Activation of prothrombin by a novel membrane-associated protease. An alternative pathway for thrombin generation independent of the coagulation cascade.

We herein report that a novel membrane-associated protease capable of activating prothrombin is present in several mammalian cells. This protease can directly convert prothrombin to active thrombin and induces blood clotting both in vivo and in vitro but is apparently different from coagulation factor Xa, which has been thought to be the only physiological activator of prothrombin. This protease activity was initially found and was very high in 8C feline kidney fibroblast cells, and we characterized its enzymological features using this cell line. Activity was detected in neither the cytosolic fraction nor the culture medium but found in the membranes and identified on the surface of intact cells. The activation of prothrombin required Ca2+ ions, and the apparent Km value for prothrombin was 0.2 microM. The activity was irreversibly inhibited by exposure to EDTA, but various inhibitors for serine proteases including antithrombin III were without effect. Based on these results, we propose that this novel enzyme, membrane-associated prothrombin activator, catalyzes an alternative pathway for generation of thrombin, which is independent of the blood coagulation cascade, and that the thrombin generated is involved in certain pathological states and/or in activation of cells that are spatially separated from the bloodstream.

tivity was irreversibly inhibited by exposure to EDTA, but various inhibitors for serine proteases including antithrombin 111 were without effect. Based on these results, we propose that this novel enzyme, membraneassociated prothrombin activator, catalyzes an alternative pathway for generation of thrombin, which is independent of the blood coagulation cascade, and that the thrombin generated is involved in certain pathological states and/or in activation of cells that are spatially separated from the bloodstream.
The serine protease thrombin is a multifunctional enzyme that participates in diverse biological processes. It is essential in hemostasis, for both blood coagulation and platelet aggregation. It also regulates vascular tone, wound healing, and inflammatory reactions by stimulating cells in vascular walls and leukocytes (1,2). In addition to these well documented actions, thrombin's effects on cells other than the hematopoietic lineage or constituents of vascular walls, e.g. neuroblasts (3,4) and osteoblasts (5), have been recognized (2). The receptor for thrombin responsible for these multiple actions on cells has recently been cloned, and a unique activation mechanism involving proteolysis and the unmasking of the agonist portion within the receptor molecule has been revealed (6). We are now gaining insight into the activation of cells by thrombin, but the way in which thrombin is delivered to these cells remains obscure.  The zymogen prothrombin is converted to the two-chain active form a-thrombin by limited proteolysis upon activation. This activation process is believed to be solely catalyzed by the enzyme prothrombinase (factor Xa in complex with factor Va, anionic phospholipids, and Ca2+ ions) at the final stage of the complex cascade of blood coagulation and occurs solely in the plasma (for reviews, see Refs. 7-9). It had also been believed that prothrombin is synthesized only in the liver and circulates in the bloodstream. Recent results have shown, however, that it is also expressed in other tissue (10). Therefore, even when a cell is not in contact with blood, it may have prothrombin in the vicinity, but the prothrombin is unable to be converted to thrombin. Note that there is nothing impaired about the ability of these cells to interact with thrombin. A number of cell types that possess thrombin receptor (e.g. fibroblasts and neuronal cells) are spatially separated from the bloodstream, at least under physiological conditions. How do these cells encounter thrombin? Extravasation of thrombin from vessels seems a simple and probable explanation. Alternatively, one can postulate that the receptor is only operative under extraordinary conditions, such as severe tissue injury, or that there exists an as yet unidentified ligand for the receptor. Another attractive hypothesis is the existence of an alternative pathway for prothrombin activation that is independent of the coagulation cascade. We herein report a novel cell-associated protease that efficiently converts prothrombin to active thrombin. We propose that this protease catalyzes the postulated alternative pathway, providing a mechanism for the supply of thrombin to these cells.
EXPERIMENTAL PROCEDURES Materials-The following proteins were prepared according to published methods: human prothrombin and factor X (ll), human a-thrombin (12), human prethrombin-l(13), bovine factor VI1 (141, bovine factor VIIa (15), bovine prothrombin, factors M and X and protein S (16), bovine factor Xa (171, bovine activated protein C (181, and bovine antithrombin I11 (19). All of these proteins were homogeneous as judged by SDS-PAGE.' Prothrombin was treated with p-amidinophenylmethanesulfonyl fluoride prior to use in order to negate the effect of possible contamination by thrombin or factor X a . Note that the a-thrombin to which we refer as the authentic standard and that obtained by incubation of prothrombin with activators is actually a-thrombin(des-1-13) (residues Th?85-Ar220 + Ile32'-G1~579 in human prothrombin). This is because human a-thrombin undergoes rapid autolysis, with removal of the N-terminal 13 residues in the A-chain, and true a-thrombin (resi-  (20). Antisera against factor X were prepared by immunizing rabbits with puritied human or bovine proteins, and IgG fractions were obtained by affinity chromatography on immobilized protein A (Protein A Superose HR 10/2; Pharmacia Biotech, Inc.). The thrombin chromogenic substrate Boc-Val-Pro-Arg-p-nitrodde (VPR-pNA) was a gift from Seikagaku Kogyo ('Ibkyo, Japan). 1,5-Dansyl-Glu-Gly-Arg chloromethyl ketone (Dns-EGRck) and heparin were products of Calbiochem and Sigma, respectively.

32441
CeW Culture and FraCtionution4C feline kidney fibroblast cells (21) and T98G human glioblastoma cells (CRL 1690,American Type Culture Collection) were maintained in 150-mm plastic culture dishes (Corning) in Eagle's minimum essential medium supplemented with 10% heatinactivated fetal calf serum and 60 pg/ml kanamycin in a humidified atmosphere of 95% air and 5% CO, at 37 "C,and cells were routinely reseeded every 3 4 days after trypsinization. Confluent cultures of cells (30 dishes) were rinsed three times with Tris-buffered saline (TBS; 50 rrmd Tris-HC1, 100 rrmd NaCl, pH 8.0), and cells were scraped off. Harvested cells were suspended in 30 ml of 50 m Tris-HC1, pH 8.0, and homogenized with a motor-driven Tenon homogenizer (15 strokes). The resultant cell lysate was first centrifuged at 600 x g for 5 min to remove nuclei and unbroken cells and then at 5,000 x g for 10 min. The supernatant was further centrifuged at 100,000 x g for 60 min. All fractionation procedures were performed at 4 "C. The 5,000 x g and 100,000 x g pellets (rich in plasma membranes and microsomes, respectively) were suspended in "BS. ' Ib assess the secretion of prothrombin activator, conditioned medium was prepared. A confluent culture of 8C cells in a 150-mm dish was rinsed once with Eagle's minimum essential medium without serum, and then 10 ml of fresh Eagle's minimum essential medium was added.
After an overnight incubation, the medium was collected and subjected to the assay for activator.
Detergent-solubilized enzyme was prepared as follows. The plasma membranes suspended in "BS were homogenized in the presence of 0.3% CHAPS, incubated for 30 min on ice, and then centrifuged at 100,000 x g for 60 min at 4 "C. The resultant supernatant was used as the solubilized activator.
Other cells tested were from American Type Culture Collection, and maintained act recommended by ATCC.
Assays for Procoagulant Activities-clotting was performed as follows. "benty pl of human citrated plasma was mixed with 40 pl of the plasma membrane frakion or the CHAPS-solubilized activator (1 mg proteidml) and incubated at 37 "C for 30 8. Then 20 pl of 25 rrmd CaCl, was added, and the time required for clot formation was measured. Tissue factor activity was evaluated by monitoring increases in amidolytic activity of factor VIIa (22, 23) using a chromogenic substrate (S-2288; Kabi) as described previously (15). Activation of factors W and IX was verified by SDS-PAGE after incubation of purified bovine proteins with the membrane. Activation of factor X was evaluated as described for thrombin (see below) except that the chromogenic substrate used was Boc-Leu-Gly-kg-p-nitrodde (Seikagaku Kogyo).
Assay for Thrombin-In typical assay condition, human prothrombin (10 in TBS containing 5 m CaCLJ was incubated with plasma membranes of 8C cells at 37 "C. After an appropriate incubation period, the reaction was terminated by the addition of 1/10 volume of 100 nm EDTA. Thrombin generated was quantXed by measuring ita amidolytic activity; 90 pl of sample (appropriately diluted with TBS containing 1 mg/ml bovine serum albumin) was mixed with 10 pl of 5 rrmd VPR-pNA and incubated at 37 "C in a microtiter plate. The initial rate of pnitmanilhe liberation was monitored at 405 n m with a kinetic plate reader (Well Reader, Seikagaku Kogyo). Activation of prothrombin by intact cells was assayed as follows. Confluent cultures of cells grown in a 96-well culture dish were rinsed three times with TBS containing 5 nm CaCl, and incubated with 80 pl of TBS contajning 10 p human prothrombin and 5 m CaC1, at 37 "C.
The reaction was terminated by the addition of 10 pl of 100 m~ EDTA. Then we added 10 pl of 5 rrmd VPR-pNA, and the amidolytic activity of thrombin was measured as above.
!l'katrnent with Protease Inhibitors-The plasma membrane was incubated with inhibitors for 30 min at ambient temperature. Then it was washed by centrifugatiodresuspension and subjected to the assay for prothrombin activator as described above. In the case of benzamidine, prothrombin was incubated with the membranes in the presence of this compound, with subsequent analysis by SDS-PAGE. Zkatment with Activated Protein C-In order to exclude the effect of possible contamination by factor VNa, we conducted treatment with temperature for 1 h. Then we added diisopropyl fluorophosphate to 10 mra so as to inactivate activated protein C and lea to stand for 1 h. The membranes were then washed and subjected to the assay for prothrombin activator. The same treatment greatly reduced factor V activity expressed on activated platelets, which was employed as the positive control, within a few minutes, as reported by Suzuki et al. (24). Protein Seqwncing-Activated prothrombin was subjected to SDS-PAGE, electroblotted onto a poly(vinylidene difluoride) membrane (Millipore Corp.) and stained with Amido Black (25). The band that comigrated with authentic a-thrombin was cut out and subjected to sequencing in an Applied Biosystems protein sequencer model 473A.
Gel Filtration-The prothrombin activator solubilized with CHAPS was applied to a column of Superdex 200pg equipped with a fast protein liquid chromatography system (Pharmacia) that had been equilibrated with TBS containing 0.3% CHAPS and eluted with the same buffer at a flow rate of 1 d m i n . One-ml fiactions were collected and subjected to the assay of prothrombin activator.
Other MethodsSDS-PAGE was performed by the method of Laemmli (26). Protein concentration was determined with BCA protein assay kit (Pierce).

RESULTS
It was found that a membrane fraction from 8C fibroblasts contained a lethal component when injected intravenously into mice. The 8C-injected mice experienced severe thrombosis? Apparent fibrin deposition in venules was seen in some tissues of injected animals; abundantly in the lung, to some degree in the liver, and less in the kidney (Fig. 1). The lethal component of 8C membranes is presumably a thrombogenic factor. Indeed, the membranes provoked clotting of normal human plasma in vitro. No direct action on fibrinogen was observed, ruling out the presence of thrombin-like substances. We next evaluated the mechanism to induce plasma clotting using various purified coagulation factom. Conversion of prothrombin to thrombin by membranes of 8C cells was clearly observed, as described below, but activation of other coagulation factors, namely fadons MI, M, and X : was not detected (see "Experimental Proce-This phenomenon was first discovered by Dr. 0. Yoshie and colleagues at Shionogi Institute of Medical Science. In the course of developing antibodies directed against cell-specific surface antigens for various cell types, they found that intravenous injection of 8C cells caused rapid death of injected mice. This information was transmitted to us, and we started the present study. Human factor X was not activated at all by the membranes, whereas activation of the bovine protein was observed. This action on bovine dures"). Tissue factor activity was also undetectable. We could not, however, exclude the possibility of the presence of small amounts of tissue factor in the membranes, because of the relatively low sensitivity of the assay. Nevertheless, it appears that the procoagulant activity found in the membranes is in part, if not all, attributable to the activity that directly activates prothrombin. We have focused our attention on the prothrombin activator. Activation of prothrombin was seen with intact, unbroken 8C cells. Prothrombin was efficiently activated, and amidolytic activity toward a chromogenic substrate specific for thrombin developed with time. Under the experimental conditions (see "Experimental Procedures"), 0.7 ndmin production of thrombin was observed. It thus appears that 8C cells constitutively express the enzyme responsible for activation of prothrombin on the surface of their plasma membranes, giving the enzyme access to extracellular substrates. Upon fractionation of 8C cells, the activator was recovered solely in membranes and was not found in the cytosolic fraction. Eighty-five percent of the activity was recovered in the 5,000 x g pellet (plasma membrane fraction) and 15% in the 100,000 x g pellet (microsomal fraction). Secretion of the activator into the medium was not detected even after 20-fold concentration of the conditioned medium. The activity was tightly associated with the membranes; washing with isotonic, hypotonic, or hypertonic buffers or a buffer with EDTA all failed to extract the activator. Extraction was achieved only with detergents, such a s 0.3% CHAPS, suggesting that the activator is a membrane-integrated protein.
factor X was significant but much weaker than that on prothrombin.
Furthermore, it is presently uncertain whether activation of prothrombin and (bovine) factor X is mediated by the same enzyme. We did not pursue activation of factor X any more in the present study.
The time course of activation of human prothrombin by isolated 8C membranes is shown in Fig. 2 4 ; more than 80% of prothrombin was converted to thrombin within 30 min under our experimental conditions. As is shown in Fig. 2B, prothrombin was processed by an enzyme(s) present in the membrane. The derivatives of prothrombin cleaved by the putative 8C protease are indistinguishable from those generated by incubation with factor Xa. The band corresponding to a-thrombin is clearly seen in the nonreducing gel, and in the reducing gel the band corresponding to the B-chain is evident (Fig. 223). The protein that comigrated with the authentic standard was shown to be a-thrombin by amino acid sequence analysis; two phenylthiohydantoin-derivatives, corresponding to the sequence of human prothrombin that starts a t T h P 5 and ne"", were found in each sequencing cycle (see "Experimental Procedures"). This result indicates that correct and selective cleavage of prothrombin has taken place. The thrombin generated had clotting activity similar to that of authentic a-thrombin. Bovine prothrombin was also activated by the membranes, and the apparent K,,, values of the human and bovine proteins were about 0.2 p~, well below the concentration in plasma (-2 PM). When the CHAPS-solubilized activator was subjected to a gel filtration column, the activity was eluted at the position of apparent molecular mass of 65-70 kDa (data not shown).
The activation of prothrombin had an absolute requirement for Ca2' ions, and physiological concentrations of Ca2' ions (in the millimolar range) were necessary (Fig. 3). Calcium ions are probably required by the substrate prothrombin and not by the activator, since prethrombin-1 (a derivative of prothrombin that lacks the major Ca*+-binding site, the N-terminal y-carboxyglutamic acid domain) was found to be a poor substrate (data not shown). We next investigated the effect of various protease inhibitors in order to characterize the nature of this (0.1 mg/ml) in the presence of heparin (5 unitdml) also gave no inhibition. The preparation of membranes was thus apparently devoid of factor Xa. We can also negate the possibility of the existence of factor VNa, which, if present, greatly amplifies the action of trace amounts of factor Xa. The activator was rather stable (it retained activity at least 1 week when stored a t 4 "C), whereas factor V is very labile; exogenous addition of factor Xa resulted in only a slight increment of the ability to generate thrombin; and finally, treatment of the membranes with activated protein C did not reduce the activity. In addition, well established low molecular weight inhibitors for trypsin-like serine proteases, i.e. diisopropyl fluorophosphate (10 mM), phenylmethanesulfonyl fluoride (1 mM), benzamidine (10 m~) , and leupeptin (1 mM), could not attenuate the activity, and pepstatin (1 mM) and iodoacetamide (10 mM) were also without effect. The activator thus appeared not to be a serine, carboxyl, or thiol protease. On the other hand, exposure of the solubilized enzyme to EDTA resulted in irreversible loss of the activity (Fig. 41, presumably by removal of the metal ion required for catalysis. Although unequivocal classification should await purification of the enzyme protein and determination of its structure, the enzyme is, therefore, most likely to be a metalloproteinase. We are in the process of isolating this protein. In order to assess whether expression of the activator is specific to 8C cells or is ubiquitous, we have screened lysates of various cells for prothrombin activator; 8C cells showed prominent activity, and activation of prothrombin was also seen in some cell types of different origins, such as MDCK cells (canine, kidney epithelial) and T98G cells (human, glioblastoma) ( Table  I). This result strongly suggests that the same enzyme is present in various cell types. Some other cells also had activities, whereas some were negative (Table I), suggesting a specificity of expressing cell.
The enzyme expressed in T98G cells was characterized in detail. Enzymological features of the prothrombin activator of T98G were indistinguishable from those found in 8C. The T98G enzyme was solely recovered in membrane fractions and is present on the surface of intact cells; showed the same cleavage pattern of prothrombin; had similar Ca2+-dependence and K,,, value for prothrombin; eluted a t a similar position on gel filtration; and was inhibited by EDTA but not by other protease  Fig. 2 A . Note that the unsolubilized enzyme was more resistant to treatment with EDTA, and prolonged incubation was necessary to attenuate its activity.

Screening for prothrombin activator in lysates of various cells
Cultured cells were washed, suspended in TBS, and lysed by repeated freezindthawing and by subsequent treatment with a Teflon homogenizer. The obtained lysates (1.0 mg of proteidml) were incubated with 10 p~ prothrombin in the presence of 5 mM CaCl, for 30 min at 37"C, and thrombin generated was quantified as described under "Experimental Procedures." Note that no amidolysis of VPR-pNA was observed in each lysate without prothrombin. -b -"Relative specific activity based on the amount of protein in the * Under detection limit (less than 5%).
lysates, expressed as percentages of that found in 8C.
inhibitors above described. The T98G prothrombin activator must be the human counterpart of the 8C enzyme. With this human cell line, we were enabled to use antisera directed to factor X and could get further evidence that eliminated the possibility of involvement of factor Xa in the activation of prothrombin. The T98G membrane was subjected to SDS-PAGE followed by immunoblot analysis with antisera against human and bovine factor X. In this experiment, the lowest detection limit was 20 ng, and the estimated factor Xa contamination in the tested sample was 400 ng, assuming the activation was solely catalyzed by factor Xa (plus Ca2+ ions and phospholipids; participation of factor VNa was unlikely as described above). Neither human nor bovine factor WXa was detected. Taking this result and the pharmacological evidences together, we get firm confidence that the activator we identified is not factor Xa that may be synthesized by these cells or mere contamination from the culture medium but is a novel enzyme completely different from factor Xa. DISCUSSION We have shown that 8C fibroblasts express a n activator of prothrombin on their surface. The action of this enzyme on prothrombin much resembles that of factor Xa, but all of the

Membrane-associated Prothrombin Activator in Mammalian Cells
pharmacological evidence argues against its participation. Furthermore, we could also rule out the possibility of involvement of another constituent of the prothrombinase complex, factor VNa. This enzyme is thus clearly disparate from the "classical" prothrombinase. Recently, Altieri (28,29) has discovered a novel membrane-integral protein, EPR-1, which serves as a cellular receptor for factor Xa and potentiates the activation of prothrombin. However, the identity of EPR-1 and the prothrombin activator herein described is not understood. Expression of EPR-1 is seen in peripheral monocytes, monocytic-myeloid cell lines (e.g. U-937 and HL-601, and some populations of T lymphocytes (281, but we scarcely observed prothrombin activation in cells of hematopoietic lineage including U-937 and HL-60 (Table I). Moreover, EPR-1 is less potent in enhancing the generation of thrombin by factor Xa (29) and cannot explain the efficient activation of prothrombin seen in the present study, in that contamination by factor Xa is, if there is any, very low. We thus conclude that the 8C enzyme is a unique and hitherto novel prothrombin activator. The presence of similar activities to generate thrombin in some other cell lines (Table I) indicates that the activator is not unique to 8C cells and also suggests its involvement in certain physiological andor pathological events in uiuo. It is true, however, that 8C cells express the activator at high levels. We currently do not know why 8C cells are so active in this regard, but it should be pointed out that this continuous cell line has been transformed by Moloney murine sarcoma virus (21). It is possible that disruption of normal gene expression by the viral oncogene results in uncontrolled overproduction of the activator. It is of interest to note that certain malignant tumors express procoagulant activities and that patients in the late stages of cancer frequently manifest a coagulation disorder, disseminated intravascular coagulation (30). Deposition of fibrin on cancer cells is thought be an important strategy for their survival against attacks by the immune system, and participation of coagulation enzymes in metastasis has been suggested (30,311. Most of these procoagulants are attributable t o tissue factor (32), and an enzyme that directly activates factor X has been reported (33). The prothrombin activator we found may also participate in the development of cancer as well as other pathological states, e.g. thrombosis and atherosclerosis.
Although we have no data on the physiological role of this enzyme as yet, its possible involvement in the central nervous system is worth considering. Thrombin has significant effects on neuronal cells, causing inhibition of neurite outgrowth and retraction of extended neurites (3, 4). Thus, generation of thrombin in the brain is now considered crucial for neural differentiation and the maintenance of the integrity of neuronal networks. Disruption of the balance between thrombin and protease nexin-1, a naturally occurring inhibitor of thrombin (341, may be involved in the pathogenesis of neurodegenerative disorders, such as Alzheimer's disease (35). Prothrombin is expressed not only in the liver but also in the brain (10). Furthermore, for stimulation of neuroblastoma cells or primary brain cells in culture, thrombin can be replaced by prothrombin even in the absence of other plasma components (36). These observations indicate that, in brain, there exist both the zymogen and, presumably, a mechanism for its activation independent of the coagulation cascade. The enzyme we identified can provide such a mechanism. Indeed, we herein also showed that a prothrombin activator with the same enzymological features was present in a cell line derived from the central nervous system, i.e. T98G. A search for the enzyme in normal human (or other animal) tissue and determination of its spatial and temporal distribution are now necessary.
In conclusion, we herein showed that a novel protease is present on certain mammalian cells. We propose to tentatively designate this enzyme as membrane-associated prothrombin activator (may be abbreviated as MAPA). A number of "exogenous" activators have been isolated from snake venoms and microorganisms (37), but, to the best of our knowledge, this is the first report of a candidate for another physiological activator specific for prothrombin in addition to factor Xa. This activator should generate thrombin independently of the coagulation cascade, and the thrombin generated would provoke various physiological events that are not restricted to hemostatic processes. Such an alternative pathway would be highly significant for activation of cells, in particular those spatially separated from the bloodstream.