Phosphatidylserine and phosphatidylethanolamine regulate the structure and function of FVIIa and its interaction with soluble tissue factor

An essential feature of blood coagulation is the formation of complexes between serine protease coagulation factors and their corresponding membrane-bound cofactors. Tissue factor (TF) is a 47-kDa protein expressed on the surface of several types of cells that do not contact blood [1]. Tissue damage exposes TF to blood; exposed TF binds to factor VIIa (FVIIa) to form the TF–FVIIa complex in which membrane-bound TF is the FVIIa cofactor [2]. The activated platelet surface provides the membrane lipids that are involved in amplification and propagation phases of coagulation, but damaged endothelial cell membranes are the primary membranes available during the initiation phase [3,4]. Both types of membranes expose the lipid phosphatidylserine (PS), but the latter membrane type is the site of action of the TF–FVIIa complex [3]. Although FX activation by TF–FVIIa is complicated by FVIIa auto-activation [5,6], the rate of FX activation by FVIIa–TF on the endothelial cell membrane surface is enhanced by ∼10⁴–10⁵-fold relative to activation by FVIIa alone [5]. Thus, the endothelial cell membrane has a key function in the complicated process of initiation of blood coagulation.

The participation of membranes and lipids in TF–FVIIa activation of FX (or FIX) has long been recognized. However, unlike the function of membranes and lipids in FXa-FVa activation of prothrombin [7] and in FIXa-FVIIIa activation of FX [8], the actions of membranes and lipids are enigmatic regarding TF–FVIIa activity. FVIIa is structurally similar to FXa and FIXa [9] and, like FXa and FIXa, FVIIa binding to membranes requires an N-terminal γ-carboxyglutamic acid-rich domain (Gla domain). However, FXa and FIXa both have reversibly bound extrinsic protein cofactors (FVa and FVIIIa), whereas TF is a Type I intrinsic protein cofactor permanently anchored to the endothelial cell membrane by a trans-membrane domain [2]. This unique feature of TF has impeded characterization of the activities of specific lipids in regulating the FVIIa (FVII)–TF complex.

The serine protease and cofactor activities of the FXa–FVa and FIXa–FVIIIa complexes are regulated mainly by the platelet PS exposed on activated platelets [8,10] and, to a lesser extent, by phosphatidylethanolamine (PE) [7]. Short chain soluble (i.e., having a high critical micelle concentration (CMC)) phospholipids have enabled dissection of membrane interactions with FXa and FVa. Dicaproyl-PS (1,2-dicaproyl-sn-glycero-3-phospho-l-serine, C6PS), a stand-in for PS, binds to specific sites on FXa to increase its activity [7] and to a PS-specific site in the C1 domain of FVa to enhance its affinity for FXa [11]; thus, C6PS enables assembly of an active FXa–FVa complex in solution [10]. C6PS and 1,2-dicaproyl-sn-glycero-3-phospho-ethanolamine (C6PE) (a PE substitute) do not occur physiologically, but they are powerful tools for locating and characterizing the specificities and the physiological functions of lipid-regulatory sites in FXa, FIXa, and their cofactors FVa and FVIIIa [7,11–13].

In contrast, TF must be studied in native or reconstituted membranes. Although investigators have recognized the important function of PS in FXa activation by FVIIa–TF, there are contradictory reports about the mechanism of the effects of PS on activation of FX by FVIIa–TF. One view is that the membrane has no activity in delivering the substrate (FX) to the complex [13,14]. Another view is that interaction of FX with a PS-containing membrane delivers FX to the complex in an appropriate conformation to maximize k_cat,app [14]. This second view is supported by observations that FXa undergoes significant conformational changes on binding to membranes [8,15] or on binding to soluble PS in solution [16]. This view also appears to be consistent with a crystal structure of FVIIa bound to a soluble recombinant form of TF (residues 1–219; rsTF) [17] that was cleaved by subtilisin to promote crystallization [18].

Besides the surface delivery of FX or removal of FXa, the endothelial membrane might also contribute to FVIIa activity by properly aligning FX with the FVIIa active site or by regulating FVIIa–TF complex formation. These mechanisms contribute to FXa–FVa activity on platelet-derived membranes [10,11,19,20]. Thus, to decipher the exact function(s) of phospholipid in regulating FVIIa activity, we used soluble short-chain phospholipids to address the following questions: (1) Do C6PS or C6PE increase the activity of FVIIa in solution? (2) Do C6PS or C6PE increase affinity of soluble TF (rsTF) for FVIIa in solution? (3) Is binding of C6PS or C6PE to rsTF required for activity of the FVIIa–rsTF complex in solution? Our results are summarized as follows:

1. Short-chain soluble PS and PE bind to FVIIa with an apparent dissociation constant of 150 ± 100 μM, as measured by FVIIa structural changes, proteolytic activity, and amidolytic activity. Fitting several datasets, we identified two lipid binding sites on FVIIa, that bind either soluble PS or PE with similar dissociation constants.

2. Saturating proteolytic activity of FVIIa was approximately 5-times greater in the presence of soluble PS compared with soluble PE, and approximately 15-times greater when both lipids were present. Thus, the two lipids appeared to act synergistically. However, the proteolytic activity of FVIIa was highest, by approximately 100 times, in the presence of rsTF and both lipids.

3. Soluble TF binds to FVIIa and Gla domainless FVIIa (des Gla-FVIIa) with similar dissociation constants of approximately 2 nM at saturating concentrations of soluble lipids. However, the saturating proteolytic activity of FVIIa was approximately 200-times higher than the activity of the des Gla-FVIIa.

In conclusion, soluble lipids (PS and PE) facilitate tight binding (K_d = 2 nM) of both FVIIa and des Gla-FVIIa to rsTF; however, the Gla domain of FVIIa is essential for the FVIIa–rsTF complex to achieve physiologically relevant activity.

Human FVIIa and des Gla-FVIIa were purchased from Enzyme Research Laboratories (South Bend, IN). rsTF was a generous gift from Professor James Morrissey (University of Michigan). The sodium salts of C6PS and C6PE were purchased from Avanti Polar Lipids (Alabaster, AL). The FXa-specific substrate N-2-benzyloxycarbonyl-d-arginyl-l-arginine p-nitroanilide dihydrochloride (S-2765) was purchased from Diapharma (West Chester, OH). All other chemicals were ACS reagent grade or the best available grade; all solvents were HPLC grade.

Intrinsic tryptophan fluorescence of FVIIa as a function of increasing concentration of soluble lipid was measured as described in [20] using a SPEX® FluoroLog-3 spectrofluorometer (Horiba Jobin Yvon Inc., Edison, NJ) with excitation wavelength at 285 nm (band-pass 4 nm) and emission recorded at 340 nm (band-pass 4 nm).

Proteolytic activity of FVIIa was followed by measuring the rate of generation of FXa from FX according to the method described in [12] with slight modification. Briefly, 100 nM FVIIa was incubated with different concentrations of C6PS in appropriate volume of buffer (20 mM Tris with 150 mM NaCl, 5 mM CaCl₂ and 0.6% PEG, pH 7.5) in the wells of a microtiter plate at 37°C for 5 min. A total of 300 nM FX and 1 mM of FX specific synthetic substrate S-2765 was added to it and absorbance at 405 nm was recorded for 5 min using a VersaMax tunable micro plate reader (Molecular Devices) at 37°C. The amount of FXa produced was determined from the initial rate of substrate cleavage using a standard curve prepared with active-site-titrated FXa as described earlier [21].

The amidolytic activity (absorbance versus time) of FVIIa in the presence of various concentrations of C6PS or C6PE was determined using a synthetic chromogenic substrate, Leu-PHG-Arg-pNA.AcOH (Aniara, West Chester, OH) according to the method described earlier with slight modifications [22]. Amidolytic activity was measured in buffer (20 mM Tris with 150 mM NaCl, 5 mM CaCl₂ and 0.6% PEG, pH 7.5) containing human FVIIa (100 nM), 1 mM chromogenic substrate, and varying concentrations of C6PS. The time course of absorbance yielded the amidolytic activity of FVIIa. A small volume of the assay buffer was first incubated in a 96-well microplate at 37°C. Human FVIIa was next allowed to bind C6PS or PE by adding both to the buffer in the well and equilibrated for 5 min at 37°C. Finally, the substrate was added to that at a final concentration of 1 mM. The absorbance at 405 nm was recorded every 20 s for 5 min using a VersaMax tunable microplate reader (Molecular Devices) to obtain the initial rate of substrate hydrolysis. Triplicate determinations were obtained for each lipid concentration in each of two independent experiments. Activity in the presence of lipids was recorded as a percent of activity in the absence of lipid.

Binding of lipid to FVIIa was assessed indirectly by recording the appropriate observable (proteolytic activity, amidolytic activity, intrinsic fluorescence) as a function of soluble lipid (C6PS or C6PE) concentration over a broad range below the CMC of the soluble lipid. The CMC was independently assessed for each conditions following pyrene fluorescence as a function of increasing lipid concentration [23]. Simple binding models were fitted to these experimental data to estimate the apparent dissociation constants for binding of FVIIa to C6PS or C6PE, as described in the Appendix, using the appropriate assumption that [C6PX]_free ≈ [C6PX] [7]. Analysis of binding curves is described in detail in the Appendix.

The stoichiometry of soluble C6PS and C6PE binding to FVIIa was estimated by equilibrium dialysis measurements. Experiments were performed using 2.0 ml Teflon dialysis cells (Harvard Apparatus, Holliston, MA) with the two cells separated by a 2000 molecular weight cut-off membrane. The total lipid concentration, [L]_tot, and total protein concentration, [P]_tot, must be large enough to allow accurate determination of differences in organic phosphorous between the two chambers but [L]_tot must be below the lipid CMC.

The details of the computational procedures are described in the ‘Results’ section.

The intrinsic tryptophan fluorescence intensity of FVIIa decreased with the addition of increasing amounts of C6PS and C6PE (Figure 1). The binding curves in both cases were reasonably well described by a single-site binding model (see Appendix), from which estimated apparent dissociation constants (K_d’s) for binding of C6PS and C6PE were found to be 160 ± 30 and 154 ± 14 µM, respectively (Appendix, Table A1). Clearly, the intrinsic fluorescence of FVIIa responds equivalently to both lipids. This equal response could mean either (1) that a common site recognizes the two lipids interchangeably (same K_d’s and responses), or (2) that multiple distinct sites recognize these two lipids with equivalent K_d’s and responses. To distinguish between these possibilities, we partially titrated FVIIa with C6PS and then further titrated with C6PE (Frame A of Figure 1) or performed the reverse experiment by titrating with C6PS after titrating with C6PE (Frame B of Figure 1). Fitting and analyses of the data from these two mixed-lipid experiments revealed comparable apparent K_d’s and changes in intrinsic tryptophan fluorences ΔF_sat for both single-lipid and two-lipid titrations for both lipids. These results support the interpretation that the fluorescence-altering site or sites is/are equally well occupied by C6PS and C6PE, with similar ensuing responses, but they do not establish the number of sites.

We also tested other soluble lipids, such as 1,2-dicaproyl-sn-glycero-3-phosphocholine (C6PC) and C6DPS, and observed very week binding (data not shown). These other lipids were also found to have no effect on the proteolytic activity of full-length FVIIa, indicating that the influence of soluble lipids on FVIIa activity is specific.

As for intact FVIIa, two type of experiments were also performed with a des Gla–FVIIa (Figure 1). C6PS and C6PE bind to des Gla–FVIIa with roughly the same apparent K_d’s (319 ± 83 µM for C6PS and 326 ± 47 µM for C6PE) (Frames C and D of Figure 1 and Table A2 of Appendix). Absolute fluorescence of des Gla-FVIIa were not much different from the full-length FVIIa, however, the mixed lipid titration presented somewhat different result. Unlike for the intact protein, elimination of the Gla domain resulted in a slight discontinuity in the fluorescence response when the concentration of one lipid was fixed but titration continued with the other lipid. The single site model introduced for describing the data in Figure 1 (Eq. 3 in the Appendix) could not describe the discontinuity in the datasets for mixed lipid titrations with des Gla–FVIIa. The failure of this model rules out the possibility that two-independent sites on des Gla–FVIIa responds equally to PS and to PE, as we found for intact FVIIa. The single-site model described well the initial portions of each titration in which only C6PS (Frame C, Figure 1) or C6PE (Frame D, Figure 1) was added to des Gla–FVIIa. The parameters defined by these single-site fits to the initial single-lipid titrations are summarized in the first and last rows in Table A2 in the Appendix.

Having demonstrated sensitivity of FVIIa and des Gla-FVIIa structure towards soluble lipid binding with apparent K_d’s of approximately 150–300 µM, (as reflected in intrinsic fluorescence), we next monitored conversion of FX (300 nM) to FXa by FVIIa (100 nM) in the presence of increasing concentrations of C6PS and C6PE (Figure 2A,B). Proteolytic activity increased with titration by C6PS or C6PE alone with apparent K_d’s of 72 ± 30 µM for C6PS and 83 ± 106 μM for C6PE (Table A3). When we performed this experiment with initial addition of 150 µM C6PE followed by titration with C6PS, the apparent K_d was essentially the same (84 ± 22 µM) with a much larger change in saturating activity (9.5 ± 1 nM/min). Titration with C6PE in Frame A at a fixed concentration of C6PS (100 µM) yielded an apparent K_d of 83 ± 40 µM with a change in saturating activity of 12 ± 2 nM/min) (Table A3). Clearly, C6PS was a tad more effective than C6PE in promoting proteolytic activity of FVIIa, but the highest proteolytic activity of FVIIa was achieved in presence of both lipids simultaneously. This indicates the cooperative influence of C6PS and C6PE in promoting the proteolytic activity of FVIIa.

Next, we considered the influence of rsTF on the response of FVIIa proteolytic activity to C6PS and C6PE. These data are shown in Figure 2C with results recorded in Table A3. Once again, the solid lines through the data represent a single-site binding model from which we estimated apparent K_d = 43 ± 3 µM, with saturating activity a = 29.1 ± 0.6 nM/min for PS; and K_d = 72 ± 8 µM, with saturating activity a = 23.9 ± 0.7 nM/min for PE. Thus, it is clear that rsTF enhanced the activation by ∼100-fold of FVIIa by both lipids and eliminated the somewhat larger influence of C6PS. As for FXa and FIXa [7,10–12,13,,20], the FVIIa cofactor is also regulated by lipids. Note that saturating activity of approximately 30 nM/min was obtained in presence of 10 nM rsTF, and 2.5 nM FVIIa (40-times lower than in frames A and B), and 150 nM FX (2-times lower than in frames A and B).

FVIIa interacts with membrane-located TF much more tightly as PS content increases [24]. The result in Figure 2C suggests a lipid-dependent interaction between FVIIa and rsTF even in solution. To estimate the strength of this interaction, we titrated the effect of rsTF on FVIIa (2.5 nM) activation of 150 nM FX in the presence of 100 μM C6PS or C6PE, with the results recorded in Figure 3A. Addition of rsTF to FVIIa enhanced the ability of both C6PS (•) and C6PE (○) to promote FVIIa proteolytic activity in activation of FX. Soluble lipids alone promoted FVIIa proteolytic activity (Figure 2, Frames A and B), reaching saturating activity of approximately 5, and 1 nM/min in presence of saturating amount of C6PS and C6PE, respectively. Addition of rsTF enhanced the lipid mediated proteolytic activities to approximately 30 and 20 nM/min, respectively; however, at 40-times lower amount of FVIIa and 2-times lower amount of FX, thus increasing saturating activity more than 100-fold. The affinity of rsTF for FVIIa in the presence of C6PE (K_d = 1.3 ± 0.5 nM) was observed to be similar to that in the presence of C6PS (K_d = 2.6 ± 0.6 nM) [25]. Next, we collected data with FVIIa and sTF in the presence of a fixed concentration of C6PS followed by the titration with increasing concentrations of C6PE and vice versa. Our result (data not shown here) indicate cooperativity of FVIIa activation of FX in the presence of rsTF similarly like in the absence of rsTF.

A reasonable interpretation of this observation is that these two amino phospholipids regulate the cofactor activity of TF by promoting its tight binding to FVIIa. Since we observed that the Gla domain of FVIIa is important in determining both its activity and structural response to both C6PS and C6PE (Figures 1 and 2), we also examined the effect of the Gla domain on binding to rsTF, as recorded in Figure 3B. The absence of the FVIIa Gla domain reduced the affinity of rsTF for FVIIa only by roughly two-fold but reduced the increase in proteolytic activity by 70–80-fold. Thus, the FVIIa Gla domain is needed to form an optimal enzyme–cofactor–substrate complex.

Proteolytic activity depends on binding of substrate (in this case FX) to both the active site and possibly to substrate-binding exosites [26]. In order to discern whether the lipid-induced increase in FVIIa proteolytic activity is due to an influence of lipid on the FVIIa active site or to binding of the substrate (FX) to an exosite, we titrated FVIIa amidolytic activity towards the synthetic substrate Leu-PHG-Arg-pNA.AcOH with both C6PS and C6PE. The results are shown in Figure 4. Analysis of the curves in Figure 4A is described in the Appendix according to a single-site model, with resulting parameters gathered in Table A5. Comparing Tables A4 and A5 makes it clear that FVIIa amidolytic and proteolytic activities respond with similar apparent K_d’s to both C6PS and C6PE binding, although the relative amidolytic response is negative (decreases upon binding soluble lipids) while the proteolytic response is positive. rsTF had essentially no effect on the amidolytic response to soluble lipid binding while the proteolytic activity was increased 5-fold and 20-fold by C6PS and C6PE binding respectively. This suggests that association of FVIIa with rsTF enhances activity by promoting formation of a substrate-binding exosite rather than solely through enhancing active-site activity.

Our data show that PS and PE cause structural changes to FVIIa with apparent dissociation constant K_d ∼ 150–300 μM (Figure 1) and affect proteolytic and amidolytic activity with apparent K_d ∼ 50–100 μM (Figures 2 and 4). Since it is not possible to determine the number of PS/PE binding sites from single datasets, we analyzed all datasets simultaneously with a global fit, where each parameter in a model, such as K_d, has to be equal for all datasets simultaneously. The details of the analysis are described below.

We used the two-independent-site model, where we assumed that (n₁ + n₂) C6PS or C6PE molecules bind to two classes of FVIIa sites, as we did previously for C₆PS binding to FIXa [12]. The model is briefly summarized here. For each class of sites, the site dissociation constants are assumed to be identical within that class (K_d,1 and K_d,2) and the sites are assumed to be independent. The data in Figures 1, 2C and 3 support this model, although these data cannot establish both the site-dissociation constants and stoichiometry of these two classes of sites when fit individually. However, a simultaneous fit of these three independent experimental datasets along with equilibrium dialysis data (Figure 5E and 6E) can establish the dissociation constants and stoichiometries. The stoichiometries of C6PS or C6PE binding to FVIIa must be natural numbers; therefore, we treated them as such in order to obtain the best fit. In addition, we also treated the stoichiometries as variable parameters, to test the appropriateness of the model.

Four processes define occupation of these sites:

1. binding of n₁ PS molecules to site 1 of FVIIa to produce the species $FVIIaPSn11$⁠,

2. binding of n₂ PS molecules to site 2 of FVIIa to produce the species $PSn22FVIIa$⁠,

3. binding of n₂ PS molecules to $FVIIaPSn11$ to produce the species $PSn22FVIIaPSn11$⁠, and

4. binding of n₁ PS molecules to $PSn22FVIIa$ to produce the species $PSn22FVIIaPSn11$⁠.

Because of site independence, there are only two distinct site-binding constants. In (eqns 1 and 2), K_d,1 is the equilibrium dissociation constant for C6PS binding to the first class of equivalent and independent PS binding sites on FVIIa, and, in (eqns 3 and 4), K_d,2 is the equilibrium dissociation constant for binding of C6PS binding to the second class of PS binding sites.

We applied the two-independent-site model to simultaneously fit four independent experimental datasets presented in (Figures 1, 2C, 3 and 5E) to determine stoichiometry of these two classes of sites and the dissociation constants.

We solved here a set of equations numerically using MATLAB, version R2011b for any given set of parameters (n₁, n₂, K₁, and K₂). Using the equations presented above, we predicted: (a) concentrations of each FVIIa species, (b) initial rates of FX activation by FVIIa, (c) amydolitic activity of FVIIa, (d) intrinsic tryptophan fluorescence of FVIIa, and (e) ΔP as a function of total soluble lipid concentration. This requires 16 additional unknown intensive parameters (molar quantities in eqns 3–7), but these can be fixed by observation as previously in the case of FIXa [12]. First, we note that only the weak class of sites affects structural properties, so F1 and F2 (= F4) are fixed as the inital and asymptotic observables for both intrinsic fluorescence and, while F3 = F1 (i.e., occupancy of the tighter class of sites does not influence intrinsic fluorescence, Figure 1). Second, only occupancy of the tighter class of sites affects activity, so k_cat,1 and k_cat,3 (=k_cat,4) are establihed as the inital and asymptotic values from Figure 3 for both proteolytic and amidolytic activity, while k_cat,2 = k_cat,1, i.e., the activity of free FVIIa was set to be negligible, i.e. 0. There are no extensive parameters associated with the observable ΔP. These predicted observables were compared with the experimental data to calculate the sum of the squares of deviations of calculated values from experimental data (χ²). We minimized this with respect to the unknown dissociation constants and stoichiometries to obtain estimates of these parameters that allow us to describe all four datasets simultaneously (i.e., globally). Dividing the minimized χ² by the number of degrees of freedom (#data points - #parameters -1) yields the reduced chi-squared (⁠$χ2¯$⁠), which should be close to 1 for an adequate description of the data by the model (if number of data points is much larger than the number of fitting parameters). For simplicity, all observables were expressed as ratios of values in the presence of C6PS to those for free FVIIa, i.e., the fluorescence recorded at the beginning of each experiment. Best fits with different assumed stoichiometries n₁ and n₂ are summarized in Table 1 (for C6PS) and Table 2 (for C6PE).

Majumder et al. have previously shown that two molecules of PS bind at two different sites on FIXa with different affinities (∼100-fold) that they appear to be independent sites [12]. A high affinity binding site (K_d ∼ 1.4 µM) regulates structure, whereas a low-affinity binding site (K_d ∼ 140 µM) regulates activity. Similarly, the activity-regulating PS binding site in FXa has dissociation constant K_d ∼ 80 μM while the other site that regulates structure has a K_d (200–600) µM [16,19].

The best global fit to all datasets, including equilibrium dialysis measurements (Figure 5 for PE, and Figure 6 for PS), indicates that FVIIa is regulated, just like FIXa, by independent binding of two molecules of PS. One binding site regulates activity (K_d ∼ 90 μM), while the other one regulates structure (K_d ∼ 160 μM). Different parameters of structure and activity regulation of factors VIIa, IXa and Xa by the soluble lipids are documented in Table 3.

In summary, there are apparent differences in structural response of FVIIa, FIXa, and FXa by soluble short-chain lipids, whereas binding of the lipids to activity regulating site seems to be preserved (Table 3). Understanding of soluble lipid (lysolipid) regulation of serine proteases might be of great physiological significance, since lysophosphatidylserine (lyso-PS) is involved in coagulation and in early stages of initiating acute inflammation as well as in its resolution [28].

The present study, which is mainly focused on exploring the details of the soluble lipid binding to FVIIa, revealed new and unexplored facts. We studied the interaction of soluble lipids (mainly PS and PE) with FVIIa and FVIIa–rsTF without the presence of any membrane surface. Mostly, previous studies described the interaction of FVIIa with phospholipid membrane in the presence of its cofactor TF. However, our study is the first to describe the interaction of soluble phospholipid with FVIIa alone and to decipher the particular roles of phospholipids in regulating FVIIa.

In addition, our study will be useful because C6PS and C6PE are a perfect tool to study the effects of lysolipids. Lysolipids are an emerging class of signaling lipids involved in inflammatory and autoimmune diseases [28]. Activated platelets secrete Lyso-PS [29] and lyso-PE [30] in vitro. Lyso-PS may be involved specifically in platelet activation in vivo, [31] yet its physiological function has not been fully elucidated. Lysolipids occur in relatively high concentrations in blood plasma. However, it is difficult to investigate lysolipid activities in vitro because they tend to aggregate due to their sub-micromolar CMCs. Conversely, short-chain soluble lipids such as C6PS and C6PE have much higher CMCs, thus, they can be used to probe the lysoPS-binding site in the Gla domain. This site binds the lyso-PS headgroup and only the first four to six carbon atoms of the lyso-PS alkyl chain. Binding involves multiple types of interactions, e.g., calcium coordination, ionic bonds, and van der Waals and hydrophobic contacts [29].

The intrinsic tryptophan fluorescence of FVIIa in the presence of both C6PS and C6PE resulted in a simple hyperbola and we used a single-site binding model (Eqn 3 in Appendix) to fit the data. However, we could not detect any noticeable change in the overall secondary structure of FVIIa in the presence of soluble lipids from the CD spectroscopic measurement (data not shown). Notably, Taboureau et al. also reported no perturbation in FVIIa due to the insertion of C6Lyso-PS [30]. Both single and two lipid titrations with des Gla-FVIIa revealed the fact that at least one of the two lipid binding sites exists in the Gla domain.

Apparently, FVIIa has two different binding sites for PS and PE that regulate the activity of FVIIa. However, both PS and PE bind with similar affinity to their respective sites. When both sites are occupied, the activity of FVIIa is greater than with only one site occupied. Thus, PE in the presence of PS was most effective in enhancing the proteolytic effect. Increased FX activation in presence of PE containing membrane with little amount of PS (as low as 3 mol%) was also reported by Neuenschwander et al.; although PE dependency slowly diminished as PS concentration was increased [31].

Our data confirmed that the K_d’s of binding to these activity-regulating sites are different from those determined by tryptophan fluorescence experiments. This indicates that the site regulating the activity of FVIIa and the site that regulates FVIIa structure are not same. Effect of PS and PE on FVIIa proteolytic activity was shown to be enhanced appreciably in presence of rsTF (Figure 2C). This might be due to stronger interaction of the enzyme (FVIIa) and cofactor (TF) in presence of the phospholipids as was observed for other coagulation factors too [10,11,20,32,33]. Removal of Gla domain, however, made FVIIa inactive with the affinity of rsTF towards des Gla–FVIIa (K_d = 2 nM) remained almost same. This clearly indicates that both PS and PE promote strong interaction between rsTF and des Gla-FVIIa by binding to rsTF rather than to FVIIa. Therefore, most of the free energy of interaction of FVIIa and rsTF is independent of the presence of the Gla domain. Lawson et al. also demonstrated that the increase in catalytic efficiency of FVIIa when complexed with TF was independent of the supporting surface [34]. A previous study showed that the rsTF–FVIIa complex did not bind to surface immobilized, mixed phosphatidylcholine-PS [17]. We made a similar observation with the soluble phospholipid.

Therefore, although the Gla domain of FVIIa is absolutely required for binding of soluble phospholipid to activate FVIIa, that domain is not necessary for binding of rsTF to FVIIa. When we determined the amidolytic activity of FVIIa in the presence of increasing concentration of PS and PE, we obtained site dissociation constants similar to those obtained from proteolytic activity measurements (Figure 4). We also observed PS to be more effective in promoting the activity of FVIIa compared with PE. When we performed the proteolytic activity assay in the presence of rsTF, we did not observe any change in apparent K_d’s (68 ± 5 μM with C6PS and 83 ± 9 µM with C6PE). This suggests that association of FVIIa with rsTF enhances proteolytic activity mostly by promoting the binding of the substrate (FXa) to FVIIa exosite, and not by regulating the active site activity of FVIIa. This was supported further by the molecular dynamics (MD) simulation studies of the membrane-bound model of FVIIa–sTF complex [34]. MD simulation studies on membrane bound TF–FVIIa revealed that sTF interacts directly with the lipid headgroups and stabilizes the large-scale structural fluctuations of FVIIa due to the presence of the hinge regions between its domains. In the process, it also positions the catalytic triad of FVIIa in a more upright position with respect to the membrane than in isolated FVIIa. This spatial stabilization of the catalytic region of FVIIa by sTF ensures optimal interaction of FVIIa with its substrate FX. Reorientation of a clotting factor with respect to the membrane, induced by a cofactor, was also reported for the FXa-FVa complex [15]. In that case, C6PS bound both FXa and FVa to form an active complex. However, unlike FVIIa, FXa has three plausible C6PS binding sites [35] Apart from the one present in the Gla domain, the other two lipid binding sites are present in the epidermal growth factor domain and the catalytic domain. Binding of lipids to these three widely separated structural domains of FXa involves linkage between the sites. Similar regulation by C6PS was also observed for another serine protease of the coagulation cascade, FIXa. FIXa has two PS binding sites, each having a stoichiometry of 1 [12]. However, whether FVIIIa, the cofactor of FIXa, binds to lipid and brings in structural changes in FIXa is yet to be explored.

In summary, we have successfully used soluble lipids as an alternative for a phospholipid membrane to establish the importance of PS and PE in blood coagulation. The present study revealed that molecular PS and PE regulate FVIIa activity, but, unlike the Xa–Va complex, neither PS nor PE influence FVIIa–rsTF complex formation. Our conclusions are enumerated below:

1. PS and PE bind to one or more sites on FVIIa with equal affinities.

2. At least one of the lipid binding sites exists in Gla doman.

3. PS and PE act cooperatively to influence the proteolytic activity of FVIIa.

4. There is huge reduction in proteolytic activity of FVIIa without its Gla domain.

5. Soluble lipids trigger a strong association of rsTF, to FVIIa,

6. Increased proteolytic activity of FVIIa in presence of rsTF can be attributed to cofactor (TF) induced conformational change in FVIIa to facilitate its binding to its substrate, FXa.

Finally, the essential findings of our study are summarized in Figure 7.

All supporting data are included in the main text and supplementary files.

The authors declare that there are no competing interests associated with the manuscript.

This work was supported by the NIH [grant numbers R01HL072827 (to B.R.L.), R01HL118557 (to R.M.)].

T.S. and T.K. performed the experiments and T.S. wrote the initial draft of the manuscript. B.R.L. and R.M. conceptualized the work and R.M. designed the experiments. R.M. and B.R.L. reviewed and edited the manuscript.

We thank Professor James Morrissey from University of Michigan for providing us the rsTF.

CMC

critical micelle concentration

C6PS

1,2-dicaproyl-sn-glycero-3-phospho-l-serine

C6PE

1,2-dicaproyl-sn-glycero-3-phospho-ethanolamine

des Gla-FVIIa

Gla-domainless factor VIIa

FVIIa

factor VIIa

FIX

factor IX

FX

factor X

Gla

γ-carboxyglutamic acid-rich domain

lyso-PS

lysophosphatidylserine

MD

molecular dynamics

PE

phosphatidylethanolamine

PS

phosphatidylserine

rsTF

recombinant soluble tissue factor

sTF

soluble tissue factor

S-2765

N-2-benzyloxycarbonyl-d-arginyl-l-arginine p-nitroanilide dihydrochloride

TF

tissue factor