Generation of a Stable Activated Thrombin Activable Fibrinolysis Inhibitor Variant*

Activated thrombin activable fibrinolysis inhibitor (TAFIa), generated upon activation of TAFI, exerts an antifibrinolytic effect. TAFIa is a thermolabile enzyme, inactivated through a conformational change. The objective of the current study was to generate a stable variant of human TAFIa. Using a site-directed as well as a random mutagenesis approach to generate a library of TAFI mutants, we identified two mutations that increase TAFIa stability, i.e. a Ser305 to Cys and a Thr329 to Ile mutation, respectively. Combining these mutations in TAFI-Ala147-Ile325, the most stable isoform of TAFIa (half-life of 9.4 ± 0.4 min), revealed a TAFIa half-life of 70 ± 3.1 min (i.e. an 11-fold increase versus 6.3 ± 0.3 min for TAFIa-Ala147-Thr325, the most frequently occurring isoform of TAFI in humans) at 37 °C. Moreover, clot lysis (induced by tissue plasminogen activator) experiments in which TAFI-Ala147-Cys305-Ile325-Ile329 was added to TAFI-depleted plasma revealed a 50% clot lysis time of 313 ± 77 min (i.e. a 3.0-fold increase versus 117 ± 10 min for TAFI-Ala147-Thr325). The availability of a more stable TAFIa variant will facilitate the search for inhibitors and allow further structural analysis to elucidate the mechanisms of the instability of TAFIa.

The antifibrinolytic effect of TAFIa is half-maximal at 1 nM, whereas the plasma concentration of total TAFI ranges between 75 and 250 nM (11)(12)(13). Therefore, only a small fraction of TAFI in plasma needs to be activated to exert the maximal antifibrinolytic effect (observed at 20 nM TAFIa). Thus, increasing the concentration (above 20 nM TAFIa) does not result in an increased antifibrinolytic effect. Consequently, the maximal antifibrinolytic effect can only be further increased by increasing the TAFIa stability (6).
TAFIa and pancreatic carboxypeptidase B (pCPB) are highly homologous members of the carboxypeptidase A family. In contrast to TAFIa, pCPB is a stable enzyme. Although TAFIa and pCPB share 48% sequence identity, the sequence identity in the 300 -330 amino acid region is only 13%.
Two polymorphisms in the TAFI coding region, resulting in two amino acid variations (i.e. A147T and T325I), have been reported (14,15). The allele frequencies of the corresponding four single nucleotide polymorphisms are 0.72, 0.28, 0.64, and 0.36 for the Ala 147 -, Thr 147 -, Thr 325 -, and Ile 325 -encoding alleles, respectively (15). However, Ile at position 325 extends the half-life of TAFIa from 8 to 15 min regardless of the residue at position 147 (6). The increased stability of the Ile 325 variants is associated with a 1.6-fold higher antifibrinolytic effect, indicating that the thermal (in)stability is an important determinant for the antifibrinolytic potential of TAFIa. To date, this T325I polymorphism is the only known amino acid variation that increases the TAFIa half-life. Attempts have been undertaken to increase the stability of TAFIa by mutagenesis (7) or by construction of chimeric variants (16). Although some variants exhibit an increased TAFIa stability (16), this effect was never associated with an increased antifibrinolytic effect. The objective of the current study was to create a more stable TAFIa variant that retains its antifibrinolytic potential.

EXPERIMENTAL PROCEDURES
Materials-Recombinant TAFI isoforms (Ala 147 -Thr 325 (TAFI-AT), Ala 147 -Ile 325 (TAFI-AI)) and monoclonal antibodies raised against TAFI purified from plasma were generated as described before (17). Oligonucleotides for mutagenesis and sequencing were purchased from Sigma-Genosys. dNTP mix was obtained from Roche Applied Science. Mutagenesis was performed using the GeneMorph TM II EZClone domain mutagenesis kit or the QuikChange TM XL site-directed mutagenesis kit (both from Stratagene). PfuTurbo TM polymerase was obtained from Stratagene. PCR reactions were carried out with the Mastercycler Gradient from Eppendorf. Plasmid DNA purification was performed with the Nucleobond TM AX500 kit (Machery-Nagel) or the FastPlasmid TM mini kit (Eppendorf). DNA was sequenced with the Automated Laser Fluorescent ALF apparatus according to the protocol for the ALFexpress TM Autoread Sequencing kit (both from Amersham Biosciences).
Citrated plasma of 23 healthy individuals was pooled for clot lysis experiments. TAFI-depleted plasma was obtained by adsorption on monoclonal antibodies T4E3, raised against human plasma-derived TAFI, and covalently coupled to CNBr-activated Sepharose 4B as described (17).
Determination of TAFIa Stability-The TAFIa activity was measured as described earlier (13). TAFI (90 nM) was activated with thrombin (20 nM), thrombomodulin (5 nM), and CaCl 2 (5 mM) in 60 l of Hepesbovine serum albumin (BSA) buffer (25 mM Hepes, 137 mM NaCl, 3.5 mM KCl, 3 mM CaCl 2 , pH 7.4, containing 0.1% BSA) for 2 min at 37°C. Thrombin-induced activation was stopped at the indicated time points by the addition of 20 l of PPACK (30 M final concentration). Samples were incubated at 37°C for different time intervals. Twenty l of the substrate Hip-Arg (5 mM final concentration) was added, and substrate conversion was allowed to proceed for 10 min at 22°C. Reactions were stopped by the addition of 20 l of 1 N HCl followed by 20 l of 1 N NaOH and 25 l of 1 M sodium phosphate buffer (pH 7.4). Subsequently, 30 l of 6% cyanuric chloride dissolved in 1,4-dioxane was added, and the mixtures were vortexed and centrifuged. One hundred microliters of supernatant was transferred into a 96-well microtiter plate, and the absorbance at 405 nm was measured using an EL808 Ultra Microplate Reader (Bio-Tek instruments Inc). The calculated activity was expressed as a fraction of the initial activity (i.e. of the sample that was not incubated at 37°C after the addition of PPACK). Data were analyzed according to a one-phase exponential decay function. Decay constants (K) were used to calculate the half-lives according to the equation K ϭ 0.693/t1 ⁄ 2 .
Inhibition of TAFIa Activity by PTCI-TAFI (90 nM) was activated with T/TM for 2 min as described above. Reactions were stopped with PPACK (30 M) followed by 10 min of incubation at 22°C with PTCI (0.05-5 M) before the addition of the chromogenic substrate (Hip-Arg) for determination of TAFIa activity as described above.
TAFI Activation by Thrombin/Thrombomodulin-TAFI (0.06 -1 M) was activated for different time intervals (ranging between 0 and 4 min) in the presence of CaCl 2 (5 mM), thrombin (4 nM), and thrombomodulin (1 nM) in Hepes-bovine serum albumin buffer. Under these conditions less than 20% of total TAFI was converted to TAFIa, and rates of activation were linear over the time of measurements. Activation was stopped by the addition of PPACK (30 M). Samples were incubated for 10 min at 25°C with Hip-Arg. The amount of TAFIa generated after each activation time was obtained using standard calibration curves (absorbance/min/nM TAFIa). The k cat and K m values for activation of each TAFI variant were determined by nonlinear regression of the data to the Michaelis-Menten equation using Graph Pad Prism 4.01.
TAFIa Hydrolysis of Hippuryl-L-arginine-Characterization of TAFIa hydrolysis of Hip-Arg was determined as described previously with some modifications (5). Briefly, TAFI (1 M) was incubated for 15 min at 22°C in the presence of CaCl 2 (5 mM), thrombin (20 nM), and thrombomodulin (80 nM) in Hepes buffer. Under these conditions, TAFI is quantitatively converted to TAFIa without conversion to TAFIai (confirmed with SDS-PAGE analysis in parallel with measurement of TAFIa activity). Thrombin activity was stopped by the addition of PPACK (30 M). Hydrolysis of Hip-Arg (0.1-2 mM) was monitored at 254 nm in a Powerwave X spectrophotometer thermostatted at 25°C (Bio-Tek Instruments, Inc). Reactions were performed in half area UV 96-well microtiter plates (Elscolab). For each substrate concentration, the rate of hydrolysis was linear over the time of measurement, and initial rates were determined under conditions where less than 10% of the substrate was hydrolyzed. The formation of hippuric acid was calculated from the change in absorbance at 254 nm using the change in extinction coefficient that occurs upon hydrolysis of Hip-Arg (⌬ Hip-Arg 3 hippuric acid ϭ 0.524 mM Ϫ1 cm Ϫ1 ) as published previously (5). Kinetic constants for the hydrolysis of Hip-Arg by the different TAFI variants were obtained by fitting the rates of substrate hydrolysis to the Michaelis-Menten equation using Graph Pad Prism 4.01.
Clot Lysis Assay-A series of identical clots was formed: TAFI (0.7-90 nM in Hepes buffer, Hepes 20 mM, pH 7.4), tissue plasminogen activator (40 ng/ml in Hepes buffer containing 0.1% Tween 20), thrombomodulin (5 nM in Hepes buffer), thrombin (20 nM in Hepes buffer), and CaCl 2 (12.5 mM in Hepes buffer) were added to TAFI-depleted plasma ( 1 ⁄ 2 final dilution). The change in turbidity was followed every 5 min at 405 nm (during a 7-h time course) with an EL808 Ultra Microplate Reader (Bio-Tek instruments Inc) to determine the 50% clot lysis time (i.e. the time from the maximum turbidity to the midpoint between maximum and return to base line). Additional experiments were performed in which clot formation and dissolution were monitored in the presence of 4 M PTCI.
Statistical Analysis-Quantitative data were summarized by the mean and S.D. Statistical analyses were performed with Graph Pad Prism 4.01 using the unpaired t test (Graph Pad Software, Inc., San Diego, USA). p values less than 0.05 were considered statistically significant.

Generation of TAFI Mutants; Targeted Strategy-
The rationale for the site-directed mutagenesis was to hamper the conformational change of TAFIa, which consequently leads to the inactivation of TAFIa followed by an increased susceptibility to cleavage at position Arg 302 . Therefore, connecting the 25-and 11-kDa moieties by disulfide bridges (formed by introducing two new Cys residues in the TAFIa molecule) seems to be a reasonable approach. Thus, 25 single and double Cys point mutants were constructed in the Glu 99 -Tyr 101 , Ala 167 -Leu 189 , Ile 294 -Ala 315 , and Glu 376 -Ile 398 regions.
These mutants were generated in pcDNA-TAFI-AT and transiently transfected in eukaryotic cells. Conditioned media of the different TAFI mutants were screened for TAFI protein expression, TAFIa activity, and TAFIa stability (data not shown). Only one TAFIa variant, carrying a Ser 305 to Cys mutation, had an increased half-life at 37°C (i.e. 5.8-fold increased versus TAFIa-AT). The introduction of all other mutations in the TAFIa molecule revealed no significant effect on TAFIa stability except for the Trp 171 to Cys mutant, which exhibited a 2.4-fold increased TAFIa half-life versus TAFIa-AT.
Generation of TAFI Mutants; Random Strategy-A library of random TAFIa-TI mutants was generated and transiently transfected in eukaryotic cells. Initially, conditioned media of 250 clones were screened for TAFI protein expression, TAFIa activity, and TAFIa stability (data not shown). This resulted in the identification of a Thr 329 to Ile mutation, revealing a 1.7-fold increased TAFIa half-life at 37°C. Subsequent to these findings the Cys 305 mutation (cfr targeted strategy) and the Ile 329 mutation (cfr random strategy) were introduced in TAFI-AI either individually or combined, resulting in TAFI-AI-Cys 305 , TAFI-AI-Ile 329 ,  TAFI-AI-Cys 305 -Ile 329 . TAFI-AT, TAFI-AI, TAFI-AI-Cys 305 , TAFI-AI-Ile (Table 1; Fig. 1).
TAFI Activation by Thrombin/Thrombomodulin-The ability of the T/TM complex to activate the different TAFI variants (0.06 -1 M) was evaluated. The initial rates of TAFIa formation were assessed by measurement of Hip-Arg hydrolysis. Double-reciprocal plots of activation rate versus substrate (TAFI) concentration were linear (r 2 ϭ 0.98), confirming Michaelis-Menten kinetics. The kinetic parameters were determined by non-linear regression of the data to the Michaelis-Menten equation. The combination of a more than 2-fold increased affinity (K m ) ( Table 2) and an almost 2-fold increased catalytic rate (k cat ) lead to 3-6-fold increased catalytic efficiency (k cat /K m ) of TAFI activation by T/TM for the TAFI variants bearing Cys 305 (compared with TAFI-AT). TAFI-AI and TAFI-AI-Ile 329 revealed a 2-fold higher catalytic efficiency for activation by T/TM compared with TAFI-AT.
TAFIa Hydrolysis of Hippuryl-L-arginine-To compare the enzymatic properties of the different TAFIa variants, the ability of the respective enzymes to hydrolyze Hip-Arg (0.1-2 mM) was evaluated. Double reciprocal plots of hydrolysis rate versus substrate (Hip-Arg) concentration were linear (r 2 ϭ 0.95), confirming Michaelis-Menten kinetics. The kinetic parameters were determined by non-linear regression of the data to the Michaelis-Menten equation. The catalytic efficiency (k cat /K m ) ( Table 3) for converting the substrate Hip-Arg by TAFIa-AI-Ile 329 , TAFIa-AI-Cys 305 , and TAFIa-AI-Cys 305 -Ile 329 was significantly lower than these for TAFIa-AT and TAFIa-AI (p Ͻ 0.05) due to a strongly decreased catalytic rate (k cat ).
Fragmentation Pattern of TAFI Variants upon Activation with Thrombin/Thrombomodulin-Upon activation of TAFI-AT, a 36-kDa band corresponding to TAFIa is generated after 5 min at 37°C (Fig. 2A). Subsequently, after 20 min, the typical degradation products of 25 and 11 kDa are formed. The activation peptide of TAFI has a theoretical M r of 19.4 kDa (with carbohydrate). In agreement with the literature, no distinct activation peptide after activation of TAFI was detected using SDS-PAGE and silver staining (8,19,20). Activation of TAFI-AI and TAFI-AI-Ile 329 reveals a similar fragmentation pattern (data not shown). Upon activation of TAFI-AI-Cys 305 (data not shown) and TAFI-AI-Cys 305 -Ile 329 (Fig. 2B), a 36-kDa band corresponding to TAFIa is generated after 5 min at 37°C. In contrast to the other TAFI variants, the formation of the 25-and 11-kDa degradation products is delayed as could be expected from their prolonged half-life.

DISCUSSION
TAFIa (36 kDa) is formed upon activation of TAFI (56 kDa) and exerts an antifibrinolytic effect. Regulated by an intrinsic temperaturedependent instability (i.e. half-life between 5 and 15 min at 37°C), TAFIa converts into an inactive conformation that is prone to further proteolytic cleavage resulting in the formation of 25-and 11-kDa fragments. The exact nature of the TAFIa instability is still unknown. However, it is shown that TAFIa variants bearing Ile at position 325 have a 2-fold increased half-life at 37°C that results in a 1.6-fold higher antifibrinolytic effect (6). Boffa et al. (7) demonstrated that replacement of Arg 302 , Arg 320 , and Arg 330 by Gln decreases TAFIa stability. We hypothesized that this region, which bears only 13% sequence identity with pCPB, a stable protease with an overall sequence identity of 48% with TAFIa, determines TAFIa stability. Marx et al. (16) created a TAFI-pCPB chimera in which the 293-401-amino acid region of TAFI was substituted with the corresponding amino acids of pCPB. Although this chimera revealed an increased TAFIa stability, it lost most of its antifibrinolytic potential. Based on the three-dimensional structure of pCPB, Barbosa-Pereira et al. (21) proposed a TAFI model and suggested that Ile 182 and Ile 183 contribute to the instability of TAFIa. However, recent data of Marx et al. (9) revealed that replacing Ile 182 by Arg and Ile 183 by Glu (as in pCPB) does not result in an increased TAFIa stability. In contrast, mutating these residues resulted in a lower rate of TAFI activation and, in accordance, in a 6-fold reduced antifibrinolytic potential.
The goal of the current study was to create a TAFIa mutant with an increased half-life at 37°C and a concomitant increased antifibrinolytic effect. Therefore, we combined two different approaches. First, sitedirected mutagenesis in well defined regions yielded a TAFI mutant (TAFI-AT-Cys 305 ) with a 5.8-fold increased TAFIa half-life upon activation. Second, random mutagenesis identified a 1.7-fold stabilizing effect of Ile at position 329. These two stabilizing mutations, either individually or combined, were introduced in the TAFI isoform, which yields the most stable TAFIa (i.e. TAFI-AI).
Combining Cys 305 , Ile 325 , and Ile 329 revealed a synergistic effect on   both TAFIa half-life (i.e. 11-fold increased at 37°C) and 50% clot lysis time (3.0-fold increased when 90 nM TAFI is added). From these data we can deduce that the introduction of Cys 305 , Ile 325 , and Ile 329 slows down the conformational change responsible for the inactivation of TAFIa (7). The delayed formation of 25-and 11-kDa products of the TAFI-AI-Cys 305 and TAFI-AI-Cys 305 -Ile 329 variants support these data (Fig. 2). The stronger antifibrinolytic effect of all TAFIa variants is in agreement with their increased functional half-life and can be explained at least in part by the threshold phenomenon (22,23). According to the threshold phenomenon, clot lysis is prevented from proceeding into the propagation phase as long as TAFIa activity remains above a threshold value. As demonstrated before by Schneider et al. (6), the increase in clot lysis time cannot be overcome by adding increasing amounts of TAFI (Fig. 3A). As expected, clot lysis time could only be further increased by increasing the stability of TAFIa (Fig. 3A). Under the current conditions, clot lysis times reached a maximum when TAFI concentrations between 45 and 90 nM were used (for all TAFI variants). This maximum could only be increased by increasing the TAFIa half-life to 40 min. Using these TAFIa variants, a maximum is reached between 270 and 320 min (Table 1; Fig. 3, A and B). Although we cannot exclude that this maximal effect is due to different activation rates by T/TM and/or to different enzyme activities, this maximal effect was also previously observed by Schneider and Nesheim (24) and Walker et al. (18), who reported maximal clot lysis times of 350 and 250 min, respectively.
In contrast to previous published results (6), we found that the catalytic efficiency for activation of TAFI-AI by T/TM is 2-fold higher than for TAFI-AT. Introduction of Cys 305 increases the catalytic efficiency for activation of TAFI by T/TM again 2-fold ( Table 2, k cat /K m ). However, introducing either Cys 305 or Ile 329 leads to a more than 2-fold lower catalytic efficiency for converting the substrate Hip-Arg compared with the two naturally occurring variants, TAFIa-AT and TAFIa-AI (Table 3, k cat /K m ). This is probably due to the influence of the introduced mutations, Cys 305 and Ile 329 , on the active site conformation.
The 3 stabilizing mutations are located in the 300 -330-amino acid region (Fig. 4), forming mainly an ␣-helix (residues 308 -325) and two loops (residues 297-307 and 326 -331, respectively) connecting the helix to the main part of the molecule. It is worth mentioning that the residues at positions 305 and 329 are in very close proximity to the ␣-helix (4.96 and 4.21 Å, respectively). Most likely this ␣-helix, the two loops, and their mutual interactions govern in part the conformational change accompanied with the inactivation process. These data are also in line with previous studies indicating that mutations in this region result in a destabilization (7,16).
In conclusion, we were able to identify, in addition to the naturally occurring Ile 325 , two stabilizing mutations i.e. Cys 305 and Ile 329 , that increase synergistically the TAFIa half-life. Subsequently, our study demonstrates that the half-life of TAFIa has an important impact on its antifibrinolytic potential. We observed that by increasing the TAFIa stability, the clot lysis time can be maximally increased to 300 min. Because the development of a TAFI(a) inhibitor is limited by the half-life of TAFIa, the availability of a more stable TAFIa variant will raise new opportunities in the search for TAFIa inhibitors. Moreover, the availability of a stabilized variant will also facilitate further structural analysis to obtain more insights in the underlying mechanisms of the instability of TAFIa.  (21). The activation peptide is shown in black, and the enzyme moiety is shown in gray. The activation peptide is cleaved after Arg 92 (green sphere) from the enzyme moiety. After a conformational change, the enzyme moiety is cleaved after Arg 302 (green sphere). The yellow spheres represent the substrate binding residues (Asn 234 , Arg 235 , Tyr 341 , and Asp 349 ), and the blue spheres represent the active site residues (Glu 363 and Arg 217 ). The red spheres represent the three stabilizing residues, i.e. Cys 305 , Ile 325 , and Ile 329 . The ␣-helix, corresponding with residues 308 -325, and the 2 connecting loops, corresponding with residues 297-307 and 326 -331, are indicated in orange.