How conformational signals initiated from one end of the integrin are transmitted to the other end remains elusive. At the ligand-binding βI domain, the α1/α1′-helix changes from a bent to a straightened α-helical conformation upon integrin headpiece opening. We demonstrated that a conserved glycine at the α1/α1′ junction is crucial for maintaining the bent conformation of the α1/α1′-helix in the resting state. Mutations that facilitate α1/α1′-helix unbending rendered integrin constitutively active; however, mutations that block the α1/α1′-helix unbending abolished soluble ligand binding upon either outside or inside stimuli. Such mutations also blocked ligand-induced integrin extension from outside the cell, but had no effect on talin-induced integrin extension from inside the cell. In addition, integrin-mediated cell spreading, F-actin stress fiber and focal adhesion formation, and focal adhesion kinase activation were also defective in these mutant integrins, although the cells still adhered to immobilized ligands at a reduced level. Our data establish the structural role of the α1/α1′ junction that allows relaxation of the α1/α1′-helix in the resting state and transmission of bidirectional conformational signals by helix unbending upon integrin activation.
Integrins are α/β heterodimeric cell surface receptors that connect the extracellular and intracellular environment (Hynes, 2002). With their large multi-domain compositions of extracellular, transmembrane and cytoplasmic regions, integrins transmit conformational signals bidirectionally across the cell membrane. Binding of talin and/or kindlin to β-integrin cytoplasmic tails propagates conformational change from the cytoplasmic domain to the ectodomains, inducing a high-affinity state for binding extracellular or cell surface ligands, known as inside-out signaling (Kim et al., 2011; Springer and Dustin, 2012). Ligand binding to the ectodomain also induces conformational rearrangements propagated from the ligand-binding site back to the cytoplasmic domains, resulting in intracellular signaling cascades that regulate cell spreading, migration and proliferation, known as outside-in signaling. How conformational signals are relayed from one end of the integrin to the other remains an active area of research.
The integrin α-subunit is composed of β-propeller, thigh, calf-1, calf-2, transmembrane, and cytoplasmic domains (Xiong et al., 2001; Zhu et al., 2008) (supplementary material Fig. S1). Half of human α-integrins contain an extra domain (I) inserted into the β-propeller, named the αI domain, that is responsible for binding ligands (Hynes, 2002). All the integrin β-subunits are composed of βI, hybrid, PSI (plexin–semaphorin–integrin), I–EGF-1–4 (integrin–epidermal-growth-factor), β-tail, transmembrane, and cytoplasmic domains (supplementary material Fig. S1). In the αI-less integrins, the α-integrin β-propeller, thigh and the β-integrin βI, hybrid, PSI and I–EGF-1 form the integrin headpiece, with the β-propeller and the βI domains coming together to form the integrin head that is responsible for binding extracellular ligands (supplementary material Fig. S1), such as fibronectin (Fn) and fibrinogen (Fg) (Luo et al., 2007). These ligands contain a common binding motif of RGD (Arg-Gly-Asp) or RGD-like sequence KQAGD (Lys-Gln-Ala-Gly-Asp, in human Fg) for a subset of integrins. The α-calf-1-2 and the β-I–EGF-2-4 and β-tail form the integrin lower legs. This multi-domain composition requires integrins to transmit long-range conformational rearrangements coupled through intra- and inter-domain communications.
Integrin conformational change was first indicated by exposing ligand-induced binding sites recognized by a subset of monoclonal antibodies (mAbs) (Humphries, 2004). Crystal structure analyses and electron microscopy (EM) revealed that integrins are in a folded, bent conformation in the resting state (Springer and Dustin, 2012). Ligand binding to the integrin head induces integrin headpiece extension from the leg domains and headpiece changing from a closed to an open conformation (supplementary material Fig. S1B,C). Binding of the talin head domain to the integrin cytoplasmic tail also induces integrin extension, visualized by EM (Ye et al., 2010). Detailed conformational changes were obtained by co-crystallization of the αIIbβ3 integrin headpiece with RGD or RGD-like antagonists or pseudoligands (Xiao et al., 2004). Compared with the unliganded structures, a striking change is the swing-out motion of the β3 hybrid away from the βI and the αIIb thigh domains, resulting in an open headpiece conformation (Fig. 1) (Xiao et al., 2004; Zhu et al., 2010). This movement is accompanied by the preceding changes at the βI domain, in which the β1-α1 loop moves toward the RGD-binding MIDAS (metal-ion-dependent adhesion site) with its coordinating ADMIDAS (adjacent to MIDAS) metal ion. This is coupled with an inward movement of the α1/α1′-helix and a piston-like motion of the β6-α7 loop and the α7-helix (Fig. 1A). The α1/α1′-helix is bent at the α1/α1′ junction in the unliganded structure. It changes to a merged and straightened α-helical conformation upon the headpiece opening. The angle between the α1′-helix and the hybrid domain changes from obtuse to straight (Fig. 1A). Crystal structures indicate that RGD-induced conformational change is relayed in the order of MIDAS-|β1-α1-loop|-|α1/α1′-helix|-|β6-α7-loop|-|α7-helix|-hybrid. It is also speculated that the changes will be in the opposite direction when the conformational signal is propagated from the cytoplasmic tail. The change of the α1/α1′-helix might play a pivotal role in bidirectional conformational transmission. The conformational regulation of the βI α1/α1′-helix is distinct from its structural homolog the αI domain, in which the α1-helix remains straight both in the low- and high-affinity state (Lee et al., 1995; Emsley et al., 2000; Shimaoka et al., 2003).
The conformational change of α1/α1′-helix in β1 integrin activation had been predicted by exposure of the 12G10 epitope, which maps to the C-terminus of the α1′-helix (Mould et al., 2002). In addition, the β1 allosteric inhibitory mAbs SG/19 also binds to the C-terminus of the α1′-helix, which restraints the movement of the α1′-helix (Nagae et al., 2012). There are also mutations in or near the α1/α1′-helix that render β1 or β3 integrins constitutively active (Barton et al., 2004; Luo et al., 2009). Soaking RGD peptides into the preformed αvβ3 ectodomain crystal (Xiong et al., 2002), or into the closed α5β1 headpiece crystal (Nagae et al., 2012), only induced an intermediate movement of the β1-α1 loop and α1-helix toward MIDAS. These results indicated limited flexibility of the β1-α1 loop and α1-helix, and the separability of conformational communication. The question is whether the bent to straight change of the α1/α1′-helix is required for high-affinity ligand binding and for bidirectional conformational transduction in the intact integrin on the cell surface. In this study, we addressed this question by introducing mutations into the α1/α1′ junction that facilitate either the bent or straightened conformation of the α1/α1′-helix. Our results demonstrate an essential role of the α1/α1′-helix reshaping in integrin affinity regulation and bidirectional signaling.
The bent conformation of the α1/α1′-helix is a common feature of all the βI domains, and the helix unbending facilitates the high-affinity RGD binding
Protein sequence alignments of the α1/α1′-helices of all the human integrin βI domains show a completely conserved glycine, a relatively conserved hydrophobic isoleucine/leucine/valine (Ile131 in β3), and a conserved hydrophobic leucine/valine/methionine (Leu134 in β3) at the α1/α1′ junction (Fig. 2A). Crystal structure analyses of β3, β1, β2 and β7 integrins at the closed headpiece conformation all revealed a bent conformation of the α1/α1′-helix (Xie et al., 2010; Zhu et al., 2010; Nagae et al., 2012; Yu et al., 2012) (Fig. 2B). The main chain hydrogen bond network of the α-helix is distorted at the α1/α1′ junction. A hydrogen bonded turn structure is formed by Ile131 and Leu134 in β3, enabling the helix bending (Fig. 1B). It is possibly facilitated by the conserved glycine residue at the N-terminus of α1′-helix (Fig. 2B). The conformations of the β6-α7 loop and α7-helix are also conserved among the βI domains.
The RGD-binding pocket of αIIbβ3 integrin is formed by αIIb Asp224 and β3 MIDAS with the surrounding residues (Fig. 2C). The RGD peptide extends its backbone in the binding pocket, with the Arg side chain fitting into the αIIb portion to hydrogen bond to Asp224, and the Asp side chain fits into the β3 portion to coordinate the MIDAS metal ion and hydrogen bonds to the backbone nitrogens of Tyr122 and Ser123 (Fig. 2C, lower panel). The unbending of the α1/α1′-helix facilitates the movements of ADMIDAS and Ser123 to approach the MIDAS, enabling the high-affinity conformation for RGD binding (Fig. 2B, β3 panel and C lower panel).
The conserved glycine maintains the bent conformation of the α1/α1′-helix
To examine the role of the conserved glycine in maintaining the bent conformation of α1/α1′-helix, we first mutated it to alanine. It is known that alanine tends to stabilize the helical structure relative to glycine when present in the internal positions of an α-helix (Serrano et al., 1992). The β3-G135A mutation is constitutively active for binding ligand-mimetic mAb PAC-1 in the presence of physiological concentrations of 1 mM Ca2+ and 1 mM Mg2+ (Ca2+/Mg2+) (Fig. 3C). This result indicates that the alanine substitution favors the straightened α1/α1′-helix conformation in the high-affinity state, whereas the native glycine favors the default bent conformation of α1/α1′-helix in the low affinity state. As a control, the glycine was mutated to proline, because proline tends to break or kink a helix owing to the lack of an amide hydrogen bond donor. As expected, the β3-G135P mutant is indistinguishable from the wild-type integrin when measuring PAC-1 binding in the presence of Ca2+/Mg2+, or 0.2 mM Ca2+ and 2 mM Mn2+ (Ca2+/Mn2+) (Fig. 3C). When combined with the active cytoplasmic mutation αIIb-R995D, which mimics the integrin inside-out activation (Hughes et al., 1996), β3-G135P also had similar levels of PAC-1 binding as wild-type β3 integrin. In contrast, the β3-G135A mutant greatly synergized the activating effect of αIIb-R995D (Fig. 3C). These data demonstrate that the conserved glycine at the α1/α1′ junction maintains the bent conformation of α1/α1′-helix in the resting state.
The β3-G135 was partially buried in the bent α1/α1′-helix conformation and buried deeper into the hydrophobic environment after the helix unbending in the high-affinity state (Fig. 3B). We rationalized that blocking the movement of G135 could restrain the unbending of α1/α1′-helix, and thus the high-affinity ligand binding. We mutated the β3-G135 to arginine and showed that the β3-G135R mutation completely abolished the soluble PAC-1 binding in the presence of Ca2+/Mn2+ (Fig. 3C). It also greatly reversed the activating effect of αIIb-R995D (Fig. 3C). The same effect was found for the β3-G135K mutation (Fig. 3C). In contrast, both β3-G135M and G135L mutations constitutively activated αIIbβ3 integrin (Fig. 3C). Lys, Arg, Met and Leu have similar helix-forming propensities in soluble proteins (Pace and Scholtz, 1998). The different effect on integrin activation is due to their hydrophobic and hydrophilic properties. These results are consistent with the idea that the positively charged arginine or lysine at the G135 position blocks its movement into the hydrophobic environment, thus restraining the α1/α1′-helix unbending, whereas the hydrophobic methionine or leucine facilitates the movement and favors the helix unbending. Thus, the α1/α1′-helix unbending is required for high-affinity ligand binding.
The β3-G135 mutations had different effects on integrin cell surface expression. The β3-G135A mutant expression level was comparable with that of the wild-type, whereas the expression of other mutant integrins was considerably reduced (Fig. 3C). The β3-G135R mutation greatly decreased the integrin expression when combined with the activating αIIb-R995D mutation. Decreased expression was also found for the β3-G135P mutation when in combination with the αIIb-R995D (Fig. 3C). In sharp contrast, the integrin expression was less affected by the combination of αIIb-R995D and β3-G135A mutations (Fig. 3C).
The mutations at the α1/α1′ junction either increased or decreased the activity of αIIbβ3 integrin
To further investigate the structural role of the α1/α1′ junction in integrin activation, we mutated β3 residues Ile131 to Leu134 (Fig. 3A). The hydrophobic Ile131 is relatively conserved and buried in the inner face of the α1-helix (Fig. 2B). It moved inward as a rigid body in its hydrophobic environment along with the α1-helix in the high-affinity conformation (Fig. 2C). We investigated whether mutating this residue to small amino acids would facilitate the inward movement of α1-helix to approach MIDAS. We found that both β3-I131G and β3-I131A mutations constitutively activated integrin for binding PAC-1 in the presence of Ca2+/Mg2+ (Fig. 3D). In contrast, the glycine substitution for another conserved hydrophobic residue at the α1/α1′ junction, β3-L134G, had the opposite effect on αIIbβ3. It greatly reduced the activating effect of Mn2+ (Fig. 3D). It also greatly reversed the activating effect of the αIIb-F993A that mimics the integrin inside-out activation (Fig. 3D).
We next examined whether further increasing the flexibility of the α1/α1′ junction would affect integrin activation. We introduced a triple glycine mutation, β3-QNL-GGG, and a double glycine mutation, β3-QN-GG or β3-NL-GG, into the α1/α1′ junction (Fig. 3A). A ligand binding assay showed that all of these mutations abolished Mn2+-induced integrin activation (Fig. 3E). The β3-QN-GG and β3-NL-GG mutations also completely reversed the activating effect of αIIb-R995D and αIIb-F993A (Fig. 3E). Thus, the increased flexibility at the α1/α1′ junction dampened high-affinity ligand binding by restraining helix unbending. The changing from a bent to a straightened helical conformation is required for high-affinity ligand binding. Consistent with this proposal, the potential helix disruption mutation β3-I131P also abolished Mn2+-mediated integrin activation (Fig. 3D).
Both the triple and double glycine mutations reduced integrin cell surface expression when combined with the αIIb activating mutations (Fig. 3E). When β3-NL-GG was combined with β3-N305T, in which a glycan wedge was introduced into the interface between the βI and the β hybrid domain to enforce headpiece opening (Luo et al., 2003), the cell surface expression was completely abolished (Fig. 3E). The combination of αIIb-F993A and β3-I131P also abolished integrin expression (data not shown). These data indicate that the open headpiece conformation of integrin is compatible with the straightened but not bent α1/α1′-helix on the cell surface.
Unbending of the α1/α1′-helix at the junction is required for the activation of β1 and β2 integrins
To test whether the α1/α1′-helix unbending is a common feature for the activation of other β-integrins, we did the same mutagenesis study on β1 integrin (Fig. 4A). Binding of the human fibronectin type III domains 9–10 (Fn9–10) fragment or activation-specific mAb 9EG7 was measured using 293FT cells transfected with EGFP-tagged wild-type α5 integrin and wild-type or mutant β1 integrins, β1-G146A and β1-KS-GG (Fig. 4B; supplementary material Fig. S2A). The β1-G146A mutation significantly induced Fn9–10 binding in the presence of both Ca2+/Mg2+ and Ca2+/Mn2+ compared with the wild-type (Fig. 4C). In sharp contrast, the β1-KS-GG mutation significantly reduced Fn9–10 binding in the presence of both Ca2+/Mg2+ and Ca2+/Mn2+ compared with the wild-type (Fig. 4C). Binding of 9EG7 (binds to β1 I–EGF-2 domain) was also significantly enhanced by the β1-G146A mutation (Fig. 4D). However, Mn2+-induced 9EG7 binding was significantly reduced by the β1-KS-GG mutation (Fig. 4D). Similar results were obtained with CHO-B2 cells transfected with wild-type α5 and wild-type or mutant β1 integrins (supplementary material Fig. S2B–D). In addition, Fn9–10 binding induced by the activating mAb TS2/16 (binds to β1 βI domain) was significantly enhanced and significantly reduced by the β1-G146A and the β1-KS-GG mutation, respectively (supplementary material Fig. S2B,C). To directly examine the effect of the β1-G146A or β1-KS-GG mutation on α1/α1′-helix conformation, we measured the epitope exposure of mAb 12G10. 12G10 binding was significantly enhanced by the β1-G146A mutation and significantly reduced by the β1-KS-GG mutation in the presence of both Ca2+/Mg2+ and Ca2+/Mn2+ (Fig. 4E). These data demonstrate that the α1/α1′-helix unbending is required for β1 integrin activation.
We next tested the same mutations on β2 integrin (Fig. 4A). Binding of ICAM-1 to 293FT cells co-transfected with wild-type or mutant β2 integrin and wild-type αL integrin was measured in the presence of Ca2+/Mg2+ or Ca2+/Mn2+ with or without the activating mAb MEM-148 (supplementary material Fig. S3A). The β2-G128A mutation significantly enhanced ICAM-1 binding, whereas the β2-KK-GG mutation significantly reduced ICAM-1 binding in all conditions compared with wild-type β2 integrin (Fig. 4F; supplementary material Fig. S3B). In addition, mutating the native Gly129 to threonine that is present in the equivalent positions of β1 and β3 integrins further enhanced the activating effect of the β2-G128A mutation in the presence of Ca2+/Mg2+ or Ca2+/Mg2+ plus MEM-148 (Fig. 4F; supplementary material Fig. S3B). Overexpression of talin1 head domain significantly induced ICAM-1 binding to wild-type β2 and the β2-G128A mutant, but failed to do so for the β2-KK-GG mutant (Fig. 4G). These data provide overwhelming evidence that the α1/α1′-helix unbending is also required for β2 integrin activation stimulated from both outside and inside the cell.
Integrin with unactivatable mutations at the α1/α1′ junction still mediate cell adhesion to immobilized ligand
αIIbβ3 integrin can mediate cell adhesion to high-density immobilized ligand without the requirement of pre-activation (Savage and Ruggeri, 1991). To investigate whether integrin with unactivatable mutations at the α1/α1′ junction still mediate cell adhesion to immobilized human Fg, we generated CHO-k1 stable cell lines expressing αIIb/β3, αIIb/β3-G135R or αIIb/β3-NL-GG (Fig. 5A). At the Fg coating concentrations above 5 µg/ml, cell adhesion of the mutant cells was comparable with that of the wild-type (Fig. 4B). However, both β3-G135R and β3-NL-GG reduced cell adhesion at the Fg coating concentrations lower than 5 µg/ml (Fig. 5B). This is not due to the reduced expression because both mutants had higher levels of cell surface expression than the wild-type (Fig. 5A). The adhered mutant cells were less resistant to hard washing than the wild-type cells, consistent with low-affinity binding (data not shown). Cell adhesion was blocked by the high-affinity αIIbβ3-specific antagonist eptifibatide in a concentration-dependent manner (Fig. 4C). The mutant cells were more sensitive to eptifibatide inhibition than wild-type cells. Eptifibatide markedly inhibited cell adhesion at concentrations as low as 3 µM for both mutant cells, compared with the requirement of more than 10 µM eptifibatide for the wild-type cells (Fig. 4C). These results demonstrate that the unactivatable α1/α1′ mutants still mediate cell adhesion, but at a reduced level. It also shows that the unactivatable α1/α1′ mutants still bind the high-affinity RGD-like antagonists.
Unbending of the α1/α1′-helix at the junction is required for ligand induced integrin extension
We used one anti-αIIb ligand-induced binding site (LIBS) mAb 370.3, and two anti-β3 LIBS mAbs, LIBS-1 and 319.4 to determine whether the α1/α1′-helix unbending is required for ligand-induced integrin extension. These antibodies recognize the epitopes at the αIIb or β3 leg domains, which are masked in the bent integrin conformation (supplementary material Fig. S1C). Eptifibatide or tirofiban markedly augmented the binding of all three LIBS mAbs to CHO-k1 or HEK293 cells stably expressing wild-type αIIbβ3 (Fig. 6A–F). In contrast, eptifibatide-induced mAb binding was greatly reduced by the β3-G135R or β3-NL-GG mutation (Fig. 6A–C). There was no difference between the wild-type and the mutant integrins in LIBS mAbs binding in the absence of ligands. Similar results were obtained for eptifibatide- or tirofiban-induced LIBS exposure with HEK293 cells (Fig. 6D–F). As a control, HEK293 cells expressing αIIb/β3-G135A had higher level of LIBS exposure than wild-type cells both in the absence and presence of eptifibatide or tirofiban (Fig. 6D–F). This is consistent with the high affinity of β3-G135A and the ability to facilitate α1/α1′-helix unbending. These data demonstrate that α1/α1′-helix unbending is required to relay the ligand-induced conformational signal to integrin leg domains.
Unbending of the α1/α1′-helix at the junction is required for talin-induced integrin ligand binding, but not for talin-induced integrin extension
Integrin conformational rearrangement is bidirectional. We showed that ligand-induced integrin conformational transmission to the leg domains was blocked by the unactivatable mutations at the α1/α1′ junction. We next investigated whether the conformational change was affected in the opposite direction, in which the conformational signal was propagated from the integrin cytoplasmic tail, induced by talin binding. Overexpression of EGFP-conjugated talin1 head domain (hereafter referred to as talin1-head) induced PAC-1 binding to wild-type αIIbβ3 (Fig. 7A). Talin1-head also greatly induced PAC-1 binding to the active αIIb-R995A mutant, indicating a synergistic effect (Fig. 7A). However, talin1-head failed to induce PAC-1 binding to the β3-NL-GG mutant (Fig. 7A). The activating effect of talin1-head was also markedly reduced on the combined αIIb-R995A/β3-NL-GG mutant (Fig. 7A). This is consistent with the data of the β2-KK-GG mutant (Fig. 4G). When the LIBS exposure was measured, the binding of LIBS-1 and 370.3 induced by talin1-head was comparable between the wild-type and the β3-NL-GG mutant integrins, although the talin1-head-induced binding of 319.4 was not significant for the β3-NL-GG mutant (Fig. 7B–D). However, talin1-head induced similar levels of LIBS exposure for the αIIb-R995A/β3 and the αIIb-R995A/β3-NL-GG mutant (Fig. 7B–D).
Similar results were obtained for β1 integrin. Overexpression of talin1-head significantly induced the binding of Fn9–10 (Fig. 7E), 9EG7 (Fig. 7F) and 12G10 (Fig. 7G) for both wild-type β1 and β1-G146A mutant. In contrast, talin1-head failed to induce the binding of Fn9–10 (Fig. 7E) and 12G10 (Fig. 7G) to β1-KS-GG mutant. However, talin1-head still significantly induced 9EG7 binding to β1-KS-GG mutant (Fig. 7F). 9EG7, LIBS-1, 319.4 and 370.3 all report integrin extension by recognizing the epitopes at the integrin leg domains (supplementary material Fig. S1C), whereas 12G10 reports the conformational change of α1/α1′-helix (Mould et al., 2002). These data demonstrate that the talin1-induced integrin extension was not affected by the unactivatable mutation at the α1/α1′ junction, but the mutants failed to relay the conformational signal to the ligand-binding site because α1/α1′-helix unbending was blocked.
Unbending of the α1/α1′-helix is required for integrin outside-in signaling
To determine the requirement of α1/α1′-helix unbending in integrin outside-in signaling, we undertook a cell spreading assay with CHO-k1 stable transfectants. The wild-type αIIbβ3 cells spread very well on immobilized Fg (Fig. 8A). In contrast, cell spreading of both αIIb/β3-G135R and αIIb/β3-NL-GG cells was abnormal (Fig. 8A). The mutant cells showed a significant (P<0.005) reduction of cell spreading area compared with wild-type cells on both immobilized Fg and PAC-1 (Fig. 8B,C). Similar results were obtained with HEK293 stable transfectants, in which the wild-type αIIb/β3 cells spread very well, whereas the αIIb/β3-G135R cells remained round or spread less on immobilized Fg (supplementary material Fig. S4B). As a control, the αIIb/β3-G135A cells spread as well as or better than the wild-type cells (supplementary material Fig. S4B). The three stable transfectants had similar levels of integrin cell surface expression (supplementary material Fig. S4A). The αIIb/β3-G135A cells were constitutively active, whereas the αIIb/β3-G135R cells were inactive in binding soluble PAC-1 (supplementary material Fig. S4C).
Integrin-mediated cell spreading is coupled with the formation of F-actin stress fibers and focal adhesions, and focal adhesion kinase (FAK) activation, which are the hallmarks of integrin outside-in signaling (Yamada and Miyamoto, 1995). Confocal microscopy of wild-type αIIb/β3 cells adhered on immobilized Fg showed obvious F-actin stress fibers anchored at the vinculin-accumulating sites (Fig. 8D). The phosphorylated FAK at Try397 was also concentrated at the anchor sites of F-actin stress fibers in the wild-type αIIb/β3 cells, demonstrating normal focal adhesion formation (Fig. 8D). In sharp contrast, rare or no F-actin stress fibers were present in the αIIb/β3-G135R and αIIb/β3-NL-GG cells, although many of the cells still changed their shape after adhering to immobilized Fg (Fig. 8D). The accumulation of vinculin or phosphorylated FAK (pY397) were also rarely visible in the mutant cells (Fig. 8D). Thus, mutations blocking α1/α1′-helix unbending dampened integrin outside-in signaling.
Crystal structures of the αIIbβ3 headpiece in the closed and open conformation revealed a transition of the α1/α1′-helix from a bent to a straight helical conformation upon the headpiece opening (Xiao et al., 2004; Zhu et al., 2010). In this study, we demonstrate that the junction of the α1- and α1′-helix regulates the conformational change of α1/α1′-helix, and thus the integrin affinity and bidirectional signaling. A completely conserved glycine at the α1/α1′ junction maintains the α1/α1′-helix in the bent conformation that is compatible with the low-affinity state of the ligand-binding site. Mutating the glycine to alanine rendered β1, β2 and β3 integrins constitutively active by favoring the unbent α1/α1′-helix conformation. In contrast, increasing the flexibility of the α1/α1′ junction by tandem glycine mutations hindered the activation of β1, β2 and β3 integrins upon stimulations from either outside or inside the cell. These mutations favor the bent α1/α1′-helix conformation. Thus, the α1/α1′-helix unbending is a common feature of integrin activation.
There are two conserved hydrophobic residues, Ile131 and Leu134 at the α1/α1′ junction of β3 integrin. Both residues are packed into the hydrophobic environment in the bent α1/α1′-helix conformation. They slide and bury deeper upon the movement of the α1-helix towards MIDAS coupled with the α1/α1′-helix unbending (Fig. 2B,C). Glycine mutations of these residues conferred two distinct effects, activating and inactivating for Ile131 and Leu134, respectively. This is consistent with their structural role in the conformational change at the α1/α1′ junction. The Ile131 residue moved rigidly in its hydrophobic environment upon α1/α1′-helix unbending. Small amino-acid substitutions for Ile131 could facilitate the movement of the β1-α1 loop and α1-helix to approach the RGD ligand for high-affinity interaction. This could compensate the potential α-helix destabilizing effect of the I131G mutation. The I131A mutation potentially had effects on stabilizing the α-helix and facilitating the movement, and therefore was more active than the I131G mutation. However, the I131P mutation might stabilize the kinked conformation and was inactivating. Leu134 directly displaced the conserved hydrophobic Val340 of the α7-helix upon α1/α1′-helix unbending at the junction, enforced the downward movement of α7-helix (Xiao et al., 2004). This was coupled with the swing-out of the hybrid domain that leads to the high-affinity open headpiece conformation. The L134G mutation inactivated αIIbβ3 integrin by both destabilizing the α-helical structure at the junction and reducing its displacement effect on α7-helix. The β3-Ile131 equivalent mutation in β1 integrin, β1-V142A, also constitutively activated α5β1 integrin, whereas the β3-Leu134 equivalent mutation, β1-L145A did not affect ligand binding (Barton et al., 2004).
The important role of β3-Ile131 in maintaining the low-affinity state by restraining the mobility of α1-helix is consistent with the inactivating effects of the triple or double glycine mutations introduced after Ile131. Presumably, glycine mutations introduced into the α1/α1′ junction should increase the mobility of α1-helix, and thus facilitate ligand binding. However, the increased flexibility of the α1/α1′ junction could not overcome the energy barrier of moving the hydrophobic Ile131. Moreover, the increased flexibility at the α1/α1′ junction strongly hindered the merging and straightening of the α1/α1′-helix that is required for high-affinity ligand binding.
Our data have important implications for integrin deactivation after ligand release from the binding site. The conserved glycine at the α1/α1′ junction would relax the straight α-helical conformation stabilized by the interacting ligand back to the bent conformation after ligand dissociation. Thus, the conserved structural features of the α1/α1′-helix act as an intrinsic allosteric property of βI domain in integrin deactivation. Antibodies or small molecules that restrain the bent to straight transition of α1/α1′-helix could allosterically block integrin ligand binding.
Mn2+ had been widely used to enhance integrin ligand binding and conformational change (Gailit and Ruoslahti, 1988). It has been suggested that Mn2+ works on ADMIDAS probably by facilitating the movement of ADMIDAS towards MIDAS (Chen et al., 2003). We found that PAC-1 binding was decreased in the presence of Ca2+/Mn2+ for β3-G135M, β3-G135L, β3-I131G and β3-I131A compared with that in Ca2+/Mg2+ (Fig. 3C,D). Although not decreased, there was also little synergetic effect of Mn2+ on β3-G135A (Fig. 3C). This is in contrast to the obviously enhanced activating effect of Mn2+ on αIIb-R995D and αIIb-F993A (Fig. 3C,D). Mn2+ might exert different effects on the metal ion binding site with the different mutations. The underlined mechanism needs further investigation.
Although the unactivatable mutations at the α1/α1′ junction greatly affected the soluble ligand binding measured by flow cytometry, the mutant integrins still mediated cell adhesion to immobilized ligands. However, the mutant cells had decreased adhesion on the surface with low coating density of Fg and were more sensitive to hard washing and antagonists than the wild-type cells, demonstrating weak integrin–ligand interaction. It has been known that, unlike soluble ligand binding, αIIbβ3 integrin can mediate cell adhesion to immobilized Fg without pre-activation (Savage and Ruggeri, 1991). The conformational change of Fg upon immobilization (Moskowitz et al., 1998), the high local ligand density (Jirousková et al., 2007), and a small subset of integrins on the cell surface that are readily accessible to bind ligand all possibly account for this phenomenon. Our data indicate that the initial or nascent adhesion to immobilized ligand does not require the high affinity of integrin. The limited flexibility of the β1-α1 loop and α1-helix, as indicated by the crystal structures with soaked RGD (Xiong et al., 2002; Nagae et al., 2012), could contribute to the low-affinity ligand binding. This low-affinity interaction might be important for integrins to sense the changes in ligand concentration and/or accessibility, for example during cell migration.
How are integrin conformational changes initiated at the ligand-binding site transmitted to other domains? Recently, by carefully soaking the closed αIIbβ3 headpiece crystals with different concentrations of RGD peptides, we observed sequential movements of the β1-α1 loop, α1-helix, ADMIDAS metal ion, α1′-helix, β6-α7 loop and α7-helix all the way to the open headpiece conformation (Zhu et al., 2013). At the final stage of headpiece opening, the α1/α1′ junction was strained by the rigid movement of α1-helix. Merging and unbending of α1/α1′-helix occurred at the junction (Zhu et al., 2013). The unbending at the α1/α1′ junction directly pushed the α7-helix downward and this levered the hybrid domain, causing it to swing out. In the full-length integrin, the large interfaces between the hybrid domain and the lower leg I–EGF domains contribute to stabilizing the bent conformation (Xiong et al., 2001; Zhu et al., 2008; Xie et al., 2010). Hybrid domain swing-out is always accompanied by integrin extension, as demonstrated by mutagenesis (Luo et al., 2003). With the full-length integrin on the cell surface, we found that favoring the unbent α1/α1′-helix by alanine mutation constitutively exposed the 12G10 epitope reporting the conformational change of the α1/α1′-helix, and exposed the epitopes at the leg domains reporting integrin extension. In contrast, mutations favoring the bent α1/α1′-helix have the opposite effect. Restraining the unbending of α1/α1′-helix at the junction dramatically blocked RGD-like ligand-induced integrin extension. Thus, the α1/α1′-helix reshaping at the junction plays a pivotal role in the ligand-induced conformational transmission from the ligand-binding site to the leg domains through hybrid domain swing-out.
Integrin conformational change is bidirectional. RGD binding to the integrin head induces headpiece opening and extension from outside. Talin and/or kindlin binding to the cytoplasmic tail induce integrin extension from inside. Talin1-head-induced integrin extension had been visualized by EM for the full-length αIIbβ3 integrin (Ye et al., 2010). Mutations that disrupt the interactions at the transmembrane or cytoplasmic domains also induce integrin extension (Luo et al., 2004a; Luo et al., 2005). However, it is not known how the conformational signal initiated from the cytoplasmic tail is relayed to the ligand-binding site. We found that mutations restraining the α1/α1′-helix unbending completely abolished integrin–ligand binding induced by the cytoplasmic mutations or by the overexpression of talin1 head domain. Talin1 head also failed to induce the 12G10 epitope exposure with the unactivatable mutations at the α1/α1′ junction. Interestingly, talin1-head-induced integrin extension was not affected. These data are consistent with the concept that integrin extension alone does not result in high-affinity ligand binding (Springer and Dustin, 2012). The inside-out integrin activation requires the conformational signal to be propagated all the way to the ligand-binding site through α1/α1′-helix unbending.
What induces α1/α1′-helix unbending in the absence of ligand in the inside-out activation? The downward movement of β6-α7 loop and α7-helix coupled with the swing-out of the hybrid domain could make way and facilitate the movement of α1/α1′-helix. Mutations that favor the downward movement of the β6-α7 loop and α7-helix constitutively activated integrin for ligand binding (Luo et al., 2004b; Yang et al., 2004; Cheng et al., 2007). Glycan wedge mutations that enforce and disulfide mutations that restrain the hybrid domain swinging out activated and deactivated integrin, respectively (Luo et al., 2003; Kamata et al., 2010). In addition, mutations enforcing integrin extension but restraining the swing-out of the hybrid domain did not activate integrin (Kamata et al., 2010). A recent study showed that talin1 alone only induces integrin extension with intermediate affinity for β2 integrin, whereas both talin1 and kindlin-3 are required to induce the open headpiece, high-affinity conformation (Lefort et al., 2012). We found that the combination of active cytoplasmic mutations with the unactivatable mutations at the α1/α1′ junction greatly reduced integrin expression. Most strikingly, integrin expression was completely abolished when combining the open headpiece mutation and the α1/α1′ mutation that restrains α1/α1′-helix unbending. This data suggest that the hybrid domain swinging out without the α1/α1′-helix reshaping is energetically unfavorable or unstable. The swing-out of the hybrid domain should always be coupled with α1/α1′-helix unbending. Thus, the inside-out allosteric signal is relayed to the ligand-binding site in the order of integrin extension, hybrid domain swinging out, α7-helix and β6-α7 loop moving downward, and α1/α1′-helix unbending. Because integrin extension could facilitate ligand binding through increasing ligand accessibility and facilitate headpiece opening through disrupting the buried interface between the hybrid and EGF domains, the headpiece opening and ligand binding induced α1/α1′-helix unbending could work synergistically during integrin activation.
Our data show that the conformational change of the α1/α1′-helix is a common feature for integrins. However, the levels of changes at the α1/α1′-helix may depend on the integrin, the ligand or the activation status. For example, α4β7 integrin mediates both cell rolling and firm adhesion on its MadCAM-1 (mucosal adhesion molecule-1) ligands (Bargatze et al., 1995; Berlin et al., 1995). It was suggested that the rolling adhesion was mediated by the closed or intermediate headpiece conformation that is intermediate affinity, while the firm adhesion was mediated by the open headpiece conformation that is high affinity (Chen et al., 2004; Yu et al., 2012). Crystal soaking studies already revealed the separability of the movements of the β1-α1 loop and the α1/α1′-helix (Xiong et al., 2002; Nagae et al., 2012; Zhu et al., 2013). It is possible that the α1/α1′-helix has different levels of conformational change in response to the requirements of different levels of affinity in the different aspects of integrin function.
Integrin outside-in signaling requires the conformational signal to be transmitted to the cytoplasmic domains. Previously, we demonstrated that the separation of transmembrane domains was required for αIIbβ3-integrin-mediated outside-in signaling (Zhu et al., 2007b). Our present study addressed the structural requirement close to the ligand-binding site that is the origin of conformational transmission in the outside-in direction. We show that α1/α1′-helix unbending is required for αIIbβ3-integrin-mediated cell spreading, F-actin stress fiber and focal adhesion formation and FAK activation. Because the mutations that restrain α1/α1′-helix unbending blocked ligand-induced integrin extension, and integrin extension and the separation of the leg, transmembrane, and cytoplasmic domains are all linked (Springer and Dustin, 2012), the defect in outside-in signaling of the mutant integrins was due to the decoupled conformational transmission all the way to the cytoplasmic tails.
Extensive mutagenesis studies were performed to understand integrin allostery upon activation. Many of them are on αIIbβ3 integrin. Supplementary material Fig. S5 summarizes the mutations of αIIbβ3 integrin that either constitutively activate or inactivate it for ligand binding. Disulfide bonds introduced into the domain interfaces inactivate integrin by restraining extension, headpiece opening, or leg separation (supplementary material Fig. S5A–D) (Takagi et al., 2002; Blue et al., 2010; Kamata et al., 2010; Wang et al., 2010; Wang et al., 2011; Kamata, 2012). In contrast, mutations that disturb the domain interfaces all constitutively activate integrin by enforcing extension, tail separation or headpiece opening (supplementary material Fig. S5) (Kashiwagi et al., 1999; Ruiz et al., 2001; Butta et al., 2003; Luo et al., 2003; Luo et al., 2009; Vanhoorelbeke et al., 2009; Mor-Cohen et al., 2012). These mutagenesis studies provide compelling evidence for the requirement of long-range conformational change for integrin activation. Our novel mutations found at the α1/α1′ junction either activate or inactivate integrin by directly altering the conformation of α1/α1′-helix (supplementary material Fig. S5D). Unlike the disulfide or glycan mutations, the α1/α1′ junction mutations have minimal direct effect on the conformations of other domains. These unique mutations provide a useful tool to understand the linkage and separation between integrin ligand binding, conformational regulation and cell signaling.
Integrin βI domain has both common and different conformational regulations with its structural homolog αI domain. Both βI and αI require a MIDAS site for ligand binding (supplementary material Fig. S6A,B). The βI domain has two extra metal-ion-binding sites, SyMBS (synergetic metal ion binding site) and ADMIDAS. The presence of ADMIDAS adds an extra affinity regulation to the βI domain. The most striking difference between βI and αI is the conformational change at the α1(α1′)-helix. In the closed to open transition of the βI domain, the α1/α1′-helix changes from a bent to a straight conformation (supplementary material Fig. S6A). However, in the closed to open transition of the αI domain, there is only a swinging motion at the α1-helix N-terminus accompanied by the inward movement of the β1-α1 loop (supplementary material Fig. S6B). The αI α1-helix remains as a straight helical structure in this process. Sequence alignments of the αI α1-helices show a completely conserved Phe residue at the position equivalent to the Gly at the α1/α1′ junction of βI domains (supplementary material Fig. S6C). This Phe residue fits into the cavity between the α1- and α7-helix, and probably contributes to squeezing the α7-helix downward upon the swung-in α1-helix (supplementary material Fig. S6B). The ADMIDAS and α1/α1′-helix reshaping add more precise regulation to the βI domain than the αI domain. The structure characters of αI α1-helix are compatible with the functions of the αI-containing integrins that require fast on/off switching for affinity regulation, for example during cell rolling, rapid cell migration and immune response.
Materials and Methods
cDNAs for human αIIb or β1 and α5 or β3 integrins were cloned into pEF/V5-HisA and pcDNA3.1/Myc-His (+), respectively. cDNAs for human αL and β2 integrins were cloned into pcDNA3.1/Hygro (−) and pcDNA3.1 (+), respectively. Constructs of human α5 integrin with a C-terminal EGFP tag and EGFP-tagged mouse talin1 head domain were as described previously (Laukaitis et al., 2001; Bouaouina et al., 2008). Mutations were introduced using site-directed mutagenesis with the QuikChange kit (Agilent Technologies).
Antibodies and ligands
PAC-1 (BD Bioscience) is a ligand-mimetic mAb (IgM) that is specific for the activated αIIbβ3 integrin (Shattil et al., 1985). AP3 is a conformation-independent anti-β3 mAb (Kouns et al., 1991). 10E5 is an anti-αIIb mAb (Xiao et al., 2004). LIBS-1 (Frelinger et al., 1990) and 319.4 (unpublished data) are LIBS mAbs that binds to β3 I–EGF domains (supplementary material Fig. S1C). 370.3 (unpublished data) is a LIBS mAb that binds to the αIIb calf-1 domain (supplementary material Fig. S1C). PE-labeled MAR4 (BD Bioscience) is non-functional anti-β1 mAb. 12G10 (AbDserotec) and 9EG7 (BD Bioscience) are activation-specific anti-β1 (I–EGF2 domain for 9EG7; βI domain for 12G10) mAbs (Lenter et al., 1993; Mould et al., 2002; Askari et al., 2010) (supplementary material Fig. S1C). TS2/16 is an activating anti-β1 (βI domain) mAb (Hemler et al., 1984; Takada and Puzon, 1993). Human fibronectin type III domains 9–10 (Fn9–10) was expressed and purified as described previously (Takagi et al., 2001). Plasminogen, von Willebrand factor and fibronectin-depleted human fibrinogen was purchased from Enzyme Research Laboratories. Human ICAM-1 with a C-terminal Fc tag of human IgG1 (ICAM-1-Fc) was purchased from R&D Systems. PE- or FITC-labeled TS2/4 (BioLegend) is a non-functional anti-αL mAb. MEM-148 is an activating anti-β2 mAb (Drbal et al., 2001). Tirofiban (Merck) and eptifibatide (Millennium Pharmaceuticals) are RGD-mimetic αIIbβ3-specific antagonists.
Soluble ligand binding assay
The soluble ligand binding assay with 293FT cells transfected with αIIbβ3, α5-EGFP/β1 or αLβ2 constructs, and CHO-B2 cells transfected with α5β1 constructs were as described previously (Zhu et al., 2007a; Weitz-Schmidt et al., 2011). 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, 1 mM Ca2+ and 1 mM Mg2+ (Ca2+/Mg2+), or 0.2 mM Ca2+ and 2 mM Mn2+ (Ca2+/Mn2+). Integrin-positive cells were acquired for calculating medium fluorescence intensity (MFI). Ligand binding is presented as a percentage of ligand MFI (after subtracting the ligand MFI in EDTA) relative to integrin MFI. 293FT cells were co-transfected with α5-EGFP and β1 constructs using Lipofectamine 2000 (Life Technologies) for at least 24 hours. CHO-B2 cells were co-transfected with wild-type α5 and wild-type or mutant β1 constructs using Lipofectamine LTX (Life Technologies). For 293FT cells, cells suspended in HBSGB buffer were incubated with 50 μg/ml Alexa-Fluor-647-labeled Fn9–10 or 10 μg/ml Alexa-Fluor-647-labeled 9EG7 in the presence of 5 mM EDTA (not for 9EG7), Ca2+/Mg2+, or Ca2+/Mn2+ at room temperature for 30 minutes. Cells were then washed and analyzed using BD Accuri C6 Flow Cytometer (BD Biosciences). EGFP-positive cells were acquired for calculating MFI. Fn9-10 or 9EG7 binding was presented as a percentage of Fn9–10 or 9EG7 MFI (after subtracting the ligand MFI in EDTA for Fn9–10) relative to EGFP MFI. For CHO-B2 transfectants, cells suspended in HBSGB buffer were incubated with 50 μg/ml Alexa-Fluor-647-labeled Fn9–10 plus 10 μg/ml PE-labeled MAR4, or 10 μg/ml Alexa-Fluor-647-labeled 9EG7 plus 10 μg/ml PE-labeled MAR4, or 10 μg/ml biotinylated 12G10 in the presence of EDTA (for Fn9–10 only), Ca2+/Mg2+, Ca2+/Mg2+ plus 10 μg/ml TS2/16 (for Fn9–10 only), or Ca2+/Mn2+ at room temperature for 30 minutes. For 12G10 binding, cells were further stained with 10 μg/ml PE-labeled MAR4 plus 10 μg/ml Alexa-Fluor-647-labeled streptavidin before flow cytometry. MAR-4-positive cells were acquired for calculating MFI. Ligand or mAb binding was presented as described above.
For talin-head-induced integrin ligand or mAb binding, 293FT (for αIIbβ3 and αLβ2) or CHO-B2 (for α5β1) cells were co-transfected with integrin and EGFP or EGFP-talin1-head for 24 hours. Cells in suspension were incubated with 5 µg/ml PAC-1 (for αIIbβ3), 20 µg/ml ICAM-1-Fc plus 20 µg/ml biotin-labeled goat anti-human IgG1 (for αLβ2), 50 µg/ml Alexa-Fluor-647-labeled Fn9–10 plus 10 µg/ml PE-MAR4 (for α5β1), or 10 µg/ml Alexa-Fluor-647-labeled 9EG7 plus 10 µg/ml PE-MAR4 (for α5β1) in HBSGB plus 5 mM EDTA (not for 9EG7) or Ca2+/Mg2+ at room temperature for 30 minutes. Cells were then washed and incubated with 10 µg/ml Alexa-Fluor-647-labeled AP3 and PE-labeled goat anti-mouse IgM (Life Technologies; for αIIbβ3), or 10 µg/ml PE-TS2/4 and Alexa-Fluor-647-labeled streptavidin (for αLβ2) in HBSGB buffer plus Ca2+/Mg2+ on ice for 30 minutes before flow cytometry. EGFP and integrin double-positive cells were acquired for MFI calculation. Ligand binding is presented as a percentage of ligand MFI (after subtracting the ligand MFI in EDTA) relative to integrin MFI.
Cell adhesion and inhibition
Generation of CHO-k1 or HEK293 cells stably expressing αIIbβ3 constructs was as described before (Zhu et al., 2007b). Integrin-expressing single cells were sorted using 10E5 staining. Adhesion of CHO-k1 stable cell lines to immobilized human Fg was as described before (Zhu et al., 2007b). Cell adhesion is presented as a percentage of bound cells relative to total input cells. For cell adhesion inhibition assay, cells were first incubated with 0–50 µM eptifibatide for 30 minutes before adhering to 100 µg/ml Fg-coated 96-well plates.
LIBS epitope exposure
LIBS epitope exposure for αIIbβ3 integrin stably expressed in CHO-k1 or HEK293 cells was described previously (Zhu et al., 2007b). For talin-head-induced LIBS epitope exposure, 293FT cells were co-transfected with αIIbβ3 and EGFP or EGFP-talin1-head for 24 hours. Cells in suspension were incubated with or without 10 µg/ml biotinylated LIBS-1, 319.4 or 370.3 at room temperature for 30 minutes. Cells were then washed and incubated with Alexa-Fluor-647-labeled AP3 and PE-labeled streptavidin (Life Technologies) on ice for 30 minutes before flow cytometry. EGFP and AP3 double-positive cells were acquired for MFI calculation. LIBS exposure is presented as a percentage of LIBS mAb MFI relative to AP3 MFI.
Cell spreading assay and microscopy
The cell spreading assay and immunostaining was as described before (Zhu et al., 2007b). CHO-k1 or HEK293 cells stably expressing wild-type or mutant αIIbβ3 integrin were allowed to adhere on Delta T dishes (Bioptechs) coated with 100 µg/ml Fg or 20 µg/ml PAC-1 in DMEM without serum at 37°C for 1 hour. Integrin, F-actin, vinculin and phosphorylated FAK were stained with Alexa-Fluor-488-labeled AP3, Alexa-Fluor-546-labeled phalloidin (Life Technologies), rabbit anti-vinculin (Sigma) and rabbit anti-phospho-FAK (pY397; Millipore), respectively. For vinculin or phosphorylated FAK detection, the cells were finally stained with Alexa-Fluor-488-labeled goat anti-rabbit IgG (Life Technologies). Cells were imaged with an EVOS digital inverted fluorescence microscope with a 20× objective or an Olympus FV1000-MPE laser scanning confocal microscope with a 40×/0.8 NA water-immersion objective plus a 2.8× digital zoom. For quantification of cell spreading, the averaged cell areas (in number of pixels) of at least 50 cells within randomly selected fields were measured using ImageJ (Schneider et al., 2012) in each of three separate experiments.
We thank Drs Daniel Bougie and Richard Aster, Barry Coller, Mark Ginsberg for providing mAbs 319.4 and 370.3, 10E5 and LIBS-1, respectively; David Calderwood for providing the DNA construct of EGFP-tagged mouse talin1 head domain; Juliano Rudolph for providing CHO-B2 cells; and Peter Newman for critical reading of the manuscript.
C.Z., J.L., X.J., N.H., C.Z. and J.Z. performed the experiments; C.Z., J.L., X.J., H.S. and J.Z. analyzed the data; J.Z. designed the study, prepared the figures and wrote the manuscript.
This work was supported by a Scientist Development Grant from the American Heart Association [grant number 12SDG12070059 to J.Z.]; and an ASH Scholar Award for junior faculty from the American Society of Hematology [to J.Z.].