The Endothelial Protein C Receptor Supports Tissue Factor Ternary Coagulation Initiation Complex Signaling through Protease-activated Receptors

Abstract

Protease-activated receptor (PAR) signaling is closely linked to the cellular activation of the pro- and anticoagulant pathways. The endothelial protein C receptor (EPCR) is crucial for signaling by activated protein C through PAR1, but EPCR may have additional roles by interacting with the 4-carboxyglutamic acid domains of procoagulant coagulation factors VII (FVII) and X (FX). Here we show that soluble EPCR regulates the interaction of FX with human or mouse tissue factor (TF)-FVIIa complexes. Mutagenesis of the FVIIa 4-carboxyglutamic acid domain and dose titrations with FX showed that EPCR interacted primarily with FX to attenuate FX activation in lipid-free assay systems. In human cell models of TF signaling, antibody inhibition of EPCR selectively blocked PAR activation by the ternary TF-FVIIa-FXa complex but not by the non-coagulant TF-FVIIa binary complex. Heterologous expression of EPCR promoted PAR1 and PAR2 cleavage by FXa in the ternary complex but did not alter PAR2 cleavage by TF-FVIIa. In murine smooth muscle cells that constitutively express EPCR and TF, thrombin and FVIIa/FX but not FVIIa alone induced PAR1-dependent signaling. Although thrombin signaling was unchanged, cells with genetically reduced levels of EPCR no longer showed a signaling response to the ternary complex. These results demonstrate that EPCR interacts with the ternary TF coagulation initiation complex to enable PAR signaling and suggest that EPCR may play a role in regulating the biology of TF-expressing extravascular and vessel wall cells that are exposed to limited concentrations of FVIIa and FX provided by ectopic synthesis or vascular leakage.

Introduction

Signaling through the G protein-coupled protease-activated receptors (PARs)² is crucial for cellular responses that sense the activation of the pro- and anticoagulant pathways. Although thrombin activation of PAR1 is governed by the direct recognition and protease-mediated cleavage of the receptor (1), other proteases rely on co-receptors of the integrin and immune receptor families for specificity and efficiency of PAR cleavage (2). The endothelial protein C receptor (EPCR), which belongs to the lipid-presenting CD1d receptor family, binds the Gla domain of protein C (PC) or activated PC (aPC) (3) and supports aPC-dependent activation of PAR1 (4) to achieve cytoprotective and anti-inflammatory activities in stroke, sepsis, and autoimmune diseases (5, 6). Cytoprotective aPC signaling is dependent on the colocalization of EPCR and PAR1 in lipid rafts (7, 8), but alternative receptors are increasingly recognized as contributing to cellular effects of aPC signaling in these microdomains (9–11).

EPCR also interacts with the highly homologous Gla domain of coagulation factor VIIa (FVIIa), but this interaction has not been linked to cell signaling (12, 13) and rather is considered to be important for transport of therapeutically administered FVIIa across the endothelium (14, 15). In addition, the interaction of EPCR with FVIIa is presumed to attenuate tissue factor (TF)-initiated FX activation, although EPCR has an opposite, rate-enhancing effect on the catalytic activity of the soluble TF-FVIIa complex (16). Human FVIIa interacts with murine EPCR (17), and the clearance of human FVIIa in mice is regulated by EPCR (14). However, murine FVIIa apparently lacks high affinity for murine EPCR (18), raising questions about the physiological significance of interactions of endogenous FVIIa with EPCR. Although FX inefficiently competes with FVIIa binding to cells (12), FXa activation of PAR1 has recently been shown to depend on EPCR expressed by endothelial or Chinese hamster ovary (CHO) cells (19), adding yet another potential mechanism by which EPCR may regulate procoagulant proteases.

The described interactions of aPC, FVIIa, and FXa with EPCR occur at protease concentrations in the nm range, which is probably not achieved in physiology. However, the concentration of zymogen forms of each of these proteases is sufficiently high to achieve at least partial saturation of EPCR in vivo. In addition, these proteases and zymogens engage in receptor-localized enzyme-substrate interactions that may synergistically enhance the overall affinity of multiprotein complexes. The apparent affinities for functionally relevant interactions of coagulation factors with EPCR are therefore difficult to predict solely based on measured affinities of individual components. For example, thrombin-thrombomodulin-mediated activation of EPCR-bound PC results in highly efficient EPCR-aPC-PAR1 signaling at a much lower concentration of aPC as compared with reactions to which aPC is added directly (20). Signaling of the ternary TF-FVIIa-FXa coagulation activation complex also occurs at lower concentrations of FXa when compared with reactions directly initiated by the addition of FXa (21, 22).

Whether EPCR plays a role in TF-dependent signaling has not been determined, but the demonstrated anticoagulant effects of EPCR on TF-initiated coagulation (16) may indicate functional interactions. It is unknown whether EPCR can regulate TF signaling by engaging the FVIIa or FX Gla domain. Distinct pools of TF support signaling of the binary TF-FVIIa versus the ternary TF-FVIIa-FXa coagulation initiation complex. TF with an unpaired Cys¹⁸⁶ can form a complex with PAR2, and in this non-coagulant TF-FVIIa binary complex, FVIIa cleaves PAR2 (22). In addition, the binary TF-FVIIa complex interacts with integrins that contribute to TF-FVIIa signaling activity in cancer cells (23). In contrast, signaling of the ternary TF-FVIIa-FXa product complex requires the allosteric Cys¹⁸⁶-Cys²⁰⁹ disulfide of TF and is mediated by FXa cleaving PAR1 or PAR2 (21, 22).

In this study, we show that EPCR is a functional component of the ternary but not the binary TF protease complex in commonly used human cell models of TF signaling. Reconstitution experiments directly demonstrate that EPCR induces more efficient cleavage of PAR1 and PAR2 by TF-FVIIa-FXa. Genetic approaches confirm that EPCR dependence of ternary complex signaling is a conserved feature in the mouse, emphasizing that mouse models are valid tools to study TF signaling pathways.

EXPERIMENTAL PROCEDURES

Mice

The following mice were extensively back-crossed into the C57BL/6J background and maintained under pathogen-free conditions: PAR1^−/−, PAR2^−/− (24), EPCR^low (25), and EPCR^flox/flox (Procr^flox/flox) (26). All animal procedures were carried out under approved protocols of The Scripps Research Institute Institutional Animal Care and Use Committee.

Proteins and Reagents

FXa, aPC, and thrombin were from Hematologic Technologies (Essex Junction, VT). Human or mouse soluble TF (sTF) was expressed in Escherichia coli and refolded (27, 28). FVIIa, FVIIa_DVQA, FVIIa Gla domain variants, and mouse FVIIa (mFVIIa) were produced as described previously (29–31). For expression of murine soluble EPCR (sEPCR), the C57BL/6-derived cDNA was modified by PCR to add six histidine codons at the 3′-end of the coding sequence. The cDNA was transfected into Drosophila melanogaster S2 cells using the calcium phosphate precipitation method, and expression was induced by addition of CuSO₄. Recombinant protein was purified by nickel-nitrilotriacetic acid affinity chromatography followed by Mono Q ion exchange chromatography and Superdex 200 (GE Healthcare) gel filtration chromatography. Purified proteins were kept in PBS, pH 7.4, containing 0.02% NaN₃ at 4 °C. Recombinant nematode anticoagulant protein 5 (NAP5) (32) and c2 (NAPc2) (33) were kindly provided by Dr. G. Vlasuk (Corvas International, San Diego, CA). ERK1/2 and phospho-ERK1/2 antibodies were obtained from Cell Signaling Technology (Danvers, MA), monoclonal anti-human EPCR-PE (RCR-252) was from BD Pharmingen, and polyclonal anti-mouse EPCR was from R&D Systems (Minneapolis, MN). Monoclonal anti-murine EPCR antibodies 1560 and 1562 (34), anti-human TF antibodies 5G9 and 10H10 (22), and cleavage blocking anti-PAR2 polyclonal antibody (22) have been described previously. Monoclonal and polyclonal antibodies to mouse TF were generated against recombinant soluble mouse extracellular domain. Anti-PAR1 antibodies WEDE15 and ATAP2 were kindly provided by Dr. L. F. Brass (35).

Cell Culture, Transfections, and Transductions

Human umbilical vein endothelial cells (HUVECs) and endothelial growth medium were obtained from Lonza (Walkersville, MD). HUVECs were maintained and transduced with TF and PAR2 adenoviral constructs as described previously (36). Human HaCaT immortalized keratinocytes were maintained in DMEM, 10% FBS. Lung smooth muscle cells (SMCs) were expanded from mouse lung endothelial cell preparations by propagation in medium lacking endothelial growth factors. Briefly, after collagenase digestion of lungs, cells were cultured in flasks coated with 0.2% gelatin (Sigma) in RPMI 1640 medium containing 20% heat-inactivated FCS, nonessential amino acids, 2-mercaptoethanol (Invitrogen), and antibiotics. For Cre-mediated deletion of the floxed EPCR allele in SMCs in vitro, cells were transduced on 2 consecutive days with 500 particles/cell adenoviral vector encoding Cre recombinase. Cells were used for experiments 48 h after transduction. CHO cells stably expressing human TF (CHO-TF) (37) were transfected with cDNA encoding human EPCR in pcDNA3.1, and CHO-TF/EPCR lines were selected for high expression of EPCR based on FACS analysis.

Surface Plasmon Resonance Analysis

The interaction between recombinant murine sEPCR and human FX, murine or human sTF, and FVIIa was characterized on a BIAcore 2000. sEPCR was immobilized by amine coupling to 200–800 resonance units on a Biacore CM5 sensor chip. Analytes at the concentrations indicated were injected in PBS, pH 7.4 supplemented with 2 mm Ca²⁺ or 2 mm Ca²⁺, 1 mm Mg²⁺. All experimental curves were normalized by subtraction of the response on a reference surface. Experimental data were fitted in the BIAevaluation 3.2 package using global fitting and a Langmuir 1:1 and two-state reaction binding model.

Functional Assays

FXa generation assays were performed in HBS (10 mm HEPES, pH 7.4, 137 mm NaCl, 5.3 mm KCl, 5.5 mm glucose, 1% BSA with 2 mm CaCl₂ or 2 mm CaCl₂ and 1 mm MgCl₂). Typically, 1 μm sTF and 10 nm FVIIa were used to activate 500 nm substrate FX, and sEPCR was added at the indicated concentrations as a competitor. For antibody inhibition experiments, 2 μm monoclonal anti-murine EPCR antibody 1560 or 1562 was added to the reaction mixture 10 min prior to addition of FX. Aliquots were quenched in 20 mm Tris, pH 7.4, 150 mm NaCl, 100 mm EDTA at defined times for chromogenic assay with Spectrozyme FXa (American Diagnostica). Amidolytic activity of the sTF-FVIIa complex was assayed in HBS, 1% BSA in the presence of 1 mm Spectrozyme FXa by recording the rate of hydrolysis in a kinetic plate reader at 405 nm. FXa generation on cells was analyzed in HBS buffer, 0.1% BSA, 2 mm CaCl₂, 1 mm MgCl₂. Samples at selected time points were quenched in 100 mm EDTA for FXa chromogenic assay.

Signaling Assays and Quantification of PAR Cleavage

Signaling assays in serum-free DMEM used ERK1/2 phosphorylation after 10 min or induction of immediate early gene mRNA levels after 90 min as readouts (38). Briefly, HaCaT cells or murine SMCs were switched for 3 h to serum-free DMEM followed by the addition of proteases in the presence of 2 μm hirudin to prevent thrombin signaling due to coagulation factor carryover from the culture medium (39). Cells were lysed in sample buffer for SDS-PAGE and Western blotting for phospho-ERK1/2, or mRNA was isolated from cells with TRIzol reagent (Invitrogen) for quantitative mRNA determinations (4). PAR cleavage was quantified in cells transiently transfected with FLAG-tagged PAR1 or PAR2 as described (38).

Imaging and Flow Cytometry Analysis

Cell surface expression of TF and EPCR was analyzed by two-color FACS of suspension cells stained with FITC-labeled anti-TF 5G9 or FITC-labeled anti-TF 10H10 and anti-human EPCR-PE on a FACSCalibur flow cytometer (BD Biosciences). Murine lung SMCs were washed three times with PBS, fixed with methanol, and stained with monoclonal anti-actin α-smooth muscle-Cy3 antibody (Sigma). Images were acquired with a Zeiss LSM 710 confocal microscope (Zeiss, Jena, Germany).

Statistical Analysis

Graphpad Prism Software (version 4.03) was used for one-way analysis of variance (ANOVA) followed by Bonferroni multiple comparison test to assess differences within groups. Two-way ANOVA followed by Bonferroni post test was used to determine differences between groups. Differences were considered significant when p was <0.05.

RESULTS

Mg²⁺-enhanced FX Activation by sTF-FVIIa Is Inhibited by EPCR

Previous studies showed that EPCR enhanced FVIIa amidolytic activity while attenuating TF-FVIIa-mediated FX activation and attributed the latter effect to inhibitory interactions of EPCR with the FVIIa Gla domain (16). Considering recent data that EPCR supports Gla domain-dependent FXa signaling (19), we revisited the effect of EPCR on the activity of the extrinsic FXase complex using sEPCR expressed in insect cells (Fig. 1A). Although initial experiments in a lipid-free system with sTF and FVIIa confirmed prior observations that sEPCR increased the amidolytic activity of FVIIa or of the sTF-FVIIa complex (16), sEPCR was without effect when BSA was added to the reaction to prevent nonspecific interactions of the FVIIa Gla domain (Fig. 1B). Therefore, all subsequent experiments were carried out in the presence of 1% BSA to prevent nonspecific protein binding. Because divalent cations have a marked effect on FX activation by TF-FVIIa (40), the effect of sEPCR on macromolecular substrate activation was characterized in the presence of 2 mm Ca²⁺ or 2 mm Ca²⁺, 1 mm Mg²⁺, a physiologically relevant ratio of these cations. sEPCR dose-dependently inhibited only the ∼4-fold faster rate of FXa generation in the presence of physiologic Ca²⁺/Mg²⁺ (Fig. 1C). The inhibitory effect of sEPCR under these conditions was significantly reversed by monoclonal anti-murine EPCR antibody 1560, which also blocks aPC binding to EPCR (34) (Fig. 1D). Borderline significant reversal of inhibition was seen with the non-blocking antibody 1562, which was much less potent than the inhibitory antibody 1560 in reversing EPCR inhibition of FX activation by soluble TF-FVIIa.

FIGURE 1.

EPCR specifically inhibits Mg²⁺-dependent activation of FX by sTF-FVIIa. A, proteins (5 μg) used shown by SDS-PAGE stained with Coomassie Blue: murine sEPCR, human (h) plasma-derived FX, recombinant human FVIIa, and sTF. B, effect of sEPCR (1 μm) on the amidolytic activity of sTF/FVIIa (1 μm/10 nm) determined with 1 mm Spectrozyme FXa in HBS, 1% BSA, 2 mm Ca²⁺, or 2 mm Ca²⁺ and 1 mm Mg²⁺ (mean ± S.D., n = 3; OD, optical density). C, inhibition of FX (500 nm) activation by sTF/FVIIa (1 μm/10 nm) by increasing concentrations of sEPCR was analyzed in HBS with 2 mm Ca²⁺ or 2 mm Ca²⁺, 1 mm Mg²⁺ (mean ± S.D., n = 3). D, inhibition of FXa generation by sEPCR (1 μm) was reversed by 2 μm inhibitory monoclonal anti-murine EPCR antibody 1560 but not the non-inhibitory anti-murine EPCR antibody 1562. Data were normalized to the non-inhibited control for each experiment (mean ± S.D., n = 6; ***, p < 0.001; *, p < 0.05; one-way ANOVA followed by Bonferroni post test). E, sEPCR inhibited FXa generation by wild-type FVIIa, Gla domain mutant FVIIa F4A/L8A, or mFVIIa in the presence of 1 μm species-compatible sTF (mean ± S.D., n = 3). F, effect of substrate concentration on sEPCR (1 μm)-mediated inhibition of FXa generation by sTF/FVIIa (1 μm/10 nm) (mean ± S.D., n ≥ 3).

The FVIIa Gla domain binds to human (12, 13, 16) and murine EPCR (17), but conflicting reports exist about the role of EPCR as a binding partner for FX/FXa (12, 19, 41). High affinity interaction of EPCR with PC requires Leu⁸ in the Ω loop of the Gla domain, and the Gla domains of human PC and FVII are highly conserved (13). To test whether interactions of FVIIa with EPCR were important for the observed functional inhibition, the major surface-exposed hydrophobic residues, Phe⁴ and Leu⁸, in the Ω loop of the FVIIa Gla domain were mutated to Ala. These mutations had no effect on the activation of FX in the lipid-free system or the sEPCR inhibition of the TF-FVIIa extrinsic FXase complex (Fig. 1E).

To exclude that the inhibitory effect of murine sEPCR on human sTF-FVIIa was a peculiarity of species mismatch, FX activation by the murine sTF-FVIIa complex was analyzed. sEPCR efficiently inhibited this reaction as well (Fig. 1E). Human sEPCR also inhibited the FX activation with human sTF-FVIIa (data not shown). Thus, EPCR inhibition was not species-specific, and contrary to previous conclusions (16, 41), the results with the mutant in the FVIIa Gla domain strongly argued that sEPCR did not primarily target the Gla domain of FVIIa but rather the Gla domain of FX to attenuate FX activation.

This conclusion was supported by sEPCR titrations shown in Fig. 1C because saturation was approached at stoichiometry of EPCR and FX. Conversely, increasing the FX concentration above stoichiometry with sEPCR dose-dependently reversed the inhibition of FXa generation by sEPCR (Fig. 1F), consistent with a primary interaction of sEPCR with FX.

FX Binding to EPCR Is Mg²⁺-dependent

Previous cell binding and surface plasmon resonance studies showed high affinity (K_D = 30–150 nm), Gla domain-dependent interaction of FVIIa with sEPCR (12, 13, 16), whereas no appreciable EPCR binding of FX (12, 16) or FXa (41) was reported. Considering the requirement for Mg²⁺ in enhanced FX activation and inhibition by sEPCR, we characterized the effect of divalent cations on FX binding to immobilized sEPCR by surface plasmon resonance. Binding of FX to sEPCR was not supported by Ca²⁺ and required the addition of Mg²⁺ (Fig. 2A), but no binding was observed with des-Gla FXa (data not shown), confirming Gla domain-dependent interactions. A concentration dependence of FX binding to sEPCR is shown in Fig. 2B, and the best data fit was obtained from a Langmuir 1:1 two-state reaction binding model yielding a calculated K_D of ∼292 nm, which is slightly higher than the physiological plasma concentration of FX (Fig. 2B).

FIGURE 2.

Surface plasmon resonance analysis of FX and TF-FVIIa binding to sEPCR. A, murine sEPCR was immobilized on a CM5 sensor chip, and binding of human FX (1 μm) in the presence of Ca²⁺ (2 mm) or Ca²⁺ and Mg²⁺ (2 and 1 mm, respectively) was monitored. B, binding of human FX to murine sEPCR in the presence of Ca²⁺/Mg²⁺ at 2-fold dilutions of FX from 2 μm to 125 nm. K_a, 6.83 × 10³ m⁻¹ s⁻¹; K_d, 1.99 × 10⁻³ s⁻¹; K_D, ∼292 nm. C, binding of human sTF (500 nm) and human FVIIa (500 nm) alone or in combination to immobilized murine sEPCR in the presence of Ca²⁺/Mg²⁺. D, dose response of 2-fold dilutions of human sTF/FVIIa (both at 500 nm) binding to immobilized sEPCR. K_a, 3.12 × 10⁴ m⁻¹ s⁻¹; K_d, 0.0022 s⁻¹; K_D, ∼70 nm. E, binding of mFVIIa (1 μm) and murine sTF (1 μm) alone or in combination to murine sEPCR in the presence of Ca²⁺/Mg²⁺. F, dose response curve of 2-fold dilutions of murine sTF/FVIIa (both at 1 μm) to immobilized murine sEPCR. K_a, 3.0 × 10⁴ m⁻¹ s⁻¹; K_d, 0.00485 s⁻¹; K_D, ∼160 nm. Colored lines represent experimental data, and black lines represent fittings. All traces are subtracted sensorgram traces (experimental sensorgram minus sensorgram on control surfaces). Each panel is representative of three to four independent experiments. RU, resonance units.

In line with previous results (16, 17), binding of FVIIa or the sTF-FVIIa complex was not strictly Mg²⁺-dependent, and human sTF-FVIIa bound to sEPCR with higher affinity (K_D ∼ 70 nm) than FX in the presence of Ca²⁺/Mg²⁺ (Fig. 2, C and D). The major difference between FX and the sTF-FVIIa complex was a slower association rate for FX, whereas the dissociation rates were similar for the human proteins. Although previous studies did not detect binding of murine FVIIa to EPCR (18), murine sTF-FVIIa (Fig. 2, E and F) bound to sEPCR albeit with somewhat lower affinity (K_D ∼ 160 nm) than human counterparts. Thus, in the presence of Ca²⁺/Mg²⁺, the TF-FVIIa binary complex and substrate FX are relevant ligands for sEPCR.

Overexpression of PAR2 in Endothelial Cells Promotes EPCR-dependent PAR2 Signaling

In addition to modulating the PC anticoagulant pathway, EPCR supports aPC-mediated PAR1 signaling in HUVECs and PAR1 or PAR2 signaling in heterologous expression systems (4). Although HUVEC-expressed EPCR binds aPC and FVIIa with similar affinity, EPCR-bound FVIIa has not been implicated in PAR signaling (12). We reasoned that low expression levels of PAR2, the signaling receptor of FVIIa (21), could explain the lack of significant FVIIa signaling responses in prior studies.

Although aPC signaling in HUVECs was PAR1-dependent, adenovirus-mediated overexpression of PAR2 (36) increased the signaling response to aPC stimulation (Fig. 3A). Inhibition of PAR cleavage by antibodies to PAR1, PAR2, or both showed that aPC activated both PAR1 and PAR2 in PAR2-overexpressing endothelial cells. Competition with active site-blocked FVIIa (FVIIai) was used to analyze whether the FVIIa Gla domain equally interacted with the binding sites that were utilized by aPC for PAR cleavage. FVIIai was a poor inhibitor of aPC-dependent PAR1 activation but reduced aPC-mediated signaling in PAR2-overexpressing cells to a similar extent as the PAR2 cleavage blocking antibody.

FIGURE 3.

Role of EPCR in signaling of TF complexes in endothelial cells. A, PAR activation by aPC in control, PAR2-, and TF/PAR2-transduced cells. Induction of the immediate early gene TR3 was measured at the mRNA level 90 min after stimulation with 10 nm aPC in the absence or presence of FVIIai, cleavage blocking anti-PAR1 antibodies (50 μg/ml WEDE15 and 25 μg/ml ATAP2), and/or anti-PAR2 polyclonal rabbit antibody (100 μg/ml) (mean ± S.D., n ≥ 3; ***, p < 0.001; **, p < 0.01; *, p < 0.05 relative to aPC alone; one-way ANOVA followed by Bonferroni post test for multiple comparisons). B, signaling of sTF/FVIIa was characterized in PAR2-transduced HUVECs. Signaling by FVIIa (10 nm) with or without sTF (1 μm) or of cofactor-independent, constitutively active FVIIa_DVQA (10 nm) was measured in the absence or presence of competition with anti-EPCR RCR-252 (20 μg/ml) or active site-blocked aPC (aPCi) (100 nm) (mean ± S.D., n > 3; **, p < 0.01; one-way ANOVA followed by Bonferroni post test for multiple comparisons). C, effect of FVIIa Gla domain mutations on signaling of sTF/FVIIa (500 nm/20 nm) in PAR2-transduced HUVECs. FVIIa_DVQA (20 nm) is shown as an additional control (mean ± S.D., n > 3). D and E, effect of anti-EPCR antibody RCR-252 (20 μg/ml) or the FXa inhibitor NAP5 (1 μm) on binary TF-FVIIa (FVIIa, 10 nm) or ternary TF-FVIIa-FXa (FVIIa, 0.5 nm; FX, 100 nm) complex signaling in TF- and PAR2-transduced HUVECs. FXa (5 nm) is used to show specificity for TF-dependent ternary complex signaling, and samples were taken after 10 min for measurements of FXa generation (E, right panel). FXa was quenched in all reactions after 10 min with NAP5, and mRNA was extracted after 90 min for measuring TR3 gene induction (mean ± S.D., n > 3; ***, p < 0.001; one-way ANOVA followed by Bonferroni post test for multiple comparisons).

We tested whether binding to TF increased the potency of FVIIai to inhibit aPC signaling. HUVECs were co-transduced to express TF in addition to PAR2. Overexpression of TF did not reveal an increased potency of FVIIai-mediated inhibition of aPC signaling, indicating that the observed inhibition of aPC signaling by high concentrations of FVIIai occurred independently of TF. Taken together, FVIIa has lower affinity for EPCR binding sites that support aPC-dependent PAR1 signaling relative to the receptors that support PAR2 cleavage.

These data suggested that a subpopulation of EPCR was localized in the same membrane microenvironment as PAR2. We next asked whether FVIIa can cleave these receptors. FVIIa is in a zymogen-like conformation that requires TF binding to switch on the catalytic activity of FVIIa (42). Consistent with previous studies (12), FVIIa alone did not induce signaling in PAR2-transduced cells. However, when soluble TF was added to induce the catalytic activity of FVIIa, EPCR-dependent signaling was observed (Fig. 3B). Signaling of sTF-FVIIa was inhibited by active site-blocked aPC, consistent with the notion that aPC and FVIIa shared the EPCR receptors that were in spatial proximity to PAR2. EPCR-dependent signaling of a cofactor-independent, superactive FVIIa_DVQA mutant (29) showed that the major function of sTF was to activate FVIIa. In addition, mutation of hydrophobic residues in the Ω loop of the FVIIa Gla domain provided evidence that PAR2 cleavage by the sTF-FVIIa complex was dependent on Gla domain-dependent recruitment to EPCR binding sites in proximity to PAR2 (Fig. 3C).

Although these data showed that soluble TF-FVIIa can utilize EPCR as a binding site to facilitate PAR2 cleavage, the membrane-proximal PAR2 (43) might be readily reached by FVIIa bound to membrane-anchored TF without a requirement for FVIIa binding to EPCR. In addition, membrane anchoring of TF could impose restrictions on the interaction of the FVIIa Gla domain with EPCR. We therefore addressed the role of EPCR in TF-dependent signaling using HUVECs that expressed membrane-anchored TF in addition to PAR2.

TF supports two signaling reactions. The binary TF-FVIIa complex directly cleaves PAR2 and requires ∼10 nm FVIIa, whereas FVIIa at subnanomolar concentrations drives coagulation. In the ternary TF-FVIIa-FXa coagulation initiation complex, only FXa cleaves either PAR1 or PAR2 (22). Antibody inhibition of EPCR had no effect on binary TF-FVIIa signaling measured in HUVECs overexpressing TF and PAR2 (Fig. 3D). In contrast, ternary complex signaling was completely inhibited by a blocking antibody to EPCR and, as expected, by NAP5, a specific inhibitor of FXa catalytic activity. Although the sTF-FVIIa complex activated PAR2 upon binding to EPCR, these data in the same cellular system showed that EPCR was not involved in signaling by the membrane-anchored TF-FVIIa binary complex. Rather, EPCR was a required component of the ternary TF coagulation initiation signaling complex.

EPCR Specifically Supports TF-FVIIa-FX Ternary Complex Signaling in Cells with Constitutive Receptor Expression

To exclude potential artifacts due to receptor overexpression, we next focused on cell types that constitutively express TF and PAR2. HaCaT cells have been used as a model for TF-PAR2 signaling (22, 44) and are known to constitutively express EPCR (45). We confirmed homogenous EPCR expression in HaCaT cells by flow cytometry (Fig. 4A). TF-FVIIa binary and TF-FVIIa-FXa ternary signaling was monitored by MAPK phosphorylation. Inhibition of EPCR with a blocking antibody to human EPCR abolished TF ternary but not binary complex signaling in HaCaT cells (Fig. 4, B and C). Although human TF is species-compatible with mouse plasma in coagulation assays (37), no binary complex signaling was induced upon addition of mFVIIa to human HaCaT cells (Fig. 4B). However, the combined addition of mFVIIa and FX induced MAPK phosphorylation that was EPCR-dependent as shown by antibody inhibition (Fig. 4C). This ternary complex signaling supported by murine FVIIa was further inhibited by the FXa-specific inhibitor NAP5, providing initial evidence that ternary complex signaling in mouse and human occurs in a similar FXa- and EPCR-dependent manner.

FIGURE 4.

EPCR-dependent ternary TF-FVIIa-FXa PAR2 signaling of HaCaT keratinocytes. A, two-color FACS analysis with anti-EPCR (RCR-252-PE, BD Pharmingen) and anti-TF (5G9-FITC) shows surface expression of TF and EPCR in HaCaT keratinocytes. B, serum-starved HaCaT cells were stimulated with human (h) FVIIa (10 nm), mFVIIa (10 nm), or SLIGRL (50 μm) for 10 min followed by Western blotting of total cell lysate for phosphorylated ERK1/2 (P-ERK1/2) and total ERK1/2 as loading control. EPCR was blocked with anti-EPCR RCR-252 added to the cells 10 min prior to stimulation. A typical experiment and quantification of repeat experiments using ImageJ are shown (mean ± S.D. of -fold induction relative to untreated controls, n ≥ 3). C, ternary signaling induced by addition of FVIIa/FX (0.5 nm/100 nm) or mFVIIa/FX (5 nm/100 nm) was measured as ERK1/2 phosphorylation after 10 min. EPCR was blocked with anti-EPCR RCR-252, or FXa was inhibited with 1 μm NAP5. A typical experiment and quantification of ERK1/2 phosphorylation and amount of FXa generated after 10 min for repeat experiments are shown (mean ± S.D., n ≥ 3; **, p < 0.01; one-way ANOVA followed by Bonferroni post test).

EPCR Expression Is Sufficient to Induce PAR Cleavage by Ternary Complex

Two independent cellular systems showed that EPCR was a functional component of the ternary but not the binary TF protease signaling complex. We considered the potential caveat that the chosen readouts for cell signaling were insensitive to detect effects of EPCR on the TF-FVIIa binary signaling complex. Therefore, we measured PAR cleavage by TF-associated proteases using surface detection of FLAG-tagged PAR constructs (38). In addition, antibody blockade of EPCR might have produced nonspecific steric hindrance of TF signaling complexes. We therefore directly addressed the role of EPCR in ternary complex signaling by a reconstitution experiment. CHO cells that expressed TF (CHO-TF) were stably transfected with EPCR (CHO-TF/EPCR) yielding clonal lines with high expression of TF and EPCR (Fig. 5A). These cells were transiently transfected with amino-terminal FLAG-tagged PAR1 or PAR2 for monitoring of PAR cleavage.

FIGURE 5.

PAR cleavage assay in TF- or TF/EPCR-transfected CHO cells. A, two-color FACS analysis of CHO cells stably expressing TF (CHO-TF) and cells subsequently transfected to stably express EPCR (CHO-TF/EPCR). Cells were stained with anti-TF (10H10-FITC) and anti-EPCR (RCR-252-PE, BD Pharmingen). B, PAR2 receptor cleavage was quantified in CHO-TF or CHO-TF/EPCR cells transiently expressing FLAG-tagged PAR2 following stimulation with FVIIa (10 nm), FVIIa/FX (10 nm/100 nm), or FVIIa/FX/NAPc2 (10 nm/100 nm/200 nm) for 60 min. aPC (10 nm) and aPC/NAPc2 (10 nm/200 nm) were included as controls (mean ± S.D., n ≥ 3; ***, p < 0.001; two-way ANOVA followed by Bonferroni post test). C, effect of FVIIa Gla domain mutations on ternary complex-mediated cleavage of PAR2 transiently expressed in CHO-TF or CHO-TF/EPCR cells following addition of wild-type FVIIa or the indicated FVIIa Gla domain mutants at two different concentrations (0.5 and 10 nm) of FVIIa in the presence of 100 nm FX for 60 min. FXa generation was characterized in an end point assay 10 min after addition of agonists (mean ± S.D., n = 2). D, cleavage of FLAG-tagged PAR1 or PAR2 by aPC (10 nm) or FVIIa/FX (0.5 nm/100 nm) with or without the FXa inhibitor NAP5 (200 nm) or sEPCR (1 μm) added as competitors. Differences between groups were analyzed by two-way ANOVA followed by Bonferroni post test (mean ± S.D., n ≥ 5; ***, p < 0.001; **, p < 0.01). E, signaling in CHO-TF and CHO-TF/EPCR cells transiently transfected with FLAG-tagged PAR1 or PAR2 was measured by ERK1/2 phosphorylation (P-ERK1/2) with total ERK1/2 and β-actin as loading controls. Cells were stimulated for 10 min with PAR agonist peptide (SFLLR or SLIGRL, 50 μm), aPC (10 nm), FVIIa (10 nm), or FVIIa/FX (0.5 nm/100 nm) with or without NAP5 (200 nm).

In a first set of experiments, cleavage of PAR2 upon addition of 10 nm FVIIa was determined. EPCR expression had no significant effect on the slow cleavage of ∼30–40% of surface-expressed PAR2 (Fig. 5B). Expression of EPCR markedly enhanced PAR2 cleavage when FX was added simultaneously. The nematode inhibitor NAPc2 stabilizes the ternary TF-FVIIa-FXa product complex, leaving the active site of FXa accessible to substrates, including PARs (21, 38). PAR2 cleavage by the NAPc2-stabilized ternary complex was similar on EPCR-expressing versus control cells. NAPc2 did not enhance aPC-dependent PAR2 cleavage in the absence of EPCR, demonstrating specificity. These results suggest that NAPc2 forces the ternary product complex to adopt a conformation that is typically stabilized by the co-receptor EPCR.

To ascertain that potential interactions of the FVIIa Gla domain with EPCR did not significantly contribute to ternary complex signaling, we further studied FVIIa Gla domain mutants. These mutants were found to be impaired in EPCR-dependent PAR2 activation by soluble TF-FVIIa in the HUVEC overexpression system (Fig. 3C). Mutation of the key hydrophobic residues Leu⁴ and Phe⁸ alone or together had marginal effects on FX activation when tested at 10 nm but showed ∼2-fold reduced substrate turnover at 0.5 nm whether EPCR was present or not (Fig. 5C). This may indicate that reduced affinity of Gla domain-mutated FVIIa for biological membranes may attenuate substrate turnover. Importantly, high level expression of EPCR in this cell model had no appreciable effect on FX activation (control, 2.8 ± 1.3 nm/min versus EPCR⁺, 3.0 ± 1.5 nm/min). In contrast, EPCR expression markedly enhanced ternary complex signaling with each of the FVIIa Gla domain mutants. These results provide an independent line of evidence that EPCR is a functional component of the ternary TF signaling complex independent of interactions with the FVIIa Gla domain.

Although PAR1 signaling by FVIIa has been reported (46), we only detected PAR1 cleavage by FVIIa when the specific thrombin inhibitor hirudin was inadvertently omitted from the reaction mixtures, indicating that addition of FVIIa to cells under serum-free conditions can cause activation of coagulation factors carried over from the tissue culture medium as shown previously (39). In contrast, FXa or the ternary complex TF-FVIIa-FXa activated PAR1 as well as PAR2 in the presence of hirudin (21, 39, 47). We asked whether EPCR supported cleavage of both PAR1 and PAR2 by the ternary complex. EPCR expression increased PAR1 and PAR2 cleavage by aPC and the ternary TF-FVIIa-FXa complex, but ternary complex-mediated cleavage of PAR2 appeared to be more efficient relative to aPC at the chosen concentration (Fig. 5D). FXa specificity of PAR1 and PAR2 cleavage was confirmed by addition of the FXa inhibitor NAP5, which was without effect on aPC-mediated cleavage of PARs.

Expression of EPCR also markedly enhanced cell signaling measured by ERK1/2 phosphorylation by aPC and the ternary complex in PAR1- or PAR2-transfected cells (Fig. 5E). The enhanced cleavage of PARs in EPCR-transfected cells was reversed by the addition of sEPCR, but inhibition of ternary complex signaling was less efficient at equivalent concentrations of soluble receptor (Fig. 5D). This may indicate that multiple interactions within the cofactor-enzyme-product complex synergistically stabilized the FX Gla domain interaction with membrane-anchored EPCR. Thus, EPCR can function as a co-receptor for the ternary TF-FVIIa-FXa complex, supporting PAR1 and PAR2 cleavage potentially by localizing and maintaining a scaffold-like signaling platform in certain membrane microdomains.

EPCR Is a Conserved Component of Ternary Signaling Complex across Species

Although tumor cell TF-FVIIa-PAR2 signaling has been demonstrated in both murine and human cells (23, 48, 49), little is known about signaling of the ternary complex in primary cells isolated from mice. Because TF is not typically expressed by endothelial cells in the absence of proangiogenic or inflammatory stimuli, we focused on murine SMCs that constitutively express EPCR and TF (50, 51). SMCs were isolated from lungs of wild-type, PAR1^−/−, PAR2^−/−, or EPCR^low C57BL/6 mice as well as mice that carried conditional alleles of EPCR for Cre-mediated gene deletion. Cell staining for smooth muscle cell α-actin demonstrated homogeneity of the primary cultures (Fig. 6A). Experiments that used MAPK phosphorylation as a signaling readout showed responsiveness of SMCs to stimulation with thrombin and FXa but not mFVIIa. Signaling was PAR1-dependent because responsiveness to protease stimulation was lost in PAR1^−/− but not PAR2^−/− SMCs (Fig. 6B). Lack of responses to mFVIIa extends the above conclusion that PAR1 is not activated by human TF-FVIIa.

FIGURE 6.

EPCR dependence of ternary complex signaling in murine SMCs. A, isolated SMCs of wild-type and EPCR^low mice are uniformly positive for smooth muscle cell α-actin (red). Nuclei were counterstained with DAPI (blue). Scale bar, 30 μm. B, ERK1/2 phosphorylation in SMCs derived from wild-type, PAR1^−/−, and PAR2^−/− mice stimulated for 10 min with thrombin (5 nm), mFVIIa (20 nm), or human FXa (20 nm). C, immunoblot analysis of total cell lysates for EPCR and TF expression levels in SMCs isolated from wild-type, EPCR^low, and EPCR^flox/flox mice with and without in vitro Cre-mediated LoxP excision of the EPCR gene (+cre). D, thrombin (10 nm) and ternary complex (mFVIIa/FX, 20 nm/150 nm) signaling was determined in wild-type, EPCR^low, EPCR^flox/flox and EPCR^flox/flox+cre SMCs using as a readout NUR77 mRNA induction after 90 min of stimulation (mean ± S.E., n = 5; **, p < 0.01; *, p < 0.05; two-way ANOVA followed by Bonferroni post test). E, ERK1/2 phosphorylation in wild-type and EPCR^low SMCs simulated with thrombin (10 nm), mFVIIa (20 nm), FXa (20 nm), or mFVIIa/FX (20 nm/100 nm) for 10 min. Where indicated, inhibitory anti-murine EPCR 1560 or non-inhibitory anti-murine EPCR 1562 antibody was added to the cells 10 min prior to agonist stimulation. Densitometry was quantified with ImageJ software, and the levels were normalized to untreated wild-type or EPCR^low controls (mean ± S.E., n = 6; ***, p < 0.001; *, p < 0.05). Statistical analysis was performed after logarithmic transformation by two-way ANOVA followed by Bonferroni post test. F, comparison of ERK1/2 phosphorylation in EPCR^flox/flox and EPCR^flox/flox+cre SMCs in the same experimental setup as described under E (mean ± S.E., n = 4; *, p < 0.05).

To address the role of SMC EPCR in TF ternary complex signaling, two independent genetic approaches were taken. Although SMCs derived from EPCR^low animals expressed constitutively very low levels of EPCR, EPCR was alternatively deleted from EPCR^flox/flox SMCs in vitro by adenoviral transduction with Cre recombinase. This second approach excluded indirect effects on SMC phenotypes due to prolonged EPCR deficiency in vivo. Western blotting showed that both approaches were equally effective to reduce EPCR protein levels in SMCs (Fig. 6C). Note that although EPCR^flox/flox mice were back-crossed into the C57BL/6 background, the floxed EPCR allele was from the 129 genetic background, which may explain the consistently higher protein expression levels compared with C57BL/6 wild-type controls.

TF-FVIIa-FXa ternary complex signaling was compared with thrombin signaling in cells with markedly reduced EPCR levels versus the respective controls using induction of the murine homologue of the nuclear orphan receptor TR3 (NUR77) as readout (Fig. 6D). SMCs with genetically reduced levels of EPCR no longer showed a response to ternary complex signaling, whereas the thrombin response was unaltered. Similarly, in the more upstream MAPK phosphorylation assay, EPCR^low SMCs (Fig. 6E) or Cre recombinase-treated EPCR^flox/flox cells (Fig. 6F) lost responsiveness to ternary complex signaling or stimulation with high concentrations of FXa but not to thrombin.

In addition, ternary complex signaling was specifically blocked by anti-murine EPCR antibody 1560, which was inhibitory in the biochemical assay shown in Fig. 1, but not by the poorly inhibitory antibody 1562. Similar to the observations in EPCR-overexpressing CHO cells, measured FXa generation was not different between wild-type and EPCR^low SMCs. However, deletion of EPCR from the higher expressing EPCR^flox/flox SMCs resulted in somewhat increased FXa generation (control, 0.19 ± 0.06 nm/min versus Cre, 0.3 ± 0.13 nm/min; p < 0.05). Thus, despite a variable effect of EPCR on procoagulant responses, these results demonstrate that EPCR is a required component of ternary complex signaling across species, validating murine models as a tool to study EPCR involvement in TF signaling pathways.

DISCUSSION

In this study, we provide new data on functional interactions of EPCR with the TF coagulation initiation complex. Prior studies showed that human EPCR binds human aPC and FVIIa with similar affinity (12, 13). We confirm these results and previous conclusions that complex formation of FVIIa with TF does not diminish the interaction of the FVIIa Gla domain with EPCR (16). Although a previous study raised doubts about the cross-species relevance of FVIIa binding to EPCR (18), our experiments demonstrate that EPCR is a receptor for FVIIa and the TF-FVIIa complex in the mouse.

Binding of human FVIIa to EPCR is not Mg²⁺-dependent (17), and in the presence of a physiologically relevant 2:1 ratio of the divalent cations Ca²⁺ and Mg²⁺, mFVIIa interacted with murine EPCR with similar affinity as the human counterpart. The FVIIa Gla domain can adopt a folded conformation in the presence of Ca²⁺ alone, but Mg²⁺ added at approximately half the Ca²⁺ concentration specifically occupies three of the Gla domain divalent cation binding sites (52). Mg²⁺ binding at these sites induces conformational changes in the Ω loop of the FVIIa Gla domain and results in closer contacts of the Gla domain of FVIIa with the carboxyl-terminal module of TF. The presented data show that these conformational changes do not adversely affect the ligand interactions with EPCR, which contacts the hydrophobic Ω loop residues Phe⁴ and Leu⁸ (13, 53, 54).

Conflicting data were reported on the role of EPCR as a cellular receptor for FX (12, 19, 41). Although human FX carries a Met⁸ instead of Leu⁸ present in human FVIIa and aPC, this substitution in the Ω loop should be compatible with EPCR binding because mouse aPC has the same amino acid replacement. The presented surface plasmon resonance experiments show that FX does not bind appreciably to EPCR unless Mg²⁺ is present. Similar to FVIIa (52), exchange of the first and seventh Gla domain Ca²⁺ by Mg²⁺ has also been demonstrated for FIX, resulting in improved ligand interactions (55). Addition of Mg²⁺ enhances aPC binding to EPCR (17), and Zn²⁺ can substitute for Mg²⁺ in markedly improving receptor binding (56). Mg²⁺ occupying the first Ca²⁺ binding position is coordinated with Gla²⁹, a residue that is shared with the second bound Ca²⁺ involved in a direct contact with Glu⁸⁶ of EPCR (53). In addition to improving Ω loop hydrophobic interactions, Mg²⁺ exchange may thus stabilize a crucial electrostatic contact required for EPCR binding (54).

Mg²⁺ binding to the FX Gla domain is known to be functionally relevant for FX activation by the soluble TF-FVIIa complex in a lipid-free system (40). EPCR specifically inhibited this reaction independently of hydrophobic interactions with the FVIIa Gla domain, demonstrating functionally relevant interactions of EPCR with the TF coagulation initiation complex. Overexpression of EPCR had no appreciable effect on FX activation on CHO cells, but deletion of high levels of EPCR from primary SMCs somewhat enhanced FX activation, indicating that the relative expression levels of TF and EPCR may be a determinant in predicting regulatory effects of EPCR on procoagulant pathways. However, in all experimental systems analyzed here, EPCR was necessary for signaling of the TF ternary complex.

Our data in overexpression systems show that EPCR is expressed in sufficient proximity to support PAR2 cleavage by either aPC or FVIIa that gained catalytic activity by soluble TF or mutagenesis. In PAR2-overexpressing HUVECs, FVIIa was a competitive inhibitor of aPC signaling, but based on competition, FVIIa had apparently low affinity for the sites utilized by aPC to cleave PAR1. These data indicate heterogeneity in cell surface EPCR with implications for ligand binding specificity.

Although soluble TF-FVIIa readily cleaved PAR2, a remarkable specificity emerged when membrane-anchored TF complexes were analyzed. EPCR expression had no detectable effect on TF-FVIIa-mediated PAR2 cleavage but markedly enhanced the activation of both PAR1 and PAR2 by the ternary TF-FVIIa-FXa complex. Experiments in human cells with constitutive TF and EPCR expression showed that antibody blockade of EPCR specifically interrupted signaling of the ternary but not the binary TF signaling complex. Genetic and antibody blocking approaches in primary murine cells further confirmed that EPCR is a novel component of the ternary complex signaling pathway. These experiments validate the use of mouse models to gain further insight into the physiological roles of EPCR in TF signaling in vivo.

The role of EPCR as a crucial co-receptor for PC activation and aPC-mediated cytoprotective signaling in vascular and immune cells is supported by in vitro and in vivo data (57–63). However, TF is normally not expressed in the vascular endothelium but co-expressed with EPCR in extravascular cells. We analyzed cell types that express TF in vivo and show that EPCR is a required co-receptor for ternary complex signaling in both keratinocytes and SMCs.

Keratinocytes in the basal layers of murine and human skin express prominently TF (64, 65). EPCR-aPC-PAR1 signaling is antiapoptotic and promotes proliferation in keratinocytes (66), and TF-FVIIa signaling has been shown to induce a repertoire of genes that support a wound healing program and angiogenesis (44). Because signaling of the ternary coagulation initiation complex is induced with high efficiency by FVIIa at pm concentrations, this signaling pathway is expected to be initiated by epithelial injury and blood loss. Supported by the extravasation of coagulation factors, ternary complex signaling may simultaneously activate EPCR-dependent cytoprotective and proliferative responses in keratinocytes and the repertoire of PAR2-dependent genes to accomplish efficient wound healing.

PAR signaling has broad extravascular roles in development, inflammation, and immunity, and these processes are frequently not dependent on coagulation proteases (67). However, perivascular TF is associated with FVIIa (68), and extravascular cells can ectopically synthesize coagulation factors as clearly demonstrated for FX in lung diseases (69, 70). FVIIa and FXa have been shown to be potent inducers of mitogenesis, migration, and inflammatory cytokines and matrix regulators in SMCs and fibroblasts (71–75). These data place upstream coagulation signaling at the center of extravascular coagulant signaling and tissue remodeling. The presented data with SMCs raise the intriguing possibility that TF and EPCR expressed in the vessel wall regulate the cellular responses to low levels of coagulation proteases that are constitutively present or ectopically synthesized in the extravascular space. Intravascular EPCR-aPC-PAR1 signaling counteracts thrombin-PAR1 signaling (8, 57, 61, 76). It is intriguing to speculate that EPCR-dependent ternary complex signaling may similarly regulate certain aspects of extravascular upstream coagulation signaling.