Reconstitution of Rabbit Thrombomodulin into Phospholipid Vesicles

The influence of phospholipid on thrombin-thrombomodulin-catalyzed activation of protein C has been studied by incorporating thrombomodulin into vesicles by dialysis  from  octyl glucoside-phospholipid mixtures. Thrombomodulin was incorporated into vesicles ranging from neutral (100% phosphatidylcholine) to highly  charged (30% phosphatidylserine  and 70% phosphatidylcholine). Thrombomodulin is randomly oriented in vesicles of different phospholipid composition. Incorporation of thrombomodulin into phosphatidylcholine, with or without phosphatidylserine, alters the Ca2+ concentration dependence of protein C activation. Soluble thrombomodulin showed a halfmaximal rate of activation at 580 PM Ca2+, whereas half-maximal rates of activation of liposome-reconstituted thrombomodulin were obtained between 500 PM Ca2+ and 2 mM Ca”, depending on the composition (pr0tein:phospholipid) of the liposomes. The Ca2+ dependence of protein C activation fits a simple hyperbola for the soluble activator, while the Ca2+ dependence of the membrane-associated complex is distinctly sigmoidal with a Hill coefficient greater than 2.4. In contrast, the Ca” dependence of y-carboxyglutamic acid  (Gla)  domainless  protein  C  activation is unchanged by membrane reconstitution ( 1/2,,,.,x = 53 f 10 NM) and fits a simple rectangular hyperbola. Incorporation of thrombomodulin into pure phosphatidylcholine vesicles reduces the K, for protein C from 7.6 2 2 to 0.7 f 0.2 PM. Increasing phosphatidylserine to 20% decreased the K, for protein C further to 0.1 2 0.02 WM. Membrane incorporation has no influence on the activation of protein C from which the Gla residues are removed proteolytically (K,,, = 6.4 2 0.5 PM). The K,,, for protein C observed on endothelial cells is more similar to the K,,, observed when thrombomodulin (TM) is incorporated into pure phosphatidylcholine vesicles than into negatively charged vesicles, suggesting that the protein C-binding site on endothelial cells does not involve negatively charged phospholipids. In support of this concept, we observed that prothrombin and fragment 1, which bind to negatively charged phospholipids, do not inhibit protein C activation on endothelial cells or TM incorporated into phosphati-

The influence of phospholipid on thrombin-thrombomodulin-catalyzed activation of protein C has been studied by incorporating thrombomodulin into vesicles by dialysis from octyl glucoside-phospholipid mixtures. Thrombomodulin was incorporated into vesicles ranging from neutral (100% phosphatidylcholine) to highly charged (30% phosphatidylserine and 70% phosphatidylcholine). Thrombomodulin is randomly oriented in vesicles of different phospholipid composition. Incorporation of thrombomodulin into phosphatidylcholine, with or without phosphatidylserine, alters the Ca2+ concentration dependence of protein C activation. Soluble thrombomodulin showed a halfmaximal rate of activation at 580 PM Ca2+, whereas half-maximal rates of activation of liposome-reconstituted thrombomodulin were obtained between 500 PM Ca2+ and 2 mM Ca", depending on the composition (pr0tein:phospholipid) of the liposomes. The Ca2+ dependence of protein C activation fits a simple hyperbola for the soluble activator, while the Ca2+ dependence of the membrane-associated complex is distinctly sigmoidal with a Hill coefficient greater than 2.4. In contrast, the Ca" dependence of y-carboxyglutamic acid (Gla) domainless protein C activation is un- The K,,, for protein C observed on endothelial cells is more similar to the K,,, observed when thrombomodulin (TM) is incorporated into pure phosphatidylcholine vesicles than into negatively charged vesicles, suggesting that the protein C-binding site on endothelial cells does not involve negatively charged phospholipids. In support of this concept, we observed that prothrombin and fragment 1, which bind to negatively charged phospholipids, do not inhibit protein C activation on endothelial cells or TM incorporated into phosphati-* This work was supported in part by National Institutes of Health Grants ROl HL29807 and R01 HL30340 and by an American Heart Association Established Investigatorship with funds contributed in part by the Oklahoma affiliate. 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. dylcholine vesicles, but do inhibit when TM is incorporated into phosphatidylcho1ine:phosphatidylserine vesicles. These studies suggest that neutral phospholipids lead to exposure of a site, probably on thrombomodulin, capable of recognizing the Gla domain of protein C.
T M has been purified to homogeneity from rabbit lung (7), bovine lung (8), and human placenta (9). T M is extracted from these tissues using detergent and cannot be extracted with high salt or low salt plus EDTA (7). This information, in conjunction with previous studies (1,2) indicating that TM remains bound to the cell surface during protein C activation, suggests that TM is an integral membrane protein. TM, however, retains some biological activity after extraction with nonionic detergent (7)(8)(9). The properties of TM and the availability of significant quantities of purified protein suggested that this might prove a useful system for studying membrane involvement in coagulation.
In recent years, considerable interest has focused on the role of membranes in coagulation, largely because of the central role they play in assembly of the activation complexes (10). In the zymogen activations which lead ultimately to clot formation, a central theme is repeated. The enzyme, a protease, interacts with a regulatory protein on a surface, usually a membrane, to rapidly activate the substrate. In several of the complexes such as prothrombinase, the enzyme, regulatory protein, and substrate all bind reversibly to the membrane and bind effectively only to negatively charged membrane surfaces (10). Binding of two of the components, factor Xa and prothrombin, requires Ca2+ and the other component, factor Va, requires Ca2+ for stability. This makes the analysis of the role of Ca2+ or membrane surface charge on the function of any single component of the complex exceedingly difficult. A system more comparable to TM involves tissue factor. Tissue factor is an integral membrane protein which interacts with the vitamin K-dependent enzyme factor VI1 (VIIa) (11, 12) to activate either factor IX or factor X (13), both of which are also vitamin K-dependent factors. Thus, even in this system, it is difficult to directly determine the role of surface The abbreviations used are: TM, thrombomodulin; PC, l-palmitoyl-2-oleoyl-phosphatidylcholine; PS, l-palmitoyl-Z-oleoyl-phosphatidylserine; TBS, tris-buffered saline, 0.1 M NaC1, 0.02 M Tris-HC1, pH 7.5; HEPES, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid; octyl glucoside, octyl-P-D-glucopyranoside; Gla, y-carboxyglutamic acid. charge since it influences both the enzyme and substrate interaction with the surface. In addition, tissue factor is devoid of significant activity in solution, making comparative studies between the soluble and membrane-bound cofactor impossible.
T M is unique in the coagulation scheme of enzyme-effector activation systems. T M binds an enzyme, thrombin, which does not interact directly with the membrane. Thus, thrombin interaction is not dependent on the membrane surface charge. The thrombin-thrombomodulin complex activates protein C, which is a vitamin K-dependent protein with relatively lower affinity for bilayer membranes (12) than the other vitamin K-dependent proteins. TM, therefore, can be used to delineate the roles of membrane composition and Ca2+ on substrate activation without the added complication of altering the binding of the enzyme or the regulatory protein to the surface. An additional question about expression of T M activity relates to the mechanism by which the endothelial cell surface promotes protein C activation. The endothelial cell surface is considered nonthrombogenic, yet T M on the endothelial cell surface saturates with respect to protein C a t a 10-fold lower concentration (2) than soluble T M (9,14). To determine whether this difference requires a specific membrane-binding site for protein C on the endothelial cell surface or whether the effect could be obtained with synthetic phospholipids, T M was incorporated into liposomes of varying phospholipid composition and the effect on protein C activation were evaluated.
Purification of Thrombomdulin--In order to purify sufficient TM for the phospholipid reconstitution studies, it was necessary to redesign the published TM preparative procedure (7) to increase the amount obtained and decrease the time required. The key step in the modified isolation procedure is the use of an immobilized murine monoclonal antibody to rabbit TM in place of the original thrombin column. The modified procedure is as follows: 80 pairs of frozen lungs driven meat grinder (International Edge Tool Co., Roseland, NJ). from young rabbits were ground while frozen in a commercial, motor- The lungs were ground three times before suspension. The ground lung was suspended in 8 liter of 0.25 M sucrose, 0.02 M Tris-HC1, 5 mM benzamidine-HCl, 0.02% NaN3, pH 7.5. All procedures throughout the preparation were performed at 4 "C. The tissue was collected by centrifugation at 4,800 X g for 30 min. The pellets were resuspended by vigorous mixing with a stirring bar in sucrose buffer and collected by centrifugation as before. This process was repeated two times. The TM was extracted from the washed pellet with 2 liters of wash buffer made 2% in Triton X-100. The extraction was done with a 250-ml glass homogenizer with the pestle attached to a drill press. Insoluble material was removed by centrifugation at 40,000 X g for 40 min. The extracts were quickly run through a 5 X 10-cm column of Sephadex G-50 to remove particulates, then adsorbed by adding 20 g of QAE-Sephadex Q-50 pre-equilibrated in 2 liters of 0.4 M NaCl, 0.02 M Tris-HC1, 5 mM benzamidine-HCl, pH 7.5. The remaining steps were as follows: 1) adsorption was for 40 min with stirring; 2) the Sephadex was allowed to settle for 30 min; 3) the supernatant was decanted; 4) the Sephadex was packed into a 5 X 30-cm column and washed with 1 liter of 0.2 M NaCl, 0.02 M Tris-HC1, 5 mM benzamidine-HC1, 0.5% Lubrol PX, pH 7.5; 5) the QAE-Sephadex (-300 ml) was removed from the column and mixed with 300 ml of 1 M NaCl in the above buffer, then mixed vigorously with a magnetic stirring bar for 30 min; 6); the Sephadex was then packed back into the column and the effluent containing the TM activity was collected. The column was eluted with an additional 300 ml of 1 M NaCl in the above buffer. Unlike the extract, the QAE-Sephadex eluate was a clear solution suitable for use with the monoclonal antibody column. The eluate (~7 0 0 ml) was adsorbed in a Teflon beaker with a monoclonal antibody to rabbit TM (70 ml coupled to Affi-Gel-10 (Bio-Rad) at a final concentration of 5 mg/ml antibody) for 2 h with gentle stirring; the gel was allowed to settle for 30 min, the supernatant siphoned off, and the beads packed into a 2.5 X 20-cm column. The column was washed sequentially with 4 liters of 1.0 M NaC1,0.02 M Tris-HC1, 5 mM benzamidine-HCl, 0.5% Lubrol PX, 0.02% NaN3, pH 7.5, at a flow rate of 250 ml/h, 200 ml of this buffer without benzamidine-HC1, and 50 ml of 1 M NaCl, 0.5 M guanidine-HC1,0.02 M Tris-HC1, 0.5% Lubrol PX, pH 7.5. The TM activity was eluted with 1 M NaCl, 1.5 M guanidine-HC1, 0.5% Lubrol PX, 0.02 M Tris-HC1, pH 7.5. The flow rate on the last two steps was reduced to approximately 40 ml/h. TM was immediately desalted on a Sephadex G-75 column into 0.1 M NaCl, 0.02 M Tris-HC1, 0.1% Lubrol PX, pH 7.5.
TM was further purified by chromatography on a diisopropylphosphoro-thrombin Affi-Gel 10 column (0.6 X 30 cm) equilibrated in 0.2 M NaCl, 0.02 M Tris-HC1, 0.1% Lubrol PX, 0.5 mM Ca2+. The TM was made 0.5 mM in Ca2+, applied to the column, and the column was washed with 15 ml of equilibration buffer, then with 15 ml of 0.4 M NaCl, 0.02 M Tris-HC1, 0.1% Lubrol PX, 0.1 mM EDTA before eluting the TM with 2 M NaCl, 0.02 M Tris-HC1, 0.1% Lubrol PX, 0.1 mM EDTA, pH 7.5. Traces of degradation products were removed and the Lubrol PX concentration reduced to 0.02% by chromatography on a Sephadex G-200 column (1.5 X 60 cm) equilibrated in 0.1 M NaCl, 0.02 M Tris-HC1, 0.02% Lubrol PX, pH 7.5, at room temperature. A total of 15 mg (5.5 ml, 2.8 mg/ml) was applied with the TM activity eluting as a clear solution in the void volume of the column. The yield from a single preparation was 8-12 mg.
Bovine TM was purified by modification of the publishedprocedure (8). After cold methanol precipitation, the supernatant was batchadsorbed with 800 ml of DEAE-Sepharose previously equilibrated in 0.075 M NaC1, 0.02 M Tris-HC1, 0.5% Lubrol PX, 25% MeOH, pH 7.5, for 2 h at 4 "C. The gel was washed with 6 liters of equilibration in a Buchner funnel, packed into a 5 X 40-cm column and step-eluted with 2 liters of 0.75 M NaCl, 0.02 M Tris-HC1, 0.5% Lubrol PX, 25% MeOH, pH 7.5. Protein-containing fractions were pooled and dialyzed against 6.5 volumes of 0.02 M Tris-HC1, 0.5% Lubrol-PX, pH 7.5, for 6 h. The sample was then batch-adsorbed with 150 ml of diisopropylphosphoro/thrombin/agarose (BioGel A-15m, Bio-Rad), washed overnight with 0.1 M NaCl, 0.02 M Tris-HC1, 0.02% Lubrol PX, pH 7.5, and step-eluted with 1 M NaCl, 0.02 M Tris-HC1, 0.02% Lubrol PX, pH 7.5. Fractions containing activity were pooled and dialyzed to 0.1 M NaCl, 0.02 M Tris-HC1, 0.02% Lubrol PX, pH 7.5. If concentration was necessary, this material was batch-adsorbed to a small amount (30-50 ml) of DEAE-Sepharose and step-eluted as before. The TM was then applied to a Mono Q HR5/5 column attached to a fast protein liquid chromatography system and developed with a 20-ml linear gradient from 0.1 to 1.4 M NaCl in 0.02 M Tris-HC1, 0.02% Lubrol PX, pH 7.5. The gradient was held at 0.4 M NaCl until 280 nm of adsorbing material fell to zero. Protein eluting at 0.6 M NaCl and above contained the TM activity and was homogeneous as judged by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (18).
Radiolabeling of Thrombomodulin-Enzymobeads (Bio-Rad) and Na"'I (ICN), were used to radiolabel TM according to the package protocol for 15 min at room temperature. After labeling, the free iodine was removed by extensive dialysis at 4 "C against TBS plus 0.02% Lubrol PX. The specific activity of the TM was 0.7 pCi/pg. The labeled TM was stored at 4 "C.
Phospholipid Preparatwn-The purity of the phospholipids was confirmed by thin layer chromatography as described by Hauser (20).
Aliquots of the PC and the PS in chloroform were transferred to borosilicate glass tubes and dried to a thin film on the wall of the tube under a stream of Nz. The phospholipid film was placed under vacuum overnight to remove residual chloroform. The phospholipids were then flushed with Nz, sealed tightly with Teflon, and stored dessicated at 4 "C.
Reconstitution of Thrombomodulin into Phospholipid Vesicle-The conditions used to incorporate purified TM into phospholipid vesicles were derived from Mimms et al. (21). All procedures were carried out at room temperature, well above the transition temperature of the phospholipids. Octyl glucoside dissolved in TBS was added to the thin film of phospholipids at a 15-fold molar excess. The detergent phospholipid mixture was mixed and allowed to completely solubilize before mixing the PC and PS. TM in TBS containing 0.02% Lubrol PX was added, producing a final concentration of Lubrol PX of ~0.001%. The initial molar ratio of phospholipid to protein was calculated to be 75,600:l (assuming M, = 787 for the phospholipids) unless stated otherwise. In some experiments, tracer amounts of lZ5I-TM and ['4C]phosphatidylcholine were added to the protein/phospholipid/detergent mixtures to quantify the protein and phospholipid in the final sample. The samples (0.2 ml of each mixture) were dialyzed against 2-3 changes, 1 liter each, of TBS at room temperature for 72 h. The phospho1ipid:protein mixtures became turbid as the detergent was removed.
Bovine TM was incorporated into liposomes by the same procedures. However, to allow direct comparison with kinetic parameters obtained using endothelial cells on tissue culture, the liposomes were formed and centrifuged in HEPES-buffered Hanks' salt solution (HyClone).
Sucrose Density Centrifugation-The reconstituted TM (0.2 ml) was layered on top of a 30-ml 5-30% discontinuous sucrose gradient prepared in TBS and centrifuged in an SW 28 rotor (Beckman) for 12-16 h at 131,000 X g. In some preparations, the samples were incorporated into the 25% sucrose layer of the gradient before centrifugation. No difference was found in the distribution of TM or phospholipid after centrifugation between the two methods (data not shown). The sucrose gradient was fractionated into 1-ml aliquots and T -T M and ["CIPC counted. The '=I-TM was counted on a Nuclear Enterprises NE 1600 instrument (Edinburgh, Scotland, UK). The ["C]phosphatidylcholine was counted on a Beckman LS-1OOC counter. Samples were prepared for scintillation counting by the addition of 150 pl of 0.1% Triton X-100 to eliminate the quenching effect of the sucrose before the addition of 2 ml of Aquasol-2. After at least 2 h in the dark, the carbon-14 counts were measured and corrected for '''I count spillover. The phospholipid content of each sample was calculated based on the specific activity of the starting protein:phospholipid mixture.
Assay of Thrombomodulin Activity-The assay for activation of protein C or Gla-domainless protein C by purified soluble or reconstituted TM was performed as previously described (14). Unless otherwise specified, all activations were performed in TBS containing 0.1% gelatin and 5 mM Ca2+ at 37 "C. All proteins were dialyzed against TBS before use. In experiments where the dependence on Ca2+ was assessed, all proteins were dialyzed against Chelex-containing buffer (Bio-Rad) before use, and the buffer was also dialyzed against Chelex. Protein C, Gla-domainless protein C, thrombin, and solubilized TM were added at the concentrations indicated in the figure legends.
Culture of Bovine Endothelial Cells-Bovine aortic and carotid artery endothelial cells were isolated from young calves as described by Schwartz (22) and grown in minimum essential medium with 50 units/ml penicillin-streptomycin, 10% fetal bovine serum (HyClone), and 1% L-glutamine. Experimental cultures were set up in 96-well flat-bottom tissue culture cluster dishes (Costar). The cultures were incubated at 37 "C in 5% CO, atmosphere in air. Confluent monolayers were washed three times with serum-free medium (D-Val minimum essential medium, 50 units/ml penicillin-streptomycin, 1% Lglutamine, 0.02 mg/ml transferrin, 0.01 mg/ml insulin, and 5 mg/ml bovine serum albumin) and assayed for protein C activation as reported by Owen and Esmon (2), except that Spectrozyme PCa C.
(American Diagnostica) was used to quantitate the activated protein Thrombin Binding to TM-reconstituted Liposomes and Correlation with Biological Activity-Liposomes, with or without TM incorporated, were mixed with thrombin and sucrose to yield 50 p1 of 30 nM thrombin and 20% sucrose in TBS in the bottom of an Airfuge (Beckman) tube. After 5 min at room temperature, each tube was carefully overlayed with 1) 75 pl of 10 nM thrombin in 10% sucrose, TBS; 2) 40 pl of 5% sucrose, TBS; and 3) 20 pl TBS. After centrifuging the tubes in an Airfuge (Beckman) at 149,000 X g for 1 h, the top 40pl liposome-rich fraction was removed. Thrombin bound to the liposomes was determined by incubating 2 pl of the fraction with 30 pl of 0.5 mM H-D-Phe-pip-ecolyl-Arg-p-nitroanilide (S2238, Kabi) in TBS, 0.1% gelatin at 37 "C for 10 min. The reaction was stopped by the addition of 1.25% acetic acid. The optical density at 405 nm was determined and converted to thrombin concentration by reference to a standard curve run simultaneously.
The ability of the liposome-incorporated TM to activate Gladomainless protein C was determined by incubating an appropriate dilution of the liposome fraction with 1 p~ Gla-domainless protein C in TBS, 0.1% gelatin, 5 mM Ca2+ (20 p1 of total volume) for 2 min at room temperature. To determine if any thrombin had dissociated from the TM due to dilution in the assay, parallel incubations containing an additional 10 nM thrombin were included. Nonspecific thrombin floated up by liposomes not containing TM was determined by incubating the liposomes with 15 nM soluble TM, 1 p~ Gladomainless protein C in the same buffer. All reactions were stopped by the addition of 20 p1 antithrombin I11 (1.7 mg/ml), 8 mM EDTA in 0.1 M NaCl, 0.07 M Tris, pH 7.5. Concentration of thrombin.TM complex in the assay was determined by relating the rates obtained to that of 5 nM thrombin, 15 nM soluble TM under the same reaction conditions.

Reconstitution of Thrombomodulin into Phospholipid Vesi-
cles-TM was reconstituted into phospholipid vesicles ranging in composition from 100% PC to 70% PC, 30% PS as described under "Experimental Procedures." To determine the relationship between the concentration of TM added and the incorporation, TM concentration was varied (5 pg/ml-1.0 mg/ml) at a constant concentration of phospholipid (23 mg/ ml). To assess incorporation, '251-TM was added as a tracer. Sucrose density gradients were utilized (see "Experimental Procedures," Methods) to separate the incorporated from the free TM. TM, determined by both Iz5I-TM incorporation and functional activity, was separated into two regions. Functional TM activity was measured using Gla-domainless protein C as the substrate to avoid overestimation of the TM concentration in the liposomes (see later sections). The top 5 ml of the gradient were rich in phospholipid and TM, while the TM that sedimented was devoid of phospholipid. As a control, TM was sedimented in the absence of phospholipid and resulted in <1% of the added TM being found in the top 5 ml. All of the detectable activity sedimented near the bottom of the gradient. The TM activity present in the top 5 ml of the gradient (16% of the total volume) was considered bound and plotted as a function of TM added (Fig. 1). TM incorporation increased as a nearly linear function of added TM up to 1 mg/ml TM. The total TM activity incorporated and the lZ5I-TM incorporated were in good agreement only if the TM activity was assayed after solubilization of the liposomes with Lubrol PX. Without detergent, the '''I incorporated was 2-2.5 times as high as the functional activity in the case of the PC vesicles and 3-3.5 times higher in the case of the PC/PS vesicles (Fig. 1). These studies imply that at least half of the TM is oriented toward the inside of the liposome. It is clear that the extent of incorporation or expression was not influenced to a substantial extent by the surface charge of the phospholipids employed. At the highest TM concentrations employed, 1 TM molecule was incorporated per 9,000 phospholipid molecules based on functional data and 1 per 10,000 based on lZ5I-TM incorporation.

Estimation of Surface Expression of TM on the Phospholipid
Vesicles-TM incorporation into membranes could alter the activity by at least three different mechanisms: 1) altering the  TM was incorporated into liposomes as described under "Experimental Procedures" at a ratio of 1 mg of TM, 23 mg of phospholipid.
Assays were performed in the presence of 0.1% Lubrol to disrupt the liposomes.
Kc.,, 2) altering K, for protein C, and 3) altering the Kd for thrombin. To estimate the influence of TM incorporation on Kc,,, we attempted to measure the TM concentration on the surface of the liposomes. From previous studies, it is clear that TM binds thrombin with very high affinity (51 nM) (2, 7, 23, 24). Since TM concentrations could be obtained that were significantly higher than the Kd, we mixed TM-containing vesicles with excess thrombin to saturate surface TM, and the TM containing vesicles were then separated from free thrombin by flotation through a sucrose density gradient as described under "Experimental Procedures." Three types of assays were performed: 1) thrombin determination by chromogenic substrate assay; 2) Gla-domainless protein C activation rate; and 3) protein C activation rate (Table I). The TM concentration based on thrombin binding and activity on Gladomainless protein C correlated well for both the PC vesicles and the PC/PS vesicles. In contrast, the rate of protein C activation was substantially greater than either of the other values. These results suggest the Gla-domainless protein C can be used to assess surface expression of TM and that the protein C activation rate overestimates TM concentration.

Influence of Membrane Insertion and Surface
Charge on Protein C Activation-To examine the mechanism by which membrane incorporation enhances the activation of protein C, we incorporated TM into vesicles ranging in composition from 100% PC to (7030) PC/PS and studied the influence of protein C concentration on the rate of activation (Fig. 2.4). Surprisingly, incorporation into 100% PC vesicles enhanced protein C activation by decreasing the K,. Increasing the phosphatidylserine content decreased the K , somewhat further. PS content did not influence the concentration dependence of Gla-domainless protein C activation (Fig. 2.4, inset). The influence of phospholipid composition on protein C activation is shown in Table 11. These results reflect pooled data from three separate experiments and were obtained by linear least squares analysis of the transformed data. One of the features we observed in attempts to measure the K , for protein C in the systems that included phospholipid vesicles (especially negatively charged vesicles) was that above 1 ~L M protein C, significant deviation from simple Michaelis-Menton kinetics occurred. This appeared to correspond to the presence of a much lower affinity component (K, = 5 p~) .
This effect was not observed if Gla-domainless protein C was used as substrate. Because the concentrations are far above physiological and because this could be caused by many factors ( i e . substrate dimerization, vesicle fusion, etc.), these values were not included in the above data analysis and no attempt was made to treat the data as from a two-site model. As can be seen from Table 11, incorporation of TM into PC vesicles decreases the K, for protein C at least 10-fold, while incorporation into PC/PS (80:20) vesicles decreases the K, an additional 3-to 7-fold. Incorporation into vesicles had  little or no effect on the Kcst of activation. T o eliminate the possibility that these results were unique to rabbit TM, or due to the use of proteins of heterologous species, the experiments were repeated with bovine T M (Fig. 2B) and virtually identical results were obtained. As indicated by the dashed line, values obtained with bovine endothelial cells in culture more closely approximate those of the TM in 100% PC vesicles than PS containing vesicles. The influence of the incorporation of TM into PC vesicles was surprising since vitamin K-dependent proteins do not bind significantly to these surfaces. Based on thin layer chromatography, it is unlikely the response results from contaminating negatively charged phospholipids.
Inhibition of Protein C Activation by Prothrombin and Prothrombin Fragment 1"Under physiological conditions, protein C circulates at <5% the concentration of prothrombin. Hence, unless protein C has a specialized site for interaction with membrane surfaces, prothrombin would effectively inhibit activation. To determine if the nature of the functional sites were different on the PC vesicles uersus the PC/PS vesicles, the influence of prothrombin and prothrombin fragment 1 was investigated. Fragment 1 inhibited protein C activation when T M was incorporated into PC/PS (70:30) vesicles with 50% inhibition occurring at 2.9 p~ fragment 1 (Fig. 3A). With either soluble T M or T M incorporated into 100% PC vesicles, fragment 1 inhibition was not detectable.
Prothrombin was also studied as an inhibitor and gave results similar to those obtained with fragment 1 (Fig. 3B). Prothrombin demonstrated slight but reproducible enhancement of protein C activation with soluble TM. As with fragment I, there was very little inhibition of protein C activation when T M was incorporated into phosphatidylcholine vesicles.

When T M was incorporated into PC/PS
(70:30) vesicles, prothrombin inhibited even more effectively than prothrombin fragment 1, with 50% inhibition occurring at 0.7 p~ prothrombin.
This allowed examination of the nature of the protein C activation site on endothelial cells. Endothelial cell surface catalysis of protein C activation is very similar to that observed with 100% PC vesicles (& = 0.7 WM) for the cell surface). To determine if the cell surface catalytic site is more similar to the PC vesicles than the PC/PS vesicles, protein C activation was studied over bovine endothelial cells in culture in the presence and absence of prothrombin or fragment 1. Inhibition of thrombin-dependent protein C activation by prothrombin (Fig. 3B), or fragment 1 (Fig. 3A) was minimal, suggesting that the endothelial cell surface site is functionally more similar to the PC vesicle than to the PC/PS vesicle.
Gla-domainless protein C was varied from 0.5 to 8 pM.
c . I

Ca2+
Dependence of Protein C Activation-The previous sections indicate that incorporation of TM into either neutral or negatively charged liposomes enhances protein C activation by decreasing the K,,, for protein C, but has no influence on T M activation of Gla-domainless protein C. Ca2+ is required for activation of either protein C or Gla-domainless protein C by the thrombin-TM complex. In addition, optimal binding of protein C to phospholipid vesicles requires Ca2+. If the interaction between protein C and the site exposed upon reconstitution into PC vesicles also required Ca", the Ca2+ concentration dependence of the activation with soluble T M and membrane-incorporated T M might differ. Comparison of the Ca2+ dependence of protein C activation with the soluble versus the membrane-incorporated form of T M did reveal substantial differences in the Ca2+ dependence (Fig. 4). With purified TM, the dependence was characterized by a simple hyperbolic relationship between Ca2+ concentration and initial rate of protein C activation. With T M incorporated into PC vesicles, the Ca2+ dependence became distinctly sigmoidal. When these data are plotted using the Hill equation (  liposomes. The basis for this variation is not known at this time. Since the Hill coefficient is greater than 2 for all cases where T M is incorporated into phospholipid vesicles, a specific phospholipid composition is not responsible for the high Hill coefficient. The activation of Gla-domainless protein C by thrombin. T M is also dependent on calcium. However, examination of Table I11 shows that half-maximal rates of activation for Gladomainless protein C are unchanged whether T M is soluble or incorporated into phospholipids. The activation curves (not shown) fit a simple hyperbola with no systematic deviation.

DISCUSSION
T M incorporation into neutral phospholipid vesicles results in substantially enhanced protein C activation due primarily to an approximate 10-to 20-fold decrease in K,. When negatively charged phospholipid is employed, the K , decreases further. Several observations suggest that it is unlikely that these negatively charged phospholipids play an important role in protein C activation by the thrombin.TM complex. First, the additional decrease in K,,, observed with the negatively charged phospholipids is almost totally reversed by the presence of physiological concentrations of prothrombin (=2 p~) .
Second, whereas protein C activation by the vesicle system containing negatively charged phospholipids is inhibited by prothrombin, endothelial cell surface activation is not. Third, the K,,, for protein C obtained with T M incorporated into PC vesicles is very similar to that obtained on the endothelial cell surface. Thus, the site responsible for protein C recognition that forms upon incorporation into PC vesicles is specific for protein C and shares many properties in common with the endothelial cell surface. This phospholipiddependent effect requires the presence of the Gla domain in protein C since no influence on the activation of Gla-domainless protein C is observed. These general characteristics of T M incorporated into PC vesicles are very similar for either rabbit or bovine TM. Since both the species and the preparative methods are very different, it is unlikely that the observed effects are due to a species specificicy or to a loosely bound contaminant in the preparation. It should be noted that the rabbit TM was washed very extensively with 1 M NaCl containing buffer in the presence of high levels of detergent, and briefly with low levels of guanidine before elution with 1.5 M guanidine-HC1, further supporting the concept that the observed effect of phospholipid incorporation is dependent on a structural component of TM.
In many respects the observed influence of T M incorporation into phospholipids suggests the formation of a substratebinding site. Support for this concept also can be derived from studies of the Ca2+ dependence. When protein C activation is studied with solubilized TM, the reaction is Ca2+-dependent but not dependent on the presence of the Gla domain (14). The Ca2+ dependence of the activation rate is represented by a simple hyperbolic function. In contrast, when T M is incorporated into vesicles, the Ca2+ dependence is sigmoidal, suggesting positive cooperativity of a t least two sites. This characteristic is dependent on the Gla domain since both the Ca2+dependence and the activation kinetics of the substrate were unaffected if this region was removed from the substrate proteolytically. In our view it is probable, but certainly not demonstrated unambiguously, that TM incorporation into vesicles results in expression of a site(s) on the TM molecule which interacts directly with the Gla domain of protein C in a Ca2+-dependent process. Further support for this concept is presented in the accompanying paper.
Although this study does not directly address the question of the mechanism of TM-phospholipid interaction, it is likely that this interaction involves hydrophobic interactions with the phospholipid. This is supported by the observation that T M incorporated equally well into phospholipids that had no net charge (PC) or a high negative surface charge (PC/PS, 70:30). These findings are compatible with our earlier observations than nonionic detergents are required to extract TM from cell cultures or tissues. The recent demonstration of circulating soluble T M (26) is probably explained by limited proteolysis of the surface-bound molecule (26, 271, and not a reversible equilibrium between the bound and free state. In a recent report by Freyssinet et al. (28), human T M in Triton X-100 was added to sonicated liposomes and the influence on protein C activation studied. They observed a 3.2-fold increase in activity which they attributed to a VmaX effect with no observed change in K,. In their study this effect was dependent on negatively charged phospholipids. Both their quantitative and qualitative results are very different from those reported here. Several differences may explain these discrepancies. No direct interaction of T M with the vesicles was demonstrated, the method of incorporation was very different, and human T M was utilized. In our experience, it is difficult to prevent generation of altered forms of human T M during isolation. The kinetics of altered forms of T M may be affected differently by the presence of a membrane surface. It is interesting that Freyssinet et al. (28) also observed that the Gla domain was essential for activation, suggesting that membrane interaction with the substrate is an essential component of the activation process.
The nature of the substrate recognition site exposed upon membrane reconstitution is uncertain. Recently, Bourin et al. (29) proposed that TM has an acidic heparin-like domain tightly associated (perhaps covalently) with the T M molecule. They felt that the ability of TM to inhibit thrombin clotting activity was mediated through this domain. Alternatively, such an acidic region could constitute the substrate-binding site. Membrane reconstitution could play a role in properly orienting this site for interaction with the substrate. Regardless of the nature of the site, it is apparent that membrane reconstitution alters the properties of the TM substantially and that moderately high affinity interaction with the substrate can occur even in phospholipids lacking a net neutral charge. These properties provide a unique mechanism which allows formation of the anticoagulant enzyme, activated protein C, on the surface of the endothelium without requiring the presence of membrane properties capable of supporting coagulation (10).