Pinedaetal.BMCGenomics2014,15:177 http://www.biomedcentral.com/1471-2164/15/177 RESEARCH ARTICLE Open Access Diversification of a single ancestral gene into a successful toxin superfamily in highly venomous Australian funnel-web spiders Sandy S Pineda1†, Brianna L Sollod2,7†, David Wilson1,3,8†, Aaron Darling1,9, Kartik Sunagar4,5, Eivind A B Undheim1,6, Laurence Kely6, Agostinho Antunes4,5, Bryan G Fry1,6* and Glenn F King1* Abstract Background: Spiders have evolved pharmacologically complexvenomsthatserve to rapidly subdue preyand deter predators.The major toxic factors in most spider venomsare small, disulfide-rich peptides. While thereis abundant evidence that snake venoms evolved by recruitmentof genes encoding normal body proteins followed byextensive gene duplication accompaniedby explosivestructural and functional diversification, the evolutionary trajectory of spider-venom peptides is less clear. Results: Here wepresent evidence of a spider-toxin superfamily encoding a high degree ofsequenceand functional diversity thathas evolved via accelerated duplication and diversification ofa single ancestral gene. The peptideswithinthistoxinsuperfamilyaretranslated asprepropeptides thatareposttranslationallyprocessedtoyield thematuretoxin. The N-terminal signal sequence, as well as the protease recognition site at thejunction ofthe propeptide and mature toxin are conserved,whereastheremainder of the propeptide and mature toxin sequences are variable. Alltoxin transcriptswithin this superfamily exhibit a striking cysteine codon bias. We show that different pharmacologicalclasses of toxins within this peptide superfamily evolved under different evolutionary selection pressures. Conclusions: Overall, this study reinforces thehypothesis that spiders use a combinatorial peptide library strategy to evolve a complex cocktailof peptide toxins thattarget neuronalreceptorsand ion channels inpreyand predators. We show that theω-hexatoxins that target insect voltage-gated calcium channels evolved under the influence of positive Darwinian selection inan episodic fashion, whereas theκ-hexatoxinsthat targetinsect calcium-activated potassium channels appear to be under negative selection. A majorityof thediversifying sites in theω-hexatoxins are concentrated onthe molecular surface ofthe toxins, thereby facilitating neofunctionalisation leading to new toxin pharmacology. Keywords: Spider toxin, Spider venom, Hexatoxin, ω-hexatoxin,κ-hexatoxin,Australian funnel-web spider, Molecular evolution, Gene duplication, Positiveselection,Negative selection *Correspondence:[email protected];[email protected] †Equalcontributors 1InstituteforMolecularBioscience,TheUniversityofQueensland,306 CarmodyRoad,StLucia,QLD4072,Australia 6VenomEvolutionLab,SchoolofBiologicalSciences,TheUniversityof Queensland,StLucia,QLD4072,Australia Fulllistofauthorinformationisavailableattheendofthearticle ©2014Pinedaetal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative CommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,and reproductioninanymedium,providedtheoriginalworkisproperlycredited.TheCreativeCommonsPublicDomain Dedicationwaiver(http://creativecommons.org/publicdomain/zero/1.0/)appliestothedatamadeavailableinthisarticle, unlessotherwisestated. Pinedaetal.BMCGenomics2014,15:177 Page2of16 http://www.biomedcentral.com/1471-2164/15/177 Background producelargemultigenefamilies.Thisprocessisanalogous Venomshaveproventobekeyevolutionaryinnovations to the birth-and-death model of evolution proposed for for many divergent animal lineages [1,2]. Although the multigene families involved in adaptive immunity, such as mostextensivelystudied venoms arefromthemedically themajorhistocompatibilitycomplexandimmunoglobulin important scorpions, snakes, and spiders, venom sys- V genes [6]. However, the evolutionary trajectory is less H tems are present in many other lineages including clear for the venoms of spiders, scorpions, and molluscs, cnidarians, echinoderms, molluscs, fish, lizards, and which are dominated by disulfide-rich peptides of mass mammals [1,2]. These venoms have evolved to serve a 2–9kDa[7-12].Thesepeptidestypicallypossesshighaffin- variety of purposes, including prey capture, competitor ity and often-exquisite specificity for particular classes of deterrence, and defense against predators. There has ion channels and other nervous system targets [13-15]. been considerable innovation both in the chemical These neurotoxic functions are perhaps not surprising composition of these venoms as well as the method of given that the primary role of these venoms is to paralyse venom delivery, which includes barbs, beaks, fangs, orkillenvenomatedprey[11,16,17]. harpoons, nematocysts, pinchers, proboscises, spines, In this study, we analysed toxin-encoding transcripts spurs,andstingers[1,2]. from five species of Australian funnel-web spider (Aranae: From a molecular evolutionary perspective, the venoms Mygalomorphae:Hexathelidae:Atracinae)fromthegenera of snakes are the best understood. There is now abundant Atrax and Hadronyche, representing a geographic spread evidence that snake venoms evolved by recruitment of of more than 2000 km (Figure 1), in order to provide genes encoding normal body proteins followed by exten- insightintotheevolutionarytrajectoryoftheω-hexatoxin- sive duplication, neofunctionalization, and in some in- 1 (ω-HXTX-1) family. ω-Hexatoxins (formerly known as stances relegation to the status of pseudogene [1,3-5]. In ω-atracotoxins) are peptides comprising ~37 residues that manycases,thesegeneshavebeenexplosivelyreplicatedto were first isolated from the venom of the lethal Blue Figure1Distribution,venomcollectionandvenom-glanddissectionofAustralianfunnel-webspiderspeciesusedinthisstudy.(A) MapoftheeasternhalfofAustraliashowingthedistributionofthefivespeciesofAustralianfunnel-webspiderusedinthisstudy.(B)Female funnel-webspider(Hadronycheinfensa)fromFraserIsland,QLD.Inresponsetoprovocation,thespiderhasadoptedatypicalaggressive/defensive posture,withfrontlegsandpedipalpsraisedandthefangsinanelevatedpositionreadytostrike.Notethedropofvenomoneachofthefang tips.(C)AsingleH.versutavenomglandthathasbeendissectedfromthesurroundingmuscletissue.Thevenomglandintheseandother mygalomorphspidersislocateddirectlybelowthedorsalsurfaceofthechelicerae. Pinedaetal.BMCGenomics2014,15:177 Page3of16 http://www.biomedcentral.com/1471-2164/15/177 Mountains funnel-web spider Hadronyche versuta [18]. Results and discussion The ω-hexatoxins are major components in the venom of Theω-hexatoxinsareexpressedasprepropeptide Australian funnel-web spiders [18-20] and they contribute precursors significantlytopreyimmobilizationbyvirtueoftheirability RACE analysis was used to amplify transcripts encoding to specifically block insect, but not vertebrate, voltage- orthologsofω-HXTX-Hv1afromfourspeciesofAustralian gated calcium (Ca ) channels [17,18,20-22]. Their po- funnel-webspider:Atraxrobustus,H.infensa,H.venenata, V tentinsecticidalactivityhasengenderedinterestinthese and H. versuta (Figure 2). Multiple ω-HXTX-Hv1a ortho- peptides as bioinsecticides [11,17,23]. Proteomic analysis logswereidentifiedineachspecies(i.e.,24paralogsencod- ofH.versutavenomrevealedanumberofω-HXTX-Hv1a ing seven distinct mature toxins were identified in H. paralogs [19],suggestingthatthispeptide toxin mightbe- infensa,18paralogsencodingsixmaturetoxinswereidenti- long to a multigene family. However, because the venom fiedintheSydneyfunnel-webspiderA.robustus,andeight used in this previousstudywaspooled from several spi- paralogs encoding two mature toxins identified in the ders,it wasunclearwhethertheseapparentparalogsare Tasmanian funnel-web spider H. venenata) (Figure 3A). simply polymorphisms resulting from allelic variation. A further eight paralogs encoding four distinct mature By using cDNA libraries obtained from a single spider, toxins were identified in the venom-gland transcriptome we demonstrate here that ω-HXTX-Hv1a is indeed part ofH.modesta(Figure3A).Thus,theaminoacidsequence of a large multigene family that appears to have arisen diversitypreviouslyreportedforω-HXTX-1basedonana- from explosive gene duplication followed by extensive lysisofpooledvenomsamples[19]isduetoexpressionof sequence divergence and neofunctionalization. Within multiple related transcripts in a single spider rather than thissuperfamily of toxins, weshowthat pharmacologic- allelic variation. The almost complete conservation of the ally distinct toxin classes are evolving under starkly signalsequence,aswellasthepatternofconservedcyste- different selection pressures, with some toxin classes ines in the mature toxin (Figure 3A), indicates that these accumulating variation under episodic bursts of adapta- ω-HXTX-Hv1a homologs arose by duplication and se- tion, while others remain constrained by negative selec- quencedivergenceoftheoriginaltoxin-encodinggene. tion. This work reinforces the idea that the chemical All of the ω-HXTXs are expressed as prepropeptide and pharmacological diversity present in spider venoms precursors that are posttranslationally processed to yield may have evolved from a relatively small number of an- thematuretoxinsequence(Figure2A).Thehighlyhydro- cestralgenes. phobic 22-residue signal sequence is of similar length to Figure2Schematicrepresentationoftoxinprecursors,overallRACEamplificationstrategyandidentificationoftoxinsuperfamilies. (A)Schematicrepresentationofatypicalspiderpeptideprecursorshowingthesignalpeptideinorange,thepropeptideinpurple,andthe maturetoxininblack.Aftertranslation,thesignalandpropeptideregionsareproteolyticallyremovedtoyieldafunctionalmaturetoxin.(B) GeneraloverviewoftheRACEprotocolforsequencinghexatoxintranscripts.Adaptorsareaddedtothe5’and3’endoftranscriptsduringcDNA librarypreparation.Inboth3’and5’RACE,gene-specificprimersareusedintheforward(3’RACE)orreverse(5’RACE)orientationtoamplify full-lengthsequences.TheresultingPCRproductsarethenclonedandsequenced.(C)SchematicrepresentationoftheShivasuperfamily highlightingthecombinatorialnatureofspider-venompeptides. Pinedaetal.BMCGenomics2014,15:177 Page4of16 http://www.biomedcentral.com/1471-2164/15/177 Figure3Sequencealignmentsofωandκ-hexatoxins.(A)Sequencealignmentofω-HXTX-Hv1aparalogsfromeachspecies:Hadronyche modesta(Hmo1a–Hmo1d),Atraxrobustus(Ar1a–Ar1g),Hadronycheinfensa(Hi1a–Hi1g),andHadronychevenenata(Hvn1aandHvn1b).Thelevelof residueconservationisgradedfromblack(fullyconservedacrossallparalogs)todarkgrey(conservedinmosttoxins)tolightgrey(conservedin amajorityoforthologs).Thelinesbelowthesequencealignmentindicatethedisulfide-bondconnectivityofω-HXTX-Hv1a.(B)Alignmentof κ-HXTX-Hv1aparalogsfromHadronycheversutaandoneorthologfromHadronychemodesta.Thelevelofresidueconservationisgradedasin panel(A)andthesignalpeptide,propeptide,andmaturetoxinregionsarehighlighted.Thelinesbelowthesequencealignmentindicate thedisulfide-bondconnectivityofκ-HXTX-Hv1c,withthevicinaldisulfidebondhighlightedinred.Notethatω-HXTX-Ar1a(UniProtPF06357), ω-HXTX-Hi1a(UniProtP0C2L5),ω-HXTX-Hi1b(UniProtP0C2L6),ω-HXTX-Hi1c(UniProtP0C2L7),andκ-HXTX-Hv1c(UniProtP82228)havebeen previouslyisolateddirectlyfromvenom. that reported for peptide-toxin precursors from spiders, study. These species are distributed along the eastern scorpions, and cone snails [8]. The 20-residue propeptide seaboard of Australia with a geographic spread of more sequenceishighlyacidic,withanetchargeof−4,afeature than 2000 km (Figure 1). The hexathelids are a group of thathasbeennotedfornumerousspider-toxinpropeptide approximately40speciesdividedintothreegenera:Atrax, sequences [24-28] but which is not characteristic of toxin Hadronyche and Illawarra [34-36]. They are adapted to precursors from other venomous animals. Moreover, forestenvironmentsbutcanalsobefoundinhabitatsthat the presence of a propeptide regioncontrasts with most range from montane herblands and open woodland to scorpion-toxin precursors in which the signal sequence closedforest[36].Conservationoftheω-hexatoxinfamily is fused directly to the mature toxin without an inter- oftoxinsoverthiswiderangeofenvironmentsanddiffer- vening propeptide [8]. The reason for the highly acidic ing prey distributions implies that there has been strong propeptide region in spider toxin precursors remains to evolutionarypressuretomaintainthesepeptidesaspartof be determined, but it may be related to specific interac- the venom arsenal, which is perhaps not surprising given tions between the toxin precursor and components of thattheyarebroadlyactiveagainstmanydifferentarthro- the secretory and/or protein folding pathway in spider pods[11,17,37]. venomglands. In addition to obvious homologs of ω-HXTX-Hv1a, ThepropeptidesequenceterminateswithadibasicArg- the RACE and transcriptomic analyses revealed add- Arg signature;dibasicsequencesare commonrecognition itional families of toxins that had almost identical signal sites for proteolytic removal of propeptide segments in sequences to the ω-HXTX-1 transcripts, but divergent neuropeptide precursors from both vertebrates [29] and propeptide and mature toxin sequences. We named one invertebrates [30]. While Arg is the terminal residue in of these families the ω/κ-HXTX family (described as U- virtuallyall knownspider-toxinpropeptidesequences,the ACTXin[38]).Theω/κ-HXTXpeptidesappeartobedis- penultimate residue is variable, though it is commonly tributed in two of the species examined (A. robustus and Asp,Glu,orLys[25,26,31-33]. H.versuta);thisreinforcestheideathatthesetoxinsmost likely arose ancestrally by duplication of a ω-HXTX-1 ω-hexatoxinsbelongtoalargetoxin-genesuperfamily gene followed by hypermutation of the propeptide and Orthologs of ω-HXTX-Hv1a were identified in all five mature-toxin regions in order to create a new function species of Australian funnel-web spider examined in this (neofunctionalization). The conservation and radiation of Pinedaetal.BMCGenomics2014,15:177 Page5of16 http://www.biomedcentral.com/1471-2164/15/177 these toxinsacrossthis family of spiders implies that they signal peptide, propeptide, and mature toxin [8]. A re- are not nonfunctional relics of an explosive radiation of vised logo analysis of the Shiva superfamily (Figure 4A) this toxin-gene superfamily, and we confirmed this by that incorporated all of the new sequences and species showing that recombinant ω/κ-HXTX-Hv1a is highly reported here reinforced the dichotomy in evolutionary insecticidal [38]. The high insecticidal potency of this forces affecting various elements of the toxin precursor. familyofpeptidesisbelievedtoresultfromasynergistic The signal peptide has clearly been highly conserved effect on insect voltage-gated calcium (Ca ) channels throughouttheevolutionofthistoxinsuperfamilyanditis V andcalcium-activatedpotassium(K )channels[38]. presumably under negative selection in order to ensure Ca RACE analysis of the venom-gland cDNA library from that these toxins are directed to the appropriate secretory H.versutaalsoledtoamplificationoftranscriptsencoding pathway.Incontrast,thereissignificantsequencevariation the insecticidal toxin κ-HXTX-Hv1c [39], and sequencing in both the propeptide and mature toxin sequences, with of the venom-gland transcriptome from H. modesta also two notable exceptions. First, in contrast to the highly uncovered an ortholog of this toxin (Figure 3B). This was variable upstream region of the propeptide sequence, the entirely unexpected since this toxin has a vastly differ- C-terminal proteolytic recognition signal (Arg-Arg) is ent primary structure to ω-HXTX-Hv1a [39]. Moreover, completely preserved (Figure 4A). Presumably there has in addition to the six conserved cysteine residues in ω- been strong selection pressure to ensure processing of the HXTX-Hv1a that form an inhibitor cystine knot (ICK) propeptide by a specific protease. Second, in contrast to motif [40,41], κ-HXTX-Hv1c contains two additional theoveralllowlevelofconservationofthematuretoxin cysteineresiduesthatformanextremelyrarevicinaldisul- sequence, the cysteine residues, which direct the three- fide bond [42-44]. Furthermore, in contrast to ω-HXTX- dimensional(3D)foldofthetoxins,are completelycon- Hv1a,whichblocksinsectCa channels,κ-HXTX-Hv1cis served (Figure 4A). The marked variation in levels of V apotentandspecificblockerofK channels[45].Never- sequence conservation between the spider-toxin signal Ca theless, the near identity of the signal sequence in these sequence and the propeptide and mature toxin regions two toxin families and the conservation of cysteine resi- isreminiscentofthatobservedforsuperfamiliesofcone dues in the mature toxin indicate that they evolved from snailtoxins[46-51]. the same ancestral toxin gene and are members of the There are two striking differences between the Shiva samegenesuperfamily. superfamily precursors and transcripts encoding human We did not find orthologs of κ-HXTX-Hv1c in any of neuropeptides and other secreted proteins. First, whereas theotherthreespeciesofAustralianfunnel-webspider(H. precursorsofhumanneuropeptidesoftenencodemultiple infensa,A.robustus,andH.venenata).However,κ-HXTX- mature neuropeptide sequences [29,52], we and others Hv1a,κ-HXTX-Hv1b,andκ-HXTX-Hv1careexpressedat have not found any examples of spider-toxin transcripts very low levels in H. versuta venom [42], and conse- that encode more than a single mature toxin sequence. quently we cannot rule out the possibility that these Secondly, in direct contrast to the toxin precursors, the toxins are present in the venom of the other three spi- sequence of the mature human neuropeptide(s) is usually ders but the transcript levels are too low to be detected strongly conserved whereas there is significantly more usingthemethodsemployedhere. variability inthe signalsequence. This isperhaps not sur- prising given that human neuropeptides usually act on a TheShivasuperfamilyofpeptidetoxins singlewell-definedmoleculartargetwhereasspidertoxins It has previously been suggested that superfamilies of typically targeta specificsubtype of receptor orionchan- spider-venom peptides evolved from a single ancestral nel that nevertheless might vary significantly in primary gene via explosive gene duplication [8]; the work de- structure between prey taxa (Note that most spiders are scribed here further supports this idea as it is clear that generalist predators that target a phylogenetically diverse the ω-HXTXs, ω/κ-HXTXs, and κ-HXTXs belong to a range of prey). Thus, expressing a family of related toxins large superfamily of toxins that arose via gene duplica- in the venom (essentially a mini-combinatorial peptide li- tion (Figure 2C). We have chosen to name spider-toxin brary) might ensurethat the desired receptor/ion channel gene superfamilies after deities of death and destruction istargeted,regardlessofpreytaxa. since the major biological role of these toxins is to paralyze and/or kill envenomated prey. Accordingly, we Position-specificcysteinecodonbias have named the ω/κ-HXTX/ω-HXTX/κ-HXTX gene Mature ω-HXTXs contain three disulfide bonds with 1– superfamily after the Hindu deity Shiva, commonly 4, 2–5, 3–6 connectivity. These disulfides form an ICK knownasthe“destroyer”. motif that provides these toxins with a high degree of Sequence logos were previously used to analyse differ- chemical, thermal and biological stability [53]. Although ences in the level of sequence conservation between the it is clear from a protein structure viewpoint why these three parts of the ω-HXTX toxin precursor, namely the six cysteine residues need to be strictly maintained in Pinedaetal.BMCGenomics2014,15:177 Page6of16 http://www.biomedcentral.com/1471-2164/15/177 Figure4SequencelogoandcodonusageanalysisfromtheShivasuperfamily.(A)Sequencelogo[54]basedonalignmentof prepropeptidesfromtheShivasuperfamily.Thereisamuchhigherlevelofsequenceconservationwithinthesignalpeptidethanwithinthe propeptideandmaturetoxinregions.Note,however,thatthecysteineresiduesthatformthecystine-knotmotifinthematuretoxinandthe Arg-Argproteaserecognitionsitethatterminatesthepropeptideregionarebothcompletelyconserved(highlightedinblueandred, respectively).(B)Codonusageforthesix-cysteineresiduesthatformthecystine-knotmotif.NotethestrongbiasforTGCatcysteinepositions1, 3,4,and6.Shownabovethehistogramisthedisulfidebridgearrangementforthesixcysteinesasinferredfromthe3Dstructuresof ω-HXTX-Hv1aandκ-HXTX-Hv1c. order to preserve the toxin’s 3D scaffold, one would not disulfidebond(Figure4B).Theobserved position-specific expecttofindapreference foreitheroneofthetwopos- codon bias is not simply a manifestation of global codon sible cysteine codons (TGT and TGC). Intriguingly, bias in thesespidersas wehaveobserved a preferencefor however, previous analysis of ω-HXTX precursors re- TGT as opposed to TGC for cysteine residues in other vealed a strong bias for TGC at four of the six cysteine hexatoxin superfamilies (data not shown). Moreover, we positions in the mature toxin region [8]. An extended did not observe extreme codon bias for any other con- logoanalysis[54]incorporatingallofthenewlydiscovered servedresidueinthematurehexatoxins. sequences reported in this study corroborated the previ- Position-specific cysteine-codon bias has also been ob- ously observed codon bias (Figure 4B). We found an ex- served in superfamilies of cone snail toxins and it has tremeTGC codon bias for the four cysteine residues that been proposed that these codons might serve as attrac- form the 1–4 and 3–6 disulfide bridges in the ω-HXTX tants for a mutator complex that includes a poorly pro- family but not for the two cysteines that form the 2–5 cessive and highly mutagenic polymerase (e.g., DNA Pol Pinedaetal.BMCGenomics2014,15:177 Page7of16 http://www.biomedcentral.com/1471-2164/15/177 V) that promotes radiation of the toxin superfamily by site-specific models (Table 1: codon numbers based on κ- facilitating hypermutation of the mature toxin region HXTX-Hv1c_2 and ω-HXTX-Ar1a_1; Additional file 1: [49,50]. However, there is currently no direct evidence Table S1–3). Model 8 estimated ω of 0.69, 1.06 and 0.78 that cysteine-codon bias plays a part in directing the for the ω-HXTXs, κ-HXTXs, and the combined Shiva evolutionofspiderorconesnailtoxins. superfamily dataset, respectively (Table 2 and Additional file 1: Table S1–3). Although the computed ω for the Molecularevolutionanalyses κ-HXTXs was >1, the assessment was not statistically We utilized various state-of-art molecular evolutionary significant (p>0.05) in comparison with the null model assessment methods to determine the influence of nat- (M7β).TheBayesEmpiricalBayes(BEB)approachimple- ural selection on the evolution of genes encoding Shiva mented in M8 was only able to identify one positively se- superfamily toxins (see Methods section for full details lected site in the combined toxin dataset (Table 2 and of the selection analyses). The one-ratio model, the sim- Additionalfile 1: Table S3). Thus, the site-specific models plest of the codon-specific models, estimated the non- failed to detect the influence of adaptive selection pres- synonymous-to-synonymous nucleotide-substitution rate sures in shaping evolution of the Shiva superfamily. In ratio (ω) to be 0.64, 1.06 and 0.69 for the ω-HXTXs, contrast,themoreadvancedFast,UnconstrainedBayesian κ-HXTXs, and combined Shiva superfamily dataset, re- AppRoximation (FUBAR) [55,56] implemented in HyPhy spectively (Additional file 1: Table S1–3). This highly detected a handful of positively selected sites in both the conservative model can only detect positive selection ω-HXTXsandthecombineddataset(Table1). when ω, averaged over all sites along the lineages in a Site-specific models for detecting positive selection phylogenetic tree, is significantly greater than one. As work best when detecting pervasive selection pressures. lineage-specific models of PAML, such as the one-ratio However, the majority of positively selected sites are model, often fail to detect positive-Darwinian selection oftensubjectedtotransientorepisodicadaptations.When thatonlyaffectscertainsitesinproteins,wealsoemployed the majority of lineages evolve under the influence of Table1Nucleotideandcomplementaryproteinanalysesforωtoxins Sitea CodeML TreeSAAP Accessible Codon AminoAcid M2ab M8c Propertyd Magnitudee surfaceareaf 21 E 0.99±0.38 0.87±0.42 - - - (0.201) (0.257) 40 V 1.41±0.49 1.47±0.37 M ,M,V0,μ 8,8,8,8 42.0 W V (0.589) (0.842) Partiallyexposed 44 S 1.25±0.37 1.28±0.41 MW,M,V0,μ 8,8,8,8 82.1 V (0.420) (0.647) Exposed 53 H 1.62±0.63 1.55±0.36 - - 0.0 (0.749) (0.929) Buried 57 G 1.39±0.48 1.45±0.38 - - 57.3 (0.570) (0.825) Exposed 60 T 0.80±0.44 0.71±0.41 - - 49.7 (0.122) (0.154) Exposed 64 N 0.53±0.40 0.51±0.30 - - 100.0 (0.031) (0.036) Exposed 69 T 1.34±0.45 1.38±0.40 - - 59.8 (0.507) (0.751) Exposed 72 R 1.10±0.33 1.01±0.42 - - 0.0 (0.26) (0.374) Buried aSitesdetectedaspositivelyselectedusingtheintegrativeapproach. bM2aBayesempiricalBayes(BEB)posteriorprobabilityandpost-meanωindicatedinparentheses. cM8BayesempiricalBayes(BEB)posteriorprobabilityandpost-meanωindicatedinparentheses. dAminoacidpropertyunderselection(M :molecularweight;M:molecularvolume;V0:partialspecificvolume;μ:Refractiveindex). W V eMagnitudeifselectionontheaminoacidproperty. fAccessiblesurfaceareaof10–20%correspondstoburiedresidues,40–50%indicatespartiallyexposedaminoacidresidues,and≥50%indicatessolvent exposedresidues. Pinedaetal.BMCGenomics2014,15:177 Page8of16 http://www.biomedcentral.com/1471-2164/15/177 Table2Molecularevolutionofωandκtoxinsfrom [59-64].Sincethesynthesisandsecretionof venompro- Australianfunnel-webspiders teins is energetically expensive [65-67], mutations that FUBARa MEMEb PAMLc disrupt the structure/function of proteins are filtered M8 M2a outofthepopulationbynegativeselectionovertime,fa- ω ω>1d:3 7 0 0 voring the conservation of catalytic and structurally im- portant residues. RAVER not only aids in generation of toxins ω<1e:5 0 0 a rapidly variable toxin molecular surface biochemistry, 0.69 0.73 but it also ensures the conservation of structurally and κ ω>1a:1 0 0 0 functionally important residues. Accumulation of varia- toxins ω<1b:0 0 0 tions on the molecular surface of the toxin is advanta- 1.06NS 1.06NS geous as the altered surface chemistry might lead to ALL ω>1a:3 8 1 0 newtoxinfunctions(neofunctionalisation). Toderivefurthersupportforthepositivelyselectedsites toxins ω<1b:7 (0+1) 0 detected by nucleotide analyses, we employed a comple- 0.78 0.83 mentaryprotein-levelapproachimplementedinTreeSAAP aFast,UnconstrainedBayesianApproximation(FUBAR). (Table1).TreeSAAPidentifiedtwopositivelyselectedsites bSitesdetectedasexperiencingepisodicdiversifyingselection(0.05 significance)bymixedeffectsmodelevolution(MEME). intheω-HXTXsthatwereincommonwiththesitesiden- cPositivelyselectedsitesdetectedusingtheBayesempiricalapproach tified by site-model 8 of PAML (Table 1). Evolutionary implementedinthesitemodelsM8andM2a.Numberofpositivelyselected sitesdetectedattheposteriorprobability≥0.99and0.95areindicatedin fingerprintanalyses(Figure5A)clearlyrevealedseveralres- parenthesis.ωcomputedusingM8andM2aarealsopresented. iduesintheω-HXTXsandthecombinedtoxindatasetthat dNumberofsitesevolvingundertheinfluenceofpervasivediversifying evolve under the influence of positive selection, while a selection,detectedbyFUBARat0.9posteriorprobability. eNumberofsitesevolvingundertheinfluenceofpervasivepurifying majorityofresiduesinthe κ-HXTXsremained under evo- selection,detectedbyFUBARat0.9posteriorprobability. lutionary constraint (Figure 5A,B). Thus, evolution of the ω=meandN/dS. ω-HXTXshasbeensignificantlyinfluencedbyshortbursts NS=notsignificantat0.05comparedtothenullmodel(M7:beta). of episodic adaptations, while the κ-HXTXs appear to be negative selection, they mask the signal of positive selec- undernegativeselection. tion that influences only a small number of lineages. In Phylogeneticanalysisrevealedthattheκ-HXTXsform such scenarios, the aforementioned analyses may fail to a separate clade to the ω-HXTXs, rendering the Shiva detecttheinfluenceofpositiveselection.Toaddressthe superfamily non-monophyletic (Figure 6). There are shortcomings of the aforementioned approaches, we also significant variations within the ω-HXTXs suggest- employed the advanced Mixed Effects Model Evolution ive of functional diversification (Figure 6). The “hybrid” (MEME) [57], which uses fixed effects likelihood (FEL) ω/κ-HXTXs exhibit functional characteristics of both along the sites and random effects likelihood (REL) theω-HXTXsand κ-HXTXs astheyblockCa channels V acrossthebranchestodetectepisodicdiversifyingselec- (like the ω-HXTXs) as well as K channels (like the κ- Ca tion.MEMEiscapableofidentifyingbothpervasiveand HXTXs). The functional activity of the ω/κ-HXTXs episodicadaptations.MEMEidentified7and8episodic- combined with their relative phylogenetic placement and ally diversifying sites in the ω-HXTXs and combined cysteine pattern indicates that they are structurally and toxin dataset, respectively (Table 2), highlighting the functionally intermediate between the ω- and κ-HXTXs. vital role of episodic diversifying selection in shaping The evolution of new cysteine residues to create the vici- theevolutionofthesespidertoxins.Sixoutofeightepi- nal disulfide bond in the κ-HXTXs potentiated toxin sodically diversifying sites (75%) were located on the activity on K channels, since mutagenesis and analogue Ca molecular surface of the toxins (Table 1 and Figure 5B) studiesindicatethatthisvicinaldisulfidebondisthemost withtheir side chains completely orpartially exposed to criticalpartoftheK pharmacophore[43,45,68]. Ca solvent, suggesting that they could act as pharmaco- logical sitesand participateinprey envenomation; these Constraintsonmutationofthematuretoxinsequence findings are also in agreement with the selection forces It is generally considered that conservation of the cyst- found on the surface of the SGTx toxin family from the eine scaffold in toxin-gene superfamilies is critical for venom of the African Baboon spider Scodra grisiepies conserving the toxin’s 3D fold [35]. However, the incred- [58].RapidAccumulationofVariationsinExposedResi- ible disparity in the amino acid sequence between ω- dues (RAVER), where the toxin molecular chemistry HXTX-Hv1a and κ-HXTX-Hv1c (Figure 7A) begs the undergoes hypervariations under the influence of posi- questionofwhetherthisisreflectedinasignificantdiffer- tiveDarwinianselectionandfocal mutagenesis[59],has enceintheir3Dstructures,despitetheircommoncystine- been documented in a plethora of venom-components knot scaffold. The 3D structure of both toxins has from a wide diversity of venomous animal lineages been determined previously using homonuclear NMR Pinedaetal.BMCGenomics2014,15:177 Page9of16 http://www.biomedcentral.com/1471-2164/15/177 Figure5Molecularevolutionanalysesofκ-andω-HXTXs.(A)Evolutionaryfingerprintofκ-andω-HXTXs.Estimatesofthedistributionof synonymous(α)andnon-synonymous(β)substitutionratesinferredfortheκ-HXTXs,ω-HXTXs,andthecombinedShivasuperfamilydataset.The ellipsesreflectaGaussian-approximatedvarianceineachindividualrateestimate,andcoloredpixelsshowthedensityoftheposteriorsampleof thedistributionforagivenrate.Thediagonallinerepresentstheidealizedneutralevolutionregime(ω=1),whilepointsaboveandbelowthe linecorrespondtopositiveselection(ω>1)andnegativeselection(ω<1),respectively.Theωforsitemodel8,alongwiththetotalnumberof positivelyselectedsitesdetectedbyitsBayesEmpiricalBayes(BEB)approachandthenumberofepisodicallydiversifyingsitesdetectedbythe mixedeffectsmodelofevolution(MEME),arealsoindicated.(B)MolecularevolutionofShivasuperfamilytoxinsfromAustralianfunnel-web spiders.3Dhomologymodelsareshownwiththeirmolecularsurfacecoloredaccordingtotheevolutionaryconservationofaminoacids(see colorkey);thelocationofpositivelyselectedsitesisshowninredinspace-fillmodelsandasredspheresinwireframemodels.Alineplotisalso providedtohighlighttherelativeaccumulationofdNversusdS,estimatedusingtheM0modelofPAML.NS:Notsignificant. spectroscopy[18,42]andtheirpharmacophoreselucidated structural differences between the two toxins are the usingalaninescanningmutagenesis[17,21,43,69]. very different orientations of loops 2 and 3. However, Figure 7B and C show schematic representations of these structural variations cannot disguise the fact that the 3D structure of κ-HXTX-Hv1c and ω-HXTX-Hv1a, the two toxins essentially conform to the same 3D scaf- respectively. The two toxins can be considered to com- fold despite their extraordinary sequence divergence prisefourinter-cystineloops,whicharelabelled1–4from (16% identity if the cysteine framework is excluded). N-toC-terminus.Althoughthere is anobvious similarity This ability to maintain a consistent molecular architec- in the disposition of the three centrally located disulfide ture despite massive variation in the inter-cystine loop bridgesthatformthecystine-knotmotifineachtoxin,the sequenceshasimportantimplicationsforthemechanism overall topology of the toxins, as well as the size and bywhich thissuperfamilyofpeptidetoxinshasevolved. relative orientation of the four inter-cystine loops, ap- pears quite different. However, the structural overlay in Conclusions Figure 7D, which was generated automatically by the Spidersandother venomousanimalsrelyontheproduc- DaliLite structural alignment program [70], reveals that tion of pharmacologically complex venoms for defense, thetwostructuresareinfactremarkablysimilar. preycapture,andcompetitordeterrence.Themajorcom- The DaliLite alignment yields a root mean square ponentsofmostspider venomsaredisulfide-richpeptides deviation of 2.4 Å over the backbone atoms of the 28 that have evolved to target a wide range of receptors and aligned residues, indicating that the two toxins are in- ion channels in the insect nervous system. The ω-HXTX deed structural homologs. The three central disulfide andκ-HXTXfamilieswerethefirstpeptidesisolatedfrom bridges and loop 1 align remarkably well. Loop 4, which Australian funnel-web spiders that were shown to be in- encompasses the β-hairpin present in both toxins, also secticidal [17]. Analysis of all transcripts encoding these aligns well except for the four-residue insertion in peptides showed that they are initially expressed as pre- ω-HXTX-Hv1a (Figure 7A), which increases the size of propeptides that are proteolytically processed to yield a the hairpin loop at the tip of Loop 4. The major 36–37residuematurepeptidethatcontainsthreedisulfide Pinedaetal.BMCGenomics2014,15:177 Page10of16 http://www.biomedcentral.com/1471-2164/15/177 Figure6BayesianphylogenetictreerepresentingthemolecularevolutionaryhistoryoftheShivasuperfamilytoxins.Thetreeshows thesplitbetweenthethreemaintoxinclasses(ω,κ,andω/κ).ω-Actinopoditoxin-Mb1afromtheEasternmousespiderMissulenabradleyiwas usedastheoutgroup.Toxinsbelongingtoeachspeciesarehighlightedinthefollowingcolours:H.versuta,red;H.modesta,black;H.venenata, green;H.infensa,magenta;A.robustus,paleblue;H.formidabilis,darkblue;ω-actinopoditoxin-Mb1afromM.bradleyi,orange.*denotesaspecies notsequencedaspartofthisstudy;thesequencewasdownloadedfromUniProtunderaccessionnumberP83588. bridgesthatformanICKmotifplusanon-canonical vici- pharmacologicalsites.Thesetoxinsmaythereforebegood naldisulfidebondintheκ-HXTXs. candidates for in vitro evolution studies designed to pro- The extreme diversity of primary structure within the duce modified peptides with desired therapeutic [14] or Shiva toxin superfamily suggests that there have been agrochemical[11]properties.Mostimportantly,thisstudy few evolutionary restraints on sequence diversification reinforces the idea that the remarkable chemical and outside of the disulfide bridges that direct the 3D fold of pharmacological complexity of spider venoms may bede- these peptides. The ω-HXTXs, in particular, seem to rivedfromarelativelysmallnumberofancestralgenes. have evolved under the influence of positive Darwinian selection in an episodic fashion, whereas the κ-HXTXs Methods appear to be constrained by negative selection pressures. Identificationofω-HXTX-1homologsviarapid Functional assessments of these toxins should shed fur- amplificationofcDNAends ther light on why they have adopted quite contrasting Venom-gland cDNA libraries were prepared from indi- molecular evolutionary regimes. ω-HXTXs were also vidual specimens of the following species of Australian found to have adopted RAVER, where a large number of funnel-web spider (Arthropoda: Chelicerata: Arachnida: the episodically diversifying sites are concentrated on Araneae: Opisthothelae: Mygalomorphae: Hexathelidae): themolecularsurface,facilitatingthegeneration ofnovel Hadronyche infensa, H. versuta, H. venenata, and Atrax