The dual structural roles of the membrane distal region of the α-integrin cytoplasmic tail during integrin inside-out activation

ABSTRACT

Studies on the mechanism of integrin inside-out activation have been focused on the role of β-integrin cytoplasmic tails, which are relatively conserved and bear binding sites for the intracellular activators including talin and kindlin. Cytoplasmic tails for α-integrins share a conserved GFFKR motif at the membrane-proximal region and this forms a specific interface with the β-integrin membrane-proximal region to keep the integrin inactive. The α-integrin membrane-distal regions, after the GFFKR motif, are diverse both in length and sequence and their roles in integrin activation have not been well-defined. In this study, we report that the α-integrin cytoplasmic membrane-distal region contributes to maintaining integrin in the resting state and to integrin inside-out activation. Complete deletion of the α-integrin membrane-distal region diminished talin- and kindlin-mediated integrin ligand binding and conformational change. A proper length and correct amino acids in α-integrin membrane-distal region was found to be important for integrin inside-out activation. Our data establish an essential role for the α-integrin cytoplasmic membrane-distal region in integrin activation and provide new insights into how talin and kindlin induce the high-affinity integrin conformation that is required for fully functional integrins.

INTRODUCTION

Integrins are cell surface molecules composed of α and β subunits, each containing a large extracellular domain, a short transmembrane (TM) domain, and usually a short cytoplasmic tail (CT). The combination of 18 α- integrin and eight β-integrin subunits results in 24 human integrins, and these play important roles in diverse biological processes including cell migration, morphogenesis, hemostasis and the immune response (Hynes, 2002). Integrins transmit signals between extracellular ligands and the intracellular cytoskeleton through bidirectional signaling across cell membrane (Kim et al., 2011b; Springer and Dustin, 2012). Intracellular stimulations impinging on the integrin CT induce conformational changes that transform resting integrin into a high-affinity ligand-binding-competent state (known as integrin activation or inside-out signaling). Ligand binding to the extracellular domain induces integrin conformational rearrangement and transmits signals to the cytoplasmic side (known as integrin outside-in signaling). The integrin TM and CT domains play pivotal roles in the bidirectional signaling process.

The large-scale integrin conformational rearrangements, including the changing from a bent to an extended, and from a closed to an open headpiece conformation, have been directly visualized by electron microscopy (EM) (Takagi et al., 2002; Nishida et al., 2006; Ye et al., 2010; Eng et al., 2011; Zhu et al., 2012), crystallography (Xiao et al., 2004; Zhu et al., 2013), and small angle X-ray scattering (Mould et al., 2003; Eng et al., 2011), and this process in now known to involve the exposure of epitopes (known as ligand-induced-binding sites, LIBS) that are masked in the inactive state (Frelinger et al., 1991; Sims et al., 1991; Humphries, 2004; Byron et al., 2009). It has been demonstrated that both integrin extension and headpiece opening are required for high-affinity ligand binding (Chen et al., 2010; Springer and Dustin, 2012), which is crucial for platelet aggregation mediated by integrin α_IIbβ₃ at the bleeding sites and neutrophil arrest mediated by integrin α_Lβ₂ at the inflamed tissues (Lefort et al., 2012).

Integrin inside-out activation is regulated through the CT. Sequence alignment shows that β-integrin CTs are relatively conserved at both the membrane-proximal (MP) and membrane-distal (MD) regions, which have common binding sites for regulatory proteins like talin and kindlin (supplementary material Fig. S1) (Anthis and Campbell, 2011; Kim et al., 2011b; Margadant et al., 2011; Morse et al., 2014). In contrast, α-integrin CTs are only highly conserved at the MP regions, which have a Gly-Phe-Phe-Lys-Arg (GFFKR) motif, whereas the MD regions vary substantially both in sequence and length (supplementary material Fig. S1). It is known that the α- integrin CT MP region is required for keeping the integrin inactive (O'Toole et al., 1991; O'Toole et al., 1994; Hughes et al., 1995; Hughes et al., 1996; Lu and Springer, 1997; Zhu et al., 2009), but the role of α-integrin CT MD region in integrin activation is not well characterized.

Mutational studies have suggested that the specific associations of α-integrin and β-integrin TM and CT domains maintain integrin in the resting state (O'Toole et al., 1994; Hughes et al., 1995; Luo et al., 2005; Partridge et al., 2005; Berger et al., 2010). This was confirmed by an analysis of the structure of the α_IIbβ₃ TM and CT domains after disulfide crosslinking using intact integrin on the cell surface (Zhu et al., 2009) (Fig. 1A) and by NMR using the isolated peptides either in lipid bicelles (Lau et al., 2009) (Fig. 1B) or in organic solvent (Yang et al., 2009) (Fig. 1C). The three structures share a similar interface at the TM domain, but differ at the CT (Fig. 1A–C). Although the disulfide-based and the NMR structure in lipid bicelles have a similar conformation at the CT MP regions (Fig. 1A,B), there are different interpretations of how the conserved residues, such as β₃-K716, contribute to integrin activation (Zhu et al., 2009; Kim et al., 2011a; Kurtz et al., 2012). The NMR structure in organic solvent has a very different conformation at the MP region as compared with the other structures (Fig. 1C). It is not known whether such structural diversity is due to the different environments used for structure determination or whether it represents a dynamic conformational change. The structure of α_IIb CT MD region is not well-defined and varies among all the structures (Fig. 1A–C). It is not known whether this region contributes to maintaining the resting integrin conformation.

Fig. 1.

Effect of cytoplasmic truncation on α_IIbβ₃ integrin activation. (A–C) The transmembrane (TM) and cytoplasmic structures of α_IIbβ₃ determined by (A) disulfide crosslinking and Membrane Rosetta (xxx reference xxx), (B) by NMR in lipid bicelles (PDB code: 2K9J; xxx reference xxx), (C) by NMR in organic solvent (PDB code: 2KNC; xxx reference xxx). TM regions are red. The cytoplasmic membrane-proximal (MP) and membrane-distal (MD) regions are cyan and gray, respectively. Interfacial residues are shown as sticks with red oxygens and blue nitrogens or as Cα spheres (for glycine), and marked with stars in the α_IIbβ₃ sequences shown. The three structures were superimposed based on the TM region and shown in horizontal separation on the page. (D) PAC-1 binding of the α_IIb CT truncations (Tr, shown schematically above the bars); (E) PAC-1 binding of the β₃ CT truncations. 293FT cells were co-transfected with α_IIb truncations and wild-type β₃ (WT), or with β₃ truncations and α_IIb WT. PAC-1 binding was measured by flow cytometry and presented as the mean fluorescence intensity (MFI) normalized to integrin expression. Data are presented as mean±s.e.m. (n=3).

In the inside-out integrin activation, the cytosolic protein talin (which has two isoforms, talin-1 and -2) binds to both the MD and MP regions of the β-integrin CT through its head domain, which is composed of F0, F1, F2 and F3 subdomains, resulting in conformational changes in the integrin that are propagated to the extracellular domains across cell membrane (Kim et al., 2011b). Kindlin (which has three isoforms, kindlin-1, -2 and -3) has been found to cooperate with talin during integrin activation through direct binding to the β-integrin CT MD region (Ma et al., 2008; Moser et al., 2008; Harburger et al., 2009; Malinin et al., 2009; Moser et al., 2009b; Svensson et al., 2009; Bledzka et al., 2012). Studies on how talin changes integrin conformation have been focused on the β-integrin TM and CT domains, and have been described in many elegant reviews (Moser et al., 2009a; Shattil et al., 2010; Anthis and Campbell, 2011; Kim et al., 2011b; Calderwood et al., 2013; Das et al., 2014). Recent structural and functional studies suggest that talin-1 binding to the β₃ CT changes the tilt angel of β₃ TM domain in the cell membrane, thus disturbing the α-integrin–β-integrin interfaces at the TM and MP regions (Kalli et al., 2011; Kim et al., 2011a; Kim et al., 2012). A crystal structure of the talin-2 F2-F3 domain bound to β_1D-CT suggests that there is an ionic interaction between a lysine residue in the talin-2 F3 domain and the conserved aspartic acid residue of the β_1D integrin CT MP region (Anthis et al., 2009). This interaction was proposed to be important in integrin activation by disrupting the putative salt-bridge (between α_IIb-R995 and β₃-D723) at the α-integrin–β-integrin MP interface that has been suggested to maintain the resting integrin state (Hughes et al., 1996). However, it is not known if or how the α-integrin CT, and especially the MD region, participates in talin- and kindlin-induced integrin activation.

In this study, we examined the role of α_IIb, α_V and α_L integrin CT MD regions in the regulation of integrin activation. We report that the α-integrin CT MD region helps maintain integrin in the resting state and is indispensible in talin- and kindlin-induced integrin conformational change and ligand binding. The proper length and correct amino acids are important for the α-integrin CT MD region to exert its effect on integrin activation. During the preparation of our manuscript, Li et al. (Li et al., 2014) reported that the α_IIb CT MD is required for α_IIbβ₃ inside-out activation and proposed a model of ‘steric clashes’ between the talin-1 head and α_IIb CT MD region that is involved in integrin activation. Our data of α_IIbβ₃ are consistent with their results, but our comprehensive analysis on multiple integrins suggests a more complicated mechanism by which α-integrin CT MD region regulates integrin activation.

RESULTS

The MP CT region is the minimum structural requirement for maintaining α_IIbβ₃ integrin in the resting state

To determine the minimum structure requirement of α_IIbβ₃ CT for maintaining the resting state, we performed a serial truncation mutagenesis. α_IIb CT truncations before F993 greatly reduced integrin cell surface expression, whereas other truncations had a minor effect (supplementary material Fig. S2A). Similarly, β₃ CT truncations before but not after I719 greatly reduced integrin expression (supplementary material Fig. S2B). The ligand-mimetic monoclonal antibody (mAb) PAC-1 was used to assess α_IIbβ₃ activation. PAC-1 binding to all the α_IIb CT truncations after R995 was comparable to that for wild-type (WT) integrin (Fig. 1D). Truncation after α_IIb-K994 only slightly increased PAC-1 binding, whereas truncations after α_IIb V990, G991, F992 and F993 greatly enhanced PAC-1 binding (Fig. 1D, truncations are denoted by the suffix Tr). For β₃ integrin, there was no detectable activation for all the β₃ CT truncations after I721 (Fig. 1E). β₃-W715Tr, β₃-K716Tr and β₃-L717Tr strongly induced PAC-1 binding, and truncation after β₃-L719 weakly induced PAC-1 binding (Fig. 1E). The combination of the α_IIb-K994Tr and β₃-I721Tr had comparable PAC-1 binding to WT α_IIbβ₃ (data not shown). These results demonstrate that the MP sequence, KVGFFK or KVGFFKR in the α_IIb CT, and the MP sequence of KLLITI or KLLIT in the β₃ CT are the minimum requirement for maintaining α_IIbβ₃ integrin in the resting state.

MD α_IIb CT truncations amplify the activating effect of β₃ mutations

Truncations after the α_IIb CT MP region might decrease the threshold for integrin activation because of structure incompleteness. To test this possibility, we performed a PAC-1 binding assay for the α_IIb CT truncations co-expressed with the constitutively active β₃ mutants, β₃-G708L or β₃-K716A. These β₃ mutations are known to activate α_IIbβ₃ by disturbing the TM interaction (Luo et al., 2005; Zhu et al., 2009; Kim et al., 2011a). Compared with WT α_IIb, truncation after α_IIb-F993, K994, R995, N996, R997 or P998 strongly enhanced PAC-1 binding to both β₃-G708L and β₃-K716A mutants (Fig. 2A,B). This is in contrast with the very subtle or undetectable activating effect of these α_IIb truncations when paired with WT β₃ (Fig. 1D). The truncation-amplified activating effect was decreased with the increase of the length of α_IIb CT MD region (Fig. 2A,B). More strikingly, the β₃-G708A mutation very weakly induced PAC-1 binding when paired with WT α_IIb, but its activating effect was greatly enhanced upon deletion of the α_IIb CT MD region (Fig. 2C). In addition, the enhanced activating effect of α_IIb CT MD deletion on the β₃-G708L mutation was close to the maximal level, given that Mn²⁺ did not further increase PAC-1 binding (Fig. 2C). Similar results were obtained with the active β₃-G135A mutant, which facilitates formation of the active conformation of β₃-I domain (Zhang et al., 2013). Complete deletion of the α_IIb CT MD region enhanced the activation of β₃-G135A (data not shown). These data demonstrate that deletion of the α_IIb CT MD region renders the integrin hyper-reactive to β₃-activating mutations. Therefore, the α_IIb CT MD region contributes to maintaining integrin in the resting state.

Fig. 2.

α_IIb CT truncations synergize or compensate the activating effect of the transmembrane and cytoplasmic mutations. PAC-1 binding of 293FT cells co-transfected with the indicated α_IIb constructs (shown schematically above the bars in A and E) and (A) β₃-G708L, (B) β₃-K716A, (C) β₃-G708A and β₃-G708L, (D) β₃-D723R or (E) β₃ WT. 293FT transfectants of α_IIbβ₃ WT and α_IIb-F993Tr plus β₃ WT are shown as controls in A, B and D. PAC-1 binding was measured by flow cytometry in the buffer containing 1♣mM Ca²⁺ and 1♣mM Mg²⁺ (Ca/Mg) (for A–E) or 0.2♣mM Ca²⁺ plus 2♣mM Mn²⁺ (Ca/Mn) (for C) and presented as the MFI normalized to integrin expression. Data are presented as mean±s.e.m. (n=3).

MD α_IIb CT truncation abolishes the activating effect of α_IIb-R995 and β₃-D723 mutations

Integrin activation induced by the α_IIb-R995 and β₃-D723 mutations has been shown to depend on talin-1 binding to the β₃ CT, which is distinct from the activation mechanism promoted by the activating mutations at the TM domain (Wegener et al., 2007). We asked whether the α_IIb CT MD truncations exert different effects on α_IIb-R995 and β₃-D723 mutations. The β₃-D723R mutant constitutively bound PAC-1 when paired with α_IIb WT (Fig. 2D). However, PAC-1 binding was reverted to a WT level when β₃-D723R was paired with the α_IIb truncations F993Tr, K994Tr, R995Tr or N996Tr, in which all or most of the MD region is deleted (Fig. 2D). Strikingly, although both α_IIb-F993Tr and β₃-D723R were constitutively active when paired with their WT partners, their combination abolished the activating effect (Fig. 2D). This is in sharp contrast with the amplified activation mediated by the α_IIb truncations for the β₃-G708L and β₃-K716A mutants (Fig. 2A,B). Retaining two or more residues of the α_IIb MD region restored or increased integrin activation induced by the β₃-D723R mutation (Fig. 2D).

We further verified our results with the α_IIb-R995A and α_IIb-R995D mutation. Both of these mutations render α_IIbβ₃ active to bind PAC-1 in the context of full-length α_IIb (Fig. 2E). Their activating effects were abolished when the α_IIb CT MD region was completely deleted (Fig. 2E). By contrast, the activation by α_IIb-F992A or α_IIb-F993A mutation was enhanced upon the α_IIb CT MD deletion (Fig. 2E). These data suggest that the α_IIb CT MD region plays a positive role in integrin inside-out activation.

The α-integrin CT MD region is required for talin-1-induced integrin activation

The deactivating effect of α_IIb CT MD truncation on α_IIb-R995 and β₃-D723 mutations raises the intriguing possibility that the α_IIb CT MD region might participate in talin-mediated integrin activation. We performed a well-established integrin activation assay by overexpressing of an EGFP-tagged talin-1 head domain (EGFP–TH) (Bouaouina et al., 2012). As expected, overexpression of EGFP–TH significantly activated WT α_IIbβ₃ as determined by the amount of PAC-1 binding (Fig. 3A). The EGFP–TH-induced activation was much more substantial for integrin bearing α_IIb-R995A or β₃-D723A mutation, indicating a synergetic effect (Fig. 3A). However, EGFP-TH failed to induce PAC-1 binding to the α_IIb-R995Tr mutant, i.e. the α_IIb CT MD region was deleted (Fig. 3A). Remarkably, the strong activating effect of EGFP–TH on α_IIb-R995A or β₃-D723A was also completely lost in the absence of the α_IIb CT MD region (Fig. 3A). The expression levels of EGFP–TH or integrins were comparable among all the transfectants (supplementary material Fig. S2C). These results clearly demonstrate the requirement of α_IIb CT MD region in talin-1-head-induced α_IIbβ₃ activation.

Fig. 3.

The α-integrin CT MD region is required for talin- and kindlin-induced integrin activation. (A–C) Complete deletion of the α-integrin CT MD region abolished talin-1-head-induced activation of (A) α_IIbβ₃, (B) α_Vβ₃, and (C) α_Lβ₂ integrins. Ligand binding was measured by flow cytometry with 293FT cells co-transfected with the indicated integrin constructs (also shown schematically above the bars) and EGFP or EGFP–TH. The EGFP and integrin double-positive cells were analyzed. (D–F) The α-integrin CT MD region is required for kindlin-induced activation of (D) α_IIbβ₃, (E) α_Vβ₃, and (F) α_Lβ₂ integrins. Ligand binding was measured by flow cytometry with 293FT cells co-transfected with the indicated integrin constructs and EGFP–TH plus mCherry (mC), mC-tagged kindlin-2 (mC–K2), or mC-tagged kindlin-3 (mC–K3). The integrin, EGFP, and mCherry triple-positive cells were analyzed. Data are presented as the MFI of the ligand normalized to integrin expression. Data are presented as mean±s.e.m. (n≥3). *P<0.05; **P<0.01; ns, not significant (unpaired two-tailed Student's t-tests).

We next examined whether the CT MD region is also important for talin-induced activation of other α-integrins. We found that overexpression of EGFP–TH significantly enhanced the binding of fibronectin type III domain 9–10 (Fn9-10) and intercellular adhesion molecule 1 (ICAM-1) to α_Vβ₃ and α_Lβ₂, respectively (Fig. 3B,C). In contrast, when the CT MD region was completely deleted for α_V or α_L integrin (α_V-R993Tr or α_L-R1094Tr), the talin-1 head failed to augment ligand binding for these integrins (Fig. 3B,C). The expression levels of EGFP–TH or integrin are comparable among the transfectants (supplementary material Fig. S2D,E). Thus, the requirement for an α-integrin CT MD region for talin-mediated integrin activation could be a generalized property of integrins.

The α-integrin CT MD region is required for kindlin-induced integrin activation

Having demonstrated the indispensable role of the α-integrin CT MD region in talin-induced integrin activation, we next asked whether the α-integrin CT MD region is also required for kindlin-mediated integrin activation. Given that integrin activation by kindlin requires the presence of talin (Ma et al., 2008), we examined the effect of deletion of the α-integrin CT MD domain on kindlin-mediated integrin activation with co-expression of the talin-1 head. Overexpression of mCherry–kindin-2 (mC–K2) significantly enhanced PAC-1 binding to WT α_IIbβ₃ (Fig. 3D). We also detected a slight increase of PAC-1 binding induced by mCherry–kindlin-3 (mC–K3), but it was not statistically significant compared with the mCherry control (Fig. 3D). In the absence of α_IIb CT MD region (α_IIb-R995Tr), mC-K2-mediated PAC-1 binding was greatly and significantly (P<0.01) reduced compared with the WT (Fig. 3D). As a control, the talin-1 head and kindlins failed to induce PAC-1 binding to α_IIb–β₃-R724Tr owing to the deletion of binding sites for talin-1 and kindlin (Fig. 3D). The expression levels of EGFP–TH, mC–K2 or mC–K3 were comparable among all the transfectants (supplementary material Fig. S3A,B). These data demonstrate that the α_IIb CT MD region is required for kindlin-mediated α_IIbβ₃ activation.

Similar results were obtained with α_Vβ₃ and α_Lβ₂ integrins. In the presence of the talin-1 head, kindlin-2 but not kindlin-3 markedly enhanced fibronectin (Fn) binding to WT α_Vβ₃ (Fig. 3E). Both kindlin-2 (P<0.01) and kindlin-3 (P<0.05) significantly increased ICAM-1 binding to WT α_Lβ₂, but the level of increase by kindlin-3 was significantly (P<0.05) lower than by kindlin-2 (Fig. 3F). Deletion of the CT MD region of α_V (α_V-R993Tr) or α_L (α_L-R1094Tr) significantly (P<0.01) and greatly reduced the kindlin-2-mediated ligand binding (Fig. 3E,F). The kindlin-3-induced ICAM-1 binding was also greatly reduced for the α_L-R1094Tr–β₂ mutant (Fig. 3F). The expression levels of EGFP–TH, and kindlin-2 or kindlin-3 were comparable among the transfectants of α_Vβ₃ (supplementary material Fig. S3C,D) and α_Lβ₂ (material Fig. S3E,F). These data strongly demonstrate that the α-integrin CT MD region is also required for kindlin-mediated integrin activation.

The kindlin-2-mediated ligand binding was not completely abolished in the absence of the α-integrin CT MD region. The low level increase of ligand binding upon addition of kindlin-2 was statistically significant for α_IIb-R995Tr–β₃ (P<0.05) and α_L-R1094Tr–β₂ (P<0.01) (Fig. 3D,F), but not for α_V-R993Tr–β₃ (Fig. 3E). In addition, we could detect a moderate increase (P=0.05) of kindlin-3-mediated ICAM-1 binding to the α_L-R1094Tr–β₂ mutant (Fig. 3F). Given that PAC-1, Fn, and ICAM-1 are all multivalent ligands, the low level of ligand binding was probably due to the integrin clustering effect of kindlin as suggested by a recent study (Ye et al., 2013).

The α-integrin CT MD region is required for PMA-stimulated integrin activation in K562 cells

We further validated our data with by expressing the α_IIbβ₃ integrin in human K562 cells. PMA was used to stimulate integrin activation, which mimics the physiological integrin inside-out activation cascade (Banno and Ginsberg, 2008). We did not detect PMA-induced PAC-1 binding to WT α_IIbβ₃ (data not shown). However, PMA induced PAC-1 binding to α_IIb–β₃-D723A in a concentration-dependent manner (Fig. 4A), indicating that the β₃-D723A mutation enhanced integrin sensitivity to PMA stimulation. Consistent with the results obtained upon overexpression of talin-1 head and kindlins, complete deletion of α_IIb CT MD region abolished PMA-induced PAC-1 binding (Fig. 4A,B).

Fig. 4.

The α_IIb CT MD region is required for PMA-stimulated integrin activation in K562 cells. (A) Dose–response curve for PMA-induced PAC-1 binding to α_IIbβ₃ integrin expressed in K562 cells. Transfected K562 cells were incubated with PMA and then with PAC-1. Flow cytometry was used to measure PAC-1 binding. Data are presented as the MFI of the ligand normalized to integrin expression. Data are presented as mean±s.e.m (n=4). (B) Representative flow cytometry plots of PMA-induced (at 1♣∀M) PAC-1 binding to K562 cells expressing the indicated α_IIbβ₃ integrins.

Retaining two residues of α-integrin CT MD region partially restores talin-1-induced activation of α_IIbβ₃, but not α_Vβ₃ and α_Lβ₂ integrins

We next asked how the α-integrin CT MD region participates in talin-mediated integrin activation. Structural superposition of talin bound to the β-CT and α_IIbβ₃ TM-CT complexes suggests that there can be steric clashes between the talin head and α_IIb MD residues that are close to β₃ and immediately follow the GFFKR motif (i.e. residues N996–R997) (supplementary material Fig. S4A,B). To test whether the potential steric clash plays a role in talin-mediated integrin activation, we made truncated α_IIb constructs that had the two native residues, NR997Tr, two small residues, GG997Tr and AA997Tr, or two bulky residues, YY997Tr and WW997Tr, at the MD region (Fig. 5A). If the steric clashes are important, we would expect to see that the mutations with bulky residues would have a greater activating effect on talin-mediated integrin activation than the mutations with small residues. All the mutants had comparable levels of PAC-1 binding to that shown by WT α_IIb when paired with β3 WT, except that the WW997Tr mutant was constitutively active (Fig. 5A). The Talin-1 head augmented PAC-1 binding to all the truncated α_IIb mutants co-expressed with WT β₃ (Fig. 5B). The WW997Tr mutant had a much higher level of PAC-1 binding than other constructs owing to its constitutive activation (Fig. 5B). We consistently observed that a lower level of PAC-1 binding was induced by the talin-1 head for the YY997Tr mutant than for the other constructs (Fig. 5B). We then examined the same α_IIb mutants in the context of β₃-D723A mutation, which is expected to enhance the sensitivity of the assay. The Talin-1 head induced similar levels of PAC-1 binding to α_IIb-NR997Tr, α_IIb-GG997Tr, and α_IIb-AA997Tr mutants co-expressed with β₃-D723A, but the binding was significantly (P<0.001) lower than for α_IIb-WT co-expressed with β₃-D723A (Fig. 5C). Strikingly, talin-1-head-induced PAC-1 binding to α_IIb-YY997Tr mutant was significantly (P<0.01) lower than with α_IIb-NR997Tr and the other constructs (Fig. 5C). When the NR997 was replaced by YY997 in the context of full-length α_IIb, the α_IIb-YY997 mutation also significantly reduced talin-1-head-mediated activation of β₃-D723A (Fig. 5C). We next extended the C-termini of α_IIb CT with a 29-residue tag of V5 epitope plus hexahistidine, which would be expected to provide more clashes with talin-1 head. We found that the response to talin-1-head-induced activation for V5–His-tagged α_IIb was indistinguishable from WT α_IIb (data not shown). This result, in accordance with the YY997 mutation data, suggests that it is not plausible to assign a simple ‘steric clash’ model to the mechanism by which α_IIb CT MD region participates in integrin activation.

Fig. 5.

Retaining two residues at the α CT MD region partially restores talin-1 head induced activation of α_IIbβ₃, but not α_Vβ₃ and α_Lβ₂ integrins. (A) PAC-1 binding of 293FT cells transfected with β₃ WT and the indicated α_IIb constructs constructs (also shown schematically above the bars). (B) EGFP–TH-induced PAC-1 binding of the 293FT cells transfected with β₃ WT and the indicated α_IIb constructs. (C) EGFP–TH-induced PAC-1 binding of 293FT cells transfected with β₃-D723A and the indicated α_IIb constructs. (D) EGFP–TH induced Fn9-10 binding of 293FT cells transfected with α_Vβ₃ constructs. (E) EGFP–TH induced ICAM-1 binding of 293FT cells transfected with α_Lβ₂ constructs. EGFP and integrin double-positive cells were analyzed by flow cytometry. Ligand binding was presented as the MFI of the ligand normalized to integrin expression. Data are presented as mean±s.e.m (n≥3). *P<0.05; **P<0.01; ***P<0.001; ns, not significant (unpaired two-tailed Student's t-tests).

We extended our study to α_Vβ₃ and α_Lβ₂ integrins. Consistent with α_IIbβ₃, the presence of β₃-D723A or the equivalent β₂-D709A mutation significantly enhanced talin-1-head-induced binding of Fn9-10 and ICAM-1 to α_Vβ₃ and α_Lβ₂, respectively (Fig. 5D,E). Complete deletion of the α_V and α_L CT MD region (αV-R993Tr and αL-R1094Tr) totally abolished the activating effect of the talin-1 head (Fig. 5D,E). However, retaining two residues in the α_V and α_L CT MD region (αV-R995Tr and αL-R1096Tr) did not restore talin-1-head-induced integrin activation (Fig. 5D,E). This is in contrast to the result that showing that retaining two residues at the α_IIb CT MD region partially restores talin-1-induced α_IIbβ₃ activation. In addition, retaining eight residues for the α_L CT MD region (α_L-A1102Tr) was still not sufficient to restore talin-1-head-induced integrin activation to the WT α_L level (Fig. 5E). Thus, different α integrins might have different requirements for the CT MD region in talin-1-mediated activation. The proper amino acids and correct length of α-integrin CT MD region are required for sufficient integrin activation.

The α-integrin CT MD region is required for talin- and kindin-induced integrin conformational change

It is generally accepted that talin activates integrin through inducing conformational change (Anthis and Campbell, 2011; Kim et al., 2011b). By using conformation-dependent LIBS mAbs, we examined the talin-1-head-induced conformational change of integrin ectodomains. Consistent with our previous results (Zhang et al., 2013), overexpression of the talin-1 head significantly increased the binding of the β₃ LIBS mAb 319.4, but not the α_IIb LIBS mAb 370.3 to WT α_IIbβ₃ (Fig. 6A,B). The β₃-D723A or α_IIb-R995A mutation markedly enhanced the talin-1-head-induced binding of both mAbs to full-length α_IIbβ₃ (Fig. 6A,B). However, deletion of α_IIb CT MD region significantly reduced talin-1-head-induced mAb binding (Fig. 6A,B). Similarly, the talin-1 head significantly increased the binding of the β₂ LIBS mAbs KIM127 and M24 to WT α_Lβ₂, but not to α_L-R1094Tr–β₂ (Fig. 6C,D). The expression levels of talin-1 head were comparable among the transfectants (data not shown). Thus, the α-integrin CT MD region is required for talin-induced integrin conformational change.

Fig. 6.

Effect of α-integrin CT truncation on talin-1 head induced integrin conformational change. (A,B) Talin-1-head-induced α_IIbβ₃ epitope exposure for LIBS mAbs 319.4 and 370.3. (C,D) Talin-1-head-induced α_Lβ₂ epitope exposure for mAbs KIM127 and M24. 293FT cells were transfected with the indicated integrin constructs (shown under B and D) and EGFP or EGFP–TH. Cells were first incubated with the biotinylated LIBS mAbs and then stained with PE-labeled streptavidin and Alexa-Fluor-647-labeled mAb AP3 (for α_IIbβ₃), or stained with Alexa-Fluor-647-labeled streptavidin and PE-labeled TS2/4 (for α_Lβ₂). The EGFP and integrin double-positive cells were analyzed by flow cytometry. The binding of LIBS mAb is presented as the MFI normalized to integrin expression. Data are presented as mean±s.e.m (n≥3). *P<0.05; **P<0.01; ***P<0.001; ns, not significant (unpaired two-tailed Student's t-tests).

We next asked whether the α-integrin CT MD region is also important for kindlin-mediated integrin conformational change. When co-expressed with the talin-1 head, kindlin-2, but not kindlin-3 significantly increased the binding of the mAbs 319.4 and 370.3 to WT α_IIbβ₃ (Fig. 7A,B). This is consistent with the data that kindlin-2 but not kindlin-3 enhanced ligand binding to WT α_IIbβ₃ (Fig. 3D). Kindlin-2 also significantly enhanced the binding of KIM127 and M24 to WT α_Lβ₂ (Fig. 7C,D). We also detected kindlin-3-induced binding of M24 to WT α_Lβ₂ (P<0.05) (Fig. 7D). The binding of KIM127 was also increased by kindlin-3, but it is not statistically significant (Fig. 7C). By contrast, both kindlin-2 and kindlin-3 failed to induce the binding of mAbs to the truncated α_IIb-R995Tr–β₃ and α_L-R1094Tr–β₂ integrins (Fig. 7). The expression levels of the talin-1 head and kindlin-2 or kindlin-3 were comparable among the integrin transfectants (data not shown). These results demonstrate that the α-integrin CT MD region is also required for kindlin-mediated integrin conformational change.

Fig. 7.

Effect of α-integrin CT truncation on kindlin-induced integrin conformational change. (A,B) Binding of LIBS mAb 319.4 and 370.3 to α_IIbβ₃ induced by the overexpression of EGFP–TH plus mCherry (mC)-tagged kindlin-2 (mC–K2) or kindlin-3 (mC–K3). (C,D) Binding of LIBS mAb KIM127 and M24 to α_Lβ₂ induced by the overexpression of EGFP–TH plus mC–K2 or mC–K3. 293FT cells were co-transfected with the integrin constructs (shown under B and D) and EGFP–TH plus mC, mC–K2, or mC–K3. The integrin, EGFP, and mCherry triple-positive cells were analyzed by flow cytometry. The binding of LIBS mAb was presented as the MFI normalized to integrin expression. Data are presented as mean±s.e.m (n≥3). *P<0.05; **P<0.01; ***P<0.001; ns, not significant (unpaired two-tailed t-tests were used for the comparison with mC control group).

Deletion of the α_IIb integrin CT MD region reduced α_IIbβ₃ transmembrane heterodimerization but did not affect the binding of the talin-1 head

To have direct evidence that the α_IIb CT MD region contributes to α_IIbβ₃ TM association, we performed a disulfide crosslinking assay with integrins on the cell surface. Consistent with our previous data (Zhu et al., 2009), disulfide bonds were formed at high efficiency between the cysteine mutations of interfacial residues α_IIb-G972C and β₃-V700C at the TM domains (Fig. 8A), indicating the close TM association of α_IIbβ₃ (Fig. 1A). However, the disulfide bond formation was significantly reduced upon the deletion of α_IIb CT MD region (Fig. 8A), suggesting a disassociation of TM domains. As a control, deletion of both the MP and the MD regions of α_IIb CT (α_IIb-G972C-V990Tr) further reduced disulfide bond formation (Fig. 8A). This is consistent with the ligand-binding assay. Deletion of α_IIb CT MD region greatly enhanced PAC-1 binding to the α_IIb-G972C mutant but not the β₃-V700C mutant, whereas deletion of the α_IIb CT MP region further enhanced PAC-1 binding due to the increased dissociation of α_IIbβ₃ transmembrane domains (Fig. 8A,B). This data clearly demonstrate that the α_IIb CT MD region contributes to α_IIbβ₃ transmembrane association, which helps to keep integrin inactive. We next asked whether deletion of the α_IIb CT MD affects the binding of talin-1 head to α_IIbβ₃ integrin. As shown in Fig. 8C, the EGFP–talin-1-head was co-immunoprecipitated with both WT α_IIbβ₃ and α_IIb-R995Tr–β3 mutant, suggesting that the loss of talin-mediated integrin activation in the absence of α_IIb CT MD region is not due to the effect on talin-1 head binding.

Fig. 8.

Deletion of α_IIb CT MD region reduced α_IIb and β₃ transmembrane domain association but did not affect the binding of the talin-1 head. (A) Disulfide crosslinking of indicated α_IIbβ₃ constructs expressed in 293FT cells. Immunoprecipitates of [³⁵S]-labeled integrins were subjected to non-reducing SDS-PAGE and autoradiography (upper panel). The quantification of crosslinking efficiency is shown in the lower panel. Data are presented as mean±s.e.m (n≥3). (B) PAC-1 binding of 293FT cells transfected with indicated α_IIbβ₃ constructs in the buffer containing 1♣mM Ca²⁺ and 1♣mM Mg²⁺ (Ca/Mg) or 0.2♣mM Ca²⁺ plus 2♣mM Mn²⁺ (Ca/Mn). (C) Co-immunoprecipitation of the talin-1 head with α_IIbβ₃ integrin. 293FT cells were co-transfected with α_IIbβ₃ constructs and EGFP or EGFP–TH. Integrins were immunoprecipitated (IP) with mAb 10E5 and subjected to western blotting (WB) with anti-EGFP and anti-β₃ antibodies.

DISCUSSION

We present evidence to support the previously unappreciated structural roles of the α-integrin CT MD region in integrin activation. In addition to the well-conserved MP region, the α-integrin CT MD region also contributes to maintaining integrin in the resting state. Deletion of the α_IIb CT MD region lowers the threshold of mutation-induced integrin activation. This is due to the reduced association of α_IIbβ₃ TM domains in the absence of the α_IIb CT MD region as demonstrated by a disulfide crosslinking assay. By contrast, we demonstrate that the α-integrin CT MD region is required for talin- and kindlin-induced inside-out activation of α_IIb, α_V, and α_L integrins, which is indicative of an essential and conserved mechanism of integrin family activation. The α-integrin CT MD region is dispensable for talin binding but is required for talin- and kindlin-induced integrin conformational change. Our data explain to a certain extent the previous observations that deletion of the α-integrin CT MD region diminished cell adhesion mediated by α₁, α₂, α₄, α_V and α₆ integrins (Kassner and Hemler, 1993; Kawaguchi and Hemler, 1993; Shaw and Mercurio, 1993; Filardo and Cheresh, 1994; Kassner et al., 1994; Kawaguchi et al., 1994; Yauch et al., 1997; Abair et al., 2008) and that α_L CT MD deletion reduced the sensitivity of α_Lβ₂ integrin to PMA stimulation (Lu and Springer, 1997).

It is not readily known how the α-integrin CT MD region contributes to maintaining the resting state because current structural information shows diverse structures for integrin cytoplasmic tails (supplementary material Fig. S4). Several models could be proposed according to the data available (supplementary material Fig. S4C). First, our data demonstrate that in addition to the MP regions, the α_IIb MD region contributes to integrin TM association on the cell surface. A recent report has shown that complete deletion of the α_IIb CT MD region reduced the interaction of isolated α_IIb and β₃ TM-CT peptides co-expressed on the cell surface (Li et al., 2014). Our disulfide crosslinking experiments (Zhu et al., 2009) and a direct interaction assay (Ginsberg et al., 2001) indicate a close association of α_IIb MD region with β₃ cytoplasmic tail. Thus, the α_IIb MD region might contribute to the structural stability of an α_IIbβ₃ TM-CT heterodimer through direct interaction with the β₃ cytoplasmic tail. This interaction might be sequence specific given that replacing the α_IIb CT with an α₅ CT rendered α_IIbβ₃ constitutively active despite α_IIb and α₅ having a very similar CT MP region (O'Toole et al., 1991). Furthermore, replacing the α_IIb CT MD region with a penta-alanine sequence reduces the interaction of isolated α_IIb and β₃ TM-CT peptides (Li et al., 2014). Second, the C-terminal α_IIb CT MD region is unique with seven acidic amino acids (supplementary material Fig. S1) that have been shown to interact with the N-terminal α_IIb CT MP region in a NMR structure of membrane-anchored α_IIb CT peptide (Vinogradova et al., 2000). This interaction might be important in keeping the resting integrin state, but the same interaction was not found in the NMR structure of the α_IIbβ₃ TM-CT heterodimer determined in organic solvent (Yang et al., 2009) (Fig. 1C). Another possible mechanism of α-integrin CT MD region contributing to the inactive state is through its interaction with the negative integrin regulators. Most of the negative regulators identified so far bind to the conserved α-integrin CT MP region (Pouwels et al., 2012; Bouvard et al., 2013; Morse et al., 2014), but the MD region might also contribute to the interaction to some extent, which helps maintain the resting integrin state.

It has been shown that membrane-permeant peptides containing the α_IIb CT MD sequences specifically blocked α_IIbβ₃ activation in platelets (Ginsberg et al., 2001; Koloka et al., 2008). A recent study has suggested that this is because the exogenous α_IIb CT MD peptides inhibit the association of talin with α_IIbβ₃ in thrombin-stimulated platelets (Gkourogianni et al., 2013). However, this is unlikely owing to competition with or steric hindrance of the talin-binding sites of β₃ CT given that deletion of the α_IIb CT MD region did not increase the binding of talin-1 head to β₃ CT (Fig. 8C and (Li et al., 2014). Talin activates integrin by disrupting the α-integrin–β-integrin TM–CT association, whereas the α_IIb CT MD region negatively regulates this process by enhancing the structural stability of the α_IIbβ₃ TM-CT heterodimer.

In addition to the negative regulation of α CT MD region in integrin activation, our data demonstrate a positive role for the α-integrin CT MD region in talin-induced integrin activation. Displacement of α-integrin CT MD region by the β CT-bound talin-1 head domain might destabilize the α-integrin–β-integrin TM–CT dimerization, leading to integrin activation. The potential steric clashes between the talin-1 head and α_IIb CT MD residues based on structure superposition have been proposed to play a role in this process (Yang et al., 2009; Zhu et al., 2009; Ye et al., 2011). This hypothesis was tested recently by replacing the α_IIb CT MD region with a penta-alanine peptide or the CT MD sequence of α₅ or α_L integrin (Li et al., 2014). It has been shown that the α_IIb chimeras remain responsive to talin-1-head-induced activation. Based on this result, it has been concluded that the α_IIb CT MD region participates in integrin activation through steric clashing with talin-1 head without the requirement of a specific sequence. We examined the potential role of steric clashes using the truncated α_IIb constructs that retain the first two residues of α_IIb CT MD region (i.e. NR997Tr) given that the structural superposition showed major steric clashes between these residues and the talin-1 head (supplementary material Fig. S4B). The steric clash model is not consistent with the results showing that the bulky YY997 mutations significantly reduced, rather than enhancing, the talin-mediated α_IIbβ₃ activation. It should be noted that both the length and sequence of α_IIb CT MD region are important to maintain the resting integrin state. Indeed, truncating the α_IIb CT or swapping the α_IIb CT MD region with other amino acids renders α_IIbβ₃ more active than WT integrin as shown in our study and by Li et al. (Li et al., 2014). In addition, the conformational flexibility of both the α_IIb and β₃ CT MD region (Zhu et al., 2009; Metcalf et al., 2010) would eliminate clashes in a talin–α_IIbβ₃ complex. Thus, the steric clash model is not the sole mechanism through which the α_IIb CT MD region participates in talin-induced integrin activation.

An interaction between α_IIb CT and talin has been reported (Knezevic et al., 1996; Yuan et al., 2006; Gingras et al., 2009; Raab et al., 2010). Although it might not have a major contribution to α_IIbβ₃ and talin interaction, a transient and weak interaction between α_IIb CT and talin might stabilize the active conformation of α_IIb CT during integrin activation. Recent biochemical and molecular dynamic simulation studies indicate that the α_IIb CT increases its membrane embedding upon integrin inside-out activation (Kurtz et al., 2012; Provasi et al., 2014). The α_IIb-YY997 mutation might affect the membrane embedding of α_IIb TM-CT and reduces integrin activation since tyrosine residues tend to locate at the membrane boundary (Killian and von Heijne, 2000). A cluster of negatively charged residues at α_IIb CT MD region may disfavor the negatively charged cell membrane and inhibit the membrane embedding of α_IIb TM-CT, while an interaction between talin-head and α_IIb CT MD acidic residues as indicated by a molecular dynamic simulation study (Provasi et al., 2014) might facilitate the membrane embedding of α_IIb TM-CT to activate integrin. In addition, the capability of talin to activate integrin depends on its binding to both β-integrin CT and cell membrane (Wegener et al., 2007; Anthis et al., 2009; Moore et al., 2012; Song et al., 2012; Kalli et al., 2013). The interaction between α CT and talin might help talin maintain a proper orientation for the favorable interaction with β-integrin CT and cell membrane in order to stabilize the active conformation of TM-CT (supplementary material Fig. S4C). Considering the significant diversity of the α-integrin CT MD region, it is tempting to propose that this diversity might regulate integrin inside-out activation in an α-integrin-specific manner as it has been suggested that the α-integrin CT MD regions determine the specificity of integrin outside-in signaling (Chan et al., 1992; Shaw et al., 1995; Sastry et al., 1999; Na et al., 2003; Goel et al., 2014).

Another potential mechanism of regulating integrin activation by α-integrin CT is through its interacting proteins. Interestingly, most α-integrin-CT-binding proteins identified so far interact with the conserved membrane-proximal GFFKR motif and function as inhibitors of integrin activation (Pouwels et al., 2012; Bouvard et al., 2013; Morse et al., 2014), but the mechanism of their negative regulation is not well-defined. A Rap1 effector, RapL, has been shown to activate α_Lβ₂ integrin through binding to the α_L CT MD sequence proximal to the GFFKR motif (Katagiri et al., 2003). It is not known how RapL cooperates with talin and/or kindlin to activate α_Lβ₂ integrin. The requirement of α-integrin CT MD region for integrin activation suggests that as-yet undiscovered regulatory proteins that interact with α-integrin CT MD region are involved in integrin activation, probably in an integrin-specific manner.

It has been shown that deletion of the α_IIb CT MD region did not affect α_IIbβ₃-mediated cell adhesion and spreading, but led to aberrant recruitment of α_IIbβ₃ into focal adhesions formed by other integrins (Ylänne et al., 1993). The underlined mechanism is not known. The same phenomenon was observed with α₂ integrin (Kawaguchi et al., 1994). However, deletion of the α CT MD region of α₁, α₂, α₄, α_V or α₆ integrin diminished cell adhesion (Kassner and Hemler, 1993; Kawaguchi and Hemler, 1993; Shaw and Mercurio, 1993; Filardo and Cheresh, 1994; Kassner et al., 1994; Kawaguchi et al., 1994; Yauch et al., 1997; Abair et al., 2008), indicating that there are different regulatory roles of α CT MD regions among different integrins.

Our data on the β₃-K716A mutation have important implications for the current model of integrin inside-out activation. It has been suggested that the β₃-K716 residue helps to maintain a proper tilt angle for the β₃ TM domain in the resting state by placing its basic side chain near negatively charged phospholipid head groups (Kim et al., 2011a). It has been proposed that the β₃-K716A mutation or talin binding to β₃ CT induces integrin activation by changing the tilt angle of β₃ TM domain to disturb the α_IIbβ₃ TM-CT interaction. It has also been shown that the tilt angle change of β₃ TM domain can occur upon adding the talin-1 head to the isolated β₃ TM-CT peptide embedded in lipid nanodiscs in the absence of α_IIb TM-CT peptide (Kim et al., 2012). We found that complete deletion of α_IIb CT MD region greatly enhanced the activation of β₃-K716A. If the talin-1 head could induce integrin activation by changing the tilt angle of the β₃ TM domain as β₃-K716A does, which seems to be independent of α_IIb CT, we would expect to see an enhanced activation mediated by the talin-1 head in the absence of the α_IIb CT MD region. However, an opposite result was obtained. The talin-1 head failed to activate integrin when the α-integrin CT MD region was completely deleted. By using a cysteine-scanning accessibility method with intact integrin on the cell surface, another group has suggested that β₃-K716 is embedded in the lipid bilayer both in the inactive and active conformation. They detected conformational change of the α_IIb TM-CT but not the β₃ TM-CT domain in activating integrin mutants (Kurtz et al., 2012). Our structural study revealed that interactions between the β₃-K716 side chain and the α_IIb F992 and K994 backbone carbonyl oxygens are important in maintaining the resting integrin state (Zhu et al., 2009). These interactions were observed in a molecular dynamic simulation study starting with the NMR structure of α_IIbβ₃ TM-CT determined in lipid bicelles (Provasi et al., 2014). The direct interaction between β₃-K716 and α_IIb CT was also observed in a NMR study (Metcalf et al., 2010). Clearly, further investigations are required to reconcile these discrepancies.

It has been proposed that the conserved α_IIb-R995 and β₃-D723 form a salt bridge that restrains α_IIbβ₃ in the resting state (Hughes et al., 1996). One of the current models suggests that talin-1 head binding to the β₃ CT breaks the putative salt-bridge to induce integrin activation (Anthis et al., 2009; Anthis and Campbell, 2011). Despite the very disruptive effect of α_IIb-R995 and β₃-D723 mutations on α_IIbβ₃ TM dimerization observed using isolated TM-CT peptides (Kim et al., 2009; Lau et al., 2009), their mutations only moderately activated full-length integrins (Peyruchaud et al., 1998; Ma et al., 2006; Ghevaert et al., 2008; Zhu et al., 2009). Remarkably, we found that the α_IIb-K993Tr and α_IIb-K994Tr mutation, in which the α_IIb-R995 was deleted, abolished rather than synergized integrin activation induced by β₃-D723 mutations and/or by the talin-1 head. In sharp contrast, it was only in the presence of α_IIb CT MD region that the α_IIb-R995 or β₃-D723 mutations greatly synergized α_IIbβ₃ activation and the conformational change induced by the talin-1 head. Thus, although α_IIb-R995 and β₃-D723 are involved in integrin activation, they might not be central in structurally maintaining integrin in the resting state. Mutagenesis data has suggested that the extracellular domains are also involved in stabilizing the α_IIbβ₃ TM dimer as well as their membrane embedding (Lau et al., 2009; Kurtz et al., 2012). Caution should be taken when interpreting data obtained using isolated TM-CT peptides.

Natural mutations of α_IIb-R995 and β₃-D723 (α_IIb-R995Q or α_IIb-R995W and β₃-D723H) have been identified in patients with thrombocytopenia (a relative decrease of platelets in blood) (Peyruchaud et al., 1998; Ghevaert et al., 2008; Kunishima et al., 2011). Although it was not studied in these patients, our data here suggest that α_IIbβ₃ integrin with α_IIb-R995 or β₃-D723 mutations might be more responsive to agonist stimulation leading to more platelet aggregation than in wild type, which might simultaneously cause a reduction in the amount of platelets in blood. An equivalent Arg to Ala mutation of α₄ integrin CT has been found to aberrantly activate α₄ integrin in genetic knock-in mice (Imai et al., 2008). We also found that the equivalent β₂-D709A mutation renders α_Lβ₂ integrin hyper-responsive to talin-1-head-induced activation. Interestingly, an equivalent Asp to Ala mutation in β₁ integrin CT did not result in obvious defect in vivo (Czuchra et al., 2006), despite the fact that the β₁ subunit forms heterodimers with 12 α-integrin subunits. Thus, the conserved α-Arg and β-Asp of the integrin CT might play different roles in regulating integrin activation among different integrins.

Among the caveats of our study is that the effect of α_IIb CT MD mutations on talin-1-head-mediated integrin activation was much more dramatic with the β₃-D723A mutant compared with WT β₃. It should be noted that overexpression of talin-1 head only moderately activates WT integrin. Increasing the DNA amount for transfection did not further increase integrin activation (data not shown). It is possible that overexpression of talin-1 head alone might not be sufficient to induce maximal integrin activation owing to the lack of the recruitment process of talin-1 head to the integrin tail. In the physiological situation, the active full-length talin might apply traction force to its bound β-integrin CT through the coupled actin cytoskeleton (Zhu et al., 2008; Schürpf and Springer, 2011), and thus exert a more disruptive effect on the integrin TM-CT interaction than that mediated by the talin head alone. The β₃-D723A mutation or other activating mutations like α_IIb-R995A and β₃-G135A might facilitate talin-head-induced integrin activation by decreasing the energy barrier. It significantly increases talin-1-head-induced integrin activation by more than 20-fold compared with WT. Moreover, PMA stimulates a significant increase in soluble ligand binding to the α_IIb–β₃-D723A mutant but not to the WT in K562 cells in suspension. Consistent with the soluble ligand-binding assay, the β₃-D723A or α_IIb-R995A mutation significantly enhanced talin-1-head-induced integrin conformational change. Thus, the combination of β₃-D723A and α_IIb CT MD mutations greatly improves the sensitivity of our assay. Similarly, the combination of the β₃ activating mutations and α_IIb CT MD truncations enabled us to reveal the contribution of α_IIb CT MD region to maintaining the resting state.

Talin-1-head-induced integrin extension, but not headpiece opening, has been directly visualized by EM using the purified intact α_IIbβ₃ embedded in the lipid nanodiscs (Ye et al., 2010). We have recently shown that talin-1-head-induced integrin conformational change needs to be propagated to the ligand-binding site, probably through headpiece opening, in order to activate integrin (Zhang et al., 2013). In this study, by using the conformation-dependent mAbs that report integrin extension (319.4 for β₃ and 370.3 for α_IIb; KIM127 for β₂) and headpiece opening (M24 for β₂), we further demonstrated that the talin-1 head induced extension and headpiece opening of integrin. In particular, we detected enhanced integrin conformation change (extension and headpiece opening) when the talin-1 head was co-expressed with kindlins. Kindlin-2 or -3 exerted more effect on M24 binding than did KIM127 binding to α_Lβ₂. This is consistent with a recent study showing that both talin-1 and kindlin-3 were required for inducing the extended open headpiece conformation of α_Lβ₂ (Lefort et al., 2012). Remarkably, the talin-1-head- and kindlin-induced integrin extension and headpiece opening require the presence of an α-integrin CT MD region. However, because how kindlins induce integrin activation remains unknown, the requirement of an α-integrin CT MD region for kindlin-mediated integrin activation can only be interpreted as a secondary effect due to the loss of effectiveness of talin according to the current data.

We found that the talin-1-head- and kindlin-induced binding of β₃ LIBS mAb 319.4 is not completely abolished in the absence of the α_IIb CT MD region. This is in contrast with the binding of α_IIb LIBS mAb 370.3. This indicates that the talin-1 head might exert some extent of conformational change on the β₃ subunit even in the absence α_IIb CT MD region, but that a fully active integrin conformation induced by talin and kindlin requires the involvement of the α_IIb CT MD region. In summary, our study provides new insights into integrin inside-out activation and suggests that further structural studies are required to understand the precise mechanism by which the α-integrin CT MD region is involved in talin- and kindlin-mediated integrin activation.

MATERIALS AND METHODS

DNA constructs

DNA constructs of human α_IIbβ₃, α_Vβ₃, α_Lβ₂, and the EGFP-tagged mouse talin-1 head (EGFP–TH) were as described previously (Zhu et al., 2007; Bouaouina et al., 2008; Zhang et al., 2013). Human kindlin-2 or kindlin-3 was cloned into the pmCherry-C1 vector (Clontech). Mutations were introduced by site-directed mutagenesis with the QuikChange kit (Agilent Technologies).

Antibodies and ligands

PAC-1 (BD Bioscience) is a ligand-mimetic mAb (IgM) specific for activated α_IIbβ₃ integrin (Shattil et al., 1985). AP3 is a conformation-independent anti-β₃ mAb (Kouns et al., 1991). 319.4 is a LIBS mAb that binds to β₃ I-EGF domains (Zhang et al., 2013). 370.3 is a LIBS mAb that binds to the α_IIb calf-1 domain (Zhang et al., 2013). PE-labeled TS2/4 (BioLegend) is a non-functional anti-α_L mAb (Sanchez-Madrid et al., 1982). KIM127 (which binds to I-EGF-2 domain) and mAb 24 (M24, which binds to βI domain) are anti-β₂ LIBS mAbs (Dransfield and Hogg, 1989; Robinson et al., 1992; Nishida et al., 2006; Chen et al., 2010). Human Fn9-10 was as described previously (Takagi et al., 2001). Human fibronectin and ICAM-1 (with a C-terminal human IgG1 Fc tag, ICAM-1-Fc) were purchased from Sigma-Aldrich and Sino-Biological, respectively.

Soluble ligand binding assay by flow cytometry with 293FT transfectants

PAC-1 binding of 293FT (Life Technologies) cells transfected with α_IIbβ₃ only were as described previously (Zhang et al., 2013). For talin-1-head-induced ligand binding, 293FT cells were co-transfected with integrin constructs and EGFP or EGFP–TH for at least 24♣hours. For kindlin-induced ligand binding, 293FT cells were co-transfected with integrin constructs plus EGFP–TH and mCherry or mCherry–kindlin for at least 24♣hours. Ligand binding was performed in HBSGB buffer (20♣mM HEPES pH♣7.4, 150♣mM NaCl, 5.5♣mM glucose, and 1% BSA) plus 5♣mM EDTA or 1♣mM Ca²⁺ and 1♣mM Mg²⁺ (Ca/Mg) at 25°C for 30♣min with 5♣∀g/ml PAC-1 for α_IIbβ₃, 20♣∀g/ml each of ICAM-1-Fc and biotin-labeled mouse anti-human IgG1 Fc for α_Lβ_2, or 50♣∀g/ml Alexa-Fluor-647-labeled Fn9-10 or Fn for α_Vβ₃. Cells were then washed and incubated with Ca/Mg on ice with the detecting reagents: 10♣∀g/ml each of biotinylated AP3, PE-labeled streptavidin and Alexa-Fluor-647-labeled goat anti-mouse IgM for α_IIbβ₃, biotinylated AP3 and PE-labeled streptavidin for α_Vβ₃, PE-labeled TS2/4, and Alexa Fluor 647-labeled streptavidin for α_Lβ₂. Integrin and EGFP double-, or integrin, EGFP, and mCherry triple-positive cells were acquired for calculating mean fluorescence intensity (MFI) by flow cytometry. Ligand binding was presented as the normalized MFI, that is ligand MFI (after subtracting the ligand MFI in EDTA) as a percentage of integrin MFI.

PMA-induced integrin ligand binding in K562 cells

Human K562 cells were transfected with α_IIbβ₃ constructs by Lipofectamine 2000 (Life Technologies). Cells were incubated in RPMI-1640 medium plus 1% BSA with PMA for 10♣minutes followed by a 10-minute incubation with PAC-1. Cells were then washed and incubated with Alexa-Fluor-488-labeled AP3 and Alexa-Fluor-647-labeled goat anti-mouse IgM on ice for 30♣minutes. Integrin-positive cells were acquired for calculating MFI by flow cytometry. Ligand binding was presented as MFI normalized by integrin expression.

LIBS epitope exposure

Talin-1-head- and kindlin-induced LIBS epitope exposure was as described previously (Zhang et al., 2013). In brief, 293FT transfectants were first incubated with or without (as background) the biotinylated LIBS mAb in Ca/Mg at 25°C for 30♣mins, and then washed and incubated with the detecting reagents: Alexa-Fluor-647-labeled AP3 and PE-labeled streptavidin for α_IIbβ₃, PE-labeled TS2/4 and Alexa-Fluor-647-labeled streptavidin for α_Lβ₂. Integrin and EGFP double-, or integrin, EGFP and mCherry triple-positive cells were analyzed for calculating the MFI. LIBS mAb binding is presented as MFI normalized to the integrin expression, that is LIBS mAb MFI (after subtracting the background MFI) as a percentage of integrin MFI.

Disulfide crosslinking and immunoprecipitation

Metabolic labeling with [³⁵S]cysteine/methionine and disulfide crosslinking using intact cells of 293FT transfectants were as described previously (Zhu et al., 2009). Formation of disulfide bonds was induced by treating the cells with CuSO₄ and o-phenanthroline on ice. Integrins were immunoprecipitated with anti-α_IIb mAb 10E5 from the cell lysate and subjected to nonreducing SDS-PAGE and autoradiography. For the talin-1-head-binding assay, α_IIbβ₃ integrins were immunoprecipitated with mAb 10E5 from cells co-transfected with EGFP or EGFP–TH. Binding of the talin-1 head was detected by western blotting with rabbit anti-EGFP polyclonal antibody (Origene). β₃ integrin was detected with rabbit anti-β₃ H-96 (Santa Cruz Biotechnology).