Comprehensive Snake Venomics of the Okinawa Habu Pit Viper, Protobothrops flavoviridis, by Complementary Mass Spectrometry-Guided Approaches

The Asian world is home to a multitude of venomous and dangerous snakes, which are used to induce various medical effects in the preparation of traditional snake tinctures and alcoholics, like the Japanese snake wine, named Habushu. The aim of this work was to perform the first quantitative proteomic analysis of the Protobothrops flavoviridis pit viper venom. Accordingly, the venom was analyzed by complimentary bottom-up and top-down mass spectrometry techniques. The mass spectrometry-based snake venomics approach revealed that more than half of the venom is composed of different phospholipases A2 (PLA2). The combination of this approach and an intact mass profiling led to the identification of the three main Habu PLA2s. Furthermore, nearly one-third of the total venom consists of snake venom metalloproteinases and disintegrins, and several minor represented toxin families were detected: C-type lectin-like proteins (CTL), cysteine-rich secretory proteins (CRISP), snake venom serine proteases (svSP), l-amino acid oxidases (LAAO), phosphodiesterase (PDE) and 5′-nucleotidase. Finally, the venom of P. flavoviridis contains certain bradykinin-potentiating peptides and related peptides, like the svMP inhibitors, pEKW, pEQW, pEEW and pENW. In preliminary MTT cytotoxicity assays, the highest cancerous-cytotoxicity of crude venom was measured against human neuroblastoma SH-SY5Y cells and shows disintegrin-like effects in some fractions.


Introduction
Since ancient times, people have been fascinated by snakes and ascribed to them a diverse set of properties and character traits. Especially, in mythologies from South America through ancient Egypt to the Asian world, snakes and snake-like creatures represent both good and evil [1][2][3][4]. Most of them appear as symbols of wisdom and protection, and healing aspects have often been attributed to them [2]. Even today, the Aesculapius, the snake-wrapped rod of the Hellenic god Asclepius, and the winged Caduceus of Hermes symbolize medical, veterinarian and pharmacological professions [3]. On the other hand, based on encounters with humans, snakes are also known and feared for their bites and their possible consequences. Most accidents with snakes were registered in the tropics and mainly flavocetin-A [34,35]. Recent studies have shown that flavocetin-A, as a well-known protein, also inhibits the collagen-binding α2β1 integrin, which is the main receptor of platelets and necessary for platelet and cell activation [33,36].
Previous studies on the venom of P. flavoviridis were transcriptomic approaches and proteomic studies, yet limited to a shot-gun analysis, which could not provide a detailed picture of the venom composition [31,37]. Here, we report the first quantitative analysis of the P. flavoviridis venom, by a combination of diverse mass spectrometric methods, to give a more accurate profile of the venom and its composition. Therefore, a combined proteomic approach of bottom-up (BU) and top-down (TD) mass spectrometry (MS), including intact mass profiling (IMP), was used for snake venomic analyses, thus ensuring a high annotation coverage of the whole venom [38][39][40].

Top-Down Analysis
In a first top-down (TD) analytical run, the venom of P. flavoviridis was analyzed by a venomic workflow to quickly identify its native peptides and proteins. The application of the intact mass profiling (IMP) revealed 80 different molecular masses ( Figure 1, Table S1). Proteins could be detected up to a size of~31 kDa: 9 proteins in a range of 13-15 kDa and 6 in the range of 21-31 kDa. Dominant mass signals of small peptides at m/z 430-617 were observed, at early gradient retention times, in peaks 1-7. The small molecular determinants could be annotated manually and identified as members of the bradykinin-potentiating peptides-related peptides (BPP-RP) (Figures S1-S9) [41,42]. The peaks 8-15 mainly exhibited masses of 7-8 kDa, putatively identified as disintegrins (DI) by IMP and later confirmed by BU annotation. While the components with molecular masses of~14 kDa, including peaks 16, 17, 18 and 22, were identified as PLA 2 s, those with molecular masses of 21-28 kDa were suspected as members of the svSP and CTL family (Table S1).
Molecules 2018, 23, x FOR PEER REVIEW 3 of 18 collagen-binding α2β1 integrin, which is the main receptor of platelets and necessary for platelet and cell activation [33,36]. Previous studies on the venom of P. flavoviridis were transcriptomic approaches and proteomic studies, yet limited to a shot-gun analysis, which could not provide a detailed picture of the venom composition [31,37]. Here, we report the first quantitative analysis of the P. flavoviridis venom, by a combination of diverse mass spectrometric methods, to give a more accurate profile of the venom and its composition. Therefore, a combined proteomic approach of bottom-up (BU) and top-down (TD) mass spectrometry (MS), including intact mass profiling (IMP), was used for snake venomic analyses, thus ensuring a high annotation coverage of the whole venom [38][39][40].

Top-Down Analysis
In a first top-down (TD) analytical run, the venom of P. flavoviridis was analyzed by a venomic workflow to quickly identify its native peptides and proteins. The application of the intact mass profiling (IMP) revealed 80 different molecular masses ( Figure 1, Table S1). Proteins could be detected up to a size of ~31 kDa: 9 proteins in a range of 13-15 kDa and 6 in the range of 21-31 kDa. Dominant mass signals of small peptides at m/z 430-617 were observed, at early gradient retention times, in peaks 1-7. The small molecular determinants could be annotated manually and identified as members of the bradykinin-potentiating peptides-related peptides (BPP-RP) (Figure S1-S9) [41,42]. The peaks 8-15 mainly exhibited masses of 7-8 kDa, putatively identified as disintegrins (DI) by IMP and later confirmed by BU annotation. While the components with molecular masses of ~14 kDa, including peaks 16, 17, 18 and 22, were identified as PLA2s, those with molecular masses of 21-28 kDa were suspected as members of the svSP and CTL family (Table S1). Figure 1. Total ion chromatogram of the P. flavoviridis venom for IMP and TD. The total ion counts in P. flavoviridis crude venom were measured by the HPLC-ESI-MS of native crude venom. The relative abundance was set to 100% for the highest peak. The peak nomenclature is based on the chromatogram fractions (shown in Figure 2). The identified molecular masses of intact peptides and proteins are listed Table S1. Accordingly, the non-reduced venom of P. flavoviridis revealed 9 different proteins belonging to five toxin families: PLA2, svMP, DI, BPP-RP and L-amino acid oxidases (LAAO) (Table S1). Interestingly, most sequences were identified as fragments belonging to svMPs, which, however, were attributed to self-digestion, e.g., by metalloproteinases, an effect described previously in vitro for bothropasin and brevilysin H6 [43]. Total ion chromatogram of the P. flavoviridis venom for IMP and TD. The total ion counts in P. flavoviridis crude venom were measured by the HPLC-ESI-MS of native crude venom. The relative abundance was set to 100% for the highest peak. The peak nomenclature is based on the chromatogram fractions (shown in Figure 2). The identified molecular masses of intact peptides and proteins are listed Table S1.
Three different BPP-related peptides were annotated, containing a characteristic pyro-glutamylated N-terminus (Table S1). Additionally, five protein masses were detected as full-length proteins. For each of these, a disintegrin and the four PLA 2 s were annotated, with different modifications (Table S1). Since these mass differences to the expected amino acid sequence were indicated as parts of a longer fragment and did not have a distinct position, we assume the identification of closely related isoforms.
In order to further increase the number of annotations and assignments achieved so far, a chemically reduced venom sample was measured ( Figure S10). The reduction of cystines, which are important post-translational modifications (PTMs) in snake venoms, breaks up the tertiary structures and leads to a better fragmentation and de novo sequencing [39]. By means of the reduced TD approach, we detected, among other fragments, proteoforms of two 14 kDa toxins. One CTL, with a monoisotopic molecular mass of 14,400.38 Da, was identified as a 31.11 Da lighter variant of the so-called IX/X-BP CTL. The second mass (13,921.36 Da) belongs to the PLA 2 family and was annotated as a proteoform of the basic phospholipase, A 2 PL-X (13,971.41 Da). The reductive workup of venom with tris(2-carboxyethyl)-phosphine (TCEP) ultimately leads to conformational changes in the proteins and thus to differences in retention time. Therefore, a peak assignment, in comparison to the native TD total ion chromatogram (TIC) nomenclature, was not possible.
Except for the BPP-RP, other isoforms of venom proteins were only detectable as fragments. It is worth mentioning that, of the 73 TD assignments of the reduced venom, only three were annotated as internal fragments, while 54 belong to the N-terminal end, and only 13 belong to the C-terminal end of the compared sequences. The de novo sequences were assigned to seven toxin families ( Table 1). The high number of observed fragments could be an effect of the aforementioned digestion by metalloproteinases, and thus an accurate peak annotation, in correlation with the native venom TD, was impeded.

Bottom-Up Analysis
The TD approach gave a first and quick overview of lower molecular mass peptides and proteins, even of less prominent components, as constituents of the venom. A severe limitation of the TD approach, however, is that proteins beyond a molecular mass of~30 kDa are hardly detectable. This requires complementary analytics by a bottom-up approach: HPLC fractionation (Figure 2A) of the venom is followed by SDS-PAGE ( Figure 2B) and a tryptic in-gel digestion of protein bands. The subsequent MS de novo sequencing and semi-quantitative analysis led to the identification of the following toxin families: 55.1% phospholipases A 2 (PLA 2 ), 31.3% snake venom metalloproteinases and disintegrins (svMP/DI), 2.8% C-type lectin-like proteins (CTL), 1.8% cysteine-rich secretory proteins (CRISP), 1.4% snake venom serine proteases (svSP), 0.7% L-amino acid oxidases (LAAO), 0.07% phosphodiesterase (PDE) and 0.02% 5 -nucleotidase (5 -N). While 6.4% of the venom was assigned to peptides, 0.3% could not be annotated (n/a) ( Figure 3).  The TD approach gave a first and quick overview of lower molecular mass peptides and proteins, even of less prominent components, as constituents of the venom. A severe limitation of the TD approach, however, is that proteins beyond a molecular mass of ~30 kDa are hardly detectable. This requires complementary analytics by a bottom-up approach: HPLC fractionation (Figure 2A) of the venom is followed by SDS-PAGE ( Figure 2B) and a tryptic in-gel digestion of protein bands. The subsequent MS de novo sequencing and semi-quantitative analysis led to the identification of the following toxin families: 55.1% phospholipases A2 (PLA2), 31.3% snake venom metalloproteinases and disintegrins (svMP/DI), 2.8% C-type lectin-like proteins (CTL), 1.8% cysteine-rich secretory proteins (CRISP), 1.4% snake venom serine proteases (svSP), 0.7% L-amino acid oxidases (LAAO), 0.07% phosphodiesterase (PDE) and 0.02% 5′-nucleotidase (5′-N). While 6.4% of the venom was assigned to peptides, 0.3% could not be annotated (n/a) ( Figure 3). According to the above findings, from components of the venom, the phospholipase A 2 class formed the biggest part and includes the three most abundant proteins in the whole venom: The highest protein content belongs to the acidic PLA 2 1 (16.6%), eluting as fraction 22. Secondly, the basic PLA 2 BP (10.4%) was identified in fraction 16 but could not be clearly assigned as PLA 2 basic protein I or II (BPI, BPII), because of their high similarity, which differs only in an N58D exchange as a single mutation [44]. The third protein is PLA-N(O) (7.9%) of fraction 17, which is a PLA-N K121N isoform (Uniprot-ID: S6BAM8), previously documented with respect to the P. flavoviridis population on Okinawa Island [14]. According to the above findings, from components of the venom, the phospholipase A2 class formed the biggest part and includes the three most abundant proteins in the whole venom: The highest protein content belongs to the acidic PLA2 1 (16.6%), eluting as fraction 22. Secondly, the basic PLA2 BP (10.4%) was identified in fraction 16 but could not be clearly assigned as PLA2 basic protein I or II (BPI, BPII), because of their high similarity, which differs only in an N58D exchange as a single mutation [44]. The third protein is PLA-N(O) (7.9%) of fraction 17, which is a PLA-N K121N isoform (Uniprot-ID: S6BAM8), previously documented with respect to the P. flavoviridis population on Okinawa Island [14].
The svMP and DI members were detected at early retention times (HPLC fractions 4 and 5), but also formed the dominating protein classes at later retention times (fractions 27, 28, 32) (Table S1). Because svMP could include DI domains, both toxin families were combined in the final composition as the svMP/DI part, thus forming the second most abundant group. The annotated sequences from fraction 4-13 belong to svMP P-II disintegrin domains, but with ~15 kDa, the observed molecular masses, determined from the SDS gel, appear too low for svMP P-II (expected molecular masses 30-60 kDa) and too high for single disintegrins (7)(8). It is conceivable that the identified disintegrin domains originate from degraded P-II or P-III metalloproteinases and contain neighboring sequence parts. This could mean that the Habu venom contains truncated or auto-digested versions of svMP P-II or P-III, which which is also known from other svMPs [43]. This autolysis was firstly observed for the two svMPs, HR1A and HR1B, isolated from the P. flavoviridis venom [45,46].
The CRISPs represent a minor part of the venom (1.8%) and were only found in fractions 20 and 21. The BU-annotated triflin revealed, in the IMP, a molecular mass of 24,767 Da, with a −16 Da shift to the expected 24,783 Da. The transcriptomic data of P. flavoviridis include another CRISP (Uniprot-ID: T2HP25), whose amino acid sequence differs from triflin (Uniprot-ID: Q8JI39) in two mutations: D110N and I113V. This isoform has an average molecular mass of 24,768 Da and would correspond to our observation. To our knowledge, this proteoform, herein termed triflin-II, has not yet been described at the proteomic level.
Another protein, validated by its molecular mass, is the flavoxobin (Uniprot-ID: P05620), which belongs to the low abundant svSP family. This thrombin-like protease is the main part of fraction 25 and, with 0.8%, represents more than half of the svSP venom content, followed by svSP2 (Uniprot-ID: O13057) with 0.3%. In summary, the combination of three mass spectrometric methods, BU, TD and IMP, led to the annotation of specific isoforms (Table 2). The svMP and DI members were detected at early retention times (HPLC fractions 4 and 5), but also formed the dominating protein classes at later retention times (fractions 27, 28, 32) (Table S1). Because svMP could include DI domains, both toxin families were combined in the final composition as the svMP/DI part, thus forming the second most abundant group. The annotated sequences from fractions 4-13 belong to svMP P-II disintegrin domains, but with~15 kDa, the observed molecular masses, determined from the SDS gel, appear too low for svMP P-II (expected molecular masses 30-60 kDa) and too high for single disintegrins (7)(8). It is conceivable that the identified disintegrin domains originate from degraded P-II or P-III metalloproteinases and contain neighboring sequence parts. This could mean that the Habu venom contains truncated or auto-digested versions of svMP P-II or P-III, which which is also known from other svMPs [43]. This autolysis was firstly observed for the two svMPs, HR1A and HR1B, isolated from the P. flavoviridis venom [45,46].
The CRISPs represent a minor part of the venom (1.8%) and were only found in fractions 20 and 21. The BU-annotated triflin revealed, in the IMP, a molecular mass of 24,767 Da, with a −16 Da shift to the expected 24,783 Da. The transcriptomic data of P. flavoviridis include another CRISP (Uniprot-ID: T2HP25), whose amino acid sequence differs from triflin (Uniprot-ID: Q8JI39) in two mutations: D110N and I113V. This isoform has an average molecular mass of 24,768 Da and would correspond to our observation. To our knowledge, this proteoform, herein termed triflin-II, has not yet been described at the proteomic level.
Another protein, validated by its molecular mass, is the flavoxobin (Uniprot-ID: P05620), which belongs to the low abundant svSP family. This thrombin-like protease is the main part of fraction 25 and, with 0.8%, represents more than half of the svSP venom content, followed by svSP2 (Uniprot-ID: O13057) with 0.3%. In summary, the combination of three mass spectrometric methods, BU, TD and IMP, led to the annotation of specific isoforms ( Table 2). In contrast to the proteomic data presented herein, the Habu is already known to have two transcriptomic compositions from an mRNA analysis of the venom glands [31,37]. They exhibit small analogies between the protein abundances and the fragments per kilobase million (FPKM %) of the main toxin families, but also remarkable differences regarding the proportional distribution ( Figure 4). Both transcription analyses (T1, T2 in Figure 4) show PLA 2 as the highest expressed gene family, followed by svMP, which coincides with our proteomic data (P* in Figure 4). On the other hand, in T1, the PLA 2 s form~30%, while the amounts of P and T2 are comparable (~55%). With 17.3%, the svMP gene expressions of T2 are lower than T1 and the protein amounts, identified in the proteome (~30%). Interestingly, the proteomic and transcriptomic compositions are correlated in less represented groups: CRISP (2-4%) and PDE (0.2-0.1%). In the other three families (CTL, svSP, LAAO), our measured proteomic levels are much lower than the mRNA level. The data for CTL show an expression up to five-fold and, for svSP, eight-fold, higher than that of the protein levels found. There is a considerable difference between the observed 0.7% of LAAOs in our study and T1 as well as T2, which show a four to 13-fold larger amount, accounting for up to 9.1% of the complete transcriptome. Due to the fact that modern analysis can detect small amounts of RNA, very low concentrated targets can usually be observed, but are not necessarily visible in proteomic approaches. All three studies identified PDEs and 5 -nucleotidases, with abundances of >0.2%. In the case of T1 and T2, further families, in a range of 0.65-0.01%, were observed, but not in the venomic proteome P: Galactose-binding lectins, nerve growth factors, phospholipases B, glutaminyl cyclases, vascular endothelial growth factor-like proteins, as well as less abundantly detected families, with <0.01% of the total transcript abundance [31,37]. This underlines that reflecting the individual proteomic and transcriptomic compositions could not easily be compared with independent venom analyses of a species. This has also been shown in studies on other members of Viperidae and Elapidae [47][48][49]. However, these two approaches in combination mutually represent a powerful aid for protein annotation and identification. The asterisked data set P is the result of this study. "Other" represents, in all three P. flavoviridis data, phosphodiesterases, 5′-nucleotidases and components that were not annotated. Additionally, in the case of T1 and T2, galactose-binding lectins, nerve growth factors, phospholipases B, glutaminyl cyclases, vascular endothelial growth factor-like proteins, as well as less abundantly detected families, with <0.01%, are included. For P. mucrosquamatus (P3, P. mucro.), "other" represents trimucrotoxin, and, for T. stejnegeri (P4, T. stejn.), "other" represents snake venom vascular endothelial growth factors. The origins of toxin ratios are marked alphabetically: a [37], b [31], c and d [50]. The phylogentic relationship is based on [15,51].
The venom of two related pit vipers show analogies in their composition to P. flavoviridis. This concerns Protobothrops (or Trimeresurus) mucrosquamatus (P. mucrosquamatus), named the Taiwan Habu, as well as the Trimeresurus stejnegeri (T. stejnegeri) (indicated with P3, P4 in Figure 4) [15,51]. As for P. flavoviridis, they can be found on island habitats, e.g., Taiwan, in the south-west of Okinawa at the end of the Japanese island chain, and the P. mucrosquamatus was also observed directly on Okinawa [15,51,52]. Like the Habu from Japan, these two vipers belong to the medically important venomous snakes in their Taiwanese habitat and are responsible for significant envenomations and deaths over the last decades [53,54].
For all three snakes, the main venom families are svMPs and PLA2s, followed by CTLs, CRISPs, svSPs, and LAAOs with abundances of <15%. While of the above-mentioned P. flavoviridis venom (P* in Figure 4) comprises a big portion, i.e., >55% PLA2s and >30% svMPs, the venoms of P3 and P4 are alike to each otther. The main protein share of both venoms (P3 and P4 in Figure 4) is contributed by svMP, with >40%, while the PLA2s (~25%) only constitute the second most abundant protein component and thus stand in contrast to P*. The lesser protein families are twofold more abundant in P3 and P4, and reflect a broader diversity of the venomous components in theses snakes. Particularly, the svSP with >10% seemed to be more abundant in the P. mucrosquamatus and T. stejnegeri, than in the P. flavoviridis.

Bradykinin-Potentiating Peptides and Snake Venom Metalloproteinase Inhibitors
Besides the previously mentioned protein families, various bradykinin-potentiating peptides (BPP) and snake venom metalloproteinases inhibitors (svMP-i) are further constituents of snake venoms. The strong vasoactive effect of bradykinin, a substrate of the angiotensin-converting enzyme, was discovered in the late 1940's in studying the Bothrops jararaca venom, and this discovery indicates that research on snake venoms can lead to impressive developments in the drug development field [55,56]. The identification of a small peptide in the same venom, which increased the effects of Kinin, was the first BPP [57]. This facilitated the design of hypertension drugs based on and two venom gland transcriptomic analyses of P. flavoviridis (T1 and T2) in fragments per kilobase million (FPKM %). The asterisked data set P is the result of this study. "Other" represents, in all three P. flavoviridis data, phosphodiesterases, 5 -nucleotidases and components that were not annotated. Additionally, in the case of T1 and T2, galactose-binding lectins, nerve growth factors, phospholipases B, glutaminyl cyclases, vascular endothelial growth factor-like proteins, as well as less abundantly detected families, with <0.01%, are included. For P. mucrosquamatus (P3, P. mucro.), "other" represents trimucrotoxin, and, for T. stejnegeri (P4, T. stejn.), "other" represents snake venom vascular endothelial growth factors. The origins of toxin ratios are marked alphabetically: a [37], b [31], c and d [50]. The phylogentic relationship is based on [15,51].
The venom of two related pit vipers show analogies in their composition to P. flavoviridis. This concerns Protobothrops (or Trimeresurus) mucrosquamatus (P. mucrosquamatus), named the Taiwan Habu, as well as the Trimeresurus stejnegeri (T. stejnegeri) (indicated with P3, P4 in Figure 4) [15,51]. As for P. flavoviridis, they can be found on island habitats, e.g., Taiwan, in the south-west of Okinawa at the end of the Japanese island chain, and the P. mucrosquamatus was also observed directly on Okinawa [15,51,52]. Like the Habu from Japan, these two vipers belong to the medically important venomous snakes in their Taiwanese habitat and are responsible for significant envenomations and deaths over the last decades [53,54].
For all three snakes, the main venom families are svMPs and PLA 2 s, followed by CTLs, CRISPs, svSPs, and LAAOs with abundances of <15%. While of the above-mentioned P. flavoviridis venom (P* in Figure 4) comprises a big portion, i.e., >55% PLA 2 s and >30% svMPs, the venoms of P3 and P4 are alike to each otther. The main protein share of both venoms (P3 and P4 in Figure 4) is contributed by svMP, with >40%, while the PLA 2 s (~25%) only constitute the second most abundant protein component and thus stand in contrast to P*. The lesser protein families are twofold more abundant in P3 and P4, and reflect a broader diversity of the venomous components in theses snakes. Particularly, the svSP with >10% seemed to be more abundant in the P. mucrosquamatus and T. stejnegeri, than in the P. flavoviridis.

Bradykinin-Potentiating Peptides and Snake Venom Metalloproteinase Inhibitors
Besides the previously mentioned protein families, various bradykinin-potentiating peptides (BPP) and snake venom metalloproteinases inhibitors (svMP-i) are further constituents of snake venoms. The strong vasoactive effect of bradykinin, a substrate of the angiotensin-converting enzyme, was discovered in the late 1940's in studying the Bothrops jararaca venom, and this discovery indicates that research on snake venoms can lead to impressive developments in the drug development field [55,56]. The identification of a small peptide in the same venom, which increased the effects of Kinin, was the first BPP [57]. This facilitated the design of hypertension drugs based on the structure of a snake toxin structure [58]. Today, a multitude of different snake BPPs are known, and, with the progress in the field of venomics, this number is still increasing.
By means of TD and IMP analytics, in total, we identified 5 different BPP-RP bearing an N-terminal pyro-glutamate (pE): pEQWMPGGRPPHHIPP ( Figure S1) and pESKPGRSPPISP ( Figure S2). Until now, the presence of both peptides was only hypothesized by transcriptomic data on P. flavoviridis [37,59].
In the peak of the BPP, pESKPGRSPPISP (peak 5, Figure S1), by means of MS/MS, two C-terminally truncated versions were identified: The 11mer peptide, pESKPGRSPPIS ( Figure S3), and the 10mer, pESKPGRSPPI ( Figure S4). Comparable to the peptide, pEQWMPGGRPPHHIPP, the pEQWSQGRPR peptide ( Figure S5) of peak 2 shows a trimeric N-terminal sequence (pEQW), which, as a tripeptide, is known for its inhibition of svMP.
To minimize the risk of a self-degradation by metalloproteinases, also present in high concentrations in the herein studied venom, snakes secrete small trimeric peptides. These svMP inhibitors are processed from the same precursor, like the BPPs, and also contain an N-terminal pyroglutamate [60]. In summary, we could identify three different svMP-is that represent the main components of the TIC: Peak 4 (pEKW, m/z 444.22), peak 6 (pENW, m/z 430.17) and peak 7 (pEQW, m/z 444.18) (Figures S6-S8). Another prominent molecular mass signal beneath the main peak, m/z 444.18, was detected at m/z 427.13 and could be identified as pEEW, which is the deaminated form of peak 7 pEQW ( Figure S9). Previously, these three svMP-is were further isolated from the closely related Taiwanese Habu and revealed their strong inhibitory activity [61].

Cytotoxicity Test
It is well-known that the snake venom of P. flavoviridis causes strong cytotoxic effects and various isolated toxins, e.g., LAAO Okinawa Habu apoxin protein-1 (OHAP-1) in glioma cells [62,63], exhibiting in vivo apoptotic activities. The Habu venom was monitored in this study against several human cancer cell lines and, therefore, the cytotoxicity was determined for cancerous (SH-SY5Y, MDA-MB-231, A549, PANC1, HeLa, PC-3) as well as non-cancerous (HEK293) cells by the MTT assay. The IC 50 values range from~1 to >50 µg/mL (Table 3). The highest inhibition of proliferation was found against HEK293 (1.02 µg/mL) and SH-SY5Y (4.7 µg/mL) cells ( Figure S11). SH-SY5Y, as the most sensitive cancer cell line, has been selected for further screenings with single RP-HPLC venom fractions. Regarding the most abundant molecular mass in a collected fraction, we indicated the IC 50 in µg/mL and µM. While various tested fractions were found to be active, fractions 4 and 7/8 were the most potent, with IC 50 values of 0.7 and 0.9 µg/mL ( Table 4). The main compound of the fraction 4 is the svMP-i pEKW, while the pEQW containing fraction 7/8 is similarly potent. According to these results, the P. flavoviridis venom, svMP-is, exhibits remarkable effects on SH-SY5Y cells, while, until now, only the correlating svMP of most Protobothrops venoms were known to have an important role in envenomation-related pathologies [46,64]. Regarding the identified families, PLA 2s , as the main venom part, are most active in fraction 15 (PLA 2 , 2.2 µg/mL), in combination with a CTF-II-like disintegrin, against neuroblastoma cells. An induced caspase-independent apoptosis by another PLA 2 (BP-II) in a leukemia cell was previously shown [63]. The PLA 2 fractions (16, 17, 18, and 19) as well as the CRISP triflin-II fraction 20 revealed a moderate growth inhibition (16 to 30 µg/mL) with respect to the mass concentration. Table 4. IC 50 values of P. flavoviridis venom fractions against SH-SY5Y cells. Single RP-HPLC venom fractions of P. flavoviridis were tested against human neuroblastoma SH-SY5Y cells, the half maximal inhibitory concentrations (IC 50 ) were determined in µg/mL, and the molar concentration in µM, regarding the most abundant molecular mass, were determined in Da. Doxorubicin was used as a reference and error mean in ±SD.   (Table S1). Future studies will focus on the mechanism of the P. flavoviridis venom action in SH-SY5Y cells, notably due to DI and svMP-i. induced caspase-independent apoptosis by another PLA2 (BP-II) in a leukemia cell was previously shown [63]. The PLA2 fractions (16, 17, 18, and 19) as well as the CRISP triflin-II fraction 20 revealed a moderate growth inhibition (16 to 30 μg/mL) with respect to the mass concentration. Table 4. IC50 values of P. flavoviridis venom fractions against SH-SY5Y cells. Single RP-HPLC venom fractions of P. flavoviridis were tested against human neuroblastoma SH-SY5Y cells, the half maximal inhibitory concentrations (IC50) were determined in μg/mL, and the molar concentration in μM, regarding the most abundant molecular mass, were determined in Da. Doxorubicin was used as a reference and error mean in ±SD.   Figure S12). Fractions 7/8 and 11-15 indicate, against SH-SY5Y cells at higher concentrations, a disintegrine-like conglomeration effect ( Figure S12), which correlates with the observation of svMP/DI in these fractions by BU, IMP and the TD, like the disintegrin CTF-II proteoform (Table S1). Future studies will focus on the mechanism of the P. flavoviridis venom action in SH-SY5Y cells, notably due to DI and svMP-i.

Discussion
In this contribution, we report on the first quantitative mass spectrometry-guided proteomic snake venom analysis of the pit viper, P. flavoviridis, one of the most feared and life-threatening snakes in Japan. The combination of all three mass spectrometric methods (BU, TD and IMP) facilitates the annotation of isoforms, respectively based on databases even without the full sequencing of every protein. The combined data reveal that PLA 2 s (55.1%) comprise the major part of the venom, with PLA 2 1, PLA 2 BP I/II and PLA-N(O) as main representatives, as well as the svMP/DI group and several minor represented toxin families, like the svSP with flavoxobin and svSP2. For the first time, a CRISP triflin-homolog, named triflin-II, was observed at the proteomics level. The top-down approach to identify proteins as well as peptides was reliable for toxins with molecular masses up to~30 kDa. Venom compounds, in a range of m/z 338.1 to 30,384.7, were detected by the intact mass profiling, with the same restrictions for high masses as in the top-down approach, while the SDS-PAGE exhibits proteins over 100 kDa.
An analytical correlation of the proteomes, with two further pit viper venoms, shows that a close relationship and similar habitats are not predictive of the venom compositions of related species. This emphasizes the importance of an individual snake venom analysis to reveal the specific components, e.g., those found for the various PLA 2 s of P. flavoviridis. Furthermore, these finding may help the development of targeted snake bite therapies.
Interestingly, venom-peptides display various bioactivities, and their antiproliferative and cytotoxic properties may also contribute to therapies in the treatment of cancer. This is corroborated by the fact that several venom-based drugs are medically assessed, e.g., chlorotoxin, against different tumor cell types and the integrin αvβ3-targeted cancer therapy [65][66][67]. In this study, we screened fractions of the P. flavoviridis crude venom for cytotoxic effects against cancer cells. Accordingly, cytotoxicity assays of crude venom and growth inhibition by venom fractions revealed promising results against SH-SY5Y neuroblastoma cells. The observed effect of svMP-i tripeptide, containing fractions on cell growth inhibition, appears very interesting from a pharmacological viewpoint as well as for a future mode of action studies, and it reveals a new facet of the bioactivity of these small venom components. Moreover, we identified the three highly potent protein fractions, 14, 15 and 20, with IC 50 values, ranging between 0.16 and 0.63 µM, which are 2-10× fold lower than the cytostatic doxorubicin. These aspects, concerning the bioactivity of components of the P. flavoviridis venom on neuroblastoma cells, might be helpful in the development of future anti-cancer treatments.
Through the wide range quantification of the Okinawa habu venom, we could extend the picture of the Japanese venomous snakes, of which only a few detailed analyses have addressed the whole venomic level.

Sample Preparation and System Setup
The pooled Habu venom of six P. flavoviridis specimens (four females, two males) was purchased from the Kentucky Reptile Zoo (Slayde, KY, USA) and kindly provided by Professor Dr. Johannes A. Eble (University of Münster, Germany). The crude venom (with a final concentration of 10 mg/mL) was dissolved in 10 µL HFo (1% (v/v) and centrifuged at 20,000× g for 5 min. Then, 30 µL of citrate buffer (0.1 M, pH 4.3) was added. One half of the sample (20 µL) was chemically reduced by adding 10 µL of 0.5 M tris(2-carboxyethyl)-phosphine (TCEP) and incubated for 30 min at 65 • C, while 10 µL ultra-pure water was added to the other half, as a non-reduced/native sample. All samples were centrifuged at 20,000× g for 5 min and submitted to IMP (native) and TD venomics (native, reduced): HPLC-high-resolution (HR) ESI-MS/MS measurements were performed on a LTQ Orbitrap XL mass spectrometer (Thermo, Bremen, Germany) coupled to an Agilent 1260 HPLC system (Agilent, Waldbronn, Germany) using a Supelco Discovery 300 Å C18 (2 × 150 mm, 3 µm particle size) column. The elution was performed by a gradient of ultra-pure water, with 0.1% formic acid (HFo) (v/v; buffer A), and acetonitrile (ACN), with 0.1% HFo (v/v; buffer B), at a flow rate of 1 mL/min. An isocratic equilibration (5% B) for 5 min was followed by a linear gradient of 5-40% B for 95 min, 40-70% B for 20 min, 70% B for 10 min and a re-equilibration with 5% B for 10 min.
ESI settings were: 11 L/min sheath gas, 35 L/min auxiliary gas, spray voltage 4.8 kV, capillary voltage 63 V, tube lens voltage 135 V, and capillary temperature 330 • C. The data-dependent acquisition (DDA) mode was used for MS/MS experiments, with 1 µ scans and 1000 ms maximal fill time. The precursor ions were selected, with a range of ±2 m/z and after two repeats within 10 s, excluded with ±3 m/z for a duration of 20 s. Three scan events were performed, with a normalized collision-induced dissociation (CID) energy of 30% and 35%, and a higher-energy collisional dissociation (HCD) with 35% collision energy.

Intact Mass Profiling (IMP)
For IMP, the mass spectrometric data were inspected via the Xcalibur Qual Browser (Thermo Xcalibur 2.2 SP1.48, Waltham, MA, USA), and the deconvolution of isotopically resolved spectra was carried out using the XTRACT algorithm of Xcalibur Qual Browser. The protein assignment was done by comparison with the retention times obtained from the HPLC runs. Sequence annotations and molecular mass comparisons with protein database entries of P. flavoviridis (taxid: 88087) were performed manually.

Top-Down (TD) Venomics
The top-down analytical data were obtained based on the protocol of Petras et al., 2016 [68], with the following alterations: Data were inspected with the Qual Browser (Thermo Xcalibur 2.2 SP1.48) and prepared based on the TopPIC workflow. The .raw data were converted to a centroided .mzXML using the MSconvert of the ProteoWizard package (http://proteowizard.sourceforge.net), version 3.0.10577. The .mzXML data were deconvoluted to an .msalign file using MS-Deconv
The analytics of tryptic peptides were performed using a reversed-phase Grace Vydac 218MSC18 column (2.1 × 150 mm, 5 µm particle size) under the control of an Agilent 1260 HPLC system (Agilent Technologies, Waldbronn, Germany). The HPLC separation operated with a flow rate of 0.3 mL/min. After an isocratic equilibration (5% B) for 1 min, the peptides were eluted, with a linear gradient of 5-40% B for 10 min and 40-99% B for 3 min, washed with 99% B for 3 min and re-equilibrated in 5% B for 3 min.
MS experiments were performed on an Orbitrap XL mass spectrometer (Thermo, Bremen, Germany), with R = 15,000 at m/z 400 and at a maximum filling time of 200 ms for the first product ion scans. MS/MS fragmentation of the most intense ion was performed in the LTQ using a collision-induced dissociation (30 ms activation time); the collision energy was set to 30% and 35%. The precursor ions were selected, with a range of ±2 m/z, and, after two repeats within 10 s, excluded with ±3 m/z for a duration of 20 s.

Relative Toxin Quantification
The percentage composition of the venom ingredients was calculated on the basis of a combination of the RP-HPLC chromatogram and SDS-PAGE evaluation [71,72]. The peak integrals at UV 214nm were measured in comparison to the total sum of peak integrals. In the case of multiple component elution in an HPLC fraction identified by SDS-PAGE staining, the integrated density ratio of the stained bands was respectively used for the emphasis of peak integrals.

Data Accessibility
Mass spectrometry proteomics data (.mgf, .raw and output files) have been deposited at the ProteomeXchange Consortium [73] (http://proteomecentral.proteomexchange.org) via the MassIVE partner repository, under the project name "Venomics of the Okinawa Habu pit viper, Protobothrops flavoviridis," and the data set identifier, PXD009414.
A modified MTT assay was used to determine the cytotoxicity of the snake venom. Therefore, 1 × 10 5 cells/mL were seeded in a 96-well microtiter plate. After 24 h of cultivation, the cells were treated for 48 h at 37 • C with crude venom, with venom fractions or doxorubicin as positive cytotoxic control drugs.
The optical density (OD) was measured in triplicates at λ = 570 nm (with a reference wavelength of λ = 690 nm) by UV/Vis spectrophotometry (Thermo, Bremen, Germany). The cell viability was determined with an absorbance of A:

Morphological Studies
The morphological changes of the cells following treatment with crude venom or single RP-HPLC venom fractions of P. flavoviridis were observed under an inverted microscope (Olympus, Toyo, Japan) and compared to the control group following a 48 h treatment.

Half Maximal Inhibition of Growth (IC 50 ) Determination
The half maximal inhibition of growth (IC 50 ) was calculated based on a sigmoidal curve fitting using a four-parameter logistic model, as compared to that of untreated controls, which was calculated using Prism 5 software (GraphPad5, San Diego, CA, USA). Values are presented at a 95% confidence interval and as the average of three independent measurements.