Crystal Structure of the GpIbα-Thrombin Complex Essential for Platelet Aggregation

See allHide authors and affiliations

Science  11 Jul 2003:
Vol. 301, Issue 5630, pp. 222-226
DOI: 10.1126/science.1083917


Direct interaction between platelet receptor glycoprotein Ibα (GpIbα) and thrombin is required for platelet aggregation and activation at sites of vascular injury. Abnormal GpIbα-thrombin binding is associated with many pathological conditions,including occlusive arterial thrombosis and bleeding disorders. The crystal structure of the GpIbα-thrombin complex at 2.6 angstrom resolution reveals simultaneous interactions of GpIbα with exosite I of one thrombin molecule,and with exosite II of a second thrombin molecule. In the crystal lattice,the periodic arrangement of GpIbα-thrombin complexes mirrors a scaffold that could serve as a driving force for tight platelet adhesion. The details of these interactions reconcile GpIbα-thrombin binding modes that are presently controversial,highlighting two distinct interfaces that are potential targets for development of novel antithrombotic drugs.

Adhesion of blood platelets to damaged vessel walls and aggregation of platelets at sites of injury are regulated by glycoprotein receptor (Gp) Ib-IX-V, a complex composed of four transmembrane subunits: GpIbα, GpIbβ, GpIX, and GpV (1). The GpIbα subunit binds the plasma glycoprotein von Willebrand factor (vWF) (1, 2), a multimeric ligand that together with GpIbα regulates adhesion of platelets to the subendothelium (3, 4). Although a major function of GpIbα is to adhere to vWF at sites of injury, GpIbα is also a receptor for thrombin, an allosteric serine protease released from damaged tissue that is essential for hemostasis and platelet activation (5, 6). Thrombin binding studies have identified sites of high, moderate, and low affinity on human platelets, and GpIbα receptors have been shown to account for high-affinity binding (dissociation constant Kd ∼100 nM) (69). Binding of thrombin to GpIbα is essential for initiation of platelet procoagulant activity, and exposure to <1 nM thrombin is sufficient to induce aggregation and activation (6, 10). In pathological states, extensive platelet aggregation can overcome the normal mechanisms that limit the size of the hemostatic plug and lead to thrombosis, the precipitating factor in most cases of life-threatening acute coronary occlusion (11).

Thrombin-mediated platelet activation requires a complex set of interactions involving multiple substrates and receptors (1, 12). Recent studies suggest that this process involves not only hydrolysis of protease-activated G protein–coupled receptors (PARs) (13), but also signaling responses through the GpIb-IX-V complex that follow GpV cleavage by thrombin (14), or an α-thrombin/GpIb–dependent pathway that does not require hydrolysis of substrate (15).

Although extensive biochemical and functional data indicate that GpIbα is a thrombin receptor (1, 16), the precise nature of GpIbα-thrombin interactions is controversial. Various data suggest that exosite I on thrombin interacts with GpIbα (1719), whereas other studies implicate exosite II in GpIbα interaction (9, 2022). These anion-binding exosites are also thought to bind to the antithrombotic agents hirudin and heparin (20, 23, 24). Both thrombin and vWF bind to the 300–amino acid N-terminal extracellular domain of GpIbα (1). Biochemical and mutagenesis data suggest that an anionic region at the C-terminal end of this domain (residues Asp269 to Asp287) binds thrombin (21, 2528). Three Tyr residues in this anionic region are subject to posttranslational sulfation (sTyr), and sulfation is necessary for optimal binding to both thrombin and vWF in vivo (2628). Synthetic peptides corresponding to this anionic region inhibit thrombin binding and platelet aggregation (25, 29). However, peptides that correspond to a distinct region of GpIbα (Phe216 to Thr240) have been shown to significantly inhibit platelet aggregation (29), although these same peptides do not appear to block thrombin binding to platelets (25).

The crystal structures of the GpIbα-binding domain of vWF (vWF-A1) and GpIbα have been determined separately (3032), as has the crystal structure of GpIbα complexed with the vWF-A1 domain (32). To gain insight into the adhesive mechanism used by platelet receptors to regulate thrombus formation, we have determined the 2.6 Å crystal structure of the N-terminal domain of GpIbα (residues 1 to 279) in complex with thrombin (light-chain and heavy-chain residues 1 to 258). The extracellular thrombin-binding domain of human GpIbα (residues 1 to 305) lacking sites of N-linked glycosylation (Asn21Asp and Asn159Asp) was expressed in Chinese hamster ovary (CHO) cells, and diisopropyl fluorophosphate (DFP)–inactivated human α-thrombin was acquired from Haematologic Technologies (33) (see details of structure determination in Table 1).

Table 1.

Statistics for data collection and refinement.

Data collection:
Space group P212121
Unit cell dimensions a = 47.1 Å, b = 111.8 Å, c = 176.3
Resolution range (Å) 30-2.6 (2.74-2.6)
Completeness (%) 97.7 (95.0)
Total observations 411,862
Unique reflections 28,711 (3,971)
Average I/σ(I) 9.0 (1.9)
RsymView inline (%) 7.1 (39.2)
Model refinement:
Maximum resolution (Å) 2.6
Number of reflections (free) 27,605 (1,903)
Rwork/RfreeView inline 20.7/23.8
No. of protein atoms 4,466
No. of waters 217
rms deviations
Bonds (Å) 0.007
Angles (°) 1.40
  • View inline* Rsym = Σ|Ih — 〈Ih〉|/ΣIh, where 〈Ih〉 is the average intensity over symmetry equivalents. Numbers in parentheses reflect statistics for the last resolution shell.

  • View inline Rwork = Σ∥Fobs| — |Fcalc∥/Σ|Fobs|, where Rfree is equivalent to Rwork but is calculated for a randomly chosen 6.4% of reflections omitted from the refinement process.

  • In the crystals, each pair of GpIbα and thrombin molecules associates as a crystallographically independent complex, and the multivalent potential of both molecules is realized by the formation of oligomeric arrays in which two distinct binding sites continuously alternate throughout the crystal. The two interfaces at these sites are evident within two adjacent complexes that are related by crystal symmetry (Fig. 1). The first interface involves interactions between thrombin exosite I (loop Lys65-Glu76 with surrounding residues) and a site on GpIbα that spans the C-terminal end of the leucine-rich repeat (LRR) domain and the conserved disulfide loop region (Cys209-Cys264). The second binding interface is between exosite II of a second thrombin molecule (region around α helix His242-Phe257) and two adjacent regions of GpIbα: the sulfated anionic region and a short α helix (Arg123-Leu129) on the convex surface of the LRR domain. The DFP molecule is located in the active-site cleft of thrombin, forming a covalent bond with the catalytic serine (Ser205).

    Fig. 1.

    Ribbon representation of two GpIbα-thrombin complexes related by crystal symmetry. The asymmetric unit contains one molecule of GpIbα and one molecule of thrombin, which associate as a crystallographically independent complex. GpIbα is colored green, except for the anionic region (Asp269 to sTyr279), which is in red, and a region corresponding to peptide Phe216 to Thr240 is in gold. The thrombin heavy chain is colored light blue, with exosite I (Lys21 to Gln24, Tyr71 to Asn74, and Lys106 to Lys107) and exosite II (Arg98, Arg123, and Arg245 to Lys252) colored dark blue (numbering for thrombin used here starts at the first residue of the mature heavy chain of human thrombin, and addition of 363 converts the numbering to that of prothrombin; see supporting online text for chymotrypsin numbering). The thrombin light chain (residues 1 to 29) is in gray.

    The binding interface between GpIbα and thrombin exosite I is predominantly electrostatic and buries a total of ∼1400 Å2 of surface area (34). Eight direct hydrogen bonds and 14 water-mediated contacts are observed at this binding interface, as well as hydrophobic interactions (Fig. 2A) (see supporting online text for list of contacts between GpIbα and thrombin exosites I and II). GpIbα residues at the N-terminal end of β strands (Lys152, regions Asn173 to Asp175, and Glu193 to Pro198) and within the disulfide loop region (Tyr215 to Tyr228, and Lys237) form contacts with residues of exosite I (region Lys21 to Lys107). Cleavage of the peptide bond Arg73-Asn74 and chemical modification of Lys residues within exosite I disrupt binding to GpIbα (35). Arg73 has a direct contact with Ser194 and water-mediated contacts with Tyr215 and Gly193 in GpIbα, and cleavage of Arg73-Asn74 could alter the recognition site for GpIbα, thus accounting for loss of binding.

    Fig. 2.

    Stereo representation of residues at the interfaces between GpIbα and thrombin exosites I and II. GpIbα is colored green, thrombin is in blue; hydrogen bonds are indicated by dashed red lines. (A) Ribbon diagram of the GpIbα–exosite I interface, including side chains of interacting residues. Extensive electrostatic and hydrophobic interactions include contacts between GpIbα residues (Tyr228, Glu225, and Ser194) and thrombin residues (Lys21, Lys106 and Arg73) (see supporting online text for list of additional contacts between Gp1bα and thrombin exosites I and II). A total of ∼1400 Å2 of surface area is buried at the GpIbα–exosite I interface [calculated with the program SURFACE (34); probe radius 1.4Å]. (B) Ribbon diagram of GpIbα–exosite II interface. Sulfate groups on sTyr276 in GpIbα form salt bridges with side-chain nitrogens on Arg123 and Lys248 in thrombin, and the main-chain carbonyl of sTyr276 forms a hydrogen bond with Arg245, which in turn is stabilized by intramolecular contacts with Asn184. Both the main-chain carbonyl and side-chain oxygen of Asp277 in GpIbα participate in a water-mediated contact with His87. The side-chain oxygen of Asp277 also forms a direct interaction with a guanidinium nitrogen on thrombin Arg98. Finally, the ring of sTyr276 in GpIbα packs against the side chain of Lys248, while its sulfate group forms a salt bridge with the side-chain nitrogen. Additional hydrophobic interactions include the packing of Phe244 from thrombin between Leu275 and sTyr276 of GpIbα, and packing of thrombin Arg89 against the ring of sTyr279 in GpIbα (not shown).

    Peptides that bind to exosite I, including those encompassing the Phe216 to Thr240 region of GpIbα and a peptide corresponding to the C-terminal end of hirudin, do not appear to block thrombin binding to platelets (17, 25). Yet, these peptides have been shown to inhibit thrombin-induced aggregation with a median inhibitory concentration (IC50 ∼ 50 μM) similar to that of anionic-region peptides (8, 17, 29). This paradox can be addressed by the GpIbα-thrombin structure.

    The extensive GpIbα–exosite II interface buries ∼2000 Å2 of surface area. About 60% (∼1200 Å2) of the total buried surface area comes from a network of hydrogen bonds between the anionic region of GpIbα (residues Asp269 to sTyr279) and the highly basic residues in thrombin exosite II (Figs. 2B and 3C). All residues from sTyr276 to sTyr279, including attached sulfate groups, participate in ionic and hydrophobic interactions with thrombin (see legend to Fig. 2B for details). Sulfate groups on sTyr276 and sTyr279 form salt bridges with side-chain nitrogens on basic residues in thrombin exosite II (Arg98, Arg245, and Lys252; see Figs. 2B and 3C and supporting online text). Point mutations at Arg98, Arg245, and Lys252 decrease GpIbα-binding affinity of thrombin by 20- to 30-fold and diminish thrombin-induced aggregation (9, 22). These residues have also been implicated in direct binding interactions with heparin (36). Overall, functional data (see supporting online text for further evidence) indicating the involvement of the anionic sulfated region of GpIbα and exosite II in binding (9, 2022, 2528) are in agreement with the observed structure.

    Fig. 3.

    (A to C) The interacting surfaces of GpIbα and thrombin. Surfaces in (A) and (B) are colored according to electrostatic potential with GRASP (41). The electrostatic potential is contoured in the range of –10 kbT (red) to +10 kbT (blue), where kb is Boltzmann's constant and T is absolute temperature (K). (A) The charge distribution on the surface of thrombin highlights the high electropositivity at exosites I and II. The GpIbα receptor is shown as green ribbons with the surface rendered transparent. (B) The charge distribution on the surface of GpIbα highlights the highly electronegative regions that bind to exosites I and II. Thrombin is shown as blue ribbons with a transparent surface. Panel (B) is rotated about the x axis to present the GpIbα binding surface to the viewer. (C) Ionic interactions between residues in the anionic region of GpIbα and basic residues in thrombin exosite II. The sulfated anionic region is shown as a ball-and-stick model (green). Hydrogen bonds are indicated by dashed white lines. The anionic region occupies a shallow depression on the surface of exosite II. The molecular surface representation of thrombin (transparent) is shown to emphasize the high geometric match between the two interacting surfaces. Tryptophan modification studies showing loss of high-affinity binding and diminished activation (42) are consistent with modification of Trp249, a solvent-accessible residue involved in interaction with sTyr279 (3.8 Å away from the sulfate group; see supporting online text). (D and E) Comparison of the GpIbα-thrombin complex, unliganded GpIbα, and the GpIbα–vWF-A1 complex. (D) Superposition of structures of GpIbα-thrombin and unliganded GpIbα (colored in red), showing the rearrangement of the C-terminal anionic region of GpIbα. sTyr residues in the anionic region are shown as a ball-and-stick representation. The structure of unliganded GpIbα (where the conformation of the anionic region may be in equilibrium with a more extended conformation) is derived from Protein Data Bank ID 1GWB (31). The overall structure of GpIbα-bound thrombin is similar to previously determined thrombin structures [e.g., hirudin-bound thrombin (23)] with no notable conformational changes observed. (E) Superposition of GpIbα-thrombin structure on GpIbα–vWF-A1 structure [determined to 2.6 Å resolution (43)]. GpIbα is colored yellow and the A1 domain is in red. The overall structure of the GpIbα–vWF-A1 complex is similar to the previously published 3.1 Å structure (32), but the 2.6 Å structure is not derived from proteins bearing the gain-of-function mutations (GpIbα-M239V and vWF-R543Q).

    In the second part of the GpIbα-exosite II interface, residues on the convex face of GpIbα, within the Arg121 to Gly170 region, make mostly hydrophobic contacts with thrombin residues from the second α helix, including Arg123 and Glu124 (supporting online text). This part of the GpIbα–exosite II binding interface represents a newly defined site, because amino acid residues that map to this interface have not been tested by mutagenesis.

    The electrostatic potential on the surfaces of GpIbα and thrombin shows a heterogeneous charge distribution and reflects striking charge complementarity between interacting surfaces of GpIbα and thrombin (Fig. 3, A to C). The largest patches of negative charge on the surface of GpIbα are localized to regions that interact with thrombin at the exosite interfaces, and the most prominent patches of positive charge on thrombin are at exosites I and II.

    The anionic region at the C-terminal end of GpIbα (Asp269 to sTyr279) adopts an entirely different conformation in the thrombin-bound and in the unliganded structures (31) (Fig. 3D). In unliganded GpIbα, the sulfate groups on sTyr residues interact with basic residues on the convex surface. In the GpIbα-thrombin complex, these sTyr residues are rotated and shifted more than 20 Å away from GpIbα to make contacts with thrombin.

    In the structure of the GpIbα–vWF-A1 complex, the vWF-A1 domain occupies a large groove on the concave surface of GpIbα, and a loop from Val227 to Ser241, termed the regulatory “β-switch,” is positioned close to vWF-A1 to form a continuous β sheet between the two proteins (32). In contrast, in the GpIbα-thrombin complex, the β-switch region extends away from the LRR domain in a conformation resembling that of unliganded GpIbα (31). This conformation favors intimate contacts with thrombin, but would be incompatible with binding of vWF-A1. Moreover, direct comparison of GpIbα-thrombin and GpIbα–vWF-A1 complexes shows that a steric clash would preclude simultaneous binding of a thrombin molecule and a vWF-A1 domain to the concave surface of GpIbα (Fig. 3E).

    Electron density for the GpIbα linker Thr266-Leu-Gly-Asp-Glu-Gly-Asp272, which tethers the C-terminal anionic region to the cysteine loop and LRR regions, is relatively weak, and these residues do not appear to have direct contacts either with thrombin or with nearby residues on the GpIbα surface. This linker could be inherently mobile and poorly ordered for functional reasons—to provide flexibility to the overall assembly of strings of receptor-ligand complexes. Structural mobility of the C-terminal anionic region (Asp269 to sTyr279) of GpIbα is limited by a disulfide bond (Cys211–Cys264) located ∼20 Å away from this region. Considering this structural limitation and the positions of exosites I and II at opposite ends of thrombin (exosite II is more than 60 Å away from Cys264), simultaneous binding of GpIbα to both exosites on the same thrombin molecule is impossible.

    In the context of cell-cell recognition and GpIbα-dependent platelet adhesion (1, 37), the high avidity of thrombin for platelets and thrombotic activity at physiologically relevant, low thrombin concentrations is likely accounted for by additive energetic contributions from individual GpIbα-exosite interfaces. In this regard, fortuitous crystal packing of the GpIbα receptor relative to thrombin provides a scaffold (Fig. 4) that can support tight, multivalent adhesive interactions between platelets. Given the mode of assembly shown in Fig. 4, a GpIbα receptor projecting from the cell membrane would interact with another GpIbα receptor, indirectly through an intervening thrombin molecule. In this arrangement, GpIbα receptors from different membranes are aligned in alternating fashion, and the active site of thrombin remains accessible for other physiological substrates (e.g. PARs, fibrinogen). The linear zipper observed in the GpIbα-thrombin lattice resembles the “cell-adhesion zipper” seen in crystals of cadherin dimers (38).

    Fig. 4.

    Schematic diagram of thrombin and GpIb-IX complexes, in which GpIbα and thrombin molecules are arranged as an adhesive ribbon structure observed in the crystals (color coding is the same as in Figs. 1, 2, 3). The region between the last residue observed in GpIbα (sTyr279) and the stalk region is shown as green dotted lines. GpIbα is covalently attached to GpIbβ by means of a disulfide bond near the extracellular surface of the membrane, and GpIb-IX-V is present as a noncovalent 2:2:1 complex on the platelet surface (1, 16). GpV is not included in the model, because recent evidence suggests that the extracellular domain of GpV (removed from the receptor complex by thrombin cleavage early in the aggregation process) functions as an inhibitor of platelet activation and aggregation (14). The model is further simplified as a 1:1 GpIb-IX complex. The orientation of multiple complexes with respect to membranes is deduced from the location of the C-terminal end of the GpIbα fragment relative to the thrombin-binding domains. Simultaneous binding of two GpIbα receptors at two polar ends of the bridging thrombin molecule stabilizes antiparallel orientations of adjacent receptors and ensures that these neighboring receptors extend their C-terminal ends in opposite directions. The long axis of each GpIbα receptor is oriented roughly normal to the platelet membrane, with the C-terminal end positioned toward the membrane and extending away from sites of thrombin attachment. A long O-glycosylated mucin-like stalk (not drawn to scale) of the GpIbα receptor places its thrombin binding domain ∼45 nm away from the platelet membrane surface (39); thus, platelet membranes that are about 100 nm apart could be linked together by binding interactions between GpIbα and thrombin.

    The multimeric assembly described above offers an explanation for the paradoxical observation that peptides corresponding to the exosite I–binding regions of GpIbα and hirudin do not inhibit thrombin binding to platelets (17, 25), yet inhibit thrombin-induced platelet aggregation (8, 17, 29). Binding of GpIbα to exosite II, not exosite I, could account for thrombin trapping on the platelet surface. After attachment of exosite II to one GpIbα receptor, platelet activation would occur as a result of subsequent binding of exosite I to another GpIbα receptor. In the proposed sequence of binding events, exosite II is responsible for thrombin attachment to the platelet surface, whereas exosite I mediates the interaction between platelets, thus explaining the inhibitory effect of these exosite I–binding peptides on thrombin-induced platelet aggregation.

    Thrombin-induced platelet activation and subsequent aggregation are a poorly understood mechanism that involves a network of cascade events including transmission of activation signals across the membrane and a number of intracellular processes associated with cytoskeletal rearrangement (39). Thrombin binding to GpIbα represents only one aspect of these multiple events. At the same time, regardless of the precise nature of the mechanisms involved, our structural data suggest that thrombin might also behave as an adhesive protein, promoting thrombus growth by a noncatalytic mechanism. Whether such adhesive interactions are required for the efficient use of thrombin (when only small amounts may be available at sites of vascular injury) or enhance the potency of GpIb-IX-V–mediated signaling by concentrating receptors and their signaling complexes at the surface of activated platelets remains to be established. The model derived from our crystal structure (Fig. 4) is in very good agreement with the recently proposed mechanism (14) by which the GpIbα-bound thrombin can itself, after GpV cleavage, induce thrombosis independent of proteolytic activity. Because the data indicate that such cleavage can occur in vivo (40), our model could represent a snapshot of GpIbα-thrombin interactions that follow GpV cleavage.

    In conclusion, the structural data presented here provide important details of GpIbα-thrombin interactions and open promising avenues for further investigation and therapeutic applications involving development of novel antithrombotic drugs to treat coronary artery disease and other hemostatic disorders.

    Supporting Online Material

    Materials and Methods

    SOM Text

    Figs. S1 and S2


    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article