Structural basis for amplifying vinculin activation by talin.

Talin interactions with vinculin are essential for focal adhesions. Curiously, talin contains three noncontiguous vinculin binding sites (VBS) that can bind individually to the vinculin head (Vh) domain. Here we report the crystal structure of the human Vh.VBS1 complex, a validated model of the Vh.VBS2 structure, and biochemical studies that demonstrate that all of talin VBSs activate vinculin by provoking helical bundle conversion of the Vh domain, which displaces the vinculin tail (Vt) domain. Thus, helical bundle conversion is a structurally conserved response in talin-vinculin interactions. Furthermore, talin VBSs bind to Vh in a mutually exclusive manner but do differ in their affinity for Vh and in their ability to displace Vt, suggesting that the strengths of these interactions could lead to differences in signaling outcome. These findings support a model in which talin binds to and activates multiple vinculin molecules to provoke rapid reorganization of the actin cytoskeleton.

Talin and vinculin are essential components and regulators of cell-matrix (focal adhesion) complexes where they direct the assembly and reorganization of the actin cytoskeleton (1). Both talin and vinculin are critical for cell growth, morphogenesis, and cell migration during the differentiation and organization of tissues, and loss of either talin or vinculin in mice results in marked defects in focal adhesions and embryonic lethality (2,3). Furthermore, both talin and vinculin play important roles in pathophysiological scenarios, including wound healing, ischemia, apoptosis, and metastasis of cancer cells (4 -7).
Focal adhesion signaling is initiated after the specific interactions of components of the extracellular matrix with the ectodomains of various integrin receptors, which are unique in their ability to direct both outside-in and inside-out signaling (8 -10). Connections to the actin cytoskeleton are mediated in part by activation of focal adhesion kinase, which phosphorylates downstream substrates on tyrosine residues and which directly associates with paxillin, which is linked to the actin cytoskeleton via its interactions with vinculin (for reviews, see Refs. 11 and 12). In addition, there are direct interactions of the cytoplasmic domains of integrin receptors with components of the cytoskeleton. For example, an NPXY/F variant of the canonical phosphotyrosine binding domain of integrin receptors mediates its interaction with the FERM 1 (four point one, ezrin, radixin, and moesin) motif present in the head domain of talin (13,14). In turn, talin binds to vinculin through vinculin binding sites (VBSs) present in its central rod domain, and this interaction contributes to reorganization of the actin cytoskeleton (15)(16)(17).
Until recently it was thought that talin largely played a passive role in focal adhesions, as a scaffold protein that simply bridged integrin receptors to the F-actin cytoskeleton via its association with other actin-binding proteins, in particular vinculin (18). However, binding of the talin FERM domain to the cytoplasmic tails of integrin receptors activates focal adhesion signaling pathways (13,14). For inside-out signaling this occurs through FERM-induced conformational changes in the receptor chains, which increase their affinities for ligands (19). Furthermore, the talin FERM domain can also interact with and activate phosphatidylinositol phosphate kinase-1␥ (20 -22), a signaling effector that generates the second messenger phosphatidylinositol 4,5-bisphosphate (PI(4,5)P 2 ). Finally, the VBSs of talin have been shown to facilitate the in vitro binding of vinculin to F-actin (23).
Vinculin contains a highly conserved head (Vh) domain (residues 1-258) that interacts, in an intramolecular fashion, with its tail (Vt) domain (residues 879 -1066; 24 -26). This interaction is thought to hold vinculin in an inactive conformation because biochemical studies have demonstrated that binding of Vh with Vt masks cryptic binding sites for talin, ␣-actinin, and ␣-catenin to Vh, and for F-actin to Vt (25,(27)(28)(29). Thus, vinculin activation is generally agreed to involve structural changes that sever the Vh-Vt interaction, yet the signals that disrupt these intramolecular contacts are still unresolved. The conventional model has suggested that these constraints are simply relieved by the binding of PI(4,5)P 2 to the Vt domain (25), which can insert into acidic phospholipid bilayers (30). Indeed, binding of PI(4,5)P 2 changes the conformation of Vt, and this facilitates the in vitro interaction of Vt with F-actin (31) and, by disrupting the Vh-Vt interaction, has been suggested to allow talin binding to Vh (25,31). However, other models are now equally plausible because talin VBS3 and the vinculin binding site of ␣-actinin can displace Vt from preexisting Vh⅐Vt complexes (32). Furthermore, the crystal structure of the Vh⅐VBS3 complex established that talin VBS3 distorts the binding site for Vt in Vh from a distance, by provoking dramatic alterations in the structure and positions of the ␣-helices of the N-terminal helical bundle of Vh, by a process coined as helical bundle conversion (32).
A long-standing mystery in focal adhesion signaling is how the structural proteins of the actin cytoskeleton undergo such rapid changes in their localization and organization. One possible mediator is talin, which harbors three separate high affinity vinculin binding sites (VBS1, VBS2, and VBS3) that are located in its central rod domain (residues 607-636 (VBS1), 852-879 (VBS2), and 1944 -1969 (VBS3); see Refs. 15,17,and 23). Each of the talin VBSs have been predicted to form an amphipathic ␣-helix (17), and larger portions of talin containing these VBSs can bind individually to Vh (17) and act as competitive inhibitors that prevent the binding of Vt to Vh (23). However, it was not known whether these interactions were indeed mediated by the refined sequences of talin VBSs. We reasoned that if all of the more highly defined VBSs of talin indeed bound to Vh in a mutually exclusive manner and each was individually capable of activating vinculin, this would al-low one molecule of talin to activate three molecules of vinculin effectively, a scenario that would amplifying integrin signaling and promote rapid changes in the actin cytoskeleton. To test this hypothesis we assessed the ability of talin VBS1 and VBS2 to displace Vt from preexisting Vh⅐Vt complexes and solved the crystal structure of the Vh⅐VBS1 complex to 2.4 Å resolution.

EXPERIMENTAL PROCEDURES
Vh⅐VBS1 Crystallization and X-ray Data Collection-The His 6tagged human Vh (residues 1-258) and Vt (residues 879 -1066) expression constructs and the purification of Vh and Vt proteins have been described previously (32).
Human talin VBS1 (residues 607-636) was synthesized and purified by high performance liquid chromatography in our in-house facility. Native Vh⅐VBS1 crystals were obtained from conditions similar to those for Vh⅐VBS3 crystals (32). These crystals belong to space group R32 with one heterodimer in the asymmetric unit, a solvent content of 59%, and a volume to mass ratio of 3.04 Å 3 /Da. The Vh⅐VBS1 crystals were cryoprotected in paratone oil. Vh⅐VBS1 x-ray data were collected at the Advanced Photon Source, SBC-CAT ID beamline. The R merge to 2.42 Å FIG. 1. Structure of the human talin Vh⅐VBS1 complex. A, stereo drawing of the Vh⅐VBS1 crystal structure. VBS1 (residues 607-631, of which 608 -626 form an amphipathic ␣-helix) is shown in ball and stick representation (oxygen atoms, red; carbon, yellow; nitrogen, blue; sulfur, green). In all figures the N terminus of each helix is marked with a plus sign, and the C terminus of each helix is indicated with a minus sign, in agreement with the helix dipole moment. B, talin VBS1 provokes helical bundle conversion of the Vh domain of vinculin. A stereo drawing superimposition of the Vh⅐VBS1 structure (Vh, pink; VBS1, teal) onto the inactive Vh⅐Vt structure (Vh, yellow; Vt, gray) is shown. Movements and helical distortions (arrows) of helices ␣1 and ␣2 of the N-terminal helical bundle which occur upon activation of Vh by talin VBS1 are indicated. Helices H1-H5 of Vt are labeled. C, stereo view of the final 2F o Ϫ F c electron density map of talin VBS1 (white bonds) in contact with ␣1 and ␣2 of the vinculin head domain (pink bonds). For clarity, only a few residues are labeled.  (38) indicates that 95.7% of all of the residues lie in the most favorable region, that the remaining 4.3% lie in the additional allowed regions, and that there are no residues in the generously allowed or disallowed regions. This structure analysis also showed that all stereochemical parameters are better than expected at the given resolution (see Table II). Talin VBS Binding Assays-0.26 nM Vh and 0.52 nM Vt were incubated for 20 min at room temperature to form the Vh⅐Vt complex. Talin VBS1, VBS2, and VBS2e peptides were then titrated into the sample at molar ratios ranging from 0 to 5:1 and allowed to incubate for 20 min. Complexes were then analyzed by native PAGE.
Binding of Talin VBS1, VBS2, VBS2e, and VBS3 to Vh-Binding studies were performed on a Biacore 3000 surface plasmon resonance (SPR) instrument. Vh was covalently attached to a carboxymethyldextran-coated gold surface (CM-5 chip, Biacore). The carboxymethyl groups on the chip were activated with EDC and N-hydroxysuccinimide to form the N-hydroxysuccinimide ester of carboxymethyldextran. Vh was attached at pH 5.0 to this activated surface by reaction of the carboxyl groups of dextran with the primary amines of Vh to form an amide linkage. Any remaining reactive sites on the surface were blocked by reaction with ethanolamine. A reference cell was prepared similarly except that no Vh was added. Binding was measured by flowing the VBS peptides in 20 mM HEPES, 0.1 mM EGTA, 0.1 mM EDTA, 0.5 mM dithiothreitol at a flow rate of 20 l/min through the reference and Vh-containing flow cells in sequence. A blank was also run consisting of only buffer. After the injection, release of the bound VBS peptides was measured by flowing only buffer through the flow cells. Regeneration of the chip surface to remove bound VBS peptides was accomplished by injecting 20 l of 10 mM glycine, pH 1.5, through the flow cells followed by a 7-min equilibration in buffer. Data reported are the difference in SPR signal between the flow cell containing Vh and the reference cell. Any contribution to the signal was removed by subtraction of the blank (buffer) injection from the reference-subtracted signal.

Structure of the Human Vh⅐VBS1
Complex-The structure of the Vh domain in its inactive conformation when complexed with Vt defined a two four-helical bundle structure linked by a long shared ␣-helix (␣4) that is remarkably similar to the structure of the ␤-catenin-␣-catenin complex (39). Strikingly, the structure of the Vh⅐VBS3 complex revealed that VBS3 binding provoked remarkable changes in the N-terminal helical bundle of Vh (helices ␣1-␣4), creating an entirely new five-helical bundle structure; hence the idiom, helical bundle conversion. In contrast, the structure of the C-terminal helical bundle of Vh (helices ␣4 -␣7) is essentially unchanged by binding of VBS3, suggesting that the C-terminal helical bundle of Vh functions as a rigid scaffold that allows the N-terminal helical bundle to bind to its various partners (32).
Talin is unique among the vinculin-binding partners because it harbors, curiously, three noncontiguous VBSs, VBS1 (residues 607-636), VBS2 (residues 852-879), and VBS3 (1944 -1969; Refs 15,17,and 23). Based upon their ␣-helical amphipathic nature (17) and in part from competition studies using larger portions of talin harboring VBS1 (residues 498 -636; Ref. 23), we reasoned that VBS1 and VBS3 would have similar or overlapping binding sites in Vh and would share in their ability to trigger structural changes in the N-terminal helical bundle of Vh. As expected from the Vh⅐VBS3 structure (32) and binding studies with VBS1 (498 -636) (23), in the Vh⅐VBS1 crystal structure one molecule of VBS1 binds to one molecule of Vh (Fig. 1A). Superimposition of the Vh⅐VBS1 structure with that of the inactive Vh⅐Vt complex established that VBS1 is very similar to VBS3 in its ability to activate vinculin. Specifically, insertion of the amphipathic ␣-helix of talin VBS1 between Vh helices ␣1 and ␣2 generates a totally new five-helical N-terminal bundle structure and stretches the loop linking Vh helices ␣1 and ␣2 by unwinding the first two helical turns of ␣1 which are present in the Vh⅐Vt structure (Fig. 1B). As in the Vh⅐VBS3 structure (32), all VBS1 interactions are restricted to the N-terminal helical bundle of Vh, with little effect on the structure of the C-terminal helical bundle (Fig. 1, A and B). Thus, binding of either talin VBS1 or VBS3 has dramatic effects on the structure of the N-terminal helical bundle of Vh, yet leaves the structure of its C-terminal helical bundle essentially unchanged.
The shared similarities in the structures of the Vh⅐VBS1 and Vh⅐VBS3 complexes were striking (Fig. 2), especially because talin VBS1 and VBS3 share only 16% identity and 46% similarity ( Fig. 3A and Ref. 17). This similarity in structure also occurs even though VBS3 has three additional ordered residues that add an extra turn to its N terminus and that VBS1 has six additional ordered residues on its C terminus which bend away at ϳ40 o from its position in the Vh⅐VBS3 structure. Nonetheless, the binding of VBS1 to Vh is similar to that of VBS3, where Van der Waals interactions between the hydrophobic face of VBS1 with the hydrophobic core of the N-terminal helical bundle of Vh ( Fig. 1C and Table I) provoke dramatic changes in the structures of Vh helices ␣1, ␣2, and ␣4 (Fig. 1B). Therefore, like talin VBS3, talin VBS1 is remarkable in its ability to provoke helical bundle conversion of the N-terminal bundle of Vh. Further, the residues of Vh interacting with VBS1, and the position of the VBS1 amphipathic ␣-helix between Vh helices ␣1 and ␣2, indicate that talin VBS1 and VBS3 must bind to the N-terminal helical bundle of Vh in a mutually exclusive fashion.
Helical Bundle Conversion of Vh by Talin Is a Conserved Response-The structure of the Vh⅐VBS1 (Table II) complex was similar in its overall architecture to that of the Vh⅐VBS3 complex. Superimposition of the Vh⅐VBS1 and Vh⅐VBS3 structures demonstrated that, despite considerable differences in their sequences, VBS1 and VBS3 fit into a similar cavity created by their insertion between Vh helices ␣1 and ␣2 (Fig. 2). The C␣ positions of VBS1-bound Vh (residues 3-18, 48 -209, and 230 -248) can be superimposed onto the 197 equivalent C␣ positions of VBS3-bound Vh with root mean square deviation of 1.06 Å. The largest differences in the structures of the Nterminal helical bundle of the two complexes were in the loop between Vh helices ␣1 and ␣2. This shift begins in the vicinity of Vh residue 19 and extends through to residue 47, with movements of more than 6 Å at residue 34 being accompanied by a rotation of the C terminus of helix ␣1 by ϳ15 o (Fig. 2).
As in the Vh⅐VBS3 complex (32) the interactions of Vh with talin VBS1 are largely hydrophobic in nature, with 24 residues of Vh participating in hydrophobic interactions with 16 residues of VBS1, and Gln 19 of Vh is within hydrogen bonding distance to Ser 620 of VBS1 (Table I). VBS1 buries 53% (more than 1,200 Å 2 ) of its entire solvent-accessible surface area upon binding to Vh. The binding of the N-terminal half of either VBS1 or VBS3 to the N-terminal portion of Vh helix ␣1 and to the C-terminal half of Vh helix ␣2 is nearly identical. However, the C terminus of each talin peptide distorts the C-terminal half of ␣1, the N-terminal half of ␣2, and the connecting loop in quite different ways (Fig. 2). Nonetheless, the striking structural alterations leading to helical bundle conversion in Vh provoked by the binding of VBS1 or VBS3, and the architecture of the binding cavity for these VBSs, are remarkably similar in the two complexes; therefore, the binding of VBS1 and VBS3 results in a similar and profound outcome in vinculin structure.
Vh⅐VBS2 Model-Human talin VBS2 (residues 852-879) has 38% identity and 57% similarity with VBS1 and 15% identity and 46% similarity with VBS3 ( Fig. 3A and Ref. 17). Based on the sequence alignment and the Vh⅐VBS1 crystal structure, we modeled the interaction of talin VBS2 (Fig. 3B) with Vh. The Vh⅐VBS2 model moved only slightly after 200 cycles of coordinate energy minimization, as the 279 C␣ positions of the Vh⅐VBS2 starting model can be superimposed onto those of the energy-minimized Vh⅐VBS2 model with a root mean square deviation of 0.367 Å. Again, the modeled structure of the Vh⅐VBS2 complex predicts little change in the structure of the C-terminal helical bundle of Vh as seen in its inactive state. However, VBS2 binding is predicted to cause alterations in Vh helices ␣1, ␣2, and ␣4 which lead to helical bundle conversion. The side chains of Vh that are predicted to contribute to the binding of all three VBSs are conserved, and the minimal binding site includes several residues of Vh helices ␣1 and ␣2 and the loop connecting ␣1-␣2, and some residues on Vh helix ␣4 (Table I).
Finally, the architecture of the modeled VBS2 structure (Fig.  3C) is remarkably similar to the structure of VBS1 (Fig. 3D) when bound to Vh. Notable features include a string of hydrophobic residues on the face of VBSs which interact with Vh (left images in Fig. 3, C and D) and a conservation of polar residues along the face of these amphipathic ␣-helices which is exposed to the solvent (right images in Fig. 3, C and D).
Binding Affinities of Talin VBSs for Vh-Chicken talin VBS3 has been reported to bind Vh with fairly high affinity (K d of 39 nM; Ref. 23). However, these experiments were performed using a larger domain, talin residues 1943-2157, and the Vh used in these assays was a glutathione S-transferase fusion protein (23). Furthermore, the binding constants for talin VBS1 and VBS2 are unknown. Given the surface areas of talin VBSs buried in the Vh⅐VBS structures we reasoned that the affinities of the refined talin VBSs would be significantly higher than appreciated previously and that this should correlate with their degree of interaction with Vh helices ␣1, ␣2, and ␣4. We therefore determined the binding affinities of talin VBS1 (residues 607-636), VBS2 (residues 852-879), and VBS3 (residues 1944 -1969) for Vh using Biacore 3000 SPR. Under these conditions, binding of VBS3 in the fit model was determined with an estimated K d of ϳ3.1 nM. In agreement with their actual or modeled structures in the binding cleft of Vh, the affinities of talin VBS1 and VBS2 were somewhat reduced relative to that of VBS3, with estimated K d for VBS1 and VBS2 of ϳ14.7 and 32.8 nM, respectively (Fig. 4).
Talin VBS1 and VBS2 Displace Vt from Vh⅐Vt Complexes-The similar talin VBSs in Vh and the ability of talin VBS1 to provoke helical bundle conversion in Vh suggested that VBS1 and VBS2 would also distort the binding site for Vt in Vh from a distance. We therefore tested whether VBS1 and VBS2 could displace Vt from preexisting Vh⅐Vt complexes. Free Vh and complexes containing Vh bound to Vt, VBS1, VBS2, or VBS3 were easily distinguishable on native gels on the basis of their migration (Fig. 5A). Vh⅐Vt complexes were formed at a molar ratio of 2 mol of Vt to 1 mol of Vh, and these complexes were then incubated with increasing concentrations of talin VBS1 or VBS2 peptide. VBS1 peptide began to displace Vt from preexisting Vh⅐Vt complexes even at ratios as low as 1:20, and Vt was displaced from all Vh complexes at a 1:2 ratio of VBS1 to Vt (Fig. 5, B and C). VBS2 was also capable of displacing Vt from preexisting Vh⅐Vt complexes, but much higher levels of VBS2 were necessary to see these effects, and Vt was not totally displaced from Vh even at a 2.5:1 ratio of VBS2 to Vt (Fig. 5, D  and E). Thus, although both VBS1 and VBS2 are capable of displacing Vt from Vh⅐Vt complexes, VBS1 was clearly more effective at displacing Vt.
Our modeled Vh⅐VBS2 structure suggests an explanation for the reduced ability of VBS2 to displace Vt versus VBS1 (Fig. 5) and VBS3 (32). First, even when considering equivalent residues (VBS1 607-626 versus VBS2 855-874), VBS1 buries 7% more (and VBS3 (1949(and VBS3 ( -1968) 15% more) of its solvent accessible surface area than VBS2 upon binding to Vh. However, more importantly, three key conserved residues intimately involved in contacts of both VBS1 and VBS3 with Vh are lacking in VBS2 (Fig. 3). Val 619 of VBS1, which is buried deep in a hydrophobic pocket and interacts with Ile 115 of Vh (Table I), is replaced by a threonine in VBS2. Next, Ser 620 of VBS1 is within hydrogen bonding distance to the N⑀ of Vh Gln 19 , and its replacement by alanine in VBS2 compromises this interaction. Finally, Leu 623 of VBS1, which is in hydrophobic contacts with  Table I), is replaced by a valine in VBS2, and the smaller side chain can only interact with Leu 23 but not with Gln 19 or Ile 20 . Collectively, these alterations would be predicted to impair some of the hydrophobic interactions of VBS2 with Vh.
Previous studies evaluating the ability of a larger domain containing talin VBS3 (residues 1943-2157) to bind to a GST-Vh fusion protein demonstrated that larger portions of talin containing VBS1 (residues 498 -636) and VBS2 (residues 727-926) could compete with talin VBS3 for binding to Vh (23). However, these experiments did not definitively pinpoint the regions that compete for binding to Vh to the more refined sequences of talin VBSs. The binding site for talin VBS1 (residues 607-626) in our Vh⅐VBS1 structure lies in the same cavity created by the binding of VBS3 in the Vh⅐VBS3 structure (Fig. 2), as does that predicted for talin VBS2 (residues 855-874) in the modeled structure of the Vh⅐VBS2 complex. Therefore, we reasoned that the binding of these refined talin VBSs should be mutually exclusive. To confirm that this was indeed the case we tested whether VBS1 and VBS2 peptides could displace the corresponding talin peptide when complexed with Vh. VBS2 (Fig. 6A) and VBS1 (Fig. 6B) peptides were both capable of displacing the corresponding VBS from preexisting Vh⅐VBS complexes, yet VBS1 was much more effective at displacing VBS2 from preexisting Vh⅐VBS2 complexes than vice versa (Fig. 6, A and B). This finding agrees with the reduced ability of VBS2 to displace Vt from Vh⅐Vt complexes (Fig. 5D).
Our model of the Vh⅐VBS2 complex and the crystal structures of the Vh⅐VBS1 and Vh⅐VBS3 complexes suggested that alterations in three key residues (T867V, A868S, and V871L; Fig. 3A) of VBS2 would enable VBS2 to displace Vt from Vh⅐Vt complexes, akin to that observed for VBS1 and VBS3. To test this notion, we synthesized this modified VBS2 (VBS2e, for "VBS2 enhanced"; Fig. 7A) and addressed its ability to displace Vt from preexisting Vh⅐Vt complexes (Fig. 7, B and C), and to displace VBS1 from preexisting Vh⅐VBS1 complexes (Fig. 7, D  and E). As predicted, these mutations in VBS2 converted this talin peptide into one that behaved as VBS1 and VBS3, thus validating the model of the Vh⅐VBS2 structure. Finally, we performed binding studies using Biacore 3000 SPR and determined the binding of VBS2e in the fit model with an estimated K d of ϳ7.65 nM (Fig. 7F), confirming the importance of these three residues in establishing hydrophobic contacts in the binding cleft of the N-terminal helical bundle of Vh.

DISCUSSION
Talin loss abolishes the formation of focal adhesions (3), consistent with its accepted role as a structural scaffold that bridges integrin receptors with components of the actin cytoskeleton, in particular with vinculin (1). However, talin has recently been shown to play an active role in inside-out signaling directed by these receptors (19). Here, our structures and binding-displacement assays establish that the three VBSs of talin bind to the same site in Vh and share in their ability to displace Vt from preexisting Vh⅐Vt complexes, suggesting that they play a direct role in activating vinculin and outside-in integrin signaling. The differences in the affinity of the three talin VBSs for Vh also establish a hierarchy in the binding of these recognition elements to vinculin (VBS3 Ͼ VBS1 Ͼ VBS2), and we propose that this property may allow talin to function as a rheostat, where its progressive binding to vinculin molecules would allow it to regulate signaling outcome.
One of the earliest events detected following the interaction of integrin receptors with the extracellular matrix is the binding of PI(4,5)P 2 to the talin head domain (40,41). In turn, binding of PI(4,5)P 2 provokes conformational changes in the talin head which allow the rapid association of its FERM domain with the NPXY/F motif present in the cytoplasmic tails of ␤-integrin receptors (14,42,43). In their resting state, the ␣ and ␤ subunits of integrin receptors are clasped together through multiple hydrophobic and electrostatic contacts within their membrane-proximal helices (44). However, binding of the talin FERM domain to ␤-integrin receptors pushes the ␣ and ␤ subunits apart (9), and this structural alteration is transferred down the length of the receptor subunits to their ectodomains,  b The free R-factor is a cross-validation residual calculated by using 5% of the native data, which were randomly chosen and excluded from the refinement.
where it changes the affinity of these receptors for extracellular matrix ligands (19). Furthermore, talin plays an active role in amplifying the inside-out signal by also binding to and triggering the activity of phosphatidylinositol phosphate kinase-1␥ (20 -23), which generates more PI(4,5)P 2 , which would then bind to talin to amplify and sustain the response.
Talin is also essential for outside-in signaling triggered by focal adhesions (3), but here its role has been generally thought to be passive, by acting as a scaffold protein that binds to other proteins of the actin cytoskeleton. Furthermore, it has been suggested that talin acts downstream of vinculin in this signaling pathway, where the binding of PI(4,5)P 2 to the Vt domain of vinculin has been proposed to "unfurl" its five-helical bundle, thus severing its intramolecular interaction with Vh and allowing Vh then to bind to talin and other partners (25,31). Our structures of the Vh⅐VBS1 and Vh⅐VBS3 complexes, the validated model of the Vh⅐VBS2 complex, and the ability of all three VBSs to displace Vt from preexisting Vh⅐Vt complexes now suggest a different model. Specifically, we propose that talin is positioned at the proximal end of the outside-in signaling pathway and acts as an effector that activates vinculin (Fig. 8).
The structures of the Vh⅐VBS1 and Vh⅐VBS3 complexes, along with the validated modeled structure of the Vh⅐VBS2 complex, suggest that helical bundle conversion may occur in other scenarios as well. In the case of vinculin the amphipathic ␣-helices of all three talin VBSs insert between Vh helices ␣1  and ␣2, causing dramatic movements and distortions of these helices, and that of helix ␣4, and collectively these alterations create a hydrophobic cavity that literally swallows the talin VBSs. In addition, in all three structures the C-terminal helical bundle of Vh is impervious to change, supporting the notion that it serves as an inflexible scaffold that supports the dynamic structural changes that occur in the Vh N-terminal helical bundle when vinculin transitions from its inactive to active state. Thus, there is a high degree of specificity to these interactions, and these structures have defined the talin ligand binding sites as residing in the cavity created by the conversion of Vh helices ␣1, ␣2, and ␣4.
Another insight coming from these studies is that there are indeed appreciable differences in the interactions of talin VBSs within the hydrophobic cleft in Vh created by helical bundle conversion. Talin VBS3 interacts most intimately within this cleft (K d of 3.1 nM), whereas the VBS2 interaction is significantly weaker (K d of 32.8 nM). Vinculin binds more efficiently to the central rod of talin than to the full-length protein (45), suggesting that talin activation is a necessary step toward linking its VBSs with vinculin. Talin appears rapidly activated during focal adhesions (40,41,43), and we suggest that these events include exposing talin VBSs to allow binding to Vh. The range in the strengths of the interactions of talin VBSs with Vh also suggests hierarchical and/or cooperative binding, which could impart talin with the ability to function as a rheostat. Here, the strengths of these interactions would be translated into strengths of signal, and this would allow the cell to finetune the signaling response (Fig. 8).
Each of the talin VBSs is individually capable of binding to the same site in Vh. When bound by lipid and/or actin talin can exist as an antiparallel homodimer (46), whereas vinculin has been proposed to form homodimers or trimers after its interaction with vasodilator stimulatory protein (47). In this scenario a vinculin trimer has been proposed to bind to the antiparallel talin homodimer to provoke changes in the actin cytoskeleton (23). However, gel filtration and electron microscopy have revealed that under physiological conditions talin and vinculin are both monomers (48 -50). These findings, together with our   7. Validation of the Vh⅐VBS2 model. A, native PAGE analysis of Vh bound to Vt, free Vh, and of Vh bound to VBS2, and VBS2e (having three substitutions, T867V, A868S, and V871L). B, VBS2e efficiently displaces Vt from preexisting Vh⅐Vt complexes. VBS2e peptide was titrated into preexisting Vh⅐Vt complexes (closed circle) at molar ratios of the respective VBS⅐Vh of 1:10, 1:3.3, 1:1, and 1.5:1 and displaced Vt to form distinct Vh⅐VBS2e complexes (gray circle). Free Vh is indicated with an asterisk. Again, free Vt is not visible in native gels because of its high pI. C, quantitation of the dissociation of the Vh⅐Vt complex (solid line and closed circle) and the formation of the Vh⅐VBS2e complex (dotted line and gray circle) is shown. D, VBS2e efficiently displaces talin VBS1 from preexisting Vh⅐VBS1 complexes. VBS2e peptides were incubated with Vh⅐VBS1 complex (gray circle) at molar ratios of 1:2, 1:1, 1. structures and biochemical studies, suggest a simpler model whereby binding of one talin molecule could serve to activate three molecules of vinculin (Fig. 8). This scenario might then explain how outside-in signaling provoked by focal adhesions induces such rapid changes in the actin cytoskeleton. The promiscuous nature of talin allows the protein to bind simultaneously to integrin receptors through its head domain and to multiple vinculin molecules through its three VBS motifs, and the multiplicity of these interactions effectively amplifies the signaling response. Thus, talin can no longer be viewed simply as a scaffold protein in focal adhesions, but rather by altering the structures of its binding partners, as a signaling effector that directs and amplifies both inside-out and outside-in integrin signaling. FIG. 8. Talin activation of vinculin in outside-in integrin signaling. In this model, talin, in its monomeric form, is proposed to function as a rheostat that activates multiple vinculin molecules to amplify integrin signaling. Talin interacts with ␤-integrin receptors through its FERM domain and binds to vinculin through any of its three noncontiguous vinculin binding sites (VBS1, VBS2, and VBS3). All three of the talin VBSs bind to the same site in Vh and activate vinculin by helical bundle conversion that displaces Vt, which is then free to bind to F-actin. The three talin VBSs would interact with and activate up to three molecules of vinculin to amplify outside-in integrin signaling in focal adhesions.