ADAMTS13 Substrate Recognition of von Willebrand Factor A2 Domain*

ADAMTS13 controls the multimeric size of circulating von Willebrand factor (VWF) by cleaving the Tyr1605–Met1606 bond in theA2 domain. To examine substrate recognition, we expressed in bacteria and purified three A2 (VWF76-(1593–1668), VWF115-(1554–1668), VWFA2-(1473–1668)) and one A2-A3 (VWF115-A3-(1554–1874)) domain fragments. Using high pressure liquid chromatography analysis, the initial rates of VWF115 cleavage by ADAMTS13 at different substrate concentrations were determined, and from this the kinetic constants were derived (Km 1.61 μm; kcat 0.14 s–1), from which the specificity constant kcat/Km was calculated, 8.70 × 104 m–1 s–1. Similar values of the specificity constant were obtained for VWF76 and VWF115-A3. To identify residues important for recognition and proteolysis of VWF115, we introduced certain type 2A von Willebrand disease mutations by site-directed mutagenesis. Although most were cleaved normally, one (D1614G) was cleaved ∼8-fold slower. Mutagenesis of additional charged residues predicted to be in close proximity to Asp1614on the surface of the A2 domain (R1583A, D1587A, D1614A, E1615A, K1617A, E1638A, E1640A) revealed up to 13-fold reduction in kcat/Km for D1587A, D1614A, E1615A, and K1617A mutants. When introduced into the intact VWFA2 domain, proteolysis of the D1587A, D1614A, and E1615A mutants was also slowed, particularly in the presence of urea. Surface plasmon resonance demonstrated appreciable reduction in binding affinity between ADAMTS13 and VWF115 mutants (KD up to ∼1.3 μm), compared with VWF115 (KD 20 nm). These results demonstrate an important role for Asp1614 and surrounding charged residues in the binding and cleavage of the VWFA2 domain by ADAMTS13.

tion with both VWF and fibrinogen. Activated platelets provide the phosphatidylserine-rich surface critical for the assembly of the procoagulant enzyme complexes that lead to the generation of thrombin (5).
VWF is expressed by endothelial cells and megakaryocytes (6). It is synthesized as 250-kDa monomers, which undergo intracellular processing-glycosylation, multimerization, and propeptide removal-that leads to formation of mature VWF multimers (7). Although much of endothelial VWF is constitutively secreted into the blood as multimers of varying size, a proportion is stored either within Weibel-Palade bodies in endothelial cells or within ␣-granules in platelets (8). These stored pools, which can be released upon specific stimulation, contain predominantly hyper-reactive "ultra-large" VWF that can exceed 2 ϫ 10 4 kDa.
A wide range of VWF multimers (500 -20,000 kDa) are found in normal plasma that differ only by the number of constituent VWF units. The largest VWF multimers unravel more readily in response to shear forces and contain more platelet binding sites. These species therefore confer the greatest hemostatic potential. The size of plasma VWF, and thus its platelet-tethering function, is regulated to prevent aberrant/ spontaneous platelet-rich thrombus formation. VWF multimeric size is modulated by the plasma metalloproteinase ADAMTS13, which cleaves at a single site in the VWF A2 domain between Tyr 1605 and Met 1606 (9). This proteolysis can only proceed once VWF has been unraveled, either by rheological forces or in vitro in the presence of denaturants, both of which induce the exposure of the A2 domain scissile bond (10). Physiologically, this occurs upon VWF secretion from endothelial cells, limiting the thrombogenic potential of newly secreted ultra-large VWF by converting it into smaller multimeric forms (11). The nature of ADAMTS13 recognition of the A2 domain is poorly understood. In part, this is because of complex folding and interdependence of the A1-A2-A3 domains. This has been illustrated by proteolysis studies conducted with isolated domain fragments and certain VWF mutations that confer type 2A (group II) von Willebrand disease (VWD) (12)(13)(14). The latter are proximate or within the A2 domain and manifest enhanced ADAMTS13-dependent VWF proteolysis. This is thought to be because of structural changes that lead to constitutive/enhanced exposure of the A2 domain cleavage site (15).
Because of the uncertainty surrounding A2 domain recognition mechanisms by ADAMTS13, we have prepared both wild-type and variant A2 domain fragments that span the Tyr 1605 -Met 1606 cleavage site. For the first time, the cleavage of these by ADAMTS13 has been evaluated kinetically under normal physiological ionic conditions. We show that critical residues in the vicinity of the cleavage site are required for optimum cleavage and form part of an ADAMTS13 docking site within the A2 domain.
surface plasmon resonance (SPR) and kinetic analysis experiments, an additional gel filtration step, using a Sephadex 26/10 column (Amersham Biosciences), was employed to obtain higher purity ADAMTS13.
The coding regions for different VWF A domain fragments (VWF76, Ser 1593 -Arg 1668 ; VWF115, Glu 1554 -Arg 1668 ; VWF115-A3, Glu 1554 -Glu 1874 ; VWFA2, Met 1473 -Arg 1668 ) were PCR amplified from the human VWF cDNA using Pfx high fidelity DNA polymerase (Invitrogen) according to the manufacturer's instructions. Fragments were cloned into the bacterial expression vector pET100/D-TOPO (Invitrogen), which fuses a His 6 and Xpress epitope tag to the N terminus of cloned sequences. As the VWF A3 domain contains two disulfide bonds, the adjacent paired cysteines at positions 1669 and 1670, at the start of the A3 domain, were mutated to alanine using the QuikChange XL kit (Stratagene). This approach improved yields of VWF115-A3 and ensured that the conserved A domain disulfide bond between Cys 1686 -Cys 1872 was correctly paired. Certain type 2A VWD mutations (R1597W, R1597Q, D1614G, I1628T, G1629R, E1638K) and specific charged residue substitutions (R1583A, D1587A, D1614A, E1615A, K1617A, E1638A, E1640A) were introduced into VWF115 and VWFA2 sequences using the QuikChange XL kit (Stratagene). All vectors were verified by sequencing.
VWF76, VWF115, and VWFA2 constructs were expressed in Rosetta Escherichia coli cells (Novagen, Nottingham, UK). VWF115-A3 was expressed in Rosetta-Gami E. coli (Novagen), which allows cytoplasmic disulfide bond formation. Transformed bacteria were grown in 2ϫ yeast tryptone cultures at 37°C with shaking. Once in log-phase growth, FIGURE 1. Purification of recombinant VWF76, VWF115, VWFA2, and VWF115-A3. A, diagram of the VWF A1A2A3 domains. A1 and A3 each contain paired cysteines, which are marked as are the amino acids at the boundary of each domain. Major ligands for each domain are highlighted. The ADAMTS13 cleavage site (Y 1605 -M 1606 ) lies in the middle of the A2 domain. Labeled arrows represent the recombinant VWF fragments (VWF76, VWF115, VWFA2, and VWF115-A3) expressed in E. coli. All had an N-terminal polyhistidine/Xpress epitope tag. The VWF numbering of each substrate is given. In VWF115-A3, the 2 adjacent cysteine residues at 1669 and 1670 (**) were mutated to alanine. B, purification of VWF76 from cytoplasmic fraction extracted from transformed Rosetta E. coli. Chromatogram of VWF76 purification is shown. The eluted peak (1) was analyzed by SDS-PAGE and Coomassie staining. VWF76 was visualized as a single 12.5-kDa band, corresponding to its predicted mass. C, purification and refolding of VWF115 from inclusion body fraction extracted from transformed Rosetta E. coli. Inclusion bodies were solubilized with 8 M urea and refolded on a nickel-chelating column using an 8 -0 M urea linear gradient. Chromatogram of VWF115 purification is shown. The eluted peak (1) was analyzed by SDS-PAGE and Coomassie staining. VWF115 migrated as a single 16.9-kDa band corresponding to its predicted mass. D, purification and refolding of VWFA2 from bacterial inclusion bodies. Inclusion bodies were solubilized with 8 M urea and refolded as for VWF115. Chromatogram of VWFA2 purification is shown. The eluted peak and stripped peaks (1) were dialysed and soluble material combined and analyzed by SDS-PAGE and Coomassie staining. As predicted, VWFA2 migrated as a 26.9-kDa band. E, purification and refolding of VWF115-A3 from Rosetta-Gami E. coli inclusion bodies. Insoluble material was solubilized and purified as for VWF115 and VWFA2. The eluted peak (1) and peaks stripped with 6 M guanidine/EDTA (2/3) were dialysed into 20 mM Tris (pH 7.8). The eluted peak (1), soluble stripped fraction (2), and insoluble precipitate after dialysis (3) were analyzed by reducing (R) and non-reducing (NR) SDS-PAGE and Coomassie staining. Under non-reducing conditions, soluble VWF115-A3 in 1 and 2 was visualized as a single 38.6-kDa band corresponding to its predicted mass. Correctly folded, monomeric VWF115-A3 was obtained within fractions 1 and 2. The presence of the disulfide bond was demonstrated by the band shift seen under reducing conditions. The insoluble precipitate contained predominantly VWF115-A3 polymers of higher molecular mass (3-NR). recombinant protein expression was induced with 1 mM isopropyl-␤-D-thiogalactoside for 16 h. Bacteria were harvested, and soluble (cytoplasmic) or insoluble (inclusion body) fractions were prepared using BugBuster reagent (Novagen). Recombinant His-tagged proteins were purified on an Ä KTA FPLC (Amersham Biosciences) using Ni 2ϩ -Hi-Trap-chelating columns (Amersham Biosciences). VWF76 was purified from the soluble fraction according to the manufacturer's instructions and eluted with 500 mM imidazole. VWF115-A3, VWF115, and VWFA2 inclusion bodies were solubilized in 8 M urea/20 mM Tris-HCl (pH 7.8)/50 mM imidazole and bound to the Ni 2ϩ -chelating column. Protein refolding was achieved using an 8 -0 M urea linear gradient (1 ml/min for 40 min). Refolded, soluble material was eluted with 500 mM imidazole. Misfolded, precipitated protein was stripped from the column with 6 M guanidine/0.5 M EDTA. All recombinant proteins (eluted and stripped) were dialyzed into 20 mM Tris (pH 7.8) overnight at 4°C. For VWFA2 and VWF115-A3, the stripped material was refolded by this process. Thereafter, soluble and insoluble protein fractions were separated by centrifugation. The purity of each recombinant protein was assessed by SDS-PAGE and Coomassie staining and quantified using a BCA total protein assay kit (Perbio, Cramlington, UK).
Cleavage of VWF Fragments by ADAMTS13-For time course reactions, recombinant ADAMTS13 (final reaction concentration 5-16.5 nM) was preincubated without substrate at 37°C for 30 min in reaction buffer (final reaction concentrations 20 mM Tris, pH 7.8/150 mM NaCl/5 mM CaCl 2 ). In preliminary experiments, VWF76, VWF115, VWFA2, or VWF115-A3 were added as final concentrations of 4 -8 M to start each reaction. For analyses of the intact A2 domain, reactions were performed in the presence and absence of 1 M urea. At different time points, 65-l sub-samples were removed and stopped with EDTA. 15 l were analyzed by SDS-PAGE (16.5% Tris-Tricine gel; Invitrogen) and Coomassie staining. The remaining 50 l were analyzed by HPLC. Samples were loaded on a BioBasic C4 column (Thermo Electron, Runcorn, UK), and the full-length and cleaved peptides were eluted/resolved using a specific acetonitrile gradient (20 -55% for VWF76, 25-45% for VWF115, 25-60% for VWFA2, 20 -65% for VWF115-A3). Mass spectrometry or N-terminal sequencing (Alta Bioscience, Edgbaston, UK) was performed to identify the polypeptide in each eluted peak and establish the precise site of proteolysis. For the determination of the specificity constant, low substrate concentrations ϽK m were used (250 -500 nM substrate/5.5 nM ADAMTS13). Unicorn analysis software was used to calculate the areas under the peaks of the full-length and cleaved fragments. The measure of the catalytic efficiency, k cat /K m , was determined using Enzfitter software by fitting the data to Equation 1 is the area under the cleavage product peak 1 curve at time t, [APf] is the area under cleavage product peak 1 curve after complete cleavage, e is ADAMTS13 concentration, t is time. For individual determination of the K m , V max, and k cat , the initial rate of proteolysis of VWF115 (after 5-10 min) was measured at varying substrate concentrations (100 nM-50 M) as outlined above. Data were transformed into Michaelis-Menten and Lineweaver-Burk plots, and the catalytic constants were derived using Prism 4 software (GraphPad, San Diego, CA). Surface Plasmon Resonance-SPR, using a dual flowcell BIAcore X biosensor system (BIAcore, Eyeworth, UK), was employed for analysis of the interaction between ADAMTS13 and different VWF115 variants. Two different SPR approaches were used. In each case, the surface of a carboxymethylated dextran (CM5) sensor chip (BIAcore) was activated with 0.4 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and 0.1 M N-hydroxysuccinimide, according to the manufacturer's instructions. First, ϳ600 response units of wild-type or variant VWF115 was covalently immobilized by amine coupling to one flow cell. To control for nonspecific binding, 600 response units of human endothelial protein C receptor was immobilized onto the other flow cell. All free reactive surface groups on both flow cells were blocked using 1 M ethanolamine. Different concentrations (10 -137 nM) of ADAMTS13 in HBS-P buffer (BIAcore) were injected over both flow cells at 30 l/min, contact time 1-2 min. After each injection, any bound ADAMTS13 was stripped with 10 mM NaOH prior to subsequent injections. Data analysis was performed using the BIAevalutaion software 3.0 (BIAcore), which allowed determination of k a , k d, and K D values by fitting the derived sensograms to 1:1 Langmuir model. For the second approach, ϳ8000 response units of ADAMTS13 or anti-endothelial protein C receptor antibody (control) were immobilized to respective flow cells of CM5 chips, as above. Various concentrations (80 nM-1 M) of VWF115, or VWF115 variants, in HBS-P buffer (BIAcore) were injected over both flow cells at 30 l/min. The chip surface was regenerated with 10 mM NaOH as before. Using this approach, data were analyzed using a heterogeneous ligand (parallel reaction) model.
Modeling of VWF A2 Domain-The VWF A2 domain was computer modeled using its sequence homology to the VWF A3 domain and the coordinates for the crystal structure of this domain (Protein Data Bank code 1atz_A), according to the method outlined by Sutherland et al. (17). Models were manipulated using Protein Explorer (www.molvis. sdsc.edu/protexpl/frntdoor.html).

RESULTS
Purification of Recombinant VWF Fragments-Different human VWF A2 domain substrate fragments (Fig. 1A) were expressed in, and purified from, bacteria. VWF76 was purified from the soluble, cytoplasmic fraction by Ni 2ϩ -chelating column affinity chromatography (Fig.  1B). SDS-PAGE and Coomassie staining demonstrated a purified product of 12.5 kDa. VWF115, VWFA2, and VWF115-A3 were purified from inclusion body preparations. Refolding of these proteins was achieved using an 8 -0 M urea linear gradient during purification. This technique produced high yields of soluble VWF115 and VWFA2 of the correct predicted size (16.9 and 26.9 kDa, respectively) that were essentially pure (Fig. 1, C and D). VWF115-A3 was initially purified and refolded similarly. However, after elution a relatively low yield of soluble material was purified (Fig. 1E, lane 1). The precipitated protein was stripped from the column and refolded by dialysis, yielding higher amounts of soluble material (lane 2) than the on-column refolded preparation. The eluted material and the refolded/solubilized stripped fraction obtained after dialysis contained a product of 38.6 kDa (lanes 1NR and 2NR). The presence of the disulfide bond between Cys 1686 and Cys 1872 was confirmed by the band shift upon reduction of this material (lanes 1R and 2R). The insoluble material after dialysis contained VWF115-A3 polymers arising through intermolecular disulfide bond formation (lane 3) and was discarded.
Proteolysis of VWF76, VWF115, and VWF115-A3-To first visualize the capability of ADAMTS13 to proteolyze each of the recombinant A2 domain fragment substrates, recombinant ADAMTS13 was incubated at 37°C with either VWF76, VWF115, or VWF115-A3 in the presence of physiological concentrations of Ca 2ϩ and NaCl. Sub-samples were stopped between 0 and 6 h and analyzed by SDS-PAGE and Coomassie staining to give a qualitative assessment of each reaction (Fig. 2, A-C). In each case, only the full-length substrate band was visible at 0 h. Thereafter, the intensity of this band diminished and two smaller cleavage product bands appeared. In each case, these bands corresponded to the predicted molecular masses of fragments generated following cleavage at the Tyr 1605 -Met 1606 bond. N-terminal sequencing was performed upon the smallest cleavage product and confirmed that proteolysis had occurred between the Tyr 1605 -Met 1606 bond. Visually, VWF76, VWF115, and VWF115-A3 appeared to be proteolyzed by ADAMTS13 at similar rates. Parallel samples were also analyzed by HPLC. The resolution of the full-length substrate from the two cleavage fragments (peaks 1 and 2) arising from proteolysis ( Fig. 2) of each different substrate allowed precise quantification of the extent of cleavage at each time point, confirming that the rates of proteolysis of VWF76, VWF115, and VWF115-A3 were, as indicated by the Coomassie staining, very similar (not shown).
To confirm specific proteolysis by ADAMTS13 and to identify each peak, cleavage product fractions were collected and analyzed by mass spectrometry. For all substrates, peak 1 corresponded to the tagged N-terminal cleavage product up to Tyr 1605 . Peak 2 corresponded to the C-terminal fragment, in each case starting at Met 1606 .
Type 2A VWD Mutations and ADAMTS13 Proteolysis-The scissile bond in VWF is predicted from a homology model (17) to be located in the center of the folded A2 domain (Fig. 3A). Given that VWF is the only known substrate for ADAMTS13 and that it is proteolyzed at just a single site, we hypothesized that other A2 domain amino acids that flank this site dictate the specificity of ADAMTS13 for the 1605-1606 bond. As an initial step toward identifying A2 domain residues that influence VWF proteolysis, we individually introduced six type 2A VWD mutations (see www.sheffield.ac.uk/vwf/) into the VWF115 construct (Fig.  3A). By mutating VWF115, which unlike the intact A2 domain naturally adopts a conformation that is permissive to ADAMTS13 proteolysis, the influence of specific mutations upon substrate specificity (as opposed to effects upon overall A2 domain structure) could be assessed under physiological ionic conditions. HPLC analysis of the proteolysis of VWF115 and VWF115 type 2A mutants (R1597W, R1597Q, D1614G, I1628T, G1629R, and E1638K) demonstrated that most of these mutations (R1597W, R1597Q, I1628T, and G1629R) had little or no effect upon VWF115 proteolysis (Fig. 3B). However, the E1638K mutant was proteolyzed slightly slower than VWF115, and the D1614G mutant was cleaved ϳ8 times slower (Fig. 3B). As Glu 1638 and Asp 1614 are charged residues that are predicted to be located in close proximity to each other on the surface of the VWF A2 domain homology model, we hypothesized that such residues might comprise part of an ADAMTS13 interaction site. Using this model, we subsequently identified additional acidic or basic amino acids predicted to lie on the surface of the A2 domain in the vicinity of Asp 1614 and Glu 1638 (Fig. 3C) and individually mutated each residue (Arg 1583 , Asp 1587 , Asp 1614 , Glu 1615 , Lys 1617 , Glu 1638 , Glu 1640 ) in VWF115 to alanine (Fig. 3D).
Kinetic Analysis of VWF76, VWF115, VWF115-A3, and VWF115 Charged Mutant Proteolysis-As a qualitative assessment of the relative ability of ADAMTS13 to proteolyze each of the charged residue mutants compared with the wild-type VWF76, VWF115, and VWF115-A3 substrates, we analyzed cleavage reactions by SDS-PAGE and Coomassie staining (Fig. 4A). From these gels, it seemed as if all of the mutants exhibited reduced rates of proteolysis. However, the concentrations of substrates needed to facilitate visualization by gel staining were high. To perform kinetic analysis of proteolysis for different substrates, it was necessary to monitor cleavage using substrate concentrations below the K m for the reaction. Consequently, we determined the individual kinetic constants (K m , k cat ) of VWF115 by measuring the initial rates of proteolysis at different concentrations (100 nM-50 M). For the ADAMTS13-dependent proteolysis of VWF115 (n ϭ 4), we derived a k cat of 0.14 Ϯ 0.06 s Ϫ1 . The K m for this reaction was determined to be 1.61 Ϯ 0.47 M, from which a value for the specificity constant (catalytic efficiency, k cat /K m ) of 8.70 ϫ 10 4 M Ϫ1 s Ϫ1 could be derived.
For separate determination of the catalytic efficiency of VWF115, repeated time course experiments were performed using low substrate concentrations (250 -500 nM). Repeated time course experiments (n ϭ 8) gave a mean value of 7.83 ϫ 10 4 M Ϫ1 s Ϫ1 for k cat /K m ( Table 1). Confirmation of the pseudo-first order conditions was obtained by deriving the same k cat /K m value for VWF115 proteolysis at both 250 and 500 nM. Determination of k cat /K m for VWF76 and VWF115-A3 gave values of 6.39 ϫ 10 4 M Ϫ1 s Ϫ1 and 6.70 ϫ 10 4 M Ϫ1 s Ϫ1 , respectively (Table 1). These kinetic data confirmed that the VWFA3 domain does not significantly contribute to the rate of A2 domain proteolysis, despite containing an ADAMTS13 binding site (18).
The contribution of each selected charged amino acid to VWF proteolysis was assessed by determination of the catalytic efficiency of all the mutant VWF115 substrates at 500 nM ( Fig. 4C and Table 1). The rate of proteolysis of the R1583A mutant was not significantly different from that of VWF115, suggesting that this residue is not important for cleavage. Mutation of Glu 1638 and Glu 1640 induced a mild (2-3-fold) reduction in catalytic efficiency. The most affected (Ͼ12-fold reduction) single mutant was D1587A, the k cat /K m of which was 0.64 Ϯ 0.10 ϫ 10 4 M Ϫ1 s Ϫ1 . The k cat /K m for K1617A, D1614A, and E1615A, which are predicted to lie on the same surface-exposed loop, were 1.90, 1.36, and 1.13 ϫ 10 4 M Ϫ1 s Ϫ1 , respectively. These results implied that this loop might play an important role in the cleavage reaction. Consequently, we generated a triple mutant DEK1614/5/7AAA) containing all 3 residues substituted for alanine. Whereas individually these mutations caused a 4 -7-fold reduction in k cat /K m , in combination the catalytic efficiency was Ͼ13fold less than wild-type VWF115, implying that the amino acids in this charged loop may act in synergy.
Influence of D1587A, D1614A, and E1615A upon VWFA2 Proteolysis-To test whether the effects of the above mutations were maintained in a fully folded A2 domain, we introduced the most affected single charged mutants into the whole A2 domain. A variety of VWF A2 domain mutations appear to alter the conformation of the domain, such that the scissile bond is more readily exposed and thus more accessible for ADAMTS13 cleavage under normal physiological ionic conditions. Therefore, this approach also served to ascertain whether the most affected charged residue mutants (D1587A, D1614A, E1615A) induced gross structural changes. Despite using increased concentrations of ADAMTS13 (16.5 nM) in the absence of a denaturant, the intact wildtype A2 domain was proteolyzed very slowly, as visualized by gel staining (Fig. 5A), when compared with the A2 domain fragments VWF115, VWF76, and VWF115-A3. When samples containing 500 nM VWFA2 were analyzed by HPLC (Fig. 5C), it was clear that the cleavage reaction did not proceed toward completion (ϳ28% after 5 h), as previously reported (13). The D1587A, D1614A, and E1615A VWFA2 mutants were proteolyzed either at a similar or slower rate (Fig. 5, A and C), suggesting that these mutations do not alter the A2 domain structure markedly. Under denaturing conditions (1 M urea), the wild-type VWFA2 was proteolyzed faster, although it still did not reach completion (ϳ50% after 5 h) (Fig. 5, B and D). Although all the mutants were cleaved at a higher rate under denaturing conditions, they were cleaved appreciably slower than VWFA2, corroborating the direct influence of these mutations upon VWF115 proteolysis.

Confirmation of ADAMTS13 Binding Site Residues in VWFA2
Domain-We hypothesized that the most likely cause of the reduced catalytic efficiency of the affected VWF115 charged mutants was impaired binding to ADAMTS13. To measure the influence of each mutation upon the interaction of VWF115 with ADAMTS13, we performed SPR analysis. Monoclonal antibody-mediated orientation-specific coupling of either VWF115 or ADAMTS13 to the sensor chip was not possible due to the relatively fast off-rates of these molecules with the available antibodies to each ligand (not shown). Consequently, we covalently bound either VWF115 or ADAMTS13 to the sensor chip by amine coupling. With VWF115 bound to the chip and ADAMTS13  Table 1. injected over the sensor, we monitored a concentration-dependent binding response indicative of a high affinity interaction (K D , 20 nM) for this interaction (Fig. 6, inset). Determination of both the association and dissociation rates for the ADAMTS13-VWF115 interaction (n ϭ 3) demonstrated a slow off-rate (1.72 ϫ 10 Ϫ3 s Ϫ1 ), which would appear to be the major determinant for the high affinity binding. A second approach, in which ADAMTS13 was coupled to the sensor chip and varying concentrations of VWF115 injected, gave a similar high affinity binding component (K D ϳ 15 nM, not shown). These data demonstrate that the affinity of ADAMTS13 for VWF115 is significantly higher than is suggested by the K m for VWF115 proteolysis.
SPR analysis of VWF115 mutants (D1587A, D1614A, E1615A, and DEK1614/5/7AAA) using both binding approaches revealed a markedly reduced affinity of ADAMTS13 for each of the mutants when compared with wild-type VWF115 (Fig. 6). Because of the low binding response measured, an accurate K D could not be derived from the VWF115 mutant data. Despite the poor fitting the estimate for the K D for the DEK1614/5/7AAA mutant was ϳ1.3 M. It was clear that for each variant the binding with ADAMTS13 was appreciably compromised and thus the likely cause of their reduced specificity constants.

DISCUSSION
ADAMTS13 cleaves a single bond within the VWF A2 domain (9). Under static or low shear conditions, proteolysis is slow because of the globular conformation adopted by multimeric VWF. Moreover, the complex folding of the A2 and surrounding domains further conceals the scissile bond (10). The isolated intact A2 domain is also not readily proteolyzed by ADAMTS13 (19). In the absence of a denaturant, the intact domain (VWF residues Met 1473 -Arg 1668 ) adopts a "non-permissive" structure that restricts access of ADAMTS13 (13). This is potentially explained by a VWF A2 domain model that suggests that the Tyr 1605 -Met 1606 bond lies buried within the center of the folded domain (17) and is thus not normally solvent exposed. However, incomplete VWF A2 domain fragments that span the ADAMTS13 cleavage site (e.g. VWF73 Asp 1595 -Arg 1668 and VWF115 Glu 1554 -Arg 1668 ) have been shown, here and elsewhere (12,20,21), to be readily and specifically  ADAMTS13 was incubated with 500 nM substrate under physiological ionic conditions at 37°C. Sub-samples from each time course reaction were taken (0 -6 h) and the extent of proteolysis measured by HPLC. The catalytic efficiency (k cat /K m ) of proteolysis of each substrate was derived using Enzfitter software. The specificity constants for each VWF substrate/VWF115 variant is given (ϮS.E.). For VWF115, VWF115-A3, and VWF76, n ϭ 8. For the least affected VWF115 variants (R1583A, D1640A, E1638A, and K1617) and the triple mutant DEK1614/5/7AAA, n ϭ 1. For the most affected single mutants (D1614A, E1615E, and D1587), n ϭ 3. Numbering as in Fig. 4 cleaved at the Tyr 1605 -Met 1606 site under normal physiological ionic conditions, suggesting that these fragments not only contain important recognition sequences for specific ADAMTS13-dependent proteolysis but also that their structure presents the cleavage site in a permissive conformation, more akin to unraveled/denatured VWF (12). In the present study, we have expressed and purified a number of related VWF A2 and A2/A3 domain fragments to delineate those regions important for A2 domain proteolysis. Previous studies have employed SDS-PAGE and/or Western blotting to monitor proteolysis, the latter to achieve high sensitivity in the early stages of the cleavage reaction (13). Cleavage of VWF fragments by ADAMTS13 was quantified here by HPLC, allowing for the first time accurate and fully quantitative analysis of the kinetics of these reactions. In our hands, most likely as a result of the smaller recombinant protein tag used in this study, VWF115 was marginally more susceptible to cleavage than the shorter substrate VWF76 (very similar to VWF73 that has been reported elsewhere (12,20)) when the cleavage reaction was fully quantified. Its K m was ϳ1.6 M, k cat ϳ 0.14 s Ϫ1 , and its specificity constant (k cat /K m ) ϳ 8 ϫ 10 4 M Ϫ1 s Ϫ1 . This estimate of specificity is quite similar to that recently reported for ADAMTS13 cleavage of fulllength VWF multimers under conditions of low ionic strength (22). This is surprising, considering those experiments were performed under denaturing conditions and without the advantage of a released peptide with which to accurately quantitate the rate of proteolysis. VWF115-A3 gave similar values for the specificity constant, suggesting that the A3 domain makes no appreciable contribution to the cleavage reaction despite its reported high affinity binding site (18). This finding corroborates previous semi-quantitative data obtained from the comparative analysis of recombinant VWFA2 and VWFA2/A3 proteolysis under denaturing conditions (13).
Residues surrounding the cleavage site are known to influence proteolysis of intact VWF, as they are reported to be dysfunctional in certain cases of type 2A (group II) VWD (15). We therefore introduced some of these mutations into the A2 domain fragment (VWF115) and evaluated proteolysis. Of course, the type 2A (group II) phenotype is generally associated with enhanced, rather than reduced, proteolysis by ADAMTS13. Nevertheless, we selected reported mutations surrounding the cleavage site, as not all had been tested extensively in terms of ADAMTS13 interaction. Interestingly, of those mutations selected, one (D1614G) was proteolyzed significantly slower when introduced into VWF115. To fully characterize how this mutation might exert a type 2A VWD phenotype, further studies examining its influence upon fulllength, multimeric VWF will be required. Type 2A group II VWD mutations are generally believed to manifest their phenotype by unfolding of the A1-A2-A3 domain region, increasing ADAMTS13 access to, and therefore cleavage of, the scissile bond. The possibility of reduced, rather than enhanced, rates of cleavage of mutants, such as D1614G, in the vicinity of the cleavage site should be taken into consideration when evaluating the complex phenotype of cases of potential type 2A group II VWD. Because of this finding, we mutated further charged residues thought to be in proximity with Asp 1614 on the basis of primary sequence and the homology model of the A2 domain. Of eight such mutants introduced into VWF115, six (D1587A, D1614A, E1615A, K1617A, E1638A, and E1640A) had clearly reduced cleavage rates. The triple mutant, DEK1614/5/7AAA, had ϳ13-fold reduction in the specificity constant for proteolysis. To confirm that the delayed cleavage of the VWF115 substrate was not merely a feature of this particular substrate, three of the mutations (D1587A, D1614A, and E1615A) were introduced into the fully folded intact A2 domain. Under denaturing conditions (1 M urea), proteolysis of all three substrates was reduced in rate and extent when compared with the wild-type VWFA2. Even under non-denaturing conditions, where under the conditions used proteolysis of the A2 domain did not exceed 30% at 300 min, proteolysis of D1587A and D1614A was appreciably impaired. These results with both VWF115 and A2 substrates suggested that specificity is appreciably influenced by a charged patch on the surface of the A2 domain.
Binding of ADAMTS13 to multimeric VWF and A2 domain fragment substrates has been reported to be of high affinity (23). Physiologically, ADAMTS13 seems to circulate in complex with VWF. However, the binding of ADAMTS13 to VWF need not necessarily be accompanied by proteolysis. Although Ca 2ϩ ions are required for efficient proteolysis (10), they are not necessary for high affinity binding (21,23). A recent report of binding of ADAMTS13 to VWF73 in the absence of Ca 2ϩ ions using a microtiter plate-based immunoassay for ADAMTS13 suggested a K D of 4.6 nM (21). Because of uncertainties surrounding the use of antibody detection and equilibrium perturbation in the detection step of binding in such assays, we used the alternative approach of SPR. Coupling VWF115 directly to the sensor chip and injecting different concentrations of ADAMTS13 yielded a K D of ϳ20 nM. This high affinity interaction appeared to be conferred by a slow dissociation rate (k d ). Confirmation of tight binding was obtained by covalently immobilizing ADAMTS13 and then injecting VWF115 over the sensor chip. In this case, a K D of 15 nM was obtained. However, there was also some evidence for heterogeneity of binding, and a lower affinity component was also apparent using this second approach. Importantly though, both SPR approaches demonstrated substantial loss (Ͼ10-fold) in binding affinity of ADAMTS13 to the mutant VWF115 fragments D1587A, D1614A, and E1615A and the triple mutant DEK1614/5/7AAA irrespective of whether SPR was performed with these fragments as analyte or ligand. This provides direct evidence that these residues form part of a docking site for ADAMTS13 on the VWFA2 domain.
The present investigation has for the first time identified critical residues within the VWF A2 domain that are important for both binding to ADAMTS13 and for proteolysis at the Tyr 1605 -Met 1606 bond. Given its specificity, proteolysis is unlikely to be determined solely by interaction with a limited number of residues surrounding the cleavage site. Previ- ous studies have shown that although VWF73 is cleaved by ADAMTS13, a 9-residue deletion at the C terminus of this fragment completely abolished cleavage (12). It remains uncertain precisely where these 9 amino acids lie in the fully folded intact A2 domain or in unraveled multimeric VWF in relation to the charged patch/residues identified in this study. Despite this, and the inability of a peptide that spans this region to inhibit proteolysis, it remains possible that residues within this 9-amino acid sequence contribute directly to high affinity binding to the protease (12). Structures remote from the cleavage site are also known to influence proteolysis. The A1 domain of VWF was recently demonstrated to inhibit access of ADAMTS13 to the A2 domain cleavage site (13). This inhibition was removed in the presence of GpIb␣, the principal A1 domain ligand, suggesting that under physiological conditions platelet binding may enhance VWF proteolysis either by increasing scissile bond access or by revealing an ADAMTS13 binding site.