Comprehensive Study of the Proteome and Transcriptome of the Venom of the Most Venomous European Viper: Discovery of a New Subclass of Ancestral Snake Venom Metalloproteinase Precursor-Derived Proteins

The nose-horned viper, its nominotypical subspecies Vipera ammodytes ammodytes (Vaa), in particular, is, medically, one of the most relevant snakes in Europe. The local and systemic clinical manifestations of poisoning by the venom of this snake are the result of the pathophysiological effects inflicted by enzymatic and nonenzymatic venom components acting, most prominently, on the blood, cardiovascular, and nerve systems. This venom is a very complex mixture of pharmacologically active proteins and peptides. To help improve the current antivenom therapy toward higher specificity and efficiency and to assist drug discovery, we have constructed, by combining transcriptomic and proteomic analyses, the most comprehensive library yet of the Vaa venom proteins and peptides. Sequence analysis of the venom gland cDNA library has revealed the presence of messages encoding 12 types of polypeptide precursors. The most abundant are those for metalloproteinase inhibitors (MPis), bradykinin-potentiating peptides (BPPs), and natriuretic peptides (NPs) (all three on a single precursor), snake C-type lectin-like proteins (snaclecs), serine proteases (SVSPs), P-II and P-III metalloproteinases (SVMPs), secreted phospholipases A2 (sPLA2s), and disintegrins (Dis). These constitute >88% of the venom transcriptome. At the protein level, 57 venom proteins belonging to 16 different protein families have been identified and, with SVSPs, sPLA2s, snaclecs, and SVMPs, comprise ∼80% of all venom proteins. Peptides detected in the venom include NPs, BPPs, and inhibitors of SVSPs and SVMPs. Of particular interest, a transcript coding for a protein similar to P-III SVMPs but lacking the MP domain was also found at the protein level in the venom. The existence of such proteins, also supported by finding similar venom gland transcripts in related snake species, has been demonstrated for the first time, justifying the proposal of a new P-IIIe subclass of ancestral SVMP precursor-derived proteins.


INTRODUCTION
Snake venoms are highly complex cocktails mainly comprising proteins and peptides involved in the immobilization and initial digestion of prey. They are also a rich source of bioactive compounds exploited by humans for the diagnosis and therapy of a variety of diseases. 1 Venom toxins have evolved from closely related body proteins that have been diversified functionally by gene duplication and adaptive evolution, generating multigene families specific for venom glands. 2 The ancestral genes were recruited from various types of tissue and usually code for the key secreted proteins involved in diverse biological processes. 3 Some time ago, it was reported by our laboratory that animal toxin multigene families have evolved under a strong positive selection that favors amino acid replacements serving to adapt the duplicated gene to a new function. 4 Venomous snakes are found in different snake families, especially those whose venom apparatus is highly developed, such as Elapidae and Viperidae. The latter, vipers, constitute a monophyletic lineage of venomous snakes comprising approximately 330 species distributed worldwide and currently divided into three subfamilies, Azemiopinae, Crotalinae, and Viperinae. 5 The nose-horned viper, Vipera ammodytes, is the most venomous snake in Europe. It is found mainly in southern Europe and partly in western Asia. Spreading from the northwest to the southeast, at least four subspecies, ammodytes (Vaa), meridionalis (Vam), montandoni, and transcaucasiana, are usually recognized. 6 V. ammodytes venom induces mainly hemotoxic and neurotoxic effects, which, in rare cases, can lead to human death. 7,8 In contrast with that from other subspecies, Vaa venom contains highly neurotoxic monomeric secreted phospholipases A 2 (sPLA 2 s), known as ammodytoxins (Atxs). 9 A comparative analysis of the Vaa and Vam proteomes revealed the presence of 38 venom components in the former. 10 Recently, we studied the proteome of the common European adder, subspecies Vipera berus berus (Vbb), and compared it with that of Vaa. 11 The Vbb proteome was shown to be much less complex than that of Vaa, in particular, possessing smaller amounts of snaclecs (snake C-type lectinlike proteins) and sPLA 2 s. The Vaa venom is rich in compounds that interfere with hemostasis, 12,13 with some that are potentially anti-tumor-active. 14,15 The main aim of the present comprehensive transcriptomic and proteomic study was to identify and build a complete library of Vaa venom proteins and peptides. The accumulated data will direct the production of a more specific and effective antivenom with which to treat venomous Vaa bites. Such antivenoms can be, namely, produced by injecting horses with a mixture of antigens stemming from the most critical toxic components of the venom only. It will also facilitate structurebased drug design, especially for the treatment of certain neurological, cardiovascular, and cancer disorders.

Venom and Reagents
Vaa venom, collected in 2005 from snakes from different parts of Croatia, was a gift from the Institute of Immunology, Zagreb, Croatia. Fibrinogen was from Hypen BioMed (France). Acetonitrile (ACN; Merck, Germany), trifluoroacetic acid (TFA; from Sigma-Aldrich, USA), and formic acid (Fluka, Germany) were of HPLC gradient grade or higher. Deionized water was purified using a Direct-Q 5 system (Millipore, Billerica, MA).

Analysis and Sequencing of cDNA
cDNAs encoding venom proteins were obtained by random screening of a representative plasmid cDNA library. Sequences encoding the complete protein-coding regions of Vaa venom gland transcripts were determined by using internal sequencing primers deduced from previously sequenced regions. The library was recently prepared from venom glands isolated 2 days after milking from a single Vaa specimen captured in the wild in the area of northeastern Slovenia. 14 The nucleotide sequences were determined by Microsynth AG (Switzerland) using the dideoxy chain-termination method. They were subsequently analyzed by free, publicly available, bioinformatics services. They were submitted to GenBank under the accession numbers KU249650−KU249656, KT148817− KT148834, and MG958491−MG958504.

Two-Dimensional Gel Electrophoresis
Two-dimensional gel electrophoresis  was performed under optimized conditions. 16 500 μg of crude Vaa venom was dissolved in 450 μL of rehydration buffer containing 7 M urea, 2 M thiourea, 30 mM Tris, 1% (v/v) ampholytes, 0.25% (m/v) ASB-14, 2.5% (m/v) CHAPS, 0.002% (m/v) bromophenol blue, and 12 μL/mL DeStreak reagent (GE Healthcare, Amersham Biosciences). A 24 cm immobilized pH gradient (IPG) strip (GE Healthcare, Amersham Biosciences), covering the pH range 3−11 NL, was rehydrated passively with the sample overnight. The first dimension separation (isoelectric focusing (IEF)) and the second dimension separation (polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate (SDS-PAGE)) were carried out using the reported experimental protocols. 16 Following reverse staining with imidazole-SDS-Zn 2+ , the gel was scanned by an Image Scanner using LabScan 5 software (GE Healthcare, Amersham Biosciences). The image was analyzed by Image Master 2D Platinum 6.0 software (GE Healthcare, Amersham Biosciences). The protein spots detected were cut out automatically using an Ettan Spot Picker (GE Healthcare, Amersham Biosciences) and kept at −20°C before analysis.

Protein Identification by Mass Spectrometry
Protein spots were destained and treated with trypsin in-gel, and the resulting peptides were analyzed using an ion trap mass spectrometer 1200 series HPLC-Chip-LC/MSD Trap XCT Ultra (Agilent Technologies, Waldbronn, Germany). 16 Spectral data were exported as Mascot generic format (mgf) files using in-house Agilent Technologies software, Data Analysis for 6300 series Ion Trap LC−MS version 3.4 (Build 175). A search against the nonredundant National Center for Biotechnology Information (NCBI) Snakes database (taxid 8750, December 2017, 159 187 entries) supplemented with our Vaa transcriptome data deposited in the GenBank NCBI database was performed using a licensed version 2 of the MASCOT program, applying the following restrictions: 2+ and 3+ peptide charge; two miscleavages allowed; peptide and fragment mass tolerance of ±1.2 and ±0.6 Da, respectively; carbamidomethyl Cys (C) as the fixed modification and oxidized methionine (M ox ) as variable; and an automatic decoy database search. The results were further validated using Scaffold software (version 2, Proteome Software, USA) with the following thresholds: protein confidence of 99% and one peptide per protein at 95% confidence. Proteins were identified at 0.1% Prophet false discovery rate (FDR), and peptides were identified at 5.27% Prophet FDR. Data are available via ProteomeXchange with the identifier PXD012752.
Low-molecular-mass peptides isolated by RP-HPLC were analyzed using a Q-TOF Premier mass spectrometer (Waters-Micromass, GB) as described. 18 activity of the Vaa snake venom metalloproteinases (SVMPs) present in the gel filtration fraction A. The vacuum-dried fraction RP-C18 containing the tripeptide was dissolved in 5 μL (0.5 μg) of fraction A supplemented with 2 mM serine protease inhibitor Pefabloc (Sigma-Aldrich, USA) and incubated for 30 min at 37°C. Five μL (30 μg) of fibrinogen in 20 mM Tris, 50 mM NaCl, 2 mM CaCl 2 , pH 7.0 was then added. The reaction was stopped after 15 min by adding 10 μL of reducing SDS-PAGE buffer. The reaction mixture was heated for 5 min at 95°C and analyzed by 12.5% SDS-PAGE. Proteins were stained with Coomassie Brilliant Blue R250.

Transcriptomic Analysis
Of the 520 randomly selected cDNA clones, 254 (48.8%) coded for precursors of Vaa venom proteins and peptides (toxic and nontoxic), whereas the remaining recombinant plasmids harbored either nonvenomous or other sequences encoding unidentified proteins. For example, those encoding phospholipase B (PLB; GenBank accession number MG958504) and leucine aminopeptidase (incomplete cDNA) were excluded from the venom-related transcripts secreted by a pair of Vaa venom glands. A search for the presence of a potential signal peptide in the deduced amino acid sequences of precursors of these two proteins and of closely similar proteins in databases did not support the assumption that these proteins are actually secreted by venom glands. Interestingly, PLB was detected at the protein level in the Vaa venom (see below). This enzyme was also reported in venoms of other snakes, either viperids, such as Pelias species, 19 or elapids, such as Pseudechis guttatus. 20 The venom-related transcriptome thus includes 254 partial and complete sequences of 45 different mRNA transcripts that are branched into 12 different groups (Table S-1). Despite the relatively small number of analyzed cDNAs, the comparison of these groups with those observed in the proteomic analysis (see below) suggests that this result is an indicative snapshot of the biosynthesis of venom proteins and of their relative distribution in Vaa venom glands. The most abundant were transcripts encoding common precursors of tripeptide inhibitors of MPs (MPis), bradykinin-potentiating peptides (BPPs), and natriuretic peptides (NPs) (25.6% of all venom transcripts), followed by those of snaclecs (13.8%), SVSPs (11.8%), P-III class SVMPs (11.0%), sPLA 2 s (10.6%), P-II class SVMPs (9.4%), and disintegrins (Dis; 5.9%) (Figure 1). These seven major groups comprise >88% of all mRNAs isolated from the Vaa venom glands. Each of the remaining five groupsSP inhibitors (SPis), vascular endothelial growth factors (VEGFs), Cys-rich secretory proteins (CRISPs), Lamino acid oxidases (LAAOs), and venom nerve growth factors (VNGFs)constitutes <5% of the transcriptome.
Interestingly, the five most abundant venom-related mRNA transcripts were also the most heterogeneous. Snaclec precursors were thus represented by nine, SVSPs by eight, P-III SVMPs by six, common MPi, BPP, and NP peptide precursors by five, and sPLA 2 s by three different mRNAs. P-II and P-III class SVMPs together form the largest group of Vaa venom-gland-encoded enzymes, comprising more than one-fifth (20.4%) of the transcriptome. A large proportion of the SVMP transcripts, ranging from 24 to 58%, has also been observed in the venom gland transcriptome of most other Viperidae species reported so far, for example, Crotalus adamanteus, 21 Protobothrops f lavoviridis, 22 Bothrops colombiensis, 23 and Echis ocellatus. 24 All of these evolved from an ancient ADAM (a disintegrin and a metalloproteinase) gene that was recruited into the venom gland of snakes and are responsible for the wide spectrum of severe local and cardiovascular pathologies observed in victims of viper envenomation. 25 3.1.1. mRNA Transcripts Encoding Precursors of MPi, BPPs, and NPs. The largest portion of Vaa mRNA transcripts, that is, about one-quarter (see above), contained information for the precursors of biologically active peptidesMPis, BPPs, and NPssimilar to those previously found in viperid snakes. 26 Six different mRNAs were recognized that can be divided into two groups, whose leading representatives we termed Vaa-MPi-1 and Vaa-MPi-2 ( Figure 2). Vaa-MPi-1 encodes a precursor protein of 180 amino acid residues, and Vaa-MPi-2 encodes a precursor of 244 residues. These two share 64% amino acid identity and possess the identical sequence of a putative signal peptide of 23 residues ( Figure  2A). Interestingly, the same signal peptide sequence has also been observed in three MPi polypeptide precursors from the Viperinae snake E. ocellatus. 27 Highly similar nucleotide and deduced amino acid sequences were also found in a genome database with whole genome shotgun data of Vbb (Viperinae) and Protobothrops mucrosquamatus (Crotalinae), enabling the presumed (first) intron position within the Vaa-MPi-2 nucleotide sequence to be deduced.
The sequences of Vaa-MPi-1 and Vaa-MPi-1′ differ in only one amino acid residue (resulting from only one nucleotide residue) at the C-terminal end, probably representing two allelic forms. In contrast, the shorter transcripts Vaa-MPi-5 and Vaa-MPi-3, with deletions of 43 and 36 amino acid residues respectively ( Figure 2B,C), could be the result of alternative splicing. In transcript Vaa-MPi-4, displaying a deletion of 106 nucleotides relative to Vaa-MPi-2, an open-reading frameshift occurs that results in a premature ending of the polypeptide chain, thus lacking the C-terminal NP sequence ( Figure 2C). This may also be due to alternative splicing. Another possibility is that Vaa-MPi-3, Vaa-MPi-4, and Vaa-MPi-5 mRNAs were transcribed from recently duplicated copies of the Vaa-MPi-2 gene. Figure 1. Relative distribution of protein groups in the transcriptome of Vaa venom glands. Percentages were calculated according to the total number, that is, 254, of venom-related transcripts. Abbreviations: PVP, the common precursor of venom peptides (NPs, BPPs, and MPis); snaclec, snake C-type lectin-like protein; SP, serine protease; MP, metalloproteinase (class II or III); sPLA 2 , secreted phospholipase A 2 ; Dis, disintegrin; SPi, serine protease inhibitor (Kunitz-type); VEGF, vascular endothelial growth factor; CRISP, Cys-rich secretory protein; LAAO, L-amino acid oxidase; VNGF, venom nerve growth factor.

Journal of Proteome Research
Article 3.1.2. Transcripts Coding for a New Ancestral SVMP Precursor-Derived Protein. Twenty-eight transcripts coding for precursors of P-III class SVMP proteins were grouped into six groups (encoded by full-length transcripts) corresponding to six different pre-pro-proteins. Two of them correspond to two previously identified and characterized Vaa hemorrhagic MPs, subunit A of heterodimeric VaH4 and homodimeric VaH3, 14,28 whose cDNAs were isolated by initial random screening, followed by PCR. The remainder encode new, previously unknown Vaa P-III SMVP proteins of a high degree of amino acid sequence identity, which were named Vaa-MPIII-2, Vaa-MPIII-3, Vaa-MPIII-4, and Vaa-MPIII-5. Unlike the others, Vaa-MPIII-3 exhibits a large deletion of 284 amino acid residues in the middle part ( Figure S-1). In the present transcriptomic analysis, no transcripts corresponding to the two previously identified P-III class SVMPs, VaF1 29 and Vaa-MPIII-1, 11 were found, but those two were then obtained by PCR amplification.
Interestingly, the protein-coding sequences of VaH3 and Vaa-MPIII-5 cDNAs of 1851 nt share a high level of nucleotide identity (95.2%). They differ only in their 579 nt pre-pro-regions (28 nt differences leading to 14 aa replacements), whereas their mature protein-coding regions of 1272 nt show 100% nucleotide and amino acid identity ( Figure S-1). These figures may reflect a recent duplication event in the evolution of their genes, opening up the possibility of a fine-tuning of their processing. A similar observation was also noted in the case of two snaclec precursors in which Vaa-snaclec-5 and Vaasnaclec-6 differ only in the signal peptide region, their mature protein regions being identical (see Figure S-7).
Four transcripts, of a total of 28 encoding P-III class SVMPs, coded for a precursor protein of 324 amino acids, designated as Vaa-MPIII-3. The Vaa-MPIII-3 mRNA encodes a mature protein without the MP domain, possessing only the Cterminal part of the Dis-like (D) domain, with a D-loop (an XXCD, i.e., an RGD-like motif), termed here the D′ domain, Figure 3. Alignment of the deduced precursor sequence of Vaa-MPIII-3 lacking the MP domain with that of the VaH4-A subunit of heterodimeric hemorrhagin VaH4 from the same venom. Important residues and motifs are highlighted in gray. Identical, conserved, and semiconserved amino acid residues are designated by asterisks, colons, and dots, respectively. Putative signal peptides are underlined once, and the N-terminus of the mature Vaa-MPIII-3 is shown by an arrow pointing right. The deletion of 284 amino acid residues in Vaa-MPIII-3, including the last part of the pro-peptide with the inhibitory Cys-switch motif, the entire MP domain, and the first part of the D domain, is indicated by dashes. The phase 0 intron is located between two codons, whereas the phase 1 intron separates codons between the first and the second nucleotides. Part of the Vaa-MPIII-3 sequence (64%), covered by Edman and MS sequencing, is underlined by a double line. In contrast with Vaa-MPIII-3, the sequence of VaH4-A includes a canonical zinc-binding active site motif, followed by a methionine turn characteristic of the metzincin superfamily of catalytically active MPs. and the Cys-rich domain (C domain). The deduced pre-proprotein sequence of Vaa-MPIII-3 is shown compared with that of VaH4-A 14 in Figure 3. The existence of Vaa-MPIII-3 mature protein in Vaa venom was confirmed by its isolation from the venom and sequencing by Edman degradation and MS (see below, Figure 6B, Table 2). The N-terminal amino acid sequence of this protein was determined to be RAGTECRPA-RSE. The Vaa-MPIII-3 mature protein of 151 amino acid residues is thus presumably preceded by a signal peptide of 20 residues and a pro-peptide of 153 residues.
Loss of the MP domain has already been observed in one of the P-III class SVMP precursors from Echis carinatus sochureki venom (GenBank No. GU012129). 30 However, in contrast with the case presented here, the deleted region was much shorter (196 vs 284 amino acid residues), resulting in a longer DC domain of the deduced mature protein. The existence of such a protein in Echis was, however, not confirmed on the protein level. These results showed that the evolutionary history of viperid SVMPs was punctuated repeatedly by domain loss, resulting in frequent alterations to the molecular scaffold. Furthermore, catalytically inactive P-II class SVMPs, with substitutions in the canonical zinc-binding motif, have also been isolated from Bothriechis lateralis venom and characterized, indicating that enzymatically inactive SVMP homologues deserve further investigation of their toxicity role in snake venoms. 31 A BLASTp search through the nonredundant NCBI database, using the 324 residue pre-pro-Vaa-MPIII-3 as a query sequence, revealed Eoc89, a translated cDNA sequence of GenBank No. AM039699 from Echis ocellatus venom glands, 32 to be one of the most closely similar structures, with ∼82% amino acid identity in the overlapping N-and Cterminal parts. The first draft of the genomic organization of a PIII-SVMP gene, that of the Eoc89-like protein, has been reported. 33 Its gene consists of 12 exons separated by 11 introns. Notably, the presumed position of intron 11 in the Vaa-MPIII-3 sequence is very close to the N-terminus of the mature protein (Figure 3), indicating that this intron could play a significant role in the evolution of the Vaa-MPIII-3 gene, in which the catalytic MP domain and subsequent first part of the D domain had been lost. The nucleotide sequence of Vaa-MPIII-3 was also used for a similarity search through the whole genome shotgun data. It appears that, at least in some of the similar gene sequences, such as those in the Vbb genome, the C-terminal domain sequence may be interrupted by additional introns, but this remains to be confirmed.
The Vaa-MPIII-3 transcript from Vaa, encoding the Nterminal signal peptide and pro-peptide, lacks the central MP domain and possesses, at its C-terminal end, a truncated D (D′) and the complete C domain, thus encoding a new type of P-III class SVMP-like proteins. We therefore suggest a new P-IIIe subclass of ancestral SVMP precursor-derived proteins. Precursors, such as those for Vaa-MPIII-3 and the P-III class SVMP from Echis carinatus sochureki venom (GenBank No. GU012129), encode mature proteins consisting of a partial (D′) or complete Dis-like domain (D), followed by a Cys-rich domain (i.e., D′C and DC proteins). The gene structure and evolution of Vaa-MPIII-3, its precise precursor processing, and actual function in the snake venom have yet to be elucidated. According to our proteome results ( Figure 6B), the mature Vaa-MPIII-3 protein, presumably possessing eight intramolecular disulfide bonds and a free Cys residue, exists in Vaa venom as a glycosylated monomer.

High-Molecular-Mass Proteome Profiling of the Vaa Venom
In the present study, the optimized 2-DE conditions 16 allowed resolution of crude Vaa venom into 208 distinct spots in the molecular mass range of 10 to 60 kDa ( Figure 4). Each spot was subjected to in-gel digestion and LC−ESI−MS/MS analysis. The MS spectra were searched against the nonredundant protein NCBI database of snake species, supplemented with the transcriptomic data obtained from our Vaa venom gland cDNA library analysis. Proteins were identified unambiguously in 176 spots ( Table 1, Table S-2). Members of different protein families were detected in certain spots. Of the . Two-dimensional gel electrophoresis of the Vaa venom. 500 μg of crude Vaa venom was separated with IEF on a 24 cm IPG strip, pH 3− 11 NL, in the first dimension. Proteins were then reduced and alkylated and separated on a 10% SDS-PAGE gel in the perpendicular dimension, according to their molecular masses, using the Tris/Taurin buffer system. The gel was stained using the imidazole-SDS-Zn 2+ method.

Journal of Proteome Research
Article 32 spots in which proteins could not be recognized, some gave poor spectrometric data, probably due to insufficient protein levels, whereas others may just represent staining artifacts. 34 The identified proteins were assigned to 16 protein families, 7 of which are typical for viperid snake venomsSVSPs, sPLA 2 s, SVMPs, LAAOs, snaclecs, CRISPs, and Dis. 35 Figure 5 shows the protein family profile of the most abundant Vaa venom high-molecular-mass proteins (∼10−60 kDa) according to the number of spots in which members of a particular family have been found. In accordance with the high hematotoxic potential of Vaa venom, 12 members of SVSP, snaclec, sPLA 2 , and SVMP protein families were found in most of the spots. In contrast, in a proteomic study of Vaa venom from Bulgaria only 139 protein spots were reported on a 2-DE gel. 10 Of these, only 38 venom components were identified, being assigned to 8 protein families. This may well be the consequence of a lower sequence identity of the Vaa venom proteins to those in protein data banks. Although the nonredundant NCBI database currently contains >150 000 protein sequences from different snake species, identification of proteins by MS is limited due to the high level of interspecies sequence variation within a particular protein family, as well as to the presence of diverse post-translational modifications. 36 For this reason, a combination of transcriptomics and proteomics was used here to obtain a more complete proteome profile of the venom. Transcriptomic data contributed to the high rate of protein identification as well as to the high sequence coverage of the identified proteins. Practically all of the proteins predicted from the transcriptome analysis were then confirmed by the venomics.
3.2.1. Protein Families in the Vaa Venom. All of the protein components described in the Vaa venom are discussed below, grouped in protein families, from the most abundant to those present only in minute amounts. We begin with enzymes and conclude with nonenzymatic venom proteins.
3.2.1.1. Serine Proteases. Despite being abundantly present in Vaa venom, representatives of the SVSP family have been poorly investigated. Two kallikrein-like enzymes with apparent molecular masses of 31.3 and 34.3 kDa were purified from Vaa venom in 1976. 37 Only recently, we reported on the 31.5 kDa fibrin(ogen)ase with the unconventional catalytic triad, VaSP1 that can also degrade prothrombin, FX, and plasminogen. 38 In the current study, cDNA sequences of 8 SVSPs (Figure S-2) were determined. With the exception of Vaa-SP-7, their presence or that of their structurally close relatives was Cysteine residues were carbamidomethylated before MS/MS analysis. Abbreviations: AP, aspartic protease; CRISP, cysteine-rich secretory protein; Dis, disintegrin; GSH, glutathione; LAAO, L-amino acid oxidase; MP, metalloproteinase; NGF, nerve growth factor; QC, glutaminyl cyclase; PL, phospholipase; SP, serine protease. Figure 5. High-molecular-mass proteome profile of the Vaa venom. The profile is showing the relative distribution of spots among the most frequently represented Vaa protein families, that is, SVSP, snake venom serine protease; snaclec, snake C-type lectin-like protein; sPLA 2 , secreted phospholipase A 2 ; SVMP, snake venom metalloproteinase; CRISP, Cys-rich secretory protein; LAAO, L-amino acid oxidase; NGF, nerve growth factor; QC, glutaminyl cyclase; and PLB, phospholipase B.

Journal of Proteome Research
Article confirmed in almost one-third (i.e., 53) of all of the proteincontaining 2-DE spots, mostly in the molecular mass range 30 to 45 kDa ( Figure 4, Table 1, Table S-2). The cDNA transcript of Vaa-SP-5 is incomplete, lacking ∼30 amino acids at the Cterminus. However, the known part of the molecule shows 96% sequence identity with Vaa-SP-3 ( Figure S-2). Of the 12 spots in which we identified peptides common to both proteins, in only three were the Vaa-SP-3-specific peptides identified.
Vaa-SP-2 is a basic protein while Vaa-SP-3, Vaa-SP-4, Vaa-SP-5, and Vaa-SP-8 are acidic and Vaa-SP-6 and Vaa-SPH-1 neutral proteins. As is usual for viperid SVSPs, 39 the Vaa-SPs also possess various numbers of consensus N-glycosylation sites in their sequences, and thus have the potential of becoming N-glycosylated. Four such sites have been found in Vaa-SP-2 and Vaa-SP-6, three in Vaa-SP-4 and Vaa-SP-8, and two in each of Vaa-SP-3, Vaa-SP-5, and Vaa-SPH-1. Nglycosylation could explain the multiple pIs and much higher apparent molecular masses of these proteins than would be expected from their primary structures. Glycosylation is known to affect the stability of SVSPs, their activity and their responses to protein inhibitors. 40,41 In spite of the wide range of substrate specificity, viperid SVSPs exhibit extensive sequence similarity. 39 Common structural characteristics of SVSPs−a C-terminal extension and 12 Cys residues that are assumed to form disulfides as in Trimeresurus stejnegeri venom plasminogen activator, TSV-PA, 42 are preserved in all full-length Vaa-SP transcripts ( Figure  S-2). The canonical active site catalytic triad, His−Asp−Ser, is preserved in all Vaa-SPs ( Figure S-2) except Vaa-SPH-1, in which it is replaced by Arg−Asp−Asn. As expected, Vaa-SPH-1 is devoid of proteolytic activity. 13 It is, however, a strong inhibitor of coagulation, acting as an antagonist of FIXa. As deduced from their respective cDNAs, the same active site replacements as in Vaa-SPH-1 are also present in SVSP homologues from Macrovipera lebetina and Bitis gabonica, 43,44 so these proteins are not expected to be enzymes as well. Vaa-SPs, other than Vaa-SPH-1, are proteases that hydrolyze fibrinogen and activate FIX and FX, but not prothrombin. 12 The highest sequence identity was observed between Vaa-SP-2/Vaa-SP-4 and fibrogenases from Macrovipera lebetina (75% both with VLAF, and 72 and 88% with VLBF) 45 and from Daboia russeli siamensis venom (71 and 74% with RVAF) 46 ( Figure S-2). Vaa-SP-3 and Vaa-SP-5 are, however, more similar to plasminogen activators (e.g., ≥ 73% identity with TSV-PA, Haly-PA, and LV-PA). 42,47,48 The very high sequence identity (95.7%) of Vaa-SP-7 to VLCTLP from M. lebetina, an angiotensin-cleaving enzyme and weak fibrinogenase with chymotrypsin-like activity, 49 was found. Similarly, Vaa-SP-6 differs in only four amino acid residues from nikobin (97.7% identity), an SVSP from V. berus nikolskii with an as yet unknown function (E5AJX2).
3.2.1.2. Phospholipases. Some years ago, we determined both protein and cDNA sequences of 5 Vaa venom sPLA 2 s (presynaptically neurotoxic ammodytoxins (Atxs) A, B, and C, nontoxic ammodytins (Atns) I1 and I2) and one enzymatically inactive myotoxic sPLA 2 homologue, AtnL. 9,50 These were also found in the present proteomic analysis ( Table 1, Table S-2). However, we could not discriminate between the Atx isoforms A and C in protein spots, since the peptides analyzed did not allow differentiation between these two proteins, whose sequences differ only in two amino acid residues at the Cterminal end.
In several 2-DE spots, instead of the expected AtnI2 sequence, 51 another AtnI2 (C) isoform 52 was detected (Figure S-5) that is usually the component of venoms of two other, eastern European V. ammodytes subspecies, Vam and V. a. montadoni. Because the Vaa venom analyzed was in fact a mixture of Vaa venom samples from different regions in Croatia, such a finding appears to be the consequence of a gene flow present in some of the viper specimens used for milking. Furthermore, besides AtnI1, found previously in Vaa, its (E) isoform was also detected in one of the 2-DE spots. This isoform has been identified so far only in another viper species, V. aspis aspis in southern France. 52 Indeed, peptides corresponding to the (E) isoform of AtnI1 were also found in a venom pool obtained from 8 Vaa specimens captured in the northwestern region of Bulgaria. 10 Again, this may also reflect a gene flow, in this case even between different viperid species.
Our bioinformatic analysis of the deduced Vaa PLB precursor sequence (MG958504) did not reveal the presence of an obvious signal peptide that would allow its secretion to the lumen of viper venom glands. However, in 2-DE spots 42,44,45,48 (Figure 4), peptides homologous to stretches of PLB were identified (Table 1, Table S-2). In fact, PLB activity was reported in snake venoms a long time ago. 53,54 Only recently, however, the first protein sequences of SV PLBs have been obtainedin two pit vipers, Protobothrops f lavoviridis and Ovophis okinavensis, 22 and in a colubrid Spilotes sulphureus snake. 55 3.2.1.3. Metalloproteinases. Vaa venom is rich in SVMPs, found here in 44 spots on the 2-DE gel (25%) (Figure 4, Table  1, Table S-2). Vaa SVMPs are approximately equally represented by P-II and P-III class SVMPs. The latter are well characterized and exhibit a wide array of biological activities, most affecting the hemostatic system. They are, for example, hemorrhagic and fibrino(geno)lytic, activating or degrading blood coagulation factors (FX, FIX, prothrombin) and inhibiting platelet aggregation. 12,14,17,28,29,56−58 These Vaa venom components belong to monomeric P-IIIa (VaH1, VaH2, VaF1, ammodytase), homo-or heterodimeric P-IIIc (VaH3, VaH4, ammodytagin) and the oligomeric P-IIId subclass of SVMPs (VAFXA-I and VAFXA-II). Additionally, new Vaa SVMPs, corresponding to transcripts Vaa-MPIII-1, Vaa-MPIII-2, and Vaa-MPIII-4 ( Figure S-3), but not to Vaa-MPIII-5, were detected at the protein level. Vaa-MPIII-1 and Vaa-MPIII-4 were identified in only one low intensity spot (spot 5), so it is not surprising that they have not yet been isolated and characterized. Their primary sequences show their high similarity to hemorrhagins from other viperid venoms ( Figure S-3). They lack Cys176, Cys132 or both, that is, the residues involved in dimerization, and therefore belong to the P-IIIa subclass of SVMPs. Judged from their position on the 2-DE gel, they are acidic proteins with an apparent molecular mass of ∼56 kDa. The discrepancy between their apparent and theoretical (∼46 kDa) molecular masses probably reflects Nglycosylation, since five potential N-glycosylation sites are present in Vaa-MPIII-1 and one in Vaa-MPIII-4. Furthermore, all other characterized Vaa P-III SVMPs are glycoproteins. Vaa-MPIII-2 (spots 6, 10, 14, 15, 22−24) is most probably a P-IIId subclass SVMP, since it exhibits a high degree of sequence identity with the partial sequence of the heavy chain of FX activator from Vaa, VAFXA-I, 58 as well as with those of P-IIId SVMPs from other snake venoms expressing the same activity, VLFXA (Q7T046) and RVV-X (Q7LZ61) (Figure S-3). Peptides arising from the MP domain of VaH4-A were also

Journal of Proteome Research
Article identified in the 2-DE spots, with molecular masses of ∼30 kDa, indicating that this SVMP is processed, increasing the structural and functional complexity of the venom. Namely, some SVMPs (P-IIIb subclass) undergo autolysis at the D domain, releasing the C-terminal DC part (DC domains), retaining their platelet binding capability, and acting as platelet aggregation inhibitors. 59 However, the presence of the DC domain of VaH4-A in the venom was not confirmed.
Vaa-MPIII-3, encoded by the unique mRNA that lacks a part coding for the MP domain, was also identified at the protein level ( Figure 6B, Table 2). It is presumably a glycoprotein with an apparent molecular mass of 21 kDa. The confirmation of the existence of such an SVMP-related protein in the venom as well led us to propose the introduction of a novel P-III subclass SVMP, a subclass P-IIIe. The function of such D′C domain proteins in the venom may be similar to that of the DC domain products of the post-translational processing of P-IIIb SVMPs, that is inhibition of platelet aggregation. 60−62 The potential platelet-binding capability of a D′C domain protein is probably related to an RGD-like motif in its D-loop.
All three Vaa-MPIIs are proteolytically processed to the P-I SVMP and Dis, as their MP domains have been found in numerous spots (93−95 and 98−110). These spots are distributed over a broad range of pIs in the region of molecular masses 26−28 kDa. Their masses are higher than those predicted theoretically. It is most probable that Vaa-MPIIs are N-glycosylated, as indicated by the existence of a potential N-glycosylation site at position 283 in the sequence of all three Vaa-MPIIs (Figure S-4) and an additional one at position 375 in the case of Vaa-MPII-3. All three Vaa-MPIIs show the highest sequence identity (89%) with fibrinogenolytic P-II SVMP, Le-3, from M. lebetina, which also undergoes processing to the MP-and Dis domain in the venom, 63 and to MPII precursors from Echis snake venoms (Figure S-4). A specific Dis domain peptide arising from Vaa-MPII-1 was detected in spot 206, together with two peptides from Vaa-Dis-2, suggesting that these two Dis form a disulfide linked dimer that was not completely reduced before the second-dimension SDS-PAGE. Because monomeric Dis have molecular masses <10 kDa, they should migrate with the electrophoretic front on a 2-DE gel, so they could not be spotted in this way. We analyzed them using a combination of liquid chromatography techniques ( Figure 6; Table 2; Disintegrins section).
3.2.1.4. L-Amino Acid Oxidases. SV LAAOs are dimeric FAD-or FMN-binding enzymes giving venoms a characteristic yellowish color. 64 In the 2-DE gel, we identified Vaa venom LAAOs in six ∼55 kDa (9, 10, 15, 22−24) and two ∼43 kDa spots (32 and 38) ( Figure 4, Table 1, Table S-2). Our cDNA library analysis revealed the presence of an LAAO precursor Vaa-LAAO-II (a 504 amino acid pre-pro-protein) that shares 92% amino acid identity with the mature form of Vaa-LAAO-I 65 (Figure S-6). Whereas Vaa-LAAO-I was the major LAAO isoform in the Bulgarian Vaa venom, Vaa-LAAO-II was in the majority in the Croatian venom that we analyzed. In addition to Vaa-LAAO-II, another LAAO isoform was identified in 2-DE spots 10 and 22 ( Figure 4, Table 1, Table S-2), which is like Ehis coloratus (JAC96580). LAAOs are present in viperid venoms in different quantities, being a minor component, as in Vaa and Vam venoms, 10 or a major one, as in Crotalus rhodostoma venom. 66 In the latter case, LAAOs comprise onethird of the venom protein content. The pathophysiological effects of LAAOs, involving induction or inhibition of platelet aggregation, induction of apoptosis, hemolysis, hemorrhage, and edema, depend mainly on their production of hydrogen peroxide. 64,67 3.2.1.5. Glutaminyl Cyclases. Glutaminyl cyclases (QCs) were detected in five 2-DE spots (39−41, 44, 135) (Figure 4, Table 1, Table S-2) with peptide sequences matching QCs from D. russelii (AFE84762) and Crotalus atrox (AFE84758) venoms. The Vaa venom QC is a ∼40 kDa protein that is most probably glycosylated, as is the QC from C. atrox. The primary structures of SVQCs, including two N-glycosylation sites, are highly conserved. 68 The ∼22 kDa protein with QC sequence in 2-DE spot 135 is probably a product of the proteolytic degradation of full-length QC. Although present in venoms in minute amounts, QCs have been found to be important in the post-translational modification of some venom proteins and peptides, for example, SVMPs and their tripeptide inhibitors, BPPs, and three-finger toxins. 27,68,69 They catalyze the formation of the N-terminal pyroglutamate residue in proteins and peptides, protecting them from degradation by exopeptidases.

Journal of Proteome Research
Article GSH peroxidase is an antioxidant enzyme that catalyzes the reduction of hydrogen peroxide to water by reduced glutathione. It was found in Vaa venom by 2-DE in two basic pI spots, 125 and 130 ( Figure 4, Table 1, Table S-2). As a minor component, GSH peroxidase has been reported in venoms of only a few other snakes. 70,71 Its possible role in the venom is to protect lipids and proteins against oxidative damage by hydrogen peroxide.
Renin-like aspartic protease was found in only one 2-DE spot (spot 47) (Figure 4, Table 1, Table S-2). Thus far, such a protease has been identified as a minor venom component of various Russian vipers 11,19 and the Indian saw-scaled viper, Echis c. carinatus. 72 The latter protease was recently purified from the venom, and its renin-like activity was confirmed. 73 Renin is a mammalian aspartic protease catalyzing the first step of the renin−angiotensin pathway in which angiotensinogen is processed to angiotensin I. This is then cleaved by angiotensinconverting enzyme to angiotensin II, a vasoconstrictor. By exerting renin-like activity, SV aspartic proteases can induce hypertensive effects, local or systemic, as was reported in the case of Vbb envenomation. 74,75 In accord with the negligible quantity of the enzyme in the Vaa venom, no such effects have so far been reported following a Vaa venomous bite.
The Vaa venom 5′-nucleotidase was identified by two peptides identical to peptides from a 55 kDa 5′-nucleotidase (BAG82602) from Gloydius blomhoff i brevicaudus (Table S-2). However, spot 139 ( Figure 4) harboring these two peptides was located at ∼20 kDa on the 2-DE gel, which suggests that the Vaa enzyme had undergone proteolytic cleavage. 5′-Nucleotidases are ubiquitous in SVs, although usually, as in the case of Vaa venom, in very small quantities. 76 They cleave 5′nucleotides to liberate adenosine, which then induces various pharmacological effects, such as vasodilation or inhibition of platelet aggregation, in this way potentiating the overall venom toxicity.
3.2.1.7. Snaclecs. Snaclecs are the largest nonenzymatic group of proteins in the Vaa venom. They are found in almost one-third (51 spots) of all identified 2-DE spots ( Figure 4, Table 1, Table S-2). The snaclec family of venom proteins comprises C-type lectin-like proteins, which do not bind sugars due to lack of the Ca 2+ /sugar-binding loop in their domains homologous to the carbohydrate recognition domain (CRD) but are still able to bind various physiologically important proteins and receptors. 77,78 Snaclecs bind to receptors on platelets, inducing either inhibition or activation of their aggregation. 78 By provoking thrombocytopenia, they contribute to the venom toxicity that was also observed in Vaa envenomed patients. 7,79,80 Some of these patients that suffered severe coagulopathy developed acute thrombocytopenia without significant changes in blood coagulation kinetics or fibrinogen level, which supports a nonenzymatic mechanism of platelet-related snaclecs' toxicity. 81−83 Snaclecs also potentiate the hemorrhagic activity of SVMPs. 84 Table 2. Low-Molecular-Mass Proteins Identified in the Vaa Venom a a Fraction B2 after gel filtration of crude Vaa venom ( Figure 6A) was separated by RP-HPLC ( Figure 6B), and the fractions were subjected to Edman sequencing. Major HPLC peaks were analyzed by nonreducing SDS-PAGE, proteins were in-gel digested with trypsin, and the resulting peptides were analyzed by tandem MS. Cys residues were carbamidomethylated before MS analysis but not before Edman sequencing. X denotes an unidentified amino acid residue, which is Cys in homologous sequences. Masses of molecular ions were determined by ESI-TOF analysis. Dis, disintegrin; MP, metalloproteinase; VEGF, vascular endothelial growth factor; M ox , oxidized Met.

Article
In venoms, snaclecs are present as heterodimers of α (14 to 15 kDa) and β (13 to 14 kDa) subunits cross-linked by a disulfide bond or as oligomers of the same or different αβ heterodimers, (αβ) 2 , (αβ) 4 , and (α 1 β 1 )(α 2 β 2 ). 77 Snaclec structures of ∼50 ((αβ) 2 ) and ∼25 kDa (αβ), have been discovered in Vaa venom. 12 Five of the nine Vaa snaclec monomers characterized in this study have sequences similar to those of α subunits (Vaa-snaclec-1, -3, -5, -6, and -9), and the other four have sequences similar to β subunits (Vaasnaclec-2, -4, -7, and -8) of snaclecs from other snake venoms ( Figure S-7). As previously noted, Vaa-snaclec-5 and Vaasnaclec-6 have identical mature amino acid sequences. The (αβ) 2 snaclec is composed of Vaa-snaclec-3 and Vaa-snaclec-2. 12 The greatest amount of amino acid sequence identity (mostly >90%) of Vaa snaclecs was found with various snaclecs from the M. lebetina venom, all of still unknown activity. Vaa-snaclec-1 and Vaa-snaclec-4 share high sequence similarity with the subunits A and B (83 and 98%) of snaclec VP12 from Daboia palestinae, which inhibits integrin α 2 β 1dependent melanoma metastasis. 85 Vaa-snaclec-8, however, shows high sequence identity to that of the partial sequence of a light chain 1 (the snaclec subunit) of VAFXA-II from Vaa, the P-IIId SVMP that activates coagulation FX. 58 In procoagulant P-IIId SVMPs, as in VAFXA-II, dimeric snaclecs are present as subunits linked to the C domain by a disulfide bond. The snaclec subunit serves to bind the substrate, FX, at its Gla (γ-carboxyglutamate residues containing) domain, to present it properly to the catalytic site at the MP domain for effective proteolytic activation. 77 3.2.1.8. Disintegrins. Disintegrins comprise another family of nonenzymatic dimeric toxins present in the Vaa venom (Tables 1 and 2; Table S-2). 12,15,86 They are common constituents of Viperinae venoms that act as integrin antagonists. 86−88 β-Subunits of dimeric Dis are derived from P-II SVMP precursors (e.g., Vaa-MPII-1, Vaa-MPII-2, and Vaa-MPII-3) in the process of post-translational proteolytic processing. α-Subunits, for example, Vaa-Dis-1, Vaa-Dis-2, and VA6, are encoded per se, by short-coding mRNAs that do not include a message for the MP domain. 86 Heterodimeric Dis are combinations of two diverse α subunits or one α and one β subunit, whereas just α subunits constitute homodimeric Dis. VA6 forms homodimers. Because sequences of Vaa-Dis-2 and VA6 differ in only four amino acid residues, three of which are similar, Vaa-Dis-2 probably also forms homodimers ( Figure S-8). Such a conclusion is also supported by the molecular ion mass of 14 027 ± 1 Da, determined for a native protein in the HPLC fraction 2 ( Figure 6B, Table 2), agreeing with the predicted masses of VA6 and Vaa-Dis-2 homodimers. However, of the other Dis molecular masses listed in Table  2, only two could be obtained by combining the theoretical masses of the known Vaa Dis monomers: 13 844 and 13 828 Da may be the masses of heterodimeric Dis that comprise Vaa-Dis-2 or VA6 as the α subunit and Vaa-MPII-1-Dis or Vaa-MPII-3-Dis as the β subunit. Many as-yet unknown Dis isoforms are therefore expected in Vaa venom. The feature common to α and β Dis subunits is that both possess 10 strictly conserved Cys residues that form the intra-and interchain disulfide bonds that define the conformation of the integrin-binding loop. 86 The specific recognition of integrins by Dis is defined primarily by the sequence of the integrinbinding motif at the tip of the integrin-binding loop (e.g., RGD, KGD, MGD, VGD, WGD, MLD) but also involves the amino acid residues flanking the tripeptide motif, where Dis display the highest level of sequence variability ( Figure S-8). At least four different integrin-binding motifs, RGD, KGD, VGD, and MLD, are present in the Vaa Dis subunits (Figure S-8). The first two are found typically in Dis, where they inhibit platelet aggregation by binding to the fibrinogen receptor, integrin α IIb β 3 . This interaction, already demonstrated in the case of VA6, 86 is additionally supported by the strong inhibition of ADP-induced platelet aggregation by crude Vaa venom as well as by the gel filtration fraction B2 that contains Dis. 12 Moreover, Dis, derived from Vaa-MPII-1 with the KGD motif, could represent a selective inhibitor of the integrin α IIb β 3 , as shown for KGD-Dis barbourin from Sistrurus barbouri. 89 Dis target integrin receptors of extracellular matrix proteins on various types of cell, in this way affecting adhesion between cells and the extracellular matrix, of the highest importance for normal tissue homeostasis. Misregulation of this process can result in the initiation and progression of a variety of diseases, such as cardiovascular, autoimmune, and cancer. 88 The receptor for fibronectin, integrin α 5 β 1 , which is involved in angiogenesis, is targeted by different viperid RGD-and VGD-Dis, including VA6. 86 The same specificity is expected from Vaa-Dis-2 with the RGD motif and from VGD-Dis that stem from Vaa-MPII-2 and Vaa-MPII-3 ( Figure S-8). Furthermore, MLD-containing Dis have been shown to bind various α 4 and β 1 integrins located on inflammatory and vascular endothelial cells, thus interfering with cell adhesion, proliferation, migration, and invasion. 90,91 For example, lebein-2 from Macrovipera lebetina and VLO5B from Macrovipera lebetina obtusa block the binding to β 1 integrins of laminin and the vascular cell adhesion molecule 1. 92 The MLD motif is also present in Vaa-Dis-1. Furthermore, its primary structure differs in only a few amino acids from those of lebein-2 and VLO5B, so the same activity can also be assumed for this molecule. As expected, a mixture of Vaa Dis significantly slowed down the migration of cancer cells. 15 3.2.1.9. Cys-Rich Secretory Proteins. Vaa-CRISP-1 homologues were identified in 15 spots on 2-DE ( Figure 4; Table 1, Table S-2) as having an apparent molecular mass of ∼26 kDa. Some of these spots (114−119) were among the most intense in the 2-DE gel. Although acidic Vaa-CRISP isoforms prevail, basic CRISPs were found in spots 112 to 114. Vaa-CRISP-1 is, like other SV CRISPs, a single-chain protein containing 16 strictly conserved Cys residues that form eight disulfide bonds. Ten of the Cys residues are clustered at the C-terminal end of the molecule (Figure S-1), which is structurally similar to the K + channel blockers. 93 SV CRISPs constitute a subfamily of the large CAP protein superfamily (pfam PF00188), whose members occur in all life kingdoms and are involved in diverse patho/physiological processes. 94 Despite the wide distribution of CRISPs in snake venoms, the biological functions of only a few have been established. Most of these inhibit the contraction of smooth muscles by blocking ion-gated, voltage-gated, or cyclic nucleotide-gated ion channels. 95,96 Vaa-CRISP-1 exhibits the highest sequence identity (∼96%) with two CRISPs from Viperinae snake venoms, Vbb (CAP74089) and V. berus nikolskii (B7FDI0), neither of whose activity is known. 97 Slightly less than identical to Vaa-CRISP-1 are ES-CRISP (∼85% identity) from Echis carinatus sochureki, 98 having antiangiogenic activity, and triflin (∼80%) from Protobothrops f lavoviridis, 99 a Ca 2+ -channel blocker ( Figure S-9). Furthermore, Vaa-CRISP-1 exhibits high amino  Figure S-10). Although the only isolated Vaa-VNGF cDNA (MG958503) codes for the C-terminally truncated protein, we were able to identify the missing sequence in two peptides, FIRIDTACVCVISR and IDTACVCVISR, in 2-DE spots, confirming that the full-length protein is expressed in the venom. VNGF, found in the venom of all venomous snake families, stimulates the growth of sensory and sympathetic nerves. 100,101 No direct toxic activity of VNGFs has been demonstrated so far, 2 but it has been suggested that they potentiate the action of certain other toxic components in venoms by binding to specific membrane receptors in a victim, increasing the vasopermeability or affecting its immune system. 100,102 VNGF from Naja kaouthia inhibits the proteolytic activity of SVMPs, to a degree comparable to that of inhibition of human MPs with the human β-NGF. 103 This suggests a further role of VNGFs in the regulation of the proteolytic activity of SVMPs.
3.2.1.11. Vascular Endothelial Growth Factor. Vaa-VEGF, or vammin, was identified in 2-DE spot 192 ( Figure 4, Table 1, Table S-2) and in RP-C18 fraction 7 ( Figure 6B, Table 2). It is a 25 kDa homodimer, a subtype of the VEGF-F molecule. 104 Vammin affects vasoconstriction by inducing hypotension and vascular permeability by a specific interaction with Tyr kinase receptor VEGFR-2 and the activation of the nitric oxide pathway. 105,106 In this way, it assists spreading of the venom from the bite site.

Low-Molecular-Mass Proteome Profiling of the Vaa Venom
The crude Vaa venom was first separated by gel filtration ( Figure 6A). Fractions containing low-molecular-mass proteins and peptides (B2, C1, C2, C3, and D) were here analyzed by RP-HPLC ( Figure 6B−F). The following low-molecular-mass proteinsDis, a new P-IIIe SVMP subclass protein (Vaa-MPIII-3), and VEGF (Table 2), together with peptides Kunitz-type SPis, NPs, and BPPs (Table 3)were identified in the HPLC fractions by Edman sequencing and MS/MS analysis.
3.3.1. Peptide Families in the Vaa Venom. In the Vaa venom, peptides were discovered that can be classified into four groups according to their structure or biological activity. They are discussed below.
3.3.1.1. Kunitz-Type Serine Protease Inhibitors. Kunitztype SPis are ∼60 amino acid long polypeptides found in the venoms of Viperidae and Elapidae. They exhibit a structural fold similar to that in bovine pancreatic trypsin inhibitor. 107 Vaa venom contains potent inhibitors of trypsin and chymotrypsin (Table 3). 108,109 Like orthologues from D. r. russelii and Pseudonaja t. textilis venoms, 110,111 trypsin inhibitor also inhibits plasmin and plasma kallikrein, thus affecting fibrinolysis and blood coagulation. SPis can form complexes with other venom components to enhance or moderate their pathophysiological activities. 107 In such a way, Vaa chymotrypsin inhibitor forms a complex with neurotoxic sPLA 2 , AtxA, thus augmenting its toxicity. 112 3.3.1.2. Natriuretic Peptides. NPs are hormones that exert diuretic, natriuretic, and vasorelaxant activities by interacting with specific receptors, thus playing an important role in cardio-renal homeostasis. 113 Many snake venoms harbor such peptides, thereby participating in prey immobilization by inducing severe hypotension. 114 The latter is one of the common symptoms following Vaa envenomation in humans, 7,8,79 so it was no surprise that NPs were found in Vaa venom (Table 3). Group I Vaa-MPi precursors ( Figure 2B) code for two 40 amino-acid-residue-long NP sequences that differ in only one amino acid residue at the C-terminus (Gly or Glu at position 38), whereas group II Vaa-MPi precursors ( Figure 2C) encode a single 36 amino-acid-residue-long NP sequence. Vaa NPs exhibit substantial sequence identity with NPs from other snake venoms and with human NPs (Figure S-11), in which two strictly conserved Cys residues form a

Journal of Proteome Research
Article disulfide bond and a 17 amino-acid-residue ring of highly conserved primary structure. 115 Besides lowering the blood pressure, Vaa NPs can inhibit platelet aggregation by analogy to the homologous lebetin-2 from Macrovipera lebetina 116 and with PNP, the NP from Pseudocerastes persicus venom. 117 3.3.1.3. Bradykinin-Potentiating Peptides. SV BPPs are Pro-rich peptides of 5 to 14 amino acid residues that induce systemic hypotension. 114 Their modular structure includes a pyroglutamic acid (pGlu or pE) at the N-terminus, the PXP motif (X is usually R, H, or G) in the middle, and the IPP sequence at the C-terminus. 118 Of the possible BPP sequences found in six Vaa-MPi precursors (Figure 2), only two, QRRPPEIPP and QRWPGPKVPP, were also detected in the venom (Table 3). Both have two Pro residues at the Cterminus, suggestive of strong bradykinin-potentiating activity. 119,120 All Vaa-MPi transcripts encode the decapeptidic BPP in different numbers of copies. In the venom, only one form of BPP was found, having a pE at its N-terminus (pERWPG-PKVPP). The message for a shorter BPP was found, however, in only two transcripts, Vaa-MPi-1 and Vaa-MPi-1′, presumably representing allelic forms. This BPP was expressed in both the N-terminally blocked (pGlu at the N-terminus) and the free forms (Glu at the N-terminus). Although the latter is expected to be more susceptible to hydrolysis by aminopeptidases, its physiological effect may be greater than that of the former, as demonstrated in the case of BPP from Gloydius halys venom. 119 3.3.1.4. Snake Venom Metalloproteinase Inhibitors. The catalytic activity of SVMPs in the venom is reduced by low pH, high concentrations of citrate ions, 121 and the presence of tripeptide inhibitors. 27,122−124 The latter are reversible, lowaffinity inhibitors, highly concentrated in the venom gland. The sequence of a tripeptide inhibitor (QKW) is encoded frequently by the Vaa MPi transcripts (Figure 2). The pyroglutamic form of the inhibitor, pEKW, is the major constituent of the Vaa venom gel filtration fraction E. 12 It effectively inhibits the fibrinogenolytic activity of Vaa SVMPs ( Figure S-12). Although the transcripts Vaa-MPi-1 and Vaa-MPi-2 code for a similar inhibitory tripeptide QNW, this, expectedly as pENW, has not been detected in the venom so far.

CONCLUSIONS
This work is the most comprehensive transcriptomic and proteomic survey of the Vaa venom to date. 45 different venom-related mRNA transcripts encoding peptide and protein precursors of 12 diverse types are characterized. More than 88% of the venom transcriptome comprises messages for MPis, BPPs, and NPs (all three on the same precursor), snaclecs, SVSPs, P-II and P-III SVMPs, sPLA 2 s, and Dis. In the venom, representatives of 16 protein families, altogether 57 different proteins, were identified. Four of themactin, calmodulin, PLB, and glutathione peroxidase are likely to be contaminants that entered the venom from damaged cells lining the venom gland. Peptides identified in the venom were NPs, BPPs, inhibitors of SVSPs, and inhibitors of SVMPs. The most abundant and diversified venom proteins were SVSPs, sPLA 2 s, snaclecs, and SVMPs, which account for 80% of all of the venom proteins and are responsible for the main toxic effects of the venom, including hemorrhage, coagulopathy, inhibition of platelet aggregation, and neurological disturbance. The production of antivenoms directed against their most toxic representatives is the way to a more effective and safer treatment of envenomed patients. Some newly discovered Vaa venom components open up novel lines of pharmacological research, for example, Vaa-LAAOs as potential antimicrobial, antitumor, and antiprotozoal agents, Vaa-snaclecs as inhibitors of melanoma metastasis, angiogenesis, and ion-channel activity, and Vaa-Dis as anticancer or antiplatelet agents. Venom peptides are also exciting; according to their structure, both Vaa-BPPs are expected to be endowed with a strong bradykinin-potentiating activity. Finally, our transcriptomic and proteomic analyses resulted in the discovery of an original SV protein, Vaa-MPIII-3. Its transcript is similar to that of P-III SVMPs but lacks the entire MP domain. The mature protein consists of just two domains, (truncated) D and C, thus defining a new subclass of SVMPs, the subclass P-IIIe. Such venom proteins presumably bind platelets and interfere with the hemostasis of the prey.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jproteome.9b00120.