INTRODUCTION

Xenopus XB/U-cadherin is a member of the classical cadherin family of calcium-dependent cell-cell adhesion molecules. As a maternal cadherin, XB/U-cadherin is inserted into all newly forming plasma membranes from maternal mRNA and protein stores (1). It contributes to interblastomere adhesion and, during gastrulation, promotes convergent-extension movements of mesodermal cells (2, 3). Dominant-negative expression of XB/U-cadherin deletion mutants in Xenopus embryos results in the malformation of neural tissues in the head region (3, 4). Sequence alignment and its tissue-specific expression pattern show that XB/U-cadherin is most closely related to mammalian P-cadherins (5).

To promote stable calcium-dependent cell adhesion, classical cadherins must be linked to cytoskeletal elements, putatively to actin microfilaments (6-8). Cytosolic catenins associate with the cadherin cytoplasmic domain, which shows the highest level of conservation on the amino acid level among classical cadherins of different types and species (9).

Previous studies showed that - and -catenin, the latter of which is most likely identical to plakoglobin, bind to overlapping domains of the cadherin tail directly (10-12). -Catenin interacts with sites shared by - and -catenin linking the - or -catenin-cadherin complex to the cytoskeleton (13, 14). Recently, other cytoplasmic proteins have been reported to interact with cadherin-catenin complexes; Knudsen et al. (15) have shown that -catenin can bind directly to -actinin, and p120^cas has been found to associate with E-cadherin and the c-src kinase (16-18). The role of these proteins in the cadherin-catenin complex is still unknown.

Using E-cadherin as a model molecule, a number of studies set out to identify the cadherin-binding site for - or -catenin. When E-cadherin mutants lacking more than 37 carboxyl-terminal amino acids were expressed in mouse fibroblasts, cadherin-catenin complex formation as well as E-cadherin-mediated adhesive function was completely abolished (19). However, although a larger portion of the E-cadherin cytoplasmic domain consisting of the carboxyl-terminal 72 amino acids was both essential and sufficient for its adhesive function and catenin binding, these terminal 37 amino acids were not (9, 20). More recently, the -catenin-binding domain of E-cadherin was assigned to a core region of 25 amino acids located 54 to 29 amino acids from the carboxyl terminus (21). At least partial phosphorylation of a cluster of eight serine residues within this binding site was found to be necessary for cadherin-catenin complex formation in transfected mouse fibroblasts (22).

In this study, we focus on the role of motifs within the 38 carboxyl-terminal amino acids in catenin complex formation of XB/U-cadherin and the induction of cell adhesion. Full-length XB/U-cadherin and four deletion mutants lacking different parts of the carboxyl-terminal 38 amino acids of XB/U-cadherin were stably expressed in murine L-TK fibroblasts, as the most established expression system for the analysis of cadherin molecules. Catenins do not associate with mutant XB/U-cadherins lacking the carboxyl-terminal 32 amino acids but bearing sequences corresponding to the -catenin-binding sites reported for murine E-cadherin (21, 22). Complex formation and adhesive function are restored when the terminal 32 amino acids are replaced by the 19 carboxyl-terminal residues alone.

However, mutant XB/U-cadherin molecules truncated by these carboxyl-terminal 19 amino acids do not induce cell adhesion, although they associate with - and -catenin at cell surfaces. Thus, cadherin-catenin complex formation itself is not sufficient to render transfectants of mutant cadherin adhesive. Nonadhesive complexes differ from functional ones in that they lack resistance to detergent. We conclude that the specific 19 amino acids of the XB/U-cadherin carboxyl terminus are required to link complexes to cytoskeletal elements. This stabilization is a prerequisite for cell-cell adhesion.

Here, we provide evidence that complex formation with catenins and cytoskeletal stabilization of complexes depend on different domains of the cadherin cytoplasmic tail. Our results complete earlier findings in that they reveal a novel aspect in the establishment of cadherin-mediated adhesion.

EXPERIMENTAL PROCEDURES

Construction of the expression vectors for the full-length cDNA of XB/U-cadherin and the cDNA of XB/U-cadherin truncated by 38 carboxyl-terminal amino acids (XBc38) have been described previously (23). Carboxyl-terminal deletion construct XBc19 was generated by SpeI-DraII cleavage, truncated XBc32 by EcoRI-TaqI cleavage of full-length XB/U-cadherin cDNA. In both cases, resulting fragments were religated in the presence of a NheI nonsense linker, causing frameshifts and producing stop codons downstream of the DraII and TaqI sites, respectively. The XBc32 construct encodes for a random sequence of 12 amino acids following the cadherin sequence. The internal deletion mutant XBi20 was generated by EspI-DraII restriction and religation of full-length XB/U-cadherin cDNA. Truncated forms of XB/U-cadherin were asymmetrically inserted into the polylinker of the mammalian expression vector pRc/CMV (Invitrogen). All constructs were controlled by DNA sequencing with a laser sequencer (Applied Biosystems, ABI 373 A).

L-TK cells (kindly provided by Dr. R. Kemler, Max Planck Institute, Freiburg, Germany) were grown, transfected, and selected as described previously (2, 23). Cadherin-positive cells were subcloned twice by limiting dilution. Several clonal cell lines expressing the full-length or the respective deleted forms, as well as pooled clones, were analyzed. Results of representative clonal cell lines are shown.

For metabolic radiolabeling of proteins, confluent monolayers of cells were preincubated with methionine-free Dulbecco's modified Eagle's medium (Life Technologies, Inc.) for 2 h. Fresh methionine-free medium supplemented with 50 µCi/ml [³⁵S]L-methionine (1100 Ci/mmol, Amersham Corp.) was added for 16 h before the cells were harvested.

Previously characterized monoclonal antibody (mAb)¹ 6D5 raised against U-cadherin (24) was a gift from Dr. P. Hausen (Max Planck Institute, Tübingen, Germany). Peptide-specific polyclonal antibodies against - and -catenin have been described elsewhere (25, 26) and were kindly provided by Dr. R. Kemler (Max Planck Institute, Freiburg, Germany). mAbs against -catenin and p120^cas were purchased from Transduction Laboratories.

Cells were grown to confluency on glass coverslips for at least 48 h before being fixed in 3% paraformaldehyde in PBS/Ca²⁺ (phosphate-buffered saline: 2.7 mM KCl, 137 mM NaCl, 0.15 mM KH₂PO₄, 0.7 mM NaH₂PO₄, and 0.75 mM CaCl₂). For detergent extraction, cells were incubated for 5 min at 4 °C in PBS/Ca²⁺ containing 0.5% Nonidet P-40 prior to fixation. Indirect immunostaining was carried out as described previously (23), except that all antibodies were diluted in PBS/Ca²⁺. Cy3-conjugated goat anti-mouse IgG (Dianova) were used as secondary antibodies. Cells were viewed with a laser scanning confocal microscope (Leitz).

Confluent monolayers of cells were rinsed three times with PBS/Ca²⁺ and harvested. Cells were lysed in PBS/Ca²⁺ supplemented with 1% each of Triton X-100 and Nonidet P-40, 2 mM each of aprotinin, leupeptin, and pepstatin, 1 mM N-ethylmaleimide, and 2 mM each of iodoacetamide and phenylmethylsulfonyl fluoride (all from Sigma). Protein concentrations were determined by Bradford protein quantification. SDS-PAGE, immunoblotting, and immunoprecipitation, as well as biotinylation and streptavidin-agarose precipitation of cell surface proteins, were carried out as described previously (23). Immunoblot signals were scanned and quantified.

Cell monolayers that had just reached confluency were used for preparing single cells. The cell aggregation assay was performed according to the method described elsewhere (27). 3-4 × 10⁵ cells were incubated at 37 °C with 1 mM CaCl₂ or with 1 mM EGTA on a rotating platform at 80 rpm. Aggregation was stopped after 90 min. Cells were fixed by the addition of an equal volume of 5% paraformaldehyde in PBS/Ca²⁺. Aggregates were observed by phase contrast microscopy.

RESULTS

Four deletion mutants of the amphibian XB/U-cadherin sequence were constructed that lack 38 (XBc38), 32 (XBc32), or 19 (XBc19) carboxyl-terminal amino acids or bear an internal deletion of 20 amino acids followed by 19 carboxyl-terminal amino acids of wild-type XB/U-cadherin (XBi20). In the construct XBc32, 12 random amino acids follow the truncated XB/U-cadherin-specific sequence. The carboxyl-terminal sequences of full-length XB/U-cadherin and the deletion mutants are shown in Fig. 1A.

To characterize the different mutants with respect to conserved sequence motifs, a sequence alignment of the last 52 carboxyl-terminal amino acids of XB/U-cadherin and murine E-cadherin was done in Fig. 1B. Although 11 exchanges were found in the region containing the -catenin-binding site, 5 of them were conservative. An accumulation of four exchanges follows directly downstream of the cluster of eight serine residues, which itself is conserved. Both proteins share a stretch of amino acids, "KLADMYGG," upstream of carboxyl-terminal glutamic or aspartic acid residues.

Cadherin protein levels did not differ by more than 18% among L-TK cell lines expressing the different cadherin constructs under the control of the human cytomegalovirus promoter (Fig. 2A) analyzed by immunoblotting using mAb 6D5 against XB/U-cadherin. Mutant forms of XB/U-cadherin showed reduced molecular weights as expected from the respective cDNA truncations. Because cadherin expression has been reported to induce up-regulation of catenins in L-TK fibroblasts (28), we assessed the steady-state levels of - and -catenin proteins in the different mutant XB/U-cadherin-expressing cell lines. 20 µg of total protein per lane were analyzed by immunoblotting and quantified. Steady-state levels of the XB/U-cadherin constructs, - and -catenin, were determined by analyzing 10 experiments. Fig. 2 shows one of these experiments as an example. Cadherin and catenin levels did not differ by more than 15% between experiments. Transfectants contained similar amounts of -catenin (Fig. 2B), although in cells expressing full-length XB/U-cadherin or XBi20, the amount of -catenin was 135 and 130% compared with the other transfectants, respectively. The -catenin content varied extensively among transfected cell lines (Fig. 2C). -Catenin protein was at the detection limit in vector-transfected controls and cells expressing XBc32 or XBc38 and did in no experiment exceed 10% of the -catenin level reached in cells bearing intact XB/U-cadherin. In XBi20 transfectants, we detected the largest amount of -catenin with an average of 133%, whereas cells expressing XBc19 exhibited only 60% both compared with the amount of -catenin with the full-length XB/U-cadherin transfectants. Other proteins we studied while we characterized our cell lines, such as integrins, fibronectin, and laminin, did not detectably alter their expression level upon transfection of XB/U-cadherin constructs (data not shown).

Protein expression of wild-type XB/U-cadherin, its deletion constructs, and catenins in cadherin-transfected L-TK cells.

Cadherin-catenin complex formation was studied by coprecipitation experiments. Cells were metabolically labeled with [³⁵S]methionine for 16 h and lysed. Transfected cadherins were immunoprecipitated from cell lysates using mAb 6D5. Immunoisolated proteins were separated by SDS-PAGE. Fluorography of all samples showed signals corresponding to the respective forms of XB-cadherin (Fig. 3A, uppermost bands). Additionally, two bands migrating at 102 and 88 kDa coprecipitated with XB/U-cadherin-specific antibodies in cells expressing the full-length XB/U-cadherin as well as in cells bearing XBc19 and XBi20. Those mutants comigrated with the 102-kDa band in a way that the two bands could not be discriminated. With XBc32 and XBc38, no signals other than those corresponding to the respective mutant forms of XB/U-cadherin were detected. When mAb 6D5 immunoprecipitates from cell extracts were immunoblotted with -catenin antibodies, the 102-kDa protein that coprecipitated with XB/U-cadherin, XBc19, and XBi20 (Fig. 3A) was stained (Fig. 3B). Similarly, the 88-kDa coprecipitating protein was identified as -catenin as determined by immunoblotting with -catenin antibodies (Fig. 3C). Neither - nor -catenin were found in a complex with XBc32 and XBc38 in these Western blots, as expected from their radioimmunoprecipitation profile (Fig. 3A). However, immunodetection of precipitated cadherins with mAb 6D5 showed that all mutants as well as wild-type cadherin were efficiently isolated from cell lysates (Fig. 3D). Thus, detection of - and -catenin as XB/U-cadherin coprecipitates did not fail in LXBc32 and LXBc38 lysates because of inefficient immunoisolation of XB/U-cadherin-containing protein complexes with mAb 6D5. In coimmunoprecipitation studies, we did not detect -catenin in a complex with any of the forms of XB/U-cadherin (data not shown).

Association of XB/U-cadherin and its deletion constructs with - and -catenin.

We also analyzed the p120^cas-binding properties to full-length and truncated forms of XB/U-cadherin. Interestingly, p120^cas was coimmunoprecipitated with wild-type as well as all four deletion mutants of XB/U-cadherin as determined by Western blotting of cadherin immunoprecipitates with antibodies against p120^cas (Fig. 4). Although the mAb used reportedly recognizes all four isoforms of p120^cas (16), only p120^cas 1A and B were found in the complex with XB/U-cadherin.

in cadherin-transfected L-TK cells.

The ability of transfected XB/U-cadherin constructs to mediate calcium-dependent cell-cell adhesion was tested using reaggregation assays. Cells were first dissociated by a 5-min trypsin treatment. Reaggregation in the presence of 1 mM Ca²⁺ resulted in the formation of stable cell-cell contacts within 90 min in L-TK cells expressing either full-length XB/U-cadherin or mutated XBi20 cDNAs (Fig. 5, A and E). Cell adhesion ability was shown to be calcium-dependent because no cell-cell aggregates were observed in the presence of 1 mM EGTA. An example of this is shown for the XB/U-cadherin-expressing cell line in Fig. 5B. In cells transfected with XBc19, XBc32, or XBc38, or with vector alone, no cell-cell contacts were established; only single cells were observed 90 min after reaggregation was initiated (Fig. 5, C, D, and F, and data not shown).

Effect of truncated XB/U-cadherin transfection on calciumdependent cell aggregation of L-TK cells.

Strikingly, the XB/U-cadherin mutant XBc19 formed complexes with both - and -catenin but did not mediate cell-cell adhesion of transfected L-TK cells. This observation differs from all earlier reports on mutant cadherins in which cadherin-catenin complex formation consistently correlated with cell-cell adhesion.

To further confirm our results on cadherin-catenin complex formation and make sure that complexes were localized to their normal site of function, we tested whether catenins were found associated with the respective XB/U-cadherin forms at plasma membranes. Cell surface molecules were biotinylated and isolated by streptavidin-agarose precipitation. The resulting streptavidin-agarose precipitated proteins and the non-streptavidin-bound cytosolic fraction were separated by SDS-PAGE. Distribution of transfected forms of XB/U-cadherin as well as - and -catenins was determined by Western blot analysis using antibodies to catenins or cadherin. Fig. 6A shows the subcellular localization of XB/U-cadherin and its deletion mutants. In all transfected cell lines, cadherin molecules were bound to streptavidin-agarose and were lacking in the soluble supernatant fractions. This provided biochemical evidence that forms of XB/U-cadherin were localized at plasma membranes of transfected cells. - and -catenin were detected in plasma membrane-associated fractions of the three cadherin-catenin complex-forming transfectants bearing full-length XB/U-cadherin, XBc19, or XBi20 (Fig. 6B, -catenin; Fig. 6C, -catenin). Because catenins do not directly bind to membranes (7), this proved that they were present in surface-associated protein complexes, presumably through interactions with the respective cadherin molecules (refer to Fig. 3). In cell lines that were transfected with vector alone or that expressed the XB/U-cadherin mutants XBc32 or XBc38, - and -catenin were detected exclusively in supernatant fractions of the biotinylated cells (Fig. 6, B and C). As expected, catenins remained cytosolic in cells bearing cadherin molecules deficient in cadherin-catenin complex formation. The fact that catenins do not appear in the streptavidin-bound fraction of complex-forming deficient constructs serves as an internal confirmation that the biotinylation was indeed restricted to cell surface molecules.

Cell surface localization of XB/U-cadherin, its deletion mutants, and bound - and -catenin in transfectants.

In agreement with the biochemical data shown in Fig. 6, immunofluorescence staining of nonpermeabilized transfected cells with XB/U-cadherin-specific mAb 6D5 showed that all forms of XB/U-cadherin were localized to the plasma membrane (Fig. 7). The complete XB/U-cadherin as well as XBi20 were predominantly detected at sites of direct cell-cell contact (Fig. 7, A and G). In contrast, XBc19, XBc32, and XBc38 were distributed over the entire cell surface in a punctate pattern (Fig. 7, C, E, and I, respectively). L-TK cells transfected with vector alone showed no staining with mAb 6D5 (data not shown).

Immunofluorescent staining of XB/U-cadherin and its deletion constructs in intact and in detergent-extracted transfected L-TK cells.

To mediate stable cell-cell adhesion, classical cadherins require linkage to cytoskeletal elements. Because this cytoskeletal stabilization renders cadherin proteins largely resistant to extraction with nonionic detergents (11), transfected L-TK cells were treated with 0.5% Nonidet P-40 prior to immunostaining with mAb 6D5. A significant amount of transfected XB/U-cadherin (Fig. 7B) as well as the internally deleted mutant XBi20 (Fig. 7H) was resistant to detergent extraction, as illustrated by the pronounced immunofluorescence in regions of cell-cell contact. On the other hand, the three carboxyl-terminal deletion mutants of XB/U-cadherin, XBc19, XBc32, and XBc38, were completely soluble in nonionic detergents. No immunofluorescence signal was obtained when cadherin staining was performed on these cells after Nonidet P-40 extraction (Fig. 7, D, F, and J).

DISCUSSION

Here we report that amphibian XB/U-cadherin associates cytoplasmically with endogenous - and -catenin and p120^cas when transfected into murine L-TK fibroblasts. This cadherin-catenin complex induces calcium-dependent cell-cell adhesion, confirming the evolutionary conservation of this adhesion mechanism. The expression of different deletion mutants of XB/U-cadherin shows that p120^cas binds upstream of the -catenin-binding site. Most interestingly, we observed that XB/U-cadherin detergent insolubility and its ability to mediate cell-cell adhesion requires the 19 most carboxyl-terminal amino acids of XB/U-cadherin, whereas cadherin-catenin complex formation does not. Thus, we provide evidence that two separate domains in the cadherin cytoplasmic tail mediate catenin complex formation and induction of adhesion, respectively.

-Catenin/plakoglobin, instead of -catenin, is able to bind to the cytoplasmic tail of E-cadherin (12). Here, XB/U-cadherin predominantly forms complexes with -catenin and -catenin, whereas a minor subpopulation of XB/U-cadherin--, -catenin complexes may exist in our L-TK transfectants. p120^cas coprecipitated with full-length or truncated XB/U-cadherin independently of their function in catenin binding or cell adhesion. This indicates that the p120^cas-binding epitope in XB/U-cadherin is located upstream of the 38 carboxyl-terminal amino acids and that the XB/U-cadherin site involved in p120^cas binding is not sterically altered by the carboxyl-terminal deletions. Because full-length XB/U-cadherin and the internal deletion mutant XBi20 did not differ from adhesion-deficient mutants XBc19, XBc32, or XBc38 with respect to their -catenin- or p120^cas-binding properties, it is not likely that these XB/U-cadherin-binding proteins contribute to the differences in formation and function between XB/U-cadherin-catenin complexes.

The most truncated of our deletion mutants, XBc38, lacks cadherin adhesive function as well as cadherin-catenin complex formation. The interpretation of earlier results from equivalent E-cadherin constructs had been that the catenin-binding domain of E-cadherin reached into the carboxyl-terminal 37 amino acids (9, 19). Our mutants XBc19 and XBc32 both contain the cluster of eight serines but differ in the amino acids following it (Fig. 1A). Because of these mutants, only XBc19 binds catenins, supporting evidence by Stappert and Kemler (22) that amino acids after the serine cluster are necessary for -catenin binding. Jou et al. (21) found that these residues were not required for -catenin binding. However, this group used the two-hybrid system, which may fail to represent in vivo binding. By attaching the 19 most carboxyl-terminal amino acids to the nonfunctional XBc38, we generated the construct XBi20. Although it lacks a part of the serine cluster as well as the following amino acids, XBi20 does form catenin complexes, suggesting that the 19 most carboxyl-terminal amino acids of XB-cadherin can substitute for the missing sequence. Taken together, our data suggest that the site of -catenin interaction is located upstream of the carboxyl-terminal 38 amino acids. Stable -catenin binding in vivo, however, requires the cadherin sequences following downstream. These can be substituted by the 19 most carboxyl-terminal amino acids in XBi20 but not by the random sequence in XBc32.

Most strikingly, we learned from our studies that catenin binding itself is not sufficient to render the cadherin-catenin complex adhesive. We could only restore complete cadherin activity by attachment of the 19 most carboxyl-terminal amino acids, which end in seven consecutive acidic residues (XBi20), to the nonfunctional XBc38. The addition of a hydrophobic sequence of equal length (XBc32) neither restored catenin binding nor cell-cell adhesion. Even the addition of the 19 amino acids that normally follow the catenin-binding site (XBc19) were not able to restore complete cadherin activity. This mutant the tail of which contains three acidic residues apart from each other forms cell surface-localized catenin complexes. However, XBc19 fails to mediate cell-cell adhesion. Our data imply that a stretch of acidic amino acids downstream of the catenin-binding site is required to confer adhesive function to the already formed cadherin-catenin complex. Importantly, this shows that the two functions of cadherins, recruitment of cytoplasmic catenins and adhesion, can be separated from each other.

Functional catenin-XB/U-cadherin and -XBi20 complexes cannot be completely solubilized by nonionic detergents, indicating that they are linked to the cytoskeleton (13). Conversely, XBc19 is extracted with mild detergents, probably because its complexes are not stabilized by components of the cytoskeleton.

Complex formation and cytoskeletal linkage seem to depend on the presence of numerous charged amino acids in the catenin-binding domain and in the cadherin tail. Negative charges in the catenin-binding region are provided by posttranslational phosphorylation of multiple serine residues, which has been shown to be essential for catenin binding (22). The binding domain of plakoglobin, a -catenin homologue, in the desmosomal cadherin desmoglein bears an accumulation of negatively charged amino acids as well (29, 30). More recently, a -catenin-binding site has been identified in the amino terminus of the transcription factor LEF-1. This domain also consists of many negatively charged amino acids (31, 32). -Catenin has been shown to bind to the product of the tumor suppressor gene apc in vivo (33-36). APC bears multiple consensus sequences for -catenin interaction, which lack any similarity with the cadherin cytoplasmic tail and are devoid of negatively charged residues (35). -Catenin also interacts with the APC--catenin complex, presumably in a manner similar to how it binds to the -catenin-cadherin complex. However, APC--catenin complexes are not induced to interact with the cytoskeleton when bound to -catenin. Thus, a protein complex including - and -catenin does not, by default, associate with cytoskeletal elements. Given that --catenin complexes serve functions in vivo as soluble complexes in the cytoplasm as well as membrane- and microfilament-associated, cytoskeletal binding of catenins must be specifically regulated. Our results suggest that this regulation is performed by cadherin cytoplasmic sequences distinct from the -catenin-binding site.