In the First Extracellular Domain of E-cadherin, Heterophilic Interactions, but Not the Conserved His-Ala-Val Motif, Are Required for Adhesion*

The classical cadherins, definitive proteins of the cadherin superfamily, are characterized functionally by their ability to mediate calcium-dependent cell aggregationin vitro. To test hypothetical mechanisms of adhesion, we have constructed two mutants of the chicken E-cadherin protein, one with the highly conserved His-Ala-Val (HAV) sequence motif reversed to Val-Ala-His (VAH), the other lacking the first extracellular domain (EC1). The inversion of HAV to VAH has no effect on the capacity of E-cadherin to mediate adhesion. Deletion of EC1 completely eliminates the ability of E-cadherin to mediate homophilic adhesion, but the deletion mutant is capable of adhering heterophilically to both unmutated E-cadherin and to the HAV/VAH mutant. These results demonstrate that the conserved HAV sequence motif is not involved in cadherin-mediated adhesion as has been suggested previously and supports the idea that in the context of the cell surface, cadherin-mediated cell-cell adhesion involves an interaction of EC1 with other domains of the cadherin extracellular moiety and not the “linear zipper” model, which posits trans interactions only between EC1 on apposing cell surfaces.

Classical cadherin proteins at the cell surface consist of five extracellular domains (ECs), 1 a transmembrane sequence, and a cytoplasmic domain (Fig. 1). Each cadherin extracellular domain shares a folding topology of seven ␤-strands arranged in two ␤-sheets forming a barrel with hydrophobic amino acid side chains largely sequestered within the interior (17)(18)(19)(20). The amino and carboxyl termini emerge from opposite ends of each folded domain, maintaining an orientation with the amino terminus directed away from the cell. Conserved amino acid residues that are capable of coordinating calcium ions are present at the ends of each barrel (17)(18)(19)(20). Binding of calcium at these sites maintains the structural integrity of the cadherin ECs, giving the protein a rigid conformation that allows it to mediate cell-cell adhesion. A single amino acid substitution in a calcium binding site between EC1 and EC2 or between EC2 and EC3 can abolish adhesion and increase motility, whereas mutations in other calcium-binding elements do not affect these behaviors (21,22).
Cadherins expressed on adjacent cells homoassociate through an antiparallel, trans, interaction of their extracellular regions. The bonds that they form across the intercellular space hold the plasma membranes of adjacent cells in close proximity. Adhesive strength is also dependent on a cis interaction between cadherins of the same cell (23) mediated at least in part by the transmembrane domain (24). These stable interactions can be rapidly disassembled or reorganized in response to extracellular signals such as growth factors (25) and can occur without extreme alterations in the makeup of the adhesive complex or the amount of complex at the junction (26).
The mechanism by which the extracellular domains mediate adhesion is poorly understood. The amino acid residues responsible for the trans interaction are currently unknown, although there is solid evidence that the first domain (EC1) is important in forming specific cadherin interactions (27,28). Several different crystal structures have yielded different interaction interfaces (18 -20, 29), and an NMR solution structure of EC1 indicates that it does not dimerize in solution (17). It has also been suggested that the highly conserved His-Ala-Val motif (HAV) of the first domain is important for binding (30,31). E-cadherin peptides containing the HAV sequence have been shown to induce epithelial cell invasion, implying that the adhesive region flanks the HAV sequence (32). However, there is doubt that this motif is crucial because many nonclassical cadherins, and some classical ones, do not have HAV but are still capable of trans adhesion (33,34).
To characterize the role of the HAV motif and of the whole EC1 in adhesion we evaluated the capability of two mutant forms of E-cadherin to mediate adhesion in a set of well defined cell-cell aggregation assays. One mutant has had the conserved HAV motif reversed to VAH, and the other mutant has the entire EC1 domain deleted. The results of these experiments, taken in conjunction with recent reports on human cadherin 4 (R-cadherin) (28) and Xenopus laevis C-cadherin (35), demonstrate that the HAV sequence is not essential for adhesion and that EC1 interacts in trans with EC domains other than EC1 to mediate cell-cell adhesion, consistent with data obtained by atomic force measurements (36 -38) and contrary to the linear zipper model, which is based on the interfaces predicted from crystal structures (18,20) (Fig. 1A).

MATERIALS AND METHODS
Cell Line-The S180 cell line was originally derived from the axial tip of a transplanted mouse sarcoma (39). S180 cells are spindle-shaped, do not undergo calcium-dependent cell-cell adhesion, and do not express any known cadherin proteins, making them ideal for observing the effects of cadherin transfection. S180 cells and all transfectants were grown in Dulbecco's modified Eagle's medium with 15% fetal calf serum in a 10% CO 2 incubator at 37°C. For passaging, cells were released from the plate by incubation with PBS (150 mM NaCl, 2 mM NaH 2 PO 4 , 10 mM Na 2 HPO 4 , pH 7.4) and 5 mM EDTA at 37°C for ϳ15-25 min, which released them from the plates as single cells, counted using a Coulter counter, and replated in culture medium at desired cell densities, usually 10 5 cells/10-cm culture dish in 10 ml of medium.
Construction of E-cadherin Expression Plasmid-The 5Ј-end of the E-cadherin reading frame was amplified from chicken liver cDNA using primers WJG1003 and WJG1004 (Table I) and digested with NheI and XhoI. An XhoI/BamHI fragment containing the majority of the Ecadherin coding region was isolated from plasmid pEC320 (40). pBK-CMV expression vector (Stratagene) was modified by digestion with EcoRI and KpnI followed by polishing with mung bean nuclease and ligation to remove part of the polylinker. The resulting vector (pBK-CMV ⅐E/K) was digested with NheI and BamHI, and a three-fragment ligation was performed. The entire insert of the resulting plasmid was sequenced completely to confirm that the plasmid had the correct sequence. A schematic drawing of the mature protein encoded by this plasmid is shown in Fig. 1B.
Construction of HAV/VAH Plasmid-The HAV/VAH mutant was created by overlapping PCR mutagenesis (41). Two fragments were amplified from unmutated plasmid template using primers WJG1034 and WJG1045, and primers WJG1058 and WJG1042 (Table I). The resulting PCR products were gel purified, and a mixture was used as a template, amplified with primers WJG1034 and WJG1042. The resulting PCR product was digested with XhoI and KpnI, ligated into the wild-type plasmid that had also been digested with XhoI and KpnI, and a cloned plasmid was isolated. The sequence of the resulting insert was determined to confirm that only the designed mutation, inversion of the HAV sequence to VAH, had been introduced into the E-cadherin sequence. A schematic drawing of the mature protein encoded by this plasmid is shown in Fig. 1B.
Construction of ⌬I Plasmid-The ⌬I mutation was created by overlapping PCR mutagenesis (41). As described for the ⌬I mutant, using primers WJG1034 and WJG1046 and primers WJG1047 and WJG1042. There were two undesigned mutations found, one converting an arginine to a proline in the propeptide sequence and the other converting valine to isoleucine in the second extracellular domain. A schematic drawing of the mature protein encoded by this plasmid is shown in Fig. 1B.
Production of Stable Transfectants-S180 cells were transfected with plasmids using an optimized calcium coprecipitation method (42) and selected in medium supplemented with 400 g/ml Geneticin (G418, Invitrogen). Two weeks after transfection, cells were selected for their abilities to express high levels of E-cadherin using magnetic activated cell sorting (Miltenyi Biotec). Cells were released from the culture plate and incubated with rabbit anti-E-cadherin antibody (40 g/ml) in PBS, 2 mM EDTA, 0.5% (w/v) bovine serum albumin, then with magnetic beads coupled to goat anti-rabbit antibody (20 l of bead suspension/10 7 cells). Cells expressing E-cadherin were retained on a high gradient magnetic separation column and then rinsed from the column upon removal of the magnetic field. Clones were isolated from the separated cells by limiting dilution. Expression of E-cadherin was confirmed by immunofluorescent staining with a polyclonal rabbit anti-E-cadherin antibody. Only a single clone from each primary transfection plate was saved, ensuring that the clones are independently derived. Cultures of stable transfectants were maintained in 200 g/ml G418 in Dulbecco's modified Eagle's medium and 15% (v/v) fetal calf serum.
Two unplanned mutations are present in the sequence of the ⌬I construct. In the precursor sequence a Pro was substituted for an Arg residue. This may have an effect on processing of the protein and transport to the membrane. However, the mutant E-cadherin is expressed at the cell surface (Fig. 1E) and is largely correctly processed by cleavage of the precursor peptide ( Fig. 2A). The second mutation is the 44th amino acid in the second domain in which Ile is substituted for Val. This mutation is unlikely to be significant because the residue that was replaced has properties similar to the one that was replaced (a large, The functional tertiary structure of classical cadherins is maintained by coordination of calcium ions between domains (small ovals between ECs). One adhesive interface is located within the EC1, which contains the highly conserved HAV peptide sequence. C-F, immunofluorescence microscopy was performed on S180L-11 (C), untransfected S180 (D), HAV/VAH-1 (E), and ⌬I-1 cells (F), using a polyclonal anti-E-cadherin. In all three transfectants the staining is concentrated at the cell surface in regions of cell-cell contact. In the unmutated E-cadherin and the HAV/VAH mutants, the contact interfaces have a crenelated appearance that has previously been found to be the result of deep intercalation of interlocked membrane processes (arrows in C and E). The ⌬I mutant cell lines do not appear to form these contact structures (arrow in F). The scale bar is 10 m. branched, nonpolar side chain). The side chain faces outward from the domain, not into the hydrophobic interior of the globular domain, so there is no effect on side chain packing in the interior of the domain, and the mutated residue is at a site that is highly variable in other domains and other cadherins (alignment not shown) and which does not correlate with adhesive specificity or subfamily identity. Antibodies-Polyclonal goat antibody was raised against the trypsinreleased extracellular fragment of chicken E-cadherin. Polyclonal rabbit antibody was raised against a fusion protein encompassing most of the E-cadherin extracellular region. FabЈ fragments were prepared by overnight digestion of the whole IgG with pepsin followed by treatment with ␤-mercaptoethanol and iodoacetamide (44).
Western Blotting-Confluent cultures were rinsed with 10 ml of TBS, 1 mM phenylmethylsulfonyl fluoride and extracted with 1 ml of 1% SDS, 1 mM phenylmethylsulfonyl fluoride, 62.5 mM Tris-HCl, pH 6.8; the extracts were scraped into 5-ml polypropylene tubes and homogenized with a Polytron homogenizer (Brinkmann Instruments) to disperse the cells completely and shear the DNA. Extracts were then stored as frozen aliquots, and the protein concentration of the extracts was determined using the BCA method (Pierce).
Aliquots of the extracts were mixed 1:1 with 2ϫ sample buffer (4% (w/v) SDS, 20% (w/v) glycerol, 10% ␤-mercaptoethanol, 0.0002% (w/v) bromphenol blue, 125 mM Tris-HCl, pH 6.8), heated in a boiling water bath for 5 min, and clarified by centrifugation for 10 min at maximum speed in a microcentrifuge. 20 l of each of the samples was resolved on a 6% polyacrylamide gel (47). Proteins were transferred electrophoretically to nitrocellulose, blocked with 3% (w/v) crude ovalbumin, 0.1% Triton X-100 in TBS, and incubated with the rabbit anti-E-cadherin (IgG final concentration, 10 g/ml) and mouse monoclonal anti-␤-tubulin ascites (Amersham Biosciences) diluted 1:5,000 in 3% (w/v) ovalbumin, 0.1% Tween 20, TBS overnight and washed five times with TTBS (50 mM Tris-HCl, 150 mM NaCl, 0.1% Tween 20, pH 8.0). The blot was then incubated with IR800 dye-labeled goat anti-rabbit IgG and IR700 dye-labeled goat anti-mouse IgG (LiCor) in 3% (w/v) ovalbumin, 0.1% Tween 20, TBS for 3 h, washed five times, and visualized using an Odyssey scanner (LiCor). The intensity of each band was determined by integrating pixel intensity over the full band with subtraction of background based on median pixel intensity along the band boundary. E-cadherin expression was normalized to the intensity of the ␤-tubulin band for each sample, and the level of E-cadherin expression was determined relative to the expression level of the S180L-11 sample.
Surface Biotinylation and Avidin Selection-Confluent cultures were rinsed once with PBS, 0.5 mM MgCl 2 , 0.5 mM CaCl 2 and then incubated with 2 ml of 0.5 mg/ml EZ-link biotinylation reagent (Pierce) in PBS, 0.5 mM MgCl 2 , 0.5 mM CaCl 2 at 37°C for 30 min. The culture was then rinsed once with TBS, 1 mM phenylmethylsulfonyl fluoride and extracted with 1 ml of 1% SDS, 1 mM phenylmethylsulfonyl fluoride, 62.5 mM Tris-HCl, pH 6.8, and the resulting extract was stored frozen at Ϫ20°C. 200 l of the SDS extract was mixed with 1 ml of HS buffer (0.1% SDS, 1% sodium deoxycholate, 0.5% Triton X-100, 20 mM Tris-HCl, pH 7.5, 120 mM NaCl, 25 mM KCl, 10 mM EDTA) and clarified by centrifugation for 30 min at maximum speed in a microcentrifuge. The clarified supernatant was transferred to a fresh microcentrifuge tube with 50 l of 1:1 slurry of avidin-coupled agarose beads (Pierce). The samples were mixed by gentle rocking at 4°C for 2 h, and then the beads were pelleted and washed once with HS buffer, once with high salt buffer (HS buffer with 1 M NaCl), and once with low salt buffer (2 mM EDTA, 10 mM Tris-HCl, pH 7.5). The beads were then resuspended in 25 l of 2ϫ Laemmli SDS sample buffer, 100 mM dithiothreitol, heated in a boiling water bath for 5 min, and 20 l of the resulting extract was resolved on a 6% Laemmli SDS gel and visualized by Western blotting.
Aggregation Assays-Aggregation assays were performed according to Hoffman (48). Cells were released from dishes with PBS, 5 mM EDTA, 2% (v/v) fetal calf serum and incubated on ice in Eagle's Spinner medium (Invitrogen) for 1 h to allow full dissociation. Aliquots of 2 ϫ 10 6 cells were incubated with anti-E-cadherin or nonimmune FabЈ fragments (300 g of FabЈ fragments/2 ϫ 10 6 cells) in HDF buffer (137 mM NaCl, 5 mM KCl, 5.5 mM glucose, 4 mM NaHCO 3 , 2 mM EDTA, pH 7.5) for a minimum of 30 min on ice. The assay was initiated by suspending the cells to a final volume of 2 ml in prewarmed Eagle's medium, transferring them to glass scintillation vials, and shaking at ϳ80 rpm at 37°C. At 0, 20, 40, and 60 min after the start of incubation, aliquots of cell suspension were fixed with 1% glutaraldehyde in PBS, pH 7.5. Aliquots of the cell suspension were counted by diluting 1:20 into Isoton II (Beckman/Coulter Electronics) and counting particles within the range of single suspended S180 cell sizes (10 -20 m) with a Z2 Coulter counter. Percentage aggregation for each vial was calculated by the formula 100 ϫ ((N 0 Ϫ N t )/N 0 ), where N 0 equals number of particles at time 0, and N t is the number of particles at the time the cells were sampled. Each point on each graph represents the average of three separate vials, treated with the same FabЈ. Each assay was performed at least three times.
Coaggregation Assays-Coaggregation was performed, with slight modifications, according to Friedlander et al. (49). Cells were loaded with a 3 mg/ml solution of either Texas Red-labeled fixable dextran or fluorescein isothiocyanate-labeled fixable dextran using the Influx method (Molecular Probes). Cells were then maintained in normal medium until harvesting for the experiment (usually ϳ8 h).
Labeled cells were released from the culture dish with PBS and 5 mM EDTA. After counting, cells were diluted to 10 5 cells/50 l in HDF with appropriate FabЈ fragments (20 g/10 5 cells). The assay was initiated by combining both cell types in a final volume of 700 l with minimum Eagle's medium and incubating in 24-well plates coated with Poly-HEMA on a rotary shaker at ϳ80 rpm. Cells were fixed after the 60-min incubation period by addition of the aliquots to 4% (w/v) paraformaldehyde PBS, pH 7.5. To calculate overall levels of aggregation, aliquots were taken from each well at times 0 and 60 min and evaluated using a Coulter counter, as above. Preliminary experiments indicated that if the level of aggregation was low (less than 20%), then there were essentially no visible aggregates to score. Also, if S180 aggregation was high, it was attributable to excess cellular debris aggregating cells in a cadherin-independent manner. When the negative control was contaminated in this way, the experiment was abandoned because any of the other wells could be contaminated in the same way.
The number of Texas Red-labeled cells in each cluster was counted and expressed as a percentage of the total number of cells in that cluster. Comparisons of percentages of the three replicate wells were performed using single factor analysis of variance. Data from 20 coaggregates from three replicate wells were pooled and graphed on a   FIG. 2. Western blot of extracts of S180 cells transfected with the various expression plasmids. Panel A is E-cadherin reactivity, and panel B is ␤-tubulin immunoreactivity from the same blot. Both the unmutated E-cadherin and the HAV/VAH mutant comigrate, indicating that the simple inversion of amino acid sequence has no effect on post-translational processing of the protein. The ⌬I mutant protein has two bands, one migrating more slowly and one more rapidly than the unmutated E-cadherin. C, Western blot of cell surface-exposed proteins isolated by surface biotinylation and avidin precipitation. Neither the HAV/VAH mutation nor the ⌬I mutation prevents transport of Ecadherin to the plasma membrane. The processed (lower molecular weight) ⌬I E-cadherin is preferentially transported to the cell surface.
histogram with bin sizes of 10%. Selected clusters were also photographed using confocal microscopy.

RESULTS
Characterization of Transfectants-S180 cells were transfected with the plasmids encoding wild-type E-cadherin (S180L), the mutant with the His-Ala-Val sequence inverted to Val-Ala-His (HAV/VAH), and the mutant with deletion of EC1 (⌬I). Three clonal cultures of unmutated E-cadherin transfectants, three clonal cultures of HAV/VAH transfectants, and two clonal cultures of ⌬I transfectants were isolated for study. PCR of genomic DNA from each line yielded the expected PCR product. The identity of the transfected construct was confirmed by direct sequencing of the PCR products (data not shown). S180 cells at low density appear spindle-shaped, usually with filopodia spreading on substrate. When the culture reaches confluence, the cells become packed together, forming more than one layer of cells. S180 cells transfected with Ecadherin (S180L) retain the tendency to form multiple layers, but the layers tend to be sheets of cells that are connected through cadherin-mediated adhesion. Immunofluorescence reveals that wild-type E-cadherin localizes to the plasma membrane at sites of cell-cell interaction (Fig. 1C). Contact regions are somewhat flattened with distinctive spikes of concentrated staining that indicate deep intercalation of the apposing adhesive membranes, as described previously for another S180 Ecadherin-transfected cell line (13). HAV/VAH-1,-2, and-4 have the same morphology as S180L cells, in that the fluorescence is present in intense spiky structures along flattened areas of contact between adjacent cells (Fig. 1E).
⌬I cadherin mutant protein is localized to the plasma membrane at regions of contact, but not as intensely as the unmutated E-cadherin and HAV/VAH clones (Fig. 1F). In particular, the spiky areas of intense staining that represent convoluted intercalations of adhering cell surfaces are absent in these cell lines, although uneven staining on free edges of the cells is observed.
Western blots were performed to confirm that protein of the correct size was synthesized and to compare protein expression levels between the various transfectants (Fig. 2). The HAV/ VAH mutant protein comigrates with the unmutated E-cadherin, indicating that the mutation did not interfere with the post-translational modification of the protein. Two distinct bands are visible in both ⌬I cadherin transfectants. One migrates faster than wild-type E-cadherin and is consistent with the expected difference in mobility of the fully processed mutant protein. Note that the difference in molecular mass does not appear to be the expected 12 kDa; however, the mobility of the cadherins in Laemmli gels is consistently anomalous (40, 50 -52), so the relative mobility cannot be taken to be an accurate reflection of molecular mass. The slower migrating species of ⌬I likely represents ⌬I precursors, with the ϳ10-kDa propeptide uncleaved. It is possible that the unintentional substitution of Pro for Arg in the precursor sequence has affected ⌬I post-translational processing. However, a similar defect in processing has recently been reported (35) in an EC1 and EC2 deletion of Xenopus laevis C-cadherin without any changes in the propeptide sequence, so this incomplete cleavage may be solely the result of an artifact fusing the propeptide to an internal site in the cadherin protein sequence. We have also previously seen increased levels of the precursor form in other constructs that have had part of the propeptide altered by replacement of the prepeptide and part of the propeptide with the amino terminus of the neural cell adhesion molecule (N-CAM) (52). Note that the presence of the uncleaved form of the full-length E-cadherin had no effect on the ability of the cleaved form to mediate adhesion (52), i.e. this uncleaved protein does not manifest a dominant negative phenotype. Also (see below), the presence of the uncleaved precursor protein does not prevent coaggregation with cells expressing full-length E-cadherin.
Surface biotinylation of the cells, followed by avidin selection and Western blotting (Fig. 2C), indicates that the HAV/VAH mutation does not prevent transport of the mutant protein to the cell surface. The ⌬I mutation also does not prevent expression of E-cadherin at the cell surface, even in the presence of the uncleaved precursor form; quantitative scanning of band intensity indicates that the surface-exposed E-cadherin is in fact enriched in the cleaved form, suggesting that the uncleaved form may be retained preferentially inside the cell until it is either cleaved or degraded.
Quantitation of band intensities from several independent Western blots demonstrated that the amount of E-cadherin did not vary by more than a factor of 2 between any of the transfected cell lines used in this study, usually less (Table II). Significantly, the levels of ⌬I mutant protein are generally equal to or higher than that of the unmutated protein, so the lack of aggregation observed between the ⌬I mutant-expressing cells (see below) cannot be caused simply by low cadherin expression levels.
Effect of the HAV/VAH Mutation on Homophilic Cell Adhesion-Short term aggregation assays are functional tests of cell adhesion activity. Although immunofluorescent staining in a monolayer can mean that the cadherin is more concentrated at the lateral plasma membrane, indicating a possible adhesive interaction, it is uninformative about the ability of the cells to initiate and maintain stable adhesive contact under the stress of shearing forces. Thus, the more stringent aggregation assay is necessary to evaluate the functional capacity of the two mutants.
HAV/VAH cells are highly adhesive, usually aggregating to the level of wild-type E-cadherin-expressing clones, or higher, within 60 min. Fig. 3A depicts the result of a typical aggregation assay of HAV/VAH cells, compared with S180L cells and untransfected S180 cells. Three independently derived clones aggregated to consistently high levels as exemplified by the graph in Fig. 3A. The observed adhesion in all three HAV/VAH clones was E-cadherin-dependent because preincubation of the cells with anti-E-cadherin FabЈ fragments consistently reduced aggregation to a level comparable with the untransfected S180 cells (ϳ10%). Table III shows the levels of aggregation at 60 min in representative assays of each clone. Pairwise comparisons of the mutant transfectants with S180 and unmutated E-cadherin were made incorporating data from the three independent replications of the experiments for each clone.
Although the HAV/VAH mutant E-cadherin is capable of mediating homotypic interactions resulting in aggregation, it is still possible that the mutation may have impaired their ability to cross-adhere with the wild-type E-cadherin. Mutation of The HAV/VAH clones were able to coaggregate with S180L lines (Fig. 4 and Table IV). They did not coaggregate with untransfected S180 cells. Coaggregation with S180L lines was inhibited by anti-E-cadherin FabЈ fragments, showing that this aggregation was cadherin-dependent. HAV/VAH coaggregated with S180L in a 1:1 ratio, indicating that the heterotypic adhesion between HAV/VAH-expressing cells and unmutated Ecadherin-expressing cells was as probable as the homotypic FIG. 3. Aggregation assays were performed to test the adhesive capacity of the mutated cadherins. Each assay was performed with the wild-type E-cadherin transfectant (S180L-11, q) as a positive control and untransfected S180 cells (ϫ) as a negative control. To demonstrate that the observed adhesion was cadherin-mediated, the cells were pretreated with nonimmune FabЈ fragments (left panel) and function-blocking epitopes of anti-E-cadherin FabЈ fragments (right panel). A, aggregation of mutant HAV/VAH-1 (OE). HAV/VAH aggregation is comparable with that of unmutated E-cadherin and is completely inhibited by specific antibodies. B, aggregation of mutant clone ⌬I-1 (OE). ⌬I fails to aggregate significantly more than untransfected S180 cells, and the aggregation in the presence of specific antibodies is comparable with that in the presence of nonimmune antibodies.  4. Coaggregation experiments to test whether the various mutants will cross-adhere to unmutated E-cadherin or the other E-cadherin mutants. Cells were differentially labeled with Texas Red-(TR) or fluorescein-coupled dextrans. The percentage of Texas Red-labeled cells in each aggregate was calculated; the bars on the histograms represent the number of clusters sharing the same ratio of Texas Red-labeled cells. A, coaggregation of differentially labeled S180L-11; most clusters have ϳ50% Texas Red-labeled cells. B, coaggregation of Texas Red-labeled HAV/VAH-1 with fluorescein-labeled S180L-11. C, coaggregation of Texas Red-labeled ⌬I-1 with fluoresceinlabeled S180L-11. D, coaggregation of Texas Red-labeled ⅐⅐-1 with fluorescein-labeled HAV/VAH-1 cells. adhesion. Also, when the three HAV/VAH mutant clones were coaggregated with S180 cells, HAV/VAH cells constituted 90 -100% of each aggregate, indicating that the presence of a functional cadherin molecule was required for the coaggregation with HAV/VAH to occur. Effect of the ⌬I Mutation on Coaggregation-Unlike HAV/ VAH, the ⌬I mutant protein does not detectably mediate aggregation in a short term suspension assay (Fig. 3B). The level of aggregation of cell lines expressing the ⌬I mutant protein was indistinguishable from untransfected S180 cells, regardless of whether the cells were preincubated with anti-E-cadherin or nonimmune FabЈ fragments. Statistical analysis of the triplicate replications of the aggregations assays of cells expressing ⌬I confirms these conclusions (Table III). The inability of ⌬I to aggregate is consistent with previous data indicating a crucial role for EC1 in cell-cell adhesion (27).
The commonly accepted model for cadherin-mediated cellcell adhesion, the linear zipper model (18), involves a trans interaction between the first domains of cadherins of adjacent cells. However, recent evidence suggests that the trans interaction is more deeply intercalated, involving interactions between EC1 and EC5, implying the possibility of two or more adhesive contact surfaces on each cadherin molecule (36 -38). The fact that the ⌬I deletion E-cadherin does not mediate cell-cell adhesion is not sufficient evidence to distinguish these two models; that result only indicates that EC1 participates in the strong interaction of adhesion, not that it necessarily binds to itself.
When cells expressing the ⌬I mutant E-cadherin are incubated with either cells expressing unmutated E-cadherin or cells expressing the HAV/VAH mutant E-cadherin they coaggregate, at levels that are significantly higher than the background levels of untransfected S180 cells (Fig. 4 and Table IV). Unlike cells expressing the HAV/VAH mutant E-cadherin, however, the clusters that formed during these experiments usually contained 10 -40% ⌬I cells. Thus, the ⌬I-mediated adhesion to fully functional E-cadherin is less favorable than the wild-type, implying that the interaction between ⌬I and wildtype E-cadherin is not as strong or stable as the interaction between two wild-type cells.
⌬I interacts with HAV/VAH to the same extent it does with wild-type E-cadherin (10 -40% ⌬I in a cluster). Coaggregation of HAV/VAH with ⌬I has a peak range that is distinct from the negative control, S180 and the positive control, HAV/VAH.

DISCUSSION
Cadherins are responsible for calcium-dependent interactions between membranes of adjacent cells. A combination of cis interactions between cadherins on the same cell and trans interactions between cadherins on apposing cells is necessary for productive cell-cell binding. However, it is unclear how the extracellular domains are involved in mediating these interactions. There is considerable evidence that isolated extracellular domains do not interact in solution in a way that mimics the trans adhesive interaction (17,35). Cells expressing mutant cadherins that have had the intracellular region that is responsible for ␤-catenin binding deleted do not aggregate (6). However, if a cadherin construct that can be inducibly dimerized through cytoplasmic domain interactions is expressed in Chinese hamster ovary cells, it will mediate adhesion in a dimerdependent manner, indicating that sufficient affinity for cell aggregation depends on cadherins having a multimeric form at the cell surface (53). Thus, the function of cadherins critically depends on the cellular context in which they are expressed.
To date, structure-function analysis of the interaction interfaces of cadherins has largely focused on EC1. This domain was shown to be important in defining the specificity of the trans interaction (27) in aggregation assays. A highly conserved tripeptide HAV, located in the first domain, has been proposed to be a cadherin-type adhesion recognition peptide, similar to the RGD sequence in fibronectin (30,31). Also, the first crystal structure of an EC1 domain, solved for N-cadherin EC1, had two extensive interfaces of interaction in the crystal form and a strand interchange of the amino terminus between adjacent domains (18). The combination of these results led to a generally accepted idea that the trans interaction between cadherins was based on an EC1-EC1 interaction between cadherins on apposed cells, the linear zipper model (18). This model was strengthened substantially by the recent determination of the crystal structure of the C-cadherin ectodomain, comprised of all five EC domains, which has a strand interchange between EC1 domains in an apparently trans orientation. However, atomic force measurement data (36 -38) have provided evidence for an alternative hypothesis, that cadherins are fully intercalated, with trans interactions between EC1 and EC5, and possibly between EC2 and EC4, and EC3 and EC3. The atomic force data also indicate that there can be significant interactions between domains EC1 and EC4, and possibly EC2 and EC3. This deep intercalation model also predicts an intermembrane distance that is much closer to that observed in vivo than that predicted by the linear zipper model.
Our experiments were designed to test these hypotheses of cadherin interaction mechanism in the context of intact cells. The cellular context is particularly important because there is ample evidence that the individual cadherin-cadherin interactions are energetically very weak and that the strong cadherindependent cell-cell interactions depend on a high level of cooperativity that is controlled by the organization of a large number of cadherin monomers at the adhesive surface. The fact that multiple interaction interfaces are found in different crystal structures suggests that there are several interaction surfaces that can mediate interactions between the cadherin monomers during the process of crystallization but that are not relevant to in vivo cadherin interactions. Therefore the various models of interaction that are derived from crystallographic studies must be validated in a cellular context. Two mutations of E-cadherin were constructed: one was a mutant with the putative trans binding site HAV (residues 78, 79, and 80 of the mature protein) inverted to VAH and the other, a clean deletion of the first extracellular cadherin domain at a highly conserved interdomain proline residue (54). These mutants and an unmutated construct were transfected into S180 cells. Both HAV/VAH and ⌬I mutant proteins were expressed at the plasma membrane and were concentrated at areas of cell-cell contact. Thus, we have created cell lines that differ only in the extracellular structure of the cadherin expressed on their surfaces.
The short term suspension aggregation assay is the definitive test for cell adhesion activity (3,55,56). Results of the aggregation and coaggregation assays revealed that the strength and specificity of adhesion of HAV/VAH-transfected cells are indistinguishable from those of the unmutated Ecadherin transfectants (Figs. 3 and 4). If HAV is an essential part of the adhesive interface, then there are two possible scenarios: it must interact directly either with the HAV of the cadherin of an adjacent cell or with some other part of the ectodomain.
In the first scenario, direct HAV-HAV interaction, HAV/VAH mutant cells could be expected to aggregate with other HAV/ VAH cells because the adhesive interface could still be complementary (i.e. His still interacts with Val, Ala with Ala, and Val with His). In this case, however, HAV/VAH would not be able to associate with unmutated E-cadherin, without the complemen-tary inversion of HAV, and cells expressing that mutation would not be able to coaggregate with cells expressing the unmutated protein. However, HAV/VAH mutants did coaggregate with wild-type E-cadherin, in a way that is quantitatively indistinguishable from the homotypic aggregation ( Fig. 4 and Table IV), contrary to this model of a direct interaction.
In the second scenario, in which HAV interacts with a different part of the E-cadherin protein, the mutation to VAH would compromise adhesion by altering the interaction interface in the homophilic adhesion assay. However, HAV/VAH aggregated homotypically at levels that are quantitatively indistinguishable from the unmutated aggregation ( Fig. 3 and Table III). Thus, these results refute the hypothesis that HAV is an important point of adhesive contact for E-cadherin (30,31). Given the similarity in structure between E-cadherin and the other classical cadherins and the fact that individual mutations of the homologous His and Val residues in human R-cadherin have no effect on adhesion competence (28), it seems likely that this highly conserved sequence is involved in a cadherin function other than primary adhesion.
If the linear zipper model of cadherin adhesion were correct, mutation deleting the first extracellular domain would be expected to abolish adhesion. As anticipated, the ⌬I mutants were unable to associate homotypically ( Fig. 3 and Table III). However, in coaggregation assays, the ⌬I mutant protein was able to mediate coaggregation with unmutated E-cadherin and with HAV/VAH mutant protein. Thus, although the ⌬I mutant lacks binding sequences present in the first domain, its binding partners, either HAV/VAH mutant or unmutated E-cadherin, must have regions within EC1 which are capable of adhering to the remaining domains of ⌬I (Fig. 4, and Table IV). ⌬I mutants cannot fully reciprocate in this binding interaction, so the adhesion is not as extensive as that mediated by the unmutated E-cadherin or HAV/VAH mutant. Thus, our results are directly contradictory to the linear zipper model; they support a model in which the trans interaction is not EC1 to EC1, but EC1 to some other domain, probably EC5 or EC4 (36).
Recently Chappuis-Flament et al. (35) have tested the adhesive function of several domain deletion mutants of X. laevis C-cadherin in assays using either aggregation of protein-coated beads or adhesion of transfected Chinese hamster ovary cells to substrate derivatized with the various mutant proteins. Their results using these techniques are fully consistent with the results that we have obtained using purely cell-cell adhesion as an assay. They also concluded, based on their data and on previously published atomic force measurements (36 -38), that the cadherin homophilic adhesion is based on a deep intercalation of cadherins on apposed cell membranes with a major energetic contribution coming from an interaction between EC1 and EC5 or EC4.
The question remains, then, why is the HAV motif so highly conserved in most of the characterized classical cadherins of vertebrates? Clearly, any selection pressure to maintain this motif is based on conservation of some function other than primary adhesion. One likely possibility is that this motif is involved in postadhesion signaling. Treatment of cell cultures with peptides including the HAV motif and adjacent sequence can inhibit a number of processes that occur consequent to adhesion, including embryo compaction and neurite outgrowth (30,57,58), and it has been suggested that the HAV motif is essential for postadhesion signaling, via a trimeric G protein and calcium channels, which cause neurite outgrowth (57,58). In addition, it has been shown recently that a peptide based on the homologous motif in desmosomal cadherins will interfere with desmosome-dependent cell-cell signaling (59).
Recently, two lines of research have indicated a role for N-cadherin EC4 domain in cell signaling. Williams et al. (60) have identified a motif in EC4 of N-cadherin which participates in N-cadherin-dependent activation of fibroblast growth factor receptors. Also, Kim et al. (61) found that EC4 of N-cadherin is essential to the ability of N-cadherin to cause an epithelial to mesenchyme transition when N-cadherin is expressed in epithelial cells; replacement of EC4 in N-cadherin with EC4 from E-cadherin abolishes the transition-inducing activity. In a deep intercalation model of adhesion, the HAV sequence of EC1 is present near EC4 in the adhering surface. Thus, it is plausible that this highly conserved motif may participate not in adhesion, but in a cadherin-dependent signaling pathway that functions in parallel to the well studied cytoplasmic domain/␤catenin pathway. That EC4 swapping from E-cadherin to N-cadherin abolishes the N-cadherin transition activity and that forced expression of E-cadherin can counteract the Ncadherin activity (11) suggest that there may be a cadherintype specificity in cell signaling which is based on structural elements that are different from those that are responsible for adhesive specificity. The cadherins were initially identified by their primary role in mediating calcium-dependent cell-cell adhesion in vitro. Subsequent analyses of their role in cell-cell interactions in tissue in vitro and in vivo have shown that a number of intercellular signaling systems include or are modulated by cadherin interactions. Elements of the extracellular portion of the cadherin protein which are highly conserved are not necessarily involved in the adhesive interaction, however. Identification of amino acid ensembles that are highly conserved but that are not essential for adhesion is a first step in defining how the extracellular portion of the cadherins may mediate cell-cell signaling.