Crystal Structure of the Platelet Glycoprotein Ibα N-terminal Domain Reveals an Unmasking Mechanism for Receptor Activation*

Glycoprotein Ib (GPIb) is a platelet receptor with a critical role in mediating the arrest of platelets at sites of vascular damage. GPIb binds to the A1 domain of von Willebrand factor (vWF-A1) at high blood shear, initiating platelet adhesion and contributing to the formation of a thrombus. To investigate the molecular basis of GPIb regulation and ligand binding, we have determined the structure of the N-terminal domain of the GPIbα chain (residues 1–279). This structure is the first determined from the cell adhesion/signaling class of leucine-rich repeat (LRR) proteins and reveals the topology of the characteristic disulfide-bonded flanking regions. The fold consists of an N-terminal β-hairpin, eight leucine-rich repeats, a disulfide-bonded loop, and a C-terminal anionic region. The structure also demonstrates a novel LRR motif in the form of an M-shaped arrangement of three tandem β-turns. Negatively charged binding surfaces on the LRR concave face and anionic region indicate two-step binding kinetics to vWF-A1, which can be regulated by an unmasking mechanism involving conformational change of a key loop. Using molecular docking of the GPIb and vWF-A1 crystal structures, we were also able to model the GPIb·vWF-A1 complex.

for the vWF-A1 domain and thrombin in its extracellular Nterminal domain and binding sites for filamin and 14 -3-3 in its cytoplasmic domain. Mutations in GPIb result in the congenital bleeding disorders, Bernard Soulier syndrome (BSS), and platelet-type von Willebrand disease (Pt-vWD). Several informative BSS point mutations result in an expressed GPIb receptor deficient in vWF binding, whereas the Pt-vWD point mutations produce a hyperactivated receptor. In addition, von Willebrand disease type IIb point mutations in the vWF-A1 domain also produce high affinity binding by inducing conformational changes in the A1 domain.
Although the principal ligand for GPIb␣ is vWf (1,2), it has also been shown to bind a growing number of other proteins including thrombin (3,4), kininogens (5), Factor XI (6), Factor XII (7), P-selectin (8), and Mac-1 (9). Proteolysis of GPIb␣ yields a 40 -45-kDa N-terminal domain containing binding sites for all the principal ligands (4). Studies on the mechanism of GPIb␣-vWf interactions are complicated by the fact that the two molecules do not normally interact under static conditions but only under shear stress (10,11). Several techniques have been used to induce interactions between these proteins including ristocetin, an aminoglycoside antibiotic (12), or botrocetin, a snake C-type lectin (13).
GPIb␣ is a member of the leucine-rich repeat (LRR) family (14,15), which includes a large number of proteins that are principally involved in mediating protein-protein interactions (16). LRRs are typically 22-28 amino acids long and occur in tandem repeats that are commonly flanked by conserved disulfide loop structures (17). The GPIb␣ N-terminal flank, LRRs, and C-terminal flanking anionic region have all been implicated in contributing to the vWF binding site (18,19), whereas GPIb␣ binding to thrombin is localized exclusively to the area of the anionic region. The anionic region includes three tyrosine residues (Tyr-276, Tyr-278, and Tyr-279) that are post-translationally sulfated in the native receptor, and this modification has been shown to be essential for binding to vWf and thrombin (20,21). We have previously determined the crystal structure of the A1 domain of vWf containing the binding site for GPIb␣ (22,23). To investigate further the molecular basis of the GPIb␣⅐vWF interaction, we now report the crystal structure of the N-terminal domain of the GPIb␣ chain.

EXPERIMENTAL PROCEDURES
Crystallization and Data Collection-The GPIb␣ N-terminal domain was prepared from platelets and in recombinant form. Glycocalicin, the GPIb␣ extracellular domain, was isolated and purified from platelets as described previously (24). This was cleaved to generate the N-terminal 45-kDa region (amino acids 1-288) and macroglycopeptide using 1:250 w/w Lys-C endoprotease treatment for 24 h at 4°C. The N-terminal domain was separated from the macroglycopeptide by linear salt gradient (0.2-1 M) on a Mono Q column equilibrated with 25 mM Tris 7.5, 0.2 M NaCl. DNA corresponding to the GPIb␣ signal peptide and resi-dues 1-291 was amplified by PCR and cloned into expression vector pCEP4 (Invitrogen), which had been modified to incorporate a C-terminal V5-His tag. This DNA was transfected into cultured cv-1 cells. The cv-1 monolayer secreted recombinant tagged GPIb␣, and the medium was harvested at 14-day intervals. GPIb␣ was purified by precipitation at 80% (NH 4 ) 2 SO 4 , and the pellet was dissolved in and dialyzed against 0.2 M NaCl, 25 mM Tris/HCl, pH 7.5 and centrifuged for 20 min at 50,000 ϫ g. This material was applied to a Q column (Bio-Rad) equilibrated with the same buffer and was eluted using a linear salt gradient. The resulting fractions were pooled and applied to a Ni 2ϩ -chelating Sepharose column (Amersham Biosciences) in 25 mM Tris/HCl, pH 8.0, 0.5 M NaCl. GPIb␣ was eluted using an imidazole gradient and gave a single 45-kDa band on an SDS-PAGE gel stained with Coomassie Blue. The protein concentration was estimated, the V5-His tag was cleaved off with Lys-C as described above, and the solution concentrated to 6 -10 mg/ml. Crystallization experiments were performed at 20°C using the sitting drop vapor diffusion method, and needle crystals grew under conditions of 0.1 M Tris/HCl, pH 6.5, 2 M (NH 4 ) 2 SO 4 over a period of 2 weeks using both platelet and recombinant GPIb␣. Crystals of 100 ϫ 100 ϫ 400 m were transferred into a cryoprotecting solution of 0.1 M Tris/HCl, pH 6.5, 2.2 M (NH 4 ) 2 SO 4 , 30% glycerol and were flashfrozen in a cryostream of nitrogen at a temperature of 100°K. The native and heavy atom data sets described in Table I were collected on station ID14 -4 of the European Synchrotron Radiation Facility. The crystals adopt space group p6 4 22 with cell dimensions a ϭ 202.1 Å, b ϭ 202.1 Å, c ϭ 128.0 Å, ␣ ϭ 90°, ␤ ϭ 90°, and ␥ ϭ 120°. Data were reduced with Denzo and scaled with Scalepack (HKL Research Inc.).
Structure Determination and Refinement-Heavy atom positions were located, and parameters were refined using SOLVE (Table I) (25). The resulting 3.5-Å phases had a 0.56 mean figure of merit. The phases were solvent-flattened (program DM, CCP4 suite) using a 50% solvent content, and the electron density showed clearly recognizable features of the GPIb␣ LRRs. Improved density was observed for space group p6 4 22 (over enantiomorphic p6 2 22) and revealed two GPIb␣ molecules in the asymmetric unit with a solvent content of 73%. These phases were applied to the 2.8-Å di-iodobis (ethylened iamine) diplatinum (II)nitrate structure factors, and using 2-fold molecular averaging and solvent flattening, the multiple isomorphous replacement phases were improved and extended to 2.8 Å. The resulting map was of sufficient quality to model 90% of the structure (program XtalView) (30). A 2F o Ϫ F c electron density map calculated at 2.8 Å was of high quality with additional side chain and main chain density observable. Several rounds of model building and refinement using CNS (26) were carried out initially using strong 2-fold non-crystallographic restraints (300 kcal/mole/Å 2 ), which were released gradually during later stages of the refinement. Individual B-factors were refined, and the addition of 166 water molecules gave a final R WORK of 0.245 and R FREE of 0.279 (5.0% of the reflections). The average temperature factors for protein atoms are 49.2 and 42.3 Å 2 for solvent atoms. Electron density is observed for GPIb␣ residues Pro-2 to Pro-280 in molecule A and Pro-2 to Thr-266 in molecule B with 99.8% of residues in allowable regions of the Ramachandran plot (67% of residue 304 in most favored regions and 32.8% of residue 163 in additional allowed regions). Residue Tyr-44 is in a disallowed region of the plot, but inspection of the electron density shows that this residue is well defined and has a stabilizing hydrogen bond. N-Acetylglucosamine residues are observed in the electron density connected to the glycosylation sites of Asn-21 and Asn-159. Additional oligosaccharide residues appear to be disordered. The model also includes one free sulfate ion and three sulfated tyrosine residues.

RESULTS
The GPIb␣ Structure-We expressed residues 1-288 from human GPIb␣ in simian cv-1 cells and the resulting purified protein crystallized in space group p6 4 22. Isomorphous crystals grow from the same GPIb␣ fragment generated by proteolysis of glycocalicin isolated from platelets. The structure was determined to 2.8 Å using four heavy atom derivatives (Table I) and refined to an R-factor of 24.5% (R FREE 27.6%) with two molecules in the asymmetric unit. The GPIb␣-fold consists of an N-terminal ␤-hairpin (residues 2-18), eight LRRs-(19 -204), a disulfide knot structure (amino acids 205-264), and a C-terminal anionic region (amino acids 265-280) (Fig. 1). The N-terminal ␤-hairpin has two anti-parallel strands with a disulfide (C4 -C17) bridge at the base. The second ␤-strand joins the convex parallel ␤-sheet in a configuration very similar to the spliceosomal complex subunit U2AЈ, which has only five LRRs (27). The two ␤-strands present on the concave face of the GPIb␣ LRRs 3 and 4 are also present in U2AЈ. The eight LRRs fold into a characteristic arc shape with a parallel ␤-sheet on the concave face. Previous sequence comparisons and/or models suggested that GPIb␣ contains seven LRRs and not the eight found here. This discrepancy is the result of the earlier definition of the sequence of an LRR, which left an overhanging sequence at each end as "flanking sequences." By changing the start and end sites of the LRR, these sequences are incorporated into the structure as an additional repeat. The convex face of LRRs 5, 6, 7, and 8 each has a novel LRR motif consisting of a tandem arrangement of three ␤-turns. They form a flattish amphipathic structure with the main chain hydrogen bonds in a linear arrangement (Fig. 2). The consensus for this motif is PXGLLXGL with leucine side chains projecting into the hydrophobic core. Adjacent ␤-turn repeat motifs form a regular parallel packing arrangement with the core leucine residues pointing alternately in and out of the plane of the structure interlocking with the corresponding leucines from the adjacent repeat. The LRR8 ␤-turn repeat motif has two phenylalanine residues in place of leucines in the central ␤-turn. These bulkier side chains serve to fill the gap between the end of the LRRs and the core of the C-terminal flank by packing against Phe-232 and Trp-235 from helix ␣1. It is unclear why the up and down repeating ␤-turn motif found here is not a more common secondary structural element. It is possible that the flattened structure is appropriate for packing in the LRRfold but would be less stable in other contexts. The GPIb␣ C-terminal Flanking Region-The GPIb␣ C-terminal flanking sequence contains two disulfide bonds, C209 -C248 and C211-C264, an ␣-helix (residues 214 -223), and a loop (residues 227-242), which extends over the interior of the GPIb␣ LRR concave face (termed the regulatory or R-loop in Fig 3, a and b). Initially, the structure continues to follow a circular coil similar to the LRRs with helix ␣1 lying parallel to the ␤-turn repeats of the concave face. Following helix ␣1 comes the triangular-shaped R-loop, two short 3 10 helices set at right angles, and a further loop (residues 251-264) extending to Cys-264. The structure of residues 269 -279 from the C-terminal anionic region is illustrated in Fig. 3c. These residues are well defined in the electron density of one GPIb␣ molecule but are disordered in the second molecule. The anionic region extends outwards from Cys-264 with Asp-269 to Asp-272 resembling a hinge that turns through 180°and is followed by residues Asp-274 to Asp-277, which form a single ␣-helical turn. because of steric hindrance from the thumb. We propose that the GPIb␣ crystal structure represents the low affinity or "closed" form of the receptor and that a conformational change in the thumb is required to unmask the A1 domain binding site. The structure of the high affinity or "open" form could either involve the thumb folding back partially and participating in an extended concave-shaped vWf-A1 binding site or, alternatively, a simple removal of the steric constraint. If the R-loop is removed from the GPIb␣ structure, the A1 domain can be satisfactorily docked against the concave surface.
BSS point mutations causing loss of vWF binding localize to the general area around the concave face ␤-strands from LRRs 5, 6, and 7 (L129P, A156V, and L179del) and also the sides of LRR2 (C65R and L57F) (Fig. 1). These mutations are all buried hydrophobic core residues with the exception of Ala-156, which lies on the surface of the concave face partially buried by surrounding side chains and the R-loop. Pt-vWD mutations G233V (28) and M239V (29), which activate the receptor, lie on the outer tips of the triangular R-loop and presumably act by favoring a more open conformation of the receptor (Fig. 3, a and  b). Further "activating" mutations D235V and K237V (30) have been identified by site-directed mutagenesis studies in the region of the R-loop. Further support for an unmasking mechanism comes from antibody 24G10 that inhibits vWf binding Secondary structure assignments made from the human GPIb␣ structure are superimposed on the alignment. Concave face LRR ␤-strands are colored gray, convex face ␤-strands are colored green, ␣-helices are colored dark blue, and 3 10 helices are colored light blue. The R-loop is colored orange, and the anionic region is colored red. Asparagine residues forming consensus LRR-buried hydrogen bonds are underlined (at residue 65, this position is occupied by cysteine). Mutations from Bernard-Soulier syndrome and platelet-type von Willebrand disease are shown as black and orange balls, respectively (note that the Leu-179 mutation is a single residue deletion). and maps to GPIb␣ residues 1-81 (20). This antibody binds to the Pt-vWd GPIb␣ mutants with increased affinity, indicating that it overlaps with a vWF-A1 binding site that is regulated by the conformation of the R-loop.
A Model of the GPIb␣⅐vWF Complex-Coordinates from the GPIb␣ structure (R-loop removed) and the vWf-A1 von Willebrand disease type IIb mutation structure (31) (residues 499 -701) were docked together using the automated search program FTDOCK (32). 2 This model was manually adjusted to optimize side chain interactions and remove steric conflicts using XtalView (33). The R-loop was then modeled back into the complex structure lying to one side of the A1 interface. In this model, loops from the top of the A1 domain-fold are oriented toward the GPIb␣ N-terminal ␤-hairpin with further extensive interactions formed down the full-length of the GPIb␣ LRR concave face ␤-strands (Fig. 6, a and b). Loops from the top of the A1 domain contribute basic residues Arg-524, Arg-629, Arg-632, which form salt bridges with Glu-14, Glu-40, and Asp-83, respectively. His-656 of A1 also forms a hydrogen bond to the main chain carbonyl from Lys-8. Gln-590 hydrogen bonds to Tyr-58/Gln-59, and Gln-628 forms a hydrogen bond to Glu-40. At the center of the interface, residues Asp-560 and Glu-596 A1 form salt bridges with Lys-152 and Lys-132 GPIb␣ and His-559, and His-563 of A1 forms hydrogen bonds to Glu-128 and Glu-175 of GPIb␣, respectively. The Asp-560 main chain carbonyl also forms a key hydrogen bond to the side chain nitrogen group of Lys-152. Toward the bottom of the interface Phe-603, Tyr-637 A1 packs against the hydrophobic residues Trp-230, Phe-199, and Phe-201 GPIb␣, and Gln-604 and Ser-607 of A1 form hydrogen bonds to the Val-299 main chain nitrogen and Asn-242 GPIb␣ side chain, respectively.
In the GPIb␣ structure, the C-terminal anionic region is located ϳ50 Å from the center of the LRR concave face, indicating that this forms a distinct binding site for vWF-A1. The model in Fig. 6, a and c, illustrates that an extended form of the anionic peptide with the Asp-274 to Asp-277 helix intact can act in concert with the GPIb␣ concave face to engage a single vWF-A1 domain. In this model the helical arrangement of three sulfate groups from GPIb␣ residues, Tyr-276, Tyr-278, and Tyr-279, form interactions with four basic residues, Lys-569, Arg-571, Lys-572, and Arg-573 from A1. The overall model is consistent with the reported 1:1 stoichiometry for the GPIb␣-A1 interaction. All A1 domain residues described in this model have been implicated by mutagenesis in binding to GPIb␣ with the exception of Phe-693 and Gln-628, which have yet to be studied (34 -38). Furthermore, as discussed in the previous sections, there are corroborative data confirming the involvement of the GPIb␣ LRRs and the anionic region in 2 The highest scoring complex with an RPscore of 5.4 was used (32). interactions with vWF-A1. The model has a similar architecture to the crystal structure of the spliceosomal U1AЈ⅐U2BЉ protein complex in that the ␣/␤-fold of the U2BЉ domain binds the concave face of the U1AЈ LRRs.
The model of the complex provides a rationale for the affinity differences observed for BSS and Pt-vWD GpIb␣ mutants and also between the wild type and activated type IIb mutant vWF-A1. In the latter case, as residue Asp-560 A1 is buried at the center of the complex interface, the small conformational differences in this residue observed between the wild type and type IIb mutant crystal structures (31) can exert a disproportionately large effect on binding affinity. In the wild type vWF-A1 structure, the Asp-60 peptide bond is rotated through ϳ180°, which results in the loss of a hydrogen bond between the Asp-560 main chain carbonyl and Lys-152 GpIb␣ NZ nitrogen. Perhaps more significantly, the Asp-560 side chain moves into a position creating a steric conflict with Tyr-130 GpIb␣ and is also closer to Glu-128 carboxylate than its predicted binding partner Lys-152. Because of the dense side chain packing on the surface of the LRR concave ␤-strands, it is difficult to alleviate these steric clashes through rotation of GPIb␣ side chain torsional angles. A similar argument may explain the complete loss of vWF binding caused by the subtle GPIb␣ A156V mutation in Bolzano variant BSS patients. Ala-156 is not predicted to interact with the A1 domain as it is partially buried by surrounding GPIb␣ side chains Lys-132, Gln-180, Leu-178, and Asn-157. However, the mutation to the bulkier valine would push the neighboring side chains out of their wild type conformation, and in particular, this would be predicted to disrupt the Lys-132 salt bridge to Glu-596 from the A1 domain. DISCUSSION The GPIb␣ structure shows how the N-and C-terminal flanking regions are intimately related with the LRRs and reveals a mechanism for regulating the affinity of the receptor via the conformation of the R-loop by unmasking the LRR binding site for vWf-A1. The role of the LRR concave face binding site is fundamental for binding the A1 domain as the mutations here or binding of antibodies with epitopes in this region block A1 binding in response to all reagents (20). Thus, the A1 domain interaction with the GPIb␣ anionic peptide probably precedes binding to the concave face in multistepbinding kinetics that are reminiscent of the thrombin-hirudin interaction (Fig. 7) (39). It is also conceivable that the GPIb␣ anionic peptide binds and stabilizes a high affinity conformation of the A1 domain. Heparin is able to activate vWf binding to GPIb, which can be explained if heparin binds to the A1 domain and stabilizes a high affinity conformation. This notion is further supported by the observation that a monoclonal antibody to the A1 domain that blocked heparin binding (mAb 724) was also able to activate vWf binding to GPIb (40). There is also the possibility that GPIb␣ anionic region could activate the A1 domain as the putative interaction site on A1 lies in the same region as the von Willebrand disease type IIB mutations (40).
The GPIb␣-unmasking mechanism may provide a general pathway whereby other members of the family of LRR proteins can regulate ligand binding by utilizing their flanking sequences in a manner similar to the GPIb␣ R-loop. How the in vivo high shear stress conditions required to stimulate GPIb␣-vWF binding affect the two vWF binding sites and the conformation of the R-loop will require further study. Platelet-sized latex beads coated with recombinant GPIb␣ N-terminal domain show the same elevated binding to immobilized vWF at increasing shear stress as platelets (41). This indicates that the shear-dependent activation is inherent within the N-terminal domain and is not induced by changes in the cytoskeleton or receptor clustering effects. It is possible that fluid shear forces act directly on the GPIb␣ N-terminal domain structure, affecting the equilibrium between open and closed conformations of the R-loop.
Binding between GPIb␣ and vWf is a central element in the regulation of hemostasis and progression to the pathological condition thrombosis. Restriction of vessel size by arteriosclerosis enhances shear and is a major factor in propagating cardiovascular disease. Recent studies have also implicated the abnormal expression of the GPIb complex on breast cancer cells in malignancy and metastasis (42,43). The GPIb␣-vWf axis is an important target for anti-thrombotic strategies and drugs, is capable of inhibiting GPIb␣-vWf binding, and would be important tools for prophylaxis and treatment of such diseases.
Thrombin and P-selectin Binding to GPIba-GPIb␣ binds constitutively to ␣-thrombin (3, 4), P-selectin (8), and integrin Mac-1 (9) through interactions with its C-terminal flanking sequences. Of these interactions, the 0.1 M K d ␣-thrombin binding to GPIb␣ is the best characterized. In particular, sulfated tyrosines 276, 278, and 279 of the GPIb␣ anionic region and the electropositive heparin binding site of thrombin were shown to be essential components for binding (23). There is an interesting similarity between the GPIb␣ Asp-274 to Asp-277 ␣-helix and the ␣-helical region in the anionic peptide from the thrombin inhibitor hirudin. Here, the sulfated Tyr-276 in GPIb␣ occupies the same position on the helical turn as the sulfated Tyr-63 in hirudin. However, the structure of the thrombin/hirudin complex showed that the anionic peptide of hirudin binds to a different region of thrombin (exosite I) than GPIb␣ (44). A recombinant fragment of vWf blocked thrombin binding to platelets (45), indicating that these binding sites in the anionic region overlap. Although the role of the thrombin binding site on GPIb␣ in platelet physiology has been controversial in the past, evidence for a critical function is growing (46).
The GPIb␣ anionic region is also strongly implicated in mediating the GPIb␣ P-selectin interaction, which is thought to play a role in the adhesion of platelets to endothelial cells lining the vessel wall. A crystal structure exists for P-selectin in complex with the anionic peptide from the PSGL-1 leukocyte receptor (47). In this case, 10 amino acid residues of the peptide were involved in binding with the sulfate groups from two modified tyrosine residues contributing through hydrogen bonding and electrostatic interactions. Unlike the interaction with GPIb␣, the P-selectin/PSGL-1 anionic peptide binding requires calcium and carbohydrate core-2 branching or ␣-(1,3)- Under high blood shear stress, GPIb␣ and vWf are activated, resulting in a high affinity adhesive interaction. Conformational changes occur in both GPIb␣ and vWf. (i) Plasma vWf undergoes a quaternary unfolding in which the A1 domain is exposed and subsequently adheres to the subendothelium (shown in green) where it becomes activated. (ii) The mobile anionic region of GPIb␣ forms an initial interaction with the A1 domain, but high affinity binding only occurs when GPIb␣ is also activated and the R-loop is displaced unmasking the second A1 binding site on the GPIb␣ concave face. fucosylation. In conclusion, these cell-adhesive anionic peptides are inherently partially disordered in the unbound state. Short elements of secondary structure and a variety of positions of charged residues and sulfated tyrosines allow for diverse specificities.