Phospholipid-independent and -dependent Interactions Required for Tissue Factor Receptor and Cofactor Function*

Membrane anchoring of tissue factor (TF), the cell receptor for coagulation factor VIIa (VIIa), exemplifies an effective mechanism to localize proteolysis at the cell surface. A recombinant TF mutant (TF1-219), deleted of membrane spanning and intracellular domains, was used to evaluate the role of phospholipid interactions for assembly of substrate with the catalytic TF-VIIa complex. TF1-219 was secreted by cells rather than expressed as a cell membrane protein. Unlike free VIIa, TF1-219 as well as the TF1-219*VIIa complex dem- onstrated no stable association with phospholipid. In the absence of lipid, kinetic evaluation of substrate factor X cleavage by free VIIa, TF-VIIa, and TF1-219*VIIa suggests that the catalytic function of VIIa rather than substrate recognition is enhanced by complex formation. Furthermore, compared with free factor X, factor X on phospholipid was preferentially cleaved as a substrate by TF1-219.VIIa. TF-dependent initiation of the coagulation protease cascades thus involves an enhancement of the activation of factor X on the cell surface by a crucial role of the TF transmembrane domain to membrane anchor the reaction, by the TF extracellular domain to provide protein-protein interactions with VIIa to enhance the activity of the catalytic domain of VIIa, and the preferential presentation of factor X as a substrate when associated with phospholipid surfaces. analyze the cleavage of solution-phase factor Xa analyze effect of phospholipid on using of phospholipid effect prothrombin on

TF, allosteric changes in VIIa enable or markedly enhance catalytic function of the bound serine protease domain of VIIa. The functional assembly of TF, VIIa, and factor X as a ternary complex of activator, enzyme, and substrate, respectively, has been characterized as an ordered addition model (6) in which substrate associates with a relatively stable activator-enzyme complex to form a transient ternary complex. The catalytic function of the binary TF . VIIa complex may be markedly reduced if the surface organization of the catalytic complex is disturbed by solubilizing TF, as deduced from reactions in the presence of nonionic detergent (7). This enhancing effect of a charged phospholipid surface could be ascribed (i) to secondary VIIa association with the phospholipid surface to enhance formation of the proteolytically active TF. VIIa complex, or (ii) to effects of association of factor X with phospholipid on its presentation as a substrate for the TF. VIIa complex.
Analysis of a soluble form of cell surface receptors provides novel insight into structure-function relationships. Proteolytic digestion and isolation of soluble domains of thrombomodulin, the anticoagulant thrombin receptor and cofactor, have been successfully applied to provide insight into the functional domains and their phospholipid interactions (8,9). A truncated form of the major histocompatibility complex has also been adopted for solution of the three-dimensional structure and function (10). Functional analysis of the isolated extracellular domain of TF should further an understanding of the cell surface interactions of TF that may be required for full receptor and cofactor function. A soluble form of the cell surface domain of T F would provide a model for analysis of the interaction of T F with VIIa independent of surface assembly and potential secondary phospholipid-induced effects. To this end, a TF mutant deleted of membrane spanning and intracellular domains was produced and used to characterize primary assembly of the TF-VIIa complex and its catalytic function towards the substrate factor X in solution in comparison with phospholipid-bound factor X. These data provide evidence for catalytic function of TF. VIIa independent of assembly on phospholipid and further demonstrate that the primary protein:protein interactions of VIIa with the surface domains of T F alone are sufficient for marked enhancement of the catalytic function of VIIa. They also indicate that association of factor X with phospholipid renders it a more effective substrate than when free in solution. The results have been incorporated into a model for assembly-dependent initiation of coagulation on cell surfaces.

EXPERIMENTAL PROCEDURES
Proteins-Factor VI1 (VII) ( l l ) , factor X (12), factor IX (13), and factor Xa (13) activated by Russell's viper venom (14) were prepared as described and were homogeneous as judged by SDS-polyacrylamide gel electrophoresis (15). Prothrombin was prepared from fractions obtained during the factor X purification using adsorption to Hepa-rin-Sepharose (16) and elution with a NaCl gradient from 0 to 0.2 M in 0.02 M MES, Tris, pH 5.9, in the presence of 2.5 mM CaCl, and 1 mM benzamidine. HCl. Prothrombin fragment 1 was prepared by the procedure of Malhotra (17) with the following modifications. Prothrombin was digested with thrombin a t a 1OOO:l (w/w) ratio for 17 h followed by isolation of uncleaved prothrombin and prothrombin fragment 1 by ion-exchange chromatography on Mono Q (Pharmacia LKB Biotechnology, Inc.) using adsorption at 300 mM NaCl and elution a t 400 mM NaCl in 20 mM Bis-Tris, pH 6.1. Prothrombin was removed from the preparation by gel filtration using Sephadex G-75 Superfine. A homogeneous preparation of prothrombin fragment 1 was obtained by this procedure as judged by SDS-gel electrophoresis. VI1 was converted to VIIa by incubation of the purified protein with factor Xa immobilized on Affi-Gel 15 beads (Bio-Rad). Conversion was monitored by SDS-polyacrylamide gel electrophoresis of reduced samples. No free factor Xa in the VIIa preparation (1.2 p M ) was detectable with the chromogenic substrate methoxycarbonyl-D-cyclohexylglycyl-glycyl-arginine-p-nitroanilide acetate (Spectrozyme FXa, American Diagnostics, New York) a t 0.2 mM final concentration in the presence of 50 mM EDTA. Alternatively, recombinant VIIa was purchased from Novo Industri (Bagsvaerd, Denmark). Hybridomas were propagated by ascites growth and the monoclonal antibodies (MAbs) were purified from ascites fluid using adsorption of MAb to immobilized protein A in the presence of 1.5 M glycine, 3 M NaCI, pH 8.9, followed by elution with 100 mM citric acid, pH 4.0. The MAb TF8-5G9 (18) was coupled to Affi-Gel-10 (Bio-Rad) a t p H 6.5 in 0.1 M MES at a concentration of 4 mg/ml of beads for affinity purification of TF and the mutant TF.
Protein Sequencing-TFl-zls was subjected to amino-terminal sequence analysis in an Applied Biosystems model 475A gas-phase sequenator at the Microchemistry Core facility of the Research Institute.
Construction of the TF1-219 Deletion Mutant and Stable Expression in Eukaryotic Cells-A 775-bp EcoRI fragment of the TF cDNA from nucleotides 1 to 775 (numbering based on Ref. 1) and encoding amino acids -32 through 216 as described (19), was ligated with the synthetic oligonucleotide fragments AATTTAGAGAATAAG and ATCTCT-TATTCTTAA. This resulted in the exchange of the triplet for Phe217 from TTC to TTT, destroyed the original EcoRI site at the 3' end of the cDNA fragment, converted the triplet for IleZz0 to a termination codon, and provided a new EcoRI cloning site at the 3' end. This fragment was ligated to EcoRI-digested pBR322. This EcoRI insert was subcloned into CDM8 (20) modified to contain an EcoRI cloning site (a gift from Dr. D. Dialynas, Research Institute of Scripps Clinic). The nucleotide changes introduced into the cDNA were confirmed by sequencing using single-stranded DNA rescued using the helper phage R408 (21) and dideoxynucleotide sequencing with Sequenase (U. S. Biochemical Corp.). Chinese hamster ovary cells (CHO-K1 cells, CCL 61, ATCC, Rockville, MD) were cotransfected with this construct and, as a selection marker, pMAM-neo (Clonetech, San Francisco, CA) a t a 2 0 1 ratio using the calcium phosphate method (22). The neomycin analogue G418 (GIBCO) was used a t 600 pg/ml to select for cells which stably incorporated foreign DNA. Stable single cellderived colonies were further selected by analyzing the medium for the TF protein by an enzyme-linked immunoassay based on two MAbs (18) to nonoverlapping epitopes on TF. The highest producing clone was expanded and cultivated in Ex-Cell 301 (J R Scientific, Woodland, CA) supplemented with 1% newborn calf serum (HyClone Laboratories, Logan, UT) for production of TF1-,,,. TF,~,,, without N-linked glycosylation was obtained from cells which were grown in the presence of a predetermined optimal concentration (1-2 pg/ml) of tunicamycin (Sigma).
Purification of TFand TF,~,,~--TF was purified by immunoaffinity chromatography as previously described (23). The same MAb, TF8-5G9, coupled to Affigel beads was used for the purification of TFI-219 from the culture supernatant. Detergent was not used during the purification of TF,.,,,. After removal of cellular debris by centrifugation and filtration (0.45 pm), the culture medium was incubated with anti-TF beads for 3 h or more. After sequential washes with T B S (20 mM Tris, 150 mM NaCI, pH 7.4), 1 M NaCl, and 0.1 M glycine at pH 4.5, T F protein was eluted from the beads with 0.1 M glycine buffer, pH 2.5. The pH was adjusted to pH >6.0 immediately upon elution and the protein was dialyzed against TBS for storage at -70 "C. Protein concentrations were determined by the BCA assay (Pierce Chemical Co.) based on standard T F or TF1-219 of known concentration which had been determined from quantitative amino acid composition of the purified protein.
Ligand Blotting of TF and TF,.,,,-The interaction of VI1 with purified TF or TF,.,,, was analyzed by ligand blotting. The purified proteins were separated by SDS-polyacrylamide gel electrophoresis on 12.6% slab gels using the buffer system of Laemmli (15) and electrophoretically transferred (24) to nitrocellulose (Schleicher & Schuell). After incubating the membranes with 5% nonfat dry milk in TBS to block residual adsorption, and three washes in TBS, the membranes were incubated with 50 nM Iz5I-VII in the presence of 5 mM CaC1, and 1% BSA for 2 h a t ambient temperature. Control reactions were performed in the presence of (i) 5 mM EDTA or (ii) the monoclonal antibody TF9-6B4 which blocks VI1 binding to TF (18,25) in a 50-fold molar excess over VII. Both controls abolished binding of VI1 to TF or TF1.219. Unbound reactivity was removed by washes in TBS, 0.1% BSA, 0.05% Tween 20, 5 mM CaC12. Bound radioactivity was detected by autoradiography and quantitated by determining emission from nitrocellulose strips of equal size in a y counter. Western Blot Analysis-Immunoreactivity was analyzed with proteins electrophoretically separated and transferred to nitrocellulose as described above. For analysis of non-N-linked glycosylated TF,.,1, produced in the presence of tunicamycin, 100 p1 of culture supernatant was applied per gel lane. After transfer, the membranes were blocked with 5% nonfat milk in TBS. The primary antibody was diluted into 5% nonfat dry milk in TBS to 1 pg/ml and reacted with the nitrocellulose membrane for 1 h at 37 "C. After washes in TBS the membrane was incubated with an appropriate (goat anti-mouse IgG or anti-rabbit IgG, Tago, Burlingame, CA) alkaline phosphatasecoupled secondary antibody a t a 1:2000 dilution for 30 min at 37 "C. Following washes, bound alkaline phosphatase was visualized by a color reaction from a substrate mixture of 330 pg/ml nitro blue tetrazolium (Sigma) and 165 pg/ml5-bromo-4-chloro-3-indolyl phosphate (Sigma) in 0.1 M Tris, 0.1 M NaCI, 5 mM MgCl, pH 9.5.
Interaction of TFl-219 and VII with Phospholipid-Proteins were labeled with carrier-free NaLZ5I (Amersham Corp.) by the coupled lactoperoxidase, glucose oxidase method (Enzymobeads, Bio-Rad) to similar specific activities of 1-4 pCi/pg. The radiolabeled proteins were functionally indistinguishable from unlabeled proteins by kinetic analysis in the factor Xa generation assay with phospholipid present (see below). Binding of the proteins to phospholipid was analyzed in 96-well microtiter plates (Falcon MicroTest 111, Becton Dickinson Labware, Oxnard, CA). Phospholipid vesicles (600 p~) prepared by the deoxycholate method (26) were adsorbed in TBS overnight to the plastic well surface (100 pl/well). After washes with TBS, the plate was incubated with 10% BSA in TBS for 1 h at 37 "C. Nonspecific binding was determined on phospholipid-free wells which were incubated only with 10% BSA. Three washes with TBS were performed before LZ51-labeled VI1 or TF1.2,, were added at indicated concentrations in the presence of 4 mM CaCl, and 0.2% BSA. Following incubation a t 37 "C for 90 min, the wells were washed rapidly 10 times with ice-coldTBS, 5 mM CaCI2, 0.5% BSA. Radioactivity bound to the wells was quantitated in an Isodata 20/20 y counter. Values were obtained from duplicate determinations and experiments were repeated a t least three times.
Amidolytic Activity of VIIa Bound to TF or TFl.21g-Catalytic activity of VIIa for peptidyl substrate was analyzed by hydrolysis of Spectrozyme FXa. Various concentrations of VIIa and nonphospholipid-associated T F (solubilized in Triton X-100) were incubated in 0.2% BSA, 5 mM CaCI,, and 1.25 mM Spectrozyme FXa. The rate of product formation was monitored at 405 nm and corrected for minor ( 4 0 % ) background hydrolysis which occurred in the absence of T F at high VIIa concentrations. The velocity of absorbance increase was determined in a kinetic plate reader and three or more periods of 1 or 2 min were used for data collection. Usually the initial reading yielded representative results for a 10-min period.
Functional Analysis of the Cleavage of the Natural Substrate Factor X-The cleavage of factor X by the TF. VIIa complex was analyzed in a coupled amidolytic assay essentially as described (25) using Spectrozyme FXa for assay of factor Xa generation. (i) T o analyze factor X cleavage in solution, factor Xa formation by VIIa in the presence of TF1.Zl9 was compared with identical concentrations of detergent (1% (v/v) Triton X-100)-solubilized T F (1.6 p~) which was diluted in TBS. VIIa (50-1000 nM) was incubated with various concentrations of T F or TF,.,,, in the presence of 5 mM CaCI, for 10 min followed by addition of the substrate factor X. Aliquots were removed every minute and the reaction stopped in 100 mM EDTA.
Initial rates of factor X formation were linear in the first 3 min and an average rate was typically calculated from individual samples obtained during that period. From these initial rates apparent Michaelis-Menten parameters were estimated using the Wilman4 com-puter program (27). To analyze the cleavage of solution-phase factor X by VIIa in the absence of TF, but presence of 5 mM CaCI2, aliquots were removed after 20, 40, and 60 min from the reaction mixture at 37 "C and stopped in 100 mM EDTA. Rates of factor Xa formation were calculated from these samples and catalytic parameters determined as mentioned above. (ii) To analyze the effect of phospholipid on the cleavage of factor X, purified human T F reconstituted into a 7800-fold molar excess of phosphatidylcholine and phosphatidylserine (70/30, w/w) (Sigma) vesicles using deoxycholate solubilization and dialysis (26) was compared to TF1.219 in the presence of vesicles of identical phospholipid composition prepared by the same method. The assay procedure was as above. (iii) To study the effect of prothrombin fragment 1 on the reaction rate of TFL-P19.VIIa cleavage of phospholipid-bound factor X, factor X, prothrombin fragment 1, and phospholipid were incubated in the presence of 5 mM CaCI2 for 3 min at 37 "C and the reaction initiated by the addition of 1 p M TFl-n19 and 10 nM VIIa. Rates were determined as above. (iv) Cleavage of factor X by the factor X activator from Russell's viper venom was studied by prewarming the enzyme (36.8 ng/ml) with 100 p~ mixed phospholipid vesicles or buffer in the presence of 5 mM CaC12 for 3 min followed by addition of the substrate factor X. Rates of factor Xa formation were determined and kinetic parameters calculated as described above.

RESULTS
Isolation of TFl-219-A truncated T F DNA coding sequence was generated by replacing the Ile220 codon by a stop codon. The entire coding sequence of the mutant was confirmed by DNA sequencing of the construct which was subsequently cloned into the vector CDM8 for expression. A stable cell line was derived from a single cell which contained the integrated DNA for TF1-219. The synthesized TF1-219 protein was secreted into the culture medium at 0.5-1 mg/liter per 24 h and less than 10% of the immunoreactive protein remained cell-associated. In contrast, no immunoreactive material was found in the media from cells which expressed the natural full-length TF. This demonstrated that TF1-219 lacked the ability to stably localize as a transmembrane cell surface protein, which is consistent with the removal of the predicted transmembrane domain in TF1-219. TF1-219 was purified from the medium by affinity chromatography on an immobilized monoclonal antibody specific for human TF. This single-step purification yielded one major band of 40 kDa and three minor bands, the most rapidly migrating having an apparent molecular mass of 34.5 kDa under nonreducing conditions (Fig. 1, lane A ) . To investigate whether the minor bands were incompletely glycosylated forms of TF1-219, cells were incubated with tunicamycin to arrest N-linked glycosylation and the secreted TF1-219 was analyzed by SDS-polyacrylamide electrophoresis and immunoblotting. A single protein band with an apparent molecular mass of 34.5 kDa was observed under these conditions, consistent with a single protein species among which the predominant 40-kDa band was the most highly glycosylated form. All molecular forms reacted (Fig. l, lane B ) on Western blots with polyclonal rabbit anti-TF antibodies affinity-purified on immobilized human brain TF (19). Unlike natural intact T F isolated from human brain (23,28), no dimeric forms of TF1-219 were observed, presumably since the cysteine residue of the intracellular domain was deleted. TF1-219 reacted with an entire panel of 24 MAbs (18) against human TF (data not shown), suggesting proper folding of TF1-219. Amino-terminal protein sequencing of TF1-z19 established Ser'-Gly2-Thr3-Thr4-Asn5-Thr6 and Thr3-Thr4-Asn5-Thr6-Va17-Alas as the amino-terminal residues with two alternative processing sites as described for natural full-length TF purified from human brain (19). The amino acid composition was concordant with the composition of the expected TF1-219 protein, thus excluding deletions at the carboxyl terminus due to intracellular processing or extracellular carboxypeptidases.
VII Binding to TF1-21g-Ligand blotting of TF and TF1-219 with '2.51-VII was used to analyze direct protein:protein association. TF or TF1-219 were separated by SDS-polyacrylamide gel electrophoresis using nonreducing conditions, then transferred to nitrocellulose membranes. Incubation of the filters with 1251-VII in the presence of 5 mM CaClz demonstrated binding of VI1 to TF and TF-dimer. VI1 bound to TFI-219 independent of the degree of glycosylation (Fig. 1, lane C ) indicating that neither full glycosylation nor phospholipid association is a requisite for the VI1 association with TF or TF1-219 under these conditions. The binding of VI1 to both proteins was dependent on calcium ions, since addition of 5 mM EDTA prevented the interaction (Fig. 1, lane E ) . Specificity was further demonstrated by the following. (i) Monoclonal antibody TF9-6B4, specific for human brain TF and demonstrated to block binding of VI1 to TF (18,25), inhibited (>80%) association of VI1 with both proteins when included during the incubation in a 50-fold molar excess over VI1 (Fig.   1, lane F). (ii) A 50-fold molar excess of factor X, factor IX ( Fig. 1, lane D ) or prothrombin did not inhibit (<20%) the association of VI1 with TF and TF1-219. (iii) TFI-Z19 included in the incubation at 2.5 pM inhibited binding of VI1 to TF and TF1-219 to more than 80% (Fig. 1, lane G). These results are consistent with specific interactions of VI1 with TF1-219 in the absence of phospholipid.
TF1-219 Interaction with Phospholipid-To determine whether the predicted TF1-219 protein interacts with relative stability with phospholipid surfaces, a direct binding assay was adopted with the following considerations. (i) The TF extracellular domain might interact with a phospholipid surface such that the surface domains of the molecule fold back onto the surface in addition to the transmembrane anchoring domain. This interactive site might be expressed on TF alone or after complex formation with VIIa. (ii) TF1-219 binding to VIIa might result in a stable binary complex which could associate with the phospholipid surface via the y-carboxyglutamic acid (Gla) domain of VIIa or other inducedxonformational site.
For analysis, we used a binding assay which employs surface-immobilized, mixed phosphatidylcholine and phosphatidylserine. In contrast to the reported low affinity ( K d = 9 PM) of VI1 binding to phospholipid (5,29), we observed binding of VIIa to phospholipid at 200 nM (Fig. 2). In addition, prothrombin fragment 1 at 200 PM effectively inhibited binding of VIIa (Fig. 2), suggesting that the observed binding of VIIa might be Gla domain mediated. Although off-rates of proteins which bind via the Gla domain to phospholipid surfaces are relatively rapid if studied at 37 "C, the use of calcium containing cold buffers for washing in the assay apparently decreased the offrate to allow demonstration of interaction of VIIa with the phospholipid surface. Under these conditions, we were not able to demonstrate binding to phospholipid for: (i) TF1-219 alone; (ii) TFt.219 in the presence of VIIa (Fig. 2); or (iii) when the catalytic complex of TF1-219.VIIa together with the substrate factor X was analyzed. At VI1 concentrations as high as 1 PM and TF1-219 as high as 2.9 PM demonstrable specific binding of TFl-,19 to phospholipid was not observed. These data indicate lack of relatively stable interaction of TF1-219. VIIa with phospholipid under conditions where binding of free VIIa was demonstrated.
Amidolytic Activity of VIIa Bound to TF1-219-Hydr~ly~i~ of the peptidyl substrate methoxycarbonyl-D-cyclohexylglycylglycyl-arginine-p-nitroanilide (Spectrozyme FXa) was used to monitor accessibility and function of the VIIa catalytic triad in the absence of the effects of extended substrate recognition of factor X. This assay is specific for VIIa in complex with TF by the following criteria. (i) Hydrolysis of Spectrozyme FXa by factor Xa was not enhanced by TF; whereas measured hydrolysis by VIIa required the presence of TF. TF alone did not hydrolyze the substrate. (ii) Addition of EDTA (50 mM) to the VIIa amidolytic assay decreased cleavage of Spectrozyme FXa to background levels; whereas cleavage by purified factor Xa continued a t approximately half the velocity, in accordance with the limited cation-dependence of peptidyl hydrolysis by factor Xa (30, 31). (iii) Results were identical with recombinant VIIa. This substantially excludes that the observed amidolytic activity in the VIIa assay is due to factor Xa generated from traces of factor X in VIIa preparations purified from plasma. (iv) The factor Xa-specific substrate S2222 (Helena Laboratories, Beaumont, TX) was not cleaved by TF.VIIa ( 4 . 5 us. 35 mOD/min (rate of absorbance increase a t 405 nm) with Spectrozyme FXa under identical conditions). (v) Inactivation of VIIa by 1 mM D-phenylanalyl-L-prolyl-L-arginine chloromethylketone (Calbiochem) (32) followed by extensive dialysis to remove residual free Dphenylanalyl-L-prolyl-L-arginine chloromethylketone abolished all activity in the assay, demonstrating that the activity originates from the VIIa and not the TF or TF1-219 preparations. Taken together, these observations support the specificity of this assay for catalytic activity of VIIa in complex with TF.
We used hydrolysis of Spectrozyme FXa to compare expres- sion of cofactor function by TF and TF1-219. Increasing concentrations of detergent-solubilized TF as well as TF1-Z19 progressively increased the catalytic efficiency of VIIa. The rate of peptidyl hydrolysis was comparable for equivalent concentrations of TF or TF1-219 (Fig. 3A). Rates were also identical when a t a given TF and TFl-219 concentration the VIIa concentration was varied from 50 to 1000 nM. We conclude that the interaction of VIIa with the TF extracellular domain appears to be sufficient to induce accessibility and function of the catalytic site of VIIa for recognition and hydrolysis of the peptidyl substrate.
Cleavage of Factor X by the TF1-219. VIIa Complex in Solution-We investigated limited proteolytic activation of the natural substrate factor X in experiments similar to those described above. With VIIa held constant at 100 nM, TF was more efficient as the cofactor for VIIa-mediated proteolytic activation of factor X than was TFI.219 (Fig. 3B). This difference could not be attributed to residual detergent in the TF preparation, since addition of equal amounts of detergent (up to 0.35% Triton X-100) to TF1-219 did not influence the catalytic rate (data not shown). Catalytic parameters for the reactions yielded a comparable Km(app) for TF.VIIa and TF1-z19. VIIa with only small differences in the Vmax(app) (

TABLE I
Kinetic parameters for the cleavage of factor X Apparent Michaelis constants (Km(app)), maximal velocities ( Vmal(app))r and calculated kat based on the assumed effective enzyme concentration ([enzyme]) for factor X cleavage by VIIa free, in complex with T F solubilized with detergent (TF in detergent), soluble TF,_,,, (TF1-ZIg), or phospholipid-reconstituted T F (relipidated TF) are given as mean f standard deviation calculated from the indicated number of experiments. The enzyme (VIIa) and activator (TF) concentrations were varied as indicated. Factor X concentrations ranged from 0.05 to 32 p~. keet for factor X cleavage by TF1-2,9.VIIa in comparison to TF-VIIa was only partially compensated by increasing concentrations of VIIa and TF1_219 (Table I). This suggests alteration of substrate recognition and cleavage by TF1-219. VIIa in fluid phase. Proteolytic Activation of Factor X in the Presence of Phospholipid-The proteolytic activation of factor X in the presence of phospholipid was studied using T F reconstituted in phospholipid vesicles. Although a 12-fold increase in the catalytic rate was observed, the 230-fold decrease in the Km(app) ( Table I) contributed most to the enhancement of the catalytic efficiency. These data therefore suggest facilitated substrate access to the binary complex if the reaction is localized on a phospholipid surface. The lack of membrane insertion of TF1_2,g allowed study of the effect of membrane association of factor X independent of membrane anchoring of the binary complex. Increasing concentrations of phospholipid vesicles resulted in a considerable rate enhancement of the factor Xa activation by TFl_219.VIIa (Fig. 4). In the experiment depicted, saturation was not achieved, presumably because the substrate factor X was still in excess of the available phospholipid binding sites (2-7 mmol of factor X/mol of phospholipid (33,34)) in the concentration range of phospholipid studied. The Kd for the interaction of factor X with mixed phospholipid vesicles containing phosphatidylserine ranges from 28 to 40 nM (33,34). Therefore, the phospholipid vesicles in Fig. 4 can be assumed to be saturated with factor X at 1 PM, which was in excess over the phospholipid binding sites. This allowed estimation of apparent Michaelis-Menten parameters for the cleavage of membrane-associated factor X by TF1-219. VIIa, since the substrate concentration was given by the binding capacity of the added phospholipid for factor X. Assuming a stoichiometry of 2 mmol of factor X/mol of phospholipid analysis of the depicted data yielded a kc,, of 1.6 mol/mol/s and a Km(epp) of 520 nM. This kc, is close to the rate observed for intact, phospholipid-reconstituted TF ( Table I). Despite the minor differences in the catalytic efficiency for the fluid-phase reaction, cleavage of membraneassociated factor X seems to be equally enhanced by T F and TF1-219, indicating that TF1-219 contains all relevant structures for catalytic enhancement of VIIa.

TF
Addition of 100 PM phosphatidylcholine/phosphatidylserine vesicles at fixed activator, enzyme and substrate concentrations enhanced the rate more than 100-fold (Fig. 4) suggesting that factor X is preferentially cleaved when associated with the phospholipid surface. If this is the case, the reaction rate should reach saturation with increasing concentrations of factor X at a fixed phospholipid concentration. Indeed, the rate of factor Xa formation reached a maximum with increasing concentrations of factor X (Fig. 5B ). However, in contrast to the reaction with phospholipid-inserted TF, a further increase in the substrate concentration resulted in a decrease in the rate (Fig. 5A). With membrane-anchored TF the rate of factor Xa cleavage did not decrease at high factor X concentrations. In addition, lower VIIa concentrations were required in the presence of T F in comparison t o TF1-219 to obtain similar rates (Fig. 5 ) , suggesting an affinity difference which becomes apparent at low VIIa concentrations.
The observed decrease in rate at high factor X concentrations in the case of TF1_2L9. VIIa can be explained in two ways.
(i) Increasing concentrations of factor X might compete with TF1-219.VIIa for binding to a limited number of identical binding sites on the phospholipid surface. This would imply that the catalytic complex must be associated with phospholipid for enhanced recognition and cleavage of factor X as substrate. (ii) Alternatively, factor X bound to phospholipid represents a preferential substrate for TFI-Z1~+VIIa. The inhibition at higher factor X concentrations could result from association of the complex with fluid-phase factor X which  5. A , formation of factor Xa by TF.VIIa inserted into a phospholipid vesicle (39 phi). 5 nM TF, 0.1 nM VIIa were assembled for 10 min at 37 "C, followed by addition of various concentrations of factor X. The initial rate of these reactions is given as mean and standard deviation calculated from three experiments. The insert depicts the substrate concentration range up to 10 times the Km(spp)r a concentration range which was typically used for calculation of kinetic parameters. B, cleavage of factor X by TF,.,,,.VIIa was studied as described above for membrane-inserted TF. 1 p~ TF,_,,,, 100 p~ mixed phospholipid vesicles, and a 10-fold higher VIIa concentrations (1 nM) as in A were assembled for 10 min at 37 "C, followed by initiation with the substrate. Means and standard deviations calculated from three experiments are given.
represents a less efficiently cleaved substrate. Since inhibition was observed at factor X concentrations approximating the Km,app) of the fluid-phase reaction, the latter is plausible. To distinguish these two hypotheses, we designed the following experiment. Prothrombin fragment 1 is devoid of the second kringle and serine protease domain of prothrombin. It is devoid of proteolytic or amidolytic activity which might interfere with the chromogenic assay used, and it is not a substrate for factor Xa. Prothrombin fragment 1 contains the entire Gla domain of prothrombin which competes with phospholipid binding of factor X (34). Addition of prothrombin fragment 1 should therefore decrease the rate of factor Xa formation by TFl-219 VIIa on phospholipid either by restricting phospholipid binding sites for factor X or for the TF1-219. VIIa complex. If assembly of TF1-219. VIIa with the phospholipid surface at factor X and prothrombin fragment 1 binding sites is required for efficient cleavage of factor X, then, in the presence of prothrombin fragment 1, increasing concentrations of factor X should further decrease the rate, since binding sites would be further restricted. However, if the binding of factor X to the phospholipid surface is critical, in the presence of prothrombin fragment 1 increasing concentrations of factor X should restore phospholipid-associated factor X and thereby increase the reaction rate. At the concentrations of factor X which produced inhibition of the reaction, cleavage of factor X in solution is demonstrable (Fig.  6). Prothrombin fragment 1 at 2 p M effectively competed the enhanced reaction at low factor X concentrations in the presence of phospholipid. At 2 p~ prothrombin fragment 1, increasing the concentration of factor X accelerated the reaction rate, approaching the rate observed in the absence of prothrombin fragment 1 at higher factor X concentrations (Fig. 6). These data demonstrate that association of factor X with the phospholipid surface is the critical event which accelerates the reaction rate. The initial increase of the reaction rate excludes that TF1-,,, . VIIa associated with phospholipid cleaves free factor X more efficiently, since no further drop of the reaction rate was observed with increasing factor X concentrations. However, the drop in reaction rate depicted in Fig. 6 for the reaction with phospholipid, but without prothrombin fragment 1 appears to be more pronounced than expected from a simple competition of the fluid-phase reaction. Based on a Km(app) of 22 p~ for the fluid-phase reaction (Table I), a 21% reduction would be expected at 6 PM factor X. This is observed in Fig. 5 at 1 nM VIIa. Since the VIIa concentration in Fig. 6 was 10-fold higher, an effect of phospholipid binding of free VIIa could influence the reaction rate.
Since VIIa binding to phospholipid was demonstrated at the given concentration (Fig. 2), the binding of VIIa to phospholipid vesicles could influence the equilibrium with TF1-219 by providing a higher local concentration of VIIa in the phospholipid shell which could facilitate assembly of TF1-219. VIIa with phospholipid-bound factor X. If the concentration of VIIa is too low to significantly allow phospholipid interaction, as in Fig. 5, the reaction rates reflect the simple competition between two substrate species more adequately. Although the data in Fig. 6 exclude that TF1-219. VIIa associated with factor X binding sites during proteolysis of factor X, a transient interaction of the binary complex at a site different from the substrate binding site cannot be excluded from these experiments. These data therefore do not exclude TF1-219.VIIa interactions with phospholipid, but emphasize the significance of phospholipid-bound substrate factor X. These data therefore suggest a conformational transition of factor X upon membrane association. Additional evidence for an altered susceptibility of membrane-associated factor X for recognition and proteolysis by a relevant protease is demonstrated by the lack of cleavage of factor X by the activator from Russell's viper venom when factor X is bound to phospholipid. In contrast, the venom rapidly activates factor X when free in solution (34). We observed an increase in the Km(app) for cleavage of factor X by the venom enzyme upon  (n = 3), while the Vmax(app) was identical with (12.2 f 5.6 nM/min) or without (12.6 -+ 3.7 nM/min) phospholipid. These data are consistent with sequestration of factor X to phospholipid where it is not cleaved, and efficient cleavage of the substrate in solution, thus confirming the prior report (34). DISCUSSION In order to gain insight into the functional properties of TF, independent of anchoring to the cell surface membrane and the known secondary membrane interactions of its ligand factor VIIa or substrate factor X, we produced a soluble recombinant T F mutant (TF1-219). TFl-219 embodies the complete surface-predicted primary structure of TF and was deleted of membrane spanning and intracellular domains. This was done by introducing a termination codon in the nucleotide sequence corresponding to amino acid 220. The DNA for this coding sequence and its natural leader sequence was cloned into a eukaryotic shuttle vector and stably expressed in mammalian CHO cells. The secreted TF1-219 was isolated from culture supernatant by a single-step immunoaffinity purification. TF1-219 was highly glycosylated and the highest as well as less glycosylated forms of TF1-219 interacted with polyclonal affinity purified antibodies to TF and also with VI1 as demonstrated by ligand blotting. TF1-219 in solution prevented ligand blotting of VI1 to TF, consistent with association with VI1 in solution to prevent interaction with the immobilized native TF. The cell-surface domain of TF embedded in TF1-219 therefore appeared to mediate association with the natural ligand in the absence of phospholipid. We were able to address three specific questions using TF1-219. (i) Does the T F extracellular domain, free or complexed with VII, stably interact with phospholipid? (ii) Is the interaction of VIIa with T F in the absence of phospholipid sufficient to induce and fully support catalytic function of VIIa? (iii) How is the cleavage of factor X by the binary TF1_219.VIIa complex affected by the presence of physically independent phospholipid surfaces?
Whereas association of VIIa with immobilized phospholipid was readily demonstrated, no association of TF1-219 with phospholipid was observed under identical conditions. Addition of VIIa to form a binary TF1-219. VIIa complex and addition of factor X to assemble a ternary complex also did not lead to observable association of TF1-219 to phospholipid. The membrane anchoring of TF by its transmembrane domain thus represents the crucial mechanism to localize the catalytic initiation of the coagulation protease cascade in the twodimensional array on the cell surface. Although the direct binding data and the functional characteristics clearly demonstrate that TF1-219 associates with VIIa, and that VIIa alone associates with phospholipid, VIIa did not mediate relatively stable association of TF1-219 to the phospholipid surface under conditions where it binds as a free molecule. The TF1-219'VIIa complex may be less stable than the TF . VIIa complex which is characterized by a long half-life on cell surfaces (3, 4). A rapid dissociation of VIIa from TFl-21g could account for the lack of association of TF1-219.VIIa with phospholipid. Alternatively, TF1-219.VIIa may dissociate faster from the phospholipid surface compared to free VIIa, or TF1-219.VIIa might not interact with phospholipid due to an eclipse of VIIa binding function when in complex with TF. When considered in the context of the functional importance of the VIIa Gla domain for binding to TF expressing cells and generation of an active binary complex (35), the latter explanation suggests that the VIIa Gla domain might not function by association with charged phospholipid, but rather by association with T F itself. The availability of TF1-21g will be useful to address this hypothesis in a more detailed study under equilibrium conditions.
TF1-219 association with VIIa enables function of the catalytic domain as evidenced by hydrolysis of peptidyl substrate. In comparison to nonionic detergent-solubilized native TF, the truncated TF1-219 formed a catalytically equivalent binary complex with VIIa. These data demonstrate that the proteinprotein interactions of VIIa and the surface domain of T F alone are sufficient for both binding and formation of a catalytically functional protease complex. Therefore, TF1-219 is a valid model for analysis of the TF. VIIa binary complex. The association of VIIa with T F or TF1-219 was sufficient to mediate an enhancement in the proteolytic activation of factor X, extending the observations with the peptidyl substrate to the natural protein substrate. The rate-enhancing effect of T F on the catalytic domain of VIIa is reflected by a 5,550fold increase in the kcat, if compared with free VIIa. There is less than a 2-fold change in the Km(app) between VIIa and TF.
VIIa, suggesting that substrate association with VIIa is not significantly influenced by the binding of VIIa to TF. Previous results have indicated a 10-fold decrease in the Km(epp) for cleavage of factor X upon VIIa complex formation with T F (7, 36), although the Km(app) (13.6 p~) in the study for factor X by free VIIa is consistent with the reported value (11.6 p M ) (36). The higher range of substrate concentration (up to 32 p~) used in our analyses may explain our estimate of a higher Km(app). In addition, we determined the initial rates immediately upon mixing reactants, whereas Bom and Bertina (36) determined rates starting 10 min after substrate addition. They also observed changes in the rate of factor Xa generation between 10 and 20 min after substrate addition with certain TF preparations, suggesting that the catalytic efficiency of the TF. VIIa complex might vary with time after the addition of substrate. Our initial measurements exhibited a linear rate of factor Xa formation during the first 3 min, suggesting that slow changes of the catalytic efficiency did not influence our rate determinations. We are aware of the limitations of the estimates for the Km(app) presented in this study, since the substrate concentrations used for the calculations did not exceed the Km(app) more than 2-fold for the fluid-phase reaction. Much larger quantities of recombinant protein, which are not available to us, would be required to obtain a more reliable estimate. Our study, however, does not support the reported decrease in the Km(app) established with an even lower substrate concentration range.
Native TF in the absence of phospholipid association was more effective than TF1-219 as a cofactor for VIIa cleavage of factor X, although no difference was observed in respect to peptidyl substrate hydrolysis. Minor changes in the extended substrate recognition embodied in secondary protein-protein interactions essential for substrate association and catalysis might account for the observed difference. This may be consistent with the finding that loci on the catalytic domain of factor X, which appear not to be in proximity to the Arg-Ile cleavage site have been implicated in the limited proteolytic activation of factor X by TF. VIIa (37). TF1-219'VIIa-mediated proteolytic activation of factor X associated with phospholipid seemed to occur with a catalytic efficiency similar to TF. VIIa, demonstrating that the alterations in the activator function of TF1_219 are minor and restricted to the fluid-phase reaction. Our data indicate that TF1-219 is not only able to induce requisite conformational transitions for VIIa catalytic triad recognition and cleavage of a peptidyl substrate, but also that protein-protein interactions are sufficient for the effec-tive cleavage of the natural substrate factor X.
To analyze the effect of membrane anchoring of T F and substrate interaction with the phospholipid surface, we compared factor Xa formation by membrane-anchored TF. VIIa complex with the fluid-phase reaction and observed a 230fold decrease in the Km(app), which significantly contributed to the catalytic enhancement. This effect could be attributed to concentrating or uniformly orienting the substrate in vicinity of the proteolytic complex on the phospholipid surface. However, characterization of the cleavage of factor X by soluble TF1-219. VIIa showed a more than 100-fold enhanced reaction rate in the presence of phospholipid. This was attributed to phospholipid association of factor X. In contrast, factor X cleavage by the factor X activating enzyme from Russell's viper venom, which requires calcium ions and the Gla domain of factor X for proteolysis of factor X (38), was markedly attenuated by the addition of phospholipid, resulting in an increase of the Km(app) with no changes in the maximal velocity of the reaction. This supports the hypothesis that conformational alterations of factor X upon phospholipid binding change susceptibility of the Arg-Ile bond to proteolytic activation, presumably by influencing recognition and peptide bond hydrolysis by enzyme. Studies by Krishnaswamy and coworkers (39) demonstrate that initial cleavage of the Arg273-T h F 4 bond in prothrombin by fluid phase-factor Xa resulted in exclusive formation of prethrombin 2 as an intermediate. Factor Xa binding to factor Va in the presence of phospholipid resulted in the preferential initial cleavage of the Ar$22-Ile323 bond in prothrombin and formation of meizothrombin as the intermediate. This sequential reaction order required the Gla domains of prothrombin since Gla-deficient prothrombin yielded prethrombin 2 as an intermediate (40). These data suggest that association of prothrombin with phospholipid via the Gla domain induces a conformational change which is essential for proper substrate presentation to the prothrombinase complex. We propose a similar conformational alteration of factor X upon assembly of its Gla domain with phospholipid resulting in a more efficient cleavage by T F . VIIa.
Our analysis demonstrated that the soluble TFl_219.VIIa complex preferentially cleaved membrane bound factor X. Forman and Nemerson (41) presented evidence that the membrane anchored TF. VIIa complex preferentially cleaves free factor X, based on kinetic evaluation of the effect of interference with membrane association of factor X by prothrombin fragment 1 and the use of differently charged phospholipid vesicles which influence the affinity of factor X binding (42).
We cannot exclude that membrane-anchored T F . VIIa allows more efficient cleavage of factor X associating from the fluid phase, thus marking a difference from the soluble TF1-219.
VIIa complex. However, the prominent decrease in the upon membrane insertion of T F demonstrated here and by others (6, 41, 43-45) suggests to us that phospholipid interaction of factor X may facilitate substrate presentation to the membrane-anchored binary complex.
In this study we used the isolated T F extracellular domain to discriminate between protein-protein interactions versus protein-phospholipid interactions in the assembly of the TF. VIIa activator complex. The slow cleavage of factor X by VIIa was markedly enhanced by protein-protein interactions with the isolated T F extracellular domain. Since the Km(app) revealed only minor changes in the presence of TF, it appears that the catalytic step rather than the substrate assembly is favored by TF. This enhancement could result from a preferential interaction between TF and the catalytically active conformer of VIIa, which occurs with low frequency in the absence of cofactor. In this manner, T F would trap the favor-able conformation of VIIa and utilize the binding energy to retain VIIa in this state rather than actively inducing a conformational transition in VIIa. An additional catalytic enhancement would be achieved by anchoring the TF.VIIa complex to the lipid surface. T F uniformly oriented due to its membrane anchor could align VIIa to assemble with the substrate factor X which, bound to the phospholipid, was shown to be presented in a favorable conformation for cleavage by the soluble complex. The decrease in the Km(app) due to assembly of the binary complex on the phospholipid surface may be critically important for the functioning of the complex under physiological concentrations of the substrate factor X.