Isolation and Characterization of Carinactivase, a Novel Prothrombin Activator in Echis carinatus Venom with a Unique Catalytic Mechanism*

The venom of the viper Echis carinatus contains a metalloprotease, ecarin, that is a potent prothrombin activator. We here show that the venom is also rich in another prothrombin activator, which does not belong to any known category of prothrombin activators. The novel enzyme, designated carinactivase-1 (CA-1), consists of two subunits held together non-covalently but very tightly. One subunit is a 62-kDa polypeptide that has metalloprotease activity and is homologous to the single-chain enzyme ecarin; the other subunit of 25 kDa consists of two disulfide-linked polypeptides of 17 and 14 kDa, and this subunit resembles the anticoagulant in the habu snake venom, IX/X-bp, that specifically binds the Gla domains of coagulation factors IX and X in a Ca 2 (cid:49) -dependent fashion. The activation of prothrombin by CA-1 requires Ca 2 (cid:49) ions at millimolar concentrations and in the absence of Ca 2 (cid:49) ions this enzyme is virtually inactive. By contrast, activation by ecarin is completely independent of Ca 2 (cid:49) ions. CA-1, unlike ecarin, does not activate prothrombin derivatives, in which binding of Ca 2 (cid:49) ions has been perturbed, namely prethrombin-1 and acarboxyprothrombin. Furthermore, the isolated catalytic subunit, although its activity is greatly reduced as compared to that of the holoenzyme, no longer requires Ca 2 (cid:49) ions for the activation of prothrombin. 25-kDa subunit level

Compounds that affect the mammalian blood coagulation system, in particular those that cause acute thrombosis, are often the major active principals of the lethal toxins in viper venoms (1,2). Thrombogenic components in these venoms exhibit considerable heterogeneity in terms of function as well as of structure. Many types of protease that convert quiescent clotting proenzymes to their active forms (or inactive procofactors to active cofactors) are known, and various prothrombin activators have been reported (2). To date, three types of prothrombin activator have been identified in venoms (3): group 1 enzymes, which are metalloproteases whose actions on prothrombin are independent of any plasma or exogenous cofactors; group 2 enzymes, which are Gla-containing, factor Xa-like serine proteases that require factor Va, anionic phospholipids and Ca 2ϩ ions, resembling in this respect the physiological activator factor Xa; and group 3 enzymes, which are hybrid proteins that consist of factor Xa-like catalytic subunits and factor Va-like regulatory subunits and require phospholipids and Ca 2ϩ ions for their action. The group 1 enzymes are widely distributed in venoms of many kinds of viper, e.g. genera Echis and Bothrops, and they are presumably the most toxic since they are resistant to the natural coagulation inhibitors (serpins) present in mammalian plasma, such as antithrombin-III. Another difference between metalloprotease-type prothrombin activators and the physiological activator factor Xa or the venom serine proteases involves the cleavage sites in the prothrombin molecule. The metallo-enzymes cleave only the bond between the A chain and the B chain (Arg 320 -Ile 321 in human prothrombin; Arg 323 -Ile 324 in the bovine protein) with the resultant production of meizothrombin, which is ultimately converted to ␣-thrombin by autolysis (4). The serine-type enzymes cleave one additional site (the junction between fragment 2 and the A chain; Arg 271 -Thr 272 in the human protein) to produce ␣-thrombin directly (5).
The venom of Echis carinatus contains a high level of a metallo-type prothrombin activator and has been widely used in laboratory studies as a convenient tool for the production of thrombin from prothrombin. The enzyme in E. carinatus venom, named ecarin, is a single-chain protein of 55 kDa that exhibits very strict substrate specificity. Prothrombin is the only protein that is cleaved by ecarin in plasma, and other structurally related coagulation factors, e.g. factors IX and X, are scarcely affected (6). The primary structure of ecarin was recently determined by molecular cloning (7). The mature protein consists of three independent motifs. From N to C terminus, there is a metalloprotease catalytic domain of approximately 200 amino acid residues, a disintegrin-like domain of approximately 90 residues, and a Cys-rich domain of approximately 120 residues. At present, however, the roles of noncatalytic regions remain unclear as to the factors that determine the strict specificity of this enzyme, and further information, in particular those about the three-dimensional structure, is necessary to clarify these issues. A significant number of proteins with the same domain organization and with unique respective functions has recently been identified in mammalian tissues (see Ref. 8, and references therein). Thus, this venom protein should serve as a good model in efforts aimed at an understanding of the biochemistry of these mammalian proteins as well as of details of the evolution of these proteins.
* 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.
During the purification of ecarin, we found a novel prothrombin activator in the same venom preparation, which could not be assigned to any of the above mentioned categories. This enzyme, designated carinactivase-1 (CA-1), 1,2 is strongly dependent on Ca 2ϩ ions for the activation of prothrombin, in sharp contrast to ecarin, whose action is unaffected by exogenous Ca 2ϩ ions. In the present report, we describe the purification of CA-1 and discuss the relationship between its structure and function and the unique mechanism by which it activates prothrombin.

EXPERIMENTAL PROCEDURES
Materials-The venom of Malian E. carinatus leucogaster used for the isolation of CA-1 and ecarin was obtained from Latoxan (Rosans, France). Gels for chromatography were from Pharmacia Biotech Inc. Bovine serum albumin (essentially fatty acid-free, ELISA grade) and phospholipids were from Sigma. Lyophilized pooled plasma from normal subjects and from those who had been taking an oral anticoagulant for long periods were obtained from Nycomed Pharma AS, Oslo, Norway. Antisera against IX/X-bp and ecarin were obtained by immunizing rabbits with the purified proteins.
Isolation of CA-1 and Ecarin-A typical isolation procedure was as follows. One hundred milligrams of lyophilized E. carinatus leucogaster venom were dissolved in 2 ml of 50 mM Tris-HCl, pH 8.0, and insoluble materials were removed by centrifugation. The supernatant was subjected to gel filtration on Superdex 200pg with elution with the same buffer. One-ml fractions were collected, and prothrombin-activating activities were assayed as described below. The active fractions were pooled and subjected to a column of Blue Sepharose CL-6B (1.0 ϫ 20 cm) preequilibrated with the same buffer. After washing with the initial buffer (approximately 1 column volume), the bound materials were eluted with a linear gradient of NaCl in the same buffer (0 -1.0 M; 100 ml each). CA-1 was recovered in the unbound fraction, and ecarin was eluted in 0.5 M NaCl. The fractions with the two peaks of activity were separately pooled and then subjected to a column of Q-Sepharose high performance (1.6 ϫ 20 cm) preequilibrated with the same buffer without NaCl. The column was developed with a linear gradient of NaCl in the buffer (0 -0.4 M; 75 ml each). CA-1 was eluted in 0.25 M NaCl, and ecarin was eluted in 0.2 M NaCl. At this step, proteins of considerable homogeneity were obtained. All chromatography was performed at 4°C with a fast protein liquid chromatography system (Pharmacia). The proteins obtained were stored at either 4°C or Ϫ80°C. The enzymes are stable for ϳ2 weeks at 4°C and for months at Ϫ80°C, but repeated freezing/thawing greatly reduced their activities.
Separation of the CA-1 Subunits-To a solution of purified CA-1, guanidine HCl was added to 4 M. This mixture was subjected to gel filtration on Superdex 200pg preequilibrated with 50 mM Tris-HCl, pH 8.0, plus 4 M guanidine HCl. Two peaks were obtained; the first peak contained a 62-kDa component, and the second one contained a 25-kDa component. Peak fractions were pooled separately and dialyzed against an appropriate buffer without denaturant.
Protein Sequencing-Purified CA-1 was subjected to SDS-PAGE under reducing condition and electroblotted onto a poly(vinylidene difluoride) membrane (Millipore) by the method of Hirano (16). The membrane was treated with 20 mg/ml dithiothreitol in 4 M urea at 50°C for 2 h and then with 100 mg/ml iodoacetamide at room temperature for 30 min. The bands of reduced, S-acetamidomethylated polypeptides were visualized by Amido Black staining, cut out, and subjected to analysis on an Applied Biosystems protein sequencer (model 473A).
Chromogenic Assay of Prothrombin Activation-Ten-l aliquots of sample (appropriately diluted with 20 mM Tris-HCl, 140 mM NaCl, pH 7.5 (TBS), containing 1 mg/ml bovine serum albumin) were mixed with 80 l of 1 M (or as otherwise indicated) bovine prothrombin in TBS containing CaCl 2 , and incubated at 37°C for an appropriate time (usually 20 min for routine assays). Then, 10 l of the chromogenic substrate for thrombin, t-butoxycarbonyl-Val-Pro-Arg-p-nitroanilide (4 mM; Seikagaku Kogyo, Tokyo, Japan) were added. The amount of thrombin generated was quantified by measuring the initial velocity of p-nitroaniline liberation at 405 nm with a kinetic plate reader (Well Reader, Seikagaku Kogyo), with pure bovine ␣-thrombin as the standard. If necessary, prothrombin activators were inactivated by EDTA prior to the addition of the substrate.
Fluorogenic Assay of Amidase Activity of CA-1-The design of the fluorogenic substrate, (7-methoxycoumarin-4-yl)acetyl-Ile-Asp-Gly-Arg-Ile-Val-Glu-Gly-(⑀-2,3-dinitrophenyl)Lys-amide, was based on the structure around the scissile site in human prothrombin (Ile 317 -Gly 324 ) and was synthesized by the Peptide Institute (Osaka, Japan). This peptide was designed so that fluorescence derived from the N-terminal coumarin derivative ( ex 328 nm, em 393 nm) was strongly quenched by the C-terminal dinitrophenyl group. Upon cleavage of the substrate, the fluorescence increases greatly with the release of the quencher (17,18). To an 800-l aliquot of a enzyme solution (20 nM) in 50 mM Tris-HCl, 100 mM NaCl, pH 8.5, 10 l of 1 mM substrate solution in dimethylformamide was added and the change with time in fluorescence was monitored at the ambient temperature.
The effects of various protease inhibitors were determined directly by this method. Each inhibitor to be tested (e.g. EDTA, Mn 2ϩ , Co 2ϩ , or other heavy metal ions at 10 mM) was incubated with the enzyme for 10 min at the ambient temperature and then assayed as described above.
Radiobinding Assay with the 25-kDa Subunit of CA-1-The 25-kDa subunit was labeled with Na 125 I (Du Pont NEN) by use of IODOBEADS (Pierce) in accordance with the manufacturer's instruction. The specific activity of the labeled protein was about 1.3 ϫ 10 6 cpm/g. The labeled protein (1 ϫ 10 5 cpm/tube) was incubated with 2 M fragment 1 in 50 mM Hepes-NaOH, pH 8.5, containing no or 10 mM CaCl 2 in a total volume of 50 l at 37°C for 60 min. Bis(sulfosuccinimidyl) suberate (BS 3 ; Pierce) was then added in 2 l of dimethyl sulfoxide to give a final concentration of 50 M, and the solution was left for 1 h at room temperature. The cross-linking reaction was terminated by the addition of 5 l of 50 mM ethanolamine, and the sample was then subjected to SDS-PAGE under non-reducing conditions. The dried gel was analyzed for the distribution of radioactivity with the BAS-2000 Bioimaging Analyzer system (Fuji Film, Tokyo, Japan).
Clotting Assay-One part (50 l) of the plasma to be tested was mixed with one part of prothrombin-deficient plasma (George King), and the mixture was incubated at 37°C for 2 min. The clotting reaction was initiated by the addition of one part of the mixture of activator/Ca 2ϩ (containing 1 nM factor Xa plus 1 mg/ml phospholipids (phosphatidylcholine/phosphatidylserine, 3:1, w/w) or 100 nM CA-1 or 100 nM ecarin in TBS containing 15 mM CaCl 2 ), which had been equilibrated at 37°C. The time required for clot formation was measured in an Amelung Coagulometer KC 4A. The amount of prothrombin in the sample was determined by reference to a standard curve that had been prepared with serially diluted pooled normal plasma; the logarithm of the clotting time was plotted against the logarithm of the amount of prothrombin. The concentrations of activators were adjusted to give a clotting time of 10 -15 s with normal plasma.
Other Methods-SDS-PAGE was performed by the method of Laemmli (19). Protein concentrations were determined with a BCA Protein Assay kit (Pierce) with bovine serum albumin as the standard.

RESULTS
The venom of E. carinatus was first fractionated by gel filtration and assayed for activation of prothrombin (Fig. 1A). Ecarin activity, which could be detected in the absence of Ca 2ϩ ions, was found in the first protein peak. When the assay was conducted in the presence of a millimolar concentration of Ca 2ϩ ions, the extent of activation of prothrombin was considerably enhanced. Since the activity of purified ecarin did not show any Ca 2ϩ dependence (see below), it was clear that this fraction also contained another, hitherto unidentified prothrombin activator(s) whose activity was dependent on Ca 2ϩ ions. We isolated this activity, as described below.
This fraction was applied to a column of Blue Sepharose. As is shown in Fig. 1B, two peaks of the activity of a Ca 2ϩ -dependent prothrombin activator were identified, and we designated these activities CA-1 and CA-2, respectively. Ecarin was eluted at higher concentrations of NaCl and was clearly separated from CA-1 and CA-2 at this step. Subsequent purification and characterization revealed that CA-2 was almost identical to CA-1 in terms of the molecular structure and enzymological features, and the main focus of this report is on CA-1. Isolation of CA-1 was accomplished by a third chromatography on Q-Sepharose (Fig. 1C). In a typical purification, we obtained 2 mg of CA-1 and 0.1 mg of ecarin from 100 mg of crude venom. The activity of CA-1 was irreversibly abolished by the incubation with EDTA or with heavy metals such as Co 2ϩ and Mn 2ϩ , but it was resistant to inhibitors of serine, thiol, or carboxyl proteases. Thus, CA-1 appeared to be a metalloenzyme, as is ecarin.
SDS-PAGE of CA-1 is shown in Fig. 2A (lane 1). Two bands (60/62-kDa doublet plus 25 kDa) were obtained under nonreducing conditions, and three bands (62/64, 17, and 14 kDa) were obtained under reducing conditions. The doublet appearance of the larger polypeptide was probably due to microheterogeneity, as discussed below. We were unable to separate these polypeptides by any subsequent chromatographic procedures under nondenaturing conditions. The polypeptides also comigrated as a single band in native-PAGE. Dissociation of the polypeptides required rather extreme conditions. For example, guanidine hydrochloride (Ͼ4 M) or SDS (Ͼ0.1%) was effective, but urea was ineffective up to 8 M. Fig. 2B depicts the chromatograms after gel filtration in either the absence or the presence of 4 M guanidine hydrochloride. Without the denaturant, a single, symmetrical peak that contained all three polypeptides was obtained. By contrast, two peaks were obtained in the presence of the denaturant; the first peak contained the 60/62-kDa polypeptide, and the second peak contained the 25-kDa component (see Fig. 2A). Thus, it appeared that CA-1 was a protein that consisted of two heterogeneous subunits held together non-covalently but very tightly, and that the 25-kDa subunit consisted of two different disulfidelinked polypeptides. The stoichiometry of the two subunits was 1 to 1. When the holoenzyme was subjected directly to protein sequence analysis, almost equimolar amounts (after correction for recovery of each amino acid) of three amino acids, corresponding to residues in the sequences of each of the three polypeptides (see below), were found after each sequencing cycle. Note that the second peak in the lower panel has a shoulder, but the peptides in this peak were homogeneous at least as far as could be judged by SDS-PAGE.
The N-terminal amino acid sequence of each polypeptide in CA-1 was analyzed (Fig. 3A). The 60/62-kDa chain had a sequence highly homologous to that of ecarin (7), and this subunit did, indeed, have metalloprotease activity, as described below. The 60-kDa polypeptide had an identical sequence to that of the 62-kDa polypeptide but lacked two N-terminal residues (Ser-Arg), suggesting the existence of a microheterogeneity (hereafter, we refer the 60/62-kDa polypeptide simply as the 62-kDa polypeptide). The two chains of the 25-kDa subunit resembled one another; furthermore, these sequences were rather similar to those of the snake venom protein IX/X-bp (20). IX/X-bp is a protein isolated in this laboratory from the venom of the habu snake Trimeresurus flavoviridis. It has strong anticoagulant activity and acts by neutralizing factors IX/IXa and X/Xa through binding to the Gla domains of these factors in a Ca 2ϩ -dependent fashion (21)(22)(23). As is shown in Fig. 3B, the apparent molecular masses of the respective polypeptide chains of CA-1 also resemble those of ecarin and IX/X-bp. The structural similarity of CA-1 to these venom proteins was confirmed by an immunochemical method. In immunoblotting analysis, the 62-kDa polypeptide of CA-1 cross-reacted with antiserum raised against ecarin and the 25-kDa subunit crossreacted with antiserum against IX/X-bp (data not shown).
We examined the enzymological features of CA-1. First, we evaluated the specificity of CA-1. Although it very efficiently activated prothrombin of bovine or human origin, it had no effect on any other vitamin K-dependent coagulation proteins. We incubated factors VII, IX, and X and protein C (all of bovine origin) with CA-1 for a long period and then analyzed by SDS-PAGE. Little if any cleavage was detected (data not shown). Thus, the specificity of CA-1 appears to be very strict, as is that of ecarin. Furthermore, the activation of prothrombin (0.1 M) was unaffected by high concentrations of factor IX or factor X (up to 10 M).
The most prominent enzymological difference between CA-1 and ecarin was the former enzyme's requirement for Ca 2ϩ ions for the activation of prothrombin. As shown in Fig. 4A, Ca 2ϩ ions at millimolar concentrations were necessary for CA-1 (half-maximal and maximal activation occurred at 0.2 and 2 mM Ca 2ϩ , respectively), while purified ecarin had no such requirement. In the absence of Ca 2ϩ ions, the generation of active thrombin by CA-1 was extremely slow (less than 1/100 of that in the presence). The Ca 2ϩ -dependent thrombin generation was confirmed by SDS-PAGE; when the incubation mixture of bovine prothrombin with CA-1 was subjected to SDS-PAGE under reducing condition, the band corresponding to the B chain of thrombin appeared only when Ca 2ϩ ions had been present (Fig. 4B). In a non-reducing gel, the band corresponding to meizothrombin but not ␣-thrombin was observed (data not shown). It appears that CA-1 cleaves only the bond between the A chain and the B chain, as does ecarin. By contrast to the  (7) and IX/X-bp (20). Identical residues are shaded. Two N-terminal residues of the 62-kDa polypeptide, indicated by dots, are absent from the 60-kDa polypeptide. B, schematic representation of the structures of CA-1, ecarin, and IX/X-bp. The apparent molecular mass of each polypeptide is shown. reaction with the natural substrate, hydrolysis by CA-1 (and also by ecarin) of a peptidyl fluorogenic substrate, which was synthesized to resemble the scissile site in prothrombin, was unaffected by exogenous Ca 2ϩ ions (Table I). Thus, it seemed likely that Ca 2ϩ ions were required for the recognition of prothrombin by CA-1; prothrombin is capable of binding Ca 2ϩ ions via the N-terminal Gla domain and undergoes a dramatic change in conformation at Ca 2ϩ -concentrations around 1 mM. It appeared possible that CA-1 might recognize the Gla domain of prothrombin with bound Ca 2ϩ ions, primarily via the IX/X-bplike 25-kDa subunit, which is not present in ecarin.
To examine this possibility, we investigated the activation of prothrombin in detail using prothrombin derivatives and isolated CA-1 subunits. First, we tested prethrombin-1, which lacks prothrombin fragment 1, the N-terminal portion of prothrombin that includes the Gla domain. As is shown in Table I, removal of the Ca 2ϩ -binding site from the prothrombin molecule had a striking negative effect on the action of CA-1. Moreover, the activation of this derivative no longer required Ca 2ϩ ions, even though the rate of activation was greatly diminished, and this rate was close to that for intact prothrombin without Ca 2ϩ ions. The steady-state kinetic parameters for the activation of prothrombin were obtained. The apparent K m and V max values of CA-1 for the activation of bovine prothrombin in the presence of 5 mM Ca 2ϩ ions were 1.1 M and 33 mol of prothrombin/min/mol of CA-1, respectively, while those of ecarin were 0.44 M and 17 mol/min/mol. In the absence of Ca 2ϩ ions, the K m of CA-1 was greatly increased (to Ͼ100 M) but that of ecarin was essentially unchanged. Human prothrombin was also a good substrate, and results similar to those with bovine protein were obtained. Apparently, the action of CA-1 was strongly dependent on both of Ca 2ϩ ions and the Gla domain of prothrombin, while that of ecarin was influenced by neither of these factors.
Next, we evaluated the role of the subunits of CA-1 (Table II). When we examined the amidolysis of the fluorogenic substrate, we found that the isolated 62-kDa subunit had enzymatic activity almost equivalent to that of the intact CA-1, but the rate of activation of prothrombin, the natural substrate, was much reduced. The 25-kDa subunit had no enzymatic activity (data not shown). Removal of the 25-kDa subunit also led to the loss of the dependence on Ca 2ϩ ions. It appeared, therefore, that the 62-kDa component is the metalloprotease catalytic subunit, while the 25-kDa non-catalytic component is an accessory, regulatory subunit. Reconstitution with the 25-kDa subunit restored both the high potency and the dependence on Ca 2ϩ ions. By contrast, the rate of activation of prethrombin-1 was unaffected by reconstitution, and the rate of the reaction was very close to that for prothrombin activation in the absence of the 25-kDa subunit and/or Ca 2ϩ ions. These results strongly sug-gested that CA-1 recognized the conformation of the Gla domain of prothrombin with bound Ca 2ϩ ions via its 25-kDa subunit.
In order to obtain further evidence, we conducted three additional experiments. In the first one, we examined the effect of prothrombin fragment 1 on the activation of prothrombin. As is shown in Fig. 5, the activation by CA-1, but not by ecarin, was effectively inhibited by fragment 1. This result indicated that the low reactivity of prethrombin-1 with CA-1 was due to the absence of fragment 1 and not due to a secondary change in conformation of the protein upon liberation of the N-terminal portion.
In the second experiment, we used plasma from individuals who had taken a vitamin K antagonist, in which abnormal prothrombin with incompletely carboxylated Gla residues was present concomitant with the decreased level of normal prothrombin. A batch of plasma with a clotting activity of 20% of that of normal controls, which had been determined by a standard assay of prothrombin time, was utilized. The plasma was mixed with prothrombin-deficient plasma, the activator, and Ca 2ϩ ions, and the time required for clot formation was measured. The prothrombin content of the tested plasma was estimated with serially diluted normal control plasma as the reference. When CA-1 was used as the activator, the results were similar to those obtained with the physiological activator factor Xa (Table III). By contrast, with ecarin, higher values were obtained, probably representing the sum of normal and abnormal prothrombins since it is known that ecarin can activate abnormal prothrombin as well as normal prothrombin (24). It was apparent that CA-1 selectively recognized normal pro-

TABLE I The velocities of cleavage of different substrates by CA-1 and ecarin
The initial velocity for the activation of bovine prothrombin (1 M) or of prethrombin-1 (1 M) by CA-1 (0.5 nM) or ecarin (0.5 nM) was determined in the presence and absence of 5 mM Ca 2ϩ ions. Effects of Ca 2ϩ ions on amidolytic activity of CA-1 and ecarin toward the fluorogenic substrate were also measured. Detailed procedures are given under "Experimental Procedures." Data are means of results of duplicate determinations in one of three similar experiments.  thrombin with all the Gla residues intact, even in the presence of excess acarboxyprothrombin. Ecarin did not recognize such differences, and it seemed to recognize only the scissile site.
In the third experiment, we investigated the association between fragment 1 and the isolated 25-kDa subunit directly. We employed a cross-linking technique. The 125 I-labeled 25-kDa subunit was incubated with fragment 1 in the presence of Ca 2ϩ ions, and the complex formed was stapled by the bifunctional reagent BS 3 . The resultant stable complex was subjected to SDS-PAGE followed by analysis with a radioimaging analyzer (Fig. 6). A small amount of the 25-kDa subunit was self-associated, and a radioactive band corresponding to the apparent molecular mass of 43 kDa (presumably a dimer) was found even in the absence of the cross-linker (lane 1). This species was presumably generated during the radiolabeling procedure, which included an oxidizing reaction. In the presence of Ca 2ϩ ions, fragment 1 per se associated with one another as reported previously (25), and a ladder-like pattern was visible after Coomassie Blue staining. The radiolabeled 25-kDa subunit was indeed incorporated into these fragment 1 polymers in the presence of Ca 2ϩ ions, but in the absence of Ca 2ϩ ions no incorporation occurred (lanes 3 and 4). The association with fragment 1 was effectively blocked by the addition of an excess of the cold 25-kDa subunit (lane 5), indicating that this interaction was specific.
These results together demonstrate that the unique structure of CA-1 explains the unique mechanism of its activation of prothrombin; the 25-kDa subunit first recognizes the N-terminal Gla domain of prothrombin in a Ca 2ϩ -dependent fashion, and then the 62-kDa subunit cleaves the distal scissile site, the bond between the A chain and the B chain (Fig. 7). DISCUSSION The results described herein clearly show that the venom of E. carinatus contains a hitherto novel type of prothrombin activator. This finding necessitates reconsideration of the classification of exogenous prothrombin activators. We propose that the previously defined group 1 enzymes (3) be divided into two subgroups, i.e. the ecarin-like (Ca 2ϩ -independent) metalloproteases (perhaps termed group 1A) and the carinactivaselike (Ca 2ϩ -dependent) enzymes (group 1B).
We screened the venoms of various Viperidae snakes for carinactivase-like activity (detailed screening data will be published elsewhere). 3 All the venom preparations from Echis snakes contained both ecarin-like and carinactivase-like activators, although the total activities as well as the relative abundance of two enzymes varied depending upon the source. However, we failed to detect carinactivase-like activity in venoms of Viperidae snakes in genera other than Echis, although ecarin-like activities were found in some of them.
CA-1 appears to be a hybrid protein that is well adapted for the efficient and selective activation of prothrombin in the plasma of target animals after envenomation. The 25-kDa subunit of CA-1 exhibits striking structural similarity to the anticoagulant IX/X-bp in the T. flavoviridis venom. IX/X-bp is a heterodimeric protein that consists of two homologous polypeptide chains (20), and it recognizes Ca 2ϩ -bound conformations of the Gla domains in factors IX and X (22,23). The function of the 25-kDa subunit is also similar to that of IX/X-bp, i.e. the Ca 2ϩdependent recognition of the Gla domain of prothrombin. Thus, the non-catalytic component is the regulatory subunit (Fig. 7). The specificity of IX/X-bp is very strict; it never binds other vitamin K-dependent coagulation factors such as prothrombin 3   F IG. 7. The proposed mechanism for the recognition and activation of prothrombin by CA-1. The 25-kDa regulatory subunit first recognizes the Ca 2ϩ -bound conformation of the Gla domain of prothrombin, and then the 62-kDa catalytic subunit cleaves the bond between the A and B chains, generating meizothrombin. Participation of Ca 2ϩ ions in the exposure of the Gla domain recognition site on the regulatory subunit has not been proven and is hypothetical at present. For further details, see under "Discussion." (21). This point is of great interest, since the Gla domains in these proteins are very similar to one another in terms of primary structure and, in view of their common function (Ca 2ϩdependent binding to anionic phospholipids), their tertiary structures should be also similar. It appears that IX/X-bp can discriminate slight difference in the tertiary structures of Gla domains. A similar statement can be made about the binding specificity of the regulatory subunit of CA-1. Strong substrate specificity and the absence of inhibitory effect of factors IX and X on the activation of prothrombin indicate that the subunit selectively recognizes the Gla domain of prothrombin. In the presence of Ca 2ϩ ions, prothrombin should interact with the regulatory subunit of CA-1 via its Gla domain. Then its scissile site is presented in the proper orientation to the active center of the 62-kDa metalloprotease catalytic subunit (Fig. 7). The regulatory subunit acts as an effective condenser of the substrate and greatly reduces the apparent K m . This function is analogous to that of phospholipids in the physiological prothrombin activator, the prothrombinase complex (factor Xa plus factor Va, anionic phospholipids, and Ca 2ϩ ions). Therefore, studies with CA-1 should provide insight into the structure and function of the prothrombinase complex at a molecular level from a novel perspective.
We showed previously that IX/X-bp is also capable of binding Ca 2ϩ ions (2 ions/molecule) and that occupation of the Ca 2ϩbinding sites is a prerequisite for subsequent binding to coagulation factors (23). It is possible that the regulatory subunit of CA-1 might also bind Ca 2ϩ ions and, upon binding of Ca 2ϩ ions, the recognition site for the Gla domain of prothrombin would be exposed (Fig. 7).
The results indicate that IX/X-bp and the regulatory subunit of CA-1 are highly analogous not only in terms of structure but also in terms of function, even though they have opposite toxicological effects. IX/X-bp is an anticoagulant, and it should support the action of hemorrhagic factors that are present in the same venom (26), while CA-1 is an enzyme that causes thrombosis via the generation of thrombin. In addition to IX/ X-bp from the habu snake venom, proteins structurally related to the regulatory subunit of CA-1 are widely distributed in viper venoms. We have identified homologues of IX/X-bp in venoms of Bothrops jararaca (27) and Deinagkistrodon acutus (28). Moreover, numerous proteins with structures very similar to IX/X-bp but with totally different pharmacological actions have been found in venoms of various Viperidae snakes, and appear to constitute a unique protein family. Ligands for these proteins are very heterogeneous. For example, botrocetin from B. jararaca binds von Willebrand factor (29), bothrojaracin from the same venom binds the anion-binding exosite in ␣-thrombin (30), and alboaggregin from T. albolabris binds platelet glycoprotein Ib (31). Each of these venom proteins has two homologous polypeptide chains, and each chain constitutes the domain structure, known as CRD. The name CRD originally referred to the "carbohydrate-recognition domain" because this structure was first identified as the minimum functional motif of Ca 2ϩ -dependent animal lectins such as asialoglycoprotein receptor. It is now known that the CRD is widely distributed in the animal kingdom, from invertebrates such as sea urchins to mammals, and it seems to be a fundamental motif that acts as an important domain in the construction of proteins (32,33).
Another component in CA-1, the catalytic subunit, also has numerous relatives in the venoms of Viperidae. These relatives include ecarin in the same venom and hemorrhagic factors in T. flavoviridis and Crotalus atrox venoms (26,34). In addition, many proteins that resemble to these venom metalloproteases have recently been identified in mammalian tissue, in partic-ular in reproductive organs, e.g. the sperm protein fertilin (PH-30) (35). Together, all these proteins constitute a superfamily for which the name ADAM has been proposed (8).
It is noteworthy that the factor X activator RVV-X in Vipera russelli venom has a structure very similar to that of CA-1. This protein also has three polypeptide chains, i.e. 57.6-, 19.4-, and 16.4-kDa chains, with a stoichiometry of 1:1:1, and these chains are held together by disulfide bonds, as recently proven unequivocally by Gowda et al. (36). In an earlier report, Takeya et al. (37) determined the complete amino acid sequences of the 57.6-and 16.4-kDa chains, and showed that the 57.6-kDa chain has a structure very similar to that of ecarin while the 16.4-kDa chain has a sequence homologous to that of IX/X-bp. Furthermore, the 19.4-kDa chain also has an N-terminal sequence homologous to IX/X-bp. 4 The action of RVV-X is also dependent on both Ca 2ϩ ions and the Gla domain in factor X (38,39). Therefore, it seems likely that the catalytic mechanism of RVV-X might be similar to that of CA-1. However, isolation of the intact catalytic chain of RVV-X appears to be impossible and thus unequivocal biochemical evidence cannot be obtained, because the chain is covalently linked to IX/X-bp-like chains. Both CA-1 and RVV-X are composed of two different components, which have a totally different genetic origin, and both proteins are likely to have originated from a single ancestral hybrid protein. It is still unclear how they are synthesized and correctly folded. The topology of the polypeptides in these enzymes is also unknown. These issues require further investigations.
In conclusion, we have shown that a newly isolated novel prothrombin activator, CA-1, has a unique machinery for the recognition and subsequent processing of its substrate. CA-1 should be useful as a convenient probe for biochemical studies in vitro of prothrombin and as a good diagnostic reagent for monitoring normal prothrombin levels in plasma (cf . Table III). Furthermore, this enzyme should be a good model for attempts to elucidate details of the evolution and the biosynthesis of multi-subunit proteins.