Note: Numbering of amino-acid positions in protein C corresponds to the chymotrypsinogen nomenclature.
G. A. F. Nicolaes, Department of Biochemistry, Cardiovascular Research Institute Maastricht, Maastricht, the Netherlands. Fax: + 31 43 3884159, Tel.: + 31 43 3881539, E-mail: G.Nicolaes@Bioch.Unimaas.nl
Inactivation of factor Va (FVa) by activated protein C (APC) is a predominant mechanism in the down-regulation of thrombin generation. In normal FVa, APC-mediated inactivation occurs after cleavage at Arg306 (with corresponding rate constant k′306) or after cleavage at Arg506 (k506) and subsequent cleavage at Arg306 (k306). We have studied the influence of heparin on APC-catalyzed FVa inactivation by kinetic analysis of the time courses of inactivation. Peptide bond cleavage was identified by Western blotting using FV-specific antibodies. In normal FVa, unfractionated heparin (UFH) was found to inhibit cleavage at Arg506 in a dose-dependent manner. Maximal inhibition of k506 by UFH was 12-fold, with the secondary cleavage at Arg306 (k306) being virtually unaffected. In contrast, UFH stimulated the initial cleavage at Arg306 (k′306) two- to threefold. Low molecular weight heparin (Fragmin®) had the same effects on the rate constants of FVa inactivation as UFH, but pentasaccharide did not inhibit FVa inactivation. Analysis of these data in the context of the 3D structures of APC and FVa and of simulated APC–heparin and FVa–APC complexes suggests that the heparin-binding loops 37 and 70 in APC complement electronegative areas surrounding the Arg506 site, with additional contributions from APC loop 148. Fewer contacts are observed between APC and the region around the Arg306 site in FVa. The modeling and experimental data suggest that heparin, when bound to APC, prevents optimal docking of APC at Arg506 and promotes association between FVa and APC at position Arg306.
Activated factor V (FVa) is an essential cofactor in the prothrombin-activating complex, stimulating the activity of membrane-bound factor Xa (FXa) more than 100 000-fold [1,2]. Hence, FVa is an ideal target for the regulation of thrombin formation . Downregulation of FVa activity is achieved through proteolysis mainly mediated by the anticoagulant protein C pathway (reviewed in [4,5]). Protein C is composed of a heavy and a light chain held together by a single disulfide bond . The light chain contains the γ-carboxyglutamic acid (Gla)-rich domain and two epidermal growth factor-like domains . The heavy chain comprises a short activation peptide and a serine protease (SP) domain which contains the active site of the enzyme. Activated protein C (APC), the product of a thrombin–thrombomodulin-catalyzed activation of the zymogen protein C, proteolytically inactivates the coagulation cofactors, FVa and FVIIIa , in reactions stimulated by the APC cofactor protein S.
FVa consists of a 105 kDa heavy (A1 and A2 domains) and a 71–74-kDa light (A3, C1, and C2 domains) chain which are noncovalently associated.
During APC-catalyzed inactivation of FVa, the heavy chain of FVa is cleaved at three sites: Arg306, Arg506 and Arg679 . The cleavages at Arg306 and Arg506 appear to be crucial for inactivation, but the cleavage at Arg679 is probably less important . The cleavage at Arg506 is kinetically favored over that at Arg306 and results in the formation of an inactivation intermediate (FVaint), which retains partial FVa cofactor activity owing to its ability to bind FXa, albeit with lower affinity . The FVa activity is lost after cleavage at Arg306. In carriers of the common FVLeiden mutation, in whom the Arg506 has been replaced by a Gln, inactivation occurs via the slow Arg306 cleavage (reviewed in ). Cleavage at Arg306 results in a large reduction in FXa affinity and also the dissociation of the A2 domain, the two processes ultimately rendering FVa inactive as a cofactor of FXa [12,13]. APC-catalyzed inactivation of FVa is modulated by other plasma components. Thus, the nonenzymatic cofactor protein S promotes cleavage at Arg306, whereas FXa specifically blocks the cleavage at Arg506 .
It has recently been shown that basic residues in two surface loops (37 and 70) in the SP domain of APC (chymotrypsinogen nomenclature) form an extended binding site for FVa [15–17]. In addition, APC loop 148 also plays a role in FVa degradation [18,19]. Loop 60 is probably less important as mutagenesis of positive residues in this loop did not affect inactivation of FVa by APC [15,16].
Heparin is an important regulator of APC activity, promoting the interaction between APC and one of its inhibitors, the serpin protein C inhibitor (PCI) . This is probably mediated via a template mechanism, for which binding of heparin to basic residues in three of the four surface loops in APC (60, 37 and 70) is crucial [17,20,21]. Interestingly, heparin has also been reported to stimulate APC-catalyzed inactivation of intact FV, but not FVa [22,23].
We performed a detailed kinetic analysis of the influence of heparin on the inactivation of FVa by APC and found a specific heparin-mediated inhibition of the cleavage at Arg506, whereas cleavage at Arg306 was mildly stimulated. Structural analysis strongly suggested that the heparin-binding loops 37, 60 and 70 in APC could complement electronegative areas surrounding the Arg506 site of FVa indicating that electrostatic interactions between regions of FVa and APC could be critical for the formation of the APC–FVa complex which is involved in the cleavage at position Arg506. These electrostatic interactions are inhibited by heparin when heparin is bound to the electropositive cluster on loops 37, 60 and 70 located at one edge of the SP domain of APC. Our data further suggest that heparin can potentially bridge APC to exosites around Arg306, thereby facilitating cleavage at position Arg306.
Materials and Methods
Proteins and reagents
Human FV and FVLeiden were purified from the plasma of a normal individual and an individual homozygous for the FV Arg506Gln mutation, and FVa2 was prepared as described . Throughout the work presented here FVa2 was used. Factor Xa, α-thrombin, protein S, human activated protein C and prothrombin were purchased from Kordia Laboratory Supplies (Leiden, the Netherlands). All coagulation factors were of human origin unless otherwise stated. 1,5-DNS-GGACK-factor Xa (DEGR-FXa) was prepared as described previously . The monoclonal antibody AHV 5146 was purchased from Haematologic Technologies (Essex Junction, VT, USA). Unfractionated heparin (UFH) and low molecular weight heparin (Fragmin®) were obtained from Leo (Ballerup, Denmark); 1 IU·mL−1 UFH contains ≈ 5.7 µg UFH·mL−1. Pentasaccharide was from Sanofi-Récherche (Montpellier, France). Phospholipid vesicles [10% 1,2 dioleoyl-sn-glycero-3-phosphoserine (DOPS), 90% 1,2 dioleoyl-sn-glycero-3-phosphocholine (DOPC), mol/mol] were prepared as described . The chromogenic substrates S-2366 and S-2238 were obtained from Chromogenix (Milano, Italy), and biotrace™ poly(vinylidene difluoride) transfer membranes from Pall Gelman Laboratory (Ann Arbor, MI, USA).
Expression and purification of recombinant human protein C
Recombinant protein C variants K37S/K38Q/K39Q (37-loop mutant), K62N/K63D (60-loop mutant), K37S/K38Q/K39Q/K62N/K63D (37+60-loop mutant) were created by PCR-based site-directed mutagenesis of the eukaryotic expression vector pGT-hyg (Eli Lilly), expressed in 293 cells (CRL-1573; ATCC), purified, and characterized as described previously [21,26].
Assay of FVa
FVa activity was determined by quantification of the rate of FXa-catalyzed prothrombin activation, as described previously . Briefly, in a reaction mixture that contained 0.5 µm prothrombin, a limiting amount of FVa (83 pm FVa), 5 nm FXa, 40 µm phospholipids (10 : 90 DOPS/DOPC, mol/mol), 0.5 mg·mL−1 ovalbumin, and 2 mm CaCl2, prothrombin activation was allowed for 1 min at 37 °C. The amount of prothrombin activated was then determined using S-2238 .
APC-catalyzed inactivation of FVa
Time courses of FVa inactivation by APC were determined by following the loss of FXa cofactor activity of FVa in the prothrombinase complex as a function of time. Routinely, 0.8 nm plasma-derived human FVa or FVaLeiden was preincubated with 25 µm phospholipid vesicles (10 : 90 DOPS/DOPC, mol/mol) in the absence or presence of protein S (200 nm) and/or heparin (0.01–25 IU·mL−1) in 25 mm Hepes buffer (pH 7.5), containing 150 mm NaCl, 3 mm CaCl2, and 5 mg·mL−1 BSA, for 5 min at 37 °C. Inactivation was started by adding wild-type APC or APC variants, and the progressive loss of FVa was monitored for up to 20 min by transfer of aliquots to the FVa assay described above.
Analysis of kinetic data
Rate constants for APC-catalyzed Arg506 and Arg306 cleavage were obtained as described previously by fitting the time courses of FVa inactivation to a random-order, two-cleavage model using nonlinear least-squares analysis . In this model, FVa can be randomly cleaved at either Arg306 or Arg506. Cleavage at Arg306, with an apparent second-order rate constant of k′306, results in complete loss of FVa cofactor activity (FVai; pathway 1, Eqn 1). Alternatively, initial cleavage at Arg506, with a corresponding rate constant of k506, results in a reaction intermediate (FVaint) with ≈ 40% residual cofactor activity, which must be further cleaved at Arg306 (k306) in order to completely abolish FVa cofactor activity (pathway 2, Eqn 2).
In wild-type FVa, in which cleavage at Arg506 is ≈ 20-fold faster than cleavage at Arg306, the major part (≈ 95%) of FVa is inactivated via pathway 2. To reliably determine k′306, single exponential inactivation time courses were determined for FVaLeiden, representing cleavage at Arg306 only and the k′306 values obtained were used in the fits for normal FVa. We have verified our previous findings  that the APC-catalysed FVa inactivation time courses were second-order throughout, i.e. were directly proportional to FVa and APC concentrations between 0 and 1.5 nm FVa and between 0.05 and 5 nm APC. Thus, the second-order rate constants obtained by this method can be directly compared within the given ranges of FVa and APC. Statistical analysis using the StatGraphics Plus for Windows package was performed to determine kinetic parameter significance.
Western blot analysis of FVa inactivation by APC
Human FVa (10 nm), phospholipid vesicles (25 µm) and wild-type APC were incubated at 37 °C in 25 mm Hepes (pH 7.5), containing 150 mm NaCl, 3 mm CaCl2, and 5 mg·mL−1 BSA in the absence and presence of 25 IU·mL−1 UFH. Aliquots of 20 µL were removed at various time points, and subjected to SDS/PAGE (7.5% gel) under reducing conditions. After transfer to poly(vinylidene difluoride) membranes, heavy chain fragments were visualized using a monoclonal antibody (AHV 5146) directed against the FVa heavy chain.
Electrostatic potentials for APC and FVa A domains
The 3D structure of Gla-domainless APC  and the homology model for the three A domains of FVa  (co-ordinate file at http://www.klkemi.mas.lu.se/dahlback) were investigated using the programs insightii, biopolymer and delphi (Accelrys, San Diego, CA, USA). Electrostatic potentials were computed with DelPhi (reviewed in ) using a standard set of formal charges. The standard protocol was applied. The volume inside the APC or FVa molecular surface was assigned a dielectric constant of 4 and the outside volume was given a value of 80.
Docking heparin-like molecules on to APC
Three different methods were used to dock heparin and APC. The first method follows the script reported by Fernandez-Recio and coworkers  as integrated in the modeling package icm (Molsoft LLC, San Diego, CA, USA). This first protocol included a pseudo-Brownian rigid-body docking, an extended force field, and a soft interaction energy function precalculated on a grid. To validate the docking protocol, it was first applied to three different experimental protein–heparin complexes deposited at the Protein Data Bank (PDB) files , PDB code 1bfc, 1azx, 1e0o [32–34]. These proteins were energy-minimized to allow relaxation of side chains, and to simplify calculations, we used a negatively charged polypeptide to mimic heparin. Negatively charged groups were moved with the simulation software Discover (Accelrys) in order to reproduce the overall positioning in space of charges and overall shape, as observed on the NMR/modeled structure of heparin (file 1hpn ).
Amino acids known to be part of the heparin-binding site in these three crystallized protein–heparin complexes were given as starting point for the docking search . At least one conformation of the simulated protein–peptide complexes among the lowest five docking energy scores reproduced accurately the X-ray crystal structures of the equivalent complexes. Therefore, the above docking protocol combined with partial knowledge of the binding site for heparin at the surface of APC as defined by several mutagenesis studies [17,18,21] was deemed appropriate to reasonably predict the overall orientation of a peptide mimicking heparin at the surface of APC.
The second approach involved flexible docking of heparin, PDB files 1hpn  and 1e0o , onto a rigid APC structure using the ICM package.
Finally, a structure-based virtual screening approach (reviewed in ) was used to dock short sugar molecules in the APC loop 37 area. The heparin (PDB file 1hpn) was shortened, and 10 trisaccharides were generated. These sugar molecules were considered rigid during the docking but in order to consider some conformational flexibility, 13 conformers for each molecule were generated and all structures stored in a single data file. A shape-based Gaussian docking function as integrated in the program fred was used to position the short sugar molecules at the surface of APC .
Partial docking of APC on to FVa
The X-ray crystal structure of APC (with modifications in loop 148, see below) and a model structure for the three A domains of FVa (with some modifications at the Arg506 and Arg306 sites, see below) were used in two different automated docking procedures. Rigid-body-docking calculations with soft potentials were performed with the ICM package, as described . Alternatively, we used the approach reported by Norel et al. . In docking with ICM, FVa Arg506 or Arg306 was given as starting point for the search, whereas for the method of Norel et al. the entire surfaces of both interacting molecules were investigated and as such the search was not restricted to known binding regions (i.e. at the Arg306 or Arg506 site).
In all experimentally known complexes of serine proteases/macromolecular inhibitors/substrates, the peptide bond to be cleaved tends to be located on a loop structure that protrudes far outside the molecular surface, either because this loop is indeed ill-structured or because its conformation has to change during the interaction with proteases. A loop structure in FVa including Arg506 was predicted from the X-ray crystal structure of ceruloplasmin and does not project significantly outside the surface of FVa. Numerous initial conformations (2000–20 000) were generated for residues 500–510 using the loop prediction program of Xiang et al. . Several runs were performed, and the 10 best-energy conformations were kept. From these structures, two residues before and after Arg506 were template-forced to partially adopt the conformation of the serpin reactive loop as present in the PDB file 1l99 . Similarly, the FVa loop that contains Arg306 was built again (residues 302–320) and several conformations were generated. Because it is known that APC loop 148 plays a role in FVa inactivation [19,26] and as this loop is not well defined in the PDB file 1aut, the APC loop including residues 145–153 was rebuilt using the approach of Xiang et al., and the 10 lowest-energy conformations were selected.
One structure of FVa with the Arg506 loop sufficiently solvent-exposed to allow interactions with APC was selected for the docking procedures. Docking simulations were performed with the two best ranked structures for the FVa Arg306 loop and an APC model in which the loop 148 tends to be covering the active site (one of the lowest-energy conformations).
For final optimization of the docking procedure, it was necessary to include experimentally obtained data. Interactive docking was consequently performed starting from the best theoretical complexes (i.e. the complexes that have orientation compatible with key experimental/structural/theoretical data) using as guidelines the relative orientation of a protease associated with a substrate/inhibitor as seen in the X-ray crystal structure of a Michaelis serpin–protease complex (file 1l99 ) to optimize further the positioning of the molecules.
APC-catalyzed inactivation of FVa in the presence of heparin
Heparin strongly influenced the time course of FVa inactivation by wild-type APC (Fig. 1). In the absence of heparin (Fig. 1A, open circles) the typical biphasic inactivation curve was obtained for normal FVa (see also ). This is indicative of rapid cleavage at Arg506, which yields a partially active reaction intermediate that is subsequently fully inactivated on cleavage at Arg306. In the presence of 25 IU·mL−1 heparin, FVa inactivation was significantly reduced (Fig. 1A, closed circles). The shape of the curve suggested strong impairment of the fast first phase (Arg506 cleavage) of the reaction without influence on the second phase (Arg306 cleavage). A direct effect of heparin in the FVa assay was excluded because preincubation of FVa with 25 IU·mL−1 UFH or direct addition of heparin (1 IU·mL−1) to the FVa assay mixture resulted in assay outcomes that were identical with those obtained in the absence of heparin (data not shown).
To gain insights into which rate constants (k506 and k′306 in intact FVa and k306 in FVaint) were influenced by heparin, two additional experiments were performed. In the first, shown in Fig. 1B, FVa was incubated for 5 min with APC to allow completion of the first phase of the reaction, i.e. cleavage at Arg506 with all FVa activity converted into FVaint. Thereafter, the reaction volume was divided into two equal parts and transferred into new reaction tubes containing either heparin or buffer. Monitoring of FVa activity in these two tubes was continued for another 25 min, during which time the loss of FVa activity in the absence of heparin (open circles) did not differ from that obtained in the presence of heparin (closed circles). This indicates that the presence of heparin did not influence the k306 in the partially active FVa intermediate. In a second experiment, the effect of heparin on the direct cleavage at Arg306 (k′306) in FVa was determined from a time course of FVaLeiden inactivation in the absence and presence of heparin. From these data, an approximately 2–3-fold (2.48; standard error: 0.47) stimulation of the rate of inactivation (k′306) by heparin was calculated (Fig. 1C) by fitting to a single exponential. Table 1 summarizes all rate constants obtained by fitting all time courses to the random-order cleavage model as described in Materials and methods, under the constraints that k306 is minimally influenced by heparin and using k′306 obtained from the FVaLeiden inactivation.
Table 1. Apparent second-order rate constants for the APC-catalyzed inactivation of FVa and FVaLeiden in the presence and absence of heparin. Rate constants (m−1·s−1) for inactivation of normal FVa (0.50–1.5 nm) and FVaLeiden (0.50–1.5 nm) catalyzed by recombinant wild-type APC (0.037–1.0 nm) obtained by fitting time courses of inactivation, such as presented in Fig. 1, to an integral time course equation as described in Materials and Methods. Reactions were performed in 25 mm Hepes (pH 7.5), containing 150 mm NaCl, 3 mm CaCl2 and 5 mg·mL−1 BSA in the presence of 25 µm phospholipid vesicles (10 : 90 DOPS/DOPC, mol/mol) at 37 °C in the absence or presence of 25 IU·mL−1 UFH. FVa activity was assayed as described in Materials and Methods. Rate constants given are means from at least three experiments, with SEM being < 20%. Values for k′306 for inactivation of normal FVa by APC were obtained from time courses of inactivation for FVaLeiden as described in Materials and methods. ND, Not determined; k506 and k306 could not be determined in FVaLeiden because of the absence of the Arg506 cleavage site.
6.83 × 105
2.17 × 106
6.83 × 105
2.17 × 106
1.17 × 108
9.96 × 106
1.55 × 106
1.49 × 106
Concentration dependence of heparin effect on k506
To further characterize the influence of heparin on the APC-catalyzed inactivation of FVa and to confirm the specificity of the effects observed in Fig. 1, time courses of FVa inactivation by wild-type APC were determined in the presence of various concentrations of UFH (0.1–55 IU·mL−1), and rate constants for the inactivation were calculated. The effect of heparin on k506 was dose-dependent and saturable, with 50% inhibition observed at ≈ 2 IU·mL−1 UFH (Fig. 2).
The inhibitory effect was not specifically mediated by unfractionated heparin because we found that 25 IU·mL−1 low molecular weight heparin (which is less than 18 saccharide residues in length) displayed similar inhibitory activity to UFH. In contrast, equimolar amounts of pentasaccharide did not influence the inactivation of FVa by APC.
To relate the inactivation curves of FVa to specific cleavages in FVa, Western blotting was performed (Fig. 3). A monoclonal antibody (AHV 5146) against the FVa heavy chain was used to visualize the inactivation fragments. In the absence of heparin, the transient 75 kDa fragment representing the heavy chain fragment 1–506, typical for the partially active reaction intermediate (FVaint), can be seen (pathway 2). The 30-kDa band represents the 307–506 fragment formed on cleavage at Arg306 in FVaint. In the presence of UFH, the pattern of fragment generation was considerably different from that obtained in its absence. The amount of 75-kDa fragment was greatly reduced, and instead a 60/62-kDa fragment accumulated, representing the 307–709 fragment, which is the result of cleavage at Arg306 in FVa. These data indicate that heparin strongly influences cleavage at Arg506, and shifts the pathway towards initial cleavage at Arg306 (pathway 1), after which Arg506 can be cleaved, resulting in the formation of a 30 kDa fragment.
Kinetic analysis of cleavage of FVa at Arg506 by APC
To further characterize the inhibitory effect of heparin on the Arg506 cleavage, we performed a kinetic analysis by using nonlinear regression analysis of initial rates of FVa inactivation by APC (initial rates representing almost exclusively cleavage at Arg506) at FVa concentrations of 0.2–30 nm, in both the absence and presence of 25 IU·mL−1 heparin (Fig. 4). In the FVa assay, a lower concentration of FXa (0.5 nm) was used than in the standard FVa assay, in order to minimize the FXa cofactor activity of the reaction intermediate, FVaint. This facilitated the determination of loss of FVa cofactor activity during the initial stage of inactivation and minimized the influence of the k306 cleavage. The kinetic parameters obtained were Km-app = 2.13 ± 0.49 nm (mean ± SEM) and kcat-app = 0.68 ± 0.048 s−1 in the absence of UFH, and Km-app = 11.4 ± 1.49 nm and kcat-app = 0.26 ± 0.016 s−1 in the presence of UFH, which correspond to second-order rate constants for cleavage at Arg506 (kcat/Km) of 3.2 × 108m−1·s−1 and 2.3 × 107m−1·s−1, respectively. These values are in reasonable agreement with the rate constants obtained from fitting the time courses of FVa inactivation (cf. Figure 1, Table 1). These data suggest that, under the conditions tested, heparin increases the Km of APC for cleavage of FVa at Arg506 5.5-fold and at the same time decreases the kcat 2.6-fold, thus acting as a mixed-type inhibitor of APC. As this analysis was performed using initial rates for inactivation of FVa that are almost completely due to cleavage at Arg506, no individual kinetic parameters can be deduced for the second cleavage at Arg306.
Effects of heparin on FVa inactivation in the presence of DEGR-FXa and protein S
To further verify that heparin inhibits cleavage at Arg506 and stimulates initial cleavage at Arg306, inactivation of FVa by APC with and without heparin was performed in the presence and absence of 20 nm DEGR-FXa, which is known to completely and specifically block APC cleavage at Arg506 . The presence of DEGR-FXa resulted in abrogation of the fast phase of the reaction, as compared with inactivation in the absence of DEGR-FXa. Addition of 25 IU·mL−1 UFH in the presence of DEGR-FXa stimulated the APC-mediated inactivation approximately twofold (data not shown), indicating that the stimulatory effect of heparin on the cleavage at Arg306 is also observed in the presence of FXa. Theoretically, the addition of heparin may induce the formation of a binary FXa–FXa complex, thus abolishing the inhibitory effect of DEGR-FXa on the Arg506 cleavage. This possibility was excluded, however, because time courses of FVa inactivation in the presence of DEGR-FXa could be fitted, both with and without heparin, to a mono-exponential equation, indicating single cleavage at Arg306 only. The effect of UFH on FVa inactivation by APC was also investigated in the presence of 200 nm protein S. The resulting time course of inactivation of FVa by APC became monophasic, an effect known to be due to stimulation of the cleavage at Arg306 . Addition of 10 IU·mL−1 heparin resulted in a somewhat slower inactivation rate (data not shown), which presumably is due to the inhibitory effect of UFH on the cleavage at Arg506.
Lack of heparin effect on FVa inactivation by APC mutants deficient in heparin binding
To verify that the inhibition of APC-mediated FVa inactivation by heparin is related to the ability of heparin to bind APC, we used recombinant mutants of APC harboring mutations in the heparin-binding site. The APC variants chosen were: K37S/K38Q/K39Q with strongly reduced affinity for heparin; K62N/K63D which has a modest effect on the APC–heparin interaction; K37S/K38Q/K39Q/K62N/K63D which shows no detectable binding to heparin . Table 2 shows the apparent second-order rate constants obtained from time courses of inactivation of FVa by the various APC variants. Results obtained in the absence of heparin were consistent with those on record . APC variants carrying mutations in the 37 loop had much lower rate constants for cleavage at Arg506, whereas the rate constants for cleavage at Arg306 were similar to those obtained with wild-type APC. The mutations in loop 60 did not affect k506 and showed a modestly increased k′306 (Table 2). In the presence of heparin, the abilities of wild-type APC and the 60-loop variant to cleave the Arg506 site in FVa were strongly inhibited, with reduction of k506 of 11.7-fold and 8.4-fold, respectively, whereas only minor effects were seen on the cleavage at Arg306. The addition of heparin during inactivation of normal FVa by the loop-37 mutant resulted in a further decrease in k506 and a small stimulation in the secondary cleavage at Arg306 (k306). However, heparin did not influence Arg506 cleavage in FVa by the APC variant that completely lacked the heparin-binding capacity (37+60 loop), but a small (inhibitory) effect on k′306 and k306 was noted.
Table 2. Apparent second-order rate constants for the inactivation of FVa and FVaLeiden catalyzed by several recombinant variants of APC in the presence and absence heparin. Rate constants (m−1·s−1) for inactivation of normal FVa (0.80 nm) and FVaLeiden (0.80 nm) catalyzed by recombinant wild-type APC (0.090–0.60 nm) obtained by fitting time courses of inactivation, such as presented in Fig. 1, to an integral time course equation as described in Materials and Methods. Reactions were performed in 25 mm Hepes (pH 7.5), containing 150 mm NaCl, 3 mm CaCl2 and 5 mg·mL−1 BSA in the presence of 25 µm phospholipid vesicles (10 : 90 DOPS/DOPC, mol/mol) at 37 °C in the absence or presence of 25 IU·mL−1 UFH. FVa activity was assayed as described in Materials and Methods. Rate constants given are means from at least two experiments.
In the absence of UFH
In the presence of UFH
a The value for k ′ 306 in the presence of heparin in FVaLeiden, could not be reliably measured, implying it is < 4 × 105m−1·s−1.
These observations suggest that, for heparin to exert its inhibitory effect on the cleavage at Arg506, a normal interaction with APC is required. It is likely that the heparin-binding loop 37 of APC interacts directly with FVa in an area adjacent to the Arg506 cleavage site and that this interaction is severely hampered by heparin in the ternary heparin–APC–FVa complex.
Electrostatic potentials of APC and the A domains of FVa
Loops 37, 60, 70 and 148 in APC form an electropositive surface, while the catalytic triad is located in an electronegative environment (Fig. 5). In FVa, both Arg306 and Arg506 are in electropositive regions. However, moving towards the C-terminus, the prime side of Arg306 is first positive then neutral and finally positive, while the prime side of Arg506 is electronegative (Fig. 5). At this site, two electronegative zones that possibly interact with APC involve residues Asp513, Asp578 and Asp577 (site 1) and Glu323, Glu374, Asp373, and Glu372 (site 2). A very negatively charged segment following Cys656, which is the last residue in the homology model based on ceruloplasmin, formed by Asp659-Asp660-Asp661-Glu662-Asp663 may also play a role during the docking of APC on to the FVa Arg506 site. In contrast, the segment containing Arg306 protrudes outside the surface of FVa, and, besides the residues directly fitting into the catalytic groove, there are no obvious interacting regions for APC.
Docking heparin-like molecule onto APC
As numerous problems have been noticed when trying to dock heparin at the surface of a protein , we used three different methods. To facilitate calculations, a negatively charged peptide mimicking the overall charge distribution and shape (see Materials and Methods) of a heparin molecule was created and the validity of this approach was tested. The validated protocol was next applied to dock the heparin-like peptide on to APC. The lowest conformation energies positioned the long axis of the peptide along a small electropositive groove formed by loops 37 and 70 (Fig. 6). These orientations are compatible with known experimental data suggesting that APC loops 37, 60 and 70 are directly or indirectly involved in heparin binding . In our structural model, loop 60 had no direct contact with the negatively charged peptide, but the distance between Lys62 and Lys63 and negative groups on the peptide (6–8 Å) was compatible with electrostatic interactions and preorientation of heparin on the APC surface during formation of an encounter complex. However, it is important to note that direct contact could occur between the heparin-like peptide and loop 60 if flexibility had been allowed.
In the second docking approach, APC was maintained rigid during the simulation but a real heparin molecule (10 sugar units, length 40 Å) was used and flexibility was tolerated (Fig. 6, inset). The top ranking conformation positioned heparin against loops 37, 60 and 70. In this case, heparin also had direct contact with positively charged residues located on loop 60. The last approach followed a protocol used for virtual ligand screening, and also positioned the short sugar molecules in between loop 37 and loop 70 of APC.
Docking of APC on to FVa
To elucidate molecular interactions between APC and FVa at cleavage sites Arg306 and Arg506, two theoretical docking protocols were used [30,38]. Because the segment containing Arg306 and Arg506 had to fit into the APC active site, and because we used essentially rigid-body docking, several conformations of the loops displaying Arg306 and Arg506 of FVa as well as the APC loop 148 were generated before docking computations. Test dockings were performed, and a theoretical FVa model with structural changes at the level of loop 306 and 506 was selected for further docking simulation (data not shown). Similarly, a model of APC in which the loop 148 partially covers the active site as compared with the structure present in the PDB file 1aut was selected (Fig. 6).
When using the docking method of Norel et al. , we noticed that, in the best-ranked complex, APC was positioned very close to the Arg506 site in a conformation basically suitable for cleavage at this site. Some interactive reorientations of APC were needed in order to remove steric clashes. This was performed using the orientation of a serpin reactive loop crystallized into a serine protease as guideline . Interestingly, in none of the predicted complexes was APC positioned next to the Arg306 site, suggesting that the shape complementarity in this region is not optimal. With the ICM method, models of the complex with APC docked at FVa position 506 or 306 could be generated that are in agreement with published experimental data. To develop our two final models of the FVa–APC complex, we merged the two sets of computations reported above and performed limited interactive reorientations of the two proteins to remove minor steric clashes (Fig. 7).
When APC is docked at position 506, the electropositive loops 37, 70 and 148 of APC seem to have contact with the electronegative area in FVa formed by residues Asp513, Asp578 and Asp577. APC could also interact with FVa residues Asp659-Asp660-Asp661-Glu662-Asp663, but these residues are not present in the model as it was only possible to predict the structure of the FVa A2 domain up to residue 656. FVa residues Glu323, Glu374, Asp373 and Glu372 could facilitate the docking process, but the role of these residues is unknown. In the present model, APC loop 60 does not seem to have significant contact with FVa. When superimposing the APC–heparin complex on to the APC–FVa complex, we observed that heparin clashes against FVa and/or could be very close to negatively charged FVa residues Asp513, Asp578 and Asp577 and/or Asp659-Asp660-Asp661-Glu662-Asp663 (Fig. 7A). Contact between FVa and APC loop 148 only occurs when the 148 loop is in a partially closed conformation. In the PDB file 1aut, the 148 loop is open and with this conformation, very limited direct contacts with FVa were noted.
When APC is docked at position Arg306, only the electropositive segment containing FVa Arg306 seems to complement well the electronegative catalytic groove of APC. The loops surrounding the active site of APC do not seem to make strong contact with FVa besides the segment to be cleaved (i.e. no major contacts besides the P5 to P5′ segment were observed). When the APC–heparin complex is superimposed on APC docked at position Arg306, enough room appears available for heparin, and heparin could even bridge FVa and APC. Some positively charged residues on FVa such as Lys320, Arg321, Arg400 and Arg501 could possibly contact heparin during this reaction.
The recent demonstration of a heparin-binding site in the SP domain of APC that overlaps with a secondary binding exosite for FVa in APC [15,17,21] prompted us to reinvestigate the effect of heparin on the inactivation of FVa by APC. We find that heparin has specific effects on the inactivation of FVa by APC that have not been observed before.
Thus, APC-mediated cleavage at Arg506 in FVa is inhibited up to 12-fold by heparin, whereas heparin was found to stimulate the slower cleavage at Arg306 severalfold if the Arg506 site had not been cleaved first. Inactivation of FV and FVa occurs via different mechanisms, with the order of cleavage at Arg506 and Arg306 being reversed during FV inactivation by APC , i.e. in FV, cleavage at Arg306 precedes cleavage at Arg506. Given that Arg306-cleaved FV is no longer capable of acquiring FVa cofactor activity, subsequent inhibition of Arg506 cleavage by heparin will not further affect the activity of FV. Therefore, our data are compatible with the reported enhancement of APC-catalyzed FV inactivation by heparin [22,23]. However, in contrast with Petäja et al., we found heparin to inhibit the APC-catalyzed inactivation of FVa, i.e. of activated FV. This discrepancy may be due to the fact that the concentrations of heparin used in our study were higher and the assay conditions were different from those used by Petäja and coworkers.
The inhibitory effect of heparin was dependent on the concentration of heparin, half-maximal inhibition being observed at ≈ 2 IU UFH·mL−1, which is higher than the concentrations generally used in a clinical setting. Therefore, our data should primarily be interpreted in a structure–function context, heparin being considered as a probe for investigation of the interaction between APC and FVa. However, given the general abundance of heparin-like structures on the surface of the vascular bed, and the lack of good estimates of local concentrations of coagulation proteins or reactants during hemostatic reactions, our data may still have clinical ramifications.
There are two possible molecular explanations for the inhibition of the Arg506 cleavage by heparin: (a) direct binding of heparin to FVa close to the Arg506 site; (b) binding of heparin to APC. Heparin binds to a cluster of basic residues that are present on three conserved surface loops in the SP domain of APC (loops 37, 60 and 70) [17,21,26]. Loops 37 and 70 have been shown to contain exosites that are important for the cleavage at Arg506 but not for cleavage at Arg306 [15–17]. The observation that heparin does not inhibit FVa inactivation by recombinant APC variants that are unable to bind heparin supports the mechanism in which heparin binding to the basic cluster in APC is responsible for the observed inhibitory activity. A direct effect of heparin on APC can also explain the stimulation of the cleavage at Arg306. On the basis of molecular modeling, we propose that, similar to the mechanism by which heparin promotes PCI–APC complex formation, heparin bridges APC to positively charged protein patches in the neighborhood of Arg306, thereby facilitating its cleavage by APC.
Because cleavage at Arg506 by the 37-loop APC mutants is unaffected by heparin and as the activity of FVa in the prothrombinase complex was not influenced by the presence of heparin , we suggest no major direct influence of heparin on the quality of FVa as a substrate for APC but we do not exclude interactions of heparin with other sites in FVa.
In FVa, the Arg506 cleavage site is predicted to be located on a short loop, whereas Arg306 is positioned on a relatively extended structure. As some loops surrounding the active site of APC protrude relatively far into the solvent, it is expected that contacts between APC and FVa are not confined to those between the APC active site and the main S and S′ 506 subsites only. Exosites on both APC and FVa should be found when docking Arg506 into the active site of APC, whereas there might be fewer (if any) exosites involved when Arg306 docks into APC. Two theoretical molecular models for APC docking to the Arg506 and Arg306 sites were developed and analyzed taking into account our APC–heparin model and different conformations for APC loop 148 (Figs 6 and 7). When APC docks at the Arg506 site, the electropositive loops 37 and 70 of APC are in contact with the electronegative area in FVa formed by residues Asp513, Asp578 and Asp577. The contacts noted in our structure are also consistent with the model of APC manually positioned at the FVa Arg506 site . The role of FVa residues Glu323, Glu374, Asp373 and Glu372 with regard to APC positioning at Arg506 remains to be investigated. This also applies to a cluster of residues (Asp659-Asp660-Asp661-Glu662-Asp663; not present in the FV model) proposed by Pellequer et al.  to interact with APC. Interestingly, when we positioned our APC–heparin model on to FVa at position Arg506, heparin was very close to the FVa electronegative regions 513–578–577 and possibly Asp659-Asp660-Asp661-Glu662-Asp663 and could directly crash into FVa (i.e. fully compatible with the observed decrease in k506).
Evaluation of the FVa Arg306 site gave a different pattern from that observed for the Arg506 site. The size of the Arg306 loop and the lack of negatively charged exosites surrounding Arg306 does not allow clear contacts with APC loops 37, 60, 70 and 148. In our model, these loops point toward neutral regions of FVa, yet are relatively distant from the molecular surface of FVa. When superimposing our APC–heparin complex on the APC–FVa complex at Arg306, we noticed that heparin could bridge APC and FVa (i.e. consistent with increased k′306). The exact region of FVa involved in contacting heparin when APC is positioned at Arg306 is not known, but we hypothesize that Lys320, Arg321, Arg400 and Arg501 of FVa play a role.
In conclusion, our experimental findings, taken together with the structural bioinformatics analyses, indicate that binding of heparin to loop structures 37, 60 and 70 in APC impairs the interaction of APC with FVa during the APC-catalyzed cleavage at Arg506. Loops 37, 70, and 148 of APC are, according to our structural models, which must be regarded as speculative, proposed to contact FVa when the cleavage site at Arg506 is approached and this contact is directly perturbed by heparin. Negatively charged regions are not present in the close vicinity of Arg306, and we found no structural basis for a contact between exosites in APC and FVa during approach of the cleavage site at Arg306 which probably contributes to the lower rate of cleavage at Arg306 compared with that at Arg506.
This work was supported by grant no. 902-26-227 from the Dutch Organization for Scientific Research (NWO; to G.N.), by grants from the Swedish Research Council (07143), the Söderberg Foundation, the Österlund's Foundation, the Påhlsson Trust, a Marie Curie training grant QLK5-CT-2000–60007 (to K.W.S.) and research funds from the University Hospital, Malmö. Support from INSERM and La Region Ile-de-France are greatly appreciated. We would like to thank Dr Theo Lindhout for helpful suggestions and for reading the manuscript. We are grateful to Dr Fernandez-Recio and Molsoft scientists for suggestions for running the ICM molecular modeling package, and to Dr R. Norel for advice on running the docking package PPD. We also thank OpenEye Scientific Software for providing the fred package.