A phosphatidylserine binding site in factor V_a C1 domain regulates both assembly and activity of the prothrombinase complex

Activated coagulation factor V (FVa) is a protein cofactor that binds to membranes containing phosphatidylserine (PS) and regulates the production of thrombin by the prothrombinase complex.^1,2  The prothrombinase complex consists of FVa, the serine protease factor X_a (FXa), calcium, and a phospholipid-containing membrane. Factor V exists in plasma as an inactive, large, single-chain glycoprotein with little or no intrinsic procoagulant activity.^3,4  It is composed of repeating domains: 2 A domains (A1 and A2) at the amino-terminal end, a heavily glycosylated connecting B domain, a third A domain (A3), and 2 carboxyl-terminal C domains (C1 and C2).^1,5,6  Limiting proteolysis of factor V by thrombin^3,7  or FXa⁸  removes the B domain-producing FVa that has procoagulant activity. FVa is a heterodimer⁹  composed of a heavy chain (FVa-HC; A1A2; molecular weight = 105,000) and a light chain (FVa-LC; A3C1 C2; molecular weight = 74 000 or 71 000).^3,7,9  Heterogeneity of the light chain is seen in both bovine and human FVa^3,7,10  and is due in human FVa to alternative glycosylation at Asn²¹⁸¹ in the C2 domain^11,12  (Figure 1). The glycosylated form is termed FVa₁, whereas the other form lacking glycosylation at this site is FVa₂. Human FVa₂ differs from FVa₁ in its ability to bind tightly to FXa in the presence of PS-containing membranes or a short-chain PS (C6PS).^13-15  Human plasma FVa₂ and FXa form a fully active prothrombin-activating complex in the presence of soluble C6PS.¹⁶ 

The carboxy-terminal C2 domain of FVa (residues 2037-2196 in human FVa) contains a soluble PS (C6PS) binding pocket flanked by a pair of tryptophan residues, Trp²⁰⁶³/Trp²⁰⁶⁴¹⁷  (Figure 1). Mutating these tryptophan residues to alanine (C2 mutant) yielded FVa with impaired ability to interact with the phospholipid membrane^18,19  and results in a more than 400-fold reduction in FVa₂-membrane–binding affinity but does not block the assembly and activity of the prothrombinase in the presence of 25% PS-containing membranes²⁰  or soluble C6PS.²¹  The binding site in the C2 domain is needed for binding of FVa to PS-containing membranes,^20,21  accounts for one of 2 C6PS sites in FVa light chain,²²  and is not specific for soluble C6PS,¹⁷  even though the cofactor activity of FVa₂ is specifically enhanced 15 000-fold by C6PS.²¹  These observations suggested that a C6PS-binding site exists somewhere else in FVa light chain and might serve as a PS-specific regulatory site for assembly or activity of the rFVa₂·FXa complex.²¹  Recently, we showed that prothrombin activation in the presence of the FVa₂ mutant (Y1956,L1957)A (C1 mutant) was markedly impaired on phospholipid vesicles containing 10% or less PS but was essentially normal on vesicles containing 25% PS.²³  This observation suggests that these residues in the C1 domain may somehow be involved in an interaction with PS that could be important for FVa₂ cofactor activity. The 2 solvent exposed amino acids (Y1956,L1957) in the C1 domain are located analogously to the critical Trp residues in the C2 domain (Figure 1).

This work aims to test the hypothesis that the PS-regulatory site in FVa₂ involves these residues and is thus located in the C1 domain analogously to the C6PS-binding site in the C2 domain. Specifically, we asked whether (Y1956,L1957)A mutant rFVa₂ (C1 mutant) affects the cofactor activity of FVa₂ in the presence of C6PS and the assembly of the FXa-FVa₂ complex in the presence of C6PS. Finally, we determined the stoichiometries of binding of C6PS to the C1 mutant and C1-C2 mutant to confirm that a single C6PS site in the C1 domain requiring the Y1956 and L1957 residues is indeed the PS-binding site responsible for regulating the cofactor activity of FVa.

1,2-Dicaproyl-sn-glycero-3-phospho-L-serine (C6PS), 1,2-dicaproyl-sn-glycero-3-phosphocholine (C6PC), and L-serine were purchased from Avanti Polar Lipids (Alabaster, AL). Radioactive serine was purchased from American Radiolabeled Chemicals (St Louis, MO). Dansylarginine-N-(3-ethyl-1,5-pentanediyl) amide (DAPA) was obtained from Haematologic Technologies (Essex Junction, VT). [5-(Dimethylamino)-1-naphthalenesulfonyl] glutamylglycylarginyl chloromethyl ketone¹  was purchased from Calbiochem (San Diego, CA). All other chemicals were ACS reagent grade or the best available grade.

C6PS stocks were prepared from measured quantities of 2.5 mg/mL stock solutions in chloroform²¹  and used within 1 day of preparation. Radiolabeled C6PS was prepared enzymatically using phospholipase D, generously provided by Avanti Polar Lipids, radiolabeled serine (serine, L-[1-14C], HOCH₂CH(NH₂)COOH, specific activity 50 to 60 mCi/mmol 1.85-2.22 GBq/mmol), and C6PC. Radioactive C6PS was purified from the reaction mixture by silica gel column chromatography according to the methods as described by Comfurius and Zwaal²⁴  with minor modifications.

Wild-type and mutant recombinant human factor V_a (rFVa) were expressed in COS cells using a B domain–deleted FV construct and purified as previously described.^11,23  Activated rFV (rFVa) was prepared by incubating rFV (1.6 mg) for 30 minutes with thrombin (17 μg). After activation, 200 nM D-Phe-Pro-Arg- chloromethyl ketone was added to inactivate the thrombin. The rFVa was then purified on a Mono S HR 5/5 ion-exchange column (GE Healthcare, Piscataway, NJ) to separate factor FVa₁ and factor FVa₂ as described previously.^11,15  We have used rFVa₂ in the work described here because it is a better cofactor than rFVa₁ and because only rFVa₂ interacts tightly with FXa in solution with C6PS.¹⁴  The C1 mutant [(Y1956,L1957)A] and C2 mutant [(W2063,W2064)A] were expressed in COS-7 cells and purified as described.^20,23  The C1-C2 mutant [(Y1956,L1957,W2063,W2064)A] purification is also described elsewhere.²⁵ 

To screen 42 rFVa mutants for cofactor activity in the presence of C6PS, mutants were purified using a 1-mL High-Trap SP HP column. These preparations contained approximately 70% rFVa₂ and approximately 30% rFVa₁ based on sodium dodecyl sulfate–polyacrylamide gel electrophoresis²⁰  (data not shown).

Human factor X_a and human prothrombin and thrombin were obtained from Haematologic Technologies.

Human DEGR FXa was prepared as described.^14,26  DEGR-FXa was then dialyzed against 50 mM Tris, 0.15 M NaCl, pH 7.5, to remove free reagent.²⁷  Titration of DEGR-FXa by rFVa₂ was carried out with 1 nM DEGR-FXa (buffer: 400 μM C6PS, 20 mM Tris, 150 mM NaCl, 5 mM CaCl₂, pH 7.5), with 4-minute stirred equilibration between rFVa₂ additions. Several intensity readings were averaged after each addition and corrected, via controls, for dilution, buffer background, and any small amount of photobleaching, which was largely eliminated by closing the excitation slit between measurements.

Changes in the intrinsic fluorescence intensity of rFVa₂ in response to added C6PS were recorded with an SLM 48 000-MHF spectrophotometer (SLM Aminco, Urbana, IL) as described previously.^17,21  Key controls for the occurrence of C6PS micelles were performed by measuring the changes in pyrene fluorescence.²⁸ 

The response of native or mutant rFVa₂ intrinsic fluorescence to soluble C6PS was fitted to a single binding site model supplied with SigmaPlot (version 6.0; SPSS, Chicago, IL).

The rate of prothrombin activation was estimated from the time-dependent change of DAPA fluorescence as it bound to the activation products exactly as described.^14,16 

The lipid-induced enhancement of cofactor activity (E) of all the 42 C1 and C2 mutants was determined as described by Zhai et al²²  using the following equation:

Here, rX_a × V_a × PL is the rate of prothrombin activation by enzyme in the presence of FVa and lipid, as measured at 400 μM C6PS as described.^14,16  Other terms are defined analogously. The quantity on the left is the ratio used experimentally to evaluate E. The ratio rXa·Va/rX_a was determined as the average from 5 independent experiments, whereas rX_a· rXa·Va·PL/rX_a·PL was the average of 3 independent measurements. The ratio on the left reflects the meaning of this quantity, as the ratio in the numerator reflects the effect of C6PS on the FXa-FVa complex, whereas that in the denominator corrects for the effect of lipid on FXa.

The stoichiometry of soluble C6PS binding to wild-type, C1-mutated, C2-mutated, and C1-C2–mutated rFVa₂ in the presence of 5 mM Ca²⁺ was determined by equilibrium dialysis as described previously,^17,21,22  but with the modification of using radiolabeled C6PS to obtain accurate differences in C6PS concentrations between chambers using the low concentrations of recombinant rFVa₂ that we had available. Experiments were performed at different rFVa₂ concentrations and a fixed concentration of C6PS (0.8 and 1 μM for wild-type at 70 μM C6PS, 2 and 1.6 μM for the C1 and C2 mutants, and 2 and 2.5 μM for the C1-C2 mutant at 100 μM C6PS). These conditions ensured that at least 80% of the protein was bound to C6PS.

Table 1 gives the cofactor activity (determined as described in the equation in “Lipid-dependent cofactor activity”) of the wild-type and mutant rFVa₂. The results clearly show that the prothrombinase activity of C2-mutated rFVa₂ is the same as wild-type, as reported previously,²¹  whereas that of C1 mutant and the C1-C2 mutant were both reduced by approximately 14 000- and 15 000-fold, respectively, relative to wild-type rFVa₂.

The initial rates of prothrombin activation in the presence of wild-type rFVa₂, the C2 mutant, the C1 mutant, and the C1-C2 mutant are plotted vs C6PS concentration in Figure 2. The data for wild-type rFVa₂ (red circles) and the C2 mutant (green circles) were fitted to hyperbolas as befitting a single-site binding model. As described in Figure 2, the K_d^app's for these fits were 3.8 (± 0.5 [SEM]) μM and 4.1 (± 0.5) μM for wild-type and C2 mutant rFVa₂, respectively. Previously reported values of K_d^app for bovine plasma-derived FVa (9 μM²² ) and FVa₁ (11 μM) and FVa₂ (10 μM) to C6PS,¹⁴  and for human rFVa₂ (20 μM) determined by intrinsic fluorescence²¹  are somewhat larger when measured by intrinsic fluorescence changes. We note that these fluorescently determined K_d^app's reflect binding to multiple sites, 2 presumably nonphysiologic sites in the heavy chain and 2 in the light chain.²²  The activity-derived K_d^app's reflect binding to whatever site(s) is (are) responsible for C6PS-triggered cofactor activity. These results demonstrate that this (these) site(s) is (are) in the C1 domain and demonstrate a K_d^app of approximately 4 μM.

Two aspects of our analysis of these data indicate that C6PS binding to rFVa₂ alone triggers assembly of a prothrombinase complex with nearly full activity. First, if we fit the data from 0 to 30, 40, 50, or 400 μM C6PS (Figure 2A, inset, up to 50 μM C6PS), the best-fit parameters did not change appreciably for either the wild-type (K_d^app from 3.64-3.8 ± 0.5 μM and saturating activity from 176 to 179 ± 10 nM/sec) or C2 mutant (K_d^app from 4.0 to 4.1 ± 0.5 μM and saturating activity from 172 to 174 ± 8 nM/sec) rFVa₂. Because the effect of C6PS on the activity of FXa does not saturate until approximately 300 μM C6PS (K_d^app of 70-90 μM), the steep rise in FXa activity (as seen in Figure 2) on adding C6PS must be because of lipid binding to a regulatory site on rFVa₂ having a K_d^app of approximately 4 μM. This means that the dramatic effect of C6PS on the rate of prothrombin activation must be the result primarily of its binding to a regulatory site on FVa₂ to trigger high-affinity binding to FXa to produce a nearly fully active prothrombinase complex. This demonstrates that the large C6PS-induced cofactor activity recorded in Table 1 is overwhelmingly the result of the ability of C6PS to produce a FVa₂ conformation that binds with very high-affinity to FXa. Second, the saturating activities for both the wild-type (179 ± 7 nM/sec) and C2 mutant (177 ± 12 nM/sec) rFVa₂ correspond to a k_cat/K_M of approximately 2.3 (± 0.1) × 10⁸ M⁻¹s⁻¹, assuming a K_M of 0.82 μM for the FVa₂-FXa complex (average of 0.65 μM for C6PS¹⁶  or to 1 μM for membrane-assembled²⁹ ). We have previously reported a k_cat/K_M of approximately 2.2 × 10⁸ M⁻¹s⁻¹ for plasma-derived human FVa₂-Xa complex.¹⁶  Consistent with this, the dashed curve in Figure 2A shows clearly that fitting the wild-type data from only 0 to 50 μM of C6PS underestimates only very slightly the activity at high C6PS activity. We note here and have noted previously¹⁶  that the activity of the C6PS-rVa₂-FXa-C6PS complex expressed as a k_cat/K_M (2.2 × 10⁸ M⁻¹s⁻¹) is surprisingly roughly an order of magnitude greater than 1 × 10⁷ M⁻¹s⁻¹²⁹  and others³⁰  have reported for the membrane-assembled prothrombinase obtained from a natural mixture of FVa₁ and FVa₂ (∼10⁷ M⁻¹s⁻¹). This observation is reproducible, and the reasons for it are under investigation.

Figure 2B shows the activity of the samples containing the C1 (red triangles) and C1-C2 mutants (green triangles) in response to increasing C6PS concentration. The lines through these data were also obtained using a single site binding model with K_d^app's = 60 (± 4) μM and 56 (± 3) μM and saturating activities of 0.0079 (± 0.0003) nM IIa/s and 0.0078 (± 0.0002) nM IIa/s, respectively. The effect of C6PS on FXa alone occurs with a comparable but somewhat larger K_d^app (70 ± 10 μM³¹ ; 86 ± 5 μM³² ). The saturating activities (∼0.0079 and ∼0.0078 nM/sec for C1 and C1-C2 mutants) are significantly greater than previously estimated for the effect of C6PS on FXa (0.0042 nM/sec³³  or 0.0036 nM/sec, R. Chattapadhyay et al, 2008, submitted for publication). This difference is probably the result of the presence of C1 and C1-C2 mutant rFVa₂, which we have previously shown still enhance prothrombinase activity slightly at increasing concentrations in the presence of PS-containing membranes.²³  These observations suggest that, even though C1 and C1-C2 mutant rFVa₂ bind FXa very weakly, the resulting C1 and C1-C2 mutated rFVa₂-FXa complexes are still more active than FXa alone.

Because mutating the (Y1956,L1957)A residues in the C1 domain alters C6PS-stimulated prothrombin activation, it is reasonable to ask whether this mutation alters the binding of C6PS to rFVa₂. Figure 3 shows the change in the intrinsic fluorescence of 0.2 μM of the C1 mutant rFVa₂ resulting from titration with C6PS. The critical micelle concentration of C6PS under these conditions was determined by changes in pyrene fluorescence²¹  to be 700 (± 5) μM. This method is documented to detect micelles with aggregation numbers as low as 10.³⁴  This critical micelle concentration is well above the maximum C6PS concentration used in our titration, guaranteeing that the effects reported are the result of monomeric C6PS and not micelle formation. The binding data were fit by a simple single-binding-site model assuming a stoichiometry of one (Figure 3), yielding a K_d^app of 47 (± 5) μM. Wild-type rFVa₂ binds C6PS with a K_d^app of 20 (± 1) μM.²¹  Comparing these results, we see that elimination of the aromatic and hydrophobic side chains Tyr¹⁹⁵⁶/Leu¹⁹⁵⁷ from the C1 domain had little effect on the affinity of C6PS binding to rFVa₂. This probably reflects the fact that FVa₂ has 4 sites of roughly equivalent affinity for C6PS, so that the effect of eliminating any one would be difficult, if not impossible, to detect with a simple binding assay. Probably for the same reason, the affinity of C2-mutated rFVa₂ for C6PS was also only slightly reduced (apparent K_d ∼37 μM²¹ ) compared with wild-type, even though we know that the C2 PS site is largely responsible for membrane binding of rFVa.²⁰  We conclude that we cannot use simple binding affinity as a means to detect the effect of the (Y1956,L1957)A mutation on C6PS binding.

Because binding affinity could not reveal the effect of the (Tyr1956/Leu1957)A mutation on C6PS binding, we turned to direct measurement of binding stoichiometry. There are 4 C6PS-binding sites on FVa, with 2 of these residing in the light chain containing the C1 and C2 domains.²²  The stoichiometries for the binding of C6PS to rFVa₂, the C2 mutant, the C1 mutant, and the C1-C2 mutant were determined by equilibrium dialysis, with values given in Table 1. These results demonstrate that mutating the Tyr¹⁹⁵⁶/Leu¹⁹⁵⁷ residues in the C1 domain blocks binding of a second C6PS molecule to the rFVa₂ light chain. We conclude that these residues are essential for expression of the second PS-binding site of rFVa₂ light chain.

Because the cofactor activity of FVa₂ requires binding to FXa, we asked next whether C6PS binding triggers tight binding of C1-mutated rFVa₂ to DEGR-labeled human FXa, as we have previously demonstrated for wild-type rFVa₂ and the C2 mutant.^16,21 Figure 4 presents isotherms, obtained in the presence of 400 μM C6PS, for the binding of wild-type rFVa₂, C2 mutant, C1 mutant, and C1-C2 mutant of rHFVa₂ to human DEGR-FXa. We have previously shown that the complex formed under these conditions with plasma-derived human FVa₂ is a fully functioning prothrombin-activating complex.^14,16,35  With the stoichiometry fixed at 1:1, we obtained the best fit description of the data shown in Figure 4 with K_d (and saturating F/F₀ ± SEM) values of 3.1 (± 0.5) nM (1.20 ± 0.015), 4.0 (± 0.6) nM (1.42 ± 0.03), 564 (± 35) nM (1.45 ± 0.02), and 624 (± 40) nM (1.45 ± 0.02) for the wild-type rFVa₂, C2 mutant, C1 mutant, and C1-C2 mutant, respectively. These results clearly show that alanine mutation of the hydrophobic (Y1956,L1957) residues of the C1 domain of rFVa₂ reduces the affinity of FVa₂ for FXa in the presence of 400 μM C6PS by approximately 200-fold. Note that this reduction is still less than the 1000-fold increase in affinity of FVa₂ for FXa seen in the presence of C6PS,¹⁶  indicating that mutation of the Y1956,L1957 residues does not completely eliminate the effect of C6PS. This could account for the observation that FXa appears to bind C6PS with slightly more affinity in the presence of the C1 mutant rFVa₂ than it does in the absence of rFVa₂ (Figure 2B). We note that all 3 rFVa₂ mutants produced a larger change in DEGR-FXa fluorescence than observed for the wild-type rFVa₂. Although we can offer no certain interpretation of this, it is conceivable that any alterations in the C domains of FVa can be transmitted to the A2 domain that appears to be involved in interactions with FXa. Indeed, a linkage between the C1 domain and sites involved in interaction with FXa is demanded by the sum total of the results presented here.

We show here that stimulation of FVa cofactor activity by C6PS requires (Y1956,L1957); we showed previously that it is blocked by glycosylation at Asn²¹⁸¹, which is located in the FVa C2 domain.^11,14  We asked next whether other amino acid residues within the FVa C1 and C2 domains might also contribute directly or indirectly to expression of cofactor activity in the presence of C6PS by investigating the effect of C6PS on a panel of 42 previously characterized FVa mutants^18,23  (Figure 5). These results indicate that Tyr¹⁹⁵⁶ and Leu¹⁹⁵⁷ contribute approximately equally to the expression of FVa cofactor activity in the presence of C6PS. Several other amino acid side chains within the C1 domain appeared to make small but statistically significant contributions to the expression of cofactor activity in the presence of C6PS. These include (Lys¹⁹⁵⁴/His¹⁹⁵⁵), Glu¹⁹⁶⁴, and Arg²⁰²³. Each of these residues is located within or in close proximity to an apparent “pocket” that is structurally (Figure 1) related to the C6PS binding pocket that we have demonstrated in the C2 domain.¹⁷  In contrast, none of the 22 C2 domain mutants had a significant effect on rFVa cofactor activity in the presence of C6PS.

The C domains of FVa contribute both to membrane binding and to regulating the ability of FVa to bind with high affinity to FXa, with both processes being dependent on binding of PS molecules to these domains. Support for this conclusion derives both from our previously published results and the new observations presented here on the role of PS binding to the C1 domain in regulating FVa binding to FXa.

The conclusion that the C2 domain is primarily responsible for membrane binding rather than cofactor activity follows from several previous reports. First, the structure of the C2 domain of human FVa is a β-sandwich that caps an apparent “pocket” at one end of the β-sheet structure.³⁶  Two partially hydrophobic loops that protrude (one in a β-loop and the other in a near β-loop) from the “bottom” end of the β-sandwich have hydrophobic residues at their tips (Figure 1). In the first loop, the hydrophobic residues are a pair of tryptophans (W2063,W2064), and in the other a leucine (L2116). These and another β-loop form a “pocket.” Second, the free energy of rFVa₂ binding to 25% PS-containing membranes differs from that of a C1-mutated rFVa₂ by only approximately 1 k_BT.²¹  Third, use of a (W2063,W2064)A mutant rFVa₂ and expressed C2 domain showed that the (W2063,W2064) residues are critical for both C6PS (K_d > 300 μM¹⁶ ) and tight membrane binding (K_d ≫ 100 nM^16,20 ). However, these tryptophans are not needed for formation of a fully active prothrombinase complex in solution.¹⁶  They are also not required for prothrombinase activity on a 25% PS-containing membrane as long as concentrations of rFVa₂ and FXa are used that are 10-fold and 4-fold higher, respectively, than required to reach maximal activity with wild-type rFVa₂.²⁰  Fourth, these Trp residues insert into the bilayer interface during membrane binding.²¹  Fifth, labeling of a Cys at the top of the C2 pocket blocked binding of C6PS to rFVa-C2 domain.¹⁷  Thus, the (W2063,W2064) indole side chains, which form the rim of the pocket on the C2 domain, define the PS-binding pocket and account for nearly all of the free energy of binding of rFVa₂ to 25% PS-containing membranes.

The results summarized in “The C2 domain, Previous work” suggest that another PS site on rFVa₂ must account for the clear regulatory role of C6PS in triggering assembly of a fully active prothrombinase complex.²¹  Two C6PS sites exist in the FVa light chain and 2 in the heavy chain,²²  with the heavy chain sites quite probably being of no physiologic significance because of the large distance that probably separates the heavy chain from the membrane.³⁷  In addition, the C1 domain is similar to the C6PS-binding C2 domain. Finally, a double C2¹⁸  and C1²³  mutant interfered with both membrane binding and appearance of an active prothrombinase complex.²⁵  For these reasons, we undertook the current study to examine the effects of C6PS on the C1 domain.

Because the isolated human C1 domain has yet to be expressed, isolated, and crystallized, somewhat less is known about its properties than about the C2 domain. However, a recent crystal structure of APC-inactivated bovine FVa³⁸  (Figure 1) shows that the C1 domain is structurally analogous to that of the C2 domain. It too has a partially hydrophobic β-loop that extends down from the β-sandwich, with a hydrophobic pair (Y1956,L1957) at its tip, and with structural similarity to the Leu²¹¹⁶ loop of the C2 domain. At least in this structure, the tip of the C1 loop is roughly in the same plane as the tips of the 2 C2 loops with the 2 C domains roughly perpendicular to this plane, so that one could imagine the entire loop tips located at the hydrophobic/hydrophilic interface of a membrane lipid bilayer (Figure 1). Two other, non-β loops help form a “pocket” in the C1 domain.

If this x-ray-derived model has relevance to the structure of FVa on a membrane, it is probable that only the C6PS-binding sites in the light chain are physiologically important. We have previously suggested that the 2 C6PS sites located in the FVa heavy chain are anomalous and correspond to protein recognition sites.²²  Here we focus on the C1 domain PS-binding sites and show by alanine mutagenesis that (1) the (Y1956,L1957)A mutation is unique in its ability to affect prothrombinase activity (Figure 5); (2) the (Y1956,L1957)A mutation eliminates the PS-triggered cofactor activity of rFVa₂ (Figure 2; Table 1); (3) the (Y1956,L1957)A mutation reduces the C6PS-induced affinity of rFVa₂ for FXa by approximately 200-fold (Figure 4); and (4) the (Y1956,L1957)A mutation eliminates one of 2 binding sites for C6PS in the light chain of rFVa₂ (Table 1).

All these conclusions support the hypothesis that the Y1956 and L1957 residues line a PS-binding pocket that regulates FVa cofactor activity. Because we have been unable to isolate expressed rHFVa₂ C1 domain, we do not have the cystine labeling data that we used to block and thus locate with certainty the PS-binding site in the C2 domain.¹⁷  However, the structural similarity between the C1 and C2 domains, along with the mutational analysis presented here, strongly supports this hypothesis.

These observations and other previously published observations on the C2 domain^{16,20-23,25,33,39}  suggest a model for FVa membrane binding and activation by PS, as summarized in Figure 6. In this model, binding of wild-type rFVa₂ to a PS-containing membrane (top panel) occurs primarily (but not exclusively^16,20 ) via the C2 domain. It is well known that extrinsic membrane proteins, including FXa and FVa, can bind to membranes through multiple interaction sites.^40,41  In the case of rFVa₂, the interaction that overcomes most of the unfavorable free energy of moving the protein from a 3-dimensional to 2-dimensional environment comes from the C2 domain.²¹  However, once rFVa₂ is on the surface, the less avid C1 site can be occupied. This triggers a change in rFVa₂ that causes it to bind with greatly increased affinity to FXa (Figure 4).^16,22  FXa itself is activated by approximately 60-fold through binding to C6PS or to a PS-containing membrane^33,39  but is optimally active when bound to FVa₂.¹⁶  This results in the fully active prothrombinase.

The model in Figure 6 accounts for all our current and past observations on lipid effects on factors FVa and FXa. Our previous reports have established that, despite the important functions that we have demonstrated for the C2 and C1 domains of FVa, both the C1 mutant and C2 mutant forms of rFVa₂ can form active prothrombin-activating complexes with FXa on membranes containing 25% PS.^20,23  In terms of our model, the ability of the (W2063,W2064)A rFVa₂ (C2) mutant to support formation of an active prothrombinase on 25% PS-containing membranes²⁰  can be explained by the presence of a very small amount of mutant rFVa on a membrane (eg, K_d of FVa binding to a phosphatidylcholine membrane ∼ 0.3 μM³³ ; K_d ≫ 100 nM for binding to 25% PS membranes²⁰ ). C2 mutant FVa₂ on a PS-containing membrane should still interact normally with the intact C1 domain, leading to high-affinity binding with FXa (K_d ∼0.6 nM²¹ ). Because FXa binds with reasonable affinity to PS-containing membranes (K_d ∼0.15 μM⁴⁰ ), binding of FXa to a membrane can also enhance assembly of a C2 mutant-containing prothrombinase complex on a membrane. Mass action then will drive formation of an active prothrombinase, but at slightly higher rFVa₂ or FXa concentrations than would be required with wild-type rFVa₂.

The affinity of the (Y1956,L1957)A rFVa₂ (C1) mutant for 25% PS-membranes is reduced only slightly compared with that of wild-type rFVa₂,^16,23  so it locates to the membrane surface very effectively. However, PS does not bind with normal affinity to the regulatory site in the mutated C1 domain (Table 1), meaning that it binds FXa with approximately 200-fold lower affinity than does wild-type rFVa₂ (Figure 4). Nonetheless, a solution K_d for the FXa-rFVa₂ complex of 0.5 to 0.6 μM (9.1-10.9 × 10⁻⁹ mol fraction) corresponds to a unitary free energy of association (ΔG°) of −18.5 to −18.3 k_BT. Under the conditions used in our previous paper (0.2 nM C1-mutated rFVa, and 5 μM lipid in membranes²³ ), nearly all rFVa should be membrane bound. Assuming that this same ΔG° of association holds for the membrane-associated FXa and rFVa₂, we can obtain the fraction of surface-bound FXa bound to surface-bound C1 mutant rFVa₂ by simply expressing surface concentrations in mol fraction units. If we do so, we find that it would take only 6 nM FXa to essentially saturate (99.9% bound) the membrane-bound C1 mutant rFVa and thereby obtain full prothrombinase activity. In solution, the same concentrations of FXa and C1 mutant rFVa₂ would yield approximately 2 fM prothrombinase. Thus, the profound effect of dimensionality reduction can explain why C1 mutant rFVa₂ can form a prothrombinase complex in the presence of 25% PS membranes, albeit at approximately 8- to 10-fold higher FXa concentration than is required for wild-type rFVa₂.²³  The fact that even higher concentrations of FXa or C1 mutant rFVa₂ still do not produce full prothrombinase activity²³  reflects the much weaker binding of these proteins to membranes with low PS content.^40,41  Taken together, the current results along with our previous observations²³  indicate that the C6PS-binding pocket defined by the (Y1956,L1957)A mutation regulates the affinity of rFVa₂ for FXa but not the effect of FVa on FXa activity.

It is worth noting that the turnover number of this membrane-assembled C1 mutant rFVa₂ complex is within error limits the same as that seen with wild-type rFVa₂ complex.^23,25  Thus, although the (Y1956,L1957)A mutation clearly affects the assembly of the prothrombinase complex,²¹  it does not affect significantly the activity of the prothrombinase complex once it is assembled. This suggests that the tight-FXa-binding conformation induced by PS binding to the C1 domain has little effect on cofactor activity other than being a conformation that promotes formation of the FXa-FVa complex.

The double-mutant (Y1956,L1957,W2063,W2064)A rFVa₂, should not bind tightly to membranes and should not be activated to a form with high affinity for FXa. According to our model, we expect that it will have little, if any, activity in the presence of PS-containing membranes. This is what we have observed.²⁵ 

Our results indicate that PS not only helps localize FVa to the membrane surface but also serves to enhance cofactor activity through enhanced FXa binding affinity. Both the C1 and C2 domains are most probably important during the initial phases of hemostasis when concentrations of FXa and exposed PS may be limiting.⁴²  Under these conditions, the C2 domain is needed to locate FVa to a platelet membrane. In addition, conformational changes in the C1 domain resulting from the binding of PS presumably result in conformational changes elsewhere in FVa that lead to enhanced FXa binding. FXa-binding sites in FVa have been localized by peptide inhibition studies to the A1/A2 domain interface,^43,44  by loop swapping to the A2 domain,⁴⁵  by limited proteolysis to the A2 domain,⁴⁶  and finally by site-directed mutagenesis to the A2 and A3 domains of FVa.⁴⁷  The latter mutagenesis study identifies His¹⁶⁸³ as the only A3 domain residue key to FXa binding. This is a surface-located histidine not far from the A2/A3 interface. Because all other studies identify residues in the A2 domain (or A1/A2 interface) as involved in FXa binding, it seems probable that the binding of C6PS to the C1 domain may lead to conformational changes in the adjacent A3 domain that then transfer the signal to the A2 domain to produce a high-affinity FXa binding configuration.

In conclusion, if mutations in the C1 domain that seriously inhibit binding of PS to this domain fail to block prothrombinase complexes on membranes containing 25% PS, what can be the importance of this domain physiologically? The answer lies in the fact that the platelet membrane contains only approximately 10% PS,⁴⁸  and it is probable that the activated platelet vesicles that support coagulation^49,50  expose only 5% to 10% PS on their surfaces. On membranes containing only 5% PS, C1 mutant rFVa is dramatically ineffective in supporting prothrombin activation.²³  From this observation and the results presented here, it is probable that PS binding to the C1 domain of FVa plays a key regulatory role in blood coagulation.

The authors thank reviewers of previous versions of this manuscript for anonymous comments.

This work was supported by National Institutes of Health, Bethesda, MD (grants HL72827, B.R.L.; and HL43106, W.H.K.).

National Institutes of Health

Contribution: R.M. and B.R.L. designed experiments; R.M. performed experiments, collected, analyzed, and interpreted data, performed statistical analysis, and wrote the initial manuscript; M.A.Q.-A. expressed, purified, and characterized the recombinant proteins; and W.H.K. and B.R.L. sponsored the research, helped interpret the data, and revised the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Barry R. Lentz, Department of Biochemistry and Biophysics, CB#7260, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7260; e-mail: uncbrl@med.unc.edu.