β Integrin Tyrosine Phosphorylation Is a Conserved Mechanism for Regulating Talin-induced Integrin Activation*

Integrins are large membrane-spanning receptors fundamental to cell adhesion and migration. Integrin adhesiveness for the extracellular matrix is activated by the cytoskeletal protein talin via direct binding of its phosphotyrosine-binding-like F3 domain to the cytoplasmic tail of the β integrin subunit. The phosphotyrosine-binding domain of the signaling protein Dok1, on the other hand, has an inactivating effect on integrins, a phenomenon that is modulated by integrin tyrosine phosphorylation. Using full-length tyrosine-phosphorylated 15N-labeled β3, β1A, and β7 integrin tails and an NMR-based protein-protein interaction assay, we show that talin1 binds to the NPXY motif and the membrane-proximal portion of β3, β1A, and β7 tails, and that the affinity of this interaction is decreased by integrin tyrosine phosphorylation. Dok1 only interacts weakly with unphosphorylated tails, but its affinity is greatly increased by integrin tyrosine phosphorylation. The Dok1 interaction remains restricted to the integrin NPXY region, thus phosphorylation inhibits integrin activation by increasing the affinity of β integrin tails for a talin competitor that does not form activating membrane-proximal interactions with the integrin. Key residues governing these specificities were identified by detailed structural analysis, and talin1 was engineered to bind preferentially to phosphorylated integrins by introducing the mutation D372R. As predicted, this mutation affects talin1 localization in live cells in an integrin phosphorylation-specific manner. Together, these results indicate that tyrosine phosphorylation is a common mechanism for regulating integrin activation, despite subtle differences in how these integrins interact with their binding proteins.

Integrins play a fundamental role in cell adhesion and migration, linking the extracellular matrix to the actin cytoskeleton.
In the adult, integrins are essential for a variety of biological processes including wound healing, leukocyte trafficking, and angiogenesis, and are thus increasingly attractive therapeutic targets for a variety of conditions, notably cancer. Mammals express 18 different ␣ subunits and 8 different ␤ subunits, which form 24 unique ␣␤ heterodimers (excluding splice variants). Each ␣ and ␤ chain of the integrin heterodimer consists of several linked globular extracellular domains, a single membrane-spanning domain, and a short cytoplasmic tail (1). The cytoplasmic tails of the ␤ subunits exhibit conserved regions (Fig. 1A), and act as a hub for protein-protein interactions (2), modulating a variety of signaling processes in both directions across the membrane. Extracellular adhesion can be activated from within the cell by the cytoskeletal protein talin (1,(3)(4)(5) via a direct interaction between the talin F3 domain and the membrane-proximal (MP) 3 portion of the integrin tail (6 -8).
The different ␤ integrins found in mammals occupy unique biological niches (1) and show significant diversity in their protein-protein interactions. Of the six ␤ integrin tails that display a high level of sequence homology to one another (␤1, ␤2, ␤3, ␤5, ␤6, and ␤7), all contain two NPXY or NPXY-like motifs, a near membrane-distal (nMD) site and a far membrane-distal (fMD) site, that bind to phosphotyrosinebinding (PTB) domains (including the talin F3 domain) (9) and are potential phosphorylation sites. In the ␤2 tail, however, both tyrosine residues are substituted with phenylalanine, and in the ␤7 tail this substitution occurs in the fMD site. Uniquely, the ␤7 tail also exhibits two additional MP tyrosine phosphorylation sites that are not part of an NPXY motif (10).
This study explores three of these integrins: ␤3, ␤1A (the most common splice variant of ␤1), and ␤7 (see Fig. 1A). Previous studies on integrin tyrosine phosphorylation have generally focused on two of these: ␤3 and ␤1. When the ␤1 integrin was first isolated and sequenced, a potential tyrosine phosphorylation site was proposed (11), and an early study observed ␤1 tyrosine phosphorylation in response to transformation of cells with viral Src (v-Src) (12), followed by the demonstration of tyrosine phosphorylation of ␤1 by v-Src in vitro (13). These studies involved a viral protein, but it was later found that ␣IIb␤3 is tyrosine-phosphorylated in response to platelet acti-vation, and that cellular Src (c-Src) and related tyrosine kinases phosphorylate ␤3 in vitro (14). Studies suggest that ␤1 interacts with various Src family kinases and that these kinases are activated by integrin engagement with the extracellular matrix (15)(16)(17)(18)(19)(20)(21)(22). There is also evidence for a direct and constitutive interaction between ␤3 and c-Src (23).
Various studies have demonstrated that tyrosine phosphorylation of ␤1 influences integrin localization and activity, as well as cell morphology. An early study showed that transformation of cells with Rous sarcoma virus, which expresses v-Src, leads to ␤1 integrins adopting a more diffuse distribution on the cell surface (12), rather than being localized in focal adhesions, which are large macromolecular complexes composed of integrins and intracellular proteins, such as vinculin, talin, and paxillin, which serve as sites of tight attachment to the extracellular matrix (24,25). Transformed cells also display rounding, decreased fibronectin matrix assembly, and decreased cell migration (12). Another study found that whereas unphosphorylated ␤1 integrins localize to focal adhesions, phosphorylated integrins localize to podosomes (26). Further insight has come from studies using non-phosphorylatable integrins with Tyr to Phe mutations (supplemental Fig.  S1A). The Y783F mutation in the ␤1 nMD site reverses the effects of v-Src (27). Fibroblasts expressing ␤1 Y783F, Y795F, or Y783F/Y795F display impaired directed cell migration but increased fibronectin binding (28), and Y783F/Y795F also causes slowed cell spreading and decreased focal adhesion kinase activation (29). In ␤3, the nMD mutation Y747F disrupts adhesion and clot retraction in hematopoietic cells (30). More tellingly, mice with the knock-in double Y747F/Y759F mutation in ␤3 display a severe bleeding defect (31), although, surprisingly, mice with the analogous ␤1 knock-in mutation show no significant phenotype (32,33).
The role of tyrosine phosphorylation in outside-in ␤3 signaling is relatively well established (30, 31, 34 -41). The role in inside-out integrin activation is less clear, but various lines of evidence indicate that phosphorylation negatively regulates activation. Tyrosine phosphorylation of ␣V␤3 decreases the affinity of live cells for fibronectin (42), and phosphorylation of ␤1 decreases its affinity for fibronectin in vitro (13).
Dok1 is a signaling protein with a PTB domain capable of binding integrins (9). Dok1 negatively regulates ␤3 integrin activation (8), an observation initially difficult to explain due to the very weak interaction observed between these proteins. We subsequently reported that tyrosine phosphorylation greatly increases Dok1 affinity for short ␤3 peptides while slightly decreasing talin1 affinity, observations that led to an initial structural explanation for this phenomenon (Fig. 2) (43). However, these findings did not clarify the specific roles of the different NPXY motifs or indicate whether this mechanism could be generalized across different integrins. Here, we explore the phosphorylation dependence of the talin1 and Dok1 interactions with full-length ␤3, ␤1A, and ␤7 tails. We show that tyrosine phosphorylation is a common mechanism for regulating the affinity of these proteins for integrin tails, and a recent crystal structure of a talin⅐␤1 complex (44) allows us to explain subtle differences between different integrins. We also describe the structural basis of this phosphorylation state specificity in detail and generate a talin mutant that shows preferential binding to phosphorylated integrin tails.

EXPERIMENTAL PROCEDURES
Expression and Purification of Proteins-Proteins and peptides were expressed in Escherichia coli and purified as reported previously (44), unless otherwise indicated. Integrin tail constructs corresponding to the entire predicted cytoplasmic region were produced in pET16b using the following boundaries: ␤3 Lys 716 -Thr 762 , ␤1A Lys 752 -Lys 798 , and ␤7 Arg 747 -Leu 798 . The talin1 F3 domain (Gly 309 -Ser 405 ) and the Dok1 PTB domain (Gln 154 -Gly 256 ) were produced in pGEX-6P-2 as reported previously (8). For experiments in mammalian cells, the cDNA encoding full-length mouse talin1 was amplified by PCR and subcloned into pEGFP-C1. Mutations in pET and pGEX vectors were introduced with the QuikChange kit, and mutations in pEGFP-C1 were introduced with the QuikChange II XL kit (Stratagene).
Cell Culture-SYF cells (mouse embryonic fibroblasts (MEFs) deficient in c-Src, c-Fyn, and c-Yes) and SYF ϩ Src cells (SYF MEFs reconstituted with c-Src) (19) were obtained from the American Type Culture Collection. Cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, L-glutamine, and antibiotics at 37°C with 6% CO 2 . Transient transfections were carried out with Lipofectamine Plus (Invitrogen) as described by the manufacturer. , and ␤7 integrin tails. The two NPXY motif tyrosine residues are indicated, with ␤3 numbering. The MP, nMD, and fMD regions are denoted. Secondary structure is based on the ␤1D⅐talin2 complex structure (44), with ␣ helices denoted in blue and 3 10 helices in green. B, sequences of the PTB domains of Dok1, talin1, and talin2 aligned by secondary structure, with secondary structure elements from the Dok1 PTB domain structure (PDB 2V76) (43) shown. Notable residues are indicated with Dok1 numbering.
Tyrosine Phosphorylation of Integrin Tails-The kinase domain of c-Src (Gln 251 -Leu 533 ) in pET28 was co-expressed with Yersinia pseudotuberculosis YopH in pCDFDuet-1 and purified by immobilized metal affinity chromatography as previously reported (45). Tyrosine phosphorylation was performed overnight at 30°C with 20 M integrin tail and 0.015 mg/ml of Src in 50 mM Tris, 20 mM MgCl 2 , 10 mM MnCl 2 , 1 mM ATP, 1 mM dithiothreitol, pH 7.0. Phosphorylated tails were separated from unphosphorylated tails by C 4 reverse phase high performance liquid chromatography and identified by mass spec-trometry and NMR (supplemental Fig. S2). Preliminary experiments observed phosphorylation by autoradiography (supplemental Fig. S3).
NMR Spectroscopy-All NMR experiments were performed on spectrometers equipped with Oxford Instruments superconducting magnets (500, 600, 750, and 950 MHz 1 H operating frequencies) and GE/Omega computers. Unless otherwise indicated, samples were prepared in NMR buffer (50 mM sodium phosphate, 100 mM NaCl, 1 mM dithiothreitol, pH 6.1) with 5% D 2 O and Complete protease inhibitors (Roche Applied Science). Experiments were performed at 25°C. The 1 H and 15 N resonances of the 15 N-labeled ␤7 integrin tail were assigned using a 0.2 mM sample in 20 mM sodium acetate, pH 4.5, and employing three-dimensional NOESY-HSQC and three-dimensional TOCSY-HSQC spectra. Resonance assignments were transferred to pH 6.1 conditions through pH titrations. Spectra were referenced in the direct dimension to 2,2-dimethyl-2-silapentane-S-sulfonate at 0 ppm, with indirect referencing in the 15 N dimension using a 15 N/ 1 H frequency ratio of 0.101329118 (46). Data were processed using NMRPipe (47) and spectra were visualized using the program SPARKY or CCPN Analysis (48 where W H and W N are weighting factors for the 1 H and 15 N amide shifts, respectively (W H ϭ 1, W N ϭ 0.154) (49), and ⌬ ϭ ␦ bound Ϫ ␦ free . Dissociation constants were determined by fitting changes in backbone chemical shifts upon increasing talin concentration to the following equation, where K d is the dissociation constant, ⌬(H,N) is the weighted shift change, ⌬(H,N) max is the shift change at saturation, and [L] and [U] are the concentrations of the labeled and unlabeled proteins, respectively. Data from peaks that were well resolved, had a significant change in position, and were discernable throughout the titration were fit simultaneously to this equation with the program OriginPro 8, extracting a single K d and multiple ⌬(H,N) max values. Values for ⌬G were calculated from the K d value.
Data are presented throughout as the K d value Ϯ S.E. This error only takes into account the uncertainty from the fitting procedure. The other source of error in these experiments would be concentration errors, but they are not reported due to difficulty in estimating them. However, experience indicates that errors in K d stemming from talin1 or Dok1 concentration determinations would at most be about 10% (corresponding to a maximum error in ⌬G of 0.25 kJ/mol). For interactions with K d values less than about 100 M, this would be compounded by ␤ integrin concentration determination errors, leading to a maximum total K d error due to concentration errors of ϳ20% (corresponding to a maximum error in ⌬G of 0.5 kJ/mol).
Some K d values are reported as approximate because binding was too weak for the generation of a binding curve. In these cases, the value of ⌬(H,N) max was estimated by comparing maximum ⌬(H,N) values for relevant peaks to the ⌬(H,N) of peaks in a corresponding quantifiable titration. The fitting procedure was then carried out as before, but with the value of ⌬(H,N) max restrained. No errors are reported for these values, as they are only estimates.
Immunofluorescence Imaging-After transfection, MEF cells were plated on 7.5 g/ml of fibronectin-coated coverslips, allowed to adhere for 90 min in Dulbecco's modified Eagle's medium, rinsed once in phosphate-buffered saline, and fixed with 3.7% formaldehyde in phosphate-buffered saline. After fixation, cells were permeabilized in 0.1% Triton X-100 for 5 min, blocked with 3% bovine serum albumin, 2% normal goat serum for 1 h, and then incubated with the appropriate primary antibody in blocking solution overnight at 4°C. Bound antibodies were detected by the corresponding fluorescein isothiocyanate-conjugated goat secondary antibodies (Santa Cruz Biotechnology). Coverslips were subsequently mounted in Prolong Gold antifade reagent (Invitrogen) on slides. Epifluorescent images of cells were acquired with a ϫ60 oil immersion objective on a Nikon Eclipse TE2000-U microscope equipped with the appropriate excitation and emission filter sets (Semrock). Additional post-acquisition processing of images was performed using ImageJ (www.rsb.info.nih.gov/ij) and Adobe Photoshop.
Anti-green fluorescent protein (GFP) rabbit polyclonal antibody was obtained from Clontech. Anti-vinculin mouse monoclonal antibody was purchased from Sigma. Anti-phosphotyrosine (Tyr(P) 100 ) mouse monoclonal antibody was purchased from Cell Signaling Technology. Anti-paxillin rabbit polyclonal antibody (RB4536) was developed in house.

Production of Tyrosine-phosphorylated Integrin Tails for
NMR-Before the current study, structural work on integrin phosphorylation had involved short chemically synthesized integrin tail fragment peptides (43). However, we recently reported a robust system for studying integrin tail protein-protein interactions by NMR using full-length 15 N-labeled integrin tails produced in E. coli. This system is cost-effective and versatile, but using such a system to produce tyrosine-phosphorylated peptides presents additional difficulties; modified residues can be incorporated directly during chemical peptide synthesis, but not in E. coli. Although glutamate can be introduced by mutagenesis to make an acceptable mimic for phosphoserine or phosphothreonine, phosphotyrosine has no natu-ral analogue (supplemental Fig. S1B). Thus, the integrin tails would have to be phosphorylated directly.
After exploring other options (supplemental Fig. S2), we were able to produce NMR scale quantities of tyrosine-phosphorylated 15 N-labeled integrin tails by in vitro phosphorylation with c-Src, which has been used for in vitro phosphorylation of integrin tails in previous studies (14,39), although not previously on a scale large enough for structural biology. Because tyrosine kinases can be toxic to bacteria, we produced the Src kinase domain in E. coli by coexpressing it with YopH phosphatase, as previously described by Seeliger et al. (45).
To phosphorylate specific tyrosine residues selectively, single and double Tyr to Phe mutants were made: ␤3 Y747F (for Tyr(P) 759 ) and Y759F (for Tyr(P) 747 ), ␤1A Y783F (for Tyr(P) 795 ) and Tyr 795 (for Tyr(P) 783 ), and ␤7 Y753F/Y758F (for Tyr(P) 778 ). The ␤7 tail contains two MP tyrosine residues, but phosphorylation of just the nMD tyrosine residue was explored (the fMD site contains a natural Tyr to Phe substitution). Tyrosine phosphorylation caused localized perturbations in the HSQC spectra of the integrin tails (Fig. 3). Phosphorylation of  DECEMBER 25, 2009 • VOLUME 284 • NUMBER 52 both ␤3 Y747F and Y759F was observed, as was phosphorylation of ␤1A Y795F and ␤7 Y753F/Y758F. However, phosphorylation of ␤1A Y783F was not observed. When the phosphorylation reaction was performed on ␤1A wild type (WT), chemical shift perturbations were only observed near Tyr 783 , indicating that Tyr 783 , but not Tyr 795 , was phosphorylated in this system. Thus, for further studies, ␤1A Tyr(P) 783 was produced from WT peptides.

Integrin Tyrosine Phosphorylation Regulates Activation
Tyrosine Phosphorylation Decreases Integrin Affinity for Talin-Binding of proteins to integrin tails was assayed by observing chemical shift perturbations in integrin tail HSQC spectra. Upon the addition of the talin1 F3 domain to the ␤3, ␤1A, or ␤7 tails, significant perturbations were observed in the MP and nMD portions of the tail (Fig. 4). The MP perturbations were greatest in the ␤3 tail, but they were present in all tails tested. The affinities of these interactions were quantified, giving K d values that ranged from 142 M for ␤7 to 273 M for ␤3 to 491 M for ␤1A (Table 1). Smaller shifts were also observed in the fMD portion of the ␤3 integrin tail (and to a lesser extent in the ␤1A tail), but these perturbations are likely due to a weak competing integrin/talin interaction, with a K d of several millimolar for the fMD portion.
The introduction of Tyr to Phe mutations employed for specific phosphorylation in ␤3 and ␤7 had negligible effects on talin affinity in these integrins. The greatest effect was seen with Y747F in ␤3, which increased the K d to 366 M (see Table 1). Such a decrease in affinity upon a mutation in the talin-binding site is not surprising, and it is notable that this effect is very small compared with the much greater changes in affinity observed elsewhere in this study.
Tyrosine phosphorylation of the nMD NPXY motif decreased the magnitude of chemical shift perturbations observed upon talin binding (Fig. 4) and decreased the affinity of these interactions. This effect was most pronounced in ␤7, less so in ␤1A, and least in ␤3. Phosphorylation of the fMD NPXY motif in ␤3 abrogated chemical shift perturbations upon talin binding in that region, but had little effect on the much tighter interaction with the MP and nMD regions or the overall affinity (see Table 1).
Tyrosine Phosphorylation Increases Integrin Affinity for Dok1-Unlike the talin1 F3 domain, the PTB domain of Dok1 only caused small perturbations in the HSQC spectra of unphosphorylated integrin tails. Such perturbations were localized to the nMD region of ␤3 and the fMD regions of ␤1A and ␤7 (supplemental Fig. S5). However, these interactions were so weak (K d greater than several millimolar) that it is unlikely that any of these interactions are physiologically relevant (Table 2). It is of note that although the K d value for ␤3 (12.6 mM) is only an estimate, as described under "Experimental Procedures," it agrees well with the 14.3 mM value reported for a short fragment of ␤3 (43).
Upon phosphorylation, the affinities of these interactions increased substantially. For phosphorylation in the nMD regions, K d values ranged from 20.9 M for ␤3 to 36.8 M for ␤7 to 78.7 M for ␤1A (see Table 2 and supplemental Fig. S4). Phosphorylation of ␤3 Tyr 759 also increased Dok1 affinity to a K d of 226 M. In each case, the interaction as observed by NMR Interaction studies were performed on unphosphorylated integrin tails and tails phosphorylated at the nMD site (␤3, ␤1A, and ␤7) and at the fMD site (␤3). Gray bars correspond to residues that could not be tracked due to exchange broadening. Note that the y axis scale differs between panels. was localized to residues near the site of phosphorylation. No MP perturbations were observed in ␤3 or ␤1A, and only minor MP perturbations in ␤7 (Fig. 5). Tyrosine phosphorylation thus greatly increases the affinity of Dok1 for integrin tails (adding 16 kJ/mol of binding energy to the interaction in the case of ␤3 Tyr(P) 747 ), making the interaction tight enough to be physiologically relevant and significantly tighter than the competing talin/integrin interaction. Experiments were also conducted on a ␤3 Y747E mutant. Glutamate is occasionally reported as a phosphotyrosine mimic, despite very little chemical similarity to phosphotyrosine (supplemental Fig. S1B). The chemical shift perturbation pattern of this mutant upon interaction with Dok1 closely resembled that of ␤3 Y747A (supplemental Fig. S6). In both cases, interaction with the nMD NPXY region was abrogated. This stands in stark contrast to the significant increase in chemical shift perturbations observed upon phosphorylation of Tyr 747 ; this strongly suggests that glutamate is an unwise choice as a phosphotyrosine mimic.
Positively Charged Residues in the NPXY-binding Pocket Make Dok1 Specific for Phosphorylated Integrin Tails-When we solved the crystal structure of the human Dok1 PTB domain (Protein Data Bank 2V76) (43), we found a sulfate anion in the NPXY-binding pocket, surrounded by a collection of positively charged residues: Arg 207 , Arg 208 , Arg 222 , and Arg 223 . Arg 207 and Arg 222 make the most direct contact (Fig. 2A). We hypothesized that these positively charged residues may explain the higher affinity of Dok1 for phosphorylated ␤3; we tested this hypothesis by individually mutating these residues to alanine. These mutations only had minor effects on the affinity of Dok1 for unphosphorylated ␤3 tail, but the effect was much more pronounced on the interaction with ␤3 tail phosphorylated at Tyr 747 (Table 3). R207A had the greatest effect (increasing the K d of the interaction with phosphorylated ␤3 from 20.9 to 1,485 M), followed by R222A (increasing the K d to 398 M).
The degree to which each residue contributes to Dok1 phosphotyrosine specificity was calculated by subtracting the change in ⌬G caused by the Arg to Ala mutation for the inter-action with the unphosphorylated tail from the change in ⌬G for the interaction with the phosphorylated tail (yielding the value ⌬⌬G RA/pY in Table 3); the additive loss of phosphotyrosine-specific affinity for all four mutations (20.8 kJ/mol) is more than enough to explain the 15.9 kJ/mol increase in binding affinity of Dok1 upon ␤3 phosphorylation at Tyr 747 . Arg 207 and Arg 222 are the greatest contributors to phosphotyrosine specificity (44 and 30%, respectively), but Arg 223 and Arg 208 also play a role (17 and 9%, respectively).
Residues Arg 207 and Arg 208 correspond to positively charged residues in talin, but Arg 222 and Arg 223 do not. Residue Arg 223 corresponds to Tyr 373 in talin1 and Tyr 376 in talin2, a residue that plays a significant role in binding integrins, particularly ␤1 integrins. 4 Interestingly, however, Arg 222 corresponds to a negatively charged residue (Asp 372 in talin1 and Glu 375 in talin2, see Fig. 2). Thus, Arg 222 was chosen as a particularly suitable candidate for exploring differences in specificity for phosphorylated integrin tails between Dok1 and talin.

Engineering Talin to Bind Preferentially to Phosphorylated
Integrin Tails-As discussed above, Dok1 Arg 222 has a reversed charge in talin (Asp 372 in talin1 and Glu 375 in talin2) and is a key residue for determining Dok1 specificity for phosphorylated integrins. Consistent with this, introduction of the mutation D372R in talin1 substantially increased its binding affinity for integrin tails phosphorylated at the nMD site: 33.1 M for ␤3 Tyr(P) 747 and 128 M for ␤1A Tyr(P) 783 (Table 4).
Interestingly, this mutation also affected talin1 binding to unphosphorylated integrin tails in different ways, although to a lesser degree. Talin1 D372R binds more tightly to unphosphorylated ␤3 than does talin1 WT (86 versus 273 M), but binds more weakly to unphosphorylated ␤1A than talin1 WT (793 versus 491 M). These differences can be explained based on recent structural data (see Fig. 2) and could influence the biological activity of this mutant.
Talin D372R Localizes to Phosphorylated Integrins in Live Cells-Due to its dramatic effect on talin binding to phosphorylated integrins, we hypothesized that the D372R mutation would affect talin activity in vivo; we investigated this by examining talin localization to focal adhesions. We transiently expressed GFP-talin1 WT or D372R in SYF MEFs (deficient for the tyrosine kinases Src, Yes, and Fyn), which were then plated on fibronectin (an extracellular ligand for ␣5␤1) and stained for the focal adhesion marker vinculin. Talin1 D372R was abundantly expressed (Fig. 6D) but was not seen at the sites of focal adhesions, whereas talin1 WT co-localized with vinculin and was therefore present in focal adhesions (Fig. 6A). Focal adhesions were also more prominent in cells expressing talin1 WT compared with those expressing talin1 D372R.
In SYF MEFs stably reconstituted with c-Src (SYF ϩ Src), talin1 WT was also targeted to focal adhesions, but these adhesions were less prominent than in cases where c-Src was absent (Fig. 6B). However, talin1 D372R in SYF ϩ Src MEFs was targeted to focal adhesions, and these were more prominent than those observed in SYF MEFs expressing talin1 D372R or SYF ϩ Src MEFs expressing talin1 WT. To confirm the levels of phosphorylation in SYF MEFs and SYF ϩ Src MEFs, fixed cells were stained with an anti-phosphotyrosine antibody (Tyr(P) 100 ), and an anti-paxillin antibody to mark focal adhesions (Fig. 6C). Greater phosphorylation was observed in the focal adhesions of SYF ϩ Src MEFs when compared with SYF MEFs. In agreement with our structural model, these data suggest that talin1 D372R is capable of competing with endogenous talin for integrin binding only when integrins are tyrosine phosphorylated. Furthermore, tyrosine phosphorylation appears to affect cell morphology in a manner that is reversed by the D372R mutation.

DISCUSSION
Here, we have demonstrated that key interactions involving the ␤3, ␤1A, and ␤7 integrin cytoplasmic tails are similarly affected by tyrosine phosphorylation. Phosphorylation at the nMD site in each tail decreases its affinity for talin; in contrast, phosphorylation greatly increases the affinity for Dok1 (by 15.9 kJ/mol in the case of ␤3). The interaction of Dok1 is localized to the NPXY region of the integrin tail, even when the affinity is relatively high (20.9 M for ␤3 Tyr(P) 747 ). Talin, on the other hand, also binds to the MP region, a unique interaction that is essential for integrin activation (8). Thus, tyrosine phosphorylation acts to decrease integrin activation both by decreasing WT (a positive value denotes a decrease in affinity upon mutation). d ⌬⌬G pY (kJ/mol) is the ⌬G value for the phosphorylated integrin binding to Dok1, minus the ⌬G value for the unphosphorylated integrin binding to Dok1 (a negative value denotes an increase in affinity upon phosphorylation) (pY, Tyr(P)). e ⌬⌬G RA/pY (kJ/mol) is the ⌬⌬G pY value for the integrin binding to mutant Dok1, minus the ⌬⌬G pY value for binding to Dok1 WT. f ⌬⌬G mut (%) is the percentage of phosphorylated integrin-specific binding energy lost by the given mutation. g Approximate K d values were estimated by comparing magnitude of chemical shift perturbations to those in a relevant titration, as described under "Experimental Procedures." a K d values are given Ϯ S.E. b ⌬G is given for binding and calculated from K d . c ⌬⌬G DR is the ⌬G value for integrin binding to talin1 D372R, minus the ⌬G value for binding to talin1 WT (a negative value denotes an increase in affinity upon mutation). d ⌬⌬G pY is the ⌬G value for the phosphorylated integrin binding to talin1, minus the ⌬G value for the unphosphorylated integrin binding to talin1 (a negative value denotes an increase in affinity upon phosphorylation) (pY, Tyr(P)).
talin affinity and increasing the affinity of competing proteins incapable of activating the integrin. This is consistent with our recent reports on ␤3 (43) and an early report on ␤1 (13), although our current report is the first detailed analysis of tyrosine phosphorylation across different integrins using structural biological methods. By performing these studies on intact full-length integrin tails, we have been able to show definitively where these interactions are localized within the cytoplasmic tails, and we can look at the effect of independently phosphorylating different tyrosine residues within the same peptide. In ␤3, we could phosphorylate either Tyr 747 or Tyr 759 ; in either case, the interaction with Dok1 was localized only to the site of phosphorylation. Whereas phosphorylation at Tyr 747 disrupted the interaction with talin, phosphorylation at Tyr 759 did not. Thus, the role of phosphorylation at this fMD site is probably not related to integrin activation; the integrin tail is very flexible in this region (50), and binding of Dok1 here would not be expected to compete with talin binding. The effect of phosphorylation on Dok1 binding was an order of magnitude higher at the Tyr 747 site in ␤3, also demonstrating that the nMD site may play a more central role in Dok1 signaling.
This is the first publication of the talin-induced chemical shift perturbation map for ␤7, a lymphocytespecific integrin (51), expanding our studies performed on ␤3 and ␤1A (44). Although showing some subtle differences to the interaction of talin with ␤3 and ␤1A, the interaction with ␤7 is broadly similar in that it consists of both MP and nMD interactions. In fact, the chemical shift perturbation map appears intermediate between those of ␤3 and ␤1A. In studying talin binding, it was found to be particularly important to use full-length peptides, given the large interaction surface. For example, in our previous study we reported the K d of the interaction between talin and a short ␤3 peptide to be 3.49 (unphosphorylated) and 6.53 mM (phosphorylated) (43), whereas the values for the full-length ␤3 peptide reported here are, respectively, 0.273 and 1.03 mM. With respect to Dok1, which engages a more limited interaction surface, the values are more comparable between truncated and full-length tails.
These experiments could only be carried out on phosphorylated integrin tails, as a suitable phosphomimetic mutation does not exist for studying tyrosine phosphorylation. In fact, we show here that mutating Tyr 747 in ␤3 to glutamate has the same effect as mutating that residue to alanine, which effectively abrogates protein-protein interactions in that region FIGURE 6. Talin D372R preferentially localizes to focal adhesions that are tyrosine phosphorylated. A, SYF MEFs and B, SYF ϩ Src MEFs transiently expressing GFP-talin1 WT or D372R were allowed to adhere to fibronectin-coated coverslips and stained to visualize vinculin. Depicted are the localization of talin (green) and vinculin (red). C, SYF MEFs and SYF ϩ Src MEFs were stained to visualize phosphotyrosine (Tyr(P) 100 , red) and paxillin (green). D, SYF cells expressing GFP-talin1 and GFP-talin1 D372R were lysed and analyzed by Western blotting to confirm comparable expression of D372R and WT talin1.
(supplemental Fig. S6). This is observed for interactions with Dok1 and talin1, but in the case of Dok1 phosphorylation of Tyr 747 strongly enhances binding, demonstrating that in this system at least, glutamate does not mimic phosphotyrosine. Any use of glutamate or aspartate as a phosphotyrosine mimic should therefore be extensively validated, at the very least by comparing it to the effect of an alanine substitution. The results of a rudimentary search through recent literature suggest that this is not common practice, and recent studies that did compare a "phosphomimetic" Tyr to Glu (or Tyr to Asp) mutant with a Tyr to Ala mutant found that the two different mutations had the same effect on the system under study (52,53), meaning that the observed effects cannot necessarily be interpreted as being phosphorylation specific. This should not be a surprising result, given the lack of structural similarity between glutamate and phosphotyrosine (Fig. 1B).
Here, we show that mutation of positively charged residues in the NPXY-binding pocket of Dok1 ( Fig. 2A) (43) disrupts Dok1 binding to the ␤3 integrin tail in a phosphorylation-specific manner. It has been reported that the mutation of two of these residues to alanine (R207A/R208A) decreases Dok1 binding to ␤3 and disrupts Dok1 signaling (54). We can now explain this observation in terms of interference with the positively charged pocket necessary for specific binding of phosphorylated integrins, particularly Arg 207 , which explains 44% of Dok1 phosphotyrosine specificity ( Table 3). The analogous residues in talin1 and talin2, however, are also positively charged, so this does not explain why Dok1 binds specifically to phosphorylated integrins, whereas talin does not. Another residue in the Dok1 NPXY-binding pocket, Arg 222 , is oppositely charged in talin (Asp 372 in talin1 and Glu 375 in talin2). Mutation of this residue in Dok1 to alanine significantly decreases the affinity of the interaction with the ␤3 integrin in a phosphorylation-dependent manner, and is the largest contributor to Dok1 phosphotyrosine specificity after Arg 207 .
We hypothesized that this understanding of the structural basis of the phosphorylation dependence of Dok1 could be used to engineer a talin variant that would bind specifically to phosphorylated integrins. Indeed, this effect was observed for talin1 D372R. This mutation increased talin affinity for phosphorylated ␤3 by 8.5 kJ/mol and ␤1A by 7.4 kJ/mol. Interestingly, this mutant had differing effects on the interaction with unphosphorylated integrins: it increased talin1 affinity for unphosphorylated ␤3 by 2.9 kJ/mol but decreased affinity for ␤1A by 1.2 kJ/mol. An examination of the structures of ␤3 and ␤1D in complex with talin (see Fig. 2) explains this difference. In the ␤3/talin1 structure (PDB 1MK9) (55), the portion of the tail C-terminal to Tyr 747 does not make extensive contacts with talin. However, this region contains a negatively charged glutamate residue (Glu 749 ) that could make favorable electrostatic contacts with the mutant D372R. ␤1A, on the other hand, has an uncharged serine residue (Ser 785 ) in this position. In the ␤1D/talin2 structure (PDB 3G9W) (44), Ser 785 forms a hydrogen bond with talin2 Glu 375 . In the similar ␤1A⅐talin1 complex this mutation would disrupt such an interaction if present and would not be expected to make the interaction more favorable as in the case of ␤3. This interaction between the ␤1 Tyr 783 side chain and a negatively charged talin residue may also explain why the introduction of a phosphate group here has a slightly greater disruptive effect on the ␤1A/talin1 interaction than on the ␤3/talin1 interaction (see Table 1). In addition, the Glu 749 -Ser 785 substitution may explain why Dok1 has a higher affinity for ␤3 than for ␤1A or ␤7 (which also has an uncharged residue, Ser 780 , at this position), as Glu 749 in ␤3 would form favorable electrostatic interactions with the positively charged Dok1binding pocket.
We tested aspects of our structural model by observing the behavior of talin1 WT and D372R in live cells. In MEFs that do not express Src or related kinases, talin1 WT localized to focal adhesions, but talin1 D372R did not. As these experiments were carried out using fibronectin, and ␣5␤1 is the primary fibronectin receptor in MEFs (56), this effect correlates with the decreased affinity observed between mutant talin and unphosphorylated ␤1A integrins. When c-Src was introduced into these cells, both talin1 WT and D372R co-localized with phosphorylated integrins at focal adhesions. Thus, the effects observed in in vitro binding experiments translate into observable effects in live cells. Interestingly, tyrosine phosphorylation correlates with reduced focal adhesion formation in the presence of talin1 WT but not talin1 D372R. We did not assess integrin inside-out activation levels directly, but the increased prominence of focal adhesions in c-Src-expressing cells in the presence of talin1 D372R versus talin1 WT provides evidence that this talin mutant disrupts the integrin activation phosphorylation switch, causing integrins to remain active even after integrin tyrosine phosphorylation. Thus, these results provide additional evidence that integrin tyrosine phosphorylation down-regulates integrin activation by inhibiting talin binding.
Phosphorylation of the nMD tyrosine residue has been shown to modulate inside-out integrin activation in ␤3 (42) and in ␤1 (13). In ␤3, phosphorylation of both nMD Tyr 747 (30, 31, 34 -36) and fMD Tyr 759 (38,39,41) have been associated with outside-in signaling. This is consistent with a report that the nMD region of ␤3 in general is associated with both inside-out and outside-in signaling, whereas the fMD portion primarily engages only in outside-in signaling (57). This makes sense from a structural standpoint, in that talin only interacts tightly with the nMD NPXY. For ␤1, studies point to a role for phosphorylation at nMD Tyr 783 in signaling in both directions. Evidence for a signaling role for Tyr 795 phosphorylation is sparser, but also exists (28,29). However, under the conditions used in this study, c-Src did not phosphorylate this residue, so a different kinase may be involved in vivo. Interestingly, in ␤7 this fMD residue is a non-phosphorylatable phenylalanine.
Despite the substantial evidence for a role for tyrosine phosphorylation in ␤1 integrin signaling, mice with the ␤1 Y783F/ Y795F knock-in mutation do not display any apparent developmental abnormalities (32,33). The ␤3 Y747F/Y759F mouse, however, displays a severe phenotype (31). ␤1 tyrosine phosphorylation is central to the pathological effect of v-Src on cells (12,26,27), but given that these residues are so highly conserved across different integrins and across different species, it is unlikely that such conservation would exist if the residue only participated in pathological conditions. In the case of ␤3, at least, mutation of either NPXY tyrosine to phenylalanine had little effect on the interaction with talin1, so it is reasonable to hypothesize that these residues are conserved as tyrosine because of a role for phosphorylation, although the side chains of unphosphorylated ␤1D 4 Tyr 783 and ␤3 Tyr 747 (55) do also participate in intermolecular hydrogen bonds in complex with talin. The ␤1 Y783F/Y795F mutation affects cell behavior in tissue culture conditions (28,29), so the lack of a phenotype in ␤1 Y783F/Y795F knock-in mice may be the result of compensation by other integrins. On the other hand, the more extreme effect of tyrosine phosphorylation on Dok1 binding to ␤3 correlates with the more definitive biological role observed for tyrosine phosphorylation in that integrin.
Several studies have provided evidence for a biological role for a direct interaction between Dok1 and ␤3 integrins (8,54,58,59), but what role Dok1 plays in the signaling of other integrins remains more of an open question. A variety of other PTB domains have been identified that also interact with integrins (9), and it is possible that one of these other proteins acts as a phosphorylation-dependent activation switch for ␤1 or other integrins. The PTB domain of integrin cytoplasmic domainassociated protein 1 (ICAP-1), for example, has been identified as a phosphorylation-independent down-regulator of ␤1 activation, but given the large number of PTB domains that bind preferentially to phosphorylated substrates (60), other phosphorylation-dependent down-regulators may also exist. Given this uncertainty, however, and the results of the ␤1 Y783F/ Y795F knock-in mouse (32,33), this is an area that will require further study.
From the current study, and from the bulk of structural and biological data on the topic, integrin tyrosine phosphorylation appears to be involved in a wide variety of integrin signaling processes, particularly for ␤3, but for other integrins as well. We demonstrate here a conserved structural mechanism in ␤3, ␤1A, and ␤7 integrins for regulation of integrin activation by nMD tyrosine phosphorylation. We have tested our predictions by engineering a talin mutant that is specific for phosphorylated integrins, and we have shown that this influences talin localization in live cells. Given that the literature on integrin tyrosine phosphorylation is substantial but sometimes ambiguous, our results add weight to the idea that tyrosine phosphorylation plays a significant role in integrin signaling.