Activation of Factor VIII by Thrombin Increases Its Affinity for Binding to Synthetic Phospholipid Membranes and Activated Platelets*

Membrane-bound thrombin-activated factor VIII (fVIIIa) functions as a cofactor for factor IXa in the factor Xase complex. We found that binding of heterotrimeric fVIIIa (A1·A2·A3-C1-C2) to synthetic vesicles with a physiologic content of 4% phosphatidylserine (PS), 76% phosphatidylcholine, and 20% phosphatidylethanolamine occurs with a 10-fold higher affinity than that of factor VIII (fVIII). The increased affinity of fVIIIa for PS-containing membranes resulted from the reduced rate of fVIIIa dissociation from the vesicles compared with that of fVIII. Similar affinities of A3-C1-C2, A1·A2·A3-C1-C2, and A3-C1-C2·heavy chain for interaction with PS-containing membranes demonstrate that removal of the light chain (LCh) acidic region by thrombin is responsible for these increased affinities of fVIIIa and its derivatives. Similar kinetic parameters of fVIII and its LCh and C2 domain for binding to PS-containing membranes and to activated platelets indicated that the C2 domain is entirely responsible for the interaction of fVIII with membranes. We conclude that the increased fVIIIa affinity for PS-containing membranes is a result of conformational change(s) within the C2 domain upon removal of the acidic region of the LCh. This conclusion is based on the finding that binding of the monoclonal antibody ESH8 to the C2 domain, which is known to prevent this conformational transition, resulted in fVIIIa binding to PS/phosphatidylcholine/phosphatidylethanolamine vesicles (4/76/20) with a lower affinity similar to that of fVIII. In addition, stabilization of the low affinity binding conformation of the C2 domain of fVIIIa by this antibody led to an inhibition of the fVIIIa activity in the factor X activation complex.

The plasma glycoprotein factor VIII (fVIII) 1 functions as a cofactor for the factor X activation complex (factor Xase) in the intrinsic pathway of blood coagulation (1). Within the factor Xase complex, thrombin-activated factor VIII (fVIIIa) associated with membranes of activated platelets (2,3) or with synthetic phospholipid vesicles (4) binds to factor X (5) and to activated factor IX (4,6). Activation of factor X by the fVIIIa⅐ activated factor IX complex assembled on a membrane is 100,000-fold more efficient than in the absence of phospholipid (7). Assembly of the factor Xase complex in vivo is localized to sites of vascular damage where the activated platelets are adherent (2,3).
The fVIII protein consists of homologous A and C domains and a unique B domain, which are arranged in the order A1-A2-B-A3-C1-C2 (8). It is processed to a series of Me 2ϩlinked heterodimers produced by cleavage at the B-A3 junction (9), generating a light chain (LCh) consisting of the acidic region (AR) and A3, C1, and C2 domains and a heavy chain (HCh), which consists of the A1, A2, and B domains (Fig. 1).
The site involved in fVIII binding to synthetic phospholipid vesicles or platelets was localized to the C2 domain residues 2303-2332 (10). The presence of at least 8% phosphatidylserine (PS) is required for fVIII binding to synthetic PS/phosphatidylcholine (PC) membranes (11). The additional presence of phosphatidylethanolamine (PE) induces high affinity binding sites for fVIII on membranes with physiologic (Ͻ8%) molar fractions of PS (12). Expression of fVIII binding sites on platelets occurs only upon their activation by thrombin or other agonists (2,3). This leads to the reorientation of PS and PE from the inner to the outer layer of the plasma membrane (13,14) to provide sufficient concentrations of PE and PS for the formation of fVIII binding sites. Maintenance of a normal fVIII level in the circulation is dependent on its complex formation with vWf, which prevents fVIII from binding to PS-containing membranes (15,16) and to activated platelets (17). Cleavage of the LCh at Arg 1689 releases the acidic region residues 1649 -1689 and leads to the dissociation of fVIIIa from vWf (18 -20). Activation of fVIII by thrombin cleavage at Arg 372 , Arg 740 , and Arg 1689 (21) results in at least a 100-fold increase of cofactor activity. The product, fVIIIa, is a A1⅐A2⅐A3-C1-C2 heterotrimer (22) in which domains A1 and A3 retain the metal ion linkage (Fig. 1) and the stable dimer A1⅐A3-C1-C2 (23) is weakly associated with the A2 subunit mainly through electrostatic forces (22). Spontaneous dissociation of the A2 subunit from the dimer results in nonproteolytic inactivation of fVIIIa.
A conformational change(s) occurs within the C2 domain upon removal of the acidic region of the LCh (residues 1649 -1689), leading to the loss of the optimal binding conformation of the vWf site within the C2 domain of thrombin-cleaved LCh (A3-C1-C2) (24). We hypothesized that this change(s) could also affect fVIIIa affinity for phospholipid. In this report, we determined the effect of activation of fVIII by thrombin on its binding to synthetic phospholipid PS/PC surfaces with a high content of PS, as well as its binding to phospholipid vesicles containing physiological mole fractions of PS, PE, and PC and to membranes of activated platelets. To elucidate the role of the conformational changes occurring in the C2 domain upon removal of the acidic region of the LCh (24), we used fragments of fVIII and fVIIIa for measurement of their binding to phospholipids.
Preparation of fVIIIa-Purified fVIIIa, generously provided by Baxter Biotech Group (Duarte, CA), was prepared from recombinant fVIII (31). Activated plasma-derived fVIII was prepared by activation of fVIII (2.6 M) by thrombin (0.1 M) for 10 min at 37°C in 20 mM HEPES, pH 7.4, 0.15 M NaCl (HEPES-buffered saline) containing 5 mM CaCl 2 . Activation was stopped by hirudin (0.15 M), and the pH was adjusted to 6.0 by the addition of 0.2 M MES.
Factor Xase Assay-Activity of fVIIIa was measured using a chromogenic factor Xase assay (25). In the assay, the limiting concentration of fVIIIa (0.2 nM) (25) was added to a solution containing PS/PC vesicles (20 M), factor IXa (2 nM), and factor X (300 nM), and the reaction was initiated by the addition of 5 mM CaCl 2 . Aliquots were withdrawn after 15, 30, 45, and 60 s, and factor X activation was stopped with 0.05 M EDTA. Factor Xa generation was measured by cleavage of 0.3 mM synthetic substrate S-2765 (Amersham Pharmacia Biotech) using a V max microplate reader (Molecular Devices). A purified factor Xa standard (Enzyme Research Laboratories) was used to convert absorbance (410 nm) into factor Xa concentration.
Radiolabeling of fVIII-FVIII was iodinated as described (24). The specific radioactivity of fVIII was 4 Ci/g of protein. The activity of 125 I-fVIII determined in the one-stage clotting assay (3800 units/g) was similar to that of unlabeled fVIII.
Preparation of Phospholipid Vesicles-Phospholipids PS, PC, and PE were purchased from Sigma, and biotin-LC-dipalmitoyl-PE was purchased from Pierce. Phospholipid vesicles with various PS, PC, and/or PE content and vesicles containing biotin-LC-dipalmitoyl-PE (0.5%) were prepared as described (32).
Protein-Phospholipid Binding Measurements Using Biosensor Tech-nology-The kinetics of protein-phospholipid interaction were determined by surface plasmon resonance using the Biacore biosensor instrument (Amersham Pharmacia Biotech, Sweden) or IAsys biosensor (FISONS, United Kingdom), which measure protein binding and subsequent dissociation in real time (33). Binding and subsequent dissociation was measured in HEPES-buffered saline, pH 7.4, 5 mM CaCl 2 at 22°C for all ligands. A supported PS/PC or PC monolayer was formed on the surface of an HPA hydrophobic chip (Amersham Pharmacia Biotech) by incubation of unilamellar PS/PC (25/75) or 100% PC vesicles at 400 g/ml in HEPESbuffered saline for 20 min at 22°C, which produced a signal of 1100 resonance units (RU). The phospholipid-coated chip was blocked by 0.1 mg/ml bovine serum albumin for 20 min. Binding of fVIII and its derivatives to a supported PS/PC or PC monolayer was measured using the Biacore biosensor instrument, where 1 ng of protein bound per mm 2 of the biosensor chip produces a resonance signal of 1000 RU. To regenerate the cuvette, complete dissociation of bound ligands was achieved by the addition of 10 mM NaOH for 30 s.
Binding to intact vesicles was measured using the IAsys biosensor, where 1 ng of protein bound per mm 2 of the biosensor chip produces a signal of 600 RU. Biotin-coated cuvettes (Affinity Sensors) were incubated with 20 g/ml streptavidin (Sigma) for 10 min, followed by the addition of biotinylated PS/PC/PE (4/76/20) or PC/PE (80/20) vesicles at 200 g/ml for 10 min. The resonance response due to binding of biotinylated phospholipid vesicles was 1000 Arc seconds. To regenerate the cuvette, complete dissociation of the biotin-streptavidin complex was achieved by the addition of 5 M NaOH for 2 min.
Calculation of the Kinetic Parameters from Biosensor Kinetic Data-The values of the rate constants for the dissociation (k off ) of fVIII and its derivatives from monolayers or immobilized intact vesicles were determined by fitting the dissociation kinetics data to the following equation describing a 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. Ref. 34 showed the following, where R max is the maximal binding capacity of the immobilized ligand surface expressed in resonance units (Arc seconds) and C is the concentration of polypeptide in solution.
In Figs. 2-5, the values of k on were determined from individual association kinetics data using the integrated form of the rate equation The values of k on and R max were derived from nonlinear regression analysis fitting R versus t to Equation 3. 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 .
In Figs. 6 and 7, the values of k on and K d were determined from multiple association curves as follows. A linear fit of dR/dt versus R yields an apparent first order association rate constant corresponding to each concentration of fVIIIa or ESH8⅐fVIIIa, The value of k on was determined from the linear fit of k s versus C (concentration of fVIIIa or ESH8⅐fVIIIa) to Equation 4. The K d was derived from the best fit of the resonance response corresponding to an equilibrium ligand binding (R e ) achieved at different ligand concentrations versus F, the concentration of unbound ligand, to Equation 5, Under the experimental conditions, the concentration of bound ligand was much less than the concentration of added ligand, indicating that the concentration of unbound ligand (F) is similar to the concentration of added ligand. The k off value was calculated as K d ϫ k on using the values of K d and k on determined as above. All the fitting procedures were performed using Sigmaplot 1.02 (Jandel Scientific).
Determination of the Affinity for fVIII, LCh, and A3-C1-C2 Binding to Activated Platelets-The binding affinities of fVIII or its fragments for activated platelets were determined by homologous or heterologous ligand displacement assays. Platelets were isolated from platelet-rich plasma and activated by thrombin as described (3). 125 I-fVIII (0.25 nM) and increasing concentrations of unlabeled fVIII, C2, LCh, or A3-C1-C2 were incubated with activated platelets in Tyrode's solution (35) at 22°C for 30 min. Aliquots (75 l) from the samples were loaded onto 20% sucrose (350 l) and centrifuged for 1 min at 10,000 ϫ g. Negative controls contained unactivated platelets and all the other components. The negative control values (Յ5% of the maximal binding values) were subtracted from average values of triplicates of all other samples. For heterologous ligand displacement studies, the data were analyzed assuming two distinct equilibria, R ϩ L 1 i RL 1 and R ϩ L 2 i RL 2 , where R is the concentration of fVIII binding sites on activated platelets, L 1 is 125 I-fVIII, and L 2 is C2, LCh, or A3-C1-C2. These equilibria are described by the equilibrium constants K d and K i , respectively. The data of homologous and heterologous displacements were fitted to a model assuming a single class of fVIII binding sites on activated platelets using the LIGAND (36) program.

RESULTS
The C2 Domain within the LCh of fVIII Is Entirely Responsible for fVIII Binding to Phospholipid Monolayers and Vesicles-Since in most previous fVIII binding studies, the synthetic phospholipid membranes were composed of 15-25% PS and PC (4,11,15,37), in the present studies we also determined parameters for binding to immobilized PS/PC (25/75) monolayers formed on a hydrophobic surface, which represents a stable and structurally defined lipid environment resembling cell membranes (38).
To determine whether LCh is entirely responsible for the high affinity fVIII interaction with phospholipid membranes, we compared the kinetics of fVIII and LCh binding to and dissociation from immobilized phospholipid monolayers or intact phospholipid vesicles. Values of second order association rate constants (k on ) ( Table I) were determined from the representative curves ( Fig. 2A) showing the resonance response of fVIII or LCh association with immobilized PS/PC monolayer over time. The half-lives for the dissociation of fVIII or LCh from phospholipid monolayers were approximately 8 min and 7 min, respectively (Fig. 2B). The k off values calculated for fVIII and LCh using the equation k off ϭ ln2/half-life were 1.44 ϫ 10 Ϫ3 s Ϫ1 and 1.65 ϫ 10 Ϫ3 s Ϫ1 , correspondingly, similar to the respective values calculated for the best fits of the dissociation curves to Equation 1 (Table I).
In a control experiment, we examined whether binding characteristics of PS/PC (25/75) monolayers are similar to those of intact vesicles of identical composition. We determined fVIII and LCh parameters for binding to immobilized small unilamellar PS/PC vesicles, containing 0.5% PE-biotin for immobilization to an SA-coated chip via biotin (kinetic data not shown). The k on and k off values for fVIII binding to such vesicles were 6.0 ϫ 10 5 M Ϫ1 s Ϫ1 and 1.22 ϫ 10 Ϫ3 s Ϫ1 , while for LCh binding the values were 6.2 ϫ 10 5 M Ϫ1 s Ϫ1 and 1.38 ϫ 10 Ϫ3 s Ϫ1 and are similar to the corresponding values for fVIII and LCh interaction with the PS/PC monolayer (Table I).
To elucidate if LCh is also entirely responsible for fVIII binding to membranes containing PS, PC, and PE at molar fractions similar to those in the membranes of activated platelets (12), we compared kinetic parameters for fVIII and LCh interaction with PS/PC/PE (4/76/20) membranes. In preliminary experiments, we found that a planar phospholipid monolayer containing physiologic ratios of PS, PC, and PE (4/76/20) formed on the hydrophobic biosensor chip surface was not able to support binding of fVIII and its fragments. This is consistent with the observation that fVIII binds in solution only to small PS/PC/PE (4/76/20) vesicles but does not bind to large vesicles, indicating that the greater degree of curvature of the membrane is a critical requirement for fVIII binding to membranes with low content of PS. 2 Therefore, parameters of fVIII and its derivatives for binding to membranes with physiologic concentrations of PS, PC, and PE were determined in our study using immobilized small intact PS/PC/PE vesicles. Kinetics of fVIII and LCh association with and dissociation from immobilized small sonicated PS/PC/PE (4/76/20) vesicles ( Fig. 2, C and D, respectively) were optimally described by assuming one single class of binding sites on the vesicle for each ligand. The k on values for fVIII and LCh derived from association curves (  Table II were 2.7-and 2.6-fold lower than the respective values for fVIII and LCh binding to a PS/PC monolayer (Table I). As seen from Tables I and II, the decrease of PS content in the membrane from 25 to 4% reduces only the rate of fVIII or LCh association with the membrane and does not affect the dissociation rates determined from Fig.  2, B and D. The K d values for fVIII and LCh binding to intact PS/PC/PE vesicles were similar (Table II). This is consistent with the lack of binding of HCh (up to 3 M) to PS/PC monolayers and PS/PC/PE vesicles (data not shown). In order to verify that the association rate constants for fVIII binding to the phospholipid PS/PC monolayer and small PS/PC/PE vesicles were not underestimated as a result of the diffusion-controlled kinetics of fVIII association with phospholipid membrane, we measured association kinetics using a 4-fold higher mixing speed ( Fig. 2A) and a 5-fold higher flow rate (Fig. 2C) to minimize the diffusion limitations (39). Kinetic data obtained under these conditions were similar to those obtained at the lower mixing speed or lower flow rate.
If regions of fVIII LCh other than the C2 domain are involved 2 G. Gilbert, personal communication.
in membrane binding, the affinity of the isolated C2 domain for binding to phospholipid would be lower than that of LCh. Therefore, we compared parameters for LCh and C2 interactions with the phospholipids. Representative kinetic data for C2 association with and dissociation from PS/PC (25/75) monolayers are shown in Fig. 3, A and B, respectively, for three different C2 domain concentrations. The k off and k on values (Table I) are the mean Ϯ S.D. of k off and k on derived from the three association and dissociation sets of kinetic data shown in Fig. 3, A and Tables  I and II).  (Tables I and II) were 4.4 or 7.8 times lower than the respective values for the intact LCh. This is in a good agreement with the observed longer half-lives of the A3-C1-C2 complexes with phospholipids than those of LCh. The k on values for A3-C1-C2 association with the monolayer or intact vesicles determined from Fig. 4, A and C, were similar to those for LCh (Tables I  and II). The K d values for A3-C1-C2 binding to the monolayer or intact vesicles were 6 -8 times lower than those for LCh.
To examine if HCh has any effect on A3-C1-C2 binding to the phospholipids, the kinetic parameters for A3-C1-C2 and the HCh⅐A3-C1-C2 heterodimer interaction with the phospholipid monolayers and with the intact vesicles were determined from the data shown in Fig. 4. Similar kinetic parameters for A3-C1-C2 and HCh⅐A3-C1-C2 (Tables I and II) suggest that HCh does not affect the affinity of A3-C1-C2 binding to the phospholipids.
Since the uncleaved HCh, consisting of A1, A2, and B domains, does not have any effect on the affinity of A3-C1-C2 interaction with the phospholipids, one might expect that isolated A1 or A2 domains would also be unable to affect the interaction of A3-C1-C2 with phospholipids and, therefore, that the affinity of heterotrimeric fVIIIa (A1⅐A2⅐A3-C1-C2) would be similar to that of A3-C1-C2. To test this hypothesis, we meas-ured the affinity of fVIIIa for the monolayers and vesicles.
Inactivation of A1⅐A2⅐A3-C1-C2 due to the equilibrium A1⅐A2⅐A3-C1-C2 i A2 ϩ A1⅐A3-C1-C2 (equilibrium 1) occurs to a significantly lower extent at pH 6.0 than at pH 7.4 (31, 40) due to a higher affinity for A2 interaction with A1⅐A3-C1-C2 at pH 6.0 than at pH 7.4. Since addition of the exogenous A2 subunit prevents this inactivation (40,41), association of fVIIIa with and dissociation from PS/PC/PE (4/76/20) vesicles was studied at pH 6.0 in the presence of 200 nM A2. We examined the stability of fVIIIa in the presence of 200 nM A2 using the factor Xase assay for determination of fVIIIa activity. We found that Ͼ85% of the initial fVIIIa activity was retained within 90 min at fVIIIa concentrations of 1-10 nM in the presence of 200 nM A2 (data not shown). This is consistent with the percentage

FIG. 3. Determination of the kinetic parameters for C2 domain binding to phospholipid monolayers and intact vesicles.
The association (A) of C2 with PS/PC (25/75) monolayer and corresponding dissociation (B) kinetic data were obtained at 6 (Ⅺ), 24 (E), and 60 nM (‚) of C2. The original kinetic data and fitted curves (solid lines) were obtained as in Fig. 2. In the control experiment (ƒ), binding of C2 (60 nM) to a control 100% PC monolayer was measured. In C and D, C2 association with and dissociation from PS/PC/PE (4/76/20) vesicles was measured at 6 (Ⅺ), 24 (E), and 60 nM (‚) of C2 as described in Fig. 2C. The solid lines show the fitted curves obtained as above. In the control experiment (ƒ), binding of C2 (60 nM) to PCPE (80/20) vesicles was measured.
of heterotrimeric fVIIIa (88%) calculated according to equilibrium 1 using a previously reported value of the A2 affinity for A1⅐A3-C1-C2 at pH 6.0 (K d ϭ 28 nM (40)). This indicates that decay of fVIIIa during the time required for equilibrium binding was not significant and therefore that the resonance response curves shown in Fig. 5A reflect the binding of integral heterotrimeric fVIIIa. The association rate constants determined for fVIIIa (Table II) and for fVIII (k on ϭ 2.36 ϫ 10 5 M Ϫ1 s Ϫ1 ) from the association kinetics (Fig. 5A) were similar. In contrast, kinetics of fVIIIa dissociation from vesicles were significantly slower than that for fVIII (Fig. 5B). The half-life for the dissociation of fVIIIa (50 min) was 8 times longer than that for fVIII, which is consistent with a 7.4-fold lower k off value derived from dissociation fVIIIa kinetics than the k off value for fVIII. The K d value for fVIIIa binding to PS/PC/PE vesicles (Table II) calculated as k on /k off ratio was 8 times lower than the K d determined for fVIII at pH 6.0. Parameters for the binding of fVIII and its fragments to the PS/PC/PE vesicles at pH 6.0 (kinetic curves are not shown) are similar to those determined at pH 7.4 (see Table II). This result predicts that the affinity of fVIIIa for synthetic phospholipid membranes at pH 7.4 would be similar to that determined at pH 6.0 and higher than that of fVIII. Since purified fVIIIa was prepared by chromatography of thrombin-activated recombinant fVIII (31), in a control experiment affinities of recombinant fVIIIa and fVIIIa prepared by activation of purified plasma-derived fVIII (as described under "Experimental Procedures") were compared under the above conditions. The values of k on and k off , (3.55 Ϯ 0.13) ϫ 10 5 M Ϫ1 s Ϫ1 and (2.12 Ϯ 0.08) ϫ 10 Ϫ4 s Ϫ1 determined for activated plasma fVIII (data not shown) were similar to those of recombinant fVIIIa (Table II), indicating that the measured affinities were not affected by the origin of fVIII or the fVIIIa isolation procedure.
Since even at pH 6.0 a spontaneous dissociation of A2 from fVIIIa may contribute to the observed kinetics of fVIIIa dissociation from phospholipid, we confirmed the parameters for fVIIIa binding to PS/PC/PE vesicles determined from association and dissociation kinetics using an alternative method. In this case, the binding parameters were derived from the equilibrium fVIIIa binding data obtained in the presence of exogeneous A2. In this experiment, the value of K d was determined by fitting the values of equilibrium binding (B e ) corresponding to increasing fVIIIa concentrations using Equation 5. Representative association curves corresponding to various concentrations of fVIIIa are shown in Fig. 6A. The k on value of (3.05 Ϯ 0.08) ϫ 10 5 M Ϫ1 s Ϫ1 was derived from the curves 1-6 ( Fig. 6A) using Equation 4. The K d value of 0.6 Ϯ 0.12 nM was determined from B e corresponding to different fVIIIa concentrations (Fig. 6B). The k off value calculated as K d ϫ k on from the above K d and k on values, was 1.8 ϫ 10 Ϫ4 s Ϫ1 . The similarity of the K d and k off values determined from equilibrium kinetics to those presented in Table II, determined from the dissociation kinetics (Fig. 5B), validates the finding that fVIIIa binds to PS/ PC/PE vesicles with higher affinity than fVIII.

The Increased Affinity of fVIII Derivatives Lacking the Acidic Region of the Light Chain Is Related to Conformational
Changes within the C2 Domain-It was previously demonstrated that the acidic region of the LCh is required to maintain an optimal binding conformation of the C2 binding site for vWf, which has some overlap with the PS binding site (24,25,42). Removal of the acidic region of the LCh produces a conformational change within C2 (24), leading to a loss of vWf binding. This loss of fVIIIa affinity for vWf, however, can be prevented if the ESH8 antibody, which recognizes the C2 domain epitope within residues 2248 -2285, is bound to fVIII during the thrombin activation step (25). If an increase in A3-C1-C2 affinity for binding to phospholipid is related to conformational change(s) within C2 leading to the loss of vWf binding, one might expect that in the presence of ESH8 these changes would not occur and therefore that the affinity of ESH8⅐fVIIIa for the phospholipids would be similar to that of fVIII.
To test this hypothesis, we prepared the ESH8⅐fVIIIa complex and determined the parameters for its binding to PS/ PC/PE (4/76/20) vesicles at pH 6.0. As seen from Fig. 7A, the resonance response (2260 RU) corresponding to equilibrium binding of ESH8⅐fVIIIa at saturating concentration (64 nM) was 1.95 times higher than that at saturating concentration (16 nM) of fVIIIa (Fig. 6A), as expected due to the 1.9-fold higher molecular mass of the ESH8⅐fVIIIa (316 kDa) than fVIIIa (160 kDa). The maximal binding capacities of vesicles for fVIIIa and ESH8⅐fVIIIa calculated from the above resonance data were similar (38.3 and 35.7 fmol/mm 2 , respectively), indicating that the ESH8⅐fVIIIa complex binds on the same sites as fVIIIa. The K d value shown in Table II was determined from the B e corresponding to different ESH8⅐fVIIIa concentrations using Equation 5, as shown in Fig. 7B. Similar kinetic parameters for ESH8⅐fVIIIa and unactivated fVIII interaction with the vesicles (see Table II) suggest that ESH8 antibody blocks the conformational changes within C2 that are responsible for the increase in fVIII affinity for the phospholipids upon thrombin activation.
Our finding that in the presence of ESH8, the affinity of fVIIIa for binding to PS/PC is lower than that of fVIIIa suggests that under conditions where both the concentration of membrane binding sites and of fVIIIa or ESH8⅐fVIIIa is below the K d for membrane binding, the concentration of membranebound ESH8⅐fVIIIa will be lower than that of fVIIIa. If this is true, the formation of membrane-bound factor VIIIa-factor IXa complex will also be inhibited in the presence of ESH8. We therefore examined the formation of factor VIIIa-factor IXa complex in the presence and absence of ESH8. We used a fVIIIa concentration below K d for its binding to PS/PC (25/75) membranes with decreasing concentrations of vesicles. The assembly of fVIIIa-factor IXa complex was determined by measuring the ability of the complex to activate factor X to factor Xa (7). As seen from Fig. 8, a reduction of concentration of PS/PC vesicles below 10 M results in more than a 65% reduction in enzymatic activity of fVIIIa-factor IXa complex. This is consistent with a lower affinity of ESH8⅐fVIIIa complex than fVIIIa for PS/PC membranes. In contrast, at PS/PC concentrations Ն 10 M, there was no inhibition of factor Xa formation in the presence of ESH8 (Fig. 8), suggesting that a saturating concentration of PS/PC sites for ESH8⅐fVIIIa binding was achieved.
Determination of fVIII, LCh, and A3-C1-C2 Binding Affinities for Platelets in Fluid Phase-Since the membrane proteins and carbohydrates may be involved in formation of fVIII and fVIIIa binding sites on platelets, the effect of thrombin activation on fVIII binding to platelets sites may not be similar to that on binding to synthetic phospholipid membranes. To examine if our observations made using synthetic phospholipid membranes are valid for membranes of platelets, we determined the affinities of fVIII, LCh, C2, and A3-C1-C2 for activated platelets using a heterologous displacement assay. In this assay, binding of 125 I-labeled fVIII to activated platelets was competed by unlabeled fVIII and its fragments in the fluid phase (Fig. 9). The K d value for fVIII binding to platelets (8.4 Ϯ 0.62 nM) was calculated from the best fit of 125 I-fVIII homolo-gous displacement by unlabeled fVIII. Values for inhibition constants (K i ) for LCh, A3-C1-C2, and C2 domain were calculated from heterologous displacement data using the LIGAND program and were 7.7 Ϯ 1.2, 1.1 Ϯ 0.16, and 8.8 Ϯ 0.56 nM, respectively. Since the heterologous displacement experiments were performed using 125 I-fVIII concentrations that were 30 times below the K d for fVIII binding to platelets, the K i values determined for fVIII derivatives should be similar to K d values for their direct binding to the activated platelets. DISCUSSION We have demonstrated that the activated heterotrimeric fVIII (A1⅐A2⅐A3-C1-C2), lacking the acidic region of the LCh, binds to phospholipid surface with approximately 10-fold higher affinity than that of unactivated fVIII. Similar kinetic parameters for A3-C1-C2, A1⅐A2⅐A3-C1-C2, and HCh⅐A3-C1-C2 interaction with PS-containing membranes indicate that only A3-C1-C2 is involved in this interaction. We conclude that removal of the LCh acidic region upon thrombin activation is solely responsible for the increased affinity of fVIIIa for binding to phospholipid membranes.
We hypothesized that this increased affinity of fVIIIa is related to conformational change(s) that occurs within the C2 domain upon removal of the acidic region of the LCh (24). To determine if this is correct, we used monoclonal antibody ESH8, which binds to the C2 domain of fVIIIa and has the ability to maintain the C2 conformation similar to that within fVIII (25). Binding of ESH8 to fVIIIa resulted in fVIIIa binding to PS/PC/PE vesicles (4/76/20) with a lower affinity similar to that of fVIII, suggesting that ESH8 may block the conformational change(s) within C2, which is responsible for the increase in fVIIIa affinity for phospholipid.
We were able to demonstrate that the increase in affinity of fVIII for phospholipid upon thrombin activation is critical for the assembly of fVIIIa⅐factor IXa⅐PS/PC complex (factor Xase) under conditions when the concentration of phospholipid binding sites is below saturation for fVIIIa⅐PS/PC binding. It would be expected that under such conditions the concentration of fVIIIa bound to phospholipid will be higher than that of ESH8⅐fVIIIa, since the affinity of ESH8⅐fVIIIa for PS/PC/PE vesicles is similar to that of fVIII. Indeed, we demonstrated that at the concentration of PS/PC vesicles below saturation for fVIIIa⅐PSPC binding (lower than 10 M), the ability of the fVIIIa⅐factor IXa⅐PS/PC complex to activate factor X is significantly less in the presence of ESH8. This observation also suggests that upon thrombin activation an increase of fVIII affinity for membranes with a physiologic content of PS, PC, and PE to a value comparable with the fVIII concentration in plasma (1 nM (43)) may be critical for assembly of the factor Xase complex in vivo when both the concentration of fVIIIa and of the binding sites on phospholipid membrane are limiting for factor Xa formation. The above observation, however, does not contradict with a previously reported lack of inhibition of the factor Xase assay by ESH8 (25) when the concentration of vesicles is saturating (20 M) for ESH8⅐fVIIIa binding.
To elucidate the physiological relevance of our findings, we examined whether thrombin activation of fVIII increases its affinity for activated platelets, which are providing a surface for assembly of the factor Xase complex in vivo (44). The affinity of fVIII for activated platelets determined in our study was 8.4 nM. This is in a good agreement with the values reported by others for fVIII binding to activated platelets (3 nM (3) and 10.4 nM (45)) and for platelet-derived microparticles (5-10 nM (2)). The affinity of fVIIIa for activated platelets predicted by our measurements of the affinity of A3-C1-C2 for activated platelets (K d 1.1 nM) is similar to the affinity of fVIIIa for activated platelets (K d 1.7 nM) determined using an electrophoretic quasielastic light scattering technique for detection of fVIIIa binding (45). The difference between the affinity of fVIIIa and fVIII for activated platelets in our study (8-fold) was similar to that previously observed (6-fold) (45).
Our data are consistent with a model in which the phospholipid binding site within C2 has two conformations: one that is capable of binding phospholipids with lower affinity in fVIII, LCh, or C2 and another with higher affinity that exists in fVIIIa or in A3-C1-C2. Our experiments, however, do not exclude the possibility that the reduction of the overall negative charge of LCh upon removal of the acidic region may be partially responsible for the increased affinity of A3-C1-C2 for phospholipids as compared with LCh. Electrostatic forces are known to be involved in fVIII binding to PS/PC vesicles (46), and removal of the acidic region 1649 -1689 containing a high number of negatively charged residues would reduce the overall charge of LCh and the electrostatic repulsion between A3-C1-C2 and negatively charged PS. Since it is likely that the acidic region of the LCh and the C2 domain are in close proximity (24,26), it would be expected that removal of the LCh acidic region would increase the energy of C2 interaction with PS-containing membranes.
In contrast to the homologous factor V, containing two different regions within the A3 and C1-C2 domains that participate in binding to PS and PC, respectively (47), the fVIII C2 domain was solely responsible for fVIII binding to PS-containing membranes. This was deduced from the similarity of the kinetic parameters of fVIII and its LCh and C2 domain for binding to phospholipid membranes. Similar affinities of fVIII and LCh for binding to PS/PC (25/75) bilayers were previously observed by Spaargaren et al. (48) using ellipsometry. However, the K d values we determined for LCh and fVIII interaction with immobilized PS/PC monolayers were 5-fold higher than those of Spaargaren. As the k on and k off values were not reported by Spaargaren, the difference between the results is difficult to interpret.  Fig. 7. The factor X activation reaction was started by the addition of 5 mM Ca 2ϩ and 0.06 nM of fVIIIa (E) or ESH8⅐fVIIIa (q). After 7 min, the reaction was stopped by the addition of 0.05 M EDTA, and factor Xa generation was measured as described under "Experimental Procedures." Preliminary experiments demonstrated that formation of factor Xa was proportional to reaction time for at least 10 min within the range of concentrations of PS/PC vesicles used. In the control experiment, amounts of factor Xa generated by the mixture of factor IXa, Ca 2ϩ , and PS/PC vesicles without added fVIIIa in the absence (ƒ) or presence (‚) of ESH8 were measured.
FIG. 9. Displacement of fVIII from platelets by its fragments. 125 I-fVIII (0.25 nM) and activated platelets (10 8 /ml) were incubated in the presence of varying concentrations of unlabeled fVIII (OE), LCh (q), A3-C1-C2 (E), C2 (‚), or HCh (ϫ), followed by the determination of 125 I-fVIII binding (see "Experimental Procedures"). 125 I-fVIII binding to activated platelets in the presence of unlabeled competitor is expressed as the percentage of 125 I-fVIII binding in the absence of competitor. 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 for a single class of binding sites using the computer program LIGAND.
As the kinetic measurements performed in our study used phospholipid monolayers or vesicles immobilized to biosensor chips, the kinetic parameters derived from these experiments may deviate from those determined in a fluid phase assay. The major sources of such deviations are possible modification of a binding partner due to immobilization (49) and potential introduction of a rate-limiting diffusion step that affects the ability to determine rate constants for fast binding reactions (39). Our determinations were independent of fVIII diffusion to membrane binding sites, since increased mixing speed or flow rate of the fluid-phase ligand did not influence the association kinetics. Furthermore, parameters for fVIII binding to immobilized small unilamellar PS/PC (25/75) vesicles and PS/PC (25/75) monolayers were identical, indicating that fVIII binding sites on the monolayer are equivalent to those on the intact vesicles. In addition, the association rate constant determined in our study using plasmon resonance for fVIII binding to PS/PC monolayer (25/75) (6.3 ϫ 10 5 M Ϫ1 s Ϫ1 ) was similar to the value of 4.2 ϫ 10 5 M Ϫ1 s Ϫ1 determined by Bardelle et al. (37) for fVIII association with small vesicles in fluid phase using a stoppedflow technique. We calculated the k off value for fVIII dissociation from small unilamellar PS/PC vesicles, using the K d for fVIII⅐PS/PC binding (50) and the above k on value. This k off value (1.68 ϫ 10 Ϫ3 s Ϫ1 ) was similar to the corresponding value determined in our study (1.22 ϫ 10 Ϫ3 s Ϫ1 ). The value of K d determined in our study for fVIII binding to immobilized PS/ PC/PE (4/76/20) vesicles with physiologically relevant molar fractions of phospholipids (7.4 nM) is in a good agreement with the K d values determined in solution for fVIII binding to the synthetic PS/PC/PE vesicles (10.2 nM (12)). This suggests that parameters determined in the present study for interaction of fVIII derivatives with synthetic phospholipid surfaces are likely to be physiologically relevant. The similarity of the parameters for fVIII⅐phospholipid binding determined by others and those determined in our study using plasmon resonance also indicate the validity of this technique for studying interactions of fVIII and its derivatives with phospholipids.
Since phospholipid binding of activated factor V (factor Va) is required for the assembly of the prothrombinase complex, which is analogous to the factor Xase complex, the affinity of factor Va for binding to phospholipid membranes may also be higher than that of factor V. Interestingly, the association rate constant for factor V binding to phospholipid vesicles determined by Bardelle et al. (37) is similar to that for fVIII, whereas the respective values for factor Va (51) or its light chain (52) are considerably higher than that determined for factor V (37). However, the K d values for factor V (K d ϭ 2.9 nM (50) and factor Va (K d ϭ 3 nM (39)) calculated as k off /k on ratios are similar, since the dissociation rate constant for factor Va (39) was also higher (51,52) than that for factor V (37). Thus, there is no evidence that factor V affinity for phospholipid increases upon activation by thrombin.