The Minimal Essential Unit for Cadherin-mediated Intercellular Adhesion Comprises Extracellular Domains 1 and 2*

N-cadherin comprises five homologous extracellular domains, a transmembrane, and a cytoplasmic domain. The extracellular domains of N-cadherin play important roles in homophilic cell adhesion, but the contribution of each domain to this phenomenon has not been fully evaluated. In particular, the following questions remain unanswered: what is the minimal domain combination that can generate cell adhesion, how is domain organization related to adhesive strength, and does the cytoplasmic domain serve to facilitate extracellular domain interaction? To address these issues, we made serial constructs of the extracellular domains of N-cadherin and produced various cell lines to examine adhesion properties. We show that the first domain of N-cadherin alone on the cell surface fails to generate adhesive activity and that the first two domains of N-cadherin form the “minimal essential unit” to mediate cell adhesion. Cell lines expressing longer extracellular domains or N-cadherin wild type cells formed larger cellular aggregates than those expressing shorter aggregates. However, adhesion strength, as measured by a shearing test, did not reveal any differences among these aggregative cell lines, suggesting that the first two domains of N-cadherin cells generate the same strength of adhesive activity as longer extracellular domain cells. Furthermore, truncations of the first two domains of N-cadherin are also sufficient to form cisdimerization at an adhesive junction. Our findings suggest that the extracellular domains of N-cadherin have distinct roles in cell adhesion, i.e. the first two domains are responsible for homophilic adhesion activity, and the other domains promote adhesion efficiency most likely by positioning essential domains relatively far out from the cell surface.

Cadherins exist as a superfamily of calcium-dependent cell adhesion molecules. They play important roles in cell sorting in embryonic development and the maintenance of normal tissue architecture (1). The classical (type I) subfamily of cadherins (N-, E-, P-, and R-cadherin) comprises an extracellular domain (EC) 1 which consists of five similar tandemly arranged domains, a single pass transmembrane domain, and a conserved cytoplasmic segment (1). Homophilic cell-cell adhesion is mediated by the cadherin extracellular domains (2)(3)(4). The conserved cytoplasmic domain interacts with intracellular catenins that anchor cadherins to the cytoskeleton (5)(6)(7)(8)(9).
Structural studies have revealed certain common features of classical cadherins that are likely responsible for adhesive engagement in the extracellular space (10 -17). The first model for N-cadherin activation derived from crystal structure analysis suggested that the first extracellular domain (EC1) forms cis-and trans-dimers and so contributes to N-cadherin homophilic associations (10,13). It has also been shown that with increasing calcium concentration, the N-terminal domain of E-cadherin first forms a "ring-like" tip to tip lateral cis-dimer with an adjacent partner. Two cis-domains then interact to form trans-dimer associations (12,14,15,18). Consistent with these observations, the crystal structure of a five domain Ccadherin (CCADEC1-5) showed that symmetric interactions between molecules were generated by tip-to-tip association, but also implicated EC2 in homophilic interactions (16). These data are supported by prior data obtained from antibody blocking (19,20), domain switching (4,21), and mutagenesis experiments (4,13,22). Together, the structural evidence strongly favors the conclusion that EC1 and EC2 of the classical cadherins are responsible for cis-and trans-interactions between identical cadherin molecules, and it is this interaction that yields homophilic adhesion between cells.
However, studies of C-cadherin interactions by surface force measurement and bead adhesion assays have yielded a different conclusion about the roles of the cadherin EC domains in generating adhesive activity, implicating multiple cadherin domains and a "ratchet" mechanism to effect adhesion, i.e. all domains function directly in generating cell adhesion (23)(24)(25)(26).
In order to examine the minimum requirement for EC domains in cell adhesion, to identify possible domain configurations that may enhance adhesive strength, and to distinguish between the functional contributions to adhesion from the extracellular and intracellular domains of N-cadherin, we generated a series of cell lines in which the serially arranged extracellular domains of N-cadherin were expressed with the wild type or with the myelin protein zero (Po) transmembrane and cytoplasmic domain. We chose Po to replace the native N-cadherin cytoplasmic domain, because Po is a transmembrane calcium-independent adhesion molecule with a single immunoglobulin-like extracellular domain and a cytoplasmic region with no catenin-binding sites (27). By substituting the Po cytoplasmic domain for the cadherin wild type domain, we effectively "neutralize" the contribution to cadherin cell adhesion mediated by an intracellular regulator domain.
Our findings demonstrate the following: 1) that the first two domains of N-cadherin (NEC1-2) form the minimal essential unit for classical cadherin to generate cell adhesion; 2) that although the entire extracellular domain of N-cadherin is required for the most efficient cell aggregation, EC length does not influence the adhesive strength generated between cells; and 3) that the NEC1-2 domains of N-cadherin play critical roles in cis-dimerization during cell adhesion.

MATERIALS AND METHODS
Antibodies-A polyclonal antibody against the cytoplasmic domain of N-cadherin was generated in rabbits. A rat monoclonal antibody against the EC domain of mouse N-cadherin (MNCD2) was kindly provided by Prof. M. Takeichi (see Ref. 28). Rabbit polyclonal or mouse monoclonal antibodies against GFP were purchased from Clontech. Rabbit polyclonal or mouse monoclonal antibodies against the Myc tag were purchased from Upstate Laboratory (Charlottesville, VA).
Constructs-An N-cadherin cDNA was as described previously (4). The complete rat Po coding region cDNA was provided by Dr. Mika Yoshida (29). Chimeric constructs encoding the various EC domains of N-cadherin were fused with six copies of the Myc epitope at the C terminus. An XhoI or AgeI site was created by PCR to substitute for the natural N-cadherin or Po stop codon. The XhoI site was used for ligation of the N-cadherin chimeric DNA to the 6ϫ Myc tag, which was isolated from the pCSϩ2 vector. The AgeI site was used for ligation of Po chimeric DNA with EGFP, which was subcloned from the pEGFP-N1 vector (Clontech). Finally, the chimeric cadherin DNAs were cloned into the pCXN2 vector. All PCR products and ligation sites were sequenced using the ABI computerized automated DNA sequencing.
Ncad-Fc Fusion Protein Generation-A construct of an N-cadherin Fc fusion protein (Ncad-Fc) was made as described previously (30). Briefly, a fragment from the 5Ј end to EcoRV site (residues 1-2126: AB008811 accession number in GenBank) was excised from the full-length Ncadherin cDNA and inserted into pRK5 fused with a cDNA encoding the human Fc region of IgG. The fragment of N-cadherin corresponds to the EC1-4 domains of the extracellular region, including its peptide sequence and cleavable prodomain region. The construct was cotransfected with a pSV2neo plasmid into HEK293 cells using Lipofectamine (Invitrogen), and stable transfectants were isolated by selection with G418 (0.8 mg/ml). Because Ncad-Fc does not have a transmembrane domain, the protein was secreted into the culture medium. Ncad-Fc was collected by immunoprecipitation from the culture supernatant with protein A beads (Zymed Laboratories Inc.). The positive clones expressing Ncad-Fc were selected by Western blot. The selected clone was plated in DMEM with 10% fetal calf serum and changed to DMEM with 2% fetal calf serum after confluence was reached. The culture supernatant was collected 1 week later and kept at 4°C. Purification of Ncad-Fc was carried out as described previously (31). The collected medium was subjected to ammonium sulfate precipitation and was applied to DE52 columns (Whatman) equilibrated in 17.5 mM NaH 2 PO 4 (pH 6.3). Bovine IgG contained in the culture medium does not bind to this column, but the Fc fusion protein bound well and was retained. After this procedure, Ncad-Fc protein was purified by using protein A columns. SDS-PAGE was carried out to estimate the purification. Purified protein was dialyzed against PBS with 1 mM CaCl 2 and stocked at Ϫ70°C until use.
Cell Culture and cDNA Transfection-L-cells were cultured in a humidified atmosphere of 10% CO 2 and in 90% DMEM containing 10% fetal bovine serum. To obtain single or double transfectants, the cells were transfected by using Superfect (Qiagen; 1 g of each cadherin in pCXN2 vector in single or double transfection), and then cultured in selective medium containing 800 g/ml of G418 (Invitrogen). Colonies were isolated and examined for N-cadherin-myc or Ncad-Po-GFP expression by immunocytochemistry and Western blot. Positive cells were expanded and used for further studies.
Immunocytochemistry-Cells were fixed in 4% paraformaldehyde, delipidated in 100% methanol, permeabilized with 0.1% Triton X-100, and blocked with 5% normal goat serum in PBS. After incubation for 1 h at 37°C with primary cadherin antibody, cells were treated (30 min) with fluorescent-conjugated secondary antibodies (Jackson Immuno-Research, West Grove, PA) and washed. Coverslips were then mounted and examined by confocal laser microscopy.
Aggregation Assays-Monolayer cultures were treated with 0.01% trypsin in HCMF (HEPES-buffered Ca 2ϩ ,Mg 2ϩ -free Hanks' solution) supplemented with 1 mM CaCl 2 for 30 min at 37°C. The trypsinized cells were washed gently in HCMF containing 1 mM calcium and 1% BSA at 4°C. This low trypsin procedure dissociates cell layers into single cells but leaves cadherins intact on the cell surface. After the cells were thoroughly dissociated, 5 ϫ 10 5 cells were transferred to 24-well dishes for a final volume of 0.5 ml of HCMF containing 1% BSA with or without 1 mM Ca 2ϩ . The plates were rotated at 80 rpm at 37°C for time course observation of aggregate formation. Cell aggregation was calculated by the index (N o Ϫ N t )/N o , where N t is the total particle number after incubation at a certain time and N o is the total particle number at the initiation of incubation (13,32).
For mixed aggregation analysis, cells expressing different cadherin constructs were labeled with lipophilic dyes prior to mixing. We used 1,1Ј-dioctadecyl-3,3,3Ј,3Ј-tetramethylindocarbocyanine (DiI) and 3,3dioctadecyloxacarbocyanine perchlorate (DiO) (Molecular Probes, Eugene, OR). A stock solution of DiI was made by dissolving 2.5 mg of DiI in 1 ml of 100% ethanol, and a stock of DiO was made by dissolving 2.5 mg of DiO in 1 ml of 90% ethanol and 10% dimethyl sulfoxide. These solutions were sonicated and filtered before use. To label cells with these dyes, incubation was carried out for 8 h at 37°C in serumcontaining DMEM at final concentrations of 15 and 10 g/ml for DiI and DiO, respectively. The cells were washed extensively with HCMF containing calcium. After single cell suspensions were obtained (as described above), 2.5 ϫ 10 5 cell/well of each of types were transferred to a 24-well dish. After rotating at 80 rpm at 37°C for 15-30 min, 50 l of the aggregates was removed, placed on a slide, and covered with a coverslip and examined by confocal microscopy.
Adhesion Strength Assays-To study adhesion strength in chimeric N-cadherin-transfected cell lines, plastic Lab-Tek chamber slides (Nalge Nunc International, Naperville, IL) were first coated with 10 g/ml protein A in a coating buffer (0.2 carbonate-bicarbonate (pH 9.4)), incubated at 4°C overnight, washed with HCMF several times, and blocked with 1% BSA/HCMF for 1 h room temperature. Ncad-Fc protein (10 g/ml in PBS) was coated on the immobilized protein A for 1 h at room temperature and then washed in HCMF four times. 5 ϫ 10 4 of suspended cells in 1% BSA/HCMF were loaded on to the coated area with 1 mM Ca 2ϩ and allowed to adhere to the substrate for 30 min. For adhesion strength assays on the cultured monolayer, the N-cadherin-GFP cell line was typsinized and plated on 18-mm coverslips (Fisher) (overnight). 5 ϫ 10 5 of the suspended cells were loaded on the coverslips and incubated (37°C for 30 min) with or without Ca 2ϩ (1 mM). The centrifugation experiments were modified from previous study (33). The Lab-Tek chamber slides coated with N-cad-Fc were placed vertically in 50-ml tubes. The 18-mm circular coverslips with monolayer N-cadherin-GFP growth were attached to the lids of the 50-ml tubes with Krazy Glue. All Falcon tubes contained 1 ml each of PBS before centrifugation and were centrifuged at 1,500 -4,500 ϫ g in a swinging rotor (Eppendorf, Centrifuge 5804R) for 10 min. After centrifugation, the cells retained on the coverslip, and those collected at the bottom of the Falcon tube were counted under a fluorescent microscope (Fig. 1).
Immunoblotting and Immunoprecipitation-For membrane fractionation, cell lines transfected with N-cadherin truncated constructs were harvested in ice-cold PBS and centrifuged at 1000 ϫ g, and the sediment was washed in PBS. After centrifuging again, sedimented cells were resuspended in hypotonic buffer (20 mM Tris-HCl (pH 7.4), 1 mM EDTA, 1 mM dithiothreitol, and protease inhibitors), placed on ice for 20 min, Dounce-homogenized (50 strokes), and placed on ice for 20 min. Cells were centrifuged at 1,000 ϫ g for 5 min, and the supernatant was collected and centrifuged again; these steps were repeated five times. The first pellet was taken as the nuclear fraction, and the last collected supernatant was centrifuged at 100,000 ϫ g for 60 min. This pellet was treated as the membrane fraction. For immunoblotting and immunoprecipitation experiments, 5 ϫ 10 6 cells were cultured in 100-mm tissue culture dishes at 37°C for 3 days. The confluent monolayer was then washed with PBS, scraped off the dish, pelleted for 5 min at 1000 rpm, and extracted in 1.0 ml of lysis buffer (150 mM NaCl, 20 mM Tris-HCl (pH 7.5), 1% Nonidet P-40, 1% Triton X-100, Ca 2ϩ free, leupeptin 5 g/ml, Pefabloc 0.4 mM, aprotinin 1 g/ml) for 30 min. After cells were lysed, the samples were centrifuged (20 min, 15,000 rpm), and the supernatants were separated for immunoblotting and immunoprecipitation. For immunoblotting, protein concentrations were determined (BCA protein assay, Pierce), and samples were run on 10% SDS-PAGE, transferred to nitrocellulose, blocked with 5% milk protein, and incubated overnight with primary antibodies. After secondary antibody incubation and routine washing, blots were developed with the ECL system (PerkinElmer Life Sciences). To detect immunoprecipitates, the supernatants were incubated with specific antibody (2 l per sample) for 1 h at 4°C. The antigen-IgG complexes were precipitated by sequential incubation with protein A-Sepharose 4B. The immunoprecipitates were washed extensively and boiled with SDS sample buffer, and the proteins were detected by immunoblotting with specific antibodies. Relative intensity of each Western band was measured and compared with control (wild type N-cadherin) by NIH image program.

RESULTS
Truncated N-cadherins Are Appropriately Expressed at the Cell Surface-We generated serial constructs of N-cadherin, which included different EC domains linked to the natural transmembrane and cytoplasmic domains (NEC). The signal peptide and prodomain of N-cadherin were preserved in the EC domain constructs to ensure proper expression of the constructs on the cell surface. In order to exclude possible homophilic adhesive interactions influenced by the intracellular domain of N-cadherin, segments of myelin protein zero (Po), including its transmembrane and cytoplasmic domains, were employed in our experiments (NECsPo Cyto ). The complete set of constructs is shown in Fig. 2A. These were transfected into L-cells to generate permanent cell lines. Chimeric proteins were detected by immunocytochemistry, which demonstrated that all constructs were expressed appropriately on the cell membrane. Western blots of total proteins were performed to confirm that the expressed protein sizes were as predicted (Fig.  2, B and C) and to compare levels of protein expression in cell lines prior to aggregation experiments. In order to analyze levels of truncated N-cadherin proteins on the cell surface, membrane fractions were prepared from these cell lines, and Western blots with Myc antibody revealed eventually the same levels of truncated N-cadherin are expressed on the cell surface (Fig. 2, D and E). The relative intensities as compared with N-cadherin wild type are 98% in NEC1, 105% in EC1-2, 92% in NEC1-3, and 107% in NEC1-4. For cell binding studies, an N-cadherin-Fc fragment (Ncad-Fc) of four EC domains (EC1-4) was generated and expressed as described (30, 34) (Fig. 2F).
The First Two Domains of N-cadherin Form the "Minimal Essential Unit" to Initiate Cell Adhesion-In cell aggregation assays, NEC1 cells as well as control untransfected L-cells were unable to form aggregates, strongly implying that the EC1 domain of N-cadherin is insufficient by itself to mediate adhesion (Fig. 3, A a-h, and B). NEC1-2 cells started forming cell aggregates at early time points, but the sizes of the aggregates remained virtually constant during adhesion assay (Fig.  3, A, i-l and B). In contrast, the cell aggregate sizes increased over time with NEC1-3, NEC1-4, or NcadWT cell lines (Fig. 3,  A, m-x, and B). Aggregate sizes for the NEC1-4 and NcadWT were ϳ5 times larger than the NEC1-2 cell line (Fig. 3, A and   B). Our results indicate that the first two domains of N-cadherin, rather than EC1-3 as reported previously (23)(24)(25)(26), together function as the minimal essential unit to generate cell-  cell adhesive activity, and additional extracellular domains promote rapidity and extent of adhesive cell-cell interactions, as revealed by aggregation times and cell number contained within aggregates.
The Catenin-binding Cytoplasmic Domain Is Not Required for N-cadherin to Generate Cell Adhesion-The notion has been presented that the catenin-binding cytoplasmic domain is essential for cadherin binding activity (5,32,35,36). However, it has also been shown that the cadherin extracellular domains can generate adhesive properties in the absence of catenin linkages (37)(38)(39). Here we replaced the native N-cadherin cytoplasmic domain with that of myelin Po to isolate the function of the extracellular domain in cell adhesion. The different NECPo Cyto cell lines were found to behave identically to their wild type cytoplasmic domain counterparts. The NEC1Po Cyto cell line also failed to form cell aggregates during adhesion assays (Fig. 4, A, a-h, and B), whereas NEC1-2Po and NEC1-4Po Cyto cell lines formed cell clusters of different sizes (Fig. 4, A, i-p, and B). Furthermore, NECPo Cyto molecules, in which the N-cadherin extracellular domains were linked to the transmembrane and cytoplasmic domains of Po, displayed adhesion properties similar to the cadherins. These cell lines showed calcium-dependent adhesion properties (Fig. 4, A  and B). Taken together, our findings reveal that the cadherin extracellular domain seems to be the sole contributor to the "purely" adhesive properties of the molecule, i.e. surface-to-surface adhesion is independent of the nature of the cytoplasmic domain. Because they link with cytoplasmic components, the catenins have important intracellular signaling roles but seem not to be necessary to allow the EC domains to generate intercellular adhesion.
We considered the possibility that a single EC1 domain might be able to enter into an adhesive bond with a wild type N-cadherin across an intercellular gap. However, mixed aggregation assays showed that neither NEC1 nor NEC1Po Cyto cells could form trans-interactions with cells expressing wild type N-cadherin (Fig. 5, a-c, and aЈ-cЈ). In contrast, NEC1-2 or NEC1-2Po Cyto cells interacted well with wild type N-cadherin cells (Fig. 5, d-f and dЈ-fЈ), consistent with the conclusion that the first two domains of N-cadherin form the minimal unit to trigger trans-interactions with full-length N-cadherin.

Strength of Cell Adhesion Seems Not to be Associated with Length of Extracellular Domain of N-cadherin-It has been
suggested that C-cadherin engages in homophilic adhesion through the interactions of its multiple EC domains, in which the overlapping of extracellular domains can generate weak (minimal overlap) or strong (five domain overlap) adhesive forces (23)(24)(25)(26)40). We wondered whether successively longer extracellular segments of N-cadherin might generate stronger intercellular bonds than NEC1-2 cadherin. Increasing centrifugation force was used to assess the strength of adhesion between EC domain transfected cells and a substrate-adherent monolayer of Ncad-GFP cells. The monolayer of Ncad-GFP cells was cultured on coverslips overnight, and NEC cells were treated with DiI (Fig. 6, red). After 30 min of incubation of dissociated NEC cell suspensions on the NcadWT-GFP monolayer, the coverslips were centrifuged under increasing g forces, and the numbers of detached cells after centrifugation were counted. The results showed that, as expected, NEC1 cells have null or minimal adherence to Ncad-GFP cells (Fig. 6, A,  a-d, and B). Most surprisingly, NEC1-2, NEC1-4, or NcadWT cells generated very similar and strong binding profiles to the monolayer (Fig. 6, A, e-p, and B), which could be abrogated by EDTA (2 mM) treatment (Fig. 6, A, q-t, and B). The relative strength of cell adhesion was assessed by increasing the force of centrifugation (from 1,500 to 4,500 ϫ g). No differences were detected between cell lines with longer extracellular domains, NEC1-4 or NcadWT cells, and cell lines with shorter extracellular domains, NEC1-2 cells (Fig. 6, A, e-p, and B).
Possibly, the uneven distribution of N-cadherin molecules on cell membranes could cause differential adhesion between cells. To get around this problem, we coated Ncad-Fc protein directly onto coverslips. In order to orient the Ncad-Fc protein on the coverslip, protein A (which binds Fc) was coated on the coverslips before binding recombinant Ncad-Fc protein. The NECs cells were dissociated by conventional methods, and cell suspensions were plated on Ncad-Fc-coated coverslips, incubated with or without calcium (1 mM) for 30 min. The results (Fig. 7, A, a-x, and B) showed that the NcadWT cell line generated relatively strong adhesion to Ncad-Fc (Fig. 7, A, q-t, and B) but was unable to generate adhesive activity with control human IgG Fc (Fig. 7, A, u-x, and B), demonstrating that adhesion between EC domains of N-cadherin and Ncad-Fc was via specific homophilic interactions. In contrast, NEC1 cells could not adhere to Ncad-Fc, displaying as low an affinity as untransfected L-cells for substrate-bound Ncad-Fc (Fig. 7, A,  a-,h and B). As expected, NEC1-2 and NEC1-4 cell lines generated strong adhesion to the Ncad-Fc substrate, virtually identical to NcadWT (Fig. 7, A, i-t, and B). A larger force of centrifugation (4,500 ϫ g) detached NEC1-2, NEC1-4, or NcadWT cells from the Ncad-Fc substrate, but even so more than 40% of the cells remained adherent (Fig. 7, A, i-t, and B). The numbers of cells detached from the Ncad-Fc substrate did not display significant differences among the three cell lines (Fig. 7B). Similar results were observed with respect to adhesive activity of NEC1Po Cyto , NEC1-2Po Cyto , or NEC1-4Po Cyto cell lines with the Ncad-Fc substrate (data not shown). These results indicate that adhesion strength between cell surfaces, or between cells and an adhesive matrix, does not depend on the number of extracellular domains of N-cadherin beyond EC1-2 or on the presence of the catenin-binding region of the wild type N-cadherin intracellular domain.
The First Two Domains of N-cadherin Are Sufficient to Form Cis-dimerization at Adhesion Junction-It has been demonstrated that lateral cis-dimerization is a prerequisite step for classical cadherins to generate homophilic adhesion activity (4 ,  10, 22, 39, 41, 42). An interesting question is whether the first two domains alone, which functionally mediate cell adhesion, also suffice to form cis-interactions with adjacent partner molecules. To investigate the potential for cis-dimerization between short NEC1-2 domains, several cotransfected cell lines were established. We found that the NEC proteins and NECPo Cyto displayed different distribution patterns on cell membranes. NECPo Cyto proteins (except NEC1Po Cyto ) preferred to concentrate at cell-cell junction sites; NEC proteins, however, were distributed more equally over the entire cell membrane (Figs. 5, a and d and aЈ and dЈ, and 8A, a-f).
In cell lines with coexpressing NEC1 and NEC1-2Po Cyto , NEC1-2Po Cyto protein was concentrated at the junction of adhesion but not the NEC1 protein, which was evenly distributed on cell surface membranes. Most surprisingly, when NEC1-2 and NEC1-2Po Cyto were cotransfected into L-cells, NEC1-2 protein was localized at the same sites as NEC1-2Po Cyto protein, suggesting that association between the two proteins might bring both together at the same cell surface sites. Cis-dimerization was detected by immunoprecipitation of the three cotransfection cell lines (NEC1 with NEC1Po Cyto , NEC1 with NEC1-2Po Cyto , and NEC1-2 with NEC1-2Po Cyto ). After dissociation of the cells with EDTA (2 mM), which releases "trans" adhesive bonds, cell lysates were incubated with either Myc or GFP antibodies. The proteins were detected by Western blot. NEC1-2-myc fusion protein coprecipitated with NEC1-2Po-GFP protein in the double transfected cell line (Fig. 8B, a and b), indicating that the proteins of  b and c, e and f,  and bЈ and cЈ, and eЈ and fЈ). NEC1 and NEC1Po Cyto cells treated with DiO (green) generate adhesion neither with itself nor with NcadWT cells incubated with DiI (red). In contrast, NEC1-2 and NEC1-2Po Cyto cells treated with DiO (green) can generate adhesion function either with itself or with NcadWT cells.
NEC1-2 and NEC1-2Po Cyto form cis interactions on the same cell membrane. In contrast, this interaction was not detected in the cell line with NEC1 and NEC1Po Cyto or the cell line with NEC1 and NEC1-2Po Cyto (Fig. 8B, a and b). Our result suggest that the first two EC domains are also essential for cis-dimer formation at a cell adhesion junction, although we do not know how this interaction takes place and which amino acids are involved in cis-dimer formation. DISCUSSION Cadherins seem to engage in multiple molecular interactions, depending on the experiments and methodologies used to assess adhesive performance (4, 10 -13, 15-17, 23-26). The evidence now supports the notion that these are versatile molecules, which utilize a repertoire of molecular interactions to produce and regulate cell adhesion. Our data support the idea that there is a minimal essential unit that plays the pivotal role in homophilic adhesion, comprising the first two domains of a classical cadherin (Fig. 9). Although it does not appear to exist by itself in nature, this smallest functional unit exerts adhesive interactions that are indistinguishable in terms of "strength" from the wild type N-cadherin. In our view, the other EC domains probably serve as accessory "spacers" to optimize the interactive distance from the cell surface and provide flexibility for molecular interactions.
Previous studies have given us a generally accepted view of cadherin homophilic interactions, in which the N-terminal domains are required and provide the essential region for the molecule to mediate homophilic adhesion (4, 10 -18). Domain switching studies have suggested that the EC1 domain of classical cadherins controls homophilic interactions between cadherin molecules (4,21). The structures of N-, E-, and C-cad- herin observed by crystallography and electron microscopy have shown end-and-end interactions that involve either EC1 or EC1-2 (4, 10 -18). Moreover, the recent interpretation of desmosomal cadherin interactions as revealed by electron microscopic tomography clearly reveals flexible tip-to-tip associations between the N termini of the molecules (17). Our findings are consonant with these studies, but in addition we have now demonstrated that the first two domains of N-cadherin are the minimal essential elements for the molecule to generate homophilic adhesion. This is in contrast to previous model, which identified EC1-3 as the minimal essential motif for adhesion (25). The "overlapping" model suggests that multiple extracel-lular domains of cadherin are involved in homophilic adhesion (23)(24)(25)(26). In a study of multiple domain associations, it has been concluded that the three N-terminal domains of C-cadherin form the smallest unit to effectively generate adhesion and that these three domains also generate similar adhesive strength as the five-domain C-cadherin (CEC1-5) (25). The difference between this previous work and our results may possibly be related to the fact that we tested cell adhesion in a cellular context, which provided a more physiological and natural situation for cell interaction. This allowed us to detect functional differences among cadherin domain-deleted cell lines. More subtle functions for the EC3-5 domains still need to be studied; Homophilic cadherin interactions at cell junctions are illustrated with respect to the model structure of C-cadherin (16). A, the first domain of N-cadherin cannot generate any adhesion function. B, the first two domains of N-cadherin form the minimal essential unit to generate cell adhesion, but only yield small aggregates in transfected cell lines. C and D, full-length of N-cadherin can mediate larger aggregates easily either with itself or with the minimal essential unit. The strength of adhesion between cell-cell in wild type cadherin is the same in comparison to that in a shorter essential EC domain cell line. however, these domains have been implicated in promoting cell migration, increasing neurite outgrowth, and cell adhesion (43)(44)(45).
The flexibility of the extracellular domains seems to be one of the most important properties for a cadherin to mediate cell adhesion (15,17). Desmosomal cadherins form a variety of shapes (W, S, and shapes) as the molecules interact, indicating distinct associations between flexible cadherin molecules in the intercellular space (17). Observation of E-cadherin by electron microscopy showed that high concentrations of calcium are needed to rigidify E-cadherin and orient it to engage in molecular adhesive interactions (14,15). It might be expected that the shorter domain of N-cadherin in our experiments might severely limit the degrees of freedom of the molecule, so that contacts between short molecules form less readily (Fig. 9). The reason that the EC1 domain alone is not sufficient to mediate cell-cell adhesion might be due to the lack of calcium-binding sites to stabilize protein conformation, forming cis-dimer associations, or both (Fig. 9). It is not the case that NEC1 is adhesive, but the cell surfaces cannot approximate close enough contact to engage these short molecules across the extracellular space, because NEC1 is not adhesive either with itself or with wild type N-cadherin molecules that presumably protrude at a greater distance from a cell surface. Furthermore, small adhesion molecules, like Po, can induce competent adhesive contacts across a very small intercellular gap (4.6 nm) (27,46).
If the minimal essential unit is functional, then what might be the functions for EC3-5, which are not by themselves adhesive? (27,46). Our data support the idea that EC3-5 domains serve to make more accessible the minimal essential adhesion unit in the extracellular milieu by distancing it further out from the cell surface, where it can more readily encounter a cognate partner molecule protruding from an apposing cell surface. This conclusion is supported by the finding that wild type cadherin is more effective at mediating the formation of larger cell aggregates than the minimal essential unit (Fig. 9). Taking our data together with previous studies (23)(24)(25)(26), it appears clear that the in vitro derived overlapping model for generation of variable adhesive force cannot readily account for cadherin behavior in intact cellular systems.
Although the cytoplasmic domain of classical cadherins has been reported to participate in producing and regulating cell adhesion through binding its intracellular partners, the catenins (32,47,48), our data reveal that cell adhesion generated by classical cadherins is unrelated to these cytoplasmic protein linkages. The notion is supported by the results of replacement of the cadherin cytoplasmic domain in our experiments. Also, several studies have implicated that the cytoplasmic domain is not necessary for homophilic cadherin interactions. T-cadherin, which is without a cytoplasmic domain and is anchored to the extracellular aspect of the cell membrane via a glycosylphosphatidylinositol link, generates cell adhesion (38). Liver-intestine cadherin (LI-cadherin), whose short cytoplasmic domain does not associate with catenins, also functionally mediates cell adhesion (38). Moreover, a tail-less E-cadherin in K563 cells was found to generate the same strong adhesion as wild type E-cadherin (49). Nevertheless, the catenins, in the natural situation, act as a bridge for cadherin, anchoring it to the cytoskeleton, and are involved in signaling processes for the production and recruiting of certain proteins to a cell junction (5,32,35,36).