In contrast to “modern” spiders (infraorder Araneomorphae), spider families in the infraorder Mygalomorphae have many characteristics that are considered primitive including large paraxial fangs and a complete reliance on book lungs for atmospheric gas exchange [1,2]. The family Barychelidae comprises 44 genera and 307 extant species [3] accounting for 11% of all mygalomorphs, making it one of the most taxonomically diverse mygalomorph spider families. However, in striking contrast with their over 200 million year old sister family Theraphosidae (commonly termed “tarantulas”), which is probably the most thoroughly studied spider family with venom gland transcriptomes of several species already examined [4,5,6,7], barychelids have so far been entirely neglected by toxinological research. Over 200 peptide toxins have been described from theraphosid venoms (accounting for 25% of all described spider toxins while representing only 2% of all spider species) [8], whereas not even a single barychelid venom has been examined to date.

Spider venoms in general are complex mixtures of salts, low molecular mass compounds, acylpolyamines, peptide toxins, and enzymes [9,10]. Primitive spider venoms such as those of theraphosids are dominated by cystine-knotted peptide toxins (“knottins”) with molecular masses of 3–7 kDa that mainly act as modulators of ion channels to provide rapid incapacitation of prey [5,11]. Since they are closely related to theraphosids with a common ancestor dating back to the Triassic [12], one might assume a similar overall composition for barychelid venoms. However, it remains to be examined whether barychelid venoms have evolved novel peptide-toxin scaffolds and/or novel venom components with new modes of action during the 200 million years of independent evolution since their divergence from theraphosids.

Recently, spider-venom peptides have been suggested as leads for the development of novel therapeutics [13] and insecticides [14]. Hence, the identification of novel structural scaffolds from barychelid venoms might add to the pool of already known spider-toxin scaffolds, thereby increasing the chances of selecting useful lead candidates for insecticide and/or drug development. In order to unravel some of these hidden treasures, we have studied the venom gland proteome of the Australian barychelid Trittame loki using a combination of transcriptomic, peptidomic and bioinformatic methods. We present the first toxin sequences from barychelid venom, which includes novel cysteine scaffolds, thus providing new insights into the evolution of spider venoms.

Magnetic resonance imaging (MRI) revealed that T. loki retained the primitive characteristic of large fangs and venom glands contained entirely within the massive chelicerae (Figure 1). Transcriptomics recovered six toxin types: 42 diverse isoforms of the classic ICK/knottin spider-venom peptides (Figure 2); variants of the previously characterised prokineticin family (Figure 3) [15,16], CAP (CRiSP/Allergen/PR-1) (Figure 4) and kunitz (Figure 5) domain proteins, with CAP sequenced for the first time from any mygalomorph spider, and the first discovery of the enzymes acetylcholinesterases (Figure 6) and neprilysin (Figure 7) in spider venom. Proteomics confirmed that each toxin type was translated and present in the secreted venom (Supplementary Table 1), except for kunitz which was not detected by mass spectrometry.

The most abundant transcript type was of the ICK/knottin neurotoxic peptide type which showed significant evidence of accelerated evolution and diversification. This diversity was accompanied by extensive variation in cysteine pattern including multiple independent de novo evolution of novel cysteines (Figure 2). These peptides were phylogenetically diverse, with the T. loki ICK/knottin toxins forming eight clades, with two of them being particularly large and divergent (Figure 2). Notably the first of the major clades of ICK peptides also had a pair of newly evolved cysteines, as did less diverse sets. We assessed the molecular evolutionary history of genes encoding these toxins using state-of-the-art selection assessment methods The one-ratio model, the simplest of the lineage-specific models, estimated an ω of 1.36 and 0.46 for the major ICK/knottin toxin clades 1 and 2, respectively (Supplementary Tables 2.1 and 2.2). This highly conservative model can only detect positive selection when the ω ratio, averaged over all sites along the lineages in a phylogenetic tree, is significantly greater than one. Hence, the ω of 1.36 estimated for the major ICK/knottin toxin clade 1 clearly highlights the dominant role of positive selection in shaping their evolutionary history. Since lineage-specific models often fail to detect episodic diversifying selection that only affects certain sites in protein-encoding genes, we further employed site-specific models (Table 1 and Supplementary Tables 2.1 and 2.2). Model 8 estimated an ω of 1.81 and 0.67 for the major ICK/knottin toxin clades 1 and 2, respectively, highlighting the significant influence of positive selection on the evolution of the major ICK/knottin toxin clade 1 (Table 1 and Supplementary Tables 2.1 and 2.2). In contrast to the rapidly evolving clade 1, an ω of 0.67 estimated for the ICK/knottin toxin clade 2 indicated that this toxin clade has evolved under significant constraints of negative selection pressure and that the major part of its coding sequence has remained extremely well conserved over evolutionary time. The Bayes Empirical Bayes (BEB) method identified as many as 17 (33% of sites) positively selected sites in the major ICK/knottin toxin clade 1 and detected only 5 (11% of sites) positively selected sites in toxin clade 2 (Table 1 and Supplementary Tables 2.1 and 2.2), further highlighting the differential influence of evolutionary selection pressures on these two toxin clades. Other methods of evaluating the nature of selection, such as FEL, REL, MEME, FUBAR and integrative approach, conclusively supported these findings and indicated that in contrast to the rapidly evolving clade 1, clade 2 remains extremely well conserved (Table 1).

With a combined nucleotide and amino acid-level approach, we were able to identify seven and five positively selected sites in the major ICK/knottin toxin clades 1 and 2, respectively (Supplementary Table 3). Interestingly, all five hypermutational sites identified in clade 2 were extremely varied in terms of biochemical and/or structural properties, relative to their ancestral residues. Radical mutations are more likely to influence the fitness of the organism, since they are capable of altering the structure and/or function of the protein, in contrast to neutral mutations that result in the introduction of residues with similar biochemical properties. Evolutionary fingerprint analyses clearly depicted several residues in the major ICK/knottin toxin clade 1 as evolving under the influence of positive selection, while a majority of sites in the toxin clade 2 were found to be evolving under the regime of negative-selection (Supplementary Figure 1). The branch-site REL test detected 4 and 3 branches respectively in the two major ICK/knottin toxin clades as evolving under episodic bursts of adaptation (Supplementary Figure 2).

Thus, the evidence provided by various analyses, such as codeml site and lineage-specific models (M8, M2a, M3 and M0), various models of HyPhy (SLAC, FEL, REL, MEME and FUBAR), the amino acid-level approach implemented in TreeSAAP, the evolutionary fingerprint analyses and branch-site REL (Table 1; Supplementary Tables 2.1, 2.2 and 3; and Supplementary Figures 1 and 2), clearly highlighted the differential rate of evolution adopted by the major ICK/knottin toxin clades and is indicative of extreme venom diversification in the barychelid spider T. loki. The major ICK/knottin toxin clade 1 in particular was found to have evolved under the significant influence of positive selection, while most codon sites in clade 2 evolved under the regime of negative selection and thus remained well conserved. We speculate that the lack of variation in toxin clade 2 might be the result of selection pressures exerted by their target receptors in the form of steric requirements that influence binding efficiency. The conspicuous rapid accumulation of variations under the influence of positive Darwinian selection in the major ICK/knottin toxin clade 1 is strongly suggestive of its role in prey envenoming. We propose that these toxins are highly likely to play a central role in prey-envenoming and capture, and are likely to be involved in a co-evolutionary predator-prey arms race with their molecular targets in prey animals. A handful of positively selected sites were detected in the coding sequences of ICK/knottin toxin clade 2, indicating that even these highly conserved toxins accumulate variations, probably to fine-tune their biochemical activities. Not-surprisingly, we identified a large number of novel cysteine variants in this rapidly evolving ICK/knottin peptide toxin class (Figure 2), pointing towards potential wide scale neofunctionalisation. Experimental examinations are warranted to precisely understand the stark difference in the evolutionary regimes adopted by the aforementioned toxin clades.

Our identification of enzymes like acetylcholinesterase and neprilysin in the venom of spiders points to a more complex spider-venom arsenal than previously realised. It is noteworthy that, as indicated by the proteomic data (Supplementary table 1), these enzymes are not just present in trace amounts in the venom. This, together with the lack of identification of other proteins with potentially metabolic roles in the venom, suggests that both acetylcholinesterase and neprilysin are likely to play a role in envenomation. While acetylcholinesterase is known for reducing the amount of the neurotransmitter acetylcholine at synapses, neprilysin is a metalloprotease that might be involved in the degradation of extracellular matrix around synapses [9]. Thus, both enzymes might participate in prey envenoming and capture, with neprilysin aiding the action of acetylcholinesterase and neurotoxic peptides by facilitating their access to synaptic targets.

In conclusion, our multidisciplinary approach has revealed: (i) the differential rate of evolution adopted by T. loki toxins: some clades are extremely well conserved under the influence of negative selection, likely as a result of the steric requirements that influence binding efficiencies, while others rapidly accumulate variations under positive selection; (ii) the evolution of novel molecular scaffolds through the acquisition of new cysteine residues; (iii) the important role of positive selection in the evolution and diversification of ICK/knottin toxin clades; (iv) the presence of novel enzymes (acetylcholinesterase and neprilysin) for the first time in spider venom; and (v) the presence of CAP for the first time in the mygalomorph spider venom. Venom components employed for predation often experience a greater influence from positive selection and accumulate variations that are not only necessary for acquiring different types of prey [34,35,36], but also in escaping host immune responses [37]. The likelihood of identifying novel venom components for drug design is theoretically higher in organisms with extremely variant and rapidly evolving venom arsenals. The results of this study reinforce the significance of studying even obscure spider lineages. These results also add to the growing body of knowledge regarding structural convergence of venom components as both acetylcholinesterase and neprilysin have been recruited into the venoms of other animals [36,37].