β-Catenin N- and C-terminal Tails Modulate the Coordinated Binding of Adherens Junction Proteins to β-Catenin*

β-Catenin plays a central role in the establishment and regulation of adherens junctions because it interacts with E-cadherin and, through α-catenin, with the actin cytoskeleton. β-Catenin is composed of three domains: a central armadillo repeat domain and two N- and C-terminal tails. The C-tail interacts with the armadillo domain and limits its ability to bind E-cadherin and other cofactors. The two β-catenin tails are mutually inter-regulated because the C-tail is also necessary for binding of the N-tail to the armadillo domain. Moreover, the N-tail restricts the interaction of the C-tail with the central domain. Depletion of either of the two tails has consequences for the binding of factors at the other end: deletion of the C-tail increases α-catenin binding, whereas deletion of the N-tail blocks E-cadherin interaction to the armadillo repeats. As an effect of the interconnection of the tails, the association of α-catenin and E-cadherin to β-catenin is interdependent. Thus, binding of α-catenin to the N-tail, through conformational changes that affect the C-tail, facilitates the association of E-cadherin. These results indicate that different cofactors of β-catenin bind coordinately to this protein and indicate how the two terminal ends of β-catenin exquisitely modulate intermolecular binding within junctional complexes.

␤-Catenin is a multifunctional protein that exerts two essential functions in epithelial cells. It is necessary for the maintenance of adherens junctions because it binds E-cadherin and connects it, through ␣-catenin, to the actin cytoskeleton (1,2). On the other hand, when not bound to E-cadherin, ␤-catenin can move to the nucleus and act as a transcriptional co-activator, through the interaction with members of the T cell transcription factor family (3). In addition to T cell transcription factor proteins, several other transcriptional factors have been reported to interact with ␤-catenin, presumably modulating its positive activity on the transcription of several target genes (4). On the other hand, a complex of proteins containing the product of the adenomatous polyposis coli, axin, glycogen synthase kinase 3␤, and ␤-TrCP/slimb is involved in the targeting of cytosolic ␤-catenin to the proteasome and subsequent degradation, therefore precluding its transport to the nucleus and its transcriptional activity (4).
It is widely accepted that tyrosine modification of ␤-catenin regulates formation or disassembly of adherens junctions. Data from different groups have indicated a relationship between adherens junction disruption and tyrosine phosphorylation of ␤-catenin (reviewed in Ref. 5), although some discrepant data have also been reported (5,6). By using recombinant proteins, our group has demonstrated that phosphorylation of a specific ␤-catenin tyrosine residue, Tyr-654, decreases its interaction with E-cadherin (7). Experiments in which transfections of ␤-catenin Tyr-654 mutants in different cell lines were performed further support the relevance of the phosphorylation of this tyrosine residue (7,8). Additional evidence has been provided by the elucidation of the crystal structure of the complex formed by the ␤-catenin-E-cadherin binding domains; ␤-catenin Tyr-654 is one of the key residues involved in the stabilization of this complex through the interaction with E-cadherin Asp-665 (9).
␤-Catenin can be divided in three different sub-domains. The central domain is composed of 12 repetitions of 42 amino acids each, which have been named armadillo repeats, after the ␤-catenin ortholog in Drosophila, armadillo. This region, with a basic pI, has been crystallized; it forms a very rigid structure, a super helix composed by 36 small ␣-helices (3 per each armadillo repeat) (10). Tyr-654 lies on the last repeat of this domain. However, the structure of the entire protein with the two ␤-catenin terminal domains, the N-and the C-tails, is still unknown. Contrary to the armadillo repeat domain, these terminal tails are mainly acidic (in both the pI is 4.5), and their degree of conservation among different species is much lower than the central core. Recently, we have described (11) that ␤-catenin C-tail interacts with the armadillo domain and that this interaction is controlled by Tyr-654 phosphorylation. This association limits the ability of ␤-catenin to bind E-cadherin and other cofactors involved in its transcriptional activity, such as the TATA-binding protein (TBP) 1 (11). We describe here that * This work was supported in part by Grants 01/045-00 from Fundació La Caixa (to A. G. H.), PM-99-0132 and PM99-0064 from Ministerio de Ciencia y Tecnología (to A. G. H. and M. D., respectively), 2FD97-1491-C02-01 and 2FD97-1491-C02-02 from FEDER-Fondo Nacional IϩD Funds (to A. G. H. and M. D., respectively), and 2001SGR00410 and 2001SGR00197 from Direcció General de Recerca. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  the N terminus also interacts with the armadillo domain. This interaction is potentiated by the C-tail. The functional relevance of this interconnection between the terminal tails is evidenced by the fact that ␣-catenin, which binds to the N-tail of ␤-catenin, also affects the association of E-cadherin to the armadillo repeats.

EXPERIMENTAL PROCEDURES
Reagents-The following ␤-catenin monoclonal antibodies were used in this study: ␤-catenin C terminus (Transduction Laboratories, Lexington, KY); the epitope recognized by this antibody has been mapped to residues 696 -750 using different deletion mutants of this protein (not shown); ␤-catenin C-end (Calbiochem), epitope recognized, 769 -781 (according the manufacturer); ␤-catenin armadillo core, epitope recognized, 422-685 (according the manufacturer) and ␤-catenin N terminus, raised against the first 100 amino acids (both from Alexis Biochemicals, San Diego, CA).
Expression of Recombinant Proteins-The preparation of all the plasmids codifying the different ␤-catenin forms (deletion and point mutants) as glutathione S-transferase (GST) fusion proteins has been described previously (7,11) except for the following cases: GST-arm-(120 -683) and GST-␤-cat-(120 -750); DNA fragments corresponding to amino acids 120 -683 or 120 -750 of murine ␤-catenin were amplified by PCR using oligonucleotides corresponding to the nucleotide sequences 358 -372 and 2035-2047 or 358 -372 and 2260 -2242 and containing BamHI and XhoI sites at their ends. The 1.7-or 1.9-kbp amplification fragments were digested with BamHI and XhoI and cloned in the same sites of pGEX-6P3 plasmid and expressed in Escherichia coli; GST-⌬N␤Ϫcatenin was prepared cutting pGEX-␤-cat with SacI/NotI, purifying the 0.66-kbp fragment and inserting it in pGEX-arm-(120 -683) cut with the same enzymes; GST-⌬C␤Ϫcatenin was prepared cutting pGEX-␤-cat with BamHI/SacI, purifying the 1.75-kbp fragment and inserting it in pGEX-arm-(120 -683) cut with the same enzymes; GST-⌬C␤Ϫcatenin (Y86E) was prepared identically to GST-⌬C␤Ϫcatenin using the BamHI/SacI fragment obtained from pGEX-␤-catenin(Y86E); GST-C-tail-(696 -750), a DNA fragment corresponding to amino acids 696 -750 was amplified from entire ␤-catenin cDNA by PCR using oligonucleotides corresponding to the nucleotide sequences 2086 -2103 and 2260 -2242 and containing BamHI and XhoI sites at their ends. The 0.16-kbp amplification fragment was digested with BamHI and XhoI, cloned in the same sites of pGEX-6P3 plasmid, and expressed in E. coli. GST fusion proteins were prepared and purified as described (7). When indicated, GST was removed cleaving with PreScission protease (Amersham Biosciences). The ␤-catenin deletion mutants used in this study are presented in Fig. 1, indicating which part of the molecule they comprise.
␤-Catenin Binding Assays-Pull-down assays using purified proteins were performed as described (7,11). A peptide comprising the last 21 residues of ␤-catenin was synthesized and attached to Sepharose-4B matrix by Lipotech SA (Barcelona, Spain). Binding assays were performed as described (7,11), and protein complexes bound to the Sepharose were analyzed by Western blot using specific mAbs against ␤-catenin (see above), ␣-catenin, E-cadherin (cytosolic domain), TBP (the three from Transduction Laboratories) or GST (Amersham Biosciences) as control. Overlay assays were performed as described (12). Briefly, recombinant proteins were separated by SDS-PAGE and electroblotted onto nitrocellulose membranes. Free membrane-binding sites were blocked in TTBS buffer (25 mM Tris-HCl, pH 7.5, 135 mM NaCl, 0,2% Triton X-100) plus 0.1% bovine serum albumin. Membranes were incubated with 5-10 g/ml of recombinant protein in TTBS with 1% bovine serum albumin for 90 min at room temperature. After extensive washing with TTBS, the nitrocellulose membranes were incubated with the corresponding specific antibody, followed by a peroxidase-conjugated secondary antibody. Immunoblots were developed with peroxidase-conjugated secondary antibody followed by enhanced chemiluminescence detection system (ECL, Pierce). The autoradiograms were scanned, and the values obtained were either compared with known amounts of recombinant proteins included as reference in the same blot (␤-catenin binding assays) or with the value obtained for wild-type full-length ␤-catenin (pull-down assays). Duplicate samples were always introduced in the gels, and each experiment was performed at least three times. Values obtained from the scanning of the gels were compared with a control and the average percentage of binding calculated Ϯ S.D.
Protease Sensitivity of ␤-Catenin-5.5 pmol of the different forms of ␤-catenin were incubated in the presence of trypsin (30 ng, when analyzing the C terminus; 0.1 ng for the N terminus analysis) at 24°C in a final volume of 100 l, in a buffer containing 90 mM Tris-HCl, pH 8.5, 2 mM CaCl 2 , and 4 mM dithiothreitol. Alternatively, 17 pmol of ␤-catenin were incubated with 51 pmol of cytoE-cadh, ␣-catenin, or GST as a control in a final volume of 50 l in 50 mM Tris-HCl, pH 7, 150 mM NaCl, 3 mM MgCl 2 , 0.1% Triton X-100, 1 mM dithiothreitol. Complexes were digested with carboxypeptidase Y (Roche Molecular Biochemicals) (17.5 ng) at 25°C. Reactions were stopped at the indicated digestion times from 1 to 90 min with electrophoresis loading buffer and boiled for 4 min. Extent of the digestion was determined analyzing the samples by SDS-PAGE and Western blot with mAbs against ␤-catenin C terminus, N terminus, or C-end. A quantification of the reaction was performed scanning the autoradiograms and representing the amount of fulllength ␤-catenin remaining at the different times of incubation relative to the initial time. The experiments, performed in duplicate, were repeated twice, and the average ␤-catenin (Ϯ range) remaining at different times was represented. In all the cases, the results were reproduced when a higher concentration of protease was used, although all the forms presented an accelerated rate of degradation.

RESULTS
␤-Catenin protein is composed of three well differentiated domains: a central core, denominated armadillo repeat domain, and two N-and C-tails (see Introduction and Fig. 1). We have reported recently that the C-tail can interact with the armadillo domain, and this interaction restricts the ability of Ecadherin to associate with the armadillo repeats (11). Because the crystal structure of the armadillo region of ␤-catenin complexed with cytosolic domain of E-cadherin has recently been elucidated (9) and the amino acids involved in this interaction identified, we looked for sequence analogies between this Ecadherin region and the ␤-catenin C-tail. As described in Ref. the last part of ␤-catenin C-tail (amino acids 760-end) (Table I). Therefore, it is possible that the interaction of the C-tail with the armadillo domain is mainly mediated through the last amino acids of the molecule, 760 -781. To check this possibility, binding of different C-tail fragments to the GST-arm was checked. As shown in Fig Analysis of this association using overlay assays rendered the same results. Moreover, as evidenced in Fig. 2D, binding of the C-tail to the armadillo domain was more effective than the binding to a fusion protein containing the N-tail as well (⌬Ctail). This result suggested that the N-tail was restricting binding of the C-tail to the armadillo domain and prompted us to examine the interconnection between both tails.
Although binding of the N-tail is weaker than that of the C-tail, association of the N-terminal tail to the complete armadillo domain was also detected (Fig. 3A, upper panel). We could not observe a direct interaction between the two tails in the absence of the arm domain (Fig. 3A, last lane). However, the association of the N-tail to a recombinant protein consisting of both the armadillo and the C-terminal domains (⌬N-tail) was greater than to just the armadillo domain, which suggests that, although the C-tail is not sufficient for binding to the N-tail, it enhances the interaction of the N terminus to the armadillo domain. The last 31 amino acids of ␤-catenin do not seem to be involved in the interaction with the N-tail because this tail interacts similarly with ⌬N-tail-␤-catenin proteins containing the entire C-tail and with those missing the last 31 amino acids (⌬N-tail-(120 -750)) (Fig. 3A).
We have reported previously (7) that tyrosines 86, located in the N-tail, and 654, in the last armadillo repeat, can be phosphorylated in vitro by pp60 c-src tyrosine kinase. In order to individually analyze the relevance of these phosphorylation in the N-tail-armadillo domain interaction, we have generated different fusion proteins with Tyr-86 and Tyr-654 replaced by Glu (Y86E and Y654E, respectively). The mutations are considered to mimic the effect of the phosphorylation. As shown in Fig. 3A, Y654E mutation in ⌬N-tail-␤-catenin did not modify N-tail binding. This mutation, located in the last armadillo repeat, has been shown to alter the association of last amino acids of the C-tail-(760 -781) with the armadillo domain (see Ref. 9 and data not shown). Therefore, this result reinforces our conclusion that the end of the C-tail does not play an important role in the potentiation of N-tail binding to the central core. On the other hand, N-tail (Y86E) interacted better than the wild-type N-tail, both with the armadillo domain and with ⌬N-tail ␤-catenin (Fig. 3B). This result indicates that an increase in the negative charge of the N-tail potentiates the interaction to the arm domain and suggests that establishment of ionic pairs are probably relevant in the interaction between this two domains.
Our binding assays indicate that the association of the N-tail to the armadillo domain is enhanced by the C-tail and suggest that the conformation adopted by both tails in the native protein is interdependent. To verify this hypothesis, experiments of limited trypsin proteolysis were performed, and the disappearance of epitopes located at the N terminus or at the C-tail was followed with specific mAbs. As shown in Fig. 4, left panels, the loss of the N-terminal epitope occurred much faster than the loss of another epitope mapped between amino acids 696 and 750 of the C-tail (Fig. 4, right panels), suggesting that the C-tail is much less accessible to the protease. This is more evident if we take into consideration the different concentrations of trypsin used in the two assays (300 times greater for the C terminus). Depletion of the C-tail (⌬CϪtail mutant) increased the sensitivity of the N terminus to trypsin (Fig. 4, left bottom panel), indicating that the N-tail is not as closely bound to the armadillo domain as in the wild-type form, and confirming our hypothesis that binding of the N-tail requires the presence of the C-tail. The presence of Tyr-654 3 Glu mutation retarded the degradation of the N terminus (Fig. 4, left middle panel), suggesting that, although the binding of the N-tail to armadillo is not modified (see Fig. 3A), the conformation of this N-tail is altered by a modification in the C-terminal part of ␤-catenin. Tyr-86 3 Glu mutation, which increases binding of the N-tail to the armadillo repeats, also protected this epitope (Fig. 4, left middle panel).
Similar analysis performed with an anti-C-tail antibody gave complementary information. Analysis of trypsin digests of wild-type ␤-catenin with this antibody rendered more than one band (Fig. 4, right panels). These lower bands did not appear when the ⌬NϪtail mutant was digested, suggesting that they were produced by the attack of trypsin on the N-terminal region. The wild-type ␤-catenin and the ⌬N-tail mutant were digested with similar rates; a slight but reproducible retard was observed in the mutant, suggesting that absence of the N-tail is favoring binding of the C-tail to the armadillo domain. As expected, Tyr-654 3 Glu mutation accelerates degradation of the C-tail, because it hinders interaction of the C-tail with the armadillo domain. On the other hand, ␤-catenin Tyr-86 3 Glu mutant was clearly more sensitive to trypsin than the wild-type form (Fig. 4, right middle panel), further supporting our conclusion that modifications in the N-tail domain affect the C-tail conformation.
These results indicate that the two tails are interconnected and differently regulate their binding to the armadillo domain; although the C-tail is necessary for the association of the N-tail TABLE I Analogy between the armadillo-binding sequences of E-cadherin and the ␤-catenin C-tail The sequence of region II of the cytosolic domain of E-cadherin is shown. The amino acids of this sequence that interact with relevant amino acids of the armadillo domain of ␤-catenin (indicated below) are indicated in bold, according to Huber and Weis (9). Equivalent amino acids in the C-tail of ␤-catenin to those present in E-cadherin are shown in boldface type in the upper row. Other amino acids pairs showing similar properties in these two regions are underlined.
to the armadillo repeats, N-tail hinders C-tail-armadillo interaction. In this regard it is particularly illustrative of the fact that a mutation that increases binding of the N-tail to the armadillo and, therefore, retards the degradation of the N terminus accelerates the proteolysis of the C-end. This inter-dependent binding of both tails suggested to us that one of them could modulate the ability of ␤-catenin to attach other cofactors, even though they do not interact through this domain.
We first analyzed the relevance of this domain in E-cadherin A, C-tail fragments (83 pmol) were incubated with 3.5 pmol of GST-arm or GST as a control as indicated under "Experimental Procedures." Protein complexes were affinity-purified with glutathione-Sepharose and analyzed by Western blot (WB) with a mAb against ␤-catenin C-tail. The numbers below the lanes indicate the amount of bound protein (in fmol); these values were calculated comparing the results of the scanning of the corresponding lanes with known amounts of ␤-catenin proteins that were included as internal references (St) in the same blots. B, 8.3 pmol of the arm domain were incubated with 1 g of C-tail (760-end) bound to glutathione-Sepharose or with the same amount of glutathione-Sepharose alone. The amount of associated ␤-catenin armadillo domain was determined using a mAb specific for this domain. C, binding assay was performed in the conditions described in A with the presence of a 100-fold molecular excess of C-tail-(760-end) when indicated. Addition of a similar excess of an irrelevant peptide did not modify binding of the C-tail-(696-end) to GST-arm (not shown). D, recombinant proteins (8.3 pmol) containing the indicated domains of ␤-catenin were separated by SDS-PAGE and transferred to nitrocellulose membranes. Binding of recombinant proteins was determined by overlay assays as described under "Experimental Procedures" incubating blots with ␤-catenin C-tail-(696-end) (0.52 nmol/ml). Binding was analyzed incubating with a mAb against the C-tail that recognized this overlaid ␤-catenin domain. A, C, and D, in order to verify that equal amounts of proteins were loaded in the gel, blots were stripped and reblotted with mAbs detecting the proteins used as bait (lower panels). The figure shows a representative experiment of three performed in each case, and the graphics at the left of A and D show the average (Ϯ S.D.) percentage of binding to GST-arm of C-tail-(696 -750) relative to C-tail-(696-end) (A), or to the C-tail of ⌬C-tail ␤-catenin relative to arm domain (D). These values were obtained after the densitometry of the three experiments performed.
binding. As mentioned previously, E-cadherin binds exclusively to the armadillo domain (see Ref. 9, and data not shown). As observed in Fig. 5A, deletion of the C-tail greatly increased the amount of the cytosolic domain of E-cadherin (cytoE-cadh) capable of attaching to ␤-catenin (compare results for fulllength and ⌬C-tail proteins, lanes 3 and 4); further deletion of the N-tail (resulting in arm protein, lane 2) did not modify this binding. On the other hand, a mutant protein lacking the N-tail (lane 5) bound very little E-cadherin; this effect was partially reversed when the C-tail was also deleted (lane 2). Therefore, the repression by the C-tail of E-cadherin binding to the armadillo domain is modulated by the N terminus.
Binding of ␣-catenin was also studied. The region of ␤-cate-nin involved in the interaction with ␣-catenin has been identified between amino acids 118 and 146 (13), residues located at the end of the N-tail and the first armadillo repeat. The integrity of this sequence seems to be required because neither the arm domain (amino acids 120 -683) nor the ⌬N-tail mutant significantly bound ␣-catenin (Fig. 5B). On the other hand, the ⌬C-tail mutant associated with ␣-catenin slightly better than the wild-type form (lanes 2 and 3). This result also demonstrated an interconnection between the two tails and suggests that loosening of the N-tail interaction with the arm repeats favors ␣-catenin binding. No significant differences were found between binding of ␣-catenin to wild-type and Y86E forms of ␤-catenin (data not shown). Similar results were obtained when the binding of ␤-catenin to a third partner, TBP, was analyzed (Fig. 5C). This protein binds to ␤-catenin through different sequences, although the main interaction was localized at the last repeats of the armadillo domain and the C-tail (see Ref. 14 and data not shown). Our results clearly indicate that deletion of N-tail facilitates binding of the protein, suggesting that this tail modulates the association of TBP to the C-terminal part of the molecule (Fig. 5C, compare  lanes 1 and 4). Accordingly, further deletion of the C-tail markedly decreases binding of TBP (compare lanes 2 and 4).
These results confirm our hypothesis that both tails coordinately modulate the interaction of ␤-catenin partners either to the central armadillo domain or to the other tail of ␤-catenin. Effects of the tails were detected on the interaction with E- FIG. 4. N-and C-tails modulate ␤-catenin sensitivity to proteolysis. 5.5 pmol of either wild-type ␤-catenin, Tyr-654 3 Glu mutant (Y654E), Tyr-86 3 Glu mutant (Y86E), ⌬N-tail, and ⌬C-tail mutants were incubated with 30 (right side) or 0.1 ng (left side) of trypsin at 24°C. Trypsin digestion was stopped with electrophoresis sample buffer at the indicated times, and samples were analyzed by Western blot with mAbs against ␤-catenin C-tail (right side) or N-tail (left side). The numbers below the lanes correspond to the percentage of ␤-catenin epitope remaining after trypsin treatment. Numbers were calculated scanning the autoradiograms. The experiment, performed in duplicate, was repeated twice, and the average ␤-catenin (Ϯ range) remaining at different times is shown. cadherin and ␣-catenin, two proteins that can simultaneously associate with ␤-catenin in the adherens junctions. We examined the possibility that the binding of one of these ␤-catenin cofactors was modifying the binding of the other and thus affecting the interaction of the tails with the armadillo repeat domain. Pre-association of ␤-catenin to E-cadherin did not increase the amount of ␣-catenin that can be bound to the complex (Fig. 6B). However, binding of ␤-catenin to ␣-catenin raised the affinity of the armadillo repeats to E-cadherin (Fig.  6A). This effect required the presence of the C-tail, because it was not detected when the ␤-catenin ⌬C-tail mutant was used (Fig. 6A). Similar stimulation by ␣-catenin in E-cadherin binding was observed when a Y86E ␤-catenin mutant was analyzed (Fig. 6A).
Conformational changes of the ␤-catenin C-tail induced by E-cadherin binding were also studied by controlled proteolysis assays. ␤-Catenin was incubated with cytoE-cadh, ␣-catenin, or equivalent amounts of irrelevant proteins, and resistance to preoteolysis of an epitope situated in the C-end (amino acids 769 -781) was monitored. As shown in Fig. 7 binding of E-FIG. 5. Effect of ␤-catenin tails on E-cadherin, ␣-catenin, and TBP binding. A, 0.7 pmol of the different ␤-catenin GST fusion proteins containing the indicated domains were incubated with 2.8 pmol of cytoE-cadh. Protein complexes were affinity-purified with glutathione-Sepharose and analyzed by Western blot with anti-E-cadherin mAb. B and C, 6 pmol of the different ␤-catenin GST fusion proteins were incubated with 100 g of total cell extracts from RWP-1 cells. Protein complexes were pelleted down with glutathione-Sepharose, and proteins bound to the complex were analyzed by Western blot with anti-␣-catenin (B) or anti-TBP (C) mAbs. The numbers below the lanes indicate the amount of bound protein (in fmol); these values were calculated comparing the results of the scanning of the corresponding lanes with known amounts of E-cadherin that were included as internal reference (St) in the same blot (A) or presented relative to the value obtained for wild-type full-length ␤-catenin (B and C). In the graphics, the relative binding to the three proteins analyzed (E-cadherin, ␣-catenin, and TBP) of the ␤-catenin fragments assayed is represented with respect to the control (binding to full-length ␤-catenin). The figure shows the results of one representative experiment of three performed for the three proteins and the graphics the average Ϯ S.D. of the relative binding obtained in the three experiments. cadherin to ␤-catenin significantly favored the degradation of this epitope by carboxypeptidase, indicating that the C-end became more accessible to the protease. A similar digestion pattern was obtained when ␤-catenin was previously incubated with ␣-catenin (Fig. 7). These results further support our previous conclusion that ␣-catenin binding promoted the unfolding of the C-tail and, therefore, facilitated the association of E-cadherin. DISCUSSION ␤-Catenin has been shown to interact with many different proteins. In most cases, binding takes place totally or partially through the central part of the protein, the armadillo repeat domain (3,(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23). When analyzed in vitro, this central domain has a high affinity for various molecular partners, but the interaction with the complete protein is weaker (11,(23)(24). Regarding E-cadherin, the ␤-catenin C-tail is responsible for the restriction of its association to the full-length protein. The C-tail end-fragment (amino acids 762-781) is very similar to the E-cadherin sequence involved in the interaction with the armadillo domain (amino acids 657-677) (9) (see Table I). As shown above, a peptide corresponding to the last 22 amino acids of ␤-catenin can bind the armadillo domain. Further support for the relevance of amino acids 750 -781 in this interaction is indicated by experiments showing that a C-tail fragment corresponding to amino acids 696 -750 binds less efficiently to armadillo than the complete C-tail-(696 -781), which covers up to the end of the protein.
Our results also clearly indicate that the two tails are closely located in the conformation adopted by the entire ␤-catenin molecule. For instance, the C-tail is necessary for an efficient binding of the N-tail to the armadillo domain. In the absence of C-tail, binding of the N-tail to the central domain is low. However, no direct binding of the two tails was detected in our assays. These results indicate that the two tails interrelate but only when the armadillo domain is present. The sequence of the C-tail responsible for this effect could be allocated to amino acids 683-750, because a ␤-catenin protein consisting of the armadillo domain (amino acids 120 -750) binds the N-tail similarly to a protein covering the complete C-tail (amino acids 120-end). Therefore, in the C-tail two different sub-domains can be distinguished: segment 683-750, which enhances N-tail binding to the armadillo domain, and the last residues of the protein (comprising at least residues 760 -781) involved in the direct binding to the arm repeats.
The N-tail also affects C-tail binding but in a distinct fashion. In the absence of N-tail, the C-tail is more closely packed against the arm domain as indicated by data showing that the C-tail from ⌬N-␤-catenin is less sensitive to trypsin treatment than the same domain from wild-type ␤-catenin. Therefore, it can also be concluded that the two sub-domains of the C-tail interact with each other negatively; thus, a better interaction of 683-750 with the N-tail would involve a lower association of 750 -781 with the armadillo domain.
Different consequences can be inferred from this mutually dependent association of the tails to the central arm domain. First, binding of ␤-catenin to factors that associate with primary sequences located far from one end might be affected by this tail. We have proved this hypothesis by analyzing the binding to ␤-catenin deletion mutants of three proteins: Ecadherin, ␣-catenin, and TBP. These three proteins attach to different amino acid sequences within ␤-catenin, but in all the cases deletion of the tails more distant from the place of interaction affects the binding.
Another consequence is that the association of two different proteins to separate binding sites on the ␤-catenin molecule might be interdependent. Thus, binding of the first partner might modulate the association of the second, through conformational changes induced in the tails. We have examined this possibility analyzing the binding of E-cadherin and ␣-catenin to ␤-catenin. These two proteins are present in the same complex with ␤-catenin. As mentioned above, binding of E-cadherin competes with the association of the C-tail to the armadillo domain. However, disruption of the binding of the last amino acids of the C-tail to the armadillo domain does not promote a better association of ␣-catenin to the other end of ␤-catenin (to amino acids 118 -146). We think that the interaction of the C-tail sub-domain 683-750 to the N-tail and to the arm domain is not modified after association with E-cadherin. One possible explanation would be that the E-cadherin sequence 655-677 replaces almost perfectly the C-tail amino acids 760 -781 in the structure of the complex; therefore, binding of E-cadherin would only unfold the last amino acids of ␤-catenin without altering amino acids 683-750. Thus, substitution of one peptide for the other does not bring about a conformational change in the N-tail. On the other hand, total deletion of the C-tail (depletion of 683-750) causes the unfolding of the N-tail and the acquisition of a new conformation of this domain that binds ␣-catenin slightly better.
The influence of ␣-catenin on E-cadherin binding to ␤-catenin was also examined. As repeatedly observed, previous association to ␣-catenin increases the affinity of ␤-catenin for Ecadherin. This effect was only detected when the ␤-catenin C-tail was present. This is consistent with results showing that ␣-catenin binding accelerates carboxypeptidase proteolysis of the C-tail. According to our model, ␣-catenin association would modulate the C-tail conformation, inducing a tighter attachment of the 683-750 sub-domain to the N-tail and a looser binding of the C-end to the arm repeats. Therefore, regarding this system, changes in the N-tail elicited by binding of cofactors do transduce efficiently into alterations in the C-tail structure. The influence of the changes in the N-tail on the C-tail conformation versus changes in the C-tail on the N-tail is also probably related to the much more evident effect of ␤-catenin deletions in the binding of E-cadherin versus ␣-catenin. For instance, deletion of the C-tail only increases ␣-catenin binding 2.5-fold, whereas deletion of the N-tail practically blocks Ecadherin association (Fig. 5A, lane 5). Thus, E-cadherin binding is much more responsive to changes in the C-tail than ␣-catenin to those in the N-tail. We cannot rule out the possibility that other factors interacting through the N-tail are much more sensitive to conformational changes in this domain than ␣-catenin and that their binding is more pronouncedly affected by proteins interacting with the C-tail.
This study supports a mechanism of molecular interactions that may help to explain the association of ␤-catenin to Ecadherin and ␣-catenin that occurs in the cell. It is accepted that after its synthesis ␤-catenin interacts with E-cadherin, and the complex migrates up to the adherens junction where it binds ␣-catenin (25). The fact that the affinity of ␤-catenin for ␣-catenin is not modified by E-cadherin binding may explain why the catenin complex does not assemble prematurely. Moreover, once it is formed, the complex is disrupted upon phosphorylation of ␤-catenin Tyr-654, which diminishes its interaction with E-cadherin without affecting the interaction with ␣-catenin (7). This effect is reversible, and several tyrosine phosphatases can dephosphorylate this residue. 2 According to our results, when bound to ␣-catenin, the dephosphorylated ␤-catenin would present a high affinity for E-cadherin, and the adherens junction complex will be rapidly reformed. This fast assembly and dis-assembly of adherens junction is necessary for the epithelial cells to move properly in their target tissue. It has been demonstrated that blocking the dis-assembly of the complex using E-cadherin-␣-catenin fusion proteins impairs the normal movement of epithelial cells (26). On the other hand, when ␣-catenin is not bound to ␤-catenin, binding of this protein to E-cadherin would not be so rapidly reshaped and ␤-catenin could in some circumstances migrate to the nucleus.
In any case, these results also suggest that the coordinated binding of cofactors to ␤-catenin, other than those mentioned here, might also occur. It is possible that some of the many known cofactors that bind ␤-catenin, especially those interacting with the N-tail or the first armadillo repeats, may help to unfold the molecule and facilitate the association of other cofactors to the C-tail or to the armadillo repeats hidden by (or interacting with) this terminal domain. Because this part of the protein contains most of the transactivator elements characterized in ␤-catenin, the possibility that the transcriptional activity of ␤-catenin is controlled in this way warrants further study. FIG. 7. ␣-Catenin binding regulates ␤-catenin C-tail sensitivity to protease. 17 pmol of wild-type ␤-catenin were preincubated in the presence of 51 pmol of either GST, cytoE-cadh or ␣-catenin with 17.5 ng of carboxypeptidase Y at 25°C. Digestions were stopped at the indicated times with electrophoresis loading buffer, and samples were analyzed by Western blot with a mAb against the last 20 amino acids of the ␤-catenin C-end. The numbers below the lanes correspond to the percentage of ␤-catenin epitope remaining after carboxypeptidase treatment. Numbers were calculated scanning the autoradiograms. The experiment, performed in duplicate, was repeated twice and the average ␤-catenin (Ϯ range) remaining at different times is shown. helpful comments on the manuscript and Dr. Berta Ponsati (Lipotech, SA) for synthesizing the peptides used in this work.