Structure and function of the platelet integrin α_IIbβ₃

Abstract

The platelet integrin α_IIbβ₃ is required for platelet aggregation. Like other integrins, α_IIbβ₃ resides on cell surfaces in an equilibrium between inactive and active conformations. Recent experiments suggest that the shift between these conformations involves a global reorganization of the α_IIbβ₃ molecule and disruption of constraints imposed by the heteromeric association of the α_IIb and β₃ transmembrane and cytoplasmic domains. The biochemical, biophysical, and ultrastructural results that support this conclusion are discussed in this Review.

Integrins are ubiquitous transmembrane α/β heterodimers that mediate diverse processes requiring cell-matrix and cell-cell interactions such as tissue migration during embryogenesis, cellular adhesion, cancer metastases, and lymphocyte helper and killer cell functions (1). Eighteen integrin α subunits and 8 integrin β subunits have been identified in mammals that combine to form 24 different heterodimers. The resulting heterodimers can then be grouped into subfamilies according to the identity of their β subunit (1). Platelets express 3 members of the β₁ subfamily (α_IIβ₁, α_vβ₁, and α_vIβ₁) that support platelet adhesion to the ECM proteins collagen, fibronectin, and laminin, respectively (2–5), and both members of the β₃ subfamily (α_vβ₃ and α_IIbβ₃). Although α_vβ₃ mediates platelet adhesion to osteopontin and vitronectin in vitro (6, 7), it is uncertain whether it plays a role in platelet function in vivo. By contrast, α_IIbβ₃, a receptor for fibrinogen, vWF, fibronectin, and vitronectin, is absolutely required for platelet aggregation. Consequently, inherited abnormalities in α_IIbβ₃ number or function preclude platelet aggregation, resulting in the bleeding disorder Glanzmann thrombasthenia (8). Conversely, thrombi that arise in the arterial circulation result from the α_IIbβ₃-mediated formation of platelet aggregates (9). Because α_IIbβ₃ plays an indispensable role in hemostasis and thrombosis, it is among the most intensively studied integrins. Thus, there is a wealth of new information relating α_IIbβ₃ structure and function, the subject of this Review.

Expression of α_IIbβ₃ is restricted to cells of the megakaryocyte lineage. In megakaryocytes, α_IIbβ₃ is assembled from α_IIb and β₃ precursors in the endoplasmic reticulum (10) and undergoes posttranslational processing in the Golgi complex, where α_IIb is cleaved into heavy and light chains (11). There are approximately 80,000 copies of α_IIbβ₃ on the surface of unstimulated platelets (12), and additional heterodimers in the membranes of platelet granules are translocated to the platelet surface during platelet secretion (13). A critical feature of α_IIbβ₃ function is that it is modulated by platelet agonists. Thus, while α_IIbβ₃ can support the adhesion of unstimulated platelets to many of its ligands when they are immobilized in vitro, platelet stimulation is required to enable α_IIbβ₃ to mediate platelet aggregation by binding soluble fibrinogen and vWF (14). EM images of rotary-shadowed α_IIbβ₃ reveal that the heterodimer consists of an 8-by-12-nm nodular head containing its ligand-binding site and two 18-nm flexible stalks containing its transmembrane (TM) and cytoplasmic domains (15).

Crystal structures for the extracellular portions of αvβ3 and αIIbβ3

A major advance in understanding the structure and function of α_IIbβ₃ resulted from the reports of crystal structures for the extracellular portions of α_IIbβ₃ (16) and the closely related integrin α_vβ₃ (17). Xiong and coworkers prepared crystals of a presumably activated conformation of the α_vβ₃ extracellular region grown in the presence of Ca²⁺ (17). Surprisingly, the crystals revealed that the head region was severely bent over 2 nearly parallel tails (Figure 1). When the structure was extended, its appearance and dimensions were consistent with rotary-shadowed EM images of α_IIbβ₃. The structure itself revealed that the amino terminus of α_v was folded into a β-propeller configuration, followed by a “thigh” and 2 “calf” domains, constituting the extracellular portion of the α_v stalk. The α_v “knee” or “genu,” the site at which the head region bends, was located between the thigh and first calf domain. The β₃ head consists of a βA domain whose fold resembles that of integrin α subunit “I-domains” and contains a metal ion–dependent adhesion site (MIDAS) motif, as well as a hybrid domain whose fold is similar to that of I-set Ig domains. The interface between the α_v β-propeller and the β₃ βA domain, the site at which the α_v head interacts with the β₃ head, resembles the interface between the Gα and Gβ subunits of G proteins. The β₃ stalk consists of a PSI (plexin, semaphorin, integrin) domain, 4 tandem EGF repeats, and a unique carboxyterminal βTD domain. A cyclic Arg-Gly-Asp–containing (RGD-containing) pentapeptide, soaked into the crystal in the presence of Mn²⁺ (18), inserted into a crevice between the β-propeller and βA domains with the Arg side chain located in a groove on the upper surface of the propeller and the Asp carboxylate protruding into a cleft between loops on the βA surface, implying that the crevice constitutes at least a portion of the binding site for RGD-containing α_vβ₃ ligands.

Figure 1.

Ribbon diagram of the structure of the extracellular portion of α_vβ₃. (A) Bent conformation of α_vβ₃ as it was present in the crystal. (B) Extension of the structure to reveal its domains. Adapted with permission from Annual Review of Cell and Developmental Biology (97).

Subsequently, Xiao et al. reported 2 crystal structures of a complex consisting of the α_IIb β-propeller and the β₃ βA, hybrid, and PSI domains (16). The structures revealed an open, presumably high-affinity conformation, similar to EM images of the α_vβ₃ extracellular domain–containing ligand, with a 62° angle of separation between the α and β subunits due in part to a 10-A downward movement of the α7 helix of the βA domain relative to the hybrid and the PSI domain. Reorganization of hydrogen bonds in the interface between the α7 helix and βC strand of the hybrid domain allowed the hybrid domain and the rigidly connected PSI domain to swing out, causing a 70-Å separation of the α_IIb and β₃ stalks at their “knees,” a feature noted in EM images of active forms of α_IIbβ₃ in the presence or absence of ligand (19).

Ligand binding to αIIbβ3

Fibrinogen, the major α_IIbβ₃ ligand, is composed of pairs of Aα, Bβ, and γ chains folded into 3 nodular domains. Although peptides corresponding to either the carboxyterminal 10–15 amino acids of the γ chain (20) or the 2 α chain RGD motifs inhibit fibrinogen binding to α_IIbβ₃ (21), only the γ chain sequence is required for fibrinogen binding to α_IIbβ₃ (22). Nonetheless, RGD-based peptides and peptidomimetics inhibit α_IIbβ₃ function in vitro and are clinically effective antagonists of α_IIbβ₃ function in vivo (23). The structural basis for these observations is not entirely clear, but competitive binding measurements indicate that γ chain and RGD peptides cannot bind to α_IIbβ₃ at the same time (24), implying that RGD peptides inhibit fibrinogen binding by preventing the interaction of the γ chain with α_IIbβ₃.

Ligand binding to α_IIbβ₃ involves specific regions of the aminoterminal portions of both α_IIb and β₃. In the crystal structure of the α_IIbβ₃ head domain, ligand binds to a “specificity-determining” loop in the β₃ βA domain and to a “cap” composed of 4 loops on the upper surface of the α_IIb β-propeller domain (16). The α_IIb β-propeller results from the folding of 7 contiguous aminoterminal repeats (17, 25). Each blade of the propeller is formed from 4 antiparallel β strands located in each repeat; loops connecting the strands are located on either the upper or the lower surface of the propeller. A number of naturally occurring and laboratory-induced mutations distributed between α_IIb residues 145 and 224 and located in loops on the upper surface of the propeller impair α_IIbβ₃ function, implying that these residues interact with ligand (26–28). Further, Kamata et al. replaced each of the 27 loops in the α_IIb propeller with the corresponding loops from α4 or α5 (29). They found that 8 replacements, all located on the upper surface of the second, third, and fourth repeats, abrogated fibrinogen binding to α_IIbβ₃, suggesting that fibrinogen binds to the upper surface of the propeller in a region centered around the third repeat. Previous chemical cross-linking experiments suggested that the fibrinogen γ chain binds to α_IIb in the vicinity of its second calmodulin-like motif near amino acids 294–314 (30), but these residues are located on the lower surface of the propeller and are unlikely to interact with ligands such as fibrinogen (16). It is noteworthy that ligand binding itself induces conformational changes in α_IIbβ₃, most often detected by the appearance of neoepitopes for mAbs. In fact, such ligand-induced changes or LIBSs (ligand-induced binding sites) may be responsible for the immune-mediated thrombocytopenia associated with the clinical use of α_IIbβ₃ antagonists (31).

Ligand binding to α_IIbβ₃ requires divalent cations (32). Eight divalent cation-binding sites were identified in the α_vβ₃ crystal structure (17, 18). Four were located in the α_v β-propeller domain, 1 at the α_v genu, and 3 in the β₃ βA domain, but only those located in the βA domain appeared to participate in ligand binding. In the absence of ligand, only the βA ADMIDAS (adjacent to the metal Ion–dependent adhesion site) motif was occupied, but when Mn²⁺ and a cyclic RGD ligand were present, each of the βA sites contained a cation. One site was the βA MIDAS; Mn²⁺ present at this site was in direct contact with ligand. A second Mn²⁺, located 6 Å from the MIDAS, was bound to a site designated ligand-induced metal-binding site (LIMBS), but the cation at this site did not interact with ligand. It had been postulated that Mn²⁺ induces integrin activation by antagonizing inhibitory effects of Ca²⁺ (33), but the α_vβ₃ crystal structure suggests that cations bound to the MIDAS and LIMBS motifs act by stabilizing the ligand-occupied conformation of the βA domain (18).

Regulation of αIIbβ3 ligand-binding activity

Integrins reside on cell surfaces in an equilibrium between inactive and active conformations (34). In experiments where the cytoplasmic domains of α_Lβ₂ and α5β₁ were replaced by acidic and basic peptides (35, 36), purified integrins were inactive when their stalks were in proximity and active when the stalks were farther apart. This was corroborated by measurements of fluorescence resonance energy transfer (FRET) efficiency between cyan and yellow fluorescent proteins fused to the cytoplasmic domains of α_L and β₂ expressed in K562 cells (37). FRET efficiency decreased when α_Lβ₂ interacted with immobilized or soluble ligand, implying that bidirectional signaling resulted from the coupling of conformational changes in the α_Lβ₂ extracellular domain to the spatial separation of the α_L and β₂ cytoplasmic domains, a result consistent with EM images of α_IIbβ₃ in which scissor-like movements of the α_IIb and β₃ stalks differentiate active and inactive molecules (19).

Nonetheless, the relationship of these observations to the α_IIbβ₃ and α_vβ₃ crystal structures is controversial. Takagi et al., supported by negatively stained EM images of active and inactive integrins, suggested that the bent conformation of α_vβ₃ in crystals corresponds to low-affinity α_vβ₃ and the shift to a high-affinity conformation occurs when the integrin undergoes a global reorganization characterized by a “switchblade-like” opening to an extended structure and scissor-like separation of the α and β subunit stalks (34). Xiong et al., however, suggested that the bent conformation resulted from flexibility at the α_v and β₃ genua and from crystal contacts not likely to occur in nature (17). This possibility was supported by cryo-EM reconstructions of intact inactive α_IIbβ₃ molecules, which revealed a collapsed but unbent structure consisting of a large globular head and an L-shaped stalk whose axis was rotated approximately 60° with respect to the head and was connected at an angle of approximately 90° to a rod containing the TM domains of the integrin (Figure 2A) (38). They also suggested that extension at the “knees” may be a post-ligand-binding “outside-in” signaling event and that the transition of α_vβ₃ from its inactive to its active conformation results when the CD loop of the β₃ βTD domain moves away from the βA domain, allowing the latter to assume its active conformation (39). How to reconcile each of these models with the rotary-shadowed EM images of demonstrably inactive and active α_IIbβ₃ shown in Figure 2, B and C, is not obvious.

Figure 2.

Cryo-EM reconstruction and rotary-shadowed EM images of α_IIbβ₃. (A) Cryo-EM reconstruction. The resolution is 20 Å. Adapted with permission from Proceedings of the National Academy of Sciences of the United States of America (38). (B and C) Rotary-shadowed EM images. The images in B were obtained in the presence of 1 mM Ca²⁺ and the images in C in the presence of 1 mM Mn²⁺. Reproduced with permission from Blood (19).

The α_IIb and β₃ cytoplasmic domains constrain α_IIbβ₃ function

Cytoplasmic domain sequences, most convincingly demonstrated for conserved membrane-proximal sequences, constrain integrins in their low-affinity (inactive) conformations. Thus, truncation of the α_IIb cytoplasmic domain at Gly991 or the β₃ cytoplasmic domain at Leu717 or deletion of the conserved membrane-proximal α_IIb GFFKR or β₃ LLITIHD motifs (Table 1) shifts α_IIbβ₃ to its active state (40). Similarly, constitutive α_IIbβ₃ function can be induced by replacement of α_IIb residue F992, F993, or R995 or β₃ residue D723 with alanine, whereas heterodimers containing simultaneous R995→D and D723→R substitutions are inactive (41). These observations led to the suggestion that the membrane-proximal sequences form an activation-constraining “clasp,” an essential feature of which is a salt bridge between α_IIb R995 and β₃ D723. Paradoxically, replacing the α_IIb cytoplasmic domain with the cytoplasmic domain of α₂, α₅, α_6A, or α_6B, each of which contains a GFFKR motif, activates α_IIbβ₃ (40). This implies that additional cytoplasmic domain sequences modulate α_IIbβ₃ function, consistent with the inhibitory effects observed for the β₃ mutation Ser752Pro (42), β₃ truncation at Arg724 (43), and mutations involving the β₃ sequences EFAKFEEE, NPLY, and NITY (44–46) and the α_IIb sequence Pro998/Pro999 (47, 48).

Table 1.

Amino acid sequences of the TM and cytoplasmic domains of α_IIb and β₃^A

Interaction between the α_IIb and β₃ cytoplasmic domains has been studied experimentally using peptides dissolved in aqueous buffer or anchored to phospholipid micelles via aminoterminal myristoylation. Using terbium luminescence and electrospray ionization mass spectroscopy, Haas and Plow observed the formation of a cation-containing complex involving α_IIb residues 999–1,008 and β₃ residues 721–740 (49). Similarly, Vallar et al. used surface plasmon resonance to detect a weak (K_d ∼50 μM) KVGFFKR-dependent, calcium-stabilized complex between soluble α_IIb cytoplasmic domain and immobilized β₃ cytoplasmic domain peptides (50). Further, Weljie et al. determined an NMR structure for a heterodimer that formed at low ionic strength between an 11-residue GFFKR-containing α_IIb peptide and a 25-residue LLITIHD-containing β₃ peptide (51). They identified 2 conformers differing in the conformation of the β₃ backbone: one had an elongated β₃ structure; the other was bent back at D723–A728, causing the peptide to adopt a closed L shape. Nonetheless, both conformers were predominantly helical with significant hydrophobic interactions between V990 and F993 of α_IIb and L717–I721 of β₃. Although there was no NMR evidence of an R995–D723 salt bridge, modeling suggested that a salt bridge was possible if the β₃ backbone was elongated. Vinogradova et al. also used NMR to characterize complexes between full-length α_IIbβ₃ cytoplasmic domain peptides (48, 52, 53). Despite low affinity, they identified interfaces for the complexes that included hydrophobic and electrostatic interactions between membrane-proximal helices (Figure 3A) (52). When the experiments were repeated in the presence of diphosphocholine micelles, α_IIb residues 989–993 and β₃ residues 716–721 were embedded in lipid and there was interaction between β₃ residues 741 and 747 and micelle lipid (53). Talin binding to β₃ disrupted the complex of α_IIb with β₃ as well as β₃ interaction with lipid (Figure 3B). On the other hand, Li et al. were unable to detect heteromeric interaction between proteins corresponding to the α_IIb and β₃ TM and cytoplasmic domains in diphosphocholine micelles at physiologic salt concentrations using a number of biophysical techniques, perhaps because heteromeric interaction is substantially weaker than homomeric interaction (54). Similarly, Ulmer et al. did not detect heteromeric interactions of α_IIb with β₃ in an NMR analysis of a coiled-coil construct containing the α_IIb and β₃ cytoplasmic domains (55).

Figure 3.

Interaction of the α_IIb and β₃ cytoplasmic domains. (A) Backbone ribbon diagram of the α_IIbβ₃ membrane-proximal cytoplasmic domain clasp showing hydrophobic and electrostatic interactions. Reproduced with permission from Cell (52). (B) Model of the changes that may occur in the clasp following talin binding to the β₃ cytoplasmic domain. Adapted with permission from Proceedings of the National Academy of Sciences of the United States of America (53).

Proteins that interact with the αIIb and β3 cytoplasmic domains

Proteins have been identified, most often using yeast 2-hybrid screens, that bind to the cytoplasmic domains of integrin α and β subunits. These proteins include CIB (calcium- and integrin-binding protein) (56), Aup1 (ancient ubiquitous protein 1) (57), ICln (a chloride channel regulatory protein) (58), and PP1c (the catalytic subunit of protein phosphatase 1) (59), each of which binds to the membrane-proximal α_IIb sequence KVGFFKR. However, because a substantial portion of this sequence is likely embedded in the plasma membrane (60, 61), the physiologic importance of these interactions is uncertain. Proteins that interact with the β₃ cytoplasmic domain include the cytoskeletal proteins talin, α-actinin, filamin, myosin, and skelemin; various members of the Src family of kinases; the kinases integrin-linked kinase (ILK), Syk, and Shc; the adaptor Grb2; the scaffold RACK1; CD98 (62); and β₃-endonexin (63). Binding of myosin, Shc, and Grb2 requires platelet aggregation and spreading, as well as the Fyn-mediated phosphorylation of β₃ tyrosines 747 and 759, and has been implicated in post-receptor-binding cytoskeleton-mediated events such as clot retraction (64).

Binding of β₃-endonexin or talin to the β₃ cytoplasmic domain is noteworthy because it can activate α_IIbβ₃. β₃-Endonexin, a 14-kDa protein of unknown function, induces α_IIbβ₃ activation when coexpressed with α_IIbβ₃ in tissue culture cells by interacting with residues located in both the aminoterminal and the carboxyterminal regions of the β₃ cytoplasmic domain, in particular the carboxyterminal NITY motif (65–67). Nonetheless, there is no evidence as yet that β₃-endonexin regulates α_IIbβ₃ function in platelets. One explanation for the presence of 2 discontinuous β₃-endonexin–binding sites in the β₃ cytoplasmic domain has been provided by an NMR analysis of a protein encompassing the β₃ TM and cytoplasmic domains (Figure 4) (68). This analysis revealed that the β₃ TM helix extended into the membrane-proximal region of the cytoplasmic domain, ending at an apparent hinge at residues H722–D723 (Table 1). Two additional helical stretches, extending from residues 725 to 735 and 748 to 755, were also present (Figure 4). Because the latter helices can interact with each other, they can place the proximal and distal regions of the β₃ cytoplasmic domain in proximity.

Figure 4.

Model of the structure of the β₃ TM and cytoplasmic domains. Helices are shown as cylinders. Three different orientations of the β₃ TM domain in the plasma membrane are shown. The membrane-proximal region of the cytoplasmic domain is shaded. Arrows indicate possible interactions between helices. Adapted with permission from Biochemistry (68).

Talin, an abundant 250-kDa cytoskeletal protein, forms antiparallel homodimers that bind to the cytoplasmic domain of integrin β subunits as well as to other cytoskeletal proteins such as actin and vinculin (69). Talin is composed of a 50-kDa head domain containing its principal integrin-binding site and a 220-kDa rod domain that binds to integrins with lesser affinity (69). The talin head itself contains an approximately 300-residue FERM (four-point-one, ezrin, radixin, moesin) domain that folds into F1, F2, and F3 subdomains (69). F2 and F3 bind to the β₃ cytoplasmic domain, although the affinity of F3 binding is substantially greater (70). A crystal structure for a fusion protein composed of the F2 and F3 subdomains and a contiguous aminoterminal peptide corresponding to the midportion of the β₃ cytoplasmic domain, including its NPLY motif, revealed that the interaction of the β₃ peptide with F3 was mainly hydrophobic and that NPLY interacted with F3 in a manner that resembled that of canonical PTB domain ligands (71). However, studies using NMR also revealed that F3 and F2-F3 interact with the membrane-proximal region of the β₃ cytoplasmic domain (71, 72), consistent with previous observations that talin binds to peptides corresponding to this portion of β₃ (73).

Overexpressing the talin head domain in α_IIbβ₃-expressing CHO cells induces α_IIbβ₃ activation (74), either directly because talin disrupts the clasp between α_IIb and β₃ (Figure 3B) or indirectly via conformational changes induced by F3 binding to the β₃ NPLY motif (70). Conversely, reducing talin expression using short hairpin RNAs decreases ligand binding to α_IIbβ₃ in CHO cells and in ES cell–derived agonist-stimulated megakaryocytes (75). Taken together, these results imply that talin binding to the β₃ cytoplasmic domain may be a final step in α_IIbβ₃ activation. Nonetheless, how talin binding to the β₃ cytoplasmic domain is regulated remains to be determined. The integrin-binding domain in intact talin appears to be masked (76). Although the enzyme calpain can cleave talin, releasing its head domain (77), calpain activation in platelets is a relatively late step after platelet stimulation (78) and would be unlikely to contribute to integrin-activating inside-out signaling. On the other hand, talin binds to membrane-associated phosphoinositol 4,5-bisphosphate, inducing a conformational change that enables it to bind to the β₁ cytoplasmic domain (79). By analogy, talin binding to phosphoinositol 4,5-bisphosphate may enable it to bind to β₃.

Regulation of α_IIbβ₃ function by TM domain interaction

TM domain–mediated protein oligomerization is a common mechanism for the assembly of membrane proteins and regulation of protein function (80). Specificity is achieved via specific sequence motifs superimposed on more general oligomerization frameworks (81–83). For example, the sequence motif GxxxG, first recognized as a framework for the homomeric association of the glycophorin A (GpA) TM helix (81), has been identified as the most overrepresented sequence motif in TM domain databases (82).

With regard to integrin TM domains, Li and coworkers reported that peptides corresponding to the α_IIb and β₃ TM domains readily undergo homodimeric and homotrimeric association, respectively, in phospholipid micelles (54), and Schneider and Engelman found that fusion proteins containing the α₂β₁, α₄β₇, and α_IIbβ₃ TM domains undergo integrin-specific TM domain–mediated homomeric and heteromeric association in bacterial membranes (84). Subsequently, Li et al. reported that facilitating the homomeric association of the β₃ TM helix by replacing either G708 or M701 with a polar asparagine induced α_IIbβ₃ activation and clustering when the mutants were expressed in CHO cells (85). They also found that mutation of the α_IIb GxxxG motif located at residues 972–975 disrupted the homomeric association of α_IIb TM helix (86) and paradoxically induced α_IIbβ₃ activation and clustering (87). These observations suggested the “push-pull” mechanism for α_IIbβ₃ activation shown in Figure 5. Processes that destabilize the association of the α_IIb and β₃ TM helices, such as talin binding to the β₃ cytoplasmic domain, would be expected to promote dissociation of the helices with concomitant α_IIbβ₃ activation. Conversely, intermolecular interactions that either require separation of the α_IIb and β₃ TM helices, such as homo-oligomerization, or are more favorable when they separate, such as ligand-induced α_IIbβ₃ clustering (88), would be expected to pull the equilibrium toward the activated state.

Figure 5.

Diagram illustrating the “push-pull” hypothesis for regulation of the α_IIbβ₃ activation state. The white and blue cylinders represent the α_IIb TM and membrane-proximal cytoplasmic domain helices, respectively. The red and green cylinders represent the β₃ TM and membrane-proximal cytoplasmic domain helices, respectively.

The ability of homomeric TM helix interactions to induce α_IIbβ₃ activation and clustering remains controversial (89, 90), but there is compelling evidence that heterodimeric interactions constrain α_IIbβ₃ in a low-affinity state. By simultaneously scanning the α_IIb and β₃ TM helices with cysteine residues, Luo et al. detected the formation of disulfide bonds with a helical periodicity in a region corresponding to α_IIb residues 966–974 and β₃ residues 693–702, consistent with the presence of a unique α_IIbβ₃ TM heterodimer (91). They also scanned the α_IIb and β₃ helices with leucines, confirming that mutation of the α_IIb GxxxG motif induces α_IIbβ₃ activation (90). Partridge et al. used random mutagenesis of the β₃ TM and cytoplasmic domains to search for interactions constraining α_IIbβ₃ activation (92). They detected 12 activating mutations in the membrane-proximal cytoplasmic domain and 13 activating mutations in the β₃ TM helix. Nine of the latter were predicted to shorten the helix, perhaps activating α_IIbβ₃ by altering the tilt of the helix in the membrane (Figure 4). The remaining mutations were located in the carboxyterminal half of the helix and were postulated to activate α_IIbβ₃ by disrupting the packing of an α_IIbβ₃ TM heterodimer.

Despite the biochemical evidence supporting the presence of α_IIb and β₃ TM domain oligomers, their existence has not been confirmed by NMR spectroscopy or x-ray crystallography because of difficulty in obtaining high-resolution structures for TM proteins using these techniques. However, computational methods have been used to construct TM domain models incorporating the constraints imposed by mutational data. Based on cryo-EM images (Figure 2A), Adair and Yeager proposed that the TM domains of inactive α_IIbβ₃ associate in a parallel α-helical coiled coil (38). Using the R995–D723 salt bridge as the primary constraint, they found that a right-handed coiled coil based on the GpA TM dimer (93) placed more conserved residues in the helix-helix interface than a coiled coil based on the canonical left-handed leucine zipper. Gottschalk and coworkers proposed that the α_IIb and β₃ TM helices remain in close contact in the activated state and that the helix-helix interface is a GpA-like structure containing the α_IIb G972xxxG975 and β₃ S699xxxA703 motifs (94). Moreover, simulated annealing and molecular dynamics supported a model in which the α_IIb and β₃ TM domains interact weakly in a right-handed coiled coil when the integrin is in its low-affinity conformation (95). Subsequently, in order to account for both aminoterminal and carboxyterminal restraints, Gottschalk proposed that the α_IIbβ₃ TM and membrane-proximal cytoplasmic domains form a right-handed coiled coil in which the helices interact over their entire length, placing the α_IIb GxxxG motif, but not β₃ S699xxxA703, in the helix-helix interface (96). By contrast, Luo et al. used their disulfide cross-linking data to construct a model based on the GpA TM dimer; however, in this model, the α_IIb GxxxG-like motif corresponded to residues 968–972, rather than 972–975 (91). DeGrado and coworkers used a Monte Carlo–simulated annealing algorithm to obtain atomic models for an α_IIb TM homodimer (86) and an α_IIbβ₃ heterodimer (87). In each case, a family of structures was found that satisfied mutational constraints. For the α_IIb homodimer, all structures had right-handed crossing angles ranging from 40° to 60°, but with an interface rotated by 50° relative to the GpA homodimer. In the case of the α_IIbβ₃ heterodimer, initial docking identified local minima with both right- and left-handed crossing angles. However, the right-handed structures had lower energies and more extensive interactions, and the α_IIb GxxxG motif was in intimate contact with the β₃ TM domain. Lastly, Partridge et al., using a Monte Carlo simulation, obtained 2 structures for an α_IIbβ₃ TM heterodimer with helix packing near either the amino or the carboxyl termini of the helices, respectively; of the 2 models, carboxyterminal helix packing was more consistent with their mutational data (92). It is obvious that there is wide disparity among these models, making it clear that obtaining actual structures for α_IIb and β₃ TM domain hetero- and homo-oligomers will be the next major advance in our understanding of the structural basis for the regulation of platelet integrin function.