Spatial Coordination of Kindlin-2 with Talin Head Domain in Interaction with Integrin β Cytoplasmic Tails*

Background: The talin and kindlin play indispensable roles in integrin activation. Results: The C-terminal 12 amino acids of β1 and β3 integrins mediate kindlin-2 binding. Conclusion: Kindlin-2 binding to the extreme C terminus allows β subunits to accommodate both kindlin-2 and talin. Significance: Binding of talin and kindlin-2 to distinct sites in integrins regulates receptor activation, a pivotal event in cellular responses. Both talin head domain and kindlin-2 interact with integrin β cytoplasmic tails, and they function in concert to induce integrin activation. Binding of talin head domain to β cytoplasmic tails has been characterized extensively, but information on the interaction of kindin-2 with this integrin segment is limited. In this study, we systematically examine the interactions of kindlin-2 with integrin β tails. Kindlin-2 interacted well with β1 and β3 tails but poorly with the β2 cytoplasmic tail. This binding selectivity was determined by the non-conserved residues, primarily the three amino acids at the extreme C terminus of the β3 tail, and the sequence in β2 was non-permissive. The region at the C termini of integrin β1 and β3 tails recognized by kindlin-2 was a binding core of 12 amino acids. Kindlin-2 and talin head do not interact with one another but can bind simultaneously to the integrin β3 tail without enhancing or inhibiting the interaction of the other binding partner. Kindlin-2 itself failed to directly unclasp integrin α/β tail complex, indicating that kindlin-2 must cooperate with talin to support the integrin activation mechanism.

Both talin head domain and kindlin-2 interact with integrin ␤ cytoplasmic tails, and they function in concert to induce integrin activation. Binding of talin head domain to ␤ cytoplasmic tails has been characterized extensively, but information on the interaction of kindin-2 with this integrin segment is limited. In this study, we systematically examine the interactions of kindlin-2 with integrin ␤ tails. Kindlin-2 interacted well with ␤ 1 and ␤ 3 tails but poorly with the ␤ 2 cytoplasmic tail. This binding selectivity was determined by the non-conserved residues, primarily the three amino acids at the extreme C terminus of the ␤ 3 tail, and the sequence in ␤ 2 was non-permissive. The region at the C termini of integrin ␤ 1 and ␤ 3 tails recognized by kindlin-2 was a binding core of 12 amino acids. Kindlin-2 and talin head do not interact with one another but can bind simultaneously to the integrin ␤ 3 tail without enhancing or inhibiting the interaction of the other binding partner. Kindlin-2 itself failed to directly unclasp integrin ␣/␤ tail complex, indicating that kindlin-2 must cooperate with talin to support the integrin activation mechanism.
Activation of integrins is a key mechanism to control many cellular adhesive responses including platelet aggregation, leukocyte trafficking, and endothelial cell migration (1,2). At a molecular level, the integrin activation process is controlled tightly by the heterodimeric ␣/␤ cytoplasmic tail (CT) 3 complex formed between the membrane-proximal regions of each integrin subunit, and this complex extends into their trans-membrane domains (3,4). The ␣/␤ CT complex restrains the integrin in a resting, low affinity state for extracellular ligands, and a process to unclasp the complex is required to initiate integrin activation (5).
Mechanistically, integrin activation is balanced by multiple CT-associating proteins, including both activators and inhibitors, and the coordination of these positive and negative regulators, together with post-translational modifications of the ␤ CT, defines the clasping state of integrin CT, and thereby integrin activation (6 -11). Among many integrin CT binding proteins, the head domain of the cytoskeletal protein talin (talin-H) is a well characterized integrin activator (12,13). Talin-H resides at the N terminus of talin and resembles a FERM domain, which contains four subdomains (F 0 , F 1 , F 2 , and F 3 ). The integrin binding sites in talin is auto-inhibited but can be exposed by calpain cleavage or binding partners, which allows its F3-phosphotyrosine binding subdomain to directly interact with integrin ␤ CT and unclasp the CT complex to induce integrin activation (14 -21). However, quantitative analysis indicated that talin-H alone is insufficient to induce efficient integrin activation, and a co-activation model was proposed (6,22). Within recent years, it has been found that kindlins could dramatically enhance talin-mediated integrin activation in model cells and were also essential to support efficient integrin activation in vivo, verifying a co-activation mechanism (23)(24)(25). Indeed, deficiencies of specific kindlins are associated with diseases in human, which reflect an inability to effectively activate integrins (26 -29). The kindlin family contains three members in mammals, and among them, kindlin-2 is expressed ubiquitously and enriched in the integrin-containing adhesion complexes (30,31). Interestingly, kindlin family members also contain a FERM-like domain, which is preceded by a distinct F 0 subdomain at the N terminus and with a pleckstrin homology domain inserted in the middle of its F 2 subdomain, both of which are potential binding sites for membrane phosphoinositides (32)(33)(34) While sharing a structurally similar FERM domains, mutations in the integrin ␤ CTs have distinct effects on kindlin and talin-H binding (23-25, 35, 36).
To further understand the regulatory role of kindlin in integrin activation, we have measured the interaction of kindlin-2 to integrin and detected an unexpected binding preference for different integrin ␤ CTs. Further detailed mapping using synthesized peptides defined a 12-amino acid kindlin-2 binding core in the C termini of integrin ␤ CT. Residues at the extreme C terminus of this binding core establish the preference of kindlin-2 for different integrin ␤ CTs. Kindlin-2 alone could not unclasp the integrin CT complex, demonstrating that kindlin-2 is dependent upon talin-H to support integrin activation. Finally, despite the relatively large size of kindlin-2 and talin-H, we demonstrate that the short integrin ␤ CTs can accommodate both co-activators simultaneously to achieve integrin coactivation, a transition that governs many cellular responses, including platelet aggregation inflammatory cell recruitment and tumor metastasis (37)(38)(39).

EXPERIMENTAL PROCEDURES
Protein Preparation and Peptide Synthesis-GST-fused integrin ␤ CT, full-length kindlin-2 and His-tagged talin-H(1-429) were expressed, purified, and quantified as described previously (10). Talin-F 2 F 3 (206 -405) was prepared as described previously (40). All synthetic peptides were prepared, purified, and authenticated by HPLC and MALDI TOF-TOF (matrixassisted laser desorption ionization time-of-flight) mass spectrometry in the Molecular Biotechnology Core of the Cleveland Clinic.
Pulldown Assays and Western Blotting-Pulldown assays were performed using GST fusion proteins. Equal amounts of GST-fused integrin ␤ CTs were added together with glutathione-Sepharose 4B (GE Healthcare) to aliquots of the cell lysates from either EGFP-kindlin-2-expressing CHO cells or HUVECs. In peptide inhibition experiments, the indicated peptide was added to the slurries at the selected concentrations. After overnight incubation in a cold room, the precipitates were washed and boiled in Laemmli sample buffer. The eluates were analyzed on gradient acrylamide gels under reducing conditions, and interactions of the integrin ␤ CT with kindlin-2 were determined by Western blotting. The gels were also stained with Coomassie blue to verify that sample loadings were similar.
Surface Plasmon Resonance (SPR)-Real time protein-protein interactions were analyzed using a Biacore3000 instrument (Biacore, Uppsala, Sweden). Purified proteins were immobilized on the carboxymethyl dextran of CM5 biosensor chip using the standard amine coupling chemistry, and biotinylated peptides were captured on a SA sensor chip (Biacore) according to the manufacturer's instructions. Experiments were performed at room temperature in 10 mM HEPES (pH 7.4) buffer containing 150 mM NaCl and 0.005% surfactant P20 at a flow rate of 25 l/min. SPR sensograms were obtained by injecting various concentrations of analyte. The chip surfaces were regenerated by injecting a short pulse of 5 mM NaOH. The resulting sensograms were analyzed in overlay plots using BIAevaluation software (version 4.01, GE Healthcare). The association (k a ), dissociation (k d ) rates constant and the affinity constant (K A and K D ) were calculated by global fitting (1:1 Langmuir model).
Isothermal Titration Calorimetry-Purified talin-H and kindlin-2 proteins were dialyzed into 50 mM phosphate buffer (pH 8.0). Talin-H at the concentration of 1.4 mM was titrated into the solution of kindlin-2 at a 70 M concentration in a Microcal iTC200 instrument (GE Healthcare). Isothermal titration calorimetry titration curves were obtained by tirating talin-H in 10 3-l injections with 70 M kindlin-2 solution in the cell. Titration curves were collected to estimate heat changes associated with the kindlin-2/talin-H interaction.
NMR Spectroscopy-Two-dimensional transferred NOESY experiments were performed with mixing time of 400 ms on Bruker Avance 600 MHz at 25°C as described (16). The spectra were processed and visualized using nmrPipe (41). Two-dimensional HSQC experiments used to examine the integrin ␤ CT-target binding were described previously (16).
Cell Spreading-To test the function of the kindlin-2 binding core in a cellular system, a series of chimeric PSGL-1/␤3 constructs were expressed in ␣ IIb ␤ 3 -CHO cells, and their effects of cell spreading were assessed. The chimera consisted of the extracellular and transmembrane domains of PSGL-1 (mPSGL-1) fused to selected segments of the ␤ 3 CT. This approach, described previously (23), allows expression of functional segments or mutants of the ␤ 3 CT in the cytosol while tethered to PSGL-1, which allows for monitoring of expression levels. The specific chimeric proteins were , and these were transfected into ␣ IIb ␤ 3 -CHO cells. In short, the transiently transfected cells were allowed to adhere and spread on fibrinogen-coated (20 g/ml) glass coverslips. After incubation at 37°C for 2 h, the wells were washed three times with PBS, and the adherent cells were fixed with 4% paraformaldehyde and stained with Alexa Fluor 647 phalloidin (Invitrogen). To identify PSGL-1-expressing cells, the fixed cells were stained by anti-PSGL-1 mAb, KPL-1 (BD Biosciences), followed by goat anti-mouse IgG conjugated with Alexa Fluor 488 (Invitrogen). As controls, nontransfected cells were included in each experiment and always showed no PSGL-1 staining. The positively stained (green) cells were visualized with a ϫ40 oil immersion objective using a Leica TCS-NT laser scanning confocal microscope (Imaging Core, Cleveland Clinic). Cell area was analyzed with ImageJ software.
Statistical Analysis-Two-tailed Student's t tests were performed using SigmaPlot (version 11). Differences were considered to be significant with p Ͻ 0.05.

Kindlin-2 Exhibits Selective Binding to Integrin ␤ Cytoplasmic Tails-Previous
studies have demonstrated that the ability of kindlin-2 to bind and activate ␤ 1 and ␤ 3 integrins depends on its interaction with the membrane-distal NxxY motif in the ␤ subunit of these integrins (23)(24)(25). Because the membrane distal NxxY/F motif is conserved across different integrin ␤ subunits (Fig. 1A), we postulated that kindlin-2 might bind to other integrin ␤ CTs similarly. Surprisingly, we failed to detect interaction of expressed kindlin-2 in cell lysates with GST-fused integrin ␤ 2 CT in pulldown assays (Fig. 1B). As expected, kindlin-2 could be precipitated by integrin ␤ 1 or ␤ 3 CTs under the same condition. The sequence of all three ␤ CTs, including the ␤ 2 CT, was confirmed, and this preference in kindlin-2 binding was observed with multiple preparations of the ␤ 2 CT. To verify these results, we measured the direct binding of purified kindlin-2 protein to integrin ␤ CTs. Kindlin-2 was expressed as a GST fusion protein in Escherichia coli and purified as described previously (Fig. 1C) (10). Real time binding of kindlin-2 to fulllength integrin ␤ CTs was monitored by SPR in which kindlin-2 protein was injected and allowed to flow over immobilized fulllength ␤ 1 , ␤ 2 , or ␤ 3 CT. As shown in Fig. 1, D and F, kindlin-2 bound to the ␤ 1 and ␤ 3 CTs. Consistent with the pulldown results, only minimal binding of kindlin-2 to immobilized ␤ 2 CT was detected (Fig. 1E). The dissociation constants for kindlin-2 with the full-length ␤ 1 and ␤ 3 CTs were similar at 1.77 ϫ 10 Ϫ7 and 1.36 ϫ 10 Ϫ7 M, respectively. Based on the concentrations used, we estimated that the affinity of kindlin-2 for the ␤ 2 CT was at least 100-fold lower than for the ␤ 1 or ␤ 3 CT. These results suggest that kindlin-2 can interact selectively with different integrin ␤ CTs and indicate that the conserved motifs in the integrin ␤ CTs are insufficient to determine kindlin-2 recognition.
Identification of Kindlin-2-binding Site in Integrin ␤ Cytoplasmic Tails-We previously implicated residues in the C-terminal region of the integrin ␤ 1 and ␤ 3 CTs in kindlin-2 binding ( Fig. 2A) (10,23). To more precisely define the requirements for their interaction, we directly immobilized these ␤ C-terminal peptides conjugated with biotin at their N termini onto SA biosensor chips for SPR analysis. As shown in Fig. 2B, kindlin-2 bound to the immobilized ␤ 1 and ␤ 3 C-terminal peptides, but the association with the ␤ 2 C-terminal peptide was significantly lower, confirming the interactive selectivity of kindlin-2 to integrin ␤ CT. The similar kindlin-2 dissociation constants for the C-terminal peptides and their full-length versions (1.77 ϫ 10 Ϫ7 M for full-length ␤ 1 and 1.96 ϫ 10 Ϫ7 M for the ␤ 1 C-terminal peptide; 1.36 ϫ 10 Ϫ7 M for full-length ␤ 3 CT and 1.81 ϫ 10 Ϫ7 M for the ␤ 3 C-terminal peptide) verified that the C termini of the integrin ␤ 1 and ␤ 3 CT served as the main binding region for kindlin-2.
To further define the kindlin-2-binding site, a series of peptides corresponding to the C termini of the ␤ 1 CT with sequential deletion of residues from both the C-and N-terminal sides were synthesized. The ␤ 1 peptides synthesized are shown in Fig.  2C, and they were tested as inhibitors of the ␤ 3 -GST/kindlin-2 co-precipitation. As shown in Fig. 2C, the ␤ 1 CT peptides containing the highlighted residues (AVTTVVNPKYEG, designated as the ␤ 1 core) were effective blockers of kindlin-2 binding to the ␤ 3 CT, whereas the ␤ 1 CT peptides lacking these highlighted residues were ineffective inhibitors. The counterpart peptide of ␤ 3 CT (ATSTFTNITYRG, designated as ␤ 3 core) also effectively blocked kindlin-2 binding to the ␤ 1 or ␤ 3 CT, as did the ␤ 1 core (Fig. 2, D and E). Thus, the 12 amino acid residues in the ␤ CTs identified in Fig. 2A constitute the basic binding core for kindlin-2 recognition with high affinity.
Intriguingly, the kindlin-2 binding core in the integrin ␤ CTs contains both conserved and variable residues (Figs. 1A and 2A). To further test the contribution of these non-conserved residues in the integrin ␤ CTs for kindlin-2 interaction, we generated chimeric ␤ peptides in which the last three variable residues in the ␤ 2 and ␤ 3 CTs were replaced with each other (Fig.  2F). By SPR analysis, we found substitution of the last three residues RGT in the ␤ 3 peptide with the ␤ 2 counterpart residues AES nearly abolished the interaction of ␤ 3 CT peptide with kindlin-2. Surprisingly, replacement of AES in the ␤ 2 CT peptide with RGT significantly enhanced kindlin-2 binding to the ␤ 2 CT peptide (Fig. 2G). These results demonstrate that the variable residues in integrin ␤ CTs also contribute to kindlin-2 binding, and they play key roles in determining binding selectivity.
Kindlin-2-binding Core in Integrin ␤3 Cytoplasmic Tail Inhibits Cell Spreading-The functional significance of kindlin-2 binding core, ␤ 3 (Ala 750 -Gly 761 ), the ␤ 3 CT(CORE), was assessed in a cellular system. This segment, as well as control segments of the ␤ 3 CT, was expressed as PSGL-1 chimera in CHO cell line stably expressing ␣ IIb ␤ 3 (␣ IIb ␤ 3 -CHO). The ␤ 3 CT segments were expressed as fusion proteins linked to extracellular and transmembrane portions of PSGL-1 but with the ␤ 3 CT replacing the PSGL-1 CT. As described previously (23), this strategy allows for intracellular expression of a selected ␤ 3 CT segment, where it competes with the ␤ 3 CT of the intact integrin and inhibits cell spreading on fibrinogen, a ligand for ␣ IIb ␤ 3 , Monitoring of expression levels can be assessed in parallel with a PSGL-1 antibody. As assessed by flow cytometry, PSGL-1 expression differed by Ͻ10%, and the ␣ IIb ␤ 3 expression levels, monitored with a different antibody, were not affected by the transfections. Cell spreading was determined microscopically by measuring the surface area of 200 PSGL-1-expressing cells. Compared with cells expressing PSGL-1 alone, expression of the full-length ␤ 3 CT fully inhibited cell spreading (Fig. 3). The ␤ 3 CT(CORE) expressing only the ␤ 3 CT kindlin-2 binding core, ␤ 3 (749 -762), was sufficient to significantly suppress cell spreading. The ␤ 3 CT(⌬748), which only expresses the talin but not the kindlin-2 binding site also significantly inhibited cell spreading (Fig. 3). These data are consistent with the contribution of both the talin-H and kindlin-2-binding domains in integrin-mediated functions and indicate that the kindlin-2binding core itself plays an important regulatory role.
Kindlin-2 Is Unable to Directly Unclasp Integrin ␣ IIb ␤ 3 CT Complex-Unclasping the CT complex is the key step to induce the conformational change in the extracellular domain to trigger high affinity ligand recognition during the integrin activation (5,16,42). As a co-activator, we wanted to know whether kindlin-2 could directly unclasp the integrin CT complex. To test this possibility, NMR transferred NOE experiments were performed as we described previously to detect the integrin ␣/␤ CT clasp (16,22). The transferred NOEs between ␣ IIb and MBP-␤ 3 were not reduced in the presence of kindlin-2 (Fig. 4A). In fact, the transferred NOEs are more pronounced in the presence of kindlin-2 indicating formation of a ternary complex involving ␣ IIb , MBP-␤ 3 , and kindlin-2. This pattern contrasts with the dramatic reduction of the transferred NOEs observed for the ␣ IIb /MBP-␤ 3 in the presence of talin-H (Fig. 4B). These data indicate that kindlin-2 is unable to dissociate the integrincytoplasmic tails complex, which is consistent with the results of previous functional analyses showing that kindlin-2 enhanced talin-induced integrin activation but alone has only a minimal effect on integrin activation (23,24).
Kindlin-2 and Talin Head Domain Can Bind Simultaneously to Integrin ␤ 3 Cytoplasmic Tail-Because talin-H and kindlin-2 are capable of co-activating integrins by interaction with integrin ␤ CT, it is conceivable that these two proteins might bind to one another or sterically influence the binding of each other to the ␤ CT. To test these possibilities, we performed a series of SPR experiments. First, we considered whether talin-H and kindlin-2 could interact with each other. As shown in Fig. 5A, we failed to detect any binding signal when kindlin-2 protein at concentrations ranging from 87 nM to 700 nM was passed over immobilized talin-H. We verified that talin-H which had been immobilized on the biosensor chip was functional by its ability to bind the ␤ 3 CT (data not shown). In addition, we failed to detect the binding of kindlin-2 to the talin-H subdomains including F 0 F 1 and F 2 F 3 (data not shown). As an independent approach, isothermal titration calorimetry experiments were performed and also showed no detectable interaction between kindlin-2 and talin-H with both constituents in solution (Fig.  5B). Thus, within the limits of sensitivity of SPR and isothermal FIGURE 2. Kindlin-2 interacts with the C termini of integrin ␤ CTs. A, amino acid sequences of the ␤ 1 , ␤ 2 , and ␤ 3 C-terminal peptides. B, representative binding of one concentration (1 uM) of kindlin-2 to ␤ C-terminal peptides. N-terminally biotinylated peptides were captured on a SA biosensor chip, and kindlin-2 binding signals were recorded after injection (red for ␤ 2 , green for ␤ 3 , and blue for ␤ 1 peptides) as described in Fig. 1. C, control GST or GST-fused ␤ 3 CT was incubated with HUVEC lysate in the presence of the indicated peptide from the ␤ 1 CT (final concentration at 200 M). The endogenous kindlin-2 binding to the ␤ 3 CT was precipitated with glutathione-Sepharose and further analyzed by Western blot. GST-fused ␤1 CT (D) or GST-fused ␤3 CT (E) were used to precipitate kindlin-2 protein from HUVEC lysates in the presence of ␤1 core peptide or ␤3 core peptide as highlighted in A. F, the chimeric ␤ 2 and ␤ 3 CT with the last three-amino acid replacement. G, comparison of the binding of one concentration of kindlin-2 to N-terminally biotinylated ␤ peptides (light green for ␤ 3 -AES, red for ␤ 2 -RGT, and green for the ␤ 3 peptide as shown in A).
titration calorimetry as performed, we could not detect interaction between talin-H and kindlin-2, although interaction of these proteins with ␤ 3 CT indicates the functionality of the protein preparations.
The undetectable interaction between talin-H and kindlin-2 allowed us to test whether they can bind simultaneously to the integrin ␤ CT by SPR experiments. Full-length ␤ 3 CT was immobilized on the biosensor chip, and talin-H or kindlin-2 was passed over the chip. As shown in Fig. 6A, a typical progress curve was obtained for either kindlin-2 or talin-H. When a mixture of kindlin-2 and talin-H, each at the same concentration, was passed over the immobilized ␤ 3 CT, the response curve was enhanced as compared with that obtained with each individual analyte, indicating that their interaction was not mutually exclusive (Fig. 6A). We further assessed whether occupancy of the immobilized ␤ 3 CT with talin-H would limit kindlin-2 bind-   JULY 13, 2012 • VOLUME 287 • NUMBER 29

JOURNAL OF BIOLOGICAL CHEMISTRY 24589
ing. As shown in Fig. 6B, after the ␤ 3 CT-coated chip was saturated with talin-H, the binding signal induced by kindlin-2 reached the level observed in the absence of talin-H. Furthermore, we measured the binding pattern of talin-H to immobilized ␤ 3 CT by sequential injections (500-s intervals) in the presence or absence of kindlin-2. The binding signals obtained with multiple injections of talin-H onto immobilized ␤ 3 CT was unaffected by kindlin-2 in the running buffer (Fig. 6, C and D). The same results were obtained when we used ␤ 1 CT instead of ␤ 3 CT in these experiments. Together, these data support the conclusion that talin-H and kindlin-2 can interact simultaneously with integrin ␤ CTs and exert minimal effects on the binding of one another to the ␤ CT.
To provide independent evidence for the ability of talin-H, kindlin-2, and ␤ 3 CT to form a ternary complex, we again used a NMR approach. As shown in Fig. 7A, when talin-F 2 F 3 interacts with 15 N-labeled ␤ 3 CT, significant line-broadening or chemical shift changes occur, indicative of complex formation. Similarly, line-broadening and chemical shift changes were also noted when kindlin-2 was added to the 15 N-labeled ␤ 3 CT (Fig.  7B). If kindlin-2 and talin-F 2 F 3 were to compete with each other, the spectrum of 15 N-labeled ␤ 3 CT mixed with both talin-F 2 F 3 and kindlin-2 would overlay with or mimic one of the single binding partners. As an example, the chemical shifts induced by filamin binding to the ␤ 3 CT are neutralized by addition of migfilin (8). However, when talin-F 2 F 3 and kindlin-2 were added to the ␤ 3 CT, further line-broadening and spectral changes were observed (Fig. 7C), indicative of distinct perturbations reflecting formation of a ternary complex. Thus, both SPR and NMR experiments support the formation of a ternary complex involving talin, kindin-2, and the ␤ 3 CT.

DISCUSSION
Kindlin family members were recently identified as integrin co-activators that play indispensable roles in integrin activation (43). Our current study reports several novel findings to further interpret the mechanism by which kindlin-2 facilitates integrin activation. First, kindlin-2 shows selectivity in binding to integrin CT; it binds with substantially lower affinity to ␤ 2 than to ␤ 1 and ␤ 3 CT. ␤ 2 -integrins also require kindlins for activation, and this need may be met by kindlin-3 in myeloid cells (27)(28)(29). Second, the kindlin-2 binding core of 12 amino acids has been identified in integrin ␤ CTs, which includes the conserved membrane-distal NxxY region and S/T cluster as well as the variable residues flanking these two motifs. The functional importance of this kindlin-2 binding core in intact cells was supported by cell spreading assays. Expression of the C-terminal kindlin-2 binding region alone was sufficient to suppress integrin mediated cell spreading. The identification of the kindlin-2 binding core provides a context in which to interpret the effects of previously reported mutations on integrin activation. Thus, the effects of ␤ 3 ⌬755, as well single point mutations at S752P or Y759A in the ␤ 3 CT on integrin activation (44 -49) can be attributed to inhibition of kindlin-2 binding. Also, the inhibition of integrin ␣ IIb ␤ 3 activation in platelets by importable peptides corresponding to ␤ 3 (743-750) and ␤ 3 (749 -756), which overlap the kindlin-2 binding core (50,51), can now be attributed to effects on kindlin binding. Expression of the ␤ 3 CT⌬748 and ␤ 3 CT (CORE) as a PSGL chimeric protein also suppressed integrin activation, suggesting that the membrane proximal and membrane distal regions of the ␤ 3 CT contribute independently to integrinmediated responses, observations consistent with the cooperativity between talin and kindlin-2 in integrin function. We cannot exclude that the effects of these two regions may also reflect their interaction with other integrin binding partners. For example, Integrin Linked Kinase (52) and Src The lower graph shows the titration plot derived from the normalized integrated heats in kcal/mol of injectant plotted against molar ratio after background subtraction. There is no significant difference in the heat changes between different injections of talin-H indicative of no direct interaction of kindlin-2 to talin-H. (53) have been shown to bind to the C-terminal region of the ␤ 3 CT as well. However, in the ␣ IIb ␤ 3 -CHO cell system, these specific binding partners do not influence ␣ IIb ␤ 3 activation or co-activation with talin-H (54). 4 Instead of NxxY in ␤ 1 and ␤ 3 tails, the counterpart motif in the ␤ 2 CT is NxxF. However, substitution of Phe with Tyr in ␤ 2 CT did not improve the binding affinity (data not shown), suggesting that the non-conserved residues in the integrin ␤ core region might be also critical for kindlin-2 recognition. This proposition was supported experimentally. A ␤ 2 CT peptide with its C-terminal AES replaced by the ␤ 3 RGT showed higher kindlin-2 binding, and conversely, a ␤ 3 CT peptide in which it C-terminal RGT was replaced with the ␤ 2 AES sequence had blunted kindlin-2 binding (Fig. 2). Even though this binding core is sufficient for kindlin-2 binding to ␤ 1 and ␤ 3 CT, we do not exclude the possibility that more membrane proximal residues in the CT might help to stabilize the interaction and support integrin activation. More detailed structural studies are still required to further understand the binding specificity of the kindlin/␤CT interaction. Third, kindlin-2 is unable to directly unclasp the integrin CT complex, which provides a mechanistic explanation for the very limited ability of kindlin-2 alone on integrin activation; for efficient activation, talin-H also has to be present (23,24). One possibility for the talin/kindlin-2 co-activating requirement is that talin-H might directly interact with kindlin-2. However, we failed to detect their direct interaction experimentally. Indeed, talin-H and kindlin-2 can bind simultaneously to integrin ␤ CT with no evidence of mutual perturbation. The capacity of kindlin-2 and talin-H to co-assemble on the ␤ 3 CT to form a ternary complex was indicated by our SPR data, where the resonance unit signal, a reflection of mass of the bound constituents (55,56) increased upon binding of both talin-H and kindlin-2 to the immobilized ␤ 3 CT. Independent evidence of formation of a ternary complex also was provided by NMR experiments showing that the chemical shifts induced by talin-H or kindlin-2 binding to the ␤ 3 CT were further shifted by the presence of the two constituents; such changes are also indicative of simultaneous interaction of the three constituents, i.e. formation of a ternary complex. Of note, the ability to detect chemical shift changes in the ␤ 3 CT upon addition of kindlin-2 suggests that it may be feasible to locate the interactive residues in ␤ 3 CT engaged in kindlin-2 binding. In view of the relatively low affinity of kindlin-2 for the ␤ 3 CT, which we estimated from our SPR experiments to be 1.36 ϫ 10 Ϫ7 M, such NMR mapping may be more FIGURE 6. Kindlin-2 and talin head domain can associate simultaneously with the integrin ␤ 3 CT. A, kindlin-2, talin-H, or their mixture, at the same individual concentrations (700 nM), was injected over the sensor chips coated with full-length ␤ 3 CT, and their association properties were recorded. B, the ␤ 3 CT coated sensor chip was first saturated with a high concentration (2.8 M) of talin-H. Once the binding reached the maximal level, kindlin-2 (700 nM) was injected. Increasing concentrations of talin-H were injected over sensor chips coated with full-length ␤ 3 CT with (C) or without (D) kindlin-2 in running buffer. JULY 13, 2012 • VOLUME 287 • NUMBER 29 feasible for detecting and dissecting the ternary complex than other approaches that are difficult to apply for lower affinity interactions. However, because of the relatively large molecular weight involving kindlin-2 (ϳ70 kDa) and talin-H (ϳ50 kDa), future investigation will employ deuteration and transverse relaxation optimized spectroscopy-type NMR experiments to reduce the line-broadening effect (57). Although we did not detect any synergistic binding effect of kindlin-2 and talin-H on integrin ␤ CT, their simultaneous binding might be the first key step to attain synergy in integrin activation through alternative pathways: i) kindlin-2 and talin-H could jointly displace negative regulator(s) such as filamin from the integrin CTs. The site that we have mapped for kindlin-2 binding on integrin CT clearly overlaps with that identified for filamin (7); ii) kindlin-2 and talin might be linked by an as yet identified bridging molecule(s), thereby enhancing their interaction with integrin; iii) more directly, kindlin-2 could attach onto the plasma membrane and place ␤ CT in a specific orientation advantageous for talin-H recognition. Although these possible regulatory mechanisms remain to be vigorously dissected, our studies herein have provided significant insights into our understanding of how kindlin-2 and talin may be simultaneously situated on integrin ␤ CT and synergistically promote integrin activation. The capacity of talin and kindlins to associate simultaneously provides an example as to how large molecules such as talin and kindlins can associate in close proximity on a small segment of receptor, the ␤ CTs, to orchestrate a physiologically relevant cellular response, integrin activation.