Modeling of human factor Va inactivation by activated protein C
© Bravo et al.; licensee BioMed Central Ltd. 2012
Received: 15 February 2012
Accepted: 20 May 2012
Published: 20 May 2012
Skip to main content
© Bravo et al.; licensee BioMed Central Ltd. 2012
Received: 15 February 2012
Accepted: 20 May 2012
Published: 20 May 2012
Because understanding of the inventory, connectivity and dynamics of the components characterizing the process of coagulation is relatively mature, it has become an attractive target for physiochemical modeling. Such models can potentially improve the design of therapeutics. The prothrombinase complex (composed of the protease factor (F)Xa and its cofactor FVa) plays a central role in this network as the main producer of thrombin, which catalyses both the activation of platelets and the conversion of fibrinogen to fibrin, the main substances of a clot. A key negative feedback loop that prevents clot propagation beyond the site of injury is the thrombin-dependent generation of activated protein C (APC), an enzyme that inactivates FVa, thus neutralizing the prothrombinase complex. APC inactivation of FVa is complex, involving the production of partially active intermediates and “protection” of FVa from APC by both FXa and prothrombin. An empirically validated mathematical model of this process would be useful in advancing the predictive capacity of comprehensive models of coagulation.
A model of human APC inactivation of prothrombinase was constructed in a stepwise fashion by analyzing time courses of FVa inactivation in empirical reaction systems with increasing number of interacting components and generating corresponding model constructs of each reaction system. Reaction mechanisms, rate constants and equilibrium constants informing these model constructs were initially derived from various research groups reporting on APC inactivation of FVa in isolation, or in the presence of FXa or prothrombin. Model predictions were assessed against empirical data measuring the appearance and disappearance of multiple FVa degradation intermediates as well as prothrombinase activity changes, with plasma proteins derived from multiple preparations. Our work integrates previously published findings and through the cooperative analysis of in vitro experiments and mathematical constructs we are able to produce a final validated model that includes 24 chemical reactions and interactions with 14 unique rate constants which describe the flux in concentrations of 24 species.
This study highlights the complexity of the inactivation process and provides a module of equations describing the Protein C pathway that can be integrated into existing comprehensive mathematical models describing tissue factor initiated coagulation.
It is critical that once a procoagulant response to vascular injury has been initiated that an appropriate anticoagulant response is concurrently mounted. As such, the components of the prothrombinase complex are targets of multiple regulatory mechanisms to terminate the production of thrombin. First, FXa availability is regulated by formation of inhibition complexes with antithrombin (AT), α1-antitrypsin, and tissue factor pathway inhibitor (TFPI) [10, 11]. The cofactor, FVa, is a target for degradation by activated protein C (APC). APC, a serine protease derived from its plasma precursor protein C (PC) in a thrombin dependent process , plays a critical role in inactivating the non-enzymatic cofactor components of both the prothrombinase and the intrinsic tenase complexes, factors Va and VIIIa, respectively [13–15]. Additional key components of the PC pathway include: thrombomodulin and endothelial protein C receptor (EPCR) which contribute to APC formation ; protein S which functions as a cofactor enhancing APC efficacy ; and protein C inhibitor, a suppressor of APC formation [18, 19].
An increased risk of thrombotic disease has been associated with partial deficiencies or loss of function mutations in the PC pathway, including deficiencies in PC, its cofactor protein S, or proteins involved in the activation of PC . The most prevalent defect in the PC pathway is a result of a genetic mutation in FV that renders the cofactor resistant to APC inactivation . This resistance to APC was first characterized by Dahlbäck and coworkers  and derives from a mutation at one of the APC inactivation sites (Arg506 → Gln506) on FV/FVa rendering it resistant to APC cleavage .
APC catalyzed inactivation of FVa cofactor activity is a complex, membrane dependent process involving cleavage at multiple sites in the FVaHC, the generation of transient species with partial cofactor activity, and the ultimate disassociation of a fragment of the FVaHC from the molecule rendering it catalytically inactive. APC interacts with FVa or partially proteolyzed FVa species through their light chains to form enzymesubstrate complexes [23–25]. Additional studies have identified other regions of interaction, including the proteolytic target residues and the surrounding regions, in the FVa heavy chain involved in the enzymesubstrate complex formation [26–28]. Three arginine residues are targeted in human APCFVa complexes: Arg306, Arg506, and Arg679[15, 29]. A number of studies have defined the activities of the partially proteolyzed FVa species [30, 31]. Cleavage at either Arg306 or Arg506 results in cofactor species, FVai 306 and FVai 506, respectively, with reduced but similar cofactor activity in the prothrombinase complex (Figure 1) [30, 31]. Cleavage at Arg306 has been shown to be essential for full loss of cofactor activity . The significance of the cleavage at Arg679 remains undetermined. The final inactive cofactor, FVai, is a two chain molecule composed of the FVaLC and FVa1-306 fragment (Figure 1) . The inactive cofactor binds APC with the same affinity as the intact cofactor [23, 33], suggesting a potential role of product inhibition in regulating APC efficacy .
Numerous studies examining how components like FXa, prothrombin and protein S modulate APC inactivation of FVa have been reported [14, 34–40]. These studies routinely employ approaches like progress curve analysis based on curve fitting to compare initial rates, or natural and recombinant cleavage site mutants to detail mechanistic features. To date, however, an integrated approach, one combining the use of physicochemical model construction based on prior research and model verification via construction of corresponding reaction systems using purified proteins, has not been extended to understand APC regulation of the prothrombinase complex stability and function. In this study we build on the prior studies modeling APC inactivation of FVa  by incorporating the presence of additional components of the prothrombinase complex (FXa and prothrombin) to construct an empirically validated mathematical model. To test model constructs, a step-wise approach to increasing the number of components was used. Experiments were replicated multiple times with different preparations of proteins in order to generate robust data sets that included estimates of measurement error. At each level of complexity, we empirically monitored multiple analytes, including starting reactants, intermediates and final products to provide multiple points of comparison to evaluate each model’s performance.
Western blot analyses (representative blot in Figure 3, Panel B) monitoring the disappearance of intact FVaHC (residues 1–709) display the rapid proteolysis of FVa heavy chain proteolyzed at either site within the first or third minutes with 2.0 or 0.5 nM APC, respectively (Figure 3, Panel C, red and black circles, respectively). These results were each fit to monophasic exponential decays (fit not shown), and the initial slope of the fitted curve was determined after normalizing the densitometry to the maximal value (RMV). At 0.5 nM APC, the initial rate of disappearance of the heavy chain was 200 pM/s (based on 20 nM starting concentration); at 2.0 nM APC, the rate of disappearance was nearly three-fold faster (558 pM/s). The disappearance of the heavy chain is initially concurrent with accumulation of the FVa1-506 fragment (Figure 3, Panel D). The generation of the FVa1-506 fragment reaches a maximum within 30 sec (2.0 nM APC) or 90 sec (0.5 nM APC) (Figure 3, Panel D, red and black circles, respectively). At this point most of the FVaHC has been proteolyzed primarily at Arg506, and the secondary cleavage at Arg306 in the FVa1-506 species becomes the dominant proteolytic event. Analysis of the appearance of the inactivation fragment FVa307-506 (resulting from cleavage at both Arg306 and Arg506, Figure 3, Panel E), indicates that as the FVa1-506 fragment is approaching its maximum level it is being further proteolyzed to the FVa307-506 fragment. In order to isolate the kinetics of proteolysis at Arg306, FVai 506 was used as the starting material in inactivation studies with 0.5 nM APC (Figure 3, Panel F, open circles). This membrane bound partially proteolyzed species has already lost most of its cofactor activity in a clotting assay [30, 31, 33]. Tracking of the final inactivation fragment, FVa307-506, by Western blotting reveals a monophasic accumulation of this fragment (fit not shown) (Figure 3, Panel F).
Equations describing APC inactivation of factor Va
APC + FVa ← 2-1 → APCFVa
1 x 108
APCFVa −3 → APC + FVai 506
APCFVa −4 → APC + FVai 306
APC + FVai 506 ← 2-1 → APCFVai 506
1 x 108
APC + FVai 306 ← 2-1 → APCFVai 306
1 x 108
APCFVai 506 -4 → APC + FVai 306/506
APCFVai 306 -3 → APC + FVai 306/506
FVai 306 -5 → FVa1-306FVaLC + FVa307-679/709
FVai 306/506 -5 → FVa1-306FVaLC + FVa307-506 + FVa507-679/709
APC + FVa1-306FVaLC ← 2-1 → APC FVa1-306FVaLC
1 x 108
However, the predicted generation and disappearance of the FVa1-506 fragment are significantly delayed compared to the empirical data for this inactivation fragment (Figure 3, Panel D, black dashed line vs. black circles). At 0.5 nM APC, the simulation predicts that this fragment will reach a maximum at 3.4 min before slowly disappearing, with 30 % still remaining after 20 min; the empirical experiments show the peak level reached at 1.5 min with only 9 % remaining after 20 min (Figure 3, Panel D, black dashed line vs. black circles). The mathematical simulation fared better with 2.0 nM APC (Figure 3, Panel D, red dashed line vs. red circles) with respect to the timing of peak accumulation, however failed to recapitulate the rapid clearance of this fragment. Mathematical simulations of FVa307-506 using the Hockin et al.  rate constants resulted in an estimated six-fold slower accumulation of this fragment compared to empirical data (Figure 3, Panel E, dashed lines vs. circles).
Though the bovine based Hockin et al.  rate constants clearly captured the disappearance of FVaHC as measured through activity and fragment analysis, they did not accurately predict the presentation of either the FVa1-506 or FVa307-506 fragments in the human system. The rapid clearance of the FVa1-506 fragment and faster appearance of the smaller FVa307-506 fragment led us to speculate that their model rate constant for the cleavage at Arg306 was not appropriate for the human system of proteins. From a comparison of the empirical data to the initial in silico simulation we estimated that the rate of cleavage occurring at Arg306 was three times faster than the model predictions. To account for this the rate constant (Figure 2, k4; Table 1, Eqns 3 and 6) in the mathematical model was adjusted from 0.064 s-1 to 0.192 s-1 and simulations rerun for all analytes (Figure 3, solid lines). As expected the simulated predictions for FVa cofactor activity and FVaHC were minimally altered by this change and continued to correlate well with their respective empirical measures (Figure 3, Panels A and C, solid lines vs. filled circles). The adjusted rate constant significantly altered the in silico predictions for both smaller FVa inactivation fragments (Figure 3, Panels D and E, solid lines vs. dashed lines) yielding a better fit with empirical results. The revised model using the adjusted kinetic rate was extended to experiments conducted with FVai 506 and the predicted time courses for the two models compared to the empirical results (Figure 3, Panel F, solid lines and dashed lines vs. open circles). As can be seen in all simulations, those using the published bovine model values (Figure 3, dashed lines) resulted in a poorer fit . The rate constant for cleavage at Arg306, k4, was adjusted to 0.192 s-1 for all subsequent studies.
Our empirically driven adjustment to k4 results in a second order rate constant for the overall process of cleavage at Arg306 of 2.74 x 107 M-1 s-1. Previously published second order rate constants for the APC cleavage at Arg306 in human FVa (1.55 x 106 M-1 s-1, 1.7 x 106 M-1 s-1, and 2.5 x 106 M-1 s-1) are an order of magnitude lower than our currently derived estimate and lower than that used by Hockin et al. (9.1 x 106 M-1 s-1) to describe the bovine system.
Studies using human proteins on APC inactivation of FVa reported second order rate constants for the cleavage at Arg506 ranging from 3.0-4.3 x 107 M-1 s-1[31, 41]; a more recent study by Nicolaes et al.  has reported the second order rate constant to be 1.17 x 108 M-1 s-1. This value is quite similar to the value (1.42 x 108 M-1 s-1) used in the current study.
An alternative approach to fitting these data sets would have been to use an optimization method in which all rate constants may be simultaneously adjusted to generate a best fit to all the experimental data. Failure of model predictions to fit the empirical data can stem from the collective effect of small errors in the rate constants used. It is not uncommon for laboratories to report 2–3 fold differences in rate constants of biochemical reactions and interactions measured in vitro. These ranges are accepted because the findings would fall within the margin of error expected for slightly variable experiments conducted in different laboratories, or for different techniques used to measure a rate constant. Such variation can have significant effects on model outputs . However, lack of fit can arise from a fundamental mistake or omitted interaction in the reaction description. Therefore, we have not taken this approach as we developed this model.
It has been widely reported that FXa shields FVa from degradation by APC through its reversible association in the prothrombinase (FXaFVa) complex [34–38, 44]. The mechanistic basis of this protection has been conceptualized as a mutually exclusive competition for association with membrane bound FVa, so that when FVa is associated with FXa it is not a substrate for APC , at least at the Arg506 site . In order to develop a model description that can predict the extent of the protective effect exerted by FXa against APC proteolysis of FVa over a range of reactant concentrations, empirical experiments were performed at saturating and non-saturating concentrations of FXa.
Equations describing interactions between intact and partially proteolyzed factor Va with factor Xa and prothrombin
FXa + FVa ← 7-6 → FXaFVa
0.2 a 0.075 b
4 x 108 a
1.5 x 108 b
FXa + FVai 506 ← 8-6 → FXaFVai 506
4 x 108 c
1.5 x 108 b
FXa + FVai 306 ← 8-6 → FXaFVai 306
4 x 108 c
1.5 x 108 b
FXa + FVai 306/506 ← 8-6 → FXaFVai 306/506
4 x 108 c
1.5 x 108 b
FXaFVai 306 -9 → FXa + FVa1-306FVaLC + FVa307-679/709
FXaFVai 306/506 -9 → FXa + FVa1-306FVaLC + FVa307-506 + FVa507-679/709
FVa + PT ← 11-10 → FVaPT
1 x 108
Our initial mathematical simulations were carried out utilizing a KD of 0.5 nM , resulting in an initial concentration of 26 pM for prothrombinase, with off and on rates of 0.2 s-1 and 4.0 x 108 M-1 s-1, respectively (Table 2, Eqn 11) [46, 47]. The mathematical predictions for the loss of prothrombinase activity in the presence of either 0.5 or 2.0 nM APC showed an approximately 2-fold faster loss in prothrombinase activity than observed in the empirical experiments (Figure 4, Panel A, dashed lines vs. filled circles).
To generate a better fit to the empirical data we first examined the dissociation constant of FXa and FVa on PC:PS vesicles, which by literature reports varies from as low as 83 pM  to as high as 1 nM . Simulations to test the wide range of reported KD values were first conducted by altering the koff rate constants for the prothrombinase complex, while maintaining the same kon rate constant of 4.0 x 108 M-1 s-1. For each KD value tested, the starting concentrations of free and bound species (FVa, FXa, and FXaFVa) were recalculated and used as the initial conditions for a mathematical simulation. An adjustment made to the off rate for the prothrombinase complex resulting in a KD of 0.1 nM (58.8 pM prothrombinase initially) generated an in silico prediction that more closely represented the empirical results seen with both 0.5 and 2.0 nM APC (Figure 4, Panel A, black and red, respectively, dotted lines vs filled circles).
In addition to altering the koff rate constant while maintaining a kon rate constant to reach a desired overall KD, changes to both isolated rate constants were explored as a potential solution. One alternative that we explored was adjusting the kon rate constant to 1.5 x 108 M-1 s-1 and setting the koff rate constant to values that yielded the same set of KD values. Figure 4, Panel B presents these model results at 0.5 and 2.0 nM APC (black and red, respectively) for KD's of 0.5 nM (dashed lines) and 0.1 nM (dotted lines) for prothrombinase assembly. In contrast to simulations based with a kon value of 4.0 x 108 M-1 s-1 this analysis indicates that a KD of 0.5 nM yields a better fit to the data.
These results highlight an important feature of dynamic systems like this one: the magnitude of the observed protective effect may depend not only on the concentrations of the three protein components and the KD's characterizing their interactions, but also on the rate constants defining the competing KD's.
Figure 5 presents empirical experiments in which preformed prothrombinase (20 nM FVa and 30 nM active site blocked FXa (FXa*)) was treated with 2.0 nM APC and time courses of FVaHC and its APC-derived fragments FVa1-506, FVa307-679/709 and FVa307-506 visualized by Western blotting (Figure 5, Panel A, inset). Under these conditions, 95 % to 99 % of the FVa is bound to FXa* prior to addition of APC given KD values in the range of 0.5 nM to 0.1 nM, respectively. The selection of 20 nM FVa reflected the fact that in closed model systems of TF-initiated coagulation, levels of FVa reach 20 nM  and the practical requirement for a concentration suitable for Western blot analysis. Active site blocked FXa was used to prevent FXa proteolysis of FVa .
Reactions where the FVa population is highly associated with FXa* showed 70 % of FVaHC remained after 1 minute (Figure 4, Panel C and Figure 5, Panel A, red circles) compared to ~ 10 % without FXa* (Figure 3, Panel C, red circles). Comparison of initial rates showed a greater than 7-fold reduction in the initial rate of 2.0 nM APC proteolysis of FVa due to its FXa association (77 pM/s vs. 558 pM/s). Consistent with this overall suppression of heavy chain proteolysis, generation of the FVa1-506 fragment was slower in the presence of FXa*, reaching a maximum in 2 min vs ~30 s in its absence (Figure 5, Panel B vs. Figure 3, Panel D, red circles). Its clearance was also suppressed, taking ~ 4 min to decline to 50 % of its maximum as opposed to ~1 min without FXa* (Figure 5, Panel B vs. Figure 3, Panel D, red circles).
In contrast to APC inactivation studies conducted in the absence of FXa*, visual inspection of Western blot images (Figure 5, Panel A, inset) reveal the accumulation of the FVa307-679/709 fragment leading to more reliable quantitative densitometric analysis of this fragment (Figure 5, Panel C, red circles). Generation of this fragment reaches 50 % of its maximal level by 1 minute, and remains at elevated levels above 70 % of maximal level after 2 minutes (Figure 5, Panel C, red circles). Generation of the FVa307-506 fragment reaches 50 % of maximal value after 3 minutes (Figure 5, Panel D, red circles).
Initial prothrombinase concentrations given the condition of pre-incubating 20 nM FVa and 30 nM FXa* were calculated to be 19.1 nM or 19.8 nM using KD values of either 0.5 nM or 0.1 nM, respectively. A range of kon and koff rate constants for each KD were tested by varying kon values between 1 and 4.0 x 108 M-1 s-1. Time courses of the inactivation of FVa by 2.0 nM APC were produced. The results of this in silico study indicated that the disappearance of FVaHC was dependent on the resultant KD value and independent of the combination of kon and koff rate constants used to generate the KD values of 0.1 or 0.5 nM (data not shown). Assuming the KD for prothrombinase is 0.1 nM and the kon is either 1.5 x 108 M-1 s-1 or 4.0 x 108 M-1 s-1, the model predicts that after 6 min 25 % of the FVaHC is proteolyzed (Figure 4, Panel C, red dotted line), in contrast to the empirical results showing ~82 % of FVaHC proteolyzed at 6 minutes (Figure 4, Panel C, red circles). Model simulations with KD values of 0.5 nM, 0.75 nM or 1.0 nM resulted in FVaHC levels decreasing after 6 min by 75 % (Figure 4, Panel C, red dashed line), 83 % (Figure 4, Panel C, red solid line), or 87 % (data not shown), respectively. From the mathematical simulations, the initial rates of FVaHC proteolysis were 22 pM/s, 86 pM/s, and 122 pM/s for KD’s of 0.1 nM, 0.5 nM, and 0.75 nM, respectively, compared to the empirically derived rate of FVaHC proteolysis of 78 pM/s. Overall these comparisons support our eliminating 0.1 nM as the KD for the prothrombinase complex.
While the mathematical model is able to recapitulate the overall combined initial cleavages of FVaHC (Figure 5, Panel A, dotted line vs. filled circles), there is a mixed success for the predictions of the inactivation fragments (Figure 5, Panels B-D, dotted line vs. filled circles). Specifically, the mathematical model captures the rapid accumulation of the FVa307-679/709 fragment (Figure 5, Panel C, dotted line), but does not capture the clearance of the FVa1-506 fragment (Figure 5, Panel B, dotted line) or the generation of the FVa307-506 fragment (Figure 5, Panel D). The same lack of fit for the proteolysis of this fragment is observed when FVai 506 is the starting substrate (Figure 5, Panels E and F, dotted line vs. filled circles).
In the mathematical model as constructed APC proteolysis at any site is permitted only with FVa species not bound to FXa (or FXa*). The more rapid clearance of FVa1-506 observed empirically suggests two possible problems with the model construct: 1) the binding affinity between FVai 506 and FXa is weaker than the model value of 1 nM based on studies with bovine FVa  or 2) APC can cleave at Arg306 when the FVai 506 species is bound to FXa. Previous studies with recombinant human proteins have reported KD values ranging from ~3.9 nM  to ~1.35 nM . To evaluate the likelihood that a weaker binding affinity was the culprit, mathematical simulations were conducted where the KD for the FXaFVai 506 complex was varied; to fit the empirical data the KD for this complex would have to be greater than 10 nM (data not shown) and as such this adjustment was discarded as a viable option.
The alternative explanation for the empirical data presenting the clearance of the FVa1-506 fragment is that when bound to FXa, the FVai 506 species is susceptible to APC cleavage at Arg306. To explore this hypothesis, an additional reaction was added to the mathematical construct: APC cleaving (FXaFVai 506) at Arg306. To establish whether there was a value for this rate constant that would improve the fit, simulations varying its value were run using the model value (2.7 X 107 M-1 s-1) for the cleavage of Arg306 in free FVa as a point of reference. Representative examples are presented in Figure 5. The mathematical analysis suggests that if the Arg306 in (FXaFVai 506) is a target for APC, the magnitude of the rate constant regulating the event is ~10 to 20 % that for free FVa.
In addition to the protection afforded to FVa by FXa in the prothrombinase complex, the substrate of the complex, prothrombin (PT), has also been shown to inhibit APC inactivation of FVa [39, 40, 58]. Guinto and Esmon reported that prothrombin protects the intact FVa molecule by competing with APC for binding to FVa ; more recent studies have reported a KD of between 500–700 nM for FVa and prothrombin [39, 40].
An additional reaction describing the reversible formation of the FVaPT complex with a KD of 700 nM  was added to the ordinary differential equation (ODE) network (Table 2, Eqn 17). The initial concentration of preformed FVaPT complex, given 20 nM FVa and 1.4 μM PT, was calculated to be 13.3 nM. Simulated time courses for FVa heavy chain fragments generated by APC (2.0 nM or 20.0 nM, red and blue lines, respectively) proteolysis did not correlate well with the empirical data, in fact suggesting that there would be little to no protection of FVa by 1.4 μM prothrombin if the KD was 700 nM (Figure 7, Panel A, dashed with dotted lines vs. filled circles; Note, red dashed with dotted line is underneath blue dashed line). Test simulations varying the KD were performed by altering the koff constant for the equilibrium describing the prothrombin and FVa interaction. Even when a KD of 50 nM was assigned to the interaction, ~ 10-fold less than published estimates  the resulting simulations did not capture the observed level of protection (data not shown).
In the computational model one of the assumptions is the presence of unlimited binding sites located on a single surface so that competition between proteins for occupancy of the phospholipid surface is not a factor. At 20 μM PC:PS vesicles, and setting the phospholipid/protein binding site ratio between 30 and 60 [59–61], available binding sites fall in a range between 220 and 440 nM. The simultaneous solution of the equilibrium expressions for prothrombin, FVa, and APC (defined as non-catalytic) binding to phospholipid membrane with KD’s of 230 nM , 2.72 nM , and 500 nM , respectively, shows that between 77–78 % of binding sites are occupied by prothrombin, over 96 % of the FVa is bound, and only approximately 11 % of the APC is able to bind to the surface whether 2.0 or 20.0 nM APC is present. By using this bound fraction of APC as the catalytically relevant population, the mathematical simulations better captured the kinetics of FVHC proteolysis in the presence of 1.4 μM prothrombin (Figure 7, Panel A, solid lines vs. filled circles). Similarly, comparison of empirical and simulated time courses for the three generated fragments (Figure 7, Panels B-D, filled circles vs. solid lines) also show improved fits. Our findings indicate that there may in fact be a tighter association between FVa and PT, in line with the findings of Yegneswaran et al. (Figure 7. Panel A, filled circles vs. dashed lines) .
To verify that the impaired FVa proteolysis observed under the empirical reaction conditions was not solely due to competition for phospholipid binding sites, mathematical simulations were constructed using the adjusted APC concentration, but without permitting the formation of the FVaPT complex. This model construct failed to predict the time courses for FVaHC disappearance and the fragments FVa1-506 and FVa307-679/709 (Figure 7, Panel A, B and C, filled circles vs. dotted lines). Thus the computational studies indicate that the prothrombin dependent suppression of FVa proteolysis by APC observed in our empirical reactions is due both to binding site competition (prothrombin membrane association limiting APC access to the membrane surface) and to the FVaPT complex not being a substrate for APC. Comparisons of the empirical data for FVaHC proteolysis (Figure 7, Panel A) with the simulations representing unlimited binding sites (dashed with dotted lines), limited binding site adjustment but no FVaPT interaction (dotted line) and the limited binding site adjustment with FVaPT complex formation (KD = 700 nM) (solid line) indicate that under these conditions approximately half of the observed PT dependent suppression is due to FVaPT complex formation.
Addition of 2.0 nM APC to the reaction system led to only 39 % of the FVaHC being proteolyzed after 20 minutes (Figure 9, Panel B, red circles). The initial rate of disappearance was 14 pM/s, nearly 41-fold lower than the inactivation rate in the absence of both prothrombin and FXa (558 pM/s; Figure 3, Panel A, red circles). Inactivation studies with 10 times more APC resulted in a nearly 8-fold increase in the initial rate of inactivation (100 pM/s) (Figure 9, Panel B, blue circles).
Equations describing interactions between catalytically inactive prothrombinase species and prothrombin
FXaFVa + PT ← 14-13 → FXaFVaPT
1.0 x 108 a
FXaFVai 506 + PT ← 14-13 → FXaFVai 506PT
1.0 x 108
FXaFVai 306 + PT ← 14-13 → FXaFVai 306PT
1.0 x 108
FXaFVai 306/506 + PT ← 14-13 → FXaFVai 306/506PT
1.0 x 108
FXaFVai 306PT −9 → FXa + FVa1-306FVaLC + FVa307-679/709 + PT
FXaFVai 306/506PT −9 → FXa + PT + FVa1-306FVaLC + FVa307-506 + FVa507-679/709
Simulated time courses for the disappearance of the FVaHC were generated for the experiments where 2.0 nM or 20.0 nM APC was added. Based on the findings from the APC inactivation studies of FVa in the presence of prothrombin, APC effective concentrations were adjusted to 11 % of the total to account for the limited binding sites (Figure 9, Panel B, solid lines). The simulations’ initial rates of proteolysis of the FVaHC (6 pM/s) were less than half the measured empirical rates (14 pM/s). The mathematical simulations greater level of protection extends through 20 minutes with higher amounts of FVaHC anticipated compared to empirical results (2.0 nM APC: 76 % vs. 61 %, respectively; 20.0 nM APC: 26 % vs. 11 %, respectively) (Figure 9, Panel B). Empirical studies were also conducted with 50 μM PC:PS vesicles and did not result in any significant changes in the initial rate or overall time course of FVaHC proteolysis by 2.0 nM APC (data not shown), suggesting that under the current conditions, phospholipid membrane accessibility is not impeding the inactivation reaction.
Potential explanations for this disparity include: 1) accumulated effect of small errors in rate constants; 2) improved binding of APC to the surface; 3) a misestimate of the affinity of the prothrombinaseprothrombin interaction; and 4) the affinity of human APC for FVa is slightly higher, e.g. the KD is less than 7 nM. Additional experiments will be required to distinguish between these possibilities. It is important to note that phospholipid composition plays an important role in protein-membrane interactions which can extend to effects on enzyme activity (Reviewed in ). Several studies have directly highlighted the effect of membrane composition on the APC inactivation of FVa [66–68].
Human prothrombin and Factor X were isolated according to methods as described . Factor Xa was prepared as previously described using Russel’s Viper Venom Factor Xa Activator [82, 83]. Human Factor V was purified from citrated plasma according to previously described procedures . Activated human protein C was purchased from Haematologic Technologies, Essex Junction, VT. 1,2-Dioleolyl-sn-Glycero-3-Phospho-L-Serine (PS) and 1,2-Dioleoyl-sn-Glycero-3-Phosphocholine (PC) were purchased from Avanti Polar Lipids, Inc (Alabaster, AL) and phospholipid vesicles (PC:PS) composed of 75 % PC and 25 % PS were prepared as described [60, 84]. Spectrozyme TH and recombinant hirudin were purchased from American Diagnostica, Inc (Greenwich, CT) and EDTA was purchased from Sigma (St Louis, MO). D-Phe-Pro-ArgCH2Cl (FPRck) was prepared in house. Monoclonal anti-fV (αHFV#17) was obtained from the Biochemistry Antibody Core Laboratory (University of Vermont) and a goat anti-mouse IgG conjugated to HRP was purchased from Southern Biotech (Birmingham, AL). Active site blocked FXa (FXa*) was produced according to the previously published method .
Multiple preparations of FV/FVa and different lots of APC were utilized for these studies. Preparations of Factor Va were made fresh prior to all experiments. Factor V (1.0 μM) in 0.2 M HEPES, 0.15 M NaCl, 0.1 % PEG-8000, 2.0 mM CaCl2, pH 7.4 (HBS-PEG-Ca) was activated with 10 nM human thrombin for 10 min at 37°C. Thrombin activation was stopped by addition of 12 nM recombinant hirudin and placing the sample on ice. Factor Va preparations were used within 4 hours. The partially proteolyzed species of factor Va cleaved only at Arg506 was generated by incubating FVa (750 nM) in HBS-PEG-Ca with 5.0 nM APC for 20 minutes at 37°C, after 20 minutes an additional 5.0 nM APC was added. Following the inactivation reaction, the FVai 506 containing reaction was placed on ice and incubated with 1.0 mM di-isopropyl phosphate to inactivate the APC.
Solutions of factor Va (20 nM) in HBS-PEG-Ca with 20 μM PC:PS were prepared at 37°C. At the zero time point 0.5 or 2.0 nM APC was added to the FVa solution. At designated intervals samples were removed, and either quenched in denaturing sample preparation buffer (0.31 M Tris[hydroxymethyl]aminomethane, 10 % sodium dodecyl sulfate, 50 % glycerol, 0.5 % Bromophenol Blue, pH 6.8) and analyzed either for proteolytic cleavage using SDS-PAGE and subsequent immunoblot analysis (Western blotting), or for co-factor activity using a one-stage clotting assay in factor V deficient plasma. Inactivation reactions were diluted 40-fold in an appropriate volume of factor V deficient plasma and assayed immediately. Clotting activity is represented as a percentage of the clotting time observed for unproteolyzed FVa.
Rates of APC inactivation of FVa (0.2 nM) in the presence of PC:PS vesicles (20 μM) and FXa (0.1 nM) were assessed with the prothrombinase complex preformed for five minutes at 37°C in HBS-PEG-Ca. At the zero time point APC (0–2.0 nM) was added to the prothrombinase complex solution. At designated intervals aliquots were removed and added to a minimal amount of concentrated prothrombin (1.0 μM final). After 90 s, aliquots were quenched with the addition of 0.5 M EDTA (25 mM final) and subsequently monitored for thrombin activity against the chromogenic substrate SpecTH. Thrombin concentrations were determined by reference to a standard curve and the level of thrombin generated used as a relative measurement of prothrombinase concentration compared to measurements with intact FVa.
Prothrombinase complex was preformed on 20 μM PC:PS vesicles in HBS-PEG-Ca for four minutes at 37°C using 30 nM fluorescein-active site blocked FXa (FXa*) and 20 nM FVa. A sample aliquot was removed, APC added (2.0 nM), aliquots taken at selected time points and quenched into denaturing sample preparation buffer, and subsequently analyzed by Western blotting.
Reaction mixtures of FVa (20 nM) and prothrombin (1.4 μM) were prepared with 20 μM PC:PS vesicles in HBS-PEG-Ca and incubated for four minutes at 37°C. A sample aliquot was removed, APC added (2.0 or 20.0 nM), aliquots taken at selected time points and quenched into denaturing sample preparation buffer, and subsequently analyzed by Western blotting.
Reaction mixtures of FVa (20 nM), fluorescein-active site blocked FXa (FXa*, 30 nM), and prothrombin (1.4 μM) were prepared with 20 μM PC:PS vesicles in HBS-PEG-Ca and incubated for four minutes at 37°C. A sample aliquot was removed, APC added (2.0 or 20.0 nM), aliquots taken at selected time points and quenched into denaturing sample preparation buffer, and subsequently analyzed by Western blotting.
All FVa samples were run under reducing conditions. FVa heavy chain and inactivation fragments were analyzed by fractionation on 4-12 % SDS-PAGE acrylamide slab gels followed by electrophoretic transfer of proteins to Immobilon-FL membranes (Millipore Inc., Billerica, MA), and detected using a mouse monoclonal antibody (αHFV#17) that binds to an epitope between residues 307–506 of the factor Va heavy chain and a goat anti-mouse IgG conjugated to HRP (Southern Biotech). Luminescence was detected using a Fuji LAS-4000 in conjunction with ImageCapture software (Fujifilm, Tokyo, Japan). Densitometry was carried out using MultiGauge software (Fujifilm, Tokyo, Japan). For each species, densitometric values at each time point were normalized to their maximal value during the time course.
The description using ordinary differential equations of the reaction pathway for bovine APC inactivation of bovine FVa reported by Hockin et al.  was used as a starting framework (Figure 2; Table 1). The scheme describes a random order cleavage process for Arg306 and Arg506 in which the cleavages are independent of each other and characterized by substantially different rate constants. Our simulator employs a fourth order Runga-Kutta algorithm to generate a series of time-dependent concentration profiles for all reactants, intermediates, and products.
Factor Va cofactor activity will be represented by the cumulative presence of FVaHC species present at any point in time. For measurements of prothrombinase activity, the sum of all FVa species was added to the weighted (20 %) sum of FVai 306, FVai 506, and FVai 306/506. Factor Va inactivation fragments will be represented by including respective fragments both associated and dissociated from the FVaLC.
To generate an appropriate simulation of the empirical experiment where FXa and FVa are pre-incubated together on the phospholipid surface to form the prothrombinase complex prior to the addition of APC, the concentrations of free FXa and FVa, and the prothrombinase complex were solved. For example, studies with 200 pM FVa and 100 pM FXa, for each potential KD (e.g. 0.1 nM or 0.5 nM, respectively), the values of the three initial conditions ranged from 41–74 pM for free FXa, 141–174 pM for free FVa, and 59–26 pM for prothrombinase. The concentrations of the three species then served as the initial conditions in the simulations in conjunction with either APC concentration.
In silico simulations were carried out with the initial concentrations used in the empirical experiments and the relevant output calculated and compared to the appropriate analyte measured in the empirical experiments.
Activated protein C
(Ethylene-dinitrilo) tetraacetic acid
Activated factor V
Heavy chain of FVa
Light chain of FVa
A two chain molecule composed of the FVaLC and FVa1-306 fragment
Activated factor X
Active site-blocked activated factor X
20 mM HEPES, 150 mM NaCl, pH 7.4
PS vesicles: Single bilayer phospholipid vesicles composed of 75 % 1,2-dioleoyl-sn-glycero-3-phosphocholine and 25 % 1,2-dioleoyl-sn-3-glycero-3-[phospho-L-serine]
Polyethylene glycol, average molecular weight = 8000
Relative to maximal value
This paper was supported by F31 GM 81904 (MCB), R01 HL 034575 (KGM), and P01 HL 46703 (KGM) from the National Institute of Health. Additional support was provided by contract W911NF-10-1-0376 (KGM) from the Department of Defense. Additional support was provided by NEAGEP at the University of Vermont (MCB), institutional funds from the College of Medicine at the University of Vermont (MCB), and the Department of Biochemistry at the University of Vermont (MCB, SJE).