TAFI, or plasma procarboxypeptidase B, couples the coagulation and fibrinolytic cascades through the thrombin-thrombomodulin complex.

TAFI (thrombin-activatable fibrinolysis inhibitor) is a recently discovered plasma protein that can be activated by thrombin-catalyzed proteolysis to a carboxypeptidase B-like enzyme that inhibits fibrinolysis. This work shows that the thrombin-thrombomodulin complex, rather than free thrombin, is the most likely physiologic activator. Thrombomodulin increases the catalytic efficiency of the reaction by a factor of 1250, an effect expressed almost exclusively through an increase in kcat. The kinetics of the reaction conform to a model whereby thrombin can interact with either TAFI (Km = 1.0 μM) or thrombomodulin (Kd = 8.6 nM), and either binary complex so formed can then interact with the third component to form the ternary thrombin-thrombomodulin-TAFI complex from which activated TAFI is produced with kcat = 1.2 s−1. This work also shows that activated TAFI down-regulates tPA-induced fibrinolysis half-maximally at a concentration of 1.0 nM in a system of purified components. This concentration of TAFI is about 2% of the level of the zymogen in plasma, which indicates that ample activated TAFI could be generated to very significantly modulate fibrinolysis in vivo. Therefore, TAFI in vitro and possibly in vivo defines an explicit molecular connection between the coagulation and fibrinolytic cascades, such that expression of activity in the former down-regulates the activity of the latter.

TAFI (thrombin-activatable fibrinolysis inhibitor) is a recently discovered plasma protein that can be activated by thrombin-catalyzed proteolysis to a carboxypeptidase B-like enzyme that inhibits fibrinolysis. This work shows that the thrombin-thrombomodulin complex, rather than free thrombin, is the most likely physiologic activator. Thrombomodulin increases the catalytic efficiency of the reaction by a factor of 1250, an effect expressed almost exclusively through an increase in k cat . The kinetics of the reaction conform to a model whereby thrombin can interact with either TAFI (K m ‫؍‬ 1.0 M) or thrombomodulin (K d ‫؍‬ 8.6 nM), and either binary complex so formed can then interact with the third component to form the ternary thrombin-thrombomodulin-TAFI complex from which activated TAFI is produced with kcat ‫؍‬ 1.2 s ؊1 . This work also shows that activated TAFI down-regulates tPA-induced fibrinolysis half-maximally at a concentration of 1.0 nM in a system of purified components. This concentration of TAFI is about 2% of the level of the zymogen in plasma, which indicates that ample activated TAFI could be generated to very significantly modulate fibrinolysis in vivo. Therefore, TAFI in vitro and possibly in vivo defines an explicit molecular connection between the coagulation and fibrinolytic cascades, such that expression of activity in the former down-regulates the activity of the latter.
The coagulation and fibrinolytic cascades comprise a series of zymogen to enzyme conversions which terminate in the proteolytic enzymes thrombin and plasmin, respectively (1)(2)(3). These enzymes catalyze the deposition and removal of fibrin. A proper balance between the activities of the two cascades is required both to protect the organism from excessive blood loss upon injury and to maintain blood fluidity within the vascular system. Imbalances are characterized by either bleeding or thrombotic tendencies, the latter of which are manifested as heart attacks and strokes.
Thrombomodulin is a component of the blood vessel wall which binds thrombin and changes its specificity from fibrinogen to protein C, yielding anticoagulant rather procoagulant activity (4). The thrombin-thrombomodulin complex catalyzes cleavage of protein C to activated protein C, which then down-regulates the coagulation cascade by proteolytically inactivating the essential cofactors Factor Va and Factor VIIIa (5). These events are essential in the regulation of the coagulation cascade (4).
Early studies suggested that activated protein C is not only anticoagulant but also profibrinolytic, both in vitro and in vivo (6 -9). Subsequent work from our laboratory showed that activated protein C only appears profibrinolytic because it prevents the thrombin-catalyzed activation of a previously unknown fibrinolysis inhibitor (10). The precursor of this inhibitor was isolated from plasma (11) and was designated TAFI (thrombinactivatable fibrinolysis inhibitor). The zymogen is activated by thrombin to an enzyme with carboxypeptidase B-like activity. This enzyme, designated TAFIa, inhibits plasminogen activation and thereby prolongs fibrinolysis (11), presumably by removing C-terminal lysines from partially degraded fibrin, thereby attenuating the cofactor activity of fibrin and preventing the accelerated phase that ordinarily occurs during plasminogen activation (12).
Our initial studies suggested that thrombin is a weak activator of TAFI. Thus, the present work was initiated to analyze the effects of a soluble form of thrombomodulin (Solulin, Ref. 13) on the reaction. The results presented below comprise a report of both the activation of TAFI by the thrombin-thrombomodulin complex and the attenuation of fibrinolysis by TAFIa.

EXPERIMENTAL PROCEDURES
Proteins and Reagents-The human proteins fibrinogen, plasminogen, prothrombin, and antithrombin III and TAFI were isolated from plasma as described previously (11). Recombinant tissue plasminogen activator (Activase) was a generous gift of Dr. Gordon Vehar of Genentech (South San Francisco, CA). Recombinant soluble thrombomodulin (Solulin) was obtained as a generous gift from Berlex Biosciences (Richmond, CA). Recombinant human ␣ 2 -antiplasmin was isolated from culture supernatants of baby hamster kidney cells transfected with the human cDNA and grown in serum free medium, as described previously (11). The synthetic carboxypeptidase B substrate, hippuryl-arginine, was obtained from Sigma. All other reagents were of analytical quality.
Analysis of Thrombin-Thrombomodulin-dependent TAFI Activation by SDS-PAGE 1 and Activity Measurements-TAFI (1.92 M), thrombin (1.0 nM), and soluble thrombomodulin (50 nM) were incubated at 22°C in 0.02 M HEPES, 0.15 M NaCl, 5.0 mM CaCl 2 , pH 7.4. At regular intervals samples were removed and added to the irreversible thrombin inhibitor D-phenylalanylprolylarginyl chloromethyl ketone (PPAck), the final concentration of which was 5.0 M. One aliquot was then immediately assayed for carboxypeptidase B activity by adding it to hippurylarginine (400 M) in 0.02 M Tris-HCl, 0.10 M NaCl, pH 7.4, and measuring the time course of increase in absorbance at 254 nm in a Perkin-Elmer 4B spectrophotometer. Another aliquot was then prepared for SDS-PAGE in a 5-15% gradient gel under nonreducing conditions according to the method of Neville (14). The gel was stained with Coomassie Blue, destained, and scanned with an LKB densitometer. One sample, taken at 5 min and added to PPAck, was assayed at regular intervals for the next 2 h to determine the rate of spontaneous activity loss (no cleavage occurs in this interval). The kinetics of loss were first order with a half-life of 70 min. Activity measurements on samples removed over the 2-h interval were recorded both as raw data and as data corrected for spontaneous decay up to the time of sampling and assay. To make this correction, the sequence of reactions TAFI3 TAFIa3 TAFIi, where TAFIi implies the product of spontaneous decay, were both approximated as first order with rate constants k a and k d . Under these conditions [ Ϫ exp(Ϫk d ⅐t)). In this way TAFIa measured by activity levels were compensated for the activity that had spontaneously decayed (without further cleavage) from the initiation of the reaction until the time of sampling. The rate constant for activation of the zymogen TAFI was determined by analysis of densitometry scans of the TAFI bands on the SDS-PAGE gel (k a ϭ 0.0225 min Ϫ1 ), and the rate constant for spontaneous decay was determined from activity measurements made at regular intervals for up to 2 h on the sample that had been removed from the reaction at 5 min and added to PPAck (k d ϭ 0.010 min Ϫ1 ). , was added in small successive aliquots to an otherwise identical solution that lacked thrombomodulin. The additions were performed in a cuvette fitted with a magnetic stirrer in the sample compartment of a Perkin-Elmer model LS50B spectrofluorimeter. Intensity values were continuously recorded with excitation and emission wavelengths of 280 and 545 nm, respectively. A 430-nm cut-off filter was used in the emission beam. The interaction between TAFI and thrombomodulin was marked by a 20% increase in intensity as the thrombomodulin concentration was elevated sufficiently to saturably bind thrombin.

Kinetics of the Thrombomodulin-dependent Activation of TAFI-
The data were analyzed as follows. The intensity of fluorescence, I, was assumed to be the sum of intensities from thrombin-DAPA (T⅐D) and thrombin-thrombomodulin-DAPA (T⅐TM⅐D). That is, , where i 1 and i 2 are the coefficients of fluorescence for T⅐D and T⅐TM⅐D (since excitation was at 280 nm, the emission from free DAPA was negligible (16)). Because TM does not appreciably alter the K m for either protein C activation, as shown by Le Bonnniec et al. (17), or TAFI activation (as will be shown below), we assume that it does not alter the affinity of the thrombin-DAPA interaction. Thus Normalizing to the initial intensity gives (⌬I/I 0 ) ϭ (⌬I max /I 0 )⅐b. If DAPA binds T and T⅐TM with equal affinity, then TM binds T and T⅐D with equal affinity.
This is a quadratic equation in b, which when solved and substituted in the expression above for (⌬I/I 0 ) gives the equation: ). This latter equation expresses the relationship between fluorescence intensity values, the nominal concentrations of thrombomodulin and thrombin, the dissociation constant for the thrombin-thrombomodulin interaction, and the fluorescence intensity increment that signals the interaction of thrombomodulin with thrombin-DAPA. Intensity data were fit to the above equation by nonlinear regression analysis, with [TM] 0 as the independent variable and K TM and ⌬I max as best-fit parameters.
In order to determine whether TAFI binds to thrombin, TAFI was used as a competitive substrate in the thrombin-catalyzed . Aliquots (100 l) of each mixture were then immediately added to the wells of a microtiter plate reader containing small separated aliquots of thrombin (6.0 nM final, including that added with the TAFIa) and tissue plasminogen activator (441 pM, final). The samples in the wells of the plate then were monitored for turbidity at 405 nm, at 2.5-min intervals 37°C, in a Titretek plate reader operated in the kinetics mode. Under these conditions, clotting, marked by increased turbidity, occurs within 1-2 min, and fibrinolysis, marked by decreased turbidity, occurs within about 30 -90 min, depending on the contents of the well. A parameter, denoted lysis time, is defined as the time, after adding the sample to the well, at which the turbidity is one-half the difference between the plateau reached after clotting and the base-line value achieved at complete lysis. In these experiments the TAFIa concentration was varied from 0 to 50 nM. Controls were performed by executing identical experiments without TAFI.

RESULTS
The proteolytic activation of TAFI by thrombin-thrombomodulin was analyzed by both SDS-PAGE (Fig. 1) and measurements of carboxypeptidase B activity with hippuryl-arginine (Fig. 2). The zymogen (58 kDa) is progressively cleaved to yield major and minor components, with respective masses of 35 and 25 kDa. Although direct measurements of activity do not correlate with the appearance of the 35-kDa product, values corrected for the intrinsic instability of TAFIa (t1 ⁄2 ϭ 70 min) correlate very well (Fig. 2).
Initial rates of TAFI activation show saturation in the concentrations of both TAFI (Fig. 3) and thrombomodulin (Fig. 4). K m values are independent of the thrombomodulin concentration, and V max values show saturation in it (Table I). These patterns were analyzed for conformity to one or more of the seven equilibrium models described by Boskovic  Since both models have the same rate equation, they cannot be distinguished by measurements of steady-state kinetics alone. They do have different physical interpretations, however. Although both models predict the thrombin-thrombomodulin complex, one predicts the existence of thrombomodulin-TAFI, but not thrombin-TAFI, on the reaction pathway, whereas the other predicts the opposite. Because of these differences, the models can be distinguished by characterizing the binary interactions between the three components.
In order to confirm the existence of the thrombin-thrombomodulin complex and to determine which one of the other two possible binary complexes exist and thereby distinguish between the two models, the binding of thrombin to thrombomodulin was measured by perturbation of the fluorescence of DAPA. In addition, the binding of thrombin to TAFI was investigated by competition by TAFI for the thrombin-catalyzed hydrolysis of a small substrate (S-2266). Thrombin binding to thrombomodulin was characterized by K d ϭ 23 Ϯ 14 nM (Fig. 5), and competition kinetics indicated 50% inhibition at 1.2 M TAFI (Fig. 6). These values are in very good agreement with  the K d and K m values inferred independently from the kinetics of TAFI activation in both the absence and presence of thrombomodulin (Table II). These observations support the model for TAFI activation shown in Fig. 7. According to this model, the formation of the ternary TAFI-thrombin-thrombomodulin complex can proceed via two parallel paths involving, respectively, the binary thrombin-thrombomodulin and TAFI-thrombin complexes. All of the data were fit globally to the above rate equation by nonlinear regression analysis with K d , K m , and k cat as fit parameters. The lines shown on Fig. 4 are the regression lines and the fits appear to be excellent. The results, along with those obtained in the absence of thrombomodulin, are summarized in Table II. The effect of thrombomodulin is expressed almost exclusively through an increase in k cat . The catalytic efficiency (k cat /K m ) of thrombin-thrombomodulin is 1250-fold greater than that of thrombin.
In order to assess the impact of TAFIa on fibrinolysis, samples of it at increasing concentrations were added to a system of purified fibrinolytic components in which initial clotting (in-duced by thrombin) and subsequent fibrinolysis (induced by tissue plasminogen activator) are monitored over time by turbidity at 405 nm. The formation of fibrin is accompanied by a rapid increase in turbidity to a plateau value, and subsequent fibrinolysis is marked by a corresponding decrease (Fig. 8). As the TAFIa concentration increases, the time required for fibrinolysis to occur increases. The relationship between the TAFIa concentration and time required for the turbidity value to fall to one-half the maximum value is shown in Fig. 9. The value increased in a saturable manner from 30 to about 90 min, and the half-maximal effect was achieved at a TAFIa concentration of 1.0 nM. Since the zymogen TAFI is present in plasma at a concentration of about 50 nM (11), TAFIa attenuates the dissolution of fibrin via the fibrinolytic cascade at a concentration considerably below that which potentially could be generated by thrombin-thrombomodulin in vivo.

DISCUSSION
The present work shows that TAFI is activated to an inhibitor of fibrinolysis by the thrombin-thrombomodulin complex. Although thrombin at high concentrations, such as those which could potentially be generated upon complete conversion of the prothrombin in plasma to thrombin, can activate TAFI in plasma (11), the catalytic efficiency is increased about three orders of magnitude in the presence of thrombomodulin. Thus, the thrombin-thrombomodulin complex is most likely the physiologic activator of TAFI. In addition, TAFI is now the second known macromolecular substrate for the thrombin-thrombomodulin complex, the other of which is protein C, the precursor of the anticoagulant enzyme, activated protein C. The existence of these two substrates implies that the thrombin-thrombomodulin complex may contribute to the down-regulation of not only the coagulation cascade but also the fibrinolytic cascade.
In some respects the mechanisms by which thrombomodulin enhances activation of TAFI and protein C appear similar. The   (16) and implies that the enzyme, thrombin (T), can interact with either TAFI or thrombomodulin (TM), and the resulting binary complexes interact further to form the ternary T⅐TM⅐TAFI complex, from which TAFIa is formed. Independent evidence for the existence of the binary complexes is given in Figs. 5 and 6.
1.3 s Ϫ1 for protein C (17)). In addition, the enhanced catalytic efficiency is clearly expressed through an effect in k cat in TAFI activation. This is probably so also in protein C activation, although this conclusion is obscured somewhat by the complex dependence of reaction kinetics on the concentration of Ca 2ϩ . At the level of structure, however, some subtle differences in the two substrates are apparent. The sequence of amino acids corresponding to P7-P5Ј peptide surrounding the thrombin activating cleavage in protein C site is EDQVDPRLIDGK (19). Le Bonniec et al. (17) and Le Bonniec and Esmon (19) provided convincing arguments with mutants of thrombin that the aspartic acid residues at the P3 and P3Ј positions of this sequence contribute substantially to resistance to cleavage by thrombin in the absence of thrombomodulin. In TAFI, however, the corresponding P7 to P5Ј sequence is NDTVSPRASAYY (11,19). Thus, in TAFI the negatively charged P3 and P3Ј residues are not present. Determining whether the elements of structure of TAFI and protein C, which both limit cleavage by free thrombin and allow it by thrombin-thrombomodulin, are similar or grossly different clearly will require further work. In addition, TAFI provides another tool by which to gather further insights into the means by which thrombomodulin alters the activity of thrombin toward macromolecular substrates.
TAFI was discovered independently in three different laboratories. It initially appeared as an unstable carboxypeptidase B-like entity in human serum and was described by Hendriks et al. (20). Then Eaton et al. (21) discovered it as a contaminant in preparations of ␣ 2 -antiplasmin; they cloned the cDNA, deduced the amino acid sequence, described its activation by trypsin, and analyzed its enzymatic properties toward synthetic carboxypeptidase B substrates. They designated the protein pCPB, for plasma carboxypeptidase B. Wang et al. (22) independently isolated the activated material and named it carboxypeptidase U, where "U" indicates unstable. In addition, our group discovered it, showed that it is both activated by thrombin and inhibits fibrinolysis, and that it accounts for the apparent profibrinolytic effects of activated protein C. Consequently, we named it TAFI (11). Subsequently, Tan and Eaton (23) studied the trypsin activated enzyme and renamed the protein plasma procarboxypeptidase B (pro-pCPB). The identity of TAFI and pro-pCPB is established by their behavior in affinity chromatography on plasminogen Sepharose and the amino acid sequences at the activation cleavage site (11,21). In addition, Redlitz et al. (24) recently demonstrated that activated pro-pCPB and carboxypeptidase N diminish tPA-induced plasminogen binding to U937 cells and that fibrinolysis occurs more rapidly in carboxypeptidase depleted, clotted plasma than in controls.
Our previous results indicate that TAFI couples the coagulation and fibrinolytic cascades in vitro, such that the operation of the former down-regulates the activity of the latter (10 -11). The present work confirms this and further suggests that this connection is most likely established in vivo through the thrombin-thrombomodulin complex on the endothelial cell lining of the blood vessel. Because TAFI is exquisitely sensitive to activation by thrombin-thrombomodulin and, when activated, potently suppresses fibrinolysis, it very likely plays a fundamental role in mediating "cross-talk" between the coagulation and fibrinolytic cascades. Because of this, TAFI should be considered in the future, both in studies of biochemical defects that lead to bleeding or thrombosis and in efforts to therapeutically accomplish thrombolysis through the use of components of the fibrinolytic cascade.