Two Distinct Head-Tail Interfaces Cooperate to Suppress Activation of Vinculin by Talin*

Vinculin is autoinhibited by an intramolecular interaction that masks binding sites for talin and F-actin. Although a recent structural model explains autoinhibition solely in terms of the interaction between vinculin tail (Vt) and residues 1–258 (D1), we find an absolute requirement for an interface involving the D4 domain of head (Vh residues 710–836) and Vt. Charge-to-alanine mutations in Vt revealed a class of mutants, T12 and T19, distal to the V-(1–258) binding site, which showed increases in their Kd values for head binding of 100- and 42-fold, respectively. Reciprocal mutation of residues in the D4 domain that contact Vt yielded a head-tail interaction mutant of comparable magnitude to T19. These findings account for the approximately 120-fold difference in Kd values between Vt binding to V-(1–258), as opposed to full-length Vh-(1–851). The significance of a bipartite autoinhibitory site is evidenced by its effects on talin binding to Vh. Whereas Vt fails to compete with the talin rod domain for binding to V-(1–258), competition occurs readily with full-length Vh, and this requires the D4 interface. Moreover in intact vinculin, mutations in the D4-Vt interface stabilize association of vinculin and talin rod. In cells, these head-tail interaction mutants induce hypertrophy and elongation of focal adhesions. Definition of a second autoinhibitory site, the D4-Vt interface, supports the competing model of vinculin activation that invokes cooperative action of ligands at two sites. Together the D1-Vt and D4-Vt interfaces provide the high affinity (∼10–9) autoinhibition observed in full-length vinculin.

been ascribed to vinculin. Disruption of vinculin expression in mice results in an embryonic lethal phenotype, with severe cardiac and brain abnormalities (1). Moreover, heterozygotes show increased susceptibility to stress-induced cardiomyopathies (2), consistent with the paralysis and defects in muscle architecture associated with the knockout of vinculin in Caenorhabditis elegans (3). Interestingly, vinculin-null cell lines show defects in mechanical stiffness (4), Rac-mediated lamellipodial protrusion (5), cell shape, and spreading on fibronectin (1,6,7), as well as misregulation of apoptotic cues (8).
Whereas mechanistic insights into the molecular basis of these phenotypes are limited with a few notable exceptions (7,8), the aforementioned effects are generally consistent with the idea that vinculin functions as a mechanical linker between the plasma membrane and actin cytoskeleton through cross-linking several adhesion proteins. This model derives from in vitro studies that identified interactions between vinculin and numerous focal adhesions proteins including talin (9), ␣-actinin (10,11), paxillin (12), VASP (13), vinexin/ponsin family members (14,15), and F-actin (16,17). However, the binding of intact vinculin to these potential ligands is highly restricted by conformational regulation of vinculin structure. An autoinhibitory interaction between the vinculin head and tail domains directly masks binding sites for talin (18), F-actin (17), and vinexin 2 and has been correlated with changes in the vinculin affinity for ␣-actinin (11), VASP (19), and Arp2/3 (7). This has led to a proposed model of vinculin function in which cell adhesion transduces a signal that disrupts the autoinhibitory interaction and permits engagement of the vinculin cytoskeletal targets within the focal adhesion (17).
Recent crystallographic studies on the autoinhibited conformation of vinculin have revealed that the principal binding site for head-tail interaction is comprised of a large hydrophobic interface between the D1 domain of vinculin (residues 1-258) and the top of the five helix bundle structure comprising the vinculin tail (20 -22). Importantly, this interface is disrupted upon binding of a talin-derived peptide (VBS3) to the D1 domain. Binding of VBS3 to D1 results in a dramatic conformational change, termed helical bundle conversion, which cannot be competed by the addition of excess tail domain (20). This observation led to a proposed mechanism for vinculin activation in which binding of talin serves as the driving force behind conformational change (20,23). In fact, vinculin is activated by a talin-related peptide, pVR (24), or VBS3 (25), but only at concentrations of 500 -1000-fold molar excess to vinculin. In light of the fact that the D1-V t interface is nearly identical in the bimolecular complex (20) and in the structure of full-length vinculin (21,22), the unexpectedly high concentration of pVR or VBS3 required to activated full-length vinculin (24,25) has ¶ Recipient of an American Heart Association postdoctoral fellowship.
been interpreted by us as evidence that additional intramolecular components functionally lower the K d of the head-tail interaction below 10 Ϫ9 (22). The crystal structure of intact vinculin led us to speculate that the high affinity autoinhibited state derived either from the intramolecular nature of the D1⅐V t complex, or from cooperativity between multiple interdomain interaction sites, but we previously could not assess the relative contribution of these two factors.
Here we present results of biochemical and mutagenesis studies that support a model in which the autoinhibited state of vinculin is achieved by two separate interfaces; one of these comprises the D1-V t interface seen in the structures of chicken and human vinculin (22,21), and the second comprises the D4-V t interface observed in the chicken vinculin structure (Fig.  1A). We find that D1 binds weakly to V t in solution, with a K d near 10 M, whereas V h -(1-851) binds with a K d of ϳ0.1 M. We then show that the increased affinity of the autoinhibitory complex is attributable to the D4-V t interface. This analysis redefines the autoinhibitory V h domain as D1-D4, rather than D1 (20) or D1-D3 (22), and enables us to provide a detailed model for autoinhibition that accounts for the functional insufficiency of talin for activating native vinculin (22).
Site-directed Mutagenesis-The clustered charge-to-alanine mutations in pET15b/V t and the N773A;E775A mutations in pET15b/V h -(1-851) were introduced by QuikChange PCR (Stratagene) per the manufacturer's instructions. QuikChange PCR was also utilized to introduce the T12, T8/19, and N773;E775A mutations into pEGFP/vinculin. The entire vinculin cDNA coding region in these plasmids was then verified to ensure that only the desired mutations were introduced.
Protein Expression and Purification-pET expression vectors were transformed into either BL21(DE3) Gold or RIL(DE3) Codon Plus competent cells (Stratagene). For expression of YFP-tagged proteins, bacterial cultures were grown to an OD 600 of 0.6 in LB medium ϩ 1% glucose at 37°C, and then induced with 0.5 mM isopropyl-1-thio-␤-Dgalactopyranoside for 16 h at 20°C. His-tagged YFP-V h , YFP/V1-258, and V t were isolated using nickel affinity chromatography with His-Bind resin (Novagen) as described previously (24). YFP fusion proteins were cleaved with biotinylated thrombin (Novagen) to remove their N-terminal His 6 tags. Proteins were dialyzed into Fusion Protein Storage (FPS) buffer (10 mM Tris, pH 7.5, 100 mM NaCl, 0.1 mM EDTA, 0.1 mM EGTA, 0.02% azide, 0.1% ␤-mercaptoethanol), supplemented with protease inhibitor mixture (PICS) (27), and stored at 4°C. Proteolytic V h derived from V8 cleavage of chicken gizzard vinculin was prepared as described previously (18). His-tagged talin rod was purified as described previously (28). To enrich for full-length talin rod, the eluate from the His-Bind column was purified over DE52 resin using conditions described for purification of native talin from chicken gizzard (29).
Screen for Head Binding Mutants in V t -Binding of proteolytic V h to recombinant His-tagged V t mutants, 0.1 M each, was assayed as described in the figure legends. V t was recovered quantitatively on His-Bind resin, and the V h remaining in the supernatant determined by Micro-BCA (Pierce).
YFP Fluorescence Supernatant Depletion Assay-YFP fusion protein (V h or D1) at 0.2 M was incubated with His-tagged V t as indicated in the figure legends. Complexes were bound to NTA beads, and the fluorescence remaining in the supernatant was assayed by fluorometry.
Fluorometry and Data Analysis-Fluorescent samples were assayed on a FluoroMax-3 fluorometer (Horiba Jobin Yvon, Edison, NJ) equipped with a thermoregulated sample compartment (21C). YFP was excited at 490 nm, and peak emission monitored at 527 nm, with excitation and emission slit widths of 3 and 5 mm, respectively. The fraction of YFP-V h or YFP-D1 bound was measured as a function of change in the 527-nm emission relative to the maximum fluorescence observed in mock V t -treated controls. Free V t concentrations were calculated from the total V t concentration less the concentration of V h or D1 bound (autoinhibitory complexes have a 1:1 stoichiometry). Bound V h was plotted against free V t concentrations in Kaleidograph and K d values computed by non-linear fitting of the Scatchard equation: Actin Binding Assay-G-actin was purified as described (30,31). V t binding to F-actin was assayed by a cosedimentation assay as described in the figure legends. Complexes were separated on 15% acrylamide, 0.8% bis-acrylamide Laemmli gels, and V t in the pellet fractions quantitated by densitometry of Coomassie Blue staining.
Ternary Complex Pull-down Assay-Complex formation between V t -(884 -1066), D1-(1-258), and V-(259 -850) was screened with a His-Bind resin pull-down assay as described in the figure legends. Bound material was eluted from the resin in 20 l of Laemmli sample buffer containing 0.5 mM EDTA. Pellet and supernatant fractions were assayed on 12% acrylamide, 0.8% bis-acrylamide SDS gels, and proteins were detected by Coomassie Blue staining.
Native Gel Assay-V t was assayed for the ability to disrupt preformed complexes of YFP-V h (WT or N773;E775A mutant) and talin rod, as described in the figure legends. Complexes were resolved on a 7.5% acrylamide native gel, using a 40 mM triethanolamine, 50 mM Bicine, pH 8.3, running buffer. YFP fluorescence was then imaged on the Alpha Innotech system (Alpha Innotech Corp, San Leandro, CA) and intensity measurement and analysis performed in NIH Image software.
Cell Culture and Transfection-HEK293 cells were provided by Dr. Peter Devreotes at Johns Hopkins University. Vinculin-null cells were a generous gift from Dr. Eileen Adamson, The Burnham Institute. Both cell lines were cultured in DMEM ϩ 10% fetal calf serum (Hyclone, Logan, UT). For transfection, HEK293 cells were plated at 2.5 ϫ 10 6 cells on gelatin coated 10-cm culture dishes and grown overnight. Each plate was transfected with 3 g of DNA using Lipofectamine Plus (Invitrogen) in serum-free DMEM for 4 h. Vinculin-null cells were plated on poly-L-lysine-treated coverslips coated with 20 g/ml fibronectin at 0.25 ϫ 10 6 cells. Cells were transfected with 1 g of DNA using Lipofectamine Plus in serum-free DMEM for 4 h. At the end of the incubation, cells were briefly rinsed with DPBS and refed with growth media for 24 -48 h.
Pull-down Assays in Cell Lysates-Hypotonic lysates of HEK293 expressing GFP-vinculin (WT or mutant) were assayed for binding to talin rod or F-actin as described in the figure legends. Pellet fractions were resolved on a 7.5% acrylamide, 0.8% bis-acrylamide SDS-PAGE gel and transferred to nitrocellulose for Western blotting.
Immunoblotting-Western blots were developed with ECL reagents (Amersham Biosciences) per the manufacturer's instructions. 8d4 antitalin, hVin1 anti-vinculin, and horseradish peroxidase-conjugated IgG antibodies were obtained from Sigma. C4 anti-actin was a gift from Dr. Jim Lessard. Polyclonal rabbit anti-V t (Cottontail sera) was raised against His-tagged chicken V t -(884 -1066). All monoclonal antibodies were used at a 1:5000 dilution. Anti-V t was used at 1:2500.
Microscopy-Transfected cells were fixed as described in the figure legends. Coverslips were mounted in an anti-fade solution containing 12% (w/v) Elvanol and 0.12% paraphenylenediamine. Images were acquired on an upright Axioskop (Carl Zeiss, Germany) microscope attached to a SensiCam (CookeCorp, Tonawanda, NY) CCD camera using the IPLab software package from Scanalytics (Fairfax, VA). Digital images were imported into Adobe Photoshop for figure preparation.

Screen for Head Binding Mutants of V t -Prior
to determination of three-dimensional structural models, analysis of the vinculin primary sequence suggested a role for electrostatic interactions in the head-tail interaction. First, the pIs of the head and tail domains, 5.5 and 9.6, are complementary in nature, and second, a high proportion of lysine and arginine residues in vinculin tail are conserved evolutionarily. Based on these observations, we screened for head binding mutants of V t using a charge-to-alanine mutagenesis scheme outlined in Fig.  1B. Twenty charge-to-alanine mutations were designed in windows of five to seven amino acids along V t and introduced into pET15b/V-(884 -1066).
The ability of the tail mutants to bind V h -(1-851) was screened in pull-down experiments measuring loss of V h in response to His-tagged tail domains immobilized on nickelcharged agarose beads ( Fig. 2A). The large majority of mutations had little or no effect on head binding, suggesting that the mutations were not globally disruptive to tail structure and that specific side chain interactions played a more important role than overall charge character of V t . The mutagenesis was highly successful in defining a subset of mutants affecting head binding, namely T8 (K944;R945A), T12 (D974;K975;R976; R978A), T14 (K996;K1002A), and T19 (K1047;R1049;D1051A). When the structure of V t was determined (32), we saw that these mutations implicate two structurally distinct head binding regions on V t (Fig. 2B); namely a basic patch at the base of the helical bundle, defined by T12 and T19, and a cluster on the faces of helices 3 and 4 comprised of T8 and K996 of T14 (Fig.  2B). Interestingly, we also found that the effect of combining mutations from these two candidate head binding sites, such as in the T8/19 double mutant, potentiated the head binding defect, consistent with biochemically distinct contributions of the two interaction surfaces ( Fig. 2A). These mutations appear to define specific residues involved in the head-tail binding site, as they have only minimal effects on the ability of V t to cosediment with F-actin filaments (Fig. 2C). Quantitative Dissection of the Head-Tail Interaction Surface-The clustering of strong head-tail interaction mutations at the base of V t is unexpected given the recently proposed structural model for the autoinhibitory head-tail interaction, in which the top of the V t bundle exclusively contacts the D1 domain in vinculin head, and this interaction is proposed to account for the high affinity, autoinhibited state of vinculin (20). To investigate the significance of the head binding mutants of V t , we decided to compare the D1 domain (residues 1-258) with V h -(1-851) for binding to V t . We developed a quantitative, solution phase assay to measure K d values for head-tail interaction, using a fluorescent supernatant depletion of YFP-tagged V h by His-tagged V t . The results show that the interaction between D1 and V t is relatively weak in solution, with a K d of 11.5 M (Fig. 3B). In stark contrast, the observed K d for YFP-V h is 93 nM (Fig. 3A), slightly above the 50 nM binding reported for proteolytic V h (18), suggesting that the presence of a YFP tag has only modest effects on head-tail binding. Although it has been reported that D1 binds V t with nanomolar affinity in a solid-phase binding assay (33), such assays introduce systemic error in K d measurement by failing to account for the high local concentration effect of ligand immobilization. Also, solid phase assays can increase apparent affinities of protein-protein interactions by reducing protein flexibility upon immobilization, thus lowering entropic costs of binding.
The 120-fold difference in K d observed between the D1⅐V t (11.5 M) and V h ⅐V t (93 nM) complexes is best reconciled by the existence of an auxiliary head-tail interaction site. Therefore, we tested whether the T12-T19 surface on V t could account quantitatively for the additional binding energy manifest in the full-length V h ⅐V t complex. Using the fluorescence supernatant depletion assay, we measured the binding affinities of the tail mutants for V h . As shown in Table I, the T19 and T12 mutations increase the K d for head binding by factors of 42-and 100-fold, respectively. In the case of the most severe mutant, T12, the overall loss of affinity closely predicts the difference in K d values observed in Fig. 3B (Table I). In contrast, the T8 and T14 mutations, which cluster near the structurally defined D1-V t interaction site, exert relatively mild effects of 6 -8-fold increases in the observed K d for V h binding. This result is consistent with the overall hydrophobic character of this interface as described in the co-crystal of D1 and V t (20) and establishes function for the contacts noted between Lys 944 , Arg 945 (T8), and D1 in that structure.
Definition of the Role of the D4 Domain in Head-Tail Interaction-The mutagenic studies predicted the existence of an auxiliary tail binding site in V h that would make specific contacts with the T12/T19 surface on V t . Upon solution of the crystal structure of intact vinculin (22), we were able to apply structural insights to the nature of the T12/T19 recognition domain in V h . In this structure of intact vinculin, the base of V t makes a small, but well ordered contact with the D4 domain of V h (residues 710 -836). As seen in Fig. 4A, this interaction surface is characterized predominantly by hydrogen bonding and salt bridge interactions. Three of the four residues in T12 (Asp 974 , Lys 976 , and Arg 978 ) make direct side-chain contacts with the D4 domain, and Lys 975 appears to be structurally important for the presentation of the other T12 side chains. In T19, Lys 1047 and Arg 1049 are peripherally associated with this interface, consistent with the smaller magnitude of their binding defect relative to T12.
Two side chains in the D4 domain of V h , asparagine 773 and glutamate 775, appeared to participate in the principal interactions with V t (Fig. 4A). Therefore, to directly test the role of D4 as a putative second V t binding site, we designed a N773; E775A double mutant in which both side chains were mutated to alanine, and measured its affinity for V t in the fluorescence   Fig.  3, several of the head-tail interaction mutants were assayed quantitatively for loss-of-binding activity for YFP-V h . Binding defects are expressed in terms of K d values as well as fold increases relative to wild-type V t . K d values were derived from non-linear fits using the Scatchard equation, assuming that the binding curves reach saturation points equivalent to wild-type V t . assay. As shown in Fig. 4B, the N773;E775A mutation interferes strongly with head-tail interaction, increasing the K d to 6.3 M, a 68-fold change relative to wild-type V h . The N773; E775A mutant demonstrates conclusively the requirement for the D4-T12 interface to mediate high affinity interaction of the head-tail domains. Given that the D4 mutant behaves in a nearly identical fashion to the isolated D1 domain, it appears that the D4-V t interface is sufficient to account for the difference in binding affinity of full-length V h and D1 for V t . Interestingly, attempts to directly measure the K d of a bimolecular D4⅐V t complex using isolated D4 were unsuccessful. Several GST fusion constructs spanning vinculin residues 259 -850 failed to mediate a direct interaction with V t in the absence of D1 domain. The interaction could be verified, however, as a ternary complex; GST/V259 -850 was competent to bind to V t in the presence of D1 domain (Supplementary Fig. 1). This interaction appeared to be V t -dependent, as D1 did not separately interact with GST/V-(259 -850). It is conceivable that this second binding site is structurally unstable in the absence of D1, or that thermodynamic coupling to the D1-V t interaction is required to offset costly entropic components to binding. This may account for why it has not been detected by conventional mapping assays defining the head-tail interaction (34).

The D4-V t Interface Is Essential for Regulation of the Talin Binding Site in V h -
The functional consequences of tail binding do not apply equally to the D1 domain and full-length V h , according to published findings. Whereas it has been reported that talin or a talin-related peptide, pVR, bind in a competitive fashion with V t to V h (18,24), more recent studies on the D1⅐VBS3 complex show that V t fails to displace VBS3 from D1 (25,20). We therefore tested whether these differences could be accounted for by the D4-dependent high affinity bimolecular interaction with V t in full-length V h . Using a native gel assay, we tested the ability of V t to compete talin rod from YFP-tagged V h , either wild-type or bearing the N773;E775A mutation.
In the case of wild-type YFP-V h , 75% of the fluorescent signal comigrated with talin rod in the absence of V t (Fig. 5, lane 2); however, at 2-fold molar excess V t to talin rod, the V h ⅐V t complex predominates and less than 33% of the fluorescence remains in the complex with talin rod (Fig. 5, lane 3). Competition is maximal at 4-fold molar excess (lane 4 versus 5). This is consistent with roughly equivalent K i values for V t (400 nM) and talin-related peptide (300 nM) in solution-phase assays, as previously reported (24). Although the N773;E775A mutation did not perturb binding to talin rod (lane 7), it did attenuate the competition activity of V t versus talin rod. At 2-fold molar excess V t to talin rod, 56% of the YFP fluorescence of the N773;E775A mutant head remained associated with talin rod in a slow migrating complex (lane 8). Moreover, V t and V h (N773;E775A) did not form a complex that was stable enough to be maintained during migration through the gel (note the accumulation of free V h (N773;E775A) as V t is titrated to higher concentrations, lanes 8 -10). Thus in the context of the high affinity bimolecular interaction, V t is sufficient to block talin binding to the D1 domain. These results provide an additional line of evidence that the tightly autoinhibited state of native vinculin is conferred by the full-length head, residues 1-851, as defined structurally by limited proteolysis (35).
Disruption of the Auxiliary Binding Site Partially Opens Full-length Vinculin-The results of the competition experiments could explain the mechanism by which head-tail interaction can effectively suppress talin binding in the context of full-length vinculin (22). However, interpretation of our experiments with reconstituted head-tail complex was limited by assumptions of native-like structure in recombinant vinculin fragments, and by the lack of intramolecular constraints present in the intact molecule. Therefore, it was important to ad- Competition proceeded for an additional hour at room temperature, after which 10 l of 50% glycerol was added to each tube. Complexes were resolved on 7.5% native polyacrylamide gel run at 120 V at 4°C. YFP-fluorescence was visualized using a CCD camera/Alpha-Innotech system. Note that V t efficiently displaces wild-type V h from talin rod, but V t is a much weaker competitor for V h (N773;E775) and fails to form a stable, tightly migrating head-tail complex.
dress the functional relevance of the proposed second head-tail binding site in the context of full-length vinculin. One prediction of our competition studies was that disruption of the auxiliary binding site would leave vinculin susceptible to activation by talin, given that the residual D1⅐V t complex would have an expected K d of 0.1 M or higher (assuming intramolecular binding provided a maximum of 100x binding enhancement). We therefore introduced the D4 and T12 mutations into GFPtagged vinculin, and assessed vinculin activation in HEK293 lysates using a pull-down assay with His-tagged talin rod (talin-(397-2541)). As shown in Fig. 6A, wild-type vinculin does not bind appreciably to talin (lanes 1-4), demonstrating the functional insufficiency of talin to activate vinculin. In contrast, the T12 (lanes 9 -12) and N773;E775A (lanes 5-8) mutants display a dose-dependent increase in binding to talin, from 0 -1 M talin rod. The increased binding to talin rod reflects only a partial opening of vinculin; however, as the actin binding site remained fully masked in these mutants. As shown in Fig. 6C, vinculin cosedimenting with F-actin (lanes  17, 19, 21) is equivalent to baseline level of vinculin sedimentation (lanes 16,18,20) in this assay. Since the affinity of the vinculin-actin interaction is ϳ1 M (17), it is consistent that the D1-V t autoinhibitory interface is sufficient to inhibit actin binding in intact vinculin, whereas the 10 -300 nM affinity of the talin-vinculin interaction (9,25,36) requires the bipartite high affinity head-tail complex to achieve full autoinhibition. Thus, native vinculin is normally in a high affinity autoinhibited state in which both talin and F-actin binding is prevented, but the presence of the T12 and N773;E775A mutations create a low affinity autoinhibited state, which is selectively able to interact with talin but not F-actin. These results confirm the importance of the D4-T12 interface in intact vinculin and demonstrate a functional correlation between strength of the headtail interaction and ligand binding.
Activated Vinculin Induces Assembly of Hypertrophied Focal Adhesions-Next, we asked whether the increased association of talin with the head-tail interaction mutants had functional significance for organization of the F-actin cytoskeleton. Talin has been proposed to be an important mediator for transmembrane linkages between integrins and the actin cytoskeleton within the focal adhesion (37,38). One potential consequence for increased talin-vinculin association might therefore be alterations in the structure of focal adhesions. To investigate this possibility, we transfected vinculin-null cells with GFP-tagged vinculin cDNA containing the head-tail interaction mutations. Activation of vinculin correlated with a loss of diffuse, cytosolic vinculin staining, and enrichment in focal adhesions. In cells expressing N773;E775A, T12, or T8/19, focal adhesions were more numerous, with particular abundance in central regions of the cell (Fig. 7). Most strikingly, however, we observed a dramatic elongation of focal adhesions in cells expressing high levels of activated vinculin mutants, T12, T8/19, compared with wild-type controls. This fibrillar adhesion phenotype was limited to the most severe head-tail interaction mutants; its absence in cells expressing N773;E775A, T8, or T19, may derive from a stronger basal level of autoinhibition in these mutants. Thus disruption of head-tail regulation leads to both increased association with talin and subsequent alteration of focal adhesion morphology. Cells were then lysed with a Dounce homogenizer and insoluble material removed by centrifugation (16,000 ϫ g, 10 min, 4°C). Lysates were normalized to equivalent GFP-vinculin levels based on GFP fluorescence. A, 95 l of lysate were supplemented with imidazole to 10 mM final and His-tagged talin rod to concentrations of 0.1, 0.3, or 1.0 M. Binding proceeded 30 min at room temperature. Vinculin-talin complexes were recovered on NTA-agarose, washed twice in phosphatebuffered saline, and analyzed by SDS-PAGE and Western blotting using monoclonal anti-vinculin (hVin1) or anti-talin (8d4) antibodies. Note that although talin, up to 1.0 M, fails to bind any detectable wild-type EGFP or endogenous vinculin (lanes 2-4), the N773;E775A (lanes 6 -8) and T12 (lanes 10 -12) mutants co-fractionate with talin in a dose-dependent manner, indicating enhanced availability of the talin binding site in V h . Lanes 1, 5, and 9 represent no talin controls. B, lanes [13][14][15] show that equivalent levels of GFP-vinculin were present for WT, N773;E775A, and T12. Expression of the GFP-vinculin constructs did not cause any changes in the level of endogenous vinculin. C, N773; E775 and T12 mutant are unaffected for autoinhibition of the actin binding site. 1.5 M F-actin was polymerized in cell lysates by the addition of KCl to 100 mM and MgCl 2 to 2 mM final concentrations. After 1.5 h of polymerization, F-actin was pelleted in an airfuge (28 psi, 30 min). Pellets were resolved by SDS-PAGE followed by Western blotting. Under these conditions, some vinculin sediments in the no actin controls (lanes 16, 18, 20). The addition of F-actin (lanes 17,19,21) produced no increase in the amount of vinculin that could be sedimented, indicating that the actin binding site in vinculin is fully masked. D, as a positive control for actin binding in the HEK293 lysates, ectopic His-tagged V t was added to a final concentration of 0.1 M. The endogenous HEK293 actin is only present in the monomer pool, and V t does not sediment on its own (lanes 22 and 23). However, at least 90% of V t cosediments with F-actin under the conditions of this assay (lane 25 versus 24), showing that V t efficiently competes with other components of the lysate in the cosedimentation assay and the absence of specific binding of vinculin to F-actin in C is a function of autoinhibition.

DISCUSSION
We have found that the autoinhibitory site in vinculin is bipartite, involving contacts not only between D1 and V t , as described by Izard et al. (20), but also a previously uncharacterized interaction between D4 and V t . This additional autoinhibitory site provides binding energy sufficient to account for an approximately hundred fold reduction in the bimolecular K d for head-tail association over that seen for the D1⅐V t complex (see Fig. 8A). Functionally, this increased binding affinity strongly shifts the conformational equilibrium of V h toward the closed state, thereby suppressing the ability of putative activators, such as talin, to displace V t from D1 (Fig. 8C). These findings establish a link between vinculin and a select class of autoinhibited cytoskeletal molecules, typified by N-WASP, in which multiple intramolecular contacts work cooperatively to maintain the inactive state.
The presence of a bipartite autoinhibitory site has two key implications for biological function. The first is the ability to build a tightly closed/autoinhibited state from relatively low affinity sites, by thermodynamic coupling of multiple weak intramolecular interactions. Our results show that the 10 Ϫ5 M D1⅐V t bimolecular complex is transformed to a 10 Ϫ7 M V h ⅐V t complex by the presence of an auxiliary site, which has as little as 10 Ϫ2 M binding (Fig. 8A). This high theoretical value for the K d of D4-V t interaction is consistent with the absence of a detectable bimolecular interaction between the isolated D4 and V t domains in pull-down assays, the small surface area buried in this contact, 540 Å 2 (22), and the presence of a limited number of bonding interactions, largely electrostatic in nature (Asp 718 -Arg 976 , Glu 775 -Arg 978 ). Nonetheless, in the context of a bimolecular complex of V h -(1-851) with V t , once V t is bound to D1, the local concentration of residues at the D4-T12 interface will be quite high, and thus the auxiliary site serves as an effective intramolecular latch to stabilize the autoinhibited state of the V h ⅐V t complex. In the full-length molecule, intramolecular tethering of V h and V t is also likely to contribute substantially to enhanced affinity of the autoinhibitory interaction (Fig. 8B). Indeed estimates of the K d for the head-tail interaction in full-length vinculin are below 10 Ϫ9 (22).
The second noteworthy consequence of the D4-T12 interface is that the "molecular logic" encoded by a bipartite autoinhibitory site is fundamentally different from the binary activation switch in molecules that contain a single inhibitory site (39,40). As illustrated by several experiments on N-WASP, the presence of two inhibitory contact sites requires the combinatorial input of two activators, such as PIP 2 and Cdc42 (39,41), PIP 2 and Nck (42,43), or Cdc42 and Toca-1 (44). In this way, N-WASP functions as a "logical AND gate" which integrates input from two signal transduction pathways into a single output, Arp2/3 driven actin polymerization (45).
In analogous fashion, it is possible to envision two signals converging upon vinculin through ligand binding at both the D1-V t , and at the D4-V t interfaces. Integrin-based adhesion can serve to cluster talin and ␣-actinin (46), both of which have been implicated in binding and activation of the D1 domain of vinculin (20). However, likely candidates to disrupt the D4-V t interface remain elusive. A strong hypothesis is that RhoA signaling or contractile force may be involved in vinculin activation at the D4-V t interface. This would have the net effect of restricting vinculin activation to zones of highly concentrated mechanical force at the membrane, a model supported by the linear relationship between vinculin localization and intracellular tension (47,48). Although PIP 2 is an attractive Rhoresponsive (49) element proposed to regulate vinculin activation (32,34,50), we remain skeptical that an interaction that is displaced by 15 mM NaCl (51) or that otherwise requires micelles of acidic phospholipids containing at least 25% PIP 2 (24), is significant in a physiological context. Although PIP 2 may function cooperatively with other activators, PIP 2 binding sites in talin (52) and ␣-actinin (53) can allow pure PIP 2 micelles to bridge vinculin with ligands in the absence of direct proteinprotein interaction. Explicit treatment of this bridging phenomenon is lacking in the literature, and thus the definitive experiments on the role of PIP 2 as a co-activator of vinculin remain to be done.
Elucidation of a physiological activation mechanism for vinculin remains an important goal. In light of the findings presented here, it is critically important that competitors for the head-tail interaction be evaluated in the context of tail binding to full-length V h -(1-851), and tested for efficacy in full-length vinculin. Elegant work on the interaction of the talin VBS3 peptide with the vinculin D1 domain (1-258) has provided both the biochemical proof (25,20) and the structural basis (20) for the mutual exclusivity of talin versus V t binding to D1. However, the weak D1-V t interaction permits competition by talin VBS3, whereas autoinhibition in full-length vinculin renders the protein refractile to talin or pVR (a talin-related peptide) binding as shown here and in previous work (18,22,24). In disrupting the D4V t interface, the T12 and N773;E775A mutations show a gain of function activity for talin binding. These findings illustrate that the inability of talin rod to activate wild-type vinculin is not merely a consequence of enhanced binding of D1 and V t in an intramolecular context, but rather results from the bipartite nature of the head-tail interaction. Thus, studies that represent the autoinhibitory interaction exclusively as the D1⅐V t bimolecular complex are likely to overestimate the ability of potential activators to induce conformational changes in full-length vinculin.
Although the precise nature of a role for talin in activation of vinculin remains controversial, there is little doubt that if vinculin incorporates into focal adhesions in an activated conformation, it will be competent to associate with talin. In light of this, the ability of V t to disrupt existing complexes of V h with talin rod becomes significant in terms of the in vivo function of head-tail interaction. If the VBS3-D1 paradigm held in focal adhesions, activation of vinculin would be extremely stable, because V t cannot compete for VBS3-bound D1 (20). Disruption of vinculin-talin complexes would require active cellular mechanisms, independent of autoinhibitory contacts, to dissociate FIG. 7. Vinculin mutants defective in head-tail interaction induce hypertrophied focal adhesions. Vinculin-null cells grown on fibronectin-coated coverslips in DMEM ϩ 10% fetal calf serum were transfected with pEGFP/vinculin (WT, N773;E775A, T12, or T8/19 mutants) using Lipofectamine Plus. Cells were fix-permeabilized in icecold 4% paraformaldehyde ϩ 0.1% Triton X-100 for 3 min, followed by a 20 min of 4% paraformaldehyde post-fix. Images were obtained on an Axioskop microscope at ϫ63 magnification. Shown are representative cells from each population. Note the increased number of central adhesions in the head-tail interaction mutants. In the most severe mutants, T12 and T8/19, the central adhesions adopt a highly elongated morphology.
these proteins during adhesion plaque remodeling. However, in the context of the bipartite autoinhibitory interaction found in full-length vinculin, V t can effectively disrupt the interaction with talin; thus inactivation of vinculin through reassociation of the head-tail complex may be an important mechanism for regulating vinculin activities in the focal adhesion. This inactivation may be antagonized or delayed by post-translational modification of V t , such as the Src-dependent phosphorylation that disrupts the bimolecular head-tail interaction (54). The change in focal adhesion morphology in cells expressing the head-tail interaction mutants suggests that disruption of headtail interaction traps vinculin in complexes with ligands. Indeed, preliminary results indicate that increased association of vinculin and focal adhesion components interferes with turnover (exchange of bound and free pools) of vinculin, which ultimately leads to hypertrophy of the adhesion (55).
In summary, we have proposed that spatial clustering of ligands at focal adhesions provides a mechanism for combinatorial activation of vinculin if the thermodynamic additivity of binding energies exceeds the affinity of the intramolecular head-tail complex (22). Here we provide compelling biochemical evidence for that model by demonstrating that high affinity autoinhibition of vinculin is achieved by a bipartite autoinhibitory site comprised of D1-V t and D4-V t contacts. Most importantly, we show that the D4-V t interface functions to mask the talin binding site in D1, providing a mechanistic basis for the inability of talin to suffice as a primary activator of native vinculin (18,22). Combinatorial activation of vinculin is feasible only in so far as ligands are able to overcome kinetic barriers to binding the closed conformation of vinculin. Thus, it is noteworthy that the D4-V t interface is largely accessible to solvent; its orientation appears to lie directly opposite to the hinge domain, and it is not sterically occluded by other regions of V h (22). In light of these findings, we believe that the most likely mechanism for combinatorial activation of vinculin will involve coordination of specific competitors of the D1-V t and D4-V t autoinhibitory sites. . The D4-V t interface (K d 3 ) is calculated to contribute 10 Ϫ2 to binding, a value corroborated by the observed change in bimolecular V h ⅐V t complex formation when the D4-V t interface is mutagenically disrupted (N773;E775A, T12, and T19 mutants). B, intramolecularly, D4 and V t are tethered by the polyproline hinge domain (curved line in figure). As in A, dashed lines between V t and V h domains indicate bonding interactions. The intramolecular contribution of the D4-V t interface is estimated to be comparable to the bimolecular context (K d 3 ). The complete V h -V t intramolecular interaction (K d 4 ) is estimated to be below 10 Ϫ9 , based on experimental data using pVR or VBS3 as an activator (22,24,25). Intramolecular association of D1 and V t (K d 5 ) is enhanced 100 times over that seen in the bimolecular context (K d 1 ). C, we propose two autoinhibited conformations of vinculin. The tightly closed state for native vinculin cannot interact with talin or F-actin, and would require combinatorial ligand binding to achieve activation. Alternatively, the tightly closed state could be activated by an as yet undiscovered ligand whose K d for vinculin binding is lower than the activation threshold (1 nM). Talin alone suffices as an activator only when vinculin is in a low affinity autoinhibited state, in which the D4-V t interface is disrupted by mutagenesis (N773;E775A or T12) or by a putative co-activating ligand binding to D4 or the base of V t . The reported K d of actin for V t is about 1 M (17); the K d of talin or VBS3 for V h is 0.01-0.3 M (9, 25, 36).