The endothelial cell protein C receptor. Cell surface expression and direct ligand binding by the soluble receptor.

Expression of the endothelial cell protein C receptor (EPCR) gene in mammalian cells imparts the capacity to bind activated protein C (APC) or protein C. Immunochemical analysis of CCD41, apparently the murine homologue of EPCR, suggested centrosomal localization, raising questions about the location of the EPCR gene product and its role in protein C binding. In this study, we express a soluble form of EPCR, demonstrate EPCR expression on the cell surface, and direct binding between soluble EPCR and protein C/APC. Affinity purified polyclonal and a monoclonal antibody against EPCR bound to the cell surface of EPCR-transfected cells but not to control cells. A 49-kDa protein, a mass similar to soluble EPCR, was immunoprecipitated from the cell surface of endothelium and cells transfected with human EPCR but not from control cells. The FLAG™ antibody and APC bound to cells expressing an EPCR construct containing the FLAG™ epitope located in a putative extracellular domain, whereas an EPCR construct truncated just before the putative transmembrane domain produced only soluble EPCR antigen. Soluble EPCR inhibited APC binding to EPCR expressing cells in a concentration-dependent fashion, Kd(app) = 29 nM and bound to immobilized protein C in a Ca2+-dependent fashion. Thus, EPCR is a type 1 transmembrane protein that binds directly to APC.

The protein C anticoagulant pathway is an indispensable, on demand, regulatory mechanism of blood coagulation (1). Thrombin is generated by the coagulation cascade and catalyzes the formation of fibrin and also promotes the activation of many cell types, including platelets (2). Thrombin also binds to the endothelial cell surface receptor, thrombomodulin. After binding to thrombomodulin, thrombin changes enzyme specificity and catalyzes the conversion of the plasma vitamin K-dependent zymogen, protein C, to the anticoagulant serine protease, activated protein C (APC) 1 (3). APC is the terminal enzyme of the protein C pathway and catalyzes the proteolytic inactivation of the coagulation cofactors, factors Va and VIIIa (4). Defects in this pathway are associated with thrombophilia (4,5).
In addition to modulating the coagulant response, this pathway appears to modulate the inflammatory response. In vivo, APC administration prevents death and organ dysfunction in baboons challenged with lethal numbers of Escherichia coli (6). Humans suffering from severe meningococcemia and an associated acquired protein C deficiency have been reported to exhibit rapid systemic improvement following protein C administration (7,8). In vitro, APC has been shown to inhibit tumor necrosis factor elaboration by monocytes (9), and protein C has been reported to inhibit leukocyte adhesion to selectins (10). The exact mechanisms by which the protein C pathway modulates inflammatory responses remains obscure.
As one approach, we initiated studies to identify new members of the pathway. Recently, we (11) and others (12) observed specific protein C and APC binding to cultured human umbilical vein endothelium (HUVEC). Binding was Ca 2ϩ -dependent and required the presence of the Gla domain of protein C (11,12).
To gain initial information on the nature of the protein C binding site, we used an expression cloning strategy to isolate a cDNA, which, when expressed in mammalian cells, elicited the formation of specific protein C binding sites with properties similar to those of endothelium (11). A survey of cultured cells indicated that message and binding function was endothelial cell-specific, and therefore the protein was designated endothelial cell protein C/APC receptor (EPCR). Cloning of murine and bovine EPCR demonstrated conserved cellular specificity, structure, and regulation, including down-regulation by tumor necrosis factor (13). The sequence of EPCR predicts it to be a type 1 transmembrane glycoprotein (11), and the functional data would be consistent with EPCR binding directly to protein C.
The unexpected observation that EPCR is homologous, and probably identical, to a previously described intracellular murine protein, CCD41 (14), raised questions about this model of * The research was funded by Grant P01 HL54804 (to C. T. E.) by the National Heart, Lung, and Blood Institute of the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  EPCR function. The CCD41, or centrocyclin, cDNA clone was originally isolated from Ehrlich ascites tumor cells (14). Immunocytochemical analysis using antibodies against a bacterially expressed CCD41 fusion protein revealed nuclear localization and centrosome association of the CCD41 gene product (14). The nucleotide sequence of murine EPCR contains only five nucleotide differences from that of CCD41, and these differences occur in regions of the cDNA that are technically difficult to sequence, suggesting that the differences are due to cloning or sequencing errors (13). If murine EPCR and CCD41 are identical molecules, then the cellular localization studies described above suggest that the EPCR gene product is not a cell surface molecule and would suggest that EPCR elicits protein C binding sites indirectly by induction of another protein that serves as the cell surface receptor for protein C.
In this study, we demonstrate that human EPCR is expressed on the surface of endothelium and cells transfected with the EPCR cDNA and that a soluble form of EPCR can bind APC directly.

EXPERIMENTAL PROCEDURES
Cell Culture-All human cell lines were maintained as described previously (11). HUVEC were kindly provided by Dr. Craig Carson. Stable transformants expressing EPCR were established as follows. The human EPCR cDNA construct in the mammalian expression vector, pEF-BOS (15), was co-transfected into human kidney 293 cells (ATCC CRL 1573) with another plasmid, pBK-CMV (STRATAGENE), carrying the neomycin-resistance gene. After G418 selection, EPCRpositive clones were screened for the ability to bind fluorescein-labeled APC (Fl-APC) by flow cytometry as described (11). As negative controls, resistant colonies were isolated from cells transfected with the same expression vector but without the insert. EPCR-negative and -positive clones of murine NIH3T3 were also generated and selected as described above.
Expression and Purification of the Soluble EPCR Fusion Protein-A cDNA fragment coding for a soluble EPCR protein, starting from residue 16, corresponding to a potential signal peptidase cleavage site (11), and truncated immediately above the putative transmembrane spanning domain at residue 210, was amplified from a human EPCR cDNA clone by the polymerase chain reaction (PCR) and ligated into the StuI and XbaI sites of the RSV-PL4 expression vector (16). This construct codes for an EPCR fusion protein containing the transferrin signal sequence followed by the HPC4 epitope at the amino terminus (16,18). Transfection, selection, and affinity purification with HPC4 antibody were as described (16). In some experiments, the s-HPC4-EPCR was further purified by Mono-Q FPLC column chromatography. The accuracy of all EPCR constructs generated in this study were confirmed by sequencing (17).
An alternative soluble fusion protein with the HPC4 epitope site on the carboxyl-terminal of the protein was prepared by PCR methods essentially as described above. The resultant construct consisted of the native signal peptide, the EPCR coding region truncated to delete the transmembrane and cytosolic tail, a factor Xa cleavage site, and the HPC4 epitope. This construct was ligated to the pGT-h expression vector (a kind gift from Drs. Brian Grinnell and David Berg) to allow hygromycin selection in human 293 cells (19) and the resultant fusion protein isolated as described above, except that 0.6 mM Mg 2ϩ was included in wash buffers.
A soluble form of EPCR was also expressed in E. coli using the ThioFusion Expression system (Invitrogen). The cloning strategy was to insert the coding sequence of the HIS tag at the 3Ј-end of the molecule before the stop codon and a factor Xa cleavage site (IEGR) immediately before the His tag on the antisense PCR primer. After PCR amplification of EPCR cDNA with appropriate primers, the PCR product was digested by EcoRV and SalI (these sites were included at the 5Ј-and 3Ј-ends, respectively) and was subcloned into the SmaI and SalI sites of the pTrxFus expression vector. The insert codes for a fusion protein starting with thioredoxin followed by soluble EPCR, a factor Xa cleavage site, and the His tag (TFT201His). The fusion protein was expressed in GI724 cells and purified from the cell lysate by column chromatography on a TALON metal affinity column (Clontech) by elution with 0.3 M imidazole according to the manufacturer's protocol.
Construction and Expression of Recombinant EPCR Molecules Containing the FLAG Epitope-Several EPCR constructs were generated by the PCR mutagenesis method in which the FLAG epitope was fused at different sites of the EPCR cDNA sequence. The DNA sequence coding for the FLAG epitope (DYKDDDK) was included at the 3Ј-end of the PCR primers. The first construct, mFL1, contains the FLAG epitope that replaces eight amino acid residues of EPCR between residue 198 and 206 (see Fig. 6 below for a schematic of the FLAG mutants described in this section). In the second full-length mutant (mFL2), the sequence between 146 and 153 was replaced by the FLAG epitope by the same methods. In the third construct (smFL3), the sequence between 211 and 218 corresponding to the amino-terminal sequence of the putative transmembrane domain was replaced with the FLAG epitope sequence followed by a stop codon. The PCR products were digested with XhoI and NotI (included at the 5Ј-and 3Ј-ends of primers, respectively) and were subcloned into a mammalian expression pEF-BOS vector (15). The constructs were transfected into 293T cells by the calcium/phosphate method. After 48 h, the cells were analyzed by flow cytometry using M2 (an anti-FLAG monoclonal antibody, Eastman Kodak Co.) staining. F1-APC binding to the transfectants was also measured by flow cytometry as described (11).
Protein Preparation-Human protein C and APC were prepared as described (20). For binding studies, the active site of APC was labeled with fluorescein to generate Fl-APC by first inhibiting APC with N ␣ -[(acetylthio)acetyl]-D-Phe-Pro-Arg-CH 2 Cl by a modification (11) of the method of Bock (21) and then reacting the modified enzyme with 5-(iodoacetamido)fluorescein as described (11,21). The extinction coefficients (mg/ml) Ϫ1 cm Ϫ1 and molecular weights used for this study were human APC (1.45, 56,000) (22) and s-EPCR-HPCR (1.0, 42,000). For EPCR, the molecular weight was determined by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE), and the extinction coefficient was calculated based on the predicted amino acid composition.
Preparation of Antibodies-The preimmune serum was collected from a goat. The goat was immunized subsequently with 0.5 mg of the purified s-HPC4-EPCR in Freund's complete adjuvant followed by 0.25 mg of s-HPC4-EPCR in Freund's incomplete adjuvant. The serum was collected weekly after the second immunization. Polyclonal immunoglobulin was prepared by standard methods including 50% ammonium sulfate precipitation, dialysis, and ion-exchange chromatography on DE52 (Whatman). For some experiments, the ammonium sulfate precipitate was resuspended in TBS (20 mM Tris, 150 mM NaCl, 0.02% sodium azide, pH 7.5), gel-filtered and affinity-purified on bacterially expressed EPCR (TFT210HIS, 5 mg), adsorbed to a 1.5-ϫ 3-cm TAL-ON column equilibrated in 6 M guanidine HCl, 20 mM Tris-HCl, pH 8.0. The column was washed extensively with the same buffer, and bound immunoglobulin (11 mg) was eluted with 6 M urea, 20 mM Tris-HCl, pH 8.0, and subsequently dialyzed against TBS.
Monoclonal antibodies against s-HPC4-EPCR were obtained as described previously for other proteins (20). The hybridoma supernatants were screened for the ability to bind to cell surface EPCR by fluorescence-activated cell sorter (FACS) analysis of cell surface staining using the EPCR-negative and -positive stably transfected cell lines. Monoclonal antibody binding was detected using fluorescein isothiocyanatelabeled goat anti-mouse Ig (Becton Dickinson) as the second antibody at a 1:100 dilution. Antibodies were purified from ascites by ammonium sulfate precipitation, QAE-Sephadex chromatography, and gel filtration as described (6). Antibodies were biotinylated with NHS-LC-Biotin (Pierce) according to the manufacturer's protocol. 8.1 mM Na 2 HPO 4 , pH 7.1 (PBS) containing 5% fetal bovine serum and 10 mM EDTA). The bound antibody was stained at 4°C for 30 min with fluorescein-labeled affinity purified rabbit anti-goat IgG (Kirkegaard & Perry Laboratories, Inc.). FACS analysis was carried out as described previously (11) using a FACSCalibur (Becton Dickinson).
Inhibition of Fl-APC Binding with s-EPCR-HPC4 -The binding of Fl-APC to E-7 cells was performed essentially as described (11). Fl-APC (80 nM) was incubated for 30 min at room temperature with s-EPCR-HPC4 (25-1000 nM) in Hank's balanced salt solution (HBSS) (Mediatech) that was supplemented with 3 mM CaCl 2 , 0.6 mM MgCl 2 , 1 mg/ml bovine serum albumin, and 0.02% NaN 3 . The mixtures were added to E-7 cells and incubated for 30 min on ice. The samples were washed with the same buffer, centrifuged, and cell-bound fluorescence was determined on a FACSCalibur. Nonspecific binding of Fl-APC (0 -80 nM) was determined in the presence of 3.44 M protein C and was subtracted from the total bound fluorescence. Assuming the decrease in mean channel fluorescence was due to s-EPCR-HPC4 binding to Fl-APC and a resultant decrease in free Fl-APC concentration, the free Fl-APC concentration was calculated by reference to the standard curve of Fl-APC concentration versus mean channel fluorescence. The concentrations of the s-EPCR-HPC4-Fl-APC complex were obtained by subtracting the free Fl-APC concentrations from total Fl-APC (80 nM). Free s-EPCR-HPC4 concentrations were calculated by subtracting the complex concentrations from added s-EPCR-HPC4 concentrations. These values were used to calculate K d (app) at each concentration of s-EPCR-HPC4 added.
Western Blotting-Samples were electrophoresed in Laemmli buffers (23), and the gels were transferred to Immobilon-P membranes (Millipore) in a semi-dry apparatus (Bio-Rad). Membranes were blocked with non-fat milk incubated with HPC4 (0.5 g/ml) in buffer containing 5 mM Ca 2ϩ , washed, and incubated with goat-antimouse IgG conjugated with horseradish peroxidase (Pierce). The membranes were washed and developed with an enhanced chemiluminescence substrate (Pierce).
Immunoprecipitation-Cell surface labeling with biotin and immunoprecipitation was performed essentially as described (24). Harvested cells (1 ϫ 10 7 ) were washed with PBS and resuspended in 1 ml of 100 mM HEPES, 150 mM NaCl, pH 8.0, and incubated at room temperature for 30 min with 0.2 mg/ml sulfo-NHS-biotin (Pierce). After washing once in serum-free medium and twice in medium containing 10% fetal bovine serum, cells were lysed in 1 ml of 20 mM Tris-HCl, 150 mM NaCl, 3 mM MgCl 2 , 2 mM phenylmethylsulfonyl fluoride, 10 g/ml soybean trypsin inhibitor, 1 g/ml leupeptin, 0.5% Nonidet P-40, pH 7.5. Samples (100 l) were pretreated for 3 h with 100 l of a 50% slurry of Affi-Gel Protein A followed by a 16-h incubation of the supernatants (50 l) with 50 l of Affi-Gel Protein A containing nonimmune IgG (100 g/50 l packed gel). These supernatants were incubated with 50 l of Affi-Gel Protein A containing either goat anti-EPCR or goat anti-mouse IgG as control for 3 or 8 h in two separate experiments. Gels were washed five times with 1 ml of 10 mM Tris-HCl, 140 mM NaCl, 3 mM MgCl 2 , 2 mM phenylmethylsulfonyl fluoride, 0.5% Nonidet P-40, pH 8.0, and once with 1 ml of HBSS. Adsorbed protein was eluted by boiling in 50 l of electrophoresis sample buffer and analyzed by Western blotting using streptavidin-horseradish peroxidase conjugate (Amersham Corp.) and the enhanced chemiluminescence system.

RESULTS
Soluble, recombinant s-HPC4-EPCR was produced from a construct in which the putative transmembrane region of EPCR was deleted using the protocols described under "Experimental Procedures." Addition of the HPC4 epitope tag facilitates detection, purification, and analysis of recombinant proteins since the HPC4 antibody binds to this P7-P5Ј region in protein C in a calcium-dependent fashion (18). Although the peptide binds to HPC4 in the presence of calcium, the peptide itself does not bind calcium (18), and the recombinant proteins can be easily removed from an HPC4 affinity column with buffer containing EDTA. The affinity purified fractions were analyzed by SDS-PAGE and Western blotting. The mass of the major protein was 46 kDa under both nonreducing and reducing conditions (Fig. 1). Mono Q chromatography separated most of the high molecular weight material from the s-HPC4-EPCR. The major protein was also blotted by HPC4 antibody (data not shown). The amount of aggregate varied between preparations, but some aggregated, very high molecular weight protein was always detected, and this did not usually react with HPC4. This contaminating protein may be due to either the presence of free sulfhydryl(s) in EPCR (25) or the low expression level (20 g/liter). The amino-terminal sequence of the purified protein was determined as XXQVDPRLIDGKIEG which was identical to the sequence of the HPC4 epitope. The 46-kDa molecular mass is considerably higher than predicted based on protein sequence (24 kDa), possibly due to the presence of N-linked sugars at the four potential N-glycosylation sites (11). This possibility was supported by the observation that endoglycosidase F/Peptide-N-glycosidase (Oxford Glyco-Systems, Inc.) reduced the apparent molecular mass on SDS-PAGE from 46 kDa to a major band at 28.5 kDa (data not shown). In addition, s-HPC4-EPCR was transferred to a polyvinylidene difluoride membrane and treated with reagents to generate aldehydes by cleaving carbohydrate vicinal alcohol groups. The modified carbohydrate was biotinylated and the product detected with a streptavidin detection system and enhanced chemiluminescence substrate (Glycoprotein Detection System, Amersham Corp.). s-HPC4-EPCR, but not bacterially expressed TFT210HIS EPCR, was detected by this method (data not shown). The s-EPCR-HPC4 fusion protein with the HPC4 epitope on the carboxyl terminus had a similar SDS-PAGE pattern. The amino-terminal sequence of the s-EPCR-HPC4 was XQXASDGLQR which indicates a preference of cleaving the signal peptide at residue 18 rather than residue 16 as predicted (11). From this sequence, we infer that EPCR has 3 Cys residues in the putative extracellular domain and hence at least one must be unpaired.
To analyze the function and expression of EPCR, stably transfected human 293 and murine NIH3T3 cells expressing EPCR were selected. E-7 and N-1 cells were EPCR-positive and -negative clones of human 293 cells, respectively. HEN-1 (EPCR-positive) and NN-1 (negative control) were derived from murine NIH3T3 cells. Analysis by flow cytometric methods indicated that Fl-APC binding to E-7 and HEN-1 cells was saturable and calcium-dependent. Based on the recent observation that Mg 2ϩ enhances factor IX activity (26), we explored the possibility that Mg 2ϩ would influence APC binding to EPCR. We observed that in the presence of optimal Ca 2ϩ , low levels of Mg 2ϩ enhanced binding ( Fig. 2A) about 2-fold and decreased the K d (app) from 266 to 74 nM (Fig. 2B). Mg 2ϩ (2.6 mM) did not support binding in the absence of Ca 2ϩ (Fig. 2B). In the experiment shown, in the presence of both metal ions the K d (app) was slightly higher than that observed previously with HUVEC (30 -50 nM). However, other experiments with HEN-1 and E-7 cells gave K d (app) values of 21.0 Ϯ 2.9 and 25.8 Ϯ 0.48 nM, respectively. Hence, we conclude that the APC affinity for EPCR-transfected cells and endothelium are similar. The increased cell-associated fluorescence was not due to Mg 2ϩ -induced changes in quantum yield since the fluorescence emission of Fl-APC, measured as described previously (27), was insensitive to the addition of Mg 2ϩ at these concentrations (data not shown).
To determine whether EPCR antigen was expressed on the cell surface, we analyzed antibody binding to EPCR-transfected cells. To ensure specificity of the polyclonal anti-s-HPC4-EPCR antibody, the antibody was affinity-purified on the bacterially expressed form of EPCR. The cell surface of the E-7 cells stained brightly with this affinity-purified polyclonal antibody, whereas little staining of the control N-1 cells was detected (Fig. 3).
To further characterize the cell surface expression of EPCR, we utilized a monoclonal antibody, JRK-1, prepared against s-HPC4-EPCR. JRK-1 bound to the cell surface of EPCR-positive E-7 and HEN-1 cells (Fig. 4, A and C). No binding was detected to the control, EPCR-negative N-1 cells or NN-1 cells (Fig. 4, B and D). Fl-APC and JRK-1 binding to cells transiently transfected with either human, bovine, or murine EPCR were also compared (Fig. 4, E-G). Fl-APC bound almost equally well to 293T cells transfected with human, bovine, and murine EPCR. In contrast, JRK-1 bound only to cells transfected with human EPCR (Fig. 4E) and not to those transfected with bovine or murine EPCR (Fig. 4, F and G). The Fl-APC and JRK-1 pattern was complex in these experiments, probably because the transient expression system resulted in variable EPCR expression levels among the cells. It is of note that the distribution of cells with high and low levels of Fl-APC and JRK-1 binding sites are similar. Biotinylated JRK-1 bound to s-HPC4-EPCR in a saturable and specific manner in an enzyme-linked immunosorbent assay format, but JRK-1 did not bind to control recombinant proteins containing the HPC4 epitope (28) (data not shown).
The antigen distribution was examined using JRK-1 and confocal microscopy. Immunofluorescence was detected on the cell surface of HEN-1 cells and not on NN-1 cells (Fig. 5).
To analyze the requirements for cell surface expression and Fl-APC binding, we prepared the three FLAG epitope-tagged forms of EPCR illustrated in Fig. 6. The FLAG epitope was detected on the surface of cells transiently transfected with mFL-1 or mFL-2, but not with sFL-3 (Fig. 6). Fl-APC bound to mFL-1, but not to mFL-2 or sFL-3, -transfected cells. mFL-1 and wild type EPCR-transfected cells bound Fl-APC similarly. Transfection with sFL-3 resulted in the appearance of immunoreactive forms of the antigen in the cell supernatant (Fig.  6B), a result consistent with deletion of the transmembrane region of EPCR. Under nonreducing conditions 42-, 84-, 130-, 170-kDa forms were detected (Fig. 6B) These aggregates were probably due to the presence of free sulfhydryl group(s) in EPCR (25). Under reducing conditions, a 45-kDa protein was detected (Fig. 6B). No antigen was detected in the cell supernatants from cells transfected with either mFL-1 or mFL-2.
As another approach to determine the cell surface expression and properties of EPCR, the cell surfaces of HUVEC, E-7, and N-1 cells were biotinylated using a membrane-impermeable reagent as described under "Experimental Procedures." The . After transfer to a membrane filter, the protein was detected with horseradish peroxidaseconjugated streptavidin and the enhanced chemiluminescence system.

FIG. 8. Inhibition of Fl-APC binding to E7 cells by s-EPCR-HPC4.
A, binding of Fl-APC to E-7 cells was determined as a function of increasing Fl-APC concentration (0 -640 nM) as described under "Experimental Procedures." The data from 0 to 80 nM Fl-APC was used as a standard curve to calculate free Fl-APC concentrations. The complete binding curve is shown (inset). B, Fl-APC (80 nM) was mixed with increasing concentrations of s-EPCR-HPC4 (0 -1000 nM), and the mixtures were added to E-7 cells to assess Fl-APC binding. All data shown is mean channel fluorescence versus added ligand or inhibitor concentration. The binding of 80 nM Fl-APC was also determined in the presence of 3.44 M protein C (solid square). polyclonal anti-EPCR antibody precipitated a 49-kDa protein from HUVEC and EPCR-positive E-7 cells and not from control N-1 cells (Fig. 7). The mass of the protein was similar to the 46-kDa observed with s-HPC4-EPCR and consistent with a glycosylated form of EPCR. Like s-HPC4-EPCR, the apparent mass of the immunoprecipitated protein was not influenced by reduction.
The above experiments indicate that EPCR is a cell surface protein, but they do not demonstrate that EPCR binds directly to APC. To test this possibility, we examined the ability of soluble EPCR to inhibit Fl-APC binding to E-7 cells by flow cytometry. We employed both s-HPC4-EPCR and s-EPCR-HPC4 in these experiments. Initially, cells were incubated with various concentrations of Fl-APC to establish a standard curve relating mean channel fluorescence intensity to free Fl-APC concentration (Fig. 8A). Next, a single concentration of Fl-APC (80 nM) was mixed with different concentrations of s-EPCR-HPC4, and these mixtures were allowed to bind to the cell surface.s-EPCR-HPC4blockedFl-APCbindinginaconcentrationdependent fashion (Fig. 8B). Assuming that inhibition is caused by complex formation between EPCR and Fl-APC in solution and the decrease in Fl-APC binding is due to the reduction in the free Fl-APC concentration, the free Fl-APC concentration can be estimated based on the standard curve of mean channel fluorescence intensity versus Fl-APC concentration, which can be used to calculate the K d (app) . Using this approach, a K d (app) ϭ 29.2 Ϯ 5.7 nM using s-EPCR-HPC4 concentrations ranging from 100 to 1000 nM with the range from 16 to 38 nM was determined. s-HPC4-EPCR inhibited Fl-APC somewhat more weakly (K d (app) Ϸ170 nM) (data not shown), and inhibition was independent of whether or not the preparation was further purified on a Mono Q column. A recombinant prethrombin-1 fusion protein which also contains the HPC4 epitope at the amino-terminal had little effect on Fl-APC binding at concentrations as high as 5 M (data not shown).
Since Fl-APC binding to the cell surface is metal-dependent, it was not possible to use this system to examine the metal dependence of the interaction with soluble EPCR. To address this question, we examined s-HPC4-EPCR binding to immobilized protein C. A noninhibitory antibody (HPC2) was used to immobilize protein C. HPC2 is a calcium-independent, anti-human protein C IgG1 monoclonal antibody that also reacts with human APC. HPC2 does not inhibit Fl-APC binding to E-7 cells. s-HPC4-EPCR did not bind to the HPC2 column in the absence of protein C (Fig. 9A, upper panel). When human protein C was adsorbed to the column prior to application of the s-HPC4-EPCR, most of the s-HPC4-EPCR bound in the presence of metal ions and was eluted with EDTA ( Fig. 9A, lower panel). The eluted fraction contained the 46-kDa s-HPC4-EPCR protein (Fig. 9B) that bound HPC4 on Western blots (Fig. 9C). Most of the contaminating protein broke through the column (Fig. 9B). When the column was run in buffers with Ca 2ϩ , but without Mg 2ϩ , approximately 30 -50% of the EPCR eluted before application of the EDTA elution buffer (data not shown). FIG. 9. Affinity chromatography of s-HPC4-EPCR on protein C. A noninhibitory murine IgG1 monoclonal antibody to human protein C, HPC2 (5 mg), was adsorbed to 2 ml of Affi-Gel Protein A (Bio-Rad) in a 0.6 ϫ 5-cm column. A, s-HPC4-EPCR (250 g) was chromatographed on this column without protein C adsorbed (control). B, 998 I.U. (Ϸ5 mg) of protein C concentrate (Immuno Ag) was adsorbed to the column, before s-HPC4-EPCR was applied. A and B, the column was equilibrated with HBSS containing 1.3 mM CaCl 2 and 0.6 mM MgCl 2 . After s-HPC4-EPCR application, the column was washed with the same buffer and eluted (arrow) in HBSS in which the Ca 2ϩ and Mg 2ϩ were replaced with 1 mM EDTA. The column was run at room temperature with a flow rate of 0.5 ml/min, and 0.7-ml fractions were collected. The fractions were analyzed by Colloidal Gold Total Protein Stain (Bio-Rad) (B) and by Western blotting using the HPC4 antibody (C). For B and C, lanes 1 and 4 was the sample applied, lanes 2 and 5, breakthrough, fraction 6 from A bottom, and lanes 3 and 6, the column eluate, fraction 17. The lower lane numbers were unreduced samples and the higher numbers were reduced. DISCUSSION This study provides many lines of evidence that indicate EPCR is expressed as a cell surface protein and that it is directly involved in protein C/APC binding. First, a monoclonal antibody and affinity purified polyclonal antibodies raised against recombinant EPCR bound to HUVEC and to human or murine cell lines stably transfected with human EPCR but not to the same sham-transfected cell lines. Second, APC and the monoclonal antibody bound to cells transiently transfected with human EPCR, but only APC bound to cells transfected with murine or bovine EPCR. Thus, binding of the antibody is species-specific and cell line-independent. Third, when the FLAG epitope was inserted into an EPCR construct and expressed in 293T cells, the cells bound both the FLAG antibody and APC. Fourth, when HUVEC or EPCR transfected E-7 cells were surface-labeled with biotin and subsequently immunoprecipitated, a major product of 49 kDa was detected in both cell lines but not from control cells. This product was similar in mass to the soluble recombinant s-HPC4-EPCR (46 kDa) and considerably larger than predicted based on the amino acid sequence (25 kDa). The increased mass was probably due to the presence of carbohydrate on EPCR, which contains four potential N-glycosylation sites (11). Fifth, truncation immediately above the putative membrane spanning domain leads to the formation of soluble EPCR (s-HPC4-EPCR or the form with the FLAG epitope). Taken together, these studies demonstrate that EPCR is a type 1 transmembrane cell surface glycoprotein.
EPCR is homologous to the CD1/MHC class 1 family of molecules (11). Many members of this family function as heterodimers (29). It was therefore of importance to determine if EPCR was sufficient by itself to allow protein C binding. The observation that s-EPCR-HPC4 inhibits binding of APC to the cell surface in a concentration-dependent fashion supports the concept that EPCR can bind to APC in solution. The K d (app) Ϸ30 nM calculated from the competition study is almost identical to that determined by the binding to EPCR on the cell surface. This observation is surprising since APC binding to EPCR-expressing cells requires Ca 2ϩ and deletion of the Gla domain eliminates the ability to compete with intact APC for binding to cellular EPCR. These metal and Gla domain dependencies are similar to the requirements for protein C binding to liposomes that had led us to believe that the phospholipidprotein C interaction was a major contribution to the EPCRdependent cell surface interaction. The observation that the binding affinity of APC for soluble EPCR is similar to that for the cell surface EPCR indicates that the majority of the energy involved in APC binding to the cell surface is derived from protein-protein interactions between protein C and EPCR with phospholipid-protein C interactions playing a relatively minor role in the binding affinity. This situation is similar to the other integral membrane cofactors, tissue factor and thrombomodulin, where the ligand affinity is determined primarily by protein-protein interactions (1,30).
Although EPCR appears to be able to interact with APC directly (31), this does not mean that it does not associate with other proteins on the membrane surface. Indeed, immunoprecipitation from the endothelium resulted in at least one additional protein band being precipitated by the immune Ig but not by control Ig. This band was not detected in the E7 cells, and hence may be associated in a cell type-specific fashion with EPCR. The nature and function of these putative EPCR-associated proteins will require characterization.
It is known that binding of APC to EPCR-expressing cells is Ca 2ϩ -dependent, but the role played by Ca 2ϩ in this process is uncertain. It was possible that the Ca 2ϩ dependence reflected the phospholipid-protein component of the interaction and that the protein-protein interaction was not metal ion-dependent. Affinity chromatography of s-HPC4-EPCR on immobilized protein C demonstrated that protein-protein interaction required divalent metal ions. Thus, EPCR-protein C interaction depends on both the Gla domain and metal ions. During the course of these studies, we observed that Mg 2ϩ facilitated binding to EPCR-expressing cells. Mg 2ϩ , however, would not substitute for Ca 2ϩ . The exact role played by Mg 2ϩ in this process remains to be determined.
Essentially equal Fl-APC binding was observed with wild type EPCR and a mutant in which the residues 198 -206 were replaced by the FLAG sequence. Thus, the region located just above of the transmembrane domain is not involved directly in the protein C binding. Another mutant, mFL-2, does not bind to APC. The residues between 146 and 153 replaced in the mutant may either be involved in protein C binding or the replacement with the FLAG epitope may have altered the conformation. Insertion of the epitope did not, however, prevent transport to the membrane surface, a problem often associated with misfolding.
These studies demonstrate that EPCR is expressed on cell surfaces as a type 1 transmembrane glycoprotein and that EPCR binds to protein C/APC directly in a metal-dependent fashion. The development of the expression system and preparation of antibodies provide key reagents for the elucidation of EPCR function.