Reagents and materials

The following materials were obtained from the indicated sources: human fXIa, fIXa, fXa, thrombin, plasma kallikrein, and activated protein C (aPC) (Hematologic Technologies; Essex Junction, VT); recombinant tissue plasminogen activator (tPA) Alteplase® (Genentech; South San Francisco, CA); human α-fXIIa (Enzyme Research Laboratories; South Bend, IL); recombinant factor VIIa (fVIIa) Novoseven® (Novo Nordisk; Bagsværd, Denmark); Lys-plasmin and chromogenic substrates Spectrozyme FVIIa, Spectrozyme FIXa, Spectrozyme PCa, Spectrozyme tPA (Sekisui Diagnostics; Stamford, CT); S-2302, S-2366, S-2765 (Chromogenix; Milano, Italy); and Pefachrome TH 5244 (Pentapharm; Basel, Switzerland). CTI provided by Gamma (Pushchino, Russia) was extracted from corn, purified with ammonium sulfate, acetone precipitation, trypsin-affinity and ion-exchange chromatography, as described previously [17]. Luffa cylindrica trypsin inhibitor (LCTI-III) was chemically synthesized (Creative Dynamics; Shirley, NY). All other reagents were from (Sigma-Aldrich; St. Louis, MO) unless otherwise noted.

Expression of the thioredoxin-fused infestin-4 and its mutants

An infestin-4 gene coding for 56 amino acids was synthesized by the PCR fusion of five primers. The oligonucleotides AS-KAZ-1F and AS-KAZ-2R with 16-nt-long complementary 3’-termini (S1 Table), as well as AS-KAZ-3F and AS-KAZ-4R (18-nt-long complementary ends), were annealed and extended, resulting in two DNA fragments (105 bp and 115 bp, respectively). Because AS-KAZ-2R and AS-KAZ-3F had 17-nt-long overlapping 5’-termini, the 105-bp- and 115-bp-long products were fused and extended. A final DNA product (214 bp) was generated by the extension of the 3’-terminus using the primer AS-KAZ-5R. This product was amplified, cloned into the p10E plasmid using NheI and HindIII restriction sites and subcloned using NcoI and HindIII (Fermentas; Vilnius, Lithuania) into the pET32a plasmid (Merck; Darmstadt, Germany), which encodes thioredoxin I (Trx) and the thrombin cleavage site. The pET32-Inf4 plasmid was transformed into Escherichia coli BL21 (DE3) cells via heat-shock, and selected clones were grown in TB medium with 0.1 mg/ml ampicillin and 0.1% glucose. Expression of Trx-Inf4 (GenBank KJ183178) was induced with 1 mM IPTG for 3 hours at 30°C. The cells were harvested and resuspended in 10 mM Tris-HCl pH 7.5, 0.1% Triton-X100, and 10 μg/ml lysozyme, and then, the cells were incubated on ice for 30 min and sonicated. The clarified lysate was supplemented with 40 mM Tris-HCl pH 7.5, 500 mM NaCl, and 10 mM imidazole and applied at 10 ml/min on a Chelating Sepharose Fast Flow column (GE Healthcare; Little Chalfont, UK). Weakly bound proteins were eluted with a solution containing 50 mM Tris, pH 7.5, 500 mM NaCl, and 100 mM imidazole, followed by a 50 mM Tris-HCl pH 7.5 solution. Trx-Inf4 was eluted with 50 mM Tris-HCl pH 7.5, 500 mM imidazole solution and applied to a Tricorn 10/100 column (GE Healthcare) with SOURCE 30Q strong anion-exchange resin in 20 mM Tris, pH 8.0. Monomeric Trx-Inf4 was eluted with a linear NaCl gradient from 0 to 500 mM at flow rate of 8 ml/min. Trx-Inf4 (150 mM NaCl) was collected and concentrated by ultrafiltration with Vivaspin 6 PES 5 kDa (Sartorius; Goettingen, Germany) to 10–14 mg/ml. The protein concentration was determined by UV-spectroscopy. The pure Trx-Inf4 yield was approximately 60 mg per liter of cell culture.

To substitute the reactive site residues Phe9, Asn11, Tyr12, and Pro14 in wt-Inf4 and generate the Trx-fused Mutant 15 (GenBank KJ183179), Mutant A (GenBank KJ183180), Mutant B (GenBank KJ183181), and Mutant C (GenBank KJ183182), we used PCR mutagenesis from the pET32-Inf4 template, essentially as described in [18]. This procedure utilized two vector-specific primers and two specific oligonucleotides that had 30-nt complementary sequences at their 5’-termini; these overlapping parts of the oligonucleotides incorporated the designed mutations. One DNA fragment (165 bp) was amplified with a vector-specific forward primer (VectorFor) and a mutation-specific reverse primer (MutRev); the second fragment (181 bp) was produced by PCR with a mutation-specific forward primer (MutFor) and a vector-specific reverse primer (VectorRev) (S1 Table). The generated fragments were combined and fused by PCR using a pair of vector-specific primers. The resulting DNA fragment (320 bp) was cloned into pET32a using the NcoI and HindIII sites (the NcoI recognition site was located 20 bp downstream of the VectorFor sequence in the plasmid). The plasmids pET32-MutA, -MutB, -MutC, and–Mut15 were transformed into E. coli BL21 (DE3) cells. The cells were plated on ampicillin-containing medium, and clones were selected by PCR screening and were confirmed by sequencing. The mutant proteins were expressed and purified as described above for wild-type infestin.

Global coagulation assays were performed using Mutant B that was preliminarily cleaved from Trx with 1 U of bovine thrombin (MP Biomedicals; Santa Ana, CA) per 1 mg of the fusion protein. The cleaved Mutant B protein (the reaction product) was purified from both thrombin and Trx. The reaction mixture was applied on a SP Sepharose HiTrap column (GE Healthcare) at 0.5 ml/min followed by washing the column with 20 mM Tris-HCl, pH 8.0. The target product was collected in the flow-through fraction which was then applied to a Tricorn 10/100 column containing SOURCE 15Q resin (GE Healthcare). The residual thrombin level in the Mutant B preparations was below the limit of detection using the thrombin-specific chromogenic substrate. The total yield of Mutant B (13 kDa) was 25 mg per liter of cell culture.

CMTI-III expression

The most efficient system for the expression of the recombinant Cucurbita maxima trypsin inhibitor CMTI-III (29 aa) was a fusion with the polyhistidine (His)-tagged domain B1 of the Streptococcus sp. protein G (GB1) [19]. Other versions, including 1) His-tagged CMTI-III, 2) a polypeptide containing several CMTI-III domains, 3) fusion with glutathione-S-transferase (pGEX-4T-1; GE Healthcare), and 4) fusion with the chitin-binding domain (pTYB1 intein system; New England Biolabs; Ipswich, MA), were less favourable because of either protein degradation, misfolding, or low yield [20]. A DNA fragment coding for GB1, a linker peptide, and CMTI-III [21] was constructed from four oligonucleotides with 20-nt-long complementary termini (S1 Table). Two primer pairs (FuCMTI1up and FuCMTI2low; FuCMTI3up and FuCMTI4low) were annealed and extended with PCR. The reaction products FuCMTI1up-FuCMTI2low (180 bp) and FuCMTI3up-FuCMTI4low (175 bp) were annealed and extended because the primers FuCMTI2low and FuCMTI3up had 20-nt-long complementary 5’-termini. The final DNA fragment (335 bp) was cloned into pET28a (Merck) with the NcoI and EcoRI sites (New England Biolabs) that were introduced by the FuCMTI1up and FuCMTI4low primers, respectively. The generated plasmid p28FuCMTI was transformed into E. coli NovaBlue competent cells, sequenced and transformed into E. coli BL21 (DE3) cells, which were grown in LB with 0.025 mg/ml kanamycin. The expression of GB1-CMTI-III (GenBank KJ183183) was induced overnight with 0.4 mM IPTG at 18°C, and the protein was purified on Ni-NTA resin, with a yield of 2.5 mg CMTI-III per liter of cell culture.

Chromogenic assay

The selectivity profiles of the inhibitors were studied against ten major coagulation-related serine proteases. The respective protease (50 μl) was mixed with varying concentrations of an inhibitor (0.5 nM – 50 μM; 50 μl) in a buffered solution (100 mM Tris-HCl pH 7.4, 260 mM NaCl, 1.0% BSA) and incubated in a 96-well microtiter plate (Corning; Corning, NY) for 15 min at 37°C. An appropriate chromogenic substrate (100 μl) diluted with water was added to the mixture, and the kinetics of the p-nitroaniline (pNA) generation were monitored by measuring the absorbance at 405 nm using a Sunrise Microplate Reader (Tecan; Männedorf, Switzerland). The final concentration of each protease and its substrate in a total volume of 200 μl are listed below (Table 1). The initial rate of substrate hydrolysis (determined as a slope of the pNA generation curve) was normalized to the value at a zero inhibitor concentration and plotted versus the inhibitor concentration. This curve was fitted with the equation V(%) = 100%/(1 + I/IC₅₀), where I is the inhibitor concentration, and IC₅₀ is the inhibitor concentration that inhibited the protease amidolytic activity by 2-fold. The inhibitory constant K_i was calculated according to the Cheng-Prusoff equation K_i = IC₅₀/(1 + S/K_m) [22], where S is the initial substrate concentration and K_m is the Michaelis constant (Table 1). The SD value for the inhibitory constant K_i accounted for both errors of measurement for IC₅₀ and K_m and was calculated as the square root of the sum (ΔIC₅₀/(1 + S/K_m))² + n*(K_i/(1 + K_m/S) *ΔK_m/K_m)², where ΔIC₅₀ –SD value for IC₅₀, ΔK_m and n–SE for the K_m value and a number of measurements of this value, respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Coagulation-related proteases and their corresponding substrates used in the chromogenic assays.

Concentrations of the serine proteases (nM) and their substrates (mM) in a final volume of 200 μl. The K_m values (mM) are expressed as the mean ± SE (number of repeats).

https://doi.org/10.1371/journal.pone.0144940.t001

Measurement of the K_m values was performed essentially as described above. The appropriate protease (100 μl) was mixed with varying concentrations (approx. 0.5*K_m− 3*K_m) of the appropriate chromogenic substrate (100 μl) in the indicated buffer. Kinetics of the pNA generation were monitored immediately at 37°C. The substrate cleavage rate was plotted versus the concentration of substrate and fit with the Michaelis-Menten equation; the K_m value was calculated as the mean ± SE.

Molecular modeling

Blood collection and plasma preparation

This study including the blood collection procedures was approved by the institution’s Ethics Committee of the Center for Theoretical Problems of Physicochemical Pharmacology (Moscow, Russia) and was performed in accordance with the Declaration of Helsinki. Blood (9 ml) was collected from healthy volunteers who provided written consent. In the coagulation experiments, normal plasma was prepared from citrated blood drawn from the median cubital vein into Vacuette plastic tubes with 3.2% sodium citrate (Greiner Bio-One; Kremsmünster, Austria). Platelet poor plasma (PPP) was prepared by centrifugation of the citrated blood at 1,600 g for 15 min. Platelet free plasma (PFP) was obtained from PPP by a second centrifugation at 10,000 g for 5 min. Normal PPP and PFP from 5–10 healthy donors were pooled and frozen at -80°C. A sample of frozen plasma was thawed in a 37°C water bath for at least 30 min prior to use. In the experiments with coagulation factor-deficient or -depleted plasma, we used lyophilized fXII-depleted plasma and lyophilized fVIII- and fIX-deficient plasma (Renam; Moscow, Russia), which were reconstituted with water according to the manufacturer’s instructions and clarified by centrifugation at 10,000 g for 15 min.

When clotting time of the uncitrated whole blood was measured, 0.7 ml of fresh whole blood drawn using a Vacuette 21G needle (Greiner Bio-One) immediately after the collection procedure was placed into a 1.5-ml polypropylene tube (Axygen; Union City, CA) pre-filled with 78 μl of fXIIa inhibitor or a vehicle (30 mM Hepes pH 7.4). The tubes were rotated at 10 rpm, and the time of blood clotting was visually examined.

APTT and PT assays

The activity of the fXIIa inhibitors in plasma was estimated by their effects on the activated partial thromboplastin time (aPTT) and prothrombin time (PT). These assays were conducted according to standard protocols from the reagent manufacturer (Renam) using a Helena C-2 coagulometer (Helena Biosciences Europe; Gateshead, UK). Briefly, 50 μl of normal frozen-thawed PFP was incubated in the absence or presence of increasing concentrations of inhibitor (5.6 μl) at 37°C for 15 min; then, two equal aliquots of this mixture (25 μl each) were placed in two cuvettes (Helena Biosciences Europe) in the coagulometer. In the aPTT assay, a kaolin-cephalin mixture (25 μl) was added to each cuvette and incubated with plasma for 3 min to activate the contact pathway. The aPTT measurement was performed in duplicate and initiated by the addition of 25 mM CaCl₂ (25 μl). Each experiment was performed at least 3 times. In the PT assay, a mixture of rabbit brain TF and 12.5 mM CaCl₂ was added to plasma (50 μl in each cuvette) to initiate clotting. The anticoagulant activity of the inhibitor (designated as CT₃) was quantitated as the concentration required to prolong the clotting time in the aPTT assay by 3-fold compared to normal plasma without inhibitor. CT₃ was calculated by fitting the aPTT dose-dependent curve (fold increase) with a linear function, P₁*I+P₂ (where I is an inhibitor concentration in μM, P₁, P₂ –coefficients, and ΔP₁ and ΔP₂ are the corresponding errors). CT₃ was calculated as (3 –P₂)/P₁. The standard error was the square root of (ΔP₂/P₁)² + (ΔP₁*P₂/P₁²)².

Thrombin generation assay

The thrombin generation experiments were carried out essentially as previously described [34]. The thrombin fluorogenic substrate Z-Gly-Gly-Arg-AMC (Bachem; Bubendorf, Switzerland—740 μl of a 2.5 mM solution dissolved in buffer 18 mM Hepes pH 7.5, 130 mM NaCl, 10% DMSO) was added to a sample of frozen-thawed PPP (2960 μl). This mixture was incubated at 37°C with various concentrations of Mutant B, CTI, or a vehicle (30 mM Hepes pH 7.4), followed by the addition of 100 μl of this mixture into a 96-well microtiter plate. Thrombin generation was initiated by the addition of a trigger solution (20 μl), which contained CaCl₂ (16.7 mM; final concentration in a total volume of 120 μl) and phospholipids (4 μM; phosphatidylcholine: phosphatidylserine 80:20). Phospholipid vesicles were prepared as previously described [35]. For the contact-activated generation, the trigger solution also included kaolin (kaolin was 200 times more dilute in a total volume of 120 μl than in the standard aPTT assay), while for the TF-induced generation, the trigger solution included rabbit TF (Renam; final concentration 5 pM). The kinetics of the 7-amino-4-methyl-coumarin (AMC) release were continuously measured immediately after the addition of the trigger solution at 37°C for 60 min using an Appliskan Multimode Microplate Reader (Thermo Scientific; Helsinki, Finland) (excitation and emission wavelengths: λ_exc 355 nm, λ_em 460 nm). In parallel, the fluorescence was measured in calibration wells containing the plasma-substrate mixture (100 μl) and 20 μl of AMC (a final concentration 8 μM in a total volume of 120 μl). The background fluorescence level was measured in wells containing the plasma-substrate mixture (100 μl) and 20 mM Hepes pH 7.5, 145 mM NaCl (20 μl). Each measurement was performed in duplicate. The thrombin generation curve for the experimental samples was calculated by averaging the duplicates, subtracting the background, converting the fluorescence signal into the AMC concentration and differentiation of the AMC kinetic curve. Consumption of the fluorogenic substrate and the contribution of the α2-macroglobulin were also taken into account.

Thromboelastography

The assay was performed using a TEG 5000 Hemostasis Analyzer System (Haemoscope; Braintree, MA). The effects of the Mutant B dose were determined on frozen-thawed samples of normal PPP and reconstituted samples of lyophilized fXII-depleted plasma (Renam). To study contact-activated clotting, 330 μl of normal PPP was pre-incubated at 37°C with various concentrations of Mutant B, CTI, or vehicle and placed in a warmed cup containing 30 μl of 200 mM CaCl₂. In the experiments studying TF-dependent coagulation, 300 μl of fXII-depleted plasma pre-incubated with fXIIa inhibitors was placed in a cup containing both 30 μl of 200 mM CaCl₂ and 30 μl of TF (Renam; final plasma concentration 0.6 pM in a total volume of 360 μl). The measurement of the clot viscoelasticity lasted for 80 min. The values for reaction time R (min), clot formation time K (min), the α angle (°), and the maximum amplitude MA (mm) were analyzed.

Thrombodynamics assay

The impact of the fXIIa inhibitors on the spatial dynamics of clot growth was studied using a Thrombodynamics Analyzer; the device and the assay reagents were from HemaCore (Moscow, Russia). The assay was performed as previously described in detail [34,36,37]. Briefly, 120 μl of frozen-thawed PFP or reconstituted lyophilized plasma sample was mixed with the fXIIa inhibitor (Mutant B, CTI) or a vehicle (13.3 μl) and incubated at 37°C for 15 min. Plasma was placed in a tube containing lyophilic calcium acetate (final concentration in citrated plasma 20 mM that corresponded to a concentration of free Ca²⁺ ions 2 mM). After re-calcification, this mixture was placed into a flat transparent assay cuvette and the assay was started immediately since the plasma was brought in contact with an activator coated with surface-immobilized TF [38]. Clotting in the assay cuvette filled with plasma was monitored for 30 min using dark-field video microscopy. A series of the cuvette images that visualized the growing fibrin clots were processed. For a clot grown from the immobilized TF (“TF-initiated clot”), the size in the direction of growth (μm) versus time (min) was plotted, while for the TF-independent clots grown far from the activator, the cuvette area occupied by the clots (%) was plotted versus time (min).

Statistical analysis

Data were analyzed with OriginPro software (Microcal Software; Northampton, MA). The data presented are the mean of at least 3 repeats ± standard deviation (SD), unless otherwise noted. Prolongation of the whole blood clotting time by fXIIa inhibitors was analyzed by the paired-sample Wilcoxon signed rank test.

Selectivity of the fXIIa inhibitors

First, we estimated the degree of selectivity of the wild type infestin-4 against serine proteases of blood coagulation. The wt-Inf4 inhibitory activity against purified fXIIa was measured in the chromogenic assay and compared with other potent inhibitors of fXIIa: corn trypsin inhibitor (CTI) from the cereal family and the squash-family trypsin inhibitors from C. maxima (CMTI-III) and L. cylindrica (LCTI-III) (Fig 1A). Wt-Inf4 strongly inhibited fXIIa with an inhibitory constant K_i = 0.113 ± 0.023 nM, which agreed with previously published values [13,15]. CTI inhibited fXIIa approximately 4-fold less, while both CMTI-III and LCTI-III had K_i values of 10 nM or greater. The off-target activity of wt-Inf4 and CTI were measured in the chromogenic assay with a coagulation factor (XIa, IXa, Xa, thrombin, or VIIa) or a coagulation-related protease (plasma kallikrein, plasmin, tPA, or aPC). Wt-Inf4 was a strong inhibitor of plasmin (K_i 4.7 ± 1.5 nM) and a weak inhibitor of fXa (2.43 ± 0.87 μM), fIXa (1.06 ± 0.25 μM), fVIIa (4.7 ± 1.7 μM), and thrombin (33.4 ± 9.4 μM) (Fig 1B), but CTI weakly inhibited fXIa (15.4 ± 2.6 μM) and aPC (15.9 ± 3.8 μM) (Fig 1C). Furthermore, proteins tested inhibited neither plasma kallikrein nor tPA up to a concentration of 30 μM. CTI had approximately 40,000-fold selectivity, and wt-Inf4 had approximately 10,000-fold selectivity against human coagulation factors (if to ignore its inhibition of plasmin).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Wt-Inf4 and CTI are the most potent inhibitors of fXIIa and have some off-target activities.

(A) The residual amidolytic activity (%) of fXIIa upon incubation with various concentrations of wt-Inf4 (red), CTI (black), LCTI-III (cyan), or CMTI-III (magenta). (B, C) The residual amidolytic activity (%) of the coagulation-related proteases plasma kallikrein (purple pentagon), plasmin (cyan), fXIa (blue), fIXa (purple sphere), fXa (red), thrombin (black), fVIIa (magenta), and aPC (orange) upon incubation with various concentrations (logarithmic scale) of wt-Inf4 (B) and CTI (C).

https://doi.org/10.1371/journal.pone.0144940.g001

Design of infestin-4 mutants

Prior to testing infestin-4 as a reagent to block in vitro contact activation, we attempted to eliminate its off-target activity by introducing mutations in its reactive site region. First, to design the mutants, we analyzed the individual amino acids in the reactive site sequences of 89 inhibitors from the Kazal family [39], including infestin-4 (P² –P^4’ sequence FRNYVP). The comparison revealed that the Phe9 residue found at the P² reactive site position of wt-Inf4, very rarely appeared at this position in the Kazal-type inhibitors (Fig 2A). In addition to wt-Inf4, Phe was found at the P² position in only the sequences of dipetalogastin_6 and brasiliensin_8, which are predicted domains with unknown activities. The P² position was more frequently occupied by either Pro or Thr; the latter is known to form hydrogen bonds with the P^1’ residue thus promoting the canonical conformation of an inhibitor [40]. Moreover, P² Thr was contained in an infestin-4 Mutant 15 (P² –P^4’ sequence TRRFVA), which was previously reported [16]. To elucidate whether the uncommon P² Phe9 residue affected the structure of Inf4, we applied a 150 ns molecular dynamics (MD) simulation to the crystal structure of wt-Inf4 (PDB ID 2ERW). The simulation revealed that the side chains of the Phe9 and Arg10 residues appeared in close proximity to each other; the distance between these groups (4 Å) is shown as a yellow dashed line in Fig 2B and indicates a π-cation interaction between the Phe9 and Arg10 residues. It was unusual that there was no indication of the typical interaction between the P² and P^1’ residues in wt-Inf4; this interaction occurs in a majority of the canonical inhibitors [40]. Hence, we designed mutants of Inf4 where Phe9 was substituted by a more common Thr residue (Mutant B) or by Asn, another hydrogen forming residue (Mutant C). In Mutant B, Asn11 (P^1’ position) was conserved, while Asn11 was substituted by Arg in Mutant C because Arg11 was present in a majority of fXIIa-binding peptides from the study [16], including Mutant 15. Phe9 was conserved in Mutant A, whereas Asn11 was substituted by Arg in this mutant. Additionally, the P^2’ Tyr and P^4’ Pro residues of wt-Inf4 were substituted by Phe and Ala residues, respectively, in Mutant A and Mutant B (for a full list of the P² –P^5’ sequences see Fig 2A). Arg10 at the P¹ position, known to determine the general specificity for the trypsin-like proteases, was preserved in all Inf4 variants.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Design of mutations in the reactive site region of Inf4.

(A) A web-logo representation [41] of the P²–P^5’ binding loop sequences of 89 proteinase inhibitors from the Kazal family [39]. The amino acid positions are numbered by the wt-Inf4 sequence Phe9–Val15. Sequences of the corresponding reactive site fragments from wt-Inf4 and its mutants are shown below. (B) The MD simulation indicated a π-cation interaction (yellow dashed line) in wt-Inf4 between the phenyl group of Phe9 and the guanidinium group of Arg10. The nitrogen atoms are shown in blue, oxygen in red, and hydrogen in white. The torsion angles ψ, ϕ, χ₁, and χ₂ that describe the scissile bond conformation and rotations of the Arg10 side chain are drawn. (C) Superposition (main chain atoms) of wt-Inf4 (the most abundant conformation over the MD trajectory) with the structures of Inf4 mutants whose reactive loop solely adopted the canonical conformation. A close-up view of the reactive site region is indicated with a rectangle (wt-Inf4 –cyan, Mutant A–yellow, Mutant B–orange, Mutant C–green, Mutant 15 –magenta; this coloring is also used in the following Figs). The scissile bond of wt-Inf4 is indicated with a wavy line; a star indicates the Arg10 carbonyl oxygen. (D) A Coomassie-stained 12.5% SDS-PAGE showing the one-step purification procedure for one of the Inf4 versions expressed in E. coli BL21(DE3) cells (Trx-Mutant B is shown as an example). The protein samples applied to the gel are as follows: M–molecular weight marker (kDa); Induction 0 h and 4 h–cell lysates before 1 mM IPTG induction and 4 h later, respectively; Debr and Sol–insoluble and soluble fractions, respectively; Fl-thr–flow-through of the Sol fraction applied to the column; Wash–washing the column with 50 mM Tris, pH 7.5, 500 mM NaCl, 100 mM imidazole; Resin–Chelating Sepharose with bound and washed Trx-Mutant B (26 kDa); Elt–Chelating Sepharose after Trx-Mutant B elution with 50 mM Tris-HCl, pH 7.5, 500 mM imidazole; thereafter, the protein sample was applied on the SOURCE 30Q column. (E) Following the two-step purification, the Trx-fused protein was cleaved with bovine thrombin. Thrombin (mU)–indicated amounts of bovine thrombin (0, 1, 2, 3, 4, 6, and 10 milliunits) were added to 10 μg of Trx-Mutant B resulting in two separate bands of Mutant B (13.1 kDa) and Trx (12.9 kDa). (F) Mutant B–indicated quantities of the Mutant B protein (10, 1, and 0.5 μg loaded into gel) that was purified after the cleavage reaction, as described in the sub-section “Expression of the thioredoxin-fused infestin-4 and its mutants”.

https://doi.org/10.1371/journal.pone.0144940.g002

Expression of thioredoxin-fused infestin-4 and its mutants

The mutants were analyzed using the MD simulation, and it showed that in Mutant B, ‘canonical’ hydrogen bonds were formed between the P² Thr9 and P^1’ Asn11 residues (the distance between the Thr9 side chain oxygen and the P^1’ Asn11 main chain nitrogen was mostly in a range of 3–5 Å over the MD trajectory). Similar interactions between the P² and P^1’ residues were proposed in Mutant C and Mutant 15. Even in Mutant A, where Phe9 was conserved, this residue interacted with the P^1’ Arg11 (the distance between their side groups approached 4–5 Å) than with P¹ Arg10 (distance was ≈8 Å). Then, structures of the Inf4 mutants overlapped with the model structure of wt-Inf4 (Fig 2C). The scissile bond region of wt-Inf4 (labeled with a wavy line) differed significantly from the canonical conformations of the Inf4 mutants, as well as from the other inhibitors CTI (the MD trajectories were from [10]), CMTI-III, and LCTI-III. As shown in Fig 2C, the Arg10 carbonyl oxygen of wt-Inf4 (labeled with a star) was oriented in the opposite direction compared to the other structures.

After the analysis was performed, the Inf4 variants were expressed in E. coli BL21 (DE3) cells from a pET32a vector as thioredoxin (Trx)-fused proteins. These recombinant proteins were purified using nickel-chelating chromatography (Fig 2D). Some additional bands were present in those preparations, and an ion-exchange chromatography step was used to further purify the fusion-proteins to homogeneity (Fig 2E; right half). The approximate yield of the homogeneous proteins was 20–60 mg per liter of cell culture. In some coagulation experiments, Mutant B was cleaved from the Trx-fusion with thrombin (Fig 2E; left half) and purified to homogeneity (Fig 2F).

Inhibition of fXIIa by the mutants

We characterized the inhibitory activities of the expressed mutants. To assess the inhibitory activity of recombinant Inf4 mutants against fXIIa, they were incubated with fXIIa in a buffered solution, followed by the addition of the S-2302 chromogenic substrate (Table 1) and measurement of the residual amidolytic activity of fXIIa. To address the ability of the Inf4 mutants to inhibit contact-activated clotting in blood plasma, they were incubated with citrated pooled plasma followed by 3-min incubation with kaolin and phospholipids, re-calcification, and measurement of the activated partial thromboplastin time (aPTT). The clotting time was dose-dependently prolonged by the inhibitors, and the CT₃ value was defined as the concentration of inhibitor that prolonged the clotting time by 3-fold compared to the clotting time in plasma without inhibitor (47 s in normal plasma). The CT₃ value was drawn on the plot for each inhibitor versus its K_i value measured in saline (Fig 3A). Mutant B was the most active of the mutants–its CT₃ concentration was 19 ± 3 μM, similar to wt-Inf4; nevertheless, its K_i value for fXIIa (0.72 ± 0.17 nM) was 6 times greater than that of wt-Inf4. S1 Fig shows the concentration-dependent inhibition of fXIIa amidolytic activity by Trx-fused Mutant B that was purified by two-step chromatography, as described in Materials and Methods. The Lineweaver-Burk plot (Fig 3B) was created by plotting the inverse rate of S-2302 cleavage by fXIIa (1/V, nM^-1*sec) versus the inverse concentration of S-2302 (1/S, μM^-1). This plot yielded three lines at Mutant B concentrations of 0 nM, 0.5 nM, and 1 nM; these lines had the same Y-intercept. It indicated that Mutant B had a competitive mode of action against fXIIa, which was previously reported for wt-Inf4 [15] and CTI [17]. Other Inf4 mutants had the CT₃ concentrations between 25 and 32 μM and K_i values in the range 1–10 nM (Fig 3A). For comparison, CTI had a CT₃ of 8.5 ± 1.6 μM and a K_i towards fXIIa of 0.41 ± 0.10 nM. The corresponding values for the squash-type inhibitors CMTI-III and LCTI-III were also plotted. Moreover, none of the inhibitors prolonged the prothrombin time.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Analysis of fXIIa binding by Inf4 variants.

(A) The CT₃ values (μM) measured in the aPTT assay in normal plasma are plotted against the K_i values for the Inf4 variants and the inhibitors of fXIIa from other families (CTI, CMTI-III, LCTI-III). The mean values ± SD are presented (n = 3). (B) Lineweaver-Burk plot (1/V, nM^-1*sec, versus 1/S, μM^-1) of fXIIa inhibition by 0, 0.5 and 1 nM of Mutant B at various concentrations (100, 200, 300, and 400 μM) of S-2302 substrate (mean + SD values, n ≥ 3). The data fitting with linear functions is shown with dots. (C, D) Complex of fXIIa (light violet) docked with wt-Inf4 (C) and Mutant B (D). The P¹ Arg10 residue of these inhibitors formed a salt bridge with S¹ Asp185 of fXIIa and the scissile bond was close (3.2–3.9 Å) to the catalytic residues Ser191 and His40. (E) A diagram representing the mean values of ln(1/K_i) (left axis) (n = 3; SD was approx. 0.2, i.e., 1%), where K_i is the inhibitory constant of the Inf4 variants for fXIIa; this value is proportional to the energy term. The values of the energy-like scoring function SF (right axis) were calculated with the ClusPro server.

https://doi.org/10.1371/journal.pone.0144940.g003

We also analyzed the binding of the Inf4 variants to fXIIa using flexible protein-protein docking; the homology model of fXIIa had been previously constructed [10]. In the complex of wt-Inf4 (cyan) and fXIIa (violet), the P¹ Arg10 side chain protruded and formed a salt bridge with the Asp185 of the S1 specificity pocket of fXIIa (Fig 3C). The catalytic atoms Ser191 O^γ and His40 N^ε2, which are involved in a nucleophilic attack on the scissile bond, were placed 3.2 and 3.9 Å from C and N atoms of the scissile bond, respectively (red dashed lines). The protease-binding loop was correctly positioned in the catalytic pocket in relation to the triad Ser191, His40, and Asp89; this position corresponds to a strong inhibition of fXIIa by wt-Inf4. Similar correct positions of the protease-binding loops were obtained in the complexes of fXIIa with Mutant B (Fig 3D) and other Inf4 mutants. Values of the scoring function (SF), which is dependent on the association energy, were calculated for the docked complexes (Table 2). The SF values were correlated with the experimentally measured K_i values for fXIIa (Fig 3E), which may confirm our biochemical data.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Association of the Inf4 variants with FXIIa.

Characterization of the anti-fXIIa activity of wt-Inf4, its mutants and the canonical inhibitors CTI, LCTI-III, and CMTI-III. The reactive site P²–P^4’ sequences are shown; K_i values (nM) against fXIIa and CT₃ concentrations (μM) estimated with the aPTT assay are presented as the mean ±SD for n ≥ 3. Dimensionless values of the scoring function SF are provided as a result of flexible protein-protein docking of these inhibitors with the model fXIIa structure by ClusPro service.

https://doi.org/10.1371/journal.pone.0144940.t002

Off-target activities of the Inf4 mutants

We estimated the impact of the mutations on the off-target activities of infestin-4. In contrast to wt-Inf4, none of the Inf4 mutants inhibited fXa (Fig 4A). Additionally, the Inf4 mutants had only minor, if any, activity against plasmin: their K_i values were 3–4 orders of magnitude greater than that of wt-Inf4 (4.7 ± 1.5 nM). None of the tested versions of Inf4 inhibited fXIa, tPA, plasma kallikrein, and aPC. Mutant A and Mutant 15 did not inhibit thrombin at a concentration up to 30 μM (Table 3), while Mutant B was more active against thrombin (K_i 2.9 ± 1.2 μM) than wt-Inf4 (K_i 33.4 ± 9.4 μM). Nevertheless, compared to wt-Inf4 and the other mutants, Mutant B had the least off-target activity against fIXa and fVIIa (Tables 3 and 4). In general, anti-fXa activity was fully eliminated by mutation of the reactive site of Inf4. Moreover, off-target activities against fIXa, fVIIa, thrombin, and plasmin were significantly reduced in (a subset of) these mutants.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Inf4 mutants did not inhibit FXa.

(A) Residual amidolytic activity of fXa at various concentrations of the Inf4 variants – 100% activity of fXa corresponds to a k_cat value of 240 s^-1. The mean ± SD values are shown (n = 3). Data fitting with a hyperbola is shown with dots. (B, C) Complexes of either wt-Inf4 (B; cyan) or Mutant B (C; orange) docked with fXa (secondary structure ribbons and C atoms are beige). Hydrogen bonds and salt bridges between charged functional groups are shown as yellow dashed lines. Red dashed lines show distances between the atoms involved in the cleavage reaction: 1) direction of the nucleophilic attack of Ser195 O^γ on the Arg10 carbonyl carbon and 2) direction of the proton transfer from His57 N^ε2 to the Asn11 carbonyl nitrogen (during the formation of the tetrahedral intermediate, a proton is acquired by His57 from Ser195, then a proton is transferred to the scissile bond to effect the cleavage). The scissile bond (shown as wavy line) of wt-Inf4 (B) was located in the fXa pocket properly for formation of the intermediate. P¹ Arg10 was bound to S¹ Asp189 and Gly218 by salt bridges. In addition, the Asn11 side chain of Mutant B (C) interfered with the catalytic Ser195 O^γ and His57 N^ε2 atoms.

https://doi.org/10.1371/journal.pone.0144940.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. K_i (μM) values of Inf4 versions against fXa, thrombin, plasmin, and fIXa.

K_i values (μM) are presented as the mean ±SD for n ≥ 3. N.I.–not inhibited at concentrations of up to 30 μM of the inhibitor.

https://doi.org/10.1371/journal.pone.0144940.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. K_i (μM) values of wt-Inf4, Mutant B, Mutant 15, and CTI against coagulation-related serine proteases.

K_i values (μM) of wt-Inf4, Mutant B, Mutant 15, and CTI against the serine proteases: fXIa, fVIIa, aPC, plasma kallikrein, and tPA, are shown as the mean ±SD for n ≥ 3.

https://doi.org/10.1371/journal.pone.0144940.t004

We attempted to understand why mutants of Inf4 did not inhibit fXa (Fig 4A). Complexes of fXa with Inf4 variants were studied using flexible protein-protein docking. In the complex of wt-Inf4 (cyan) with fXa catalytic domain (beige; PDB ID 2JKH), the P¹ Arg10 residue formed a salt bridge with the Asp189 residue of the S1 specificity pocket (Fig 4B). The scissile bond of wt-Inf4 was adjacent to the catalytic triad Ser195, His57 and Asp102, 4.4–5.8 Å apart from the Ser195 O^γ and His57 N^ε2 atoms. In the complex of Mutant B (orange) and fXa (beige), where the P¹ Arg10 was also bound to S¹ Asp189, the Asn11 residue was able to form hydrogen bonds with the catalytic Ser195 and His57 residues (Fig 4C). We proposed that the hydrogen bonding of Asn11 with the catalytic residues might prevent the nucleophilic attack on the scissile bond, thus interfering with the standard mechanism of inhibition [42]. Such interference could result in a weak noncovalent binding of Mutant B with fXa and a loss of the inhibitory activity. Similar ‘interfering’ interactions with the catalytic residues of fXa were also proposed for the remaining Inf4 mutants (data not shown).

Specific repression of contact-activated clotting in blood and plasma by Mutant B

Finally, we addressed the potential value of Mutant B, one of the most potent Inf4 mutants, as a new anticoagulant. Using a set of the global assays, we investigated whether this mutant, which weakly inhibited thrombin in saline, could be used in plasma to repress the contact pathway without affecting TF-dependent coagulation (i.e., the extrinsic pathway). For better reliability, Mutant B was cleaved and purified from Trx for these experiments (Fig 2E and 2F). The thrombin generation assay triggered by either kaolin, an activator of the contact pathway (Fig 5A), or 5 pM TF (Fig 5B) revealed that Mutant B did not affect TF-dependent generation in normal plasma, whereas it repressed the contact-dependent generation of thrombin. According to the competitive mode of action, Mutant B dose-dependently prolonged the time to reach the thrombin peak (Fig 5C) and reduced the thrombin amplitude (Fig 5D) when coagulation was activated by kaolin. The effective concentration of Mutant B was approx. 5 μM, similar to that of CTI. Similar effects were observed in the thromboelastography assay. Mutant B as well as CTI, dose-dependently delayed clotting in normal plasma triggered by contact with the assay cuvette (Fig 5E). These inhibitors dose-dependently prolonged the ‘R-time’ value in normal plasma without the addition of an activator (Fig 5F, dark bars) while both inhibitors had no effect on clotting in fXII-depleted plasma triggered by 0.6 pM TF (Fig 5F, light bars). In another experiment, when blood was collected into plastic tubes pre-filled with either Mutant B or the vehicle, whole blood clotting time was visually measured. Clots in the whole blood might appear due to the contact with tube walls. Clotting became visible at 29.1 ± 8.9 min after the collection of blood into tubes containing Mutant B (similarly to the effect of CTI; data not shown) compared to at 12.4 ± 1.8 min for tubes without an inhibitor (Fig 5G). In this experiment, the calculated concentration of Mutant B in plasma was 20 μM (near its CT₃ value), and no other anticoagulant (like sodium citrate) was used.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Mutant B selectively inhibits the contact pathway of coagulation.

(A–D) Thrombin generation in normal platelet-poor plasma was triggered with a mixture of calcium chloride and either kaolin (A) or TF (5 pM final concentration) (B) and was monitored for 90 min. Frozen-thawed plasma was preincubated at 37°C with various concentrations of Mutant B as indicated in the legend (0.2, 0.5, 1, 2, and 5 μM), CTI (0.5 and 5 μM) or vehicle (control). The assay parameters including the time to the thrombin peak (C) and the peak amplitude (D) were calculated for the experiments with the activation by kaolin (black bars) or 5 pM TF (gray bars). The mean values ± SD are shown; each experiment was performed in duplicate and repeated twice. (E, F) The thromboelastography assay was carried out in plasma preincubated with Mutant B (1, 2, 5, 10, and 20 μM), CTI (5 and 15 μM), or vehicle (control). (E) Thromboelastogram representing clotting in frozen-thawed platelet-poor plasma from healthy volunteers, drawn into a flask without any activator (coagulation was triggered by the contact pathway from flask walls). (F) Diagram showing the mean R-time values (n = 2) for thromboelastograms, obtained as a result of the assay either in normal platelet-poor plasma without activator (black bars), or in fXII-depleted plasma activated with 0.6 pM TF (gray bars). (G) Prolongation of the whole blood clotting time by Mutant B. The time to the visually detected clotting was measured in whole blood collected into tubes, which were prefilled with either vehicle (final concentration in blood 30 mM Hepes pH 7.4, empty squares) or Mutant B (10 μM, filled squares), without any other anticoagulant. Individual data from 5 donors and the median values are presented; for each donor the experiment was performed in duplicate. P-value (* < 0.005) was estimated using the Wilcoxon rank test.

https://doi.org/10.1371/journal.pone.0144940.g005

The use of Mutant B as a reagent for in vitro assays was also demonstrated with the thrombodynamics assay. In this assay, re-calcified plasma was placed in the assay cuvette, where a frontal growth of fibrin clot occurred from the surface with immobilized TF (Fig 6A). After re-calcification when no inhibitor of contact activation was applied, the problem was fXIIa autoactivation, which, as we proposed, occurred on walls of the assay cuvette and represented an artifact. It resulted in generation of TF-independent clots, which appeared in a bulk of normal plasma far from the TF localization site. We tested whether the inhibition of fXIIa by Mutant B would repress these fXIIa-initiated, artificial clots. The addition of either 20 μM Mutant B or 10 μM CTI into normal plasma successfully prevented the appearance of these contact-activated clots during a 30 min experimental time (Fig 6B). Moreover, Mutant B did not affect the growth of the TF-initiated clot both in normal and fXII-depleted plasma samples (Fig 6C) (as well as in fVIII- and fIX-deficient plasma; data not shown). However, in a patient’s plasma some TF-independent clots were formed even in the presence of a sufficient amount of the fXIIa inhibitor [37]. It was proposed that in this case, TF-independent clots were initiated by both fXIIa (generated during plasma handling) and circulating fXIa; the latter could be a marker of a hypercoagulant state [35]. Therefore, we used 20 pM fXIa added into normal plasma as a model of hypercoagulation. Neither Mutant B nor CTI prevented the appearance of the fXIa-initiated, TF-independent clots in this modeled plasma (Fig 6D). To visualize these experiments, we present images of the assay cuvette in 30 min after assay start in Fig 6E. If no inhibitor was added (upper panel), TF-independent clotting in normal plasma (initiated by autoactivation of fXIIa) was poorly distinguished from the clotting initiated by fXIa. In contrast, when plasma was pre-incubated with Mutant B (20 μM), clotting in ‘hypercoagulant’ plasma could be clearly distinguished from clotting in normal plasma, where no ‘artificial’ fXIIa activation occurred. From our point of view, all of these data provide a body of evidence regarding the sufficient selectivity of Mutant B, an improved variant of infestin-4, in plasma. These results particularly indicate that Mutant B can be used as a novel reagent to repress in vitro contact activation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Application of Mutant B in the thrombodynamics assay.

The thrombodynamics assay in platelet free plasma. (A) A representative dark-pole image of the assay cuvette filled with recalcified normal plasma (dark background) 30 min after the start of the thrombodynamics assay. TF was localized on the surface of a rectangular insert at the upper side of the cuvette (“Surface-immobilized TF”) and triggered a frontal growth of the bright fibrin clot (“TF-initiated clot”) in the downwards direction. Additional TF-independent artifact clots appeared far from the activator. (B) A percentage of the cuvette area occupied by fXIIa-initiated, TF-independent artifact clots in frozen-thawed normal plasma versus time (min). If no fXIIa inhibitor was added, then these clots appeared (occupied 5% of the cuvette area) 10 min after the assay was initiated and occupied the entire cuvette (100% of the cuvette area) 30 min later. Plasma samples were pre-incubated at 37°C with a vehicle (empty squares) or Mutant B (filled squares) and CTI (filled triangles) at their CT₃ concentrations: 20 μM and 10 μM, respectively. The mean values are presented for n = 10. (C) The size of the TF-initiated fibrin clot in a downward direction (μm) versus time (min) in fXII-depleted plasma. The plasma was preincubated at 37°C with Mutant B (20 μM, filled triangles) or vehicle (empty squares; mean values for n = 3 are shown). (D) A percentage of the cuvette area occupied by fXIa-initiated, TF-independent clots in frozen-thawed normal plasma versus time (min). These clots appeared approximately 10–12 min after the assay was initiated and occupied the entire cuvette 20 min later. The plasma samples were preincubated with a vehicle, Mutant B (20 μM) or CTI (10 μM). The mean values are presented for n = 10. (E) Images of the assay cuvettes 30 min after the assay was initiated, filled with normal frozen-thawed plasma that was spiked with 20 pM fXIa (“hyper”) or vehicle (“normal”) and preincubated with Mutant B (“MutB”) or vehicle (“no inhibitor“).

https://doi.org/10.1371/journal.pone.0144940.g006