Habu (Protobothrops flavoviridis) crude venom was provided by the Kagoshima Prefectural Public Health Office in Japan. P. flavoviridis [Asp49] PLA2 and BPII were prepared from crude venom as previously reported with some modifications (Tanaka et al., 1986; Liu et al., 1990). Sephadex G-75, a Hi-Trap NHS-activated HP column (1 ml) and a HiTrap benzamidine column (1 ml) were obtained from GE Healthcare Ltd. (Buckinghamshire, United Kingdom). CM-52 was purchased from Whatman (Maidstone, Kent, United Kingdom). POROS HS was obtained from Life Technologies (Carlsbad, CA, United States). Phenylmethylsulfonyl fluoride (PMSF) and RPMI-1640 were purchased from Sigma-Aldrich Co. (St. Louis, MO, United States). Dishes and 96-well microtiter plates were purchased from Becton Dickinson Co. (Franklin Lakes, NJ, United States). Centrifugal filter devices and ZipTip C18 pipette tips were obtained from Merck Millipore (Darmstadt, Germany). The Silver Stain MS Kit and V8 protease were purchased from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). The XL-trypKit was purchased from APRO (Osaka, Japan). Boc-Val-Pro-Arg-MCA was purchased from Peptide Laboratory (Osaka, Japan). SkMCs (human skeletal muscle origin cells) and their basal medium (SkBM) were purchased from Lonza Ltd. (Visp, Switzerland). SW839 (human kidney origin) cells were obtained from the Cell Resource Center for Biomedical Research (Institute of Development, Aging and Cancer, Tohoku University, Sendai, Japan). The Cell Counting Kit-8 was purchased from DOJINDO Lab, Inc., (Kumamoto, Japan). 3,3′,5,5′-tetramethylbenzidine was purchased from Tokyo Chemical Industry Co. Ltd. (Tokyo, Japan). All other reagents were of commercial grade and were purchased from Wako Pure Chemicals or Nacalai Tesque (Kyoto, Japan).

Two types of antivenom against habu venom, one without anti-myonecrotic activity (lot no.34 (manufacture’s serial number)) and one with anti-myonecrotic activity (lot no.1469), were obtained from the Kagoshima Prefecture government’s pharmaceutical affaires section and KM Biologics Co., Ltd. (formerly Kaketsuken or The Chemo-Sero-Therapeutic Research Institute, Kumamoto, Japan), respectively. Conventional antivenom (lot no.34) without anti-myonecrotic activity, which is a mixture of some horse sera immunized by different venoms derived from Okinawa habu or Amami one, is a product approved by the Japanese Pharmacopoeia (approved number: 61E1174, expiry dates: January 20, 2013, >300 units/ml for anti-hemorrhage activity (anti-HA)) as pharmaceutical agent for habu snake bite, Freeze-dried Habu Antivenom “Kaketsuken.” On the other hand, antivenom with anti-myonecrotic activity (lot no.1469) was newly prepared from serum of a female horse immunized by antigen injection as follows; preliminary twice immunization by toxoid (formalin-inactivated crude venom) with incomplete Freund’s adjuvant, subsequently, after for 7 weeks breaking, eight times booster immunization by increasing the dose of Amami habu crude venom (3–320 mg). To assess antibody titers of antivenoms against myonecrotic [Lys49] PLA2, ELISA test was conducted by using BPII ([Lys49] PLA2) and [Asp49] PLA2 as antigens. Briefly, purified antigens BPII and [Asp49] PLA2 (625 ng/well each) were respectively immobilized on the wells of a microtiter plate for 2 h at room temperature (RT), and blocked with 1% bovine serum albumin (BSA). After washing with phosphate buffered saline (PBS: 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, 2.7 mM KCl, 137 mM NaCl, pH7.4), 100 μL of serial diluted antivenoms (500–64,000 dilutions) were added on the microtiter plate wells, and were incubated at 4°C overnight. After washing with PBS three times, 100 μL of secondary antibody, horseradish peroxidase (HRP)-labeled goat anti-equine IgG (H + L) (Proteintech Japan, Tokyo), was added on plate wells, and incubated for 2 h at RT. After washing the plate four times with PBS, 100 μL of 3,3′,5,5′-tetramethylbenzidine substrate in 0.2 M citrate buffer (pH 4.0) including 0.01% H₂O₂ (substrate for HRP) was dispensed on the plate wells, and then monitored sufficient color development absorbance at 450 nm after adding 100 μL of the stop solution (1 M H₂SO₄).

DDY mice (6 weeks-old) were purchased from CLEA Japan, Inc. (Tokyo, Japan). The mice were maintained for 1 or 2 weeks and had free access to the laboratory animal standard diet and tap water. Protobothrops flavoviridis crude venom (CV, 20 μg) with/without 10 μL of antivenoms (lot no. 34 or lot no. 1468) (total volume: 15 μL) were injected into musculus gastrocnemius of mice left hindlimb, and the control group was injected with only saline. After 24 h, the mice were euthanized, and the hindlimbs were dissected for tissue preparation. In vivo experiments using BPII and TLf2 were also conducted same as crude venom and antivenoms described above. Thereafter, the tissue samples were dehydrated in a graded ethanol series and embedded in paraffin. 5 μm sections were cut in a microtome and stained with hematoxylin and eosin to be examined under the light microscope. The experimental plans used in this work were reviewed and approved by the Animal Resarch-Animal Care Committee of the Institute of Medical Science, the University of Tokyo or Graduate School of Agricultural Sciences, Tohoku University (approved number: 21BA-7). These procedures are in accordance with standards outlined by Japanese laws for use of experimental animals (Act on Welfare and Management of Animals), and with ethical principles adopted by the University of Tokyo and Tohoku University.

The IgGs from the two habu antivenoms, which were obtained by equine immunization with habu venom, were individually purified using protein G columns and then coupled overnight at 4°C using HiTrap NHS-activated HP columns with coupling buffer (0.2 M NaHCO₃ containing 0.5 M NaCl, pH 8.3). After washing with 4 ml blocking buffer (0.5 M ethanolamine containing 0.5 M NaCl, pH 8.3), followed by 4 ml washing buffer (0.1 M sodium acetate containing 0.5 M NaCl, pH 4.0), the columns were equilibrated with 5 ml PBS. The binding capacity of each column (2.0 mg/ml) was estimated by monitoring the adsorbed peak area measurement for serial concentrations of crude venom proteins.

Crude venom (2 mg) was dissolved in 2 ml PBS (pH 7.4) and applied to both habu antivenom IgG (lot no. 34)- and antivenom IgG (lot no. 1469)-conjugated columns. The columns were washed with PBS (pH 7.4) buffer and eluted with 0.1 M glycine-HCl (pH 2.7) buffer at a flow rate of 1 ml/min. The eluted protein fractions were neutralized by adding the 1/10 volume of 1 M Tris-HCl (pH 8.5). Flow-through and adsorbed peak fractions were individually dialyzed against distilled water and lyophilized.

For proteomics analysis using two-dimensional polyacrylamide gel electrophoresis (2D-PAGE), the proteins separated by the antivenom IgG-conjugated affinity columns were analyzed by isoelectric focusing (IEF) using the IPGphor System (Amersham Biosciences, Piscataway, NJ, United States) with immobilized pH gradient (IPG) strips (pI 3–11), and then subjected to the sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions. SDS-PAGE was performed on 15% slab gels according to the method described by Laemmli. The Auto 2D BM-100 device (SHARP CO., Ltd., Osaka, Japan) was also used to conduct 2D-PAGE analysis using pI 3–10 and 12.5% gels. A typical condition for IEF on Auto 2D BM-100 device was 5 steps (Step 1: 200 V for 5 min, Step 2: 200–1000 V linear gradient for 5 min, Step 3: 1000 V for 5 min, Step 4: 1,000–6000 V linear gradient for 10 min, Step 5: 6000 V for 5 min) at 100 μA, at 10°C. After running 2D-PAGE, the protein spots were stained with Oriole fluorescent gel staining (Bio-Rad Lab. Inc.) or using the Silver Stain MS Kit. Proteins were quantified and compared using CS analyzer ver. 3.0 and ImageSaver5 for EZ-Capture MG (ATTO Co., Tokyo, Japan).

The Silver-stained spots on the gel were cut into 1 mm³ pieces, washed with 100 μL of De-staining solution (in Silver Staining MS Kit, Wako Pure Chemicals), incubated with same solution at RT for 15 min with gentle agitation, washed twice with 500 μL of 0.1% trifluoroacetic acid (TFA) (15 min each), and then dehydrated by incubating the gel pieces with 100 μL acetonitrile at RT for 5 min. After removing the acetonitrile, the gel pieces were dried for 15 min in vacuo, and then reduced with 10 mM DTT in 25 mM ammonium bicarbonate buffer for 2 h at 56°C. After cooling to RT, the gel pieces were washed with 100 μL of 25 mM ammonium bicarbonate buffer and treated with 100 μL of 55 mM iodoacetamide. The gel pieces were then washed with 100 μL of 0.1% TFA for 10 min and twice dehydrated by treatment with 50% acetonitrile in 25 mM ammonium bicarbonate buffer. The gels were dried for 15 min in vacuo, and subsequently digested overnight at 37°C with trypsin using the XL-trypKit or V8 protease in 0.1 M NH₄HCO₃ containing 4 M urea. Then, 50 μL of extraction buffer solution (50% acetonitrile containing 5% TFA) was added to the gel pieces, which were incubated for 30 min at RT with shaking to extract the digests. After concentrating the digests, the samples were desalted using ZipTip C18 pipette tips according to the manufacturer’s instructions. The desalted samples (1 μL) were mixed with α-cyano-4-hydroxycinnamic acid matrix and then applied to a MALDI-TOF MS plate. Samples were analyzed by MALDI-TOF MS/MS reflector positive mode using the TOF/TOF™ 5800 Analyzer (ABSCIEX, Framingham, MA, United States) with mass range from 800.00 to 4,000.00 Da. In gel digested samples were also analyzed by nanoLC-MALDI TOF MS/MS using a DiNa Nano LC system equipped with a DiNa MALDI spotting device (KYA Technologies Co., Tokyo, Japan). Molecular masses were calibrated using the Sequazyme Peptide Mass Standards Kit (Applied Biosystems). Protein identification was performed by searching of each MS/MS spectrum against the protein sequence databases derived from the RNA-seq data of P. flavoviridis snake venom by using Protein Pilot software (version 3.0; AB Sciex) with the Paragon method.

Publicly available datasets were analyzed in this study. Protein profiling data can be found in the JPOST repository, a member of ProteomeXchange, with ID numbers of JPST001309/PXD028215 and JPST001307.

To isolate the two thrombin-like serine proteases, TLf1 and TLf2, the habu snake crude venom was separated by cation exchange chromatography, as previously reported with some modifications (Shieh et al., 1985). Briefly, habu crude venom was loaded onto CM-52 column (ϕ 15 mm × 870 mm) equilibrated with 20 mM ammonium acetate (pH 6.8) and eluted with a linear gradient to 500 mM ammonium acetate (pH 6.8) (Supplementary Figure S2). Flavoxobin (TLf1) obtained in the acidic flow-through peak fraction of the CM-52 column was further purified by affinity chromatography on the Hi-trap benzamidine Sepharose FF column (Supplementary Figure S3). The most basic peak fraction, which included TLf2, was further purified by cation-exchange chromatography on a POROS HS column (7.5 × 200 mm), and eluted with a gradient of 0–1 M NaCl in 50 mM Tris-HCl (pH 8.8) (Supplementary Figure S4). The N-terminal amino acid sequence of purified TLf2 was confirmed by gas-phase protein sequencing using the Shimadu PPSQ-10 Protein Sequencer (Shimadzu, Kyoto, Japan). Purified TLf1 was confirmed by MALDI-TOF MS/MS analysis with trypsin digestion (Supplementary Figure S5).

The hydrolytic activity of TLf2 was determined by incubation with fibrinogen (8.6 mg/ml) in 20 mM Tris-HCl buffer (pH 7.5) containing 0.15 M NaCl for 30 min and 24 h at 25°C. To assess fibrinogen degradation, SDS-PAGE was performed using a 15% separating gel and the protein bands were stained with Coomassie brilliant blue R-250. Furthermore, the hydrolytic activity of TLf2 was also analyzed using a synthetic fluorogenic substrate, Boc-Val-Pro-Arg-MCA (200 μM), in a 96-well black microtiter plate. Fluorescence changes, based on the amount of amino methyl coumarin (AMC) released, were monitored using a Gemini XPS-TON fluorescence microplate reader (Molecular Devices Japan, Tokyo, Japan). The excitation and emission wavelengths were 365 and 440 nm, respectively. Human α-thrombin (Mochida Pharmaceutical Co. Ltd., Tokyo, Japan) was used as a positive control.

SW839 and SkMC cells were cultured in 90 mm dishes containing 10 ml of RPMI-1640 supplemented with 10% fetal bovine serum (for SW839) and SkBM medium (for SkMC) at 37°C under an atmosphere of 5% CO₂. After reaching 80% confluence, the cells were detached from the dish by trypsin-EDTA treatment and plated on a 96-well microtiter plate at approximately 1.0 × 10⁵ cells/well. These cells were cultured under the same conditions described above for 24 h to approximately 70% of confluence prior to the cytotoxicity assay.

The proteins were dissolved in PBS buffer, and subsequently passed through a sterile 0.22 μm filter. The effects of TLf2 and TLf1 on the cytotoxic activity of [Lys49] PLA2 (BPII) were analyzed by incubating SW839 and SkMC cells with various concentrations of TLf2 and TLf1 mixed with 2.5 μg/ml and 25 μg/ml BPII. After incubation for 24 h, cell viability was analyzed using the Cell Counting Kit-8, measuring the absorbance at 450 nm using an iMark microtiter plate reader (Bio-Rad Laboratories, Hercules, CA, United States). All measurements were performed in triplicate. The data were evaluated by using Student’s t-test.

The BIAcore™ surface plasmon resonance (SPR) instrument (Amersham Biosciences) was used to determine the interaction between the TLfs and BPII. Purified TLf2 and TLf1 were diluted to a concentration of 100 μg/ml in 10 mM HEPES buffer (pH 7.5) and immobilized on the carboxymethylated dextran-modified gold surface of a CM5 sensor chip via primary amino groups using carbodiimide chemistry as described in the manufacturer’s protocol. Briefly, the carboxyl groups on CM5 were activated by 10 mM N-hydroxy succinimide and 400 mM N-ethyl-N’-(3-diethylaminopropyl)-carbodiimide, followed by the addition of 20 μL of TLf2 or TLf1. The remaining activated groups were blocked by addition of ethanolamine. BPII and [Asp49] PLA2 were individually dissolved in HBS (10 mM HEPES buffer (pH 7.4) containing 0.15 M NaCl, 3 mM EDTA, and 0.005% (v/v) surfactant P-20) at a concentration of 200 μg/ml. Then, 2–16 serial solutions were injected onto the TLf1 or TLf2-immobilized surfaces and unmodified sensor chip CM5 (control) at a flow rate of 20 μL/min. The sensor chip surface was regenerated with 10 mM HCl (40 μL) at the end of each measurement. Sensor-grams were analyzed using nonlinear least-squares curve fitting with BIAevaluation software (Amersham Biosciences).

Interactions between TLfs (TLf1 and TLf2) and BPII were estimated by the protein-protein docking simulation using the HDOCK server that based on hybrid docking strategies, template-based modeling, and ab initio docking (Yan et al., 2020). The 3D structures of TLf1 and TLf2 were predicted by homology modeling based on the complex 3D structure of trypsin and its inhibitor (PDB 1TX6). Predicted 3D structures were used as receptor molecules, while [Lys49] PLA2 (BPII, PDB 6AL3) (Matsui et al., 2019) was used as a ligand molecule. The top 10 predicted models with the lower docking energy scores were obtained for each interaction.

Statistical analysis was carried out using Student’s t-test at a 99% confidence level.

The pathological diagnosis and autopsy in mice including anti-myonecrosis activities of two types of habu antivenoms were determined against habu crude venom (Table 1). Newly prepared anti-venom (lot no. 1469) showed anti-lethal activity (anti-LA) and anti-HA of immunized serum with 676 units/ml (for anti-LA), 550 units/ml (for anti-HAI (HR1a/HR1b)), and 233 units/ml (for anti-HAII (HR2a/HR2b)), respectively (unpublished data). Although both anti-venoms had anti-hemorrhage activities (>300 units/ml), they showed very different effect on myonecrosis caused by [Lys49]PLA2 myotoxins. Figure 1 presents histochemical evidence supporting the effectiveness of the anti-myonecrotic antivenom in mouse skeletal muscle treated with habu venom, in which lot no. 1469 (with anti-myonecrotic activity) prevented the necrosis of mouse fascia lata tissues, while lot no. 34 (without anti-myonecrotic activity) exerted no protective effects. First, the antibody titration of two different antivenoms, lot no. 1469 with anti-myonecrotic activity and lot no. 34 without anti-myonecrotic activity, were assessed by using ELISA against BPII, one of the [Lys49]PLA2 myotoxins in habu venom, indicated that the antivenom (lot no. 1469) with anti-myonecrotic activity showed the higher (about ten times) sensitivity against BPII compared with conventional antivenom, lot no. 34, although both antivenoms showed the similar reactivity against [Asp49]PLA2 (Supplementary Figure S6). Next, to elucidate the different properties of the two antivenoms, including their target venom proteins, comparative proteomics analysis of habu venom proteins was conducted using antivenom IgG-conjugated affinity columns. 2D-PAGE analysis demonstrated similar profiles for both venom proteins purified by antivenom-columns in both adsorbed and flow through fractions (Figure 2). Both antivenom-specific-conjugated columns bound necrotic [Lys49] PLA2 at similar level (indicated by arrows labelled 2 in Figure 2A), of which partial amino acid sequences were determined by MALDI-TOF MS/MS analysis as shown in Supplementary Figure S7. However, some distinct differences were detected in the adsorbed fraction of the antivenom with anti-necrotic activity (lot no. 1469), while disappeared in the flow-through fraction, including a ∼30 kDa protein with a basic pI ranging from 9 to 10 (Figure 2). Interestingly, this protein was detected in the flow-through fraction of the antivenom without anti-myonecrotic activity (lot no. 34) (Figure 2B), and confirmed by reproducibility experiment using different 2D-PAGE condition (pI 3–11, 15% gel) (Supplementary Figure S8). These results indicated that the basic ∼30 kDa protein showed specific binding to the antivenom that prevented necrosis (lot no. 1469).

To investigate the venom proteins in more detail, we separated the basic 30 kDa protein by combined cation-exchange chromatography on a CM-52 column (Supplementary Figure S2) and further purified the protein by affinity chromatography (Supplementary Figure S9A). The SDS-PAGE profiles of the fractions from the two antivenom-specific-conjugated columns (Supplementary Figure S9B) indicated the 30 kDa protein more effectively bound to anti-myonecrotic antiserum than that without anti-myonecrotic activity.

The N-terminal sequence analysis of the purified basic 30 kDa protein was determined to be IIGGDEXNINEHRFLVALYXF, and MALDI TOF-MS/MS analysis with V8 protease and trypsin digestion, respectively, covered 47.9% of the full-length amino acid sequence of the 30 kDa protein (Supplementary Table S2 and Supplementary Figure S10). These results indicated the purified 30 kDa protein was identical to the basic serine protease homolog, TLf2, of which the amino acid sequence was previously determined from cDNA cloning (Deshimaru et al., 1996). The amino acid sequence of TLf2 suggested it to be an inactive form of serine protease with a mutation at one of the catalytic triad residues, His43, by arginine (Figure 3).

To confirm whether TLf2 was inactive, the hydrolytic activity of TLf2 was assayed using fibrinogen and a synthetic fluorogenic substrate, and compared with that of human α-thrombin. The SDS-PAGE profiles of fibrinogen treated with/without TLf2 or human α-thrombin indicated that TLf2 displayed no catalytic activity against fibrinogen after incubation for 24 h (Figure 4A). Additionally, TLf2 displayed no enzymatic activity against the synthetic fluorogenic substrate (Figure 4B). Thus, replacement of the catalytic triad amino acid, His43 in TLf2 with Arg rendered the serine protease inactive.

TLf2 was shown to bind specifically to anti-myonecrotic antivenom, which efficiently suppressed the myonecrosis of skeletal muscle caused by habu crude venom (Figure 1). This result suggested that TLf2 possessed necrotic activity by itself and/or affected necrosis factors, such as [Lys49] PLA2 (BPII). Therefore, we aimed to confirm the cytotoxicity of TLf2 on SW839 and SkMC cells; however, TLf2 exerted no toxicity on either cell line (Figure 5). Next, we analyzed the effect of TLf2 on necrosis factor [Lys49] PLA2 (BPII). Interestingly, TLf2 drastically decreased the survival rate of BPII-treated cells compared to those treated with BPII alone (Figures 5A,B). The number of viable cells treated with both TLf2 and BPII was significantly decreased by ∼50 and 25% in SW839 and SkMC cells (p-values <0.01), respectively, compared with cells treated with BPII alone (Figures 5A,B). These results indicated that TLf2 exerted a synergistic enhancing effect on myonecrotic [Lys49] PLA2, representing a novel function. TLf1 (flavoxovin), a major thrombin-like serine protease in habu snake venom, was further assessed for its ability to enhance [Lys49] PLA2 myotoxicity and elucidate whether the myonecrosis-enhancing activity was unique to TLf2 or common to other thrombin-like serine proteases. As shown in Figure 5C, TLf1 alone had no effect on cells, but it exerted a weak synergistic effect on myonecrosis caused by BPII compared with TLf2. These results indicated that the synergistic pharmacological activity of TLf2 was unique.

Further, synergistic cytotoxic effects of TLf2/TLf1 against BPII were provided a preliminary in vivo assessment using mice. Although both TLf2 and TLf1 showed the synergistic effects compared to the treating with BPII alone, noticeable difference between TLf2 and TLf1 could not be observed (Supplementary Figure S11).

To elucidate the mechanisms underlying the synergistic effect of TLf2 on myonecrotic [Lys49] PLA2, direct interactions between TLfs and PLA2s were assessed via SPR. Figure 6 illustrates typical sensor grams of SPR for TLf2 and TLf1 with BPII or [Asp49] PLA2. TLf2 interacted with BPII at a nanomolar level (K _d = 9.79 × 10⁻⁸ M), while TLf1 desplayed weaker binding affinity (K _d = 1.52 × 10⁻⁷ M). Further, TLf2 and TLf1 displayed much weaker binding affinities for [Asp49] PLA2 than for BPII (Figure 6, Supplementary Table S3). These results indicated that the binding properties of the TLfs for BPII were correlated with their myonecrosis-enhancing activities. Furthermore, interactions between TLfs and BPII were assessed by simulating protein-protein docking in silico using HDOCK server. Figure 7 shows the typical predicted interaction mode between BPII and TLf2 (Figure 7A) or TLf1 (Figure 7B), demonstrating that both interactions predicted similar complex structures. However, TLf2 and TLf1 differed significantly with respect to their docking energy scores and their top 10 prediction models (Supplementary Figure S7). The docking energy scores for the top 10 models of the TLf2-BPII complex (average value 220.33) were more stable than those for the TLf1-BPII complex (average value 199.10). The top 10 models of the TLf2-BPII complex predicted the same sites of interaction between TLf2 and BPII (Supplementary Figure S7B), while those of the TLf1-BPII complex predicted two interaction sites that located on the opposite side (Supplementary Figure S7A). Further, structural model of complex of TLf2 and BPII indicated the obvious emphasized basic properties composed of positively charged amino acids, while TLf1-BPII showed the different electric charged properties compared with TLf2-BPII (Figure 7C). Thus, these in silico predictions explained the difference in binding affinities for TLf1 and TLf2, and the different properties such as surface charged distribution of complexes.