alpha -Catenin binds directly to spectrin and facilitates spectrin-membrane assembly in vivo.

The anchorage of spectrin to biological membranes is mediated by protein and phosphoinositol phospholipid interactions. In epithelial cells, a nascent spectrin skeleton assembles in regions of cadherin-mediated cell-cell contact, and conversely, cytoskeletal assembly is required to complete the cell-adhesion process. The molecular interactions guiding these processes remain incompletely understood. We have examined the interaction of spectrin with alpha-catenin, a component of the adhesion complex. Spectrin (alphaIIbetaII) and alpha-catenin coprecipitate from extracts of confluent Madin-Darby canine kidney, HT29, and Clone A cells and from solutions of purified spectrin and alpha-catenin in vitro. By surface plasmon resonance and in vitro binding assays, we find that alpha-catenin binds alphaIIbetaII spectrin with an apparent K(d) of approximately 20-100 nm. By gel-overlay assay, alpha-catenin binds recombinant betaII-spectrin peptides that include the first 313 residues of spectrin but not to peptides that lack this region. Similarly, the binding activity of alpha-catenin is fully accounted for in recombinant peptides encompassing the NH(2)-terminal 228 amino acid region of alpha-catenin. An in vivo role for the interaction of spectrin with alpha-catenin is suggested by the impaired membrane assembly of spectrin and its enhanced detergent solubility in Clone A cells that harbor a defective alpha-catenin. Transfection of these cells with wild-type alpha-catenin reestablishes alpha-catenin at the plasma membrane and coincidentally recruits spectrin to the membrane. We propose that ankyrin-independent interactions of modest affinity between alpha-catenin and the amino-terminal domain of beta-spectrin augment the interaction between alpha-catenin and actin, and together they provide a polyvalent linkage directing the topographic assembly of a nascent spectrin-actin skeleton to membrane regions enriched in E-cadherin.

The spectrin-actin cytoskeleton contributes to membrane structure and provides molecular linkages between organized membrane domains and the filamentous cytoskeleton (reviewed in Refs. [1][2][3][4]. Its assembly on any given membrane is guided by interactions with membrane proteins and phosphoinositol phospholipids (5)(6)(7)(8). Best understood are linkages involving the adapter protein ankyrin. Ankyrin binds spectrin with high affinity (Ϸ10 -100 nM), linking spectrin to a variety of transmembrane proteins including ion channels or pumps, such as the anion exchanger AE1, the voltage-gated Na ϩ channel, Na ϩ /K ϩ -ATPase, H ϩ /K ϩ -ATPase (9 -13), and cell adhesion molecules of the Ig superfamily such as neuroglian, neurofascin, or NrCAM (7, 14 -16). Other recognized adapter proteins linking spectrin to the membrane include adducin and members of the protein 4.1 superfamily, 4.1R, 4.1B, and 4.1G (reviewed in Refs. 4 and 17). Direct interactions of spectrin with the membrane are mediated by at least three distinct regions of ␤-spectrin, termed membrane association domains 1, 2, and 3 (MAD1, 1 MAD2, and MAD3). MAD1 is located in repeat unit 1 of both ␤I and ␤II spectrin (5) and contains a constitutive targeting signal that together with sequences in region 1 of spectrin can direct spectrin to either the Golgi or plasma membrane of MDCK cells (18). 2,3 MAD2, found in the COOH-terminal pleckstrin homology domain of all spectrins except ␤I⌺1 (4), binds both a protein ligand (5,6) as well as phosphatidylinositol-4,5-P 2 phospholipid (20,21). MAD2 may participate in bestowing G-protein control on the assembly process (21)(22)(23). MAD3 involves a region within ␤-spectrin repeats 3-9 (6) 3 ; its function remains uncharacterized, although it has been proposed to interact with a membrane calmodulin-binding protein (24). Other modes of attachment and other adapter proteins almost certainly exist; their identification remains central to a complete understanding of spectrin assembly and function.
In epithelial cells, assembly of the nascent cortical spectrin skeleton occurs at zones of cell-cell contact, regions where there is productive Ca 2ϩ -mediated homophilic adhesion between surface E-cadherin molecules (8,25,26). Associated with E-cadherin is a group of cytoplasmic proteins that include ␣and ␤-catenin (or ␥-catenin) and p120 (for review see Refs. 25 and 27). ␤-Catenin (or in some cells ␥-catenin) directly binds the cytoplasmic domain of E-cadherin; ␣-catenin joins the membrane complex via a direct association with ␤-catenin or ␥-catenin (28). ␣-Catenin binds and bundles F-actin, an activity that presumably facilitates the attachment of actin filaments to the adhesion complex (29). ␣-Catenin also binds ␣-actinin, a dis-tant member of the spectrin gene superfamily; this interaction may facilitate the docking of actin at the adhesion complex (30). The catenins, and their assembly with the cortical cytoskeleton, are closely linked to the regulation of cadherin function (25,31,32). Spectrin assembles with the adhesion complex soon after productive cell-cell contact is established (26). The interactions guiding this process remain undefined. We now report a direct interaction of spectrin with ␣-catenin, and we demonstrate that ␣-catenin is required for spectrin assembly at the plasma membrane in Clone A cells, a human intestinal cell line. These studies thereby identify ␣-catenin as a novel adapter protein mediating spectrin-membrane association, suggest that this association is necessary for maturation of at least some types of cadherin-mediated junctions, and provide insight into the molecular mechanisms by which spectrin participates in the establishment of specialized membrane domains in polarized cells.

EXPERIMENTAL PROCEDURES
Construction of Expression Plasmids-Molecular biological procedures followed standard protocols (33). Recombinant peptides were prepared in Escherichia coli as before using one of the pGEX vector systems (Amersham Pharmacia Biotech, also see Refs. 29 and 34). Inserts of cDNA encoding selected regions of ␣(E)-catenin were isolated from a human colon cDNA library (GenBank TM accession number L23805, see Ref. 35). The desired region of the cDNA was digested with appropriate restriction endonucleases; the inserts were purified from agarose gels with the Qiaex gel extraction kit (Qiagen) and subcloned into pGEX-2T. The alignment with the ␣-catenin constructs used in relation to the full sequence is shown in Fig. 2 and as used previously (29).
Protein Purification-Recombinant proteins were expressed in the E. coli strains HB101 and DH5␣. For peptides particularly sensitive to proteolysis, strain CAG-456 was used. The expressed peptides were purified from bacterial lysates on glutathione-agarose columns (36). Purified ␣-catenin-GST fusion proteins were utilized as antigens for the generation of polyclonal antibodies that were subsequently affinity purified by absorption to immobilized immunogen. ␣II␤II spectrin was purified by low ionic strength extraction of fresh demyelinated bovine brain membranes, followed by gel filtration on Sephacryl-S-500 HR (37).
Cell Culture-A subcloned line of high resistance type II strain of Madin-Darby canine kidney (MDCK) cells were cultured as before (38). The human colon carcinoma cell line Clone A (39) was provided as a gift by Dr. A. Mercurio, Harvard Medical School. Clone A cells were maintained in RPMI-H 1640 supplemented with 10% fetal calf serum and 0.3% L-glutamine. The human colon carcinoma cell line HT-29 was provided by Dr. X-Y. Fu (Yale University). HT-29 cells were maintained in McCoy's modified medium with 10% fetal calf serum and 0.3% L-glutamine. Cell cultures were incubated at 37°C in a humidified atmosphere of 5% CO 2 .
Immunoprecipitation-Confluent monolayers of MDCK cells (35-mm plates) were washed once in phosphate-buffered saline and lysed at 4°C in 600 l of RIPA buffer (150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 50 mM Tris, pH 7.5) by gentle rocking for 30 min. The lysate was spun for 10 min at 10,000 ϫ g, and the supernatant was cleared by treatment with normal rabbit serum (50 l/ml lysate) and protein A-Sepharose for 30 min on ice. The lysate (100 l) was then incubated with 1 l of anti-spectrin antibody (RAF-A) for 1 h at 4°C and precipitated with 50 l of a 50% suspension of protein A-Sepharose. Immunoprecipitates were washed briefly 2ϫ with RIPA buffer, followed by 2 washes with 400 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 50 mM Tris, pH 7.5. Pellets were resuspended in 50 l of SDS-PAGE sample buffer and evaluated by SDS-PAGE.
Precipitation Assays-Affinity purified anti-GST antibodies were used to precipitate quantitatively 125 I-labeled spectrin bound to GST-␣-catenin fusion proteins. Binding assays were performed in BSA-coated polypropylene tubes (41). Recombinant ␣-catenin peptides N576 or C447 (1 g) were incubated with increasing concentrations of 125 Ilabeled ␣II␤II spectrin for 2 h at room temperature in binding buffer (50 mM NaCl, 10 mM HEPES, 0.5 mM EGTA, 0.5 mM dithiothreitol, 1 mM NaN 3 , 0.2 mM phenylmethylsulfonyl fluoride, pH 6.8) containing 3% BSA. Anti-GST antibody (10 g) was added, and the incubation was continued for 1 h. Protein A-Trisacryl beads (100 l of a 10% final suspension) were added for an additional hour. The bound and free samples were separated by centrifugation through a 20% sucrose cushion at 5,000 rpm. Bound spectrin was quantified by ␥-counting using a 1282 Compugamma CS (LKB instruments). Control experiments ensured that the GST fusion proteins were completely precipitated. Binding to GST alone or to protein-A beads in the absence of fusion protein was used as controls for nonspecific binding and were subtracted from all results to obtain specific binding data.
Overlay Binding-GST was cleaved from recombinant ␣-catenin by thrombin digestion (28). GST-spectrin peptides (1, 0.5, and 0.1 g) were analyzed by SDS-PAGE on 7.5% gels and transferred to PVDF membranes. Membranes were blocked for 1 h in 5% BSA in Tris-buffered saline (TBS) , pH 7.5, and overlaid for 1 h with 0.5 mg/ml recombinant ␣-catenin. After five brief rinses with TBS, a 1:1000 dilution of monoclonal antibody 3H4 was overlaid for 1 h at RT. After five additional washes, goat anti-mouse antibody at a 1:10,000 tagged with horseradish peroxidase was used with enhanced chemiluminescence (Amersham Pharmacia Biotech) to detect the bound ␣-catenin.
Surface Plasmon Resonance-Binding detection by surface plasmon resonance was implemented using a Biacore TM 1000 or 2000 instrument (Biacore AB). This technique of detecting protein-protein interactions is fully described in several publications (e.g. see Refs. [42][43][44]. Purified ␣II␤II spectrin was immobilized onto a carboxymethylated dextran gold surface of the Biacore TM chip in 100 mM acetate buffer, pH 4.5, 50 mM N-hydroxysuccinimide, and 0.2 M N-ethyl-NЈ-(dimethylaminopropyl) carbodiimide hydrochloride. Several chip surfaces were prepared, ranging from 350 to 1500 RUs of bound spectrin. Purified recombinant human ␣-catenin or expressed ␣-catenin subdomains were injected onto the spectrin surface, and the binding was measured as an increase in the resonance units (RU). The kinetic constants, k a and k d , for the binding of ␣-catenin to spectrin were determined as described (45) from a plot of dRU/dt versus R (s Ϫ1 ) versus concentration. The slope of these plots are equal to k a and the abscissa intercept equal to k d . The equilibrium dissociation value is determined from the equation Alternatively, the sensograms were fit using the nonlinear fitting algorithms for multisite binding provided by Biacore AB. Residuals were evaluated for systematic divergences from the fitting algorithms as a measure of the appropriateness of the binding model. The spectrin binding surfaces were regenerated between determinations with 10 mM HCl or 10 mM NaOH. Control studies established the stability of the binding surfaces over the course of these experiments. Binding buffer conditions were 10 mM HEPES, 50 mM NaCl, 1 mM dithiothreitol, 0.5 mM EGTA, 1 mM NaN 3 , pH 6.8.
Transfection Procedure-Clone A cells (Ϸ5 ϫ 10 5) were plated in 60-mm Petri dishes, incubated overnight, and transfected using the LipofectAMINE reagent (Life Technologies, Inc.) with 3-6 g of the plasmid pcDNA3 (Invitrogen) carrying full-length human ␣ 1 (E)-catenin (pcDNA3-␣-catenin) (35). The manufacturer's transfection protocol was followed without modification. When stable lines were desired, neomycin-resistant clones were isolated by selective growth in medium containing 0.6 mg/ml of G418 (Life Technologies, Inc.). Subclones were identified on the basis of their assumption of a more sheet-like morphology, a phenotypic change in Clone A cells characteristic of fulllength wild-type ␣-catenin expression (46). Alternatively, since stable lines often proved unstable, transiently expressing cells were used for most experiments. The expression of ␣-catenin was verified by immunofluorescence and Western blotting.
Immunofluorescence-Cells were plated onto glass chamber slides and grown to subconfluence. Cells were washed five times in TBS, pH 8.0, and fixed in methanol for 20 min at 4°C. After fixation, cells were washed again and blocked with 0.3% BSA/TBS for 1 h at RT. After blocking, cells were washed and incubated with various primary antibodies (diluted in 0.3% BSA/TBS) for 1 h at RT. Antibodies were used at the following dilutions: monoclonal antibodies ␣-catenin (3H4) and 7A11 were used as undiluted culture supernatants; polyclonal antibody (RAFA) to ␣II spectrin was used at 1:500 dilution. After incubation with primary antibodies, cells were washed and incubated with CY3-or CY2-conjugated secondary antibodies (Jackson ImmunoResearch) diluted 1:500 in 0.3% BSA/TBS for 1 h at RT. Cells were finally washed, mounted, and viewed by phase and epifluorescence using an Olympus AX-70 microscope.
Other Methods-All reagents were from Sigma unless otherwise stated. Restriction enzymes were obtained from New England Biolabs. Ampicillin and HEPES were from U. S. Biochemical Corp. Synthetic oligonucleotides were prepared by the Yale Critical Technologies Laboratory. Other reagents and their suppliers were Pfu polymerase from Stratagene; Taq polymerase from Cetus Corp.; Sephacryl S-500 HR and DEAE-Sepharose chromatography media from Amersham Pharmacia Biotech, Na 125 I from Amersham Pharmacia Biotech; Enzymobeads from Bio-Rad; sucrose from ICN; and protein-A Trisacryl-2000 beads from Pierce. Calf brain was obtained immediately after death from a local abattoir, transported at 0°C, washed in 0.32 M sucrose, frozen in liquid N 2 , and stored at Ϫ80°C until use. ␣II␤II spectrin was labeled with 125 I by immobilized lactoperoxidase and glucose oxidase using Enzymobead reagent as per the manufacturer's instructions. Final specific activity of the labeled proteins ranged between Ϸ0.01 and 1 mCi/mg. Protein concentrations were determined by the Coomassie Blue Protein Assay reagent (Pierce) using BSA as a standard. When needed for purposes of precise quantitation, protein concentrations were verified by triplicate amino acid analysis at the Keck Biotechnology Laboratory (Yale University).

RESULTS
␣II␤II Spectrin and ␣-Catenin Associate in Vivo-Several studies have established the coincidence at the light microscopic level of plasma membrane-associated spectrin and the cadherin-catenin adhesion complex along zones of cell-cell contact. A comparison of the pattern of spectrin staining in Clone A cells and HT29 cells (a closely related colonic epithelial cell line) suggests that ␣-catenin may mediate the linkage of spectrin to the adhesion complex. In poorly confluent HT29 cells, spectrin is largely cytoplasmic, as is ␣-catenin (Fig. 1, A-C). When these cells become more confluent and establish effective intercellular cadherin-mediated junctions, ␣-catenin is recruited to zones of cell-cell contact (Fig. 1D). Coincident with this recruitment of ␣-catenin, there is recruitment of spectrin from cytoplasmic to membrane pools, in a distribution indistinguishable from ␣-catenin (Fig. 1, D and E; Fig. 2). Conversely, in Clone A cells, an internal deletion of exons 4 and 5 in the transcribed cDNA of ␣-catenin generates a shortened and mutated ␣-catenin transcript (Fig. 3, also see Refs. 46 and 47). Clone A ␣-catenin is unstable and fails to associate with the cadherin-based adhesion complex (39,46), even though its ability to associate with ␤-catenin and actin appear qualitatively unimpaired (46). As a result, intercellular adhesion in Clone A cells is impaired. Also impaired is the assembly of spectrin to the plasma membrane. This is evident both by its  ). B, anti-␣II␤II spectrin (spec) Western blot of the Triton X-100 soluble and insoluble extracts from HT29 cells versus Clone A cells. In cell monolayers at confluence, the majority of spectrin is typically Triton X-100-insoluble. In the experiment shown, 85% of the spectrin is insoluble in HT29 cells versus 52% in Clone A cells. The lane marked (Sp) is purified ␣II␤II spectrin. s, soluble; p, pellet. C, immunoprecipitates of confluent MDCK, HT29, or Clone A cells with antispectrin antibodies, Western-blotted with the anti-␣-catenin monoclonal antibody 3H4. Note the presence of ␣-catenin in the whole cell lysates (lys) and in the immunoprecipitate with ␣II␤II spectrin antibodies but not in the immunoprecipitate using antibodies to ␣I␤I spectrin or nonreactive antibodies (nrs). Also apparent is the lower molecular weight of the Clone A ␣-catenin (␣-cat). Note that whereas sparse cultures display high cytoplasmic concentrations of both ␣-catenin and spectrin, these proteins assemble together along zones of productive cell-cell contact, with reduction of the cytoplasmic concentrations as cells grow to confluence. This is most evident in the merged images (C and F). Bar ϭ 10m. persistent cytoplasmic intracellular distribution even in confluent monolayers of Clone A cells ( Fig. 2A), as well as by its reduced resistance to extraction with Triton X-100 (compared with HT29 cells, an intestinal line with normal ␣-catenin, Fig.  2B). ␣-Catenin can also be directly demonstrated in immunoprecipitates of ␣II␤II spectrin solubilized from confluent MDCK, HT29, and Clone A cells (Fig. 2C); MDCK cells are an epithelial line with well documented spectrin association at cadherin-based junctions (e.g. see Ref. 8). Collectively, these in vivo observations indicate a tight and possibly direct link-age between spectrin and ␣-catenin and also suggest that these two proteins may associate both in the cytosol and at the membrane.
Spectrin Binds Directly to the NH 2 -terminal 228 Residues of ␣-Catenin-To obtain a more precise analysis of the interaction between ␣-catenin and spectrin, the ability of recombinant ␣-catenin peptides to bind to ␣II␤II spectrin was investigated by surface plasmon resonance. A series of recombinant ␣-catenin peptides were prepared as fusions with GST (Fig. 3A). Also prepared was an ␣-catenin of the type found in the Clone A cells, incorporating an internal deletion. These peptides were used in binding assays either as a GST fusion peptide or as the recombinant proteins alone after removal of the GST by thrombin. Bovine ␣II␤II spectrin was immobilized on the Biacore TM sensor chip surface, and the changes in resonance units were monitored for different concentrations of ␣-catenin (Fig. 3B) or for different ␣-catenin subdomains (Fig. 3C). Sensograms (not shown) were also obtained for different surface loadings of spectrin, to evaluate artifacts arising from limitations of mass transport to the sensor surface. In each analysis, the earliest portions of the association and dissociation phases were analyzed. These are the regions where mass transport (during association) and rebinding (during dissociation) do not dominate the sensogram. Each sensogram was fit to several different binding models, as provided in the Biacore software package, including simple one-to-one Langmuir binding isotherms and two-exponential models. All models generated reasonable fits, with apparent values of k a , k d , and K D within a factor of Ϸ3 of each other for a given peptide. However, no model generated fits with fully random residuals, indicating that the binding of ␣-catenin to immobilized spectrin, although real and of high affinity, is complex and does not conform to any simple binding model. A summary of the apparent kinetic values and derived apparent K D values for ␣-catenin binding to ␣II␤II spectrin is presented in Table I. In this analysis, wild-type ␣-catenin bound spectrin with an apparent K D of 19 -80 nM, GST-␣catenin bound with an apparent K D of 19 -24 nM, and the mutant Clone A ␣-catenin, devoid of GST, bound with an apparent K D of 15-25 nM. The differences in apparent binding affinity between the GST-␣-catenin versus peptides without GST presumably reflects the propensity of GST to induce homodimerization. Oligomers of ␣-catenin generated by this mechanism would bind with enhanced affinity, a phenomenon that has now been well documented (19). A surprise is the apparently greater affinity of GST-free Clone A ␣-catenin for spectrin versus wild-type ␣-catenin. The genesis of this effect is unknown and was not further studied, although such a finding does suggest the possibility that the Clone A mutation might affect the oligomerization pathway of native ␣-catenin (28).
To identify the site in ␣-catenin that interacts with spectrin, additional recombinant peptides derived from the NH 2 and COOH termini of ␣-catenin (Fig. 3A) were prepared and assayed qualitatively for their ability to bind directly to ␣II␤II spectrin using surface plasmon resonance (Fig. 3C). As before, full-length ␣-catenin (peptide a907) bound avidly. Peptide N576, representing the NH 2 -terminal half of ␣-catenin, bound in a similar way, achieving approximately half of the RUs of peptide a907. A peptide representing the NH 2 -terminal 228 residues of ␣-catenin also bound. Since the Biacore TM instrument measures a mass change at the sensor surface, the decrease in RU values seen with the various NH 2 -terminal peptides are in proportion to their relative molecular weights and indicate that all of these peptides, loaded at equal concentrations, are saturating the spectrin surface to the same extent. Conversely, the C447 peptide, encompassing the COOH-terminal half of ␣-catenin, did not bind to spectrin any better than FIG. 3. ␤II spectrin binds directly to the NH 2 -terminal 228 amino acids (aa) of ␣-catenin. A, schematic representation of the structure of ␣-catenin, the mutation found in Clone A cells (shaded box (46)), and the recombinant ␣-catenin peptides used in this study. On the right is shown an SDS-PAGE analysis, Coomassie Blue-stained, of the purified ␣II␤II spectrin (S) and each of the ␣-catenin peptides (with GST removed). Their M r Ϭ 1000 is shown. B, surface plasmon resonance analysis of the interaction of GST-␣-catenin with immobilized ␣II␤II spectrin. A representative analysis is shown. Binding was evaluated on sensor chips containing three different concentrations of spectrin (see "Experimental Procedures"). Each curve (from bottom to top) represents the sensogram trace of the binding of 0.15, 0.30, 0.60, 1.2, and 2.4 M recombinant GST-␣-catenin, respectively. Similar experiments carried out with different levels of immobilized spectrin, and with recombinant ␣-catenin in which the GST moiety had been removed by thrombin treatment, gave comparable results. Interestingly, analysis of recombinant ␣-catenin representing the form found in Clone A cells, with the internal deletion of sequences encoded by exons 4 and 5, did not change its affinity for spectrin (data summarized in Table I). C, comparison sensogram of the binding of various GST-␣-catenin peptides at 5.0 M to ␣II␤II spectrin. The ␣-catenin peptides are as indicated. Note that only those peptides containing the NH 2 -terminal 228 residues demonstrate appreciable binding and that the level of binding achieved in the sensogram at saturation (for active peptides) is roughly proportional to the mass of each active peptide. D, solution binding of ␣II␤II spectrin to ␣-catenin peptides. Increasing concentrations of 125 I-labeled ␣II␤II spectrin were incubated with ␣-catenin peptides N576 (f) or C447 (q). The binding of spectrin to GST alone or to the protein A beads in the absence of fusion protein was taken as a measure of nonspecific binding. did GST or BSA alone. Thus, it appears that a site near the NH 2 terminus of ␣-catenin fully accounts for its interaction with spectrin.
Finally, it was of interest to determine whether ␣-catenin would also bind ␣II␤II spectrin in vitro in solution (versus immobilized spectrin on the Biacore sensor surface). Increasing concentrations of purified 125 I-labeled ␣II␤II spectrin were mixed with either of the recombinant GST fusion proteins N576 or C447. Antibodies to GST were used to co-precipitate GST-catenin along with bound 125 I-labeled spectrin, which was quantified by ␥-counting (Fig. 3D). Nonlinear regression analysis of the binding isotherm (fitted line) revealed K D values of 164 Ϯ 86 (2 S.D.) nM for spectrin binding to GST-N576, with an estimated K max of 0.50 Ϯ 0.12 (2 S.D.) mol of spectrin dimer bound per mol of GST-N576. In these assays, there was no binding of spectrin to GST-C447 or GST alone. The K D determined from this assay agreed reasonably well with those from the Biacore studies, especially considering the differences in technique. Collectively, they demonstrate a strong and direct interaction between ␣II␤II spectrin and the NH 2 -terminal 228 residues of ␣-catenin. Consistent with this binding locus, no differences in spectrin binding by Clone A ␣-catenin were detected. Clone A ␣-catenin deletes residues 197-354 of the native protein, suggesting that the actual interaction site in ␣-catenin for spectrin is proximal to residue 197.
␣-Catenin Binds to the First 313 Residues of ␤II Spectrin-The site to which ␣-catenin binds in ␤II spectrin was identified by gel-overlay assay (Fig. 4). Recombinant GST fusion peptides representing all regions of human ␤II spectrin were transferred to PVDF membranes and overlaid with ␣-catenin (from which GST had been removed) (Fig. 4). Of the peptides examined, only those (␤II N-1 , ␤II N-4 , and ␤II N-6 ) that included the NH 2 -terminal region of ␤II spectrin bound ␣-catenin (Fig. 4B,  center). To assess further the relative affinities of ␣-catenin for this region of ␤II spectrin, overlay experiments were designed using a range of peptide concentrations (Fig. 4B, right). Regardless of concentration, ␣-catenin did not bind to GST alone or to ␤II 9-C . Conversely, strong binding was detected at every Center, PVDF transfer of spectrin recombinant peptides, overlaid with ␣-catenin and developed with the 3H4 antibody to ␣-catenin. Note the strong binding of ␣-catenin to ␤II N-6 , ␤II N-4 , and ␤II N-1 and the absence of binding to the other spectrin peptides or to GST alone. Data from peptides prepared as a fusion with GST or as the peptide alone after removal of the GST are shown. Right, PVDF transfer of ␤II 9-C , ␤II N-1 , or GST alone at three loadings (1.0, 0.5, or 0.1 g) overlaid with ␣-catenin (GST-free). Note the strong binding to ␤II N-1 with no detectable binding to ␤II 9-C or GST at any concentration.

TABLE I
Kinetics and derived dissociation constants as measured by surface plasmon resonance The values presented represent only an apparent K D , since in neither case (simple bimolecular Langmuir binding or two-exponential fit) were the fitting residuals completely random. Several mechanisms may contribute to this complexity, including surface microheterogeneity, limitations in mass transport to and from the binding surface, and complex conformational dependent binding mechanisms inherent to the proteins themselves. Available data are insufficient to discriminate between these possibilities. The data do demonstrate unequivocal high affinity binding of ␣-catenin to spectrin. concentration of the ␤II N-1 peptide. This active peptide, the smallest one examined in these experiments, spans residues 1-313 of ␤II spectrin and places the ␣-catenin-binding site within this region. ␣-Catenin Facilitates Spectrin Membrane Assembly in Vivo-Clone A cells are defective in cell-cell adhesion and harbor an internal deletion in the expressed ␣-catenin (39,46). This mutation leads to the loss of ␣-catenin associated with the plasma membrane, reduced cell-cell adhesion, and coincidentally, reduced assembly of ␣II␤II spectrin at the membrane (Figs. 2 and 5). These cells do, however, form epithelial-like sheets at confluence (albeit with altered morphology and highly refractile membranes) and display surface E-cadherin and ␤-catenin staining (46). To test whether the failure of spectrin assembly in these cells was due to the defect in ␣-catenin, Clone A cells were transiently transfected with wild-type ␣-catenin, and the assembly of ␣-catenin and ␣II␤II spectrin at the membrane was monitored (Fig. 5). The transfected wild-type ␣-catenin was fully competent for assembly with the cadherin adhesion complex at the plasma membrane (Fig. 5, A and C), and coincident with its appearance, ␣II␤II spectrin was restored to its plasma membrane location (Fig. 5, B and D). Thus, wild-type ␣-catenin is fully competent to rescue the impaired membrane assembly of spectrin in Clone A cells. DISCUSSION These findings establish that ␣-catenin can bind directly to spectrin, that these proteins are associated in cultured epithelial cells in vivo, and that ␣-catenin facilitates the plasma membrane assembly of spectrin in regions of direct cell-cell contact. These conclusions are supported by several lines of evidence as follows. (i) In vitro coprecipitation, surface plasmon resonance, and gel overlay assays detect a direct interaction of moderate to high affinity between the NH 2 -terminal domain of ␣-catenin and the first 313 residues of ␤II spectrin. (ii) Spectrin and ␣-catenin co-localize and co-precipitate in confluent monolayers of Clone A, HT29, and MDCK cells. (iii) Spectrin and ␣-catenin do not assemble into a detergent-insoluble matrix at the plasma membrane in clone A cells, a defect restored by transfection of wild-type ␣-catenin. Collectively, these findings suggest that in addition to its other roles, ␣-catenin acts as a novel adapter protein directing the assembly of a nascent cor-tical spectrin membrane skeleton to zones of productive cellcell adhesion. It is also likely that the binding of spectrin to the adhesion complex stabilizes the complex itself and facilitates adhesion by linking adjacent adhesion complexes into macromolecular membrane mosaics centered at cadherin-based junctions. In this respect, the interaction of spectrin with the adhesion complex is but a specific example of the more general role of spectrin as an organizer of linked membrane mosaics (2,4).
The demonstration of a direct interaction between ␣-catenin and spectrin is reminiscent of the binding of ␣-catenin to actinin (30). A member of the spectrin gene superfamily, ␣-actinin shares the repeat structure of spectrin and binds F-actin. However, unlike for spectrin, a sequence in the two central ␣-actinin repeat units appears to interact with a region in ␣-catenin that is downstream of the spectrin-binding site identified here. This result is a bit surprising given the similarity of ␣-actinin to spectrin and suggests that spectrin and ␣-actinin, despite their similarities, play distinct roles in the physiology of the adhesion complex. The presence of distinct binding sites for both ␣-actinin and spectrin (and actin, Ref. 29) in ␣-catenin, as well as independent binding sites for actin in both spectrin and ␣-actinin, suggests that these molecules can bind simultaneously and independently (although this premise has not been formally examined). Thus, it is likely that a cooperative and redundant interaction of ␣-catenin with spectrin, F-actin, and ␣-actinin guides the assembly of a spectrin-actin skeleton to regions of cell-cell contact. The findings in Clone A cells lend support to this notion. Although ␣-catenin from Clone A cells binds spectrin normally (as it does F-actin and ␤-catenin (46)), the deletion in this catenin (residues 197-354) overlaps a region previously demonstrated to bind ␣-actinin (residues 325-394 (30)). Perhaps a loss of ␣-actinin binding leads to an impairment of actin assembly at the membrane, with consequential impairment of spectrin-actin assembly in zones of cell-cell contact.
FIG. 5. Wild-type ␣-catenin restores spectrin assembly at the membrane in Clone A cells. Clone A cells were transiently transfected with wild-type ␣-catenin and stained with the monoclonal antibody 7A11 which only recognizes wild-type ␣-catenin (46) (A and C) or with RAF-A, an antibody to ␣II␤II spectrin (B and D). Note the cluster of transfected cells that express ␣-catenin at the membrane and assume a more epithelial morphology. As shown in higher power in the bottom panels (C and D), the spectrin in these cells shifts from a largely cytoplasmic distribution in the untransfected Clone A cells (arrows, D) to a predominantly plasma membrane association (arrowheads, D).