Single Amino Acid Substitutions in Proteins of the armadillo Gene Family Abolish Their Binding to -Catenin

Analysis of the calcium-dependent cell adhesion molecule E-cadherin has led to the identification of catenins, which are necessary for cadherin function. Growing evidence that cadherins and catenins are subjected to genetic alterations in carcinogenesis makes it especially important to understand protein-protein interactions within the cadherin-catenin complex. Here we report the identification and analysis of the α-catenin binding site in plakoglobin (-catenin). Using N- and C-terminal truncations of plakoglobin, we identified a domain of 29 amino acids necessary and sufficient for binding α-catenin. The α-catenin binding site is fully encoded within exon 3 of plakoglobin but only partially represented in Armadillo repeat 1. This suggests that exons rather than individual Arm repeats encode functional domains of plakoglobin. Site-directed mutagenesis identified residues in the α-catenin binding site indispensable for binding in vitro. Analogous mutations in β-catenin and Armadillo had identical effects. Our results indicate that single amino acid mutations in the α-catenin binding site of homologs of Armadillo could prevent a stable association with α-catenin, thus affecting cadherin-mediated adhesion.

Analysis of the calcium-dependent cell adhesion molecule E-cadherin has led to the identification of catenins, which are necessary for cadherin function. Growing evidence that cadherins and catenins are subjected to genetic alterations in carcinogenesis makes it especially important to understand protein-protein interactions within the cadherin-catenin complex. Here we report the identification and analysis of the ␣-catenin binding site in plakoglobin (␥-catenin). Using N-and C-terminal truncations of plakoglobin, we identified a domain of 29 amino acids necessary and sufficient for binding ␣-catenin. The ␣-catenin binding site is fully encoded within exon 3 of plakoglobin but only partially represented in Armadillo repeat 1. This suggests that exons rather than individual Arm repeats encode functional domains of plakoglobin. Site-directed mutagenesis identified residues in the ␣-catenin binding site indispensable for binding in vitro. Analogous mutations in ␤-catenin and Armadillo had identical effects. Our results indicate that single amino acid mutations in the ␣-catenin binding site of homologs of Armadillo could prevent a stable association with ␣-catenin, thus affecting cadherin-mediated adhesion.
Cadherins comprise a family of calcium-dependent, homophilic cell adhesion molecules that function in establishing tissue integrity and cell polarity (1). E-cadherin is expressed on the basolateral surfaces of epithelial cells and is concentrated at adherens junctions. Three proteins termed ␣-, ␤-, and ␥-catenin are complexed with the cytoplasmic domain of E-cadherin (2). Biochemical evidence suggests that ␥-catenin is identical to the desmosomal component plakoglobin (3)(4)(5). ␤-Catenin and plakoglobin seem to play a central role in the architecture of two independent cadherin-catenin complexes (CCC) 1 by linking E-cadherin to ␣-catenin (4 -6). The catenins are thought to be involved in connecting E-cadherin to the cortical actin cytoskeleton (7). For proper adhesive function of the CCC, both the homophilic interaction of the extracellular domain and the binding of the catenins to the cytoplasmic domain are essential. If one component is missing or nonfunctional, E-cadherindependent adhesion is abolished (7)(8)(9).
The primary sequences of ␤-catenin and plakoglobin contain a 42-amino acid motif repeated 12 or 13 times, originally identified in the Drosophila segment polarity gene product Arma-dillo (10). Proteins with these repeats have been grouped together as the Armadillo (Arm) repeat family. Arm repeats are present in a variety of proteins with diverse cellular functions (11). This gene family can be further divided into two subclasses, true Armadillo homologs (Armadillo, ␤-catenin, plakoglobin) (12) and more distantly related proteins (p120 cas , band 6 protein, APC, smgGDS, SRP1) (11,13).
Disruption of the CCC can destabilize intercellular junctions, leading to altered cell morphology and increased invasiveness (14,15). For instance, nonadherent PC9 lung carcinoma cells and invasive colon carcinoma cells which do not express ␣-catenin recovered adhesiveness after transfection with ␣-catenin cDNA (16,17). Cells of the human gastritic cancer cell line HSC-39 express a truncated ␤-catenin that does not associate with ␣-catenin and barely form cell aggregates, but made epithelial-like structures after transfection with full-length ␤-catenin (18,19).
Genetic studies support a possible role of the CCC in invasion suppression. For instance, E-cadherin mutations are found in gynecologic cancers and in 50% of diffuse-type gastritic carcinomas (20,21). The gene encoding human plakoglobin has been mapped to chromosome 17q21, a complex genomic region which is subjected to genetic alterations in both sporadic and familial breast carcinomas (22). Plakoglobin has been shown to be subjected to loss of heterozygosity in breast and ovarian tumors (23). Because plakoglobin links ␣-catenin to E-cadherin, mutations in important domains of plakoglobin mediating protein-protein interactions could inhibit cadherin function.
Therefore, we analyzed the interaction between plakoglobin and ␣-catenin. We demonstrate here that a 29-amino acid domain (Q29A) located at the N terminus of plakoglobin (aa 109 -137) is necessary and sufficient for high affinity binding to ␣-catenin in vitro. Single amino acid substitutions in the Q29A sequence abolished the binding of plakoglobin, ␤-catenin, and Armadillo to ␣-catenin in vitro. These amino acids could represent mutational hot spots during tumor progression, as disruption of the cadherin system leads to decreased epithelial adhesion.

MATERIALS AND METHODS
Chemicals and Reagents-Restriction endonucleases, Pwo DNA polymerase and other molecular biology reagents were purchased from Boehringer (Mannheim, Germany) or New England Biolabs (Beverly, MA). ECL detection kit and x-ray films were from Amersham (Braunschweig, Germany). Oligonucleotides were synthesized on an Applied Biosystems model 394A synthesizer with ␤-cyanochemistry. Peptides were synthesized on an Applied Biosystems model 431A with Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry and purified by reversedphase high performance liquid chromatography on a Vydac TP218 column. Peptides were biotinylated at the N terminus with biotinamidocaproate-N-hydroxysuccinimide-ester (Sigma) directly after synthesis.
Cell Culture and Antibodies-The human colon carcinoma cell line SW480 (ATCC CCL-228) was grown in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) heat-inactivated fetal calf serum at 37°C in a 10% CO 2 atmosphere. Affinity-purified polyclonal antibod-* 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 U.S.C. Section 1734 solely to indicate this fact.
Construction of GST-Plakoglobin Fusion Proteins with Deletions of the C Terminus-The plasmid pGEXPlako, encoding a GST-fusion protein for full-length human plakoglobin, has been described (4). C-terminal deletion constructs N110, N125, N129, N137, and N144 were generated by PCR; all PCR reactions in this study were performed using Pwo DNA polymerase with proofreading activity. PCR was carried out with the forward primer BamHI.ATG (ATATGCCGGCGGATC-CATGGAGGTGATGAACCTGATGG) and a reverse primer which added a stop codon followed by an EcoRI restriction site after the last desired codon, e.g. ATTGAATTCTTAGGCCTGCCCCTCCACCTG (Pla-koN110). PCR products were purified using the JetPure Kit (Genomed, Bad Oeynhausen, Germany) and cloned into pGEX4T1 (Pharmacia, Freiburg, Germany) using the BamHI and EcoRI restriction sites. The N139 construct was created by digesting PlakoR142H, which carries a unique SpeI site at codon 139 (see site-directed mutagenesis) with SpeI and NotI, filling with Klenow polymerase and religating. The N233 construct was created by digesting pGEXPlako with SphI and NotI and religating after treatment with Klenow polymerase. All PCR-derived constructs in this study were sequenced to confirm their sequences using Sequenase 2.0 (U. S. Biochemical Corp.).
Construction of MBP-Plakoglobin Fusion Proteins with Deletions of the N Terminus-The MBP-Plako construct was created by recloning the full-length plakoglobin cDNA from pGEXPlako into pMalC2 (New England Biolabs) using BamHI and SalI. Constructs 109C and 120C were created by PCR using a forward primer which added a BamHI site 5Ј of the first desired amino acid: GATGGATCCCAAGCCACCAACCT-GCAGCGA (109C), AGTGGATCCCAGCTGCTCAAGTC (120C), and the reverse primer 2096R (GTCTGGGTTCTTGTCCTC). The purified PCR product was subcloned into pGEXPlako using the BamHI and XmaI restriction sites. Construct ⌬128/139 was created by joining a PCR-derived N-terminal plakoglobin fragment with a C-terminal plakoglobin fragment. For the N-terminal fragment we used forward primer BamHI.ATG with reverse primer 127MunIR (GCACAATTGC-CGACTTGAGCAGCTGGGACG), which introduced a MunI site at amino acid 127. The second fragment was amplified with forward primer 140MunI (CCTCAATTGTCACTCGCGCCCTGCC) and reverse primer 2096R. The N-terminal and the C-terminal fragment were digested with BamHI/MunI or MunI/XmaI, respectively, and ligated into pGEXPlako digested with BamHI and XmaI. Construct 139C was created by digesting PlakoR142H (see "Site-directed Mutagenesis of Plakoglobin") with the enzymes NcoI and SpeI, filling in 5Ј overhangs with Klenow polymerase and religating. In all cases, inserts were recloned from pGEX4T1 into pMalC2 using BamHI and SalI.
Construction of Plakoglobin Peptides Fused to GST-GST-peptide fusion proteins were generated by PCR using the same primers used for the N-and C-terminal truncations. The N-and C-terminal amino acid numbers of the peptide refer to the forward and reverse primer used in the above mentioned N-or C-terminal deletion construct. Purified PCR products were digested with BamHI and EcoRI and ligated into the corresponding sites of pGEX4T1. Colonies were selected by PCR typing (26).
Construction of GST-ECT Chimeric Proteins-The construct pGEXUC1 encoding the cytoplasmic domain of mouse E-cadherin (ECT) has been described (27). All ECT chimeras were obtained by simultaneously ligating a BamHI/XhoI ECT fragment and a XhoI/EcoRI fragment containing the desired domain of plakoglobin into pGEX4T1 digested with BamHI and EcoRI. The BamHI/XhoI fragment of ECT was obtained by amplifying pGEXUCI with forward primer ATAGGATC-CAGAACGGTGGTCAAAGAG and reverse primer ACTACTC-GAGATCGTCCTCGCCACCGCCGTACA. The XhoI/EcoRI fragment of plakoglobin Arm repeat 1 was generated by PCR using forward primer TCACTCGAGCTCAAGTCGGCCATTGTGC and reverse primer AC-CGAATTCTCAAATCATGCCCGCCTTGGTC. The XhoI/EcoRI fragment of plakoglobin exon 3 was generated by PCR using forward primer GTTCTCGAGCTTGAGTACCAGATGTCC and reverse primer ACT-GAATTCTCACGGGTCCTCGTCGTTG. The ECT109 -144 construct was created by subcloning the insert of Plako109 -144 into pMalC2 using BamHI and EcoRI, removing this insert with SalI and EcoRI, and ligating it with the BamHI/XhoI-digested ECT fragment into pGEX4T1 as described above.
Site-directed Mutagenesis of ␤-Catenin and Armadillo-To facilitate recloning of ␤-catenin cDNAs with single amino acid substitutions, a XbaI site was introduced into the ␤-catenin cDNA by silent mutation using the Clontech Transformer kit (Clontech, Palo Alto, CA). The ␤-catenin cDNA was recloned from pGEX␤ (4) into pUC19 using BamHI. Mutagenesis was performed according to the instructions of the manufacturer using the SspI/EcoRV trans oligonucleotide (CTCT-TCCTTTTTCGATATCATTGAAGCATT) and a mutagenic oligonucleotide (CATGATGGCATGTCTAGAGGCTTCCTTTTTGGA). Clones positive for the XbaI mutation were isolated, and the insert was recloned into pGEX4T1 using BamHI. The resulting plasmid (pGEX.M␤X) was used for all subsequent mutagenesis. For primer extension, we used the primer M␤733.Tag1 (GGTGGTGGGACTCGGACTGCTTCACGAGCT-GTCTCTACATCATTTG), containing 23 nucleotides of mouse ␤-catenin on its 3Ј end and the artificial Tag1 sequence on its 5Ј end, and the following mutagenic primers: T120A (CAGCCAAGCGCTGAACGTTT-GCGGGATGAGCAGCGT, Psp1406I); Y142A (CCGCGTCATCCTGAG-CATTAATCAAATTGACA, AsnI); Y142F (CCGCGTCATCCTGAAAAT-TAATCAAATTGACA, AsnI). The mutagenized strand was selectively amplified with forward primer pGEX 903 and the reverse primer Tag1 and subcloned into pGEX.M␤X using NdeI and XbaI.
Expression and Purification of Recombinant Fusion Proteins-All GST-and 6 ϫ His-tagged proteins were expressed and purified as described previously (4). MBP-fusion proteins were isolated according to the instructions of the manufacturer (New England Biolabs). Briefly, bacteria were resuspended in lysis buffer (20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 2 mM EGTA, 1 mM MgCl 2 , 1 mM NaF, and 10 g/ml of DNase I, leupeptin, and phenylmethylsulfonyl fluoride). Cells were lysed by two passages through a French pressure cell (18,000 lbf/in 2 ; 1 lbf/in 2 ϳ 6.9 kilopascals). MBP-fusion proteins from cleared lysates were purified with amylose-agarose beads at 4°C. Beads were washed with lysis buffer. Proteins were eluted with 10 mM maltose in lysis buffer and dialyzed against 20 mM Tris-HCl, pH 7.5. Protein solutions were stored in 50% glycerol at Ϫ20°C.
In Vitro Reconstitution and Affinity Purification-All plastic materi-als were coated with 0.1% (v/v) Triton X-100. Recombinant proteins were mixed in association buffer (10 mM HEPES-NaOH, pH 7.4, 100 mM KCl, 1 mM MgCl 2 , 0.1% Triton X-100) to a final concentration of 10 g/ml of each protein. The mixture was incubated for 1 h at room temperature. Protein complexes were isolated by adding 60 l of glutathione-agarose beads (50% slurry). After 20-min gentle agitation at 4°C the beads were pelleted (2 min, 800 ϫ g) and washed four times with association buffer at room temperature. Bound proteins were eluted with Laemmli sample buffer, separated by SDS-PAGE, and analyzed by Coomassie Blue staining or immunoblotting. Affinity precipitations for MBP-tagged proteins were carried out with amyloseagarose using the same protocol, except that the Triton X-100 concentration was lowered to 0.05%, and all washes were performed at 4°C. SW480 cells (1 ϫ 10 7 ) were metabolically labeled with 50 Ci/ml [ 35 S]methionine (1000 Ci/mmol) (Amersham) for 4 -12 h, washed twice with phosphate-buffered saline, and extracted with 1 ml of lysis buffer (10 mM HEPES-NaOH, pH 7.4, 100 mM KCl, 1 mM MgCl 2 , 2 mM EGTA, 0.2% Triton X-100) phosphatase inhibitors (1 mM each of NaVO 4 and NaF) and protease inhibitors (10 g/ml of leupeptin and phenylmethylsulfonyl fluoride, 0.1 unit/ml of ␣ 2 -macroglobulin). Crude extracts were clarified by centrifugation (4°C, 16, 000 ϫ g, 10 min). The soluble fraction was precleared with 10% (v/v) glutathione-agarose beads; 200 l of precleared supernatant (2 ϫ 10 6 cells) was mixed with 800 l of association buffer and incubated with 2 g of recombinant protein for 1 h at room temperature. Protein complexes were purified with 60 l of glutathione-agarose beads for 20 min at 4°C. Beads were washed five times with association buffer at room temperature. Proteins were eluted with Laemmli sample buffer, separated by SDS-PAGE, and analyzed by fluorography, autoradiography, or immunoblotting. Precipitated ␣-catenin was quantified with the aid of a Fujix BAS 1000 PhosphorImager.

RESULTS
The primary sequence of plakoglobin can be subdivided into three domains according to sequence homology among homologs of Armadillo. Depending on the alignment, the central region consists of a modular structure of 12 or 13 hydrophobic Armadillo repeats (Fig. 1) (11,12). In order to study the interaction between plakoglobin and ␣-catenin, various domains of plakoglobin were expressed as bacterial fusion proteins carrying at the N terminus either GST or MBP. Recombinant ␣-catenin carries six histidine residues (6 ϫ His) as a Tag sequence at the C terminus. The fusion partner was used as a Tag for both purification from crude lysates and affinity isolation of reconstituted protein complexes. The presence of a GST-or MBP-Tag alters the electrophoretic mobility of the recombinant proteins in comparison with their native counterparts (by 26 or 42 kDa, respectively). The identity of the fusion partner in a given deletion mutant had no effect on the interaction with ␣-catenin. Fusion proteins seemed to be folded correctly, because they were recognized by specific antibodies in immunoprecipitation experiments (not shown) and they interacted with their endogenous binding partners in lysates of SW480 human colon carcinoma cells (Fig. 2).
N-and C-terminal Mapping of the ␣-Catenin Binding Site in Plakoglobin-SW480 cells have cytoplasmic pools of catenins which are not associated with cadherins (29,30). In particular, a pool of free ␣-catenin exists, which is not bound to ␤-catenin or plakoglobin. Triton X-100 extracts from radioactively labeled SW480 cells were incubated with full-length and C-terminally truncated plakoglobin proteins. As can be seen in Fig. 2A, PlakoN144, PlakoN233, and full-length plakoglobin bind specifically to a 102-kDa protein. Recombinant ␣-catenin could specifically compete for this binding (not shown). The 102-kDa protein was identified as ␣-catenin by immunoblotting the samples with ␣-catenin-specific antibodies (Fig. 2B). The GSTfusion proteins PlakoN110 and PlakoN125 did not interact with endogenous ␣-catenin ( Fig. 2A). To further narrow down essential amino acids involved in the interaction, various Cterminally truncated fusion proteins were examined in reconstitution assays (Fig. 2C). ␣-Catenin interacted with plakoglo-bin polypeptides containing amino acids 1-137, 1-139, and 1-144, but not with 1-125 and 1-129. Results with C-terminal deletion mutants suggest that amino acids located in a highly conserved 8-amino acid cluster (aa 129 -137) are necessary for binding ␣-catenin.
In order to identify an N-terminal boundary of the ␣-catenin binding site, deletion mutants were constructed lacking parts of the N terminus but extending to the C terminus of plakoglobin. The pMAL-C2 vector, encoding the MBP, was chosen for these constructs, because a flexible spacer of 20 amino acids renders the MBP unlikely to overlap the putative binding site at the N terminus of the truncated plakoglobin proteins. Purified MBP-plakoglobin fusion proteins were incubated with 6 ϫ His-tagged ␣-catenin and affinity-isolated with amylose-agarose beads. The precipitates were immunoblotted with plakoglobin-specific or ␣-catenin-specific antibodies (Fig. 3, A and B). Plako109C bound strongly to ␣-catenin. In contrast, Plako120C, Plako139C, and Plako⌬128/139 (a construct carrying an in-frame deletion of the amino acid cluster mapped with C-terminal truncated proteins) were unable to interact with ␣-catenin. The results with N-terminal deletion mutants of plakoglobin suggest that amino acids between 109 and 120 are necessary for interaction with ␣-catenin. In summary, N-and C-terminal mapping narrowed down the N-and C-terminal borders of the ␣-catenin binding region to amino acids 109 -137 of plakoglobin (Q29A). 168). To determine whether Arm repeat 1 or the sequence encoded by exon 3 is sufficient for binding to ␣-catenin as an independent binding domain, fusion proteins with the cytoplasmic domain of E-cadherin (ECT) were generated. Since ␣-catenin cannot bind directly to ECT (4,30), the ␣-catenin binding site of plakoglobin fused to ECT should permit direct binding of ␣-catenin to the chimeric protein. As can be seen in Fig. 4, A and B, neither ECT nor (more importantly) the chimeric protein ECT fused to Arm repeat 1 of plakoglobin (ECT.R1) interacted with ␣-catenin in reconstitution assays. In contrast, the chimeric proteins of both exon 3 (ECT.E3) and amino acids 109 -144 (ECT.109 -144) of plakoglobin fused to ECT bound tightly to ␣-catenin. The interaction between E-cadherin and ␤-catenin was not affected by the additional association of ECT.E3 with ␣-catenin, strongly suggesting that the ECT fusion partner did not interfere with proper folding of the sequence encoded by exon 3 (not shown). These results strongly suggest that an amino acid sequence encoded by exon 3 of plakoglobin, but not Arm repeat 1, can serve as an independent binding domain for ␣-catenin.

Amino Acid Sequences Encoded by Exon 3 and Not Arm
The Synthetic Peptide Q29A Competes with Plakoglobin for Binding to ␣-Catenin-To see whether the binding site could be further narrowed down, short peptides of plakoglobin were expressed as GST-fusion proteins (Fig. 5A). In reconstitution assays Plako109 -129 fused to GST could not associate with recombinant ␣-catenin, whereas Plako109 -137 and Plako109 -144 could. Plako109 -137 and Plako109 -144, but not Plako109 -129 or Plako120 -144, also co-precipitated native ␣-catenin from metabolically labeled SW480 cells (not shown). These binding assays demonstrate that the amino acid se-

FIG. 4. A sequence encoded by exon 3 of plakoglobin but not
Arm repeat 1 is sufficient to mediate the interaction with ␣-catenin. A, GST-tagged chimeric proteins were incubated with recombinant ␣-catenin as indicated. Protein complexes were purified with glutathione-agarose beads, separated by SDS-PAGE, and stained with Coomassie Blue. B, duplicate gels were blotted and developed with antibodies specific for ␣-catenin. C, schematic localization with respect to fulllength plakoglobin of Arm repeat 1, the ␣-catenin binding site Q29A, and amino acids encoded by exon 3. quence 109 -137 of plakoglobin is sufficient to bind to ␣-catenin. A biotinylated synthetic peptide (peptide Q29A; plakoglobin, aa 109 -137) competed with the binding between fulllength plakoglobin and recombinant ␣-catenin in a dose-dependent fashion (Fig. 5, B and C). Increasing amounts of the peptide Q29A (0, 0.1, 1, and 10 g/ml) were incubated with recombinant ␣-catenin and plakoglobin (1 g/ml each). The ability of 6 ϫ His-tagged ␣-catenin to interact with plakoglobin was analyzed by affinity isolation of the GST-fusion protein followed by immunoblotting. The results shown in Fig.  5, B and C, demonstrate that a constant amount of plakoglobin was associated with decreasing amounts of ␣-catenin due to increasing concentrations of the peptide competitor. No bound ␣-catenin was observed at a peptide concentration of 3 ϫ 10 Ϫ6 M (Fig. 5C). Precipitation first with glutathione-agarose beads and then with avidin-agarose beads demonstrated that ␣-catenin not bound to GST-plakoglobin indeed was associated with the peptide Q29A (not shown). A biotinylated control peptide S19A comprising only the C-terminal half of the binding site (aa 125-143) did not associate with ␣-catenin, nor could it compete with plakoglobin for the formation of heterodimers.
The interaction between the synthetic peptide Q29A and ␣-catenin was also studied using surface plasmon resonance detection (31). This technique measures real-time association and dissociation of molecules on a sensor surface and allows estimates of kinetic binding constants. The biotinylated peptide Q29A was immobilized on a streptavidin-coated sensor surface. Concentration-dependent binding of ␣-catenin was observed and used to calculate kinetic rate constants (not shown). The association rate constant k a was 2.0 ϫ 10 4 Ϯ 2.9 ϫ 10 3 M Ϫ1 s Ϫ1 (n ϭ 5). The dissociation phase was biphasic, probably due either to heterogeneity in the biological material or to rebinding of ␣-catenin during the dissociation phase. Therefore, two dissociation rate constants were obtained: k d 1 1.4 ϫ 10 Ϫ2 Ϯ 2.3 ϫ 10 Ϫ4 s Ϫ1 and k d 2 2.1 ϫ 10 Ϫ3 Ϯ 4.6 ϫ 10 Ϫ5 s Ϫ1 (n ϭ 5). The apparent dissociation equilibrium constant (K D ϭ k d /k a ) was 7.0 ϫ 10 Ϫ7 M for k d 1 and 1.1 ϫ 10 Ϫ7 M for k d 2. The values for the K D obtained from the biosensor agree well with the peptide concentration needed to inhibit binding of ␣-catenin to plakoglobin (3 ϫ 10 Ϫ6 M Ϸ 10 ϫ K D ).
Site-directed Mutagenesis-To identify amino acids that are indispensable for stable complex formation, each of the residues within the binding motif was mutated individually to alanine (alanine mapping). The ability of full-length plakoglobin carrying single point mutations to interact with recombinant ␣-catenin was analyzed by in vitro reconstitution assays (Figs. 6A and 7A). Interestingly, no exchange of a charged amino acid to alanine abolished the binding (e.g. E118A, K124A, D135A, D136A). The mutation of the evolutionarily conserved proline P119 also had no effect. This was somewhat unexpected, as computer predictions suggested this should alter the secondary structure. The mutation Y133A abolished the interaction with ␣-catenin. Interestingly, binding could be restored if the tyrosine was replaced by phenylalanine. Similarly, most mutations of hydrophobic amino acids in the sequence led to loss of binding. This observation suggests that hydrophobic interactions stabilize the heterodimeric complex. The binding activities of a representative group of mutants were unchanged in assays performed in HEPES, phosphate-buffered saline, or imidazole buffers. Reconstitution assays at various temperatures (4, 20, and 37°C) and salt concentrations (20 -200 mM KCl) showed no significant differences. However, the stability FIG. 5. A peptide corresponding to amino acids 109 -137 of plakoglobin is sufficient to bind to ␣-catenin. A, GST-tagged peptides of plakoglobin were incubated with recombinant ␣-catenin in association buffer as indicated. Protein complexes were purified with glutathione-agarose beads, separated by SDS-PAGE, and stained with Coomassie Blue. GST migrates at the same molecular weight as GST-Plako109 -129, because the polylinker of pGEX4T1 adds additional amino acids. B and C, the synthetic peptide PlakoQ29A competes with fulllength plakoglobin to bind ␣-catenin. GST-tagged plakoglobin was incubated with recombinant ␣-catenin and the indicated amounts of the peptide Q29A (numbers indicate the protein concentration in g/ml). Protein complexes of plakoglobin were purified with glutathione-agarose and immunoblotted with antibodies specific for plakoglobin (B) and ␣-catenin (C). D, sequence comparison of the ␣-catenin binding site in homologs of Armadillo. The position of the sequence within the full-length protein is given at the left. Residues identical to those of human plakoglobin are shaded in black. Alignment was performed by the Clustal alignment algorithm.
of the ␣-catenin-plakoglobin dimers could be slightly improved by stabilizing hydrophobic interactions with high salt concentrations (1 M KCl) during the washing procedure.
To demonstrate that the ␣-catenin binding site is conserved in members of the armadillo gene family, point mutations were introduced in the ␣-catenin binding site of mouse ␤-catenin and Drosophila Armadillo (Fig. 6B). Wild type ␤-catenin and Armadillo were able to interact with ␣-catenin from lysates of metabolically labeled SW480 cells. As with plakoglobin, the mutants ␤T120A, ␤Y142A, armT128A, and armY150A showed strongly reduced binding. Again, replacement of the tyrosine with phenylalanine restored the binding. The amount of metabolically labeled ␣-catenin in the precipitates was used to quantify the ability of a given mutant protein to interact with ␣-catenin. The results of these measurements are summarized in Fig. 7A. The data suggest that identical amino acids in all homologs of Armadillo are required for the stable interaction with ␣-catenin via a not yet known hydrophobic interaction mechanism. DISCUSSION Biochemical analysis of eukaryotic cells has already provided important information on the molecular organization of the cadherin-catenin complex (CCC). To investigate the interactions of catenins, which are necessary for the adhesive function of E-cadherin, we expressed components of the CCC as fusion proteins. We have previously assembled the CCC in vitro with recombinant proteins (4). Here we have extended this study and add new information about the interaction between plakoglobin and ␣-catenin. N-and C-terminal truncated proteins defined a 29-amino acid sequence in plakoglobin (Q29A, aa 109 -137) necessary and sufficient for binding to ␣-catenin.
The corresponding motif in other homologs of Armadillo also serves as a binding domain for ␣-catenin. The homologous sequence in ␤-catenin is located between amino acids 118 and 146, which fits well with our previous study identifying amino acids between 120 and 151 of ␤-catenin as necessary for interaction with ␣-catenin (4). The human gastritic cancer cell line HSC-39 has a homozygous in-frame deletion in ␤-catenin that removes amino acids 28 -134, the N-terminal part of the ␣-catenin binding site. This cell line shows extremely weak cadherinmediated adhesion, because the truncated ␤-catenin cannot interact with ␣-catenin (18,19). Using amino acid exchanges in mouse ␤-catenin and Drosophila Armadillo, we confirm that this motif mediates the interaction with ␣-catenin in homologs of plakoglobin.
Although highly conserved in homologs of Armadillo, the ϫ His-tagged ␣-catenin have approximately the same molecular weight. B, extracts of metabolically labeled SW480 cells were incubated with GST-fusion proteins of plakoglobin, ␤-catenin, and Armadillo, either wild type or carrying identical mutations at corresponding amino acids. Protein complexes were affinity-purified with glutathione-agarose beads. Bound proteins were separated by SDS-PAGE and processed for fluorography. FIG. 7. A, quantitative analysis of the alanine mapping of the ␣-catenin binding site in plakoglobin, ␤-catenin, and Armadillo. In the chart above, only the amino acids differing from the wild type sequence are shown. The relative percentage of a given mutant protein associating with metabolically labeled ␣-catenin in comparison with wild type plakoglobin is shown on the right. The values represent the mean of four independent measurements using a PhosphorImager. Individual values obtained from a given mutant varied less than 20%. Numbers higher than 100 indicate an increased stability of the heterodimeric complex during the washing procedure. B, schematic representation of a partial sequence (aa 121-133) of the ␣-catenin binding site as a helical wheel. All hydrophobic amino acids which affect the binding (black circles) are located on one side of the putative helix.
Q29A motif is absent in more distantly related proteins containing Armadillo repeats, indicating that such proteins (e.g. p120 cas , APC, band 6 protein) (11,13) cannot be associated with ␣-catenin by the same mechanism as are homologs of Armadillo. Indeed, previous studies have shown that APC and p120 cas do not directly interact with ␣-catenin (32,33). Based on the alanine mapping, it is also worth noting that the part of the Q29A sequence derived from Arm repeat 1 does not match the consensus sequence for Armadillo repeats (11) and cannot be substituted by any other published Armadillo repeat to reconstitute a functionally ␣-catenin binding site.
We have established the partial genomic structure of the human plakoglobin (23) and mouse ␤-catenin genes (34). The region of plakoglobin encoded by exon 3 (aa 72-156) constitutes a defined protein domain sufficient for interaction with ␣-catenin. In contrast, Arm repeat 1 (aa 123-165) contains only the C-terminal portion of the binding region and is not by itself sufficient to bind to ␣-catenin. This could also be observed in vivo with adhesion-deficient HSC-39 cells, in which the truncated ␤-catenin has an almost complete copy of Arm repeat 1 (19). This finding argues that exon-encoded protein domains rather than individual repeats could be carriers of function in plakoglobin. It favors the idea that individual Arm repeats could assemble together into a scaffold-like structure, which might serve to place functional amino acids at the right place.
Alanine mapping of the ␣-catenin binding site identified mostly hydrophobic amino acids as indispensable for interaction with ␣-catenin in vitro. The implicated residues are spaced appropriately for an ␣-helical interaction mechanism (Fig. 7B). Further analysis of the potential secondary structure using three different software packages (DNAstar, MacVector, and GCG) resulted in different predictions, probably due to the conserved proline residue in the center of the ␣-catenin binding site. Assuming a stabilization of ␣-catenin-plakoglobin dimers through hydrophobic forces, one should expect a similar motif in ␣-catenin.
Exchanges of critical hydrophobic amino acids into hydrophilic residues could affect the interaction between Armadillo homologs and ␣-catenin in vivo. This, in turn, would probably disturb cadherin-mediated adhesiveness, leading to a destabilization of intercellular junctions which might play an important role in the progression of cancer from an adhesive to an invasive state.