Contribution of Conserved Lysine Residues in the α2-Antiplasmin C Terminus to Plasmin Binding and Inhibition*

α2-Antiplasmin is the physiological inhibitor of plasmin and is unique in the serpin family due to N- and C-terminal extensions beyond its core domain. The C-terminal extension comprises 55 amino acids from Asn-410 to Lys-464, and the lysine residues (Lys-418, Lys-427, Lys-434, Lys-441, Lys-448, and Lys-464) within this region are important in mediating the initial interaction with kringle domains of plasmin. To understand the role of lysine residues within the C terminus of α2-antiplasmin, we systematically and sequentially mutated the C-terminal lysines, studied the effects on the rate of plasmin inhibition, and measured the binding affinity for plasmin via surface plasmon resonance. We determined that the C-terminal lysine (Lys-464) is individually most important in initiating binding to plasmin. Using two independent methods, we also showed that the conserved internal lysine residues play a major role mediating binding of the C terminus of α2-antiplasmin to kringle domains of plasmin and in accelerating the rate of interaction between α2-antiplasmin and plasmin. When the C terminus of α2-antiplasmin was removed, the binding affinity for active site-blocked plasmin remained high, suggesting additional exosite interactions between the serpin core and plasmin.

␣ 2 -Antiplasmin is the physiological inhibitor of plasmin and is unique in the serpin family due to N-and C-terminal extensions beyond its core domain. The C-terminal extension comprises 55 amino acids from Asn-410 to Lys-464, and the lysine residues (Lys-418, Lys-427, Lys-434, Lys-441, Lys-448, and Lys-464) within this region are important in mediating the initial interaction with kringle domains of plasmin. To understand the role of lysine residues within the C terminus of ␣ 2 -antiplasmin, we systematically and sequentially mutated the C-terminal lysines, studied the effects on the rate of plasmin inhibition, and measured the binding affinity for plasmin via surface plasmon resonance. We determined that the C-terminal lysine (Lys-464) is individually most important in initiating binding to plasmin. Using two independent methods, we also showed that the conserved internal lysine residues play a major role mediating binding of the C terminus of ␣ 2 -antiplasmin to kringle domains of plasmin and in accelerating the rate of interaction between ␣ 2 -antiplasmin and plasmin. When the C terminus of ␣ 2 -antiplasmin was removed, the binding affinity for active siteblocked plasmin remained high, suggesting additional exosite interactions between the serpin core and plasmin.
When tissue injury occurs, fibrinolysis and coagulation are activated in concert. Tissue plasminogen activator secreted by the injured endothelium activates plasminogen to plasmin, which in turn degrades the fibrin clot. Fibrinolysis remains localized because tissue plasminogen activator and plasminogen co-localize on fibrin strands, dramatically improving catalytic efficiency (1). In addition, free plasmin in plasma is rapidly inactivated by its principal regulator, ␣ 2 -antiplasmin. Dysregulation of either fibrinolysis or coagulation has the potential to cause thrombosis, whereas on the other hand, therapeutic manipulation of these pathways can be used to treat thrombotic diseases (2). Currently, most therapeutic agents used to promote fibrinolysis directly activate plasminogen, but their use is limited by the risk of excessive bleeding. However, there is increasing evidence that ␣ 2 -antiplasmin is a useful alternative target in the development of new therapeutics for thrombotic diseases (3). ␣ 2 -Antiplasmin is unique among the serpin family due to the presence of N-and C-terminal extensions beyond its inhibitory core. The 42-amino acid N terminus is cross-linked to fibrin by factor XIIIa, localizing its activity to the clot (4,5). The ␣ 2 -antiplasmin C terminus extends 55 amino acids from Asn-410 to Lys-464 and is believed to be important in the initial interaction with kringle domains of plasmin (6). A form of ␣ 2 -antiplasmin that lacks the C terminus interacts with plasmin much more slowly compared with the intact molecule (7). The crystal structure of murine ␣ 2 -antiplasmin lacking the first 43 amino acids has been determined; however, the C terminus could not be modeled, which suggests that the C-terminal region of ␣ 2 -antiplasmin is highly flexible (8).
The C terminus of ␣ 2 -antiplasmin contains six lysines, five of which are conserved between species (Lys-427, Lys-434, Lys-441, Lys-448, and Lys-464), and it is known that lysines are key in mediating binding to kringle domains. Using a synthetic peptide corresponding to the final 26 amino acids of the C terminus, a previous study showed that the ␣ 2 -antiplasmin/plasmin association was reduced, indicating that this region has high affinity for plasmin kringle domains (9). However, the relative importance assigned to the C-terminal lysine compared with other internal lysines differs in published studies. Frank et al. (6) and Gerber et al. (10) studied the binding of kringle domains to C-terminal peptide containing Lys-to-Ala mutations. They showed that the C-terminal lysine (Lys-464) was the most important contributor to binding affinity for plasmin kringle domains (6,10). However, kinetic inhibitory studies by Wang et al. (11) indicated that Lys-448 in full-length ␣ 2 -antiplasmin was most important in the interaction with plasmin.
At present, there are no comprehensive studies investigating the role of lysine residues using inhibitory rates as an end point. To fully understand the role of the C-terminal lysine residues in the interaction with kringle domains of plasmin, we present a systematic and sequential study on the mutagenesis of ␣ 2 -antiplasmin C-terminal lysines (Lys-427, Lys-434, Lys-441, Lys-448, and Lys-464) and truncation of the C terminus. This is the first report describing the plasmin inhibition rate (k a ) for fulllength human ␣ 2 -antiplasmin and intact plasmin. Furthermore, we investigated the binding affinity (K D ) of wild-type ␣ 2 -antiplasmin and various mutants for active site-blocked plasmin using plasmon surface resonance. These studies suggest additional exosite interactions between the serpin core domain and plasmin. ␣ 2 -antiplasmin cDNA was isolated from a liver cDNA library  using PCR with primers 5Ј-GGA TCC ACC CCA GGA GCA  GGT GTC CC-3Ј and 5Ј-GGA TCC TCA CTT GGG GCT  GCC AAA C-3Ј. The product was cloned into the pET-His(3a) expression vector and sequenced for authenticity (12). The QuikChange site-directed mutagenesis kit (Stratagene) was used on human WT ␣ 2 -antiplasmin template in which alanine or stop codons were introduced at various lysine residues along the C terminus. Several mutations within the C terminus of human  Fig. 1). 3 All constructs were nucleotide-sequenced to confirm the mutations introduced.

Construction of ␣ 2 -Antiplasmin Variants-Human WT
Expression and Purification of ␣ 2 -Antiplasmin Variants-Recombinant human ␣ 2 -antiplasmin (WT and mutants) was expressed in Escherichia coli BL21(DE3)pLysS cells. Cells were grown overnight at 37°C with shaking at 220 rpm in 2ϫ tryptone-yeast culture medium supplemented with 50 g/ml ampicillin. The cell cultures were then diluted 1:10 in fresh 2ϫ tryptone-yeast medium supplemented with 50 g/ml ampicillin and grown for a further 2 h at 37°C and 220 rpm. Human ␣ 2 -antiplasmin was then induced with a final concentration of 0.02 mM isopropyl ␤-D-thiogalactopyranoside and incubated for 4 h at 30°C and 220 rpm. Cells were harvested by centrifugation at 3500 rpm (Beckman JS-4.2 rotor) for 20 min and stored at Ϫ80°C until used.
To obtain soluble recombinant ␣ 2 -antiplasmin, the cell pellet was lysed in lysis buffer (50 mM NaPO 4 (pH 8.0), 500 mM NaCl, 20 mM imidazole, and 5 mM ␤-mercaptoethanol) containing 1 mg/ml lysozyme, 0.2 mg/ml DNase, 1:1000 protease inhibitor mixture (Sigma), and 0.01% phenylmethanesulfonyl fluoride per liter of cell culture. The cells were frozen in liquid nitrogen and then completely thawed three times in a 37°C water bath. The lysate was centrifuged (Sorvall SS34 rotor) at 18,000 rpm for 20 min. To purify recombinant ␣ 2 -antiplasmin, the supernatants were loaded onto a HisTrap column (GE Healthcare) and eluted with a linear imidazole gradient (0.02-0.5 M) in lysis buffer. The proteins were further purified on a Mono Q column (GE Healthcare) with a linear NaCl gradient (0 -0.5 M) in 20 mM Tris (pH 8.0) and 0.1 mM EDTA. All recombinant human ␣ 2 -antiplasmin was electrophoresed on 12.5% SDS-polyacrylamide gel to confirm the purity of peak fractions. Aliquots of recombinant protein were stored at Ϫ80°C until used.
The raw data were fitted using nonlinear regression in GraphPad Prism (Equation 1), where P is the concentration of product at time t, V 0 is the initial velocity, and k obs is the apparent first-order rate constant (14). The k obs value for each serpin concentration was calculated by Equation 1. Each k obs value was plotted against the respective ␣ 2 -antiplasmin concentration. Linear regression analysis was performed to obtain the uncorrected rate of inhibition (kЈ). Equation 2 was used to account for fluorogenic substrate competition, where the rate of inhibition was adjusted with the SI, substrate concentration ([S]), and Michaelis constant (K m ) of the substrate to give the second-order rate constant (k a ) (14).
Binding of ␣ 2 -Antiplasmin to Active Site-blocked Plasmin by Surface Plasmon Resonance-The interactions between active site-blocked plasmin and various recombinant ␣ 2 -antiplasmin proteins were determined via surface plasmon resonance using the Biacore T100 system (GE Healthcare). Active site-blocked plasmin was produced by incubating human plasmin with a 1000-fold molar excess of D-Val-Phe-Lys chloromethyl ketone (Calbiochem) at 37°C for 1 h. Active site-blocked plasmin was then dialyzed overnight at 4°C in running buffer (0.01 M HEPES (pH 7.4), 0.15 M NaCl, 50 M EDTA, and 0.05% surfactant P20 (GE Healthcare)). To confirm that the active site was completely blocked, the activity of active site-blocked plasmin was checked against active plasmin in the presence of 200 M H-Ala-Phe-Lys-AMC fluorogenic substrate.
To measure the binding affinity and rate of plasmin association of ␣ 2 -antiplasmin with active site-blocked plasmin, we immobilized hexahistidine-tagged WT or mutant ␣ 2 -antiplasmin on the nitrilotriacetic acid (NTA) surface of a NTA chip (GE Healthcare) following the manufacturer's instructions. All experiments were carried out in running buffer at a flow rate of 10 l/min and performed in triplicates. Briefly, the NTA sur-face was activated by injecting 500 M NiCl 2 for 1 min. 20 nM WT or mutant recombinant ␣ 2 -antiplasmin was immobilized onto the surface on one flow cell for 1 min. A reference flow cell containing no ␣ 2 -antiplasmin was used to account for nonspecific binding to the NTA surface. Six different concentrations of active site-blocked plasmin (analyte) were injected for 1 min of association time, followed by 10 min of dissociation time. The range of active site-blocked plasmin concentrations (2-120 nM) was adjusted for each ␣ 2 -antiplasmin variant. At the end of each concentration cycle, the NTA surface was completely stripped with regeneration buffer (0.01 M HEPES (pH 7.4), 0.15 M NaCl, 0.35 M EDTA, and 0.05% surfactant P20) at a flow rate of 30 l/min for 2 min. NTA surface activation and ␣ 2 -antiplasmin immobilization were performed at each active site-blocked plasmin concentration. The injection needle was cleaned with an extra wash of running buffer after each subsequent step.
Real-time binding curves were monitored on a sensorgram as resonance units over time. For kinetic and binding affinity analysis of recombinant ␣ 2 -antiplasmin with active site-blocked plasmin, we used the two-state reaction model provided in the Biacore T100 evaluation software (Version 1.1.1) (Equation 3).
This model describes a 1:1 binding of analyte to immobilized ligand, followed by a secondary interaction that stabilizes the two molecules. 2 analysis supported this model. The overall equilibrium dissociation constant (K D ) was calculated using Equation 4 (15).

Stoichiometry of Inhibition between Recombinant Human
␣ 2 -Antiplasmin and Plasmin-To examine the role of lysine residues in the ␣ 2 -antiplasmin C terminus, we employed sitedirected mutagenesis and produced a series of mutant recombinant proteins (Fig. 1). Mutations within the C-terminal region of ␣ 2 -antiplasmin would not be expected to affect the inhibitory mechanism because this property resides in the serpin core domain. However, to verify that the recombinant proteins were correctly folded, the SI for human plasmin was determined for WT and mutant ␣ 2 -antiplasmin.
The serpin/protease ratio that resulted in complete loss of protease activity was determined as the SI of the serpin (Fig. 2). The SI of WT ␣ 2 -antiplasmin was found to be 1.0 (Fig. 2, A and E) and corresponds with published values (16). Fig. 2E shows the SI values obtained for each recombinant protein produced. All mutants were shown to have an SI of between 0.9 and 1.5 (Fig. 2, B-E), indicating that the efficiency of mutant protein had not been structurally perturbed by the introduction of various amino acid substitutions in the C terminus. To confirm these observations, WT recombinant ␣ 2 -antiplasmin and mutants were also assessed by CD spectrometry. Mutant ␣ 2 -antiplasmin proteins produced CD spectra similar to WT ␣ 2 -antiplasmin, demonstrating that the mutant proteins retained their native fold (data not shown).
Rate of Plasmin Inhibition of Recombinant Human ␣ 2 -Antiplasmin Variants-The primary hypothesis being tested in this work is that lysine residues at the extreme C terminus of ␣ 2 -antiplasmin as well as internally within the C-terminal extension accelerate inhibition of plasmin. Therefore, kinetic studies using a protease inhibition (progress curve) assay were performed to measure the rate of plasmin inhibition by WT and mutant ␣ 2 -antiplasmin. Fig. 3 shows examples of progress and fitted curves obtained from the analysis of WT and P414stop (Cterm⌬) ␣ 2 -antiplasmin.
Because the reductions in the inhibition rate observed with individual Lys-to-Ala mutants were small compared with the effect of removal of the complete C terminus, we examined the effect of progressive mutations of the Lys residues in this domain. Compared with ␣ 2 -antiplasmin K464A, each additional mutation (K448A/K464A, K441A/K448A/K464A, and K434A/K441A/K448A/K464A) resulted in an ϳ2-fold reduction in the plasmin inhibition rate (Fig. 5). Therefore, the overall reduction in the rate of plasmin inhibition observed with the four-residue substitution (K434A/K441A/K448A/K464A) was 45-fold, which is comparable with the effect of removing the entire ␣ 2 -antiplasmin C terminus (P414stop). It is conceivable that substitution of four lysine residues within the C terminus of ␣ 2 -antiplasmin could produce a structural perturbation of the domain. We therefore measured the kinetic effect of adding the lysine analog ⑀-aminocaproic acid to the inhibitory reaction. When 1 mM ⑀-aminocaproic acid was added to the inhibitory reaction, a k a of (2.3 Ϯ 0.2) ϫ 10 6 M Ϫ1 s Ϫ1 was observed, which was comparable with the mutant with a deletion of the C terminus (P414stop) (k a ϭ (9.2 Ϯ 0.2) ϫ 10 5 M Ϫ1 s Ϫ1 ) (Fig. 4). This suggests that most of the reduction in the plasmin inhibition rate caused by the removal of the C terminus can be accounted for by lysine-specific interactions.
The relative importance of the C-terminal Lys-464 compared with the internal lysine residues was further addressed by measuring the rate of inhibition of a mutant in which Lys-464 was preserved while the internal lysines were mutated to alanine (K427A/K434A/K441A/K448A/K464WT) (Fig. 5). This mutant demonstrated a 5.2-fold rate reduction (k a ϭ (7.1 Ϯ 0.1) ϫ 10 6 M Ϫ1 s Ϫ1 ) compared with WT ␣ 2 -antiplasmin.
The importance of the C-terminal lysine was further evaluated by producing a mutant in which a stop codon was intro-duced after Lys-448 (L449stop). Despite this mutant possessing a C-terminal lysine, it still showed a rate reduction in plasmin inhibition (k a ϭ (1.1 Ϯ 0.1) ϫ 10 5 M Ϫ1 s Ϫ1 ) of 3.4-fold compared with WT ␣ 2 -antiplasmin. Additionally, when the L449stop mutant was modified by substituting Lys-448 with Ala (K448A/L449stop), there was a further decrease in the plasmin inhibition rate (k a ϭ (2.8 Ϯ 0.1) ϫ 10 6 M Ϫ1 s Ϫ1 ).
Binding Affinity of Recombinant Human ␣ 2 -Antiplasmin Variants for Active Site-blocked Plasmin-The association rate constants described above were derived from measurements of inhibition of plasmin amidolytic activity. To corroborate these results and to partition the ␣ 2 -antiplasmin/plasmin interaction between the serpin core versus the C-terminal extension, we used surface plasmon resonance to directly measure the binding affinity of recombinant human ␣ 2 -antiplasmin for human plasmin. WT ␣ 2 -antiplasmin and various mutants were reacted with active site-blocked plasmin, and binding was observed in real time. The association and dissociation constants were calculated to obtain the binding affinity (K D ). Examples of sensorgrams for the interaction between active site-blocked plasmin and recombinant ␣ 2 -antiplasmin are shown in Fig. 6. To achieve similar response units for WT ␣ 2 -antiplasmin (Fig. 6A), a higher concentration of active site-blocked plasmin was used with ␣ 2 -antiplasmin mutants, which accounts for the difference in the shape of the binding curves observed in the sensorgrams (Fig. 6).
The K D of WT ␣ 2 -antiplasmin for active site-blocked plasmin was determined to be 1.6 nM, indicating a high affinity interaction. Single Lys-to-Ala mutants at positions 448 and 464 showed decreases in K D of 1.3-and 3.3-fold, respectively (Fig.  7), compared with WT ␣ 2 -antiplasmin. Removing the C terminus (P414stop) resulted in a 31-fold reduction in K D (50 nM). Overall measurements of binding affinity were consistent with association rates observed using the progress curve plasmin inhibition assay.
Sequential mutation of the Lys residues within the C terminus resulted in a progressive decrease in the K D , which corresponds to the observations made previously. The rate of plasmin association (k a1 ) obtained using surface plasmon resonance was very similar to the rate of plasmin inhibition (k a ) obtained using the kinetic assay described previously. The dissociation rate constant (k d1 ) and the forward and reverse rate constants (k a2 and k d2 ) remained relatively unchanged for WT ␣ 2 -antiplasmin and mutants with active site-blocked plasmin.

DISCUSSION
In this work, we have reported the first comprehensive description of the kinetics of ␣ 2 -antiplasmin/plasmin interactions employing two different methods. By incorporating a fluorogenic substrate with high affinity for plasmin, we have been able to use the "progress curve" method to accurately measure the association rates for WT ␣ 2 -antiplasmin and mutants. In addition, we have independently observed the ␣ 2 -antiplasmin/plasmin interaction via surface plasmon resonance. In both methods, we employed full-length ␣ 2 -antiplasmin (WT and mutants) and intact plasmin containing the protease and all kringle domains. It is important to recognize that the two methods measure different kinetic rate constants. Using Fig. 8 as a reference schematic, surface plasmon resonance measures the initial rate of interaction (k 1 ) between ␣ 2 -antiplasmin and plasmin. The protease inhibition assay (progress curve) measures the overall rate at which the irreversible covalent ␣ 2 -antiplasmin⅐plasmin complex is formed, resulting in complete inhibition; therefore, this takes into account k 1 , k 2 , and k 4 (Fig. 8). We were able to obtain comparable rates of plasmin inhibition (k a ) and asso- . Nonlinear regression analysis using Equation 1 was used to determine the first-order rate constant (k obs ). B, the k obs is plotted against the WT ␣ 2 -antiplasmin concentration, and linear regression analysis was performed to determine the uncorrected secondorder rate constant (kЈ). To account for substrate inhibition, Equation 2 was applied to determine the corrected second-order rate constant (k a ). C, progress curves of plasmin inhibition using plasmin (0.5 nM) and P414stop (Cterm⌬) ␣ 2 -antiplasmin (80 -140 nM) in the presence of H-Ala-Phe-Lys-AMC (1 mM). the k obs was obtained as described above. D, the k obs is plotted against the Cterm⌬ ␣ 2 -antiplasmin concentration to determine the kЈ. The k a of Cterm⌬ ␣ 2 -antiplasmin was calculated using Equation 2.
ciation (k a1 ) despite the fact that both methods measure different values. This indicates that the rapid interaction is predominantly due to the formation of the initial reversible encounter ␣ 2 -antiplasmin⅐plasmin complex, thus suggesting that the rate-limiting step occurs when the covalent ␣ 2 -antiplasmin⅐plasmin complex is formed.
Using protease inhibition and binding affinity data, we were able to demonstrate a progressive decrease in the rate of plasmin association, inhibition, and binding affinity with consecutive Lys-to-Ala mutations within the C terminus of ␣ 2 -antiplasmin. We showed that all conserved Lys residues (Lys-427, Lys-434, Lys-441, Lys-448, and Lys-464) play a role in the interaction with kringle domains of plasmin, with Lys-464 being the main initiator, followed by Lys-448, which corresponds with previously published data (6,10). Individually, the internal Lys residues appear to have a minor function in the interaction with plasmin. However, as demonstrated by several of our ␣ 2 -antiplasmin mutants, primarily K434A/K441A/K448A/K464A and P414stop, we were able to show that when five of the lysines were mutated, the rates of plasmin inhibition and binding were reduced to those of the C-terminally truncated ␣ 2 -antiplasmin protein. This indicates that the Lys residues within the C terminus of ␣ 2 -antiplasmin are the primary mediators in the binding to the kringle domains and that removal of these residues will result in the loss of C-terminal binding. Furthermore, K427A/K434A/K441A/K448A/K464WT demonstrated that even with the presence of the most C-terminal lysine with all the internal lysines mutated, the plasmin inhibition rate obtained was not comparable with WT ␣ 2 -antiplasmin. Therefore, each conserved Lys in the ␣ 2 -antiplasmin C terminus participates in the binding and inhibition of plasmin.
Previous studies by Frank et al. (6) using individual recombinant plasmin kringle domains (K1, K3, K3mut, K4, and K5) showed that the isolated ␣ 2 -antiplasmin C terminus had the highest affinity for K1, followed by K4, K5, and K2. In further experiments, Gerber et al. (10) examined the affinity of the recombinant plasmin kringle domains (K1, K1-3, K4, and K4-5). They demonstrated that progressive mutations of lysine residues within the ␣ 2 -antiplasmin C terminus decreased the affinity for K1-3, although the greatest contribution to binding was attributable to Lys-464 and Lys-448. The apparent lack of effect on the affinity of mutations of Lys-418, Lys-427, Lys-434, and Lys-441 may be explained by the fact that only two lysine-binding kringle domains were present in the K1-3 protein. The seeming discrepancy between our results and those of Gerber et al. can therefore be  accounted for by differences in experimental approaches. They measured the association constants of isolated kringle domains (K1, K4, K1-3, and K4-5) with the C-terminal por-tion of ␣ 2 -antiplasmin. By contrast, this study describes the rate of plasmin inhibition and binding affinity of full-length ␣ 2 -antiplasmin with intact plasmin. One striking observation made in this study is that the association rate and binding affinity of plasmin with C-terminally truncated ␣ 2 -antiplasmin (P414stop) were relatively high (k a1 ϭ 5.1 ϫ 10 5 M Ϫ1 s Ϫ1 ; K D ϭ 49 nM). It is important to note that surface plasmon resonance studies were performed with active site-blocked plasmin, suggesting that the rapid association and high affinity observed in the absence of the C-terminal extension were mediated by exosite interactions between ␣ 2 -antiplasmin and the plasmin protease domain. These important interactions between the ␣ 2 -antiplasmin core domain, outside the immediate vicinity of P1-P1Ј, and the active site cleft of plasmin are also likely to contribute specificity to the serpin/protease interaction. Having additional exosite interactions is not uncommon in the serpin inhibition mechanism, as it may aid in the recognition of its target protein (18,19). Together with the specific binding of the ␣ 2 -antiplasmin C terminus to plasmin kringle domains, this leads to an exquisite specificity of the interaction minimizing off-target inhibition of non-cognate proteases.
In summary, we have performed detailed kinetic and binding studies of the interaction between ␣ 2 -antiplasmin and plasmin. Within the ␣ 2 -antiplasmin C terminus, we have measured the contribution of the conserved lysines to the interaction with the plasmin kringle domains. This study demonstrates that the C-terminal lysine (Lys-464) is the single most important amino acid in this domain. The remaining conserved lysine residues within the C terminus of ␣ 2 -antiplasmin individually contribute less, but together significantly enhance the rate of association of serpin with plasmin. These data support the zipper model of interaction whereby Lys-464 binds initially to plasmin (most likely at K1), followed by progressive binding of the other conserved lysines (Lys-448, Lys-441, and Lys-434) to the remaining lysine-binding kringle domains (K4, K5, and K2) (6). Our data also highlight the importance of exosite interactions between the ␣ 2 -antiplasmin core serpin and the plasmin  . Mechanism of serpin inhibition. I represents serpin, and E represents the protease. The forward rate constant of serpin with protease is represented as k 1 . The association rate constant (k a1 ) measured in surface plasmon resonance studies is equal to k 1 . k Ϫ1 is the reverse rate constant. The rate at which conformational change (EI*) occurs is denoted as k 2 . k 3 occurs during the substrate reaction, which results in cleaved serpin and release of active protease (E ϩ I*). Formation of the covalent protease⅐serpin complex (E-I*) results in complete inhibition, and the rate constant is represented as k 4 . protease domain, which provide an additional mechanism of specificity in the serpin/protease interaction.