The interaction of xKaiso with xTcf3: a revised model for integration of epigenetic and Wnt signalling pathways

The BTB/POZ transcriptional factor xKaiso is a bimodal DNA-binding protein that is reported to specifically bind methyl-CpGs, or a CTGCNA consensus DNA sequence (Ruzov et al., 2004; Park et al., 2005). Our previous work has established the essential and global role of xKaiso in regulating the timing of zygotic gene activation at the mid-blastula transition (MBT) (Ruzov et al.,2004). Other work proposes a model in which xKaiso specifically binds CTGCNA sequences present in the promoter region of Siamois (and also xWnt11) and interacts with the Wnt effector molecule xTcf3 to promote its stable repression (Kim et al.,2004; Park et al.,2005).

In recent work that is not in keeping with the latter model, we demonstrated that the CTGCNA motifs derived from the promoters of Siamois and xWnt11 are not sequence specific xKaiso-binding sites and these genes are not mis-expressed in xKaiso morphants (Ruzov et al., 2009). Although our loss-of-function experiments did not identify a role for xKaiso in regulating Wnt target genes, two observations suggest a potential role for it in canonical Wnt signalling: the xKaiso protein can be co-immunoprecipitated with xTcf3, and over-expression of xKaiso can suppress axis-duplication that is induced by over-expression ofβ-catenin in the ventral regions of a four-cell embryo(Park et al., 2005). We wished to determine the molecular basis of the Kaiso/Tcf3 interaction as it has profound implications for the intersection of two important regulatory pathways in amphibian development: regulation of transcriptional silencing in pre-MBT Xenopus embryos and in Wnt signalling pathways(Heasman, 2006). We find that the interaction surfaces for both proteins correspond to their previously identified DNA-binding domains. Our data squarely rules out a model for xKaiso repression through stabilisation of xTcf3 binding to DNA. Instead, our analysis suggests that the xKaiso and xTcf3/4 interaction results in their mutual delocalisation from chromatin, a prediction that we demonstrate at the cellular and the DNA levels.

The xKaiso expression constructs (XKaiso/pCS2+MT and HA-tagged)(Kim et al., 2004) were provided by Pierre McCrea. The dKaiso expression construct was from(Ruzov et al., 2004). The Myc-xTcf3, xTcf3dn and β-catenin expression vectors were provided by Randall Moon. The VP16 fusions with the ZF1-3 region of xKaiso (amino acids 447-635) and the HMG domain of xTcf3 were from Ruzov et al.(Ruzov et al., 2009). All reporter assays using SuperTOP/FOP, Tex19 and Siamois luciferase reporters were performed as described previously(Houston et al., 2002; Ruzov et al., 2009).

Embryos were manipulated as described previously(Houston et al., 2002; Ruzov et al., 2009). At the two-cell stage, the embryos were injected into the animal half with 200-750 pg of sense capped RNA (c-myc-xKaiso mRNA) synthesized in vitro (T3/T7 Cap-Scribe kit, Boehringer).

The Kaiso GST fusions were from Ruzov et al.(Ruzov et al., 2009). The TCF4 constructs were provided by Vladimir Korinek. A coupled transcription/translation kit (Promega) was used for in vitro translation/labelling with ³⁵S-Met. GST pull-down assays and immunoprecipitations were performed according to standard protocols in the presence of a 10- to 100-fold excess of recombinant full-length xKaiso or GST where indicated. Samples were visualized by phosphoimaging. EMSA was performed as described previously (Ruzov et al.,2009).

The ChIP assay was performed as described previously(Ruzov et al., 2009) using myc tagged xTcf3.

Immunostaining was performed according to standard techniques using P53^-/- (Trp53^-/-) mouse embryonic fibroblasts. Cells were analysed 24 hours after transfection. Mouse monoclonal anti-T7 tag (Novagen), anti-HA-tag (Sigma), rabbit polyclonal anti-myc(Upstate) and Alexa secondary antibodies were used.

Quantitative real-time RT-PCR of Siamois was evaluated as described previously (Houston et al.,2002).

The TCF family proteins have a multi-domain organisation with a central HMG box DNA-binding region recognizing the sequence A/TA/TCAAA; an N terminus containing the β-catenin-interacting domain adjacent to a Groucho-binding region; and a C-terminal CtBP1 interaction domain(Roose et al., 1998). We demonstrate using recombinant proteins that the DNA-binding, zinc-finger domain of xKaiso (xZF1-3) is sufficient for direct interaction with full-length xTcf3 and dominant-negative xTcf3 (xTcf3dn), which lacks theβ-catenin-interacting region (Fig. 1A,B). The ability of Kaiso to bind xTcf3 is conserved, as comparable zinc-finger regions from zebrafish Kaiso (dZF1-3) and chicken Kaiso(gZF1-3) can also interact with xTcf3 (Fig. 1B). The same pattern of interaction is seen between mouse TCF4 and xZF1-3. Deletion analysis suggests that the interaction occurs through the HMG domain of TCF4 (Fig. 1C,D). The experiment was performed in the presence of high concentrations of ethidium bromide to exclude the possibility that the interaction was mediated by non-specific binding to DNA. This suggests that the interaction between Kaiso and TCF3/4 is mutually exclusive of their binding to DNA, and that inhibition of β-catenin activation by Kaiso is not through competitive binding of a shared interaction domain on xTcf3 and TCF4. We tested the first possibility by performing an EMSA with xTcf3 and its target DNA binding site(ATCAAA) in the presence of GST-Kaiso fusions. Binding of xTcf3 to its target sequence was abolished in the presence of the full-length xKaiso, GST-xZF1-3 and GST-dZF1-3. GST alone or GST-xZF1-2 had no effect(Fig. 1E). Increasing the amount of xTcf3 overcomes the inhibitory effect of GST-xZF1-3. Moreover, the interaction between xTcf3 and β-catenin in vitro is not compromised by xZF1-3 (Fig. 2A). We also examined the cellular location of myc-tagged xTcf3 in the absence and presence of Kaiso. xTcf3 localisation in mouse cells is nuclear, with about 40% of cells with a homogenous, as opposed to a speckled, pattern (60%)(Fig. 2B); the latter is reminiscent of the discrete foci observed for endogenous mouse TCF4(Valenta et al., 2006). xKaiso in mouse fibroblasts exhibits homogenous nuclear staining(Fig. 2C). In the presence of either xKaiso or dKaiso, the pattern of xTcf3 staining was uniformly homogenous, suggesting they alter its nuclear sublocalisation(Fig. 2D).

The interaction data suggests that overexpression of xKaiso can displace xTcf3 from its genomic binding sites. Using the chromatin immunoprecipitation(ChIP) technique we could localise myc-xTcf3 to the Siamois promoter in A6 cells (Fig. 3A). However,myc-xTcf3 was delocalised in the presence of either Ha-xKaiso or T7tag-dKaiso(Fig. 3A). In contrast to the model of Park et al. (Park et al.,2005), this suggests that Kaiso can disengage xTcf3 from its target genes and alter their expression state. Under the same conditions, we could not localize dKaiso or xKaiso alone to the Siamois promoter; this is not surprising in view of the absence of high-affinity CTGCNA or methyl-CpG binding sites in it (Ruzov et al.,2009).

The SuperTopflash Wnt reporter plasmid contains the luciferase gene under the control of multiple TCF3/4-binding sites, which is activated by the addition of exogenous β-catenin and endogenous TCF3/4 in HeLa cells(Veeman et al., 2003)(Fig. 3B). This activation was inhibited between four- to sevenfold by co-expression of either xKaiso or dKaiso (Fig. 3B). The transcriptional activity of a control reporter with mutated TCF3-binding sites is unresponsive to β-catenin induction and unaffected by the presence of either Kaiso (Fig. 3B). An obvious explanation of these results is that the Kaiso interaction with the HMG domain of endogenous TCF3/4 blocks its ability to bind DNA and inhibitsβ-catenin-dependent activation of transcription(Fig. 3B). Conversely,overexpression of xTcf3 interferes with transcriptional activation of a methylated reporter template by an xKaisoZFVP16 fusion(Fig. 3C) in a dose-dependent manner suggesting the Kaiso/TCF3 interaction results in a mutual disengagement of these factors from chromatin-binding sites. This model can account for the inhibition of dorsal axis formation when Kaiso mRNA is co-injected withβ-catenin mRNA (Park et al.,2005) as excess xKaiso will displace the xTcf3/β-catenin complex and inhibit axis duplication. Similarly, xKaisoZFVP16 directly inhibits xTcf3HMGVP16 activation of a Siamois reporter in a dose-dependent manner (Fig. 3C).

Loss-of-function studies in Xenopus suggest the primary role of xTcf3 is as a repressor of organiser genes, such as Siamois, throughout the early embryo (Houston et al.,2002). This repression is inactivated on the dorsal side by a maternally encoded Wnt signalling pathway. Based on the above results, we predicted that overexpression of xKaiso in developing Xenopus embryos would mimic certain aspects of xTcf3 depletion such as ectopic Siamois expression (Houston et al.,2002). To test this, we injected xKaiso mRNA into two-cell embryos and allowed them to develop until gastrulation (stage 10.5-11)(Fig. 3D). In comparison with controls, this results in an exogastrulae phenotype that is distinct from the developmental delay phenotype observed in xKMO morphants, but is similar to the phenotype of embryos that are maternally depleted of xTcf3(Fig. 3D,E)(Houston et al., 2002; Ruzov et al., 2009). The normalised (relative to histone H4 or GAPDH levels) expression levels of Siamois are increased in the xKaiso mRNA-injected embryos(Fig. 3F). This observation suggests that high levels of xKaiso can interfere with xTcf3 function,resulting in a relative relief of xTcf3/Groucho-mediated repression of Siamois in gastrulating embryos.

These new results call for a revision of the model connecting the xKaiso repressor function and the xTCF3 repression/activation of Wnt-target genes(Kim et al., 2004; Park et al., 2005). The mode of interaction between xKaiso and xTCF3 mutually prevents binding to their cognate DNA sites, which can potentially inhibit the two pathways in which these transcription factors are major participants(Heasman, 2006). However, the recent observation that neither xWnt11 or Siamois expression is altered in xKMO morphants suggests that there is no intersection between these pathways during gastrulation (Ruzov et al.,2009).

The role of Kaiso expressed in adult somatic tissues is not clear as there is no obvious mis-expression of normally silent genes in Kaiso-null mice(Prokhortchouk et al., 2006). In cancer cells, gene expression patterns are highly disturbed and this is mirrored by alterations in the level and genomic distribution of DNA methylation, as well as histone modifications(Ohm et al., 2007). Interestingly, Kaiso levels are increased in colon cancer cells and enhance polyp formation in Min mice (Apc^Min/+)(Prokhortchouk et al., 2006). This is juxtaposed by positional differences in TCF variant expression in colon cancer (Clevers, 2006). In this case, there is a possible intersection of Kaiso with Wnt signalling pathways, where overexpression of Kaiso could attenuate constitutive Wnt signalling, while at the same time promote cancer progression through silencing of de novo methylated tumour suppressor genes. By the same token,there is a potential for TCF3/4 to displace Kaiso from its cancer target genes and promote their re-expression, which may be coincident with alterations in the tumour environment, such as the transition to metastases phenotype(Fig. 4E). There are two related Kaiso-like proteins, Zbtb4 and Zbtb38, that may also mediate an intersection between epigenetic and cellular signalling pathways in cancer via protein-protein interactions (Filion et al., 2006; Weber et al.,2008). Zbt4 has been shown to inhibit MIZ1 regulation of p21CIP1 expression possibly by displacing this transcription factor from the p21CIP1 promoter (Weber et al., 2008). This suggests that the distributions of these Kaiso-like proteins and their relative abundance within nuclear sub-compartments has the potential to determine their transcriptional output of many signalling pathways.