-4 transcriptional activity. This model agrees with data from Daniel and co-workers indicating that p120-catenin nuclear location sequences are required for relieving Kaiso transcriptional repression (Kelly et al., 2004).

Binding to p120-catenin does not alter the Kaiso interaction with methylated DNA sequences, given that it poorly competes with this association. This is probably a consequence of the different Kaiso amino acid sequences involved in DNA and Tcf-4 binding; whereas, the interaction with methylated DNA is confined to zinc-finger domains (ZF) 1 and 2, the association with Tcf-4 requires ZF1–ZF3 (Ruzov et al., 2009a; Ruzov et al., 2009b). According to our model, p120-catenin is very efficient in blocking the Kaiso–Tcf-4 interaction because both p120-catenin and Tcf-4 share the same binding motif in the Kaiso protein (Daniel 2007; Ruzov et al., 2009a). However, the conformation of the complex formed with Kaiso is probably different in the two cases, and Tcf-4 effectively affects the Kaiso interaction with the DNA, whereas p120-catenin does not, and is removed from Kaiso in the presence of methylated CpG sequences. Therefore, disruption of the Tcf-4–Kaiso complex by p120-catenin, besides enabling Tcf-4 to activate its transcriptional targets, also facilitates Kaiso interaction with promoters containing CpG islands, such as CDKN2A (Fig. 9B). However, this binding requires the previous methylation of this sequence, which is cell dependent. In any case, our results indicate that, at least in some cells, Kaiso-binding genes are regulated by Wnt3a.

Besides binding to methylated CpG, Kaiso might also interact with KBS sequences, an aspect that we have not studied and is not depicted in Fig. 9. According to our results (Fig. 3), methylated CpG sequences present a higher binding affinity for Kaiso than KBS elements. This different affinity might mean that the regulation by p120-catenin is different, and, by contrast to the Kaiso interaction with methylated CpG, p120-catenin can efficiently prevent Kaiso association to KBS, relieving the repression caused by this factor on the expression of target genes, such as those encoding matrilysin or Wnt11 (Daniel et al., 2002; Kim et al., 2004).

Our results also bring new light on the role of Kaiso in Wnt signaling. As discussed above, in addition to enhancing Kaiso binding to specific promoters, Wnt3a also releases the restriction created by Kaiso on the formation of the β-catenin–Tcf-4 complex. The demonstration that Kaiso binds either to Tcf-4 or β-catenin adds new complexity to the system because the extent of β-catenin upregulation in the nucleus, determined by Wnt stabilization of the protein, might also affect the Tcf-4–Kaiso binding. Therefore, in some cells where Kaiso levels are low, the upregulation in nuclear β-catenin might be sufficient for it to bind Kaiso, disrupting the Kaiso–Tcf-4 interaction, and also for it to associate with Tcf-4, stimulating the activity of this transcriptional complex. In other cells, with higher levels of Kaiso, mobilization of p120-catenin to the nucleus would be required to eliminate the Kaiso restriction. In any case, our results show that Wnt signaling is not exclusively mediated by its role in enhancing the stability of β-catenin, and other proteins mobilized by these extracellular factors also contribute to this important pathway.

A subpopulation of SW-480 epithelial cells, established from a primary colon adenocarcinoma, was obtained from our Institute Cell Bank. These cells (SW-480-ADH) contain a truncated form of APC and express higher levels of E-cadherin than the parental population. Previous use of these cells has been reported (Pálmer et al., 2001; Casagolda et al., 2010). Alternatively HT-29 M6 intestinal tumor cells, also containing a truncated APC, or HEK-293 cells were used. Control L fibroblasts or L fibroblasts stably transfected with a plasmid encoding Wnt3a were obtained from the ATCC. The Wnt3a-expressing cell line was cultured in medium containing 0.4 mg/ml G-418. When indicated, cells were treated with conditioned medium from control or Wnt3a-expressing cells for 6 hours. Leptomycin B (LMB; Sigma) was added at 2 ng/ml to the culture medium for 6 hours when indicated.

The generation of the bacterial expression plasmid pGEX-6P encoding the GST protein fused to wild-type p120-catenin isoform 3 (amino acids 102–911) or the point mutations in Ser268 and Ser269 (to alanine or asparagine residues) has been described previously (Casagolda et al., 2010). The cDNA fragments encompassing amino acids 1–911 and 350–911, corresponding to p120-catenin isoforms 1 and 4, respectively, were obtained as described previously (Castaño et al., 2007). To generate the expression plasmid pGEX-Kaiso, the full-length Kaiso cDNA was generated from murine NIH-3T3 cells using RT-PCR and two oligonucleotides containing restriction sites for BamH1 and XhoI, and cloned into BamHI-XhoI-linearized pGEX-6P. The absence of mutations was checked by sequencing. The PCS2 Kaiso-myc plasmid was a gift from Pierre D. McCrea (MD, Anderson, Houston, TX).

The eukaryotic expression plasmid pEBG-2T encoding the GST protein fused to wild-type p120-catenin isoform 3 (amino acids 102–911), and the point mutants S268A, S269A and S268D, S269D were generated by PCR amplification of p120-catenin (102–911) inserts from pGEX plasmids, introducing 5′ KpnI and 3′ NotI restriction sites. These fragments were cloned into the KpnI-NotI-linearized pEBG-2T. The generation of the eukaryotic plasmid encoding Tcf-4–HA and the preparation of GST–Tcf-4 (1–53) have been described previously (Miravet et al., 2002).

GST fusion proteins were expressed in E. coli and purified by affinity chromatography on glutathione–Sepharose as described previously (Castaño et al., 2007). When required, GST was removed by cleaving with Pre-Scission protease (Amersham Biosciences). When indicated, GST–p120-catenin (7.5 pmol) was phosphorylated using 300 milliunits of protein kinase CK1 (Sigma) in a final volume of 50 μl kinase buffer (9 mM MgCl₂, 0.5 mM EGTA, 1 mM DTT, 0.5 mM EDTA, 2.5 mM 2-glycerophosphate, 20 mM Tris-HC1 pH 7.0 and 0.1 mM ATP). Reactions were performed for 40 minutes at 30°C. Pull-down assays were performed using purified recombinant proteins fused to GST as bait and SW-480 cell extracts. Proteins bound to the glutathione–Sepharose were analyzed by western blotting with specific monoclonal antibodies (mAbs) against p120-catenin, E-cadherin, CK1ε (all from BD Biosciences), Kaiso and Tcf-4 (from Upstate, Millipore), HA (Roche) or p120-catenin phosphorylated Ser268 (Xia et al., 2004). The polyclonal antibody against GST was from GE Healthcare. Immunoblots were analyzed using the SNAP protein detection system (Millipore). All binding assays were repeated three times.

Human shRNA for p120-catenin and CK1ε were obtained from Sigma. For stable expression of p120-catenin shRNA, SW480 cells were infected with lentiviral particles containing an shRNA targeting p120-catenin. Infected cells were selected with puromycin at 1 μg/ml. Control cells were infected with lentivirus bearing a non-targeting shRNA (clone SCH002, Sigma). Stable cell populations of human CK1ε shRNA was generated by transfecting SW480 cells with shRNA targeting CK1ε using Lipofectamine 2000 (Invitrogen) as described by the manufacturer. After incubation with the reaction mix for 5–6 hours, cells were washed twice with Opti-MEM without antibiotics and were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal calf serum (FCS) for 72 hours before selection with 2 μg/ml puromycin. Transient expression of ectopic proteins was achieved in cells at 80% confluency, transfecting the indicated eukaryotic plasmid using Lipofectamine 2000. Cells were analyzed after selection.

Cell extracts were prepared by homogenizing cells in lysis buffer (LB1) (20 mM Tris-HCl pH 7.6, 100 mM NaCl, 10 mM MgCl₂, 1 mM EDTA and 0.5% Nonidet P-40) supplemented with protease inhibitors (0.3 μM aprotinin, 1 μM leupeptin, 1 μM pepstatin, 1 mM pefabloc) and phosphatase inhibitors (10 mM NaF, 0.1 mM sodium orthovanadate and 2.5 mM 2-glycerophosphate). After passing cells ten times through a 20-gauge syringe, extracts were left on ice for 20 minutes and centrifuged at 14,000 g for 5 minutes at 4°C. Supernatants constituted the cell extracts.

To separate the nuclear fraction from the membrane and cytosol fraction, cells were lysed in 1% TX-100 buffer (25 mM Tris-HCl pH 7.6, 1% Triton X-100, 137 mM NaCl, 10% glycerol) supplemented with protease inhibitors (0.3 μM aprotinin, 1 μM leupeptin, 1 μM pepstatin, 1 mM pefabloc) and phosphatase inhibitors (10 mM NaF, 0.1 mM sodium orthovanadate and 2.5 mM 2-glycerophosphate), resuspended carefully with a micropipette and incubated on ice for 10 minutes. Cell homogenates were centrifuged at 500 g for 10 minutes at 4°C, and the supernatant was considered to be the cytosolic fraction. As shown in Fig. 8A, this fraction also include membrane proteins (such as E-cadherin) but was free of nuclear markers. Pellets were resuspended in half the volume of LB1 buffer supplemented with protease and phosphatase inhibitors, passed ten times through a 20-gauge syringe and centrifuged at 20,000 g for 10 minutes at 4°C. Supernatants constituted the nuclear fraction. Equivalent amounts of protein extracts were loaded in each condition. Antibodies against lamin β-1 (Abcam) and pyruvate kinase (Sigma) were used as a control for nuclear and cytosol fractions, respectively.

Proteins were immunoprecipitated from cell extracts (300 to 600 μg) using 2 μg/ml of the appropriate antibody or an irrelevant IgG, as a control, for 16 hours at 4°C. Tcf-4–HA was immunoprecipitated using 20 μl of anti-HA affinity matrix (Roche) for 16 hours at 4°C. Precipitated material was removed by centrifugation at 12,000 g and the resulting supernatant was incubated for 120 minutes with 20 μl of γ-bind G-Sepharose (GE-Healtcare). Immunoprecipitates were washed three times with 0.1% NP-40 and bound proteins eluted with electrophoresis sample buffer. Immunoprecipitated proteins were analyzed by western blotting using specific mAbs.

RNAs were obtained as previously reported (Solanas et al., 2008) and analyzed by quantitative RT-PCR using the QuantiTect SYBR Green RT-PCR (Qiagen) in triplicate using oligonucleotides specific for CDKN2A or HPRT.

Kaiso-DNA binding assays were performed using the biotin-labeled oligonucleotide MTS1, 5′-CAGCAGCCGCGCCCAACGCTGGGA-3′, corresponding to the S100A4 (metastasin-encoding) promoter (Prockhortchouck et al., 2001), and the corresponding complementary oligonucleotide (oligo CpG). When indicated, the assay was performed using an oligonucleotide where the three guanines labeled in bold were methylated (oligo ^MeCpG). A biotin-labeled oligonucleotide corresponding to a fragment of the Kaiso-binding sequence, 5′-GTGCTTCCTGCCAATAACG-3′, in the MMP7 promoter (Daniel et al., 2002) was also used. In Tcf-4-binding assays biotin-labeled TBE2 oligonucleotide (5′-CTCTTGATCAAAGCGCGG-3′), corresponding to Tcf-4-binding site within the Myc promoter was used. Oligonucleotides were purchased from Invitrogen. Assays were performed by incubation of 200 ng of the annealed oligonucleotides with 200 μg of cell extracts and the indicated amounts of recombinant proteins. Incubations were performed as indicated (Miravet et al., 2002). DNA-bound protein complexes were isolated by incubation with 40 μl of 50% (w/v) streptavidin–agarose (Sigma). Beads were collected, washed three times with 0.1% NP40 buffer and analyzed by western blot with specific monoclonal antibodies against Tcf-4, Kaiso or p120-catenin.

ChIP assays were performed as described previously (Escrivà et al., 2008). Briefly, 1×10⁷–1.5×10⁷ cells were crosslinked with 1% formaldehyde, lysated (50 mM Tris-HCl pH 8, 10 mM EDTA and 1% SDS for 10 minutes at room temperature) and sonicated to generate 200 to 1500 bp DNA fragments. Samples were diluted 1:10 (16.7 mM Tris-HCl pH 8, 1.2 mM EDTA, 167 mM NaCl, 1.1% Triton X-100 and 0.01% SDS), pre-cleaned with protein-G–agarose for 3 hours and immunoprecipitated twice with antibodies against Kaiso (Abcam) or rabbit IgG, as a control. DNA–protein complexes were pelleted with protein-G–agarose. Then, samples were treated with elution buffer (100 mM NaCO₃ and 1% SDS) and incubated at 65°C overnight to reverse formaldehyde crosslinking. After proteinase K and RNAse treatments, DNA was purified using the GFX PCR DNA and Gel Band Purification Kit (Amersham) and promoter regions were detected by PCR amplification. The following primers, specific for the CDKN2A promoter were used: 5′-GGCGGGGAGCAGCATGGAG-3′ (sense) and 5′-GGCCTCCGACCGTAACTATT-3′ (antisense).

The analysis was performed as described previously (Lopes et al., 2008), taking advantage of the the sequence in CDKN2 promoter sensitive to the HpaII restriction enzyme (which only cleaves it only when it is not methylated); the isoschizomere MspI, which is insensitive to methylation, was used as a control. Genomic DNA was purified from the indicated cells lines according to standard methods, divided in three parts and subjected to enzymatic digestion with MspI, HpaII or without restriction enzyme for 1 hour at 37°C. Enzymes were inactivated by heating for 20 minutes at 80°C. Each product was then purified by GFX purification columns and amplified by SYBR green qRT-PCR on a Light Cycler 480 (Roche) in triplicate using the ChIP primers indicated above. Results were represented as a percentage of the undigested DNA control.

We thank Albert B. Reynolds for generously supplying the antibody against p120-catenin phosphorylated at Ser268. We also thank Neus Ontiveros, for technical support, and Santiago Roura and Pierre McCrea for reagents. This study was supported by grants awarded by the Ministerio de Ciencia e Innovación (BFU2009-07578 to M.D. and SAF2006-00339 to A.G.H.) and La Fundació La Marató de TV3 to A.G.H. Partial support from the Instituto Carlos III (RD06/0020/0040) and the Generalitat de Catalunya (2009SGR-121 and 2009SGR-585) is also appreciated. N.D. was supported by a postdoctoral contract awarded by La Fundación Científica de la Asociación Española contra el Cáncer; D.C. and G.V. were supported by predoctoral fellowships from the Ministerio de Educación, and E.L. was supported by a predoctoral fellowship from the UAB.