Background & Summary

Natural toxins comprise a vast group of compounds that are diverse in their chemical nature, the most numerous of which are polypeptide molecules. As a rule, peptide toxins are present in animal venoms, and these venoms usually contain a diverse array of bioactive peptides affecting various receptor targets. A number of polypeptide toxins are currently used in fundamental research as biological tools for the study of various receptor systems and for the development of pharmaceutical agents used in medicine1–5. Among venomous animals, marine snails and spiders can be considered the leaders in venom component diversity6–10, with venoms containing up to several hundred distinct components. Taking into account that the number of spider species (~ 44,500) surpasses that of all other known species of venomous animals, spider venoms represent the largest naturally edited library of biologically active molecules, which is predicted to include more than several million compounds9,11. It is likely that peptide components with unique specialization to any given ionotropic receptor can be found in spider venoms12–15.

Peptidomic and proteomic approaches can be practically applied for the identification and characterization of venom components and as high-throughput screening techniques for bioactivity testing10,16. Such approaches make it possible to mine rare components in venoms that were not examined by earlier researchers and to determine their structure. In previous proteomic studies of Australian spider venoms, approximately 1,000 individual components for Hadronyche versuta and more than 600 for Atrax robustus were identified with a combination of chromatographic and off-line mass spectrometry procedures17,18. A total of 378 venom polypeptides were found in the venom of the spider Dolomedes sulfurous by means of off-line HPLC/MALDI-TOF-MS analysis19. The presence of 400 different compounds, many of which overlapped between species, was detected by means of mass spectrometry analysis for three related Brazilian species: Phoneutria nigriventer, Phoneutria reidyi, and Phoneutria keyserlingi20. Using a combination of a multidimensional chromatographic approach and tandem mass spectrometry, 286 components were identified in the venom of the spider Cupiennius salei21. However, the investigation of other spider venoms has not revealed such component variety. A proteomic analysis of the venom of the Chinese tarantula Haplopelma huwenum (Selenocosmia huwena) indicated the presence of 133 polypeptides22, and the venom of another Chinese spider Chilobrachys guangxiensis (Chilobrachys jingzhao), comprised 120 polypeptides with a molecular weight of 2,000–8,000 Da10. The venom of the tarantula Psalmopoeus cambridgei included 150 polypeptides23 and that of the tarantula Theraphosa leblondi included 65 polypeptides24 and was the least diverse venom reported among tarantulas. Furthermore, the Central Asian species Agelena orientalis25 and the South American Loxosceles intermedia26 included 21 and just over 30 components, respectively.

The diversity of polypeptides in a particular venom can be successfully estimated by gene sequencing. Until recently, the most popular procedure involved cDNA library construction generated on the basis of known toxin structures such as ω- HXTX -Hv1a11, several huwentoxins10,27, and hainantoxins28, but today, researchers pay more attention to EST techniques29–32. Parallel proteomic studies of the same spider venom serve as reliable proof for the analysis of EST data18,28,33–35. It should be noted that transcriptomic analyses generally identify fewer polypeptide components in spider venoms compared to proteomic analyses. Although the detection of more than 200 individual polypeptides is uncommon29,35, proteomic analyses usually establishes dozens of structures10,32,33,36. Therefore, these two approaches produce unequal predictions of polypeptide variety in spider venom. Apparently, this difference is the result of overestimation by the proteomics approach caused by a large number of false positive data.

Methods

Dolomedes fimbriatus (Clerck, 1,757) spiders were collected in the south European region of Russia. D. fimbriatus, of the family Pisauridae, is also known as the raft or fishing spider. The raft spider is semi-aquatic: it hunts on the water surface and can submerge itself if necessary. Venom glands were dissected from several specimens and frozen in liquid nitrogen until sample preparation. To obtain a sufficient amount of mRNA from the venom gland, a preliminary (one week) milking procedure was performed to activate massive toxin expression in accordance with a previous study31.

Total RNA was extracted with an SV Total RNA Isolation System (Promega, USA). The yield and purity were assessed using a Nanodrop ND-1,000 spectrophotometer (Thermo Scientific, USA), with the RNA integrity determined by the RNA Integrity Number (RIN) using a Bioanalyzer 2,100 (Agilent Technologies, USA). The PCR-based cDNA library was created following the instructions for the SMART cDNA library construction kit (Clontech, USA). Competent E. coli One ShotTOP10 cells (Invitrogen, USA) were transformed with the cDNA library plasmids to amplify the cDNA. Plasmid DNA was purified with alkaline lysis and sequenced in both directions using an ABI Prism 3730xl automatic DNA sequencer (Sanger technique) with the BigDye Terminator version 3.1 cycle sequencing kit (Applied Biosystems, USA).

Single-residue distribution analysis (SRDA)37,38 was used to search the polypeptide structures in the crude EST bank. Basic sequence transformation for the search was performed using the key residue Cys and the termination translation symbol -SRDA (‘C.’). The deduced proteins were retrieved from a translated database by 9 structural motifs that consider all the major structural features of spider venom polypeptides (Table 1).

Table 1 The structure of the search motifs used. The last column shows the impact of each motif on the number of overall sequences retrieved.
Full size table

The presence of some features specific to spider toxin precursor proteins was assessed to identify incorrect data with a high degree of reliability. Because toxin-like polypeptides are secretory proteins that are synthesized together with the preceding signal peptide, the deduced protein structure precursors were evaluated for the presence of the correct eukaryotic signal peptide using a Phobius predictor39. The second validation criterion was based on the specificity of a maturation process that is typical for cysteine-containing and linear spider venom polypeptides; this criterion is also known as the presence of the processing quadruplet motif (PQM)31,40. The last validation criterion was based on the 5’ and 3’-read identity of the clones in a first ATG to termination translation codon range. To determine the novelty of the identified structures, BLASTX was used against a non-redundant protein sequence database.

Data Records

The generated database includes 11,712 ESTs from 5,952 individual clones, with almost all of the clones being sequenced from both ends. Raw data were submitted to GenBank of National Center for Biotechnology Information after verification by VecScreen (http://www.ncbi.nlm.nih.gov/tools/vecscreen/) in according to dbEST sequence deposition requirements (Data Citation 1). The data after verification involved only the deduced sequences of secreted venom gland polypeptides. In total, 7,169 translated sequences were identified as possible polypeptide compounds of D. fimbriatus spider venom. Rejected sequences corresponded to partially defined polypeptides (fragments) or secreted molecules with enzymatic, regulatory, structural, and other functions that were beyond the goal of the investigation. Only one exception was made, for linear cysteine-free polypeptides called cytotoxins or antimicrobial peptides (AMPs) GO:0003795.

For comprehensive spider venom characterization, we performed step-by-step transcriptome processing that reckons the growth of the deduced polypeptide number to the size of the EST sequenced (Fig. 1a). We considered all clones coding for identical mature sequences as one polypeptide toxin contig to estimate the amount of compounds in the natural venom. A total of 163 trusted sequences with structures confirmed by at least two repeats according to our results could be completely determined, even from 10,000 sequences (curve with closed squares in Fig. 1a). We assumed that 163 different polypeptides represent the diversity level of D. fimbriatus venom. The diversity of mistrusted mature sequences encoded in dbEST was estimated at 420 polypeptides (curve with closed circles in Fig. 1a). One third of all sequenced clones were assembled into 6 contigs, each including more than 100 initial clones, though one contig included 1,063 clones (see Fig. 1b). Approximately 50 contigs have moderate representativeness (groups 6–19 and 20–99 in Fig. 1b). The largest number of deduced structures was assembled into 256 singletons and 113 small-sized contigs (2–5 clones). Overall the toxin structure representation was approximately 61%, which was in good agreement with other spider venom ESTs10,31. To determine all moderately present sequences (represented by 6–99 clones), it was sufficient to generate a dbEST from 3,500 sequences (Zone 2 in Fig. 1a), whereas it was only necessary to analyze 300 sequences for the detection of the six major toxins (Zone 1 in Fig. 1a).

Figure 1: The databank representativeness.
figure 1

(a) Relationship of the deduced polypeptide number (ordinate) with the size of the analyzed bank (abscissa). The curve with open circles corresponds to the growth of total toxin-like sequences, the curve with closed circles reflects the growth of validated sequences, and the curve with close squares represents the growth of true sequences. The marked area denotes the bank size that would be sufficient to identify all major compounds (Zone 1) or all moderately distributed compounds (Zone 2). (b) The distribution of clones in the dbEST by contig size. The number of unique structures is shown in brackets for each group of contigs. The contigs were generated for exact mature polypeptide sequence deducing. A singleton corresponds to a singular sequence, and the out of range group included all sequences that were not retrieved by the motif search together with the sequences that were not proven by verification.

Full size image

In total, 163 mature polypeptides were identified in the investigated venom, but many more nucleotide sequence variants were measured by genes. A total of 451 different sequences encoding polypeptide precursors were counted when comparing structural variation in whole-protein precursor sequences. Another 344 sequence variants with synonymous mutations were found after an alignment of the original nucleotide sequences. Thus, the analysis of the studied database revealed the presence of 795 unique nucleotide sequences encoding 451 different polypeptide precursors.

Structural information about the deduced venom polypeptides is available in Supplementary Files (Supplementary File 1 contains all nucleotide sequence variants in FASTA format; Supplementary File 2 contains polypeptide structures expected in crude venom also in FASTA format; Supplementary File 3 includes information about both nucleotide and polypeptide sequences).

Technical Validation

Superfamily features

Because spider venoms commonly contain a large number of homologous sequences with point substitutions6, it is convenient to designate a group of related sequences as a superfamily. Some peculiarities of the superfamily organization can be illustrated using the first superfamily, consisting of 1,467 ESTs in the total database. After verification, only 644 valid mature sequences were recognized. In the superfamily, the number of errorless nucleotide sequences coding for full-size precursor proteins was 563, with 25 mature venomous polypeptides. For each mature polypeptide, additional transcriptome variants were detected possessing substitutions in the signal and propeptide regions (Table 2). For example, 11 transcripts were found with nonsynonymous substitutions and 16 transcripts with synonymous substitutions that coded for the same mature polypeptide LTDF 01-01. As a result, this peptide is expressed in venom glands from 27 different mRNA sequences and can be described by 11 precursor proteins. For superfamily 01, 78 unique precursors and 136 transcriptome variants were found. The same diversity of nucleotide sequences was observed for other superfamilies and singleton sequences.

Table 2 Consolidation table of superfamily 01 precursors.
Full size table

The dominance of one nucleotide sequence in each superfamily was observed by inspecting the distribution of unique transcripts. The sequence named as the preferable nucleotide sequence (PNS) was that present in the EST database at a much higher frequency than the other variants. Very rarely, a second transcriptome variant was also moderately represented at a frequency up to 2/3 of that of the PNS. To estimate the superiority of the PNS over other EST variants, we calculated the variability level as the percentage of the total number of sequences encoding each polypeptide represented by other variants. Similar variabilities of approximately 24% were obtained for the well-distributed polypeptides LTDF 01-01, 02, 03, and 04, whereas other members of the superfamily exhibited variability dispersion due to an insufficient amount of data for analysis. We assessed the variability of all polypeptides represented in the database by precursors encoded by more than 5 variants. In Fig. 2a, we divided the obtained results into several groups according to the variant’s quantity. The variability value was similar at approximately 20–25% for all groups, and an increase in the number of variants only led to a deviation decrease. Therefore, we can assume the presence of uniform variability machinery for entire spider polypeptides based on one major gene that is available for further mutagenesis. It is obvious that active mutation machinery can also produce substitutions in mature chains. These point mutations correlate to the size of a superfamily. We separated the most frequent genes that are transcribed predominantly in PNS inside each superfamily, evaluated their representation in the analyzed EST bank, and calculated the variability of the superfamily as a whole (Fig. 2b).

Figure 2
figure 2

Variability of venomous polypeptides. (a) Analysis of the 36 precursor proteins represented in the EST bank by more than 5 transcriptome variants. Standard deviations are shown. (b) Variability into 24uperfamilies’ for a set of best-represented genes. (c) Consensus disagreement for an alignment of nucleotide sequences encoding superfamily 01 protein precursors except the sequences encoding toxins 01–05 and 01–22.

Full size image

Most superfamilies demonstrated differences in the number of major genes but were comparable in terms of the variability of minor transcripts. Superfamilies 01 and 03 have the largest number of major genes. These superfamilies each have four major genes with total PNS contents of 387 for superfamilies 01 (see data in Table 2) and 145 (for superfamily 03), corresponding to calculated variabilities of 31 and 40%. Two superfamilies, 09 and 15, did not include a PNS, and we were unable to identify their major genes. The presence of 1 or 2 major genes was common for the other analyzed superfamilies. The estimated average variability excluding superfamilies 09 and 15 was approximately 29%. To summarize the obtained data, we conclude that there is a core set of 28 major genes in 14 superfamilies that are intensively transcribed in the venom glands of the spider D. fimbriatus. There are approximately 2 major genes in each superfamily. The structural abundance of polypeptides is derived from rare transcripts; thus, the total variations inside the superfamily are approximately 30% EST.

We speculate that in contrast to a normal protein’s expression, the diversification of major genes into a wide variety of transcripts is an attribute of toxin expression, not only for spiders but for other venomous animals as well. Because the leading roles of gene duplication and diversifying selection have been demonstrated for the formation of functionally variable conotoxins41, gene duplication is assumed to be the driver of animal toxin diversity42. This structural diversity can not be explained by errors of sequencing and sample preparation, such as dubious data (the curve with open circles in Fig. 1a) have been thoroughly eliminated at the stage of data verification.

A sharp consensus sequence for superfamily 01 was achieved after the elimination of two sequences encoding peptides LTDF 01–05 and LTDF 01–22 (Fig. 2c). The transcripts for these polypeptides were distinct from the other members by 3 insertions of 4, 11, and 4 bp into the mature chain region and by a large unusual 3’ region. The remaining members of superfamily 01 exhibited a high level of homology. Strong consensus disagreements were found only for the region encoding the propeptide that had a partial triplet insertion, and for one nucleotide insertion into the mature chain region. This nucleotide insertion of approximately 300 bp leads to a reading frame shift and the production of the longer toxins LTDF 01–06, 07, 21, 23, 24, and 25. For other positions, occasional point mutations were observed. By type, a nucleotide transition occurred 4 times more frequently than a transversion. The more conservative and evolutionarily stable region was located in the area coding for the mature polypeptide, which correlated with previously described variability differences. To date, it has been thought that mature chain sequences should be the predominant mutation sites. In contrast, we found that the most variable region in the analyzed transcripts was the entire propeptide, a large part of the signal peptide, and part of the N-terminal sequence of the mature polypeptide (Fig. 2c). Based on the number of substitutions across the full-length protein precursor sequence, a group of polypeptides (LTDFs 01, 02, 06, 12, 13, 20, and 23) with approximately 20 mutations per molecule was clearly distinguished, in contrast to the main group of sequences, which showed 1–8 mutations per molecule.

The analysis of Fig. 2c suggests the probable presence of one or two introns in the major genes of superfamily 01. Such a hypothesis is put forward by us on the basis of a greater variability in the precursor sequence preceding the mature chain, which might indirectly indicate the presence of different splice variants. Alternative splicing is often described for genes coding polypeptide toxins. For cone snails, polypeptide toxins revealed the presence of a quite extended intron inside a propeptide sequence located upstream from the mature chain43. Similarly, one or two intronic fragments were found near the end of a signal peptide in scorpion toxin genes44–46. In the case of spiders, the situation remains yet undefined, because for some short polypeptide toxins10,27,47 and for macromolecular latrotoxins14,48, no introns were found. In contrast, several introns were detected in the genes for long insectotoxins from the venom of the spider Diguetia canities49 and in the genes for sphingomyelinase D from several species of Loxosceles and Sicarius50,51.

Polypeptide toxins

All mature sequences were distributed into 16 superfamilies and 19 orphan proteins that did not have homologous polypeptides in the dbEST. The spider D. fimbriatus belongs to the superfamily Lycosoidea; thus, the deduced toxin-like polypeptides were named lycotoxins-Df, abbreviated as LTDFs. To distinguish the polypeptides, each was assigned either a number corresponding to a superfamily (01 to 16) or an ‘S’ for an ungrouped protein together with an ordinal number. The first number was assigned to the most represented sequence and the last number to the rarest one. Because the identity level between superfamily members was rather high, we applied a BLASTX homology search only to the first member in each superfamily. These results are summarized in Table 3. As would be expected, almost all derived polypeptides were found to be homologous to toxins that have previously been identified in spider venoms. However, there were no complete homologies, and all derived polypeptides were found to be novel polypeptide toxin-like molecules. Important levels of homology were found with the cysteine knot toxins predicted from the related spider species Dolomedes mizhoanus and to one polypeptide isolated from a natural venom of Cupiennius salei21,32.

Table 3 Polypeptide homologies by BLASTX.
Full size table

In the analyzed transcriptome bank, we found predominant sequences encoding ‘inhibitor cysteine knot’ (ICK) toxin-like polypeptides (147 out of a total of 163 trusted sequences). The alignment of the discovered toxins among themselves indicated their significant differences from each other in amino acid composition, polypeptide chain length, number of cysteine residues, and distance between cysteines (Fig. 3a). One obvious biological function of ICK toxins is interaction with ion channels.

Figure 3: Derived sequences of polypeptides.
figure 3

(a) Mature ICK toxin structures. The * symbol after the C-terminal amino acid residue indicates an amidation. The key cysteine residues are highlighted in accordance with the two main motifs of spider toxins, PSM (blue) and ESM (pink). (b) Alignment of non-ICK toxins. Identical residues are highlighted. Amino acid residues identified in the structure of Ca-channels blocker ω-agatoxin Ia as removable by maturation are underlined. Venom protein-7 from the scorpion Mesobuthus eupeus presented a partial structure without 25 C-terminal residues. (c) Comparison of linear peptide sequences. Similar amino acid residues are highlighted. For LTDF S-19 peptide, only the structure of a mature chain is shown. The * symbol in LTDF 16-01 sequence indicates amidated C-terminal residue.

Full size image

In addition to ICK toxin-like structures, some polypeptides with other spatial folds were detected in the transcriptome (Fig. 3b). Superfamily 14, with seven members, was larger than many of the ICK polypeptide superfamilies by EST count, but superfamily 15 and LTDF S-18 were rare. The primary feature particular to LTDF 14-01 is the presence in the protein precursor of two propeptides that are removed during maturation. The first propeptide is located between the signal peptide and the heavy protein chain, and a second small one with a length of 7 amino acid residues is found at the C-terminus. Such processing was described early for the structural homolog of ω-agatoxin Ia, which was detected in the natural venom as a double-stranded protein52. We can suppose that the biological function of polypeptides from superfamily 14 is connected with blocking Ca-channels.

The alignment of the amino acid sequence for another nonstandard protein precursor, LTDF 15-01, indicated moderate homology to a number of known proteins. Unfortunately, no biological function of these homologues has been determined experimentally thus far. Therefore, the functions of the two poorly represented polypeptides from superfamily 15 cannot be predicted. Moderate homology to the spider venom protein PN16C3 (Uniprot ID P84032) was found for LTDF S-18, encoding an extended protein precursor with an estimated molecular weight of greater than 14. The biological function of LTDF 15-01 cannot be predicted.

The investigated transcriptome belongs to a Lycosoidea spider, the venoms of which typically contain AMPs, but the amount of deduced linear peptides was negligible. Orphan protein LTDF S-19 and the well-represented superfamily 16 consisted of 5 different sequences were found (Fig. 3c). All identified linear peptides were of small size. A homolog search in the Uniprot database found a certain similarity of the LTDF 16-01 mature chain to filamentous proteins from the fungus Penicillium chrysogenum53. For the linear peptide LTDF S-19 homologs were detected only in a large number of short fragments from hypothetical protein.

Usage Notes

One peculiarity of a spider venom combinatorial library is the large number of genes with point mutations, which can lead to sequencing procedure limitations. If the reads obtained during the analysis are too short, further attempts at full-length gene reconstruction from several pieces will distort the fine details. As a result, a sequence can be obtained only for the most well-represented transcripts, and the number of coded toxins will be underestimated. The same underestimate of the number of components by a database analysis may occur when contig assembly does have not sufficiently strict parameters. The method based on SRDA and toxin primary structure motifs is more effective for the thorough analysis of combinatorial libraries, as previously confirmed31,37. The investigation of nucleotide sequences treated by any enzyme raises the question of how much error was introduced in the sequence array. The technique of double sequencing a clone from forward and reverse primers helps to eliminate such errors, as the probability of a polymerase mistake in the same place several times is negligible. The double sequencing cleared a most part of validated sequences (major reduction from 420 to 163 as shown on Fig. 1a). The BLASTX algorithm can also be considered as a tool for error elimination during EST database analysis. The true homology can be found in other reading frames in the case of a probable non-homologous protein retrieved by the bank screening. We discarded several sequences that thoroughly satisfied the other criteria but showed sufficient homology to a known protein in an alternative reading frame.

The most important point is how the real quantity of polypeptide structures in spider venom can be estimated. The dissimilarity of the compounds identified in the transcriptome from the number of components detected in natural venoms is common and not confined to spiders54,55. It has been reported that the most structurally rich spider venoms can contain approximately 500 individual components18, but recent proteomic studies on the basis of combinations of various separation and detection methods measured approximately 200 components in some spider venoms21,35. It is clear that the number of components in venom varies between spider species; moreover, venom composition is observed to vary within species when the venoms of several individuals are compared25.

A thorough variation search in the transcriptome allowed us to discover important features of spider polypeptide organization. First, there was a strong dominance of ICK folds over other structures. For disulfide-stabilized polypeptides, three alternative folds were found. There were two uncommon spatial folds: the first is similar to ω-agatoxin 1a (superfamily 14) and the second a novel one for toxin LTDF S-18. Another rare fold appears to resemble the well-known ‘three-finger’ fold of snake neurotoxins (superfamily 15). We assume the possibility of the presence of a toxin-like polypeptide that is most likely folded in the same way as ‘three-finger’ neurotoxins. In fact, the linear homology of the two polypeptides from superfamily 15 with the snake neurotoxin56 is quite low, but the arrangement of cysteine residues shows a common peculiarity (Fig. 4). The number of cysteine residues and the distances between the four C-terminal cysteines are identical to the three-finger neurotoxin. These differences occur in other parts of the polypeptides: the distance between cysteines 1 and 2 (first finger domain) is longer by 6 amino acids, and the distance between cysteines 3 and 4 (second finger domain) is shorter by 7 amino acids. We assume the possibility of a ‘three-finger’ fold for spider polypeptides with altered sizes of finger 1 and finger 2 and the same disulfide bond bridging as in snake neurotoxins, but this assumption requires further prove. The presence of the ‘three-finger’ motif suggests that, originally, venomous terrestrial animals had similar sets of genes for polypeptides with different folds. However, over the course of evolution, ICK polypeptides became predominant in spiders, reaching a large variety of structures, while the development of non-ICK polypeptide diversity was eliminated.

Figure 4: ‘Three-finger’ neurotoxin.
figure 4

(a) Alignment of LTDF 15-01 to king cobra neurotoxin haditoxin. The pattern of cysteine bridges and finger regions is shown in accordance with 3D data56. Equal residues are boxed, and the probable finger region is drawn for LTDF 15-01 based only on similarity. For LTDF 15-01, a partial sequence is shown without the 12 N-terminal residues. (b) 3D structure of haditoxin.

Full size image

Additional information

How to cite this article: Kozlov, S. A. et al. Comprehensive analysis of the venom gland transcriptome of the spider Dolomedes fimbriatus. Sci. Data 1:140023 doi: 10.1038/sdata.2014.23 (2014).