Leucine-rich Repeats 2-4 (Leu60-Glu128) of Platelet Glycoprotein Ibα Regulate Shear-dependent Cell Adhesion to von Willebrand Factor*

Glycoprotein (GP) Ib-IX-V binds von Willebrand factor (VWF), initiating thrombosis at high shear stress. The VWF-A1 domain binds the N-terminal domain of GPIbα (His1-Glu282); this region contains seven leucine-rich repeats (LRR) plus N- and C-terminal flanking sequences and an anionic sequence containing three sulfated tyrosines. Our previous analysis of canine/human and human/canine chimeras of GPIbα expressed on Chinese hamster ovary (CHO) cells demonstrated that LRR2-4 (Leu60-Glu128) were crucial for GPIbα-dependent adhesion to VWF. Paradoxically, co-crystal structures of the GPIbα N-terminal domain and GPIbα-binding VWF-A1 under static conditions revealed that the LRR2-4 sequence made minimal contact with VWF-A1. To resolve the specific functional role of LRR2-4, we compared wild-type human GPIbα with human GPIbα containing a homology domain swap of canine for human sequence within Leu60-Glu128 and a reverse swap (canine GPIbα with human Leu60-Glu128) for the ability to support adhesion to VWF under flow. Binding of conformation-specific anti-GPIbα antibodies and VWF binding in the presence of botrocetin (which does not discriminate between species) confirmed equivalent expression of wild-type and mutant receptors in a functional form competent to bind ligand. Compared with CHO cells expressing wild-type GPIbα, cells expressing GPIbα, where human Leu60-Glu128 sequence was replaced by canine sequence, supported adhesion to VWF at low shear rates but became increasingly ineffective as shear increased from 50 to 2000 s-1. Together, these data demonstrate that LRR2-4, encompassing a pronounced negative charge patch on human GPIbα, is essential for GPIbα·VWF-dependent adhesion as hydrodynamic shear increases.

Binding of platelet glycoprotein (GP) 2 Ib-IX-V to von Willebrand factor (VWF) in plasma, subendothelial matrix, or on endothelium initiates thrombus formation at high shear stress in normal hemostasis and thrombotic diseases, such as heart attack or stroke (1)(2)(3)(4)(5). GPIb-IX-V consists of four members of the leucine-rich repeat (LRR) family: GPIb␣ disulfide-linked to GPIb␤ and associated with GPIX and GPV (2:2:2:1) (1). VWF binds to the N-terminal domain of GPIb␣ (His 1 -Glu 282 ), consisting of seven leucine-rich repeats (Leu 36 -Ala 200 ), N-and C-terminal flanking sequences (His 1 -Ile 35 and Phe 201 -Gly 268 ), and an anionic sequence (Asp 269 -Glu 282 ) containing three sulfotyrosines at positions 276, 278, and 279. We previously expressed a series of human/canine and canine/human chimeras of GPIb␣ and mapped binding sites for VWF and a panel of inhibitory anti-GPIb␣ antibodies to precise structural domains (6,7). This approach is based on the specificity of human VWF and murine antibodies for human (not canine) GPIb␣ (6). Chimeras consisted of human sequence incrementally replaced by canine sequence from the N terminus at domain boundaries and canine sequence His 1 -Glu 282 rehumanized from the N terminus (human replacing canine sequence). LRR2-4, spanning residues Leu 60 -Glu 128 , was identified as crucial for GPIb␣-dependent adhesion to VWF under shear conditions. Paradoxically, co-crystal structures subsequently reported for GPIb␣ N-terminal domain and VWF-A1 domain fragments (8 -10) revealed major contact sites clustered N-and C-terminally to LRR2-4, whereas the Leu 60 -Glu 128 sequence made minimal contact with VWF-A1 (with the exception of a single water-mediated contact between Asp 63 of GPIb␣ and Arg 571 of VWF-A1) (10). This discrepancy between structure and function requires resolution to understand the molecular basis for shear-dependent platelet adhesion, especially if the GPIb␣-VWF interaction is considered as an anti-thrombotic target (4,5). Notably, LRR2-4 in human but not canine GPIb␣ has a pronounced negative charge patch at the concave surface of the repeats, complementary to a positive patch on VWF, implying electrostatic interactions are critical for GPIb␣mediated adhesion to VWF (1,11), even though this * This work was supported in part by the National Health and Medical Research Council of Australia and by the National Heart Foundation of Australia. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. sequence makes minimal contact with VWF-A1 under static conditions used for crystallography (8 -10).
Here we analyzed GPIb␣ human/canine homology domain swaps designed to maintain N-and C-terminal contact sites (based on crystal structures) (9, 10) but altered the intervening electrostatic region (6), to establish the specific functional role of Leu 60 -Glu 128 . Comparing wild-type (WT) human GPIb␣ and a homology swap with canine instead of human sequence within Leu 60 -Glu 128 (HUM CAN60 -128 ) expressed on CHO cells shows that LRR2-4 of GPIb␣ is essential for GPIb␣⅐VWF-dependent adhesion as hydrodynamic shear increases.
Molecular Modeling-Models of canine GPIb␣ N-terminal domain and human/canine homology swaps, based on the GPIb␣ crystal structure, were built as described previously (6,11).
GPIb-IX-transfected CHO Cells-Human/canine GPIb␣ Leu 60 -Glu 128 homology domain swap constructs were gen-  (9,10). In the underlying ribbon representation, residues Leu 60 -Glu 128 of GPIb␣ are in green, and the rest of the molecule is in magenta. In the right-hand panel, the VWF-A1 domain is in cyan. The model of the HUM CAN60 -128 homology domain swap, containing canine Leu 60 -Glu 128 sequence, shows the loss of the negative patch (compare bottom left and top left; yellow circles). The HUM CAN60 -128 model was produced using the coordinates of GPIb␣⅐VWF-A1 domain complex (Protein Data Bank code 1SQ0) and previously described methods (11). The panel was constructed using PyMOL (DeLano Scientific LLC, San Carlos, CA). erated by PCR using previously described human/canine chimeras as templates (6) and cloned into F460 vector (13) with puromycin selection. WT/mutant GPIb␣ vectors were transfected into CHO cells stably transfected with GPIb␤ and GPIX (CHO␤IX cells) (6, 7, 14 -16) and selected using WM23 (6). GPV is not required for GPIb-IX expression. Flow cytometry of cells (10 6 /ml) in 0.01 M phosphate, 0.15 M NaCl, pH 7.4, probed with anti-GPIb␣ (WM23, AN51, 6D1, or VM16d) or control (CR1) antibodies (2 g/ml) and fluorescein isothiocyanate-conjugated secondary antibodies (Silenus, Hawthorn, Australia) was performed as described previously (6,7).
Adhesion Assays-Flow-based CHO cell adhesion assays were performed as described elsewhere (14 -16) using glass microslides (VitroCom) coated with 150 g/ml human . Cells in Tyrode's/EDTA buffer (12 mM NaHCO 3 , 10 mM HEPES, 137 mM NaCl, 2.7 mM KCl, 5.5 mM glucose, pH 7.5) containing 2 mM EDTA were perfused at 50 s Ϫ1 followed by incremental increases in shear rate every 30 s. In this assay, cells interact with VWF in a strictly GPIb␣-dependent manner (14 -16) and roll rather than stably adhering to the VWF-coated surface, because the presence of 2 mM EDTA precludes any stationary integrin-dependent adhesion. Once cells have tethered, they continue rolling and do not become stationary. The term "adherent" is used to describe any cell that has tethered to surface-coated VWF from the bulk flow. Cell rolling velocity was analyzed for 25 cells over five separate fields for 30 s at each shear rate. The shear rates used for GPIb␣-expressing CHO cells are based on previous studies and discussed elsewhere (14 -16). Statistical analyses using one-way analysis of variance and Student's unpaired t test, with p Ͻ 0.05 considered significant, were performed using Prism software (Graphpad, San Diego, CA).

RESULTS
To establish the functional role of Leu 60 -Glu 128 of GPIb␣, especially the prominent negative charge patch that this region contributes to the concave face of the repeats (complementary to a positive charge patch on the interacting face of VWF-A1) (1, 9 -11), we expressed human/canine homology swaps of GPIb␣ (Fig. 1A). This approach enables functional analysis of the N-terminal domain of GPIb␣ that is conformationally sensitive and not amenable to analysis by short peptides or scanning mutagenesis (6,7). Furthermore, expressing mutants on CHO cells enables antibody binding and adhesion to VWF under flow to be evaluated. Several lines of evidence confirm that GPIb␣ homology swaps are expressed in a functional form without significant conformational disruption. First, modeling of constructs as described previously (6,11) suggests no significant structural disorder because all core-structural residues are conserved across the species. The models also illustrate conspicuous electrostatic differences in surface charge between human and canine GPIb␣ (Fig. 1B). A negative patch on human GPIb␣ is predominantly centered on Asp 63 and other residues comprising this surface. In canine GPIb␣, Asp 63 is substituted by Arg, which together with His 61 results in loss of the negative patch. There is no direct contact between the human GPIb␣ sequence, Leu 60 -Glu 128 (LRR2-4), and VWF-A1 (Fig. 1B), although the co-crystal structure (10) reveals a single water-mediated contact between Asp 63 and Arg 571 of VWF-A1 (an interaction that is predicted to be abolished in the HUM CAN60 -128 swap). Second, GPIb␣ homology swaps expressed as a GPIb-IX complex on CHO cells are recognized by conformation-sensitive antibodies (6,7,11): epitopes for AN51 mapping to the N-terminal sequence; His 1 -Ile 35 and VM16d mapping to the C-terminal flank sequence; Val 226 -Gly 268 are both present in WT-GPIb␣ and HUM CAN60 -128 ; and 6D1 maps to Leu 104 -Glu 128 , present in WT-GPIb␣ and CAN HUM60 -128 (Fig. 2, cf.  Fig. 1A). WM23, with an epitope C-terminal of Glu 282 , recognizes WT-GPIb␣, HUM CAN60 -128 , and CAN HUM60 -128 confirming equivalent expression on cells used for functional analysis. Third, all of the mutants bind human VWF in the presence of the modulator botrocetin, which does not discriminate between species (6), suggesting that the recep- tors are expressed in a functional form competent to bind ligand (Fig. 3). HUM CAN60 -128 retaining the VWF-contacting residues also retains the capacity to mediate VWF-dependent adhesion at low shear (Fig. 4).
Adhesion of GPIb␣-expressing cells to VWF in a flow chamber, mimicking pathophysiological shear (14 -16), confirms that CHO cells expressing WT-GPIb␣ adhere to human VWF (Fig. 3), whereas CHO␤IX cells or cells expressing canine GPIb␣ do not bind human VWF (6, 7, 14 -16). In contrast, the HUM CAN60 -128 swap (Leu 60 -Glu 128 replaced by canine sequence) showed a substantial decrease in the number of cells adhering from flow at 50 s Ϫ1 and a significantly reduced ability to maintain adhesion with increasing shear (Fig. 4, A and C). This is despite the fact that the two major contact sites for VWF-A1 predicted by crystal structures were preserved. The interaction at low shear suggests that although HUM CAN60 -128 was competent to bind, shear-dependent adhesion was severely compromised without human Leu 60 -Glu 128 (and almost nonexistent at high shear, where WT-GPIb␣ still remained func-tional). HUM CAN60 -128 cells that did roll on VWF rolled 2.5-6-fold faster than WT-GPIb␣; this difference became more pronounced with increasing shear rate (Fig. 4, C  and D). The lack of binding of CAN HUM60 -128 (Fig. 4A) suggests human Leu 60 -Glu 128 alone is not sufficient to support adhesion to VWF. In this regard, the co-crystal structures of GPIb␣ N-terminal domain and VWF-A1 fragments (9, 10) demonstrate that elements flanking LRR2-4 make contact under the static conditions used for crystallography, and these regions are evidently required for optimal VWF recognition at low or high shear. However, the combined functional data show that the relative functional importance of specific structural elements within Leu 60 -Glu 128 increases as the shear force increases.

DISCUSSION
The aim of this study was to reconcile discrepancies between structural and functional analyses of binding of platelet GPIb␣ (the major ligand-binding subunit of GPIb-IX-V) to VWF, an interaction that initiates pathophysiological thrombus formation under shear stress (1)(2)(3)(4)(5). Our previous studies using human/canine chimeras of GPIb␣ suggested that the LRR2-4 sequence, Leu 60 -Glu 128 , is required for GPIb␣-dependent adhesion to VWF under flow conditions (6), whereas co-crystal structures of GPIb␣ and VWF fragments under static conditions reveal interactive sites, predominantly N-and C-terminal to LRR2-4, and a bidentate mode of GPIb␣ binding to ligand (8 -10). All of the chimeras showing impaired VWF binding, however, not only lack human Leu 60 -Glu 128 sequence (6) but also lack either N-or C-terminal contact sites (8,9).
To establish the functional role of Leu 60 -Glu 128 , we expressed human/canine homology swaps of GPIb␣ on CHO cells. Three lines of evidence (molecular modeling, conformation-specific antibody binding, and botrocetin-dependent VWF binding) support the correct folding of the mutant receptor as published previously (6,7); in addition, the HUM CAN60 -128 mutant still supports adhesion to VWF at low shear rates (see above). These functional data, on adhesion of GPIb␣-expressing cells to VWF in a flow chamber mimicking pathophysiological shear (14 -16), however, show how the relative functional importance of elements within Leu 60 -Glu 128 increases as the shear force increases. Binding of fluorescein isothiocyanate-labeled human VWF (5 g/ml, final concentration) to CHO␤IX cells or CHO␣␤IX cells expressing WT or mutant GPIb␣ (10 6 /ml) measured by flow cytometry using a method similar to that reported previously for platelets (18) in the absence (gray histograms) or presence (red histograms) of botrocetin (6). This is not to imply that regions outside LRR2-4 are not important for the interaction of GPIb␣ with VWF but rather that the elements within LRR2-4 become increasingly critical for binding as shear rate increases. The lack of binding of CAN HUM60 -128 to VWF suggests that human Leu 60 -Glu 128 alone is not sufficient to support adhesion to VWF at low or high shear.
The structure of the N-terminal domain of GPIb␣ has been described as a cupped human hand with LRR domains as the palm and fingers, the extended loop of the C-terminal flanking sequence as the thumb, and the anionic/sulfated sequence as the wrist (17). The dimensions of VWF-A1 preclude contact with a ligand-binding surface at the palm of GPIb␣ because of steric hindrance from the thumb, but Uff et al. (17) propose that "the GPIb␣ crystal structure [may represent] the low affinity or "closed" form of the receptor and that a conformational change in the thumb [may be] required to unmask the A1 domain binding site." In contrast, the GPIb␣⅐VWF-A1 complex crystal structure, with only the N-and C-terminal projections (but not the concave surface) in direct contact (8), would appear more consistent with a low-affinity structure postulated by Uff et al. (17). However, the VWF-A1 domain could be satisfactorily docked against the concave surface of GPIb␣ if the "thumb" is moved (17). In the absence of VWF modulators ristocetin and botrocetin, shear stress may provide the necessary impetus to open the structure sufficiently to allow interaction between VWF-A1 and the concave surface of the LRR domain, mediated predominantly by complementary electrostatic patches on GPIb␣ and VWF-A1 (9 -11). This model would predict that although the contact sites may confer specificity, there is an increasing requirement for the electronegative surface encompassing Leu 60 -Glu 128 as the shear rate increases, consistent with the functional data (6) (Fig. 4). Definitive resolution of the contact surface between receptor and ligand will undoubtedly require further structural information. However, in the context of shear this will be almost impossible to achieve using current technologies. Instead, the functional analysis of homology domain swaps of GPIb␣ identifies shear-dependent regions of the receptor involved in binding VWF.