Coincidence of Actin Filaments and Talin Is Required to Activate Vinculin*

Vinculin regulates cell adhesion by strengthening contacts between extracellular matrix and the cytoskeleton. Binding of the integrin ligand, talin, to the head domain of vinculin and F-actin to its tail domain is a potential mechanism for this function, but vinculin is autoinhibited by intramolecular interactions between its head and tail domain and must be activated to bind talin and actin. Because autoinhibition of vinculin occurs by synergism between two head and tail interfaces, one hypothesis is that activation could occur by two ligands that coordinately disrupt both interfaces. To test this idea we use a fluorescence resonance energy transfer probe that reports directly on activation of vinculin. Neither talin rod, VBS3 (a talin peptide that mimics a postulated activated state of talin), nor F-actin alone can activate vinculin. But in the presence of F-actin either talin rod or VBS3 induces dose-dependent activation of vinculin. The activation data are supported by solution phase binding studies, which show that talin rod or VBS3 fails to bind vinculin, whereas the same two ligands bind tightly to vinculin head domain (Kd ∼ 100 nm). These data strongly support a combinatorial mechanism of vinculin activation; moreover, they are inconsistent with a model in which talin or activated talin is sufficient to activate vinculin. Combinatorial activation implies that at cell adhesion sites vinculin is a coincidence detector awaiting simultaneous signals from talin and actin polymerization to unleash its scaffolding activity.

Vinculin (V), 3 a 116-kDa soluble protein, is recruited to membrane-associated protein complexes that link the actin cytoskeleton to the extracellular matrix or neighboring cells (1). These adhesive sites have an organization visible in the light microscope and include skeletal and cardiac muscle costameres (2)(3)(4), cardiocyte intercalated discs, smooth muscle dense plaques, the zonula adherens between epithelial cells (5), myotendinous junctions (4), as well as focal adhesions and focal complexes in lamellipodia of migrating cells in culture (1). Cell adhesion structures consist of three domains; a transmembrane receptor such as an integrin, cadherin, or IgCam, extracellular matrix proteins or cell surface proteins that bind to the extracellular portion of these receptors, and a cytoplasmic plaque containing signaling and cytoskeletal proteins, such as focal adhesion kinase, Src, vinculin, talin, and actin, that assemble on the cytoplasmic domain of the receptors and regulate adhesion site turnover and strength. Cell adhesion is essential for motility in embryogenesis, wound healing, inflammation, and metastasis; therefore, it is important to discover the regulatory mechanisms.
Previous work shows that vinculin both strengthens cell adhesions and regulates signaling. Nematodes that lack vinculin fail to develop connections between muscle myofibrils and the sarcolemma, arresting at the 2-fold stage of development with muscles that do not contract (6). Mice heterozygous at the vinculin locus are predisposed to stress-induced cardiomyopathy and show abnormal intercalated discs and misaligned Z lines (7). In humans, a vinculin mutation in Vd2 has been linked to hypertrophic cardiac myopathy (8), and mutations in metavinculin, the muscle-specific form of vinculin (9), have been linked genetically to both dilated cardiomyopathy (10,11) and hypertrophic cardiomyopathy (12). Mice that are null for the vinculin gene die at embryonic day ϳ9.5 with massive defects in brain, heart, and somite development (13). Although vinculin is not required for viability of embryonic cells in culture or for formation of focal adhesions in tissue culture cells, it increases the size of focal adhesions and regulates cell spreading, motility, and adhesion to extracellular matrix (13)(14)(15). Activated vinculin transiently recruits Arp2/3 to adhesive sites in the leading edge of spreading cells where new dendritic actin polymerization is required (16), strengthens the linkages of integrins to extracellular matrix (17)(18)(19), and lengthens the residence time of talin in focal adhesions (20). In some cells vinculin is required for the effects of Rac on lamellipodia formation (21) and for signaling pathways employing paxillin (22) and PTEN (23) and has tumor suppressor-like action on cell growth and motility (24,25).
Vinculin likely regulates cell adhesion by forming complexes with other proteins because no enzymatic activity for vinculin has been reported. Purified domains of vinculin bind talin, ␣-actinin, ␥-catenin, F-actin, paxillin, VASP, Arp2/3, ponsin, vinexin, Cbl-associated protein, and acidic phospholipids with K d values in the 10 Ϫ5 -10 Ϫ7 M range (26). The established roles of these vinculin ligands in integrin function, actin dynamics, cell adhesion, and cell motility suggest that vinculin regulates these same processes by forming complexes with these ligands. Importantly, the protein-protein interactions required are not available to purified vinculin in solution because a high affinity (K d Ͻ 10 Ϫ9 ) intramolecular interaction between proteolytic head (1-851) and tail (852-1066) domains of vinculin (27) inhibits binding of most of these ligands (16,(27)(28)(29)(30)(31)(32). Therefore, to understand how vinculin works to regulate adhesion, motility, and transduction of force across cell membranes, it is essential to learn how it is autoinhibited and how it is activated.
Recent structural and biochemical studies of vinculin show how the protein is maintained in its autoinhibited state (33)(34)(35)(36). Vinculin is maintained in its high affinity autoinhibited state by cooperation between two much lower affinity interfaces, Vd1/Vt and Vd4/Vt (34,36) (Fig. 1). This inhibition is so strong that talin rod (Tn rod) cannot bind vinculin even when both proteins are present in solution at 14 M (34). These findings suggested that two ligands might cooperate to activate vinculin by simultaneously disrupting both low affinity interfaces (34,36). Indeed, whereas vinculin is autoinhibited and cannot bind talin, vinculin point mutants with 40 -100-fold reduced affinity of head-tail interaction (HTI) bind talin in cell lysates (36), recruit talin to ectopic sites in living cells, and prolong the residence time of talin in focal adhesions (20). The HTI mutants are still autoinhibited with respect to binding F-actin when that is the only ligand presented, consistent with the much greater strength of the remaining HTI compared with the affinity of Vt for F-actin (36). These results are both predicted by and consistent with the combinatorial activation model but do not provide direct proof of that mechanism.
In contrast to the combinatorial activation hypothesis for vinculin activation, crystal structures of Vd1 and VBS3 (33) and of the 4-and 5-helix domains of talin that contain the VBS peptides (37)(38)(39) suggest an alternative model in which talin undergoes activation to present a VBS peptide to vinculin (40). In this model activated talin is capable by itself of activating vinculin. A variety of biochemical evidence has also been obtained to support this model (33, 40 -42).
In this paper we demonstrate activation of vinculin by talin and actin filaments, whereas neither ligand alone is competent. We then present measurements of the affinities of vinculin, Vd1, and Vh for Tn rod domain in solution and of VBS3 for vinculin, Vh, and Vd1, which support the conclusion that reduction of intramolecular head-tail interaction in vinculin is the rate-limiting structural change required for vinculin and talin interaction. We interpret the structures of vinculin (34,35), Vd1/VBS3 (33), Vt-F-actin (43), and of talin domains (37)(38)(39) to provide a rationalization for combinatorial activation by talin and actin filaments. Next, we reconcile the two different models for vinculin activation by showing that whereas published data is not necessarily inconsistent with the combinatorial activation model, some evidence for the combinatorial model is inconsistent with the talin or ␣-actinin alone model. Finally, we present our view of the potential significance of combinatorial activation of vinculin in cells.

EXPERIMENTAL PROCEDURES
Reagents and Proteins-Actin was extracted from chicken skeletal muscle acetone powder, processed through one cycle of polymerization and depolymerization, and gel-filtered through a Sephadex G-150 column according to Pardee and Spudich (44). Recombinant His 6 -tagged chicken vinculin was purified by sequential affinity chromatography on His-Bind resin, DE52 anion exchange chromatography, and Mono S chromatography. Recombinant His 6 -tagged vinculin 1-258 and His 6 -tagged vinculin 1-851 were purified by affinity chromatography. Recombinant His 6 -taggged Tn rod was purified by sequential affinity chromatography, Mono Q chromatography, and Sephacryl S-200 gel filtration. TAMRA 5 (6)-aminohexanoic acid-KELIESARKVSEKVSHVLAALQA-NH 2 (TAMRA-VBS3) was synthesized at the PeptideSynthesis Facility at Johns Hopkins.
Fluorescence Depletion Assay-TAMRA VBS3 at 100 nM was incubated with His-tagged vinculin 1-258, His-tagged vinculin 1-851, or His-tagged vinculin at concentrations from 0 to 800 nM (in 3.75 mM Tris, pH 7.5, 0.9 mM KH 2 PO 4 , 5 mM Na 2 HPO 4 , 125 mM NaCl, 1.7 mM KCl, 10 mM imidazole, 0.0375 mM EDTA, 0.0375 mM EGTA, 0.1 mg/ml bovine serum albumin, 0.1% ␤-mercaptoethanol, 0.0075% NaN 3 , and 0.1% Triton X-100) for 1 h at room temperature in the dark. His tagged protein was captured using Ni-NTA agarose (Qiagen) with inversion for 30 min at room temperature followed by centrifugation at 13,000 ϫ g for 5 min at room temperature. The fluorescence of unbound TAMRA VBS3 in the supernatant was assayed with a Fluoromax-3 spectrofluorimeter (Jobin Yvon) at 21°C. TAMRA emission was recorded at 582 nm with excitation at 555 nm. The integration time was 0.2 s. The excitation and emission slit widths were 3 and 5 mm, respectively. The amount of unbound TAMRA-VBS3 was derived from its fluorescence intensity in the linear range.
The dissociation constant (K d ) defined in Equation 1 was transformed to the quadratic form (Equation 2). K d was obtained by curve-fitting the data to Equation 2 using Kaleidagraph (Synergy Software). In these equations, R and L denote total concentration, R f and L f denote concentration of unbound, and RL denotes the concentration of complex.
Cell Culture and Transfection-Cells were cultured on 0.1% gelatin-coated tissue culture plates in DMEM with high glucose and glutamine (Media Tech; catalog # 10 -017-CV), supplemented with 10% fetal calf serum in a 5% CO 2 incubator at 37°C. HEK 293 cells were seeded on 0.1% gelatin-coated 100-mm dishes at 3 million per plate; transfection was performed the next day with 3 g of plasmid DNA using Lipofectamine/Plus reagent (Invitrogen). HEK 293 cells were lysed 2 days after transfection. Vin Ϫ/Ϫ murine embryonic fibroblasts were seeded on 35-mm tissue culture dishes coated with 20 g/ml fibronectin at 300,000 cells; transfection was performed the next day with 1.5 g of plasmid DNA using Lipofectamine/ Plus reagent.
Cell Spreading Assay-Vin Ϫ/Ϫ were transfected with GFP, GFP-vinculin, or GFP-vinculin-A50I. Twenty-four hours posttransfection ϳ400,000 cells in Dulbecco's modified Eagle's medium (MediaTech 10-017-CV) containing 10% fetal calf serum were seeded into 6-well plates containing coverslips coated with polylysine and then 250 l of 20 g/ml of human fibronectin. Cells were incubated at 37°C. At 2 and 3 h after plating cells were fixed in 4% paraformaldehyde in phosphatebuffered saline (PBS) for 20 min, washed twice with PBS, and mounted on a slide with Prolong Gold antifade reagent (Molecular Probes). Areas of transfected cells were measured using the segmentation and quantitation tools in IPLab (Scanalytics). Multinucleated cells and cells that had not spread at all were excluded from the analysis.
FRET Assay of Cell Lysates-Cell lysate was prepared as described (49). The emission spectrum of fluorescent proteins in the lysate was acquired with a Fluoromax-3 spectrofluorimeter (Jobin Yvon) maintained at 21°C. CFP and FRET emissions were traced from 460 to 600 nm with excitation at 440 nm. YFP fluorescence was measured at 525 nm with excitation at 490 nm. The increment was 1 nm, and the integration time was 0.2 s. The excitation and emission slit widths were 3 and 5 mm, respectively. The concentration of head probe in cell lysates was quantified from its YFP fluorescence using purified YFP for generating a standard curve (0 -100 nM).
Lysate of HEK 293 cells expressing HP3 was treated first with either 5 M G-actin or G buffer (2 mM Tris, pH 8.0, 0.2 mM CaCl 2 , 0.5 mM dithiothreitol, and 0.2 mM ATP) and then with increasing concentrations of Tn rod or VBS3. A series of buffer controls was performed with lysates containing HP3 treated first with G buffer and then with the sequential addition of equivalent amounts of either Tn rod storage buffer (10 mM Tris, pH 7.5, 100 mM NaCl, 0.1 mM EDTA, 0.1 mM EGTA, and 0.02% NaN 3 ) or VBS3 storage solution (water). The emission ratio (fluorescence at 525 nm/fluorescence at 475 nm upon excitation at 440 nm) was obtained, and the raw change in emission ratio (raw ⌬ER) was calculated by subtracting the emission ratio after each treatment from the emission ratio without treatment. The series of buffer controls generates a slight apparent change in raw ⌬ER. The raw ⌬ER due to dilution was then subtracted from each raw ⌬ER obtained from the treatment with Tn rod or VBS3 to obtain the real change in emission ratio (⌬ER).
The real change in emission ratio was converted to the concentration of activated vinculin or ligand-bound vinculin assuming a linear relationship between the two; ⌬ER ϭ m2 ϫ RL, and m2 is a linear coefficiency factor. The maximum change in emission ratio can be extrapolated from curve-fitting the data by Equation 3, a transformed form of Equation 2 and ⌬ER ϭ m2 ϫ RL. The maximum change in emission ratio is the multiplication of m2 and total concentration of HP3 or R. m2 is 0.00195, and HP3 is 100 nM in this study. Fig. 1 defines nomenclature of the proteins, their domains, and derivative peptides that are used in the following work and depicts the relationship of structure to the biochemical findings that we present. SDS gel analysis of the recombinant proteins used is shown in supplemental Fig. S1.

RESULTS
To assay activation of vinculin, we developed a FRET probe (HP; Fig. 1) that reports on binding of the Shigella virulence protein, IpaA, to vinculin. The binding of IpaA to vinculin is a high affinity interaction that results in exposure of vinculin actin binding activity (50). We characterized HP using spectrofluorimetry of lysates prepared from HEK 293 cells expressing HP. HP has a strong FRET signal that is unaffected by the addition of 5 M actin filaments (Fig. 2a). The addition of a submaximally activating amount of IpaA to HP reduces the FRET, and the addition of actin filaments to this IpaA-treated vinculin further reduces the FRET (Fig. 2, a, c, and e), indicating a synergistic effect. To confirm that loss of FRET reports on activation of vinculin, we assayed co-sedimentation of HP with actin filaments. HP was found in the high speed pellet only when both IpaA and actin filaments were present in the lysate (Fig. 2b). Significantly, actin alone neither induces a FRET change nor binds to the FRET probe (Fig. 2, a-c), whereas IpaA alone both binds to the FRET probe (Fig. 2d) and induces a FRET change (Fig. 2, a, c, and e) that signifies exposure of the actin binding activity (Fig. 2b). These data show that loss of HP FRET reports on structural changes that accompany exposure of the actin binding activity of vinculin, i.e. vinculin activation.
In subsequent experiments we use HP3 in which EYFP was replaced with monomeric (A206K) (51) citrine (47), and ECFP was replaced with monomeric cerulean (48) to give HP3, which we anticipated would perform better in live cell FRET studies. 4 The change in fluorescent proteins was made to improve brightness, stability to photobleaching at physiological pH and chloride concentration, and rate of folding at 37°C. These changes affect only the fluorescent proteins and do not affect the sequence of vinculin or of the linkers between vinculin and the GFP variants. HP3 localizes in focal adhesions, rescues cell spreading and lamellipodia formation in vin Ϫ/Ϫ embryonic fibroblasts, and responds similarly to HP to activation by sub-saturating amounts of IpaA Ϯ actin (supplemental Fig. S2, a-d).
Having established that loss of FRET in HP reports on conformational changes that occur when vinculin is activated, we used the directly analogous HP3 to assay the ability of Tn rod domain to activate HP3 in the presence and absence of actin filaments. Neither actin filaments nor Tn rod alone was able to activate HP3. But in the presence of 5 M actin filaments, Tn rod induced dose-dependent activation of vinculin, detected by loss of FRET (Fig. 3a) and confirmed by co-sedimentation of the probe with the F-actin (Fig. 3b). Although Tn rod contains the conserved actin binding I/LWEQ module at its C terminus (52, 53), recombinant Tn rod or Tn rod prepared by proteolysis of gizzard talin co-sedimented poorly with actin under the conditions used for the FRET assay (supplemental Fig. S3). This is because the pH (7.5), ionic strength (100 mM KCl added to lysate), and Tn rod fragment itself are unfavorable for demonstrating the interaction of talin and actin (54 -56). Because Tn rod does not bind actin significantly in the conditions used for the FRET experiments, our co-sedimentation results show that activated vinculin can indeed bind Tn rod and actin filaments at the same time.
Others have proposed that talin must be activated before it can interact with vinculin (40,41). To assess whether activated talin can activate HP3 in the absence of actin filaments, we used the talin peptide, VBS3, to mimic the postulated activated form of talin (40 -42). Strikingly, even at 10 M, VBS3 alone was unable to activate HP3. However, in the presence of 5 M F-actin VBS3 induced dose-dependent activation of HP3 (Fig. 4, a and b). The stoichiometry of VBS3 binding to vinculin is 1:1; therefore, we can estimate a K d for VBS3 activation of vinculin in the presence of actin of ϳ481 Ϯ 26 nM. As a specificity control for both the FRET and co-sedimentation assays, we incorporated the talin binding point mutation, A50I (20, 34), into HP3. VBS3 was unable to activate HP3-A50I even in the presence of actin filaments (Fig. 4, a and c).
The inability of VBS3 alone to activate HP3 is explained by the relative affinities of VBS3 for vinculin, Vd1, or Vh when all reactants are in solution phase. Using TAMRA-VBS3, we measured dissociation constants for VBS3/Vd1 and VBS3/Vh of ϳ85 nM, whereas binding of VBS3 to vinculin could not be   (20,34). d, Vh, vinculin residues 1-851, corresponding to proteolytic head domain (76). e, Vd1, vinculin residues 1-258, corresponding to the talin binding subdomain of vinculin head (61,76). f, VBS, a vinculin binding sequence (61) composed of ϳ25 amino acids that forms an amphipathic basic helix in the context of talin 4-helix bundles (37)(38)(39). g, Vd1-VBS, structure of VBS3 bound to Vd1 (PDB code 1RKC) (33). The green space-filled residue marks the N terminus of Vd1. Note the large increase in distance between the first and second N-terminal helices, colored green, when VBS binds (33). h, schematic structure of talin consisting of the N-terminal FERM, central rod, and C-terminal I/LWEQ domains. Eleven VBSs, marked in red, are found within Tn rod domain (residues 397-2541) (59). detected (Fig. 5a). The assay permits accurate detection of 7% of TAMRA-VBS3 in complex. Because we did not detect 7% complex formation using 800 nM vinculin and 100 nM VBS3, we can calculate from the quadratic form of the binding equation that the K d of VBS3 for vinculin must be Ͼ10 M. These data show that whereas the VBS3 binding site on Vh and Vd1 is readily accessible, the VBS3 binding site on vinculin is inaccessible and, therefore, cryptic.
To determine whether autoinhibition of vinculin is the primary regulatory event in talin-vinculin interaction, we assessed the interaction of Vh and Tn rod and found a K d of ϳ100 nM (Fig. 5b), assuming one binding site as indicated by a Hill plot of the data. 4 In contrast, no complex could be detected between 14 M vinculin and 14 M talin by isothermal titration calorimetry (34). If the K d for complex formation is 14 M, at least 38% complex should have been detected in this assay, as estimated from the quadratic form of the binding equation (Equation 3). Equivalent amounts of total sample before spin (T), supernatant (S), and pellet (P) fractions were subjected to SDS-PAGE and immunoblotted (IB) with hVIN1 and C4 monoclonal antibodies (Sigma) to vinculin and actin, respectively. Head probe co-sedimented with actin filaments only when IpaA was also present, demonstrating that loss of FRET induced by IpaA reports on exposure of the actin binding activity of the probe. Doublet bands in head probe reflect SDS-resistant structure in GFP proteins causing aberrant migration in SDS gels (49). c, the mean FRET efficiencies of HP before and after IpaA and/or actin treatment were obtained as described in Chen et al. (49). Error bars are S.E., n ϭ 3. d, IpaA binds to head probe. HEK 293 cell lysate containing head probe was mixed with a colloidal suspension of IpaA or buffer for 30 min at room temperature. IpaA and its bound proteins were pelleted at 16,000 ϫ g for 10 min. Equivalent amounts of supernatant and pellet were resolved by SDS-PAGE followed by blotting with monoclonal hVIN-1 for vinculin. e, IpaA alone can activate vinculin. HEK lysates containing HP were supplemented with 5 M of G-actin or G buffer and incubated for 1 h at room temperature to allow actin polymerization.
HPϩ actin alone is represented by the first point (zero IpaA) in the curve labeled HP ϩ actin ϩ IpaA. Then the reactions were treated with sequential addition of IpaA up to 1350 nM. At low concentrations of IpaA, F-actin, and IpaA acted synergistically to activate vinculin. But at the highest concentration IpaA alone was capable of inducing the maximum FRET change observed for IpaA in the presence of actin filaments. Equivalent amounts of total sample before spin (T), supernatant (S), and pellet (P) fractions were subjected to SDS-PAGE and immunoblotted (IB) with 8d4, hVIN1, and C4 monoclonal antibody to talin, vinculin, and actin, respectively. Activation of vinculin also occurs when Tn rod is held at 1 M and F-actin is varied from 0 -9 M. 4 These data show that when Vt is separated from Vh and is free to diffuse away, interaction of Vh with Tn rod domain occurs readily and with high affinity. But in vinculin the intramolecular effect on the affinity of the HTI blocks interaction with talin.

DISCUSSION
The mechanism of vinculin activation is controversial; for a recent review, see Ref. 26. In this report we have provided direct evidence for combinatorial activation of vinculin by talin and actin acting simultaneously to reduce HTI, thereby permitting stable ternary complexes between vinculin, talin, and actin filaments to accumulate. By drawing on previous structural and biochemical work we can rationalize why and how such activation occurs. From the structures of vinculin it can be seen that the head domain of vinculin forms a vise that clamps the tail domain at the top and bottom of the five-helix bundle (Fig. 1). It is clear that the major interface between Vh and Vt is at the Vd1/Vt juncture (34,35). This interface is a typical proteinprotein interaction domain; it buries Ͼ2000 Å 2 and has a substantially hydrophobic center and polar perimeter. Significantly, the binding constant for Vd1 and Vt is only ϳ10 Ϫ5 M, whereas the K d for Vh and Vt is ϳ10 Ϫ7 M (36), and the autoinhibition of vinculin is estimated at K d Ͻ 10 Ϫ9 . In addition to the Vd1/Vt interface, there are several minor interfaces, Vd4/Vt and Vd3/Vt, that bury only a few hundred Å 2 and are largely min. Equivalent amounts of total sample before spin (T), supernatant (S), and pellet (P) fractions were subjected to SDS-PAGE and immunoblotted (IB) with hVIN1 and C4 monoclonal antibodies to vinculin and actin, respectively. c, vinculin mutant A50I, defective in talin binding, was not activated by VBS3 and actin. Lysate from HEK 293 cells expressing HP3-A50I was mixed with either actin (5 M) alone or actin and VBS3 (10 M). Actin co-sedimentation assay was performed under the same conditions as in b. HP3-A50I did not co-sediment with actin filaments even when both VBS3 and actin filaments were present. FIGURE 5. Binding constants of vinculin, Vh, and Vd1 for VBS3 and Tn rod. a, VBS3 binds tightly to Vd1 and Vh but not to vinculin in solution. TAMRA-VBS3 at 100 nM was incubated with 0 -800 nM His 6 -tagged Vd1, Vh, or vinculin for 1 h at room temperature and then incubated with Ni-NTA agarose. The unbound TAMRA-VBS3 was separated from bound fraction by centrifugation and assayed by spectrofluorimetry. The concentration of free TAMRA-VBS3 was plotted against the total concentration of His-tagged vinculin or its domains. Error bars represent S.D. from two independent experiments. b, Tn rod has high affinity for Vh. YFP-Vh at 100 nM was incubated with 0 -800 nM His 6 -tagged Tn rod for 1 h at room temperature and then incubated with Ni-NTA agarose. The unbound YFP-Vh was separated from the bound fraction by centrifugation and assayed for concentration by spectrofluorimetry. The fraction of YFP-Vh in complex with Tn rod was plotted against unbound Tn rod. Data were fitted to the equation m1 ϫ x/(m2 ϩ x) using Kaleidagraph software, where m1 ϭ maximum fraction of receptor capable of binding to ligand, and m2 ϭ K d . Error bars represent S.D. from two independent experiments. polar (34). Although such small polar interfaces are not found to be protein-protein interaction domains in the crystallographic data base (57), biochemical studies show that the Vd4/Vt interface is essential for vinculin high affinity autoinhibition. In the context of vinculin, where the intramolecular effect is present, i.e. Vt is not free to diffuse away from Vh, mutagenesis data shows that the Vd4/Vt interface provides the additional binding energy to maintain vinculin in the tight autoinhibited conformation (20,36).
How Might Combinatorial Activation of Vinculin Work?-Two models come to mind. In the absence of external input, the equilibrium dissociation constant for HTI, estimated at Ͻ10 Ϫ9 (34) dictates that there will be a small number of vinculin molecules in which Vt is dissociated from Vh (Fig. 6). Talin will bind readily to Vh that is not in contact with Vt, as shown here. But because the K d for binding to talin is only ϳ100 nM, talin will dissociate from Vh ϳ100-fold more readily than Vt will dissociate from Vh, assuming that dissociation is rate-limiting in both cases. Furthermore, because dissociated Vt remains tethered to Vh, it has a kinetic advantage in reassociating with Vh after the talin leaves. Consequently, very little talin-vinculin complex will accrue. But if at the same time that talin binds to Vh, F-actin binds to Vt, then reassociation of head and tail will be reduced, allowing formation of a relatively stable ternary complex, as shown in the experiments reported here.
Alternatively, because the binding sites for the talin peptide (33-35) and F-actin (43) are partially exposed on the autoinhibited structure of vinculin and because the tight HTI is generated by the synergistic effect of two low affinity interfaces (ϳ10 Ϫ5 and ϳ10 Ϫ2 ) that are coupled by the intramolecular context, high concentrations of talin and actin might bind with very low affinity (undetectable with assays used here) to the partially exposed sites. When HTI dissociates transiently, the simultaneous presence of weakly bound talin and actin gets converted to strong binding at the now fully exposed binding sites.
For the previous arguments, it is important to note that although the binding site on Vd1 for talin peptides (the groove between helix 1 and 2) is visible on the surface of the vinculin structure (33), it is too narrow to accommodate the peptide (Fig. 1, a and e-g). Substantial expansion of the distance between helices 1 and 2 (compare Fig. 1, e and g) and movements of helices 3 and 4 occur to accommodate the talin peptide in Vd1 (33). In our studies the talin peptide alone is incapable of inducing these rearrangements in vinculin as shown by the lack of TAMRA-VBS3 binding to vinculin and inability of VBS3 alone to induce FRET change in HP3. In contrast, the peptide can readily induce the rearrangements required to bind to Vh or Vd1. To explain these data, we propose that the interactions of Vt with helices in Vd1 (33,34) in the context of the entire vinculin molecule restrict the molecular motions required to allow the groove between helix 1 and 2 of Vd1 to expand (33). When Vt is dissociated, these helices are free to move. Indeed purified Vd1 is conformationally dynamic (i.e. is sensitive to proteases), whereas the VBS3-containing domain of talin is stable until Vd1 binds, after which the protease stabilities of the two domains reverse (39). These considerations enable us to understand how the VBS binding site can be visible on the structure of vinculin yet unable to bind the ligand.
In both scenarios for combinatorial activation of vinculin, release or reduction of HTI is the gated event for interaction of vinculin and talin. Although it is clear from x-ray and NMR data that the structure of isolated talin domains is disrupted when Vd1 binds (37)(38)(39), it is not obvious that the structure of talin must change before vinculin can bind. Our analyses show that reduction of vinculin HTI (e.g. by removal of Vt, as done here or, less drastically, by point mutations at the Vh/Vt interface, which reduce HTI by 40 -100-fold (20,36)) is sufficient to permit vinculin to engage Tn rod and talin. Moreover, the nearly identical high affinity K d values that we find for Vh binding to Tn rod and for Vh binding to VBS3 suggest that the initial site on Tn rod for binding to free Vh is easily accessible. Interestingly, one of the VBS sequences in talin may be more easily accessed by released Vh as a result of low stability of the helix bundle in which it resides (58).
A Competing Model for Activation of Vinculin Holds That Talin Itself Is Sufficient to Activate Vinculin-Comparing the crystal structure of a Vd1⅐VBS complex to the crystal structure of Vd1 and Vt (33) shows that the talin peptide induces structural rearrangements of the Vd1 N-terminal four-helix bundle that disrupt the binding interface of Vt and Vd1. The 11 VBS sequences in talin (59) are amphipathic basic helices, and those that have been studied in situ by crystallography or NMR comprise a segment of a 4-or 5-helix bundle (37)(38)(39). The hydrophobic face of the VBS peptides are buried in the helix bundles. But in crystals of VBS peptides bound to Vd1, the hydrophobic face of the VBS is buried in the hydrophobic core of a Vd1 FIGURE 6. A model illustrating how the combination of talin and actin filaments can activate vinculin, whereas neither ligand alone is competent. Blue, Vd1; cyan, Vd2; green, Vd3; yellow, Vd4; red, Vt; orange, proline-rich region. Native vinculin in solution is tightly autoinhibited (ϳK d Ͻ 10 Ϫ9 ). This tight autoinhibition is achieved through two relatively low affinity contacts between head and tail domain (Vd1:Vt and Vd4:Vt) that are amplified by the intramolecular effect (that is, the reactants are not free to diffuse away) (36). The equilibrium constant dictates that some small fraction of vinculin will be in the non-inhibited state. Because free Vh binds readily to talin, a talin-vinculin complex can form. But talin affinity for Vh is only ϳ10 Ϫ7 M, and when talin dissociates, it is free to diffuse. Tail domain will rapidly reclose the vinculin, and little talin-vinculin complex can accrue. However, if actin, a ligand for tail domain, is present at the same time that head domain binds talin and if actin binding to Vt reduces head-tail interaction, then a talin-vinculin-actin complex can accrue, as it does in our experiments. Theoretically, there are other ligand combinations that might activate vinculin. Both phosphatidylinositol 4,5-bisphosphate binding to Vt (77) and phosphorylation of Vt (78,79) have been shown to block bimolecular interactions of Vh and Vt. Possibly, these effectors can also cooperate with talin to activate vinculin. Interestingly, phosphatidylinositol 4,5-bisphosphate (PIP2)-bound Vt cannot bind actin filaments directly (62), suggesting a different role for PIP2-Vt interaction in vinculin biology, perhaps as a lipid sensor (80) involved in regulating focal adhesion turnover (80,81). (33). Therefore, it was further proposed that talin has to be activated (40) to expose the VBS sequence and that this sequence is then sufficient to activate vinculin (42). A similar proposal was made for the VBS in ␣-actinin (41), which is buried within a triple helical, coiled coil bundle (60).

4-helix bundle
NMR experiments with both vinculin and VBS peptide at 200 M or greater concentrations show evidence for interaction (42), but it is uncertain whether non-VBS amphipathic basic peptides could have a similar effect at these concentrations and, thus, whether the interaction observed is truly with the VBS binding groove on Vd1. Even if the interaction is specific, it is clearly of low affinity and would not be sufficient by itself to displace the HTI. Previous observations that VBS (61) and VBS-like peptides (62) induce co-sedimentation of vinculin with actin filaments were confirmed (42). This data is completely consistent with the combinatorial activation model because the VBS peptides are present with the F-actin. In the absence of an assay for vinculin activation that is independent of vinculin binding to actin, it cannot be determined that the VBS alone activates vinculin.
The affinity of VBS for vinculin was measured by surface plasmon resonance (SPR) (42). This is an important parameter for deciding between the models because activation of vinculin by a single ligand requires that the ligand binds with significantly more energy than the HTI, estimated at K d Ͻ 10 Ϫ9 . In SPR assays the affinities of VBS1, -2, and -3 for immobilized Vd1 have been estimated to be 15, 33, and 3 nM, respectively (40), and the affinities of these peptides for immobilized vinculin are 77, 530, and 74 nM (42).
Using solution-based assays, we find that both Vh and Vd1 bind to TAMRA-VBS3 with a K d of ϳ80 nM, whereas binding to vinculin is undetectable. As noted earlier, the conditions of the assay allow us to set a lower limit for the K d of VBS3-vinculin interaction at Ͼ10,000 nM. This estimate for VBS3-vinculin interaction is far different from the 74 nM K d obtained by SPR. The difference might be explained by structural distortion of vinculin when it adsorbs to a surface. In electron micrographs of vinculin adsorbed to mica, heterogeneity of vinculin structures is observed (63), and the tail domain of vinculin is often displaced from Vh (63)(64)(65). In this case Vh is no longer inhibited by Vt and can freely bind VBS3. Indeed a VBS-like peptide was found to bind to Vh but not vinculin in solution, although this same peptide could bind readily to vinculin adsorbed to nitrocellulose (62,66). In this regard, it is interesting that the value (74 nM) obtained for K d of VBS3-vinculin by SPR is nearly identical to the value we obtain for VBS3⅐Vh in solution assay. Based on these considerations, the K d values of VBS for vinculin obtained by SPR might reflect binding of VBS to vinculin that is already partially activated by covalent coupling or other type of adsorption to the surface of the plasmon resonance chip. Loss of degrees of freedom in SPR might also contribute to apparent stronger interactions than found in solution.
Although there are no reported observations that are necessarily inconsistent with combinatorial activation of vinculin, two observations are inconsistent with the conclusion that talin or ␣-actinin or their postulated activated forms are sufficient to activate vinculin. Two-dimensional crystals of actin filaments, ␣-actinin, and vinculin reveal that ␣-actinin does not unfold to bind vinculin and that its buried VBS is not involved in the interaction (67). Additionally, the 74 nM K d measured by SPR for the vinculin-VBS3 interaction (42) is inconsistent with the present observations that 10 M VBS3 fails to activate the vinculin FRET probe and that TAMRA-VBS3 fails to bind vinculin yet shows high affinity for Vh.
To date, only the Shigella virulence protein IpaA has been demonstrated to activate vinculin in the absence of another ligand (herein), and its K d for vinculin has been estimated in the ϳ5 nM range (50). We hypothesize that the pathogenic IpaA protein is capable of coordinately disrupting both the Vd1/Vt and Vd4/Vt interface, thereby circumventing the normal requirement for combinatorial activation of vinculin. A recent structural analysis shows that a small fragment of IpaA containing two VBS sequences dimerizes Vd1, and this has been suggested to be the basis of IpaA activation of vinculin (68). It will be interesting to see if this model applies to activation of vinculin by full-length IpaA.
Significance of Combinatorial Activation of Vinculin-We originally reported that vinculin is autoinhibited (27)(28)(29) and proposed that vinculin gets activated when it is recruited from cytoplasm to focal adhesions and other sites of actin-membrane organization (29). Significantly, recent FRET studies show that recruitment of vinculin to a focal adhesion is not always correlated with maximal activation (49), suggesting that FIGURE 7. Hypothesis: activation of vinculin at cell contacts depends on spatial and temporal coincidence of growth factor and integrin signaling. We have shown that vinculin can be activated by the simultaneous presence of actin filaments and Tn rod but not by either ligand alone. In cells, polymerization of actin is tightly regulated in time and space by arp2/3 and mDia pathways under the control of growth factors (69), and integrins recruit components to adhesion sites in a hierarchal manner depending on receptor occupancy and clustering (82,83). Thus, potentially vinculin could be recruited to an adhesion site but not activated until actin polymerization is initiated in the vicinity in response to additional environmental cues. Indeed, FRET studies in living cells indicate that in some focal adhesions the average vinculin conformation is autoinhibited (49). ECM, extracellular matrix. a second signal is involved. Our current results suggest that one such signal could be actin polymerization, an event that is tightly regulated in time and space (69). At cell membranes, assembly of unbranched, fine actin filaments is controlled by the action of RhoA on mDia1 (70,71). RhoA activity is, in turn, regulated by environmental cues from growth factors (72), substrate adhesion (73), and by shear stress (74). In the lamellipodia of motile cells, where vinculin and talin are co-localized in Racdependent focal complexes, control of arp2/3-dependent, branched actin polymerization is governed indirectly by Rac, Cdc42, and phosphatidylinositol 4,5-bisphosphate; for a recent review, see Ref. 75. These molecules are second messengers for multiple receptors that control motility, including epidermal growth factor receptor, platelet-derived growth factor receptor, and integrins. We suggest that vinculin is recruited to focal complexes and focal contacts where it is a coincidence detector for simultaneous signals from integrin-talin complexes and nearby actin polymerization that has been stimulated by growth factors, adhesion, or force (Fig. 7). In response to spatial and temporal coincidence of talin and actin filaments, vinculin is activated to form a link between talin and actin that could regulate the function or strength of talin-integrin linkages to extracellular matrix.