The Acidic Region of the Factor VIII Light Chain and the C2 Domain Together Form the High Affinity Binding Site for von Willebrand Factor*

A binding site for von Willebrand factor (vWf) was previously localized to the carboxyl terminus of the C2 domain of the light chain (LCh) of factor VIII (fVIII). The acidic region of the LCh, residues 1649–1689, also controls fVIII·vWf binding by an unknown mechanism. Although anti-acidic region monoclonal antibodies prevent formation of the fVIII·vWf complex, the direct involvement of the acidic region in this binding has not been demonstrated. By limited proteolysis of LCh with Staphylococcus aureus V8 protease, we prepared 14- and 63-kDa LCh fragments, which begin with fVIII residues 1672 and 1795, respectively. Using surface plasmon resonance to measure binding interactions, we demonstrated that the 14-kDa fragment binds to vWf, but its affinity for vWf (K d 72 nm) was 19-fold lower than that of LCh. This was not due to an altered conformation of the acidic region within the 14-kDa fragment, since its affinity for an anti-acidic region monoclonal antibody was similar to that of LCh. All LCh derivatives lacking the acidic region (thrombin-cleaved LCh, recombinant C2, and 63-kDa fragment) had also greatly reduced affinities for vWf (K d 564–660 nm) compared with LCh (K d 3.8 nm). In addition, the similar affinities of these derivatives for vWf indicated that apart from its acidic region, the LCh contains no vWf binding site other than the one within C2. The reduced affinities of the LCh derivatives lacking the acidic region for monoclonal antibody NMC-VIII/5 (epitope, C2 residues 2170–2327) indicated that removal of the acidic region leads to a conformational change within C2. This change is likely to affect the conformation of the vWf binding site in C2, which overlaps the epitope of NMC-VIII/5; therefore, the acidic region also appears to be required to maintain the optimal conformation of this vWf binding site. Our results demonstrate that the acidic region and the C2 domain are both directly involved in forming a high affinity binding site for vWf.

The plasma glycoprotein factor VIII (fVIII) 1 functions as a cofactor for factor IXa in the factor X activation enzyme complex of the intrinsic pathway of blood coagulation (1). fVIII internal protein sequence homology has led to the designation of six domains arranged in the order A1-A2-B-A3-C1-C2 (Ref. 2; Fig. 1). The heavy chain (HCh) of fVIII consists of the A1, A2, and B domains, whereas the light chain (LCh) consists of the A3, C1, and C2 domains. Three regions rich in acidic amino acids are located at the carboxyl termini of the A1 and A2 domains and the amino terminus of the A3 domain.
Maintenance of a normal fVIII level in the circulation is dependent on its complex formation with von Willebrand factor (vWf) since patients with severe von Willebrand's disease, who have a complete deletion of the vWf gene or mutations which reduce binding between fVIII and vWf, have a secondary deficiency of fVIII. When fVIII is bound to vWf, a stable association of its HCh and LCh is maintained (3), and fVIII is prevented from binding to phospholipid vesicles (4), platelets (5), or factor IXa, functions required for its procoagulant activity. In addition, vWf protects fVIII from activation by activated factor X (6) and from protein C-catalyzed inactivation (4). fVIII binds to vWf through the LCh (7,8). The reported binding stoichiometry is 1 fVIII molecule/vWf monomer (9,10).
Thrombin, the principal physiological activator of fVIII, cleaves the protein at Arg 372 and Arg 740 in the HCh and at Arg 1689 in the LCh (11). Activated fVIII (fVIIIa) is a heterodimer of 50-, 43-, and 73-kDa subunits, all of which are required for procoagulant activity (12). Cleavage of the LCh at Arg 1689 , which releases the acidic region (fVIII residues 1649 -1689), is responsible for dissociation of fVIIIa from vWf (8,13). The importance of the LCh acidic region for fVIII binding to vWf was suggested by the observations that several anti-acidic region monoclonal antibodies (mAbs) with epitopes within residues 1670 -1689 (14 -17) inhibit fVIII binding to vWf, as does complete deletion of the acidic region (18). In contrast, deletion of the fVIII B domain and part of the acidic region (1649 -1669) did not abolish vWf binding (18), suggesting that the acidic region residues 1669 -1689 are critical for fVIII⅐vWf binding. Mutants of fVIII with partial or complete deletions of the acidic region have normal procoagulant activity (18), which suggests that the loss of this region has no effect other than the elimination of vWf binding. The presence of post-translationally sulfated Tyr 1680 was shown to be essential for vWf binding. However, synthetic peptide 1673-1689 failed to inhibit fVIII binding to vWf, regardless of whether Tyr 1680 was sulfated (18). Thus, the exact function of the acidic region is not known, since its direct binding to vWf has not been demonstrated.
We demonstrated that a glutathione S-transferase-C2 fusion protein binds to immobilized vWf in a dose-dependent, saturable fashion (19), which indicates that the C2 domain also con-tains a binding site for vWf. Synthetic peptide 2303-2332, corresponding to the previously identified fVIII phosphatidylserine binding site (20), prevented this interaction (19,21), demonstrating that the C2 region 2303-2332 is involved in fVIII interaction with both phosphatidylserine and vWf.
In this study we describe methods for isolating several proteolytic LCh fragments. We have used these fragments for quantitative measurements of their binding to vWf to further elucidate the roles of the acidic region and the C2 domain.
Quantification of Proteins-Concentrations of antibodies and vWf were determined by the method of Bradford (25) and those of purified recombinant C2 and proteolytic fragments of LCh by the Quantigold assay (Diversified Biotech). The concentrations of plasma-derived fVIII, LCh and HCh were determined by absorbance at 280 nm, using extinction coefficients of 1.2 (26), 1.34 (8), and 1.0 (8), respectively. The molar concentrations of fVIII, LCh, A3-C1-C2, and vWf were calculated using molecular masses determined by SDS-PAGE of 300 kDa (2), 80 kDa (27), 73 kDa (28) and 270 kDa (29). Since the B domain is partially proteolyzed, we used the average molecular mass 136 kDa for the HCh (27). The molecular mass of 209 kDa for HCh/A3-C1-C2 heterodimer was calculated as a sum of molecular masses of A1-C1-C2 and HCh.
Enzyme-linked Immunosorbent Assay-For ELISA 1, detection of residual LCh in the HCh preparation was performed using a previously described ELISA with two anti-C2 mAbs, ESH4 and ESH8 (21), which detects LCh at Ն 0.025 nM. For ELISA 2, detection of residual uncleaved LCh in the A3-C1-C2 preparation was performed as follows. Immulon I plates (Dynatech) were coated with 5 g/ml mAb C4, blocked with 200 l/well TBS, 2% bovine serum albumin, and LCh (0.025-1.25 nM) standards or A3-C1-C2 were added. All incubation steps were at 37°C for 1 h with shaking, and wells were washed three times with 200 l of TBS, 0.05% Tween 20 (Bio-Rad). Bound LCh was detected with biotinylated mAb ESH8 followed by streptavidine-alkaline phosphatase. The color developed by alkaline-phosphatase cleavage of p-nitrophenyl phosphate was read at 410 nm on an MR5000 microtiter plate reader (Dynatech). Wells coated with mAb C4 and containing all components except LCh were used as negative controls. The average values of these controls were Յ10% of the maximal signal, and they were subtracted from the values for all other samples.
Protein Purification-Plasma fVIII was purified from therapeutic concentrates of Method M, American Red Cross (30). LCh and HCh were purified as described previously (12,21). Residual HCh was removed from the LCh preparation by its passage over an immobilized mAb 8860 column. The purity of the LCh was confirmed by SDS-PAGE, where it migrates as a single band of 80 kDa (Fig. 2, lane 1). Residual LCh present in the HCh preparation was removed by passage over an immobilized mAb ESH8 column equilibrated with 20 mM Tris, pH 7.4, 0.15 M NaCl (TBS), 5 mM CaCl 2 . The final HCh preparation (2650 nM) contained Ͻ0.025 nM LCh as tested by ELISA 1. A3-C1-C2 was prepared by cleavage of LCh with thrombin (31) and purification by ion exchange chromatography on a Resource S column (Pharmacia Biotech Inc.) (32). Traces of uncleaved LCh were removed by incubation for 18 h at 4°C in TBS, 5 mM CaCl 2 , 0.01% Tween 20 with the anti-acidic region mAb C4 immobilized on CNBr-activated Sepharose 2B resin at 1.4 mg/ml. The final A3-C1-C2 preparation (3700 nM) contained Ͻ0.03 nM LCh, as measured by ELISA 2. The purity of A3-C1-C2 was also confirmed by SDS-PAGE, where it migrates as a single band with a molecular mass of 73 kDa (Fig. 2, lane 2). The fVIII HCh/A3-C1-C2 heterodimer was prepared from HCh and A3-C1-C2 and purified by ion exchange chromatography on a Resource S column (31). Its activity was determined in the one-stage clotting assay (33) to be 3300 units/mg, which is similar to a previously reported value (31). The recombinant C2 domain was produced as a soluble, secreted protein in Spodoptera frugiperda Sf9 insect cells (34), and it was purified by gel filtration and ion-exchange chromatography (21). vWf was purified from cryoprecipitate (Cutter Biological) (21). The final vWf preparation (0.68 mg/ml) contained Ͻ0.2 g/ml fVIII, as determined by an immunoradiometric fVIII antigen assay (35).
Radiolabeling of fVIII-Prior to iodination fVIII was dialyzed into 0.2 M sodium acetate, 5 mM calcium nitrate, pH 6.8 (iodination buffer). Five g of fVIII in 30 l of iodination buffer were added to 5 l of lactoperoxidase beads (50% suspension, Worthington Biochemical Corp.), 5 l of Na 125 I (100 mCi/ml, Amersham), and 5 l of 0.03% H 2 O 2 (Mallinckrodt) and incubated for 3 min. Free Na 125 I was removed by chromatography on a PD10 column (Pharmacia) in 0.1 M MES, 0.15 M NaCl, 5 mM CaCl 2 , pH 6.3. The specific radioactivity of fVIII was 4 Ci/g of protein. The activity of 125 I-fVIII determined in the one-stage clotting assay (3600 units/g) was similar to that of unlabeled fVIII.
Preparation of LCh Proteolytic Fragments-LCh (250 g/ml) in TBS, 5 mM CaCl 2 was digested with Staphylococcus aureus V8 protease (ICN Biochemicals Inc.) at an enzyme/substrate ratio of 1:100 for 40 h at 25°C. The digested mixture was treated with 1 mM protease inhibitor Pefabloc (Boehringer Mannheim), lyophilized, and subjected to fast protein liquid chromatography on a Superdex G75 HR (Pharmacia) column under denaturing conditions (9 M urea, TBS). Peaks corresponding to 14-and 63-kDa fragments were eluted at V e /V t ratios of 0.35 and 0.42, respectively, and identified by SDS-PAGE. The pooled fractions containing each fragment were concentrated, individually repurified on the Superdex G75 column, and further purified by chromatography on a Resource S column. The 14-or 63-kDa fragments in TBS, 9 M urea, were diluted 1:4 with 25 mM His, 9 M urea, pH 6.0 (Buffer A), loaded on a Resource S column equilibrated with Buffer A, and eluted with a linear gradient of 0 -1.2 M NaCl in Buffer A as single peaks at 0.4 and 0.3 M NaCl, respectively. For refolding the fragments were dialyzed against 0.05 M glycine, pH 3.5, 5 mM CaCl 2 at 4°C. The pH was adjusted to 7.4 with 0.5 M glycine, pH 9.5, followed by dialysis in TBS, 5 mM CaCl 2 , 0.01% Tween 20. The yields of the 14-and 63-kDa fragments were 19 and 28%, respectively. The final purity of each fragment was determined by SDS-PAGE in 8 -25% gels (Fig. 2).
The association of the 14-and 63-kDa fragments was determined by analytical size-exclusion chtomatographic measurements performed at room temperature using fast protein liquid chromatography (FPLC) on a Superdex G75 HR column (Pharmacia) at flow rate 0.5 ml/min. Elution was monitored by the absorbance at 280 nm. All experiments were performed in TBS, 5 mM CaCl 2 , 0.01% Tween 20.
Analysis of 14-kDa Fragment and LCh Cleavage by SDS-PAGE-The kinetics of 14-kDa fragment and LCh thrombin cleavage at Arg 1689 was analyzed by electrophoresis by 8 -25% SDS-PAGE, followed by staining with silver nitrate. Protein distribution among the bands was quantitated by an HP ScannerJet IIcx scanner (Hewlett Packard) and the NIH Image computer program. Cleavage of LCh or 14-kDa fragment at Arg 1689 was determined as the ratio of cleaved product migrating as 73or 11.5-kDa bands, respectively, to total protein in the lane.
Protein-Protein Binding Using Biosensor Technology-The kinetics of protein-protein interaction were determined by surface plasmon res- onance using the IAsys biosensor (FISONS, Cambridge, UK) which measures protein binding and subsequent dissociation in real time (36). vWf (50 g/ml), NMC-VIII/5 (50 g/ml), and NMC-VIII/10 (50 g/ml) in 10 mM sodium acetate, pH 5.0, were covalently coupled to the activated carboxymethyldextran-coated biosensor cuvette (FISONS) via amino groups using succinimide ester chemistry (36) (FISONS). Binding was measured in 200 l of TBS, pH 7.4, 5 mM CaCl 2 , 0.01% Tween 20 at 37°C for all ligands. Dissociation was initiated by substitution with the same buffer (200 l) lacking ligand, which was changed with fresh buffer at a flow rate 10 l/min. To regenerate the cuvette, complete dissociation of ligands bound to immobilized vWf was achieved by addition of 20 mM Tris, 0.35 M CaCl 2 , 0.6 M NaCl, 0.01% Tween 20, pH 7.2, for 2 min. Complexes of mAb NMC-VIII/5 or mAb NMC-VIII/10 with fVIII derivatives were completely dissociated by addition of 0.1 M glycine, pH 3.0, for 3 min. Identical signals for reference binding of LCh or its derivatives to immobilized vWf, NMC-VIII/5, and NMC-VIII/10 were obtained before and after a series of 25 regenerations.

Calculation of the Kinetic Parameters from Biosensor Kinetic Data-
The values of the rate constants for dissociation (k off ) of fVIII and its derivatives from immobilized vWf or from mAbs were determined by fitting the dissociation kinetics data to the following equation describing single phase dissociation process, where the surface plasmon resonance signal observed, R, is proportional to the formation of a complex between immobilized component and added ligand.
It was shown (37) that R max is the maximal binding capacity of the immobilized ligand surface expressed in resonance units (Arc s), and C is the concentration of polypeptide in solution.
The k on values were determined from individual association kinetics data using the integrated form of the rate Equation 2.
The values of k on and R max were derived from nonlinear regression analysis fitting R versus T to Equation 3 by Sigmaplot 1.02 (Jandel Scientific). The value of the k off constant used in Equation 3 was derived from the dissociation kinetics data fitted to Equation 1. The values of equilibrium dissociation constants (K d ) were calculated as k off /k on . Determination of fVIII, LCh, 14-kDa Fragment, and C2 Affinity for vWf in Fluid Phase-Binding affinity of fVIII or its fragments for vWf was determined by homologous or heterologous ligand displacement assays. 125 I-fVIII (0.04 nM), 5 nM vWf, and varying concentrations of unlabeled fVIII or LCh, 14-kDa fragment, or C2 were incubated in TBS, 5 mM CaCl 2 , 0.01% Tween 20 at 37°C for 1 h. Aliquots (100 l) from the samples were added to triplicate microtiter wells (Immulon I, Dynatech) coated with 2.5 g/ml anti-vWf mAb 2.2.9, which does not interfere with fVIII binding to vWf (19). Preliminary kinetic studies of the binding of 125 I-fVIII⅐vWf complex to immobilized mAb 2.2.9 demonstrated that binding reaches equilibrium within 3 min at 37°C. Following incubation for 4 min at 37°C, wells were washed with TBS, 0.1% Tween 20, and bound radioactivities were determined. Negative controls were mAb 2.2.9-coated wells containing all components except vWf, and these values were subtracted from average values of triplicates of all other wells. The values of these controls were Յ5% of the maximal binding values. For heterologous ligand displacement studies, in which displacing ligand differs from the labeled ligand, the data were analyzed assuming two distinct equilibria, R ϩ L 1 u RL 1 and R ϩ L 2 uRL 2 , where R is vWf, L 1 is 125 I-fVIII, and L 2 is LCh, 14-kDa fragment, or C2. These equilibria are described by the equilibrium constants K d and K i , respectively. Since there is insufficient information in a single heterologous displacement curve to solve for both unknown parameters (K d and K i ), the K d value was determined from a homologous 125 I-fVIII displacement by unlabeled fVIII. The data of homologous and heterologous displacements were fitted to a model assuming a single class of fVIII binding sites on vWf molecules using the computer program LIGAND (38).

RESULTS
The LCh of fVIII Binds to vWf with Lower Affinity than fVIII-To determine whether LCh is solely responsible for the high affinity fVIII interaction with vWf, we compared kinetics of fVIII and LCh binding to and dissociation from vWf immobilized on a biosensor cuvette using an IAsys biosensor instrument. The detection of bound protein is based on the surface plasmon resonance phenomenon, which measures the change in the refractive index due to association of a fluid phase ligand with an immobilized ligand or due to dissociation of the formed complex (36) by generating a signal of 163 Arc s/1 ng of protein bound/mm 2 of the biosensor cuvette. Representative curves in Fig. 3A show the resonance response reflecting time course of fVIII or LCh association interaction with immobilized vWf. Similar binding to a control cuvette lacking vWf by maximal concentrations of fVIII or LCh gave resonance signals Յ4% of those in the presence of vWf (Fig. 3A), indicating the absence of nonspecific interactions with the cuvette.  The k off values presented in Table I were derived from dissociation curves for fVIII and LCh (Fig. 3B) using Equation 1 describing single exponential dissociation to the base line (see "Experimental Procedures"). The values of second order association rate constants (k on ), shown in Table I, were determined from the best fit of association kinetics data to Equation 3 describing single phase association (see "Experimental Procedures"). Equilibrium binding was achieved for fVIII and LCh after 1 and 3 min (Fig. 3A), respectively. Since the concentrations of fVIII or LCh were Ͼ10 times the corresponding K d values for their interaction with vWf (Table I), the amounts of fVIII and LCh bound to vWf at equilibrium (3.8 fmol/mm 2 and 3.62 fmol/mm 2 , respectively) represent the maximal binding capacity of immobilized vWf. The binding capacity of immobilized vWf, which is only 20% of the concentration of immobilized vWf (18.5 fmol/mm 2 ). This is a typical percent for dextran-immobilized protein molecules accessible for ligand binding (39).
The k off for the fVIII⅐vWf complex was 14.8-fold lower than that for the LCh⅐vWf complex, whereas the k on values for these complexes were similar. The K d for fVIII⅐vWf binding in our experiments (0.4 nM) was similar to that of 0.2-0.46 nM determined in previous studies (5, 10, 17, 40) and 9.5-fold higher than that of LCh (3.8 nM), suggesting that HCh is involved in fVIII⅐vWf binding. Since HCh up to 2 M did not bind vWf (data not shown), its participation in fVIII binding to vWf is indirect.
Purification and Characterization of LCh Proteolytic Fragments-The V8 protease from S. aureus cleaves LCh into two fragments with molecular masses of 14 and 63 kDa, as determined by SDS-PAGE (Fig. 2, lane 3), which remain associated and are designated below as the 14-kDa/63-kDa complex. Since these fragments contain either the acidic region or C2, we purified them further for studies of the role of each region in vWf binding. Isolation of the individual fragments required denaturing conditions (9 M urea) and refolding as described under "Experimental Procedures." The refolded 14-and 63-kDa fragments exhibited fluorescent spectra with max at 332 and 335 nm, respectively, which indicate that Trp residues are partially buried, and therefore it is likely that fragments had regained a compact structure. Size exclusion chromatography on a Superose 12 HR (Pharmacia) column revealed that the fragments eluted as expected for polypeptides of 14 and 63 kDa, indicating the absence of intermolecular aggregates (data not shown).
Amino-terminal sequencing (10 cycles) of the purified 14and 63-kDa fragments demonstrated that they begin with residues 1672 and 1795, respectively (data not shown), indicating that S. aureus V8 protease cleaves LCh after Glu 1671 and Glu 1794 (Fig. 1). The molecular masses of the fVIII sequences 1672-1794 and 1795-2332 calculated from the deduced protein sequence of fVIII (41) are 14.4 and 63.5 kDa, respectively. Although the carboxyl termini of the fragments were not determined, the similarity of the calculated masses for the proposed fVIII sequences to that determined for the fragments by SDS-PAGE analysis and silver staining (Fig. 2, lanes 4 and 5), suggests that the termini (Fig. 1) are close to the expected residues 1794 and 2332.
If the isolated 14-and 63-kDa fragments are in a native conformation, it would be expected that the 14-kDa/63-kDa heterodimer can be reconstituted from the isolated fragments. As shown in Fig. 4C, the equimolar mixture of 14-and 63-kDa fragments elutes as a single peak by size-exclusion chromatography with a molecular mass Ͼ66 kDa. Analysis of the material from the peak by SDS-PAGE confirmed the presence of both 14and 63-kDa fragments (data not shown). The binding between 14-and 63-kDa fragments is apparently of high affinity, since a very little material contained in the peak corresponding to 14-kDa fragment in the equimolar mixture of 14-and 63-kDa fragments. Increasing proportion of the 63-kDa fragment in the mixture caused a steady reduction of the area of the peak corresponding to 14-kDa fragment while increasing that of the complex. A plot of the peak area of the 14-kDa fragments versus the molar ratio of the 63-kDa and 14-kDa fragments in the mixture revealed the formation of an equimolar complex (Fig. 4C, inset).
LCh Derivatives Lacking the Acidic Region Bind to vWf with the Low Affinity of C2-If no LCh regions other than C2 and the acidic region are involved in vWf binding, affinities of A3-C1-C2 and recombinant C2 domain (Fig. 1, lines 4 and 5) for vWf should be similar. As seen from the kinetic curves measured using the biosensor technique (Fig. 5), association of A3-C1-C2, C2, or 63-kDa fragment with vWf approached equilibrium within 1.5 min. The observed association and dissociation kinetics were optimally described by assuming one single class of vWf binding sites for each of the above ligands. The k on values for A3-C1-C2, C2, and 63-kDa fragment derived from the association curves (Fig. 5A) using Equation 3 (see "Experimental Procedures") were 6.4 ϫ 10 4 , 6.2 ϫ 10 4 , and 5.0 ϫ 10 4 M Ϫ1 s Ϫ1 (Table  I), respectively. The similar K d values for each ligand demonstrate that removal of the LCh acidic region leads to a similar reduction in vWf affinity for the remaining part of the LCh.
Since Ͻ0.003 nM LCh was present in 400 nM A3-C1-C2 used in these experiments (see "Experimental Procedures"), we measured binding of 0.003 nM LCh to vWf as a control, but none was observed during 2 min (data not shown). Preincubation (30 min, 37°C) of 400 nM A3-C1-C2 with 1200 nM anti-acidic region mAb NMC-VIII/10 (which inhibits fVIII⅐vWf binding; Ref. 15) had no effect on its binding to vWf (data not shown), which also indicates that A3-C1-C2 binding to vWf was not due to its contamination with residual acidic region in uncleaved LCh. In contrast, preincubation of A3-C1-C2 with anti-C2 mAb NMC-VIII/5 resulted in Ն95% reduction of its binding to vWf (Fig. 5), which demonstrates that A3-C1-C2 binding to vWf is specific and occurs only through the C2 domain.
To determine if HCh has any effect on A3-C1-C2 binding to vWf, the affinities of A3-C1-C2 and HCh/A3-C1-C2 het-  5A) and dissociation ( Fig. 5B) kinetics data are shown in Table I. The similar affinities of A3-C1-C2 (K d 625 nM) and HCh/A3-C1-C2 (K d 640 nM) for vWf suggest that HCh did not affect the affinity of A3-C1-C2 binding to vWf. The k on values for all fVIII derivatives lacking the acidic region were similar within 30% and 10 -13-fold lower than that for fVIII association with vWf ( Table I). The k off values for the above ligands were also similar and 143-170-fold greater than that for fVIII. The half-lives for dissociation of the ligand⅐vWf complexes, calculated as ln2/k off , were 17-21 s. These data suggest that removal of the acidic region results both in a slower rate of association of the ligands with vWf and in a much higher rate of dissociation of the formed complexes. The similar K d values for A3-C1-C2, 63-kDa fragment, C2, and HCh/A3-C1-C2 binding to vWf indicate the C2 domain contains the only vWf binding site within fVIII derivatives lacking the acidic region.
The 14-kDa LCh Fragment Binds to vWf with Low Affinity-The interaction of the 14-kDa fragment with immobilized vWf was tested at varying concentrations. The representative association and dissociation kinetics data are shown in Fig. 6 for three different concentrations of 14-kDa fragment. The k off and k on values (Table I) are the mean Ϯ standard deviation of k off and k on derived from the three association and dissociation sets of kinetic data (Fig. 6), using Equations 1 and 3, respectively (see "Experimental Procedures"). The values of k on and k off derived from separate curves were similar within 6% variation and did not depend on the concentration of the 14-kDa fragment. To determine if the K d value calculated as the k on /k off ratio is correct, we calculated the values of equilibrium binding corresponding to 50 and 200 nM of 14-kDa fragment from the equation where B is the equilibrium binding, F is the concentration of unbound ligand, and B max is the maximal binding capacity of immobilized vWf achieved at the concentration of 14-kDa fragment Ͼ Ͼ K d . The equilibrium binding values corresponding to 50, 200 and 1000 nM of 14-kDa fragment, were 8.7, 17.1, and 25.8 Arc s, respectively. These binding values correspond to 0.054, 0.108, and 0.158 ng of bound protein/mm 2 of the cuvette surface. Since the amounts of 14-kDa fragment bound to vWf immobilized on 1.8-mm 2 cuvette surface were 0.0968, 0.187, and 0.284 ng, corresponding to 50, 200, and 1000 nM of added 14-kDa fragment are much less than the amounts of added fragment (140, 560, and 2800 ng, respectively), the concentration of free ligand is similar to the concentration of added ligand. Assuming that B max is achieved at 1000 nM of 14-kDa fragment, the equilibrium binding values achieved at 50 and 200 nM 14-kDa fragment, calculated from the above equation are 0.055 and 0.11 ng/mm 2 , respectively, and similar to the experimentally observed values, 0.054 and 0.108 ng/mm 2 .
It was demonstrated in the above experiments that the LCh derivatives lacking the acidic region bind to vWf only via the C2 The fitted curves (solid lines) were obtained as in Fig. 3. In the experiment shown by É, binding of thrombin-cleaved 14-kDa fragment (1000 nM) was measured. In the experiment shown by छ, binding of the 14-kDa fragment (1000 nM) to a cuvette without vWf was measured. binding site. Therefore, it is expected that thrombin cleavage of the 14-kDa fragment at Arg 1689 completely abolishes its binding to vWf. The kinetics of cleavage of the 14-kDa fragment (1000 nM) or LCh (1000 nM) by 100 nM thrombin (see analysis under "Experimental Procedures") were similar (data not shown). The Ͼ95% conversion of the 14-kDa fragment into the 11.5-kDa product was achieved after 30 min at 25°C. The 11.5-kDa fragment (1000 nM) did not bind to immobilized vWf (Fig. 6), indicating that the acidic region binding site is destroyed by the cleavage. In control experiments, no binding of 14-kDa fragment to a biosensor cuvette without vWf was observed (Fig.  6). The 14-kDa fragment (1000 nM) also did not bind to immobilized anti-C2 mAb NMC-VIII/5 (data not shown), which indicates that it was not contaminated with 63-kDa fragment or LCh.
LCh and 14-kDa/63-kDa Complex Bind Similarly to vWf-We have shown that the individual 14-and 63-kDa fragments bind to vWf with low affinity. To determine if this is due to the separation of the fragments from each other or to the proteolytic cleavages by S. aureus V8 protease at positions 1672 and 1794 within the LCh, we also compared affinities of LCh and 14-kDa/63-kDa complex for vWf. The kinetic curves for LCh and 14-kDa/63-kDa interaction with vWf ( Fig. 7) demonstrated that these ligands bind similarly to vWf. The resonance signal produced by 14-kDa/63-kDa binding to vWf was approximately 5% lower than that for LCh, as expected due to the 4% lower molecular mass of the 14-kDa/63-kDa complex (77 kDa) compared with the LCh (80 kDa). The k off and k on values derived for 14-kDa/63-kDa complex and LCh from dissociation and association kinetics data using Equations 1 and 3, respectively, are similar (Table II). In a control experiment, the complex did not bind to a biosensor cuvette without immobilized vWf (Fig.  7). These results demonstrate that proteolytic cleavage within the LCh does not by itself lead to a lower affinity for vWf.
To test if the reconstituted 14-kDa/63-kDa complex has the same vWf binding properties as the original complex, its binding to immobilized vWf was measured as above. Binding of the both complexes was identical (data not shown), indicating that irreversible changes did not occur in the 14-and 63-kDa fragments as a result of their purification under denaturing conditions.
Determination of fVIII, LCh, 14-kDa Fragment, and C2 Affinity for vWf in Fluid Phase-In the above experiments vWf was covalently immobilized to a dextran-coated biosensor cuvette through amino groups. Since this procedure may potentially alter its binding properties, it is possible that the kinetic parameters derived from biosensor kinetic measurements may show deviations from those determined in a fluid phase assay. In order examine this possibility, binding between 125 I-labeled fVIII and vWf was competed by unlabeled fVIII or its fragments in the fluid phase (Fig. 8). The K d value for fVIII⅐vWf binding calculated from the best fit of the 125 I-fVIII homologous displacement by unlabeled fVIII was 0.38 Ϯ 0.04 nM. The values of inhibition constants (K i ) for LCh, 14-kDa fragment and C2 domain calculated from heterologous displacement data using the LIGAND program were 4.1 Ϯ 0.6 nM, 84 Ϯ 16 nM, and 470 Ϯ 94 nM, respectively. Since the heterologous displacement experiments were performed using 125 I-fVIII concentration 10 times below K d for fVIII⅐vWf binding, it would be expected that K i values determined for fVIII derivatives are similar to K d values for their direct binding to vWf. The similarity between the K i values derived from the fluid phase equilibrium binding data to the K d values (Table I) determined as the ratios of k off and k on kinetic rate constants derived from the association and dissociation biosensor measurements indicate that binding properties of immobilized vWf are similar to that in solution. In addition, thrombin-cleaved 14-kDa fragment was not inhibitory in the above assay (Fig. 8), which is also consistent with lack of its binding to immobilized vWf in biosensor experiments (Fig. 6).
Dissociation of the 14-kDa/63-kDa Complex Leads to a Conformational Change within the C2 Domain but Not within the Acidic Region-The high affinity of the 14-kDa/63-kDa complex for vWf may be due to a simple addition of the affinities of the individual fragments. Alternatively, at least one of the fragments may have the appropriate conformation required for maximal vWf binding affinity only when the other fragment is present. To distinguish between these two possibilities, we tested the conformations of the acidic region and the C2 domain in the individual fragments as well as in the 14-kDa/63-kDa complex to determine if they are the identical.
The affinity of the 14-kDa fragment for anti-acidic region mAb NMC-VIII/10 (epitope, residues 1674 -1684; Ref. 15) was compared with those for LCh and for the 14-kDa/63-kDa complex. The dissociation and association rate constants were derived for each ligand as before (data not shown). The similar k on , k off , and K d values for the interaction of the 14-kDa fragment, 14-kDa/63-kDa complex and LCh with mAb NMC-VIII/10 (Table II) suggest that their acidic region conformations are identical regardless of the presence of the 1795-2332 residues. To test if the presence of the acidic region is required for the optimal vWf binding conformation of the C2 domain, the affinity of the LCh and the 14-kDa/63-kDa complex for immobilized anti-C2 mAb NMC-VIII/5 (epitope, residues 2170 -2327), which is known to inhibit fVIII⅐vWf binding (22), was compared with that of A3-C1-C2, 63-kDa fragment, and C2. The k off , k on , and K d values for each ligand are shown in Table  II. The K d values for LCh and 14-kDa/63-kDa complex binding to NMC-VIII/5 were similar and approximately 10-fold lower than those for A3-C1-C2, 63-kDa fragment, and recombinant C2. The lower affinity of NMC-VIII/5 for LCh derivatives lacking the acidic region than that for intact LCh or 14/63-kDa fragment suggests that removal of the acidic region produces a conformational change within the C2 domain regardless of whether C2 is attached to other regions of the LCh.
To exclude the possibility that the acidic region of the LCh is a part of the NMC-VIII/5 epitope and the reduced affinity of A3-C1-C2, C2, or 63-kDa fragment for NMC-VIII/5 is simply due to lack of this region, we tested the effect of NMC-VIII/5 on binding of biotin-labeled NMC-VIII/10 to fVIII immobilized on an anti-HCh antibody 413 (epitope within A2 domain residues 373-616). Lack of any effect of NMC-VIII/5 on NMC-VIII/10 binding suggests, that the epitope of NMC-VIII/5 has no overlap with the NMC-VIII/10 epitope (acidic region residues 1675-1684). In addition, the acidic region synthetic peptide 1670 -1684, which completely inhibited binding of biotin-labeled NMC-VIII/10 to immobilized fVIII at 1 M, did not have any effect on binding of biotin-labeled NMC-VIII/5 up to 100 M (data not shown).

DISCUSSION
Several previous studies had identified mAbs that recognize epitopes in the acidic region of the fVIII LCh (residues 1649 -1689) and that also prevent the formation of the fVIII⅐vWf complex (14 -17). Deletion of the entire acidic region (1649 -1689) eliminated vWf binding, whereas deletion up to 1668 did not (18). However, synthetic peptide 1673-1689 at a molar excess of 2,500 over fVIII was not able to inhibit its binding to vWf (18). Our approach to this apparent contradiction was to assume that the synthetic peptide did not have the appropriate conformation for vWf binding and that a larger proteolytic LCh fragment might be more suitable. We were able to cleave the fVIII LCh with S. aureus V8 protease to generate a 14-kDa fragment consisting of amino acids 1672 to approximately 1794, and we have shown in the present study that it does indeed bind to vWf.
In our study we used a surface plasmon resonance phenomenon for direct real time measurement of association and dissociation of unlabeled proteins and subsequent determination of the corresponding rate constants (42). Since one binding partner must be immobilized to a matrix, the kinetic parame-ters derived from optical biosensor kinetic measurements may potentially show deviations from those determined in a fluid phase assay. We were able to demonstrate, however, that the K d values determined for fVIII, LCh, 14-kDa, and C2 binding to vWf using the biosensor method (Table I) were similar to the K i values of 0.38, 4.1, 84, and 470 nM for the respective fragments derived from the fluid phase assay in which binding between 125 I-fVIII and vWf was competed by unlabeled ligands. This result indicates that vWf was not altered by immobilization and therefore the kinetic parameters for binding of fVIII and its derivatives to vWf derived in our experiments are valid. The high sensitivity of this technique and its capability to register fast dissociation kinetics has allowed us to determine up to 2,000-fold lower affinities of fVIII proteolytic fragments for vWf than that of fVIII.
All LCh derivatives lacking the acidic region (A3-C1-C2, C2, and 63-kDa fragment) had greatly reduced affinities for vWf (K d 564 -660 nM) compared with LCh (K d 3.8 nM) (Table I). In addition, the similar affinities of A3-C1-C2, 63-kDa fragment, C2, and HCh/A3-C1-C2 for vWf indicated that the LCh contains no vWf binding site other than the one within C2. The 14-kDa fragment lacking the C2 domain also had a lower affinity (K d 72 nM) for vWf than LCh. In contrast, the proteolytically cleaved but undissociated complex containing both the acidic region and C2 (14 kDa/63 kDa) had an affinity identical to that of the intact LCh. Reassociation of the 14-kDa/63-kDa complex from the individual 14-and 63-kDa fragments and the demonstration that its binding to vWf is similar to that of the original 14-kDa/63-kDa complex indicated that the above fragments were not altered by purification under denaturing conditions. The previous hypotheses that the LCh acidic region contains a vWf binding site and that both this region and the C2 domain are essential for high affinity binding are therefore confirmed. Our data also demonstrate that the two binding sites must be simultaneously present, although not necessarily covalently linked, for maximal affinity vWf binding to occur. Our data also demonstrate that the residues 1649 -1671 are not involved in vWf binding, which is consistent with previous findings (18). We demonstrated that neither the acidic region peptide 1672-1689 nor the COOH-terminal 11.5-kDa fragment derived by the thrombin cleavage of 14-kDa fragment were able to bind vWf in biosensor experiments or inhibit fVIII⅐vWf interaction in fluid phase assays. We hypothesize, therefore, that the acidic region vWf binding site extends COOH-terminal to the thrombin cleavage site at residue 1689 and that it is destroyed by thrombin cleavage. This hypothesis could explain why synthetic peptide 1673-1689 (18) did not bind vWf. The decreased affinity of the 14-kDa fragment affinity for vWf is due to a lower association rate constant (k on ) than that for LCh (Table I) since the dissociation rate constants determined for the 14-kDa fragment and LCh complexes with vWf were simi-  8. Displacement of fVIII by its fragments from vWf in fluid phase. 125 I-fVIII (0.04 nM) and vWf (5 nM) were incubated in the presence of varying concentrations of unlabeled fVIII (å), LCh (q), 14-kDa fragment (E), C2 (Ç), or thrombin-cleaved 14-kDa fragment (ϫ), which was followed by determination of 125 I-fVIII⅐vWf binding (see "Experimental Procedures"). 125 I-fVIII⅐vWf binding in the presence of unlabeled competitor is expressed as the percentage of 125 I-fVIII binding when no competitor was added. Each point represents the mean value Ϯ S.D. of triplicates. The curves show a best fit of the data to a model describing heterologous ligand displacement from a single class of binding sites using the computer program LIGAND. lar. For the C2 domain the k off value was 10-fold greater than that for LCh or 14-kDa fragment. These results suggest that the interaction of the LCh acidic region with vWf is the ratedetermining step for the dissociation of LCh⅐vWf and fVIII⅐vWf complexes in the absence of thrombin activation. The loss of the acidic region leads to a 160-fold increased k off and a Ͼ1000-fold increased K d for HCh/A3-C1-C2 heterodimer binding to vWf compared with that of fVIII. This would predict a similar reduction of fVIII affinity upon thrombin activation, allowing efficient fVIIIa binding to the phospholipid surface required for its maximal activity in the factor Xase enzyme complex.
We demonstrated that the LCh acidic region not only directly participates in vWf binding, but it is probably also required to maintain the normal conformation of the C2 binding site. We used anti-C2 mAb NMC-VIII/5, which prevents fVIII⅐vWf binding (19,22), as a probe to detect possible conformational changes in C2 upon removal of the acidic region. The reduced affinities for NMC-VIII/5 of LCh derivatives lacking the acidic region indicated that such a change occurs. Similar affinities of NMC-VIII/5 for the LCh and the 14-kDa/63-kDa complex demonstrated that noncovalent association of the amino-terminal 14-kDa fragment with the carboxyl-terminal 63-kDa fragment appeared to be sufficient to maintain the C2 conformation similar to that within intact LCh. In contrast, the conformation of the acidic region does not depend on the presence of C2, since the affinity of the anti-acidic region mAb NMC-VIII/10 for the 14-kDa fragment and the LCh was similar.
Our results are consistent with the hypothesis that the light chain acidic region and C2 are in close proximity and together form one high affinity binding site for vWf. The computer modeling of the three-dimensional structure of the fVIII A domains, based on their structure in ceruloplasmin, predicts that the carboxyl and amino termini of A3 are in close proximity (43). In addition, the disulfide bond determined between Cys 2021 and Cys 2169 and that proposed between Cys 2174 and Cys 2326 (44) demonstrate that the amino and carboxyl termini of the C1 (residues 2019 -2172) and C2 (residues 2173-2332) domains, respectively, are also spatially close. These findings suggest that the acidic region located at the amino terminus of the LCh and the carboxyl terminus of the C2 domain may also be close together in the three-dimensional structure of the LCh. Our findings that the LCh acidic region and C2 together form the high affinity vWf binding site is consistent with this model. The inhibition of high affinity fVIII binding to mature vWf residues 1-116 (45) by both anti-LCh acidic region and anti-C2 mAbs (21) would also fit this model.
The K d for LCh binding to vWf (3.8 nM) is 9.5 times higher than that for fVIII (0.4 nM), demonstrating that the HCh is required for the maximal affinity of fVIII for vWf. The higher affinity of fVIII⅐vWf interaction than that of LCh⅐vWf is mainly due to the lower dissociation rate of the former complex (Table  I). The participation of the HCh in fVIII⅐vWf binding is indirect because the HCh itself did not bind to vWf in our experiments or as previously demonstrated by ultracentrifugation (8). Recently Sudhakar and Fay observed that dissociation of the HCh⅐LCh complex led to conformational changes in each chain (46). It is therefore possible that the LCh has to be associated with HCh to have the optimal conformation for its binding to vWf. Once the LCh acidic region is removed by thrombin cleavage, the HCh cannot enhance the binding of A3-C1-C2 to vWf.