Plasma and Recombinant Thrombin-activable Fibrinolysis Inhibitor (TAFI) and Activated TAFI Compared with Respect to Glycosylation, Thrombin/Thrombomodulin-dependent Activation, Thermal Stability, and Enzymatic Properties*

Thrombin-activable fibrinolysis inhibitor (TAFI) is a human plasma zymogen similar to pancreatic pro-carboxypeptidase B. Cleavage of the zymogen by thrombin/thrombomodulin generates the enzyme, activated TAFI (TAFIa), which retards fibrin clot lysis in vitro and likely modulates fibrinolysis in vivo. In the present work we stably expressed recombinant TAFI in baby hamster kidney cells, purified it to homogeneity from conditioned serum-free medium, and compared it to plasma TAFI (pTAFI) with respect to glycosylation and kinetics of activation by thrombin/thrombomodulin. Although rTAFI is glycosylated somewhat differently than pTAFI, cleavage products with thrombin/thrombomodulin are indistinguishable, and parameters of activation kinetics are very similar with k cat= 0.55 s−1, K m = 0.54 μm, and K d = 6.0 nm for rTAFI and k cat = 0.61 s−1,K m = 0.55 μm, andK d = 6.6 nm for pTAFI. The respective TAFIa species also were prepared and compared with respect to thermal stability and enzymatic properties, including inhibition of fibrinolysis. The half-life of both enzymes at 37 °C is about 10 min, and the decay of enzymatic activity is associated with a quenching (to ∼62% of the initial value at 60 min) of the intrinsic fluorescence of the enzyme. Stability was highly temperature-dependent, which, according to transition state theory, indicates both high enthalpy and entropy changes associated with inactivation (ΔH o ‡ ≅ 45 kcal/mol and ΔS o ‡ ≅ 80 cal/mol/K). Both species of TAFIa are stabilized by the competitive inhibitors 2-guanidinoethylmercaptosuccinic acid and ε-aminocaproic acid. rTAFIa and pTAFIa are very similar with respect to kinetics of cleavage of small substrates, susceptibility to inhibitors, and ability to retard both tPA-induced and plasmin-mediated fibrinolysis. These studies provide new insights into the thermal instability of TAFIa, a property which could be a significant regulator of its activity in vivo; in addition, they show that rTAFI and rTAFIa are excellent surrogates for the natural plasma-derived species, a necessary prerequisite for future studies of structure and function by site-specific mutagenesis.

The balance between the activities of the coagulation and fibrinolytic cascades is essential to protect the organism from excessive blood loss upon injury as well as to maintain the fluidity of blood within the vasculature. Imbalances lead to a tendency toward either bleeding or thrombosis, the latter of which is manifested as heart attacks and strokes.
The coagulation and fibrinolytic cascades consist of a series of zymogen to enzyme conversions, terminating in the proteolytic enzymes thrombin and plasmin, which, respectively, catalyze the deposition and removal of fibrin. However, when in a complex with the endothelial cell-surface cofactor thrombomodulin, the specificity of thrombin is changed from fibrinogen to protein C (1), thus changing thrombin to an anticoagulant rather than a procoagulant enzyme.
When formed in the context of a fibrin clot, activated protein C was found to be profibrinolytic as well as anticoagulant (2)(3)(4)(5), an observation that results from the ability of activated protein C to prevent the formation of a previously uncharacterized fibrinolysis inhibitor (6). Further investigations in our laboratory led to the isolation of this factor, a 60-kDa glycoprotein present in human plasma at a concentration of approximately 50 nM, which we termed TAFI 1 (thrombin-activable fibrinolysis inhibitor (7)). Upon activation of TAFI by cleavage with thrombin, an active enzyme is formed (TAFIa) that possesses carboxypeptidase B-like activity and inhibits fibrinolysis, probably by removal from partially degraded fibrin the C-terminal lysines that contribute to the development of positive feedback in the fibrinolytic cascade (8). Although thrombin itself is a weak activator of TAFI (8), the thrombin-thrombomodulin complex activates TAFI with a 1250-fold higher catalytic efficiency than thrombin alone, suggesting that thrombin/thrombomodulin is the physiological activator of TAFI (8).
We have shown that TAFIa inhibits tPA-mediated fibrinolysis in vitro half-maximally at a concentration of approximately 1 nM (8). This concentration is about 2% the level of the zymogen in plasma, indicating that enough TAFIa could be produced in plasma to have a significant effect on fibrin clot lysis in vivo. In support of this scenario, recent studies utilizing a canine model of thrombolysis indicated that inducible carboxypeptidase activity (probably TAFIa) is increased during thrombosis and thrombolytic therapy (9); in addition, a significant positive correlation is observed between inducible carboxypeptidase activity and the time required for restoration of blood flow (9). These studies strongly imply that regulation of TAFIa activity may play a significant role in modulating hemostasis and thrombolysis in vivo.
The protein that we have termed TAFI has been described previously by other groups. Hendriks et al. (10) detected an unstable carboxypeptidase B-like activity in human serum; the enzyme corresponding to this activity was subsequently isolated and named CPU ("unstable" carboxypeptidase (11)). In addition, Eaton et al. (12) discovered the zymogen as a contaminant in preparations of ␣ 2 -antiplasmin and termed it plasma procarboxypeptidase B (pro-pCPB). These investigators also cloned a cDNA corresponding to pro-pCPB, the deduced amino acid sequence of which showed significant homology with pancreatic procarboxypeptidase B (12). Amino acid sequence analysis of TAFI and fragments derived from TAFI showed that TAFI and pro-pCPB are the same protein (7).
One of the striking features of the enzyme TAFIa is its inherent instability, a property for which it was given the name carboxypeptidase U (11). The relatively short half-life of the enzyme at body temperature suggests that this inherent instability might be relevant in the down-regulation of the enzyme in vivo. The following studies were undertaken to both investigate the instability of TAFIa and to compare thoroughly the properties of recombinant and plasma-derived TAFI and TAFIa.
Cloning of a TAFI cDNA and Construction of a TAFI Expression Vector-Based on the cDNA sequence for plasma procarboxypeptidase FIG. 1. Cloning of a TAFI cDNA and construction of a TAFI expression plasmid. Using three sets of primers based on the cDNA sequence of plasma procarboxypeptidase B published by Eaton et al. (12), three overlapping PCR products encompassing the TAFI coding region were amplified from a first-strand human liver cDNA library. The TAFI cDNA was reconstructed in the plasmid pBluescript SKϩ using the unique restriction sites BglII and SphI and then inserted into the SmaI site of the mammalian expression vector pNUT. Nucleotide numbering is based on that of Eaton et al. (12). B, BglII; Sp, SphI; Xb, XbaI; Xh, XhoI; Sm, SmaI; MP, mouse metallothionein I promoter; PA, human growth hormone polyadenylation sequence. B (pro-pCPB) published by Eaton et al. (12), three pairs of oligonucleotide primers were designed (see Fig. 1). The sequences of the primers pairs were as follows: primer 1, 5Ј-CTGTTGGGATGAAGCTTTGC-3Ј and primer 2, 5Ј-TCGTTGGAAATCTGCTGTTG-3Ј; primer 3, 5Ј-CTTG-CTGGCAGACGTGGAAG-3Ј and primer 4, 5Ј-GCTGGGAGTATGAAT-GCATG-3Ј; primer 5, 5Ј-GCATACATCAGCATGCATTC-3Ј and primer 6, 5Ј-CAATGATTTGGTCTTGCTGG-3Ј. By using these primers, three overlapping PCR products were obtained by PCR amplification using a first-strand human liver cDNA library (1 g) as the template. The PCR cycling conditions were as follows: 5 min at 95°C followed by 35 cycles of 30 s at 95°C, 30 s at 52°C, and 30 s at 75°C, and a final 15 min at 75°C. The three primer pairs gave rise to PCR products of 356 base pairs (nucleotides 10 -366 of pro-pCPB sequence; Eaton et al. (12)), 665 base pairs (nucleotides 318 -983), and 425 base pairs (nucleotides 953-1378), respectively. The three PCR products were individually inserted into the EcoRV site of pBluescript SKϩ (Stratagene) and analyzed by DNA sequence analysis; the sequence of the cloned PCR fragments was identical to that reported by Eaton et al. (12).
The full-length TAFI cDNA was reconstructed in the pBluescript vector using the unique BglII and SphI restriction sites that are located in the regions of overlap between the individual fragments (see Fig. 1). The reconstructed TAFI cDNA was excised by digestion with XbaI and XhoI; the ends were made blunt using the Klenow fragment of Escherichia coli DNA polymerase I, and the fragment was inserted into the pNUT plasmid that had been digested with SmaI. The resultant expression plasmid (TAFI-pNUT) contains nucleotides 10 -1378 of the TAFI cDNA under control of the mouse metallothionein I promoter and the human growth hormone polyadenylation sequence. The plasmid also contains a mutant version of the dihydrofolate reductase gene to select stable cell lines in the presence of a high concentration of methotrexate (see below). An expression plasmid containing the TAFI cDNA in the reverse orientation (revTAFI-pNUT) was constructed in a similar fashion.
Culture and Transfection of Mammalian Cells-BHK cells were cultured in 100-mm dishes in DMEM/F-12 containing 5% newborn calf serum in a 37°C humidified incubator (95% air/5% CO 2 atmosphere). Cells were transfected by the method of calcium phosphate co-precipitation (14) using 10 g of TAFI-pNUT plasmid per 100-mm plate. Six hours following transfection, the cells were washed and fed fresh DMEM/F-12 containing 5% newborn calf serum. The cells were allowed to recover overnight, after which the medium was replaced with DMEM/F-12 containing 5% newborn calf serum and 400 M methotrexate. After a 2-week selection period, surviving foci were picked and analyzed for expression of TAFI by Western blot analysis.
For Western blot analysis of TAFI expressed from transiently transfected cells, BHK cells were transfected, as described above, with 10 g/100-mm plate of TAFI-pNUT or revTAFI-pNUT. Following transfection and an overnight recovery period, the medium was replaced with serum-free medium (Opti-MEM; 5 ml/100-mm plate) and culture continued for a further 48 h. At this time, conditioned medium was harvested from the cells and subjected to SDS-PAGE followed by Western blot analysis using a TAFI-specific monoclonal antibody (13).
Recombinant Protein Expression-BHK cell lines stably expressing TAFI were seeded in triple flasks (500 cm 2 ; Nunc, Roskilde, Denmark) in DMEM/F-12 containing 1% (v/v) UltroSer G and 200 M methotrexate. Once the cells had become confluent, the medium was changed to Opti-MEM (100 ml/flask) containing 1% (v/v) PSF and 50 M ZnCl 2 . Conditioned medium was harvested at 24 -48-h intervals and replaced with fresh Opti-MEM; the harvested medium was supplemented with Tris, pH 8 (to 5 mM), reduced glutathione (to 0.5 mM), and dEGRck (to 2 M) and stored at Ϫ20°C.
Purification of Recombinant TAFI-Two liters of conditioned medium was concentrated 10-fold by ultrafiltration using an LP-1 pump and S1Y30 (30-kDa cutoff) spiral cartridge (Amicon, Oakville, Ontario). The concentrated medium was then dialyzed against two changes of 4 liters of 20 mM HEPES, pH 7.4 (HB). The retentate was passed at 22°C over a 20-ml Q-Sepharose column that had been washed with HB containing 0.5 M NaCl and then equilibrated with HB. The column was then washed with 5 column volumes of HB, and rTAFI was eluted with HB containing 0.2 M NaCl. Fractions containing activity (assessed by Hip-Arg hydrolysis following treatment of the fractions with thrombin/ thrombomodulin; see below) were pooled and applied to a 5-ml plasminogen-Sepharose column that had been equilibrated at 22°C with HB containing 0.2 M NaCl. The column was washed with 50 ml of HB containing 0.2 M NaCl and 25 ml of HB containing 20 mM NaCl and 0.01% (v/v) Tween 80, and rTAFI was eluted in this buffer containing 100 mM ⑀-ACA. Protein containing fractions were then passed over a 1-ml DEAE-Sepharose Fast Flow column equilibrated with HB at 22°C. After a wash with 40 ml of HB, rTAFI was eluted with HB containing 0.15 M NaCl and stored at Ϫ70°C in this buffer.
Treatment of TAFI with N-Glycosidase F-Ten micrograms of rTAFI or plasma-derived TAFI (pTAFI) (each in 20 mM HEPES, pH 7.4, 0.15 M NaCl (HBS), 1 mM diisopropyl fluorophosphate) in a volume of 50 l was denatured by boiling for 5 min in the presence of 1% (w/v) SDS. The reactions were then diluted 10-fold with HBS, and Nonidet P-40 was added to a final concentration of 0.5% (v/v). The reactions were initiated by the addition of 1 unit of N-glycosidase F. Reactions were incubated at 37°C; aliquots were removed at various times and the reactions stopped by the addition of SDS-PAGE sample buffer (1% (w/v) (final) SDS, 5% (v/v) (final) glycerol, 0.05 mg/ml (final) bromphenol blue) followed by boiling for 5 min. The samples were then subjected to electrophoresis on a 10% polyacrylamide Tris/Tricine gel (15); protein bands were visualized by silver staining (16).
Activation of TAFI and TAFIa Activity Assays-rTAFI or pTAFI (1 M) was incubated for 10 min at 22°C in the presence of 5 mM CaCl 2 , 20 nM thrombin, and 80 nM thrombomodulin (Solulin) in HBS containing 0.01% (v/v) Tween 80. Under these conditions, the zymogen is quantitatively converted to the active enzyme, and the inactivating cleavage of the active enzyme at Arg-330 is undetectable. Thrombin activity was then quenched by the addition of the irreversible chloromethyl ketone inhibitor PPAck (100 nM). Hydrolysis of the substrates Hip-Arg and Hip-Lys was monitored at 254 nm in a Perkin-Elmer Lambda 4B spectrophotometer (thermostatted to 22°C) in 20 mM Tris, pH 7.65, 100 mM NaCl. Extinction coefficients for the two substrates and the change in extinction coefficient that occurs upon hydrolysis of the substrates were determined by titrating Hip-Arg, Hip-Lys, and HA and were as follows: Hydrolysis of FA-Ala-Lys was monitored at 340 nm in a Titertek Twin kinetic plate reader (22°C) in HBS containing 0.01% (v/v) Tween 80. Reactions were performed in a 200-l volume in Immulon 1 Removawell strips (Dynatech Laboratories Inc., Chantilly, VA). To determine the change in absorbance per mol of FA-Ala-Lys that occurs upon hydrolysis of this substrate, various concentrations of FA-Ala-Lys (in 198 l) were placed in successive wells, and measurement of absorbance was initiated. Two microliters of rTAFIa (60 nM final concentration) was then added to each well, and the reactions were allowed to proceed until the substrate was fully consumed. From these data, the molar change in absorbance upon hydrolysis of FA-Ala-Lys was determined to be Ϫ0.390 mM Ϫ1 .
Measurement of the Thermal Stability of TAFIa-TAFIa was formed from pTAFI and rTAFI by incubation in the presence of thrombin/ thrombomodulin as described above. Following the addition of PPAck, TAFIa preparations were transferred to wet ice (0°C) or to water baths thermostatted at 22, 30, or 37°C. Timed aliquots were removed, and the TAFIa activity in them was assessed by measurement of the rate of FA-Ala-Lys hydrolysis as described above. In some experiments, saturating concentrations of the competitive inhibitors ⑀-ACA and GEMSA were added to the TAFIa preparations incubated at 37°C. To assess the stability of the respective zymogens, pTAFI and rTAFI were incubated at 37°C for 0, 10, and 60 min, after which CaCl 2 , thrombin, and thrombomodulin were added, and the incubation was continued for a further 10 min at room temperature. At this time, PPAck was added, and the TAFIa activity was measured as described above.
To measure changes in the intrinsic fluorescence of the respective enzymes, pTAFIa and rTAFIa were formed as described above, and the thrombin was quenched with PPAck. The respective enzymes were then diluted 20-fold (to 50 nM, final), and their intrinsic fluorescence ( ex ϭ 280 nm; em ϭ 340 nm) was measured continuously in a Perkin-Elmer LS50B Luminescence Spectrometer with the cuvette thermostatted to 37°C; the excitation and emission slits were 10 and 5 nm, respectively, and a 290-nm cutoff filter was placed in the emission beam. Timed aliquots were removed from the incubations, and the TAFIa activity remaining in them was measured using FA-Ala-Lys as the substrate.
In Vitro Fibrinolysis Assays-To assess the effect of rTAFIa and pTAFIa on tPA-mediated fibrin clot lysis, 194 l of a solution containing fibrinogen (2.9 M), Glu-plasminogen (0.66 M), recombinant ␣ 2 -antiplasmin (0.5 M), antithrombin III (0.96 M), and various concentrations of rTAFIa or pTAFIa (in HBS containing 0.01% (v/v) Tween 80) was added to microtiter wells containing individual 2-l aliquots of CaCl 2 (10 mM, final), thrombin (7.7 nM, final), and tPA (442 pM, final). Lysis of the resultant clots was monitored turbidometrically at 405 nm in a Titertek Twin reader thermostatted at 37°C as described previously (7). To assess the effect of rTAFIa and pTAFIa on plasminmediated clot lysis, a solution of 194 l containing fibrinogen (2.9 M) and various concentrations of rTAFIa or pTAFIa was added to micro-titer wells containing individual 2-l aliquots of CaCl 2 (10 mM, final), thrombin (7.7 nM, final), and plasmin (2 nM, final). Lysis of the resultant clots was monitored turbidometrically as for the tPA-containing clots.

Expression and Purification of Recombinant TAFI (rTAFI)-
Conditioned serum-free medium (CM) harvested from BHK cells transiently transfected with the TAFI-pNUT expression plasmid was subjected to SDS-PAGE followed by Western blot analysis using a TAFI-specific monoclonal antibody (13) (Fig.  2). Also included in this blot was 10 ng of plasma-derived TAFI (pTAFI) and CM harvested from BHK cells transiently transfected with the pNUT plasmid containing the TAFI cDNA in the reverse orientation (revTAFI-pNUT). An intense immunoreactive band similar in mobility to pTAFI was present in CM harvested from cells transfected with the expression plasmid containing the TAFI cDNA in the forward, but not the reverse, orientation indicating that the former cells were expressing a protein corresponding to rTAFI. Inspection of the blot reveals that rTAFI migrates marginally slower on SDS-PAGE than pTAFI (Fig. 2).
To obtain preparative amounts of rTAFI, stable BHK cell lines expressing rTAFI were grown in serum-free medium (Opti-MEM) in triple flasks. Typically, 2 liters of pooled CM was used as the starting material for purification of rTAFI, the scheme for which was adapted from the method for purification of TAFI from plasma (7). rTAFI was isolated from concentrated, dialyzed CM by anion exchange chromatography over Q-Sepharose followed by affinity chromatography over plasminogen-Sepharose, and anion exchange chromatography over DEAE-Sepharose (see "Experimental Procedures"); the results of a typical purification are shown in Table I. As measured by Hip-Arg hydrolysis following treatment with thrombin/thrombomodulin, rTAFI was present in the CM at a concentration of 0.5-1 mg/liter, and homogeneous preparations of rTAFI were typically obtained with a yield of 20 -30%.
Comparison of the Biochemical Properties of rTAFI and pTAFI- Fig. 3 shows the results of SDS-PAGE analysis of purified rTAFI and pTAFI. As predicted by Western blot anal-ysis (Fig. 2), purified rTAFI migrated marginally slower than pTAFI (Fig. 3). However, SDS-PAGE analysis of rTAFI and pTAFI activated by thrombin/thrombomodulin (Fig. 4, upper panel) revealed that the M r ϳ35,000 active enzyme (TAFIa) derived by cleavage of the zymogen at Arg-92 (7) and the M r ϳ25,000 and M r ϳ12,000 fragments derived from cleavage of TAFIa at Arg-330 (7) were of identical size for rTAFI and pTAFI. In addition, the rates of appearance of the M r ϳ35,000, M r ϳ25,000, and M r ϳ12,000 bands (Fig. 4, upper panel) and the rates of appearance of TAFIa activity (as measured by hydrolysis of hippuryl-L-arginine (Hip-Arg)) ( Fig. 4, lower  panel) were similar for rTAFI and pTAFI.
To assess if the difference in mobility on SDS-PAGE of rTAFI and pTAFI might be due to differences in N-linked glycosylation, rTAFI and pTAFI were treated with N-glycosidase F, an enzyme that removes N-linked glycans from protein substrates (17). Analysis of samples treated for various times with Nglycosidase F by electrophoresis on a Tris/Tricine gel (15), followed by silver-staining (Fig. 5), revealed that both rTAFI and pTAFI gave rise to identically sized (presumably fully deglycosylated) terminal products that corresponded closely in molecular weight to the predicted peptide molecular weight of TAFI (M r 45,999 (12)). In addition, both rTAFI and pTAFI gave rise to three intermediately glycosylated species, indicating that all four potential N-linked glycosylation sites in the activation peptide of both rTAFI and pTAFI (12) are utilized. Both deglycosylated rTAFI and pTAFI species migrated as doublets on the gel, which may be the result of heterogeneity in O-linked glycosylation in the respective proteins.
Comparison of the Activation of rTAFI and pTAFI by Thrombin/Thrombomodulin-Although previous studies have shown that TAFI can be activated by thrombin (7), plasmin (12), and trypsin (12), these are all relatively poor activators of the FIG. 2. Expression of rTAFI in BHK cells. BHK cells were transiently transfected with the expression plasmids TAFI-pNUT or revTAFI-pNUT. The indicated volumes of conditioned serum-free medium (CM) from these cells were subjected to SDS-PAGE on a 5-15% polyacrylamide gradient gel followed by Western blot analysis using a TAFI-specific monoclonal antibody. Following incubation of the blot with a secondary antibody (goat anti-mouse IgG conjugated to horseradish peroxidase), immunoreactive bands were visualized by chemiluminescence. Also shown on the blot is 10 ng of purified plasma-derived TAFI (pTAFI). The positions of molecular mass markers are shown to the left of the blot.  zymogen. Furthermore, at concentrations of these activators required to achieve significant cleavage at Arg-92, TAFIa itself is subsequently cleaved rapidly at Arg-330 to inactivate the enzyme. Recent studies from our laboratory, however, have shown that thrombin complexed with thrombomodulin activates TAFI with a 1250-fold higher catalytic efficiency than thrombin alone without appearing to enhance significantly the rate of the inactivating cleavage (8), suggesting that the physiologic activator is thrombin/thrombomodulin. Therefore, the ability of thrombin/thrombomodulin to activate rTAFI was investigated. The concentrations of substrate (rTAFI or pTAFI) and cofactor (Solulin) were systematically varied, and the initial rates of TAFIa formation were assessed by measurement of  N-[3-(2-furylacryloyl)]-L-alanyl-L-lysine (FA-Ala-Lys) hydroly-sis. Kinetic data were fit to an equation developed to describe the kinetics of pTAFI activation by thrombin/thrombomodulin (8). Presented in Fig. 6 are the rates of pTAFIa (Fig. 6A) or rTAFIa (Fig. 6B) formation plotted as a function of TAFI concentration. The solid lines in both panels represent the rates calculated from the fit parameters k cat , K m , and K d . These kinetic constants are presented in Table II and are very similar for rTAFI and pTAFI.
Comparison of the Enzymatic Properties of rTAFIa and pTA-FIa-To compare the enzymatic behavior of rTAFIa and pTA-FIa, the ability of the respective enzymes to hydrolyze the synthetic substrates Hip-Arg (50 -600 M), Hip-Lys (100 -1200 M), and FA-Ala-Lys (0.1-2 mM) was assessed. Kinetic constants for rTAFIa and pTAFIa for these substrates were obtained by fitting the rates of substrate hydrolysis to the Michaelis-Menten equation using nonlinear regression; the results are presented in Table III. Inhibition constants (K i ) for three competitive inhibitors of TAFIa, ⑀-aminocaproic acid (⑀-ACA (19)), 2-guanidinoethylmercaptosuccinic acid (GEMSA (11)), and potato carboxypeptidase inhibitor (PCI (18)) were also determined (Table III) Table III), and the data were fit to a modified form of the Michaelis-Menten equation which describes competitive inhibition. All of the kinetic constants determined were comparable for rTAFIa and pTAFIa (Table III).
Comparison of the Stability of rTAFIa and pTAFIa-Previous studies have shown that TAFIa activity is not stable (8, 11). Thus, we measured and compared the stability of rTAFIa and pTAFIa (Fig. 7). The respective zymogens were activated by thrombin/thrombomodulin; the thrombin was irreversibly inhibited with PPAck, and the activated enzymes were placed at 0, 22, 30, or 37°C. Aliquots were removed at various times, and the FA-Ala-Lys hydrolytic activity in each sample was assessed. Whereas rTAFIa and pTAFIa are both stable at 0°C, their activities decay with similar half-lives of about 120 -150, 40 -50, and 8 -9 min at 22, 30, and 37°C, respectively (Fig. 7A). The data of Fig. 7A  imply that inactivation is not enthalpically favored, possibly because of the need to break numerous noncovalent bonds. This is offset, however, by a highly favorable entropy change associated with inactivation. The high enthalpy change also accounts for the high sensitivity of inactivation to temperature. According to this interpretation, the process of TAFIa inacti-vation involves both the energetically unfavorable disruption of numerous noncovalent interactions within the protein and the energetically favorable assumption of a less ordered structure.
We also investigated the stability of rTAFIa and pTAFIa at 37°C in the presence of saturating concentrations of the competitive inhibitors ⑀-ACA and GEMSA. Interestingly, we found that the activity of both enzymes was stable in the presence of the inhibitors (Fig. 7B). In addition, the respective zymogens were stable at 37°C; the potential TAFIa activity of the zymogens remained constant at this temperature over a 60-min period, as represented by the dashed (pTAFI) and dotted (rTAFI) lines in Fig. 7B.
To assess structural changes that may accompany the decay of enzymatic activity of the respective enzymes, the intrinsic fluorescence of pTAFIa and rTAFIa was measured at 37°C (Fig. 8). The data show that incubation of the enzymes at this temperature elicits a marked quenching of the intrinsic fluorescence signal that correlates with the decline in enzyme activity and that reaches ϳ62% of the initial value by 60 min. In control experiments, the intrinsic fluorescence of the respective zymogens was found to remain essentially stable over 60 min at 37°C (data not shown).
Comparison of Inhibition of Fibrinolysis by rTAFIa and pTA-FIa-We initially identified TAFI as a factor in human plasma which, in the context of sustained thrombin generation, was capable of generating in response to thrombin an activity that retards tPA-mediated fibrin clot lysis in vitro. Thus, the ability of rTAFIa and pTAFIa to inhibit fibrinolysis was directly compared. Fibrin clots containing plasminogen, tPA, recombinant ␣ 2 -antiplasmin, antithrombin III, and rTAFIa or pTAFIa at various concentrations were formed in the wells of microtiter  7. Comparison of the stability of rTAFIa and pTAFIa. A, rTAFIa (closed symbols) and pTAFIa (open symbols) were formed by incubation of the respective zymogens with thrombin/thrombomodulin, and the activated species were placed at 0°C (circles), 22°C (triangles), 30°C (diamonds), or 37°C (squares). At various times, aliquots were removed and the TAFIa activity measured using FA-Ala-Lys as the substrate. B, the effects of ⑀-ACA (5 mM) and GEMSA (100 M) on the stability of rTAFIa and pTAFIa at 37°C were determined. The controls (no inhibitor) are indicated by circles, whereas results with GEMSA and ⑀-ACA are indicated by squares and triangles, respectively. The dashed and dotted lines represent the total potential activity of pTAFI and rTAFI, respectively, incubated for 0, 10, and 60 min at 37°C. plates, and lysis of the clots was monitored turbidometrically. A plot of the time required to achieve 50% clot lysis versus TAFIa concentration is presented in Fig. 9A. Both rTAFIa and pTAFIa are capable of retarding clot lysis time up to 2-fold, with both enzymes achieving their half-maximal effect at a concentration of approximately 1 nM (Fig. 9A).
In addition to the inhibitory effect of TAFIa on tPA-mediated fibrinolysis, which is likely due to removal from fibrin of the C-terminal lysines that are important for stimulating tPAmediated plasminogen activation, we recently observed that TAFIa also directly attenuates plasmin-mediated fibrin breakdown (21). This effect is apparently due to direct inhibition of plasmin activity by TAFIa. To compare the ability of rTAFIa and pTAFIa to inhibit plasmin-mediated fibrinolysis, fibrin clots containing plasmin and rTAFIa or pTAFIa at various concentrations were formed in the wells of microtiter plates and lysis of the clots monitored turbidometrically; the results are shown in Fig. 9B. The data indicate that rTAFIa and pTAFIa have an identical effect on plasmin-mediated fibrinolysis, with little inhibition occurring below 50 nM TAFIa but with a rapid increase in lysis time above 50 nM TAFIa which then appears to tend toward saturation. DISCUSSION We have expressed a recombinant form of TAFI (thrombinactivable fibrinolysis inhibitor; also known as CPU (11) and plasma procarboxypeptidase B (12)) in BHK cells. We chose to express rTAFI in mammalian cells primarily because a number of other proteins involved in the coagulation and fibrinolytic cascades have been successfully expressed in BHK cells. These include human prothrombin (22), human factor VII (23), hu-man factor VIII (24), human antithrombin III (25), human plasminogen (26), and human ␣ 2 -antiplasmin (26). Unlike bacterial expression systems, recombinant proteins expressed in mammalian cells can be secreted in their native conformation and are subjected to post-translational modifications such as glycosylation.
In the present study, we found that while the recombinant version of TAFI migrates marginally more slowly on SDS-PAGE than its plasma-derived counterpart, rTAFI is virtually indistinguishable from pTAFI in terms of its ability to be activated by thrombin/thrombomodulin and the stability and enzymatic properties of TAFIa including the ability to inhibit tPA-and plasmin-mediated fibrin clot lysis in vitro.
The mobility difference between rTAFI and pTAFI on SDS-PAGE (Fig. 3) is likely due to differences in the size and/or composition of N-linked glycans. Inspection of the cDNA sequence for plasma procarboxypeptidase B reveals the presence of four potential N-linked glycosylation sites (Asn-22, Asn-51, Asn-63, and Asn-86), all of which are located within the 92amino acid activation peptide (12). Indeed, there is no apparent difference between the mobility of the M r ϳ35,000 TAFIa species derived from rTAFI and pTAFI (Fig. 4). Furthermore, treatment of rTAFI and pTAFI with N-glycosidase F, which specifically removes N-linked glycans, gave rise to terminal products of identical electrophoretic mobility (Fig. 5). Three intermediately glycosylated species are present, indicating that all four potential N-linked sites are utilized in rTAFI and pTAFI.
Differences in the size and/or composition of N-linked glycans are frequently observed in the comparison of a recombinant protein and its naturally occurring counterpart (25,(27)(28)(29), which reflect the cell-, tissue-, and species-specificity of glycosylation (30). In some cases such differences have consequences for the functional properties of the recombinant protein. For example, BHK cells secrete a glycoform of ATIII that is not found in plasma and that differs in its affinity for heparin and in its rate of proteinase inhibition, in addition to two other glycoforms that are functionally similar to plasma-derived ATIII (25). However, in other cases differences in glycosylation have no apparent effect on the functional behavior of the recombinant protein. Human factor VIII expressed in BHK cells contains differences in the composition of its N-linked glycans relative to plasma-derived factor VIII (28), yet the two factor VIII preparations are similar with respect to cleavage by thrombin, factor Xa, and activated protein C (24), subunit association and dissociation (24), and pharmacokinetic parameters in baboons (28). Although we did not observe any functional differences between rTAFI and pTAFI attributable to differences in glycosylation, differences in, for example, the pharmacokinetics of rTAFI and pTAFI or in their binding to as yet undescribed substrates cannot be ruled out.
It should be noted that the fully N-deglycosylated TAFI species migrated as doublets, with both members of the doublet in similar proportions in both rTAFI and pTAFI (Fig. 5). These doublets are most likely the result of heterogeneity in usage of O-linked glycosylation sites. It is unlikely that the doublet is due to N-terminal sequence heterogeneity, since TAFI derived from plasma has been shown to possess a unique N-terminal sequence (7,12).
We found that rTAFI is virtually indistinguishable from pTAFI in terms of its ability to hydrolyze small peptide substrates and to be inhibited by the competitive inhibitors ⑀-ACA, GEMSA, and PCI (Table III). These data demonstrate the integrity of the active site of rTAFI and show that, when produced under the conditions reported in this study, rTAFI possesses a specific activity similar to that of its plasma-derived counterpart. Of note is the similarity in the K i for PCI, a small (39 amino acid) inhibitor that has been shown to bind to carboxypeptidase A through regions distinct from those involved in substrate binding. Together with the similar kinetic constants obtained for the activation of rTAFI by thrombin/ thrombomodulin (Table II), these data indicate that the overall structure of TAFI is likely to be faithfully represented by the recombinant protein.
A striking feature of TAFIa is that its activity is unstable, decaying rapidly at 37°C both in vitro in the absence of detectable proteolysis and in the serum milieu. In the present study, we found the stability of rTAFIa and pTAFIa to be very similar; although both enzymes were stable at 0°C, their activities decayed with half-lives of about 10 min at 37°C, about 40 -50 min at 30°C, and about 120 -150 min at 22°C (Fig. 7). The structural basis for the instability of TAFIa is not clear at present, but a thermodynamic interpretation suggests a highly unfavorable enthalpy change associated with inactivation (which may reflect the requirement to disrupt many noncovalent bonds), offset by a corresponding highly favorable entropy change (which in turn may reflect the adoption of a less ordered structure). Consistent with such changes in structure, we found a marked quenching of the intrinsic fluorescence of TA-FIa at 37°C which correlated temporally with the decay of enzymatic activity (Fig. 8). The quenching of the fluorescence signal is presumably attributable to the exposure to the solvent of residues (largely tryptophans) previously buried in the hydrophobic core of the enzyme. It is noteworthy that this presumptive structural change is not detectable by electrophoresis; the mobility of TAFIa species incubated for 60 min at 37°C on SDS-PAGE is identical to that of TAFIa species incubated at 0°C for 60 min (data not shown).
Whether an endogenous inhibitor of TAFIa exists is not known at present, although data exist that indicate that TAFIa interacts with ␣ 2 -macroglobulin and pregnancy zone protein (31). The consequences of these interactions are not known, although binding of TAFIa to these proteins does not affect its enzymatic activity (31). Because carboxypeptidase U activity (TAFIa) in serum has a half-life similar to that of purified TAFIa shown here (10), the presence of a fast-acting endogenous inhibitor in vivo is not indicated. Therefore, the intrinsic instability of TAFIa at 37°C may be physiologically relevant in the down-regulation of TAFIa in vivo. Furthermore, that the activity of TAFIa is stable at 37°C in the presence of saturating concentrations of competitive inhibitors suggests that in the clot milieu, TAFIa activity may be maintained as long as there is sufficient substrate available.
Although the mechanism by which TAFIa inhibits fibrinolysis has yet to be conclusively determined, it likely involves removal of the C-terminal lysine residues from partially degraded fibrin that are required for maximal stimulation of tPA-mediated plasminogen activation (8). However, additional mechanisms for the antifibrinolytic effect of TAFIa may also be possible, since we have recently observed that TAFIa (albeit at higher concentrations) is capable of attenuating plasmin-mediated fibrinolysis (i.e. when plasminogen activation has been bypassed). Although the mechanism underlying this effect is unclear at present, plasmin itself may be the substrate for TAFIa in this instance (21). Based on the data presented in Fig.  9, which show that rTAFIa is capable of inhibiting both tPAand plasmin-mediated fibrinolysis in a manner identical to pTAFIa, it is clear that rTAFIa and pTAFIa are comparable in their ability to participate in the reactions involved in the antifibrinolytic effect. These reactions may include hydrolyzing C-terminal basic residues in fibrin and plasmin, as well as inactivation of TAFIa by plasmin cleavage at Arg-330.
In conclusion, these studies show that although plasma and recombinant TAFI exhibit minor differences in glycosylation, recombinant TAFI and TAFIa are excellent surrogates for the natural species and thus their properties can be used to interpret further differences obtained in structure-function studies utilizing site-directed mutagenesis.