Thrombin Activatable Fibrinolysis Inhibitor, a Potential Regulator of Vascular Inflammation*

The latent plasma carboxypeptidase thrombin-activable fibrinolysis inhibitor (TAFI) is activated by thrombin/thrombomodulin on the endothelial cell surface, and functions in dampening fibrinolysis. In this study, we examined the effect of activated TAFI (TAFIa) in modulating the proinflammatory functions of bradykinin, complement C5a, and thrombin-cleaved osteopontin. Hydrolysis of bradykinin and C5a and thrombin-cleaved osteopontin peptides by TAFIa was as efficient as that of plasmin-cleaved fibrin peptides, indicating that these are also good substrates for TAFIa. Plasma carboxypeptidase N, generally regarded as the physiological regulator of kinins, was much less efficient than TAFIa. TAFIa abrogated C5a-induced neutrophil activation in vitro. Jurkat cell adhesion to osteopontin was markedly enhanced by thrombin cleavage of osteopontin. This was abolished by TAFIa treatment due to the removal of the C-terminal Arg168 by TAFIa from the exposed SVVYGLR α4β1 integrin-binding site in thrombin-cleaved osteopontin. Thus, thrombin cleavage of osteopontin followed by TAFIa treatment may sequentially up- and down-modulate the pro-inflammatory properties of osteopontin. An engineered anticoagulant thrombin, E229K, was able to activate endogenous plasma TAFI in mice, and E229K thrombin infusion effectively blocked bradykinin-induced hypotension in wild-type, but not in TAFI-deficient, mice in vivo. Our data suggest that TAFIa may have a broad anti-inflammatory role, and its function is not restricted to fibrinolysis.

Thrombin has long been regarded as a multifunctional procoagulant enzyme important in hemostasis. At sites of vascular injury it converts fibrinogen to fibrin, amplifies the clotting cascade by activating factor XI (FXI) and the cofactors FV and FVIII, stabilizes the fibrin clot by activating FXIII, and activating platelets via the protease-activated receptors (PAR). It can also function as an anticoagulant by binding to thrombomodulin (TM) 1 on the surface of endothelial cells and activating protein C, which inhibits FVa and FVIIIa, effectively localizing clot formation to the site of vascular injury (1). TM-bound thrombin can also activate thrombin-activable fibrinolysis inhibitor (TAFI; Refs. 2 and 3), a latent plasma procarboxypeptidase also known as carboxypeptidase R, carboxypeptidase U, and procarboxypeptidase B (4 -6). TAFI is now recognized as the second physiological substrate for the thrombin/TM complex. It is important for dampening fibrinolysis by removal of plasmin-exposed lysines on partially digested fibrin clots thereby restricting tissue plasminogen activator binding and further activation of plasminogen (2,7). Thus, activated protein C (aPC) and TAFIa may play complementary roles in hemostasis, with aPC dampening the clotting cascade and preventing excessive thrombin generation, while TAFIa serves to protect the clot already formed at the site of injury.
Thrombin can also act as a proinflammatory molecule by the activation of PAR on monocytes, smooth muscle cells and endothelial cells, thus providing a direct link between coagulation and inflammation (8). On the other hand, thrombin/TM activation of PC could also function as a negative regulator of inflammation along with its defined role as an anticoagulant (1). aPC suppressed the expression of proinflammatory cell adhesion molecules and augmented the expression of anti-apoptotic molecules in cultured endothelial cells (9,10). This anti-inflammatory role of aPC may contribute to its clinical efficacy in the treatment of severe sepsis (11). It has been proposed that TAFIa could also play a role in regulating inflammation by inactivation of kinins and anaphylatoxins through its arginine/ lysine specific carboxypeptidase activity (12,13). However, the efficiency of TAFIa in inactivating these biologically active mediators in comparison with carboxypeptidase N, the constitutively active plasma anaphylatoxin inhibitor, has not been defined.
Recently, a new mechanism by which thrombin can regulate inflammation and tissue repair is revealed by its interaction with osteopontin (OPN, 14 -16). OPN is a multifunctional RGD-containing phosphoprotein with adhesive and cell-signaling functions involved in cell-cell and cell-matrix interactions important in inflammatory responses (17). It is present as an extracellular matrix component in mineralized tissues and in the subendothelial matrix of blood vessels involved by atherosclerosis. OPN also circulates as a soluble proinflammatory cytokine and is widely expressed by many inflammatory cells in culture, including T cells, macrophages, and NK cells. Its expression is enhanced in response to inflammation, tissue injury and stress. It interacts with many cells via RGD-dependent (␣ v ␤ 1 , ␣ v ␤ 3 , or ␣ v ␤ 5 ) and RGD-independent integrins (␣ 4 ␤ 1 , ␣ 5 ␤ 1 , ␣ 8 ␤ 1 , or ␣ 9 ␤ 1 ) and also CD44. It is chemotactic for various cell types, notably monocytes and macrophages and stimulates cell motility and cell survival (17). Of interest, thrombin cleavage of OPN increases the adhesion, spreading, and migration of a variety of cells significantly in vitro, and this can occur in vivo suggesting an important functional role of OPN cleavage at sites of thrombin generation (15,18). Thrombin cleavage of OPN generates an N-terminal fragment that exposes a cryptic integrin-binding motif 161 SVVYGLR 168 on its C terminus, allowing the specific interaction to cells bearing the integrins ␣ 4 ␤ 1 (19 -22) or ␣ 9 ␤ 1 (23,24). In certain cell types, such as melanoma cells, cell binding only occurs with the thrombincleaved, but not the intact OPN, suggesting that thrombin cleavage is critical for certain OPN-cell interactions (15,24).
Here we address the potential role of TM-dependent thrombin activation of TAFI, and the subsequent TAFIa inactivation of bradykinin, C3a, C5a, and thrombin-cleaved OPN. Our data suggest that along with aPC, TAFIa serves to counterbalance the prothrombotic and proinflammatory effects of thrombin generation and thus its biological role is not limited to fibrin clot stabilization.
TAFIa-mediated Hydrolysis of C-terminal Peptides-Soluble recombinant TM (40 nM) was complexed with 1 nM thrombin. TAFI (100 nM) was added and incubated at 25°C for 20 min. The reaction was terminated by the addition of PPACK (5 M). The concentration of TAFIa was determined using an Actichrome TAFI activity kit (American Diagnostica). The peptides based on BK, OPN 160 -168 , C3a 69 -77 , C5a 66 -74 , FB␣-Arg 96 -104 , FB␤-Lys 125-133 , FB␥-Lys 54 -62 , and FB␥-Lys 77-85 with concentrations ranging from 15 M to 2 mM were digested with either 5 nM TAFIa or CPN for 2-5 min at 37°C in assay buffer. Reactions were stopped by boiling for 5 min. Cleaved peptides were resolved by HPLC and the area of the cleaved peptide peak was calculated using the Waters HPLC software then converted to nmoles of peptide using a calibration curve determined with the purified des-Arg or des-Lys form of each peptide. The values for K m and k cat were determined by plotting the initial velocities of cleavage against the different substrate concentrations, then fitting to the Michaelis-Menten equation by non-linear regression analysis. Experiments were performed in duplicate and the data pooled for analysis.
Escherichia coli Expression of Recombinant Osteopontin-The cDNA sequences for mature full-length human OPN (OPN-FL: amino acids 17-314), N-terminal OPN mimicking thrombin-cleaved OPN (OPN-Arg 168 : amino acids 17-168) and the N-terminal OPN mimicking thrombin-cleaved and TAFIa-treated OPN (OPN-Leu ⌬Arg : amino acids 17-167) were inserted as C-terminal in-frame fusions to GST using the E. coli expression vector pGEX-6P-3 (Amersham Biosciences). The constructs pOPN-FL, pOPN-Arg 168 , and pOPN-Leu ⌬Arg also incorporated the 159 RGD 161 to 159 RAA 161 substitutions by site-directed mutagenesis using the Quick-change site directed mutagenesis kit (Stratagene, La Jolla, CA) to form the constructs pRAA/OPN-FL, pRAA/OPN-Arg 168 , and pRAA/OPN-Leu ⌬Arg . Constructs were sequenced and then transformed into E. coli BL21 Gold (Stratagene, La Jolla, CA). For large scale expression of the constructs, transformed cells were grown in 1 liter of LB broth supplemented with ampicillin (100 g/ml) at 37°C until an OD 600 of 0.6 -0.7. Isopropyl-1-thio-␤-D-galactopyranoside (0.2 mM) was added and grown for a further 2 h at 37°C. Cells were pelleted and washed twice with PBS, pH 7.2. Cells were lysed with Bugbuster™ reagent supplemented with the protease inhibitor III mixture set and 2 mM DTT. The cell lysate was clarified, diluted 5-fold in low salt buffer (50 mM Tris, 50 mM NaCl, 2 mM DTT, pH 7.5, and the protease inhibitors) and loaded onto a hiPrep QXL column utilizing an FPLC system (Amersham Biosciences) at 4°C. The column was washed extensively with wash buffer (50 mM Tris, 200 mM NaCl, 2 mM DTT, pH 7.5, the protease inhibitors), then eluted with high salt buffer (50 mM Tris, 600 mM NaCl, 2 mM DTT, pH 7.5) containing 1 mM benzamidine. Fractions containing GST-OPN fusion proteins were pooled and loaded onto a 10-ml glutathione-Sepharose column equilibrated with low salt buffer with 2 mM DTT and 1 mM benzamidine. The column was washed with low salt buffer and OPN was eluted with low salt buffer containing 10 mM glutathione. Fractions containing GST-OPN fusion proteins were pooled and dialyzed against 50 mM Tris, 100 mM NaCl, pH 7.5. The recombinant proteins were greater than 95% pure as judged by SDS-PAGE and Coomassie Blue staining.
Adhesion Assays-Jurkat cells were grown in RPMI 1640 media supplemented with 10% fetal calf serum. Cells with greater than 95% viability as demonstrated by Trypan Blue exclusion assay were washed with Hanks balanced salt solution (HBSS) with 50 mM Hepes, pH 7.5 and adjusted to a concentration of 2 ϫ 10 6 cells/ml in buffer containing 0.2 mM MnCl 2 (21). Recombinant wild-type and mutant OPN fusion proteins were diluted in PBS (concentrations ranging from 0, 0.1, 1, 10, 100 g/ml) and coated onto high protein binding 96-well microtiter plates (Greiner Labortechnik, Ocala, FL) with a final volume of 100 l per well, then incubated overnight at 4°C. For experiments studying the effect of thrombin and TAFIa on OPN-FL mediated Jurkat cell adhesion, the plates were washed three times with thrombin assay buffer (25 mM Hepes, 150 mM, 5 mM CaCl 2 , 0.1% PEG 6000 , pH 7.5). To some of the wells, 100 l of thrombin (100 nM final concentration) was added, and incubated for 1 h at 37°C. Wells were washed three times with thrombin assay buffer. To some of the wells that had been treated with thrombin, 100 l of TAFIa (1.7 nM final concentration) was added and incubated for 60 min at room temperature (RT). The plates were washed three times with PBS, then blocked for 1 h with 3% BSA in PBS. The wells were then washed three times with HBSS-Hepes buffer with 0.2 mM MnCl 2 . To each well, 100 l of 2 ϫ 10 6 cells/ml were added in the same HBSS-Hepes buffer with 0.2 mM MnCl 2 , and cells were allowed to adhere for 60 min at 37°C. Cells were washed twice with PBS, once with absolute ethanol and fixed for 20 min with absolute ethanol. Wells were then washed three times with PBS, stained with 0.1% crystal violet, and again washed three times with PBS. Cells were then lysed with 0.5% Triton X-100. Lysates were read at A 570 nm using a Spectro-MAX plate reader (Molecular Dynamics, Sunnyvale, CA). For studies using recombinant GST-OPN fusion proteins (OPN-Arg 168 and OPN-Leu ⌬Arg with or without the RGD 3 RAA substitution), protein adsorption to the microtiter wells was carried out and Jurkat cell binding determined as described above. For antibody inhibition of Jurkat cell binding to OPN-Arg 168 and OPN-Leu ⌬Arg , cells were preincubated with 10 g/ml of the following anti-integrin antibodies: HP2/1, AMF7, SAM1, 7E4, 25.3.1, 4B4, and BEAR1 for 10 min on ice. Cells were then allowed to attach for 60 min and the amount of cell binding determined.
Neutrophil Myeloperoxidase Release Assay-Neutrophils were prepared from buffy coat concentrates obtained from Stanford Blood Bank following a published protocol (26). Neutrophils (4 ϫ 10 6 cells/ml) were resuspended in Hanks solution with 0.25% bovine serum albumin. To test the effect of TAFIa on C5a-mediated myeloperoxidase release, 60 nM recombinant soluble TM was complexed with 3 nM thrombin for 10 min. TM-complexed thrombin was used to activate TAFI (200 nM). Human C5a (10 M) was hydrolyzed by 0, 1, 10, 100 nM TAFIa at RT. 10-l aliquots were removed at 1, 5, 10, 30, and 60 min and the TAFIa inhibited with potato carboxypeptidase inhibitor. The samples were then diluted 10-fold such that the concentration of C5a/C5a desArg was 1.0 M. 1 ml of neutrophils (4 ϫ 10 6 cell/ml) was treated with 5 g/ml cytochalasin B for 5 min at 37°C. 1 l of each diluted C5a/TAFIa reaction was added to the cell suspension and incubated for 15 min at 37°C, then centrifuged. Myeloperoxidase released to the supernatant was measured by adding 20 l of supernatant to 130 l of 33 mM phosphate buffer pH 6.2, 0.002% H 2 O 2 , 6.7% o-dianisidine-HCl. Reaction velocities were followed at A 460 for 30 min.
Measurement of Mouse aPTT-Mice were anesthetized with ketamine followed by isoflurane. A midline incision was made and the internal jugular vein cannulated. Saline, wild-type, or E229K thrombin solutions were infused through the cannulated jugular vein at rates ranging from 0 to 50 g thrombin/kg/min. Heart and respiratory rates were monitored and blood samples drawn after a 10-min infusion time. Blood was taken through a syringe inserted into the left ventricle and collected in 0.32% citrate. Blood samples were spun down for 1 min, 100 l of plasma was warmed to 37°C then mixed with 200 l of prewarmed Thrombomax-HS reagent, and the aPTT determined using a BBL fibrometer. Fibrinogen levels were taken from similarly processed plasma while platelets counts were performed on whole blood collected into citrate anticoagulant tubes by the Stanford Animal Laboratory Facility (Stanford, CA). At each E229K thrombin dose, at least three mice were used. Results are presented as the mean a PTT Ϯ S.E. (n ϭ 5 for each data point in E229K, n ϭ 3 for each data point in wild type thrombin). The study was approved by the Stanford Panel on Laboratory Animal Care and was conducted in compliance with university regulations.
The Effect of E229K on Bradykinin-induced Hypotension in Vivo-Male C57BL/6 mice (8 -12 weeks old, 25-32 g) or TAFI-deficient mice and their wild-type littermates (in a mixed background of C57BL/6 and 129/Sv) were anesthetized with isoflurane 2%. A midline incision was made in the upper thorax to expose the carotid artery and jugular vein. The carotid artery was cannulated with a pressure transduction catheter connected to a computerized pressure monitor (Powerlab, Colorado Springs, CO) to record blood pressure. The transduction system was calibrated using a sphygmomanometer. Another catheter (PE-10) was inserted into the left jugular vein serving as a route of delivery for experimental treatments. After the level of isoflurane was reduced to 1%, the blood pressure and respiratory rate of the mouse was allowed to stabilize for several minutes. Then 100 l of either E229K thrombin (40 g/kg) or a saline control was administered to the jugular vein. Immediately afterward, 50 l of saline was injected to flush out the catheter, then 100 l of BK in saline (10 nmol/kg) or the equivalent molar dosage of des-ArgBK was given to the mouse through the same jugular vein catheter. Blood pressure tracings were monitored and recorded. Data are presented as maximum drops in mean arterial pressure after administration of BK or des-Arg BK.

Bradykinin, C5a and Thrombin-cleaved Osteopontin Were
Good Substrates for TAFIa-The established view on TAFIa function is that it modulates clot fibrinolysis by removing exposed C-terminal lysines from partially plasmin-degraded clots. Removal of the lysines reduces t-PA and plasminogen binding thereby reducing t-PA enhanced activation of plasmin (2, 3, 5-7). Recent studies also imply the possible role of TAFIa inactivation of kinins and anaphylatoxins (4,12,13). To show that kinins such as BK, anaphylatoxins such as complement C3a and C5a, and the thrombin-cleaved cytokine, OPN, are potential substrates of TAFIa, we determined their Michaelis-Menten constants for hydrolysis by TAFIa. These were compared with the hydrolysis of peptides based on four major plasmin cleavage sites on fibrin clots (27), the fibrin ␣-chain Arg 104 -Asp 105 , the ␤-chain Lys 133 -Asp 134 , and the ␥-chain Lys 62 -Ala 63 and Lys 85 -Ser 86 sites (Table I). TAFIa is thermolabile (5); however, using reaction conditions in the presence of excess substrate with short incubation times at 37°C, no significant TAFIa thermal instability and inactivation was observed (data not shown). The K m and k cat values for hydrolysis of the fibrin peptides ranged from 14.3 M and 13.6 s Ϫ1 for FB␤Lys 125-133 to 361.4 M and 1.5 s Ϫ1 for FB␣Arg 96 -104 . Differences were both in K m and k cat for the fibrin peptides with FB␤Lys 125-133 and FB␥Lys 54 -65 being the best substrates for TAFIa while FB␣Arg 96 -104 and FB␥Lys 77-85 were the worst. By comparison, the overall specificity constants for BK (2.8 ϫ 10 5 M Ϫ1 ⅐s Ϫ1 ), OPN 159 -168 (4.1 ϫ 10 5 M Ϫ1 ⅐s Ϫ1 ), C5a 66 -74 (1.3 ϫ 10 5 M Ϫ1 ⅐s Ϫ1 ) and C3a 69 -77 (2.3 ϫ 10 5 M Ϫ1 ⅐s Ϫ1 ) were similar to the two best fibrin substrates FB␤Lys 125-133 (9.5 ϫ 10 5 M Ϫ1 ⅐s Ϫ1 ) and FB␥Lys 54 -62 (7.6 ϫ 10 4 M Ϫ1 ⅐s Ϫ1 ). The in vitro data suggests that BK, C5a, C3a, thrombin-cleaved OPN and the surface exposed basic residues on partially degraded clots are all good substrates for TAFIa in vivo. It is notable that CPN, commonly regarded as the physiological inactivator of kinins and anaphylatoxins, was less efficient for hydrolysis of BK TAFIa Reduced Jurkat Cell Adhesion to Thrombin-cleaved Osteopontin-Recombinant full-length OPN (OPN-FL) was adsorbed to 96-well microtiter plates at various concentrations and then treated with either thrombin or thrombin followed by TAFIa and compared with untreated OPN-FL for Jurkat cell binding (Fig. 1A). Thrombin-treated OPN-FL at a concentration of 1.0 g/ml supported a 4.7-fold increase in Jurkat cell adhesion (A 570 ϭ 0.47 Ϯ 0.12, n ϭ 6) compared with untreated OPN-FL (A 570 ϭ 0.10 Ϯ 0.03, n ϭ 13) (Fig. 1, A and B). These results are consistent with previous studies showing that the increased cell adhesion is due to exposure of a cryptic integrinbinding motif SVVYGLR at the C terminus of the thrombincleaved OPN fragment (15)(16)(17)(18). TAFIa treatment of thrombincleaved OPN-FL reduced Jurkat cell adhesion to levels seen with OPN-FL (A 570 ϭ 0.12 Ϯ 0.03, n ϭ 6), suggesting a potential role for TAFIa regulating integrin-mediated cell adhesion to thrombin-cleaved OPN. At higher concentrations of OPN (Ͼ10.0 g/ml), cell adhesion between the differently treated OPN were similar to OPN-FL, suggesting saturation of secondary lower affinity binding sites. To confirm that the TAFIa effect was mediated by hydrolysis of the C-terminal arginine from the newly exposed SVVYGLR motif and not due to some other effects of TAFIa, we expressed and purified recombinant GST-OPN-Arg 168 (which mimicked the thrombin-cleaved N-terminal fragment of OPN) and GST-OPN-Leu⌬ Arg (which mimicked the TAFIa treatment). Dose response studies showed that the EC 50 for Jurkat cell binding was 2.3 Ϯ 0.5 g/ml for OPN-FL while OPN-Arg 168 showed higher affinity at 0.5 Ϯ 0.1 g/ml ( Fig. 2A). OPN-Leu⌬ Arg mimicked TAFIa treatment of OPN-Arg 168 as seen by 7-fold decrease in cell adhesion with an EC 50 value of 3.4 Ϯ 0.6 g/ml. The differences in cell adhesion between OPN-FL, OPN-Arg 168 and OPN-Leu ⌬Arg closely paralleled the differences in adhesion between thrombin-treated and thrombin/TAFIa-treated OPN at 1.0 g/ml (Figs. 1B and 2B).
The Roles of RGD and SVVYGLR in Supporting Cell Binding to Thrombin-cleaved Osteopontin-The RGD site in OPN has been shown to have enhanced RGD-dependent cell binding following thrombin cleavage, presumably due to either exposure and/or increased affinity at the RGD site to RGD-dependent integrins (16). There is also evidence of a third integrinbinding site, 131 ELVTDFPTDLPAT 143 (22). In order to delineate the relative importance of the RGD site and the SVVYGLR site in OPN-cell interaction, we mutated the RGD motif to RAA in , indicating that the reported ELVTDFPTDLPAT binding site is not functionally active, at least in this Jurkat cell binding assay (22). Interestingly, while the single SVVYGLR site in the cleaved OPN (OPN-RAA-Arg 168 ) was able to support substantial cell binding (A 570 ϭ 0.29 Ϯ 0.04, n ϭ 6), the magnitude of cell binding is clearly less than that when both sites were exposed in OPN-Arg 168 (A 570 ϭ 0.53 Ϯ 0.04, n ϭ 14). Previous studies have shown that different integrins bind the RGD motif (␣ v ␤ 1 , ␣ v ␤ 3 , ␣ v ␤ 5 , ␣ 5 ␤ 1 , or ␣ 8 ␤ 1 ) and SVVYGLR motif (␣ 4 ␤ 1 , ␣ 4 ␤ 7 , and ␣ 9 ␤ 1 ). The close proximity of the 159 RGD 161 and the 162 SV-VYGLR 168 sites suggests that these two sites cannot bind different integrins simultaneously (26). Therefore the observed enhanced Jurkat cell binding in OPN-Arg 168 may be due to binding of the different OPN molecules to different integrins, resulting in increased overall cell adhesion (Fig. 3). Alternatively it is possible that the RAA mutation or the des-Arg mutation alters the conformation of the other binding site, leading to decreased cell binding affinity. Antibody blocking studies indicate that the anti-␣4 and anti-␤ 1 antibodies blocked the cell binding to OPN-Arg 168 by ϳ75% (Fig. 4), while anti-␣ V , anti-␣ 5 , anti-␣ L , anti-␣ M , and anti-␤ 2 antibody had no effect, consistent with the hypothesis that the binding is partly mediated by ␣ 4 ␤ 1 integrin (on Jurkat cells) binding to the exposed SVVYGLR site on cleaved OPN. The data indicate that in thrombin-cleaved OPN, both the RGD site and the exposed SVVYGLR site are functionally active, capable of supporting Jurkat cell adhesion.
TAFIa Inactivates C5a-mediated Myeloperoxidase Release from Cytochalasin-primed Neutrophils-Kinetic studies on TA-FIa hydrolysis of peptides based on the C termini of C3a and C5a revealed that both anaphylotoxins were potentially good substrates for TAFIa (Table I). The effect of TAFIa on C5ainduced myeloperoxidase release from cytochalasin-primed neutrophils was investigated using purified human neutrophils in a functional assay. TAFIa inactivation of C5a was dose-dependent and effectively inhibited the ability of C5a to activate neutrophils at a molar ratio of 1:1000 within 30 min (Fig. 5). This suggests that TAFIa could potentially regulate the proinflammatory effects of anaphylatoxins in vivo at sites of inflammation.
The Anticoagulant Thrombin E229K Activated Mouse Protein C and TAFI in Vivo-In order to investigate the inactivation of BK by TAFIa in mice, we assessed the ability of the human anticoagulant thrombin E229K to activate mouse TAFI in vivo. Previous studies have shown that E229K thrombin has largely lost all its procoagulant functions but maintains ϳ50% wild-type activity to activate protein C and TAFI when bound to TM (25; data not shown). Intravenous infusion of E229K thrombin produced a dose-dependent reversible anticoagulation in mice as monitored by prolongation of aPTT, an indirect measure of protein C activation (Fig. 6). A 2-fold prolongation of aPTT was achieved with E229K thrombin infused at 20 g/kg/min for 10 min, without any decrease in plasma fibrinogen level or platelet count (Table II). Higher doses of E229K thrombin caused a progressive shortening of the aPTT. It is possible that with such a large dose of E229K thrombin, even with a small residual fibrinogen clotting activity, disseminated intravascular coagulation (DIC) occurred. In contrast to E229K, wild-type thrombin did not prolong the aPTT and at 20 g/kg/min, caused severe depletion of platelets and fibrinogen, consistent with widespread DIC (Table II). The E229K thrombin also activated TAFI in mice in vivo, as demonstrated by TAFIa activity in the mouse plasma as well as by Western blot analysis of the mouse serum (Fig. 7). It is notable that there appeared to be Ͼ90% conversion of plasma TAFI to TAFIa after 10 min of infusion of E229K at 5 g/kg/min. At this infusion rate of E229K, the aPTT was only modestly prolonged (ϳ13 s versus ϳ9 s baseline, Fig. 6), suggesting that E229K thrombin was more efficient in activating TAFI than PC in mice in vivo.
The data indicate that human E229K thrombin is a useful tool for investigating the functional role of TAFIa since it can bind mouse TM and activate mouse TAFI in vivo.
The Effect of E229K Thrombin on Bradykinin-induced Hypotension in Mice in Vivo-To investigate the role of TAFIa inactivation of BK in vivo, we used the E229K thrombin to activate mouse TAFI and determine if TAFIa could modulate BK-induced hypotension. Administration of BK caused a mean decrease of arterial blood pressure of 19.2 Ϯ 2.7 mm Hg (n ϭ 5) in mice while E229K thrombin infusion effectively blocked BK-induced hypotension with a decrease in blood pressure of only 6.4 Ϯ 4.1 mm Hg (n ϭ 6) (67% inhibition, p Ͻ 0.001, Fig.  8A). Des-Arg BK did not produce similar acute hypotension in mice (2.7 Ϯ 2.0 mmHg, n ϭ 5), strongly suggesting that the E229K thrombin blockage effect is mediated by generation of TAFIa in the mouse plasma, with its subsequent inactivation of BK. To confirm this, we treated TAFI-deficient mice and their wild-type littermates with E229K thrombin, followed by BK. E229K thrombin reduced BK-induced hypotension in the wildtype littermates, with a decrease in mean arterial blood pressure of 5.8 Ϯ 1.0 mm Hg (n ϭ 4, Fig. 8B), similar to the wild-type mice response above. In contrast, E229K thrombin had no effect on the BK-induced hypotension in TAFI-deficient mice where the decrease of mean arterial blood pressure of 13.9 Ϯ 4.8 mmHg (n ϭ 6, p Ͻ 0.01, Fig. 8B) was similar to wild-type mice infused with saline followed by treatment with BK. The data indicate that TAFIa is more effective than the endogenous CPN in inactivating BK in this experimental setting. DISCUSSION A major premise of this study is that TAFIa has broad substrate specificity, and by cleaving C-terminal arginine or lysine from a number of biologically active peptides, it functions as an anti-inflammatory molecule. Initial studies by Campbell and co-workers (4,12,13) showed that TAFIa could have a role as a regulator of inflammation by inactivation of kinins and anaphylatoxins. To test this, we studied the kinetics of TAFIa hydrolysis of BK and peptides based on the thrombincleaved OPN SVVYGLR motif and the C-terminal residues of C3a, C5a and compared these to peptides based on exposed C-terminal arginine or lysine sites on plasmin-degraded fibrin clots (27). In general, BK, C3a 69 -77 , C5a 66 -74 , and the thrombin-cleaved OPN motif SVVYGLR (OPN 160 -168 ) were equally good substrates for TAFIa as those based on the fibrin peptides (FB␣Arg 104 , FB␤Lys 133 , FB␥Lys 62 , FB␤Lys 85 ), which are com-monly regarded as the physiological substrates for TAFIa in vivo. It is interesting that plasma CPN, generally thought to be the physiological inactivator of the kinins and anaphylatoxins, hydrolyzed the fibrin peptides quite efficiently, suggesting a possible role of CPN in the maintenance of fibrin clot stability. CPN is more efficient in hydrolyzing C3a than C5a, while TAFIa shows the opposite preference, consistent with previous findings (13). On the other hand, in addition to C5a, TAFIa is also much more efficient than CPN in cleaving BK and OPN 160 -168 . Our data confirm and extend the initial observations by Campbell and co-workers (4,12,13). The thrombincleaved OPN fragment demonstrated in this study, in addition to kinins such as BK and kallidin, the anaphylatoxins C3a and C5a, can all be hydrolyzed by TAFIa efficiently and are thus potential substrates for TAFIa in vivo.
To further investigate the functional role of TAFIa on OPN, kinins (BK), and anaphylatoxins (C5a), we employed several in vitro and in vivo assays. The role of TAFIa on integrin binding by thrombin-cleaved OPN was assessed using a Jurkat cell adhesion assay. Previous studies have shown that several Tlymphocyte derived cell lines, including Jurkat, HL-60, and Ramos cells that express ␣ 4 ␤ 1 integrin, have increased adhesion to OPN following thrombin cleavage and this is mediated by binding to the exposed integrin-binding motif SVVYGLR (19 -22). We confirmed that thrombin cleavage of OPN enhanced Jurkat cell binding to the SVVYGLR motif and this was mediated by ␣ 4 ␤ 1 and was independent of the RGD motif (Figs.  1-4). TAFIa treatment of thrombin-cleaved OPN and recombi-nant proteins mimicking the des-Arg N-terminal thrombincleaved OPN fragment had markedly reduced adhesion indicating the critical importance of the exposed C-terminal arginine on SVVYGLR. Of interest, both the RGD motif and the SVVYGLR motif, which are contiguous and capable of binding to different integrins, appear to be functional in the thrombincleaved OPN fragment and both contribute to the enhanced cell binding (Fig. 4) with minimal contribution by the ELVTDFPT-DLPAT binding site (22). It has been previously suggested that due to their close proximity, the RGD and SVVYGLR binding sites on a thrombin-cleaved OPN fragment may not bind to two integrins simultaneously (28). One interpretation of our data is that, while each thrombin-cleaved OPN fragment can bind only one type of integrin via either binding motif, Jurkat cells utilized different integrin molecules to bind to different OPN fragments, resulting in enhanced cell adhesion. Alternatively it is possible that the RAA mutation alters the conformation of   7. Activation of TAFI in vivo by E229K thrombin. Mice were infused with saline, wild-type thrombin or E229K thrombin at 5 g/kg/ min for 10 min. Plasma samples at the end of infusion were obtained, and TAFIa activity determined (mean Ϯ S.E., n ϭ 5). Western blot results show in the first lane the saline control, the second lane shows mice treated with wild-type thrombin (5 g/kg/min), and the third lane shows mice treated with E229K (5 g/kg/min), and is representative of three separate experiments.

FIG. 8. The effect of E229K on bradykinin-induced hypotension in vivo.
Each mouse was anesthetized and the carotid artery was cannulated with a pressure transduction catheter connected to a computerized pressure monitor to record blood pressure. Either E229K thrombin (40 g/kg) or a saline control was administered via the jugular vein. Saline was injected to flush out the catheter then BK or des-Arg BK (10 nmol/kg) was administered. Blood pressure tracings were monitored and recorded. Panel A shows the results for maximum drops in mean arterial pressure after administration of BK or des-Arg BK (BK (n ϭ 5) versus E229K followed by BK (n ϭ 6), p Ͻ 0.001; BK (n ϭ 5) versus des-Arg BK (n ϭ 4), p Ͻ 0.0001). Panel B shows the drop in mean arterial pressure for wild-type mice (littermates for knockout mice) treated with E229K followed by BK (n ϭ 4) versus TAFI knockout mice treated with E229K followed by BK (n ϭ 6) p Ͻ 0.01. either or both the SVVYGLR and ELVTDFPTDLPAT binding sites thus reducing its cell binding affinity. Functional deletion of both sites resulted in minimal Jurkat cell binding, indicating that the third integrin binding site ELVTDFPTDLPAT does not play a significant functional role, at least in this in vitro assay system (22). ␣ 4 and ␣ 9 are sole members of a subfamily of integrin ␣ subunits (29). The ␣ 4 ␤ 1 integrin (VLA-4) is predominantly expressed on leukocytes, especially lymphocytes, monocytes and eosinophils (30). On the other hand, ␣ 9 ␤ 1 is predominantly expressed on neutrophils and is also expressed on epithelial cells, smooth muscle cells and skeletal muscle. Both ␣ 4 ␤ 1 and ␣ 9 ␤ 1 can bind to the SVVYGLR site on thrombin-cleaved OPN (31). TAFIa cleavage of the C-terminal arginine largely abolished ␣ 4 ␤ 1 -mediated Jurkat cell binding (Figs. 1, 3, and 4), while its effect on the OPN fragment binding to ␣ 9 ␤ 1 -bearing cells has not been tested. Since both thrombin and OPN are present at high concentrations at sites of tissue injury and tissue repair, thrombin cleavage of OPN with the subsequent exposure of the SVVYGLR site for ␣ 4 ␤ 1 and ␣ 9 ␤ 1 binding is likely of physiological significance. Thus thrombin cleavage of OPN followed by TAFIa treatment may sequentially up-and down-modulate the pro-inflammatory properties of OPN.
An important part of the inflammatory response is the generation of BK that leads to an increase in vascular permeability and vasodilatation (32). CPN is generally regarded to be the physiological inactivator of BK (33). However cleavage studies showed that TAFIa is in fact 10-fold more efficient than CPN in hydrolyzing BK in vitro (Table I). To investigate the role of TAFIa inactivating BK in vivo, we used a mouse model where we could activate endogenous TAFI by infusing the mice with the anticoagulant thrombin E229K (25). Infusion with E229K thrombin rapidly ablated the BK-induced hypotension, presumably by activating TAFI leading to the conversion of BK to its des-Arg form (Fig. 8). This is supported by our findings that E229K thrombin infusion effectively converted TAFI to TAFIa in mice (Fig. 7) and that des-Arg BK did not produce a significant hypotensive effect (Fig. 8A). Furthermore, the protective effect of E229K thrombin was lost in the TAFI-deficient mice (Fig. 8B). Our data suggest that plasma CPN, which is constitutively active, is important for acting as a "housekeeping" carboxypeptidase whereas TAFIa, generated at the site of tissue injury, is important for inactivation of excessive BK produced under inflammatory conditions. Previously, using a kaolin-induced writhing model, in which kaolin was administered intraperitoneally to activate factor XII and prekallikrein, TAFI-deficient mice failed to show any increased writhing response as compared with the wild-type mice, suggesting that either TAFIa does not play a significant role in modulating the activity of BK or that TAFI is not activated in this model (34). The kaolin writhing model is very different from the BK-induced hypotension model used in this study. In the former, BK was generated indirectly in an extravascular compartment and the exact amount of BK generated undetermined. In addition there is no data on whether TAFI is activated in this model. In our model, a specific amount of BK was directly infused intravenously and TAFI was activated in the same compartment. Whether the effect of TAFIa on BK is limited to the vascular compartment needs to be explored in future studies.
Recently a single nucleotide polymorphism (SNP) in the coding region of the CPB2 gene, which encodes TAFI, was found in its homozygous form to be associated with lower diastolic blood pressure in a study of aboriginal people in Canada (35). This SNP, designated as 1057C 3 T, results in an amino acid change at TAFI residue 325 from isoleucine to threonine, and the [I325]TAFIa variant is twice as stable as the wild-type TAFIa at room temperature. The efficiency of this TAFIa variant in hydrolyzing BK and whether it plays a role in accounting for the lower diastolic blood pressure remain to be studied.
Also important in the inflammatory response is the generation of C3a and C5a that are potent leukocyte chemoattractants. Interestingly, C5a blockade has a protective effect in sepsis (36). Kinetic analysis of TAFIa revealed both C3a and C5a to be good substrates for TAFIa (Table I), and an in vitro assay showed that TAFIa can efficiently inactivate C5a and reduce its ability to activate neutrophils (Fig. 5). This suggests that in vivo, TAFIa could inactivate both C3a and C5a by hydrolysis of their C-terminal arginines thereby reducing the proinflammatory effects of these potent anaphylatoxins.
Current interest in TAFI is mainly focused on its role as a fibrinolysis inhibitor. While this notion is amply supported by in vitro studies, the in vivo role of TAFI remains uncertain. TAFI-deficient mice had no gross phenotypic abnormalities and no differences in the rate of endogenous clot lysis could be demonstrated using a variety of acute or subacute clot lysis models (34). On the other hand, enhanced pulmonary clot lysis was found in TAFI deficiency superimposed on a partial plasminogen deficiency setting (37). It is interesting that in these compound deficient mice, increased number of leukocytes was demonstrated in thioglycollate-induced peritoneal inflammation. While the increased leukocyte influx could be due to enhanced cell migration secondary to unimpeded cell surface fibrinolysis, it is also possible that it is the result of increased inflammation because of TAFI deficiency. These two possibilities are not mutually exclusive. Since E229K thrombin can effectively activate TAFI in mice, it becomes a powerful tool to study the role of TAFI in vivo by comparing its effect in wildtype and TAFI-deficient mice.
Taken together, our data suggest that TAFIa has broad substrate specificity and its function may not be restricted to inhibition of fibrinolysis. At sites of inflammation, thrombin not only has proinflammatory effects but also can act as an indirect anti-inflammatory molecule by binding TM and activating PC and TAFI. TAFIa then inactivates kinins, anaphylatoxins and thrombin-cleaved OPN. Hence, thrombin plays a key role in regulating coagulation and inflammation through the intricate interplay between thrombin activation of proinflammatory molecules and thrombin/TM-dependent activation of anticoagulant and anti-inflammatory effectors.