Quantitative high-throughput profiling of snake venom gland transcriptomes and proteomes (Ovophis okinavensis and Protobothrops flavoviridis)

Metalloproteases

Snake venom MPs are presently classified into four groups, according to domain structure and size: P-I MPs possess a metalloprotease domain only and are largely hemorrhagic; P-II MPs are larger, with metalloprotease and disintegrin domains; P-III enzymes have metalloprotease, disintegrin, and cysteine-rich domains; and P-IV enzymes have a lectin-like domain linked by disulfide bonds to a P-III structure [33]. The structural complexity of P-III enzymes has resulted in greater functional diversity. They promote hemorrhage, inflammation, apoptosis, and prothrombin activation, while inhibiting platelet aggregation. As a general rule, P-III enzymes are more potent hemorrhagins than P-I enzymes [33]. In addition to degrading vascular endothelial basement membrane (hemorrhagins), collectively, MPs exhibit diverse and variable combinations of activities. Some anticoagulant metalloproteases degrade only the fibrinogen Aα chain [34], while others degrade one or more chains of both fibrinogen and fibrin with varying specificity [34–36]. Still others release histamine [37], antagonize platelet aggregation by different mechanisms [38–41], or activate [42] or digest plasminogen [43]. Some are procoagulant, possessing Factor Xa-like activity [44]. Few laboratories have exhaustively assayed MPs for potential biological and biochemical activities; thus, inferring such functions from structure is almost impossible. The same may be said of SPs.

The Protobothrops transcriptome contained transcripts for twelve P-II MPs and nine P-III MPs. One of the P-II enzymes (MP 01) constituted 11.06% of all toxin transcripts and collectively P-II transcripts accounted for barely ~11.1% of the transcriptome (Additional file 9: Figure S2; Additional file 1: Tables S1, Additional file 5: Table S3, and Additional file 2: Table S4). P-III transcripts were more abundant, comprising 15.8% of all transcripts. Three sequences were homologous to hemorrhagic proteases HR1A and B [45, 46]. The Ovophis transcriptome included seven P-II transcripts and three P-III transcripts. In Ovophis, P-II transcripts represented only 1.6% of all transcripts (Additional file 3: Table S2 and Additional file 4: Table S5). P-III transcripts added another 2.6%. Thus MPs comprised a mere 4.2% of the Ovophis transcriptome, compared to 26.9% in Protobothrops (Figure 1, Additional file 9: Figure S2 and Additional file 10: Figure S3; Additional file 3: Table S2, Additional file 5: Table S3, and Additional file 4: Table S5).

Of the 21 Protobothrops MPs, peptides were sequenced by mass spectrometry for 15, with coverage ranging from 31.1-91.4% of the respective transcripts (Additional file 1: Table S1). Peptide coverage of Ovophis MPs ranged from 26.9-80.6% (Additional file 3: Table S2).

Phospholipases A ₂

The Protobothrops transcriptome contained four transcripts for PLA₂s (Additional file 1: Table S1), including a Lys-49 myotoxin (PLA₂ 4) [Accession #: AB851948] and a weak neurotoxin similar to trimucrotoxin [47] (PLA₂ 2) [AB848131]. PLA₂ 1 [AB848154] accounted for 26.7% of all transcripts, while PLA₂ 2 amounted to an additional 5.5% (Additional file 1: Table S1 and Additional file 2: Table S4). The Ovophis transcriptome contained two PLA₂ transcripts; however, the more abundant transcript, PLA₂ 1, comprised only 0.65% of the transcriptome (Additional file 3: Table S2 and Additional file 4: Table S5) [AB848270]. Peptides sequenced by mass spectrometry covered 98.3% of PLA₂ 1, but no peptides were found for the minor transcript.

Serine proteases

Of the 18 SP transcripts in the Protobothrops library (Additional file 1: Table S1 and Additional file 2: Table S4), only two (SP01 and SP 12) (FPKM% 2.6 and almost zero, respectively) can be confirmed as complete (Additional file 11: Figure S4). Several transcripts appear to encode dysfunctional SPs. For instance, SP16 encodes 36 residues and is bracketed on both ends by stop codons (Additional file 1: Table S1; Additional file 11: Figure S4). Given that it was expressed at a very low level and that no peptides were sequenced by mass spectrometry, we think it is unlikely to play any role in envenomation.

SP01, the most abundant SP transcript, corresponds to a protein that appears in the literature under the names of habutobin [48–51] and flavoxobin [52–56], a weakly thrombin-like enzyme of 242 amino acids that specifically releases fibrinopeptide A from fibrinogen [55]. No information is available with regard to possible kallikrein-like activity. However, Yamamoto et al. [56] found that flavoxobin is an active C3-convertase that selectively releases C3b and C3a. It remains active in blood containing endogenous protease inhibitors, and promotes massive C3 consumption, and to a lesser extent, C5 cleavage. A kinin-releasing enzyme, flavoviridiobin, is also known from this venom [57]; however, since no sequence data are available, we cannot identify it among our transcripts. Enzymatic digests of crude venom effected with trypsin, chymotrypsin, and Glu-C (Pierce) yielded peptides that accounted for 94.6% of the primary structure of SP01 (Additional file 1: Table S1; Additional file 11: Figure S4). Reasonable peptide coverage of transcripts as minor as 0.24% (SP11, 50.0%) was achieved (Additional file 1: Table S1).

In contrast to the Protobothrops library, the Ovophis library contained transcripts for 26 different SPs (Additional file 1: Table S1 and Additional file 2: Table S4; Additional file 12: Figure S5). Peptide coverage of 36% or above was achieved for 22 of these, with coverage above 70% for 11 of them. Two transcripts (SP09 and 10) appear to be plasminogen activators, while SP20 is most similar to a kinin-releasing enzyme from the venom of Bothrops jararaca. Serine proteases display numerous amino acid substitutions, and the structural determinants that specifically account for kinin-releasing activity are unknown [58]. The difficulty in assigning pharmacological activities to specific sequence variations is immediately apparent upon a cursory examination of Additional file 11: Figure S4 and Additional file 12: Figure S5.

Wu et al. [59] reported a novel class of inactive serine protease homologs (SPH) that displayed an arginine substitution for His-43 of the catalytic triad. SP13 was the only serine protease in our Protobothrops library that showed this His → Arg mutation (Additional file 12: Figure S5, position 126); however, the Ovophis library contained eight transcripts with His → X substitutions (Additional file 12: Figure S5, position 101). Two of these, SP08 and SP22 showed His → Lys substitutions; two putative thrombin-like enzymes, SP16 and SP17 displayed His → Asn substitutions, and SP07 had a His → Ala substitution. Numerous other sequence differences appear in that transcript as well (Additional file 12: Figure S5). SPHs from other sources have been shown to possess diverse activities, so it is possible that inactive SPs in venoms have developed other unknown functions, some of which may be specialized for particular prey types.

An inactive catalytic triad is but one of many structural differences manifested by Ovophis SPHs (Additional file 12: Figure S5). Almost all of the cysteine residues are in different positions as well (Cys-102 has only moved to position 100, but most have shifted substantially more), although within the group, most residues are conserved across most sequences. SP07 is a marked exception in the latter regard (Additional file 12: Figure S5). Another oddity among these sequences is that four of them (SP01, 07, 23, and 26) are truncated C-terminally with stop codons, despite the fact that SP01 and 07 display expression levels of 9.6 and 7.1%, respectively. Wang et al. [60] reported that a Kentucky population of Crotalus horridus lacks an acidic PLA₂ because the codon for Tyr²² has mutated into a stop codon. They concluded that low PLA₂ expression levels in most Crotalus horridus venoms can be attributed to translation blockage. At this point, it is difficult to know how widespread this phenomenon might be, but it is apparent that these two Ovophis SPs are translated effectively since they had ample peptide coverage (Additional file 3: Table S2).

L-amino acid oxidase

The Protobothrops transcriptome included two transcripts for L-amino acid oxidase [AB848142, AB848143], comprising 2.3% and 6.8% of all transcripts, respectively (Figure 1; Additional file 1: Table S1). A single LAO transcript was present in Ovophis glands [AB848269], representing 0.6% of the transcriptome (Additional file 3: Table S2). Peptides accounting for 84.6% and 70.8% of Protobothrops LAO 1 and LAO 2, respectively, and 78.7% of the Ovophis LAO transcript sequence was identified by mass spectrometry (Additional file 1: Table S1 and Additional file 3: Table S2).

Cysteine-rich secretory proteins

Two CRISPs were identified in the Protobothrops transcriptome (Additional file 1: Table S1 and Additional file 2: Table S4). CRISP 1 [AB848115], (FPKM = 3.9%) for which a complete transcript was obtained, is identical to triflin [61], but CRISP 2 [AB851959] aligns best with a CRISP bearing an EGF-like calcium-binding domain from the venom of Crotalus adamanteus[62] (Additional file 2: Table S4). However, the putative 39-residue EGF domain in the C. adamanteus toxin does not align well with the corresponding region of the Protobothrops transcript. The latter contains only four acidic residues, compared with nine in the C. adamanteus sequence. Only three of the five C. adamanteus cysteine residues match, and the two sequences require a two-residue gap to achieve even this poor alignment. Therefore, we think it unlikely that there is a functional EGF-like calcium binding domain in the Protobothrops toxin. Moreover, no peptides were sequenced for this odd CRISP, whereas 84.6% of CRISP 1 was sequenced.

A single, complete CRISP transcript (FPKM = 0.2%) was identified in the Ovophis transcriptome (Additional file 2: Table S2) [AB848276], but sequenced peptides accounted for 89.0% of its primary structure. It was most similar to a CRISP from the venom of Bothriechis schlegelii [GenBank: ACE73559.1].

CRISPs are generally not abundant components of snake venoms, but they are widely distributed taxonomically. Ablomin (Gloydius blomhoffii), triflin (Protobothrops flavoviridis) and latisemin (Laticauda semifasciata) are L-type Ca²⁺ channel antagonists of depolarization-induced arterial smooth muscle contraction, but they do not affect caffeine-induced contraction [61]; thus they promote vasodilation and hypotension. Tigrin from “venom” of the Japanese colubrid, Rhabdophis tigrinus, affected neither. This is probably because Rhabdophis venom glands are not secretory in nature. Instead, Rhabdophis glands sequester toxins from the blood stream that are derived from the toads that Rhabdophis eats [63]. Thus, tigrin is most likely an amphibian toxin, intended for oral or gastric activity, and not a snake toxin, designed for direct vascular action. In contrast, patagonin, a CRISP isolated from the venom of the colubrid, Philodryas patagoniensis, damaged murine skeletal muscle [64].

Nerve growth factor

Both habu transcriptomes contained a single, complete transcript for nerve growth factor [Pf: AB848144; Oo: AB848271] (Additional file 1: Table S1 and Additional file 3: Table S2). The Protobothrops transcript accounted for 0.7% of all transcripts while the Ovophis transcript accounted for 0.5%. Both transcripts are translated and peptides were isolated by mass spectrometry. NGFs function as arginine esterases [65, 66], so they probably contribute to venom hypotensive activity via nitric oxide liberation and histamine release [67, 68]. Mouse salivary NGFs activate plasminogen, their only known action upon a biologically important, non-neural substrate [69, 70], but it is not clear whether snake venom NGFs can also do this. If so, they would hinder blood clotting.

C-type lectins

Snake venom C-type lectins, or snaclecs [71] are commonly found in pit viper venoms. These proteins differ from classical C-type lectins in that they lack the calcium and sugar-binding loop and instead bind to a large variety of proteins and receptors involved in hemostasis, including coagulation factors IX and X and various blood platelet receptors [72]. They may consist of one, two, or four αβ heterodimers, and in some cases, the heterodimer is incorporated into a metalloprotease [73]. In many CTLs, dimers are formed by domain swapping between subunits [73].

CTL pharmacology is quite complex. Taniuchi et al. [74] found that flavocetin A actually induces formation of small platelet aggregates, but the dose-dependency is bell-shaped, with a maximum effectiveness at 1-2 μg/mL. Clemetson [72] lamented that because so much venom research is now done at the transcriptional level, the protein chemistry and pharmacology necessary to understand CTL diversity has lagged way behind. In reality, the same could also be said of any other toxin family that shows significant diversification, such as 3FTxs, SPs, MPs, and PLA₂s.

Venom C-type lectins may activate platelets or inhibit platelet activation, but either mechanism serves the function of inducing thrombocytopenia. Because C-type lectins are non-enzymatic, a 1:1 stoichiometry exists between these toxins and their targets. Clemetson [72] noted that for this reason, it is much more efficient to clear platelets by activating them than by inhibiting them. However, different species of snakes employ both strategies, and it is probably necessary to look at all the toxins in a given venom that impact hemostasis, before drawing any conclusions.

Twelve Protobothrops CTL transcripts included three α-chains and three β-chains homologous to flavocetin A, an (αβ)₄ inhibitor of von Willibrand factor-induced, GP1B-mediated platelet aggregation [75, 76] and convulxin, a potent (αβ)₄ inducer of platelet aggregation that binds to GPVI [73] (Additional file 13: Figure S6; Additional file 1: Table S1 and Additional file 2: Table S4). One of the flavocetin A-like α-chains (CTL03α) and CTL07 F IX/X displayed a number of sequence differences, including an unusual C-terminus (CKFLRPR). Whether these have any pharmacological significance is unknown. In addition to toxins that target blood platelets, there were five A chains and one B chain for proteins that bind to coagulation Factors IX/X (Additional file 1: Table S1 and Additional file 2: Table S4). Factor IX/X binding proteins inhibit blood coagulation by blocking the host clotting cascade [77].

Seven Ovophis CTL transcripts apparently all encode proteins that affect platelet activation (Additional file 3: Table S2 and Additional file 4: Table S5; Additional file 13: Figure S6). They are homologous to flavocetin A and convulxin. We did not discover any Ovophis transcripts that encode anticoagulant Factor IX/X-binding proteins. Our Ovophis cDNA library contained one α-chain, CTL1α, similar to the α-chain of flavocetin A (Protobothrops flavoviridis) and the convulxin A- and C-chains (Crotalus durissus terrificus) (Additional file 13: Figure S6). CTL1α is most like crotacetin (Crotalus durissus terrificus. It represented 0.16% of all transcripts. In addition, there were six β-chains, homologous to the flavocetin A β-chain and the convulxin B- and D-chains (Additional file 13: Figure S6; Additional file 3: Table S2). Together these seven CTLs represented 0.47% of all transcripts.

Bradykinin-potentiating peptides

A single bradykinin-potentiating peptide (BPP) was sequenced from Protobothrops venom using mass spectrometry (QSKPGRSPPISP) (Additional file 1: Table S1), confirming the existence of a BPP proposed by Higuchi et al. [78], on the basis of a cDNA transcript. A second possible BPP (VVVQPHESPAGGTTA) was also sequenced, but to date, no other BPPs have been found with proline immediately after the N-terminal pyroglutamic acid, making this sequence suspect. Moreover, the VVV-sequence, N-terminal to the glutamine, and the C-terminal AGGTTA-sequence are highly questionable. Possibly this peptide could be processed to QPHESP. This possible BPP is located at the C-terminus of our BPP transcript; however, our BPP transcript is incomplete, since it lacks a stop codon and it does not include the C-type natriuretic peptide-coding region reported by Higuchi et al. [78] (Figure 3).

Our Protobothrops transcript [AB851926] also contains the second BPP sequence reported by Higuchi et al. [78], although this BPP was not identified by mass spectrometry. They posited the existence of two BPPs based on the assumptions that such sequences should possess glutamine at the N-terminus and proline at the C-terminus, and should be about 11 residues in length. In fact, BPPs from three to 14 residues have been reported [79, 80] (Additional file 14: Figure S7). Both the Higuchi Protobothrops transcript and ours suggest another probable BPP with the sequence QWMPGGRPPHHIPP (Figure 3). The Gloydius transcript of Higuchi et al. [78] also contains a tripeptide (QWS) that occurs in five locations at the end of the BPPs that they predicted (Figure 3). Two tripeptides from Bothrops insularis venom having pyroglutamic acid at the N-terminus (QQK, QKW) were sequenced by Cintra et al. [79], and these peptides were shown to have bradykinin-potentiating activity on guinea pig ileum (Additional file 14: Figure S7). It is possible that the peptide QWS is likewise biologically active. Other tripeptides are found in the Higuchi Protobothrops and Gloydius transcripts and in our Ovophis transcript. These have the sequences QER (Protobothrops) and QAR (Gloydius and Ovophis) (Figure 3, residues 281-283). All of these are immediately N-terminal to nonapeptides that could also be BPPs (Figure 3, residues 284-292). These sequences are as follows: Pf, QKWGRMVQP; Gb, QNWARMVNP; Oo, QKWGRMVPP.

In addition to being truncated on the C-terminal end relative to the Higuchi transcript, our transcript displays a significant N-terminal extension, containing three additional possible BPPs (Figure 3). These have the sequences QRRVHGGERIWP, QSARLDSTRLGSAP, SRPPSLPAPAQP; however, additional work will be necessary to determine whether these sequences are actually hypotensive and whether they are actually expressed in habu venom.

Our Ovophis BPP transcript [AB852004] displayed a C-terminal stop codon, but was incomplete on the N-terminal end. However, the Ovophis transcript did contain a sequence for a C-type natriuretic peptide (GVAKGCFGLKLDRIGTMSGLGC) that was identical to that reported for Gloydius blomhoffii venom [78] (Figure 3). It differed at five residues from the Higuchi Protobothrops transcript (Figure 3).

When mass spectrometry was used to analyze crude Ovophis venom for the presence of BPPs, the sequence RPPGPPIPP, and derivative forms thereof (PPGPPIPP and GPPIPP) were isolated. This sequence does not occur in our truncated transcript; however, it is nearly identical to a proposed BPP from the N-terminal end of a BPP-CNP transcript from Gloydius blomhoffii (RPPGPPIPR) [78, 81] and from Bothrops jararaca venoms [80] (Figure 3 and Additional file 14: Figure S7).

Potency of bradykinin-potentiating peptides (BPPs) increases ~200-fold if the C-terminal proline residue is doubled [82]. While the C-terminal tripeptide of a BPP from Gloydius halys venom was shown to be essential for its activity, removal of the N-terminal pyroglutamate residue made it twice as potent [82]; thus, while the N-terminal pyroglutamate common to BPPs (Additional file 14: Figure S7) may prevent their rapid degradation by prey aminopeptidases, it is actually an impediment to bradykinin potentiation. Interestingly, bradykinin-potentiating activity is not correlated with inhibition of angiotensin-converting enzyme (kininase II) activity [82, 83], which is much too slow to be relevant to envenomation.

Various studies have shown that bradykinin potentiation and inhibition of somatic angiotensin-converting enzyme (sACE) by pit viper hypotensive peptides are independent biochemical activities [84–89]. The presence of paired proline residues at the C-terminus and a pyroglutamic acid residue at the N-terminus are not the only requirements for bradykinin-potentiating activity or sACE inhibition. Guerreiro et al. [86] have shown that argininosuccinate synthetase is activated by a BPP from Bothrops jararaca venom, indicating that nitric oxide formation represents yet another means by which BPPs promote hypotensive shock to limit prey flight [1].

Phospholipase B

Phospholipase B (PLB) activity was first reported in snake venoms by Doery and Pearson [90], who confirmed its presence in the venoms of Naja naja, Pseudechis porphyriacus, and Agkistrodon piscivorus. In 1987, PLB from Pseudechis colletti venom was characterized for the first time [91]. No venom PLB sequences were reported until 2011, when transcripts were isolated from venoms of Drysdalia coronoides[92] and Crotalus adamanteus[62]. While PLB accounted for only 0.06% of all transcripts in those species, it represented 0.14% of Protobothrops [AB848155], and 0.15% of Ovophis transcripts [AB848284, AB848285] (Additional file 1: Table S1, Additional file 3: Table S2, Additional file 5: Table S3). Peptides covering 26.1% of the Protobothrops sequence and 50.5% and 61.6% of the two Ovophis sequences, respectively, were isolated by mass spectrometry (Additional file 1: Table S1 and Additional file 3: Table S2; Figure 4). To the best of our knowledge, these are the first protein sequence data for any snake venom PLB.

Feola et al. [93] found that in rabbits, i.v. injections of phosphatidylethanolamine (PE) and phosphatidylserine (PS) caused significant hypotension, cardiac arrhythmias, bronchospasm, activation of intravascular coagulation, complement, platelets, and leukocytes with release of histamine, serotonin, and thromboxane at a dose of 0.10 mg/kg and caused cardiac arrest and death at a dose of 0.30 mg/kg. All of these effects are consistent with snake venom envenomation strategies [1]; however, it is not clear whether intact PE and PS are released from cell membranes by pit viper venoms. Kinoshita et al. [94] found that PS and PE were not released from membranes by purified Protobothrops flavoviridis phospholipase A₂; however, one would not really expect this, and venoms contain many other components in addition to phospholipase A₂. What is more, prey tissue destruction by venom components liberates many endogenous compounds, further complicating the picture. At present, the role of PLB in envenomation remains unclear, beyond its generalized hydrolysis of cell membrane phospholipids.

Phosphodiesterase

The Protobothrops transcriptome contained four phosphodiesterase (PDE) transcripts, ranging from 0.33-0.56% of all transcripts (Additional file 1: Table S1), which comprised, in aggregate, 0.2% of the transcriptome [AB848150, AB848151, AB848152, AB848153]. Peptides covering 53.4-56.8% of the four PDE sequences were sequenced by MS. PDE was less diversified in Ovophis (Additional file 3: Table S2). Two PDE transcripts accounted for a negligible portion of the Ovophis transcriptome [AB851989, AB851990]. Sequenced peptides accounted for only 7.8-13.0% of the two PDE sequences.

Vascular endothelial growth factor-like proteins

Five VEGF isoforms comprised just over 0.008% of all Ovophis transcripts [AB852007, AB852008, AB852009, AB852010, AB848274], while three Protobothrops transcripts totaled 0.32% of that transcriptome [AB848141, AB851940, AB851941] (Additional file 1: Table S1, Additional file 2: Table S4, Additional file 3: Table S2, Additional file 4: S5, Additional file 5: Table S3). Fourteen unique peptides were isolated for Protobothrops VEGF 1, accounting for 81.1% of its sequence. Fourteen peptides were also sequenced from Ovophis VEGF 5, amounting to 60.3% coverage (Additional file 1: Table S1 and Additional file 3: Table S2).

Both venomes contain transcripts for several structural subclasses of VEGFs, although owing to the great diversification of these sequences, classification is difficult. For instance, Ovophis VEGF 1 possesses a 24-residue insert seen in no other sequence (Figure 5). Ovophis VEGF 1 and 2 and Protobothrops VEGF 2 all possess long C-terminal extensions and align well with human VEGF-A₁₆₅ (Figure 5). Ovophis VEGF 2 is the most heavily expressed VEGF in that venome, at 0.222% (Additional file 3: Table S2). Human VEGF-A binds to fms-like tyrosine kinase-1 (VEGF Receptor-1) (VEGFR-1) and to kinase insert domain-containing receptor (VEGFR-2), but not to VEGFR-3 (fms-like-tyrosine kinase-4) [95–98]. VEGF-A induces vasodilation mediated by nitric oxide [99] and increases vascular permeability 50,000-fold more potently than histamine [100]. In addition, VEGF-A promotes tachycardia, hypotension, and diminished cardiac output when injected i.v. in rats [101]. It is likely that Ovophis VEGF 1-2 and Protobothrops VEGF 2 have similar pharmacology, as these symptoms are consonant with snake envenomation strategies [1].

Ovophis VEGF 5 [AB848274] and Protobothrops VEGF 1 [AB848141] are homologous to vammin, from the venom of Vipera ammodytes. All three of these display short C-terminal extensions of 16-17 residues that bind heparin [102] (Figure 5). Vammin specifically recognizes VEGFR-2 [98]. Both vammin and VR-1, a VEGF from Daboia russellii venom, enhance vascular permeability with greater potency than does VEGF-A₁₆₅[98]. Additionally, Yamazaki et al. [103] have shown that a Lys-49 PLA₂ without catalytic activity further enhances the vascular-permeability promoting capacity of vammin.

Ovophis VEGF3-4 and Protobothrops VEGF3 comprise a subclass with no C-terminal extension, or an extremely short extension corresponding to the C-terminus of Ovophis VEGF 1-2 and Protobothrops VEGF2 (Figure 5). These are significantly shorter than barietin from the venom of Bitis arietans[98], and they do not align well with it or with vammin (Figure 5).

5’-Nucleotidase

Both transcriptomes included a single transcript for 5’-nucleotidase (Additional file 1: Table S1 and Additional file 3: Table S2) [Pf: AB848147; Oo: AB851991]. In both transcriptomes 5’-nucleotidase was a negligible constituent. Mass spectrometry identified 51 venom peptides accounting for 63.3% of the expected sequence of the mature Protobothrops protein, while 65 unique peptides were detected in Ovophis venom, accounting for 12.9% of the 5’-nucleotidase in that venom.

5’-nucleotidase is ubiquitous in snake venoms [104–107], suggesting a central role in envenomation. This enzyme is known to cleave a wide variety of ribose- and deoxyribose-containing nucleotides [108–110]. It is most active against AMP [109, 110] supporting the central role of adenosine in envenomation proposed by Aird [1]. 5’-nucleotidase does not cleave flavin mononucleotide, or cAMP; however, these are hydrolyzed by venom PDE.

Galactose-binding lectins

In contrast to C-type lectin-like proteins (CTL), galactose-binding lectins (GBLs) possess intact calcium and galactose-binding loops [72]. GBLs are similar in size to CTL-like proteins and are also dimeric. However, instead of interacting with platelets, GBLs aggregate erythrocytes [111, 112]. For this reason, most authors, starting with Gartner et al. [112], have assumed that the presence of GBLs in venom is related to envenomation; however, several lines of evidence raise the possibility of a role unrelated to prey immobilization or digestion [1].

GBLs have been shown to be strongly mitogenic [113–115]. Their mitosis-inducing effects on lymphocytes were found to be comparable to those of concanavalin A [115]. Fry and Wüster [116] noted that GBLs appear to be basal phylogenetically among venomous snakes, whereas CTL-like proteins appear only in the Viperidae. Unlike CTL-like proteins, GBLs display very little sequence variability, suggesting that they are not under selective pressure to diversify, as CTL-like proteins are [117]. Lectins with similar sugar specificity are found in many tissues [118]. In Protobothrops and Ovophis, GBLs are expressed at very low levels (Additional file 1: Tables S1 and Additional file 3: Table S2) [Pf: AB848130; Oo: AB848277]. Ogilvie et al. [119] likewise found low expression levels for GBLs in Bothrops atrox (0.2%) and Dendroaspis jamesonii (0.4%) venoms, with a somewhat higher level (0.8%) in Lachesis muta venom. Lomonte et al. [120] found that the GBL from Cerrophidion godmani venom exhibited edema-forming activity in mice, but concluded that with its low potency and low abundance, it probably plays relatively little role in envenomation. The aforementioned data suggest that GBLs may exist in venom as mitogens to regulate synthetic activity in the glandular epithelium itself. If this view is correct, hemagglutinating and edematogenic activities would be fortuitous, but of secondary importance. Nonetheless, the relative importance of such activities might vary among taxa.

Aminopeptidases

Aminopeptidase N plays a significant role in preventing hypertension by degrading Angiotensin III to Angiotensin IV [121]. The role of aminopeptidase A in blood pressure regulation appears to be more complex. APA degrades Angiotensin II to Angiotensin III [122]. When acting at peripheral sites, Angiotensin III is less potent hypertensive than Angiotensin II [122], but in central sites, Angiotensin III raises blood pressure more effectively than Angiotensin II [122, 123].

Various lines of evidence suggest a role for APA in promoting hypotension in situations analogous to envenomation. Systemic administration of APA in spontaneously hypertensive rats [124] or hypertensive rats infused with angiotensin II [125] reduced their blood pressure. Administration of amastatin, an APA inhibitor, raised blood pressure in normotensive rats [126].

To date nominal aminopeptidases A and N have been isolated from pit viper venoms, although expression levels appear to be generally low, and many venoms apparently may not contain either. In the present study, Ovophis venom contained a complete transcript for Aminopeptidase A [AB848288], while Protobothrops venom contained two APA transcripts [AB848148, AB848149]. However, the Ovophis Aminopeptidase A transcript comprised only 0.002% of all transcripts, while the two more abundant Protobothrops transcripts together comprised 0.073%; hence both are very minor venom constituents. Ovophis APA and Protobothrops APA 1 were closely related to that reported from Gloydius brevicaudus venom [127], differing at only 24 and 22 residues out of 953, respectively.

Tu and Toom [128] found that nearly all venoms hydrolyze L-leucyl-β-naphthylamide, but that there exists great variation in activity levels. Aird [1] suggested that the principal function of leucine aminopeptidase (arylamidase) (LAP) is digestive and that it links the hemorrhagic venom metalloproteases and other venom and endogenous prey peptidases, to L-amino acid oxidase in order to potentiate H2O2 liberation, resulting in hypotension and anticoagulation. It is probable that numerous other amino- and carboxypeptidases in plasma also pass free amino acids to LAO. Clearly the release of Leu from circulating peptides is not solely dependent upon venom LAP. This may partly explain the variation in LAP levels that exists among different venoms [107]. If LAP is abundant in prey tissues, there may not be great selection pressure governing its level of expression in venoms. In the two transcriptomes, LAP was a very minor component [Pf: AB851938; Oo: AB851994] (Additional file 2: Table S4 and Additional file 4: Table S5).

The Protobothrops transcriptome possessed two aminopeptidases that show similarity to Aminopeptidase N [AB851954, AB851955] (Additional file 2: Table S4), but some of these did not manifest much similarity to the two Gloydius brevicaudus enzymes [127]. They also showed similarity to Aminopeptidase A, so without careful biochemical analyses it is impossible to classify them precisely. Furthermore, it may be that the aminopeptidase nomenclatural system devised for use with human enzymes, may not be applicable to snake venom aminopeptidases.

Dipeptidyl peptidase IV

Dipeptidyl Peptidase IV (DPP IV) was first discovered in venoms of various Micrurus species by Jorge da Silva and Aird [107]. It was also detected in the venoms of two other elapids, Bungarus multicinctus, Naja naja, and in that of the Brazilian crotaline, Bothrops moojeni. DPP IV titers varied by more than 4x among the different venoms. DPP IV is believed to function in envenomation by blunting a hypertensive response on the part of envenomated prey [1]. Ogawa et al. [129] published the first snake venom DPP IV primary structures, a pair of isomeric sequences derived from cDNA libraries of Gloydius brevicaudus venom glands. They determined that the signal peptide was not removed from these sequences. Later Ogawa et al. [130], showed that DPP IV, is actually secreted membrane-bound in exosomes. These micro-vesicles probably account for the “pre-peak” that elutes well ahead of the largest proteins when snake venoms are fractionated using gel filtration chromatography [131, 132]. Exosomes were later shown to be present in human saliva as well [133]. DPP IV is nearly ubiquitous among elapid and viperid venoms, but it exhibits great quantitative variability even among full siblings [134].

The Protobothrops flavoviridis DPP IV sequence [AB851922] comprises 751 residues, like those from Gloydius, while the Ovophis sequence has 752 [AB848286]. Nonetheless, the Protobothrops and Ovophis sequences are more similar to each other than to the Gloydius sequences (Additional file 15: Figure S8). The Protobothrops sequence is missing one of a pair of asparagine residues present in the other three sequences, but both the Protobothrops and Ovophis sequences have a leucine residue that is missing in the Gloydius sequences (Additional file 15: Figure S8). No DPP IV peptides were discovered with mass spectrometry following enzymatic digestion of Protobothrops venom; however, three unique peptides accounting for 4.6% of the Ovophis DPP IV sequence were isolated. Venoms were well centrifuged before sample digestion, which probably pelleted the exosomes; thus it is surprising that any Ovophis peptides were identified.

Glutaminyl cyclase

QC cyclizes, and thereby protects the N-termini of biologically active peptides, such as the BPPs [135], some metalloproteases [136–138], and the B and C chains of the acidic subunit of crotoxin homologs [139, 140]. No direct role in envenomation has been suggested for QC to date. However, while cyclization protects these peptides against degradation by prey plasma aminopeptidases, in the case of BPPs, bradykinin-potentiating potency is reduced by half [82].

A total of five snake venom QC cDNAs have been sequenced to date. Two of these belong to colubrids of the Genus Boiga[141] and the other three have been sequenced from crotalids on three different continents (Gloydius blomhoffii, Bothrops jararaca, and Crotalus adamanteus).

The present study adds eight additional sequences, of which a couple are distinctly different from those previously published. The Protobothrops sample contained four QC transcripts for two pairs of toxins [AB848133, AB848134, AB851933, AB851934]. The two identical long Protobothrops transcripts show near identity with other published crotalid sequences (Figure 6). However, as confirmed by the presence of stop codons, two other identical short sequences are missing the N-terminal 37 residues of the longer sequences. The next eight residues of the short sequences are unique, but thereafter they are identical to the long sequences (Figure 6). Pawlak and Kini [141] reported a similar, though less extensive deletion in the Boiga dendrophila QC; thus it is clear that this sort of alternate splicing/post-translational modification is characteristic of snake venom QCs. Ovophis venom also contains four QC sequences [AB852014, AB852015, AB851985, AB851986], but because all are incomplete, no conclusions can be drawn regarding their length. The most highly expressed of these four represented only 0.008% of all transcripts (Additional file 3: Table S2), consistent with an indirect role in envenomation. Peptides were isolated for all four Protobothrops QCs, but only one of the Ovophis isoforms.

Hyaluronidase

Hyaluronidase is not a major constituent of either venom. A single complete transcript was found in the Protobothrops library [AB851937], while two complete Ovophis transcripts were sequenced [AB851977, AB851978]. No hyaluronidase transcript was more abundant than the cutoff for contaminants and no peptides were isolated from either venom. Venom hyaluronidase has been deemed a “spreading factor” because its degradation of the extracellular matrix enables other venom constituents, such as metalloproteases and phospholipases, to attack additional tissues [142, 143]. As such, hyaluronidase probably serves primarily to digest the prey.

Three-finger toxins

Protobothrops venom, but apparently not that of Ovophis, contains a three-finger toxin (3FTx) [AB851958]. This sequence is most closely related to a transcript reported from Sistrurus catenatus edwardsi venom [144] and to candoxin isolated from the venom of an elapid, Bungarus candidus[145] (Figure 7). 3FTxs were not detected in an earlier study of Sistrurus catenatus barbouri venom [146], and they have not been observed in many other venomics studies of pit vipers [62, 147–152]. Other studies have located 3FTxs by transcriptomic means, but not by proteomics approaches [15]. This is not surprising, given their low expression levels in many taxa (0.8% in Sistrurus catenatus venom [144]). While 3FTxs are minor components of most pit viper venoms, relatively high expression levels have been reported in some species. In a study of Caribbean pit vipers, using Roche 454 sequencing technology, Durban et al. [32] reported considerable variability (Crotalus simus, 12.7%, western Bothrops asper, 4.7%; Bothriechis schlegelii, 3.6%; eastern Bothrops asper, 0.8%; Cerrophidion godmani, 0.6%; and Atropoides picadoi, 0.4%).

The Protobothrops 3FTx differs slightly in its disulfide bond structure from all known 3FTxs (Figure 7). It shares a cysteine residue in position 18 with the 3FTx from Sistrurus catenatus edwardsi venom; however, Cys-11, which is linked to Cys-18 in the Sistrurus toxin, in the Deinagkistrodon acutus short neurotoxin, and in candoxin, occurs at position 9 in the Protobothrops toxin (Figure 7).

Enzymes involved in purine and pyrimidine biosynthesis

Aird [1] explained the neuromodulatory and hypotensive roles of purine nucleosides in the pharmacology of snake envenomation. A later study quantified purine and pyrimidine nucleosides in a wide variety of elapid, viperid, and crotalid venoms [31]. Possible roles of uridine and cytidine in envenomation are less clear than those of purine nucleosides. Because nucleosides are endogenous regulatory substances in all vertebrates, it is impossible for any prey species to develop resistance to them; thus they represent the perfect predatory biochemical weapon. However, their endogenous nature also means that the enzymes involved in nucleoside biosynthesis would be expected in any venom gland transcriptome, regardless of whether nucleosides are actually secreted into the venom in quantities relevant to envenomation. As a result, no venomics studies to date have specifically looked for the presence of nucleoside biosynthetic enzymes. Instead they have been treated as “housekeeping” genes. In fact, only Rokyta et al. [62] have reported the sequences of adenylosuccinate synthetase, adenylosuccinate lyase, IMP dehydrogenase, GMP synthetase, nucleoside monophosphate kinase, nucleoside diphosphate kinase, or CTP synthetase.

In both transcriptomes, we found transcripts for all four of the enzymes required to synthesize AMP and GMP from IMP [adenylosuccinate synthetase, Pf: AB851944; Oo: AB851992, AB851995; adenylosuccinate lyase, Pf: AB851928; Oo: AB851974; IMP Dehydrogenase, Pf: AB848116; Oo: AB851975, AB851979, AB852003; GMP synthetase, Pf: AB851932, AB851936, AB851946, AB851952; Oo: AB851972, AB851983, AB851999] (Additional file 1: Table S1 and Additional file 2: Table S4). The monophosphates may then be dephosphorylated by a variety of non-specific phosphatases or by venom or endogenous prey 5’-nucleotidase. Regarding pyrimidine biosynthetic enzymes, nucleoside diphosphate kinase [Pf: AB851920, AB851929; Oo: AB851996, AB852013] and CTP synthetase [Pf: AB851957, AB851927, AB851931; Oo: AB851984, AB851993] were found in both transcriptomes, but nucleoside monophosphate kinase was detected only in Protobothrops [AB851951](Additional file 2: Table S4). All of these sequences were identical or nearly so to those reported by Rokyta et al. [62, 153].

Unfortunately, because both species in the present study are crotalids, the confirmation of nucleoside biosynthetic enzymes in the venome was less interesting than it might have been. The crotalid envenomation strategy involves liberation of endogenous prey purine nucleosides, but the venoms themselves have a minimal nucleoside content [1, 31]. In contrast, some viperid venoms and mamba venoms may contain nearly 9% purines by dry weight [31]. Thus in crotalid venomes, nucleoside biosynthetic enzymes probably are largely metabolic in function. It would be interesting to examine the transcript levels of these enzymes in Bitis or Dendroaspis venoms by comparison. Direct analysis of venom nucleoside levels would be required to determine what level of mRNA expression corresponds to a departure from metabolic function to envenomation.

Acid Phosphomonoesterase

Acid PME comprised a negligible percentage of all transcripts in both venoms (Additional file 2: Table S4 and Additional file 4: Table S5) [Pf: AB851930; Oo: AB851982]. The sequences were most closely related to a tissue PME from Anolis carolinensis. To the best of our knowledge, these are the first snake acid PME mRNA sequences reported.

Acetylcholinesterase

The Ovophis transcriptome included five acetylcholinesterase (AChE) transcripts that collectively amounted to less than the contaminant cutoff for venom gland transcripts, so its presence in the transcriptome may be accidental (Additional file 4: Table S5) [AB852000, AB852001, AB852002, AB851981, AB851973]. AChE activity is considered characteristic of most elapid, but not viperid venoms. AChE transcripts have been reported recently in selected colubrid and dipsadid venoms [11, 154]. These are the first reported crotalid transcripts.

Homologs of crotamine, GAP and crotasin

Crotamine, a highly basic 42-residue myotoxin was first reported 75 years ago [155] in the venom of Crotalus durissus terrificus. Homologs were later discovered in various other rattlesnake venoms [156–160]. These proteins display perplexing geographic distributional patterns [161] and individual quantitative variation [131], and they are products of duplicated loci [162]. Their physiological targets have remained controversial [163–169] and new biochemical activities continue to be discovered [170–176]. Myotoxin a, a crotamine homolog from the venom of Crotalus viridis viridis, was shown to undergo temperature-sensitive conformational transitions owing to cis-trans isomerization of Pro-20 [177, 178]. It is unknown whether the isomers bind to different physiological targets.

Marquardt et al. [179] patented a crotamine homolog called GAP (growth arresting peptide) with mitosis-arresting activity. It was isolated from the venom of Crotalus atrox, which, to date, has not been reported to contain a small myotoxin. GAP seems to have gone unnoticed by the toxinological community for the past 24 years, but crotasin, a crotamine homolog with some of the structural features of GAP was reported by Radís-Baptista et al. [180].

The present study isolated two GAP/crotasin-like transcripts from the Ovophis transcriptome (Figure 8) [AB852005, AB852006], but no crotamine- or crotasin-like sequence was found in the Protobothrops transcriptome. Crotasin/GAP-like proteins are significantly less basic than the crotamine-like proteins, and they lack a Phe-Pro dipeptide (crotamine residues 12-13), as well as the N-terminal Tyr of the latter. The two Ovophis transcripts differ very significantly from each other and from both GAP and crotasin (Figure 8). Though the exact location of the N-terminal residue cannot be determined with certainty, they both apparently possess the N-terminal disulfide bond present in crotamine and GAP, but absent in crotasin, and they are comparable in length to crotamine and GAP. Crotasin lacks the N-terminal eight residues of crotamine homologs. However, the signal peptide sequence for various crotamine isomers [181] exactly matches the signal peptide sequences of our Ovophis crotasin/GAP homologs. Both Ovophis transcripts manifested near-zero transcription levels, so it seems unlikely that these are functional venom components, but it is clear that the sequence diversification that Oguiura et al. [182] reported, applies to these transcripts as well.

Waprins

Waprins belong to a family of proteins with diverse activities that are structurally related to whey acidic protein [183]. Other members of the family have anti-bacterial activity and protease inhibitory activity [184, 185]. Waprins discovered to date are small proteins of about 50 amino acids, containing four disulfide bonds [186]. Clauss et al. [187] identified a segment of human chromosome 20, displaying 14 genes for proteins related to whey acidic protein. They postulated that the resulting gene products could potentially serve an anti-microbial function against pathogenic bacteria, or that they might participate in the regulation of endogenous proteases. They also opined that kallikrein-like proteases are of particular interest.

The protease inhibitory capacity of members of this family suggests possible roles in envenomation, though to date, no evidence has been presented for any of these functions. Snake venom proteins belonging to the Kunitz/BPTI family have been modified to serve as ion channel inhibitors [188] and to chaperone neurotoxic PLA₂s [189]. BPPs inhibit angiotensin I-converting enzyme to promote hypotension [83, 190], but also may act directly upon other physiological targets to induce hypotension [191, 192]. Some of the bradykinin-potentiating peptides serve an interesting dual role by inhibiting hemorrhagic metalloproteases in the venom gland [193].

Pahari et al. [144] reported the first viperid waprin-like protein in the venom glands of Sistrurus catenatus edwardsi. However, the putative Sistrurus toxin comprised a waprin domain fused to a Kunitz/BPTI domain. The function of the encoded protein is unknown. It was represented by only a single transcript, so it is difficult to say whether this toxin is biologically significant. This non-enzymatic toxin was expressed at near-zero levels.

Rokyta et al. [62] reported a full-length waprin transcript in the venom of Crotalus adamanteus. Both the Protobothrops and Ovophis transcriptomes contained transcripts that were strongly homologous to the Crotalus waprin [Pf: AB851939; Oo: AB852011] (Figure 9). Interestingly, the Ovophis waprin has a C-terminal Pro-Met, instead of the usual Pro-Leu/Val-Pro. One peptide representing 28% of the transcript sequence was isolated (Additional file 3: Table S2).

Both venoms also contained sequences that are related to the Kunitz serine protease inhibitor domain of the novel ku-wap hybrid toxin from Sistrurus catenatus edwardsi venom [62] (Figure 9). All of these transcripts are incomplete and the three N-terminal transcripts show relatively little overlap with the region of fusion in the Sistrurus ku-wap toxin; however, all three of the putative ku-wap homologs show the acidic and basic residues (positions 83-84) and other features of the Kunitz domain of the Sistrurus toxin (Figure 9). They do not show strong homology to either α-dendrotoxin [194] or to bovine pancreatic trypsin inhibitor (Figure 9). They may be additional examples of the ku-wap family; however, they appear to be most closely related to vertebrate inhibitors of the tissue factor pathway.

Putative inhibitors of tissue factor pathway

In vertebrates, blood coagulation is initiated by the tissue factor (TF) pathway. This pathway is regulated primarily by tissue factor pathway inhibitor (TFPI), a Kunitz serine protease inhibitor that inhibits Factor Xa and thrombin at concentrations as low as 2.5 nM, thus controlling the generation of thrombin and ultimately, of fibrin [195, 196]. Platelet TFPI is believed to modulate intravascular coagulation [197].

The Protobothrops transcriptome contained a single, partial transcript [AB851921] and the Ovophis transcriptome contained two, very short, identical transcripts [AB851997, AB851998] that align well with a predicted Anolis TFPI, and less well with the Ku-Wap fusion toxin from Sistrurus catenatus edwardsi venom glands and with bovine pancreatic trypsin inhibitor (Figure 10; Additional file 2: Table S4 and Additional file 4: Table S5). The Protobothrops TFPI transcript aligns well with both the acidic N-terminus and the highly basic C-terminus of human TFPIα [198] (Figure 10).

All three transcripts are expressed at vanishingly low levels (~0.001% of all transcripts) and it seems extremely unlikely that they function in envenomation; however, peptides ranging from 6.3% to 11.9% of the Protobothrops and Ovophis sequences were isolated. Most likely, these are tissue transcripts related to snake vascular homeostasis. If they serve any additional roles, they might inhibit venom SPs in the gland, or they might inhibit prey thrombin, allowing venom SPs to clot fibrinogen improperly, resulting in its rapid clearance by the prey's anti-clotting cascade.

Paraoxonase

Paraoxon hydrolytic activity has been reported only in the venom of Daboia russellii to date [199]. Venoms of Naja naja, Crotalus adamanteus, and Agkistrodon contortrix contortrix showed only trace level activity by comparison. Three genes comprise the paraoxonase gene family in humans. PON1 is largely associated with high-density lipoprotein, but has organophosphatase, arylesterase, or lactonase activities, and it hydrolyzes a wide array of substrates [200]. PON2 and PON3 are not well studied, but PON2 is known to be a widely-distributed cellular enzyme. Two transcripts were found in the Protobothrops transcriptome, but none in Ovophis. Both Protobothrops transcripts were expressed at near-zero levels, suggesting that paraoxonase is not a venom component in either of these species. The Protobothrops paraoxonase isozymes share diagnostic residues with all three human isozymes and are not clearly related to any one of them [AB851924, AB851925].

Vespryns

Pung et al. [201] isolated a novel 12 kDa toxin from the venom of the king cobra that acts centrally to induce hypolocomotion and pain in mammalian prey. A toxin from Lachesis muta venom [202] was the first crotalid vespryn and a second was sequenced from Crotalus adamanteus venom [62]. The Protobothrops transcriptome contained a partial, 70-residue vespryn transcript [AB851949], but the Ovophis transcriptome had none (Additional file 16: Figure S9). No vespryn peptides were sequenced. The Protobothrops vespryn is most closely related to that from Lachesis, which also displays a four-residue gap from positions 25-28. Only three of the first 70 residues differ between these two toxins. The three crotalid vespryns are all 28-32 residues longer at the N-terminus than the two corresponding toxins from Ophiophagus hannah and Pseudechis australis venoms [203].