Contextual Binding of p120ctn to E-cadherin at the Basolateral Plasma Membrane in Polarized Epithelia*

E-cadherin-catenin complexes mediate cell-cell adhesion on the basolateral membrane of epithelial cells. The cytoplasmic tail of E-cadherin supports multiple protein interactions, including binding of β-catenin at the C terminus and of p120ctn to the juxtamembrane domain. The temporal assembly and polarized trafficking of the complex or its individual components to the basolateral membrane are not fully understood. In Madin-Darby canine kidney cells at steady state and after treatment with cycloheximide or temperature blocks, E-cadherin and β-catenin localized to the Golgi complex, but p120ctn was found only at the basolateral plasma membrane. We previously identified a dileucine sorting motif (Leu586-Leu587, termed S1) in the juxtamembrane domain of E-cadherin and now show that it is required to target full-length E-cadherin to the basolateral membrane. Removal of S1 resulted in missorting of E-cadherin mutants (EcadΔS1) to the apical membrane; β-catenin was simultaneously missorted and appeared at the apical membrane. p120ctn was not mistargeted with EcadΔS1, but could be recruited to the E-cadherin-catenin complex only at the basolateral membrane. These findings help define the temporal assembly and sorting of the E-cadherin-catenin complex and show that membrane recruitment of p120ctn in polarized cells is contextual and confined to the basolateral membrane.

functions in adhesion as a multiprotein complex. The extracellular domain of E-cadherin binds in a homophilic Ca 2ϩ -dependent fashion to juxtaposed E-cadherin molecules on adjacent cells. In addition, ␤-catenin or plakoglobin binds with high affinity to the C terminus of E-cadherin, and ␣-catenin binds indirectly via ␤-catenin to link the complex to the actin cytoskeleton (reviewed in Ref. 6). A number of other cytoplasmic proteins, including Hakai (7), G␣ 12 (8), and presenilin-1 (9), have also been shown to interact with E-cadherin. In addition to their roles in cell-cell adhesion, members of the catenin family are also involved in cell signaling and cell dynamics. Notably, ␤-catenin can alternately act as a transcriptional coactivator in the Wnt signaling pathway (reviewed in Ref. 10). p120 ctn is an Armadillo repeat protein originally identified as Src kinase substrate (11). p120 ctn can interact with a number of receptor protein-tyrosine phosphatases at the plasma membrane or translocate to the nucleus, suggesting a role in regulating cell-cell adhesion and gene expression (12)(13)(14)(15). Although a series of studies have elucidated roles for p120 ctn in stabilizing and regulating the adhesive properties of cadherins, it has not yet been established when and where p120 ctn and E-cadherin interact. The p120 ctn -binding site has been mapped to a core region of 15 amino acids in the juxtamembrane domain (JMD) 1 of the E-cadherin cytoplasmic tail (16).
Delivery of E-cadherin to the basolateral surface and its incorporation into adherens junctions are key events in cell polarization and tissue morphogenesis. E-cadherin itself, or as a presumptive complex, has to be sorted and targeted to the basolateral membrane in polarized cells. Previous studies have demonstrated that ␤-catenin binds with high affinity to Ecadherin early in the biosynthetic pathway (17) at a core distal region of 30 amino acids (18). Further evidence suggested that this cadherin-catenin pair first forms at the level of the endoplasmic reticulum (19). E-cadherin and ␤-catenin are therefore structurally and functionally linked from the outset. The temporal sequence for assembly of other components of the complex is currently less clear. Several reports show that ␣-catenin with its attendant actin does not interact until the complex is delivered to the plasma membrane (17). A recent study now reinvigorates the issue of cadherin-catenin complex assembly by suggesting that N-cadherin binds early to p120 ctn and at a later stage to ␤-catenin (20).
Recently, we described a dileucine targeting signal required for the sorting and polarized trafficking of E-cadherin chimeras to the basolateral surface in epithelial cells (21). A chimeric Tac/E-cadherin reporter molecule is sorted by utilizing the cytoplasmic tail of E-cadherin for basolateral trafficking. Mu-* This work was supported in part by grants from the National Health and Medical Research Council of Australia (to J. L. S., R. D. T., and A. S. Y.). The Institute for Molecular Bioscience receives infrastructural support as a Special Research Center for Structural and Applied Genomics of the Australian Research Council. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  tation of the so-called S1 motif in the tail results in missorting and delivery to the apical membrane (21). We have now gone on to explore the sorting function of the S1 motif in the context of full-length E-cadherin. Dileucine motifs are commonly found as basolateral or sometimes endocytic sorting signals (22). Mechanically, they bind to adaptor subunits such as ␤-adaptin at the level of the trans-Golgi network (TGN) to sort proteins during polarized trafficking (23). The close proximity of the S1 motif and the p120 ctn -binding site in the JMD of E-cadherin prompted us to jointly investigate sorting, polarized trafficking, and the temporal binding of p120 ctn based on the assumption that steric hindrance might possibly preclude simultaneous binding of a putative sorting adaptor and p120 ctn . The results presented herein suggest that the assembly and sorting of E-cadherin with catenins are indeed orchestrated in both time and space.

EXPERIMENTAL PROCEDURES
Cell Culture-Madin-Darby canine kidney (MDCK) cells (strain II) were grown and passaged as described previously (21) in Dulbecco's modified Eagle's medium (Invitrogen) with 10% fetal calf serum and 2 mM L-glutamine in 5% CO 2 and 95% air. For experiments, cells were plated at subconfluent density on glass coverslips or at confluent density on semipermeable polycarbonate filters (Transwell, Corning Costar, Cambridge, MA) and maintained for 1-3 days before being used. Chinese hamster ovary (CHO) cells were grown and passaged in Ham's F-12 medium with L-glutamine and 10% fetal calf serum in 5% CO 2 and 95% air. Cells were plated onto glass coverslips and dishes and maintained for 2-3 days before being used.
In some experiments, cells were incubated in medium containing 10 M cycloheximide (Sigma) to prevent protein synthesis and to deplete proteins from the biosynthetic pathway. Some cells were incubated in CO 2 -independent medium (Invitrogen) at either 18 or 20°C for 3 h prior to use. The 18°C temperature block accumulates proteins in early or recycling endosomes (24). The 20°C temperature block prevents exit of proteins from the TGN, causing accumulation of proteins within the Golgi complex (25). Antibodies-Mouse monoclonal antibody HECD1 (provided by Dr. M. Takeichi, Kyoto University) and a rabbit polyclonal antibody raised against the extracellular domain of human E-cadherin (provided by R. G. Ali, S. Verma, and A. S. Y., University of Queensland) were used to detect human E-cadherin. A mouse monoclonal antibody that recognizes dog E-cadherin (3B8, obtained from Dr. W. Gallin, University of Alberta) was used in experiments using MDCK cells. Other primary antibodies used were a mouse anti-human p120 ctn monoclonal antibody and an anti-p230 monoclonal antibody (Transduction Laboratories, Lexington, KY) and rabbit polyclonal antibodies raised against ␤-catenin (Sigma) and green fluorescent protein (GFP; Molecular Probes, Inc., Eugene, OR). Secondary antibodies used included Cy3-conjugated sheep anti-mouse IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA), Alexa 488-conjugated goat anti-rabbit IgG (Molecular Probes, Inc.), and horseradish peroxidase-conjugated sheep anti-mouse IgG and horseradish peroxidase-labeled goat anti-rabbit IgG (Amrad, Victoria, Australia). Alexa-488 and Texas Red-conjugated phalloidin were used to stain F-actin (Molecular Probes, Inc.).
Vector Transfection and Expression-MDCK or CHO cells were plated at subconfluent density 24 h before transfection. The Lipo-fectAMINE TM Plus system in combination with Opti-MEM (both from Invitrogen) was used for transfections according to the manufacturer's guidelines. Cells were typically used 24 -48 h later, in some cases after reseeding on filters at confluent density. Stably expressing cell lines were generated as described previously (21).
Indirect Immunofluorescence-Cells grown on filters or glass coverslips were fixed in 4% paraformaldehyde in phosphate-buffered saline for 60 min and then permeabilized in 0.1% Triton X-100 for 5 min and stained as described previously (21). Cells were viewed using an Olympus Provis AX-70 microscope, and images were captured on a CCD300ET-RCX camera (Dage-MTI, Inc., Michigan City, IN) using NIH Image software or a Bio-Rad Radiance 2000 confocal system (Bio-Rad, Hertfordshire, UK) mounted on a Nikon E600 microscope. All confocal hardware was driven by a workstation running LaserSharp 2000 imaging software (Bio-Rad). Images were collected using a ϫ60 objective (numerical aperture of 1.4), giving an optimal section depth of 0.6 m. Cy3-conjugated secondary antibodies were excited using a 543-nm helium/neon laser, whereas Alexa-488-conjugated secondary antibodies were excited with a 488-nm argon laser. Post-capture image analysis and processing of confocal image stacks were performed using the LaserVox software package (Bio-Rad).
Immunoprecipitation and Immunoblotting-Cells were solubilized in cold precipitation buffer (10 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, and 0.5% Triton X-100) containing Complete TM protease inhibitors (Roche Applied Science, Mannheim, Germany) on ice and extracted for 30 min at 4°C. Post-nuclear supernatants were obtained by centrifugation at 17,000 ϫ g for 10 min. For immunoprecipitation, extracts were precleared with protein G or protein A for 30 min, followed by incubation with anti-human E-cadherin antibody complexed to Ultralink-immobilized protein G (Pierce) or anti-GFP antibody complexed to protein A (Pierce) for 2 h. Precipitates were recovered by centrifugation and then washed several times with cold precipitation buffer and with 10 mM Tris-HCl (pH 7.4) prior to solubilization in concentrated SDS-PAGE sample buffer. Immunoprecipitates were separated by SDS-PAGE and analyzed by immunoblotting with Supersignal West Pico (Pierce) as described previously (21).

E-cadherin Complexes with ␤-Catenin and p120 ctn at the Lateral Plasma Membrane in MDCK Cells-E-cadherin,
␤-catenin, and p120 ctn are known members of cadherin-catenin complexes at adherens junctions in polarized epithelia (6). Confocal microscopy of MDCK cell monolayers confirmed the colocation of all three molecules on lateral cell membranes in these cells (Fig. 1A). Antibodies to E-cadherin co-immunoprecipitated both ␤-catenin and p120 ctn (Fig. 1B), confirming that, at steady state, all three proteins form a stable complex at adherens junctions.
⌭-cadherin-GFP and ␤-Catenin, but Not p120 ctn , Are Found on Intracellular Membranes-In MDCK cells, transiently expressed GFP-tagged E-cadherin (Ecad-GFP) (21) produced the same staining pattern as endogenous E-cadherin; it was found predominantly at the basolateral membrane, and newly synthesized Ecad-GFP was also seen in a perinuclear Golgi compartment ( Fig. 2A). Co-localization with the TGN marker p230 (29) confirmed the Golgi localization of the intracellular pool of Ecad-GFP. Both endogenous ␤-catenin and p120 ctn co-localized with Ecad-GFP at the basolateral cell surface ( Fig. 2A). Costaining of ␤-catenin with Ecad-GFP also occurred at the level of the Golgi complex, consistent with previous studies showing that ␤-catenin associates with E-cadherin early in the biosynthetic pathway (19). In contrast, we noted that p120 ctn staining was not seen in the region of the Golgi complex ( Fig. 2A).
To rule out the possibility that p120 ctn was not detected at intracellular sites because of epitope masking or any deficiencies of the immunolabeling process, we performed experiments on cells overexpressing p120GFP (28). In confluent cells, p120GFP localized to the lateral plasma membrane; it was not on intracellular membranes and appeared as faint diffuse staining only in the cytoplasm (Fig. 2B). Endogenous E-cadherin co-localized with p120GFP at the lateral membrane; however, there was no co-staining of p120GFP on intracellular membranes. Notably, p120GFP-transfected cells showed increased E-cadherin staining on lateral membranes, an observation consistent with the demonstration that increased p120 ctn levels lead to increased stabilization of E-cadherin at the cell surface (30). Thus, in MDCK cells at steady state, ␤-catenin and Ecad-GFP appear together in the Golgi complex, but p120 ctn is found with Ecad-GFP only at the plasma membrane.
To further investigate whether endogenous p120 ctn staining ever appears in the biosynthetic pathway, we treated cells with cycloheximide to deplete all newly synthesized proteins from the biosynthetic pathway and then analyzed staining patterns after washout of the drug, when newly synthesized proteins reappear inside cells (Fig. 3A). Immediately after incubation with cycloheximide, there was a significant decrease in Ecadherin plasma membrane staining and a complete disappearance of any intracellular staining; the cells also became flattened and more irregular in shape (Fig. 3A, 0hr). This corresponded to diminished p120 ctn plasma membrane staining and increased cytoplasmic staining. Two hours after washout of the cycloheximide, E-cadherin staining reappeared in the Golgi complex as determined by co-localization with ␤-COP (data not shown). By 4 h, there was prominent Golgi and vesicular staining, representing E-cadherin accumulation in secretory or endocytic compartments (Fig. 3A, 2hr and 4hr). Throughout this recovery, staining of E-cadherin on the plasma membrane also increased steadily, as did staining of p120 ctn , which increased on the plasma membrane while decreasing in the cytoplasm. At no time did p120 ctn staining appear on intracellular membranes other than on the plasma membrane. These results suggest that p120 ctn does not appear in biosynthetic compartments on its way to the cell surface, whereas E-cadherin complexes can be stained in transit through the Golgi complex and secretory pathway. The staining in these cells suggests that p120 ctn is translocated directly to the plasma membrane from the cytoplasm.
As another approach to investigate whether p120 ctn can be found on intracellular compartments, we used temperature blocks to accumulate E-cadherin in the TGN or in endosomes. Cells subjected to a 20°C temperature block for 3 h typically accumulate newly synthesized proteins in the TGN (25). In keeping with this, E-cadherin staining of the perinuclear Golgi region was increased in cells incubated at 20°C (Fig. 3B). Under the same conditions, there was no staining of p120 ctn in the perinuclear area. We have previously shown that recycling FIG. 1. E-cadherin complexes on the lateral plasma membrane in MDCK cells. A, confocal sections of MDCK cells were immunolabeled for E-cadherin (antibody 3B8), p120 ctn , or ␤-catenin (␤-cat) to localize endogenous proteins by immunofluorescence. The upper panels show XY cross-sections of monolayers, whereas the lower panels show XZ sections of filter-grown cells. All proteins were localized to the lateral membranes. B, extracts of MDCK cells were immunoprecipitated (IP) for E-cadherin (antibody 3B8) and immunoblotted (IB) for E-cadherin (antibody 3B8), p120 ctn , and ␤-catenin. Both p120 ctn (120 kDa) and ␤-catenin (94 kDa) coprecipitated with E-cadherin.
FIG. 2. Differential staining of ␤-catenin and p120 ctn with Ecad-GFP. A, shown are epifluorescence images of MDCK cells transiently transfected with Ecad-GFP and co-immunolabeled for p230, ␤-catenin (␤-cat), and p120 ctn . Intracellular Ecad-GFP co-localized with the TGN marker p230. Ecad-GFP co-localized with ␤-catenin both at intracellular sites and the plasma membrane. p120 ctn staining was present only on the basolateral plasma membrane. B, MDCK cells were transiently transfected with p120GFP and co-stained for endogenous E-cadherin (antibody 3B8). The merged image shows that p120 ctn colocalized with E-cadherin at the plasma membrane, but not at internal sites.
surface E-cadherin is accumulated in endosomes in cells treated with cycloheximide (to deplete biosynthetic proteins) and then incubated at 18°C for 3 h (31). This treatment resulted in enhanced vesicular endosomal staining of E-cadherin, but not of p120 ctn (Fig. 3B). This staining provides further evidence to suggest that p120 ctn does not co-accumulate with endogenous E-cadherin complexes in the biosynthetic pathway or in endosomes. In all of these experiments in polarized cells, the only membrane domain that p120 ctn localized to was the lateral plasma membrane, regardless of the presence of Ecadherin complexes on other membranes.
Cadherin Binding and Membrane Recruitment of p120 ctn -The association of the E-cadherin complex with p120 ctn was then investigated using transfected cells expressing Ecad-GFP and a targeting mutant of E-cadherin. We have previously identified the dileucine motif (S1) as a basolateral sorting signal required for targeting of an E-cadherin chimera, Tac/Ecad-(574 -724)⌬S1, consisting of the extracellular and transmembrane domains of Tac (interleukin-2 receptor ␣-subunit) fused to the cytoplasmic tail of E-cadherin (21). This S1 mutation (Leu 586 -Leu 587 mutated to alanines) was now introduced into full-length human E-cadherin cDNA, creating Ecad⌬S1, and into GFP-tagged E-cadherin, creating Ecad⌬S1-GFP (Fig. 4A), which were used to probe the sorting of E-cadherin and its association with catenins; this is also of interest since the S1 motif is closely adjacent (within 15 amino acids) to the p120 ctnbinding site in the JMD of E-cadherin (16).
CHO cells endogenously express p120 ctn , but not E-cadherin (32). Transient expression of Ecad-GFP or E-cadherin in CHO cells produced plasma membrane staining and resulted in the recruitment of p120 ctn and the expression and recruitment of ␤-catenin to the plasma membrane (Fig. 4B). Transient expression of ⌬S1 targeting mutants in the context of full-length E-cadherin (Ecad⌬S1-GFP) or a chimera (Tac/Ecad⌬S1) (data not shown) also successfully recruited p120 ctn and ␤-catenin to the plasma membrane, showing that both catenins can interact successfully with the targeting mutants. Immunoprecipitation was used to demonstrate directly that p120 ctn can bind to the tail of both wild-type Ecad-GFP and Ecad⌬S1-GFP (Fig. 4C). An anti-GFP antibody was used to immunoprecipitate protein complexes from extracts of cells transfected with GFP alone, with Ecad-GFP, or with Ecad⌬S1-GFP. p120 ctn coprecipitated efficiently with both human Ecad-GFP and Ecad⌬S1-GFP. Thus, mutation of the S1 motif does not affect the binding of p120 ctn to the E-cadherin tail, providing further definition of the E-cadherin p120 ctn -binding site and allowing for speculation about the nature of interactions between E-cadherin and sorting adaptors or p120 ctn .
Targeting of E-cadherin and ⌬S1 Mutants in Polarized Cells-To study a possible relationship between sorting and p120 ctn binding, we constructed and expressed full-length Ecadherin with the ⌬S1 mutation. Overexpressed E-cadherin is faithfully sorted and targeted to the basolateral membrane of polarized MDCK cells. This is shown here in cells overexpressing untagged human E-cadherin, which could be detected with a human-specific antibody (Fig. 5), but is also true of Ecad-GFP (21). In contrast, Ecad⌬S1-GFP or Ecad⌬S1 expressed in polarized MDCK cells was mistargeted and could be stained on both the apical and basolateral membrane domains (Fig. 5). Missorting of these ⌬S1 mutants is not a result of excessive overexpression, as wild-type E-cadherin expressed at similar or higher levels was not missorted. These results suggest that the S1 dileucine motif operates as a basolateral sorting determinant in full-length E-cadherin. E-cadherin molecules can dynamically form dimers (33), introducing the potential for heterodimerization of Ecad⌬S1 and endogenous E-cadherin during trafficking. To study this, MDCK cells transiently transfected with human Ecad⌬S1 were co-labeled with both human-specific and canine-specific antibodies to detect recombinant and endogenous E-cadherin, respectively (Fig. 6). Immunofluorescence labeling and confocal analysis showed there was no missorting of endogenous E-cadherin; however, trans-

FIG. 3. Staining of E-cadherin in the biosynthetic pathway:
lack of p120 ctn co-localization. A, MDCK cells were incubated with cycloheximide for 12 h to deplete E-cadherin from the biosynthetic pathway and the cell surface. Cycloheximide was washed out, and cells were returned to normal medium for various times before being fixed. Cells were then immunostained for E-cadherin (antibody 3B8) or p120 ctn . In cycloheximide-treated cells, staining of both proteins appeared only on the plasma membrane. At 2 and 4 h of recovery, Ecadherin staining reappeared inside cells as Golgi staining (arrowhead), and vesicular p120 ctn staining still appeared only on the plasma membrane. B, MDCK cells were incubated at 20°C to block exit from the TGN or at 18°C in the presence of cycloheximide (CHX) to accumulate proteins in endosomes. Cells were fixed and labeled for either E-cadherin (antibody 3B8) or p120 ctn . Under both conditions, E-cadherin accumulated in intracellular compartments, particularly the Golgi complex (arrows); but p120 ctn did not. p120 ctn remained as diffuse cytoplasmic or plasma membrane staining. fected cells did have reduced membrane staining of endogenous E-cadherin on the lateral membrane, suggesting that there is competition for membrane binding at this site or compensatory changes in the amount of endogenous E-cadherin synthesized in transfected cells. Wild-type E-cadherin was not missorted along with Ecad⌬S1, making it unlikely that the two molecules dimerize, suggesting that they are handled separately by the sorting machinery or that the S1 sorting signal is dominant.
Expression and Mistargeting of Ecad⌬S1 Change Epithelial Cell Morphology-Transfected MDCK cells were put under antibiotic selection to produce cell lines stably expressing Ecad⌬S1 or Ecad-GFP. By immunostaining, the resulting cell lines still showed heterogeneous expression of recombinant proteins. Groups of cells expressing detectable levels of Ecad⌬S1 or Ecad-GFP were stained with the anti-human Ecadherin antibody. Cells overexpressing wild-type Ecad-GFP grew as typical cobblestone monolayers, with individual cells maintaining their shape and polarized phenotype. Cells overexpressing Ecad⌬S1 showed a markedly altered morphology, taking on a spindle shape and extending cellular processes consistent with a loss of cell polarity and epithelial organization. In a double-blind analysis, cells overexpressing Ecad⌬S1 were recognized as having distinctly altered morphology (data not shown). This change in morphology is shown in Fig. 7b, where we noted that even cells adjacent to the highly expressing stained cells also exhibited altered morphology. Cell lines expressing Ecad⌬S1 also grew noticeably slower than the parent MDCK cells or Ecad-GFP-expressing cells. Expression of missorted Ecad⌬S1 on the apical cell surface disrupted cell polarity and epithelial morphology, showing that correct sorting of E-cadherin is essential for these processes.
␤-Catenin, but Not p120 ctn , Is Mistargeted to the Apical Membrane with Ecad⌬S1-MDCK cells transiently expressing human E-cadherin or Ecad⌬S1 were co-stained to detect ␤-catenin or p120 ctn . Introduction of Ecad⌬S1 resulted in missorting of ␤-catenin, which was then found by confocal imaging on the apical membrane, along with Ecad⌬S1 (Fig. 8A). The correct basolateral targeting of ␤-catenin, as part of a preformed cadherin-catenin complex, was also dependent on the S1 sorting signal. This provides a new line of evidence to suggest that

FIG. 4. E-cadherin induces membrane recruitment in CHO cells. A, shown is a diagrammatic representation of E-cadherin constructs.
Ecad is wild-type E-cadherin with the transmembrane (tm) indicated. Ecad⌬S1 has Leu 586 -Leu 587 mutagenized to alanines in the cytoplasmic tail. Ecad-GFP has GFP fused to its C terminus. Ecad⌬S1-GFP has the targeting mutation and the GFP tag. Extracts of CHO cells transiently transfected with cDNAs were probed by immunoblotting with anti-human E-cadherin antibody HECD1 to demonstrate expression of the appropriate recombinant proteins. B, shown are epifluorescence images of CHO cells transiently transfected with E-cadherin (panel a) or Ecad⌬S1 (panel b) and immunolabeled with HECD1 to detect recombinant proteins. Transfected cells show clear plasma membrane labeling. Untransfected cells (panels c and d) and cells transfected with either Ecad-GFP (panels e and f) or Ecad⌬S1-GFP (panels g and h) were immunolabeled for p120 ctn (panels c, e, and g) or ␤-catenin (panels d, f, and h). p120 ctn recruited to the plasma membrane was stained in transfected cells (panels e and g). ␤-Catenin expression was induced in transfected cells, and it was also stained on the plasma membranes (panels f and h). C, an anti-GFP antibody was used to immunoprecipitate recombinant proteins from extracts of cells transfected with GFP alone, with Ecad-GFP, and or with Ecad⌬S1-GFP. Cell extracts (E) and immunoprecipitates (IP) were immunoblotted for either GFP (to detect tagged E-cadherins) or p120 ctn . Note that only the high molecular mass region is shown. p120 ctn was coprecipitated with both Ecad-GFP and Ecad⌬S1-GFP.
␤-catenin interacts with E-cadherin prior to sorting, early in the biosynthetic pathway.
The same result was not found when cells were doublelabeled to localize Ecad⌬S1 and endogenous p120 ctn . p120 ctn remained localized at the basolateral membrane even in cells with mistargeted Ecad⌬S1 (Fig. 8B). p120 ctn never appeared at the apical membrane, despite the findings in Fig. 6, which confirmed that Ecad⌬S1 has the ability to interact with p120 ctn . We therefore conclude that p120 ctn targeting to the basolateral membrane is not dependent on the Ecad⌬S1 motif. p120 ctn arrives independently at the lateral cell surface, or it may bind to the basolaterally destined cadherin-catenin complex after sorting has occurred. Most importantly, the recruitment of p120 ctn to the plasma membrane in polarized cells can happen only in the context of the basolateral plasma membrane.

DISCUSSION
The cadherin-catenin complex is a large multiprotein unit that functions at adherens junctions to mediate cell-cell adhesion. The cytoplasmic tail of E-cadherin supports many protein interactions necessary for assembly of the complex, for its association with the actin cytoskeleton, and for modulating adhesive function and signaling. Just where and when individual components join the cadherin-catenin complex have profound implications for its function at adherens junctions and for its trafficking to and from the plasma membrane. In this study, we used several approaches to explore the association of ␤-catenin and p120 ctn with E-cadherin, and we have shown that they bind in a temporal fashion to E-cadherin, before and after delivery to the cell surface, respectively. Importantly, we found that the sorting and basolateral targeting of the E-cadherin-␤-catenin complex is mediated by the S1 motif of Ecadherin. S1-mediated sorting is necessary for (but precedes) the binding of p120 ctn , which does not occur until at or near the basolateral plasma membrane. Several pieces of evidence support this conclusion, including the lack of intracellular membrane localization of endogenous p120 ctn or p120GFP at steady state and then after manipulations to highlight or accumulate E-cadherin inside cells. Finally, the finding that it can bind to, but is not missorted by, Ecad⌬S1 shows that p120 ctn recruitment is an obligate process of the basolateral membrane environment.
The localization of endogenous E-cadherin and overexpressed untagged human E-cadherin or Ecad-GFP was consistent in all our experiments with synthesis and trafficking through the biosynthetic pathway and polarized delivery to the basolateral plasma membrane. At steady state, ␤-catenin   (a and b). Ecad⌬S1 co-localized with endogenous E-cadherin only at the basolateral membrane (c and d); no endogenous E-cadherin was missorted to the apical membrane with Ecad⌬S1 (see insets).  (HECD1; b). Only a few cells with high expression levels in each case were stained for recombinant proteins. Using actin staining as a guide, an outline around each cell was drawn by hand to emphasize cell shape and pattern throughout the monolayer. Cells expressing high levels of Ecad⌬S1 in b have altered morphology; this altered cell shape can also be seen immediately adjacent (asterisks).
staining coincided with that of cell-surface or newly synthesized E-cadherin or Ecad-GFP on the basolateral membrane and in the perinuclear Golgi compartment, suggesting that it, too, is jointly trafficked via the biosynthetic pathway. Our results concur with those derived from pulse-chase labeling and immunoprecipitation experiments in previous studies, which established that ␤-catenin binds to a C-terminal domain of E-cadherin soon after biosynthesis (17,19,34), forming the nexus of a complex that is trafficked to the cell surface. Early binding of ␤-catenin was deemed necessary for the efficient endoplasmic reticulum exit and processing of some overexpressed E-cadherin chimeras (19); although more recently, using different chimeras, we found that, although ␤-catenin may facilitate processing of E-cadherin, it is not essential for its trafficking (21). Expression of the E-cadherin targeting mutant Ecad⌬S1 and the conjoint mistargeting of ␤-catenin to the apical cell surface now further demonstrate that the two molecules are trafficked together. The joint targeting and efficient binding of ␤-catenin, shown by membrane recruitment and immunoprecipitations (data not shown), with Ecad⌬S1 or with Ecad⌬S1-GFP serve to verify that the cytoplasmic domains of both recombinant proteins are correctly folded. The polarized trafficking of ␤-catenin, a cytoplasmic protein, is thus wholly dependent on its association with E-cadherin, which provides the necessary sorting signal in the form of the S1 dileucine motif. This is a prime example of how a single sorting signal dictates the polarized trafficking of multiple complexed proteins, and it provides further insight into one of the many possible fates of ␤-catenin in epithelial cells. p120 ctn is another cytoplasmic catenin that binds to E-cadherin, but that appears to function as part of cadherin-catenin complexes only at the plasma membrane, rather than at intracellular sites. The reported roles of p120 ctn focus mainly on its positive or negative modulation of adhesive strength and on stabilization of surface E-cadherin (reviewed in Ref. 35). Indeed, our current evidence, together with previous studies, indicate that p120 ctn is only loosely associated with E-cadherin while it is at the cell surface and that it dissociates from the cadherin-catenin complex during trafficking to and from the plasma membrane. For instance, surface E-cadherin induced to internalize by Hakai binding and ubiquitination does so at the expense of p120 ctn binding (7). In our current study, although ␤-catenin staining was frequently seen in association with Ecadherin at the level of the Golgi complex, similar co-staining of p120 ctn was not found, prompting the further analysis of p120 ctn localization during experimental manipulations of the biosynthetic traffic. De novo synthesized E-cadherin was localized in the perinuclear Golgi complex and in vesicles in cells pretreated with cycloheximide, without any similar staining of p120 ctn . Under these conditions, p120 ctn clearly translocated between diffuse cytoplasmic staining and lateral plasma membrane staining, implying that it moves to the cell surface directly from the cytoplasm. p120 ctn could not be accumulated along with E-cadherin in the TGN at 20 or 18°C in endosomes, further suggesting that it is not associated with intracellular compartments.
E-cadherin expression was previously found to be necessary and sufficient for plasma membrane recruitment of p120 ctn (16). Expression of Ecad-GFP or Ecad⌬S1-GFP in CHO cells produced the same result, efficiently recruiting endogenous p120 ctn to the cell surface. However, in polarized MDCK cells, expression of E-cadherin randomly on the cell surface was not sufficient to recruit p120 ctn to the membrane. Instead, p120 ctn was recruited only by E-cadherin that was targeted to the lateral cell surface. In MDCK cells overexpressing Ecad⌬S1, p120 ctn was not missorted or recruited to the apical membrane, but bound only to the lateral cell membrane. This finding is clear in showing that the membrane association of p120 ctn is highly regulated and occurs only in the presence of E-cadherin within a defined polarized subdomain of the plasma membrane. One of two scenarios could produce this effect: first, a cofactor or modification of p120 ctn required for its recruitment may be present at the lateral membrane, but not at the apical membrane; or second, p120 ctn is actually tethered to or constrained at the lateral membrane. Further studies will now have to address these issues. Recently, members of the p120 ctn family of Armadillo proteins (other than p120 ctn itself) such as p0071 (plakophilin-4), ␦-catenin, and ARVFC have been shown to have PDZ-binding domains, which can mediate interaction with the LAP proteins such as ERBIN (36,37). ERBIN localizes to the lateral membrane in MDCK cells and may have a role in promoting the final assembly of cadherin-catenin complexes at FIG. 8. ␤-Catenin (but not p120 ctn ) is sorted to the apical membrane in Ecad⌬S1-expressing cells. Shown is the immunofluorescence labeling of MDCK cells transiently overexpressing Ecad⌬S1 (panels a-f) or E-cadherin (panels g-l). Recombinant proteins were colabeled with ␤-catenin (A) or p120 ctn (B). Panels show confocal XY sections taken at apical (AP), lateral (LT), and basal (BA) levels of monolayers grown on filters. There was basolateral membrane staining of all proteins. In cells overexpressing Ecad⌬S1, ␤-catenin staining was found on the apical membrane along with Ecad⌬S1 (A, panels a-d), but not in cells overexpressing E-cadherin (A, panels g-j). In Ecad⌬S1overexpressing cells and in E-cadherin-expressing cells, there was no apical staining of p120 ctn (B, a-d and g-j). this location (38). Finally, confining p120 ctn to the lateral domain in epithelial cells may be uniquely important for controlling its function in the context of polarized cells.
A recent study (20) on the complexing of catenins to Ncadherin found temporal relationships among N-cadherin, ␤-catenin, and p120 ctn quite different from those emerging from our current data and previous studies on E-cadherin. ␤-Catenin was found to bind later in the biosynthetic pathway after phosphorylation of pro-N-cadherin, whereas p120 ctn bound to the cadherin both before and after these modifications (20). These conflicting results might reflect differences in the formation of cadherin-catenin complexes stemming from the different cadherin molecules themselves or from their expression and function in epithelial versus non-epithelial cells.
The JMD of E-cadherin supports multiple functions associated with adhesive strength and cadherin clustering (39) and multiple protein interactions, including those of p120 ctn (16), Hakai (7), and presenilin-1 (9). In addition, by showing that the S1 motif functions as a sorting signal, the binding of a sorting adaptor to the JMD of E-cadherin is also invoked. The interplay between sorting adaptor and p120 ctn binding is of special interest due to the close proximity of both binding sites. Our results show that mutation of Leu 586 -Leu 587 at S1 does not disrupt binding of p120 ctn to the JMD at residues 600 -616, helping to further define the structural requirements for p120 ctn binding. Similarly, another mutation in this region was also shown to have no effect on p120 ctn binding (16). We were also able to show that sorting occurs prior to p120 ctn binding, suggesting that there may be sequential and dynamic interactions: first of adaptor complexes, presumably at the level of the TGN, followed by p120 ctn at the plasma membrane. E-cadherin undergoes three major sorting events in its trafficking life. First, in the biosynthetic pathway, E-cadherin is sorted in the TGN for transport directly to the basolateral membrane (40). Second, from there it can be endocytosed with several possible fates, one of which is a third sorting event for recycling back to the plasma membrane (31,41). Confocal analysis over a time course beginning soon after transfection showed that newly synthesized Ecad⌬S1 is transported directly, separately, and simultaneously from the TGN to the apical membrane and to the basolateral membrane (data not shown); and thus, unlike some other membrane proteins in MDCK cells (e.g. the polymeric IgA receptor) (42), there is no transcytosis of newly synthesized Ecad-GFP or of Ecad⌬S1. Thus, S1-dependent sorting for basolateral trafficking occurs in the TGN, which is now further supported by the lack of p120 ctn missorting. To date, dileucine-based signals have been found to bind to ␥-adaptin (43); 1, 2, and 3 adapter subunits (44); and to ␤-adaptin (23). The sorting adaptor responsible for binding to S1 in the TGN has yet to be identified.
As expected, mutation of S1 in the context of full-length E-cadherin resulted in missorting to the apical membrane; and in addition, there was also basal and lateral localization of Ecad⌬S1. This is not a surprising result, as other dileucine-dependent basolateral proteins are similarly distributed when their dileucine signals are removed. For example, mutation of the dileucine sorting signals in both the basal cell adhesion molecule and nucleotide pyrophosphatase-1 results in 70% apical and 30% basolateral distributions (45,46). This non-polarized trafficking could be a result of overloading and spillover from the apical into the basolateral pathways, piggybacking of mutant proteins to other basolateral proteins, the presence of other basolateral signals, or translocation from apical to basolateral membranes. Our data generally discount most of these possibilities, leaving general, bulk, default, or unsorted membrane flow to all areas of the cell as the likely explanation for the presence of Ecad⌬S1 in both membrane domains. The missorting of Ecad⌬S1 had striking effects on the morphology and growth of epithelial cells, more so in fact than equivalent overexpression of wild-type Ecad-GFP. We would predict that Ecad⌬S1 is adhesion-competent, to some degree, when expressed on the apical membrane; and its participation in aberrant homophilic cadherin interactions or in cadherin-based cell motility could account for the loss of cell polarity and epithelial organization. These morphological changes highlight the importance of S1 function in the patent polarized targeting of E-cadherin and suggest that S1-dependent sorting may be particularly critical for epithelial patterning and morphogenesis during development. Our data now show another important consequence of polarity and that is the correct and complete assembly of the full E-cadherin-catenin complex including p120 ctn .