ABSTRACT

Antifibrinolytic drugs provide important pharmacological interventions to reduce morbidity and mortality from excessive bleeding during surgery and after trauma. Current drugs used for inhibiting the dissolution of fibrin, the main structural component of blood clots, are associated with adverse events due to lack of potency, high doses and non-selective inhibition mechanisms. These deficiencies warrant the development of a new generation highly potent and selective fibrinolysis inhibitors. Here we use the 14-amino acid backbone-cyclic sunflower trypsin inhibitor-1 scaffold to design a highly potent (Ki=0.05 nM) inhibitor of the primary serine protease in fibrinolysis, plasmin. This compound displays a million-fold selectivity over other serine proteases in blood, inhibits fibrinolysis in plasma more effectively than the gold-standard therapeutic inhibitor aprotinin and is a promising
candidate for development of highly specific fibrinolysis inhibitors with reduced side effects.

Keywords: Antifibrinolytics; Fibrinolysis; Inhibitors; Peptides; Plasmin

INTRODUCTION

The physiological process of fibrinolysis regulates the dissolution of blood clots and thrombosis. At the core of the fibrinolytic system (also known as the ‘plasminogen-plasmin’ system), is the serine protease plasmin, that degrades fibrin, the principal structural protein of blood clots. Plasmin is produced as the zymogen plasminogen which binds to the surface of fibrin via lysine binding sites, before activation primarily by tissue plasminogen activator (tPA). Activation of the plasminogen-plasmin system during surgery or traumatic injuries results in excessive bleeding, the need for blood transfusions, and the use of antifibrinolytic drugs. The most commonly used antifibrinolytic drug is the lysine analogue tranexamic acid (TXA), which prevents binding of plasminogen to fibrin and thereby its activation by tPA, but it does not inhibit plasmin once active. In traumatic injuries, the use of TXA is reported to give a small but significant reduction in mortality, although with no reduction in the need for blood transfusions, therefore limiting its clinical applications.1

In bypass surgery TXA is given pre-operatively and shows good efficacy in reducing the need for blood transfusions, but has been associated with risks of seizures.2,3 Aprotinin is a reversible (Laskowski mechanism) plasmin active site inhibitor used for decades with good efficacy in reducing blood loss and blood transfusions,4 highlighting the advantage of using plasminactive site inhibitors as antifibrinolytics. However, the use of aprotinin is hampered by its lack of specificity, since it inhibits virtually every S1 family serine protease present in blood.5 A large clinical trial found no survival benefits of using aprotinin in surgery (despite its ability to reduce the need for blood products)4 and it has been withdrawn from general use. Consequently, there is a pressing need to develop potent and specific plasmin inhibitors to reduce bleeding after trauma and during surgery more specifically and effectively.

A number of engineered plasmin inhibitors have been reported over the last two decades, but most of the more potent inhibitors suffer from poor selectivity.6,7 However, after aprotinin was withdrawn from clinical use there have been intensified efforts to design plasmin inhibitors, and a number of inhibitors with higher potency and selectivity,8,9 as well as allosteric inhibitors10 have emerged in recent years. One recent strategy that has emerged for designing potent and selective plasmin inhibitor is by producing substrate-analogues by cyclisation between the P2 and P3 residues.8 For example, a series of peptidomimetic inhibitors cyclized between the P2 and P3 residue side chains has been reported, with the most promising variant having a Ki of 0.2 nM for plasmin, but with low micromolar inhibition of pKLK, FXIa and uPA.9 This lead compound was further optimized by modifying the P1 residue and the N-terminal group to produce a new series of plasmin inhibitors, some of which are more potent for plasmin. The most promising lead compound shows potent plasmin inhibition (Ki=0.56 nM) and greatly increased selectivity over several other blood coagulation proteases.11 Although this series of inhibitors is highly promising, further selectivity optimization and evaluation is needed for therapeutic development.

In an alternative approach we have been using the sunflower trypsin inhibitor-1 (SFTI-1) scaffold as a template for design of plasmin inhibitors. SFTI-1 is a 14-amino acid backbone cyclic peptide (Figure 1A) that inhibits serine proteases by the Laskowski mechanism, trapping the target protease in a futile cycle of cleavage and re-ligation of the scissile peptide bond.12,13 SFTI-1 inhibits several S1 family serine proteases such as trypsin14 (Figure 1B) and plasmin.15 Its cyclic peptide backbone and bisecting disulfide bond makes SFTI-1 highly stable and thus an attractive scaffold for design of therapeutic compounds targeting serine proteases, as well as cell surface receptors and protein-protein interactions.16 SFTI-1 has been used to engineer potent and/or selective inhibitors of a number of serine proteases, including thrombin,15 chymotrypsin,13 matriptase-1/-2,17,18 cathepsin G19 and several
kallikrein-related peptidases,15,20,21 highlighting the versatility of the scaffold.

In this study we use substrate-guided and structure-based design methods to engineer a series of highly potent and selective inhibitors of plasmin based on the SFTI-1 scaffold. The most promising plasmin inhibitor (Ki=0.05 nM) has a million-fold selectivity over otherserine proteases in blood and blocks fibrinolysis in human plasma with higher efficacy than aprotinin. This inhibitor is a promising lead for the development of a new generation of antifibrinolytics with higher selectivity than those currently used in the clinic.

RESULTS

We previously reported that plasmin has a comparable, but sequence dependent (P2-P4), cleavage preference for Genetic alteration P1 Lys or Arg in peptide substrates.22 In the current study we found that substituting the P1 residue Lys5 for Arg in SFTI-1 resulted in a compound, (1), with 13-fold reduced potency, indicating that Lysis preferred in the context of the SFTI-1 scaffold (Table 1). We have also shown that plasmin has a substrate preference for aromatic residues at the P2 position (SFTI-1 residue 4), and particularly Tyr in combination with P1 Lys.22 Substituting Thr4 with Tyr in SFTI-1 produced an inhibitor (2) with more than a 60-fold increased potency for plasmin (Ki=0.140 nM) and over 6000-fold reduction in inhibition of trypsin. Compound 2 inhibited the neutrophil serine protease cathepsin G, coagulation factor XIa (FXIa) plasma kallikrein (pKLK), thrombin and matriptase in the micromolar range, while showing no inhibition of coagulation factors FIXa, FXa, FXIIa or the urokinase-/tissue-plasminogen activators (uPA/tPA) at 50 μM.

At a concentration of ~1.6 µM23, plasminogen is one of the most abundant serine protease zymogens in blood and a highly potent and specific inhibitor is required for desirable clinical outcomes. We previously used a SFTI-based inhibitor library to show that plasmin has a P2ʹ preference for Lys (SFTI residue 7).15 Substituting Ile7 for Lys in compound 2 produced a more potent inhibitor of plasmin (compound 3; Ki=0.051 nM) with no detectable inhibition of thrombin, FIXa, FXa, FXIa, FXIIa, tPA, uPA, pKLK or matriptase, and with improved selectivity over trypsin and cathepsin G. The structure and binding kinetics for compound 3 are shown in Figure 2 and Figure S1A,respectively.

To gain a further understanding of the molecular mechanisms underpinning the high potency of compounds 2 and 3 for plasmin we produced crystal structures of these inhibitors in complex with the catalytic domain of plasmin (µ-plasmin). Our results revealed that the structures of µ-plasmin/SFTI-variant complexes adopt the typical Bowman-Birk inhibitor and type I serine protease assembly (Figures 2 and 3A and 3B), similar to the first crystal structure of µ-plasmin domain in complex with a peptide-chloromethylketone (Cα RMSD<0.5 Å).24The µ-
plasmin shows a typical trypsin-like serine protease fold consisting of two subdomains (N-terminal and C-terminal domains) which are connected by loops. Each domain is made of a 6-stranded β-barrel. The catalytic domain of µ-plasmin is well ordered except for residues 14 (plasmin numbering in brackets, 560) and 15 (561),which results from cleavage between residues 15 (561) and 16 (562) during the activation of µ-plasminogen to µ-plasmin by tPA. Upon cleavage, Val16 (562) moves more than 12 Å away to the activation pocket and the -amino group of Val16 (562) forms ionic bond with the side chain of Asp195 (740). This leads to a major shift of the loops at the C-terminal domain and more importantly the formation of functional catalytic triad at the interface of the two subdomains. In our structures, the catalytic triad [His57 (603), Asp102 (646) and Ser195 (741)] adopts the active conformations, similar to the structures of trypsin and matriptase complexed with SFTI-1 (PDBID 1SFI and 3P8F, respectively) and that of the µ-plasmin structures in the PDB (PDB ID 5UGD and 5UGG; Figure S2 and S3). (Cα RMSD<0.5 Å)

The µ-plasmin/compound 2 complex diffracted to 1.43 Å (space group C121) with one binary complex in an asymmetric unit (Table S1). The µ-plasmin/compound 3 was crystallized with two protease:inhibtor complexes per asymmetric unit at 1.8 Å (space group P212121) which aligned with an RMSD of 0.121 Å over 1525 atoms. Here we focus on the analysis of the structure of μ-plasmin/compound 3 for monomer A and its cognate inhibitor. The total buried surface area at the interface is 1239 Å2, with the main-chain of 3 forming backbone hydrogen bond interactions with µ-plasmin via residues Cys3, Lys5, Ser6, Lys7 and Asp14, whereas the Lys5 NZ (P1 position) forms hydrogen bonds with Asp189 (735) OD1 and Ser190 (736) OG deep inside the S1 pocket. Furthermore, in the µ-plasmin/compound 3 structure Tyr4 of the SFTI scaffold plays key roles in both an intramolecular interaction within 3 and intermolecular interactions with the catalytic site of µ-plasmin (Figure 3 and 3B; Table S2). Specifically, Tyr4 OH forms a hydrogen bond with Arg2 NH2 in 3 which further constrains the compact structure of the inhibitor. Consistent with this, substituting Tyr4 in compound 3 with Phe (4) or Trp (5) results in over 10-fold reductions in potency. These findings are in agreement with our previous studies showing that intermolecular hydrogen bonds within the SFTI scaffold promotes high potency of inhibition.13,25 Furthermore, the side chain of Tyr4 forms a π-stacking interaction with Trp215 (761) and an aromatic dipole interaction with the negatively charged S2 pocket containing His57 (603), Asp102 (646) and Ser195 (760).

Another intermolecular key interaction between µ-plasmin and 3 is the positively charged side chain of Lys7 (P2′), which is positioned on top of the negatively charged S2′ pocket formed by Glu73 (623) and Glu143 (687) (without forming any hydrogen bonds) Comparing the crystal structures of 3 (Lys7) with 2 (Ile7) (Figures 2, 3A and 3B) we found that 2 forms backbone hydrogen bonds as well as hydrophobic interactions with Phe41 (587) of plasmin, similar to that of 3, except without forming any electrostatic interaction with the S2′ pocket. These findings confirm that Lys7 stabilizes the plasmin-inhibitor complex. Accordingly, we generated an Arg7 variant (6). The inhibitory potency of compound 6 for plasmin was comparable to that of 3 and without detectable inhibition of thrombin, FIXa, FXa,FXIa,FXIIa, tPA, uPA, pKLK or matriptase, but with reduced selectivity over trypsin and cathepsin

In a previous plasmin screen against a P2′ library of SFTI-1 (GTCTRSXPPCNPN, X=variable residue), we found that Lys7 was preferred over Arg7, most likely because this SFTI-based library had Arg5 rather than Lys5 as the P1 residue (considerable cooperativity appears to occur between the two subsites).15 To confirm this, we synthesized Arg5 variants of compound 3 (7) and compound 6 (8) and found that the combination of Arg5/Arg7 (8) was less preferred than Arg5/Lys7 (7) with a 2-fold reduction in potency. Further, substituting Arg7 in compound 6 with homo-Arg7 produced an inhibitor (9) with 10-fold lower potency for plasmin. Combined, these findings indicate that only the combinations of Lys5/Lys7 (3) or Lys5/Arg7 (6) ideally place the basic side chains for maximum potency of inhibition.

Despite the contribution of the basic P2′ residue,it appears that the intermolecular and intramolecular interactions of Tyr4 are essential for the potency. To confirm this, we substituted Tyr4 with the wild-type residue Thr in one of the lead compounds (6) to produce 10, and this single substitution resulted in nearly a 50-fold reduction in potency. To understand this phenomenon, we crystalized the µ-plasmin/compound 10 complex which diffracted to 1.32 Å, and as for the µ-plasmin/compound 2 complex, belong to space group C121 with one binary complex in an asymmetric unit (TableS1). Comparison of the crystal structures of 10 (Thr4) and the lead inhibitor 3 (Tyr4) in complex with µ-plasmin (Figures 2, 3B and 3C) provides an insight into this preference. Whereas the mainchain of 10 Thr4 O (P2) mediates an additional intermolecular interaction with Gln738 NE2 (Figure 4C, Table S2) and the binding surface area remains similar (1206 Å2),the aromatic π-stacking interactions between the inhibitor and plasmin were lost as were the Tyr4-Arg2 stabilizing intramolecular hydrogen bond.

The µ-plasmin/compound 3 complex revealed other specific interactions where Asp14 OD2 forms a salt-bridge with Arg175 (719) NE while stabilizing the inhibitor by a hydrogen bond between Asp14 OD1 and the Gly1 backbone N. We have shown previously by molecular dynamics, NMR, crystallography (PDB 4K1E and 4KEL) and binding assays that Asp14 in SFTI can be substituted with Asn to stabilize the peptide backbone (through hydrogen bonds) and improve the inhibition constants for various proteases.13,21,25,26 To evaluate the contribution of the SFTI Asp14, we substituted it with Asn in the most potent inhibitor (6) and the resulting compound 11 lost 29-fold binding affinity for plasmin. To examine this, we crystalized the µ-plasmin/compound 11 complex, which diffracted to 2 Å and belongs to the P63 space group with one binary complex in an asymmetric unit (Table S1). The µ-plasmin/compound 11 complex revealed loss of the intramolecular Tyr4 OH Arg2 NH2 hydrogen bond as well as the salt bridge between Asp14 OD2 and Arg719 NE (Figure 4D), indicating these interactions as key for the potency of compounds 3 and 6.

The crystal structure of the µ-plasmin/compound 3 complex (Figures 2 and 3B) showed that Glu60 (606) was located at the S5’ pocket of plasmin and that the P5’ Ile10 side chain was perfectly aligned and spaced to allow for substitutions of Ile10 with residues to interact with Glu60 (606) (Figure S4). Ile10 in compound 3 was substituted with Lys (12) or Arg (13) with the aim of introducing an intermolecular salt bridge with Glu606, but these compounds lost 60-fold and 116-fold in potency, respectively. Similarly, Ile10 in compound 3 was substituted with Gln (14) or Asn (15) with the aim of introducing an intermolecular hydrogen bonds with Glu 60 (606). These substitutions resulted in a loss of potency of 44-fold (Gln10) and 16-fold (Asn10), although less than when substituting Ile10 with basic residues.

Plasmin has a preference for Arg at the P4 position (SFTI residue 2) in peptide substrates22 and in the µplasmin/compound 3 complex Arg2 forms a perfectly aligned T-stacked cation-π interaction with the pyrrole ring of the Trp215 (761) side chain of plasmin. Substituting Arg2 for homo-Arg in one of the lead inhibitors (6) resulted in compound 16 with a 5-fold reduction in inhibition, indicating the importance of this interaction. Since the Arg2 side chain is known to interact with Asp14 to stabilize the backbone in wild-type SFTI-1,13,25 we explored the effects of combining the homo-Arg2 substitution with a Asp14 to Glu substitution; however, the resulting inhibitor (17) displayed a 27-fold loss of potency compared with compound 6, again highlighting the importance of the salt bridge between Asp14 OD2 with Arg175 (719) NE of plasmin. The µ-plasmin/compound 3 structure revealed that near Arg175 (719), another Arg residue [(Arg221 (767)] is located at the S5 binding site near Gly1 in 3. Gly1 in compound 6 was substituted with Asp with the aim to promote the formation of another intramolecular salt bridge, but the resulting inhibitor 18 displayed a 390-fold loss of activity. While we have previously shown that a Gly1 to Ala substitution in SFTI-1 may be tolerated by trypsin (3-fold loss in activity),27 if other functional substitutions could affect the canonical binding mode and thus the mechanism of Laskowski inhibitors is yet to be systematically explored. These mutational studies based on µ-plasmin/inhibitor structures combined with previous substrate and inhibitor library screening studies15,22 suggested that further substitutions in compound 3 were unlikely to significantly improve inhibition of plasmin.

Compound 3 has a 300-fold selectivity over all serine proteases examined and over a million-fold selectivity over family S1 serine proteases from blood and was selected for further evaluation of fibrinolysis inhibition in an ex vivo fibrinolysis assay. Pooled human plasma was diluted with buffer 1:4 to allow for measurement of the fibrin content by its light scattering properties28 and the level of inhibition was compared to the gold-standard plasmin inhibitor aprotinin. Under these conditions the maximum available plasminogen concentration is expected to be around 300 nM.23 Incomplete inhibition of fibrinolysis was achieved with 400 nM (70%) and 200 nM (25%) aprotinin (Figure 5 and Figure S5), respectively, consistent with aprotinin being a promiscuous inhibitor of most blood serine proteases, and thus also binding to other protease targets. Conversely, 400 nM of compound 3 completely attenuated fibrinolysis, while 200 nM achieved 90% inhibition, consistent with a 1:1 binding to activated plasmin in the presence of other blood serine proteases.

In this study we used the SFTI-1 scaffold to design plasmin inhibitors with high potency and exquisite selectivity over other serine proteases in the coagulation pathways. This was achieved by combining information from plasmin substrate and inhibitor library screens15,22 as well as structural studies using µ-plasmin/SFTI-variant complexes. The most selective inhibitor (3) inhibited fibrinolysis in plasma with high efficacy and provides a promising lead compound for development of an antifibrinolytic agent as well as a powerful tool to study the diverse roles of plasmin in normal physiology and disease.

Plasmin’s preference for aromatic residues in the S1 pocket is clearly the most defining feature for potent and selective inhibition of this serine protease. The side chain of Tyr4 forms a perfect π-stacking interaction with Trp215 (760) and a dipole interaction with the negatively charged S2 pocket, containing His57 (603), Asp102 (646) and Ser214 (760). This finding aligns with the strong preference for aromatic P2 residues by plasmin in combinatorial29 and non-combinatorial 22 peptide substrate libraries. Whereas the residues in the S2 binding pocket of plasmin are conserved among the serine proteases examined in this study (excluding Tyr215 in cathepsin G), the S2 pockets of the other serine proteases are too small to accommodate an aromatic residue because of steric hindrance from the 93-100 loop absent in plasmin (Figure S6). Indeed, the most potent and selective peptidomimetic plasmin inhibitors produced by others also have aromatic groups at the P2 position.9,11 and it is likely that capitalizing on this feature will be essential for design of highly potent and selective inhibitors in the future.

Another important defining feature of plasmin is its negatively charged S2′ pocket due to the flanking Glu73 (623) and Glu143 (687) residues and the alignment of serine proteases in this study shows that Glu73 and Glu143 are only present in plasmin (Figure S6). In another study the P2 ′ residue in Kunitz domain-1 of TFPI-2 was mutated to from Leu to Arg, resulting in a plasmin inhibitor (Ki=0.9 nM) with high selectivity over tPA, aPC, pKLK, and factors IIa, VIIa and XIa.30 However, substituting the P2 ′ residue with Lys in compound 2 resulted in a modest 2.8-fold increase in potency but increased the selectivity for all off-targets. This aligns with our previous findings using an inhibitor library based on SFTI-1 showing that whereas S1 family serine proteases generally prefer the wild-type P2 ′ residue (Ile7) in the SFTI scaffold, most proteases also have different additional preferences that may be used to modulate inhibitor specificity.15

Considering the general success of substituting the P2 ′ residue in the SFTI scaffold we thought that substituting the P5 ′ residue might further improve potency and/or specificity. The Glu60 (606) in plasmin appears perfectly positioned to interact with a substituted P5 ′ residue (Ile10) in (3), and the only protease examined with an acidic residue at this position was FXIa, which is not inhibited by compound 3. However, substitutionsof Ile10 with residues for potential of forming intermolecular salt bridges (Arg/Lys) or hydrogen bonds (Gln/Asn) resulted in major loss of potency. This may be due to steric hindrance, since the level of loss of potency correlated with the size of the substituted residue. Medication non-adherence Another possible explanation is that introducing a residue at position 10 in SFTI that forms intermolecular interactions with plasmin has a negative effect on the reversible (Laskowski) mechanism of inhibition, trapping the protease in a futile cycle of cleavage and re-ligation of the peptide bond.12,31 We have previously shown that intramolecular hydrogen bonds in the SFTI-scaffold are important for positioning the cleaved N-terminus for resynthesis of the peptide bond.13,25,26 Thus, it is possible that favorable interactions between the P5′ site of SFTI-scaffold and the S5′ site of the protease mis-aligns the cleaved N-terminus, reducing the efficiency of peptide bond re-ligation.

Compound 3 is the most potent and selective plasmin inhibitor engineered to date and provides a promising leadcompound for development of the next generation fibrinolysis inhibitors with higher efficacy and selectivity, and thereby potentially reduced side effects. While several challenges remain to be overcome in turning a promising lead into a drug, the nature of 3 as a disulfide-stabilized head-to-tail cyclic peptide places it in one of the most favorable classes of peptides for drug development.32,33 Its intrinsic advantages include high chemical stability, resistance to protease digestion and, given its small size, ease of synthesis. The latter Voxtalisib feature bodes well for the important commercial consideration of a low cost-of-goods. The plasminogen-plasmin system is involved in many physiological processes in addition to homeostasis, including inflammation,34 neurobiology,35 clearance of misfolded proteins,36 cell migration/tissue remodeling37 and wound healing.38 These processes involve activation of plasminogen via interaction with numerous plasminogen receptors present on the surfaces of most cell types and have been implicated in a number of diseases related to dysregulated inflammation, autoimmunity and cancer. 39-42 Therefore, highly potent and selective plasmin inhibitors also provide invaluable tools to further define the roles of the plasminogen-plasmin system in cell-based and in vivo models of physiology and disease in the future.

EXPERIMENTAL SECTION

Peptide synthesis, purification and validation

SFTI-1 and its variants were synthesized using standard 9-fluorenylmethoxycarbonyl (Fmoc) solid phase peptide synthesis on 2-chlorotrityl chloride resin (0.8 mmol eq. per gram) before cyclization and disulfide bond formation as previously described.19 Briefly, coupling reactions were performed using Fmoc N-protected amino acids (4 eq.) activated with 4 eq. O-(6-chlorobenzotriazole-1-yl)-1,1,3,3-tetramethylaminium hexafluorophosphate (HCTU) and 8 eq. N,N-diisopropylethylamine DIPEA in N,N-Dimethylformamide. Cyclization was performed in solution as above, with 4eq. 1-[Bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxid hexafluorophosphate (HATU) as the activator over 3 hours. Disulfide bonds were formed by oxidation in 10% ammonium bicarbonate (pH 8.3) with 10 µM oxidized glutathione. Colorimetric peptide-pNA (para-nitroanilide) substrates were synthesized on para-phenylenediamine derivatized 2-chlorotrityl chloride resin followed by solution oxidation of the free para-amine to a para-nitro group by a modified version of the method by Abbenante et al.43 as previously described.19 Peptides were purified using reversed-phase HPLC (Shimadzu Prominence) using a 5 µm ZORBAX Extend-C18 PrepHT column (21.2 × 250 mm) and a linear gradient of 10% acetonitrile/0.05% TFA to 90% acetonitrile/0.05% TFA. SFTI variants were purified both before and after formation of the disulfide bond. Peptide purity (> 95%) was confirmed by reversed-phase UPLC using a 5 µm Agilent 300 SB C18 column (2.1 × 50 mm) at 40°C with mobile phases as above (Figure S7). Peptide masses (Table S3) were determined by electrospray ionization mass spectroscopy (Shimadzu Prominence).

Enzyme assays

Native activated and purified proteins were obtained from Sigma-Aldrich (human plasmin, bovine cationic trypsin) and Molecular Innovations (human thrombin, pKLK, beta-FIXa, FXa, beta-FXIIa, two chain-tPA, HMW uPA and cathepsin G). The concentration of plasmin was determined by active site titration using bovine aprotinin (SigmaAldrich). Inhibition constants for the cyclic peptide inhibitors were determined in three independent triplicate experiments with 100 µM substrate in assay buffer (100 mMTris pH 8.0, 100 mM NaCl and 0.005% (v/v) Triton™ X-100), following equilibration of proteases and inhibitors for 30 minutes in a 96-well plate (low binding microplate wells, Corning). Protease concentrations, buffer additives and substrates are given in Table S4. Fluorescent peptide-MCA (7-methoxycoumarin-4-yl acetyl) substrates were obtained from Peptide Institute Inc. Substrate hydrolysis was monitored by following absorbance at 405 nm for 7 minutes (peptide-pNA substrates) (excluding FXIa which was monitored for 60 minutes) or by fluorescence at λex=360 nm/λem=460 nm for 10 minutes (peptide-MCA substrates) using a Tecan M1000 Pro microplate reader. Inhibition constants were calculated with the Morrison equation for tight binding inhibitors using the substrate KM values given in Table S4 by non-linear regression using Prism 7 (GraphPad). The results are reported as the mean ± SEM from three independent experiments performed in triplicates. The dissociation constant (koff) for 3 was calculated with the method derived by Baici and Gyger-Marazzi44 based on the lag phase between the steady state of hydrolysis rates preincubated protease/inhibitor versus simultaneous addition of substrate and inhibitor. A graphical representation of the method and the equation are given in Figure S1B.

Protein Expression and Purification

The catalytic domain of human μ-plasminogen (residues 542-791) and its active-site mutant Ser195(741)Ala were expressed and purified from Pichiapastoris as previously described,45 Both the wild-type and active-site mutant μ-plasminogen was activated into μ-plasmin with tPA in the presence of SFTI inhibitor (1:1.2 molar ratio) to generate μ-plasmin/SFTI complexes which was further purified by size exclusion chromatography.

Purified complexes at 10 mg/ml were used for crystallisation trials by hanging drop vapour phase diffusion method at 20℃, and crystals were obtained in the following conditions. Complex crystals of μ-plasmin with compound 3 were formed with 100 mM sodium citrate pH 4-6, 100 mM MgCl2, 15-21% PEG-4000 as the mother liquor, and the condition for μ-plasmin/compound 2 and μ-plasmin/ compound 10 was 100 mM MES, 150 mM ammonium sulphate, and 13-17% PEG-4000. Lastly, the μ-plasmin (S741A)/compound 11 complex was crystalized in 100mM sodium acetate pH 4.5, 1 M sodium formate.

Crystals were flash cooled in liquid nitrogen in the presence of 15% glycerol. Datasets were collected at Australia Synchrotron MX beamlines, and processed using XDS.46 The crystal structures of complexes were solved by molecular replacement (using 5UGG as staring search model for μ-plasmin) and the program PHASER from CCP4.47 After refinement using the REFMAC program,48 SFTI-1 (PDBID 1SFI) was fitted into the Fo-Fc electron density using COOT,49 and inhibitor sequences were correspondingly modified to SFTI variants. Model refinement
and building was carried out using PHENIX50 and COOT.

Fibrinolysis inhibition assays

Human pooled citrated plasma (GeneTex) was diluted with buffer (Tris-HCL, pH 7.4, 150 mM NaCl, 30 mM CaCl2) and with various concentrations of aprotinin or compound 3. The assay was initiated by addition of tPA (1.5 nM) and thrombin (2.5 nM) in buffer (final plasma concentration 1:5). Fibrin formation and fibrinolysis was monitored by light scattering at λ 405 nM using a Tecan M1000 Pro microplate reader over one hour as previously described.51 The results are reported as the mean from three independent experiments performed using 12 technical
replicates.