Inhibition of Tissue Factor-Factor VIIa-catalyzed Factor X Activation by Factor Xa-Tissue Factor Pathway Inhibitor

The physiological inhibitor of tissue factor (TF)·factor VIIa (FVIIa), full-length tissue factor pathway inhibitor (TFPIFL) in complex with factor Xa (FXa), has a high affinity for anionic phospholipid membranes. The role of anionic phospholipids in the inhibition of TF·FVIIa-catalyzed FX activation was investigated. FXa generation at a rotating disc coated with TF embedded in a membrane composed of pure phosphatidylcholine (TF·PC) or 25% phosphatidylserine and 75% phosphatidylcholine (TF·PSPC) was measured in the presence of preformed complexes of FXa·TFPIFL or FXa·TFPI1–161 (TFPI lacking the third Kunitz domain and C terminus). At TF·PC, FXa·TFPIFL and FXa·TFPI1–161 showed similar rate constants of inhibition (0.07 × 108 m −1 s−1 and 0.1 × 108 m −1 s−1, respectively). With phosphatidylserine present, the rate constant of inhibition for FXa·TFPIFL increased 3-fold compared with a 9-fold increase in the rate constant for FXa·TFPI1–161. Incubation of TF·PSPC with FXa·TFPIFL in the absence of FVIIa followed by depletion of solution FXa·TFPIFL showed that FXa·TFPIFL remained bound at the membrane and pursued its inhibitory activity. This was not observed with FXa·TFPI1–161 or at TF·PC membranes. These data suggest that the membrane-bound pool of FXa·TFPIFL may be of physiological importance in an on-site regulation of TF·FVIIa activity.

The physiological inhibitor of tissue factor (TF)⅐factor VIIa (FVIIa), full-length tissue factor pathway inhibitor (TFPI FL ) in complex with factor Xa (FXa), has a high affinity for anionic phospholipid membranes. The role of anionic phospholipids in the inhibition of TF⅐FVIIacatalyzed FX activation was investigated. FXa generation at a rotating disc coated with TF embedded in a membrane composed of pure phosphatidylcholine (TF⅐PC) or 25% phosphatidylserine and 75% phosphatidylcholine (TF⅐PSPC) was measured in the presence of preformed complexes of FXa⅐TFPI FL or FXa⅐TFPI 1-161 (TFPI lacking the third Kunitz domain and C terminus). At TF⅐PC, FXa⅐TFPI FL and FXa⅐TFPI 1-161 showed similar rate constants of inhibition (0.07 ؋ 10 8 M ؊1 s ؊1 and 0.1 ؋ 10 8 M ؊1 s ؊1 , respectively). With phosphatidylserine present, the rate constant of inhibition for FXa⅐TFPI FL increased 3-fold compared with a 9-fold increase in the rate constant for FXa⅐TFPI  . Incubation of TF⅐PSPC with FXa⅐TFPI FL in the absence of FVIIa followed by depletion of solution FXa⅐TFPI FL showed that FXa⅐TFPI FL remained bound at the membrane and pursued its inhibitory activity. This was not observed with FXa⅐TFPI 1-161 or at TF⅐PC membranes. These data suggest that the membrane-bound pool of FXa⅐TFPI FL may be of physiological importance in an on-site regulation of TF⅐FVIIa activity.
Blood coagulation in vivo is initiated when circulating factor VII(a) binds in a calcium-dependent way to its cofactor, tissue factor (TF) 1 (see Refs. 1 and 2 for a review). This complex formation results in enhanced catalytic activity of factor VIIa (FVIIa), which via limited proteolysis, activates factors X (FX) and IX (FIX) (3). TF is a transmembrane glycoprotein, which under normal conditions is expressed only in extravascular tissues (4,5).
The main physiological regulator of TF-induced blood coag-ulation is tissue factor pathway inhibitor (TFPI) (6, 7), a single chain glycoprotein of 42 kDa and a member of the Kunitz family of serine protease inhibitors. TFPI contains an acidic N terminus, followed by three tandemly repeated Kunitz-type inhibition domains, and a basic C-terminal tail (8). Site-directed mutagenesis has revealed that the first Kunitz domain binds to FVIIa and that the second Kunitz domain interacts with the active site of FXa (9). No such functions could be attributed to the third Kunitz domain (10). Yet, various interactions have been ascribed to this domain, e.g. with lipoproteins and heparin, but their importance for the inhibitory function of TFPI is not clear (11,12). On the other hand, the basic C-terminal region of TFPI (residues 240 -276) has been shown to play a crucial role in the anticoagulant activity of this inhibitor (13,14). Despite numerous studies, it remains unclear how this basic C terminus modulates the anticoagulant activity of TFPI (15)(16)(17)(18)(19). TFPI inhibits the generation of FXa and FIXa by the TF⅐FVIIa complex in a unique, two-step reaction (20). First, TFPI binds Ca 2ϩ independently to FXa, thereby inhibiting the FXa catalytic activity (9). In a second step, the FXa⅐TFPI complex binds in a Ca 2ϩ -dependent way to TF⅐FVIIa. This results in the formation of the quaternary complex TF⅐FVIIa⅐FXa⅐TFPI, in which the proteolytic activity of the TF⅐FVIIa complex is fully neutralized. The effect of TFPI on TF⅐FVIIa activity in the absence of FXa is negligible (21,22), implying that the true inhibitor of TF⅐FVIIa activity is the FXa⅐TFPI complex. The rate of complex formation of FXa and TFPI is enhanced by negatively charged phospholipids for full-length TFPI (TFPI FL ) but not for TFPI 1-161 , a truncated variant lacking the third Kunitz domain and the potential phospholipid binding C-terminal tail (16,23).
Recently (24), we demonstrated that TFPI FL in complex with FXa has a much higher affinity for anionic phospholipid membranes compared with that of either protein alone. It is well recognized that the binding of blood coagulation enzymes as well as their cofactors and substrates to membranes containing anionic phospholipids may result in an immense increase of the catalytic efficiency of these enzymes. On the other hand, excess binding sites could cause a lowering of both solution concentration and surface density of the reactants (25,26), resulting in a decrease of reaction rates. Previous studies (17,22,23) did not reveal a stimulatory or interfering effect of anionic phospholipids on the complex formation between FXa⅐TFPI and TF⅐FVIIa. These studies, however, were performed in the presence of excess phospholipid vesicles. It remains unclear what the role is of lipid-protein interactions and of the TFPI C terminus when TF is embedded in a macroscopic phospholipid membrane. Answering these questions could provide insight in the role of (TF-bearing) cell membranes in the regulation of the TF⅐FVIIa-catalyzed initiation of the blood coagulation process.
To mimic the processes occurring at TF-bearing cell surfaces as closely as possible, we made use of TF embedded in a macroscopic phospholipid surface to which the reactants were supplied under well defined flow conditions. This approach allows a quantitative study of the kinetics of inhibition of TF⅐FVIIa-catalyzed FX activation as a function of the phospholipid composition of the membrane and the phospholipid-binding properties of the FXa⅐TFPI complex. Our study showed that negatively charged phospholipid surfaces act as a safety net in the sense that FXa⅐TFPI FL complexes readily bind to the surface. These phospholipid-bound FXa⅐TFPI FL complexes then show a potent inhibitory activity against TF⅐FVIIa present at the same membrane.

Proteins
Human FX was purified according to Mertens et al. (27). Human FXa was prepared by activation of purified FX with the FX activating protein from Russell's viper venom (Sigma) and isolated as was described for bovine FXa (28). The molar concentration was determined by active site titration with p-nitrophenyl p-guanidinobenzoate hydrochloride (29). Recombinant human TFPI FL , produced in Escherichia coli (30), was kindly supplied by Searle/Chiron (Emeryville, CA). This recombinant TFPI FL preparation, not being glycosylated, is very similar to the native highly glycosylated protein expressed in mammalian cells with respect to anticoagulant activity (16,30). The C-terminal truncated TFPI preparation was kindly supplied by Dr. O. Nordfang (Novo Nordisk, Bagsvaerd, Denmark). TFPI  was expressed in Saccharomyces cerevisiae and purified as described previously (31). The molar concentrations of the TFPI preparations were determined by titration with known amounts of FXa (18

Relipidation of Recombinant Human TF
Recombinant human TF, 20 and 100 nM, was reconstituted with 2 mM PSPC (25 mol% PS/75 mol% PC) and 2 mM PC (100 mol% PC), respectively, using n-octyl-␤-D-glucopyranoside (Calbiochem, La Jolla, CA) solubilization and dialysis as described (32). Because the TF is randomly oriented in the reconstituted vesicles, we considered the effective TF concentration to be 50% of the total concentration present in the reconstituted preparation (32). Calculations were therefore based on a TF:phospholipid ratio of 1:20 ϫ 10 4 and 1:4 ϫ 10 4 for PSPC and PC, respectively.

The Rotating Disc Device
The set-up used in the present study is a modification of the one described previously (33). In short, a cylinder (height 12 mm, width 16 mm) made of polyethylene terephthalate (PETP; AKZO Plastics, Amersloont, The Netherlands), containing four tiny magnets, was mounted on a motor with servo-controlled rotation speed. A cylindrical vessel is placed on top of the motor-containing holder, such that the cylinder rotates close to the glass bottom of the vessel. A glass coverslip (Menzel Glä ser, Braunschweig, Germany) with a diameter of 20 mm is attached to another PETP cylinder (height 2 mm, width 12 mm), also containing four magnets. This second cylinder fits into a thin PETP ring, mounted in the vessel, which guides the movements of the rotating disc. A stationary baffle was immersed in the solution (3 ml) during the experiments to establish a well defined flow pattern.

Preparation of Phospholipid-coated Discs
The glass coverslips of the rotating discs were made hydrophilic as described previously (34). In order to apply a TF-containing lipid bilayer, the rotating discs (63 rad/s) were exposed for 30 min to reconstituted vesicles in Tris buffer (50 mM Tris-HCl, pH 7.9, and 175 mM NaCl), with a final phospholipid-concentration of 20 M. Fluid phase vesicles were removed by flushing for 5 min with TBSA-CaCl 2 buffer (Tris buffer containing 3 mM CaCl 2 and 1 mg/ml BSA) at a rate of 10 ml/min using a Minipuls II pump (Gilson, Villiers-le-Bel, France). Finally the disc was transported, without exposing the lipid surface to an air-buffer interface, to a second vessel that had been pretreated for 1 h with 20 mg/ml BSA in Tris buffer. Before placing the coated disc in this second vessel, the high content BSA Tris buffer was replaced by TBSA-CaCl 2 buffer. All preparations were performed at 37°C. The effective TF surface density was 2.5 fmol/cm 2 and 12.5 fmol/cm 2 for PSPC and PC, respectively, as calculated on the basis of a phospholipid surface density of 0.5 nmol/cm 2 and TF:phospholipid ratios of 1:20 ϫ 10 4 or 1:4 ϫ 10 4 . The disc surface area was 3.14 cm 2 ; the theoretical amount of TF available at the surface is therefore 7.8 and 39 fmol for PSPC and PC, respectively. Throughout this paper, TF⅐PSPC and TF⅐PC refer to the planar surfaces thus obtained.

Measurement of FX Activation
Discs coated with TF⅐PSPC or TF⅐PC were spun at 63 rad/s in 3 ml TBSA-CaCl 2 buffer. FVIIa (1 nM) was added and incubated with the spinning disc for 5 min. The TF⅐FVIIa activity was then assessed from FXa generation following the addition of FX (100 nM). Timed samples were taken from the reaction mixture and collected into polystyrene cuvettes containing 50 mM Tris-HCl, pH 7.9, 175 mM NaCl, 20 mM EDTA, and 0.5 mg/ml BSA, (total volume 450 l). To determine the amount of FXa present, 50 l of 3 mM chromogenic substrate S2765 was added to the cuvette, and the conversion of the chromogenic substrate was followed at 37°C on a dual wavelength spectrophotometer at 405 nm (reference wavelength 500 nm). The FXa concentrations were calculated from standard curves obtained with known amounts of the enzyme. The TF containing phospholipid surface of the rotating disc could be reused after flushing with TBSA-EDTA buffer (20 mM EDTA) for 5 min at 10 ml/min, followed by a rinse with TBSA-CaCl 2 buffer (5 min at 10 ml/min). All procedures were performed at 37°C and unless otherwise mentioned, at an angular velocity of 63 rad/s.

Inhibition of TF⅐FVIIa Activity
All inhibition experiments were carried out with preformed FXa⅐TFPI complexes. These complexes were prepared by incubating FXa and TFPI at equimolar concentrations (50 nM) for 30 min at 37°C in TBSA-CaCl 2 buffer. This incubation time was sufficiently long to obtain an equilibrium in complex formation (95% of FXa in complex), as was established by measuring the free FXa concentration with chromogenic substrate S2765. The inhibitory activity of FXa⅐TFPI was assessed in two different ways. The first method was by preincubating TF⅐PSPC or TF⅐PC with FVIIa (1 nM) for 5 min, then adding FX (100 nM) in the presence of various concentrations of FXa⅐TFPI and subsequently assaying the timed samples taken from the vessel for FXa activity. Under these conditions, fluid phase inhibitor complex is assumed to be in equilibrium with surface-bound FXa⅐TFPI. The second method was by preincubating TF⅐PSPC or TF⅐PC for 10 min with 0.4 nM FXa⅐TFPI, in the absence of FVIIa, and then depleting the solution of inhibitor complex by flushing for various time intervals with TBSA-CaCl 2 buffer. Subsequently, FX activation was started by adding FX (100 nM) and FVIIa (1 nM). Timed samples were taken and assayed for FXa activity as described above.

Data Analysis
FVIIa Concentration-dependent FX Activation-As the catalytic efficiency of FVIIa in complex with TF is several orders of magnitude higher than that of free FVIIa, the observed FX generation rate, V obs , reflects the formation of the binary TF⅐FVIIa complex and is described by the formula shown in Equation 1, which relates V obs to V max the conversion rate at saturating FX concentrations, [FX] the FX concentration, and K m(app) the apparent Michaelis constant. The parameters V max and K m(app) were determined by least squares fitting of Equation 2 to the measurements of V obs . It should be noted that the substrate transport from the bulk solution to the disc surface depends on the concentration gradient between bulk and the solution, immediately adjacent to the catalytic surface. To what extent this substrate depletion near the surface causes an underestimation of the true K m value depends on the ratio of substrate conversion over the transport-limited rate of substrate supply (33).

Inhibition of TF⅐FVIIa-mediated FX Activation by Preformed FXa⅐TFPI Complexes-
The decrease in FX activation rate reflects the formation of quaternary TF⅐FVIIa⅐TFPI⅐FXa complexes and the corresponding decline of TF⅐FVIIa activity, which for any fixed concentration of inhibitor, can be described as a pseudo-first order process (22,23). Therefore, the time-dependent FXa generation indicating a mono-exponential decay of TF⅐FVIIa activity can be expressed as shown in Equation 3, with [FXa] t the observed FXa generation at time t, V 0 the rate of FXa generation in the absence of inhibitor, k the pseudo-first order rate constant of inhibition, and t the reaction time. In principle, the rate constants V 0 and k can be determined by a least squares fit of Equation 3 to the FXa generation curve. This approach, however, is subject to the drawback that, because of the rapid decline of the FXa generation rate, the estimates of V 0 and k become increasingly correlated and inaccurate for increasing concentrations of the inhibitory complex. To avoid this source of error, we employed the excellent reproducibility of repeated measurements of FXa generation on the same TF⅐PSPC/TF⅐PC disc after regeneration by an EDTA wash (see "Results"). Each measurement of FX activation in the presence of FXa⅐TFPI complex was preceded by a measurement of the corresponding FXa generation in the absence of inhibitor complex. The rate constants V 0 and k were estimated by a simultaneous least squares fit of Equation 3 to the experimental data, with an individual parameter k for the uninhibited and the inhibited FX activation and a shared parameter V 0 for the initial rate of FXa generation.

Initial Rate of FX Activation at the Surface of a Rotating Disc with TF Embedded in a Phospholipid
Bilayer-Typical results of FX activation experiments using a rotating disc are shown in Fig. 1. The disc with TF embedded in a planar phospholipid bilayer composed of PSPC or PC was incubated first with FVIIa (1 nM) during 5 min before FX (100 nM) was added to start the reaction. Apparently, for both TF⅐PC and TF⅐PSPC, the FXa activity increases linearly as a function of time. However, the inset to Fig. 1 reveals that, in contrast to the TF⅐PC surface, it takes for the TF⅐PSPC surface about 1 min before FXa activity starts to increase linearly. This delayed rise in solution FXa activity most likely reflects binding of FXa to the TF⅐PSPC surface (35). Therefore, the initial rate of FX activation is defined as the linear increase in solution phase FXa 1.5 min after the start of the reaction. When the rotation of the disc was stopped FXa generation halted indicating that under the conditions of the experiment convective transport is essential both for delivery of FX to the catalytic surface and for the transfer of FXa from the surface to the solution. No FXa was produced in the absence of TF or FVIIa.
The mean value Ϯ S.D. of the initial rate of FXa generation at TF⅐PSPC after repeated regeneration (n ϭ 5) of the same disc (see "Experimental Procedures") was 2.99 Ϯ 0.12 nM/min. For TF embedded in PC, the initial rate of FXa generation was 0.82 Ϯ 0.07 nM/min (mean value Ϯ S.D., n ϭ 5). A single TF⅐PSPC or TF⅐PC disc could thus be regenerated and used several times, allowing repeating experiments under different conditions with an unchanged catalytic active surface. The inter-disc variations were larger with mean values for the initial rate of FX activation Ϯ S.D. of 3.24 Ϯ 0.70 nM/min (n ϭ 30) and 0.77 Ϯ 0.37 nM/min (n ϭ 30) for TF⅐PSPC and TF⅐PC, respectively. It should be noted that the TF⅐PC membrane preparation contained 5-fold more TF than the TF⅐PSPC membrane.
Dependence of the Rate of FX Activation on the Angular Velocity of the Rotating Disc- Fig. 2 shows the relation between the angular velocity of the rotating disc and the initial rate of FXa formation. For a rotating disc with TF⅐PSPC, the initial rate of FXa formation increases when the angular velocity increases from 0 to 60 rad/s. An additional increase of the angular velocity does not result in a concomitant increase of the initial rate. Thus, an angular velocity of more than 60 rad/s appears to be sufficient to approach a kinetically controlled situation. That is, the rate at which FX is activated at the surface is smaller than the transport limited rate of FX delivery to the catalytic surface. A similar experiment was also performed with TF embedded in PC (Fig. 2). Obviously, an apparent kinetically controlled FX activation is achieved at a lower angular velocity (30 rad/s) than observed with TF⅐PSPC.  2. Dependence of the initial velocity of FX activation on the angular velocity of the rotating discs. At the indicated angular velocities, the initial rate of FX activation was measured on TF⅐PSPC (q) and TF⅐PC (OE) as described in the legend to Fig. 1. A single rotating disc was used that was regenerated for each measurement by flushing (10 ml/min) the reaction vessel and spinning disc for 5 min with TBSA-EDTA and subsequently for 5 min with TBSA-CaCl 2 buffer.
The 5-fold lower rate of FX activation at the TF⅐PC surface likely reduces the transport-limited supply of reactants. All further experiments were performed at an angular velocity of 63 rad/s.
Kinetic Parameters for FX Activation at the Rotating Disc-Based on a previously reported K d value of 41 pM for FVIIa binding to TF in PSPC vesicles (36), we reckoned that a FVIIa concentration of 1 nM, as used in the experiments described above, was sufficient to saturate all the TF exposed at the catalytic surface. To verify whether this assumption applied for the rotating disc system, we measured the initial rate of FX activation as a function of the solution phase FVIIa concentration. The titration curves obtained (data not shown) were analyzed using Equation 1 to yield binding as well as kinetic parameters. For TF⅐PSPC the fit yielded a K d of 18 pM, and for TF⅐PC the K d value was 240 pM. Thus, whereas the K d of FVIIa for TF⅐PSPC is in good agreement with the value previously reported (36,37), the affinity of FVIIa for TF⅐PC is 10-fold higher than values reported by others (32,36). Hence, a FVIIa concentration of 1 nM is indeed sufficient to saturate more than 80% of the TF present at the PC and PSPC surfaces. Furthermore, the fit yielded a V max of 4.4 nM/min (TF⅐PSPC) and 0.85 nM/min (TF⅐PC), corresponding to a turnover number of 1677/ min for TF⅐PSPC and of 65/min for TF⅐PC.
Likewise, the relationship between the initial rate of FX activation at the spinning surface and the fluid-phase FX concentration was determined for TF⅐PSPC and TF⅐PC. By fitting Equation 2 to the experimental data (not shown), we estimated values of 110 nM for the apparent K m (K m(app) ) and of 5.5 nM/min for V max . Calculation on basis of an estimated TF density of the PSPC surface (2.5 fmol/cm 2 ) gave a k cat value of 2098/min for TF⅐PSPC. For TF⅐PC a linear relation between the FXa generation rate and the FX concentration was found despite a much higher (up to 400 nM) FX concentration employed (data not shown). This indicated for the TF⅐PC a K m(app) value of Ͼ 5 M, thus at least 1 order of magnitude greater than the highest FX concentration used.
Inhibition of Membrane-associated TF⅐FVIIa as a Function of the Concentration of FXa⅐TFPI-In a successive set of experiments, we assessed the inhibitory activity of preformed FXa⅐TFPI FL and FXa⅐TFPI 1-161 complexes toward TF⅐FVIIacatalyzed FX activation at TF⅐PC and TF⅐PSPC membranes deposited on rotating discs. Following a preincubation with FVIIa for 5 min, the reaction was started by the simultaneous addition of FX (100 nM) and inhibitor complexes in varying concentrations. All inhibition experiments were preceded by a control experiment in the absence of inhibitor to assess the precise catalytic activity of that particular rotating disc. To correct for inter-disc variations in catalytic activity (Ϯ 20%, as shown before), the FXa generated in the presence of inhibitor is expressed as a percentage of the corresponding amount of FXa generated after 10 min in the absence of inhibitor.
It is seen from Fig. 3A that the rate of FXa generation on TF⅐FVIIa⅐PC in the presence of FXa⅐TFPI 1-161 gradually decreases in time and that this progressive inhibition becomes more pronounced at higher inhibitor concentrations. The very same observations were made with FXa⅐TFPI FL as inhibitor (Fig. 3B). The FXa generation curves were analyzed according to a model that assumes that complex formation between TF⅐FVIIa and FXa⅐TFPI results in a quaternary complex in which the catalytic activity of TF⅐FVIIa is completely blocked (Equation 3, see "Experimental Procedures"). The hypothesis that the initial FXa generation rate did not alter by addition of the FXa⅐TFPI complex was checked by fitting Equation 3 to the individual curves of the pairs of uninhibited and inhibited FXa generation. The value obtained for V 0 of the inhibited curve, expressed as percentage of V 0 of the uninhibited curve, was 96 Ϯ 6% and 105 Ϯ 9% (mean Ϯ S.D.; n ϭ 6) for FXa⅐TFPI  and FXa⅐TFPI FL complexes, respectively. These data thus justify our analysis of the FXa generation curves by a simultaneous fit of Equation 3 to the uninhibited and inhibited curves with a shared rate constant V 0 . The solid lines in Fig. 3 (A and  B) show the result of this fitting procedure, which yielded pseudo-first order rate constants of inhibition (k) as a function of the concentration of the inhibitory complex (Fig. 3C). It is clear that k increases linearly with the concentration of the inhibitory complex. This indicates that, in the concentration range employed, the bimolecular association of TF⅐FVIIa with preformed FXa⅐TFPI is the rate-limiting step of the formation of the quaternary TF⅐FVIIa⅐FXa⅐TFPI complex. The second order rate constants of inhibition (Table I) (Fig. 4B), a significant FXa generation is still observed. The FXa generation curves as shown in Fig. 4 were analyzed as described for those in Fig. 3. The pseudo-first order rate constants of inhibition thus obtained are presented in Fig. 4C as a function of the concentration of the inhibitory complex. This plot clearly reveals the difference in inhibitory activity of FXa⅐TFPI 1-161 and FXa⅐TFPI FL on TF⅐PSPC. According to Table I, the second order rate constant of inhibition for FXa⅐TFPI 1-161 and FXa⅐TFPI FL increases when PS is present in the TF-bearing membrane. However, the increase for FXa⅐TFPI 1-161 is about 9-fold, whereas a 3-fold increase is seen for FXa⅐TFPI FL . The rate constant of 0.94 ϫ 10 8 M Ϫ1 ⅐s Ϫ1 found for FXa⅐TFPI 1-161 is in good agreement with the previous reported value of 1.1 ϫ 10 8 M Ϫ1 ⅐s Ϫ1 (23), whereas the value for FXa⅐TFPI FL (0.20 ϫ 10 8 M Ϫ1 ⅐s Ϫ1 ) is 5-10 fold lower than the values reported previously (22,23). Thus at TF⅐PSPC, FXa⅐TFPI FL apparently has a lower inhibitory activity than FXa⅐TFPI 1-161 . In view of our earlier finding that FXa⅐TFPI FL complexes have a considerably higher affinity for PS containing lipid membranes than complexes of FXa⅐TFPI 1-161 , these data suggest that upon binding to TF⅐PSPC, FXa⅐TFPI FL complexes lost their inhibitory action.
Inhibition of TF⅐FVIIa-catalyzed FX Activation by Membrane-bound FXa⅐TFPI Complex-FXa⅐TFPI FL and FXa⅐ TFPI 1-161 complexes (0.4 nM) were incubated for 10 min with TF⅐PSPC-coated discs spinning at 63 rad/s. In some cases FXa⅐TFPI complexes were, after the incubation, removed from the solution by flushing the reaction vessel with a TBSA-CaCl 2 buffer during 5 min at 10 ml/min. FXa generation was initiated by the simultaneous addition of FX (100 nM) and FVIIa (1 nM). Fig. 5A shows the time courses of FXa generation for TF⅐PSPC surfaces that were preincubated with FXa⅐TFPI 1-161 with inhibitor complexes remaining in solution during the FXa generation and with inhibitor complexes removed from solution at the end of the preincubation. For comparison we also show the FXa generation in the absence of inhibitory complexes and the FXa generation as observed when the inhibitory complexes were added simultaneously with FX. It is apparent from Fig.  5A that, as long as the FXa⅐TFPI 161 is not depleted from the solution, identical inhibition is seen with or without preincubation, whereas depletion of the inhibitory complex from the solution completely removes the inhibitory activity.
Similar experiments performed with FXa⅐TFPI FL complexes gave a completely different outcome, as shown in Fig. 5B. The largest extent of inhibition of FXa generation is observed when the TF⅐PSPC membrane was preincubated with FXa⅐TFPI FL complexes. Even more strikingly, it turned out that depletion of the solution from FXa⅐TFPI FL complex by flushing the reaction vessel at the end of the preincubation did not diminish the inhibitory activity. This indicates that under the conditions of this experiment, the contribution of TF⅐PSPC-bound FXa⅐TFPI FL complexes to the inhibitory activity is dominant. To ascertain that the inhibitory activity that remained manifest after flushing originated from FXa⅐TFPI FL bound to the TF⅐PSPC surface and not from inhibitory complex nonspecifically bound to the wall of the reaction vessel, we performed the following control experiment. The reaction vessel was preincubated for 10 min with 0.4 nM FXa⅐TFPI FL and flushed for 5 min with TBSA-CaCl 2 buffer, after which an unused TF⅐PSPCcoated disc was placed in the vessel. Subsequently, the rate of FXa formation at that disc was measured following the addition of FVIIa (1 nM) and FX (100 nM). No inhibition was observed, showing that the retained inhibitory activity in the previous experiment solely originated from TF⅐PSPC-bound FXa⅐TFPI FL complexes.
The different effects of flushing on the inhibitory activity (Fig. 5, A and B) of the FXa⅐TFPI 1-161 and FXa⅐TFPI FL complexes at TF⅐PSPC presumably reflect the differences in desorption rates of these complexes that were previously reported (24). As has been shown, at a macroscopic PSPC surface (no TF  present) the residence time of membrane-bound FXa⅐TFPI FL was about 45 min, compared with a residence time of about 1-2 min for complexes of FXa with TFPI 1-161 . When a TF⅐PC surface was preincubated with FXa⅐TFPI FL (0.4 nM) and subsequently washed for 5 min as described above, an uninhibited FXa generation was observed (data not shown). Thus, omitting the negatively charged phospholipids from the membrane resulted in loss of FXa⅐TFPI FL membrane binding and thus of the exerted inhibitory activity. Additional support was obtained when, following the preincubation of the TF⅐PSPC surface with FXa⅐TFPI FL , the time period of flushing with buffer was increased (from 5 to 90 min). This resulted in a gradually decreasing inhibitory activity (Fig. 6). It is noteworthy, however, that even after a rinse for 90 min, a significant inhibition of FXa generation is still observed.
Further analysis of inhibited and uninhibited FXa generation curves, as presented in Figs. 5B and 6, reveals that preincubation of TF-PSPC with FXa⅐TFPI FL results in a major decrease of the initial rate of FX activation. Correspondingly, the simultaneous fit of Equation 3 to the inhibited and the matching uninhibited FXa generation curves with a shared rate constant V 0 , as used in the fits of Figs. 3 and 4, resulted in an inadequate fit with large systematic deviations. Using individualized parameter values of V 0 , however, resulted in excellent fits (solid lines in Figs. 5B and 6). The estimated parameters V 0 and k, listed in Table II, show that the initial rate of FXa generation of the preincubated curves is instantaneously decreased to about 25% of the uninhibited V 0 when FXa generation is initiated shortly (0 -5 min) after the preincubation. With increasing flushing time, a gradual increase of the estimated V 0 is observed, from 27% after 5 min flushing to 59% after 90 min of flushing. The rate constants found for the exponential decay of the TF⅐FVIIa activity ranged from 0.08 to 0.15 min Ϫ1 , dependent on the flushing time. These values are 3-5-fold lower than the decay rate of 0.42 min Ϫ1 , found for the inhibition of TF⅐FVIIa when the same concentration of FXa⅐TFPI FL was added to the reaction mixture (Fig. 4). DISCUSSION Rationale behind Experimental Design-In this study we addressed the question of whether the previously reported (24) high affinity of FXa⅐TFPI FL for negatively charged phospholipids is reflected in its inhibitory activity against TF⅐FVIIa. Because phospholipids were also shown to enhance complex formation between FXa and TFPI (16,23), preformed FXa⅐TFPI FL and FXa⅐TFPI 1-161 complexes were used in this study to simplify the interpretation of the experimental data on the effect of phospholipids on the TF⅐FVIIa inhibition by FXa⅐TFPI. Furthermore, a macroscopic phospholipid membrane (with embedded TF) was used rather than unilamellar vesicles because this model mimics the plasma membrane of TF-bearing cells better than small unilamellar vesicles. Moreover, several studies have shown that the kinetics of activation and inactivation of blood coagulation enzyme complexes are dependent on the characteristics of the phospholipid surface like the radius of the phospholipid vesicle (35,38), microscopic homogeneity (39), and ratio of reactant-bearing vesicles over non-bearing vesicles (25). Additionally, with macroscopic surfaces it is easier to separate physically the surface-bound and fluid-phase reactions. The use of a macroscopic phospholipid membrane at the surface of a rotating disc (40) has been extensively characterized for immobilized enzymes (33,41,42). Its main advantage over flow systems like that of a capillary flow reactor (43,44) is the uniformly accessibility of the surface (40).
Determination of the Kinetics of FX Activation Using the Rotating Disc System-Our study confirms the role of anionic phospholipids in the binding of FVIIa to TF embedded in a phospholipid bilayer: the K d of FVIIa for TF⅐PC (240 pM) is 13-fold higher than the K d of FVIIa for TF⅐PSPC (18 pM).
The k cat (35 s Ϫ1 ) and K m(app) (110 nM) values for TF⅐FVIIacatalyzed FX activation at PSPC are in reasonable agreement with the k cat ϭ 12 s Ϫ1 and K m ϭ 70 nM measured at TF-bearing PSPC vesicles (23,45). However, others have reported much higher values (46). A trivial explanation for this apparent discrepancy could be the high vesicle concentration in the latter study. The estimated occupancy of the TF⅐FVIIa by FX, as determined from K m values in PSPC and PC is 47% and less than 2%, respectively, suggesting that some competition between substrate and inhibitor for TF⅐FVIIa might be expected at PSPC but not at PC.
Inhibition Kinetics of TF⅐FVIIa-catalyzed FX Activation as Studied at a Rotating Disc Surface-The similar kinetics of TF⅐FVIIa inhibition that were observed for FXa⅐TFPI FL and FXa⅐TFPI 1-161 on a TF⅐PC surface, which does not bind FXa (47) nor TFPI FL (48), are consistent with other studies which indicated that a direct interaction of FXa⅐TFPI with TF⅐FVIIa does not require the binding of FXa⅐TFPI to negatively charged phospholipids (22,43,49). Kazama (50), however, concluded from studies using TF lacking the trans-membrane and cytoplasmic domains that binding of FXa to anionic phospholipids via its Gla domain may be an absolute requirement for TFPImediated regulation of full-length membrane-inserted TF.
Interestingly, the addition of PS (25%) to a neutral membrane brings about a significant stimulation of the rate of inhibition by both FXa⅐TFPI FL (k ϭ 0.20 ϫ 10 8 M Ϫ1 ⅐s Ϫ1 ) and FXa⅐TFP 1-161 (k ϭ 0.94 ϫ 10 8 M Ϫ1 ⅐s Ϫ1 ). We note that the marked difference in fractional saturation of TF⅐FVIIa on a PC membrane and that of TF⅐FVIIa on a PSPC membrane as aforementioned could mask the stimulatory effect of PS. When FX and FXa⅐TFPI compete for TF⅐FVIIa at PSPC, then it can be calculated according to the relation k true ϭ k obs (1 ϩ S/K m ) that the true rate constants of inhibition (FXa⅐TFPI 1-161 : 1.79 ϫ 10 8 M Ϫ1 s Ϫ1 , FXa⅐TFPI FL : 0.38 ϫ 10 8 M Ϫ1 s Ϫ1 ) are about 2-fold higher then the observed rate constants (Table I). Because for TF⅐PC the K m is much larger than the substrate concentration [FX], competition is in this case negligible. As a result, the differences in rate constants of inhibition at a membrane with PSPC and pure PC become even more pronounced.
The stimulatory effect of PS on the inhibition of both FXa⅐TFPI 1-161 and FXa⅐TFPI FL might be related to the FXamediated binding of the complexes to the phospholipid membrane (24). Several mechanisms have been proposed to explain the membrane-mediated acceleration of reactions of the blood coagulation system. Next to a proper juxtaposition of the FXa⅐TFPI complexes toward the active site of the FVIIa, anionic planar phospholipid membranes may also allow lateral diffusion of the substrate toward the membrane associated enzymatic complex as shown in experiments on prothrombin activation (33,34,38). As a result, a much lower solution phase substrate concentration is needed to saturate half of the immobilized enzyme. If this mechanism would be operational for the inhibition of TF embedded in the PSPC membrane, one would predict the highest rate constant of inhibition for the inhibitory complex with the highest affinity for the membrane. Our experiments, however, show the opposite; the presence of PS stimulated FXa⅐TFPI 1-161 9-fold compared with the only 3-fold stimulation of FXa⅐TFPI FL . Apparently the high affinity binding of FXa⅐TFPI FL with PSPC membranes is accompanied by a lower inhibitory activity. These data thus suggest that solution phase and membrane-bound FXa⅐TFPI FL contribute to the overall inhibition reaction but to a different extent. Therefore, the description of the inhibition reaction by a bimolecular association reaction (Equation 3), although it adequately describes the experiment of Figs. 3 and 4, presents a gross simplification. Indeed, the experiments in Figs. 5 and 6 demonstrate that inhibition of TF⅐FVIIa at PSPC by preadsorbed FXa⅐TFPI FL is in fact a multistep process. After a rapid inactivation of Ϯ 75% of the TF⅐FVIIa activity, as suggested by the reduced initial rate of FXa generation after the addition of FVIIa and FX, a much slower elimination of the remaining activity is observed. This rapid first phase suggests that the membrane-bound pool of FXa⅐TFPI FL may, also in the absence of FVIIa, associate with TF (51) and that this ternary complex upon addition of FVIIa and FX rapidly converts to the quaternary complex. Alternatively, FXa⅐TFPI FL binding to TF could interfere with the interaction of TF with FVIIa. However, the same study on TF⅐PC showed no persistent inhibitory activity. This indicates that the high affinity binding of FXa⅐TFPI FL to TF⅐PSPC membranes is the result of a protein (FXa⅐TFPI)-lipid interaction rather than a protein (FXa⅐TFPI)-protein (TF) interaction. The decreasing extent of the initial inhibition with increasing flushing time as shown in Table II, thus probably reflects the decrease of the membrane-bound pool of FXa⅐TFPI FL .
Taken together, our data indicate efficient inhibition both by fluid phase FXa⅐TFPI FL and by phospholipid-bound complex. The complex between full-length TFPI and FXa is so tightly bound to the catalytic TF⅐PSPC surface that it represents a highly effective pool of inhibitory activity that remains available for hours after exposure of the membrane to FXa⅐TFPI FL complexes. This property may be of physiological importance in allowing an on-site regulation of the TF⅐FVIIa activity. That is, this phospholipid-bound pool of inhibitor complex, which is immediately available, may represent an effective control of newly formed catalytic units when newly synthesized TF arrives at the plasma membrane of TF producing cells. , as determined from TF⅐FVIIa-mediated FX activation initiated by simultaneous addition of FVIIa and FX, after preincubation of the TF:PSPC surface during 10 min with FXa⅐TFPI FL complexes (0.4 nM) followed by no flushing or a flush of the reaction vessel with Ca 2ϩ -containing buffer (10 ml/min) during the indicated period. V 0 is presented as percentage of the initial rate of FXa generation of the paired uninhibited curve obtained at the same disc.