Two naturally occurring variants of TAFI (Thr-325 and Ile-325) differ substantially with respect to thermal stability and antifibrinolytic activity of the enzyme.

Thrombin-activable fibrinolysis inhibitor (TAFI) is a carboxypeptidase B-like zymogen that is activated to TAFIa by plasmin, thrombin, or the thrombin-thrombomodulin complex. The enzyme TAFIa attenuates clot lysis by removing lysine residues from a fibrin clot. Screening of nine human cDNA libraries indicated a common variation in TAFI at position 325 (Ile-325 or Thr-325). This is in addition to the variation at amino acid position 147 (Ala-147 or Thr-147) characterized previously. Thus, four variants of TAFI having either Ala or Thr at position 147 and either Thr or Ile at position 325 were stably expressed in baby hamster kidney cells and purified to homogeneity. The kinetics of activation of TAFI by thrombin/thrombomodulin were identical for all four variants; however, Ile at position 325 extended the half-life of TAFIa from 8 to 15 min at 37 degrees C, regardless of the residue at position 147. In clot lysis assays with thrombomodulin and the TAFI variants, or with pre-activated TAFI variants, the Ile-325 variants exhibited an antifibrinolytic effect that was 60% greater than the Thr-325 variants. Similarly, in the absence of thrombomodulin, the Ile-325 variants exhibited an antifibrinolytic effect that was 30-50% greater than the Thr-325 variants. In contrast, the variation at position 147 had little if any effect on the antifibrinolytic potential of TAFIa. The increased antifibrinolytic potential of the Ile-325-containing TAFI variants reflects the fact that these variants have an increased ability to mediate the release of lysine from partially degraded fibrin and suppress plasminogen activation. These findings imply that individuals homozygous for the Ile-325 variant of TAFI would likely have a longer lived and more potent TAFIa enzyme than those homozygous for the Thr-325 variant.

thrombomodulin complex (5) to the carboxypeptidase B-like enzyme, TAFIa. When exposed to a fibrin clot, TAFIa catalyzes the removal of carboxyl-terminal lysines, thereby diminishing the cofactor activity for plasminogen activation (6). Less efficient plasminogen activation on the fibrin clot corresponds to prolongation of fibrinolysis, and in this way TAFIa can serve as a potent antifibrinolytic enzyme. Studies performed using an in vitro human plasma model have found that clot lysis times can be attenuated up to 3-fold in the presence of TAFIa as compared with clots lysed in the absence of TAFIa (5).
Activation of TAFI is catalyzed only slowly by thrombin alone; however, in the presence of thrombin/thrombomodulin, the efficiency of activation is increased 1000-fold. Despite the large thrombomodulin dependence of TAFI activation, in vitro clot lysis assays done in the absence of thrombomodulin still exhibit prolonged clot lysis times as compared with similar assays performed with TAFI-depleted plasma, indicating that thrombin alone is still potentially significant with respect to TAFI activation. To form the active enzyme, TAFI is cleaved after Arg-92 to form the 37-kDa enzyme and the activation peptide. Even though TAFIa is a powerful antifibrinolytic enzyme, there are no known inhibitors of TAFIa in human plasma. Instead, TAFIa is highly unstable, with reported halflives at 37°C of 8 -9 (7,8) and 15 min (9). Whereas TAFIa can also be cleaved by thrombin at Arg-302, this cleavage is subsequent to the spontaneous structural changes corresponding to inactivation of the enzyme (8,10) and is therefore probably not relevant to the regulation of the enzyme.
The antifibrinolytic potential of TAFIa has been found to be concentration-dependent. In vitro human plasma models have measured a half-maximal antifibrinolytic effect at roughly 1 nM of the enzyme. Even though plasma TAFI concentrations have been reported as ranging between 75 and 250 nM (11)(12)(13), a maximum antifibrinolytic effect is reached at about 20 nM TAFIa. Therefore, only a small fraction of the potential plasma TAFI pool needs to be activated to mediate a pronounced antifibrinolytic effect. Whereas the maximum antifibrinolytic effect cannot be altered by increased concentrations of TAFIa, it can be substantially altered by less thermally stable variants of human TAFI (8).
Previously, we generated variants of TAFI with Arg to Gln substitutions at positions 302, 320, and 330 (8). Each of these variants was found to be less thermally stable than the wild type, highlighting the importance of this region for the stability of the active enzyme. Moreover, these TAFI variants exhibited a decreased ability to attenuate lysis of a fibrin clot. Because less stable variants had less antifibrinolytic potential, thermal stability was shown to correspond to the antifibrinolytic potential of TAFIa. At concentrations of TAFIa above 20 nM, the determinant of its antifibrinolytic effect seems to depend on the * 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.
half-life of the TAFIa rather than the concentration of the enzyme.
In this study we have identified a naturally occurring variation in human TAFI at position 325 (Thr-325 or Ile-325). It is particularly interesting that this variation lies within the region of TAFI shown to influence thermal stability (8). Previous work has also identified a variation at position 147 (Ala-147 or Thr-147), but the functional properties of the two variants were indistinguishable (14). Our collaborators (15) have recently determined the prevalence of the Ile-325 allele in a population of 152 individuals and found that Ile/Thr-325 is a common polymorphism in human TAFI. Here, we report that the Ile/Thr variation at position 325 corresponds to TAFI variants with substantial differences in their thermal stabilities and antifibrinolytic potentials. We have also introduced anisylazoformyllysine (AAFK) as a substrate for measuring TAFIa activity. Previously, AAFK has been shown to be cleaved by pancreatic carboxypeptidase B (16), and we found that it is similarly hydrolyzed by TAFIa. Thrombin was prepared from human plasma-derived prothrombin as described previously (17). Fibrinogen was prepared from fresh frozen human plasma as described previously (17). Recombinant tPA (Activase) was generously provided by Dr. Gordon Vehar (Genentech, Inc., South San Francisco, CA). Recombinant soluble thrombomodulin (Solulin) was a generous gift of Dr. John Morser (Berlex Biosciences, Inc., Richmond, CA). For immunoadsorption chromatography, a monoclonal antibody raised against purified TAFI (monoclonal antibody 16) (11) was coupled to CNBr-activated Sepharose 4B. Synthetic 75% phosphatidylcholine, 25% phosphatidylserine (PCPS) vesicles were prepared as described previously (18).

Materials
Reverse Transcriptase-PCR Analysis-Human liver biopsy specimens were obtained with institutional approval from 9 individuals undergoing liver resection. All samples were snap-frozen in liquid nitrogen and stored at Ϫ70°C. RNA was isolated from the liver tissue using previously published methods (19,20), and first strand cDNA was synthesized using random hexanucleotide primers and avian myeloblastosis virus reverse-transcriptase according to the manufacturer's specifications (Promega). Approximately 1 g of the resultant cDNA from the libraries was subjected to PCR using Platinum Pfx polymerase according to the manufacturer's specifications (Invitrogen). The primer pair used corresponded to primers 5 (5Ј-GCATACATCAGCATGCATTC-3Ј) and 6 (5Ј-CAATGATTTGGTCTTGCTGG-3Ј) described by Boffa et al. (7) and resulted in the amplification of a 425-bp product. The cycling conditions were as follows: 5 min at 94°C, followed by 40 cycles of 20 s at 94°C, 20 s at 52°C, and 30 s at 68°C, followed by a final 5 min at 68°C. The PCRs were extracted with phenol/chloroform/isoamyl alcohol (25:24:1) and then with chloroform and then were precipitated with ethanol. Approximately 0.2-1 g of the precipitated PCR product was subjected to digestion with SpeI (New England Biolabs) (5 units) for 2 h at 37°C, and the digestion products were resolved on a 1% agarose gel containing 0.5 g/ml ethidium bromide.
Expression and Purification of Recombinant TAFI Variants-TAFI expression plasmids containing cDNA clones corresponding to the four TAFI variants were assembled in the mammalian expression vector pNUT using methods analogous to those described previously (7).
Briefly, a cDNA fragment encoding the Ala variant at position 147 (11) was amplified from a human liver first strand cDNA library using primers 3 (5Ј-CTTGCTGGCAGACGTGGAAG-3Ј) and 4 (5ЈGCTGG-GAGTATGAATGCATG-3Ј) (7). A cDNA fragment encoding the Ile variant at position 325 was amplified from a human liver first strand cDNA library using primers 5 and 6 (see above). The PCR products were individually cloned into pBluescript II SKϩ (Stratagene, La Jolla, CA) and subjected to DNA sequence analysis. The clones were identical in sequence to the TAFI cDNA reported by Eaton et al. (21) with the exception of the respective single nucleotide substitutions at the codons corresponding to amino acid positions 147 and 325. The variant fragments were inserted into the wild-type TAFI cDNA in the context of the TAFI-SK plasmid (7) using the BglII and SphI and the SphI and XhoI restriction sites, respectively. The resultant three variant full-length TAFI cDNA sequences were then excised by digestion with XbaI and XhoI. The ends were made blunt using a Klenow fragment of Escherichia coli DNA polymerase I and were inserted into pNUT that had been digested with SmaI. The nomenclature for the four TAFI variants is as follows: TAFI-TT, TAFI (Thr-147/Thr-325) (corresponds to the original cDNA isolated by Eaton et al. (21)); TAFI-AT, TAF1 (Ala-147/ Thr-325); TAFI-TI, TAFI (Thr-147/Ile-325); TAFI-AI, TAF1 (Ala-147/ Ile-325). Baby hamster kidney cells were transfected with the respective expression plasmids, using calcium-phosphate precipitation (22). Stably expressing lines were selected by culturing cells in the presence of 400 M methotrexate. For recombinant TAFI expression, stably expressing lines were cultured in triple flasks (500 cm 2 ; Nunc, Roskilde, Denmark) in Opti-MEM containing 1% (v/v) penicillin/streptomycin/ fungizone and 40 M ZnCl 2 . Conditioned medium was harvested at 48-h intervals and replaced with fresh medium. Harvested medium was centrifuged at 1000 ϫ g for 10 min, supplemented with Tris-HCl, pH 8.0 (to 5 mM), reduced glutathione (to 0.5 mM), and dEGRck (to 2 M) and stored at Ϫ20°C.
To isolate the TAFI variants, typically 1.5 liters of conditioned, supplemented medium was thawed at room temperature, passed through a 0.22-m filter, and passed over a 3.0-ml monoclonal antibody 16-Sepharose 4B column (6 mg of antibody/ml) at 4°C. The column was washed extensively with 20 mM HEPES, 150 mM NaCl, 0.01% Tween 80, at pH 7.4 (HBS/Tween 0.01%). TAFI was eluted with 1.0-ml fractions of 0.2 M glycine, pH 3, into 1-ml aliquots of 1 M Tris-HCl, pH 8. Protein-containing fractions were pooled and concentrated 10-fold in the Ultrafree-4 centrifugal filter device at 4°C. Concentrated TAFI was diluted 10-fold with HBS/Tween 0.01% and re-concentrated as before. This process was repeated 5 times before the TAFI was removed and passed through a 0.22-m filter. Purified TAFI was quantified by measurement of absorbance at 280 nm (⑀ 1%,280 ϭ 26.4; M r ϭ 60,000) (1) and stored at Ϫ20°C.
Activation of TAFI and Characterization of TAFIa Hydrolysis of AAFK-For routine activation of TAFI to TAFIa, TAFI (1 M) was incubated with thrombin (25 nM), Solulin (100 nM), and CaCl 2 (5 mM) in HBS/Tween 0.01% at 24°C for 10 min. Where appropriate, the thrombin was quenched with PPAck (1 M) before the mixture was placed on ice.
AAFK was dissolved in HBS/Tween 0.01%, protected from light, and stored at 4°C until use. The concentration of AAFK was quantified by absorbance at 349 nm (⑀ ϭ 18,400 M Ϫ1 cm Ϫ1 ) (16). To determine the kinetics of TAFIa hydrolysis of AAFK, TAFIa was formed as described, diluted, and placed on ice. TAFIa (25 nM) was then incubated with AAFK at various concentrations (0 -4000 M) at 24°C in a microtiter plate that had been presoaked in HBS with 1% Tween 80 (HBS/Tween 1%) and thoroughly rinsed with deionized distilled water. The absorbance of the mixtures at 350, 375, and 400 nm were monitored over time using a Spectramax Plus plate reader (Molecular Devices, Sunnyvale, CA). Wavelengths longer than 350 nm were required for higher concentrations of AAFK to compensate for absorbances that exceed the measurable absorbance range of the instrument. Initial rates were determined using the first 20% of the total potential change in absorbance that would have occurred with complete substrate hydrolysis and were converted to moles of AAFK hydrolysis/s/mol of TAFIa (s Ϫ1 ). The k cat and K m values for each variant were determined by nonlinear regression of the data to the Michaelis-Menten equation using Sigmaplot 4 (SPSS Inc., Chicago IL).
Activation of TAFI by Thrombin and Solulin-TAFI at various concentrations (0 -2.0 M) was incubated with Solulin at various concentrations (0 -50 nM) in the presence of CaCl 2 (5 mM) and thrombin (0.5 nM). Reactant mixtures (40 l) were incubated at 24°C for 10 min in a microtiter plate that had been presoaked with HBS/Tween 1% and thoroughly rinsed with deionized distilled water. Reactions were then quenched with AAFK (120 M) and PPAck (1 M) to a final volume of 200 l. The reactions were monitored at 350 nm, and the initial rate of hydrolysis was determined from the initial slope of the absorbance versus time relationship. To quantify the TAFIa formed, standard curves with TAFIa and TAFI for each variant were measured with AAFK (120 M) and PPAck (1 M). The TAFI zymogen shows modest activity toward the AAFK substrate (2% compared with TAFIa). Thus, the following analysis was used: K 1 is the slope determined from the rate of hydrolysis of AAFK versus the concentration of TAFI, and K 2 is the slope determined from the rate of hydrolysis of AAFK versus the concentration of TAFIa. The initial rate of hydrolysis (r ϭ dA/dt) is shown in Equation 1, Because the total concentration of TAFI, , this relation can be re-arranged as shown in Equation 2, Calculated [TAFIa] was corrected for dilution and converted to (TAFIa formed per s per thrombin (s Ϫ1 )). The complete data set was fit globally to the model (Equation 2) described previously (5), where [TM] is the concentration of Solulin, and K d represents the interaction between thrombin and thrombomodulin as shown in Equation 3, The calculated k cat , K m , and K d values are reported with the S.E. of the regression calculated by the regression algorithm (Nonlin module of SYSTAT 9, SPSS Inc., Chicago, IL). TAFI-deficient Plasma-To make TAFI-deficient plasma, 150 ml of fresh-frozen, citrated human plasma was thawed and passed over a 1-ml monoclonal antibody 16-Sepharose 4B column at room temperature. The plasma was passed over the column 3 times, and between each pass the column was washed copiously with HBS/Tween 0.01%; bound TAFI was eluted with 0.2 M glycine, pH 3, and the column was re-equilibrated with HBS/Tween 0.01%. To be certain the plasma was TAFI-deficient, a clot lysis assay (see below) was performed in the absence of TAFI and either in the absence or presence of Solulin (10 nM). There was no difference between the clot lysis times, and the plasma was considered to be TAFI-deficient.
Clot Lysis Assays-All clot lysis assays were performed as described previously (8) in a final volume of 120 l in a plastic microtiter plate presoaked with HBS/Tween 1% and thoroughly rinsed with deionized distilled water. In each reaction, TAFI-deficient plasma was diluted 1:3 in HBS/Tween. Reactions were initiated by the addition of thrombin (5 nM), CaCl 2 (10 mM), and tPA (0.3 nM). In lysis assays containing TAFIa, the zymogen was activated for 10 min as described previously and placed on ice, with all dilutions made on ice. In some lysis assays, TAFI zymogen was added in the presence of Solulin (10 nM), whereas other lysis assays were performed in the presence of TAFI zymogen, in the absence of Solulin, and the presence of PCPS (20 M). Clot lysis was monitored by the change in turbidity of each reaction at 400 nm in a Spectramax Plus plate reader (Molecular Devices), and the time to 50% lysis was determined graphically as the midpoint between maximum turbidity of a clot and the minimum absorbance when a clot was completely lysed. All reactions were performed at 37°C, and the microtiter plates were covered to minimize evaporation.
Effect of TAFIa Variants on the Kinetics of Plasminogen Activation-To determine whether the prolongation in clot lysis time caused by the variation at position 325 in TAFI reflects differences in plasminogen activation, we compared the ability of TAFIa-TT and TAFIa-TI in terms of their ability to suppress plasminogen activation. Plasminogen activation was monitored by measuring fluorescence changes of Gluplasminogen with fluorescein-labeled cysteine in place of the active site serine (23). Fluorescent measurements were made using a Spectramax Gemini XS fluorescence plate reader (Molecular Devices) with excitation and emission wavelengths set to 490 and 535 nm, respectively, with a 530 nm cut-off filter in the emission path. Fluorescein-plasminogen (f-plasminogen, 0.22 M) and fibrinogen (3.33 M) were incubated with various concentrations of activated TAFI-TT or TAFI-TI (0 -11.1 nM) in the absence or presence of 11.1 nM native Glu-plasminogen in 90 l of HBS/Tween. The temperature was set to 25°C. Fluorescence intensities were monitored for 10 min at 1-min intervals to obtain a starting value, after which clotting and plasminogen activation were initiated by the addition of 10 l of 100 nM thrombin and 3 nM t-PA, respectively. Decreases in fluorescence were monitored for 6 h and converted to plasminogen concentrations. Rate constants for plasminogen activation were calculated from the plasminogen concentrations at successive 1-min intervals as described by Wang et al. (6) using Equation 4, where [Pg] 1 and [Pg] 2 are the concentrations of plasminogen at time 1 (t 1 ) and time 2 (t 2 ), respectively. Determination of Free Lysine Released by TAFIa Variants-To determine whether activated TAFI-TT and TAFI-TI differ with respect to their ability to release lysine from partially degraded fibrin, free lysine was measured during plasminogen activation in the presence of one of these TAFI variants. Plasminogen activation was initiated and monitored, as described above, in the presence of 0.5 nM activated TAFI-TT or TAFI-TI, and the samples were thermostated to 25°C. At intervals, clots were solubilized, and reactions were quenched with acetic acid (0.1 M final). Solutions were deproteinated with 0.2 M perchloric acid (final concentration) followed by centrifugation at 16,000 ϫ g for 5 min. The supernatants were neutralized with potassium hydroxide, and the insoluble potassium perchlorate was removed by placing the samples on ice for 10 min followed by centrifugation. The concentration of free lysine was determined enzymatically using methods described by Wang et al. (6). Briefly, 100 l of deproteinated clot supernatant was incubated with 80 l of 25 M NADH (Roche Molecular Biochemicals) and 2.5 mM ␣-ketoglutaric acid (Sigma) in 0.05 M HEPES, 150 mM NaCl, pH 7.0. The reaction was initiated by the addition of 0.1 units of saccharopine dehydrogenase (Sigma). Oxidation of NADH was monitored using a Spectramax Gemini XS fluorescence plate reader with excitation and emission wavelengths set to 340 and 450 nm, respectively, with a 435 nm cut-off filter in the emission beam. The concentration of free lysine in clot supernatants was determined by comparing decreases in NADH fluorescence after 4 h with those obtained in experiments using known concentrations of L-lysine (Sigma).

Identification of a Novel Amino Acid Sequence Variant of
TAFI-DNA sequence analysis of reverse transcriptase-PCR products obtained from a number of different human liver cDNA libraries revealed a single nucleotide difference from the cDNA sequence published by Eaton et al. (21). This was a C to T mutation at position 1057 of the TAFI cDNA (numbering is as per Boffa et al. (24)) and would result in the conversion of a Thr codon (ACU) to an Ile codon (AUU) at amino acid position 325. Because the mutation also resulted in the destruction of a SpeI restriction site (ACTAGT to ATTAGT), we amplified a fragment of the TAFI cDNA from liver cDNA libraries corresponding to nine individuals and subjected the PCR products to SpeI digestion followed by agarose gel electrophoresis to determine the genotype of these individuals with respect to the mutation. In five cases, the 425-bp PCR product was completely digested with SpeI resulting in 298and 127-bp fragments. In four cases, however, ϳ50% of the PCR product was digested under identical conditions; a representative agarose gel containing digestion products for seven individuals is shown in Fig. 1. These findings indicate that five of the individuals are homozygous for the wild-type (Thr-325) allele, whereas four individuals were heterozygotes. In this small sample, no individuals were identified that are homozygous for the mutant (Ile-325) allele. However, Brouwers et al. (15) have recently determined the frequency of this allele in a much larger population of 152 blood donors and found that 17 individuals were homozygous for the Ile-325 allele, whereas 76 individuals were heterozygous.
Isolation of the Recombinant TAFI Variants-If we consider the Ile/Thr polymorphism at position 325 and the previously described Thr/Ala polymorphism at position 147, there are four possible variants of TAFI. Therefore, these variants were constructed, stably expressed in baby hamster kidney cells with serum-free medium, and were isolated from the medium by affinity chromatography on a column with covalently attached monoclonal antibody to TAFI. The isolated variants all migrated as a single band at ϳ60 kDa in SDS-PAGE. In addition, each was fully cleaved to yield the 35-kDa activated enzyme described previously (1, 5, 7) after a 10-min incubation with thrombin/thrombomodulin (data not shown).
AAFK Hydrolysis by TAFIa-AAFK is a substrate shown previously to be hydrolyzed by pancreatic carboxypeptidase B (16). We found that the substrate can also be hydrolyzed by TAFIa, resulting in greater than 95% loss of absorbance at its 348.5 nm absorbance maximum. Because the rate of substrate hydrolysis is linear with respect to TAFIa concentration, AAFK can be used to quantify TAFIa activity in a sample.
To determine the kinetics of AAFK hydrolysis by the TAFIa variants, each variant was activated, diluted, and incubated with AAFK at various concentrations. The estimated Michaelis-Menten parameters for the variants are summarized in Table I. Each TAFIa variant hydrolyzed AAFK similarly, with k cat values ranging between 27 and 30 s Ϫ1 , and K m values ranging between 1.0 and 1.3 mM. This indicates that the conformations of the active sites of the variants are probably not altered by the variation.
We also found that AAFK can be hydrolyzed by the TAFI zymogens with roughly 2% of the activity of the enzyme. AAFK is stable under the conditions used in these carboxypeptidase assays, and we measured no hydrolysis of the substrate (loss of absorbance) in the absence of the TAFI zymogen. We do not believe that this activity is due to background TAFIa contamination because activity of the zymogen remains even if a sample of TAFI is incubated at 37°C for an hour, which would inactivate most of the TAFIa present (8). Also, we do not see activity of the zymogen when using N- [3-(2-furylacryloyl)]-Lalanyl-L-arginine as a substrate. Therefore, to quantify TAFIa activity in the solutions containing both TAFI and TAFIa, we measured activity of both enzyme and zymogen independently and calculated the [TAFIa] formed using Equation 2, see under "Experimental Procedures." Thermal Stabilities of the TAFIa Variants-We compared the thermal stability of each variant at four temperatures. Representative plots for TAFIa-TT and TAFIa-TI are shown in Fig. 2. Table II summarizes the thermal stability of the four variants. Although the variation at position 147 does not have a significant effect on the thermal stability of TAFIa, the presence of an Ile at position 325 nearly doubles the thermal stability of TAFIa over the range of temperatures indicated in Table II. At 37°C the half-lives of TAFIa-TI and TAFIa-AI are 15 min compared with 8 min for TAFIa-TT and TAFIa-AT. The difference in thermal stability conferred by this variation in TAFI may also resolve some discrepancy in the half-lives for plasma-derived TAFI which have been reported as both 8 -9 (7, 8) and 15 min (9) at 37°C.
Activation of the TAFI Variants by Thrombin and Solulin-To determine whether the variation at position 325 could affect the kinetics of activation, each variant was activated by thrombin and Solulin. The k cat , K m , and K d parameters were determined by fitting the data to the model for TAFI activation described previously (5). Fig. 3 shows representative plots of the activation of TAFI-TT and TAFI-TI. Similar experiments were performed with each of the variants, and the estimated kinetic parameters are summarized in Table III. The model derived for TAFI activation closely fits the data for each variant activation, and neither the variation at position 147 nor the variation at position 325 has a substantial effect of the kinetic parameters. We found that k cat /K m values ranged between 0.9 and 1.3 M Ϫ1 s Ϫ1 .
Attenuation of Clot Lysis by the TAFI Variants-Our previous work (7,8) indicated that differences in thermal stability correspond to differences in the ability of TAFI to attenuate clot lysis. Because we have identified a natural variant that has an increased thermal stability, it seemed likely that there would be a corresponding increase in the ability of the more stable variants to attenuate clot lysis.
To determine whether the difference in the stability of TAFIa corresponds to a change in clot lysis time, we measured the attenuation of clot lysis by each variant in vitro. Fig. 4 shows the prolongation of clot lysis time by each variant at various concentrations in the presence of 10 nM Solulin. Under these conditions, the attenuation of clot lysis is half-maximal at a concentration of about 2 nM TAFI zymogen. TAFI-TI and TAFI-AI can increase the maximum attenuation of clot lysis by about 60% over that obtained with TAFI-TT and TAFI-AT. The difference in lysis times persist even at high concentrations of the variants; thus, the lower antifibrinolytic activity of the Thr-325 variants compared with the Ile-325 variants cannot be overcome by using high concentrations of the Thr-325 variants. As predicted by the half-life data, the variation at position 147 has little effect on the ability of the TAFI variants to attenuate clot lysis. Fig. 5 presents a similar experiment using TAFIa instead of the zymogen for each variant. TAFIa has a similar effect on the attenuation of clot lysis in that a half-maximal effect is observed at about 2 nM TAFIa. At higher concentrations, TAFI-TI and TAFI-AI can prolong clot lysis 60% more than TAFI-TT and TAFI-AT. Each variant reaches a maximum attenuation of clot lysis with respect to TAFI variant concentration, and this maximum depends on the residue at position 325. Whereas the tendency to reach a maximum time to clot lysis with TAFIa is similar to the tendency with TAFI activated in situ by throm-  1. Demonstration of the existence in the human population of the Thr/Ile-325 polymorphism. First strand human liver cDNA libraries were subjected to PCR to amplify a fragment encompassing the codon for the amino acid at position 325. The PCR products were digested with SpeI and were resolved by electrophoresis on a 1% agarose gel. Ethidium bromide was included in the gel to allow visualization of DNA bands. A 100-bp ladder was included on the gel as a size marker. The results show that three of seven individuals were homozygous for the wild-type Thr-325 allele, whereas four were heterozygotes. An additional two individuals (not shown on this gel) were also found to be homozygous for the wild-type allele.
bin/thrombomodulin (Fig. 4), the lysis times in the presence of the TAFI variant enzymes are marginally smaller. This probably reflects the lag time to complete TAFI activation in the results of Fig. 4, which is not a factor in this case. Nevertheless, the similarity between the behavior of pre-activated TAFI as compared with TAFI zymogen added in the presence of thrombin and Solulin demonstrates that the attenuation of clot lysis is not a result of differences in the rate of activation of the variants by thrombin/thrombomodulin. Fig. 6 shows clot lysis experiments performed in the absence of thrombomodulin, using the variant TAFI zymogens and PCPS (20 M) to enhance endogenous prothrombin activation and thereby partially activate TAFI (albeit to a lesser degree than in the presence of thrombomodulin). TAFI-TI and TAFI-AI had a 30 -50% greater antifibrinolytic effect than either TAFI-TT or TAFI-AT. The concentration of TAFI required for half-maximal effect is increased 15-fold (30 nM) for each variant as compared with lysis assays done in the presence of activated TAFI, which is in agreement with similar studies conducted previously (8) using TAFI-TT.
Finally, Fig. 7 shows a series of clot lysis assays performed in the presence of mixtures of TAFI-TT and TAFI-TI. With increasing amounts of the more stable TAFI-TI variant, there was a corresponding increase in the antifibrinolytic potential of Timed aliquots were removed, and the remaining TAFIa activity was determined using AAFK. The rates are reported as a fraction of the initial activity and were fit by nonlinear regression of the data to the equation for first order decay to determine the half-lives. Half-lives were estimated for each TAFI variant at least twice at 37 and 30°C, and the variation in the estimated values between trials was always less than 10%.

Half-lives of the TAFIa variants
The TAFI variants were activated with thrombin and Solulin at room temperature, and the thrombin was quenched with PPAck (1 M), and the TAFIa variants were placed in waterbaths at the indicated temperatures. Timed aliquots were removed, and the TAFIa activity was quantified with AAFK. Half-lives were determined by nonlinear regression of the data and indicated along with the S.E. provided by the regression algorithm.  Fig. 7 that plots the times to 50% clot lysis at 90 nM TAFI with respect to the proportion of each variant present. Therefore, in plasma with a mixture of TAFI mutants, the fraction of each mutant present may be more important in determining the antifibrinolytic potential than the total concentration of TAFI. Effect of TAFIa Variants on the Kinetics of Plasminogen Activation-Plasmin potentiates plasminogen activation in the presence of fibrin by partially degrading fibrin, thereby exposing new plasminogen-or t-PA-binding sites on fibrin (25). Recently, we demonstrated that high concentrations of TAFIa eliminate plasmin-mediated up-regulation in plasminogen activation (6). To determine whether variation at position 325 of TAFI affects its ability to suppress plasminogen activation, various concentrations of activated TAFI-TT or TAFI-TI were added to reactions containing fixed concentrations of fplasminogen, t-PA, fibrinogen, and thrombin. Concentrations of f-plasminogen were calculated from decreases in fluorescence, and rate constants for plasminogen activation were determined from these data. To determine the effect of native plasmin on plasminogen activation, experiments were performed in the absence or presence of Glu-plasminogen. Under the conditions outlined under "Experimental Procedures," plasmin increases the rate constant for plasminogen activation 3.5-fold (from 4 ϫ 10 Ϫ5 to 1.4 ϫ 10 Ϫ4 s Ϫ1 ), with a half-maximal effect 21 min after the TABLE III Kinetic parameters for the activation of the TAFI variants by thrombin and solulin Each variant was activated with Solulin at various concentrations and TAFI at various concentrations in the presence of thrombin (0.5 nM) and CaCl 2 (5 mM). The k cat , K m , and K d values were determined by nonlinear regression of the data as described previously (5). The values are reported Ϯ S.E. for each constant determined by the regression algorithm. initiation of plasminogen activation. This increase is consistent with that determined by Wang et al. (6). For comparison purposes, rate constants for plasminogen activation in the presence of increasing concentrations of activated TAFI-TT or TAFI-TI were determined at 21 min after the initiation of plasminogen activation. As can be seen in Fig. 8, each TAFIa variant decreases the rate constant in a concentration-dependent fashion and, at a high concentration, eliminates the increase in rate constant caused by plasmin. However, the variants differ with respect to their potencies of plasminogen activation suppression. Whereas 5.0 nM TAFIa-TT abolishes the increase in rate constant, only 0.5 nM TAFIa-TI is needed for the same effect.
Determination of Free Lysine Released by TAFIa Variants-To determine whether differences in plasminogen activation reflect the release of lysine from partially degraded fibrin, we compared the generation of free lysine by TAFIa-TT with that by TAFIa-TI during plasminogen activation. Fig. 9 demonstrates the generation of free lysine at various times when plasminogen was activated under the same conditions as those outlined for rate constant determination in the presence of 0.5 nM TAFIa-TT or TAFIa-TI. At each time point, TAFIa-TI mediated the release of ϳ2-fold higher concentrations of lysine than TAFIa-TT. In the presence of TAFIa-TI, free lysine concentrations maximized at 2.7 M, 150 min after the initiation of plasminogen activation. In contrast, free lysine concentrations reached a maximum of 1.4 M at 210 min in the presence of TAFIa-TT. DISCUSSION While screening human cDNA libraries, we found a single nucleotide polymorphism in TAFI cDNA clones that corresponds to a substitution of Ile for Thr at amino acid position 325 of the protein. This is the second variation to be found in human TAFI, with the first being a variation at position 147 (14) that was found to have little effect on its thermal stability or it antifibrinolytic potential. Previously we found that mutations in residues at positions 302, 320, and 330 can have a marked effect on the stability of TAFI, albeit a negative effect. Because the present variation lies in the middle of this region, we were particularly interested in characterizing the effect of the Ile-325 variation with respect to thermal stability and antifibrinolytic potential.
As expected from previous work (14), the variation at position 147 did not affect the behavior of TAFI; however, the presence of an Ile at position 325 nearly doubles the thermal stability of the enzyme. Again, we have found that altering a residue in the 302-330 region of TAFI can affect thermal stability, and the variation at position 325 could be particularly relevant since it occurs naturally. Most notably, the presence of the Ile at position 325 results in a 30 -60% greater antifibrinolytic effect for TAFIa than the Thr at this position.
The increased antifibrinolytic potential of Ile-325-containing TAFIa variants reflects the fact that these variants have an increased ability to suppress the up-regulation in plasminogen activation that normally occurs during fibrinolysis. In turn, differences in plasminogen activation are explained by TAFIamediated lysine release. Thus, TAFIa variants containing Ile at position 325 release twice as much lysine during fibrinolysis as variants containing Thr at this position. It is possible in vivo that if the ability to suppress plasminogen activation and the antifibrinolytic potential of TAFI were to be increased, the dynamic balance between accretion and lysis of a clot could be shifted toward accretion and therefore thrombosis.
Why this portion of TAFIa is so important for stability is not immediately clear; however, this stretch of amino acids differs substantially from that of the homologous carboxypeptidase B of the pancreas (21), which is stable. Previously, we found (8) that the TAFIa inactivation process is highly unfavorable enthalpically but highly favorable entropically. Possibly, the Arg to Gln mutations in TAFI eliminate charge interactions, which decrease the enthalpic component of the active/inactive transition and thereby destabilize TAFIa. In contrast, in the case of the Ile-325 variant of TAFI, changing the hydrophilic Thr residue to a hydrophobic Ile residue may diminish the effect of the favorable entropic change that accompanies the inactivation process, thereby stabilizing TAFIa.
Most likely, when tissue is damaged, physiologically relevant amounts of TAFIa can be generated from the plasma zymogen, either by small amounts (several nanomolar) of thrombin complexed with thrombomodulin or by large amounts (several hundred nanomolar) of thrombin generated after clotting via the intrinsic pathway. We were able to see a change in antifibrinolytic potential under conditions that simulate both proposed activation pathways. The main difference between the two was that in the absence of Solulin, the half-maximal effect occurred at much higher concentrations of the initial zymogen. Nevertheless, there was a significant prolongation in clot lysis time The turbidity of each clot was monitored at 37°C, and the times to 50% clot lysis are plotted as a function of the initial TAFI concentration. The inset graph shows the times to 50% clot lysis at 90 nM TAFI plotted as a function of fraction of the TAFI mixture consisting of TAFI-TI. Each point represents the mean of two independent experiments, and all measured values varied by less than 10% between trials. mediated by the more stable variants (TAFI-TI or TAFI-AI) as compared with the less stable variants (TAFI-TT or TAFI-AT). Thus, the more stable variants would be more antifibrinolytic regardless of whether thrombin or the thrombin-thrombomodulin complex is the activator. Under both types of conditions, a difference in the antifibrinolytic potential between the Thr/Ile-325 variants persists throughout a range of TAFI concentrations used.
All of the clot lysis experiments presented here demonstrate that increased stability of TAFI results in the increased ability to maintain a fibrin clot. This increase in clot lysis time cannot be overcome by adding increasing amounts of TAFI. Both zymogen and pre-activated TAFI tend toward saturation, and the lysis time when saturation is reached depends on the thermal stability of the TAFIa variant used. This phenomenon is not restricted to this study. Previously, we found (8) that less stability corresponded to a smaller maximum in the time of clot lysis, while similar to this study, increased stability conferred an increase in the maximum time to clot lysis. Because the kinetics of AAFK hydrolysis were similar for all of the variants, active site differences between variants are probably not significant in mediating the variant-dependent change in the saturation of time to clot lysis. To resolve why all TAFI variants do not reach the same maximum time to clot lysis even at higher concentrations, a thorough understanding of the dynamics of lysine residue exposure during clot lysis by plasmin, and subsequent removal by TAFIa, is required.
Finally, here we have used AAFK as a small substrate to quantify TAFIa activity. A limitation of the furylacryloyl substrate analogues is their meager loss of absorbance upon hydrolysis, typically about 15%. The large loss of absorbance by AAFK (about 95%) can improve the signal to noise ratio when quantifying TAFIa. In addition, the substrate has an absorbance maximum in the visible light range, and even wavelengths as high as 400 nm or greater are adequate for consistent measurement of carboxypeptidase B activity. An unexpected property of this analogue is the activity of the TAFI zymogen, which we found to have one-fiftieth the activity of TAFIa. Among the pancreatic carboxypeptidases that are evolutionarily related to TAFI, activity of the zymogen against small substrates has been observed for procarboxypeptidases A1 and A2 but not for procarboxypeptidase B (26,27). Based on x-ray crystallographic analysis of the respective zymogen structures, the interactions between the activation peptides and the latent active sites of procarboxypeptidase A1 and A2 result in conformations of the active site residues that are compatible with carboxypeptidase activity against substrates small enough to enter the partially occluded active site cleft (26,27). The activation peptide of procarboxypeptidase B, by contrast, completely blocks access to the critical S1 and S1Ј subsites of the latent active site (27). The activation peptide of TAFI may therefore primarily serve to occlude macromolecular substrates from the latent active site while still allowing a latent active site of reduced catalytic efficiency that is accessible to some small substrates.
In this study we have found that human TAFI can have either a Thr or an Ile residue at position 325, and this variation can significantly alter the behavior of the active enzyme. Ile confers stability to TAFIa, and this increased stability has an effect on the antifibrinolytic potential. Because this variability affects antifibrinolytic potential in vitro, the Thr/Ile variation at position 325 may account for a proportion of the variable risk for the thrombogenic disorders observed in human populations.