Defining the Factor Xa-binding Site on Factor Va by Site-directed Glycosylation*

Activated Factor V (FVa) functions as a membrane-bound cofactor to the enzyme Factor Xa (FXa) in the conversion of prothrombin to thrombin, increasing the catalytic efficiency of FXa by several orders of magnitude. To map regions on FVa that are important for binding of FXa, site-directed mutagenesis resulting in novel potential glycosylation sites on FV was used as strategy. The consensus sequence for N-linked glycosylation was introduced at sites, which according to a computer model of the A domains of FVa, were located at the surface of FV. In total, thirteen different regions on the FVa surface were probed, including sites that are homologous to FIXa-binding sites on FVIIIa. The interaction between the FVa variants and FXa and prothrombin were studied in a functional prothrombin activation assay, as well as in a direct binding assay between FVa and FXa. In both assays, the four mutants carrying a carbohydrate side chain at positions 467, 511, 652, or 1683 displayed attenuated FXa binding, whereas the prothrombin affinity was unaffected. The affinity toward FXa could be restored when the mutants were expressed in the presence of tunicamycin to inhibit glycosylation, indicating the lost FXa affinity to be caused by the added carbohydrates. The results suggested regions surrounding residues 467, 511, 652, and 1683 in FVa to be important for FXa binding. This indicates that the enzyme:cofactor assembly of the prothrombinase and the tenase complexes are homologous and provide a useful platform for further investigation of specific structural elements involved in the FVa·FXa complex assembly.

The formation of thrombin from prothrombin is a key event in the coagulation process. In this reaction, the activated form of coagulation Factor V (FVa) 1 functions as a cofactor to the enzyme Factor Xa (FXa). The two proteins are assembled on the surface of negatively charged phospholipid membranes to the highly efficient prothrombinase complex (1)(2)(3)(4).
Factor V (FV) is a single-chain procofactor (330 kDa), having the domain organization A1-A2-B-A3-C1-C2 in common with the homologous Factor VIII (FVIII) (5)(6)(7). FV circulates in human plasma at a concentration of 7 g/ml but is also present in platelets from where it is released and activated during hemostasis (8). FV is converted to its activated form FVa upon proteolytic cleavage mediated by either thrombin or FXa. During this reaction, the B domain of FV is released from the active FVa molecule (9 -12). FVa is composed of the 105-kDa heavy chain (A1-A2 domains) and the 71/74-kDa light chain (A3-C1-C2 domains), the two chains being held together by noncovalent, calcium-ion dependent forces (1,11,12). The activity of the prothrombinase complex is regulated by activated protein C (APC), which inhibits FVa by proteolytic cleavage at Arg-306, Arg-506, and Arg-679 (13).
Factor X (FX) is a zymogen to a serine protease. It is composed of a Gla domain, two epidermal growth factor-like domains, an activation peptide, and a serine-protease domain. During blood coagulation, FX is activated by either the extrinsic pathway Factor VIIa⅐tissue factor complex or the intrinsic pathway tenase complex that comprises the phospholipidbound Factor IXa (FIXa) and activated FVIII (FVIIIa) (14). FVIIIa is a cofactor to FIXa in the activation of FX, a reaction that in many respects is very similar to the prothrombin activation (2). After activation of FX, the light chain (18 kDa) containing the Gla and two epidermal growth factor-like domains remains associated via a disulfide bridge to the heavy chain (27 kDa) composed of the serine-protease domain (15,16).
Even though the kinetics of the prothrombinase complex have been well characterized, surprisingly little is known about structure-function relationships of the complex and the sites of protein-protein interaction between FVa and FXa. Two regions in the A2 domain of FVa, 493-506 (17) and 311-325 (18), have, based on peptide studies, been suggested to be important for binding of FXa. In addition, removal of the C-terminal region of the A2 domain, 684 -709, has been shown to result in reduced FXa affinity (19). In contrast to the FVa⅐FXa complex, extensive interaction studies have been carried out on the structurally and functionally homologous FVIIIa⅐FIXa complex. Various interaction sites for FIXa have been identified on FVIIIa. Three of these binding sites, comprising residues 511-530, 558 -565, and 692-710, are located in the A2 domain (20 -22). Furthermore, a FIXa binding sequence, including residues 1811-1818 in the A3 domain of the FVIIIa light chain, has also been identified previously (23).
To map the regions on FVa that are important for binding of FXa, targeted glycosylation created by site-directed mutagenesis was used as the experimental strategy. It has been shown that mutations aiming at introduction of N-linked glycosyla-tion at selected residues can be a useful way to probe molecular regions involved in protein-protein interaction (24 -26). The size of a carbohydrate chain is convenient for searching binding site regions, because it probes larger areas than can be achieved via amino acid scanning, while it is more directed than peptide or antibody inhibition studies. The targeted glycosylation was achieved by introduction of the sequence signaling for N-linked glycosylation, Asn-X-Thr. Threonine rather than serine was chosen at the third position of the consensus sequence of glycosylation, because it has been shown to result in a higher degree of glycosylation than when serine occupies the third position (27,28).
A recently created three-dimensional model of FVa was very helpful for the selection of adequate sites for the novel carbohydrate side chains (29,30). We focused on introducing the glycosylation sites at residues that are surface-exposed in the three-dimensional model of FVa. Unfortunately, in the model, the end of the A2 domain (after residue 656) is missing. In addition to the computer model of the A domains of FVa, we designed our mutations taking into account documented binding sequences of FVIIIa and corresponding homologous sequences in FVa (20 -23). The results of this study show that regions surrounding residues 467, 511, 652, and 1683 of FVa are important for binding of FXa and thus for the assembly of the prothrombinase complex.

EXPERIMENTAL PROCEDURES
Materials-Bln1 was from Roche Molecular Biochemicals (Germany), and Bsu36I and BspEI were from New England BioLabs (Beverly, MA). Pwo polymerase and T4 DNA ligase were purchased from Roche Molecular Biochemicals (Penzberg, Germany). A doublestranded DNA sequencing kit was obtained from PerkinElmer Life Sciences. Cell culture media (Optimem Glutamax) were from Invitrogen. LipofectAMINE 2000 was from Invitrogen. Tunicamycin was from Sigma Chemical Co. (St. Louis, MO). Benzamidine was purchased from Acros Organics. Phe-Pro-Arg-chloromethylketone (PPACK) was obtained from Calbiochem (La Jolla, CA). A monoclonal antibody (Mk1) directed to the B domain of Factor V has been described before (31). Monoclonal HV-1 reacting with the C2 domain of Factor V was from Sigma. Monoclonal antibody AHV-5101 was from Hematologic Technologies Inc. A polyclonal antibody (A299) against FV was from Dako (Copenhagen, Denmark). Polyclonal anti-human FV antibody (8806) was raised by our group and showed epitopes to both the heavy and light chain as well as to the B domain. Egg extracts of phosphatidylcholine (PC) and phosphatidylethanolamine (PE) and brain extracts of phosphatidylserine (PS) were purchased from Avanti Polar Lipids (Birmingham, AL). The chromogenic substrate D-Phe-(pipelocyl)-Arg-pnitroaniline (S-2238) was a kind gift of Chromogenix (Milano, Italy). Streptavidin-coated magnetic beads (Dynabeads, Streptavidin) were purchased from Dynal (Oslo, Norway). Biotinylated PE was purchased from Molecular Probes (Eugene, OR) and [ 14 C]PC from Amersham Biosciences (Uppsala, Sweden).
Proteins-Human prothrombin and human FXa were purchased from Kordia (Leiden, Netherlands). Human ␣-thrombin was obtained from Hematologic Technologies Inc. Bovine serum albumin (BSA), ovalbumin, soybean trypsin inhibitor, and leupeptin were purchased from Sigma. Aprotinin was from Bayer (Leverkusen, Germany). Bovine FX (bFX) was purified as described (32) and activated using Russel's viper venom (33). Bovine-Factor Xa (bFXa) was labeled with 125 I using the chloramine T method. The specific activity was 20,000 cpm/ng, equal to 0.4 -0.5 MBq/g of protein. The labeled bFXa was characterized on SDS-PAGE and functionally analyzed using a prothrombinase assay and appeared unaffected by the labeling procedure.
Site-directed Mutagenesis-Mutations were introduced into the expression vector pMT2 containing the full-length cDNA of human FV using the QuikChange site-directed mutagenesis kit (Stratagene). For each mutant two complementary oligonucleotides containing the appropriate mutations were used. The mutated fragments were then isolated by restriction enzymes and used to replace corresponding fragments in the template. For each mutant, the sense primers used for mutagenesis and the restriction enzymes used for digestion are shown in Table I. The sequences of the fragments were confirmed by DNA sequencing.
Expression of Recombinant FV (rFV)-Expression plasmids containing the various FV cDNA constructs were transfected into COS 1 cells using both the DEAE-dextran method as described (34) and the Lipo-fectAMINE 2000 (Invitrogen) method as described by the manufacturer. Proteins were collected 72 h after transfection in serum-free media (Optimem Glutamax). For the expression in the presence of tunicamycin, cells were pretreated with 10 g/ml tunicamycin prior to protein collection in serum-free media in the presence of 10 g/ml tunicamycin.
Concentration and Quantification of rFV-Media containing the recombinant proteins were concentrated ϳ40-fold using Vivaspin 100,000 molecular weigh cut off, on the day of harvest, and the FV concentration was measured using an ELISA, after minor modifications (35). In brief, microtiter plates were coated overnight with 10 g/ml monoclonal Mk1 and then quenched for 30 min. Samples were diluted in ELISA buffer supplemented with 10 mM benzamidine and 2 mM CaCl 2 and incubated in 4°C overnight. Standard curves were created using pooled normal citrated plasma, assuming the FV concentration to be 7 g/ml (8). Biotinylated monoclonal HV-1 (0.1 g/ml) was used as secondary antibody, and after 2 h streptavidin-peroxidase was added. After 30 min, the plates were developed for 3 min, and the absorbance was measured at 490 nm. To rule out that the cell culture medium interfered with the ELISA, plasma-purified FV was diluted in mock medium or in buffer. No difference was detectable, indicating that the medium does not interfere.
Pulse Labeling of rFV-A pulse Labeling experiment was performed TABLE I Sense primers used for mutagenesis and restriction enzymes employed for isolating the fragments containing the mutations, which were then used to replace the corresponding fragment in WT FV cDNA Changes introduced are in boldface; antisense counterpart primers are not shown. The size of each cleaved fragment, which was isolated and ligated into the WT FV cDNA lacking the corresponding fragment, is shown in parentheses. The three amino acids given before the number represent WT FV sequence; the number, referring to the amino acid residue after mutagenesis, represents the carbohydrate attachment site.

Mutant
Primer Enzymes 24 h after transfection by radioactive labeling with [ 35 S]methionine and [ 35 S]cysteine as previously described with some modifications (36). In brief, cell extracts were prepared by lysis in a cell lysis buffer supplemented with inhibitors. Cell lysate was precleared with protein A-Sepharose (Amersham Biosciences) for 2 h at 4°C and immune precipitated overnight at 4°C using a polyclonal anti-FV antibody, 8806 (20 g). The immune complexes were precipitated and washed with immune precipitate washing buffer. After elution by boiling, samples were subjected to 2 units/ml thrombin for 30 min in an activation buffer. Proteins were deglycosylated by endoglycosidase H digestion (1 unit/ml) overnight at 37°C in an endoglycosidase H buffer, supplemented with inhibitors. Boiling the sample terminated the digestion. Proteins were separated by electrophoresis on a 7.5% SDS-PAGE (37). The gels were then exposed in a cassette and finally scanned using a PhosphorImager (Amersham Biosciences).
Purification of rFV-The purification was performed as described by Heeb et al. (38) after some modifications. The supernatant of conditioned medium of rFV (100 ml) was concentrated to 3 ml using the Vivaspin 100,000 molecular weight cut off. Concentrated medium was treated with 10 mM benzamidine and 2 M PPACK and incubated on ice for 30 min. The supernatant was loaded onto a 5-ml Hi-trap column (Amersham Biosciences) coupled with 2 mg of anti-(FV light chain) monoclonal antibody AHV 5101 at a flow rate of 0.18 ml/min. The column was then washed with 20 ml of buffer at a flow rate of 0.18 ml/min. Finally, the column was then eluted with buffer B at a flow rate of 0.5 ml/min and collected in 1-ml fractions. Fractions were analyzed for FVa activity and the SDS-PAGE profile. Pooled fractions were desalted using PD-10 (Amersham Biosciences) and stored at Ϫ80°C.
Phospholipid Vesicle Preparation-Phospholipids dissolved in chloroform/methanol (9:1, v/v) were dried in a glass tube under a mild flow of nitrogen. The phospholipids were suspended in 25 mM Hepes, 150 mM NaCl (pH 7.5), vigorously vortexed for 2 min, and then sonicated for 10 min at room temperature at amplitude 3, using an XL 2020 sonicator (Misonix). For the binding experiments, the phospholipid composition of the vesicles was PE/PS/PC at molar ratio of 10/20/70. A trace amount of 14 C-labeled PC was included, and 10% of the PE was biotinylated. For the prothrombinase assay, the phospholipid composition was PS/PC, at molar ratio of 10/90.
Prothrombinase Assay-Recombinant FV proteins were incubated with 0.5 unit/ml of ␣-thrombin (Hematologic Technologies Inc.) for 10 min at 37°C prior to analysis in the prothrombinase assay. Typically, if not specifically noted, the conditions for the prothrombinase assay were 50 pM FVa, 5 nM FXa, 0.5 M prothrombin, and 50 M phospholipid vesicles (10/90 of PS/PC). Thrombin generation was allowed to continue for 1 min before the reaction was stopped by dilution with ice-cold EDTA buffer. The amount of generated thrombin was quantified using the chromogenic substrate S-2238. In control experiments, plasmapurified FVa and recombinant wild-type (WT) FVa in conditioned medium behaved similarly with respect to FXa cofactor activity and FXa affinity. Moreover, in the absence of added FV, there was no appreciable thrombin generation.
Determination of Apparent K d of FXa for FVa Using the Prothrombinase Assay-The prothrombinase assay was used to determine the apparent K d for the binding of FXa to the thrombin-activated FV variants. The formation of membrane-bound FXa⅐FVa complexes was measured by determining the rates of prothrombin activation in the presence of phospholipid vesicles at increasing concentrations of FXa and a fixed concentration of FVa. FVa (50 pM) was preincubated for 4 min with FXa (0.1-50 000 pM) and phospholipid vesicles (50 M of 10/90, PS/PC). Thrombin generation was started by addition of 0.5 M preheated prothrombin and allowed to continue for 1 min. The apparent K d for the membrane-bound FXa⅐FVa complex was obtained from plots of the rate of thrombin generation as a function of the FXa concentration. The K d was obtained by fitting the data to the following equation for a singlesite binding isotherm using non-linear least squares regression analysis, The free concentration of FXa [FXa] was assumed to be equal to the added concentration of FXa. The binding maximum, B max , set as the value where the binding of FXa to WT FVa, was saturated. Determination of K m for Prothrombin Activation by FXa Using the Prothrombinase Assay-The K m for prothrombin activation by the different prothrombinase complexes was determined by varying the prothrombin concentrations. FVa (200 pM) was preincubated for 4 min with increasing concentrations of prothrombin (5-2500 nM) and 25 M phospholipid vesicles (5/95, PS/PC). Thrombin generation was started by addition of preheated FXa (5 nM final concentration) and allowed to proceed for 1 min before being stopped.
Binding of FXa to Membrane-bound FVa Using Magnetic Beads-A magnetic bead-based assay for FXa binding to membrane-bound FVa was performed as described by Rudolph, after some modifications (39). In this assay, biotinylated phospholipid vesicles were immobilized on the surface of streptavidin-coated magnetic beads. The beads were washed twice with the assay buffer and then quenched during constant shaking with BSA-containing buffer for 4 h at 37°C. The beads (2.5 mg/ml) were mixed with 2 mM small unilamellar phospholipid vesicles (biotinylated-PE/PE/PS/PC/[ 14 C]PC, with the molar ratio 1/9/20/70/ trace amounts) and incubated for 2 h at 37°C during constant shaking. The beads were then washed twice and resuspended in assay buffer. The concentration of phospholipid was determined by counting [ 14 C]PC and typically found to be around 50 M.
To measure the binding of FXa to membrane-bound FVa, FVa (0.5 nM) was incubated for 10 min with the phospholipid-coated beads. The final concentration of the phospholipid was 625 nM (this concentration was experimentally found to give the best signal-to-noise ratio). Increasing concentrations of 125 I-bFXa (200 -6000 pM) were incubated with the beads for 30 min. The binding reaction was stopped by isolation of the magnetic beads using an MPC-96 plate (Dynal). The collected beads were washed with ice-cold assay buffer to reduce nonspecific binding. The amount of FXa that was associated with the beads and the FXa remaining in the supernatants were measured. Nonspecific binding determined from reactions containing a 100-fold excess of unlabeled bFXa was subtracted from the total binding. In addition, binding that was not dependent on FVa was determined in parallel from reactions lacking added FVa. To estimate the K d of the FXa binding to FVa, the amount of bound FXa was plotted as a function of added FXa concentration. The data were fitted to the above equation for a single sitebinding isotherm via non-linear least squares regression analysis. In control experiments, plasma-purified FVa and recombinant WT FVa in conditioned medium behaved in an identical manner.
Statistical Analysis-All results are presented as mean values Ϯ S.E. of three independent experiments performed in duplicate.

RESULTS
Mutagenesis Strategy-To identify potential interaction sites on FVa for FXa and/or prothrombin, the consensus' sequence Asn-X-Thr for N-linked glycosylation was introduced at thirteen different positions in FV. We took advantage of previously identified interaction sites for FIXa in FVIIIa when choosing probing sites, assuming the FIXa⅐FVIIIa and FXa⅐FVa complexes to have similar molecular architectures. Four different FIXa interaction sites in FVIIIa have been identified; three in the A2 domain comprising regions 511-530, 558 -565, and 692-710 and one in the A3 domain composed of residues 1811-1818 (20 -23). Corresponding regions in FVa include residues 455-474, 502-509, and 638 -655 in the A2 domain and 1676 -1683 in the A3 domain. Therefore, carbohydrate side chains were introduced at FV residues 461, 467, 511, 652, 655, 1677, and 1683 to probe the different regions for FXa interaction sites. In addition, six other carbohydrate side chains were introduced in areas surrounding these segments. These sites were at residue 172, 319, 345, 373, 435, and 450. In all cases, the consensus sequence for glycosylation was introduced within solvent exposed regions in areas in which substitutions and glycan grafting should be structurally tolerated.
FV Variants with Attenuated Procoagulant Activity-The different FV variants were expressed transiently in COS 1 cells, and their concentrations were determined with ELISA. All of the variants yielded expression levels similar to that of WT FV, around 200 ng/ml. The procoagulant activity was tested in a prothrombinase-based system at two different conditions, one with limiting concentrations of FV (5 pM) and saturating concentrations of FXa (5 nM), and the other with saturating concentrations of FV (100 pM) and limiting concentration of FXa (5 pM). Four of the mutants, having their glycosylations located at positions 467, 511, 652, or 1683, demonstrated decreased FXa cofactor activity (less than 50% of WT) and were chosen for further functional characterization (Table II). These FV vari-ants were analyzed by Western blotting before and after activation by thrombin and detected using a polyclonal antibody (A299). The intact FV variants migrated identically to single chain WT FV, and the thrombin-activated FV similarly to WT FVa (data not shown). The resolution of this system was insufficient to detect the molecular weight differences that were due to carbohydrate variations.
SDS-PAGE Analysis of the Pulse-labeled FV Variants-To assess whether the FV variants had the additional carbohydrate side chain or not, the mobility on SDS-PAGE of pulselabeled thrombin-activated FV variants was compared. In this experiment, pulse-labeled FV immunoprecipitated from the cell lysates was chosen rather than FV secreted in the medium, because the proteins in the cell lysate carry relatively homogenous carbohydrate side chains, which facilitate the analysis. As judged by the electrophoretic analysis, an extra glycan was present in all mutants having impaired procoagulant activity (Fig. 1A). Thus, the apparent molecular weights of the heavy chains of FV variants 467 NNT, 511 NAT, and 652 NDT were increased as compared with that of WT FV. The light chain of the 1683 variant demonstrated altered mobility compatible with the presence of an extra carbohydrate side chain. After deglycosylation of the FV variants with endoglycosidase H, no differences were observed between the mutants having an extra carbohydrate in the heavy chain, i.e. 467 NNT, 511 NAT, and 652 NDT, and WT FV. After endoglycosidase H treatment, the heavy chains migrated as multiple closely spaced bands, suggesting partial deglycosylation (Fig. 1B). After the deglycosylation, the light chains migrated as a single band indicating that the partial glycosylation disappeared. The light chain of the 1683 NAT variant migrated as the light chain of WT FV, demonstrating release of the extra carbohydrate side chain. The additional nine FV mutants that had full biological activity were also pulse-labeled and analyzed on SDS-PAGE after activation with thrombin to investigate whether the mutagenesis indeed resulted in the attachment of an extra carbohydrate side chain. The heavy chains of two of these mutants, 435 NET and 655 NCT, migrated like the heavy chain of WT FVa, indicating that they did not carry an extra carbohydrate, whereas the other seven mutants appeared to be glycosylated at the sites of mutation (Fig. 2, A and B). The sequence for glycosylation at position 655 (NCT) is not optimal for the glycosylation reaction, which might explain why position 655 was not glycosylated. Only a fraction of the 319 NKT variant demonstrated increased molecular weight on the SDS-PAGE anal-ysis, suggesting partial glycosylation at position 319.
Apparent K d for FXa Binding to FVa as Determined by Prothrombinase Assay-Impaired binding of FXa or decreased interaction with prothrombin may be the cause of the poor FXa cofactor activity of the FV variants that was observed in the prothrombinase assay. In addition, a decreased catalytic activity of the bound FXa, as reflected by a low k cat value, could potentially result in decreased biological activity of the FV variants. To characterize the nature of the attenuated cofactor activity of the FV mutants, their ability to support prothrombin activation at increasing FXa concentrations was studied (Fig.  3). Three of the FV mutants, 467 NNT, 511 NAT, and 1683 NAT, demonstrated poor FXa cofactor activity, which was particularly pronounced in the case of variants 467 NNT and 1683 NAT. The 511 NAT variant demonstrated a different pattern with the highest concentrations of FXa tested yielding around 70% of the maximum activity of WT FVa. In contrast, the 652 TABLE II Screening to find mutants with attenuated procoagulant activity The procoagulant activity of the mutants were tested in a prothrombinase-based system. The activity of the mutants were determined at two different conditions: with limiting concentrations of FVa (5 pM) and saturated concentration of FXa (5 nM) and with saturated concentrations of FVa (140 pM) and limiting concentration of FXa (5 pM FIG. 1. SDS-PAGE analysis of the mutants with attenuated procoagulant activity. COS 1 cells expressing the WT, 467 NNT, 511 NAT, 652 NDT, and 1683 NAT FV variants were radiolabeled for 1 h (pulse). The radiolabeled FV in the cell lysates was immunoprecipitated and incubated with thrombin (2 units/ml) for 30 min at 37°C (A). The positions of the heavy (HC) and light (LC) chains are indicated. In B, the thrombin-treated immune precipitates were subjected to endoglycosidase H (1 unit/ml) digestion overnight at 37°C. The sharp upper band represents the deglycosylated B domain. The HC is partially deglycosylated as illustrated by the multiple closely spaced bands, whereas the LC appears to be fully deglycosylated. Proteins were separated on 7.5% SDS-polyacrylamide gel electrophoresis and detected using a PhosphorImager.
NDT variant reached a plateau of maximum thrombin generation, which was ϳ50% of that of WT FVa. The results of this experiment allowed calculation of apparent K d for WT FVa and the 652 NDT and 511 NAT variants (Table III). Wild-type FVa demonstrated an apparent K d of 0.2 nM, which agrees with results on record from similar experiments (19,40,41). The 652 NDT variant yielded a similar K d as WT FV, but the maximum rate of thrombin generation was only half of that of WT FV. The K d for the 511 NAT variant was estimated to be 6 nM. The experiment illustrated in Fig. 3A was performed at 50 pM FVa. To ensure that the low maximum thrombin generation rate observed for several of the variants was not the result of low FVa levels, the FXa concentration was kept constant at 5 pM and the FVa increased up to a maximum of 2 nM (Fig. 4). In this case, the thrombin-generation rate of the 652 NDT variant was still around 50% of that of WT FVa. The 511 NAT variant also reached around 60% rate of the WT FVa thrombin generation, whereas the 467 and 1683 variants were less efficient in supporting prothrombin activation.
In the described experiments, the recombinant FV were not purified from the condition medium. To ensure that components of the medium did not affect the results, control experiments in which plasma-purified FVa was added to mock me-dium were performed. These experiments demonstrated that the medium did not affect the ability of FVa to express FXa cofactor activity. In addition, the recombinant mutants demonstrating low activity and WT FVa were analyzed after purification (Fig. 3B). Because the concentrations of the recombinant mutants were considerably lower after purification (0.1 nM), the FVa variants were allowed to generate thrombin for longer duration of time (8 min). The results obtained were similar to those derived from experiment with non-purified FV, suggesting that it is adequate to study the FV variants without prior purification.
Restored Procoagulant Activity upon Synthesis in the Presence of Tunicamycin-To ensure that the attenuated FXa affinity of the FVa variants was the result of the added carbohydrate side chain rather than a direct effect of the mutations, the recombinant FV variants were expressed in the COS 1 cells in the presence of tunicamycin, which inhibits N-linked glycosylation. The expression levels obtained in the presence of tunicamycin were lower than in its absence but enough FV was recovered for analysis in the prothrombinase assay (Fig 5). The rate of thrombin generation was measured at increasing concentrations of FXa, which allowed estimation of the binding ability of FXa to the recombinant FVa. Of the investigated FV variants, only the 511 NAT variant demonstrated reduced ability to bind FXa. All the other variants yielded similar profiles as WT FV, suggesting tunicamycin to restore the procoagulant activities of the mutants. Thus, the deteriorated FXa affinity of the FV variants observed with FV expressed in the absence of tunicamycin was caused by the added carbohydrate side chains rather than by the amino acid replacements per se.
Prothrombin Titration to Determine K m -To estimate the interaction between prothrombin and the prothrombinase complexes formed by FXa and the various FV variants, the rate of thrombin generation was determined at increasing concentrations of prothrombin (Fig. 6). To ensure that the prothrombin interacted specifically with the FV⅐FXa complex rather than with the phospholipid surface, high concentrations of Va (200 pM) and suboptimal vesicles were used. The phospholipids had a composition of 5/95 PS/PC and was used at a concentration of 25 M. Control experiments showed that, in the absence of Va, the K m for prothrombin activation was severalfold higher. This indicates that prothrombin interacted specifically with the FVa⅐FXa complex. The results allowed determination of the K m for each reaction. All the FV mutants, when assembled in the prothrombinase complex, yielded similar values of K m for prothrombin (Table III).
FXa Binding to FVa Variants in the Presence of Phospholipid Vesicles-The functional studies were performed in the presence of prothrombin and therefore reflected the ability of FXa to interact with FVa/prothrombin rather than with FVa alone. A direct binding assay was therefore established to investigate the specific interaction between FXa and the phospholipid membrane-bound FVa variants (Fig. 7). In the assay, phospholipid-coated magnetic beads were incubated with the thrombintreated recombinant FV variants and then with radiolabeled bFXa. To estimate the level of nonspecific binding and to allow calculation of binding affinities, increasing amounts of unlabeled bFXa was added. The direct binding assay demonstrated no binding of FXa to the 467 NNT, 511 NAT, and 1683 NAT variants. The binding of FXa to the 652 NDT mutant was only slightly lower than that of WT FVa and the K d was estimated to be around 4-fold lower than that of WT FVa (Table III). DISCUSSION To date, three regions in FV involved in the binding of FXa have been identified, including residues 311-325, 493-506, and 684 -709, all of which are located in the A2 domain (17)(18)(19). The first two regions were identified using synthetic peptides, whereas the last was the result of specific proteolysis that liberated the 684 -709 peptide. To enhance our understanding of the FVa⅐FXa interaction, we screened putative binding regions in FVa by introducing N-glycosylations using site-directed mutagenesis. With this approach, we have now identified three previously unknown interaction sites for FXa. The A2 domain carried two of the sites, at residue 467 and 652, whereas the third was localized in the A3 domain at residue 1683.
To investigate whether the method we have used, i.e. to introduce novel N-linked glycans to probe binding regions, was able to confirm regions suggested by peptide studies, the 319 NKT and 511 NAT variants were created. The 511 NAT variant  Fig. 3. b K m values were calculated from data of Fig. 6  demonstrated mobility shift on SDS-PAGE compatible with the presence of a carbohydrate side chain at the 511 position. This FV variant demonstrated loss of procoagulant activity due to impaired FXa binding and supports the conclusion that the 493-520 region is involved in FXa binding (17). Moreover, this is consistent with the localization of a predominant FIXa interaction site in the A2 domain of FVIIIa, at residues 556 -569, which is equivalent to the 500 -513 region in FVa (42). The 319 NKT mutant, having a potential carbohydrate site located in the predicted binding region 311-325 of FVa (18), did not show attenuated FXa cofactor activity. However, upon SDS-PAGE analysis, only a small portion of the recombinant protein appeared to carry an extra carbohydrate. Thus, a likely explanation for the lack of inhibitory effect could be that the region is not properly masked, because the site is only partially glycosylated. It is also possible that the carbohydrate is directed away from the main binding surface for FXa, whereas the entire 311-325 segment or part of it is indeed involved in the interaction. It is also noteworthy that a homologous binding site in FVIIIa has not been identified.
There is no consensus at present as to whether FXa interacts with the light chain of FVa. The observation, that a monoclonal antibody directed against the A3 domain of FVa inhibits the FVa⅐FXa interaction, supports this direct interaction (43). However, the membrane-bound light chain of FVa has been proposed to not interact with FXa (44). The now observed inhibitory effect on FXa binding of the introduction of a carbohydrate side chain at 1683 supports the idea that the light chain carries an interaction site for FXa. It is interesting to note that a homologous FIXa interaction site on the light chain of FVIIIa is suggested to confer a most important interaction between FIXa and FVIIIa. The light chain of FVIIIa binds FIXa with a K d of about 15 nM, whereas a K d of around 300 nM characterized the binding of FIXa to the A2 domain (45)(46)(47). The proposed FIXa-binding region in FVIIIa includes residues 1811-1818 (the equivalent residues in FV are 1676 -1683), a region that is rich in basic Lys residues (46). FV does not carry such a cluster of Lys residues, but it is still possible that binding sites are present at this location in both cofactors. In this context it is noteworthy that the mutation introducing an N-linked glycosylation at 1677 did not affect the FXa binding. Residue 1677 is about 20 Å away from amino acid 1683 and could thus be outside the key interaction area.
The 652 NDT variant demonstrated an interesting functional pattern in the prothrombinase complex with a decreased k cat , whereas the K d for FXa and the K m for prothrombin seemed to be unaffected. This suggests that the extra carbohy- The three-dimensional model for the three A domains of FV is presented as a solid surface (the A1 domain is blue, A2 is white, and A3 is yellow). Only two of the three cleavage sites for APC can be visualized on this structure, namely at positions Arg-306 and Arg-506. The solvent-exposed surface of the residues expected to be glycosylated are colored in red. The residues identified to be important for the interaction with FXa in the present study are enclosed in boxes (magenta). The overall dimension of the surface delineated by residues 467, 511, 652, and 1683 is compatible with the size of the binding region identified on the FXa side. Our data are crucial for the future building of models of the prothrombinase complex, because we defined important regions for the contact as well as areas that should not be directly involved. drate side chain affects the catalytic activity of the bound FXa. The exact underlying mechanism is unknown. Proteolytic removal of residues 683-709 in FVa results in impaired FVa⅐FXa interaction, and it is possible that the region surrounding 652 is part of an interaction site that also includes residues 683-709. It is not possible to estimate how close residue 652 is to the 683-709 region, because the present three-dimensional models end at residue 656. Therefore, the region around residue 652 is difficult to analyze structurally. However, it is important to note that the equivalent region of the A2 domain of FVIIIa, encompassing residues 698 -712, has been reported to play a crucial role for the binding of FIXa (22).
A third A2 domain region of FVIIIa involved in FIXa binding was found after identification of a naturally occurring substitution at residue 527. This FIXa-binding site was confirmed by peptide inhibition studies, demonstrating a peptide comprising 511-530 to inhibit FIXa binding (20). In the present investigation, introduction of a carbohydrate at residue 467, which is located at the equivalent segment of FV, showed a remarkably attenuated FXa binding. In fact, the 467 NNT mutant appeared to have lost almost all FXa-binding ability. In the threedimensional model of FVa, residue 467 is strikingly close to the reported FXa binding region at 493-520 and as such could be part of an extended FXa interaction site.
Our ultimate goal is to understand the structure-function relationships of the fully assembled prothrombinase complex and to create a three-dimensional model for the complex. This requires information about the binding site on FXa for FVa. Recently, an FVa-binding site in FXa was identified and found to be a homologous counterpart to the FVIIIa binding site of FIXa (48). In FVIIIa the two regions at 517-527 and 556 -569 are suggested to bind to a surface loop at 330 -339 in the catalytic domain of FIXa (45,49). If the situation would be similar for the FVa⅐FXa complex, the FVa region surrounding residue 467 (residues 461-471) would, together with FV residues 500 -513, bind to a surface loop of FXa, including amino acids 344 -352.
Although the identified FXa binding regions in FVa are far apart in the primary amino acid sequence, i.e. residues 467, 652, and 1683 and regions 311-325, 493-520, and 683-709, they are relatively close in space according to the three-dimensional model predictions (Fig. 8). The three-dimensional model has some inherent limitations, e.g. some segments are missing and the resolution of some loops is low. Despite this, the model suggests that the identified multiple binding sites form an extended interaction site for FXa. This is compatible with the extended surface predicted to be involved in the interaction from the FXa side (48). The residues that were probed in this investigation but with negative results, i.e. with no measurable effect on FXa binding, also help define the binding surface.
In conclusion, we have defined three previously unidentified interaction sites for FXa on FVa, which are located close to residues 467 and 652 in the A2 domain and to residue 1683 in the A3 domain. This information provides a useful platform for further investigation of the specific structural elements that are involved in the FVa⅐FXa complex assembly. Our data suggest the enzyme:cofactor assembly of the prothrombinase and the tenase complex to be homologous, demonstrating the rational of using comparisons between the two complexes when mapping interaction sites.