Identification of Plasmin-interactive Sites in the Light Chain of Factor VIII Responsible for Proteolytic Cleavage at Lys36*

We have recently reported that plasmin likely associates with the factor VIII light chain to proteolyze at Lys36 within the A1 domain. In this study, we determined that the rate of plasmin-catalyzed inactivation on the forms of factor VIIIa containing A1-(1–336) and 1722A3C1C2, reflecting Lys36 cleavage, was reduced by ∼60%, compared with those containing 1649A3C1C2 and 1690A3C1C2. SDS-PAGE analysis revealed that Lys36 cleavage of factor VIIIa with 1722A3C1C2 was markedly slower than those with 1649A3C1C2 and 1690A3C1C2. Surface plasmon resonance-based assays, using active site-modified anhydro-plasmin (Ah-plasmin) showed that 1722A3C1C2 bound to Ah-plasmin with an ∼3-fold lower affinity than 1649A3C1C2 or 1690A3C1C2 (Kd, 176, 68.2, and 60.3 nm, respectively). Recombinant A3 bound to Ah-plasmin (Kd, 44.2 nm), whereas C2 failed to bind, confirming the presence of a plasmin-binding site within N terminus of A3. Furthermore, the Glu-Gly-Arg active site-modified factor IXa also blocked 1722A3C1C2 binding to Ah-plasmin by ∼95%, supporting the presence of another plasmin-binding site overlapping the factor IXa-binding site in A3. In keeping with a major contribution of the lysine-binding sites in plasmin for interaction with the factor VIII light chain, analysis of the A3 sequence revealed two regions involving clustered lysine residues in 1690–1705 and 1804–1818. Two peptides based on these regions blocked 1649A3C1C2 binding to Ah-plasmin by ∼60% and plasmin-catalyzed Lys36 cleavage of factor VIIIa with A1-(1–336) by ∼80%. Our findings indicate that an extended surface, centered on residues 1690–1705 and 1804–1818 within the A3 domain, contributes to a unique plasmin-interactive site that promotes plasmin docking during cofactor inactivation by cleavage at Lys36.

Factor VIII circulates as a complex with von Willebrand factor and functions as an essential cofactor in the tenase complex responsible for anionic phospholipid surface-dependent conversion of factor X to Xa by factor IXa (1). Molecular defects in factor VIII result in the congenital bleeding disorder, hemo-philia A. Factor VIII is composed of 2,332 amino acid residues with a molecular mass of ϳ300 kDa and contains three types of structural domain, arranged in the order of A1-A2-B-A3-C1-C2 (2,3). Mature factor VIII is processed to a series of metal ion-dependent heterodimers by cleavage at the B-A3 junction, generating a heavy chain consisting of the A1 and A2 domains, together with heterogeneous fragments of a partially proteolyzed B domain, linked to a light chain consisting of the A3, C1, and C2 domains (2)(3)(4).
Factor VIII is converted into an active form, factor VIIIa, by limited proteolysis catalyzed by either thrombin or factor Xa (5). Cleavages at Arg 372 and Arg 740 in the heavy chain produce 50-kDa A1 and 40-kDa A2 subunits. Cleavage of the 80-kDa light chain ( 1649 A3C1C2) at Arg 1689 produces a 70-kDa A3C1C2 subunit ( 1690 A3C1C2). Additional cleavage by factor Xa at Arg 1721 produces a 67-kDa A3C1C2 subunit ( 1722 A3C1C2). Proteolysis at Arg 372 and Arg 1689 is essential for generating factor VIIIa cofactor activity (6). Cleavage at the former site exposes a functional factor IXa-interactive site within the A2 domain that is cryptic in the unactivated molecule (7). Cleavage at the latter site liberates the cofactor from its carrier protein, von Willebrand factor (8), and contributes to the overall specific activity of the cofactor (9, 10). APC 2 (5), factor Xa (5), and factor IXa (11) are serine proteases that inactivate factor VIII(a) by cleavage at Arg 336 within the A1 subunit. This inactivation appears to be associated with an altered interaction between the A2 subunit and truncated A1 and is coupled with an increase in the K m value for the substrate, factor X (12,13), reflecting loss of a factor X-interactive site within residues 337-372 (14). In addition, a second specific cleavage site for factor Xa, Lys 36 , was identified within the A1 subunit (13). Attack at this site also results in factor VIII inactivation mediated by an altered conformation of the A1 subunit limiting productive interaction with the A2 subunit (13).
Plasmin is a potent fibrinolytic protease and is composed of a heavy chain consisting of five kringle domains and a light chain containing the catalytic domain. The protease associates with numerous proteins via the LBS on the exposed surface (15). Several reports have shown that plasmin proteolytically inacti-vates several coagulation proteins, including factors Va (16,17), VIII (18), IXa (19), and X (20). In detail, we have demonstrated that plasmin rapidly inactivates factor VIII by cleavage at Arg 336 (18). Direct binding assays using Ah-plasmin, a catalytically inactive derivative of plasmin, revealed that plasmin interacts with the factor VIII heavy chain, predominantly the A2 domain, with high affinity (K d , ϳ6 and ϳ20 nM, respectively), in a mechanism largely independent of LBS (21). Our findings demonstrated, in particular, that Arg 484 in the A2 domain significantly contributes to a unique plasmin-interactive site within the heavy chain that promotes plasmin docking during cleavage of the heavy chain (21).
In contrast, plasmin interacts with the factor VIII light chain with moderate affinity (K d , ϳ70 nM), predominantly through LBS-dependent mechanisms (21). Our previous data suggested that plasmin cleavage at Lys 36 within the A1 domain appears to be selectively regulated by the light chain (18). In this study, we have expanded our studies using truncated light chain, recombinant factor VIII subunits, and synthetic peptides, and identified plasmin-interactive sites within the light chain responsible for cleavage at Lys 36 . Our results indicate that an extended surface centered on lysine residues involving the 1690 -1705 and 1804 -1818 regions in the A3 domain contributes to a unique plasmin-interactive site that promotes plasmin docking during cofactor inactivation by cleavage of the heavy chain at Lys 36 .

MATERIALS AND METHODS
Reagents-Purified recombinant factor VIII preparations and the monoclonal antibody (mAb58.12) recognizing the N terminus of the A1 domain (22) were generous gifts from the Bayer Corp. (Osaka, Japan). The mAb NMC-VIII/5 recognizing the C2 domain was purified as described previously (23). Purified human plasmin (Lys-plasmin) devoid of factor Xa or APC was purchased from Sigma. Ah-plasmin, a catalytically inactive derivative of plasmin in which the active site serine is replaced by dehydroalanine, was prepared as described previously (21). The modified product demonstrated Ͻ1% plasmin activity, and its molecular weight was similar to that of native plasmin. Factor IXa, Glu-Gly-Arg active site-modified factor IXa (EGR-factor IXa), factor Xa, and thrombin were obtained from Hematologic Technologies Inc. (Essex Junction, VT). Pefabloc and horseradish peroxidase-labeled streptavidin were purchased from Roche Applied Science and Chemicon (Melbourne, Australia), respectively. Phospholipid vesicles containing 10% phosphatidylserine, 60% phosphatidylcholine, and 30% phosphatidylethanolamine (Sigma) were prepared using N-octyl glucoside (24). The mAb IgG preparations were biotinylated using N-hydroxysuccinimido-biotin reagent (Pierce). Synthetic peptides corresponding to factor VIII A3 residues 1690 -1705 and 1804 -1818 were prepared by BioSynthesis, Inc. (Lewisville, TX).
Expression and Purification of Recombinant A3 (rA3) and C2 (rC2) Domains-The rA3 domain of factor VIII (residues 1690 -2019) was expressed in Escherichia coli using the pET expression system (Novagen, Madison, WI). The pMT2/factor VIII plasmid containing the B domain-deleted factor VIII gene (29) was obtained from Dr. Pipe (University of Michigan, Ann Arbor, MI). DNA fragments encoding the A3 domain were generated by PCR using pMT2/factor VIII as a template and a pair of corresponding primers. The amplified fragments were ligated with pET-20b(ϩ) expression vectors. Plasmid DNA was purified, and the sequence was confirmed by direct sequencing in both directions using Applied Biosystems technology (Foster City, CA). The plasmid was used for transformation of Origami(DE3)pLysS E. coli cells (Novagen), the host strain for the protein expression. The protein was expressed and subsequently purified using a His-Select affinity cartridge (Sigma). Proper folding of the rA3 fragment was confirmed by determination of the affinities for conformationally sensitive anti-A3 mAb CLB-CAgA (30). The cDNA coding the C2 domain sequence of human factor VIII was constructed, transformed into Pichia pastoris cells, and expressed in a yeast secretion system as described previously (31). The rC2 protein was purified by ammonium sulfate fractionation and CM-Sepharose chromatography (Amersham Biosciences) as described previously (31).
Factor Xa Generation Assay-The rate of conversion of factor X to factor Xa was monitored in a purified system (13). Plasmin-catalyzed inactivation of factor VIIIa was performed in HBS buffer (20 mM HEPES, pH 7.2, 0.1 M NaCl, 5 mM CaCl 2 , 0.01% Tween 20) containing 0.1% bovine serum albumin and phospholipid vesicles (10 M). Samples were removed from the mixtures at the indicated times, and plasmin reaction was immediately quenched by the addition of 0.2 mM Pefabloc and dilution. All reactions were performed at 37°C. Factor Xa generation was initiated by the addition of factor IXa (20 nM) and factor X (400 nM) in the presence of phospholipid (10 M). The reaction was quenched by addition of EDTA (100 mM). Rates of factor Xa generation were determined at 405 nm using a microtiter plate reader after the addition of chromogenic substrate, S-2222 (0.46 mM final concentration). A control experiment showed that the presence of plasmin and Pefabloc in the diluted samples did not affect this assay (data not shown). Factor VIIIa activity was determined as the amount (in nanomoles) of factor Xa generated per min and converted into the amount (in nanomoles) of factor VIIIa.
Experiments assessing the stability of factor VIIIa were performed in the absence of plasmin to determine the rates of factor VIIIa activity loss resulting from the A2 dissociation. At the concentrations employed, ϳ5% loss of the initial activity was observed over a 20-min time course. Thus, for each time point in this experiment, including plasmin, the observed residual activity was corrected for the contribution of activity loss from the plasmin-independent mechanism.
Cleavage of Factor VIIIa Forms by Plasmin-Human plasmin (4 nM) was added to the reconstituted factor VIIIa forms (100 nM) at 37°C in HBS buffer containing phospholipid vesicles (10 M). Samples were obtained at the indicated times, and the reactions were immediately terminated and prepared for PAGE by adding SDS and 2-mercaptoethanol and boiling for 3 min.
Electrophoresis and Western Blotting-SDS-PAGE was performed using 8% gels by the procedure of Laemmli (32). Electrophoresis was carried out at 150 V for 1 h. For Western blotting, the proteins were transferred to a polyvinylidene difluoride membrane at 50 V for 2 h in buffer containing 10 mM CAPS, pH 11, and 10% (v/v) methanol. Proteins were probed using anti-A1 mAb58.12, followed by goat anti-mouse peroxidase-linked secondary antibody (MP Biomedicals, Aurora, OH). The signals were detected using the enhanced chemiluminescence system (PerkinElmer Life Sciences).
Kinetics Measurements Using Real Time Biomolecular Interaction Analysis-The kinetics of factor VIII light chain and plasmin interaction were determined by SPR-based assays using a BIAcore X instrument (BIAcore AB, Uppsala, Sweden) as reported previously (21). Ah-plasmin was covalently coupled (ϳ7 ng/mm 2 ) to a CM5 sensor chip surface. Association of the ligand was monitored at a flow rate of 20 l/min for 4 min. The dissociation of bound ligand was recorded over a 4-min period by replacing the ligand-containing buffer with buffer alone. The level of nonspecific binding corresponding to ligand binding to the uncoated chip was subtracted from the signal. The reactions were run at 37°C. The rate constants for association (k a ) and dissociation (k d ) were determined by nonlinear regression analysis using the evaluation software provided by BIAcore AB. Equilibrium dissociation constants (K d ) were calculated as ELISA Binding Assays Using Immobilized Ah-plasmin-These assays were performed as reported previously (21). Briefly, Ah-plasmin (200 nM) was coated onto microtiter wells overnight at 4°C. The wells were blocked with 5% bovine serum albumin for 2 h at 37°C, and various concentrations of the A3C1C2 subunits were added and incubated for 2 h at 37°C. Biotinylated anti-C2 NMC-VIII/5 mAb IgG (1 g) was added to each well, and bound IgG was detected by addition of horseradish peroxidase-labeled streptavidin. The absorbance was measured at 492 nm with a Labsystems Multiskan Multisoft microplate reader (Labsystems, Helsinki, Finland). The amount of nonspecific binding of biotinylated IgG, observed in the absence of A3C1C2, was Ͻ5% of the total signal, and the amount of specific binding was obtained by subtracting the amount of nonspecific binding of biotinylated IgG.
Data Analysis-All experiments were performed at least three separate times. The parameters and their standard errors are shown. Nonlinear least squares or linear regression analysis was performed by KaleidaGraph (Synergy Reading, PA). The rates of the slope of first several points (within 5 min) in the time course of the plasmin-catalyzed inactivation of factor VIIIa form were fitted to a straight line from linear regression, and the obtained values were expressed as the inactivation rate. All correlation values (r) were Ͼ0.99.
Analyses of the interactions between the different forms of A3C1C2 and Ah-plasmin in ELISA were performed by a singlesite binding model using Equation 1, where [S] is the concentration of A3C1C2 form in the solidphase binding assay; K d is the dissociation constant; and A max represents maximum absorbance signal when the site is saturated by the A3C1C2 form. Data from studies assessing the EGR-factor IXa or A3 synthetic peptide-dependent inhibition of plasmin interaction with A3C1C2 form were fitted by nonlinear least squares regression by using Equation 2, where L represents the concentration of EGR-factor IXa or A3 peptide; B max represents maximum binding; K d is the dissociation constant for the interaction between the A3C1C2 form and Ah-plasmin; K i is the apparent inhibition constant for L; and C is a constant for binding of the A3C1C2 form and Ah-plasmin that was unaffected by L.

Plasmin-catalyzed Inactivation of Factor VIIIa
Reconstituted with the A1 1-336 and Various A3C1C2 Forms-We have recently reported that plasmin appears to associate with the light chain of factor VIII to regulate the proteolytic cleavage at Lys 36 within the A1 subunit by its protease (18). To investigate whether the light chain contributes to plasmin-catalyzed factor VIIIa inactivation because of Lys 36 cleavage, we first examined the effect on plasmin-catalyzed inactivation of factor VIIIa forms reconstituted with various A3C1C2 forms. Intact A1-(1-372) subunit is proteolyzed at Arg 336 and Lys 36 by plasmin, and the former cleavage significantly contributes to the factor VIIIa inactivation (18). Therefore, the A1-(1-336) subunit was utilized instead of the A1-(1-372) to observe the effect of Lys 36 cleavage alone. Factor VIIIa forms were reconstituted in twostep procedures. The A1-(1-336)/A3C1C2 dimer forms were prepared by reacting the equimolar concentrations (500 nM) of the A1-(1-336) and isolated A3C1C2 subunits ( 1649 A3C1C2, 1690 A3C1C2, and 1722 A3C1C2) in the presence of Ca 2ϩ . The resultant dimers were diluted 5-fold and reacted with the A2 subunit (300 nM) to generate factor VIIIa heterotrimers. The A2 subunit was used at higher concentration to minimize its inactivation because of A2 dissociation from A1/A3C1C2 dimer. Inactivation of various factor VIIIa forms were then monitored over time in the presence of plasmin (1 nM) and phospholipid vesicles (10 M) using a factor Xa generation assay as described under "Materials and Methods." The presence of plasmin and Pefabloc in the diluted samples did not affect this assay (data not shown). These data are illustrated in Fig. 1 and Table 1.
In the absence of plasmin, the maximal generated factor Xa obtained at saturable levels of reconstituted factor VIIIa with A1-(1-336) and the loss (decay) of its activity were similar (V max , ϳ55 nM/min and ϳ5% activity loss at 20 min, respectively), independent of various A3C1C2 forms (data not shown). The activity of factor VIIIa forms with 1649 A3C1C2 subunit was rapidly reduced in a time-dependent manner by the addition of plasmin, similar to that of factor VIIIa with 1690 A3C1C2, with similar inactivation rates. The reduced activity reached ϳ55% of initial activity at 20 min. Of interest, slower inactivation of factor VIIIa with 1722 A3C1C2 by plasmin was observed, compared with 1649 A3C1C2 and 1690 A3C1C2. The activity was reduced by ϳ30% at 20 min, and the inactivation rate was ϳ45% that observed with 1649 A3C1C2 and 1690 A3C1C2. These results suggested that the N terminus of the light chain might be associated with plasmin-catalyzed inactivation of factor VIIIa with A1-(1-336) mediated by cleavage at Lys 36 .
To further examine that inactivation of factor VIIIa with A1-(1-336) subunit by plasmin attributed to Lys 36 cleavage, we repeated the same experiments using factor VIIIa forms reconstituted with the A1-(37-336) subunit, deleting the A1-(1-36). Factor VIIIa forms with the A1-(37-336), and various A3C1C2 subunits were reconstituted by the same approach. Independent of A3C1C2 forms, the maximally generated factor Xa and the activity loss (decay) of factor VIIIa forms were similar (V max , ϳ25 nM/min and ϳ5% activity loss, respectively) in the absence of plasmin (data not shown). Three factor VIIIa forms with A1-(37-336) were, however, little inactivated (by ϳ5%) by plasmin even at over a 20-min reaction ( Fig. 1, inset). Inactivation rates were similar, supporting the view that plasmin-catalyzed inactivation of factor VIIIa with A1-(1-336) subunit was regulated by Lys 36 cleavage.
Addition of Various A3C1C2 Forms on Plasmin-catalyzed Inactivation of Factor VIIIa with A1 1-336 -To confirm that the N terminus of the A3C1C2 domain is responsible for plasmincatalyzed inactivation of factor VIIIa, we examined the inhibitory effects by the addition of various A3C1C2 forms on these reactions. The A1-(1-336)/ 1649 A3C1C2 dimer was reconstituted with excess amounts of A2 subunit, and factor VIIIa inactivation was then monitored by the addition of plasmin (1 nM) in the presence of various A3C1C2 in a factor Xa generation assay as described under "Materials and Methods." The presence of competitors, A3C1C2 forms, did not affect this assay (data not shown). These data are illustrated in Fig. 2 and Table  2. The addition of the 1649 A3C1C2 and 1690 A3C1C2 subunits, rA3 domain (residues 1690 -2019) (75 nM), each similarly inhibited plasmin-catalyzed inactivation of factor VIIIa with A1-(1-336), with an ϳ55% decrease in inactivation rate. Furthermore, the addition of a higher concentration (250 nM) of these A3C1C2 forms showed ϳ90% decreases in inactivation rates. In contrast, in the presence of 1722 A3C1C2, the inactivation rates were reduced by ϳ30 and ϳ55% at both concentrations, respectively, and the inhibitory effects were less than those of the other three A3C1C2 forms. The presence of rC2 domain (residues 2174 -2332) had little effect. Taken together, these findings indicated that the 1690 -1721 region in the A3 domain contributed to plasmin-catalyzed inactivation of factor VIIIa through Lys 36 cleavage.
Effects of the A3C1C2 Forms on Plasmin-catalyzed Cleavage at Lys 36 within the A1 Domain-To evaluate visually the effect of the various A3C1C2 forms on cleavage by plasmin at Lys 36 in  the A1 subunit, the A1-(1-372)/A3C1C2 dimer forms ( 1649 A3C1C2, 1690 A3C1C2, and 1722 A3C1C2) were reconstituted with the A2 subunit by the same approach. Factor VIIIa forms were then incubated with plasmin (4 nM), followed by Western blotting analyses using anti-A1 mAb58.12 for detection (Fig. 3A). This antibody recognizes the N-terminal region of A1, and the failure to detect the A1-(1-336) fragment indicates complete cleavage at Lys 36 and conversion to A1-(37-336) (18). In addition, the ratio of the A1-(1-336) product (at each time point) to the A1-(1-372) substrate (at time zero) was evaluated by scanning densitometry of the bands (Fig. 3B). In control experiments using isolated A1-(1-372) alone, the A1-(1-336) fragment was generated in a time-dependent manner within 20 min, indicating relatively slow cleavage by plasmin at Lys 36 (Fig. 3A, panel a). In contrast, with A1/ 1649 A3C1C2 (Fig. 3A, panel b) and A1/ 1690 A3C1C2 (panel c), the A1-(1-336) fragments were very weakly visualized, sug-gesting almost complete cleavage by plasmin at Lys 36 . Interestingly, with the A1/ 1722 A3C1C2 (Fig. 3A, panel d), the A1-(1-336) fragment appeared to be derived in a time-dependent manner within 10 min and then subsequently diminished. These findings suggested that cleavage at Lys 36 in the A1/ 1722 A3C1C2 was much slower than that with A1/ 1649 A3C1C2 or A1/ 1690 A3C1C2 but was faster than that with A1 alone. The data indicated that residues 1690 -1721 in the A3 domain contain a site contributing to plasmin-catalyzed cleavage at Lys 36 in the A1 domain.
Binding of the A3C1C2 Forms to Ah-plasmin-A series of experiments were designed to assess plasmin binding to the A3C1C2 subunit to obtain direct evidence that the A3 domain contains a plasmin-interactive site that contributes to enzyme docking and facilitates catalysis of cleavage at Lys 36 in the A1 domain. In these experiments, Ah-plasmin, an active site-modified plasmin lacking enzymatic activity, was utilized, and this interaction was evaluated in a real time SPR-based assay. We have recently examined the interaction of Ah-plasmin and the factor VIII heavy chain using this technique (21). Varying amounts of A3C1C2 forms were applied to Ah-plasmin immobilized on a sensor chip. Fig. 4A shows a representative signal corresponding to association and dissociation of immobilized Ah-plasmin with 1649 A3C1C2 (panel a), 1722 A3C1C2 (panel b), and rA3 subunits (panel c). Binding parameters are summarized in Table 3. The data could be comparatively well fitted by nonlinear regression using a 1:1 Langmuir binding model. The rA3 domain and 1649 A3C1C2 and 1690 A3C1C2 subunits bound to Ah-plasmin. The resulting kinetic constants derived from the association and/or dissociation kinetic curves showed that the rA3 domain bound to Ah-plasmin with similar affinity to those of the 1649 A3C1C2 and 1690 A3C1C2 subunits (K d , 44.2, 68.2, and 60.3 nM, respectively). The rC2 domain failed to bind. However, the 1722 A3C1C2 subunit bound with an ϳ3-fold lower affinity (K d , 176 nM) than that obtained with 1690 A3C1C2. These data demonstrated that the A3 domain, especially residues in the 1690 -1721 region, plays a significant role in the interaction with plasmin.
Our experiments utilizing the factor VIIIa with various A3C1C2 forms as substrate suggested that the N terminus of the A3 domain contains the predominant region contributing to plasmin-catalyzed cleavage at Lys 36 . To confirm that the inhibition of plasmin cleavage at Lys 36 was mediated by the association between the dimer and plasmin, binding experiments with A1/A3C1C2 dimers were further examined. Both of the A1 dimers with 1649 A3C1C2 and 1690 A3C1C2 bound to Ahplasmin with similar affinities (K d , 26.9 and 32.9 nM, respectively), and these affinities were ϳ2.5-fold higher than those with either form of A3C1C2 alone. This somewhat higher affinity may have been derived from a synergistic effect of the two binding domains and/or a conformational change resulting from interaction with the two domains. However, the A1/ 1722 A3C1C2 dimer bound with an ϳ4-fold weaker affinity (K d , 124 nM) than with the other dimers, and the isolated A1 alone bound very poorly (K d , ϳ200 nM). The findings therefore suggest that residues 1690 -1721 are indeed involved in a plasmin-binding site for cleavage at Lys 36 .

Plasmin-interactive Sites in the Factor VIII Light Chain
We further evaluated the interaction between the A3C1C2 subunit and plasmin using a solid-phase binding assay in which Ah-plasmin was immobilized onto microtiter wells. For these experiments, varying amounts of the A3C1C2 subunits were reacted with 200 nM immobilized Ah-plasmin. Bound factor VIII was detected using biotinylated anti-C2 NMC-VIII/5 mAb. Control experiments confirmed that this mAb did not affect the reaction between plasmin and the light chain (data not shown). Results are presented in Fig. 4B. Reactions between the A3C1C2 forms and Ah-plasmin yielded saturable binding curves, well fitted using a single-binding site model. This method is not based on a true equilibrium binding assay, however, and the K d values obtained represent an apparent K d value for the interactions. The results obtained for the 1649 A3C1C2 and the 1690 A3C1C2 subunits binding to Ah-plasmin were 97 Ϯ 10 and 89 Ϯ 9 nM, respectively, similar to those obtained in the SPR-based assays. However, the binding affinity (265 Ϯ 21 nM) for the 1722 A3C1C2 subunit was ϳ3-fold lower than that for the two other forms. Again, the rC2 domain failed to bind. Overall, the affinities determined using the ELISA-based assays were in good agreement with those obtained in the SPR-based analyses, and the findings were mutually supportive.
Effect of Factor IXa on A3C1C2 Form Binding to Ah-plasmin-Our solid-phase binding assays demonstrated that the 1722 A3C1C2 subunit bound to Ah-plasmin, albeit with relatively weak affinity. These data led us to speculate on the presence of another plasmin-interactive site(s) in the A3 domain. We have recently shown that factor IXa inhibited plasmin-catalyzed inactivation of factor VIIIa, and we identified overlapping binding sites for plasmin and factor IXa in the A2 domain of factor VIII(a) (21). It is known that residues 1804 -1818 in the A3 domain of factor VIII interact with factor IXa on phospholipid surfaces (30), and we therefore investigated the inhibitory effect of factor IXa on the light chain binding to Ah-plasmin in our ELISA method. The 1649 A3C1C2 subunit (120 nM) was mixed with varying amounts of active site-modified EGR-factor IXa for 1 h prior to incubation with Ah-plasmin (200 nM) immobilized onto microtiter wells. Bound 1649 A3C1C2 was detected using biotinylated anti-C2 mAb. EGR-factor IXa blocked 1649 A3C1C2 subunit binding to Ah-plasmin by ϳ40% at the maximum concentrations employed (500 nM), and this effect was dose-dependent (Fig. 5). The apparent K i value for factor IXa obtained from curve fitting was 160 Ϯ 51 nM. The association between the factor VIII light chain and factor IXa is surface-dependent, however (25), and hence the effects of factor IXa in our current binding studies were also examined in the presence of phospholipid. EGR-factor IXa blocked binding in a dose-dependent manner, and the inhibitory effect (ϳ60%) was greater than that in the absence of phospholipid. Furthermore, the K i value (10.5 Ϯ 1.5 nM) obtained from curve fitting was ϳ15-fold lower in the presence of phospholipid than that in its absence. These values were similar to the K d values for light chain and factor IXa association in the presence and absence of phospholipid determined in a fluid-phase model (25).
It was evident, however, that inhibition of the interaction between 1649 A3C1C2 and plasmin mediated by EGR-factor IXa was only partial (ϳ60%), and to exclude the possibility that the 1690 -1721 region in A3 contributed to the binding reaction in these experiments, we repeated the assays using the 1722 A3C1C2 subunit (240 nM), instead of the nontruncated A3C1C2. EGR-factor IXa blocked the binding of 1722 A3C1C2 subunit to Ah-plasmin by ϳ95 and ϳ80% in the presence and absence of phospholipid, respectively, at the maximum concentrations employed (Fig. 5). The calculated K i values for factor IXa obtained from curve fitting were 9.8 Ϯ 1.1 and 179 Ϯ 23 nM, respectively. This significant competitive reaction between factor IXa and 1722 A3C1C2 for binding to plasmin therefore suggested that an alternative plasmin-interactive site in the factor VIII light chain, within residues 1804 -1818 in the A3 domain, might overlap or juxtapose the factor IXa-binding site.
Binding of the 1690 -1705 and 1804 -1818 Peptides to Ah-plasmin-Further experiments focused on two distinct regions, residues 1690 -1721 and 1804 -1818, in the A3 domain, responsible for plasmin docking. We have recently demonstrated, in binding-inhibition assays using the plasmin-specific competitor 6-aminohexanoic acid, which directly binds to LBS (21), that the association between plasmin and the factor VIII light chain is mediated by an LBS-dependent mechanism. The known amino acid sequence of the A3 domain indicates that the clustered lysine residues are located in residues 1690 -1705 (Lys 1693 and Lys 1694 ) and 1804 -1818 (Lys 1804 , Lys 1808 , Lys 1813 , and Lys 1818 ), and these two regions are highly conserved in other species (Fig. 6). Therefore, to confirm that these lysine residues confer interactive sites for plasmin, two synthetic peptides derived from sequences 1690 -1705 and 1804 -1818 were prepared and examined with Ah-plasmin in competitive inhibitory ELISA.
The 1649 A3C1C2 subunit (120 nM) was incubated with immobilized Ah-plasmin (200 nM) in the presence of increasing concentrations of A3 peptides as described under "Materials and Methods." The results are shown in Fig. 7A. Both the 1690 -

TABLE 3 Binding parameters of factor VIII A3C1C2 forms and Ah-plasmin in an SPR-based assay
Reactions were performed as described under "Materials and Methods." Parameter values were calculated by nonlinear regression analysis in Fig. 4A using the evaluation software provided by BIAcore AB.  (Fig. 7B). As expected, the 1690 -1705 peptide had little inhibitory effect on this binding. To further investigate the importance of the lysine residues and/or structural alignments in the two sequences (1690 -1705 and 1804 -1818), we prepared peptides with scrambled sequences of the same composition and synthesized peptides where the lysine residues were replaced by alanine ( 1690 -1705 Ala and 1804 -1818 Ala, respectively). An equimolar mixture of the scrambled peptide (200 M) did not inhibit the 1649 A3C1C2 binding to Ah-plasmin (data not shown). Furthermore, 1690 -1705 Ala and 1804 -1818 Ala did not significantly inhibit the binding of 1649 A3C1C2 or 1722 A3C1C2 subunits to Ah-plasmin, respectively (Fig. 7B, inset). These results indicated that the lysine residues in the A3 domain, within sequences 1804 -1818 and 1690 -1705, contribute to the plasmin-interactive sites in the light chain.

Ligands
Effects of A3 Peptides on Plasmin-catalyzed Cleavage at Lys 36 in A1-To further confirm the functional role of residues 1690 -1705 and 1804 -1818 in plasmin binding, we examined the effects of the A3 peptides on Lys 36 cleavage by plasmin. The A1-(1-336)/ 1690 A3C1C2 dimer (100 nM) was reconstituted with the A2 subunit (300 nM), followed by the addition of plasmin (4 nM) and phospholipid vesicles (10 M) in the presence of A3 peptides (150 M). Lys 36 cleavage in A1-(1-336), representing the disappearance of A1-(1-336) fragment, was analyzed by Western blotting using an anti-A1 (58.12) IgG in a timed course reaction (Fig. 8A). Change of band density of A1-(1-336) was evaluated by scanning densitometry (Fig. 8B). Compared with the absence of A3 peptide (Fig. 8A, panel a), both individual 1690 -1705 and 1804 -1818 peptide slightly delayed the disappearance of A1-(1-336) (panels b and c, respectively). Furthermore, in the presence of equimolar amounts of mixture of both peptides, the disappearance of A1-(1-336) was markedly slow, and the band could be observed by ϳ50% even at 15 min after adding of plasmin, supportive of significant inhibition of plasmin-induced Lys 36 cleavage.  Furthermore, the inhibitory effect of the mixture of 1690 -1705 and 1804 -1818 peptides showed the dose-dependent manner, and the inhibition effect was ϳ80% at maximum concentration employed (300 M) (Fig. 8C). A control experiment using mixture of 1690 -1705 Ala and 1804 -1811 Ala peptides showed no significant inhibition (by ϳ10%). These findings were in keeping with the concept that both the 1690 -1705 and 1804 -1818 regions in A3 were essential for plasmin docking during factor VIIIa inactivation induced by Lys 36 cleavage, although each region interacted separately with the protease.

DISCUSSION
Plasmin inactivates factor VIIIa by proteolysis at specific sites within the heavy and light chains of the activated molecule. We have recently demonstrated that Arg 484 in the A2 domain of factor VIII significantly contributes to plasmin docking for proteolytic cleavage at Arg 336 in the A1 subunit during enzyme-catalyzed factor VIIIa inactivation (21). Our present study further revealed that proteolytic cleavage at Lys 36 in the A1 domain is supported by plasmin-interactive sites located in the A3 domain of the light chain. This conclusion is based on several novel findings using well established models. (i) The rate for plasmin-catalyzed inactivation of factor VIIIa reconstituted with A1-(1-336) and 1721 A3C1C2, reflecting cleavage at Lys 36 , was reduced by ϳ55% compared with those with 1649 A3C1C2 and 1690 A3C1C2. Furthermore, plasmin-catalyzed inactivation of factor VIIIa with A1-(1-336) was significantly inhibited by the addition of exogenous 1690 A3C1C2 and rA3, but to a much lesser extent by 1721 A3C1C2. (ii) Plasmin cleavage at Lys 36 in factor VIIIa with A1/ 1722 A3C1C2 dimer was significantly slower than that with A1/ 1690 A3C1C2. (iii) The 1722 A3C1C2 subunit bound to Ahplasmin with an ϳ3-fold weaker affinity than the 1649 A3C1C2 or 1690 A3C1C2. The rA3 domain bound to Ah-plasmin with similar affinity as 1649 A3C1C2, whereas the rC2 domain failed to bind, although a contributory role for the C1 domain in this plasmin binding remains to be completely excluded. (iv) Factor IXa (that binds to the 1804 -1818 region) or the A3 peptides (residues 1690 -1705 and 1804 -1818) competed for light chain binding to Ah-plasmin. (v) The presence of both A3 peptides inhibited plasmin-catalyzed Lys 36 cleavage of factor VIIIa with A1-(1-336) by ϳ80%. These identified amino acid residues 1690 -1705 and 1804 -1818 within the A3 domain as essential to plasmin docking for proteolytic cleavage at Lys 36 .
We observed the partial inactivation of reconstituted factor VIIIa activity with A1-(1-336) or A1-(37-336) by plasmin in this study. Factor Xa generation assay showed that the reconstituted factor VIIIa with truncated A1 forms (A1-(1-336) and A1-(37-336)) retained significant activity, whereas the onestage clotting assay completely lost these activities (13,18). This discrepancy can be explained as we described in an earlier report (13). In the factor Xa generation, a K m value for factor X using factor Xase with native factor VIIIa is ϳ40 nM, whereas this value is increased 5-fold (K m , ϳ200 nM) for that with factor VIIIa with A1-(1-336) or A1- . Because the typical factor Xa generation assay uses concentrations (400 nM) of substrate that yield near V max reaction rates, the rates are inde- pendent of factor X concentration. However, because the factor X concentration in plasma is ϳ120 nM and this plasma is diluted 4-fold in one-stage clotting assay, the limiting amount of factor X (ϳ15% of K m for factor Xase with factor VIIIa with truncated A1 forms) markedly depresses the rate of factor Xase activity, and consequently the A1-(1-336)/A2/A3C1C2 or A1-(37-336)/A2/A3C1C2 loses the factor VIIIa activity in onestage clotting assay.
Both factor Xa and plasmin inactivate factor VIIIa by cleavage at Arg 336 and Lys 36 in the A1 domain (33). An earlier study by Nogami et al. (13,33) revealed that proteolysis by factor Xa at these sites in A1 correlated with inactivation of cofactor function. Close analysis of A1-(1-336) and A1-(37-336) subunits demonstrated inactivation of native factor VIIIa activity by ϳ30 and ϳ60%, respectively, in factor Xa generation assay. Loss of activity resulting from cleavage at Lys 36 is associated with an altered molecular conformation that markedly affects the affinity of the A1 subunit for A2 (13). Similarly, factor Xa maximal generation with 1721 A3C1C2 subunit in the presence of A1-(37-336) was observed to be ϳ50% that in the presence of A1-(1-336) (ϳ55 and ϳ25 nM/min, respectively), supporting that the N terminus of the A3 domain is unlikely related to associate with the A1 domain. The cleavage rate by factor Xa at Lys 36 was Ͼ10-fold lower than that at Arg 336 (33). Factor Xacatalyzed cleavage at Lys 36 , however, is governed by the A1-(337-372) region, in particular by Asp 361-363 residues (34), and proteolysis at Arg 336 might interfere with factor Xa docking to the 337-372 region. Cleavage at the identical site by plasmin is regulated by the A3 domain, not by A1. Therefore, we can speculate that Lys 36 cleavage by plasmin may be faster than that by factor Xa, although the kinetics on cleavages at both sites by its protease remains to be determined.
The LBS in plasmin consists basically of a cationic center (Lys 35 and Arg 71 ), an anionic center (Asp 55 and Asp 57 ), and a hydrophobic core (Trp 62 , Phe 64 , Trp 72 , and Tyr 74 ) (35). The LBS facilitates interaction with substrates and proteins by hydrogen bond and/or ion pair interaction with the cationic or anionic center and van der Waals electronic interaction with the hydrophobic core (36). Recently, we reported that 6-aminohexinoic acid, a specific inhibitor of LBS, blocked interaction between plasmin and the factor VIII light chain by ϳ90% (21), strongly suggesting that this mechanism is LBS-dependent. Two lysine-rich regions reside in the A3 domain at the N terminus containing Lys 1693 and Lys 1694 and in the mid-section containing Lys 1804 , Lys 1808 , Lys 1813 , and Lys 1818 . We have now identified plasmin-interactive sites in these two regions by competition experiments using synthetic peptides composed of residues 1690 -1705 and 1804 -1818. Random peptides with scrambled sequences and peptides in which lysine was substituted with alanine failed to compete in the light chain-Ah-plasmin interactions. Our results therefore suggested that lysine residues and structural arrangements within both regions contributed significantly to plasmin docking in the A3 domain.
As noted above, both individual A3 peptides (1690 -1705 and 1804 -1818) partially inhibited (by ϳ50%) the binding of 1649 A3C1C2 to plasmin, and these effects appeared to be additive. In contrast, however, only mixtures of both peptides significantly inhibited plasmin-catalyzed cleavage of functional factor VIIIa at the Lys 36 site. These findings were similar to those observed using functional and binding experiments with the 1722 A3C1C2 subunit. The reasons for the discrepancy between functional and binding assays remain to be fully explored, and studies using modified A3C1C2 in which the 1804 -1818 sequence is deleted could be informative. Nevertheless, our data indicate that although each of these regions within the A3 domain interacts separately with plasmin, both are required for plasmin docking and inactivation of procoagulant factor VIIIa mediated by cleavage at Lys 36 . The factor VIII domain model based on homology with ceruloplasmin (37) describes putative lysine residues, Lys 1693 , Lys 1694 , Lys 1804 , Lys 1808 , Lys 1813 , and Lys 1818 , within residues 1690 -1705 and 1804 -1818 of the A3 domain, arranged spatially adjacent and exposed on the A3 surface (Fig. 9). This provides an extended surface for plasmin binding, but it is far removed from the cleavage site of Lys 36 within the A1 domain. Glu-plasminogen contains five kringle domains and a catalytic domain in a closed form with a radius of gyration of ϳ40 Å. Conversion to Lys-plasmin by plasminogen activator, however, induces a marked conformational change, termed an open form, with greater flexibility and increased radius of gyration (ϳ60 Å). The dramatic structural alteration enhances high affinity enzyme interactions with protein ligands (38,39). Molecular mechanisms of this nature could explain the functional relationship between the remote plasmin-interactive sites in A3 and the proteolytic cleavage site at Lys 36 . In addition, structural modification of the factor VIII light chain bound to plasmin might preferentially support catalysis at Lys 36 . Interestingly, comparison of amino acid sequences among human, porcine, murine, and canine factor VIII molecules reveals that the two A3 regions are well conserved, supporting the concept that both binding domains are fundamental for protein interactions.
Our study indicated that these functionally essential subunits, involving residues 1690 -1705 and 1804 -1818 in the A3 domain of factor VIII, constitute a highly basic spacer region exposed on the surface contributing to interaction with plasmin. The 1804 -1818 sequence is known to participate in the interaction with at least three proteins, including factor IXa (30), alloantibody inhibitors from multiply transfused hemophilia A patients (40), and low density lipoprotein receptorrelated protein that mediates clearance of factor VIII from the circulation (41). Therefore, it is clear that these particular residues play a significant role in the modulation of coagulation reactions by up-and down-regulation of factor VIIIa cofactor function. The 1690 -1705 sequence does not seem to have been reported to be involved in other protein interactions.
EGR-factor IXa inhibited factor VIII light chain binding to Ah-plasmin in a solid-phase assay. Furthermore, the presence of phospholipid enhanced the inhibitory effect of EGR-factor IXa and resulted in an ϳ15-fold decrease of the K i value for EGR-factor IXa binding. These K i values were consistent with the K d values obtained for the light chain-factor IXa interactions observed using a steady-state fluorescein energy transfer fluid assay (25). However, binding stoichiometry for A3C1C2 form and EGR-factor IXa in the presence of phospholipid was unexpectedly different in competition assays. Several possibilities for this reason may be raised. A solid-phase ELISA is a limited assay and is not based on an equilibrium binding. Because the binding affinity for the A3C1C2 of Ah-plasmin is much lower (ϳ15-fold) than that of factor IXa, this difference may affect the competitive inhibition. Furthermore, the A3C1C2 itself bound to factor IXa on the phospholipid membranes may partially affect competitive inhibition, for instance the influence of plasmin binding because of conformational alteration, etc. In addition, the possibility of steric hindrance because of the EGR molecule of its competitor cannot be excluded. However, the precise reason is unclear at present.
A plasmin-interactive site has been recently identified in the A2 domain within and/or close to a factor IXa-interactive site (21), and in our studies, peptide 1804 -1818, representing a factor IXa-interactive site, similarly inhibited the interaction between light chain and Ah-plasmin. Moreover, EGR-factor IXa blocked plasmin-catalyzed inactivation of factor VIIIa in a clotting assay. Therefore, overall the data support that factor IXa, bound to factor VIIIa on the phospholipid surface, might restrict plasmin-induced inactivation of factor VIIIa by occupying the plasmin-interactive sites in the A2 and A3 domains. This mechanism by which factor IXa protects factor VIIIa from plasmin-catalyzed inactivation appears to be similar to that observed with APC-catalyzed inactivation of factor VIIIa (26).
In conclusion, the extended surface of the factor VIII A3 domain, centered on lysine residues in both the 1690 -1705 and 1804 -1818 regions, contributes to a unique plasmin-interactive site that facilitates plasmin docking during cofactor inactivation by cleavage at Lys 36 . Our present results further suggest that factor VIII cofactor function is regulated by more complex mechanisms than previously anticipated.