Identification of the amino acid residues of the platelet glycoprotein Ib (GPIb) essential for the von Willebrand factor binding by clustered charged-to-alanine scanning mutagenesis.

At the site of vascular injury, von Willebrand factor (VWF) mediates platelet adhesion to subendothelial connective tissue through binding to the N-terminal domain of the alpha chain of platelet glycoprotein Ib (GPIbalpha). To elucidate the molecular mechanisms of the binding, we have employed charged-to-alanine scanning mutagenesis of the soluble fragment containing the N-terminal 287 amino acids of GPIbalpha. Sixty-two charged amino acids were changed singly or in small clusters, and 38 mutant constructs were expressed in the supernatant of 293T cells. Each mutant was assayed for binding to several monoclonal antibodies for human GPIbalpha and for ristocetin-induced and botrocetin-induced binding of 125I-labeled human VWF. Mutations at Glu128, Glu172, and Asp175 specifically decreased both ristocetin- and botrocetin-induced VWF binding, suggesting that these sites are important for VWF binding of platelet GPIb. Monoclonal antibody 6D1 inhibited ristocetin- and botrocetin-induced VWF binding, and a mutation at Glu125 specifically reduced the binding to 6D1. In contrast, antibody HPL7 had no effect for VWF binding, and mutant E121A reduced the HPL7 binding. Mutations at His12 and Glu14 decreased the ristocetin-induced VWF binding with normal botrocetin-induced binding. Crystallographic modeling of the VWF-GPIbalpha complex indicated that Glu128 and Asp175 form VWF binding sites; the binding of 6D1 to Glu125 interrupts the VWF binding of Glu128, but HPL7 binding to Glu121 has no effect on VWF binding. Moreover, His12 and Glu14 contact with Glu613 and Arg571 of VWF A1 domain, whose mutations had shown similar phenotype. These findings indicated the novel binding sites required for VWF binding of human GPIbalpha.

Platelet adhesion is an initial step in thrombus formation and is dependent on the binding of the glycoprotein (GP) 1 Ib-V-IX complex on the platelet surface to von Willebrand factor (VWF) (1,2). VWF has been demonstrated to bind to the extracellular N-terminal domain of the ␣ chain of GPIb (GPIb␣) (3), but the precise molecular mechanisms responsible for VWF-GPIb␣ binding remain to be elucidated.
The VWF binding domain of human GPIb␣ consists of the N-terminal flank region, domains of leucine-rich repeats (LRRs), the C-terminal flank region, the anionic sulfated tyrosine region, and the macroglycopeptide region (4). There are two potential N-glycosylation sites, Asn 21 and Asn 159 , 2 which may not affect the VWF binding in the absence (5) or in the presence of ristocetin and botrocetin (6,7). By several experiments using synthetic oligopeptides, VWF binding sites first have been assigned in the sulfated tyrosine region between Asp 235 and Tyr 279 (8 -10).
GPIb␣ contains 7-8 LRRs (4), and the LRR motif often works as the functional domain for the ligand binding in many proteins (11)(12)(13)(14). However, few studies have been conducted for the LRR domain of GPIb␣. Because of the complex tertiary structure of LRRs, functional analysis conserving the native folding of GPIb␣ should be conducted to estimate the role of the LRR domain of GPIb␣.
In this study, we perform charged-to-alanine scanning mutagenesis of the N-terminal fragment between His 1 and Asp 287 of human GPIb␣ and analyze the VWF binding to mutant GPIb␣ fragments in the presence of ristocetin or botrocetin. We also study the epitopes of several monoclonal antibodies (MAbs) for GPIb␣ that inhibit the VWF binding. Finally, the role of LRRs and of the novel N-and C-terminal regions of the VWF binding domain of GPIb␣ is interpreted by the crystallographic observation of the position of the mutated amino acids.

EXPERIMENTAL PROCEDURES
Materials-Vent DNA polymerase and restriction endonucleases were purchased from New England Biolabs (Beverly, MA). Ristocetin was obtained from Helena Laboratory (Beaumont, TX). Purified botrocetin was purchased from Pentapharm (Basel, Switzerland). Polyclonal antibody MWCL-1 was developed against human platelet GPIb. GPIb␣-CaM is a chimeric protein consisting of amino acid residues His 1 -Val 289 of GPIb, two Ala residues, and calmodulin and was purified as described previously (15). Both were kindly provided by Chester Q. Li and Mark R. Wardell (Washington University, St. Louis, MO). MAb 6D1 against human platelet GPIb␣ was kindly provided by Dr. Barry Collar, Mt.
Sinai Medical Center, New York. MAb HPL7 was purchased from Sapphire (Alexandria, Australia), AN51 from DAKO (Carpinteria, CA), and SZ2 from Immunotech (Westbrook, ME). Polyclonal anti-human VWF antibodies D-082 and P-226 were obtained from DAKO. Anti-FLAG M2 monoclonal antibody and FLAG-bacterial alkaline phosphatase were obtained from Sigma. Purified VWF was kindly provided by Dr. J. Evan Sadler (Washington University). ECL-Western blotting detection reagents, Hyperfilm-ECL, Na 125 I, and peroxidase-conjugated goat anti-mouse immunoglobulin were from Amersham Biosciences. LipofectAMINE and other cell culture reagents were purchased from Invitrogen. IODO-GEN was from Pierce. Fixed human lyophilized platelets were from Biodata, Hatboro, PA.
Plasmid Construct-The full-length cDNA for human GPIb␣ was kindly provided by Dr. Gerald J. Roth (University of Washington, Seattle, WA). The EcoRI fragment containing the signal sequences and the full coding sequence of human GPIb␣ was cloned into pBluescriptII/ KS(ϩ), and the 1-kb EcoRI/XbaI fragment containing amino acid residues 1-318 of GPIb␣ was constructed (pBS/hGPIb/EX). To this template, a primer pair was designed: G652, CCAAAGGGCTTTTTTG-GGTC (652-671 of human GPIb␣ cDNA (16)); and GFLAGa, TGTCT-AGATCACTTGTCATCGTCGTCCTTGTAGTCGGCTTTGGTGGGGAA-CTTGAC (976 -1031; underlined sequence represents FLAG, stop codon, and an XbaI site). The amplified fragment was digested by PstI and XbaI, and the resulting 280-bp fragment was inserted into PstIand XbaI-digested pBS/hGPIb/EX. Thus, in the resulting plasmid pBS/ hGPIbFLAG, the corresponding amino acid residues 303-318 were replaced by FLAG sequence (DYKDDDDK) followed by the stop codon TGA. Plasmid pBS/hGPIbFLAG was used as a template for further alanine scanning mutagenesis by using a PCR-based method as described earlier (17). Plasmid pBS/hGPIbFLAG has unique HindIII, BsmBI, PstI, and NotI sites that enable mutagenesis of 287 amino acids by a cassette replacement method. Thirty-eight mutated GPIb fragments were produced and cloned into pBS/hGPIbFLAG by using two of these four enzymes. By the automated sequencer (Applied Biosystems 310, Foster City, CA), DNA sequence analysis was performed for ϳ150 -200 bp of the amplified fragments between each enzyme site. Then, the HindIII/NotI fragment was cloned into pcDNAI/neo (Invitrogen), and the expression plasmid pcPNAI/hGPIbFLAG was produced. It contained the N-terminal fragment between Met 16 and Ala 302 of human GPIb␣ and would be expressed as a soluble fragment containing a FLAG tag at the C-terminal end.
Expression and Characterization of Recombinant GPIb␣ Chain-Human 293T cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Invitrogen). Cells were transfected with pcDNAI/hGPIbFLAG containing either wild type or mutant GPIb␣ by using LipofectAMINE according to the manufacturer's recommendation. In mock transfection, pcDNAI/neo was used. Twenty-four h after transfection, cells were washed with phosphatebuffered saline (PBS) and then incubated with the serum-free medium (Optimem-1, Invitrogen). After 48 h, recombinant GPIb␣ (rGPIb␣) secreted in the medium was concentrated using Centriprep-30 and Centricon-100 devices (Millipore, Billerica, MA). The concentration of each mutant was determined by the two-step ELISA. For the first ELISA, diluted MAb 6D1 (7.5 g/ml) in bicarbonate buffer (pH 9.6) was coated onto microtiter plates with U-shape bottoms (Costar, Cambridge, MA) at 4°C overnight. The wells were washed five times with PBS containing 0.1% Tween 20 (PBST) and then incubated for 120 min at room temperature with 15 l of rGPIb␣ diluted in PBS containing 3% bovine serum albumin (BSA). The wells were washed again and incubated with diluted MWCL-1 for 90 min at room temperature. Washed wells were then incubated with diluted peroxidase-conjugated goat anti-rabbit IgG (Amersham Biosciences) followed by the color development of the absorbance at 490 nm. GPIb␣-CaM (7) was used as ELISA standard. After the first ELISA, the cell supernatant was concentrated, and PBS containing 0.5% BSA was used to dilute to 10 mg/ml. The sample was subjected to the second ELISA using anti-FLAG M2 MAb as the capturing antibody, and then the amount of each mutant was arbitrarily determined by comparing with wild type rGPIb␣ with known concentrations. The two ELISAs were performed in four distinct experiments. Mutants were aliquoted for the following ristocetin-induced or botrocetin-induced binding assays, and each aliquot was used for one experiment. In the VWF binding assays, an aliquot was also subjected to the simultaneous ELISA (same as the second ELISA). This additional ELISA was performed in duplicate. In the first ELISA, several mutants showed remarkably low absorbance, suggesting that they decreased 6D1 binding specifically or nonspecifically. For these mutants, the second ELISA was performed alternatively twice by using wild type rGPIb␣ with known concentrations as a standard.
Western Blotting Analysis of Wild Type Recombinant GPIb-Approximately 25-30 ng of wild type rGPIb␣ was diluted 2-fold with gelloading buffer and separated by SDS-polyacrylamide gel electrophoresis in 10% acrylamide gels under non-reduced and reduced conditions. After transfer onto polyvinylidene difluoride membrane (Bio-Rad), proteins were incubated with anti-FLAG M2 MAb or various anti-GPIb MAbs (6D1, HPL7, AN51, and SZ2) and then incubated with peroxidase-conjugated anti-mouse IgG. Blots were developed by using ECL-Western blotting detection systems.

125
I Labeling of VWF-Purified human VWF (1 mg) and 1 mCi of Na 125 I were added to a vial containing 5 Iodobeads. The reaction mixture was incubated for 10 min with gentle agitation, and then 1 mM NaI buffer was added. Labeled proteins were separated from free Na 125 I on Centriprep-10 (Millipore). VWF antigen levels in the reaction mixture were determined by VWF-specific ELISA as described previously (18).
VWF Binding Assay-Polystyrene microplate wells with round bottom (Nalge Nunc, Rochester, NY) were coated with 50 l of anti-FLAG M2 MAb (10 g/ml in 50 mM bicarbonate buffer, pH 9.6) at 4°C for 24 h. The plates were washed three times with 150 l of PBST, pH 7.4, and then blocked with PBS containing 3% BSA (PBSA) at 4°C for 48 h. After blocking, all incubations were performed at room temperature. For ristocetin-induced VWF binding to rGPIb␣, varying concentrations of 125 I-VWF and rGPIb␣ were incubated with PBSA and 2 mg/ml (final) ristocetin in a total volume of 25 l for 1 h. Ristocetin causes nonspecific agglutination of VWF (19), and it is prevented by addition of 3-4% BSA (17). The mixture was transferred to anti-FLAG M2 MAb-coated plates, incubated, and washed six times with PBST. Bound VWF was solubilized by adding 2% SDS and 2-mercaptoethanol, and the radioactivity was quantitated by a ␥-scintillation counter. For botrocetin-induced VWF binding, varying concentrations of VWF and rGPIb␣ were incubated with PBS, 10 g/ml botrocetin (10 g/ml, final), and 0.3% BSA (final) in a total volume of 50 l for 30 min. Forty l of the solution was transferred to anti-FLAG M2 MAb-coated plate wells and incubated; the solution was then washed, followed by measurement by the ␥-scintillation counter as in the ristocetin-induced binding assay. Negative control assays were performed by using concentrated conditioned media from mock-transfected 293T cells. For both assays, nonspecific binding was determined by subtracting the total radioactivity from that of mock control.
Inhibition of VWF Binding to GPIb by Anti-GPIb MAb-To analyze MAb inhibition for the ristocetin-induced VWF binding of wild type rGPIb␣, 4 g/ml VWF, 0.7 g/ml wild type rGPIb␣, and 10 g/ml each anti-GPIb antibody were incubated for 1 h in the presence of 2 mg/ml ristocetin and 3% BSA in a total volume of 25 l at room temperature. For botrocetin-induced binding, 1.25 g/ml VWF, 0.7 g/ml wild type rGPIb␣, and 10 g/ml each antibody were incubated for 1 h with 10 g/ml botrocetin and 0.3% BSA in a total volume of 25 l at room temperature. For both assays, rGPIb␣ in the mixture was captured on anti-FLAG M2 MAb-coated plates, and the bound radioactivity was quantitated as described above. The specific inhibition ratio of each antibody was normalized to corresponding values obtained without the antibodies. The MAb inhibition assay for the ristocetin-or botrocetininduced platelet GPIb binding of VWF was performed as for recombinant VWF platelet binding assay as described previously (17), except that 10 g/ml each anti-GPIb antibody was added. Bound VWF was calculated subtracting percent unbound VWF from 100%, and the percent inhibition was expressed by normalizing to the values obtained without antibodies (17).
Mutant rGPIb␣ Binding to Anti-GPIb MAbs-Polystyrene microplate wells with U-shape bottoms (Coster, Cambridge, MA) were coated with 30 l of various anti-GPIb MAbs (6D1, AN51, SZ2, and HPL7) at 4°C for 24 h. Antibodies were diluted at 10 g/ml in 50 mM bicarbonate buffer, pH 9.6. Plates were washed three times with PBST. Because each antibody appeared to have a different dissociation constant for GPIb, the appropriate rGPIb␣ doses applied on the wells should vary, and appropriate concentrations were determined before the comparison of mutants. Several different concentrations of wild type rGPIb␣ were tested, and the specific range was set to give linear concentration-dependent signals for each antibody (data not shown). As a result, for 6D1 and HPL7, 5 ng/ml rGPIb␣ was applied on the wells. For AN51 and anti-FLAG M2 MAb, 500 ng/ml rGPIb␣ was used. For SZ2, 150 ng/ml rGPIb␣ was applied on the wells. Given concentrations of wild type or mutant rGPIb␣ were incubated in a total volume of 20 l at 4°C for 24 h. The wells were washed, incubated with diluted MWCL-1, washed, and then incubated with diluted peroxidase-conjugated goat anti-rabbit IgG followed by the color development of the absorbance at 490 nm. At least two independent duplicate assays were performed, and the percent binding was calculated by normalizing to the value obtained for wild type rGPIb␣. Negative control assays were performed by using concentrated conditioned media from mock-transfected 293T cells.
Crystallographic Structure Representations-The PDB file of the structural coordinate was downloaded from the Protein Data Bank (www.pdb.org). The coordinate of isolated human GPIb␣ (1GWB) (20) was obtained from Dr. Jonas Emsley. 3 Protein ribbon representation of the complex of GPIb␣ was prepared with the program PyMol (www.pymol.org).

Design and Expression of rGPIb␣ Mutants-A functional
domain that retains VWF binding activity resides in the Nterminal 45-kDa fragment between residues His 1 and Arg 293 of GPIb␣ (21,22). The segment that was targeted in this study consists of 287 amino acid residues between His 1 and Asp 287 of human GPIb␣ (16). This segment contains several regions including the N-terminal flank, LRRs, the C-terminal flank region, the anionic sulfated tyrosine region, and a part of the macroglycopeptide region ( Fig. 1) (4,23).
To simplify the production and analysis of mutant constructs, a clustered charged-to-alanine scanning strategy was chosen. Charged amino acids usually are at or near the protein surface, and alanine was selected as the replacement residue because it is the most common amino acid in proteins, is compatible with all types of secondary structures, and does not impose new structural effects related to hydrogen bonding, unusual hydrophobicity, or steric bulk (24). The recombinant GPIb␣ fragment 1-302 contained 62 charged residues, includ-3 J. Emsley, personal communication.
FIG. 1. Amino acid residues of human GPIb␣ targeted for charged-to-alanine scanning mutagenesis. The amino acid sequence shown includes residues 1-287 of the ␣ chain of human GPIb. The functional elements of GPIb␣ are indicated below the sequence. Seven short LRRs have been predicted in human GPIb␣ (28), but the crystal structure (5) suggested that amino acid sequences from Phe 201 contain LRR-8 and the C-terminal flank region. Charged residues His, Arg, Lys, Glu, and Asp were targeted for the mutagenesis. Segments that contain mutations in each construct are indicated by boxes. For convenience, the mutant proteins were named according to the residue number of the mutated amino acids in the mature GPIb␣. If more than one charged amino acid was mutated, the range of residue numbers and the number of alanine substitutions are indicated. For example, in mutant (172-175)2A, the two residues Glu 172 and Asp 175 were changed to alanine. As a result, 38 mutants were constructed, including 19 single mutations and 19 clustered mutations. Only one mutant D106A was not secreted, and D106K was constructed and analyzed instead. ing arginine, lysine, aspartate, glutamate, and histidine, and was changed singly or in small clusters to alanine. The 62 charged residues in the targeted region were covered in a total of 38 constructs of rGPIb␣, including 19 clustered mutants and 19 single mutants (Fig. 1). For convenience the mutants were named according to the residue number of the mutated amino acids in the mature GPIb␣. If more than one charged amino acid was mutated, the range of residue numbers and the number of alanine substitutions are indicated. For example, in construct E128A, one glutamate at position 128 was changed to alanine; in construct (172-175)2A, the two residues Glu 172 and Asp 175 were changed to alanine (Fig. 1).
Mutants were expressed in human kidney 293T cells and secreted as soluble proteins with FLAG tag at the C-terminal end. Cells were transfected with each construct, and serumfree media were analyzed for the expression of mutant rGPIb␣. Because a construct containing the single mutation D106A was not secreted, D106K, where negatively charged Asp was changed to positively charged Lys, was constructed and analyzed. All other mutant proteins were expressed and secreted efficiently (data not shown).
Western Blotting Analysis of Wild Type Recombinant GPIb␣-The present study employed four MAbs, 6D1 (25), HPL7, AN51 (26), and SZ2 (27), directed against human GPIb. We performed Western blotting analysis of wild type rGPIb␣ by using the four MAbs (Fig. 2). The binding of anti-FLAG M2 antibody was studied simultaneously, and it bound both reduced and non-reduced forms of wild type rGPIb␣ with ϳ45-kDa molecular masses (Fig. 2). Under non-reduced condition, all anti-GPIb antibodies reacted with wild type rGPIb␣, and ϳ45-kDa bands were visible ( Fig. 2A). When proteins were reduced, 6D1, HPL7, and AN51 lost the binding, whereas SZ2 bound normally (Fig. 2B). Therefore, the binding of the antibodies 6D1, AN51, and HPL7 was sensitive to reduction, suggesting that these antibodies are protein conformationdependent. The binding of SZ2 was not sensitive to protein reduction.
Effect on VWF Binding by Monoclonal Antibodies for GPIb-The four MAbs, 6D1, HPL7, AN51, and SZ2, were tested for their ability to inhibit the ristocetin-or botrocetin-induced binding by 125 I-VWF to rGPIb␣ (Fig. 3A). We also tested the inhibition for plasma VWF binding to fixed human platelets (Fig. 3B). MAb 6D1 completely inhibited both rGPIb␣-VWF and platelet-VWF interaction in the presence of ristocetin or botrocetin (Fig. 3, A and B). SZ2 inhibited botrocetin-induced rGPIb␣-VWF interaction by 30% and botrocetin-induced plate- FIG. 4. Binding of monoclonal antibodies to mutant rGPIb␣. Each monoclonal antibody (indicated at the top of the histograms) was coated onto polystyrene microtiter wells. For each MAb, binding was determined at a fixed concentration of rGPIb␣ (5 ng/ml for 6D1 and HPL7, 150 ng/ml for SZ2, and 0.5 g/ml for AN51) and normalized to the value obtained for wild type rGPIb␣ as described under "Experimental Procedures." The mutant proteins are indicated at the left. Each column represents the mean and S.D. obtained for at least two independent duplicate assays.
Binding of Monoclonal Antibodies to Mutant rGPIb␣-Mutant binding to each MAb was measurable by a specific ELISA. The MAb epitope might be related to the important amino acid residues for binding to VWF, and we analyzed the binding of the mutants to a panel of the four MAbs (Fig. 4).
For each antibody, the relative absorbance value was expressed as the percentage for wild type GPIb␣ (Fig. 4). Mutants D83A, H86A, D106K, and (149 -152)3A constitutively decreased binding to MAbs 6D1, HPL7, and AN51, whereas these mutants normally bound to SZ2. From the Western blotting analysis, the binding of 6D1, AN51, and HPL7 for wild type rGPIb␣ was lost by reduction, whereas SZ2 binding was not affected (Fig. 2), suggesting that the binding of the former three MAbs is dependent on protein conformation. Therefore, it is suggested that Asp 83 , His 86 , Asp 106 , Lys 149 , Glu 151 , and Lys 152 are included in amino acids that are important for protein conformation of GPIb␣.
In contrast, MAb 6D1 specifically reduced the binding to mutant E125A by 6.3% of wild type rGPIb␣, and MAb HPL7 showed 9.7% binding to a mutant R121A (Fig. 4). These mutants showed normal binding to other antibodies. It is thus suggested that the epitope of 6D1 and HPL7 is included in Glu 125 and Arg 121 , respectively.
SZ2 showed decreased binding to (274 -277)2A (13.2%) and to (281-283)3A (14.8%) (Fig. 4), suggesting that the binding of SZ2 is dependent on the C-terminal side of the rGPIb␣ fragment between Asp 274 and Asp 283 and that the epitope is included in residues Asp 274 , Asp 277 , Glu 281 , Glu 282 , and Asp 283 . This result is consistent with the study on human-canine chimera proteins by Shen et al. (28) that mapped the epitope of SZ2 into the anionic sulfated tyrosine region. Fig. 4 does not clearly demonstrate epitopes of MAb AN51.
Binding of rGPIb␣ to VWF-Thirty-eight mutants were tested for the VWF binding induced by ristocetin or botrocetin.
To determine the assay condition, we first evaluated the specific 125 I-labeled VWF binding of varying doses of immobilized rGPIb␣. For ristocetin-induced binding, wild type rGPIb␣ was tested in the presence of 2 mg/ml ristocetin. Wild type rGPIb␣ (1.4 g/ml) immobilized on anti-FLAG M2 MAb-coated plates appeared to give half-maximum binding of VWF (Fig. 5A). Thus, varying VWF concentrations were tested for immobilized wild type rGPIb␣ (1.4 g/ml), and 50 g/ml 125 I-VWF gave half-maximum binding (Fig. 5B). Based on these control experiments, the binding of 50 g/ml 125 I-VWF to 1.4 g/ml each mutant GPIb␣ was tested and compared.
The botrocetin-induced VWF binding assay was tested in the presence of 10 g/ml botrocetin. Wild type rGPIb␣ (0.7 g/ml) immobilized on anti-FLAG M2 MAb-coated plates appeared to give optimum binding of VWF (Fig. 5C). Varying VWF concentrations were tested for 0.7 g/ml wild type rGPIb␣, and 1.25 g/ml 125 I-VWF gave the optimum binding (Fig. 5D). This condition was used for comparison of the mutants.
To summarize binding data in graphical form, values obtained for each rGPIb␣ mutant were normalized to the corresponding values obtained for wild type rGPIb␣ in the presence of ristocetin or botrocetin (Fig. 6). Twenty-six mutants, representing 40 charged amino acids residues, had essentially normal VWF binding. Mutants E128A and (172-175)2A decreased the VWF binding in the presence of ristocetin or botrocetin, whereas they showed normal bindings to the MAb panel (Fig.  4). For ristocetin-induced binding, E128A and (172-175)2A reduced the binding by 9.2 and 12.7%, respectively. For botrocetin-induced binding, E128A and (172-175)2A decreased by 38.9 and 39.8%, respectively. Fig. 4 indicates that mutation at Glu 125 , located only three amino acids N-terminal side of Glu 128 , was an epitope of MAb 6D1 that remarkably inhibited ristocetin-and botrocetin-induced binding (Fig. 3). These facts may propose that the VWF binding site is included in residues Glu 128 , Glu 172 , and Asp 175 , which are located in the fourth and the fifth LRR domains.
Mutations at His 12 , Glu 14 , Asp 83 , His 86 , Asp 106 , Lys 149 , Glu 151 , and Lys 152 reduced the ristocetin-dependent VWF binding by less than 50% of wild type rGPIb␣ but showed normal botrocetin-dependent VWF binding (Fig. 6). Among these mutants, only mutations at His 12 and Glu 14 bound the MAb panel normally (Fig. 4). Such a phenotype may imply the possibility that His 12 and Glu 14 participate in the VWF binding induced by ristocetin.
Four mutants, D83A, H86A, D106K, and (149 -152)3A, showed constitutive decrease in binding to the three conformation-sensitive MAbs (Fig. 4). It is therefore possible that residues Asp 83 , His 86 , Asp 106 , Lys 149 , Glu 151 , and Lys 152 are important for protein conformation, and the decreased ristocetininduced VWF binding of the four mutants may be due to protein conformation of each rGPIb␣ mutant. However, botrocetin-induced binding of these mutants was normal (D106K and (149 -152)3A) or rather increased (D83A and H86A) (Fig.  6B), suggesting the possibility that the botrocetin-VWF complex may bind GPIb regardless of the protein conformation of GPIb␣. Also in these mutants, botrocetin may bypass the requirement of ristocetin for GPIb binding to VWF. These obser-  (Fig. 4), and the mutated amino acid residues are located in the C-terminal flank region and the sulfated tyrosine region. On the other hand, Fig. 4 indicates that mutations at Asp 274 , Asp 277 , Glu 281 , Glu 282 , and Asp 283 included epitopes of MAb SZ2 that partially inhibited botrocetin-induced VWF binding of GPIb (Fig. 3). Taken together, it is suggested that botrocetin-induced VWF-GPIb␣ interaction is dependent on the C-terminal side of the GPIb␣ fragment.
In addition to D83A and H86A, mutants H203A, (222-225)2A, and (269 -272)3A also increased the botrocetin-induced VWF binding by near 200% of wild type rGPIb␣ (Fig. 6B). Such a gain-of-function phenotype has been also observed in platelettype von Willebrand disease (29,30), and the location of the mutations found in this disease was recently explained by the crystallographic analysis of VWF A1-GPIb␣ complex (5). The crystallographic interpretation is required to assess whether the similar mechanism underlies the phenotype of our mutants.

DISCUSSION
Interaction between GPIb␣ and VWF-Crystallographic structure of human GPIb␣-VWF complex was recently determined by Huizinga et al. (5). This recombinant GPIb␣ was expressed in BHK cells and harbored the mutations N21Q and N159Q to remove N-glycosylation sites. In addition to the previous studies by way of total deglycosylation experiment (6) or by expression in insect cells (7), the successful co-crystallization implies that the two N-glycosylation sites may not necessarily be required for GPIb␣ binding to VWF. The Protein Data Bank data file (1M10) was converted into graphical form and is shown in Fig. 7, providing a framework for interpreting the effects of mutations on VWF binding of recombinant human GPIb␣.
From the VWF binding experiments, two mutants, E128A and (172-175)2A, reduced both the ristocetin-and botrocetininduced VWF binding (Fig. 6). In Fig. 7, Glu 128 is located at the central position of the concave LRR domain, and it appears to FIG. 6. Histogram of rGPIb␣ binding to VWF. Ristocetin-induced (A) and botrocetin-induced (B) binding of mutant rGPIb␣ was determined as described under "Experimental Procedures." A, assays were performed in the presence of 50 g/ml 125 I-VWF, 1.4 g/ml rGPIb␣, and 2 mg/ml ristocetin. B, assays were performed in the presence of 1.25 g/ml 125 I-VWF, 0.7 g/ml rGPIb␣, and 10 g/ml botrocetin. For ristocetin-induced binding the average number of independent determinations was three (range of 3-4); for botrocetin-induced binding the average number of independent determinations was four (range of 4 -6), and each bar represents the mean and S.D. 125 I-Labeled VWF binding is expressed as a percentage relative to that determined for wild type rGPIb␣. Asterisks indicate values that were significantly different from 100% (p Ͻ 0.005).
be associated with Lys 1371 (608) 4 of globular VWF A1 domain (5). Glu 172 and Asp 175 belong to LRR-6; Asp 175 faces to Phe 1366 (603) of VWF, although Glu 172 is projected outside the binding surface (Fig. 7, A and B). These facts suggest that Asp 175 may represent (172-175)2A phenotype, and Glu 128 and Asp 175 appear to be functionally important residues for VWF binding of human GPIb␣. The importance of GPIb␣ LRRs for VWF binding has been recently suggested (5,26,28), and our study identified the functionally important amino acid residues in the LRR domains of human GPIb␣. Fig. 4 indicates that Glu 125 is an epitope of MAb 6D1 that inhibited both ristocetin-and botrocetin-induced VWF binding. Glu 125 is located in close proximity with the VWF binding surface, although it is projected ϳ90°toward the binding surface (Fig. 7, A and B). Indeed, the VWF binding of Glu 125 was normal (Fig. 6), but the binding of 6D1 may interfere the GPIb␣-VWF interaction. In contrast, the location of Arg 121 (Fig.   7), which is a HPL7 binding site (Fig. 4), provides the explanation why Arg 121 had no effects on VWF-GPIb binding (Fig. 3). Arg 121 is located at the fourth LRR domain, but its location is on the opposite side of the VWF binding surface (Fig. 7).
Mutant (12)(13)(14)2A selectively decreased ristocetin-induced binding with normal botrocetin-induced binding (Fig. 6). The current crystal structure indicated that His 12 and Glu 14 , which are located at the N terminus of GPIb␣, contact closely with Glu 1376 (613) and Arg 1334 (571) of VWF A1 domain, respectively (Fig. 7, A and B). As indicated by Huizinga et al. (5), this binding involves the electrostatic interaction between His 12 and Glu 14 of GPIb␣ and Glu 1376 (613) and Arg 1334 (571) of VWF, respectively. Our previous alanine scanning mutagenesis of VWF A1 indicated that VWF mutants E613A and R571A did not bind platelet GPIb in the presence of ristocetin but bound in the presence of botrocetin (31). Taken together, ristocetin-induced VWF-GPIb interaction may require N-terminal sequences of GPIb␣ and the loops located at the bottom face of the VWF A1 domain (Fig. 7, A and B).
Decreased Binding of Mutants D83A, H86A, D106K, and (149 -152)3A to VWF or the Panel of MAbs-Mutants D83A, 4 Amino acid residues of human von Willebrand factor are numbered from the starting methionine as ϩ1. The previous numbering is indicated in parentheses. For old numbering of mutants, the N-terminal Ser of the mature processed VWF subunit was used as ϩ1. . The proposed counterpart of these GPIb␣ residues in VWF A1 (blue, Lys 608 and Phe 603 ; orange, Arg 571 and Glu 613 , respectively) is shown in the same color. On GPIb␣, Glu 172 , mutated along with Asp 175 , is not located at the important position and is shown by light blue. The epitope of MAbs 6D1 (Glu 125 ) and HPL7 (Arg 121 ) is shown with gray. Four clustering amino acid residues of GPIb␣ in yellow, magenta, green, and salmon pink are not indicated in number but are indicated in C. B, 150°clockwise rotation from bottom of the view in A is shown, and the location of residues Lys 149 , Glu 151, and Lys 152 is indicated, whose mutations showed decreased binding to the conformation-dependent panel of MAbs (Fig. 4) and to VWF in the presence of ristocetin (Fig. 6A). The coloring of other amino acid side chains is the same as in A. A close contact between Glu 125 and Lys 149 is expressed by a dotted line. C, close-up view of 45°clockwise rotation of the view in A. A transparent surface image of VWF A1 is depicted, and several residues involved in VWF-GPIb␣ interaction are shown as well as A. Mutations at residues Asp 83 , His 85 , and Asp 106 decreased the binding to conformationdependent MAbs (Fig. 4) and to VWF in the presence of ristocetin (Fig. 6A). They are shown along with Ser 85 , which appears to be located in the center of these residues. Location of these residues is also indicated without numbering in A but not in B. See "Discussion" for details.
H86A, D106K, and (149 -152)3A constitutively decreased the binding to conformation-dependent MAbs (Fig. 6), whereas these bound normally to SZ2 (Fig. 6), which also binds the reduced form of GPIb␣ (Fig. 2). Fig. 7B indicates that Lys 152 is projected toward VWF and may participate in VWF binding. This finding may be consistent with the decreased ristocetin-induced binding of (149 -152)3A (Fig. 6A). Indeed, Huizinga et al. (5) have suggested that Lys 152 has close contact with Asn 604 of VWF A1, and thus the importance of Lys 152 of GPIb␣ for VWF binding is not excluded (Fig. 7B). In contrast, Lys 149 is located close to Glu 125 , which contains the 6D1 epitope in 2.89 Å distance (Fig. 7B), suggesting the possibility that Lys 149 may also be included in the 6D1 epitope. However, Arg 121 is remote from Lys 149 , Glu 151 , or Lys 152 (Fig. 7B), and it is unlikely that HPL7 shares an epitope with the three amino acids. Taken together, (149 -152)3A may contain the 6D1 epitope but not the HPL7 epitope, and thus decreased HPL7 binding may be possibly due to abnormal conformation. Further study employing single mutations is required to unveil the role of each amino acid. Fig. 7C indicates that Asp 83 , His 86 , and Asp 106 appear to be clustered along with hydrophilic Ser 85 . This area is separated from the 6D1 epitope region (Glu 125 ) or HPL7 (Arg 121 ) (Fig. 7, A and C), suggesting that decreased binding to the MAb panel is due to conformation of the mutants. Because Ala mutation at Asp 106 resulted in no protein secretion, negatively charged Asp 106 may be important for protein conformation. Lys mutation at 106 was secreted normally but decreased in the binding to the MAb panel (Fig. 4). These facts may indicate that integrity of negatively charged side chains of these residues is important for protein conformation. In fact, Fig. 7C indicates that these residues are apart from the VWF binding surface.
Interaction among GPIb␣, VWF, and Botrocetin-Although the structure by Huizinga et al. (5) did not cover the C-terminal amino acid residues beyond Leu 267 , another crystal structure of human GPIb␣ (20) covered the amino acid residues between 1 and 280, and it contained the three sulfated tyrosines at 276, 278, and 279. On the other hand, the crystal structure of botrocetin complexed with VWF A1 mutant I546V was published recently (32), and molecular interaction of botrocetin-A1-GPIb␣ complex could be interpreted. To address the functional role of GPIb␣ for the interaction with botrocetin or VWF A1, the coordinate by Uff et al. (5) was first superimposed onto the structure of the VWF-GPIb␣ complex (Protein Data Bank code 1M10). The position of VWF was used to adjust the VWFbotrocetin complex (Protein Data Bank code 1JK) (32), thus virtually representing A1-botrocetin-GPIb␣ trimeric structure (Fig. 8).
Botrocetin is a non-physiological modulator to initiate VWF to bind GPIb. VWF A1 forms a complex with botrocetin via Arg 1399 (636), Lys 1430 (667), Arg 1329 (629), and Arg 1395 (632) (31), and botrocetin-A1 complex binds GPIb tightly. However, botrocetin alone dose not bind GPIb (33), and it has been thought to introduce the conformational changes of VWF. Mutant (217-218)2A decreased the botrocetin-induced binding, and (285-287)2A selectively decreased the botrocetin-induced binding, suggesting the possible role of the C-terminal regions of the GPIb␣ fragment for association with the botrocetin-VWF complex. Fig. 8 indicates that the C-terminal flank region including Arg 217 and Arg 218 appears to interact with the ␤ subunit of botrocetin. Remarkably, Arg 218 is closely projected on the surface of subunit ␤, partly explaining the (217-218)2A phenotype. On the other hand, the amino acid stretch from Asp 274 is projected toward the B chain of botrocetin (Fig. 8).
Mutations at Glu 285 and Asp 287 , both of which are not mapped on Fig. 8, may also participate in the association with botrocetin, interpreting the (285-287)2A phenotype. In the C-terminal anionic region, tyrosine residues 276, 278, and 279 are sulfated and have been proved important for botrocetin-mediated VWF-GPIb interaction (34,35). These facts may propose another role of the C-terminal regions of the GPIb␣ fragment for association with the botrocetin-VWF complex.
In terms of sulfation of three tyrosines, studies on other glycoproteins have shown that 293 cells, parents of 293T cells, had considerable ability of sulfation (36,37). The important feature of tyrosine sulfation consensus sequences has been the amino acid residue directly before a tyrosine to be sulfated, and it can be acidic or neutral (38). Bundgaard et al. (39) recently indicated that the charge of position Ϫ1 of the tyrosine is critical and can be Asp, Asn, or Ala. However, the degree of sulfation was partly influenced by the residues in positions Ϫ2 and Ϫ3, although the C-terminal side has not been studied (39). In mutant (269 -273)3A, positions Ϫ7, Ϫ6, and Ϫ4 of Tyr 276 are mutated to Ala from Asp, Glu, and Asp, respectively. In mutant (274 -277)2A, positions ϩ1 and Ϫ2 of Tyr 276 are mutated from Asp to Ala. These facts suggest that our mutations do not completely disrupt the consensus sequences, but deficit in sulfation is still possible.
MAb SZ2 partially inhibited botrocetin-induced VWF binding to GPIb but had no effect on ristocetin-induced VWF binding (Fig. 3). Fig. 4 indicates that SZ2 showed reduced binding to mutations at several residues in the C-terminal anionic region of the GPIb␣ fragment: Asp 274 , Asp 277 , Glu 281 , Glu 282 , and Glu 283 ; whereas mutations at Asp 269 , Glu 270 , and Asp 272 showed normal SZ2 binding. Ward et al. (27) reported previously that SZ2 immunoprecipitated a peptide Tyr 276 -Glu 282 (27). A later study using recombinant human-canine chimeric GPIb suggested that the SZ2 epitope is between Gly 268 and Glu 282 (28). Dong et al. (40) reported that SZ2 failed to bind mutants of the sulfated three tyrosines, suggesting that the sulfated tyrosines are also the key residues for SZ2. A possibility remains that (274 -277)2A is not completely sulfated, and decreased SZ2 binding of (274 -277)2A may be related to the sulfation. Above all observations suggest that in addition to sulfated tyrosines, the reactivity of SZ2 may be dependent on the wide range of C-terminal amino acids between 274 and 283, rather than between 268 and 273, thereby possibly expressing inhibitory activity for interaction between GPIb␣ and VWFbotrocetin. Although direct association between botrocetin and GPIb has not been proven, botrocetin complexed with VWF A1 may interact weakly with the C-terminal regions of the VWF binding domain of GPIb␣, and this interaction may aid the tight trimer complex formation. Further studies are required for the involvement of this region in GPIb␣ VWF binding.
Mutant (269 -272)3A showed normal binding to SZ2, but Fig.  6 indicates that it displayed increased botrocetin-induced VWF binding. In Fig. 8, the C-terminal regions contain two disulfide bonds, Cys 209 -Cys 248 and Cys 211 -Cys 264 , and residues beyond Asp 269 extend outward from Cys 264 resembling a hinge that turns through 180°. Asp 269 and Glu 270 are located at the first hinge, while Asp 272 is located before the second hinge comprising residues Asp 274 -Asp 277 (Fig. 8). In solution, this structure may be in equilibrium with a more extended conformation through flexibility around the hinges. In this context, it is possible that mutation at three acidic amino acids may affect the conformation of the flexible hinges, thus shifting the Cterminal amino acids, probably enhancing the interaction with botrocetin B chain.
On the basis of structure of VWF-GPIb␣, the flexible disordered loop from Val 227 to Ser 241 (␤ switch) undergoes a conformational change on complex formation resulting in the gainof-function phenotype of platelet-type von Willebrand disease (5). Asp 222 and Glu 225 are located at the bottom of the ␤ switch (Fig. 8B) and thus may be shifted by the conformational change of the ␤ switch. Asp 222 and Glu 225 are closer to botrocetin than VWF A1, and the interaction between botrocetin and mutant GPIb␣ (222-225)2A may be influenced (Fig. 6B). On the other hand, His 203 has close contact with Val 229 located in the ␤ switch (Fig. 8A). Substitution by alanine may disrupt this contact, probably influencing the ␤ switch movement. However, it is not known how the conformational shifting introduced by H203A affects the interaction with botrocetin; this has to be clarified by the crystallographic analysis of the VWF-GPIb␣-botrocetin complex.
Although mutants D83A, H86A, D106K, and (149 -152)3A decreased the binding to conformation-dependent MAbs and the ristocetin-induced VWF binding (Figs. 4 and 6), botrocetininduced binding was normal (D106A and (149 -152)A) or increased (D83A and H86A) (Fig. 6B). Usually, recombinant proteins with defective conformation may lead to the binding defect for their ligand. In the case of VWF A1, such recombinant proteins had shown reduced binding both to GPIb and botrocetin, probably because the formation of the GPIb binding site and the formation of the botrocetin binding site are conformationally related (31). However, rGPIb␣ with apparently defective conformation also abolished ristocetin-induced binding to VWF but did bind the botrocetin-VWF complex. Such botrocetin-mediated "rescue" might be supported by the sulfated anionic region to interact with botrocetin, and the rescue might occur regardless of the protein conformation of GPIb␣. Particularly in the case of D83A, H86A, D106K, and (149 -152)3A, the region of LRRs is mutated, and the function of the C-terminal regions may not be affected. Additionally, interpretation of the tertiary structure of the VWF-GPIb␣-botrocetin complex should be awaited to explain the enhanced botrocetininduced VWF-GPIb␣ interaction.