Recombinant Human Extrinsic Pathway Inhibitor PRODUCTION, ISOLATION, AND CHARACTERIZATION OF ITS INHIBITORY ACTIVITY ON TISSUE FACTOR-INITIATED COAGULATION REACTIONS*

shown that extrinsic pathway inhibitor (EPI) is an effective inhibitor of factor Xa alone or factor VIIa-tissue factor complex in the pres- ence of factor Xa. Since tissue factor exposure is im-plicated

EPI and HeLa EPI. The ability of rEP1 to inhibit factor X activation by a complex of factor VIIatissue factor was then examined in the presence and absence of plasma concentrations of human factors VIII and IX. Using relipidated human brain tissue factor apoprotein, rEP1 inhibited the factor VIIa-mediated activation of factor X half-maximally at 2.5 and 1 nM in the presence and absence of factors VIII and IX, respectively.
Using monolayers of a human bladder carcinoma cell line (582) as the source of tissue factor, the activation of factor X by cell-bound factor VIIa was inhibited half-maximally by 5 nM rEP1 in the presence of factors VIII and IX, and at 0.8 nM EPI in the absence of factors VIII and IX. The proteolytic activity of 582 cell-bound factor Xa toward prothrombin was inhibited half-maximally at -5 nM rEPI, while the amidolytic activity of factor Xa in solution was inhibited by rEP1 with a I& of 130 PM. Recombinant EPI also inhibited the amidolytic activity of factor VIIa half-maximally at 10 nM rEP1 in the presence of relipidated tissue factor apoprotein and calcium. These results indicate that, in the presence of plasma concentrations of factors VIII and IX, at least 10 times the plasma concentration of EPI is required to reduce fac-* This work was supported in part by Research Grant HL 35246 from the National Institutes of Health and a grant from the Blood Systems, Inc. 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 U.S.C. Section 1734 solely to indicate this fact.
Y To whom correspondence should be addressed.
tor VIIa-dependent factor X activation one order of magnitude in vitro. In the absence of functional factor VIII and IX, rEP1 at plasma levels was a potent inhibitor of factor VIIa-mediated factor X activation, and this activity presumably accounts for the inability of hemophiliacs to initiate hemostasis via the extrinsic pathway.
The extrinsic pathway of mammalian blood coagulation is initiated when factor VII or factor VIIa binds to its cofactor, tissue factor. The factor VIIa-tissue factor complex then activates either factor IX or factor X by limited proteolysis which eventually leads to thrombin formation and a fibrin clot. In contrast to every other serine protease involved in blood coagulation, factor VIIa-tissue factor does not appear to be regulated to any significant extent in uivo by antithrombin III, the principal regulator of coagulation proteases. Based in part on studies reported by Hjort in 1957 (l), two different laboratories have identified a novel plasma protein that recognizes the factor VIIa-tissue factor complex (2,3). This plasma protein, designated as extrinsic pathway inhibitor (EPI)',' by Rapaport's laboratory (5), has now been purified to homogeneity from human plasma (6) and the conditioned, serum-free medium of the HepGZ hepatoma cell line (7,8). Purified preparations of plasma EPI were heterogeneous when examined by SDS-PAGE and exhibited major bands at 40 and 46 kDa and minor bands at 55,65,75,90,and 130 kDa (6). The 46-kDa form of EPI and higher M, forms were shown to be associated with apolipoprotein A-II in mixed disulfide linkages (6). All forms of plasma EPI were recognized by antibodies directed against EPIs amino and carboxyl termini (6). The amino acid sequence of the first 20 residues of plasma EPI were identical to that observed for HepG2 EPI with the exception of residue 5 where leucine was identified in plasma EPI and glutamic acid was found in HepG2 EPI (6). The majority of plasma EPI has been shown to circulate in complex with plasma lipoproteins, although the functional significance of this interaction is not fully understood. The amino acid sequence of human EPI has recently been deduced from the nucleotide sequence of a cDNA coding for EPI (9). The EPI cDNA codes for a mature protein consisting of 276 amino acids (M, 32,000) with 18 cysteines and three canonical N-linked glycosylation sites (9). The sequence of EPI revealed a highly positive amino terminus, three tandemly repeated Kunitz-type serine protease inhibitory domains, and a highly positively charged carboxyl terminus (9). Site-directed mutagenesis experiments indicated that the Kunitz domain 1 interacts with the factor VIIa active site while the Kunitz domain 2 binds to the factor Xa active site (10). Mutation of the reactive-site residue of the Kunitz domain 3 had no effect with respect to factor VIIa/Xa inhibition ( 10).
The mechanism whereby EPI inhibits factor VIIa-tissue factor is currently thought to occur in two discrete steps (4,11). In the first step, EPI binds to factor Xa through the interaction of its Kunitz domain 2 and the reactive site of factor Xa. In the second step, EPI Kunitz domain 1 interacts with the active site of factor VIIa. This mechanism appears to be operative on cell-surface tissue factor as well as soluble tissue factor generated from relipidated tissue factor apoprotein (5,12).
In the present study, we have expressed human EPI in stably transfected baby hamster kidney cells and purified the protein from cell culture media. The molecular properties of the recombinant EPI were compared with that observed for EPI synthesized by HepG2 and HeLa cells. Finally, we have investigated the inhibitory activity of EPI toward factor VIIatissue factor in the presence of physiological levels of factor VIII and factor IX using soluble and cell-surface expressed tissue factor. Our data indicate that EPI may be responsible, in part, for the severity of bleeding episodes observed in hemophiliacs.
Our findings also provide support for the use of recombinant EPI as adjunct therapy in the treatment of some thromboembolic diseases, although doses in excess of 10 times the plasma concentration of EPI may be required to achieve a therapeutic effect.  Bell and Alton (13). All other reagents were the best grade commercially available. Cell Culture-Human bladder carcinoma cells (582) were cultured as previously described (14). Factor X activation studies on 582 monolayers were performed within 24 h after reaching confluence in 12-well plates.
Proteins-Human plasma-derived factor IX, factor IXa, factor X, factor Xa, prothrombin, and thrombin were purified to homogeneity as described (15)(16)(17). Recombinant human factors VII and VIIa were purified from baby hamster kidney cell culture medium as described by Pedersen et al. (18) and Thim et al. (19), respectively. Recombinant human tissue factor apoprotein, produced in Escherichia coli and purified to homogeneity by immunoaffinity chromatography (20), was kindly provided by Dr. Gordon Vehar, Genentech, Inc., South San Francisco, CA. Human brain tissue factor was partially purified from acetone brain powder through the 2% Triton X-100 extraction step essentially as described (15) Table I. In this  purification procedure, heparin-Sepharose chromatography was selected as the initial step since earlier studies demonstrated that HeLa cell EPI possessed a relatively high affinity for heparin-Sepharose. 3 For recombinant EPI, the heparin-Sepharose step results in more than loo-fold purification from culture media. The recovery of EPI activity in each step was 55-86% with the exception of the reverse-phase HPLC where virtually a quantitative recovery was observed. Recombinant EPI eluted from the Pro RPC reverse-phase column as three closely spaced peaks (data not shown). As EPI-specific activity in each peak was identical, all three peaks were pooled for further characterization.
The structural differences between the three recombinant EPI isoforms eluting from the Pro RPC column are unknown, although it is probable that the observed heterogeneity reflects different degrees of glycosylation. A similar heterogeneity has been observed by HPLC separation of HeLa cell-derived EPI." Sodium dodecyl sulfate-polyacrylamide gel electrophoresis of recombinant EPI from each step in the purification indicated that recombinant EPI was essentially pure after the Mono S-FPLC step. As shown in Fig. 1, purified recombinant EPI migrated as a single band in SDS gels with an apparent molecular weight of 40,000 in the absence of reducing agent and 42,000 in the presence of 10% P-mercaptoethanol.
SDS-PAGE immunoblot analysis of recombinant EPI indicated a single band that migrated with essentially the same molecular weight as HeLa cell and HepG2 cell-derived EPI (Fig. 2). Although not proved in this study, small differences in the apparent M, values of these proteins are probably due to differences in glycosylation. Amino-terminal amino acid sequence analysis of recombinant EPI indicated a single sequence of Asp-Ser-Glu-Glu-Asp-Glu-Glu-His-Thr-Ile-Ile-Thr-Asp. This sequence is identical to that found for HeLa cell EPI,3 and HepG2 EPI (9).  4) that was maximal at equimolar concentrations of factor VIIa and tissue factor apoprotein (data not shown). Recombinant EPI inhibited the amidolytic activity of an equimolar complex of factor VIIa-tissue factor in a dose-dependent manner with half-maximal inhibition at 10 nM EPI (Fig. 5).

Recombinant
EPI had little, if any, effect on the amidolytic activity of factor VIIa in the absence of tissue factor, suggesting that EPI recognized a conformation in factor VIIa induced through its association with tissue factor.
As factor VIIa-tissue factor activates both factors IX and X, and EPI inhibited the amidolytic activity of factor VIIatissue factor, we next examined whether or not recombinant EPI inhibited factor VIIa-tissue factor proteolytic activity toward factor IX using factor VIIa bound to 582 cell surfaceexpressed tissue factor. Inconsistent with its effect on factor VIIa-tissue factor amidolytic activity, but consistent with earlier reports, no inhibition of factor IX activation by 582 cell-bound factor VIIa was observed in the presence of several concentrations of recombinant EPI and in the absence of exogenous factor Xa. Precisely how EPI can inhibit the active site of factor VIIa toward a small chromogenic substrate and step Reduced and unreduced samples were electrophoresed in a 10% polyacrylamide slab gel according to Laemmli (31). . Approximately 1 pg of protein was loaded in each well. Following electrophoresis, proteins were detected with rabbit IgG against the amino terminus of EPI. have no effect on a large protein substrate is unknown and merits further investigation. With the exception of its ability to inhibit factor VIIa-tissue factor amidolytic activity, recombinant EPI appears to exhibit the same molecular properties and inhibitory specificity as that reported for EPI isolated from HepGP media (7,8).
Only a limited amount of information is currently available concerning the activation of factor X by factor VIIa-tissue factor in the presence of EPI, factor VIII, and factor IX (37). Thrombin-stimulated human umbilical vein endothelial cells express cell-surface tissue factor that is functionally equivalent to purified relipidated tissue factor apoprotein (12). Rapaport and co-workers (12) have demonstrated that endothelial cell-surface tissue factor activity is inhibited by dilute human plasma in a reaction that is blocked by polyclonal antibodies directed against plasma EPI. In addition, EPI isolated from HepGP-conditioned media has recently been shown to inhibit factor VIIa bound to tissue factor expressed on subcultured fibroblasts in a reaction that depended upon exogenous factor Xa (38). In order to gain insight into the Relipidated, recombinant human tissue factor apoprotein (10 nM effective concentration) was incubated with various concentrations of factor VIIa (O-10 nM) in Tris-NaCl-CaClz-BSA buffer (pH 7.5) for 15 min at 25 "C. Factor VIIa amidolytic activity was then assessed using S-2288 chromogenic substrate.
inhibitory efficacy of EPI during normal hemostasis, we investigated the effect of various concentrations of recombinant EPI on factor X activation by factor VIIa-tissue factor in the presence and absence of factors VIII and IX. In these studies, relipidated human brain tissue factor apoprotein was used as the tissue factor. Preliminary studies performed in the absence of EPI indicated that more factor Xa was formed in the copresence of factors VIII and IX than in their absence, suggesting that trace amounts of factor Xa generated early in the reaction activated factor VIII (39). Factor IXa generated by factor VIIa-tissue factor, in complex with factor VIIIa, presumably activates additional factor X. When increasing amounts of recombinant EPI were added to the above reaction mixtures, an inhibition of factor X activation was observed (Fig. 6). The inhibition of factor X activation by recombinant EPI was considerably more effective in the absence of factors VIII and IX, and half-maximal inhibition was observed at 1 nM recombinant EPI (Fig. 6). In the presence of physiological levels of factors VIII and IX, 2.5 nM recombinant EPI was required to effect half-maximal inhibition. In all of these experiments, factor Xa activity was measured using a chro- mogenic substrate (S-2222) and the activity averaged over the first 5 min of assay. These results indicate that EPI, considered to be a slow-acting inhibitor, has a profound effect on the generation of factor Xa activity even when the activity is assessed in the initial phase of the reaction.
We next investigated the effect of recombinant EPI on the activation of factor X by factor VIIa in complex with tissue factor constitutively expressed on the cell surface of a human bladder carcinoma cell line, 582. Initial experiments performed in the absence of EPI demonstrated a linear time course for the generation of factor Xa activity in the absence of factors VIII and IX, whereas a sigmoidal time course was observed in the presence of factors VIII and IX (data not shown). All subsequent experiments were performed as singletime point experiments at 10 min. This time point was in the linear portion of the time course in the absence of factors VIII and IX and close to the inflection point of the sigmoidal time course in the presence of factors VIII and IX. As shown in Fig. 7 of factor X activation in the absence of factors VIII and IX than in the presence of these clotting factors. Recombinant EPI completely blocked factor X activation by cell-bound factor VIIa in the absence of factors VIII and IX at a concentration of -10 nM, or roughly 4 units/ml of EPI assuming a plasma concentration of 2.5 nM EPI (7). In sharp contrast, the rate of factor X activation was reduced by -70% at 5 units/ml EPI in the presence of factors VIII and IX (Fig. 7).
The above results suggested that factor Xa in the cell experiments was still capable of activating factor VIII in the presence of fairly high concentrations of recombinant EPI. Recent studies performed in our laboratory indicated that 582 cells possess -28,000-binding sites for human factor Xa (& = 1.6 nM factor Xa), and that cell-bound factor Xa readily activates prothrombin (36). Accordingly, we investigated the ability of recombinant EPI to inhibit the activation of prothrombin by cell-bound factor Xa. Fig. 8  These small differences in M, may be due to differences in glycosylation of EPI by HepG2, HeLa, and BHK cells. Purified recombinant EPI preparations exhibited a specific activity of 30,000 units/mg in a tissue factor inhibition assay. This value is approximately 2-fold higher than that reported for purified plasma EPI and EPI isolated from HepG2 conditioned media (6). HeLa cell-derived EPI exhibited a specific activity of 20,000 units/mg by the same method used to assay recombinant EPI." The reason(s) for the significant differences in specific activity values between recombinant EPI and that reported for plasma and HepG2 EPI are unknown, but may reflect subtle differences in the assays and reagents employed.
According to earlier reports, EPI does not inhibit the proteolytic activity of factor VIIa-tissue factor in the absence of factor Xa (2,3). This finding was confirmed in the present study using factor IX as the substrate for factor VIIa-tissue factor. However, in separate studies, we observed that tissue factor augmented the amidolytic activity of factor VIIa toward S-2288. The amidolytic activity of factor VIIa, in the presence of tissue factor but not in its absence, was inhibited halfmaximally by recombinant EPI at 10 nM EPI. This finding confirms earlier data that determinants present in factor Xa, other than its active-site serine residue, are important for the formation and stabilization of the double protease-inhibitor complex (11). In this regard, it would be informative to examine the proteolysis of an active site serine mutant of human factor X by factor VIIa-tissue factor in the presence of rEPI, and these experiments are currently underway in our laboratory.
Previous studies revealed that EPI inhibited a complex of factor VIIa-tissue factor on subcultured fibroblasts (38) and thrombin-stimulated human endothelial cells (12). These studies, however, did not examine the influence of the copresence of factors VIII and IX on the formation and expression of factor Xa proteolytic activity. In this study, we observed that, in the absence of EPI, more factor Xa activity was produced by factor VIIa-tissue factor in the presence of factors VIII and IX, supporting earlier data concerning the activation of factor X by factor VIIa in complex with perturbed bovine aortic endothelial cell-surface tissue factor (40). Recombinant EPI inhibited the activation of factor X by factor VIIa-tissue factor in a dose-dependent manner in the absence of factors VIII and IX. This inhibitory activity was, however, markedly reduced in the copresence of factors VIII and IX irrespective of the source of tissue factor used, i.e. soluble tissue factor or cell-surface tissue factor provided by the human bladder carcinoma cell line, 582. Additional experiments revealed that substitutitig factor VIIa with zymogen factor VII did not influence the EPI inhibition pattern (data not shown). Thus, activation of tissue factor-bound zymogen factor VII is apparently not inhibited to any appreciable extent by recombinant EPI. This observation may stem from the relatively slow action of recombinant EPI on factor Xa, thus allowing rapid activation of factor VII bound to tissue factor before EPI inhibits factor Xa (14,41,42). This hypothesis is supported by the putative incipient activation of factor VIII by factor Xa in our incubation mixtures even in the presence of relatively high levels of EPI. These observations seemingly have great relevance concerning the inability of hemophiliacs, who presumably have ample tissue factor and normal levels of factors VII and X but lack either functional factor VIII or factor IX, to initiate hemostasis via the extrinsic pathway of blood coagulation.
Thus, in the absence of either functional factor VIII or factor IX, sufficient factor Xa to initiate clot formation would not be generated in a hemophiliac, whereas normal individuals with intact factor VIII/IX should be able to overcome the inhibition of factor VIIa-tissue factor through the formation of factor Xa by the intrinsic pathway. Our findings do not shed light on the mechanisms associated with the observed therapeutic effectiveness of infused plasma-derived or recombinant factor VIIa in arresting bleeding episodes in hemophiliacs with circulating antibodies against factor VIII (43-45). Presumably, these individuals have normal circulating levels of EPI which, as noted in a single patient, do not appear to change following infusion of recombinant factor VIIa (46). One explanation for the beneficial effect of recombinant factor VIIa in the treatment of inhibitor patients may be that factor VIIa directly activates factor X in the absence of tissue factor but in the presence of calcium and an appropriate phospholipid membrane as proposed by several groups (47-49). EPI appears to recognize only that conformation in factor VIIa generated by its association with tissue factor. The rate of factor X activation by factor VIIa-phospholipid would appear to be insignificant given the relatively low affinity of factor VII/VIIa for acidic phospholipids (50, 51). On the other hand, factor X has a relatively high affinity for mixed phospholipid vesicles (52), and conceivably mechanisms are operative that would allow high affinity binding of factor VIIa to the phospholipid membrane through its affinity for factor X. This possibility is currently under investigation in our laboratory.