Regulation of β-Catenin Structure and Activity by Tyrosine Phosphorylation*

β-Catenin plays a dual role as a key effector in the regulation of adherens junctions and as a transcriptional coactivator. Phosphorylation of Tyr-654, a residue placed in the last armadillo repeat of β-catenin, decreases its binding to E-cadherin. We show here that phosphorylation of Tyr-654 also stimulates the association of β-catenin to the basal transcription factor TATA-binding protein. The structural bases of these different affinities were investigated. Our results indicate that the β-catenin C-terminal tail interacts with the armadillo repeat domain, hindering the association of the armadillo region to the TATA-binding protein or to E-cadherin. Phosphorylation of β-catenin Tyr-654 decreases armadillo-C-terminal tail association, uncovering the last armadillo repeats. In a C-terminal-depleted β-catenin, the presence of a negative charge at Tyr-654 does not affect the interaction of the TATA-binding protein to the armadillo domain. However, in the case of E-cadherin, the establishment of ion pairs dominates its association with β-catenin, and its binding is greatly dependent on the absence of a negative charge at Tyr-654. Thus, phosphorylation of Tyr-654 blocks the Ecadherin-β-catenin interaction, even though the steric hindrance of the C-tail is no longer present. These results explain how phosphorylation of β-catenin in Tyr-654 modifies the tertiary structure of this protein and the interaction with its different partners.

sequence in the first armadillo repeat of ␤-catenin (3). It has been proposed that the interactions of ␤-catenin with these two proteins are regulated by tyrosine phosphorylation (4,5). In the case of E-cadherin, we have recently demonstrated that phosphorylation of tyrosine residue 654 diminishes the association of ␤-catenin to this protein by a factor of 10 (6). This residue is modified in vivo by effectors that concomitantly decrease ␤-catenin-E-cadherin binding (6). On the other hand, there is no direct evidence so far that modification of any Tyr residue on ␤-catenin inhibits its interaction with ␣-catenin.
In addition to its structural role in cellular junctions, ␤-catenin is a critical component of the wnt-signaling pathway that governs cell fate in early embryogenesis (7,8). Activation of this pathway induces the stabilization of free ␤-catenin, its translocation to the nucleus, and its binding to members of the LEF-1/TCF family of transcription factors (7,8). Interaction of ␤-catenin with these factors converts them to transcriptional activators (9) and stimulates the expression of several genes containing Tcf-4-responsive sequences in their promoter (10 -14). In the absence of wnt stimulus, cytosolic ␤-catenin is degraded through a mechanism requiring its binding to the tumor suppressor gene product adenomatous polyposis coli (8). The exact role of adenomatous polyposis coli in the regulation of ␤-catenin levels has not been perfectly explained, although it is thought to facilitate the formation of a complex between ␤-catenin and axin/axil, glycogen synthase 3-B, and ␤-TCRP/slimb (8).
The domains of ␤-catenin involved in transcriptional activation have been localized in the N-and C-terminal parts of this molecule (15,16). The C-terminal tail of ␤-catenin, when fused to LEF-1, has been shown to be sufficient to promote transactivation (15). Although the mechanism underlying this activation is not totally known, the N-and C-terminal transactivation domains of ␤-catenin interact with a growing list of nuclear factors that include the TATA-binding protein (TBP) 1 (16), Pontin (17), Teashirt (18), Sox17 and 13 (19), histone deacetylase (20), SMAD4 (21), the retinoic acid receptor (22), and the CREB binding protein and related proteins (23)(24)(25)(26). One of the essential roles of ␤-catenin-Tcf-4 complex consists in recruiting the basal transcriptional machinery to the promoters of wntsensitive genes. A key component of this transcriptional complex is TBP, which interacts with two different domains of ␤-catenin necessary for transactivation (16).
As mentioned above, phosphorylation of ␤-catenin Tyr-654 severs ␤-catenin-E-cadherin binding (6). Since this residue is located in a domain involved in TBP binding (16) investigated the possible role of this phosphorylation in the interaction between ␤-catenin and TBP.

EXPERIMENTAL PROCEDURES
Expression of Recombinant Proteins-Expression and purification of full-length ␤-catenin, fragments 1-106 and 696-end, and ␤-catenin point mutants Tyr-86 3 Glu, Tyr-86 3 Phe, Tyr-654 3 Glu and Tyr-654 3 Phe have been previously described (6). A DNA fragment corresponding to the complete 12 armadillo repeats (amino acids 138 -683) was amplified from entire ␤-catenin cDNA by polymerase chain reaction using oligonucleotides corresponding to nucleotide sequences 358 -372 and 2035-2047. The 1.7-kilobase amplification fragment was inserted in the BamHI-SmaI sites of a pGEX-6P-1 plasmid and expressed in Escherichia coli as a glutathione S-transferase (GST) fusion protein. Armadillo fragments comprising repeats 7-12 (amino acids 422-683) and 10 -12 (amino acids 575-696) were generated cutting the entire armadillo domain cDNA with EcoRI-EcoRV or EcoICRI-EcoRV and inserting in pGEX 6P-2 digested with EcoRI-SmaI or pGEX 6P-3 digested with SmaI. The ␤-catenin deletion mutants used in this study are presented in Fig. 1, indicating which part of the molecule they comprise. The 1-80-amino acid fragment of Tcf-4 was generated from pcDNA3-hTcf-4 cutting with BamHI and SmaI and inserting in pGEX-6P-1 plasmid. Phosphorylation of ␤-catenin mutant forms by recombinant pp60 c-src protein kinase (from Upstate Biotechnology, Inc.) was performed as described (6). To avoid a possible interference of this kinase in the binding assay, once phosphorylated the GST-␤-catenin protein was purified by chromatography on glutathione-Sepharose 4B as indicated below.
␤-Catenin Binding Assays-The indicated amounts of ␤-catenin proteins or the 12 armadillo repeats were incubated with different concentrations of N-and C-terminal-GST-␤-catenin tails (or GST as a control) at ratios from 1:1 to 5:1 (GST protein versus ␤-catenin) for 30 min at 20°C. Incubations were performed in binding buffer: 50 mM Tris-HCl, pH 7.3, 150 mM NaCl, 3 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol, and 0.1% (w/v) Triton X-100 in a final volume of 200 l. In some experiments binding to the cytosolic domain of E-cadherin (cytoE-cadh) or to the Tcf-4 ␤-catenin binding domain (Tcf-4-(1-80)) was performed in these same conditions. Protein complexes were isolated by incubation with 40 l of a 50% (w/v) suspension of glutathione-Sepharose 4B for 30 min at 20°C. Beads were collected by spinning in a microcentrifuge and washed three times with binding buffer. Samples were separated by SDS-polyacrylamide gel electrophoresis, and the presence of bound proteins in the complex was analyzed by Western blot with specific monoclonal antibodies (mAbs) against ␤-catenin C terminus (Transduction Laboratories, Lexington, KY), ␤-catenin armadillo core (Alexis Biochemicals, San Diego, CA), E-cadherin cytosolic domain (Transduction Labs), or Tcf-4 N terminus (Santa Cruz Biotechnology). Lysate pull-down assays were performed incubating 12 pmol of GST or GST-␤-catenin with 50 g of SW-480 total cell extract in the conditions mentioned above. Samples were purified by glutathione-Sepharose chromatography and the presence of TBP or Tcf-4 in the complex was determined by Western blot with specific mAbs from Transduction Laboratories and Santa Cruz Biotechnology, respectively. Immunoblots were developed with peroxidase-conjugated secondary antibody followed by enhanced chemiluminiscence detection system (ECL, Pierce). The autoradiograms were scanned, and the values obtained were either compared with known amounts of recombinant proteins included as reference (␤-catenin binding assays) or with the value obtained for wild-type full-length ␤-catenin (pull-down assays).
Transient Transfections and Analysis of Transfectants-Assays were performed in SW-480 cell line, which although it contains high levels of ␤-catenin (like most intestinal epithelial cells), it is deficient in Ecadherin (27). Absence of E-cadherin precludes that the observed differences in TBP binding by the different ␤-catenin mutants could be attributed to impaired transport to the nucleus due to a distinct association to E-cadherin. Cells were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum (Life Technologies). When 80% confluent, cells were transfected with the indicated plasmids using LipofectAMINE (Life Technologies) according to the instructions of the manufacturer. After transfection, cells were incubated for 48 h in Dulbecco's modified Eagle's medium plus 10% fetal calf serum. Cell extracts were prepared in radioimmune precipitation buffer (20 mM Tris-HCl, pH 7.6, 150 mM NaCl, 1% sodium deoxycholate, 1% Nonidet P-40, 0.1% SDS, 1 mM EDTA) supplemented with 10 g/ml aprotinin, 20 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, and 0.25 mM Na 3 VaO 4 . Lysates were centrifuged at 13,000 rpm in a microcentrifuge for 5 min at 4°C. 250 g of extract were incubated in a final volume of 0.3 ml with 20 l of a 50% (w/v) suspension of nickel nitrilotriacetic acid-agarose (Qiagen, Hilden, Germany) for 30 min at 4°C. Beads were washed with radioimmune precipitation buffer, and bound proteins were eluted with electrophoresis sample buffer. Samples were separated by 10% SDS-polyacrylamide gel electrophoresis and analyzed by Western blot. To reprove the membranes, blots were stripped as described (28). The absence of signal after stripping was always checked by incubating with the correspondent secondary antibody and ECL reagent.
Analysis of ␤-Catenin-mediated Transcriptional Activity-␤-Catenin-mediated transcription was performed transfecting NIH-3T3 fibroblasts, SW-480 cells, or E-cadherin-deficient MiaPaca-2 pancreas cells with a plasmid containing three copies of the Tcf-4 binding site upstream a firefly luciferase reporter gene (plasmid TOP-FLASH) as described (29). The activity of the product of the Renilla luciferase gene under the control of a constitutive thymidine kinase promoter (Promega) was used as control. Assays were always performed in triplicate; the average of the results of 3-4 independent transfections Ϯ S.D. is given.
Protease Sensitivity of ␤-Catenin-1 g of the different forms of ␤-catenin, phosphorylated or not by pp60 c-src , were incubated in the presence of trypsin (60 ng) 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. Reactions were stopped at different digestion times from 1 to 90 min with electrophoresis loading buffer and boiled for 4 min. The extent of the digestion was determined analyzing the samples by SDS-polyacrylamide gel electrophoresis and Western blot with a mAb anti-␤catenin C terminus, which recognizes an epitope situated between amino acids 696 and 781 of this protein. A quantitation of the reaction was performed scanning the autoradiograms and representing the amount of full-length ␤-catenin at the different times of incubation relative to the initial time.

RESULTS
␤-Catenin is a good substrate of pp60 c-src tyrosine kinase in vitro; this kinase modifies specifically Tyr-86 and Tyr-654, located in the N-terminal domain and in the last armadillo repeat of ␤-catenin, respectively (6) (see Fig. 1). Although Tyr-86 is phosphorylated with a higher stoichiometry, only modification of Tyr-654 alters the interaction of ␤-catenin with E-cadherin. Since Tyr-654 is located in the domain of interaction with TBP, we examined whether tyrosine phosphorylation of ␤-catenin influences the association with this factor. As shown in Fig. 2, phosphorylation of ␤-catenin by pp60 c-src greatly increased its interaction with TBP in pull-down assays (by 6-fold). To analyze the specific influence of Tyr-654 phosphorylation, ␤-catenin mutants were used in which Tyr-86 and Tyr-654 were replaced by Phe. The same amounts of pulled down TBP were obtained when phosphorylated wild-type ␤-catenin or phosphorylated Tyr-86 3 Phe mutant were used as bait (Fig. 2). In this case the amount of Tyr(P) incorporated to the ␤-catenin form was greatly reduced, since only Tyr-654 was phosphorylated (Fig. 2). On the other hand, binding of TBP to the Tyr-654 3 Phe mutant was not increase after phosphorylation, demonstrating that phosphorylation of this residue is involved in the augmented interaction of TBP and ␤-catenin (Fig. 2).
To confirm these results, binding of TBP to ␤-catenin mutants Tyr-86 3 Glu and Tyr-654 3 Glu was determined. These forms were generated to mimic the effect of phosphorylation in these two residues. ␤-Catenin Tyr-654 3 Glu interacted much better with TBP than the wild-type form; the amount of pulled down TBP was eight times greater (Fig. 2). Therefore, it could be demonstrated that the introduction of a negative charge in Tyr-654 enhances ␤-catenin binding to TBP.
We also noticed that phosphorylation of Tyr-86 exerted an opposite effect on ␤-catenin association to TBP. ␤-Catenin Tyr-86 3 Glu consistently pulled down a lower amount of TBP than wild-type ␤-catenin (approximately a 30% less); phosphorylation of Tyr-86 in ␤-catenin Tyr-654 3 Phe exerted a similar action (Fig. 2). Although consistently detected, this negative effect of Tyr-86 phosphorylation on TBP binding was clearly less important than the positive effect observed after Tyr-654 phosphorylation. Probably for this reason, no significant differences were observed in the interaction of TBP to the Tyr-654 3 Glu mutant or to the double mutant Tyr-86 3 Glu/Tyr-654 3 Glu (data not shown).
The effects of ␤-catenin phosphorylation on its association to a well known co-factor, Tcf-4, were determined. No differences in the amount of this protein pulled down by GST-␤-catenin were observed after phosphorylation of this molecule or when ␤-catenin Tyr-86 3 Glu and Tyr-654 3 Glu mutants were analyzed (Fig. 2). The same results were obtained when in vitro binding of recombinant ␤-catenin and Tcf-4 was determined (not shown).
The in vivo association between ␤-catenin and TBP was also investigated. SW-480 cells were chosen for these assays because they contain very little E-cadherin, and most of the ␤-catenin is not retained in the membrane by this molecule. Cells were transfected with wild type or Tyr-654 3 Glu ␤-catenin labeled with polyhistidine and the X-Press® tag to facilitate their purification and identification. Transfected forms were purified by Ni 2ϩ -agarose, and the amount of associated FIG. 2. Phosphorylation of Tyr-654 enhances binding of ␤-catenin to TBP but not to Tcf-4. 11 pmol of GST or GST-␤-catenin fusion proteins were phosphorylated by pp60 c-src in the conditions indicated under "Experimental Procedures." Pull-down assays were performed incubating the GST proteins with 50 g of total cell extracts prepared from SW-480 cells. Protein complexes were pelleted down by affinity on glutathione-Sepharose beads, and proteins bound to the complex were analyzed by SDS-polyacrylamide gel electrophoresis and Western blot with anti-TBP mAb. Membranes were stripped and re-analyzed with mAb against Tyr(P), ␤-catenin, or Tcf-4. wt, wild-type ␤-catenin; Y86F, Y654F, Y86E, and Y654E correspond to ␤-catenin mutants Tyr-86 3 Phe, Tyr-654 3 Phe, Tyr-86 3 Glu, and Tyr-654 3 Glu, respectively. The estimated molecular weights of the bands detected with each antibody are shown. The autoradiograms were scanned in a densitometer, and the results obtained (numbers below the lanes) are presented relative to the value obtained for wild-type ␤-catenin (or phosphorylated wild-type ␤-catenin in the case of the analysis with Tyr(P) mAb). Only the upper band in the analysis of ␤-catenin was employed for this analysis; the lower band corresponds to a degradation product of this protein occasionally observed in our preparations that does not interfere in the assay.
FIG. 1. Diagram of ␤-catenin. The three different domains that form this protein are shown. The 12 armadillo repeats of ␤-catenin are represented with numbered boxes, and the two tyrosine residues phosphorylated by pp60 c-src are also indicated. The deletion mutants used in this article are depicted, indicating which parts of the molecule they comprise. wt, wild type. TBP was determined. As shown in Fig. 3A, TBP associated in vivo better with ␤-catenin mutant Tyr-654 3 Glu than with the wild-type form (2.5-fold better). This higher association correlated with a greater stimulation of ␤-catenin-Tcf-4-mediated transcription. Overexpression of wild-type ␤-catenin in SW-480 cells induced a significant increase (60% stimulation) in the activity of a reporter gene placed under the control of a ␤-catenin-and Tcf-4-sensitive promoter (TOP plasmid) (30). Expression of Tyr-654 3 Glu ␤-catenin mutant raised the activity of this promoter to a higher extent (194% stimulation) (Fig. 3B). Similar stimulations of TOP activity were obtained in other cell lines (Fig. 3B).
Our results indicate that phosphorylation of ␤-catenin Tyr-654 regulates not only the interaction with E-cadherin but with TBP as well. The structural basis of these differences was investigated. Three different regions can be distinguished in ␤-catenin with distinct charge distributions: the N-and Cterminal tails, with pIs close to 4.5, and the armadillo repeat domain, which presents a basic pI of 8.3 (31). It has been proposed that ␤-catenin C-terminal region directly interacts with the armadillo domain (32)(33)(34). The association between the complete armadillo domain (amino acids 138 -683) and the N-and C-terminal regions of ␤-catenin was studied using binding assays with recombinant proteins. Both N-tail (amino acids 1-106) and C-tail (amino acids 696-end) interacted with the armadillo domain. Binding of the C terminus to the armadillo domain requires sequences upstream of the last six armadillo repeats, since a recombinant protein comprising only repeats 7-12 (amino acids 422-683) associated to the C-terminal tail much worse than the complete armadillo domain (Fig. 4B). A possible effect of phosphorylation of ␤-catenin Tyr-654, located in the last armadillo repeat, on armadillo-C-tail association was studied. Phosphorylation of this residue decreased armadillo-C-tail interaction (Fig. 4A) but did not modify the binding of the armadillo domain with the N-tail (Fig. 4C). Consequently, ␤-catenin C-terminal tail also interacted better with the wild-type armadillo domain than with an armadillo form containing the Tyr-654 3 Glu mutation (Fig. 4A).
These results suggest that, in its native conformation, ␤-catenin is folded with its C-tail interacting with the armadillo repeats. Phosphorylation of Tyr-654 disrupts this interaction and releases the C-terminal tail. To prove this model, experiments of limited trypsin proteolysis of ␤-catenin were performed, and the extent of unfolding of the C-tail was followed by measuring the rate of disappearance of ␤-catenin reactivity using an antibody that recognizes only the intact C-tail. As shown in Fig. 5, the ␤-catenin mutant Tyr-654 3 Glu presented a higher susceptibility to proteolysis than the wild- type form. A faster degradation of the wild-type protein was also observed when it was phosphorylated by pp60 c-src . In this case, differences in sensitivity to trypsin proteolysis were less evident, probably due to the incomplete phosphorylation of Tyr-654 in our conditions (6). Phosphorylation of either wild type or Tyr-86 3 Phe mutant produced the same patterns of trypsin digestion (not shown), discarding possible effects due to phosphorylation of Tyr-86 in our assay.
We have also analyzed whether binding of the armadillo repeat domain to the C-tail affected the interaction of ␤-catenin with E-cadherin or Tcf-4. Interaction of the armadillo repeats with the cytosolic domain of E-cadherin (cytoE-cadh) was disrupted by the addition of the C-tail (696-end), indicating that both protein domains interact within the same region of the armadillo domain (Fig. 6A). On the contrary, the addition of ␤-catenin N-tail (1-106) did not modify armadillo-cytoE-cadh association (Fig. 6A). It is remarkable that the armadillo domain bound cytoE-cadh significantly better than full-length ␤-catenin (Fig. 6A), supporting the conclusion that removal of the C-tail facilitates the interaction with cytoE-cadh. On the other hand, binding of the armadillo domain to a Tcf-4 fragment containing the ␤-catenin-binding site was not modified by the addition of both ␤-catenin terminal tails (Fig. 6B).
The binding site for TBP to the ␤-catenin C-terminal domain has been ascribed to amino acids 630 -729, with residues 630 -675 contributing critically to this association (16). In our hands, TBP binds uniquely to the armadillo domain (amino acids 138 to 683) and not to the C-tail (amino acids 696-end) (Fig. 7A). As in the case of cytoE-cadh, ␤-catenin armadillo domain also bound TBP significantly better than full-length ␤-catenin (Fig.  7A), indicating that the C-tail restricted the interaction with TBP. The association of TBP to the 12 armadillo repeats was also competed by preincubation with ␤-catenin C-tail ( Fig. 7A) but not with Tcf-4 (data not shown).
At this point, we also considered the possibility that phosphorylation of Tyr-654 might be inducing alterations in TBP binding independent of the presence of the C-tail. As shown in Fig. 7A, this is not the case; either the wild-type armadillo domain as well as the phosphorylated form of this protein or the Tyr-654 3 Glu mutant pulled down similar amounts of TBP. Thus, these results indicate that changes in TBP binding upon phosphorylation of Tyr-654 are basically due to the release of ␤-catenin C-terminal tail from the armadillo domain, allowing a better interaction of the last armadillo repeats with TBP.
Our results on the binding of the armadillo domain to TBP and to E-cadherin differ in their sensitivity to tyrosine phosphorylation. As shown in Fig. 6 and 7A, whereas phosphorylation of Tyr-654 decreases binding of E-cadherin, it does not modify the interaction of TBP to the armadillo domain. This result suggests that both proteins interact with this domain in a different way. One possibility is that TBP is not binding through an interaction based in the establishment of ion pairs, FIG. 5. Phosphorylation of ␤-catenin tyrosine residue 654 modifies its sensitivity to proteolysis. 1 g of either wild-type (wt) ␤-catenin, Tyr-654 3 Glu mutant (Y654E) or phosphorylated ␤-catenin by pp60 c-src were incubated with 60 ng of trypsin at 24°C. Trypsin digestion was stopped with electrophoresis sample buffer at the indicated times, and samples were analyzed by SDS-polyacrylamide gel electrophoresis and Western blot with a mAb against ␤-catenin C-tail.
The arrowheads indicate the migration of GST-␤-catenin (120 kDa); the lower bands represent degradation fragments generated by trypsin. The numbers below the lanes correspond to the percentage of full-length ␤-catenin remaining after trypsin treatment. Numbers were calculated scanning the autoradiograms. as it has been proposed for E-cadherin. Another possibility is that both proteins interact with different surfaces of the armadillo domain. To explore these possibilities, binding of TBP to full-length ␤-catenin or to the armadillo domain was performed in the presence of an excess of cytoE-cadh. As shown in Fig. 7B, the addition of a 10-fold molar excess of cytoE-cadh did not modify the amount of TBP bound to the armadillo domain, whereas it increased the amount of TBP pulled down by fulllength ␤-catenin. This result suggests that, although TBP and E-cadherin interact with overlapping armadillo repeats, both proteins bind to different faces of ␤-catenin.

DISCUSSION
␤-Catenin has been shown to act both as a regulator of E-cadherin-dependent cell-to-cell adhesion and as an essential mediator in the wnt-signaling pathway (8,35). Experimental data indicate that the presence of ␤-catenin in the cellular junctions is controlled by tyrosine phosphorylation (5, 36 -40). We have previously demonstrated that phosphorylation of Tyr-654, a residue located in the 12th and last armadillo repeat of ␤-catenin, modifies the association of this protein to E-cadherin (6). The armadillo repeat domain has been shown to be essential for the binding of ␤-catenin to its many binding partners, as E-cadherin and the transcription factor Tcf-4. However, binding of both proteins does not show the same requirements; whereas Tcf-4 associates mainly to repeats 3-8 (41), E-cadherin requires the last 8 repeats (2, 9, and 42). Therefore, it makes sense that, as we show in this article (Fig. 2), modification of a residue placed at the 12th armadillo repeat does not affect Tcf-4 binding.
Armadillo repeat 12 has also been characterized as part of the C terminus-transactivating element required for activation of gene expression (16). Our results indicate that phosphorylation of ␤-catenin Tyr-654 increases binding of this protein to TBP both in vitro and in vivo, and this greater association correlates with a higher stimulation of Tcf-4-␤-catenin transcriptional activity. This higher stimulation of Tcf-4 transcriptional activity observed in vivo by ␤-catenin Tyr-654 3 Glu mutant is not a consequence of its impaired association to E-cadherin, since it is observed in cells that present very low levels of E-cadherin. In any case, our data suggest that phosphorylation of Tyr-654 is relevant not only for disruption of ␤-catenin-E-cadherin binding but for stimulation of the interaction of ␤-catenin to the basal transcriptional machinery as well. These results are consistent with the fact that the nonjunctional pool of ␤-catenin is preferentially phosphorylated on tyrosine (37). According to our results, phosphorylation of Tyr-654 affects binding of ␤-catenin to TBP by releasing the restriction created by the C-tail. This restriction is evidenced by the fact that the armadillo domain interacts better with TBP than the complete ␤-catenin and also by the inhibitory effect of the C-tail on the binding of TBP to the armadillo domain. These data have suggested a working model, presented in Fig. 8, which proposes that, when not phosphorylated and not bound to any ligand, ␤-catenin would adopt a folded conformation in which the Cterminal tail and the N-tail interact with the armadillo repeat domain. This conformation would prevent the binding to armadillo repeats of low affinity ligands and would select those (such as E-cadherin) presenting high association constants. Phosphorylation of tyrosine residue 654 would remove the C-tail and allow a better access of TBP to the last armadillo repeats. As indicated, association of Tcf-4, which takes place mainly through armadillo repeats 3-8, would not be affected by the C-tail.
Thus, the presence or absence of a negative charge at Tyr-654 would act as a key for opening or closing ␤-catenin and would affect the association of this protein to factors binding to the last armadillo repeats and, possibly, to the C-tail as well. In some cases, as for TBP (Fig. 7), after removal of the C-tail the presence or not of a phosphate in Tyr-654 does not modify the interaction of proteins with the armadillo domain. However, in other cases the introduction of a negative charge at this position might hamper the binding of ␤-catenin with factors like E-cadherin that interact mainly by charge complementarity (31) (see Fig. 6). Thus, for these proteins the negative effect caused on the establishment of ion pairs would predominate over the removing of the steric interference caused by the C-tail.
We have also demonstrated that, although E-cadherin and TBP associate to the same armadillo repeats, they do not appear to compete directly for binding to the armadillo domain. As mentioned before, these results suggest that E-cadherin and TBP are interacting with different surfaces of this domain, as depicted in Fig. 8. The interference of the C-tail on the binding of these two proteins indicates that the C-tail can interact with, and possibly hide, both binding surfaces. Accordingly, interaction of E-cadherin with ␤-catenin displaces the C-tail from its binding to the armadillo domain and facilitates the interaction of factors as TBP that associates to the other side of this domain (see Fig. 8). Although a simultaneous interaction of E-cadherin and TBP with ␤-catenin is evidently not physiological (E-cadherin and TBP are localized in different cellular compartments), it is possible that a similar role to E-cadherin might be played by other factors interacting with the same binding surface.
Nevertheless, a definitive validation of our model of ␤-catenin regulation by tyrosine phosphorylation would require the determination of the complete structure of this molecule and the characterization of the effect of the two terminal tails in the interaction of ␤-catenin with its numerous protein partners (35). In ␤-catenin, the cylinders represent the ␣-helices that constitute the armadillo domain, and the red dot stands for the phosphate that modifies Tyr-654. In the absence of phosphorylation (left) ␤-catenin adopts a closed conformation with the C-tail interacting with the armadillo repeats (middle left). E-cadherin binds strongly to these repeats through the establishment of ion pairs between its acidic residues and the positively charged groove in the armadillo domain and can displace the C-tail (bottom left). On the other hand, TBP is not able to overcome this restriction while interacting with the last armadillo repeats (upper left). Phosphorylation of Tyr-654 (symbolized by the introduction of a red dot in the armadillo domain, right side) hampers binding of the C-tail, opening ␤-catenin and making the last armadillo repeats more accessible (middle right). Therefore, separation of the C-tail makes the binding of TBP much easier (upper right). However, although the armadillo domain is much more accessible, E-cadherin does not bind to phosphorylated ␤-catenin since the presence of a negative charge makes it difficult to establish the correct ion pairs required for the interaction between these two proteins (lower right). Binding of Tcf-4 (upper) does not require the last armadillo repeats of ␤-catenin and, thus, is not affected by Tyr-654 phosphorylation. Notice that the associations of TBP and E-cadherin with ␤-catenin take place on different surfaces of the molecule and, thus, are not exclusive, whereas the C-tail interacts with both surfaces.