Soluble Phosphatidylserine Triggers Assembly in Solution of a Prothrombin-activating Complex in the Absence of a Membrane Surface*

Factor Xa (FXa) binding to factor Va (FVa) on platelet-derived membranes containing surface-exposed phosphatidylserine (PS) forms the “prothrombinase complex” that is essential for efficient thrombin generation during blood coagulation. There are two naturally occurring isoforms of FVa, FVa1 and FVa2. These two isoforms differ by a 3-kDa polysaccharide chain (at Asn2181 in human FVa1 (Kim, S. W., Ortel, T. L., Quinn-Allen, M. A., Yoo, L., Worfolk, L., Zhai, X., Lentz, B. R., and Kane, W. H. (1999)Biochemistry 38, 11448–11454)) and have different coagulant activities. We examined the interaction of the two bovine isoforms with active site-labeled FXa, finding no significant difference. A soluble form of PS (C6PS) bound to FVa1 and FVa2 with comparable affinities (K d = 11–12 μm) and changes in FVa intrinsic fluorescence. At concentrations well below its critical micelle concentration, C6PS binding to bovine FVa2 enhanced its affinity for FXa in solution by nearly 3 orders of magnitude (K d eff = 40–2 nm over a C6PS range of 30–400 μm) but had no effect on the affinity of FVa1for FXa (K d = 1 μm). This results in a soluble complex between FXa and FVa2, whose expected molecular weight was confirmed by calibrated native gel electrophoresis. This complex behaved as a normal Michaelis-Menten enzyme in its ability to produce thrombin from meizothrombin (apparentk cat/K m ≅ 109 m −1 s−1). The ability of soluble PS to trigger formation of a soluble prothrombinase complex suggests that exposure of PS molecules during platelet activation is likely the key event responsible for the assembly of an active membrane-bound complex.

The final step in the blood coagulation cascade involves the activation of prothrombin to thrombin, which is the central enzyme of the coagulation system. This activation requires assembly of an enzyme complex, called prothrombinase (1), which consists of blood coagulation factors X a (a serine protease) and V a (a cofactor), Ca 2ϩ , and membranous vesicles derived from stimulated platelets (2). Several studies (3)(4)(5) have suggested that phosphatidylserine (PS) 1 might play a specific role in prothrombin activation. PS is asymmetrically distributed to the cytoplasmic surface of resting platelet membranes (6) but is exposed when human platelets are activated (7). It has become clear only very recently that PS regulates the structure and function of factors X a and V a (8,10). 2 Here we explore further the extent of this regulation.
Factor V exists in plasma as an inactive, single chain glycoprotein with a molecular mass of 330 kDa. The active form of factor V, FV a , has a central domain removed to yield a heterodimer composed of two chains, a heavy chain (M r ϭ 94,000 in the bovine species; 105,000 in human) and a heterogeneous light chain (M r ϭ 74,000 in FV a1 or 71,000 in FV a2 ). The heavy and light chains form a tight complex in the presence of a calcium ion (11). The heterogeneity in the light chain is seen in both the bovine and human molecules. In the human form, it appears to arise from glycosylation of Asn 2181 at the C-terminal end of the light chain (12). Prothrombinase complexes assembled from the two molecular species derived from human plasma are observed to have somewhat different cofactor activities (13,14). This has been attributed to substantially different affinities for binding to membranes (13). However, our lab has reported that the two forms of both bovine and human FV a bind to membranes with only ϳ3-fold different affinities (12,14). This suggests that differences in the ability to support prothrombinase activity must reflect either different binding between factors FX a and FV a1 versus FV a2 or different intrinsic activities of the FX a ⅐FV a1 and FX a ⅐FV a2 complexes. Although our results have favored the former possibility (14), it has been difficult to prove this unambiguously because it is difficult to measure precisely the interaction between FX a and FV a on a membrane surface (15).
The presence of FV a in a reaction mixture is critical to obtaining a maximal and physiologically significant rate of prothrombin activation (1). As a result, intense effort has been devoted to discerning the role of this cofactor in thrombin generation. The picture that emerges is that factor V a binds tightly to acidic lipid membranes (K d Ϸ2 nM) (16) and then tightly binds factor X a already bound to that surface (K d Ϸ1 nM) (15), making it effectively an anchor to assemble the prothrombinase at the low concentrations of factor X a and membranes expected to be found in vivo (17). Because the interaction between factors V a and X a in solution is much weaker (K d Ϸ1-3 M) (18) than that reported on a membrane (15), it has been speculated that either the membrane alters these two proteins so as to enhance their interaction or the presence of both proteins in the reduced dimensionality of the membrane surface leads to tighter association than would be seen in three dimensions (15). Again, the difficulty of assessing interactions between proteins on a membrane surface makes it hard to distinguish between these two possibilities.
We have shown previously that a soluble form of PS, C6PS, binds to and FX a (8). Binding is accompanied by conformational changes in FX a that alter protein function (8). 2 Four C6PS molecules also bind to bovine FV a to induce conformational and functional changes (10). We present here compelling evidence that PS binding to these two proteins regulates their assembly into an active complex and thus regulates the functioning of the prothrombinase complex. In doing so, we answer five questions. First, do the two isoforms of factor V a (F Va1 or F Va2 ) form complexes of different affinities with FX a ? Second, does soluble C6PS bind with similar affinities to these two isoforms? Third, does C6PS binding alter the affinities of the complexes between the two FV a isoforms and factor FX a ? Fourth, can differential binding of the two FV a isoforms to FX a account for the apparently different cofactor activities of these two isoforms? Fifth, if C6PS induces tight association between FV a and FX a in solution, is this sufficient to trigger formation of a prothrombinase complex in solution? If so, our ability to study the structure and function of this critical enzyme complex would be dramatically enhanced.

Methods
Phospholipid and Protein Preparation-C6PS stock solutions, bovine FV a , FV a1 , and FV a2 were prepared as reported previously (8,14). Prothrombin, FX, and FXa were purified from bovine plasma (19). The final purity of the proteins was ascertained by SDS-PAGE to be greater than 90%. DEGR-X a was prepared by sequential addition of 5-l aliquots of DEGR-CK (1 mg/ml in 0.02 M Tris, 0.1 M NaCl, pH 7.5) to 1 ml of about 17 M purified factor X a until complete loss of enzymatic activity, as monitored by the S-2765 assay (20). DEGR-X a was then dialyzed against 50 mM Tris, 0.1 M NaCl, pH 7.5, to remove free reagent (21). DEGR-X a was analyzed by SDS-PAGE and visualized under UV light. Bovine MzII a was generated with E. carinatus venom linked to Affi-Gel active ester-agarose (Bio-Rad) as described previously (22).
Fluorescence Titration of DEGR-X a and FV a1 and FV a2 -Changes in the fluorescence intensity of DEGR-X a in response to FV a addition were measured at 37°C using a SLM 48000 TM spectrofluorometer (Spectronic Instruments, Inc., Rochester, NY). Slits (8 ϫ 8 and 4 ϫ 4 nm for excitation and emission) were closed between measurements to avoid photo-degradation of the sample. Buffer solutions were filtered using 0.2-m filters (Nalge Co., Rochester, NY). A stirred microcuvette (Hellma Cells, Jamaica, NY) was initially charged with a solution of 1 nM DEGR-X a and 5/50 nM FV a1 /FV a2 and then rinsed with buffer. This procedure prevented the formation of aggregates on the cuvette walls. To 0.95 ml of DEGR-X a solution (1 nM DEGR-X a in 30, 60, and 400 M C6PS, 50 mM Tris, 0.1 M NaCl, 5 mM Ca 2ϩ , pH 7.5), FV a1 or FV a2 were added from stock solutions with 4 min of equilibration, and fluorescence intensity was recorded using an excitation wavelength of 340 nm and an emission wavelength of 550 nm. Several intensity readings were averaged after each addition, and corrected, via controls, for dilution, buffer background, and any small amount of photobleaching.
Changes in FV a1 /FV a2 intrinsic fluorescence in response to C6PS was measured similarly, except that samples were excited at 295 nm (slits: 8 nm), and emission was recorded at 345 nm (slits: 8 nm).
Fluorescence Stopped-flow Measurements-Rates of thrombin formation from MzII a or of thrombin plus MzII a formation from prothrombin at 37°C were estimated from time-dependent changes in the fluorescence of DAPA, an active site inhibitor of the activation products (1). Stopped-flow measurements were performed on the SLM 48000 TM spectrofluorometer with 280 nm excitation (4-nm slits) and a 515-nm cut off filter and 8-nm slits in the emission path. Reactions were initiated by rapidly mixing equal volumes (200 l) of the contents of the two driving syringes of a SLM-Aminco Milliflow TM stopped-flow reactor. Syringe A contained substrate solution and DAPA in 50 mM Tris, 150 mM NaCl, 5 mM CaCl 2 , pH 7.5, and syringe B contained a mixture of factors X a and V a and C6PS pre-equilibrated 37°C for 2-3 min in the same buffer. The final concentration of FX a in the reaction chamber was always maintained as 1 nM, and the concentrations of FV a and C6PS were varied. The substrate/DAPA ratio was 1:5. The initial fluorescence intensity (F 0 ) was obtained from mixing experiments with substrate and DAPA in syringe A and only lipid in syringe B. All fluorescence intensities were corrected for background light scattering. Fluorescence intensity at the completion of the reaction (F ϱ ) reflected complete conversion of a given substrate to thrombin. The initial rate of thrombin generation was then determined from the initial rate of fluorescence change (5-10% completion) normalized to the intensity at completion times the concentration of substrate in the reaction mixture (23).
Native Gel Electrophoresis-Polyacrylamide gels were prepared at five different percentages of cross-linking (5, 6, 6.5, 7, and 8% total acrylamide with bisacrylamide/acrylamide being 1:29 in all cases). A mixture of 1 nM factor X a and 1 nM factor V a was incubated at 37°C for 2 min in the presence and absence of 400 M C6PS, and these two samples were run together with known marker proteins (270 to 14.2 kDa; Sigma) on these five gels in a Bio-Rad Mini-Protean II TM minigel apparatus (Bio-Rad) and stained with colloidal Coomassie Blue (24). Factor X a and factor V a , at these same concentrations and in the presence of 400 M C6PS, were run separately as controls. R f values in each gel were measured relative to tracking dye. For each protein, a plot of log(R f ϫ 100) against the percent gel concentration gave a straight line, the negative slope of which is the retardation coefficient (25). A log-log plot of the retardation coefficient against the molecular weights of the marker proteins produced a linear curve from which the molecular weight of the FV a ⅐FX a complex was determined (26).

Data Analysis
FX a -FV a Interaction-There were two approaches to data analysis. The first recognized that FX a and FV a bind two and four C6PS molecules, respectively, and that their interaction with each other will depend on how many sites are occupied in each protein. This approach, described under the "Appendix" in its simplest form, requires knowledge of the mechanisms of C6PS binding to the two proteins as well as adjustment of the values of some unknown parameters, but provides physical insight into our results if one is willing to accept some uncertainty in parameter values. Alternatively, one can simply acknowledge that the interaction between DEGR-X a and FV a2 depends on C6PS but make no assumptions about the mechanistic details of this dependence. In this instance, several equilibria are replaced by a single effective DEGR-X a ϩ V a binding equilibrium with an apparent K d eff that depends on C6PS concentration (Equation 1), The total concentration of DEGR-X a ⅐V a complex at any given C6PS concentration ([DEGR-X a ⅐V a ] (C6PS) ) is then given by the familiar expression for two-species binding (Equation 2), where the concentration terms all are "total" concentrations of each species in a reaction mixture. The change in fluorescence signal of DEGR-X a was taken as proportional to the fraction of DEGR-X a bound to FV a (Equation 3) where the fluorescence parameter F 0 was fixed as the DEGR-X a fluorescence before titration began and F SAT is the fluorescence at saturation, which was obtained along with K d eff by regression of the data to Equations 2 and 3.
The interaction between FX a and FV a1 or FV a2 was also detected by an increase in the rate of prothrombin activation. In this case, the same analysis was used except that the increase in initial rate of activation (V) and [X a ] replaced F and [DEGR-X a ] in Equation 3.
The kinetics of MzII a activation were parameterized according to the normal Michaelis-Menten model (Equation 4), where the enzyme concentration ([X a ⅐V a ] (C6PS) ) was calculated using the K d eff values and Equation 2 described above. In all cases, regression of data sets to these models was performed using SigmaPlot 2000 (SPSS Inc., Chicago, IL).

Effect of C6PS on Prothrombin Activation by FX a in the
Presence of FV a1 /FV a2 -Initial rates of FX a -catalyzed prothrombin activation (see "Methods") were determined as a function of added soluble C6PS concentration (Fig. 1). In the presence of 5 nM FV a2 , the initial rate saturated at about 100 M C6PS (squares), but for 50 nM FV a1 (higher concentration needed to record rates comparable with those seen with FV a2 ) the initial rate increased rapidly up to a value larger than seen for FV a2 by 100 M C6PS and then slowly increased well beyond this rate up to and beyond 400 M PS (triangles). For 5 nM FV a (the natural mixture of these two isoforms), the initial rate saturated at about 100 M C6PS (circles) at about 2/3 the rate seen with FV a2 . Because FV a is roughly 2/3 FV a2 , it seems that the cofactor activity of the natural mixture is dominated under our conditions by a complex containing the lighter FV a2 isoform.
It is important to know whether the effects recorded in Fig.   1 are due to individual molecules of C6PS or to an aggregated form of this moderately soluble amphipath. We have characterized previously (8) the formation of C6PS micelles and have reported the CMC at 5 mM Ca 2ϩ to be 950 M in a similar buffer. However, both FX a 2 and FV a (10) lower the CMC of C6PS to Ͼ800 M and 800 -900 M, respectively. Thus, we determined by QELS the mean size of any C6PS aggregate or micelle that might form under our experimental conditions (i.e. in the presence of 1 nM FX a and 5 nM FV a ) at increasing C6PS concentrations ( Fig. 1, inset). No aggregates could be detected below ϳ650 M C6PS, which is lower than the CMC in the presence of either FX a or FV a individually but also greater than the highest lipid concentration we have used in our experiments.
Response of FV a1 and FV a2 Intrinsic Fluorescence to C6PS-C6PS and other short chain phospholipids bind to bovine FV a to induce conformational changes detected by intrinsic fluorescence changes and CD spectral changes (10). Fig. 2 shows the effect of increasing C6PS concentration on the tryptophan region of FV a1 (circles) or FV a2 (squares) intrinsic fluorescence. The observed decrease indicates that C6PS binds to both FV a1 and FV a2 with similar affinities and structural responses. We assumed that the change in intrinsic fluorescence should be proportional to the fraction of protein bound and fitted the titration data using a model that assumed four identical and independent sites (10). We have shown that there is a 3-fold difference in the affinity of both human (12) and bovine (14) FV a1 and FV a2 for PS-containing membranes. Unlike their interactions with membranes, we conclude that bovine FV a1 and FV a2 bind to soluble C6PS with comparable affinities (11 Ϯ factor X a , along with either 50 nM factor V a1 (triangles), 5 nM FV a2 (squares), or 5 nM of the natural mixture of these two isoforms (circles). The C6PS CMC under the condition of this experiment (except that prothrombin was omitted) was determined by QELS to detect C6PS aggregates, as indicated in the inset. Prothrombin has been shown previously to bind C6PS very weakly (K d Ͼ5 mM) and thus to have no effect on its CMC at a concentration of 2 M (8).

FIG. 2.
Binding of C6PS to factor V a1 or V a2 as detected by intrinsic fluorescence response. Integrated intrinsic fluorescence intensities of 0.33 M FV a1 (q) or FV a2 (f) in 20 mM Tris, 150 mM NaCl, 0.5 mM CaCl 2 , pH 7.5, were measured as a function of C6PS concentration at 22°C to follow C6PS binding. The data were analyzed according to a simple stoichiometric binding model with the fraction of sites occupied expressed as the fraction of total fluorescence change, where F 0 is the fluorescence before addition of C6PS and [C6PS] is the concentration of free C6PS (well approximated for this experiment as total C6PS concentration). K d and F ϱ were adjusted during regression of the hyperbolic form shown to the data to minimize the sum of squared residuals. The inset shows the mean particle diameter detected by QELS as C6PS was added to 100 nM FV a2 with the sudden appearance of measurable particles indicating the C6PS CMC. 0.6 M for FV a1 and 12 Ϯ 0.7 M for FV a2 ) and changes in intrinsic fluorescence.
We determined the CMC of C6PS (Fig. 2, inset) under exactly the conditions used for our titration (0.33 M FV a ) to be ϳ750 M, which is consistent with what we have found elsewhere in the presence of FV a (10) and far above the lipid concentrations used for our C6PS titrations of FV a1 /FV a2 . This confirms that the responses we have seen are to a molecular rather than aggregated form of C6PS.
Binding of FV a1 and FV a2 to DEGR-X a in the Presence and Absence of C6PS- Fig. 3 shows the increase in DEGR-X a fluorescence with addition of FV a1 (squares) or FV a2 (circles) in the absence of C6PS (open symbols). Both FV a1 and FV a2 bind similarly to DEGR-X a in the absence of C6PS (K d ϭ 1.0 Ϯ 0.07 M for FV a2 and 1.3 Ϯ 0.07 M for FV a1 ). These K d values are in essential agreement with those derived from prothrombin activation kinetics, fluorescence experiments (18), affinity chromatography studies (27), and ultracentrifugation experiments (28). The presence of 400 M soluble PS (filled symbols) dramatically enhanced the interaction of FV a2 with FX a (K d eff ϭ 2.0 Ϯ 0.02 nM; inset) but had little effect on the interaction of FX a with FV a1 (K d eff ϭ 1.1 Ϯ 0.07 M). Thus, at least part of the different effects of C6PS on the activation of prothrombin by FX a in the presence of FV a2 or FV a1 (Fig. 1) reflects a dramatic difference in how C6PS influences the FV a2 -or FV a1 -enzyme interaction.
Binding of FV a1 and FV a2 to FX a Increases Proteolytic Activity-Because the presence of C6PS influenced the interactions of FV a1 and FV a2 with FX a so differently, we could not compare the cofactor activities of FV a1 and FV a2 under a common set of conditions. For this reason, initial rates of active site formation from prothrombin were monitored by DAPA fluorescence in the presence of 400 M C6PS and are shown in Fig. 4 and its inset on different cofactor concentration scales. The rate saturated upon addition of very low concentrations of FV a2 (inset) to yield an initial rate of catalysis by the FX a ⅐FV a2 complex of 100 Ϯ 2 nM/s. However, the initial rate of prothrombin activation did not saturate with addition of FV a1 (circles), and an estimate of the rate catalyzed by the FX a ⅐FV a1 complex had to be made from the hyperbolic fit shown by the solid line in Fig. 4 (540 Ϯ 5 nM/s). Just as for titration of DEGR-X a fluorescence, the K d eff of FV a2 interaction with FX a obtained from these hyperbolic fits (2.8 Ϯ 0.3 nM) was ϳ350-fold smaller than that for FV a1 interaction with FX a (1.0 Ϯ 0.15 M). It seems from these results that FV a1 is the more effective cofactor but that it has limited ability to bind FX a in a PS-dependent fashion. This explains why the behavior of the FX a ⅐FV a2 complex seemed to dominate the whole FV a experiment presented as open circles in Fig. 1.
Binding of DEGR-X a to FV a2 in the Presence of Varying C6PS-Next, we titrated DEGR-X a with FV a2 at fixed C6PS concentrations of 30, 60, 150, and 400 M (circles, squares, triangles, and diamonds, respectively, in Fig. 5). Each titration curve was fitted as described under "Experimental Procedures" to obtain effective dissociation constant, K d eff , for the X a ⅐V a2 complex. This yielded the variation with C6PS concentration of K d eff shown in the inset in Fig. 5 as well as fluorescence parameters (F 0 and F SAT , Equation 3) characterizing the complexes at different C6PS concentrations (not recorded in Fig. 5). Note that K d eff obtained in this way is not related to the K d of interaction between FX a and FV a in solution that is known to be ϳ1-3 M (18). This is an effective K d of interaction between DEGR-X a and FV a in which all C6PS-bound species of DEGR-X a and FV a2 are in equilibrium with all possible C6PSbound complex species (DEGR-FX a ⅐FV a ⅐C6PS) (see Equation 1 and "Appendix"). Because the species present in this equilibrium vary with C6PS concentration, K d eff also varies with C6PS concentration.
Characterization of the FX a ⅐FV a2 Complex in the Presence of C6PS by Native Gel Electrophoresis-Native PAGE was performed with FX a /FV a2 samples in the presence and absence of 400 M C6PS (see "Methods"). The inset to Fig. 6 shows one such gel. FX a and FV a2 samples in the presence of C6PS were also run as controls. A log-log plot of the retardation coefficient against the molecular weights of the marker proteins (circles), FX a (inverted triangle), or FV a2 (4-pointed open star) in the absence of C6PS, and of the FX a ⅐FV a2 complex (5-pointed star) in the presence of C6PS is shown in Fig. 6. From this, we estimated the molecular mass of the complex in the presence of C6PS as 217 Ϯ 2.5 kDa. In the absence of C6PS, no complex appeared (Fig. 6, inset, lane 1), and the measured molecular masses were 45.5 Ϯ 1.0 and 167 Ϯ 2 kDa for FX a and FV a2 , respectively. These compare with literature values of 46 kDa (29) for FX a and 168 kDa (30) for FV a , making the expected molecular mass of the complex 214 kDa. Controls gave measured molecular masses of FX a and FV a individually in the presence of C6PS as 45.8 Ϯ 1.2 and 167 Ϯ 2 kDa. These results demonstrate unequivocally the complex that is implied by our DEGR-X a fluorescence and FX a activity measurements and establish the stoichiometry of this complex as 1:1 as expected (1). An additional control (lane 3) confirmed that FV a1 (measured molecular mass 170 Ϯ 1.92 kDa) did not form a complex with FX a even in the presence of C6PS.
Activation of Meizothrombin to Thrombin in the Presence of FV a2 -The results in Fig. 5 show that the ability to form a FX a ⅐FV a2 complex clearly increased with C6PS concentration. We next asked whether the ability of that complex to function as a "prothrombinase" also increases with C6PS concentration. Activation of prothrombin to thrombin involves cleavage of two bonds. This means that activation proceeds via two possible intermediates and that four distinct proteolytic reactions must be characterized to define the process (22). We showed that, even in the absence of FV a , PS-containing membranes as well as C6PS 1) alter the rates of all four reactions, 2) alter the preferred pathway of activation, and 3) cause some intermediate to be converted processively to thrombin without release from the prothrombinase complex (22,31). In order to avoid these complications in characterizing the activity of the soluble FX a ⅐FV a2 complex, we focused here on only one of the four possible reactions, conversion of the intermediate MzII a to thrombin. Because the activation of this intermediate to thrombin reflects the prothrombinase-catalyzed cleavage of only one peptide bond, it is reasonable to expect proteolysis to behave according to a Michaelis-Menten formalism. Fig. 7 shows the initial rates of thrombin formation from MzII a measured at different lipid concentrations (50, 100, and 200 M) and as a function of increasing concentrations of MzII a in the presence of FV a2 . We fit these three curves individually to the Michaelis-Menten model (Equation 4) to obtain the K m and V max parameters that define the kinetics of MzII a activation. As reported in Fig. 7 (upper inset), K m did not change within the error of the experiment (0.47 Ϯ 0.03). However, V max increased with lipid concentration (circles in lower inset). By using Equation 2 and K d eff extrapolated from values measured at four C6PS concentrations (Fig. 5), the effective or empirical concentration of the assembled prothrombinase enzyme, E bound , was calculated. As noted under the "Appendix," this enzyme species is actually a mixture of species (PS⅐X a ⅐V a⅐ PS 4 , PS 2 ⅐X a ⅐V a ⅐PS 4 , PS 2 ⅐X a , and PS⅐X a ), although not all of these species are equally active. Without a more complex analysis, it is not possible to calculate k cat values for all these species, so we have calculated instead an effective k cat as V max /E bound and plotted this versus C6PS concentration in the lower inset to Fig.  7. The fact that V max /E bound decreased with added C6PS confirms that more than one enzyme species must be present and that the mix of species changes with C6PS concentration. It also suggests that species saturated in C6PS may be less active than species with a suboptimal amount of C6PS bound. We will comment further on this below. DISCUSSION We set out in this work by asking five questions posed in the Introduction, whose answers have led to the following significant new conclusions.
1) The two isoforms of factor V a (F Va1 or F Va2 ) bind with nearly identical affinities to FX a (Fig. 3).
2) Soluble C6PS binds with the same affinity to the two isoforms of FV a (Fig. 2).
3) In the presence of 400 M C6PS, the binding of FV a2 with factor FX a is 300 -350 times tighter than is the binding of FV a1 , FIG. 4. Initial rate of active site formation as a function of added FV a2 and FV a1 in the presence of C6PS. The initial rate of active site formation, monitored by DAPA fluorescence, was determined by stopped-flow fluorescence measurements, performed at 37°C with final concentrations in the mixing chamber of 1 M prothrombin, 5 M DAPA, 1 nM factor X a , 400 M C6PS, and various concentrations of factor FV a1 or FV a2 (inset) in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM CaCl 2 . The lines drawn through the data were obtained by fitting the data to a simple stoichiometric binding model (Equations 1-3) to determine K d eff values given in the text. whose affinity for FX a was not significantly influenced by C6PS (Fig. 3). 4) C6PS stimulates the assembly of FV a2 , as opposed to FV a1 , into a V a2 ⅐X a complex. This means that at least some of the difference in the cofactor activities of the two isoforms of FV a derives from their different abilities to bind FX a . However, it appears also that the FX a ⅐FV a1 complex is more active than the FX a ⅐FV a2 complex. 5) Perhaps the most significant conclusion from this study is that the prothrombinase complex assembled in the presence of soluble C6PS ϩFV a2 , without a membrane surface, catalyzes MzII a activation to II a at a rate (apparent k cat /K m ϳ1 ϫ 10 9 M Ϫ1 s Ϫ1 ) that is comparable with the rate reported in the presence of a PS-membrane surface and FV a .
The first three conclusions are directly evident from our data, as referenced above, but still deserve brief comment here. These three conclusions require that the reader be convinced that C6PS binds to individual sites on FX a and the FV a isoforms rather than condensing with them to form macromolecular complexes or micelles. That FX a bound to and was activated by C6PS well below the CMC of this short chained phospholipid has been thoroughly documented (8). Another paper from our lab (32) has confirmed this and shown by equilibrium dialysis that only two C6PS bind to FX a as follows: one to a site that regulates activity and is located in the epidermal growth factor domains linked to the Gla domain at the N terminus of the molecule, and one that competes with synthetic substrate binding and probably binds in a functionally insignificant fashion to the amine binding locus of the S2-S3 substrate recognition motif (33). Recent work from our lab has documented by equilibrium dialysis, intrinsic fluorescence, CD spectroscopy, and functional analysis that FV a binds four C6PS, also below the C6PS CMC, with resulting structural and C6PS-specific functional changes (10). One of these four C6PS binds to a single site in the C2 domain (34); the other three sites have not been located.
One might argue that some lipid complex smaller than a micelle could be responsible for the influence of C6PS on the structure and function of FV a and FX a , i.e. that the effects of C6PS on FV a and FX a might be due to locally organized surfaces rather than to individual C6PS molecules binding to individual sites on these two proteins. This is unlikely for two reasons. First, although it is well documented that amphipaths can form both small and large micelles of different shapes (35), the smallest micelle we could detect for C6PS under our conditions was roughly 11 nm in diameter and contained only about 220 lipids (8). This is a very reasonable limiting micelle size for a two-chain amphipath, because such amphipaths are unlikely to be able to form smaller, spherical micelles (35). Any smaller structure would leave hydrocarbon chains exposed to water, a situation that is thermodynamically very unfavorable (35). The smallest C6PS micelle in the presence of FX a was 26 nm in diameter, whereas the smallest we have detected in the presence of FV a was about 70 -90 nm in diameter (10). Our results indicate that the mix of FX a and FV a considered here lowered the CMC a bit more than either FX a or FV a did individually, but led to aggregates about the same size (70 -90 nm) as those recorded previously in the presence of FV a alone. This means that many more lipid molecules were needed to accomplish efficient lipid packing in the presence of these proteins. This makes it very unlikely that very small and hydrodynamically undetectable C6PS aggregates could explain our observations. The second, and even stronger, argument against this possibility is provided by the observations that only two and four C6PS molecules were bound at saturation per FX a and FV a , respectively (10,32). This small amount of lipid could not form even a local surface for the binding of these proteins, and we must conclude that specific sites on both FX a and FV a bind C6PS in a way that regulates the structure, function, and assembly of the components of the prothrombinase complex.
C6PS Promotes the Assembly of a FX a ⅐FV a2 Complex, although It Is Less Active than the FX a ⅐FV a1 Complex-The experiment that originally stimulated this investigation is shown in Fig. 1. The response of a mixture of FX a and either FV a isoform to C6PS is clearly biphasic, with the initial rapid rise in activity seen for both FV a isoforms up to about 20 M C6PS probably due to the tight association of FV a with C6PS (Fig. 2). The affinity of FX a for C6PS (K d ϳ110 M) 2 is such that very little FX a can be activated by C6PS at such low C6PS concentrations, so it must be that this rise is due to C6PS binding to FV a to stimulate its association with FX a . This must mean that binding of FV a to FX a in the absence of the activation of FX a by binding of C6PS significantly enhanced the activity of the enzyme. The second phase of the rise in prothrombinase activity occurs over a C6PS concentration range of 10 -100 M, consistent with this being due to binding of C6PS to FX a (8). 2 We conclude that C6PS binding to factor X a still has a significant role in fully activating factor X a even when it is already partially activated by binding to FV a .
It is evident from Fig. 1 that the rates of prothrombin activation in the presence of FV a1 or FV a2 are quite different. To observe good activity in the presence of FV a1 in the experiment shown in Fig. 1, we had to use a concentration of this isoform (50 nM) ten times greater than that of FV a2 (5 nM). Even so, the rate of prothrombin activation did not saturate in the presence of FV a1 even in the presence of 400 M C6PS, whereas in the presence of FV a2 , saturation occurred at roughly 200 M C6PS, a concentration for which both FV a2 and the tight site on FX a were saturated with C6PS. When bound to C6PS, the apparent affinity of FV a2 for FX a was 2 nM (Fig. 3), meaning that all FX a was bound to FV a2 under the conditions of our experiments. This means that the observed rate at saturation represents the activity of the FV a2 ⅐FX a complex. However, because the interaction between FV a1 and FX a remained weak in the presence of 400 M C6PS (Fig. 3), titration of the FV a1 /FX a mixture with C6PS in Fig. 1 did not result in complete incorporation of FX a into FV a1 ⅐FX a complex, consistent with the observed failure of prothrombin activation to saturate. These observations imply that the activity of the FV a1 ⅐FX a complex must be much more than 30% larger than that of the FV a2 ⅐FX a complex. This is supported by the FV a titration presented in Fig. 4, which indicates that the prothrombin-activating ability of the FV a1 ⅐FX a complex is at least twice that of the FV a2 ⅐FX a complex. Why might both these isoforms, with different abilities to bind to and activate FX a , be present in plasma? At this point, an answer is not known.
The Solution Prothrombinase Complex Is Optimally Active with Binding of One C6PS to Fx a -Under the "Results," we treated the assembly of the prothrombinase complex in solution by an empirical approach that ignored the fact that several species exist in solution. Here we present a detailed treatment in terms of the actual prothrombinase species likely to be present (see "Appendix"). We have shown that C6PS binding to the tight site on factor X a enhances the activity of the enzyme (8), 2 and binding of C6PS to factor V a2 enhances its affinity for factor X a (Fig. 3 and Ref. 10). Thus, the observed decrease of V max /E bound for MzII a activation with increasing C6PS concentration (Fig. 7) suggests that binding of a second C6PS to factor X a might actually inhibit binding to Va⅐PS 4 and/or inhibit the activity of the assembled prothrombinase complex. In an effort to test quantitatively this prediction, we adopted a second approach to analyzing the data in Figs. 5 and 7. This approach acknowledges that FX a binds two molecules of C6PS (32), the first with a k d XaPS of 110 M. 2 To a reasonable approximation, the second site is occupied sequentially, i.e. only after the first site is occupied, 2 with a site dissociation constant k d XaPS2 of 150 -1500 M. 2 The value of this constant is uncertain because it depends on the aggregation state of FX a , being roughly 150 -250 M when FX a exists as a dimer and roughly 1500 M when FX a is at low concentration, 2 as it is in our studies. Because of the existence of two C6PS sites on FX a , it can bind to FV a2 ⅐PS 4 in any of three putative forms, FX a , FX a ⅐PS, or FX a ⅐PS 2 , with stoichiometric dissociation constants that we cannot determine independently. We estimated these dissociation constants and the activities of the putative species as described under the "Appendix." This analysis predicts that the presumed species X a PS⅐V a PS 4 (k cat ϭ 2360 Ϯ 66 s Ϫ1 or 711 Ϯ 20 s Ϫ1 ) formed with a K d of 42 Ϯ 8 or 6.7 Ϯ 0.1 nM, depending on whether we used k d XaPS2 ϭ 150 or 1500 M, respectively. The putative XaPS 2 ⅐V a PS 4 complex (k cat ϭ 333 Ϯ 13 or 167 Ϯ 7 s Ϫ1 ) formed with a K d of 2.2 Ϯ 0.1 or 1.6 Ϯ 0.04 nM, again depending on the value assigned to k d XaPS2 . A third species (X a ⅐V a PS 4 ) was present in our model but appeared to be formed with so weak an affinity (K d ϭ 82 Ϯ 13 or 0.13 Ϯ 0.02 M) that our activity data (Fig. 7) were adequately described without accounting for it. For comparison to the literature, we calculated pseudosecond order rate constants, k cat /K m , for the species X a PSV a2 PS 4 (1.5 to 5.0 ϫ 10 9 M Ϫ1 s Ϫ1 ) or for the species X a PS 2 V a2 PS 4 (3.5 to 7.1 ϫ 10 8 M Ϫ1 s Ϫ1 ) at 37°C. These rate constants can be compared with a rate constant (9.2 ϫ 10 7 M Ϫ1 s Ϫ1 ) reported for the proteolysis of bovine MzII a at 22°C by a FV a FX a complex assembled on a membrane containing PS plus phosphatidylcholine (23). The difference between these pseudosecond order rate constants could derive from a variety of differences between our experiments and earlier kinetic studies. For instance, most previously published kinetic data are for a natural mixture (ϳ2:1) of FV a2 and FV a1 , whereas our solution prothrombinase contains only FV a2 . Very little is known about the abilities of these two isoforms to form a productive prothrombinase on a PS-containing membrane, although there seems not to be as dramatic a difference (13,14) as we have observed in solution. Until these and other issues are addressed, it would be premature to conclude that the prothrombinase assembled in solution is more active than that assembled on a membrane. To make this comparison properly, the rates on a membrane and in the presence of C6PS would need to be compared in the same lab under the same experimental conditions. A detailed kinetic analysis of the solution-assem-bled human prothrombinase is underway and will accomplish this comparison.
Whereas the uncertainty in k d XaPS2 leads to considerable uncertainty in k cat estimates for the X a PSV a2 PS 4 or X a PS 2 V a2 PS 4 species and in the K d values for forming these species, the results still clearly indicate that occupancy of the second C6PS site on FX a promotes formation of the prothrombinase complex but inhibits its proteolytic activity. At first glance, this is quite surprising, but it is actually not an unreasonable possibility, because we have shown that binding of C6PS to the second FX a site competes with synthetic substrate binding to the active site of FX a (32). Although this does not seem to interfere with the FX a proteolytic activity in the absence of FV a (8), 2 our results indicate that C6PS binding to the second site on FX a bound to FV a2 may interfere with formation of a productive enzymesubstrate complex.
Implications for the Role of Platelet PS in Regulating Blood Coagulation-An ATP-dependent, amine-phosphatide-specific pump maintains PS asymmetry across the resting platelet membrane (36). However, PS is exposed on membrane vesicles that appear when human platelets are activated (2,7,37). It has been widely believed that membrane vesicles, either those released from activated platelets or synthetically produced membranes that model these (1), were essential for assembly of an active prothrombin-activating complex. Our results show that a soluble form of PS can assemble a fully active prothrombinase in solution. This implies that it is the exposure of PS on the surface of activated platelet membranes, not the membrane surface itself, that is crucial to assembly of the prothrombinase complex. It also suggests that PS exposure acts not only to locate factor X a to the membrane surface but also to activate this enzyme (8,31). 2 Similarly, PS exposure both locates factor V a to the membrane and activates it (10) to bind factor X a (Fig.  3). Because the factor VIII-dependent factor X-activating complex is highly homologous to the prothrombinase (38,39), it seems likely that platelet PS also triggers factor X a formation at the platelet plug. Thus, a single upstream event (platelet activation by thrombin or collagen) can trigger activation of key downstream steps in blood coagulation via exposure of otherwise buried PS. In this sense, platelet membrane PS can be thought of as a second messenger in regulating blood coagulation. APPENDIX C6PS binds to two different binding sites of FX a in roughly a sequential fashion to elicit different structural responses from the two sites 2 (Equation A1), XaPS 2 (Eq. A1) The first of these sites seems to regulate activity and has a reasonably well defined site-dissociation constant of k d XaPS Ϸ110 M, whereas the site-dissociation constant of the second site (k d XaPS2 ) is less well defined. 2 Because binding of four C6PS to FV a2 is quite tight (K d VaPS ϭ 12 M) and follows a simple hyperbolic functionality (Fig. 2), we can treat this in terms of a simple stoichiometric binding equilibrium that is close to saturation under almost any C6PS concentration that we have considered (Equation A2): 4 (Eq. A2)