Inhibition of Plasminogen Activation by Lipoprotein(a)

Similarity between the apolipoprotein(a) (apo(a)) moiety of lipoprotein(a) (Lp(a)) and plasminogen suggests a potentially important link between atherosclerosis and thrombosis. Lp(a) may interfere with tissue plasminogen activator (tPA)-mediated plasminogen activation in fibrinolysis, thereby generating a hypercoagulable state in vivo. A fluorescence-based system was employed to study the effect of apo(a) on plasminogen activation in the presence of native fibrin and degraded fibrin cofactors and in the absence of positive feedback reactions catalyzed by plasmin. Human Lp(a) and a physiologically relevant, 17-kringle recombinant apo(a) species exhibited strong inhibition with both cofactors. A variant lacking the protease domain also exhibited strong inhibition, indicating that the apo(a)-plasminogen binding interaction mediated by the apo(a) protease domain does not ultimately inhibit plasminogen activation. A variant in which the strong lysine-binding site in kringle IV type 10 had been abolished exhibited substantially reduced inhibition whereas another lacking the kringle V domain showed no inhibition. Amino-terminal truncation mutants of apo(a) also revealed that additional sequences within kringle IV types 1–4 are required for maximal inhibition. To investigate the inhibition mechanism, the concentrations of plasminogen, cofactor, and a 12-kringle recombinant apo(a) species were systematically varied. Kinetics for both cofactors conformed to a single, equilibrium template model in which apo(a) can interact with all three fibrinolytic components and predicts the formation of ternary (cofactor, tPA, and plasminogen) and quaternary (cofactor, tPA, plasminogen, and apo(a)) catalytic complexes. The latter complex exhibits a reduced turnover number, thereby accounting for inhibition of plasminogen activation in the presence of apo(a)/Lp(a).

Elevated plasma levels of lipoprotein(a) (Lp(a)) 1 have been identified as a risk factor for the development of atherosclerotic disorders including coronary artery disease (1). Originally described by Berg (2), Lp(a) is similar to low density lipoprotein in that it contains a cholesteryl ester core surrounded by a monolayer of unesterified cholesterol and phospholipid in which apolipoprotein B-100 (apoB-100) is embedded. Lp(a) is distinct from low density lipoprotein in that it contains a single molecule of an additional glycoprotein moiety called apolipoprotein(a) (apo(a)), which is covalently linked to apoB-100 by a single disulfide bond (3,4). The cloning of apo(a) revealed a striking similarity between apo(a) and the fibrinolytic serine protease zymogen plasminogen (5).
The apo(a) moiety of Lp(a) lacks sequences corresponding to the plasminogen kringle (K) 1, 2, and 3 domains, yet contains multiple repeats of the plasminogen K4-like domain (apo(a) KIV) followed by plasminogen K5-like and protease-like sequences (apo(a) KV and protease-like (P) domains, respectively) (5). There are 10 distinct classes of apo(a) KIV repeats (designated KIV 1 -KIV 10 ) that differ from one another by only a small number of amino acid substitutions (5). There is a single copy of each KIV repeat except for KIV 2 , which is known as the major repeat kringle (6,7). This sequence can be present in 3 to greater than 30 identical copies that gives rise to Lp(a) isoform size heterogeneity in the population (8). Apo(a) KIV 5-8 each contain a weak lysine binding site, which mediate the initial non-covalent interaction between apo(a) and apoB-100 in Lp(a) assembly (9 -11). Apo(a) KIV 9 contains an unpaired cysteine residue that forms the disulfide linkage with apoB-100 in Lp(a) assembly (3,4). Apo(a) KIV 10 contains a stronger lysine binding site that is believed to mediate binding of apo(a)/Lp(a) to lysine residues exposed on the surface of biological substrates such as fibrin (12,13). The apo(a) protease-like domain exhibits no catalytic activity (14).
The similarity of Lp(a) to both low density lipoprotein and plasminogen suggests that the pathogenic mechanism of Lp(a) likely involves both proatherogenic and prothrombotic/antifibrinolytic effects (1). Several in vitro and in vivo studies have shown that apo(a)/Lp(a) can inhibit fibrinolysis (15)(16)(17). Moreover, binding interactions have been demonstrated between apo(a)/Lp(a) and fibrin(ogen), plasmin-modified fibrin(ogen), plasminogen, and tPA (13, 18 -22). Early studies demonstrated that apo(a)/Lp(a) can inhibit the binding of plasminogen and tPA to fibrin (13,18), but the mechanism by which apo(a) inhibits plasminogen activation remains controversial as both competitive (23) and uncompetitive (18,24) mechanisms have been reported. Efficient activation of plasminogen requires the formation of a ternary complex between substrate (plasminogen), enzyme (tPA), and cofactor (fibrin) (25,26). Kinetic studies of plasminogen activation are complicated by the existence of plasminmediated positive feedback reactions. First, limited plasmin cleavage of fibrin results in the exposure of carboxyl-terminal lysine residues in the cofactor that enhance plasminogen binding (27). Second, single chain tPA is cleaved by plasmin to generate the two-chain form. Third, plasmin cleaves native Glu 1 -plasminogen after Lys 77 to form Lys 78 -plasminogen, which is a better substrate for tPA (28). Formation of Lys 78plasminogen is, in turn, stimulated by partially degraded, but not native, fibrin.
In the present study, we adapted a methodology previously described by Horrevoets and co-workers (29) that allows plasminogen activation to be measured directly in a fibrin clot and in the absence of positive feedback reactions catalyzed by plasmin. Briefly, a recombinant variant of human Glu 1 -plasminogen, in which the active site serine has been replaced by cysteine, was employed such that the zymogen could be labeled with a thiol-specific fluorescein tag. The labeled protein remains catalytically inactive upon cleavage by tPA. Using a series of recombinant apo(a) (r-apo(a)) variants in this system, we have delineated the domains in apo(a) required for inhibition of plasminogen activation and have described the inhibition mechanism as an equilibrium template model.
Construction, Expression, and Purification of Recombinant Apo(a) Variants-The construction and expression of the various r-apo(a) variants utilized in our study have been described previously (10,11,21,32,33), with the following exception. For construction of the 17K r-apo(a) variant lacking the KV domain (17K(ϪV) variant), the pRK5ha17 plasmid (32) was digested with AvrII and BstEII, thereby generating three fragments of 9974, 1018, and 660 bp. The 660-bp fragment (encoding the apo(a) KIV 10 , KV, and P domains) was replaced with a 425-bp PCR product containing the apo(a) KIV 10 and P domain, but lacking the kringle V domain, as follows. Poly(A) ϩ RNA was extracted from baboon liver as previously described (34), and reverse transcribed using random hexanucleotide primers to generate first-strand cDNA. Using the baboon liver cDNA as template, PCR amplification was performed using the following primers: primer A (5Ј-CCAAGCCTAGGGGCTCCT-TCTGAACAA-3Ј) flanking the AvrII site at nucleotide position 324 in the human KIV 10 sequence (5) and primer B (5Ј-TCCGGTCGGTGAC-CACATAATTTGGGG-3Ј) flanking the BstEII site in the apo(a) protease-like domain (5). A 425-bp PCR product was obtained and cloned into pBluescript SKϩ for DNA sequence analysis. The resultant sequence was found to correspond to baboon apo(a) and contained the 3Ј end of the sequence encoding apo(a) KIV 10 , followed by the apo(a) protease-like domain. 2 The 425-bp PCR product was digested with AvrII and BstEII and ligated with the 9974-and 1018-bp pRK5ha17derived BstEII/AvrII fragments. The final expression plasmid encoding the 17K(ϪV) r-apo(a) variant was designated pRK5ha17(ϪV). It should be noted that the AvrII site resides in the region encoding the interkringle region. Thus, KIV 10 in the 17K(ϪV) construct consists entirely of human sequence. The identity of all r-apo(a) variants, including 17K(ϪV), have been verified by DNA sequence analysis.
All r-apo(a) derivatives were purified from the conditioned medium of stably expressing human embryonic kidney (HEK 293) cell lines by affinity chromatography as described previously (11,21). Briefly, conditioned medium (Opti-MEM) harvested every 3 days from the stably transfected cells was loaded over a 50-ml lysine-Sepharose column pre-equilibrated in PBS. The column was washed extensively with PBS containing 500 mM NaCl and r-apo(a) was eluted with 200 mM ⑀-ACA in the same buffer. Protein-containing fractions were pooled, dialyzed extensively against HBS, and concentrated using PEG 20000. Aliquots of the purified proteins were stored at Ϫ70°C.
All r-apo(a) protein concentrations were determined spectrophotometrically (corrected for Rayleigh scattering) using the following molecular weights and extinction coefficients (21)  Preparation of Fluorescently Labeled Recombinant Plasminogen-The construction and expression of a variant of native Glu 1 -plasminogen, containing an active site serine to cysteine mutation (rplasminogen(S741C)), has been described previously (29). Likewise, r-plasminogen(S741C) protein was purified from the conditioned medium of stably expressing baby hamster kidney (BHK-21) cells as described previously (29), but with modifications. Briefly, conditioned medium (Opti-MEM) containing 50 M ZnCl 2 was harvested from the stably transfected cells every 2 days and treated with reduced GSH (1 mM final), Tris, pH 8.0 (10 mM final), and dEGR-CK (1 M final). Pooled harvests were loaded onto a 40-ml lysine-Sepharose column pre-equilibrated in PBS. The column was washed extensively with PBS containing 1 mM EDTA and r-plasminogen(S741C) was eluted with 20 mM ⑀-ACA in the same buffer. Protein-containing fractions were pooled and then fluorescein-labeled by incubation with a 50-fold molar excess of 5Ј-iodoacetamidofluorescein at 4°C overnight in the dark. Free 5Јiodoacetamidofluorescein label was subsequently removed by passage over an 8-ml DEAE-cellulose Fast Flow column pre-equilibrated in PBS containing 0.02% (v/v) Tween 80 (PBST). Labeled protein (Flu-plasminogen) recovered in the flow-through was dialyzed extensively against 10 mM Tris-HCl, pH 8.0, containing 0.02% (v/v) Tween 80 and then loaded over a similar DEAE column (30 ml) pre-equilibrated in the same buffer. The column was washed extensively with 20 mM HEPES, pH 7.4, and Flu-plasminogen was eluted in HBS containing 0.02% (v/v) Tween 80 (HBST). Protein-containing fractions were pooled and aliquots were stored in the dark at Ϫ70°C. All chromatographic and dialysis steps were performed in the dark at 4°C.
The amount of fluorescein incorporated was determined spectrophotometrically as previously described (29); labeling efficiency was typically greater than 80%. The Flu-plasminogen protein was assessed for purity by SDS-PAGE under non-reduced and reduced (containing 10 mM dithiothreitol) conditions, followed by Coomassie Blue staining. All purified Flu-plasminogen preparations exhibited single bands under both non-reduced and reduced conditions and migrated at the expected molecular weight.
Purification of Human Fibrinogen and Non-cross-linked FDPs-Fibrinogen was purified from units of citrated, fresh frozen, human plasma as described previously (35), but with modifications. Briefly, all steps were performed at room temperature with the exception of the final dialysis performed at 4°C. Initially, 80 ml of 1.0 M BaCl 2 was added to 1.0 liter of plasma over 5 min with stirring. The solution was stirred for an additional 30 min and then centrifuged at 5000 ϫ g for 30 min. A 1.0 M stock of diisopropyl fluorophosphate was added to the supernatant (to 1 mM final) and stirred for 10 min. A 4.0 M ␤-alanine stock in TSC buffer (50 mM sodium citrate, pH 6.5, 150 mM NaCl) was then added to the solution over 5 min with stirring (to 1.0 M final), stirred an additional 30 min, and centrifuged at 5000 ϫ g for 30 min. ␤-Alanine was added to the supernatant again (to 2.0 M final), stirred, and centrifuged as before. The pellet was dissolved in 1.0 liter of TSC buffer and ␤-alanine added again (to 2.0 M final), stirred, and centrifuged as before. The pellet was dissolved in 250 ml of TSC buffer and PEG 8000 (40% (w/v) stock in water) was added to the solution over 15 min with stirring (to 1.2% final), stirred an additional 30 min, and centrifuged at 5000 ϫ g for 30 min. PEG 8000 was added to the supernatant again over 15 min with stirring (to 5.0% final), stirred an additional 30 min, and centrifuged at 5000 ϫ g for 30 min. The pellet was dissolved in 100 ml of 20 mM HEPES, pH 7.4, 500 mM NaCl by rapid agitation on an orbital shaker table for 2 h. Subsequently, the 100-ml solution was diluted in 20 mM HEPES, pH 7.4 (to 25 mM NaCl final), before passage over 5-ml lysine-Sepharose and 100-ml DEAE-cellulose fibrous form columns. Both columns were pre-equilibrated in 20 mM HEPES, pH 7.4, 25 mM NaCl and were linked in tandem. The columns were washed with 50 ml of the same equilibration buffer and the DEAE column was disconnected and washed further with 300 ml of the same buffer. Fibrinogen was eluted from the DEAE column using 20 mM HEPES, pH 7.4, 100 mM NaCl and further fractionated to yield high molecular weight fibrinogen (containing intact ␣-chains) by precipitation with 19% ammonium sulfate as described previously (36). The purified fibrinogen was dialyzed extensively against HBS, passed through a 5-m syringe filter, and aliquots were stored at Ϫ70°C.
Non-cross-linked FDPs were prepared by plasmin-mediated lysis of fibrin clots as described previously (37), but with modifications. Briefly, all steps were performed at room temperature with the exception of the final dialysis performed at 4°C. In HBS containing 0.02% (v/v) Tween 80 (HBST), 50 mg of fibrinogen was clotted with 5 mM CaCl 2 and 20 nM thrombin in the presence of 20 nM plasmin. Clot formation (occurred within 1 min) and subsequent fibrinolysis were monitored spectrophotometrically at 600 nm. The reaction was allowed to proceed to ϳ75% total lysis and then terminated by the addition of VFK-CK (5 M final) and PPA-CK (1 M final). Remaining aggregates were removed by centrifugation at 13,000 rpm for 5 min. The resultant FDPs were finally dialyzed extensively against 20 mM HEPES, pH 7.4, 500 mM NaCl, passed through a 5-m syringe filter, and stored at 4°C.
Fibrinogen and FDP protein concentrations were determined spectrophotometrically (corrected for Rayleigh scattering) using the following molecular weights and extinction coefficients (38,39) for fibrinogen (M r ϳ 340,000; ⑀ 180 nm 1% ϭ 16.0) and FDP (M r ϳ 250,000; ⑀ 180 nm 1% ϭ 16.0). The proteins were assessed for purity by SDS-PAGE under non-reduced and reduced (containing 10 mM dithiothreitol) conditions, followed by Coomassie Blue staining. Purified high molecular weight fibrinogen (M r ϳ 340,000) exhibited a single, homogeneous band under non-reduced conditions and characteristic A␣-(M r ϳ 64,000), B␤-(M r ϳ 57,000), and ␥-(M r ϳ 48,000) chain bands under reduced conditions. Subsequent conversion of fibrinogen into non-cross-linked FDPs showed that the final mixture exhibited characteristic fragment X (M r ϳ 260,000), Y (M r ϳ 160,000), and D (M r ϳ 100,000) bands under non-reduced conditions and characteristic ␤and ␥-chain bands only under reduced conditions. Fluorescent Plasminogen Activation Assays-Cleavage of Flu-plasminogen by tPA was followed at room temperature in 96-well Mi-crofluor2 plates (Dynex), pre-equilibrated with HBS containing 1% (v/v) Tween 80 for 1 h to prevent protein adsorption to the plastic. LinbroR plate sealers (ICN, Montreal, PQ, Canada) were placed over the 96-well plates to prevent evaporation in the wells and fluorescence was monitored using a SpectraMax Gemini XS plate reader (Molecular Devices, Sunnyvale, CA). For all studies, the method parameters were as follows: excitation wavelength, 495 nm; emission wavelength, 535 nm; cut-off wavelength, 530 nm; sensitivity setting, normal; PMT setting, low; run time, 1 h; interval, 36 s.
In all cases, raw fluorescence measurements were corrected for the buffer blank (HBS containing 0.02% (v/v) Tween 80) and corrected for internal filter effects as follows. First, standard Flu-plasminogen curves were generated (0 -3 M) for each purified preparation and fit by nonlinear regression (SYSTAT, Evanston, IL) to Equation 1, where I is the relative fluorescence intensity, i is the amount of fluorescence per mole of Flu-plasminogen, [P] is the concentration of Fluplasminogen, a is the exponential coefficient, and I is the raw fluorescence measured. The input data for the non-linear regression were [P] and I and the best-fit parameters returned were a and i. Second, all raw fluorescence measurements were converted to corrected values according to Equation 2 using the appropriate best fit a and i parameters, where I corr is the corrected fluorescence intensity, I raw is the measured fluorescence intensity, a is the exponential coefficient, and i is the amount of fluorescence per mole of Flu-plasminogen. Initial rates of fluorescence decrease were then determined by linear regression analysis and converted to rates of plasminogen activation according to Equation 3, where d[Pn]/dt is the rate of plasmin formation per mole of tPA (s Ϫ1 ), dI/dt is the initial rate of fluorescence decrease, r is the relative maximum change in fluorescence intensity (0.5 for fibrin surface (29) and 0.4 -0.5 as determined empirically for FDP solution), I 0 is the initial fluorescence intensity, [P] 0 is the initial Flu-plasminogen (substrate) concentration, and [A] 0 is the initial tPA (activator) concentration. In all cases, rates of plasminogen activation reported represent the average of duplicate measurements. A Model for the Modulation by Apo(a) of Fibrin-dependent Plasminogen Activation-Our data indicate that apo(a) both inhibits and promotes plasminogen activation, depending on the concentrations of plasminogen, fibrin, and apo(a) (see "Results"). At most concentrations studied, inhibition was observed, but at very low concentrations of clotted fibrinogen, stimulation occurred. Efforts were made to rationalize these observations by taking into account the known interactions between the various components and the template mechanism for fibrindependent plasminogen activation (26). Among the known interactions are fibrin/plasminogen, fibrin/tPA, fibrin/apo(a), apo(a)/plasminogen, fibrin/plasminogen/apo(a), and apo(a)/tPA. Numerous models were constructed that included only binary interactions of apo(a) with fibrin, tPA, or plasminogen. None of these could adequately account for all of the experimental observations, particularly the stimulation of plasminogen activation by apo(a) at very low concentrations of clotted fibrinogen. Eventually, however, a model was found that accounts for all of the experimental observations. In the absence of apo(a), the model comprises the template model described by Horrevoets and co-workers (26). In this model, either the plasminogen substrate (P) or the tPA activator (A) can bind to the fibrin cofactor (F) to form the binary FP or FA complexes. These can then bind the third component to form the ternary FAP complex, from which plasmin is generated with a first-order rate constant k 1 . When the apo(a) (inhibitor) (I) is included, the assumption is made that I can bind F to form a IF complex, and this can function as a template analogous to F, such that P can bind IF to form IFP, and A can bind FI to form IFA. These ternary complexes can then bind the fourth component to form the quaternary complex, IFAP, from which plasmin is generated with first-order rate constant k 2 . These concepts, along with the equilibrium binding expressions and conservation equations for each of the species involved in the process, allow derivation of the rate equation for the model as follows.
Expressed in terms of the concentrations of the FAP and IFAP complexes, the rate of plasminogen activation (r) is given by Equation 4, The binding interactions and equilibrium expressions associated with assembly of FAP and IFAP are defined as follows.
In this equation k cat1 , K m1 , k cat2 , and K m2 are as defined above. K A is the dissociation constant for the binding of tPA to fibrin, and K IA is the dissociation constant for the binding of tPA to the fibrin-apo(a) complex (K IA ϭ K 6 K 1 /K 10 ).
To evaluate Equation 24, the free concentrations of F, P, and I are needed. These are found from the equilibrium binding expressions and conservation equations for total F, P, and I. The conservation equations, except for terms containing A, which are negligibly small because A was used only in trace concentrations, are Equations 25-27. were provided that were equal to the average of the calculated value and the previously assigned value. This process was repeated until the calculated and assigned values differed by no more than 1 part in 10 8 . This typically required 10 to 20 iterations and provided solutions to all three equations accurate to 1 part in 10 8 . Convergence was always rapid and robust because if an assigned value were larger than the "true" value, the calculated value was smaller, and vice versa. Thus, averaging guaranteed rapid convergence. The value of K 1 (binding of apo(a) to fibrin) was 1.4 M (20). The value for K 3 (binding plasminogen to fibrin) was 30 M when intact fibrin was used (29) and 0.225 M when fibrin degradation products were used (40). The value for K 11 (the binding of apo(a) to plasminogen) was 20 nM (21). The value for K 7 (the binding of plasminogen to the apo(a)-fibrin complex) was 25-fold greater than the value for binding of plasminogen to fibrin or fibrin degradation products, which reflects the observation the apo(a) negatively influences the binding of plasminogen to fibrin (21  The effects of the different r-apo(a) variants on tPA-mediated plasminogen activation, in the presence of fibrin or FDPs, is summarized in Fig. 3. For the most part, there was little difference in the magnitudes of the inhibitory effects in the presence of the respective cofactors. Notable differences in-cluded a minor inhibitory effect of 6K in the presence of fibrin but none in the presence of FDPs, a greater inhibitory effect of 12K in the presence of FDPs than in the presence of fibrin, and a greater inhibitory effect of Lp(a) in the presence of fibrin than in the presence of FDPs.

Analysis of the Mechanism by Which Apo(a) Inhibits
Plasminogen Activation-The kinetics of plasminogen activation are complicated by the role of the fibrin/FDP cofactor. Therefore, kinetic analysis of inhibition of plasminogen activation by apo(a) required us to perform initial rate experiments in which the concentrations of cofactor, plasminogen, and r-apo(a) were systematically varied. In these studies, apo(a) was titrated over a higher concentration range (0 -3.0 M) than that employed in the domain studies (0 -1.4 M). At fixed cofactor concentrations, 12K r-apo(a) inhibited plasminogen activation in a dosedependent manner in the presence of both fibrin (Fig. 4A) and FDPs (Fig. 5A).
At fixed Flu-plasminogen concentrations, apo(a) was then titrated over variable fibrin and FDP concentration ranges. Interestingly, 12K r-apo(a) inhibited plasminogen activation in a dose-dependent manner at fibrin concentrations of 0.56 -3.0 M (Fig. 4B) while below 0.5 M fibrin 12K r-apo(a) in fact stimulated plasminogen activation in a dose-dependent manner. On the other hand, 12K r-apo(a) inhibited plasminogen activation in a dose-dependent manner over all FDP concentrations tested (Fig. 5B). The data (Figs. 4 and 5) were first analyzed by non-linear regression according to Michaelis-Menten kinetics (data not shown). This approach did not provide good global fits to any of the four standard types of inhibition (competitive, non-competitive, uncompetitive, or mixed). Thus, non-linear regression of the rate data to an equilibrium template model of plasminogen activation (Equation 24) was performed. In this model (Fig. 6), all three fibrinolytic components (cofactor, tPA, and plasminogen) are allowed to interact with the apo(a) inhibitor. Global fits of the fibrin and FDP data to this model resulted in regression lines that correlated well with the respective experimental data (Figs. 4 and 5). Of special note, both cofactor series fit equally well to the same model and the model predicted the profibrinolytic behavior of apo(a) that was observed at low fibrin concentrations (Fig. 4B).
Examination of the kinetic parameters predicted by the equilibrium template model (Table I) reveals several notable findings. Michaelis constants (K m ) for the apo(a)-containing quaternary catalytic complexes (IFAP) were significantly lower than those for the normal fibrinolytic ternary (FAP) complexes in the presence of fibrin (0.003 M versus 1.35 M, respectively) and FDPs (0.001 versus 0.034 M, respectively). Turnover numbers (k cat ) for the IFAP complexes were also lower than those for the FAP complexes in the presence of fibrin (0.050 s Ϫ1 versus 0.112 s Ϫ1 , respectively) and FDPs (0.038 s Ϫ1 versus 0.045 s Ϫ1 , respectively). Dissociation constants for the interac-tion of tPA with fibrin or FDPs were similar in the presence (K IA ) or absence (K A ) of apo(a). Specifically, the dissociation constants for the interactions of tPA with fibrin or FDPs (K A ) were 0.055 and 1.93 M, respectively. The first value is in reasonable agreement with the value reported by Horrevoets and co-workers (26) and the latter agrees with values reported by Walker and Nesheim (40) for FDPs with relatively lower molecular weights below 1 ϫ 10 6 . Finally, the dissociation constants for the solution-phase apo(a)-tPA interaction (K 10 ) predicted by the model were 3.65 M with fibrin and 4.55 M with FDPs.

DISCUSSION
It is generally accepted that apo(a)/Lp(a) is capable of inhibiting fibrinolysis, but the mechanism by which this inhibition occurs remains controversial. In addition, the domains in apo(a) that mediate inhibition of fibrinolysis are not known. We addressed these issues in the current study using a series of recombinant apo(a) variants and a system to measure tPAmediated Glu 1 -plasminogen activation kinetics in the absence of plasmin-mediated positive feedback reactions.
In the absence of apo(a), the observed rates of plasminogen activation agreed well with values reported in the literature.  (40). The latter finding is significant as it indicates that our preparation of non-cross-linked, soluble FDPs possesses similar intrinsic cofactor activity to the cross-linked, soluble FDPs described by Walker and Nesheim (40). These latter species display similar tPA cofactor activities over a range of FDP sizes (weight-average molecular weight 0.48 ϫ 10 6 to 4.94 ϫ 10 6 ). In agreement with the findings of Walker and Nesheim (40), we found that the FDPs show a similar k cat but a substantially reduced (ϳ10-fold) K m relative to fibrin, which accounts for the enhanced cofactor activity of the FDPs. As such, our non-cross-linked FDPs likely represent an excellent surrogate for the large soluble FDPs as well as for insoluble, partially degraded, fibrin.
The carboxyl-terminal lysine residues present in partially degraded fibrin serve to mediate positive feedback in the fibrinolytic cascade by (i) promoting the binding and activation of plasminogen, (ii) promoting the plasmin-mediated conversion of Glu 1 -plasminogen to Lys 78 -plasminogen, the latter of which is a better substrate for plasminogen activators, and (iii) by binding plasmin and thus protecting it from consumption by its major plasma inhibitor ␣ 2 -antiplasmin (41). Activated thrombin-activable fibrinolysis inhibitor is a recently described enzyme that removes the carboxyl-terminal lysines from partially degraded fibrin and thus attenuates fibrinolysis through suppression of positive feedback in the fibrinolytic cascade (reviewed in Ref. 42). It has been shown that partial plasmin digestion of fibrin increases the capacity of fibrin for Lp(a) and plasminogen, but not its affinity for these molecules (13). Thus, we would predict that the presence of activated thrombin-activable fibrinolysis inhibitor in our system would decrease the amount of both apo(a) and plasminogen bound to fibrin. Further experimentation will be required to determine the impact of the activated thrombin-activable fibrinolysis inhibitor pathway on inhibition of plasminogen activation by apo(a).
Our data do not indicate that there is direct competition of plasminogen and apo(a) for binding to fibrin or FDPs. This is in keeping with the results of Sangrar and co-workers (21), who found that plasminogen and apo(a) bind to distinct sites on partially degraded fibrinogen and proposed that the solutionphase interaction between plasminogen and apo(a) results in a complex that binds less avidly to plasminogen binding sites. Our data also indicate that apo(a) would not compete with plasmin for fibrin binding and thus would not potentiate plasmin consumption by ␣ 2 -antiplasmin. However, we cannot rule out that the presumptive solution-phase interaction between apo(a) and plasmin might influence ␣ 2 -antiplasmin interaction with plasmin.
It was found that the apo(a) domains required for inhibition of plasminogen activation with fibrin were identical to those required for inhibition with FDPs. We discovered that the lysine-binding KIV 10 domain and an intact KV domain were both required for maximal inhibition. Decreased inhibition upon mutation (17K(D56A)) or deletion (17K(ϪV)) of the respective domains supports the notion that the strong lysinebinding site in apo(a) KIV 10 alone cannot account for the interaction of Lp(a) with fibrinogen; this is in keeping with reports by Edelstein and co-workers (19,43) that provide evidence for a fibrinogen-binding domain of apo(a) that is outside the lysinebinding site in KIV 10 . A variant lacking the protease domain (17K(ϪP)) was also a potent inhibitor-like wild type 17K indicating that the high affinity, solution-phase, apo(a)-plasminogen binding interaction mediated by this domain (21) is not ultimately responsible for inhibition of plasminogen activation in the presence of fibrin or FDPs. Our model, however, does take this solution-phase interaction into account as a determinant of the concentration of free plasminogen.
Interestingly, KV is a kringle domain that is missing in apo(a) from some species including rhesus monkey (44) and baboon. 3 The kringle V-protease domains have also been recently identified as a fibrinogen-binding region within apo(a) because a KV-P fusion construct was shown to compete for the binding of 125 I-labeled Lp(a) to plasmin-modified fibrinogen (45). These studies, along with our identification of the critical role for KV in inhibition of plasminogen activation, suggest 3 Equation 24 by non-linear regression. k cat1 is the turnover number for the ternary catalytic complex in the absence of apo(a); k cat2 is the turnover number for the quaternary catalytic complex in the presence of apo(a); K A and K IA are the dissociation constants for binding of tPA to the cofactor in the absence or presence of apo(a), respectively; K m1 is the Michaelis constant for the formation of the ternary catalytic complex; K m2 is the Michaelis constant for the formation of the quaternary catalytic complex; and K 10 is the dissociation constant for the solution-phase apo(a)-tPA interaction. Results are presented as best fit values Ϯ the asymptotic standard error returned by the regression algorithm.  that rhesus monkey and baboon may possess a less pathogenic, less antifibrinolytic form of Lp(a) in the absence of the KV domain. Indeed, rhesus monkeys normally exhibit almost no atherosclerosis unless challenged with a high fat and high cholesterol diet (46,47), and such animal studies have detected apo(a) in atherosclerotic plaques but without co-localization with fibrin(ogen) (48,49).
Because 17K, 12K, and 10K r-apo(a) all inhibit plasminogen activation to similar extents in our study, the number of KIV 2 repeats present (i.e. apo(a) isoform size heterogeneity) did not appear to be critical to the inhibition process. Because smaller apo(a) isoforms have been shown to have a greater inhibitory effect on plasmin generation (50,51), we speculated that 10K r-apo(a) might be a more potent inhibitor than either 12K or 17K r-apo(a). Although this was not observed in our study, it should be noted that all of the r-apo(a) variants that we examined correspond to "small apo(a) isoforms" (M r Ͻ 580,000) in the study of Falco and co-workers (50). In addition, it has been demonstrated that the isoform size dependence of fibrin binding is a property of Lp(a) but not apo(a) (52).
Despite the presence of intact KIV 10 and KV domains, deletion of the amino-terminal apo(a) KIV 1-4 domains (especially KIV 3 and KIV 4 ) resulted in dramatically decreased inhibition with either the fibrin or FDP cofactors (compare 17K to 6K r-apo(a); Fig. 3). Whereas no functional lysine-binding sites are predicted to exist in the KIV 1-4 domains (53), other non-lysinedependent interactions involving these domains may help stabilize or orient apo(a) at the site of a fibrin clot to elicit inhibition of plasminogen activation through the KIV 10 and KV domains. Likewise, the present study has shown that the high affinity, solution-phase, apo(a)-plasminogen binding interaction (mediated by the apo(a) protease domain and originally hypothesized to result in decreased ternary complex formation and decreased plasmin generation (21)) is not sufficient to inhibit plasminogen activation alone. Therefore, apo(a)-mediated inhibition of plasminogen activation in the presence of fibrin/FDPs is likely a complex function of contributions from several apo(a) structural domains. Although our kinetic model does not specifically identify the components in the quaternary catalytic complex that are involved in direct binding interactions with apo(a), it is conceivable that the requirement for multiple domains in apo(a) for inhibition reflects the interaction of apo(a) with multiple components in the quaternary complex.
Analysis according to Michaelis-Menten kinetics failed to provide good global fits for the experiments in which the concentrations of plasminogen, fibrin/FDPs, and 12K r-apo(a) were systematically varied (Figs. 4 and 5). This is probably because of the large number of components and interactions involved in plasminogen activation and its inhibition. The traditional Michaelis-Menten model does not strictly apply to a threecomponent template mechanism. Accordingly, earlier attempts to describe apo(a)/Lp(a)-mediated inhibition of plasminogen activation by Michaelis-Menten kinetics may account for the apparent variable modes of inhibition (i.e. competitive, uncompetitive) reported in the literature to date (18,23,24). However, the uncompetitive inhibition kinetics described previously (18,24) are compatible with the interaction of apo(a) with the catalytic complex that our equilibrium template model implies.
Because plasminogen activation can be adequately described by a template mechanism, the data were analyzed according to the model of Horrevoets and co-workers (26). In this model (Fig.  6), activator (A) and substrate (P) can bind to the cofactor surface (F) in either order; subsequent ternary complex formation (FAP) leads to efficient plasmin generation. Accordingly, the template model was revised in the present work to allow apo(a) to interact with all three fibrinolytic components simultaneously. Plasminogen (P) and tPA (A) are still envisaged to bind to the apo(a)-bound cofactor surface (IF) in either order; subsequent quaternary complex formation (IFAP) still results in plasmin generation.
We developed a rate equation (Equation 24) to describe the revised, equilibrium template model. Non-linear regression of the data to this equation resulted in very good global fits for both the fibrin and FDP data (Figs. 4 and 5). Interestingly, Michaelis constants (K m ) for the apo(a)-bound quaternary (IFAP) complexes were significantly lower than those for the corresponding ternary (FAP) complexes lacking apo(a) with the fibrin and FDP cofactors. These very low values are likely because of the very high affinity interaction of apo(a) with plasminogen and the fact that K m reflects the free rather than the total concentration of plasminogen. On the other hand, turnover numbers (k cat ) for the IFAP complexes were lower than those for the FAP complexes on the fibrin and FDP surfaces. Essentially, more stable but less catalytically robust catalytic complexes are formed in the presence of apo(a). Despite the nominal increase in catalytic efficiency of apo(a)containing catalytic complexes, inhibition of plasminogen activation in the presence of apo(a) is observed at physiological concentrations of fibrinogen because of the lower k cat value. Conversely, the apparent stimulation of plasminogen activation by apo(a) at low fibrinogen concentrations would be explained by the lower K m for IFAP complexes than FAP complexes. Interestingly, one study reported that Lp(a) increased the rate of plasma clot lysis (54); however, in this work the plasma was significantly diluted (to 13.3% of the final volume) which, in the context of our finding of enhanced plasminogen activation at low fibrin concentrations (Fig. 4B), might account for these apparently paradoxical results. Dissociation constants for the binding of tPA to fibrin or FDPs remained the same in the presence (K IA ) or absence (K A ) of apo(a), indicating that tPA binding was unaltered by the presence of apo(a) bound to the cofactor. This might imply that there is no direct interaction between apo(a) and tPA in the quaternary complex. Finally, relatively large dissociation constants were predicted by the model for the apo(a)-tPA interaction (K 10 ) in solution, indicative of a relatively weak interaction between apo(a) and tPA in these settings.
In summary, we have defined, for the first time, the structural domains in apo(a) that mediate its ability to inhibit plasminogen activation. Furthermore, our studies are the first to investigate the mechanism of apo(a) inhibition of plasminogen activation in the presence of native, intact fibrin and soluble FDPs derived from plasmin digestion of fibrin. We have developed an equilibrium template kinetic model that, unlike models based on Michaelis-Menten kinetics, holds at all concentrations of substrate, cofactor, and inhibitor concentrations tested. Our results imply a common mechanism for apo(a) inhibition of plasminogen activation in the presence of native fibrin or degraded fibrin (FDPs) cofactors. In addition, our findings demonstrate that the ability of apo(a) to inhibit plasminogen activation is a complex function of the contribution of numerous apo(a) structural domains. Indeed, our kinetic analysis implies that interactions between apo(a) and potentially each of plasminogen, fibrin/FDPs, and tPA determine the functional outcome.