Identification of a Binding Site for Blood Coagulation Factor IXa on the Light Chain of Human Factor WII*

The interaction between human factor Ma and factor VI11 or its constituent units was investigated. Equilib- rium binding studies were performed employing factor VI11 light chain that was immobilized on a monoclonal antibody. Factor VI11 light chain was observed to bind factor IXa with high affinity (ICd = 14.8 * 3.2 nm) and approximately 1:l stoichiometry. Optimal interaction required NaCl concentrations below 0.2 M and the pres- ence of Ca2+ ions. Factor W I light chain in solution effectively inhibited binding of factor IXa to the immo- bilized light chain (Ki = 10.9 1.9 a). The isolated factor VI11 light chain and the factor VI11 heterodimer were equally effective in factor IXa binding, demonstrating that this interaction did not require the factor VI11 heavy chain. Factor Xa and activated Protein C were found to be inefficient (Ki 1 1.2 p ~ ) in competing with factor Ma, indicating that the high affinity for factor VI11 light chain was unique for factor Ma. The factor &-factor VIII light chain interaction was inhibited by von Willebrand factor, but this effect was abolished by cleavage of the factor VI11 light chain by thrombin. An antibody that inhibits von Willebrand factor-factor VIII complex formation did not compete for factor Ma bind- ing. In contrast, association of factor Ma with the factor VI11 light chain was inhibited by an antibody directed against the factor VI11 region Gln1778-Asp1”o.

The interaction between human factor Ma and factor VI11 or its constituent units was investigated. Equilibrium binding studies were performed employing factor VI11 light chain that was immobilized on a monoclonal antibody. Factor VI11 light chain was observed to bind factor IXa with high affinity (ICd = 14.8 * 3. antibody that inhibits von Willebrand factor-factor VIII complex formation did not compete for factor Ma binding. In contrast, association of factor Ma with the factor VI11 light chain was inhibited by an antibody directed against the factor VI11 region Gln1778-Asp1"o. W e propose that this sequence provides a factor Ma binding site and that its exposure requires dissociation of the factor VIII-von Willebrand factor complex. In the intrinsic pathway of blood coagulation, factor X (FX)l is activated by a complex consisting of the serine protease factor IXa (FIXa), factor VI11 (FVIII), calcium ions, and phospholipids (1). Within this complex, FVIII functions as a nonenzymatic cofactor. The fact that FVIII deficiency or dysfunction is associated with the severe bleeding disorder known a s hemophilia A underscores that this cofactor is indispensable for appropriate blood coagulation (2). FVIII is synthesized as a single chain polypeptide with the domain structure Al-A2-B-A3-C1-C2 (3,4). Due to endoproteolytic processing, FVIII circuzation for Scientific Research (NWO) . The costs of * This study was financially supported by the Netherlands Organipublication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. In plasma, M I 1 is present as an inactive precursor, which is tightly associated with its carrier, von Willebrand factor (vWF) (2). FVIII activation is required for appropriate cofactor function in the FX activating complex (9,lO). Activation is achieved by limited proteolysis in both the heavy and light chain by FXa or thrombin (8), and results in a labile heterotrimeric activation product, M I I a (11,12). FVIII activity is subject to downregulation by the serine protease activated Protein C (13,14). This serine protease, which binds to a site located at the C terminus of the A3 domain in the FVIII light chain (FVIH-LC) (13,15), inactivates FVIII by at least two cleavages in the factor VI11 heavy chain (FVIII-HC) (14).
Although less effective than activated Protein C, FKa is also capable of inactivating FVIII by limited proteolysis (16,17).
The involvement of FIXa in both stabilization and inactivation of FVIII seems to imply that intrinsic FXa formation is regulated by interactions between FKa and FVIII.
The aim of the present study was to identify the binding sites involved in assembly of the binary enzyme-cofactor complex.
For this purpose, the interaction between FVIII and FKa was assessed in equilibrium binding experiments using immobilized FVIII-LC. The same technique was used for competition studies employing other FVIII-binding proteins, including activated Protein C, FXa, vWF, and monoclonal antibodies with known epitopes. This approach allowed the identification of a FVIII-LC region associated with FIXa binding.
Antibodies-The monoclonal antibodies CLB-CAg A, CLB-CAg 12, CLB-CAg 69, and CLB-CAg 117 against FVIII-LC and CLB-CAg 9 against FVIII-HC have been described previously (22)(23)(24). The antibody CLB-FIX 2 is a Ca2+-dependent murine antibody against human FIX. Hybridomas were prepared using standard procedures. Cell supernatants were screened for binding to immobilized FIX in the presence of 20 n u CaC12, and positive clones were rescreened with EDTA replacing CaC12. One cell line, designated CLB-FIX 2, was found to produce an IgG,, antibody displaying FIX binding that was strictly Ca2+-dependent. All monoclonal antibodies used were purified from culture medium, employing Protein A-Sepharose as recommended by the manufacturer.
Polyclonal antibodies against human F M were raised in rabbits and purified employing immobilized FIX. Antibodies were conjugated with horseradish peroxidase as described (25).
Other Proteins-Human FMap and FXa were prepared as described previously (26). Human activated Protein C was obtained essentially as described (27). Recombinant human vWF was expressed in AtT-20 cells and consisted of normally processed and multimerized vWF (28). vWF was isolated from the culture medium by S-Sepharose FF chromatography in 0.1 M NaCl, 10 m~ benzamidine, 10 m~ imidazole (pH 6.5) using a linear NaCl gradient for elution. After Sephacryl S-300 chromatography, vWF containing fractions were pooled and stored a t -20 "C in 0. After blocking with HSA-containing buffer, the immobilized antibodies were incubated with various concentrations FVIII-LC for 16 hours a t 4 "C in 0.15 M NaCI, 1% (w/v) HSA, 25 m~ Tris (pH 7.2). Concentrations of total and nonbound FVIII-LC were determined, and Equation 1 (see below) was used to estimate the Kd and B,,,=. The koff was calculated from dissociation experiments in which dissociation of FVIII-LC was monitored by analyzing subsamples from the supernatant during a 4-h incubation a t 37 "C in a buffer containing 0.15 M NaCI, 5 m~ CaCI,, 0.1% (v/v) Tween-20, 1% (w/v) HSA, 25 m~ histidine (pH 6.2).
Remaining binding sites were blocked with 2% (w/v) HSAin 20 m~ Tris (pH 8.0) in a volume of 250 pl. FVIII-LC (62.5 n~) was then incubated for 16 h at 4 "C in 0.1 M NaCI, 5 m~ CaCI,, 0.1% (v/v) Tween-20,1% (w/v) HSA, 20 m~ histidine (pH 6.2) in a volume of 100 1. 11. After this incubation, total and nonbound FVIII-LC were determined, and the concentration of bound FVIII-LC was calculated. Prior to binding experiments, FIXa was inactivated by a 200-fold molar excess of the irreversible inhibitor EGR-CK for 15 min in 0.1 M NaCI, 1% (w/v) HSA, 10 m~ Tris (pH 7.4). Subsequently, EGR-FIXa was incubated with the immobilized FVIII-LC in the histidine buffer in a volume of 100 1. 11 for 4 h at 37 "C. Total and nonbound FIXa were determined, and the fraction of FIXa bound was calculated. In some experiments, the antibody CLB-CAg A was used instead of CLB-CAg 12. Nonspecific binding was determined for each separate incubation by performing the experimental procedure in the absence of FVIII-LC and found to be 3 4 % of the concentration of FMa added. All data were corrected for nonspecific binding.
Calculations-Dissociation constants (Kd) and the maximum number of bindings sites (B, , , -) were estimated from equilibrium binding assays, by fitting experimental data into a model describing the interaction of FIXa with a single class of binding sites. This model is given by the following equation (31).
The concentrations of FIXa initially added (total) and determined in the supernatant of immobilized FVIII-containing wells (nonbound) were subtracted to obtain concentrations of bound FIXa. Data were fitted employing EnzFitter software (Elsevier, Amsterdam, the Netherlands). The same nonlinear regression program was used for the analysis of experiments in which potential inhibitors of FVIII-FMa interaction (FVIII in solution, activated Protein C, etc.) were evaluated. Binding data obtained in the presence of a varying concentrations of competitor were analyzed using the following equation (32).
Herein, represents the ratio of FIXa bound in the presence and absence of competitor, [El is the concentration of FIXa added, while [Il is the concentration of inhibitor added. This leads to the direct estimation ofKj,,,,p, the apparent inhibition constant. Correcting this value for the contribution of the direct interaction between FIXa and immobilized FVIII-LC then reveals the constant K, (32) as follows.
where Kd and E,, are obtained from equation (1). The inhibition constant K, reflects the dissociation constant Kd of the interaction between inhibitor and FIXa or immobilized FVIII-LC.

Design of Equilibrium Binding Studies-The interaction be-
tween FMa and FVIII has been studied employing an experimental approach based on the immobilization of the FVIII-LC by monoclonal antibodies. Previous studies from our laboratory have identified several monoclonal antibodies that are particularly effective in immobilizing FVIII (22) and resistant to elution under moderately chaotropic conditions (23). Two of these antibodies, called CLB-CAg A and CLB-CAg 12, were further evaluated in the present study. The antibody CLB-CAg A is known to be a strong inhibitor of FVIII activity and has its epitope within the A3 domain on the FVIII-LC (24). The antibody CLB-CAg 12 differs from CLB-CAg A in that it is noninhibitory and binds to another, so far unidentified FVIII sequence (22). The FVIII binding properties of these two antibodies were assessed in more detail. The results are summarized in Table I. Both antibodies proved effective in binding the isolated FVIII-LC, with Kd values in the KIM range and low dissociation rates. These parameters allowed FVIII binding sites to be easily saturated in a virtually irreversible manner using an excess of FVIII-LC. In each individual series of experiments, the amount of antibody-bound FVIII-LC was   Table I).
Since dissociation of the FVIII-antibody complex was negligible under these conditions, the amounts of immobilized FVIII-LC were precisely known and constant over time. This provided the basis for equilibrium binding studies to assess the parameters of FIXa binding to the immobilized FVIII-LC. Binding of FlXa to Immobilized FVZZI-LC-The analysis of the interaction between FMa and FVIII may be complicated by the notion that FIXa may cleave both the heavy and light chain of FVIII under particular conditions (16,17). To exclude any interference by limited proteolysis, all binding studies were performed using FIXa that had been inactivated by the irreversible serine protease inhibitor, EGR-CK. In preliminary experiments, the time dependence of the association between inactivated FIXa and FVIII was examined using FVIII-LC that had been immobilized on the antibody CLB-CAg 12. FMa binding was observed to increase until a maximum was reached after 2 h (results not shown). In all further experiments, 4-h incubation periods were employed to ensure that equilibrium had been attained. This interaction was explored in more detail by varying the concentration of FIXa. As shown in Fig. 1, a saturable, dose-dependent binding was observed. These data were in agreement with a model describing the interaction of FMa with one single class of binding sites (see Fig. 1). The dissociation constant was calculated to be 14.8 k 3.2 n~ (average & S.D.), while the maximum number of FIXa binding sites was found to be 0.9 2 0.1 pmol/well. Similar values could be derived from the Scatchard analysis of the same data ( Fig. 1, inset 1. Using the given amount of immobilized FVIII-LC (Table   I, Bobs), the stoichiometry was estimated to be 0.8 f 0.2 mol of F W m o l of FVIII-LC.
As shown in Table I, the antibody CLB-CAg A is equally effective in immobilizing the FVIII-LC as the antibody CLB-CAg 12. However, the same binding experiments failed to reveal any FIXa binding to FVIII-LC immobilized on this antibody (Fig. 1). Apparently, interaction of the antibody CLB-CAg A with the FVIII-LC is incompatible with FIXa binding. As this may be due to the inhibitory nature of this particular antibody (see below), the noninhibitory antibody CLB-CAg 12 was used for further studies.

Competition Studies with FIXa-related Serine Proteases-
The interaction between the FVIII-LC and FIXa was further studied by competition studies using related vitamin K-dependent proteases. For this purpose, human FXa and activated Protein C were inactivated using EGR-CK and subsequently tested for inhibition of FIXa binding to the immobilized FVIII-LC (Fig. 2). With regard to FXa, no competition was observed, even at concentrations as high as 1.2 p~, representing a 40-fold molar excess over FIXa. Similar experiments using activated Protein C displayed no inhibition at concentrations up to 0.5 p~. At higher concentrations, a slight inhibition became apparent, which corresponds with a Ki in the p~ range.

Effect of Ionic Strength and Ca2+ Ions on FIXa Binding to the FVIIZ-LC-To further characterize the interaction of FIXa with
FVIII-LC, the effect of ionic strength was examined. A constant amount of FMa was incubated with immobilized FVIII-LC in the presence of various concentrations of NaCl. Between 50 and 150 mM, FIXa binding was found to be constant. However, at higher NaCl concentrations FMa binding was markedly decreased (Fig. 3). As the interaction of FVIII-LC to the immobilized antibody is unaffected by high concentrations of NaCl (23), the number of FIXa binding sites is constant under the conditions of Fig. 3. The observed effect of ionic strength, therefore, directly reflects the binding of FIXa to the FVIII-LC. The complex NaCl dependence suggests that both hydrophobic and electrostatic interactions contribute to the association between FVIII and FIXa.
Since virtually all biological functions of vitamin K-dependent proteins require Ca2+ ions, it was of further interest to examine the effect of these divalent cations on the interaction between FIXa and the FVIII-LC. As shown in Fig. 3 (see inset), some FVIII binding did occur in the absence of Ca2+, but optimal binding required Ca2+ concentrations of a t least 2 m. This suggests that Ca2+ indeed contributes to the interaction between FIXa and FVIII, although the Ca2+ dependence apparently is not absolute.

Znteraction between FZXa and FVIZI in Solution-
Our experimental approach also provided the possibility to examine the interaction between FIXa and FVIII in solution. For this purpose, known amounts of immobilized FVIII-LC (see Table I) were incubated with a mixture of FMa and FVIII or its purified subunits in varying concentrations. As shown in Fig. 4, the isolated FVIII-LC was found to effectively inhibit the binding of FIXa to the immobilized FVIII-LC in a dose-dependent manner. The experimental data were in agreement with the model of free FVIII-LC competing with its immobilized counterpart for FIXa binding (Fig. 4). The calculated inhibition constant (10.9 2 1.9 IMI) was similar to the Kd of FMa binding to the immobilized FVIII-LC (cfi Fig. 1). This finding demonstrates that free and immobilized light chain are equivalent with respect to FMa binding. In contrast, the isolated heavy chain did not affect FIXa binding to the FVIII-LC (Fig. 4). Furthermore, the heterodimeric complex comprising light and heavy chain was found to compete for FIXa binding with the same effectiveness as the isolated light chain alone (K, = 13.4 2 2.0 IMI) (see Fig. 4). Neither as the free subunit nor as part of the heterodimer does the FVIII-HC seem to support FMa binding to the FVIII-LC.

Effect of vWF on the Znteraction between FIXa and the FVZZI-LC-
The FVIII-LC is known to contain binding sites for a number of components involved in FVIII function, including phospholipids (33) and activated Protein C (15). As these interactions are regulated by complex formation between FVIII and its physiological carrier vWF (34, 35), we examined whether vWF also affects FMa binding. For this purpose, immobilized FVIII-LC was allowed to interact with varying amounts of purified recombinant vWF. After incubation, residual nonbound vWF was quantified, and the amount of vWF bound was then calculated. At the highest concentration tested (45 IMI) (see Fig. 5), about 0.6 pg of vWF was adsorbed, which is equivalent to 2 pmol of vWF monomers. These data demonstrate that a significant proportion of the immobilized FVIII-LC (-1.2 pmol/well) (see Table I) was indeed complexed to vWF. Subsequently, FMa was added, and binding to FVIII-LC was assessed. As shown in Fig. 5, the association between FMa and the FVIII-LC was effectively inhibited by the presence of vWF in a dose-dependent manner. The inhibitory effect of vWF was also evaluated using thrombin-cleaved FVIII-LC, which lacks the acidic region involved in high affinity vWF binding (36-38). As shown in Fig. 5, FMa binding to thrombin-cleaved FVIII-LC was retained but no longer subject to inhibition by vWF. This demonstrates that vWF and FIXa have distinct requirements for binding to the FVIII-LC.
Effect of Monoclonal Antibodies against the FVZZI-LC on FZXa binding--Two monoclonal antibodies against the FVIII-LC were considered to be of particular interest with regard to FIXa binding. The first antibody, called CLB-CAg 69, is known to compete with vWF for FVIII binding and is directed to the FVIII-LC sequence Lys1673-Arg1689 (37). As shown in Fig.  6 , this antibody did not interfere in the interaction between FVIII and FIXa. This confirms the observation that FIXa binding does not require the N-terminal acidic region, which is  Table I absent in the thrombin-cleaved FVIII-LC (Fig. 5). The second antibody, CLB-CAg A, was of interest, because it is a strong inhibitor of FVIII activity (24). Moreover, as is apparent from Fig. 1, FIXa did not bind to FVIII-LC that was immobilized on this antibody instead of antibody CLB-CAg 12. Indeed, antibody CLB-CAg A effectively inhibited FIXa binding to the FVIII-LC (Fig. 6). The observed dose dependence of this inhibition fitted well into the model of competitive inhibition as described under "Experimental Procedures." The calculated inhibition constant was 0.51 2 0.22 n~, which is similar to the Kd for the binding of the FVIII-LC to the same antibody (see Table  I). These data suggest that the monoclonal antibody CLB-CAg A and FIXa share common binding sites on the FVIII-LC.

DISCUSSION
Within the blood coagulation cascade, FVIII and factor V (FV) serve as nonenzymatic cofactors that assemble into phospholipid-bound complexes with the enzymes FIXa and FXa, respectively (1). As FVIII and FV exhibit considerable homology, it is generally assumed that these cofactors support coagulation by essentially the same mechanism (1,391. With respect to FV, it has been shown that the FVa heterodimer is more effective than its isolated heavy or light chain in FXa binding (40) and that both the light chain (41) and the heavy chain (42) contain FXa binding sites. Moreover, both subunits are capable of interacting with FXa (43), although the precise sites of interaction remain unidentified. In view of these observations, it is not unexpected that the isolated light chain of FVIII displays FMa binding (Fig. 1). A further similarity between FVIII and FV is that FMa binding to FVIII has the same Ca2+ dependence as reported for the interaction between FVa and FXa (Fig. 3) (cf Ref. 43). However, FVIII seems different from FV in that it displays a higher affinity (Kd = 13.4 nM) (see Fig. 4) than FVa for enzyme binding (0.8 p~) (see Ref. 43). We have considered the possibility that high affinity FMa binding could have been induced by FVIII-LC being adsorbed to a n antibody in our studies. However, competition studies have established that FVIII-LC in solution has the same affinity for FIXa binding (Fig. 4). The observation that the intact FVIII heterodimer has similar affinity as the isolated light chain (Fig. 4) suggests that FIXa binding may be predominantly due to the FVIII-LC.
The Kd for binding of FIXa to the FVIII-LC (Figs. 1 and 4) is about 5-fold higher than reported recently for the complete FVIII heterodimer (44). In this regard, it should be noted that those studies have addressed the interaction between FIXa and FVIII in the presence of phospholipids, whereas we have studied phospholipid-independent binding. As phospholipids contribute to the assembly of the FVa -FXa complex (1,43), it seems reasonable to assume that a similar effect on the interaction between FVIII and FIXa explains the lower affinity of the phospholipid-independent binding as revealed by our studies. It is further relevant t o note that we have evaluated FMa-FVIII interaction by direct binding, whereas Duffy et al. (44) have monitored complex formation by FVIII-induced conformational changes in the FMa active site. Employing similar techniques, it has been observed that the A2 domain of the FVIII-HC is required for maximal changes within the FIXa active site (21). This might seem in conflict with our observation that the FVIII-HC does not contribute to the affinity of FMa for the FVIII-LC (Fig. 4). We considered the possibility that this could be due to denaturation of the isolated FVIII-HC. However, this seems unlikely, as reconstitution experiments (not shown) demonstrated that the isolated FVIII-HC and FVIII-LC could be effectively assembled into biologically active heterodimers.' These findings do not necessarily mean that the FVIII-HC is not involved in FIXa-FVIII interaction. It is conceivable that the FVIII-LC is responsible for complex assembly with FMa, while FVIII-HC might induce repositioning of the FIXa active site.
In order to identify the FVIII-LC region involved in FMa binding, we have performed competition studies using ligands with previously established binding sites. One possibility that we addressed was that the observed FMa binding was associated with its known cleavage site within the N-terminal portion of the FVIII-LC, at Arg1719 (16, 17). To exclude that this enzyme substrate binding would complicate our studies on enzyme cofactor association, active-site blocked FIXa has been employed in our experiments. Therefore, proteolytic interactions are not likely to contribute to the observed binding of FIXa to FVIII. Moreover, FXa was not capable of competing with FIXa for FVIII binding, although it cleaves the light chain at almost the same position, Arg17'l (8). This supports our view that the high affinity interaction observed in our studies originated from enzyme cofactor assembly. Another component that we studied was activated Protein C, which binds to the Cterminal part of the FVIII A3 domain, within the sequence Protein C (19, 201, it would have been conceivable that these two enzymes have overlapping binding sites. However, activated Protein C proved inefficient in competing with FIXa for FVIII-LC binding (Fig. 2), indicating that FIXa primarily binds to another region in the FVIII-LC.
A positive identification of the FMa binding site on the FVIII-LC was achieved using the monoclonal antibody CLB-CAg A (Fig. 6). In previous studies, we have shown that this antibody is a strong inhibitor of FVIII activity and has its epitope within the light chain A3 domain (24). Several other antibodies have been described that have their epitope on the FVIII-LC and also strongly interfere in FVIII function. In a number of cases, studies using such inhibitory antibodies as functional probes have contributed to our current understanding of FVIII function. For instance, antibodies against the C2 domain inhibit FVIII function by interfering in the interaction between FVIII and phospholipids (45). Furthermore, an antibody against the N-terminal acidic region has been described to interfere in the proteolytic activation of FVIII at Arg1689 by thrombin (46). Our monoclonal antibody CLB-CAg A has its His2009-Va12018 (15). As FMa inhibits inactivation by activated epitope within the A3 domain sequence Gln1778-Asp1840 (24). The notion that binding to this region strongly interferes with FVIII activity suggests that an essential FVIII function is associated with this A3 domain region. This is fully consistent with the observed interference in the assembly of the enzymecofactor complex (Fig. 6). We therefore propose that the FVIII-LC sequence Gln1778-Asp1840 provides a FIXa binding site.
One intriguing finding is that the interaction between FIXa and the FVIII-LC is readily inhibited by vWF (Fig. 5). This phenomenon is probably due to steric hindrance rather than to competition with FIXa for the same site for two reasons. First, thrombin cleavage of the fragment Gln1649-Arg1689 from the FVIII-LC abolishes the inhibitory effect of vWF but not FIXa binding (Fig. 5). Second, FIXa binding is unaffected by a monoclonal antibody against the sequence Lys1673-Arg1689, whereas the same antibody does interfere in vWF binding (37). Irrespective of the precise location of the sites involved, it is of relevance to note that FIXa binding is not restricted to FVIIIa, since the nonactivated heterodimer also displays high affinity FMa binding (Fig. 4) (cc Ref. 44). As a consequence, FIXa binding would be among the very few processes within the coagulation mechanism that might escape proteolytic control. However, the aflinity of FVIII for vWF is about 0.1 nM (371, which exceeds the affinity for FIXa by two orders of magnitude. Therefore, the FIXa binding site is not likely to be accessible when vWJ? is simultaneously present. The exposure of the FIXa binding site then requires dissociation of the FVIII-vWJ? complex. As this process is a consequence of proteolytic cleavage of the FVIII-LC at position (36) (see also Fig. 5), the same proteolytic event may result in the exposure of the FIXa binding site.