Spontaneous assembly and active disassembly balance adherens junction homeostasis

Abstract

The homeostasis of adherens junctions was studied using E-cadherin and its two mutants tagged by the photoconvertible protein Dendra2 in epithelial A-431 cells and in CHO cells lacking endogenous cadherin. The first mutant contained point mutations of two elements, Lys738 and the dileucine motif that suppressed cadherin endocytosis. The second mutant contained, in addition, an extensive truncation that uncoupled the mutant from β-catenin and p120. Surprisingly, the intact cadherin and its truncated mutant were recruited into the junctions with identical kinetics. The full-size cadherin was actively removed from the junctions by a process that was unaffected by the inactivation of its endocytic elements. The cadherin’s apparent half-residence time in the junction was about 2 min. Cadherin clusters made of the truncated mutant exhibited much slower but ATP-independent junctional turnover. Taken together, our experiments showed that adherens junction homeostasis consists of three distinctive steps: cadherin spontaneous recruitment, its lateral catenin-dependent association, and its active release from the resulting clusters. The latter process, whose mechanism is not clear, may play an important role in various kinds of normal and abnormal morphogenesis.

Adherens junctions are the most prominent intercellular adhesion structures in virtually all vertebrate cells (1, 2). They form via the clustering of their key transmembrane receptor, cadherin. Similar to all other polymeric cellular structures, cadherin clusters are believed to be stabilized by several types of interactions. According to the current general model, trans- (or adhesive) intercadherin interactions establish intercellular contacts, whereas cis- (or lateral) interactions reinforce the entire adhesive structure (3, 4). Cadherin clusters, through intracellular cadherin regions, interact with catenins that, in turn, couple cadherin to the cytoskeleton (3, 5). These intracellular interactions apparently regulate major cell-type-specific junctional features, such as their morphology, dynamics, and lifetime (6). The detailed structural organization of cadherin in adherens junctions is still not completely understood.

Even less is known about the assembly and disassembly of adherens junctions. Some experiments have suggested that catenins and their interactions with the cytoskeleton are required for cadherin to be clustered into adherens junctions. For example, deletions of catenin-binding sites, α-catenin depletion, or abnormalities in the actin cytoskeleton abolish adherens junction assembly (7 –11). The very low affinity of cadherin adhesive interactions also suggests that other cellular proteins facilitate efficient cadherin clustering (3). However, recent data (12 –14) are also consistent with an alternative point of view, which holds that cadherin clustering can proceed via a diffusion-trap mechanism. By this mechanism, cadherin molecules first form a small number of adhesion bonds that maintain the resulting nascent intercellular contact long enough for new cadherin molecules to be diffused into this contact.

Our previous biochemical experiments suggested that cadherin adhesive dimers are continuously assembled and disassembled. Of these two processes, cadherin dimer disassembly is ATP dependent (15). These experiments could not, however, distinguish between two possibilities: each junction may have a constant population of cadherin molecules that are in equilibrium between the monomeric and dimeric states, or the adherens junction could be a continually renewing structure. Published FRAP experiments performed with E-cadherin–EGFP fusion proteins have supported both possibilities. In some cases adherens junctions completely exchanged their cadherin molecules in several minutes (16). In other experiments adherens junctions exchanged their cadherin more slowly (9) or contained a large immobile fraction (17, 18). The fast exchange of cadherin in adherens junctions in Madin-Darby canine kidney (MDCK) cells was blocked by dynamin 2 inhibitors, suggesting that this process is driven by endocytosis (16). In MCF cells, however, the same inhibitors were ineffective, indicating alternative mechanisms. Such notable variations suggest that the homeostasis of adherens junctions can be regulated by very different mechanisms. In the present work, we investigate this important issue in living cells that express photoconvertible versions of cadherin and its mutants. Our experiments convincingly show that adherens junctions continuously gain and lose cadherin molecules. Cadherin enters the junctions with catenin-independent kinetics. Binding to catenins is required in the following step, where cadherin dimer integrates with the adhesive cluster. Such continuous cadherin recruitment into the junctions is counterbalanced by the active removal of cadherin from these structures.

Results

Adherens Junctions Are Rapidly Renewing Structures.

To investigate adherens junction dynamics, we expressed human E-cadherin tagged by EGFP (Ec-EGFP) in A-431 cells. The advantage of these cells is that their adherens junctions are very well-defined structures, ≈300 nm in diameter. In agreement with recently published data (19), they were stable, but mobile structures. They continuously formed at the basal edge of the cell–cell contact region and moved apically (Fig. 1A and Movie S1). Interestingly, during such movement, the fluorescence intensity of an individual adherens junction remained stable (Fig. 1B). Such stability suggests that a specific mechanism controls the amounts of cadherin in adherens junctions. It is possible that once assembled, adherens junctions are unable to recruit new cadherin molecules. Alternatively, adherens junctions can exhibit a balanced turnover of their cadherin. Continuous production of new adherens junctions, however, presents clear difficulties for studying this issue by using a conventional FRAP assay: one cannot be certain whether the recovered fluorescence is caused by a recruitment of new cadherin molecules into the photobleached junction or by the assembly of a new adherens junction in the photobleached area.

Fig. 1.

(A) Time-lapse images of cell–cell contact between two Ec-EGFP-expressing A-431 cells acquired at 2-min intervals (Left Column, only frames taken at 0, 8, and 16 min are shown, see Movie S1). The contact exhibits numerous apically moving adherens junctions, one of which is marked by an arrow. Schematic representation of this contact is shown at Right. Black arrows indicate the apical and basal regions of the contact. Trajectories of four adherens junctions are shown. The numbers indicate time intervals at which corresponding junctions can be traced in Movie S1. The red trajectory represents the arrow-marked junction. (Scale bar, 10 μm.) (B) Fluorescence intensities (in arbitrary units) of the junctions colored in A over time. Note that they fluctuate insignificantly. (C) Time-lapse analysis (selected frames from Movie S2) of adherens junctions between two A-431 cells expressing Ec-Dendra. At each time point, cells were imaged in green and red channels. The green channel shows normal Dendra2 fluorescence. The red channel reveals a photoconverted Dendra2 form. Frame 0a shows the cells before photoactivation. Note that the red fluorescence is undetectable. Frame 0b was taken immediately after the photoactivation. Frame 3 is 3 min later. (D) The graph shows changes in intensity of the red fluorescence in Ec-Dendra adherens junctions after Dendra2 activation. The black lines show values for four individual junctions, one of which is marked by an arrow in C. The red line is an average of four independent experiments (n = 30). The initial red fluorescence of the junctions is considered 1.0. The same experiment was also done with ATP-depleted cells (No ATP, blue line). The error bars represent SD (n = 20). (E) Redistribution of the cadherin molecules from the middle of the cells to the adherens junctions, frames from Movie S3: (0) immediately after the activation and (16) 16 min later. (F) Time-lapse images (Movie S5) of Ec-Dendra-expressing cell before (0) and 40 min after ATP depletion (40). The intensity of cadherin fluorescence in individual junctions (one of them is marked by an arrow) is increased. (G) The quantification of individual junction fluorescence (relative to the initial fluorescence, which is considered 1.0) before (control) and 20 min after the addition of ATP depletion media (no ATP). The average values of eight junctions from two independent experiments are shown.

To overcome this problem, we replaced EGFP with the photoconvertible protein Dendra2 in our E-cadherin construct resulting in the protein Ec-Dendra. Green-to-red photoactivation of Ec-Dendra in the small pool of adherens junctions in A-431 cells allowed us to trace the fate of the junctional cadherin and the rate of its exchange for a nonactivated green form (Fig. 1C and Movie S2). These experiments showed that the individual adherens junction continuously replaces its cadherin molecules with an apparent half-residence time of about 2 min (Fig. 1D). Importantly, adherens junctions of two independent A-431 subclones, as well as basally and apically located junctions, had only minor variations in their exchange rates. In complementary experiments we tested whether extrajunctional cadherin can be recruited into the preassembled adherens junctions. Figure 1E (Movie S3) shows that cadherin activated in the middle of the cell was uniformly recruited into the preassembled adherens junctions in the course of a few minutes.

To exclude a possibility that such fast cadherin exchange is a specific feature of the Ec-Dendra protein, we corroborated our data using A-431 cells expressing Dendra-tagged β-catenin (Fig. S1). Taken together, these experiments showed that adherens junctions in A-431 cells are not static structures; they continuously lose and recruit cadherin molecules.

ATP Depletion Induced Rapid Recruitment of Cadherin into the Junctions.

The fast turnover of cadherin in adherens junctions can be an active process or based on a weak cadherin–cadherin binding affinity. To confirm one of these possibilities, we blocked active cellular processes in Ec-Dendra-expressing cells using the combination of the oxidative phosphorylation and glycolysis inhibitors (NaN₃ or antimycin and 2-deoxyglucose, correspondingly). Such treatments were shown to rapidly deplete cells for ATP (13). Junctional cadherin turnover would be ATP independent if it is based on the weak affinity of cadherin–cadherin interactions. We found, however, that ATP depletion dramatically decreased the turnover of cadherin in adherens junctions (Fig. 1D and Movie S4). Furthermore, application of these inhibitors resulted in a rapid rise in the junctional fluorescence intensity (Fig. 1 F and G and Movie S5). Taken together, ATP depletion experiments showed that active processes are needed to counterbalance cadherin recruitment into the junctions.

Cadherin Endocytosis Has No Role in Cadherin Turnover in Adherens Junctions.

Pharmacological or siRNA-mediated inactivation of clathrin-dependent endocytosis was shown to inhibit the dissociation of adherens junctions (15, 16, 20). However, because a complete arrest of endocytosis can lead to numerous side effects, we sought a way to specifically block the endocytosis of E-cadherin by point-inactivating its endocytic motifs. To this end, we first mapped the motifs that are responsible for the very active endocytosis of the tailless cadherin mutant Ec-Δ748-Myc (13). We had proposed that the vigorous endocytosis of this mutant is driven by cadherin endocytic motifs that cannot be downregulated in the mutant because it lacks both β-catenin and p120-binding sites (Fig. 2A). Inspection of the residual 17-amino-acid-long tail of the Ec-Δ748-Myc mutant showed that it does contain two potentially important endocytic elements (Fig. 2A). The first one is Lys738, which is conserved in all classic and type II cadherins. In many transmembrane proteins, lysine ubiquitination triggers protein endocytosis. The second element is a dileucine motif that had been shown to mediate clathrin-dependent endocytosis of many proteins including E-cadherin (21). To test the involvement of these two elements in the Ec-Δ748-Myc endocytosis, they both were point-inactivated (Fig. 2A). Experiments with the resulting Ec-Δ748-KL-Myc mutant showed that this double mutation nearly completely inactivated its endocytosis (Fig. 2B). Furthermore, in contrast to the parental Ec-Δ748-Myc mutant, only a negligible amount of which was present in the adherens junctions (Fig. 2D), the mutant Ec-Δ748-KL-Myc was efficiently recruited into the junctions (Fig. 2E). Double staining for the mutant and endogenous components of cadherin adhesion revealed that these proteins were highly colocalized (Fig. 2 D′ and E′). Importantly, both Lys738 and the dileucine motif must be inactivated for such strong reduction of the Ec-Δ748-Myc endocytosis. The individual inactivation of only one of those elements produced much weaker effects.

Fig. 2.

(A) Schematic representation of E-cadherin and its mutants: the extracellular cadherin-like repeats (1–5), the transmembrane domain (TM), and the p120- and β-catenin-binding domains (p and cat, respectively) are shown. The solid square is the myc or the Dendra2 tags. The intracellular portion of the mutant Ec-Δ748 consists of a short, 17-amino-acid-long fragment that is located between the transmembrane and the p120-binding domains in the intact E-cadherin. Its amino acid sequence (lane Ec) is aligned with the homologous sequences of selected classic (N-cadherin, Nc) and type II (VE-cadherin, Vc; cadherin 11, C11) cadherins. Conserved residues are capitalized. Note that they all share two elements, a conserved Lys residue (K738) and, with the exception of VE-cadherin, a dileucine motif (LL motif). The line EcKL shows the KL mutation that is incorporated into the cadherin mutants Ec-KL-Dendra, Ec-Δ748-KL-Dendra, and Ec-Δ748-KL-Myc. (B) A-431 cells expressing Ec-Δ748-Myc (Δ748) and Ec-Δ748-KL-Myc (Δ748KL) mutants were surface biotinylated and then chased in regular media for 0, 15, or 30 min. The remaining surface biotin was stripped from the surface. Internalized biotinylated proteins were recovered using streptavidin-agarose and analyzed by immunoblotting using anti-myc. To approximate the size of the internalized cadherin pool, total biotinylated proteins from the control plates were precipitated by streptavidin-agarose and the same volumes of the resulting precipitates were loaded (T). Note that the KL mutation blocks the mutant internalization. (C) The same experiment as in B with A-431 cells expressing Ec-Dendra and Ec-KL-Dendra proteins. The blots were stained with anti-cadherin antibody recognizing both the endogenous (Ec) and the recombinant (Ec-D or Ec-KLD) cadherins. The control lane T was loaded with 25% of the precipitate. Note that whereas endogenous cadherin endocytosed identically in both cell clones, the Ec-KL-Dendra mutant had a much lower rate of endocytosis. (D and D′) Double immunofluorescence microscopy of Ec-Δ748-Myc-expressing A-431 cells. The cells were stained with rabbit anti-myc (D, myc) and mouse anti-β-catenin (D′, βCat) antibodies. Only negligible amounts of the mutant were present in the β-catenin-positive adherens junctions. (E and E′) The same experiment as in D with Ec-Δ748-KL-Myc-expressing cells shows the efficient recruitment of the mutant into the endogenous adherens junctions. (F–H) Anti-Dendra staining of A-431 cells expressing Ec-Dendra (F); a low level of Ec-KL-Dendra (G) and a high level of Ec-KL-Dendra, clone EcKLD2 (H).

Next, we inactivated the same two elements in the Ec-Dendra protein and selected A-431 subclones expressing the resulting Ec-KL-Dendra mutant at the same level as the Ec-Dendra in the subclones described above (Fig. 3A). Cell-surface biotinylation showed that the inactivation of these two elements significantly reduced cadherin endocytosis (Fig. 2C). However, despite these differences, the adherens junctions of the Ec-KL-Dendra-expressing cells exhibit no serious morphological defects (Fig. 2 F and G). Furthermore, the inactivation of these two elements did not affect the ATP-dependent junctional turnover of cadherin (Fig. 3E). Taken together, these data show that cadherin endocytosis is not the sole process that mediates the active removal of cadherin from adherens junctions.

Fig. 3.

(A) The equal amounts of cell lysates of WT A-431 cells (A431), and A-431 cell subclones expressing Ec-Dendra (EcD) or Ec-KL-Dendra (EcKLD) were stained for E-cadherin. Arrows indicate endogenous cadherin (Ec) and the recombinant Dendra-tagged cadherins (EcD). Note that the levels of the recombinant cadherins are the same. (B) Total lysates of cells expressing a low level of Ec-KL-Dendra (EcKLD, the subclone is the same as in A), high level of the same protein (EcKLD2) and Ec-Δ748-KL-Dendra (Δ748) were analyzed using anti-Dendra2 (Dn), anti-E-cadherin (Ec), and anti-tubulin (Tl) antibodies. (C) Time-lapse (selected frames from Movie S6) of adherens junctions between two Ec-Δ748-KL-Dendra expressing cells. Frame 0 shows the cells immediately after photoactivation. Frame 3 is 3 min later. (D) Time-lapse images (Movie S7) acquired at 20-sec intervals of the Ec-Δ748-KL-Dendra mutant clustering during the calcium-shift assay (numbers indicate seconds after addition of calcium). Frame 0 shows cells immediately before addition of calcium. Note that the cadherin mutant forms clusters nearly instantly. (E) The average decay of red fluorescence in adherens junctions of cells expressing different Ec-Dendra mutants in A-431 cells. The error bars represent SD (n = 20). (F) Clustering kinetics of Ec-Δ748-KL-Dendra (EcΔ748KLD) and Ec-KL-Dendra (EcKLD) mutants after addition of calcium (n = 10).

The inactivation of endocytic signals in the Ec-KL-Dendra mutant, however, hampered the mechanisms controlling cadherin expression. This is evident from the fact that the subclones expressing a low level of the mutant, such as in Fig. 3A, were found only occasionally; most subclones exhibited significant (more than twofold) overexpression of this mutant (Fig. 3B). Such cadherin overexpressing cells produced giant, very abnormal cadherin junctions (Fig. 2H), which, surprisingly, still exhibited a rapid cadherin turnover (Fig. 3E).

Intracellular Cadherin Region Mediates Cadherin Turnover in Adherens Junctions.

To assess the role of catenins in the adherens junction homeostasis, we studied A-431 cells expressing the Ec-Δ748-KL-Dendra mutant. This mutant, like the Ec-Δ748-KL-Myc mutant described above, features an extended deletion encompassing both the β-catenin and p120-binding sites (Fig. 2A). Because of this deletion, the mutant’s expression did not influence the level of endogenous cadherin (Fig. 3B). As a result, both the mutant and the endogenous cadherin were coclustered in the cell–cell junctions of these cells (see Fig. 2 E and E ′). Remarkably, the turnover of this mutant in the junctions was slower than that of the Ec-Dendra (Fig. 3 C and F and Movie S6). These observations were unexpected, because they suggested that this catenin-uncoupled cadherin mutant produces junctions, which, in fact, are more stable than the endogenous adherens junctions.

Coclustering of the Ec-Δ748-KL-Dendra mutant with endogenous cadherin may be caused by two alternative mechanisms. The first possibility is that these two proteins are corecruited into the junctions by the same catenin-independent diffusion-trap mechanism. In this case both the truncated and full-size cadherins would have similar clustering kinetics. If, alternatively, cadherin clustering is facilitated by catenins, then the Ec-Δ748-KL-Dendra mutant would be recruited into the junctions much more slowly. To test these possibilities, we performed a calcium switch experiment: the cell–cell junctions of cells expressing Ec-Δ748-KL-Dendra and Ec-KL-Dendra were first disintegrated in low calcium and then reconstituted by addition of calcium. Time-lapse analyses of the junction assembly showed that the recruitment of the Ec-Δ748-KL-Dendra mutant into the junctions took seconds and was completed even slightly faster than the assembly of the junctions in the control Ec-KL-Dendra cells (Fig. 3 D and F and Movie S7). Taken together, experiments with the tailless Ec-Δ748-KL-Dendra mutant suggested that the intracellular cadherin region is not essential for cadherin clustering.

The slow turnover of the cadherin tailless mutant in the junctions of A-431 cells may be caused by the endogenous cadherin present in the same junctions. To exclude this possibility, we compared the stability of the Ec-KL-Dendra and Ec-Δ748-KL-Dendra clusters in cadherin-deficient CHO cells—a classic model in cadherin studies (22). For this work, we produced CHO cell subclones expressing comparable levels of these two mutants (Fig. 4). The adherens junctions of Ec-KL-Dendra-expressing CHO cells, were very similar in their appearance and dynamics to those of A-431 cells (Fig. 4 A and C). Also, as with A-431 cells, ATP depletion blocked the junctional cadherin turnover and raised their fluorescence intensity (Fig. 4C). The adherens junctions in Ec-Δ748-KL-Dendra-expressing CHO cells were very different. Many junctions were enormous in size, sometimes filling the entire area of the cell–cell contact (Fig. 4B). Furthermore, the rate of mutant turnover in these cell junctions was irregular, but in all cases it was between that of the control and ATP-depleted Ec-KL-Dendra-expressing cells (Fig. 4C). Such slow and erratic dynamics suggest that catenin-uncoupling somehow stabilizes cadherin clusters but makes them more dependent on various local factors. The most striking feature of these junctions was that ATP depletion, while making uniform the dynamics of the individual junctions, changed neither their average rate (Fig. 4C) nor their fluorescent intensity.

Fig. 4.

(A and B) CHO cell subclones stably expressing Ec-KL-Dendra (A) and Ec-Δ748-KL-Dendra (B) mutants and stained using anti-Dendra antibody. (C) Turnover of cadherin and its mutant in CHO cells. Black lines show the average decay of the adherens junction red fluorescence in control (filled circles) and ATP-depleted (filled squares) CHO cells expressing Ec-KL-Dendra (EcKLD). The error bars represent SD. (n = 20). Blue (control cells) and red (ATP-depleted cells) lines show the decrease of red fluorescence in individual junctions of the Ec-Δ748-KL-Dendra expressing cells. Note the big difference in the rates of decay between individual junctions in control cells, but the relatively uniform rates for ATP-depleted cells. (D) Total lysates of CHO subclones expressing Ec-KL-Dendra (EcKLD) and Ec-Δ748-KL-Dendra (Δ748) were analyzed using anti-Dendra2 (Dn) and anti-tubulin (Tl) antibodies.

Discussion

The aim of this work is to examine adherens junction homeostasis. Using a unique approach—the activation of the Dendra-tagged cadherin in a small subset of adherens junctions—we found that in stationary conditions, adherens junctions of A-431 cells continuously recruit and lose cadherin molecules. The half-residence time of cadherin in the junction was ≈2 min. This value is a rough approximation and seems too high because of two apparent factors. First, a 5-μm diameter cell–cell contact area that we activated with a 405-nm light contained a large pool of extrajunctional cadherin. This pool, once activated, competed with the nonactivated green cadherin for recruitment into the activated red junctions. Second, an activated cadherin molecule that exited from the activated junction could reenter the same or neighboring junctions during the observation time. These two factors may lead to a very significant error; the actual cadherin residence time in the junction could be much less than 2 min. Although new experimental strategies are required to determine the actual rate of cadherin turnover in individual adherens junctions, our data reveal some basic principles of this process.

Using the traditional FRAP assay, a fast exchange of cadherin in the stationary adherens junctions has recently been shown to occur in MCF and MDCK cells (16). Using pharmacological inhibitors, the authors of that study proposed that this rapid turnover is mediated by cadherin endocytosis. Whereas our work shows that the mechanism of this process is far more complex, we also found that it is driven by the active removal of cadherin from adherens junctions. These data allow us to discard an alternative possibility—that the catenin-mediated recruitment of cadherin into the junctions continuously compensates for the loss of cadherin from the junctions that is caused by the very weak affinity of cadherin adhesive bonds. Indeed, in the case of this mechanism, the deletion of catenin-binding sites would impede cadherin clustering. In contrast to this outcome, we found that the deletion of the intracellular cadherin region stabilizes cadherin in the junctions and even slightly accelerates its clustering kinetics. The active removal of cadherin from the junctions was also evidenced by the fact that ATP depletion blocked the loss of cadherin from the junctions but not its recruitment into the junctions. This imbalance between cadherin recruitment and its release in the ATP-depleted cells led to the junctional entrapment of nearly the entire cadherin pool present on the cell surface. This work together with our previous biochemical data (13, 15), compellingly show that the cadherin extracellular region is able to produce relatively stable junctional structures.

Our data suggest that cadherin turnover in the junctions is a three-step process. A diffusion-trap process, driven by cadherin adhesive dimerization, continuously recruits cadherin adhesive homodimers into the periphery of the preexisting cadherin clusters. Next, these newly formed cadherin homodimers are laterally aligned with the cadherin dimers present in the cluster via direct dimer-to-dimer cis-bonds (3) and indirect coassociation via catenins and cytoskeleton. These bonds together establish a relatively immobile cadherin-containing adhesive scaffold that we observed in the ATP-depleted cells. The important role of intracellular interactions in the formation of this scaffold is suggested by our observation that the clusters produced by the catenin-uncoupled cadherin mutant cannot be stabilized by ATP depletion in CHO cells. Future work is needed to unravel the exact protein–protein interactions that integrate cadherin dimers into the cadherin clusters.

The third step in cadherin turnover is an energy-consuming release of cadherin from such clusters, which is needed to counterbalance the continuous recruitment of cadherin. Several potential mechanisms can mediate this process. One possibility is that it is driven by actin polymerization–depolymerization cycles, which are also blocked by ATP depletion (23). However, we found that cytochalasin D in the 500-nM concentration, which was shown to inhibit actin filament dynamics by capping actin filament barbed ends (24), did not change the rate of cadherin release from the junctions (Fig. S2). Furthermore, our attempts to stabilize adherens junctions using another actin filament stabilizing agent, jasplakinolide, were also unsuccessful. Another possibility is that cadherin release from the junctions is driven by clathrin-mediated endocytosis. This possibility is suggested by recent FRAP experiments (16), a pharmacological study of junction disintegration in low calcium (20), and our previous work with the catenin-uncoupled Ec-Δ748-Myc mutant (13). In that work we showed that the Ec-Δ748-Myc cadherin is mostly cytosolic and cannot form junctions in control cells. However, the mutant is able to form junctions once its endocytosis is blocked by clathrin depletion. The present work extends that observation. We identified two elements, the dileucine motif and Lys738, which are required for the efficient internalization of this mutant from the cell surface. Their joint inactivation by point mutations stabilized the mutant on the cell surface and led to the catenin-independent formation of cadherin junctions even in CHO cells lacking endogenous cadherin.

We found, however, that clathrin-mediated endocytosis alone cannot account for cadherin turnover in adherens junctions. It is evident from our observations that the inactivation of both the dileucine motif and K748 significantly reduced endocytosis of full-size cadherin without any effect on its junctional exchange rate. Furthermore, a biotinylation assay showed that only a small fraction of the surface Ec-KL-Dendra pool is internalized over a 15-min period in which, Dendra-activation experiments showed, the complete renewal of the entire junctional pool of this mutant takes place. Finally, in complementary experiments, we found that cadherin turnover in the A-431 cell junctions is independent of the Dynamin 2 inhibitors Dynasore and MiTMAB. A similar observation was made by Beco et al. for MCF cells (16). Therefore, if any kind of endocytosis is involved in cadherin junctional turnover, it must be a specific process that returns cadherin molecules to the cell surface immediately after internalization. Only if the entire endo/exocytosis cycle took just a few minutes, would it go undetected by our biotinylation assay. An alternative possibility is that the active cadherin removal from the junctions is based on some unknown motor proteins that use the cortical cytoskeleton as a track. Finally, chaperone-like ATP-dependent proteins could be required for the dissociation of cadherin adhesive dimers (13).

In summary, our live-cell imaging experiments showed that cadherin is recruited into the junctions with catenin-independent kinetics. Previous experiments suggesting a role for intracellular proteins in cadherin clustering may not have taken into account that catenin uncoupling can activate cadherin endocytic elements such as K748 or the dileucine motif. Our data, however, do not exclude the possibility that intracellular processes, such as actin or microtubule polymerization, positively modulate the production of new junctions and their subsequent expansion by pushing membranes of adjacent cells toward each other. Once recruited into junctions, cadherin is integrated into the clusters, from which it can be released only in the course of active dissociation. Future examination of the molecular mechanisms of cadherin exchange in individual adherens junctions is extremely important because this dynamic molecular process may play a role in the regulation of a great variety of more complex morphogenetic processes that require fast and coordinated junction remodeling.

Materials and Methods

Cell Culture, Antibodies, Plasmids, and DNA Transfections.

Transfection, growth, and immunofluorescence microscopy of human A-431 and CHO cells were done as described (15). Both cell lines were cultured in DMEM with 10% FCS. The plasmid pRc-EcM-Δ748 expressing Ec-Δ748-Myc protein was described (13). After transfection and selection, the cell colonies were screened for transgene expression and only homogeneously positive clones were selected for the experiments. The point mutations inactivating K738 and the dileucine motif were incorporated into Ec-Δ748-Myc using site-directed mutagenesis resulting in the plasmid pRc-EcM-Δ748-KL. A DNA encoding the myc epitope of this plasmid and the plasmid pRc-Ec1M (15) was replaced for EGFP- or Dendra2-encoding DNAs (Evrogen). Note that Ec-Dendra contains a small internal deletion (772–792), which eliminates the epitope for the anti-E-cadherin C20820 antibody, but does not abolish the cadherin’s function (15). The following antibodies were used: anti-E-cadherin, clone HECD-1 (Zymed Laboratories), recognizing recombinant and endogenous cadherins in A-431 cells; and clone C20820 (BD Biosciences), recognizing only the endogeous cadherin; mouse anti-β-catenin (BD Biosciences); rabbit anti-Dendra2 (Evrogen); and rabbit anti-myc (Santa Cruz Biotechnology). Depletion of ATP was achieved by ATP-depletion media containing 2 mM 2-deoxy-D-glucose and 1 μM antimycin A or 5 mM NaN₃ (for details, see ref. 15).

Internalization Assay and Western Blotting.

The overall rate of cadherin endocytosis was determined as previously described (15). In brief, cells were surface-biotinylated using sulfo-NHS-SS-biotin (0.5 mg/mL, for 10 min at 4 °C), followed by a chase in normal culture media for various durations. Noninternalized biotin was then stripped from the surface by two 20-min washes with a glutation solution. The internalized biotinylated proteins were precipitated by streptavidin-agarose (Sigma) and analyzed by immunoblotting.

Live-Cell Imaging and Data Processing.

Cell suspension (0.3 mL, 2 × 10⁵ cells/mL) was plated into a homemade chamber (Φ = 1.28 cm) built on no. 1.5 cover glass. Before plating, it was coated with 804G cell-conditioned media for 1 h at room temperature. The next day, the culture media was replaced with imaging media (L-15 plus 10% FBS) and the chamber was fixed onto the stage of an Eclipse Ti-E microscope controlled with Nikon’s NIS-Elements software and Perfect Focus System. The microscope was equipped with an incubator chamber, a CoolSNAP HQ2 camera, and two different halogen and mercury arc episcopic light sources, which minimize the photo-damage to the illuminated cells. A circular region of interest (Φ = 5 μm) was photoactivated by a 3-sec-long exposure to the 402-nm light using the mercury arc light source and a pinhole insert. Time-lapse images were taken in both FITC and TRITC filter sets using halogen light source that minimized phototoxicity and photobleaching. For the calcium switch experiment, cells were first incubated for 30 min in low calcium media (L-15/20 μM Ca²⁺/10% calcium-free FBS), and then the media was replaced with regular imaging media. All images were saved as Tiff files for further image analysis.

Tiff stacks were processed using ImageJ software (National Institutes of Health). A circular region of interest (Φ = 0.65 μm) was positioned on a single AJ and the mean value of fluorescent intensity in each frame was calculated in both photoactivation and calcium switch experiments. In photoactivation experiments, the red fluorescent intensity was normalized in such a way that 0 and 1 corresponded to the background and the initial values, respectively. The background value was obtained from the image taken right before the photoactivation. The time course of intensity change was produced from 10 sets of independent experiments. The intensity changes during the calcium switch were normalized to 0 and 100% for the minimum and maximum values.