Structural Requirements for Activation in αIIbβ3 Integrin*

Integrins are postulated to undergo structural rearrangement from a low affinity bent conformer to a high affinity extended conformer upon activation. However, some reports have shown that a bent conformer is capable of binding a ligand, whereas another report has shown that integrin extension does not absolutely lead to activation. To clarify whether integrin affinity is indeed regulated by the so-called switchblade-like movement, we have engineered a series of mutant αIIbβ3 integrins that are constrained specifically in either a bent or an extended conformation. These mutant αIIbβ3 integrins were expressed in mammalian cells, and fibrinogen binding to these cells was examined. The bent integrins were created through the introduction of artificial disulfide bridges in the β-head/β-tail interface. Cells expressing bent integrins all failed to bind fibrinogen unless pretreated with DTT to disrupt the disulfide bridges. The extended integrins were created by introducing N-glycosylation sites in amino acid residues located close to the α-genu, where the integrin legs fold backward. Among these mutants, activation was maximized in one integrin with an N-glycosylation site located behind the α-genu. This extension-induced activation was completely blocked when the swing-out of the hybrid domain was prevented. These results suggest that the bent and extended conformers represent low affinity and high affinity conformers, respectively, and that extension-induced activation depends on the swing-out of the hybrid domain. Taken together, these results are consistent with the current hypothesis that integrin affinity is regulated by the switchblade-like movement of the integrin legs.

Integrins are postulated to undergo structural rearrangement from a low affinity bent conformer to a high affinity extended conformer upon activation. However, some reports have shown that a bent conformer is capable of binding a ligand, whereas another report has shown that integrin extension does not absolutely lead to activation. To clarify whether integrin affinity is indeed regulated by the so-called switchblade-like movement, we have engineered a series of mutant ␣IIb␤3 integrins that are constrained specifically in either a bent or an extended conformation. These mutant ␣IIb␤3 integrins were expressed in mammalian cells, and fibrinogen binding to these cells was examined. The bent integrins were created through the introduction of artificial disulfide bridges in the ␤-head/␤-tail interface. Cells expressing bent integrins all failed to bind fibrinogen unless pretreated with DTT to disrupt the disulfide bridges. The extended integrins were created by introducing N-glycosylation sites in amino acid residues located close to the ␣-genu, where the integrin legs fold backward. Among these mutants, activation was maximized in one integrin with an N-glycosylation site located behind the ␣-genu. This extension-induced activation was completely blocked when the swing-out of the hybrid domain was prevented. These results suggest that the bent and extended conformers represent low affinity and high affinity conformers, respectively, and that extension-induced activation depends on the swing-out of the hybrid domain. Taken together, these results are consistent with the current hypothesis that integrin affinity is regulated by the switchblade-like movement of the integrin legs.
Integrin-mediated bidirectional signaling is closely associated with the structural rearrangement of integrin itself. During inside-out signaling, talin has been shown to bind to the ␤ cytoplasmic tail and to disrupt the endogenous interaction between the ␣ and ␤ cytoplasmic tails (1,2). The dissociation of the two tails induces a structural rearrangement of the extracellular domains increasing the affinity to the ligand. During outside-in signaling, ligand binding in turn induces the structural rearrangement of the extracellular domains. This structural change propagates through the plasma membrane to separate the cytoplasmic tails, providing binding sites for numerous cytoplasmic proteins (3). Thus, the two structures flanking the plasma membrane affect each other. This structural rearrangement can be detected using a group of monoclonal antibodies (mAbs) that bind preferentially to the ligand-bound form (4). These anti-LIBS 2 (ligand-induced binding site) mAbs have been used not only to investigate the activation status of a specific integrin but also to activate it (5). However, without information on the actual three-dimensional structure, it is impossible to determine the specific conformation recognized by each anti-LIBS mAb.
The first observation of the actual three-dimensional structure of integrin was made using conventional electron microscopy (EM) studies of ␣IIb␤3 purified from platelets (6,7). Although this modality had a relatively low resolution, ␣IIb␤3 was shown to consist of a globular head with two short legs extending outward. A crystal structure analysis of the extracellular domains of ␣V␤3 integrin (PDB 1M1X) revealed that the ␣-chain consists of the N-terminal ␤-propeller domain followed by the thigh, calf-1, and calf-2 domains, whereas the ␤-chain consists of the plexin-semaphorin-integrin domain, ␤A domain, hybrid domain, four EGF domains, and ␤T domain (8). The ␤-propeller and ␤A domains non-covalently associate with each other to form the globular head that was observed in the EM images. In contrast, the thigh, calf-1, and calf-2 domains of the ␣-chain and the plexin-semaphorin-integrin, EGF, and ␤T domains of the ␤-chain form the two leg-like regions, respectively. Thus, the crystal structure is consistent with the conventional EM image described above. However, a striking difference in the orientation of the head was noted. In the crystal structure, the two legs were folded backward, with a 135°a ngle between the thigh and the calf-1 domains, unlike the straight legs observed using conventional EM. Consequently, the head region pointed downward, facing the plasma membrane. The discrepancies between these two structures were reconciled using high resolution EM images of the extracellular domains of recombinant ␣V␤3 integrin (9). This modality revealed that ␣V␤3 could adopt multiple distinct structures, including the bent and extended conformers observed in the crystal structure analysis and the conventional EM study, respectively. Because Mn 2ϩ and a ligand peptide significantly increased the number of extended forms, the extended form was suggested to represent a high affinity state, whereas the bent form was thought to represent a low affinity state. Thus, the transition from one conformer to another (or the so-called switchblade-like movement) might regulate the affinity of integrin to its ligand. Aside from this movement, substantial structural rearrangement has been observed in the head region (10). A crystal structure analysis of the ␣IIb␤3 head region when the molecule forms a complex with ligand mimetics revealed that the ␤-hybrid domain swings outward upon ligand binding (11). This movement is accompanied by the rearrangement of the ligand-and/or cation-binding loops in the ␤A domain, thereby regulating ligand binding (11).
However, contradictory reports suggest that integrin extension is not an essential event for ligand binding. Cryo-electron microscopic observations of ␣IIb␤3 purified from activated platelets revealed that this molecule adopts a rather compact structure, unlike the extended conformer (12). The crystal structure of ␣V␤3 complexed with a small peptide ligand revealed that the bent conformer was capable of binding a ligand (13). In this experiment, ␣V␤3 was understandably unable to undergo gross structural rearrangement upon ligand binding because of the constraints of the crystal lattice. However, a single particle analysis of recombinant ␣V␤3 complexed with a fibronectin fragment has shown that ␣V␤3 can bind macromolecular ligands while in a bent conformation in the presence of Mn 2ϩ (14,15). This evidence suggests that the bent conformer is capable of binding both small ligands and macromolecular ligands without requiring substantial structural rearrangements.
In this study, we examined the relationship between the three-dimensional structure of integrin and its ligand affinity. Our findings provide evidence that the extended conformer represents a highly activated state, whereas the bent conformer represents a low affinity state. These results are consistent with the view that the ligand binding activity of integrin can be regulated allosterically through the switchblade-like movement of the legs of integrin, centering on the genu region.
Cell Culture and Transfection-Chinese hamster ovary (CHO)-K1 cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum (HyClone, Logan, UT), 1% penicillin and streptomycin (Invitrogen), and 1% non-essential amino acids (Sigma-Aldrich) and maintained at 37°C in a humidified incubator supplemented with 5% CO 2 . Fifty micrograms of ␣IIb cDNA construct was co-transfected with 50 g of ␤3 cDNA construct into CHO-K1 cells using electroporation. After 48 h, the cells were detached and used for further experiments.
Flow Cytometry-Cells were detached with phosphate-buffered saline (PBS) containing 3.5 mM EDTA. After washing, the cells were incubated with 10 g/ml mAb in modified HEPES-Tyrode's buffer (5 mM HEPES, 5 mM glucose, 0.2 mg/ml bovine serum albumin, 1ϫ Tyrode's solution) supplemented with 1 mM CaCl 2 and 1 mM MgCl 2 for 30 min at 4°C. In some experiments, 1 mM GRGDS peptide was included together with the mAbs. After washing, the cells were incubated with the R-Phycoerythrin-conjugated F(abЈ)2 fragment of goat anti-mouse IgG for 30 min at 4°C. After washing, the cells were resuspended in HEPES-buffered saline (10 mM HEPES, 150 mM NaCl (pH 7.4)) containing 1 mM CaCl 2 and 1 mM MgCl 2 . Fluorescence was measured using a FACSCalibur (BD Biosciences). To compare the binding of conformation-dependent mAbs among cells expressing different ␣IIb␤3 mutants, each mAb binding was normalized by the expression of ␣IIb␤3 on the cell surface. This relative mAb binding was calculated by dividing the mean fluorescent intensity obtained for each mAb by the mean fluorescent intensity obtained for the non-conformation-dependent anti-␤3 mAb SZ21 or the anti-␣IIb␤3 complex-specific mAb A2A9.
Fibrinogen Binding Assay-FITC labeling of human Fbg was performed as described previously (20). Briefly, after adjusting the pH of human Fbg at 1 mg/ml in PBS to 8.5 using 5% Na 2 CO 3 , 1/100 volume of 10 mg/ml FITC in dimethyl sulfoxide (DMSO) was added and incubated at room temperature for 10 min. FITC-labeled Fbg was separated from free FITC on a PD-10 column (Amersham Biosciences, Uppsala, Sweden) equilibrated with HEPES-buffered saline. The concentration and fluorescence-to-protein ratio of FITC-labeled Fbg were calculated as described previously. The typical concentration and fluorescence-to-protein ratio were 3.4 mg/ml and 5.0 -6.0, respectively. Forty-eight hours after transfection, the cells were detached and washed once with HEPES-Tyrode's buffer. The ␣IIb␤3-transfected cells were incubated with non-functional anti-␣IIb mAb PL98DF6 followed by incubation with the ribulose-phosphate 3-epimerase-conjugated F(abЈ)2 fragment of goat anti-mouse IgG. In some experiments, cells were treated with dithiothreitol (DTT) prior to incubation with the mAbs, as described previously (19). After washing, the cells were incubated with 340 g/ml FITC-labeled Fbg with or without 1 mM GRGDS peptide in HEPES-Tyrode's buffer containing 1 mM CaCl 2 and 1 mM MgCl 2 or 1 mM MnCl 2 for 2 h at 4°C. In some experiments, the mAb PT25-2 was included at a concentration of 10 g/ml to activate ␣IIb␤3. After washing, fluorescence was measured using a FACSCalibur. The mean Fbg binding (FL1) to cell populations expressing high levels of ␣IIb (FL2 Ͼ 500) was calculated. Background binding in the presence of 1 mM GRGDS peptide was subtracted to obtain the specific binding.
Immunoprecipitation-Biotin labeling of the cell surface protein was done using Sulfo-NHS-Biotin (Thermo Scientific) following the manufacturer's instructions. Cells were lysed in 1 ml of lysis buffer (100 mM n-octylglucopyranoside, 20 mM N-ethyl maleimide, 1 mM PMSF, 25 mM Tris-HCl, and 150 mM NaCl, pH 7.4). After removing the insoluble material by centrifugation, the supernatant was used for further analysis. Two hundred microliters of cell lysate was precleared by adding 1 g of mouse IgG, together with 20 l of protein G-agarose beads. After centrifugation, the supernatant was recovered and further incubated with 1 g of PL98DF6 or VNR5-2, together with 20 l of protein G-agarose beads overnight at 4°C. Then, the supernatant was discarded, and the remaining protein G-agarose beads were washed three times with washing buffer (25 mM Tris-HCl, 150 mM NaCl, 0.01% Triton X-100 (pH 8.0)). The protein G-agarose beads were resuspended in 10 l of washing buffer. The TEV protease digestion of the immunoprecipitates was performed by adding 1 l of TurboTEV to the suspension with or without 1 mM DTT followed by incubation for 3 h at 30°C. The peptide N-glycosidase F digestion of the immunoprecipitates was done according to the manufacturer's instructions except that the DTT was excluded from the denaturation buffer. After digestion, the samples were subjected to 7.5 or 10% SDS-PAGE, transferred to a polyvinylidene difluoride membrane, probed with horseradish peroxidase-conjugated avidin, and detected using chemiluminescence with West Pico chemiluminescent substrate (Thermo Scientific).

Bent Conformer of ␣IIb␤3 Represents a Low Affinity Form-
In the ␣V␤3 crystal structure, in addition to the ␣-head and ␤-head, the ␤-head and ␤-tail domains create a large interface that keeps ␣V␤3 in a bent conformation (8,13). To constrain ␣IIb␤3 in this bent conformation, artificial disulfide bridges were introduced at different locations in the ␤-head/␤-tail interface. As shown in Fig. 1A, the amino acid residues Ser-367 and Gly-382 in the hybrid domain and Val-332 in the ␤A domain are localized close to Ser-551 in the EGF-3 domain, Thr-564 in the EGF-4 domain, and Ser-674 in the ␤T domain, respectively. If these residues are simultaneously mutated to Cys, a disulfide bride is expected to form. Thus, the resulting S367C/S551C, G382C/T564C, and V332C/S674C double mutations are expected to stabilize the hybrid/EGF-3, the hybrid/EGF-4, and the ␤A/␤T interfaces, respectively. In either case, these mutations should prevent the ␤3 chain from adopting an extended conformation. These mutants were expressed in CHO cells, and FITC-labeled Fbg binding to these cells was examined using FACS. As reported previously, the wild-type ␣IIb␤3 expressed in CHO cells is in a low affinity state and requires activation by the anti-␣IIb␤3 mAb PT25-2 to bind Fbg in the presence of 1 mM Ca 2ϩ /1 mM Mg 2ϩ (20). Cells expressing single Cys mutations, such as S367C, G382C, V332C, S551C, T564C, or S674C, bound Fbg in the presence of PT25-2, albeit slightly less than that observed in wild-type cells (data not shown). By contrast, cells expressing double Cys mutations were completely unable to bind Fbg unless they were pretreated with DTT to disrupt the disulfide bridges (Fig. 1B). These artificially introduced disulfides did not affect PT25-2 binding to ␣IIb␤3 (supplemental Fig. S1). These results suggest that the blocking effect of the double mutation is actually caused by disulfide bridge formation between the mutated residues, rather than a local effect of the mutation itself.
To examine whether these artificially introduced disulfide bridges stabilize the ␤-head/␤-tail interface, we introduced a TEV protease recognition site between the head and tail regions of the ␤3 chain (21). This mutation (475TEV) was designed to separate the N-terminal head region (amino acids 1-480) from the C-terminal tail region (amino acids 481-762) of the ␤3 chain upon TEV protease digestion. Because these two regions are connected by a disulfide bridge formed by Cys-473 and Cys-503, these residues were mutated to Ser (475TEVCS) to facilitate separation upon digestion. Then, we introduced S367C/ S551C, G382C/T564C, and V332C/S674C double mutations in 475TEVCS. These mutant ␤3 chains were expressed together with wild-type ␣IIb in CHO cells. The surface-expressed ␤3 was immunoprecipitated with anti-␤3 mAb and analyzed using SDS-PAGE. All the mutant ␤3 chains co-precipitated with ␣IIb, as did wild-type ␤3, with the exception that a slight difference in the electrophoretic mobility of the mutant ␤3 chains was noted (supplemental Fig. S2A). When digested with TEV protease under non-reducing conditions, only 475TEVCS generated a 70-kDa band in place of a 98-kDa intact ␤3 chain (supplemental Fig. S2B). However, all but the wild-type ␤3 generated 73-and 43-kDa bands in place of a 116-kDa band when digested with TEV protease under reducing conditions (supplemental Fig. S2C). The results indicate that 475TEVCS is indeed cleaved into a 73-kDa N-terminal head region and a 43-kDa C-terminal tail region by TEV protease, as expected. Because these two regions are still connected by a disulfide bridge in 475TEVCS 367/551, 475TEVCS 382/ 564, and 475TEVCS 332/674 as well as in 475TEV, the two fragments could not separate from each other under nonreducing conditions, although ␤3 was already cleaved by the TEV protease treatment. However, under reducing conditions, these mutant ␤3 molecules readily separated into two fragments. These results prove that in S367C/S551C, G382C/T564C, and V332C/S674C mutants, the disulfide bridge actually ligates the head and the tail regions of the ␤3 chain, thereby stabilizing the ␤-head/␤-tail interface.
␤A/␤T Interface Interaction Is Not Sufficient to Act as a Deadbolt to Maintain Integrin in a Low Affinity State-Xiong et al. (8) initially reported that the ␤A/␤T interface interaction might act as a deadbolt to keep integrin in a low affinity state by preventing the movement of the ␣7 helix of ␤A, which is associated with ligand binding. Indeed, stabilizing this interface with a disulfide bridge (V332C/S674C) completely inhibited ligand binding (Fig. 1B). To further examine this hypothesis, amino acid residues 671-676 of ␤3, composing most of the CD loop ( Fig. 2A) that participates in ␤A/␤T interface formation, was either deleted (del-CD) or replaced with an irrelevant FLAG tag sequence (FLAG-CD). When expressed in CHO cells, the del-CD or FLAG-CD mutant did not bind Fbg unless activated by PT25-2 in the presence of Ca 2ϩ /Mg 2ϩ . One mM Mn 2ϩ did not significantly increase Fbg binding in del-CD, as it did in wild-type cells. However, cells expressing the FLAG-CD mutant bound ϳ2.5 times as much Fbg as cells expressing wildtype ␣IIb␤3 (Fig. 2B). These results suggest that although the endogenous ␤A/CD loop interface interaction alone does not play a critical role in constraining ␣IIb␤3 in a low affinity state, alterations in its interactions might affect the activation status. The FLAG tag sequence (DYKDDDDK) consists of 8 amino acid residues when compared with the 6 residues in the wildtype sequence. This may imply that physical separation of the two domains can induce activation. To induce complete separation, we attempted to insert a bulky spacer between the ␤A and the ␤T domains. This would not only disrupt the ␤A/␤T interface interaction completely but would also affect the hybrid/␤T interface formation and slightly extend the integrin. For this purpose, we created an N-linked glycosylation site (NX(T/S) motif) at Val-332 and/or Ser-674 in the interface ( Fig.  2A). When the 332VLS 3 sequence in the ␤A was mutated to When the 674SGK sequence was mutated to 674NGT (S674N/K676T), slight but consistent Fbg binding was observed in the presence of Ca 2ϩ /Mg 2ϩ without any activators. The addition of PT25-2 or Mn 2ϩ induced more robust Fbg binding than the wild type. Combining these two mutations (V332N/S674N/K676T) had a synergistic effect on constitutive binding, although it did not further increase binding in the presence of Mn 2ϩ (Fig. 2C). These results suggest that the more the ␤A and ␤T domains are separated, the stronger the activation of ␣IIb␤3. In other words, integrin extension by itself may induce activation.
Extended Conformer of ␣IIb␤3 Represents a Highly Activated Form-The ␤A domain provides a part of the ligand-binding site. Therefore, any direct change imposed on the ␤A domain might affect ligand binding. To induce integrin extension without directly affecting the ligand-binding domains, we introduced N-linked glycosylation sites in ␣IIb amino acid residues Asp-589, Gln-595, and Thr-478, which are located in the proximity of the ␣-genu region where the integrin folds backwards in the bent conformation. These residues are all located in the thigh domain (Fig. 3A). We have previously shown that swapping the entire thigh domain between ␣IIb and ␣V did not have a significant impact on the ␣IIb␤3-Fbg interaction (22). Among these residues, Gln-595 is located immediately behind the ␣-genu. When 595QTR 4 was mutated to 595NTT (Q595N/ R597T), robust Fbg binding was observed in the presence of Ca 2ϩ /Mg 2ϩ , and the addition of PT25-2 did not significantly increase the binding. When the 589DTH sequence located distal to the ␣-genu region was mutated to 589NTT (D589N/ H591T), it induced moderate Fbg binding, and PT25-2 significantly increased the binding. In contrast, when 478TKT, which is located above the ␣-genu region, was mutated to 478NKT (T478N), it did not affect Fbg binding at all (Fig. 3B). It is possible that these mutations affect ligand binding by directly altering the local structure of the ␣IIb␤3, regardless of the actual N-glycan binding. To rule out these possibilities, the 595QTR sequence was mutated. Mutating 595QTR to ATT, DTT, WTT, NTR, QTT, or NTA did not induce significant activation. However, mutating 595QTR to NTS induced activation comparable with that induced by NTT (Fig. 4). These results suggest that the introduction of virtually any sequence other than the NX(T/S) sequence fails to induce constitutive activation. Thus, these results indicate that the activating effect of these mutations actually depends on the attachment of a bulky N-glycan to these sites.
To confirm whether N-glycans are actually attached to the intended sites in these mutants, we next compared their molecular sizes using SDS-PAGE. If an extra N-glycan is indeed attached to these mutants, their molecular size should be larger than that of the wild-type. When surface-expressed ␣IIb␤3 was labeled with biotin and immunoprecipitated with anti-␣IIb mAb, the ␤3 chain was always co-precipitated from the cells expressing wild-type or mutant ␣IIb␤3. However, the size of the mutant ␣IIb chain that carries an extra N-glycan-binding site (T478N, D589N/H591T, Q595N/R597T) was not remarkably different from that of the wild-type (supplemental Fig.  S3A). As the size of the N-glycan was relatively small when compared with the entire ␣IIb chain, it was difficult to discriminate such small differences in molecular weight using SDS-PAGE. To circumvent this problem, the ␣IIb leg region encompassing amino acid residues 450 -1008 was generated using a FLAG tag sequence on its N terminus. This fragment was surface-expressed and migrated as an 89-kDa band on SDS-PAGE when immunoprecipitated with PL98DF6. In contrast, a similar fragment carrying an extra N-glycan binding site migrated as a 91-kDa band, which is slightly larger than that of the wild type. The fragment that was not supposed to attach N-glycan (R597T) migrated as fast as the wild type (supplemental Fig.  S3B). Similar results were obtained when anti-FLAG M2 was used instead of PL98DF6 (data not shown). This difference in apparent molecular weight was completely lost when the fragments were digested with peptide N-glycosidase F as all the fragments migrated as 69-kDa bands (supplemental Fig. S3C). These results clearly indicate that the mutant ␣IIb that carries an extra NX(T/S) motif indeed binds N-glycan to these sites.
␣IIb␤3 Activation Induced by Integrin Extension Depends on the Swing-out of the Hybrid Domain-Takagi et al. (10) and Luo et al. (23) have reported that the outward swing of the hybrid domain is the most critical step in integrin activation. To examine whether this step is truly required for activation, the swingout of the hybrid domain was prevented by covalently ligating the ␤-propeller and the hybrid domains with a disulfide bridge. The amino acid residues Asp-319 of the ␣IIb ␤-propeller and Val-359 of the ␤3 hybrid domain are physically close in the closed head (swing-in) conformation, whereas they are separated in the open head (swing-out) conformation (Fig. 5A). If these residues are simultaneously mutated to Cys, a disulfide bridge will be formed between these domains, thereby fixing the angle between the ␤A and hybrid regions in the closed head conformation. The 2-3 loop in blade 5 of the ␣IIb propeller, where Asp-319 is located, was previously shown not to partic-   DECEMBER 3, 2010 • VOLUME 285 • NUMBER 49 ipate in ligand binding (24). As shown in Fig. 5B, single D319C or V359C mutation did not significantly affect Fbg binding. However, D319C/V359C double mutation completely abolished the Fbg binding induced by PT25-2 unless the cells were pretreated with DTT. The binding of PT25-2 was unaffected by disulfide formation (supplemental Fig. S1). Next, we examined the binding of an activation-independent ligand-mimetic mAb, OP-G2. OP-G2 has an RGD-related RYD sequence in the CDR3 and binds ␣IIb␤3 in almost the same fashion as Fbg, although it does not require integrin activation for binding (25). Unlike the V332C/S674C mutation, which keeps integrin in a bent conformation, the D319C/V359C mutation did not affect OP-G2 binding (Fig. 6, A and B). To examine the effect on the conformational change induced by ligand binding, the binding of anti-LIBS mAb was examined. The binding of anti-LIBS2 and anti-LIBS6 increased significantly in the presence of RGD peptide in cells expressing wild-type ␣IIb␤3 as well as in cells expressing the single Cys mutation V332C or S674C. However, cells expressing V332C/S674C bound significantly less anti-LIBS mAb than cells expressing wild-type ␣IIb␤3, and these cells did not respond to RGD peptide (Fig. 6, C and E). In contrast, cells expressing D319C/V359C showed a basal binding comparable with that of the wild type or single Cys mutants, although the response to RGD peptide was slightly attenuated (Fig. 6, D and  F). These results indicate that the swing-out of the hybrid domain is only required for high affinity ligand binding and that the prevention of the swing-out does not completely inhibit the conformational change associated with ligand binding. To examine whether the swing-out of the hybrid domain is required for the activation induced by integrin extension, we combined Q595N/R597T with D319C/V359C and examined its effect on Fbg binding. The resulting 595-319/359 mutant was expected to adopt an extended with a closed head conformation. The activating effect of the Q595N/R597T mutation was completely suppressed by the D319C/V359C mutation. PT25-2 was ineffective, unless the cells were pretreated with DTT (Fig. 7). These results suggest that integrin extension must be accompanied by the swing-out of the hybrid domain for it to induce activation.

DISCUSSION
By characterizing the recombinant ␣IIb␤3 integrin expressed in CHO cells, we established that ␣IIb␤3 constrained in its bent conformation represents a low affinity form, whereas ␣IIb␤3 constrained in its extended conformation represents a high affinity form. This constitutive activation depends on the swing-out of the hybrid domain because the prevention of this swing-out completely inhibited ligand binding regardless of the bent/extended state.
Integrin domains make large interdomain interfaces between the ␣-head and the ␤-head, the ␣-tail and the ␤-tail, and the ␤-head and the ␤-tail. Among these interfaces, the ␤-head/ ␤-tail interface is presumed to play a central role in keeping integrin in its bent conformation because this is the only interaction that directly connects the head with the tail region in the bent conformer but not in the extended conformer. This interface is maintained by multiple interdomain interactions. The ␤A and the hybrid domains in the ␤-head make contact with the C-terminal ␤T domain. The hybrid domain also makes con-  tact with the EGF-3 and EGF-4 domains. We attempted to stabilize this interface by introducing artificial disulfide bridges between these domains. As previously reported, stabilizing the ␤A/␤T interface (V332C/S674C) completely blocked Fbg binding (9). Likewise, stabilizing the hybrid/EGF-3 (S367C/ S551C) or hybrid/EGF-4 (G382C/T564C) interface completely abolished Fbg binding. Regardless of the positions of the disulfide bridges that were introduced, stabilizing these interfaces prevented integrin from adopting the extended conformation. However, the S367C/S551C and G382C/T564C mutations not only prevented integrin extension but also restricted the relative movements of the hybrid and ␤-tail domains. For this reason, it might be premature to conclude that integrin extension is essential for activation. However, the fact that ligating the ␣-head with the ␤-tail or limiting ␣IIb extension using intrachain disulfide bridges that do not directly restrict hybrid/␤-tail movement also prevented activation (9,26) suggests that the completely bent conformer observed in the crystal structure represents the low affinity form rather than the high affinity form. These results also indicate that the ␤-head/␤-tail interface must be disrupted all the way up to the linker region for the integrin to be activated.
The fact that the ligand-mimetic non-activation-dependent mAb OP-G2 did not bind to cells expressing the V332C/S674C mutant suggests that this mutant does not support low affinity ligand binding. In addition, anti-LIBS mAb binding to this particular mutant indicates that V332C/S674C is unable to undergo structural rearrangement in the presence of the RGD peptide. Taken together, these results suggest that the V332C/ S674C mutant is not capable of binding either macromolecular ligands or ligands as small as the RGD peptide. Because the V332C/S674C mutation ligates the ␣7 helix in ␤A with the CD loop in ␤T, the possible downward movement of the ␣7 helix required for ligand binding in the integrin A domains would be inhibited. Thus, the effect of the V332C/S674C mutation is a combination of both ␤-head/␤-tail stabilization and the inhibition of ␣7 helix movement. In agreement with these findings, the S367C/S551C mutation, which does not restrict ␣7 helix movement, only partially blocked OP-G2 binding (data not shown).
The contribution of each interface interaction in maintaining integrin in the bent conformation has not been clarified. It has been reported that replacing the ␤2 CD loop sequence with the homologous ␤3 sequence or inserting a N-glycan-binding site in the CD loop in ␣M␤2 integrin induced robust ligand binding (27). However, the fact that the deletion of the CD loop of the ␤T domain failed to activate ␣IIb␤3 in our experiment strongly argues against the deadbolt theory, in which an endogenous ␤A/␤T interface interaction plays a critical role in maintaining integrin in its low affinity state. In addition, the insertion of a bulky N-glycan at the interface only slightly activated ␣IIb␤3 in the presence of Ca 2ϩ /Mg 2ϩ . These results suggest that hybrid/EGF-3, hybrid/EGF-4, and hybrid/␤T interface interactions, rather than the ␤A/␤T interface alone, play important roles in maintaining integrin in a low affinity state. In agreement with these conclusions, a computer-assisted approach has identified key interactions in the hybrid/␤-tail interface in ␤3 integrin (28). Although the disruption of the hybrid/␤T or the hybrid/EGF-3 interaction alone only produced weak activation, disrupting multiple interactions at the same time induced significant activation. Thus, the hybrid/␤tail interface seems to be maintained by a group of several key interactions that individually are not sufficiently strong to do so. The apparently distinct role of the ␤A/␤T interface interaction in regulating activation in ␤2 and ␤3 integrins suggests that the contributions of each interdomain interaction to maintaining integrin in a low affinity state may differ among different integrin subfamilies.
A high resolution electron microscopic analysis of recombinant ␣V␤3 extracellular domains revealed that ligand-bound ␣V␤3 preferentially adopts an extended, rather than a bent, conformation (9). Based on these observations, it has been tentatively concluded that the extended conformer represents a high affinity form. The current study provides direct evidence that the highly extended conformer indeed has a higher ligand affinity than the completely bent conformer. Among the three mutants that showed constitutive Fbg binding in this study, ␤3S674N/K676T showed the lowest, ␣IIbQ595N/R597T showed the highest, and ␣IIbD589N/H591T showed an intermediate binding affinity. These results suggest that the degree of extension may be correlated with the extent of activation. These results also indicate that integrins are capable of assuming a wide range of affinity states depending on the degree of extension. Recently, it has been shown that integrin extension may not necessarily be accompanied by activation based on discrepancies between the expression of an extension-reporting epitope for KIM127 and an activation-reporting epitope for mAb24 on ␣L␤2 under flow conditions (29). It is possible that ␣A domain-containing integrin may require an additional step to achieve activation, unlike integrins without ␣A domains. It would be interesting to examine whether the introduction of a neoglycan that induces ␣L extension activates ␣L␤2.
Our results apparently contradict a report that ␣V␤3 is capable of binding fibronectin while in a bent conformation (14). However, it is not possible to tell to what degree integrin must extend to enable substantial ligand binding based on our experiments. The fact that ␣IIb␤3 can exist in a wide range of affinity states depending on the degree of the extension implies that as long as it is not completely bent, ligand binding could be observed to a varying extent. In other words, relaxation of the ␤-head/␤-tail interface interaction, but not complete extension, may be sufficient for ligand binding to occur, especially in the presence of Mn 2ϩ . As shown in Fig. 3B, Mn 2ϩ seems to lessen the requirement for integrin extension for Fbg binding. Mn 2ϩ activation alone has consistently been reported not to be accompanied by integrin extension (30). A recent report by Blue et al. (26) has provided a plausible explanation for the discrepancies in ligand binding observed under different cation conditions. Limiting ␣IIb extension using intrachain disulfides did not block Fbg binding in the presence of Mn 2ϩ , although binding was blocked in the presence of Ca 2ϩ /Mg 2ϩ . In contrast, limiting ␤-head/␤-tail movement using S367C/S551C, G382C/ T564C, and V332C/S674C double mutations blocked Fbg binding significantly in the presence of Mn 2ϩ as well as in the presence of Ca 2ϩ /Mg 2ϩ (data not shown). Taken together, these results may imply that it is not integrin extension per se, but Structural Requirements for Integrin Activation DECEMBER 3, 2010 • VOLUME 285 • NUMBER 49 the relative ␤-head/␤-tail movement (e.g. the swing-out of the hybrid domain), that is essential for activation in the presence of Mn 2ϩ . These results may explain why ligand binding was observed for the bent conformer in some studies in which Mn 2ϩ was utilized to induce ligand binding (13,14). Further study is required to determine the differences in the structural requirements for activation under different cation conditions. Springer and co-workers (10,23) have shown that ligand binding induces swing-out of the hybrid domain and that this change induces strong activation by itself, regardless of the bent/extend conformation. Our results show that the swingout of the hybrid domain is essential for activation and that extension-induced activation absolutely depends on this change. These results indicate that the affinity state of the extended conformer is controlled by the swing-out of the hybrid domain and that to down-regulate activation, integrin does not necessarily need to go back to its original bent conformation but that this can rather be accomplished by the swing-in of the hybrid domain. Interestingly, constraining the integrin head in a closed state did not prevent OP-G2 binding at all (Fig.  6B). Unlike PAC-1, which binds ␣IIb␤3 in an activation-dependent fashion, OP-G2 is less dependent on integrin activation (18). This difference indicates that the swing-out of the hybrid domain is required only for high affinity ligand binding but not for low affinity ligand binding. However, we are not sure at this point whether the swing-out of the hybrid domain alone is sufficient for high affinity ligand binding, as reported by Springer and co-workers (23). Our experiments using recombinant ␣IIb␤3 expressed on the CHO cell surface have shown that the swing-out of the hybrid domain only induced moderate activation. To induce full activation, integrin extension was required. 5 It is possible that the proximity of the integrin head domains to the plasma membrane in the bent conformation may limit the access of macromolecular ligands. Experiments utilizing cell-free binding studies should help to clarify these discrepancies. Interestingly, anti-LIBS mAb binding was still observed in the closed head mutant (D319C/V359C) in the presence of RGD peptide. Because ligand binding induces the outward swing of the hybrid domain and this movement would probably disrupt the ␤-head/␤-tail interface, it is reasonable to assume that this swing-out triggers the structural transition from a bent to an extended conformation in outside-in signaling. However, our result suggests the possibility that a structural change in addition to the swing-out may trigger the conformational change upon ligand binding. A recent report also suggests that integrin affinity may be regulated independently from the swing-out of the hybrid domain based on the expression of anti-LIBS epitope located in the hybrid domain (31). Because we do not know the specific conformation to which each of the anti-LIBS mAbs binds, further analysis is needed to address this issue.
Then, what triggers the structural transition from the bent to the extended conformation during inside-out signaling? Numerous studies have suggested the importance of integrin cytoplasmic tails in regulating integrin activation. It has been shown that integrin cytoplasmic tails undergo structural rearrangement upon ligand binding (32). It was subsequently shown that the two cytoplasmic tails separate from each other upon ligand binding (3). On the other hand, the deletion of the entire ␣ or ␤ cytoplasmic tail at the membrane-proximal sites induced significant activation (33). NMR studies on recombinant ␣IIb␤3 cytoplasmic tails have shown that talin binding to the ␤3 cytoplasmic tail disrupts the endogenous interaction between the ␣ and ␤ cytoplasmic tails (2). Because talin binding to the ␤3 cytoplasmic tail activates ␣IIb␤3, it was concluded that the separation of the two cytoplasmic tails somehow induces structural rearrangement of the integrin extracellular domains (1). We have previously shown that stabilizing the ␣-tail/␤-tail interface with artificial disulfide bridges completely abolished the activation induced by cytoplasmic tail deletion (22). Based on these observations, we hypothesized that the separation of the two extracellular tails following the cytoplasmic tail dissociation induces structural rearrangement from the bent to the extended conformation. Indeed, the separation of the ␣-tail/␤-tail interface induced robust activation. 6 The ␣-tail/␤-tail interface and the ␤-head/␤-tail interface are located next to each other, flanking the ␤-tail. Because the ␤-head/␤-tail interface, and not the ␣-tail/␤-tail interface, maintains integrin in its bent conformation, it is reasonable to assume that the separation of one interface destabilizes the other. Further elucidation of the role of these interface interactions in integrin affinity regulation will facilitate understanding of integrin-mediated bidirectional signaling.