Modeling and Functional Analysis of the Interaction between von Willebrand Factor A1 Domain and Glycoprotein Ib a *

Binding of the von Willebrand factor (vWF) A1 domain to the glycoprotein (GP) Ib-IX-V complex mediates platelet adhesion to reactive substrates under high shear stress conditions, a key event in hemostasis and thrombosis. We have now used the known three-dimensional structure of the A1 domain to model the interaction with the GP Ib a sequence 271–279, which has previously been implicated in ligand binding. Docking procedures suggested that A1 domain residues in strand b 3 and preceding loop (residues 559–566) as well as in helix a 3 (resi-dues 594–603) interact with Asp residues 272, 274, 277 and sulfated Tyr residues 278 and 279 in GP Ib a . To verify this model, 14 mutant A1 domain fragments containing single or multiple side chain substitutions were tested for their ability to mediate platelet adhesion under flow. Each of the vWF residues Tyr 565 , Glu 596 , and Lys 599 proved to be strictly required for A1 domain function, which, in agreement with previous findings, was also dependent on Gly 561 . Moreover, an accessory functional role was apparent for a group of positively charged residues, including Arg at positions 629, 632, 636 and Lys at positions 643 and 645, possibly acting in concert. There was, however, no evidence from the model that these residues a small random displacement was applied to each of the internal and external degrees of freedom of the peptide, involving translation of its center of gravity, orientation with respect to the vWF A1 domain, and rotation around each of its flexible dihedral angles. The energy resulting from the new conformation of the peptide and its interaction with the A1 domain was evaluated. This value was accepted as valid if lower than the one obtained in the preceding cycle of the simulation, resulting in a final conformation with the lowest energy of interaction between the GP Ib a peptide and the vWF A1 domain. Intermolecular contacts CONTACSYM using van der Waals radii Cut-off defining van der Waals a 1.7 Å probe radius.

Binding of von Willebrand factor (vWF) 1 to the platelet membrane glycoprotein (GP) Ib␣, a component of the GP Ib-IX-V complex (1), supports the arrest of bleeding after tissue trauma (2,3) and may also contribute to the acute thrombotic occlusion of diseased arteries (4,5). The interaction between GP Ib␣ and immobilized vWF can tether platelets to surfaces in rapidly flowing blood but is inherently short-lived, mediating transient periods of arrest that alternate with slow translocation (6). During this rolling motion, however, additional synergistic bonds can form through specific integrin receptors and establish firm platelet adhesion to extravascular matrix constituents (6,7). The rapid association of GP Ib␣ with vWF not only initiates platelet deposition at sites of vessel injury but is also essential for maintaining thrombus growth in blood flowing with high shear rates (8). A detailed explanation of the molecular mechanisms underlying these processes may, therefore, elucidate key regulatory aspects of normal and pathological platelet function as well as indicate new approaches for antithrombotic therapy.
The notion that the GP Ib␣ binding site is located between vWF residues 449 and 728 (9,10), comprising the entire A1 domain and the preceding carboxyl-terminal portion of domain D3 (11,12), was established over a decade ago. Since then, studies attempting to identify A1 domain residues involved in interacting with the receptor have yielded inconclusive and sometimes conflicting results (13)(14)(15)(16). One reason for this outcome is that the assays employed to evaluate the association between vWF and GP Ib␣ have typically used exogenous modulators such as ristocetin (17) and botrocetin (18), which possess unique mechanisms of action (19 -22) but are of uncertain significance with respect to biologically relevant events (6,7,23). Recently, however, evidence has been presented that a recombinant A1 domain fragment can function as intact vWF in tethering rolling platelets to a surface in a flow field with high shear rate (23). This method may be used to evaluate the consequences of specific vWF mutations without the confounding presence of exogenous modulators and, complemented by modeling procedures based on the available A1 domain crystal structure (24), may provide initial information on individual residues involved in receptor binding. Such an approach is valuable since the atomic structure of functional GP Ib␣ and vWF fragments in complex is not available. Here we present the results of a study based on docking a limited GP Ib␣ sequence, previously shown to contain residues involved in ligand binding (25)(26)(27), onto the A1 domain three-dimensional structure. Selected vWF residues thus predicted to establish contact with the receptor were mutated in recombinant A1 domain fragments that were then tested for the ability to support platelet tethering in flowing blood. The functional changes caused by these mutations define the location of a putative GP Ib␣ binding interface in vWF. The GP Ib␣ sequence GDTDLYDYY was selected for docking onto the vWF A1 domain because of its experimentally proven involvement in ligand binding (25)(26)(27). Molecules were constructed and visualized with INSIGHTII (release 95.0 with Builder, Biopolymer, and Discover modules; Biosym/Molecular Simulations Inc.). The presumed conformation of the YDYY sequence in the GP Ib␣ peptide was derived from analysis of known structures in the Protein Data Bank data base, namely 1GOX (residues [22][23][24][25], 1HYT (residues 81-84), and 1NPC (residues 82-85). The three-dimensional conformation of the YDYY sequence in these proteins was similar, as shown by a root mean square deviation in the position of all atoms of less than 0.3 Å. The starting coordinates for the docking procedure were taken from 1GOX. The remaining portion of the GP Ib␣ peptide was built one residue at a time, and the conformation was optimized by energy minimization using X-PLOR (28). In these calculations, each new peptide unit was treated as completely flexible, whereas only 2-4 torsion angles at a time were allowed to vary in the YDYY sequence. In subsequent cycles, torsion angles that had been kept fixed were allowed to vary. The choice of an initial conformation for the YDYY portion of the GP Ib␣ peptide reduced the complexity of calculations, but no portion of the peptide was treated as rigid, thus eliminating or reducing bias. The three Tyr residues in the sequence were constructed both with and without sulfation. The latter posttranslational modification is thought to occur on all three residues in the native protein and to be necessary for normal function (27,29). The geometry of the three sulfated tyrosine residues was refined using X-PLOR (28). Docking of the GP Ib␣ peptide onto vWF was performed with AUTODOCK, release 2.4 (30). This program uses Monte Carlo simulated annealing procedures to explore a wide range of conformational states. In our case, the vWF A1 domain molecule was kept fixed throughout the simulation while the GP Ib␣ peptide was subjected to random movement in the space around the protein accompanied by changes in its conformation. At each step in the simulation, a small random displacement was applied to each of the internal and external degrees of freedom of the peptide, involving translation of its center of gravity, orientation with respect to the vWF A1 domain, and rotation around each of its flexible dihedral angles. The energy resulting from the new conformation of the peptide and its interaction with the A1 domain was evaluated. This value was accepted as valid if lower than the one obtained in the preceding cycle of the simulation, resulting in a final conformation with the lowest energy of interaction between the GP Ib␣ peptide and the vWF A1 domain. Intermolecular contacts were generated with CONTACSYM (31) using extended van der Waals radii (32). Cut-off distances of 3.4 Å and 4.33 Å were used for defining hydrogen bonds and van der Waals interactions, respectively. The molecular surface was calculated (33) as described previously using a 1.7 Å probe radius.

Modeling of the vWF-GP Ib␣
Expression and Purification of Recombinant Fragments Containing the vWF A1 Domain-Recombinant polypeptides corresponding to residues 445-733 of the mature vWF subunit, designated rvWF 445-733 , were expressed in host Escherichia coli BL21-DE3 as described previously in detail (34 -36). All vWF fragments used in these studies had Cys 3 Gly mutations at positions 459, 462, 464, 471, and 474 to prevent formation of random aggregates during purification (23,34). These and additional amino acid substitutions described under "Results" were obtained by site-directed mutagenesis using oligonucleotide primers with the desired nucleotide change (34). The occurrence of each mutation was verified by sequence analysis. Expressed fragments were purified by reverse-phase high performance liquid chromatography and refolded after oxidation of the two Cys residues at positions 509 and 695 to create the intramolecular disulfide bond that exists in native vWF (23,37). Purified fragments were dialyzed against 2 mM acetic acid titrated to pH 3.5 with HCl and stored at Ϫ70°C. Refolding before use was achieved by slow dialysis with incremental pH increase in steps of 0.5 units each, up to a final value of 5.0 (23,34). Protein concentration was determined with the micro-BCA assay (Pierce) according to the instructions of the manufacturer. The molecular mass of each purified protein was verified by mass spectroscopy and found to be in accordance with the predicted sequence. All fragments had comparable behavior upon analysis by SDS-polyacrylamide gel electrophoresis under reducing and nonreducing conditions, indicating that they contained the Cys 509 -Cys 695 intrachain disulfide bond (34). They all migrated as a single band with apparent molecular mass of 39 kDa in oxidized form and were soluble up to a concentration of at least 0.5 mg/ml.
Functional Evaluation of the Interaction between vWF A1 Domain and GP Ib␣-Wild type and mutant recombinant vWF fragments were immobilized onto glass coverslips using solutions in 20 mM HEPES, 150 mM NaCl, pH 7.4. The coating protein concentration was 100 g/ml and was previously shown to ensure saturation of the surface with all the fragments tested (23). Blood was collected from healthy, medicationfree donors into polypropylene syringes containing the ␣-thrombin inhibitor D-phenylalanyl-L-prolyl-L-arginine chloromethyl ketone dihydrochloride (PPACK) as anticoagulant (50 M final concentration). All human subjects participating in these studies were aware of the experimental nature of the research and gave their informed consent in accordance with the Declaration of Helsinki. To eliminate potential confounding effects of plasma vWF, perfusion studies were performed with blood cells removed from plasma and resuspended in an equivalent volume of modified HEPES-Tyrode buffer, pH 7.4, containing 50 mg/ml bovine serum albumin fraction V (Calbiochem) and 5 mM EDTA (23). Moreover, the platelet count in the cell suspension was lowered to a value between 9,000 and 12,000/l to decrease interacting platelets on the surface and facilitate quantitative image analysis. This was achieved by centrifuging the cell suspension at 150 ϫ g for 15 min, removing the platelet-rich supernatant fluid, and replacing it with an equivalent volume of HEPES-Tyrode buffer. Platelets were then counted in both the whole cell and the platelet-rich suspensions, and appropriate volumes of the latter were added into the former to obtain the target count in the reconstituted blood, at the same time maintaining a normal hematocrit. Alternatively, if the platelet count was still too high, the centrifugation step was repeated. Interactions between surface-bound vWF A1 domain and platelet GP Ib␣ in flowing blood were observed in real-time by epifluorescence videomicroscopy using an Axiovert 135 inverted microscope with a 40ϫ objective (Zeiss) after labeling blood cells with 10 m of mepacrine (quinacrine hydrochloride) (6,23). Note that the fluorescence of erythrocytes is blocked by hemoglobin so that these cells are not visible in the background. Leukocytes, on the other hand, are brightly fluorescent but fail to interact with the surface under the chosen experimental conditions. Reconstituted blood was aspirated through a parallel plate chamber by a syringe pump (Harvard Bioscience) at the flow rate of 0.484 ml/min, producing a wall shear rate of 1500 s Ϫ1 at the inlet. The distance between the plates of the chamber was 147 m and was maintained by a gasket designed to produce variable shear rates along the direction of flow (6,38). All experiments were recorded on standard videotape with a Sony 9500 video cassette recorder. The interaction of selected mutant and wild type A1 domain fragments with the monoclonal antibody NMC-4 was evaluated by enzyme-linked immunoadsorbant assay (ELISA), following a procedure described elsewhere (39,40) with minor modifications.
Image Analysis-The number of individual platelets interacting with immobilized recombinant vWF fragments was measured on images obtained at selected wall shear rates (6). Each image corresponded to a single frame from real time (30 frames/s) videotape recordings, digitized and processed by computer analysis using a Matrox Image LC frame grabber and the MetaMorph software package (Universal Imaging Corp.). After 1 min from the beginning of blood flow and over the ensuing 4 min, a stack of 10 images captured randomly from the tape at 2-5-s intervals was created for each of the shear rates tested (340 -630 and 1210 -1500 s Ϫ1 ). Background fluorescence was defined as the average of 15 images of the surface without platelets in focus and was subtracted from all processed images. After adjusting brightness and contrast, an arbitrary threshold was applied to distinguish platelets from noise, and the images were binarized. Adherent platelets, regardless of the duration of their transient interaction with immobilized vWF A1 domain, were counted on an area of 25,600 m 2 .

Modeling of the vWF A1 Domain-GP Ib␣ Interaction-
The conformation of the GP Ib␣ peptide GDTDLY(SO 4 ) DY(SO 4 )Y(SO 4 ) (residues 271-279 in the mature protein) exhibiting the lowest energy of interaction with the vWF A1 domain is shown in Fig. 1. Interatomic distances between GP Ib␣ and A1 domain residues allowed the identification of putative interactions and the definition of their nature, according to established parameters (Table I). Residues Asp 560 , His 563 , Tyr 565 , and particularly Glu 596 and Lys 599 in the A1 domain appeared to play a prominent role in contacting the receptor (Fig. 2). This finding highlights the possible participation of strand ␤3 and preceding loop (residues 559 -566) as well as helix ␣3 (residues 594 -603) in mediating platelet adhesion. The main GP Ib␣ residues involved in forming a complex with the A1 domain were Asp 277 , Tyr 278 , and Tyr 279 , with a notable contribution by the sulfate group of Tyr 278 and additional participation of Gly 271 , Asp 272 , and Asp 274 (Table I and Fig. 2). Of note, the 271-279 GP Ib␣ peptide containing nonsulfated Tyr residues could also be docked onto the A1 domain with relatively low energy of interaction but not as low as measured with the peptide containing sulfotyrosine (not shown). Moreover, the nonsulfated peptide appeared to interact predominantly with the side chain of Arg 629 in the A1 domain (Fig. 2), shown not to be directly required for binding to GP Ib␣ in functional assays under flow conditions (see below). The buried surface areas in the complex were 418 Å 2 for the A1 domain and 353 Å 2 for the docked peptide.
Functional Analysis of the vWF A1 Domain-GP Ib␣ Interaction-Recombinant vWF fragments comprising residues 445-733 of the mature subunit and containing the A1 domain were expressed with single or multiple mutations of residues selected among those that appeared to be involved in the modeled interaction with GP Ib␣ (Table I). Additional residues previously shown to be involved in binding to the function blocking antibody NMC-4 (24) or thought to participate in modulatorinduced vWF binding to GP Ib␣ (24,41) were also mutated. In total, 14 different mutant fragments in addition to the one with native sequence were individually tested after immobilization onto a glass surface for their ability to interact with platelets when exposed to blood cells flowing in a plasma-free suspension at wall shear rates between 340 and 1500 s Ϫ1 . After coating onto glass, the presence of all fragments on the surface was confirmed by reactivity with specific monoclonal antibodies detected by enzyme-linked immunoassay (42). In agreement The peptide backbone appears as a thin tube to which side chains are attached. Atoms, except hydrogen, are shown as small spheres connected by short sticks. Tyrosine residues are sulfated. The conformation of this sequence results from the modeling procedure. The A1 domain structure, based on x-ray crystallographic data (24), is presented as a ribbon diagram, with ␣-helices in purple (except for helix ␣3, shown in blue, and helix ␣4, shown in red) and ␤-strands in light green (except for strand ␤3, shown in blue). Strand ␤3 and helix ␣3 contain most of the residues that form the proposed GP Ib␣ binding site; helix ␣4 contains the residues that interact with the function blocking antibody, NMC-4. Loops connecting ␣-helices and ␤-strands are represented with thin tubes. N and C indicate the amino and carboxyl terminus, respectively, of the recombinant vWF fragment utilized for crystallographic studies, corresponding to residues 508 (N) to 709 (C).

FIG. 2. Residues involved in the modeled vWF-GP Ib␣ interaction.
The GP Ib␣ sequence is presented in orange, as described in the legend to Fig. 1. The side chains of residues interacting with the A1 domain are depicted and numbered with their position in the native protein. A limited portion of the A1 domain structure is shown, with ␣-helices in red, ␤-strands in light green, and connecting loops in purple. This segment begins with a portion of the loop between strands ␤2 and ␤3 (G 561 ) and continues with strand ␤3, helix ␣2, helix ␣3, strand ␤4, helix ␣4 as well as intervening loops. The vWF A1 domain side chains are colored in blue (residues in strand ␤3) or purple (residues in helix ␣4) and numbered according to the sequence of the mature protein (24). All the A1 domain side chains shown in the figure have been mutated to document their functional role in GP Ib␣ binding (see Fig. 3) or antibody NMC-4 binding (see Fig. 4). Note that the orientation of the GP Ib␣ side chains is based on the results of modeling, whereas that of the A1 domain side chains is based on x-ray crystallographic data.
with previous results, the fragment with native sequence supported a characteristic rolling adhesion of platelets as typically mediated by multimeric vWF with similar shear rate dependence (23). For the purpose of these studies, all platelets in contact with the surface, regardless of velocity of rolling and duration of interaction, were considered as adherent. At shear rates between 1500 and 1210 s Ϫ1 , 7 of the 14 mutant A1 domain molecules supported platelet adhesion with at least 50% lower efficiency that the native control, and all with the exception of one showed a similar defect at 630 -340 s Ϫ1 (Fig.  3A). Five of the mutants contained single-residue substitutions, and two had multiple substitutions of positively charged residues. The corresponding side chains are depicted in Fig. 2. Platelet adhesion to the remaining 7 mutant A1 domain fragments was 80% or more of that supported by the native fragment at least at one of the shear rates tested (Fig. 3B). Analysis of variance by both parametric and nonparametric (Kruskal-Wallis) tests indicated that the results obtained with distinct mutant A1 domain fragments were significantly different (p Ͻ 0.001 at 1500 -1210 s Ϫ1 ; p Ͻ 0.002 at 630 -340 s Ϫ1 ). Comparisons between mutants by t test with Bonferroni correction showed that none of the single substitutions R629A, R632A, R636A, K644A, H656A, K660A, and R663A caused significant differences in platelet adhesion at the shear rate of 1500 -1210 s Ϫ1 (p Ͼ 0.05; Fig. 3B). In contrast, each of the single substitutions G561S, Y565A, E596A, K599A as well as the multiple substitutions R629A,R632A,R636A and K643A,K645A resulted in significantly lower platelet adhesion (p Ͻ 0.05; Fig. 3). Moreover, the single substitutions G561S, Y565A, E596A, and K599A caused significantly lower platelet adhesion also at the shear rate of 630 -340 s Ϫ1 .
Of note, the four single-residue mutations that essentially caused complete loss of A1 domain function (Fig. 3A) included the substitution of two residues, Glu 596 and Lys 599 , with a putative key role in the modeled interaction with GP Ib␣ (Fig.  2). The other two mutations with such a drastic effect on function were Tyr 565 3 Ala and Gly 561 3 Ser, of which the latter is the reported cause of a variant form of von Willebrand disease (43). Single mutations of the three Arg residues at positions 629, 632, and 636 in helix ␣4 had no relevant effect on platelet adhesion (Fig. 3B), but the concurrent substitution of all three resulted in markedly decreased adhesion (Fig. 3A). In contrast to their apparently negligible participation in GP Ib␣ binding, Arg 632 and Arg 636 were individually necessary for interacting with the function-blocking monoclonal antibody NMC-4 (Fig. 4), as anticipated on the basis of the solved structure of the latter in complex with the vWF A1 domain (24). DISCUSSION We have identified a putative GP Ib␣ binding site in vWF using the crystal structure of the A1 domain to model the interaction with the receptor and verifying experimentally the predicted role of specific vWF residues in mediating adhesion of flowing platelets. Previous studies using synthetic peptides (25,27) and site-directed mutagenesis of recombinant fragments (26,27,29) provide evidence that the GP Ib␣ sequence selected for the docking procedure described here contains residues with a role in ligand binding. In this regard, it has already been shown that a peptide corresponding to the GP Ib␣ sequence 251-279 can interfere with vWF binding to platelets Reconstituted blood with mepacrine-labeled platelets was perfused at 37°C through a Hele-Shaw chamber at a constant flow rate, such that the indicated wall shear rates were attained at set positions on the x, y axes (see "Experimental Procedures"). The bottom of the chamber was formed by a glass coverslip coated with recombinant vWF fragment. The number of surface-interacting platelets was counted between 1 and 5 min from the beginning of flow on single frames from real time recording and represents instantaneous (1/30th of a second) surface coverage by platelets that were mostly transiently interacting with the immobilized A1 domain. In each frame, platelets were counted in an area of 25,600 m 2 . To analyze a representative portion of the entire surface exposed to flowing blood, 10 distinct frames were evaluated in each experiment. Because the wall shear rate in the chamber used varied continuously at different x, y positions, results obtained between 1500 and 1210 s Ϫ1 and between 630 and 340 were grouped for presentation. The results for mutant vWF fragments are expressed as percentage relative to the number of platelets adhering to wild type fragment using blood from the same donor tested on the same experimental day. The average number of platelets interacting with wild type vWF fragment was 57 at 1500 -1210 s Ϫ1 and 50 at 630 -340 s Ϫ1 (n ϭ 11). The results shown represent the mean with S.E. of the mean of between 2 and 4 separate experiments performed with each mutant fragment.

FIG. 4. Binding of the monoclonal antibody NMC-4 to recombinant vWF fragments representing residues 445-733.
A standard enzyme-linked immunoadsorbant assay was used for these experiments in which the recombinant fragment had either normal sequence (wild type) or the indicated single or multiple mutations. The monovalent Fab fragment of the antibody was used in these studies (45). The results shown are expressed as percentage of the highest absorbance value obtained with the wild type fragment on any give experimental day and represent the mean with S.E. of the mean of two separate experiments. mediated by ristocetin or botrocetin, whereas a peptide with sequence 271-285 has less inhibitory activity (25). However, only the entire 45-kDa amino-terminal GP Ib␣ fragment with intact disulfide bonds could block the direct binding of asialo-vWF to platelets (25), indicating that the vWF-GP Ib␣ interaction in the absence of modulators may depend on conformational attributes that cannot be retained by small synthetic peptides. Such a requirement for function appears to be in addition to sulfation of the Tyr residues at positions 276, 278, and 279, necessary to attain maximal ligand binding efficiency of GP Ib␣ (27,29). Thus, we did not attempt to verify whether a synthetic peptide containing the GP Ib␣ sequence 271-279 modeled here could inhibit platelet adhesion to immobilized vWF under flow. On the other hand, in the absence of information on the three-dimensional structure of GP Ib␣, we elected to perform the docking studies with a short sequence to reduce the possibility of bias in the results. As a consequence, we may have obtained only a partial view of the receptor binding site in the A1 domain.
The recombinant vWF fragment used for functional evaluation contains the A1 domain with the intrachain disulfide bond between Cys 509 and Cys 695 as well as the carboxyl-terminal region of the preceding domain D3 with 5 Cys residues mutated to Ala. We have previously demonstrated (23) that such a fragment exhibits a functional behavior comparable with that of A1 domain in multimeric vWF. Thus, after immobilization onto a surface, it can tether platelets and mediate rolling, but in solution, it fails to inhibit the function of GP Ib␣ even when present at relatively high concentrations unless an exogenous modulator is added (23). Findings of this type have suggested that the A1 domain undergoes a conformational change upon becoming insolubilized onto a surface, and in addition or in alternative, is modified by shear forces to achieve the functional conformation required for interaction with GP Ib␣. There are, however, alternative explanations as to why soluble A1 domain cannot interfere with the function of immobilized A1 domain in the absence of modulators. For example, since the interaction with GP Ib␣ is rapidly reversible (6), the local high density achieved after binding to a surface may make competition by molecules in solution practically impossible to demonstrate because of limits to the concentrations that can be achieved in blood. Nevertheless, it cannot be excluded that the conformation provided by the known crystal structure of the A1 domain (24,44), which reflects that of the soluble protein, may not correspond in all details to the functional conformation capable of supporting GP Ib␣ binding. Even with this caveat, it appears that the docking procedure described here can correctly predict a number of A1 domain residues likely to play a direct role in supporting the interaction with GP Ib␣.
Our findings, in agreement with the hypothesis suggested by analysis of the A1 domain crystal structure (24), indicate that the GP Ib␣ binding interface is centered around a surface groove formed by helix ␣3 with neighboring strand ␤3 and helix ␣4. In this regard, the possible involvement of Glu 596 and Lys 599 (helix ␣3) as well as of Tyr 637 (helix ␣4) in GP Ib␣ binding has already been suggested (24,41), but the crucial role of Tyr 565 (strand ␤3) has been unknown so far. The latter residue appears to be absolutely required for normal vWF function, since removal of its side chain results in a totally inactive A1 domain. On the other hand, the present results argue against the proposed participation of Lys 644 in receptor binding (24). Of note, mutations of two residues predicted by our model to participate in A1 domain function, Gly 561 and Glu 596 , are known to cause decreased vWF binding to platelets in Type 2M von Willebrand disease (3). With respect to GP Ib␣, our findings agree with mutagenesis studies (29) and support the participation of sulfate groups on Tyr 278 and Tyr 279 in vWF binding (27). In the modeled interaction, each of these Tyr residues forms multiple contacts with Glu 596 , Lys 599 , and other residues in the A1 domain, perhaps explaining why only the concurrent obliteration of more than one sulfate group causes abnormal GP Ib␣ function (27). The latter conclusion, however, is based on the results of assays dependent on exogenous modulators and must be verified with the study of platelet adhesion under flow conditions. An attempt to explain the mechanism of vWF association with GP Ib␣ has previously been based on alanine-scanning mutagenesis of charged residues in the A1 domain combined with the measurement of soluble ligand binding to platelets in the presence of exogenous modulators (41). The results of those studies correspond at least in part to the findings presented here, notably with respect to the key functional role played by Glu 596 and Lys 599 . In contrast, the suggested participation of Arg 632 and possibly Arg 629 in receptor binding appears to be contradicted by our observation that single substitutions of these residues have limited effect on the tethering of platelets to the A1 domain. This was certainly true at the higher shear rates tested, although in both instances there was a tendency to partially decreased function at lower shear rates (Fig. 3B). It is possible, therefore, that neither Arg 632 nor Arg 629 are directly required for the association between immobilized A1 domain and GP Ib␣. These residues, however, may participate in interactions dependent on the activity of exogenous modulators (41). Whether the latter have any counterpart in vivo is still questionable; thus, the physiological significance of the finding is uncertain, but it is intriguing to speculate that the results obtained with exogenous modulators may partly reflect vWF and GP Ib␣ function at lower but not higher shear rates. Such a consideration may be taken to indicate that shear forces directly modulate this ligand-receptor interaction, with possible effects on the conformation of relevant residues in vWF, GP Ib␣, or both. We did not evaluate the involvement in GP Ib␣ binding of charged residues in strand ␤4 (Arg 616 ) and the preceding loop (Glu 613 ), both of which were indicated as possible participants in modulator-dependent vWF activity (41). The known A1 domain three-dimensional structure (24) shows that the side chains of these residues are oriented away from the receptor binding site defined by our modeling and functional studies. This observation, however, does not exclude the possibility that they may contribute to interactions with residues in GP Ib␣ other than the ones considered here.
It is of interest that the combined substitution of Arg 629 , Arg 632 , and Arg 636 with Ala, but not of each individual residue, resulted in markedly reduced platelet adhesion to immobilized A1 domain, as did the combined substitution of Lys 643 and Lys 645 . The side chains of these charged residues, as shown by the three-dimensional structure of the A1 domain (24), are oriented away from the proposed GP Ib␣ binding site involving mainly strand ␤3 and helix ␣3. In the crystal structure, moreover, Lys 643 and Lys 645 mediate a homodimeric interaction between adjacent A1 domain molecules. These charged residues on the surface of the molecule, therefore, may influence A1 domain conformation and function without participating directly in establishing contact with GP Ib␣. Of note, binding to the A1 domain of the function-blocking antibody NMC-4 involves mainly the Arg residues at positions 632 and 636, as shown previously by the solved crystal structure of the complex (24) and confirmed here by functional studies, neither of which is necessary for interaction with GP Ib␣. Thus, the inhibitory activity of the antibody may be mainly a consequence of steric hindrance. These considerations are in agreement with the concept that two contiguous but distinct sites in the A1 domain are involved in interacting with the antibody and GP Ib␣, respectively.
In conclusion, our findings provide three-dimensional modeling and functional data that define a region in the vWF A1 domain necessary for binding to GP Ib␣ on flowing platelets. These results set the stage for future work aimed at the definitive elucidation of the molecular mechanisms that regulate the association of this ligand-receptor pair and control its key influence on hemostasis and arterial thrombosis.