Mechanistic insights from structural studies of β-catenin and its binding partners

β-catenin is a multi-functional protein that plays essential roles both at adherens junctions and in Wnt signaling. The majority of β-catenin in the cell is associated with adherens junctions, where it interacts with the cytoplasmic region of cadherin (Fig. 1). Although β-catenin was long thought to link cadherin to the cytoskeleton via α-catenin, recent work indicates that β-catenin regulates the homodimerization of α-catenin, which in turn controls actin branching and bundling (Gates and Peifer, 2005; Weis and Nelson, 2006). In addition, there is a small and dynamic pool of β-catenin in the cytosol and nucleus that is responsible for the transduction of Wnt signals. Wnt proteins play critical regulatory roles in many biological processes, including embryonic development and stem cell maintenance (Moon et al., 2002), and deregulation of Wnt signaling is associated with multiple diseases, including various cancers (Bienz and Clevers, 2000; Kinzler and Vogelstein, 1996; Moon et al., 2002; Morin, 1999; Nelson and Nusse, 2004; Polakis, 2000; Reya and Clevers, 2005).

In the absence of a Wnt signal, cytosolic β-catenin is captured and phosphorylated by a cytoplasmic protein complex called the β-catenin-destruction complex, which catalyzes the phosphorylation of β-catenin and thus labels it for ubiquitin-dependent degradation by the proteasome (Kimelman and Xu, 2006) (Fig. 1). In the presence of a Wnt signal, through a mechanism that remains unclear, the phosphorylation of β-catenin is inhibited, which leads to its accumulation in the cytosol. Accumulated β-catenin migrates into the nucleus and binds to a DNA-binding protein of the Tcf/LEF family, and together they turn on the transcription of `canonical' (i.e. β-catenin dependent) Wnt-response genes. In stem cells these genes include those involved in the cell cycle, whereas during embryonic development the spectrum of Wnt response genes varies with the cell type and age of the embryo.

Abnormal activation of the Wnt/β-catenin pathway in certain stem or transit amplifying cells is caused by loss-of-function mutations in the tumor suppressors adenomatous polyposis coli protein (APC) and axin or gain-of-function mutations in β-catenin, causing overproliferation of the cells and the onset of tumorigenesis owing to a constitutive increase in the levels of nuclear β-catenin (Bienz and Clevers, 2000; Kinzler and Vogelstein, 1996; Moon et al., 2002; Morin, 1999; Nelson and Nusse, 2004; Polakis, 2000). A thorough understanding of the structural basis of the interactions between β-catenin and its numerous binding partners is thus of crucial importance both for understanding the mechanisms of cell adhesion, canonical Wnt signaling, and Wnt-independent signaling through β-catenin, as well as for the design of small molecules to manipulate β-catenin-dependent transcriptional activation, which potentially have a variety of applications. Here we discuss the interaction of β-catenin with its protein partners. Because of space limitations, we have focused on those that have experimentally determined 3D structures (for the complete set of partners, see http://www.stanford.edu/~rnusse/pathways/cell2.html).

β-catenin has a central armadillo repeat domain (residues 141-664) composed of 12 armadillo repeats, an N-terminal domain that harbors the binding site for α-catenin as well as the GSK3 and CK1 phosphorylation sites that are recognized by the β-TrCP ubiquitin ligase (Jiang and Struhl, 1998; Rimm et al., 1995; Wu et al., 2003), and a C-terminal domain (Fig. 2A) (Stadeli et al., 2006; Willert and Jones, 2006). The N- and C-terminal domains are sensitive to trypsin digestion and thus may be structurally flexible, whereas the central domain forms a relatively rigid scaffold (Huber et al., 1997). The central domain is the most conserved region of β-catenin, which is consistent with its role as the binding site for most β-catenin-binding partners. Each armadillo repeat contains approximately 42 residues, forming three helices arranged in a triangular shape. The 12 contiguous repeats form a superhelix that features a long, positively charged groove, in which the third helix of each repeat forms the floor of this groove (Fig. 2B).

There are three irregularities in the armadillo repeat domain structure. First, there is a long flexible insertion between helices 2 and 3 in armadillo repeat 10, which creates a binding site for 14-3-3ζ when β-catenin is phosphorylated in this region (Fang et al., 2007). Whether this insertion has additional roles is not known. Second, the seventh armadillo repeat contains only two helices (H2 and H3). It is uncertain whether this structural irregularity has any functional implications. Third, the first armadillo repeat is structurally more dynamic than the rest of the domain, and the first two helices of this repeat usually form a long, kinked helix.

The N- and C-terminal domains of β-catenin were proposed to interact with the armadillo repeat domain by a fold-back mechanism, which could regulate the partner-binding properties of the armadillo repeat domain (Castano et al., 2002; Cox et al., 1999; Piedra et al., 2001; Solanas et al., 2004). Quantitative ITC analysis using purified proteins does not support this model, at least for proteins that bind tightly to the armadillo repeat domain (Choi et al., 2006). In the electron density map for the crystal structure of a full-length β-catenin, the N- and C-terminal domain of β-catenin are not visible, which indicates that the N- and C-terminal domains do not interact with the armadillo repeat domain in a static conformation (Y. Xing and W.X., unpublished). Since the N- and C-terminal domains are negatively charged, and the groove of the armadillo repeat domain is highly positively charged, it is plausible that the N- and C-termini interact with the armadillo repeat domain in a highly dynamic and non-specific manner.

The 3D structures of β-catenin peptides containing the N-terminal GSK3 phosphorylation sites (S33 and S37) have also been determined (Megy et al., 2005a; Megy et al., 2005b; Wu et al., 2003) since this is the recognition site for the β-TrCP ubiquitin ligase (Hart et al., 1999; Kitagawa et al., 1999; Latres et al., 1999; Liu et al., 1999; Winston et al., 1999). These structures showed that the region covering β-catenin residues 20-31 has a tendency to form a helix, which facilitates the recognition of the DpS₃₃GθXpS₃₇ motif (θ denotes a hydrophobic amino acid and pS are the serines phosphorylated by GSK3) by β-TrCP (Megy et al., 2005a; Megy et al., 2005b; Wu et al., 2003).

The crystal structure of the β-catenin–Tcf3 complex revealed that the 51-residue β-catenin-binding domain of Tcf comprises an elongated region that runs along much of the groove formed by the armadillo repeat domain (Graham et al., 2000). Structure-based mutagenesis of β-catenin suggested that cadherins share at least part of this binding site (Graham et al., 2000). The crystal structure of a β-catenin–E-cadherin complex confirmed this and showed that the E-cadherin cytoplasmic domain has more extensive interactions with the groove than does Tcf3 (Huber and Weis, 2001).

Crystal structures of several other proteins in complexes with the β-catenin armadillo repeat domain have been determined, including axin, APC, α-catenin, BCL9 and ICAT (inhibitor of β-catenin and Tcf4) (Daniels and Weis, 2002; Eklof Spink et al., 2001; Graham et al., 2002; Graham et al., 2001; Ha et al., 2004; Pokutta and Weis, 2000; Poy et al., 2001; Sampietro et al., 2006; Xing et al., 2003; Xing et al., 2004). As in the case of Tcf and cadherin, these structures revealed that numerous β-catenin-binding partners have overlapping binding sites in the groove of the armadillo repeat domain (Fig. 2A). Biochemical analysis confirmed that many of these partners (e.g. Tcf, cadherin, ICAT and APC) cannot bind to β-catenin simultaneously (Choi et al., 2006; Hulsken et al., 1994; von Kries et al., 2000). Spatial segregation of different β-catenin-binding partners within the cell may therefore be important for the proper function of these proteins. In addition, competition between them could be important for regulating the Wnt signaling pathway.

The groove within the central domain can be divided into several sections on the basis of the way β-catenin-binding partners bind. Armadillo repeats 5-9 form the core binding site for Tcf and cadherin and an essential part of the binding sites for APC and ICAT (Fig. 2A; see below). All of these proteins interact with this region of β-catenin through conserved DxθθxΦx_2-7E binding motifs (θ and Φ are hydrophobic and aromatic residues, respectively), displaying conformations almost identical to that first observed in the β-catenin–Tcf3 crystal structure (Daniels and Weis, 2002; Eklof Spink et al., 2001; Graham et al., 2002; Graham et al., 2001; Graham et al., 2000; Ha et al., 2004; Huber and Weis, 2001; Poy et al., 2001; Xing et al., 2004). Two pairs of salt bridges exist in all of these complexes, formed between Lys435 and Lys312 of β-catenin and the Asp and Glu residues in the DxθθxΦx_2-7E motif of the β-catenin-binding partners, respectively. These two salt bridges, especially that involving Lys435, are crucial for β-catenin–partner interactions in this region – and were thus dubbed `charged buttons' since they fasten the partners to β-catenin (Graham et al., 2000) (Fig. 2B). Since the β-catenin-binding partners have no apparent evolutionary relationship, it was surprising to find they have adopted the same sequence motif to recognize these two specific lysine residues among the many lysine residues in this groove.

Another critical binding region for cadherin, Tcf, APC and axin is the groove in armadillo repeats 3-4 (Fig. 2A). Tcf and axin both bind to this region by using a single α-helix. The Tcf helix runs antiparallel and the axin helix runs parallel to the axis of the armadillo repeat domain superhelix (Graham et al., 2001; Graham et al., 2000; Poy et al., 2001; Xing et al., 2003). Both APC and cadherin contain a conserved SxxxSLSSL motif that interacts with this region, using a very similar conformation, in a phosphorylation-dependent manner (Fig. 2A, and see below) (Ha et al., 2004; Huber and Weis, 2001; Xing et al., 2004).

The majority of β-catenin molecules in the cell are associated with cadherins. The cytoplasmic domain of cadherin interacts with the entire groove of the β-catenin armadillo repeat domain (Fig. 3) (Huber and Weis, 2001). As discussed above, the `core' β-catenin-binding region of E-cadherin interacts with armadillo repeats 5-9, which include K435 and K312. E-cadherin contains an α-helix N-terminal to this region that docks in the groove formed by armadillo repeats 11-12. β-catenin residue Y654 lies within this region (Fig. 3); its phosphorylation by Src reduces the affinity of E-cadherin for β-catenin by approximately 85%, apparently by disrupting the docking of the E-cadherin helix (Huber and Weis, 2001; Roura et al., 1999).

C-terminal to the E-cadherin core region are E-cadherin regions IV and V. Region IV of the E-cadherin cytoplasmic domain contains several Ser/Thr phosphorylation sites. This region is disordered when unphosphorylated but interacts extensively with armadillo repeats 3-4 when phosphorylated (Fig. 2A) (Huber and Weis, 2001). This enhances binding of E-cadherin to β-catenin by several hundred-fold and may thus regulate the β-catenin–E-cadherin interaction in vivo (Choi et al., 2006). Region V at the very C-terminus of E-cadherin forms a two-helix motif that dynamically binds to the hydrophobic N-terminal tip of the β-catenin armadillo repeat (Fig. 3); this appears to play an auxiliary role in the β-catenin–E-cadherin interaction (Huber and Weis, 2001; Sampietro et al., 2006).

N-terminal to the first armadillo repeat of β-catenin is an extended helix (residues 120-147) that forms the binding site for α-catenin (Fig. 3). The β-catenin–α-catenin interface was first revealed by the crystal structure of a β-catenin–α-catenin fusion protein (βα-cat) (Pokutta and Weis, 2000). The binding of α-catenin to β-catenin disrupts the continuity of the first armadillo repeat around Y142 to D144, creating a hinged region (Fig. 3). This hinged region accommodates both E-cadherin (region V) and α-catenin, which allows β-catenin to bind to both α-catenin and E-cadherin simultaneously (Fig. 3) (Huber and Weis, 2001; Pokutta and Weis, 2000). Importantly, the β-catenin-binding site on α-catenin overlaps with the α-catenin homodimerization interface. α-catenin can thus be either in a homodimeric form, which is required for its interaction with actin, or bound to E-cadherin through β-catenin, which prevents the binding of α-catenin to the actin cytoskeleton (Drees et al., 2005; Yamada et al., 2005). Thus, α-catenin in the cell cannot simultaneously be bound to both actin and β-catenin (Fig. 1). Instead of linking actin to E-cadherin, α-catenin has been proposed to function as a key regulator of actin dynamics in the vicinity of E-cadherin (Drees et al., 2005; Yamada et al., 2005).

Phosphorylation of β-catenin in the β-catenin-destruction complex is the central regulatory step of canonical Wnt signaling. In this complex, CK1α phosphorylates S45, which primes the sequential phosphorylation of T41, S37 and S33 by GSK-3β. It is generally accepted that, once S33 and S37 are phosphorylated, β-catenin's fate is sealed: the phosphorylated β-catenin is recruited by the β-TrCP-containing ubiquitin ligase, which adds ubiquitins to β-catenin, causing it to be degraded by the proteasome (Fig. 1). β-catenin interacts with two proteins in the complex: APC and the scaffolding protein axin. The β-catenin-binding site in axin lies just C-terminal to the GSK-3β-binding site, apparently positioning GSK-3β close to the phosphorylation target sites in the β-catenin N-terminus. The core β-catenin-binding domain of axin (only 17 residues long) forms a helix that docks in the groove formed by armadillo repeats 3 and 4 (Fig. 4) (Xing et al., 2003).

APC is the only protein that contains multiple apparent β-catenin-binding sites. It contains two types of β-catenin-binding motif: three 15 amino acid (15aa) repeats and seven 20 amino acid (20aa) repeats. The APC 15aa repeats bind to the groove region of repeats 5-9 (Fig. 2A), displaying a binding affinity (K_d) of 0.1-1 μM (Eklof Spink et al., 2001; Liu et al., 2006), and are not regulated by phosphorylation. By contrast, the 20aa repeats each comprise a highly conserved 20-residue sequence that has potential phosphorylation sites in a consensus SXXSSLSXLS motif. Phosphorylation of these by CK1 and GSK-3β dramatically enhances the ability of APC to bind β-catenin, which suggests this plays a crucial role in Wnt signaling (Ha et al., 2004; Rubinfeld et al., 2001; Xing et al., 2004).

In vitro isothermal calorimetric (ITC) analysis showed that, despite their significant sequence similarity, the different β-catenin-binding repeats of APC have dramatically different affinities for β-catenin and thus may play different biological roles. The phosphorylated third 20aa repeat has by far the tightest binding affinity for β-catenin of all the repeats (Liu et al., 2006). Interestingly, most APC mutations associated with colon cancers have lost this repeat (Bienz and Clevers, 2000; Nathke, 2004; Polakis, 1995). The crystal structure of its phosphorylated form in complex with the β-catenin armadillo repeat domain revealed that a single APC 20aa repeat, together with its flanking residues, is packed along almost the entire groove of the armadillo repeat domain (Fig. 4) (Ha et al., 2004; Xing et al., 2004). The four phosphorylated residues help form a hairpin structure that binds in the groove region of armadillo repeats 1-5. Surprisingly, the residues N-terminal to the third 20aa repeat bind along most of the groove region of repeats 5-12 in a conformation almost identical to that of Tcf and E-cadherin, with APC residues D1486 and E1494 forming salt bridges with the charged buttons K435 and K312 on β-catenin. This significantly increases the affinity of APC for β-catenin (Liu et al., 2006).

Structural and biochemical studies demonstrated that the phosphorylated APC 20aa repeats, but not the 15aa repeats or unphosphorylated 20aa repeats, strongly compete with axin for binding to β-catenin (Xing et al., 2003). We have therefore suggested that β-catenin initially binds to both axin and the 15aa repeats, which positions it for phosphorylation by CK1α and GSK-3β. Then, the 20aa repeat 3 is phosphorylated, which causes it to displace axin from β-catenin, allowing β-catenin to leave the destruction complex. In this way APC functions as a `rejuvenator' of the β-catenin destruction complex by releasing phosphorylated β-catenin from axin and allowing the recruitment of the next β-catenin substrate (Kimelman and Xu, 2006; Xing et al., 2004).

In an alternative model (Ha et al., 2004), the phosphorylation state of the APC controls the rate of release of β-catenin from the destruction complex. In this model, the binding of axin and its associated kinases and phosphatases to APC results in heterogeneous APC phosphorylation. In the absence of a Wnt signal, the cytosolic β-catenin concentration is very low such that β-catenin will only interact with phosphorylated APC (pAPC), which binds β-catenin with high affinity compared with the unphosphorylated APC. The β-catenin–pAPC dimer is then recruited into the destruction complex owing to the ability of axin to bind APC. Because the phosphorylated APC binds β-catenin tightly, it releases β-catenin only slowly from the destruction complex, thus sequestering the β-catenin. In the presence of a Wnt signal, the cellular β-catenin concentration will increase and APC phosphorylation should decrease (a prediction of the model not yet proven); so binding of β-catenin to unphosphorylated APC becomes significant. When the β-catenin–APC dimer joins the destruction complex by binding of axin directly to β-catenin as well as its binding to APC, β-catenin is phosphorylated and then released quickly owing to the relatively weak interaction between unphosphorylated APC and β-catenin. The increased turnover of the destruction complex in the presence of a Wnt signal is proposed to be important for shutting off the Wnt signal after a certain time (Ha et al., 2004).

β-catenin is tightly regulated by a plethora of post-translational modifications, including Ser/Thr and Tyr phosphorylation at several sites, ubiquitylation and acetylation (Table 1 lists all known modifications of β-catenin, only some of which are discussed here). These modifications all occur in the flexible N- and C-terminal domains or on the surface of the armadillo repeat domain. Therefore, none is likely to cause a large scale conformational change in the armadillo repeat domain, although local structural changes in this domain are produced in some cases (Table 1). For example, phosphorylation of β-catenin at S246 by Cdk5 (Munoz et al., 2007) causes the binding of the prolyl isomerase Pin1, which prevents APC from binding to β-catenin, thereby preventing β-catenin degradation and causing the activation of Wnt target genes (Ryo et al., 2001). Intriguingly, elevated Pin1 levels are observed in breast and colon cancer tissues, and the levels of expression correlate with tumor grade and cyclin D1 expression (Kuramochi et al., 2006; Pang et al., 2004; Wulf et al., 2001). Pin1 bound to β-catenin is proposed to isomerize β-catenin residue P247 (Ryo et al., 2001), which lies right in the middle of the region that binds to phosphorylated APC 20aa repeats (Ha et al., 2004; Xing et al., 2004). Thus, a local structural change in this region could have a major effect on the affinity of APC for β-catenin.

An important question in the field is whether the two essential functions of β-catenin, cell adhesion and transcriptional activation, are related. In C. elegans, different β-catenin-like proteins perform these roles (Kidd, 3rd et al., 2005; Korswagen et al., 2000; Natarajan et al., 2001), whereas in most species one β-catenin carries out both functions. It is not clear whether there is any evolutionary advantage to this, and the relationship between the β-catenin at the membrane and β-catenin in the nucleus is still a contentious issue. In principle, any change in the cell that releases enough β-catenin from the membrane that the destruction complex cannot process it all should increase the cytosolic and nuclear levels of β-catenin. Whether this actually happens is unclear. Certain receptor tyrosine kinases (e.g. EGFR and Met) can synergize with Wnt/β-catenin signaling, possibly by directly phosphorylating tyrosine residues in β-catenin (Brembeck et al., 2004; Bustos et al., 2006; Coluccia et al., 2007; Roura et al., 1999; Zeng et al., 2006), and thus skewing its role towards transcriptional activation. β-catenin Y142 (Fig. 3) – a potential regulatory site – was initially proposed to be a site whose phosphorylation would disrupt α-catenin binding and promote the binding of the transcription factor BCL9-2 (and presumably the related factor BCL9) (Brembeck et al., 2004), which could thus play a major role switching β-catenin between its roles in adhesion and transcription. However, structural, thermodynamic and functional studies demonstrate that phosphorylation of Y142 does not promote BCL9-2 binding, although it does abolish binding of α-catenin to β-catenin (Hoffmans and Basler, 2007; Sampietro et al., 2006). Although there is no conclusive evidence that decreases in β-catenin at the membrane lead to increased stabilization of nuclear β-catenin, the possibility that this occurs cannot be ruled out.

In the nucleus, β-catenin forms multiple protein complexes with either Wnt pathway inhibitors, such as ICAT and Chibby, or transcriptional co-activators, including BCL9/BCL9-2, CBP/p300, pygopus, Brg-1, Pontin-52, MED12 and parafibromin/Hyrax (Fig. 1) (Barker et al., 2001; Bauer et al., 2000; Brembeck et al., 2004; Hecht et al., 2000; Kim et al., 2006; Kramps et al., 2002; Miyagishi et al., 2000; Mosimann et al., 2006; Nusse, 1999; Roose and Clevers, 1999; Sun et al., 2000; Takemaru and Moon, 2000; Thompson et al., 2002) and members of the Tcf/LEF family of DNA-binding proteins. It remains unclear whether different β-catenin-associated co-activators bind to β-catenin simultaneously. Some Wnt pathway inhibitors, such as ICAT, inhibit Wnt signaling by interfering with the formation of appropriate β-catenin-containing transactivation complexes (Tago et al., 2000; Takemaru et al., 2003). However, it is largely unknown how the interactions between β-catenin and inhibitory proteins are regulated in the cell.

In the presence of a Wnt signal, β-catenin is recruited to the promoters of Wnt-responsive genes through its interaction with members of the Tcf/LEF family. β-catenin–Tcf complexes are essential for transcription of a plethora of genes (Clevers, 2006). The N-terminal ∼50 residues of Tcf/LEF are required for its interaction with β-catenin. Residues 16-32 of human Tcf4 contain the DxθθxΦx_2-7E sequence that interacts with armadillo repeats 5-9 (Fig. 5). Residues 40-50 form an α-helix that docks in the groove of armadillo repeats 3-4 and engage in mostly hydrophobic interactions with β-catenin. At the β-catenin–Tcf4 interface, the salt bridge formed between β-catenin K435 (one of the charged buttons) and Tcf4 D16 is by far the most crucial interaction (Fasolini et al., 2003); the other salt bridge is formed between the second charged button, K312, and one of several acidic residues in the Tcf/LEF E₂₄GEQEE₂₉ sequence that adopts either extended or helical conformations (Graham et al., 2001; Poy et al., 2001). In addition to the N-terminal domain, Tcf/LEF may have a secondary β-catenin-binding site, which includes sequences just N-terminal to the DNA-binding HMG domain (this is suggested by the proteolysis pattern of Tcf/LEF protected by β-catenin). This secondary binding site overlaps that of the transcriptional co-repressor Groucho/TLE and is critical for displacing Groucho/TLE from Tcf/LEF in the presence of a Wnt signal (Daniels and Weis, 2005).

β-catenin promotes the transcription of Wnt responsive genes by recruiting several transcriptional co-activators, including BCL9. The crystal structure of a BCL9–β-catenin–Tcf4 complex revealed that the BCL9 β-catenin-binding domain forms an α-helix that binds to the first armadillo repeat of β-catenin so that it `caps' the N-terminus of the armadillo repeat domain (Fig. 5) (Sampietro et al., 2006). It therefore differs from most other known β-catenin partners, except region V of E-cadherin and α-catenin, which also bind in this region (Fig. 3). The β-catenin–BCL9 interface can be mutated to prevent BCL9 from binding to β-catenin without affecting cadherin or α-catenin binding, which demonstrates that BCL9 binds to β-catenin through unique, essential contacts (Sampietro et al., 2006). Note that despite a lack of deep pockets at this interface, the BCL9-binding site in β-catenin is structurally dynamic, which would allow more compounds to bind this region with high affinity and interfere with BCL9 binding since compounds can bind to a dynamic surface with an induced-fit mechanism. Therefore the BCL9–β-catenin interface is a potentially good target for drugs to block β-catenin signaling. Some other β-catenin-binding transcriptional activators, such as CBP and hyrax/parafibromin, interact with the C-terminal domain of β-catenin (Brembeck et al., 2004; Hecht et al., 2000; Kramps et al., 2002; Miyagishi et al., 2000; Mosimann et al., 2006; Sun et al., 2000; Takemaru and Moon, 2000). However, the structural basis of these interactions remains to be revealed.

In the past several years, structural and biochemical studies have laid a solid foundation for understanding some basic issues in β-catenin-regulated cell adhesion and Wnt signaling. In the coming years, structural and biochemical studies should be carried out with larger β-catenin-containing complexes. For example, a through understanding of how Wnt signaling inhibits β-catenin phosphorylation in the destruction complex will require biochemical and biophysical studies of the complete complex in different functional states.

In addition, structural studies will continue to allow the design of compounds to manipulate β-catenin-dependent signaling. The β-catenin–Tcf complex has been considered a premium drug target since it is the essential final effector of canonical Wnt signaling and since it is downstream of most cancer-associated Wnt pathway mutations (e.g. APC mutations). However, a compound that disrupts the β-catenin–Tcf interaction could also interfere with the β-catenin–cadherin and β-catenin–APC interactions, since Tcf, cadherin and APC share the core binding site in the groove formed by the armadillo repeat domain (Figs 3, 4, 5). Other protein-protein interfaces in the β-catenin-containing transcriptional complex, such as those involving β-catenin and BCL9, could therefore also be attractive targets. It will be valuable to determine the structure of β-catenin in complex with other essential transcriptional co-activators, including hyrax/parafibromin and CBP/p300, to identify other regions that might serve as useful drug targets.

This work was supported by NIH grants CA90351 to W.X. and HD27262 to D.K.