Ramón Montes and José Hermida, Laboratory of Thrombosis and Haemostasis, Division of Cardiovascular Sciences, Centre for Applied Medical Research (CIMA), University of Navarra. Avenida Pío XII 55, 31008 Pamplona, Spain. E-mail: firstname.lastname@example.org; email@example.com
Traces of activated factor VII (FVIIa) are required to maintain haemostasis. Activated factor X (FXa) is the main activator of FVII in the absence of tissue factor. However, little is known about how this mechanism is regulated. We and others reported the interaction between FVII and the endothelial cell protein C receptor (EPCR). We have analysed the role of EPCR in the FXa-dependent FVIIa generation. Activation was performed on the surface of human aortic endothelial cells in the presence or absence of a blocking anti-EPCR monoclonal antibody (mAb). Western-blot analyses revealed that FVII activation was increased twofold upon EPCR blocking. Kinetic analyses revealed that blocking doubled the catalytic efficiency for activation. Protein C was unable to mimic the effect of the anti-EPCR mAb on activation. Surface plasmon resonance experiments revealed that binding of EPCR and phospholipids to FVII were mutually exclusive. The 50% inhibitory concentration value for phospholipids to reduce the binding of FVIIa to EPCR was 57·67 ± 0·11 μmol/l. Immunofluorescence experiments showed that EPCR and phosphatidylserine are located at different regions of the cell surface. We propose that EPCR downregulates FVII activation by moving it from phosphatidylserine-rich regions. In summary, this study described a new anticoagulant role for EPCR.
Activated factor VII (FVIIa) is a serine protease that, upon binding to tissue factor (TF), triggers the coagulation cascade by activating factors X (FX) and IX (FIX) (Bom & Bertina, 1990; Komiyama et al, 1990). When the endothelial lining is intact and a thrombin clot is not required, TF is not in contact with clotting factors. However, tiny amounts of FVIIa circulate in plasma under these conditions, and are necessary to allow the coagulation cascade to proceed rapidly once the vessel is injured and TF comes into contact with circulating factors (Mackman, 2006). FVIIa generation on the unperturbed endothelium must be finely tuned to avoid overcoagulation. However, little is known about how this mechanism is regulated. FVIIa traces are the consequence of FVII activation by several serine proteases. Among these, activated factor X (FXa) plays a pivotal role (Butenas & Mann, 1996).
We and others recently reported that the endothelial cell protein C/activated protein C (APC) receptor (EPCR) bound FVII with high affinity on the endothelial cell surface (Ghosh et al, 2007; López-Sagaseta et al, 2007). As EPCR is highly expressed in the vasculature (Laszik et al, 1997; Fukudome et al, 1998), we hypothesised that EPCR, through interaction with FVII, could influence the extent of FXa-dependent FVIIa generation. The experiments presented herein provide evidence that EPCR plays a so far unsuspected anticoagulant role by preventing FVII binding to phosphatidylserine (PS) and its subsequent activation by FXa.
Materials and methods
Soluble EPCR (sEPCR) and R87A sEPCR were expressed in Pichia pastoris and purified as described elsewhere (López-Sagaseta et al, 2007). FVII, FXa and protein C (Enzyme Research Laboratories, South Bend, IN, USA) were human plasma-purified. FVIIa (Novoseven®; Novo Nordisk, Bagsvaerd, Denmark) was of recombinant origin. Antithrombin III (Kybernin®) was from ZLB Behring, Marburg, Germany. Fondaparinux (Arixtra) was from NV Organon, Oss, the Netherlands; Sanofi-Synthélabo, Paris, France. S-2366 was from Chromogenix (Milan, Italy). Anti-EPCR monoclonal antibodies (mAbs) RCR-252 and RCR-2 were kindly provided by Dr. Fukudome (Saga Medical School, Japan) and are described elsewhere (López-Sagaseta et al, 2007). A blocking mAb against TF was from American Diagnostica (Stamford, CT, USA) and a non-reactive rat IgG1 mAb was from Becton Dickinson (BD Biosciences, San José, CA, USA). Polyclonal antibody (pAb) against FVII/VIIa was from Abcam (Cambridge, UK). A fluorescein isothiocyanate (FITC)-conjugated pAb against rat immunoglobulins was from BD Biosciences. A thromboplastin reagent (Innovin, Dade Behring, Schwalbach, Germany) was used for FVII activation. Phosphatidylcholine (PC), phosphatidylethanolamine (PE) and PS were from Avanti Polar Lipids Inc (Alabaster, AL, USA). Alexa 568-conjugated annexin V was from Invitrogen (Paisley, UK) Primary human aortic endothelial cells (HAEC; Lonza, Basel, Switzerland) were allowed to grow to 90% confluence in endothelial basal medium-2 (EBM-2) enriched with the appropriate supplements provided by the manufacturer (Lonza) and used for experiments between passages 6 and 10. EA.hy926 cells were provided by C.J. Edgell, University of North Carolina, Chapel Hill, NC, USA.
Activation of FVII by FXa on the endothelial surface
To monitor the effect of EPCR on FVII activation in the unperturbed endothelium we used HAEC, because they express on their surface large amounts of EPCR as well as enough PS to permit colocalization of Gla domain clotting factors. Ten nanomole per litre FVII and 1 nmol/l FXa were incubated for 30 min with HAEC in a 96-well culture plate at room temperature (RT) in 10 mmol/l HEPES, pH 7·4, supplemented with 150 mmol/l NaCl, 5 mmol/l CaCl2, 0·6 mmol/l MgCl2, 0·1% PEG 6000 (HEP-Ca-Mg). Activation was arrested by adding 12·5 mmol/l EDTA. The effect of 20 μg/ml blocking anti-EPCR mAb (RCR-252) or 60 nmol/l protein C on FVII activation was studied. The effect exerted by 20 μg/ml non-reactive rat IgG1 mAb (isotype control) or by 1 μmol/l annexin V on activation was also studied. Activation was analysed by Western blot. The samples were submitted to sodium dodecyl sulphate polyacrylamide gel electrophoresis under reducing conditions in 4–12% gradient gels (Invitrogen, Paisley, UK), transferred to a nitrocellulose membrane (GE Healthcare, Little Chalfont, United Kingdom) and probed with the anti-FVII/VIIa sheep pAb. After subsequent addition of a biotin-labelled anti-sheep pAb and streptavidin-horseradish peroxidase (GE Healthcare), development was performed using ECL™ Western Blotting Detection Reagents (GE Healthcare). Bands of 50 and 30 kDa, corresponding to FVII and the heavy chain of FVIIa respectively, were detected and their signal intensities densitometrically determined in a Chemidoc XRS densitometer (Bio-Rad, Hertfordshire, UK). The Mann–Whitney U-test was used for data analyses.
To study the effect of EPCR blocking on the catalytic efficiency of FXa-dependent FVII activation, we used EA.hy926 cells (kindly provided by Dr. Edgell, University of North Carolina at Chapel Hill, USA), which express EPCR on their surface to an extent similar to HAEC (not shown). A quantity of 15–1000 nmol/l FVII was activated for 10 min with 1 nmol/l FXa in the presence of EA.hy926 cells, with or without 100 μg/ml RCR-252. Prior to the addition of 2 mmol/l S-2366 to monitor FVIIa generation, FXa was inhibited by a mixture of 50 nmol/l antithrombin III and 1 U/ml Fondaparinux. The kinetic parameters [Michaelis–Menten constants (Km) and maximum rates (Vmax)] were calculated using ENZFITTER (Biosoft, Cambridge, UK). For catalytic constant (Kcat) calculation purposes, a curve with known amounts of FVIIa and the corresponding maximum absorbances of proteolyzed S-2366 was constructed.
Effect of sEPCR on the activation of FVII by FXa
In order to study the role of sEPCR in the FXa and PS-dependent activation of FVII, we used a thromboplastin reagent, Innovin, because it exhibits TF on its surfaceand provides such a PS-rich environment that it is possible to set up FVII activation to be PS-dependent only. We used Innovin at 1:16 dilution in HEP-Ca-Mg and 750 nmol/l FVII to ensure maximum PS availability. Five nanomole per litre FXa was added and activation was allowed to proceed for 5 min at RT with or without 1 μmol/l sEPCR. Additional controls were performed in the presence of 1 μmol/l boiled sEPCR, 1 μmol/l R87A sEPCR, 1 μmol/l annexin V or 50 μg/ml of blocking anti-TF mAb. Immediately after activation had finished, the samples were electrophoresed to analyse FVIIa generation by Western blot as described above.
Phospholipid vesicles preparation
A phospholipid mixture consisting of 1·34 μmol PC, 1·51 μmol PE, and 0·54 μmol PS (40:40:20) in 25 ml acetone was prepared and subjected to a coaxial turbulent injection with 26 ml of a solution containing 20 mmol/l HEPES and 100 mmol/l NaCl. The sample was placed in a rotary evaporator until a final 2·6 ml volume was obtained. The resulting vesicle preparation was homogeneous and vesicles exhibited a mean diameter of 156 nm, as measured with a Malvern Nanosizer (Malvern, UK).
Surface plasmon resonance (SPR) experiments
All interaction experiments were carried out by SPR technology in a BIAcore X Biosensor (GE Healthcare).
Interaction between sEPCR and FXa
To study whether FXa was able to interact with sEPCR, we proceeded essentially as previously described (López-Sagaseta et al, 2007). Two-hundred resonance units (RU) of sEPCR were captured with the RCR-2 mAb directly immobilised on the flow channel 2 of a CM5 sensor chip (GE Healthcare). Flow channel 1, which was loaded with RCR-2 only, was used as the reference cell. Then, 100 nmol/l FXa in 10 mmol/l HEPES, pH 7·4, supplemented with 150 mmol/l NaCl, 0·005% Tween-20, 3 mmol/l CaCl2 and 0·6 mmol/l MgCl2 (HBS-P*), was injected over both surfaces and the response was monitored.
Effect of phospholipid vesicles on the interaction between sEPCR and FVIIa
Forty-five nanomole per litre FVIIa was preincubated for 20 min at RT with increasing concentrations of PC:PE:PS (40:40:20) vesicles, and FVIIa binding to sEPCR was analysed by injecting the preincubated samples over 200 RU of RCR-2-captured sEPCR in HBS-P* buffer at 30 μl/min. After each injection, the RCR-2 surface was regenerated by injecting 10 μl of 10 mmol/l glycine, pH 1·5. The responses in RU at the end of the injections were plotted versus phospholipids concentrations and the IC50 value, i.e. the phospholipid concentration needed to reduce the binding of FVIIa to EPCR to half, was obtained by nonlinear regression analysis of the curve.
To study the localization of EPCR and PS on the cell surface, HAEC were plated in 4-well chambered glass slides (Nunc Inc, Roskilde, Denmark). After 24 h the plates were incubated for 1 h at RT with 100 μg/ml anti-EPCR mAb (RCR-2) in binding buffer (10 mmol/l HEPES, pH 7·4, 150 mmol/l NaCl, 5 mmol/l KCl, 1 mmol/l MgCl2 and 2·5 mmol/l CaCl2). To prevent the internalisation of EPCR, 0·02% NaN3 was also added. After washing, they were incubated for 1 h at RT with FITC-conjugated anti-rat IgG and/or 1:50-diluted Alexa 568-conjugated annexin V for 5 min in binding buffer. After washing with binding buffer, fixation was performed for 15 min at RT with 4% paraformaldehyde (PFA) solution [4% (w/v) PFA, 10 mmol/l HEPES, 150 mmol/l NaCl, 2·5 mmol/l CaCl2, pH 8·0]. Control staining was performed using the same method without the first antibody. After washing with phosphate-buffered saline, cells were mounted in 90% glycerol with 0·02 mol/l Tris-HCl, pH 8, 0·002% NaN3, 2% 1·4 diazabicyclo (2,2,2)-octane (DABCO; Merck, Whitehouse Station, NJ, USA) and 0·5 μg/ml DAPI (4′,6′-diamidino-2-phenylindole hydrochloride) to stain the nuclei. The immunostained cells were viewed under Nikon Eclipse 80i microscope (Nikon, Tokyo, Japan) at 40× optical lens at RT. Images were acquired by scanning the cells sequentially with 200 μm increments using the DXM 1200F camera with ACT-1 version 2.63 software (Nikon). Carl Zeiss LSM 510 Processing software (Carl Zeiss, Oberkochen, Germany) was used for measuring the colocalization of EPCR and PS by determining the correlation coefficient of the green and the red fluorescence. A total of 20 cells from three experiments were randomly selected for colocalization analysis.
EPCR inhibits the activation of FVII by FXa on the endothelial cell surface
Activation of FVII was performed on the surface of HAEC. Activation was efficiently abrogated by annexin V. Residual activation in its presence was most probably due to the annexin V amount used, being insufficient to totally preclude binding of FVII and FXa to PS. Therefore, activation was, as expected, dependent on PS availability. When EPCR existing on the cell surface was inhibited by a blocking anti-EPCR mAb (RCR-252) at a saturating concentration (López-Sagaseta et al, 2007), FVIIa generation underwent a twofold increase (Fig 1). An isotype control mAb did not influence FVII activation, indicating that activation was EPCR-specific. EPCR therefore seems to play a down-regulating role towards FVII activation on the endothelial surface.
FVII and protein C interact with EPCR with similar affinity, involving the same residues of the latter (i.e. binding of one ligand would preclude the binding of the other one) (López-Sagaseta et al, 2007). For this reason, we tested whether the EPCR-dependent down regulation of FVII activation was still seen when both FVII and protein C were present at physiological concentrations, i.e. 10 and 60 nmol/l respectively. Importantly, FVII activation was not increased under these conditions (Fig 1), which means that EPCR prevented FVII activation even in the presence of protein C, and therefore its new anticoagulant role may be of physiological significance. Autoactivation of FVII was excluded because FVIIa generation was not detected in the absence of FXa (Fig 1). The activation of FVII by FXa was not dependent on the presence of TF because a blocking anti-TF mAb did not influence the activation (data not shown). To better assess the extent of the effect of EPCR towards FVII activation on the endothelial surface, we compared the catalytic efficiency of FXa in the presence and absence of RCR-252. In agreement with the result presented above, prevention of FVII binding to EPCR enhanced the catalytic efficiency of FXa twofold. This effect was achieved by a decrease in the Km upon EPCR blocking, the Kcat remaining unaltered (Table I).
Table I. Effect of EPCR blocking on the kinetic constants of FXa-mediated activation of FVII on endothelial surface.
Kcat/Km [/s/(mol/l)] × 105
15–1000 nmol/l FVII was activated for 10 min with 1 nmol/l FXa, with or without 100 μg/l RCR-252 monoclonal antibody (mAb) in the presence of EA.hy926 cells (1 × 104/μl). The chromogenic activity of the generated FVIIa towards 2 mmol/l S-2366 was subsequently determined. The values of the kinetic constants correspond to the mean ± SD of at least two determinations.
634·0 ± 194·7
0·48 ± 0·07
7·7 ± 1·2
318·8 ± 64·9
0·42 ± 0·06
13·1 ± 0·7
EPCR does not significantly interact with FXa
In order to determine the mechanism explaining the effect of blocking EPCR on FVII activation, we first analysed the interaction between FXa and EPCR. SPR experiments showed that FXa did not significantly interact with sEPCR (Fig 2A), thus indicating that the inhibitory effect of EPCR must be due to mechanisms other than interference with FXa activity.
EPCR modulates FVII activation by inhibiting its binding to phospholipids
FVII binding to PS-containing phospholipid vesicles improves the ability of FXa to generate FVIIa (Butenas & Mann, 1996). Given that both EPCR and PS bind FVII through the Gla domain, we hypothesised that the effect of EPCR on FVII activation by FXa could rely on its ability to prevent the interaction between FVII and PS. The SPR experiments directly showed that PS-containing phospholipid vesicles and sEPCR competed for binding to FVIIa in a dose-dependent manner (Fig 2B). The IC50 value for phospholipid vesicles to reduce to the half the binding of FVIIa to sEPCR was 57·67 ± 0·11 μmol/l, which is remarkably high since EPCR exhibits a much higher affinity than PS for FVIIa. These findings apply to FVII as well, since its Gla domain-dependent interactions with phospholipids and EPCR are identical to those of FVIIa (López-Sagaseta et al, 2007).
In order to confirm that EPCR was able to prevent FVII binding to phospholipids, we performed additional experiments using Innovin vesicles under conditions such that FVII activation by FXa was exclusively PS-dependent. When using FVII at 750 nmol/l, an amount enough to reach maximum PS-binding, sEPCR, at a roughly equimolar concentration, 1 μmol/l, reduced the activation of FVII by 50%. Appropriate controls [boiled sEPCR, R87A sEPCR (unable to bind FVII/VIIa)] confirmed the specificity of the inhibition (Fig 2C). Addition of annexin V and blocking anti-TF mAb confirmed that activation was PS-dependent. Taking these results all together, it is conceivable that EPCR may reduce the activation of FVII by preventing its interaction with negatively charged phospholipids.
EPCR on the endothelial surface moves FVII further away from negatively charged phospholipids
On the cell surfaces, PS molecules bring the Gla domain-containing coagulation factors near each other. EPCR is believed to be located in the caveolae (Bae et al, 2007), regions of the membrane, which are rich in sphingolipids and cholesterol while poor in phospholipids. Immunofluorescence experiments were performed to analyse the localization of EPCR and PS on the endothelial surface. FITC-labelled RCR-2 anti-EPCR mAb strongly stained the cell surface of HAEC (Fig 3, left panel); a weaker signal was obtained with Alexa 568-conjugated annexin V, according to the presence of PS molecules, rather modest in unperturbed endothelium (Fig 3, centre panel); interestingly, when cells were incubated with RCR-2 and annexin V simultaneously, the merge of the two signals was poor (Fig 3, right panel) as confirmed by the poor correlation coefficient (0·23 ± 0·02). The image analysis performed in this study enabled us to objectively state that EPCR and PS are not near each other on the endothelial cell surface. This result provides a mechanism to explain how EPCR, by interacting with FVII, prevents the latter from concentrating in the regions of the membrane where PS-bound FXa is present.
Activation of coagulation is tightly regulated. Normally, TF is not in contact with clotting factors and thus thrombin generation is not encouraged. Nevertheless, in order to allow rapid activation of coagulation upon requirement, small amounts of FVIIa must always be available even when the endothelium is unperturbed (Nemerson, 1988). Under these circumstances, the activation of FVII by FXa is the main source of FVIIa (Butenas & Mann, 1996). Little is known about the regulation of this mechanism, which must be finely tuned to avoid FVII overactivation. We provide evidence that the EPCR expressed at the surface of the endothelial cells can reduce the activation of FVII by FXa. Blocking the binding of FVII to EPCR increased the generation of FVIIa twofold on the surface of HAEC, as shown by the Western blot experiments as well as by the catalytic efficiency calculations. FVII and protein C bind to EPCR with similar affinity (Preston et al, 2006; Ghosh et al, 2007; López-Sagaseta et al, 2007). As plasma protein C levels are sixfold higher than those of FVII, one could speculate that protein C could obscure the effect of EPCR on FVII activation. However, the down-regulating effect of EPCR towards FVII activation was also seen in the presence of protein C, when protein C and FVII were at physiological concentrations. This finding supports that the anticoagulant role played by EPCR through FVII binding is relevant even in the presence of protein C. Nevertheless, the theoretical ability of protein C to displace FVII from PS, which would hamper FVII activation, could also contribute to partly explain this result. This observation leads us to suggest that EPCR does play an anticoagulant role by avoiding overactivation of FVII on the unaltered endothelial surface i.e. in the absence of TF, when only trace amounts of FVIIa are needed to maintain haemostasis.
In the search for a mechanism to explain our findings, we first performed SPR experiments to discard the possibility that the effect of EPCR was due to a direct interaction with FXa. No interaction was observed, in agreement with previous results (Fukudome & Esmon, 1994; López-Sagaseta et al, 2007), indicating that the increase of FVII activation upon EPCR blocking must rely on the prevention of the interaction between FVII and EPCR. We then examined whether such an interaction could influence the role played by PS in the activation of FVII by FXa. PS molecules on the cell surfaces interact with the Gla domain-containing factors, positioning them near each other (Zwaal et al, 1998). For this reason, PS plays a key role in enabling the activation of FVII by FXa. Although the expression of PS on the outer leaflet of the membrane increases dramatically upon vessel injury (Zwaal et al, 2005), the intact endothelium exposes enough PS to allow FVIIa generation. Given the background that EPCR inhibits factor Va inactivation by APC through preventing the interaction with PS (Liaw et al, 2000), we hypothesised that EPCR would also prevent the interaction between FVII and PS. We set up SPR experiments to answer this question and observed that PS-containing phospholipid vesicles blocked the interaction between FVII and sEPCR, confirming the hypothesis that EPCR and PS are mutually exclusive in binding to factor VIIa. The large number of PS-containing vesicles needed to prevent the interaction between FVII and EPCR (IC50∼60 μmol/l) is the consequence of the much higher affinity of FVII for EPCR than for PS and reveals that EPCR prevents binding of FVII to PS efficiently.
EPCR is believed to localise in the caveolae of the cell membrane (Bae et al, 2007), regions which are poor in phospholipids, i.e. where PS expression is scarce. Therefore, it was conceivable that EPCR and PS might not be close to each other on the cell surface. The immunofluorescence experiments confirmed this theory: EPCR was not present within the PS-containing regions. Therefore, EPCR moves FVII further away from PS, where FXa anchors, thus avoiding FVIIa generation. The results obtained in the kinetic study of the effect of blocking EPCR on FVII activation support this mechanism: the increase in the catalytic efficiency of FXa upon EPCR blocking was due to a reduction in the Km while the Kcat remained unaltered, which means that EPCR decreased the affinity between FXa and FVII, i.e. enzyme and substrate must be allocated in different regions of the cell surface.
A recent study has demonstrated the association between higher levels of circulating FVII/VIIa and reduced amounts of EPCR on the surface of endothelium when EPCR shedding is increased (Ireland et al, 2009), which suggests that binding of FVII and the modulation of its activation by EPCR herein described may have physiological significance. Therefore, one could speculate that the H3 haplotype of the gene encoding EPCR (PROCR), which is associated with increased sEPCR levels in plasma (Saposnik et al, 2004; Ireland et al, 2005; Navarro et al, 2008) and shedding in vitro (Ireland et al, 2005; Qu et al, 2006) would confer a higher thrombotic risk by reducing the ability of endothelium not only to activate protein C but also to protect FVII from activation. However, a series of epidemiological studies that aimed to assess the role of EPCR haplotypes in venous or arterial thrombosis have yielded conflicting results (Medina et al, 2004, 2005; Saposnik et al, 2004; Uitte de Willige et al, 2004; Ireland et al, 2005; Navarro et al, 2008). The fact that sEPCR decreases the ability of FVIIa to activate FX (López-Sagaseta et al, 2007), could help to explain the inconsistency among the findings of these works. Patients bearing the H3 haplotype of PROCR would exhibit higher sEPCR levels, which would partially prevent FX from activation by FVIIa, thus counteracting the procoagulant effects.
In summary, we propose that EPCR may play a hitherto unsuspected anticoagulant role by inhibiting FXa-dependent FVII activation. Pathological conditions predisposing to low EPCR exposure on the endothelial surface, such as atherosclerosis (Laszik et al, 2001), or the presence of blocking anti-EPCR autoantibodies (Hurtado et al, 2004), may disrupt the haemostatic balance by favouring not only the lower activation of protein C, but also an increase in FVIIa-dependent procoagulant activity at times and sites where clots are not required.
Supported through the Unión Temporal de Empresas project CIMA and by grants from Instituto de Salud Carlos III (PI05/1178, Red Temática de Investigación RECAVA RD/0014/0008), Fundación Mutua Madrileña and from Health Department, Gobierno de Navarra (12/2006). Jacinto López-Sagaseta was supported by a fellowship from the Education Department, Gobierno de Navarra.
We thank Eva Molina and Maider Esparza for her excellent technical assistance.