|Home | About | Journals | Submit | Contact Us | Français|
Spatial and functional organization of cells in tissues is determined by cell-cell adhesion, thought to be initiated through trans-interactions between extracellular domains of the cadherin family of adhesion proteins, and strengthened by linkage to the actin cytoskeleton. Prevailing dogma is that cadherins are linked to the actin cytoskeleton through β-catenin and α-catenin, although the quaternary complex has never been demonstrated. We test this hypothesis and find that α-catenin does not interact with actin filaments and the E-cadherin-β-catenin complex simultaneously, even in the presence of the actin binding proteins vinculin and α-actinin, either in solution or on isolated cadherin-containing membranes. Direct analysis in polarized cells shows that mobilities of E-cadherin, β-catenin, and α-catenin are similar, regardless of the dynamic state of actin assembly, whereas actin and several actin binding proteins have higher mobilities. These results suggest that the linkage between the cadherin-catenin complex and actin filaments is more dynamic than previously appreciated.
The spatial and functional organization of cells in tissues is determined by cell-cell adhesion (Takeichi, 1995). The cadherin family of Ca2+-dependent cell-cell-adhesion proteins play important roles in initiating adhesion and cell sorting in development (Takeichi, 1995; Foty and Steinberg, 2005). Disruption of cadherin function abrogates normal embryonic development (Larue et al., 1994; Tepass et al., 1996; Costa et al., 1998) and is a common occurrence in metastatic cancers (Thiery, 2002).
Regulation of cadherin-mediated cell-cell adhesion is determined by distinct protein interactions of the cadherin extracellular and cytoplasmic domains. The extracellular domain forms homo- and heterophilic bonds with cadherins on adjacent cells, which specify cell-cell recognition and sorting of mixtures of cells (Gumbiner, 2000; Foty and Steinberg, 2005). Binding between cadherin extracellular domains is relatively weak, but cell-cell adhesion may be strengthened by lateral clustering of cadherins mediated by protein linkages between the cadherin cytoplasmic domain and the actin cytoskeleton (Jamora and Fuchs, 2002). The cadherin cytoplasmic domain forms a high affinity, 1:1 complex with β-catenin, and β-catenin binds with lower affinity to α-catenin (Aberle et al., 1994; Hinck et al., 1994; Pokutta and Weis, 2000; Huber and Weis, 2001).
Several studies show that α-catenin interacts with the actin cytoskeleton. Purified α-catenin binds and bundles actin filaments in vitro, and the actin binding site of α-catenin maps to the C-terminal domain (Rimm et al., 1995; Pokutta et al., 2002). In addition to binding actin, α-catenin interacts with actin binding proteins, including vinculin (Watabe-Uchida et al., 1998; Weiss et al., 1998), α-actinin (Knudsen et al., 1995; Hazan et al., 1997), ZO-1 (Itoh et al., 1997), spectrin (Pradhan et al., 2001), Ajuba (Marie et al., 2003), afadin (Pokutta et al., 2002), and formin (Kobielak et al., 2004).
However, the key experiment showing that α-catenin binds simultaneously to the cadherin-β-catenin complex and actin, either directly or indirectly through actin binding proteins, has not been performed. Here we show that α-catenin binding to β-catenin and α-catenin binding to actin filaments are mutually exclusive, both in vitro and on isolated cadherin-containing membrane patches, and are independent of E-cadherin clustering. In these assays, we could not establish an indirect link of the cadherin-catenin complex and actin filaments through vinculin or α-actinin. The cadherin-catenin complex displays dynamics very different from actin or other actin-associated proteins in in vivo imaging experiments. These results indicate that our understanding of how cadherins interact with the actin cytoskeleton must be reassessed.
E-cadherin cytoplasmic domain (Ecyto), β-catenin, and α-catenin form a stoichiometric complex in vitro (see below). We tested whether α-catenin could bind simultaneously to the E-cadherin-β-catenin complex and actin filaments in an actin-filament pelleting assay (Figure 1A). A small amount of E-cadherin-β-catenin complex pelleted with actin filaments, which most likely is due to nonspecific protein trapping in the actin-filament network. Purified α-catenin pelleted with actin filaments (Rimm et al., 1995; Pokutta et al., 2002). α-catenin also pelleted with actin filaments in the presence of increasing concentrations of E-cadherin-β-catenin complex, whereas the E-cadherin-β-catenin complex did not pellet above the background level (Figure 1A), even though the assay was performed at protein concentrations that were sufficient to form a ternary Ecyto-β-catenin-α-catenin complex. Similar results were obtained when plakoglobin was substituted for β-catenin (see Figure S1 in the Supplemental Data available with this article online). This result indicates that the interaction of α-catenin with actin filaments significantly decreases the affinity of α-catenin for the cadherin-catenin complex. It was reported previously that purified β-catenin pelleted with α-catenin and actin (Rimm et al., 1995). In that study, both β-catenin and α-catenin were prepared as GST-fusion proteins, which raises the possibility that β-catenin was pelleted through homodimerization of the GST tag rather than a direct α-catenin-β-catenin interaction.
Since the affinity of the α-catenin-β-catenin interaction is low, we generated a chimeric β-catenin-α-catenin protein and tested its binding to actin filaments. The chimera mimics the interaction of the two proteins by covalently linking the α-catenin binding site of β-catenin to the β-catenin binding domain of α-catenin (Pokutta and Weis, 2000) and extends to the C terminus of α-catenin including the actin binding site (Figure 1B); this creates a high effective concentration of the partners that favors formation of the complex. The chimera failed to bind actin in the pelleting assay (Figure 1C), demonstrating that the interaction of β-catenin with α-catenin strongly affected the affinity of α-catenin for actin. Taken together, these results show that the interaction of α-catenin with the E-cadherin-β-catenin complex and actin filaments is mutually exclusive.
Lateral clustering of cadherins is thought to be important in the formation of strong cell-cell adhesions (Yap et al., 1997) and may increase interactions of the cadherin-catenin complex with the actin cytoskeleton. To mimic cadherin clustering, E-cadherin cytoplasmic domain was fused to the coiled-coil domain of the cartilage oligomeric matrix protein (COMP), which forms a pentamer (Tomschy et al., 1996) (Figure 2A). Gel filtration chromatography showed that GST-COMP-Ecyto eluted at a molecular weight higher than that calculated for the pentamer (S.P. and F.D., unpublished data), which is most likely due to the unstructured nature of Ecyto when not bound to β-catenin (Huber et al., 2001) and/or further protein oligomerization through the GST tag. In a pull-down assay, we compared binding of increasing amounts of α-catenin to constant amounts of either GST-COMP-Ecyto or GST-Ecyto and β-catenin and found no detectable difference (Figure 2B). This result shows that clustering of the E-cadherin-β-catenin complex by COMP did not affect α-catenin binding to β-catenin.
We next tested the effect of E-cadherin-β-catenin oligomerization on α-catenin binding in the presence of actin filaments in the actin pelleting assay. Neither GST-COMP-Ecyto-β-catenin nor GST-Ecyto-β-catenin pelleted with α-catenin and actin filaments (Figure 2C). This result demonstrates that α-catenin bound to F-actin does not associate with β-catenin bound to E-cadherin, even when oligomerization increased the local concentration of the E-cadherin-β-catenin complex.
Cadherins are transmembrane proteins that become organized into clusters in the plasma membrane in part through trans-interactions with cadherins on the opposing cell (Gumbiner, 2000). These conditions were not met by the solution biochemistry experiments described above. Therefore, we developed a method to reconstitute the catenin complex onto cadherin-containing plasma membranes isolated from mammalian epithelial cells. MDCK cells rapidly attached and spread on a substratum of correctly oriented, high-density extracellular domain of E-cadherin fused to Fc purified from HEK293 mammalian cells (Figure 3A); in this short period, few or no integrin-based adhesions were formed (Drees et al., 2004). The ventral cell surface is the interface between endogenous cadherin and the cadherin substratum. To access the cytoplasmic side of this ventral cell surface, cell monolayers were sonicated briefly to remove the dorsal plasma membrane, nucleus, and almost all intracellular membrane organelles, a treatment termed “unroofing” (Drees et al., 2004; Heuser, 2000; Figure 3A). A patchwork of ventral membranes attached to the cadherin substratum was left behind after sonication (Figure 3B). Puncta-like clusters containing E-cadherin, β-catenin, and α-catenin overlayed by actin filaments were localized at the interface with the E-cadherin substratum (Figure 3B).
In order to reconstitute catenin and actin binding at E-cadherin-mediated adhesion sites on these ventral membrane patches, membrane-associated proteins were removed with 4 M guanidine hydrochloride, including nearly all β- and α-catenin, as judged by immunostaining (Figure 3C), and other membrane-associated proteins, including actin and vinculin (F.D., unpublished data). Note that membrane patches containing E-cadherin remained attached to the substratum and that the exposed cytoplasmic domain of E-cadherin is natively unfolded (Huber et al., 2001) and, therefore, would be unaffected by 4 M guanidine hydrochloride. Subsequent incubation of the stripped membrane patches with purified recombinant β-catenin led to the readdition of β-catenin to the membranes, where it colocalized precisely with E-cadherin clusters (Figure 3C). As expected, α-catenin did not bind to stripped membranes. However, when membrane patches were preincubated with β-catenin, α-catenin accumulated with β-catenin at E-cadherin clusters (Figure 3C). Quantification of the fluorescent signal of α- or β-catenin added to the patches relative to the E-cadherin immunofluorescence signal confirmed this observation (Figure 3D). β-catenin addition to patches reached about 80% of the prestripped control levels. α-catenin addition in the presence of β-catenin was less efficient, reaching only about 25% of control levels, perhaps reflecting the weaker affinity of the α-catenin-β-catenin interaction under these conditions.
We tested whether phosphorylation of catenins affected their binding to E-cadherin. Although phosphorylation of β-catenin by casein kinase II (CKII) has been reported to increase its affinity for α-catenin in vitro (Bek and Kemler, 2002), we did not find any increase in binding of α-catenin to phosphorylated β-catenin. We found that α-catenin is also a substrate for CKII (data not shown), but we found that phosphorylation did not affect α-catenin binding to β-catenin on membrane patches (Figure 3D) or in an in vitro pull-down assay (unpublished data). In summary, specific, order-of-addition catenin binding can be reconstituted without posttranslational phosphorylation on membrane patches at E-cadherin-mediated adhesion sites.
We tested whether the reconstituted cadherin-catenin complex on membrane patches could bind actin filaments. Although fluorescein-labeled G-actin bound to membrane patches prior to stripping, we did not detect actin binding to stripped membrane patchesor to stripped membrane patches preincubated with β-catenin and α-catenin (Figure 3E). Similar results were obtained if prepolymerized filaments were added to the membrane or α- or β-catenin was phosphorylated by CKII (F.D., unpublished data). This result was surprising since actin filaments were present on the original membrane patch prior to stripping with 4 M guanidine hydrochloride. The simplest explanation, taking into account our failure to reconstitute binding of the cadherin-catenin complex to actin filaments in solution (see above), is that actin filaments do not bind the cadherin-catenin complex on membranes. The presence of actin filaments on membrane patches prior to stripping may be due to their interaction with proteins other than the cadherin-catenin complex, although it remains possible that additional proteins are required for interactions with this complex (but see below).
Several actin binding proteins interact with α-catenin and could mediate linkage of the cadherin-catenin complex to actin filaments (Jamora and Fuchs, 2002). Of these proteins, vinculin and α-actinin are considered good candidates (see Introduction).
To test whether vinculin can bridge an interaction between the cadherin-catenin complex and actin filaments, we performed actin pelleting assays in the presence of vinculin, α-catenin, β-catenin, and Ecyto. Vinculin, β-catenin, and Ecyto did not pellet with α-catenin and actin filaments above background levels (Figure 4A). In the case of vinculin, this result is not surprising since its binding to actin and other ligands is autoinhibited by an intramolecular head-tail interaction (Johnson and Craig, 1995). Therefore, pull-down assays were used to compare binding of the vinculin head region with full-length vinculin to components of the cadherin-catenin complex. The head domain of vinculin bound directly to α-catenin (Figure 4B), as shown previously (Watabe-Uchida et al., 1998; Weiss et al., 1998), whereas full-length vinculin did not (Figure 4B).
β-catenin bound to the vinculin head domain and full-length vinculin in a pull-down assay (Figure 4B; Hazan et al., 1997). To test whether binding of β-catenin to full-length vinculin released the inhibitory head-to-tail interaction and enabled vinculin to bind actin filaments, we monitored the amount of vinculin that pelleted with actin filaments in the presence of increasing amounts of the Ecyto-β-catenin complex. There was no increase in the fraction of full-length vinculin that pelleted with actin filaments, nor did the Ecyto-β-catenin complex pellet with actin filaments above background levels (Figure 4C). These results indicate that, even though β-catenin bound full-length vinculin, this interaction did not appear to disrupt the intramolecular head-to-tail interaction that masks the actin binding domain of vinculin.
We tested vinculin binding to cell-adhesion sites on membrane patches to assess whether the presence of native cellular membranes would influence vinculin binding to catenin-cadherin complex. Some vinculin staining was detected on control membrane patches prior to stripping, but there was little or no colocalization with E-cadherin (Figure 4D). Incubation with 4M guanidine hydrochloride eliminated most of vinculin staining. Readdition of full-length vinculin in the presence of β-catenin revealed some binding to membrane patches but very little full-length vinculin bound to membrane patches preincubated with β-catenin and α-catenin (Figures 4D and 4E). As expected from the pull-down assays, the vinculin head domain showed significantly higher binding than full-length vinculin to stripped membrane patches preincubated with either β-catenin or β-catenin and α-catenin (Figures 4D and 4E).
Although α-actinin has been reported to be associated with the cadherin complex (Knudsen et al., 1995), we were unable to detect any specific binding between the cadherin-catenin complex and α-actinin in pull-down assays (Figure 4F). Similarly, little or no specific binding of α-actinin was detected on stripped membrane patches preincubated with β-catenin and α-catenin (Figures 4E and 4G).
To test whether cytosolic factors might be involved in the linkage of the cadherin-catenin complex to actin filaments, cytosol was added to stripped membrane patches that had been preincubated with α-catenin and β-catenin followed by fluorescein-labeled actin. Neither cytosol from confluent MDCK cells (data not shown) nor bovine brain cytosol (Figure 4H), which is considerably more concentrated than MDCK cell lysate and is therefore potentially a better source of a missing factor, induced binding of actin filaments to membrane patches (Figure 4H). Likewise, no actin binding was observed when either vinculin or α-actinin was added to membrane patches in the presence of β-catenin and/or α-catenin and cytosol (Figure 4H).
Since a ternary complex of Ecyto-β-catenin-α-catenin but not a quaternary complex of these proteins and actin could be assembled in vitro, we re-examined the dynamics of these proteins in living cells. MDCK cells were stably transfected with E-cadherin, β-catenin, and α-catenin tagged with green fluorescent protein (GFP). In each case, the level of exogenous protein expression in stable cell lines was less than that of the endogenous protein (Figure 5A). Stably expressed E-cadherin-GFP and GFP-β-catenin localized at the plasma membrane and accumulated at cell-cell contacts with little or no protein in the cytoplasm (Figure 5B; Movies S1 and S2), whereas GFP-α-catenin localized to cell-cell contacts and throughout the cytoplasm (Figure 5B; Movie S3), as expected from the distribution of endogenous proteins (Nathke et al., 1994). Protein dynamics were measured by monitoring fluorescence recovery after photobleaching (FRAP) in dense cell monolayers in which mature cell-cell contacts had formed for 24–36 hr.
At cell-cell contacts, the mobile fractions of membrane bound E-cadherin-GFP (22.9% ± 2.2%), GFP-β-catenin (34.2% ± 2.6%), and GFP-α-catenin (33.7% ± 2.2%) were similar (Figure 5B; Movies S1–S3). The immobile fractions of these proteins may be due to trans-interactions between the extracellular domains of E-cadherin on opposing cells. Of these mobile fractions, the half-time of fluorescence recovery (τ1/2) for E-cadherin-GFP (0.54 ± 0.1 min), GFP-β-catenin (0.66 ± 0.08 min), and GFP-α-catenin (0.43 ± 0.04 min) was also similar (Figures 5B and 5D). These recovery rates are consistent with direct protein-protein interactions between E-cadherin, β-catenin, and α-catenin.
To test whether linkage of the cadherin-catenin complex to the actin cytoskeleton affects component mobilities, we expressed mutants of E-cadherin lacking the cytoplasmic domain (E-cadherinΔC-tdDsR) and α-catenin lacking the actin binding domain (GFP-α-cateninΔC). E-cadherinΔC had a mobile fraction and mobility rate similar to those of full-length E-cadherin (Figure 5D; Movie S4), which may be due to interactions with endogenous E-cadherin. GFP-α-cateninΔC localized to cell-cell contacts, as expected since it includes the β-catenin binding site, and had a mobile fraction and recovery rate similar to those of full length α-catenin (Figure 5D; Movie S5). Thus, breaking potential links between components of the cadherin-catenin complex and the actin cytoskeleton did not affect dynamics of the complex.
If the cadherin-catenin complex is bound directly and stably to actin at cell-cell contacts, we would expect that membrane- associated actin would have a significant immobile fraction with a recovery rate similar to those of proteins in the cadherin-catenin complex. MDCK cells expressing a low level of GFP-actin (<3% of the total actin level; Figure 5A) showed GFP-actin localization to stress fibers, lamellipodia, and cell-cell contacts in addition to a large cytoplasmic pool (Figure 5B; Ehrlich et al., 2002). When GFP-actin was photobleached at cell-cell contacts (Figure 5B; Movie S6), its recovery was almost complete (mobile fraction = 90.0% ± 7.2%) and rapid (τ1/2 = 0.16 ± 0.03 min), in contrast to the more immobile and relatively slow recovery properties of membrane bound E-cadherin, β-catenin, and α-catenin (Figure 5D).
The high mobility of actin observed at the membrane may be due to rapid exchange with the large pool of cytoplasmic actin. To eliminate the high fluorescence signal of cytoplasmic GFP-actin that might prevent observation of a less dynamic, membrane-associated pool, we expressed actin tagged with photoactivatible GFP (PAGFP-actin). Photoactivation of PAGFP-actin at cell-cell contacts was immediately followed by a rapid loss of the GFP signal from the activated spot at a rate (τ1/2 = 0.27 ± 0.02 min) closer to that of GFP-actin than that of the cadherin-catenin complex (Figure 5B; Movie S7). Lateral diffusion at the activated spot of PAGFP-actin was not observed. Significantly, the signal from minimally activated PAGFP-actin decayed with a single exponential function that reached complete depletion of fluorescence at the contact site (S.Y., unpublished data), indicating that a there is a single reaction step consistent with the simple exchange of actin between the membrane-associated and cytosolic pools.
Although GFP-actin polymerizes inefficiently, GFP-actin can copolymerize with endogenous actin (Westphal et al., 1997) and colocalized with endogenous actin filaments (Figures 5 and and6).6). Nevertheless, to circumvent potential problems associated with drawing conclusions from kinetics of GFP-actin, we introduced fluorescently labeled actin into MDCK cells either by microinjection or after permeabilizing cells with a low concentration of saponin. Microinjected fluorescently labeled actin (Rhod-actin) localized to cell-cell contacts (Figure 5B) and the tips of actin bundles at focal adhesions in saponin-treated cells (Figure 5C). After photobleaching, microinjected Rhod-actin had recovery kinetics similar to that of GFP-actin (τ1/2 = 0.21 ± 0.03 min, Figure 5B; Movie S8).
To test the extent of actin exchange between the cytoplasm and membrane-associated pools, we measured the fluorescence loss of GFP-actin at cell-cell contacts while continuously photobleaching a spot in the cytoplasm (fluorescence loss in photobleaching, FLIP). Different periods of photobleaching were tested, with similar results, but the period used was generally proportional to the level of expression of the GFP-tagged protein. GFP-actin was rapidly and completely depleted from cell-cell contacts when the cytoplasmic pool was continuously photobleached (Figure 6A; Movie S9), consistent with the FRAP experiments (Figure 5); in contrast, α-catenin exhibited a much slower loss from the membrane-associated pool (Drees et al., 2005 [this issue of Cell]). Note, however, that GFP-actin exhibited a much slower dissociation from stress fibers associated with focal adhesions at the base of cells to the actin cytoplasmic pool (Figure 6A; Movie S10). Thus, actin associated with cell-cell contacts is unusually dynamic compared to that associated with cell-substratum adhesion.
Together these data show that membrane-associated actin at cell-cell contacts rapidly exchanges with a cytoplasmic actin pool. Furthermore, most of the actin at cell-cell contacts is highly mobile, with a recovery rate much faster than that of either α-catenin, β-catenin, or E-cadherin, indicating that actin filaments were not stably associated with the cadherin-catenin complex.
We examined the dynamics of several actin binding proteins at cell-cell contacts that might indirectly link the cadherin–catenin complex to the actin cytoskeleton. Stably expressed GFP-vinculin localized intensely to sites of focal adhesions and the cytoplasm, but very weakly to cell-cell contacts (Figure 6B). Continuous photobleaching of the cytoplasmic pool of GFP-vinculin immediately depleted the small amount of GFP-vinculin at the cell-cell contacts (Figure 6B; Movie S11) but more slowly depleted GFP-vinculin from focal adhesions (Figure 6B; Movie S12). Stably expressed Arp3-GFP had a punctate localization along cell-cell contacts, and these spots disappeared rapidly when the cytoplasmic pool of Arp3-GFP was continuously photobleached (Figure 6C; Movie S13). Thus, although some vinculin and Arp2/3 complex localized to cell-cell contacts, the results of these experiments indicate that, like actin, they were not stably associated with the plasma membrane and that their recovery rates were very different from those of components of the cadherin-catenin complex. Similar results were obtained with mDia2-GFP and formin 1-GFP (S.Y., unpublished data).
Given that the dynamics of actin at cell-cell contacts differ from those of E-cadherin, α-catenin, and β-catenin, we tested whether disruption of actin at cell-cell contacts affected the dynamics of the cadherin-catenin complex using cytochalasin D, which binds to the barbed end of actin filaments (Cooper, 1987), and jasplakinolide, which stabilizes actin filaments (Cramer, 1999). While these approaches affect actin organization globally, they complement our more direct experiments. When an MDCK cell monolayer was treated with 10 μM cytochalasin D for 1 hr, most of the GFP-actin (Figure 7A; Movie S14) and endogenous actin (Figure 7B) redistributed and aggregated in the cytoplasm, though a small fraction remained associated with intact cell-cell contacts. In cytochalasin-treated cells, GFP-actin selectively photobleached at cell-cell contacts had a slower rate of recovery and a lower mobile fraction (Figure 7A and D) than in control cells. However, the recovery rate and mobile fraction of E-cadherin-GFP and GFP-α-catenin remained the same as in control cells (Figures 7C and 7D; Movies S15 and S16).
Treatment of cells with jasplakinolide induced the formation of extensive actin stress fibers (S.Y., unpublished data) and some actin aggregates in the cytoplasm, but cell-cell contacts remained intact (Figures 7E and 7F; Movie S17). In the presence of 0.2 μM jasplakinolide, GFP-actin became highly immobile, although the small mobile fraction had the same recovery rate as that of GFP-actin in control cells (Figure 7H). Despite the significant immobilization of GFP-actin by jasplakinolide, the mobile fraction and recovery rate of either E-cadherin-GFP or GFP-α-catenin was not statistically different from those of E-cadherin-GFP and GFP-α-catenin in control cells (Figures 7G and 7H; Movies S18 and S19). Together these results demonstrate that the mobility of the cadherin-catenin complex at cell-cell contacts is independent of the state of actin organization or dynamics.
The interaction of cadherins with cytoplasmic proteins and the actin cytoskeleton is thought to underlie many aspects of cell-cell adhesion, including clustering of cadherins, strengthening of adhesive contacts, and downstream effects on membrane and cell organization (Kobielak and Fuchs, 2004). To understand these processes, it is essential that protein assemblies that link cadherins to the actin cytoskeleton are rigorously defined. Virtually all of the candidate components of these assemblies (see Introduction) were identified through binary interactions such as yeast two-hybrid, pull-down, or coimmunoprecipitation assays. A general assumption has been that binding of a given protein to two distinct partners means that all three proteins are in the same complex. In particular, independent binding of α-catenin to β-catenin-E-cadherin and to actin filaments has led to the assumption that α-catenin binds to both simultaneously although this quaternary complex has not been demonstrated previously. A direct test of this widely believed conclusion presented here shows, in fact, that this is not the case.
Using purified proteins in solution or membrane patches containing clustered E-cadherin, we reconstituted, in the correct order of protein-protein interactions, a ternary complex of E-cadherin-β-catenin-α-catenin and the interaction between α-catenin and actin filaments in vitro. However, we were unable to bind α-catenin simultaneously to the E-cadherin-β-catenin complex and to actin filaments, even when E-cadherin was clustered in vitro (COMP-Ecyto) or on membrane patches from cells. Similarly, we could reconstitute vinculin binding to β-catenin or α-catenin but not to the cadherin-β-catenin complex and actin filament simultaneously. α-actinin could not be reconstituted into complexes with either β-catenin or α-catenin.
An additional activity might be needed to enable simultaneous binding of α-catenin or vinculin to E-cadherin-β-catenin complex and actin filaments or to relieve the head-to-tail autoinhibition of vinculin bound to β-catenin. Neither lipids in the membrane patches nor MDCK or bovine brain cytosol provided such an activity, although we cannot exclude that such a factor was missing or inactivated in our cytosol preparations. We also tested whether specific post-translational modifications shown previously to enhance complex assembly were involved (Lilien et al., 2002). Serine/ threonine phosphorylation by CKII had no effect on α- or β-catenin interaction or on binding of actin filaments to reconstituted cadherin-catenin complexes on membrane patches. The small GTPase Rac1 and PI3K have been suggested to be transiently activated during initial phases of cell-cell contact formation and cadherin ligation (Ehrlich et al., 2002; Kovacs et al., 2002; Noren et al., 2001). However, our preparations of membrane patches would not be able to capture these transient activation states that might be important for actin-filament interactions with the cadherin-catenin complex. Alternatively, linkage of the E-cadherin-β-catenin complex to actin filaments could be mediated by other α-catenin and actin binding proteins, including ZO-1 (Itoh et al., 1997), afadin (Pokutta et al., 2002), spectrin (Pradhan et al., 2001), Ajuba (Marie et al., 2003), or formin-1 (Kobielak et al., 2004). However, there is no direct evidence that these proteins can bind simultaneously to α-catenin and actin filaments. In this context, it is noteworthy that, when membrane extracts of adherent cells are immunoprecipitated with anti-cadherin antibody, only β- and α-catenin are coprecipitated stoichiometrically in the complex (Hinck et al., 1994; Ozawa and Kemler, 1992). Thus, although a number of actin binding proteins are reported to colocalize with cadherins and/or interact with α-catenin, it is unclear whether any of them represents a significant structural component of the cadherin-catenin complex in cells.
If the core cadherin-catenin complex does not bind to actin filaments directly, we would expect that interactions between this complex and underlying the actin cytoskeleton in cells might be very dynamic rather than being relatively static as has been assumed. E-cadherin, β-catenin, and α-catenin had essentially identical mobile fractions and recovery rates, consistent with an integrated complex of these proteins on the membrane. In contrast, actin, vinculin, and the Arp2/3 complex were highly mobile and had recovery rates completely different from those of the cadherin-catenin complex. These data are also inconsistent with a static linkage of the cadherin-catenin complex, either directly or indirectly, to the actin cytoskeleton and support our biochemical studies that the cadherin-catenin complex does not bind to actin filaments.
That a stable linkage does not exist between membrane-anchored cadherin cell-adhesion molecules and the underlying cytoskeleton may be surprising. However, adhesion must be a dynamic process to enable morphogenetic changes during cell and tissue development (Takeichi, 1995). The interaction of clustered cadherin extracellular domains on opposing cells may provide the necessary adhesive force as long as the underlying actin cytoskeleton is correctly organized to provide the mechanical properties required for cell and tissue function. Data presented in the accompanying paper (Drees et al., 2005) provide mechanistic evidence of why α-catenin does not bind simultaneously to the E-cadherin-β-catenin complex and actin filaments and new insights into how α-catenin may regulate actin dynamics at cell-cell contacts.
Murine E-cadherin cytoplasmic domain (Aberle et al., 1994; Huber et al., 2001), COMP-E-cadherin cytoplasmic domain, murine β-catenin, plakoglobin, and βα-catenin were expressed as N-terminal cleavable GST-fusion proteins. Full-length α-catenin (mouse α(E)-catenin) (Aberle et al., 1994; Huber et al., 2001) and vinculin head domain (a kind gift of Dr. Sue Craig, Johns Hopkins University) were expressed with a C-terminal His6 tag. Construction of expression vectors and protein purifications are described in the Supplemental Data. Full-length vinculin and α-actinin were purified from chicken gizzard as described previously (Feramisco and Burridge, 1980). Vinculin head domain was also generated by cleavage of full-length vinculin with Glu-C V8 endoproteinase from Staphylococcus aureus, as described previously (Johnson and Craig, 1995).
β- and α-catenin were phosphorylated with casein kinase II (NEB) as described previously (Bek and Kemler, 2002).
Binding assays were performed in 20 mM Tris (pH 8.0), 100 mM KCl, 10 mM MgCl2, and 1 mM DTT. Proteins were incubated at a 10 μM concentration in a total volume of 60 μl. After 1 hr, the protein mix was incubated with 60 μl glutathione agarose and incubated for 1 additional hr. After centrifugation and removal of the supernatant, beads were washed four times with 500 μl wash buffer containing 20mM Tris (pH 8.0), 20mM KCl, 0.1% Triton X-100, and 1 mM DTT. After the last wash step, the supernatant was removed and the beads were resuspended in 60 μl gel loading buffer. Binding assays with full-length and the head domain of vinculin were performed at 8 μM protein concentration. For binding assays with His-tagged α-catenin, full-length and vinculin head domain were precleared with 30 μl Ni-NTA beads and incubated with α-catenin coupled to 30 μl Ni-NTA beads in 20 mM Tris (pH 8.0), 150 mM NaCl, and 2 mM imidazole for 1 hr. Gel filtration binding assays are described in the Supplemental Data.
Skeletal actin was prepared from chicken pectoral muscle from acetone powder as described previously (Spudich and Watt, 1971). G-actin was purified on a Superdex 200 column, and aliquots were frozen at −70°C in G-actin storage buffer (5 mM Tris [pH 8.0], 0.2 mM CaCl2, 0.2 mM ATP, and 0.5 mM DTT). Polymerization was induced by addition of polymerization buffer to a final concentration of 50 mM KCl, 2 mM MgCl2, and 1 mM ATP. After incubation at room temperature for 2 hr, F-actin was mixed with the respective proteins at a 1:1 ratio at final concentration of 8 or 10 μM. Proteins were incubated at room temperature for 1 hr and then centrifuged at 100,000 rpm (TLA 100, Beckman) for 7 min. Supernatant and pellet were analyzed by gel electrophoresis and Coomassie blue staining.
80% confluent MDCK GII cells were lysed in homogenization buffer (20 mM HEPES [pH 7.2], 90 mM K acetate, 2mM Mg acetate, 25mM sucrose, 10 μg/ml leupeptin, antipain, pepstatin, and 0.1 mM Pefabloc) by sonication. Postnuclear supernatant was subjected to ultracentrifugation at 100,000 × g for 45 min at 4°C (TLA 100, Beckman). Bovine brain cytosol was prepared as described previously (Grindstaff et al., 1998).
E-cadherin:Fc substratum and lateral membranes from MDCK GII cells were prepared as described previously (Drees et al., 2004). To strip membrane- associated proteins, membrane patches were incubated with 4 M guanidine hydrochloride or 0.5 M sodium bicarbonate (pH 11.5) in Ringer’s buffer (10 mM HEPES [pH 7.5], 154 mM NaCl, 7.2 mM KCl, and 1.8 mM CaCl2) (RB) for 20 min at RT, washed with RB, and fixed or blocked with 2% BSA for 20 min at RT. Membranes were rinsed with RB and incubated with bovine brain cytosol or purified proteins for 1 hr as indicated. For actin binding to lateral membranes, patches were incubated with 5 μM actin containing 10% fluorescein-labeled actin (Cytoskeleton) in polymerization buffer for 30 min at RT (either G-actin or prepolymerized), rinsed briefly with polymerization buffer, fixed with 1.6% paraformaldehyde in RB, and processed for immunofluorescence. The following antibodies were used for characterization of membrane patches: anti-E-cadherin E24 (Marrs et al., 1993), anti-β-catenin (Transduction Lab), anti-α-catenin (Alexis Corp.), anti-actin (Chemicon), anti-vinculin 11-5 (Sigma), and anti-α-actinin (Sigma). Fluorescent intensities were analyzed in ImageJ (http://rsb.info.nih.gov/ij/).
MDCK G type II cells were transfected with Lipofectamine 2000 (Invitrogen) and selected with G418 (Invitrogen). Construction of expression vectors is described in the Supplemental Data. The antibodies used to characterize stable cell lines are the same as above except E-cadherin antibody: anti-E-cadherin 3G8 (Shore and Nelson, 1991). Relative expression levels of GFP-tagged and endogenous proteins are shown in Figure 5A. pEGFP-C1-vinculin and pEGFP-N1-Arp3 plasmids were gifts from Drs. Susan Craig (Johns Hopkins University) and Matt Welch (University of California, San Francisco), respectively. All plasmids were stably expressed in MDCK GII cells without any apparent change in phenotype.
Cells were seeded on collagen-coated coverslips for 24 hr at confluent cell density and visualized in phenol-red-free DMEM media (Sigma-Aldrich) with 10% fetal bovine serum (Atlas Biologicals) and 25 mM HEPES (Invitrogen). Live-cell imaging was performed using the Marianas system from Intelligent Imaging Innovations equipped with the MicroPoint FRAP laser system (Photonic Instruments, Inc.). Cytochalasin D (Sigma-Aldrich) or jasplakinolide (Molecular Probes) was applied to cells on the microscope stage to allow comparison between pre- and posttreatment from the same cell population. Kymographs are time (x axis) and intensity profiles (y axis) along (Figures 5 and and7)7) or across (Figure 6) the cell-cell contact. The fluorescence intensity in the kymographs was expressed in a pseudocolor scale (shown in figures) where white and black pixels denote maximum and minimum intensity levels. The intensity profiles were analyzed for the maximum intensity recovery (%) and fitted to a single exponential function up to 2.5 min after photobleaching to extract the half-time of intensity recovery (τ1/2).
Rhodamine-labeled actin (Cytoskeleton) was microinjected using an Eppendorf microinjection system, and cells were observed a few hours later as described above. FITC-labeled actin was also purchased from Cytoskeleton and was introduced into cells as described (Symons and Mitchison, 1991).
We thank J. Engel for the COMP plasmid; S. Craig, J. Lippincott-Schwartz, R. Tsien, and M. Welch for reagents; and S. Fridman for assistance with protein purifications. This work was supported by National Institutes of Health grant R01 GM35527 (to W.J.N.), National Institutes of Health grant R01 GM56169 (to W.I.W.), and a predoctoral fellowship from Boehringer Ingelheim Fonds (to F.D.).
Supplemental Data include Supplemental Experimental Procedures, Supplemental References, 1 figure, and 19 movies and can be found with this article online at http://www.cell.com/cgi/content/full/123/5/889/DC1/.