The Cytoskeletal Protein α-Catenin Unfurls upon Binding to Vinculin*

Background: α-Catenin provides links for cadherin receptors to the actin cytoskeleton at cell-cell adherens junctions. Results: Extensive α-catenin interactions with vinculin are displaced by the vinculin tail domain. Conclusion: α-Catenin-vinculin interactions are stabilized by F-actin. Significance: The data support a new model whereby vinculin activation at adherens junctions is sufficient to stabilize connections of α-catenin with the actin network. Adherens junctions (AJs) are essential for cell-cell contacts, morphogenesis, and the development of all higher eukaryotes. AJs are formed by calcium-dependent homotypic interactions of the ectodomains of single membrane-pass cadherin family receptors. These homotypic interactions in turn promote binding of the intracellular cytoplasmic tail domains of cadherin receptors with β-catenin, a multifunctional protein that plays roles in both transcription and AJs. The cadherin receptor-β-catenin complex binds to the cytoskeletal protein α-catenin, which is essential for both the formation and the stabilization of these junctions. Precisely how α-catenin contributes to the formation and stabilization of AJs is hotly debated, although the latter is thought to involve its interactions with the cytoskeletal protein vinculin. Here we report the crystal structure of the vinculin binding domain (VBD) of α-catenin in complex with the vinculin head domain (Vh1). This structure reveals that α-catenin is in a unique unfurled mode allowing dimer formation when bound to vinculin. Finally, binding studies suggest that vinculin must be in an activated state to bind to α-catenin and that this interaction is stabilized by the formation of a ternary α-catenin-vinculin-F-actin complex, which can be formed via the F-actin binding domain of either protein. We propose a feed-forward model whereby α-catenin-vinculin interactions promote their binding to the actin cytoskeleton to stabilize AJs.

Cell-cell contacts require the formation and stabilization of multiprotein cell membrane complexes coined adherens junctions (AJs). 2 These complexes are directed by homotypic interactions of single-pass transmembrane receptors (cadherins) that have tandem calcium binding ectodomains and short but functionally critical intracellular tail domains (1). Cadherin interactions promote interactions of their tail domains with ␤-catenin, an oncoprotein that both functions as a component of AJs and regulates the activity of the T-cell transcription factor (TCF) family of transcription factors (2,3). Interestingly, the cadherin-␤-catenin complex is hardwired as they bind to one another immediately following their synthesis and are co-transported to the cell membrane as a nascent complex (4). Once there, ␤-catenin binds to the cytoskeletal protein ␣-catenin to form the functional ternary cadherin-␤-catenin-␣-catenin complex, and all three components are required for development and tissue homeostasis (5)(6)(7)(8). Stabilization of AJs had long been thought to rely on the interaction of this ternary complex with F-actin, where increased pools of ␣-catenin at nascent junctions lead to the formation of ␣-catenin homodimers that bind to F-actin via a C-terminal domain (9,10). However, the isolated ternary complex, or these complexes in the context of isolated cell membranes, cannot bind to F-actin (11,12). Further, in these experimental systems, F-actin binding of the ternary complex is not restored by the addition of the cytoskeletal proteins vinculin or ␣-actinin that themselves can bind to F-actin (2). Rather, the function of the ternary AJ complex was suggested to impair the activity of the F-actin nucleating, branching Arp2/3 complex (13,14). Finally, to add more complexity, other ␣-catenin binding partners, specifically Eplin (15) or vinculin (16,17), have been suggested to play important roles in stabilizing AJs, as for example, vinculin knockdown disables cadherin receptor-actin connections and provokes marked reductions in AJs (18).
␣-Catenin is a highly conserved, 906-residue cytoskeletal protein that is composed of four functional domains: (i) an N-terminal ␤-catenin interacting motif; (ii) a vinculin and ␣-actinin interacting domain; (iii) an l-afadin binding module that lies within a dimerization domain; and (iv) a C-terminal domain that can bind to both zonula occludens-1 (ZO-1) and F-actin (19). The crystal structure of the N terminus of ␣-catenin, and of this domain (residues 57-264) in complex with an ␣-helix of ␤-catenin, revealed that this region is a homodimer composed of two antiparallel four-helix bundle domains and that ␤-catenin binds to ␣-catenin via a helix exchange mechanism (20). In addition, the crystal structure of the "M (modulation) fragment" (residues 377-633) (9, 21), which harbors the bind-ing site for l-afadin, established that this domain is composed of a tandem repeat of two antiparallel four-helix bundles that also form a homodimer and are similar in structure, as found in the N terminus of ␣-catenin.
Vinculin is a dynamic 117-kDa globular protein composed of five helix bundle domains that are held in a closed conformation via hydrophobic interactions of the N-terminal seven-helix bundle head (first vinculin head domain, Vh1) with the fivehelix bundle tail (vinculin tail domain, Vt) (22,23). These intramolecular interactions are severed by the binding of amphipathic ␣-helix vinculin binding sites (VBS) of talin or ␣-actinin to the N-terminal seven-helix bundle domain of vinculin, which bury into the four-helix bundle subdomain of Vh1 to form an entirely new five-helix bundle by a mechanism coined helix bundle conversion (22). Binding of these VBSs to vinculin first requires that they unravel from their buried positions in the helix bundles of talin or ␣-actinin, which is provoked by mechanical stress (24,25).
Given the reported roles of vinculin in stabilizing AJs via its interactions with ␣-catenin (26), we solved the crystal structure of the vinculin Vh1 domain in complex with the vinculin binding domain (VBD) of ␣-catenin at 2.7 Å resolution. Remarkably, this structure reveals a novel mechanism of helix bundle interactions where ␣-catenin is bound to vinculin in an unfurled state allowing dimerization. Furthermore, this structure and binding studies support a model whereby simultaneous binding to F-actin functions to stabilize vinculin-␣-catenin interactions and AJs.
Proteins were expressed in Escherichia coli BL21(DE3). Cells were lysed in 20 mM Tris-HCl (pH 8), 500 mM NaCl, and 20 mM imidazole by sonication for 3 min with on/off cycles of 5 s and clarified by ultracentrifugation (100,000 ϫ g; 45 min).
Expressed proteins were purified using a HisTrap chelating nickel affinity column (GE Healthcare) and eluted over a gradient to 500 mM imidazole. Affinity tags were not cleaved for ␣-catenin 82-906 and 144 -906 constructs. Further purification was performed on a Superdex 200 (GE Healthcare) in 20 mM Tris-HCl (pH 8), 150 mM NaCl, and 5 mM DTT and concentrated to about 25 mg/ml.
For crystallizations, cell pellets of His-tagged Vh1 and GSTtagged ␣-catenin VBD were lysed together in 20 mM Tris-HCl (pH 8) and 150 mM NaCl and purified on a GST-Sepharose column (GE Healthcare). The complex was eluted with 15 mM reduced glutathione, GST cleaved with PreScission protease in 20 mM Tris-HCl (pH 8), 150 mM NaCl, 1 mM EDTA, and 1 mM DTT, reloaded onto a GST-Sepharose column to remove the cleaved GST, and further purified on a preparative Superdex-200 26/60 column (GE Healthcare) in 20 mM Tris-HCl and 150 mM NaCl. The complex was concentrated to 27 mg/ml. ␣-Catenin-Vinculin Crystallization-Vh1-VBD complex crystals were obtained by hanging drop vapor diffusion equilibration against a reservoir condition containing 12% (w/v) polyethylene glycol 3350 and 0.3 M sodium acetate. The crystals were flash-frozen in liquid nitrogen and cryoprotected with 25% (v/v) glycerol.
X-ray Data Collection, Reduction, Structure Determination, and Crystallographic Refinement-x-ray diffraction data were collected at the Southeast Regional Collaborative Access Team (SER-CAT) beamline BM22 at the Advanced Photon Source, Argonne National Laboratory and integrated and scaled using autoPROC (29), which uses XDS (X-ray Detector Software) (30) and SCALA (31) as the data reduction engine. The Vh1 domain of the vinculin-IpaA complex (32) was used as the search model to obtain molecular replacement solution with MOLREP (33). The difference electron density map calculated after initial rigid body refinement followed by restrained refinement with BUSTER (34) showed clear electron density for all four ␣-catenin ␣-helices, which enabled the manual building of the remaining model using COOT (Crystallographic Objectoriented Toolkit) (35). The final refinement statistics are detailed in Table 1.
Size Exclusion and Light Scattering-Light scattering (LS) data were collected by size exclusion chromatography using a Superdex 200, 10/300, HR column (GE Healthcare) that was connected to an Agilent 1200 high performance liquid chromatography (HPLC) system (Agilent Technologies, Wilmington, DE) equipped with an autosampler. Elution was monitored using a photodiode array UV-visible detector (Agilent Technologies), a differential refractometer (OptiLab rEx, Wyatt Technology Corp., Santa Barbara, CA) and static and dynamic, multiangle laser LS detector (HELEOS II with Quasi-Elastic-Light-Scattering capability, Wyatt Technology). The size exclusion chromatography-UV/LS/refractive index system was equilibrated in 20 mM Tris (pH 8) and 150 mM NaCl, with a flow rate of 0.5 ml/min. For data collection and analysis, two software packages were used. The ChemStation software (Agilent Technologies) controlled the operation of the HPLC and data collection from the UV-visible detector, whereas the ASTRA software (Wyatt Technology) was used for data collection from the refractive index and LS detectors and to record the 280-nm UV trace from the photodiode array detector. ASTRA software was used to determine the weight average molecular masses across the entire elution profile in 1-s intervals from static LS measurements as described (36).

Structure of Unfurled VBD of ␣-Catenin in Complex with Vh1-
Binding studies have suggested that ␣-catenin residues 327-402 were necessary for binding of ␣-catenin to vinculin (38,39), and more recently, the minimal VBS was determined as residues 325-360 (40) or 326 -377 (41). We determined the ␣-catenin VBD as residues 277-382 based on secondary structure analyses and homology model prediction (42) and co-crystallized the Vh1-VBD complex. The crystals belong to the orthorhombic space group P2 1 2 1 2 with unit cell dimensions of a ϭ 57.2 Å, b ϭ 73.9 Å, and c ϭ 107.4 Å ( Table 1) with one heterodimer in the asymmetric unit (Z ϭ 4) resulting in a solvent content of 59% and a crystal volume per unit of protein molecular weight, V M (43), of 3.03 Å 3 /Da. The Vh1-VBD structure was determined to 2.7 Å resolution. The regions corresponding to Vh1 residues 219 -222 and ␣-catenin 277-289 were not modeled due to lack of discernible electron density, and although the electron density for ␣-catenin residues 354 -361 is weak, the connectivity is unambiguous (supplemental Fig. S1). The final model comprises one Vh1-VBD, nine acetates, and 82 water molecules with a crystallographic and free R-factor of 0.19 and 0.23, respectively (Table 1). All residues fall in allowed regions of the Ramachandran plot as calculated by MolProbity (44) with clash scores of 1.64 (100th percentile) and 7.48 (99th percentile), respectively.
Although two IpaA-VBSs have been shown to simultaneously bind to Vh1, where the second binding site is located between two different ␣-helices (␣5 and ␣6) in the C-terminal 4HB (supplemental Fig. S2B), these were short IpaA-VBS peptides (22 residues) that bind with 2:1 molar ratio (32) and that bury less than 1,300 Å 2 of solvent-accessible surface area. Further, no hydrogen bonds were observed for this interaction (32). In contrast, our Vh1-VBD structure is the first report of an entire helix bundle (116 residues) having good electron density that connects two individual ␣-helices that bind to vinculin. Importantly, in our Vh1-VBD structure, the binding sites for ␣-catenin on both subdomains of Vh1 are located on the surface of full-length vinculin. Thus, these regions are accessible for ␣-catenin binding to the full-length vinculin in its inactive, closed conformation (supplemental Fig. S4).
Two additional ␣-helices of the ␣-catenin VBD seen in our crystal structure, one on each terminus, engage in further and novel interactions with vinculin (supplemental Fig. S5). Although the N-terminal ␣-helix is quite short (residues 293-300) (supplemental Fig. S4) and makes crystal contacts with ␣-helices ␣5 and ␣6 of a symmetry-related Vh1 molecule (supplemental Figs. S3B and S5B), the C-terminal ␣-helix lies almost parallel (at about 15°) to ␣-helix ␣2 from a two-fold related Vh1 molecule (Fig. 2B, supplemental Fig. S4). The resulting dimer seems further stabilized by a disulfide bond that is formed between Cys-324 and its symmetry-related mate (Fig.  2, A and C). Although ␣-catenin is predominantly monomeric but also found as a dimer in solution (53), we found that our Vh1-VBD complex, as well as unbound VBD, is either monomeric or dimeric in solution (Fig. 3). Furthermore, the elution of VBD is consistent with a compact globular domain in solution. Thus, the VBD unfurls upon binding to vinculin in accord with calorimetry data showing that ␣-catenin (residues 273-510) bound to the VH domain (residues 1-718) with nM affinity and that the binding was endothermic and entropy-driven, consistent with a major conformational change (39). Collectively, our data are consistent with the notion that ␣-catenin unfurls to bind to vinculin, allowing dimerization.
Vinculin-␣-Catenin Binding Is Facilitated by Severing Vinculin Head-Tail Interaction-To avoid potential nonspecific effects on the binding, we assessed the binding properties of untagged ␣-catenin and vinculin proteins. Because deletion of residues 1-81 increases the stability of ␣-catenin (20), we thus cloned ␣-catenin residues 82-906 (hereafter referred to as fulllength) for our binding studies. ␣-Catenin was reported to bind weakly to full-length vinculin (16,39). Indeed, only a fraction of full-length ␣-catenin and vinculin formed a complex in solution, as seen in our native gel shift assays (Fig. 4A). Thus, fulllength ␣-catenin in its inactive conformation is not sufficient to activate vinculin. However, the minimal ␣-catenin-VBS is sufficient for severing the vinculin Vh1-Vt interaction to form a FIGURE 2. ␣-Catenin-vinculin complex crystal structure. Vh1 is shown in cyan or green, and its respective bound ␣-catenin is shown in white or magenta. A, ␣-catenin Cys-324 forms a disulfide bond with a two-fold related molecule. Likewise, the C-terminal ␣-catenin ␣-helix binds to the second Vh1 ␣-helix of a two-fold related heterodimer burying over 500 Å 2 of solvent-accessible surface area with a shape correlation statistic derived using the CCP4 program SC (55) of over 0.8 for this interface, a significant value where a value of 1 indicates a perfect fit versus 0.35, which indicates the mismatch of an artificial association. Vh1 ␣-helices are labeled, as well as some ␣-catenin residues 324 and 373. B, close-up view of the dimer interface. Vinculin residues are shown in cyan, and ␣-catenin are in magenta as in A and in a slightly rotated view from A. C, space-filling representation of the dimer of heterodimers (same color coding as shown in A) and rotated about 180°about the horizontal axis of the A image. Some contact residues are indicated.
(␣-catenin-⌬FABD, residues 82-634), can form a complex with a vinculin head-tail chimera (Fig. 5B) lacking the vinculin Vt2 domain shown to keep vinculin in its closed conformation (56) and F-actin (Fig. 5C), indicating that F-actin binding blocks the ability of Vt to displace ␣-catenin. Furthermore, ␣-catenin activates vinculin lacking the Vt2 domain. Collectively, these findings suggest that F-actin stabilizes the ␣-catenin-vinculin interaction by binding to the C terminus of ␣-catenin and/or vinculin.

DISCUSSION
Stabilization of AJs to the actin cytoskeleton is necessary for maintenance of cell-cell contacts (17), yet the mechanism by which this occurs is not resolved. One clear contributor to the response is ␣-catenin, which is necessary for the formation and stabilization of AJs, for recruitment of vinculin to these sites, and for linking cadherin complexes to the actin cytoskeleton (16 -19). ␣-Catenin also controls branched actin polymerization (53) and directly associates with F-actin during junction reinforcement (40). Our findings suggest that a complex interplay of vinculin, ␣-catenin, and the F-actin network are together necessary to stabilize AJs.
The structure of the vinculin Vh1-␣-catenin VBD complex shows that ␣-catenin unfurls upon complex formation, allowing dimerization (Fig. 2), and that it has a novel vinculin binding mode, where two ␣-helices in the VBD bind simultaneously to the four-helix bundle subdomains of the Vh1 domain. Further, the ␣-catenin binding sites are fully accessible in full-length vinculin (supplemental Fig. S4), and the entire globular head VH domain readily binds to ␣-catenin (Fig. 4, C-E). However, the two full-length proteins interact only weakly in solution (Fig. 4A), and competitive binding studies support a model whereby vinculin in its open conformation binds to F-actin and then binds to and unfurls ␣-catenin to form a stable ternary complex.
Vinculin recruitment by ␣-catenin is force-dependent (40). However, the intramolecular head-tail interaction of vinculin impairs ␣-catenin-vinculin interaction. Furthermore, the formation and/or stabilization of the ␣-catenin-vinculin complex is greatly stabilized by the binding of either partner to F-actin through their C-terminal domains. Here proximity rules may come into play, where F-actin-bound ␣-catenin promotes the association of vinculin with F-actin as vinculin is a highly dynamic protein and exists in equilibrium between its closed and open states. Collectively, these findings suggest that cytoskeletal tension stabilizes the ␣-catenin-vinculin complex at AJs.
␣-Catenin deletion mutants that retain the VBD but cannot bind to the cytoskeleton can recruit vinculin to AJs, which led to the hypothesis that a region C-terminal of the VBD prevents ␣-catenin binding to vinculin and that this inhibition is relieved when ␣-catenin is bound to the actin cytoskeleton (40). Furthermore, binding studies with truncated proteins suggested the hypothesis that ␣-catenin activates vinculin (41). While ␣-catenin binds vinculin weakly in native gel assays, our competition binding assays show that ␣-catenin readily binds to VH, yet is displaced by Vt. Thus, the nanomolar affinity of the intramolecular VH-Vt interaction prevents ␣-catenin binding and vinculin needs to be in its open state to bind to ␣-catenin. Collectively, the data indicate that ␣-catenin does not activate vinculin but rather that F-actin bound-vinculin unfurls and binds to ␣-catenin to stabilize AJs.
What is still not resolved is what trigger activates vinculin at AJs. One possibility is that this response simply relies on an equilibrium of vinculin in its open versus activated state, where . Binding in each case was saturable with no unbound ␣-catenin remaining. C, the ␣-catenin⌬FABD-vinculin complex binds F-actin via Vt. Left gel, lanes 1 and 2, F-actin is found in the pellet (P). Lanes 3 and 4, the Vh1-Vt chimera (asterisks) does not aggregate and remains in the soluble fraction (S). Lanes 5 and 6, the Vh1-Vt chimera does not bind to F-actin. Right gel, lanes 1 and 2, the Vh1-Vt chimera in complex with ␣-catenin⌬FABD does not aggregate. Lanes 3 and 4, the Vh1-Vt chimera in complex with ␣-catenin⌬FABD binds F-actin via Vt. Lanes 5 and 6), ␣-catenin⌬FABD does not aggregate. Lanes 7 and 8, ␣-catenin⌬FABD does not bind to F-actin. ␣-cat or ␣, ␣-catenin (residues 144 -906); F-act or F, F-actin. Molecular weight markers are shown for the left gel.
␣-Catenin-Vinculin Complex Crystal Structure the latter can bind to F-actin that then facilitates interactions with ␣-catenin. Alternatively, vinculin may bind to other partners at AJs to augment this response, for example, to the Arp2/3 complex, as Arp2/3-directed actin polymerization is suppressed by ␣-catenin (11).