Exceptional cryptic diversity and multiple origins of parthenogenesis in a freshwater ostracod

 Sonnensystem

 2 views
of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Description
Exceptional cryptic diversity and multiple origins of parthenogenesis in a freshwater ostracod
Share
Tags
Transcript
  Exceptional cryptic diversity and multiple srcins of parthenogenesisin a freshwater ostracod S.N.S. Bode a,b, * , S. Adolfsson c,d,e , D.K. Lamatsch a,b,f  , M.J.F. Martins c,d,g,h , O. Schmit g,i , J. Vandekerkhove g,h,i ,F. Mezquita i , T. Namiotko h , G. Rossetti g , I. Schön a,1 , R.K. Butlin b,1 , K. Martens a,1 a RBINS, Dept. of Freshwater Biology, Rue Vautier 29, 1000 Brussels, Belgium b University of Sheffield, Dept. of Animal and Plant Sciences, Western Bank Sheffield, S10 2TN, UK  c EAWAG, Swiss Federal Institute of Aquatic Science and Technology, Department of Aquatic Ecology, Überlandstrasse 133, 8600 Dübendorf, Switzerland d ETH-Zürich, Institute of Integrative Biology (IBZ), Universitätstrasse 16, 8092 Zürich, Switzerland e GEMI, UMR CNRS IRD 2724, IRD, 911 Avenue Agropolis, BP 64501, 34394 Montpellier Cedex 5, France f  Institute of Limnology, Austrian Academy of Sciences, 5310 Mondsee, Austria g University of Parma, Dept. of Environmental Sciences, Viale G.P. Usberti 33A, 43100 Parma, Italy h University of Gdansk, Dept. of Genetics, Lab. of Limnozoology, Kładki 24, 80-822 Gdan´sk, Poland i University of Valencia, Dept. of Microbiology and Ecology, Avenue Dr. Moliner 50, 46100 Burjassot, Spain a r t i c l e i n f o  Article history: Received 15 May 2009Revised 5 August 2009Accepted 18 August 2009Available online 22 August 2009 Keywords: Cryptic speciesSpecies complexGeographical parthenogenesisAsexual reproductionEvolution of sexCoalescencemtDNAOstracoda Eucypris virens a b s t r a c t The persistence of asexual reproduction in many taxa depends on a balance between the srcin of newasexual lineages and the extinction of old ones. This turnover determines the diversity of extant asexualpopulationsandsoinfluencestheinteractionbetweensexualandasexualmodesofreproduction.Specieswith mixed reproduction, like the freshwater ostracod (Crustacea) morphospecies  Eucypris virens , are agood model to examine these dynamics. This species is also a geographic parthenogen, in which sexualfemalesandmalesco-existwithasexualfemalesinthecircum-Mediterraneanareaonly,whereasasexualfemalesoccuralloverEurope. Amolecularphylogenyof  E. virens  basedonthemitochondrialCOI and16Sfragments ispresented. It ischaracterisedbymanydistinctclustersof haplotypeswhichareeitherexclu-sively sexual or asexual, with only one exception, and are often separated by deep branches. Analysis of thephylogenyrevealsanastonishingcrypticdiversity,whichindicatestheexistenceofaspeciescomplexwith more than 40 cryptic taxa. We therefore suggest a revision of the single species status of   E. virens .The phylogeny indicates multiple transitions from diverse sexual ancestor populations to asexuality.Although many transitions appear to be ancient, we argue that this may be an artefact of the existenceof unsampled or extinct sexual lineages.   2009 Elsevier Inc. All rights reserved. 1. Introduction Sex is the most common mode of reproduction in the animalkingdom(seefor example, Bell, 1982). It isalso the ancestral modeof all metazoa and probably all eukaryotes. Yet it is easily lost bymutation, hybridisation, or polyploidisationandit also comes withcosts,suchasthetwofoldcostofmales(MaynardSmith,1978),vul-nerabilitytosexuallytransmittedinfectionsorpredatorsandmanymore (Kondrashov, 1993). The reason behind the success of sex isone of the most intriguing paradoxes of evolution still waiting tobe solved. Many hypotheses have been proposed but, so far, noneappears capable of providing a general and inclusive explanationfor widespread sexual reproduction (West et al., 1999; Butlin,2002).Thebest-supportedhypothesesimplicatemutationaccumu-lationduetolackof recombinationinasexuals(Kondrashov, 1988;Lynch et al., 1993) or rapidly changing selection pressures due tothe ‘Red Queen’ process (Van Valen, 1973), particularly as a resultof host–parasite interactions (Hamilton, 1980). It is becomingincreasingly evident that several factors must be involvedsimulta-neously (West et al., 1999) and that their contributions may differamong organisms (Burt, 2000; Schön et al., 2009).Asexual mutants can have significant selective advantages dueto superior population growth and colonising abilities. However,they have not generally replaced their sexual relatives but insteadthey are typically short-lived in evolutionary terms (MaynardSmith, 1978; Bell, 1982). It is difficult to estimate the ages of asex-ual lineages (Butlin, 2002). Therefore apparent exceptions to therule are controversial, ironically called ‘ancient asexual scandals’( Judson and Normark, 1996). The most well-known putative an-cient asexuals are bdelloid rotifers (Welch and Meselson, 2000),darwinulid ostracods (Martens et al., 2003; Martens and Schön, 1055-7903/$ - see front matter   2009 Elsevier Inc. All rights reserved.doi:10.1016/j.ympev.2009.08.022 *  Corresponding author. Address: University of Sheffield, Dept. of Animal andPlant Sciences, Western Bank Sheffield, S10 2TN, UK. Fax: +44 (0) 1142220002. E-mail address:  s.bode@sheffield.ac.uk (S.N.S. Bode). 1 These authors contributed equally to this work. Molecular Phylogenetics and Evolution 54 (2010) 542–552 Contents lists available at ScienceDirect Molecular Phylogenetics and Evolution journal homepage: www.elsevier.com/locate/ympev  2008), and oribatid mites (Heethoff et al., 2007; Laumann et al.,2007). These are entire families that have been fully asexual formillions of years, a conclusion supported by several lines of evi-dence, such as genomic characteristics in the case of bdelloids(Welch and Meselson, 2001; Welch et al., 2004) and extensive fos-sil records without males in the case of the darwinulids (Martenset al., 2003; Martens and Schön, 2008; but see Smith et al., 2006) and oribatid mites (Maraun and Scheu, 2000).Ancient asexual lineages are genetically diverse because of anaccumulation of mutations since their srcin. In other taxa, genet-ically diverse asexual populations may be generated by a high rateoforiginresultinginastandingdiversitythatisdependent,atleastin part, on population size ( Janko et al., 2008). High rates of srcinof asexuals occur incyclical parthenogens throughrepeatedloss of the sexual phase, as in aphids for example (Delmotte et al., 2001;Loxdale and Lushai, 2003), and where there is repeated hybridisa-tion, as in  Poeciliopsis  fish (Mateos and Vrijenhoek, 2002) or  Daph-nia  (Paland et al., 2005). In some cases, frequent srcin of asexuallineages from an ancestral sexual population clearly occurs, butthe mechanism remains uncertain, as in  Potamopyrgus  snails (Nei-man et al., 2005). It is important to understand both the mecha-nism and rate of origin of asexual lineages because theyinfluence models for the maintenance of co-existing sexual andasexual populations. For example, Janko et al. (2008) show thatasexual lineages may be short-lived purely because of turn-overin a finite population, without the need to invoke declining fitnessdue to mutation accumulation or environmental change. In con-trast, in  Potamopyrgus , frequent srcin of novel asexual lineagesis necessary to maintain the asexual population in the face of co-evolving parasites (Lively et al., 2004).The freshwater ostracod (Crustacea)  Eucypris virens  is a modeltaxon containing distinct and non-cyclical sexual and asexual lin-eages. The reproductive mode is the only known difference be-tween these groups of lineages. This species belongs to theCyprididae, a family that harbours many species with either fullyasexual, fully sexual or mixed modes of reproduction (Horneet al., 1998; Martens et al., 2008). It belongs to the superfamilyCypridoidea, which is related to the superfamily Darwinuloideawhose extant representatives are the putatively ancient asexualostracod family of the Darwinulidae, mentioned above. Eucypris virens  ischaracterisedbyhighlevelsofvariability,bothat the morphological (Martens, 1998; Baltanas et al., 2002) and ge-netic level (Rossi et al., 1998; Schön et al., 2000; Schön, 2007).Schön et al. (2000) suggested that both multiple srcins of asexuallineages and hybridisation might contribute to this high diversity.They also proposed that some asexual lineages might be old, withsequencedivergenceofthemitochondrialCOIlocusupto21%indi-cating that the oldest asexual lineages might have srcinated 10million years ago. However, wider sampling of lineages wasneeded to test these ideas and fossil evidence is scarce, as this par-ticular species occupies only temporary pools. Eucypris virens  is also known for its distinct geographical distri-bution in which males are restricted to the circum-Mediterraneanarea and have never been recorded north of the Alps (Horne et al.,1998) whereas asexual lineages occur in Northern Europe as wellas in sympatry with sexual lineages. This pattern is termed ‘‘geo-graphical parthenogenesis” (Vandel, 1928) and has been recordedin a number of organisms (Law and Crespi, 2002; van Dijk, 2003).This phenomenonmaybeexplainedbythe greater colonisingabil-ity of asexuals, their potential to adapt to novel environments be-cause they are free of the retarding effects of gene flow (see Butlinet al., 2003, for further discussion), or the difference in environ-mental stability between regions (see Horne and Martens, 1999,for a review).Here, we investigate the number of sexual and asexual lineagesof   E. virens  in Europe and how they are related to each other phy-logenetically, using partial mitochondrial cytochrome oxidase I(COI) and 16S ribosomal DNA gene sequences. This phylogeneticinformation provides a framework for understanding the srcinof the genetic diversity within asexuals, the rate of appearance of asexual lineages and their ages, as well as their geographic distri-bution. Asexual organisms are thought to have originated fromsexual ancestors without reversals. This assumption is supportedbymultiplelinesofevidence(Bell,1982) andbytheargumentthatit requires more evolutionary steps to re-acquire sexual reproduc-tion than it does to lose it (but see Domes et al., 2007). Althoughmost species of the Cyprididae are asexual or include asexual lin-eages,itisverylikelythattheancestralstateofthefamilywassex-ual reproduction. Therefore, we assume that the ancestor of   E . virens  wassexualandtestwhethertransitionsfromsexualtoasex-ual reproduction are frequently encountered across the phylogenyor whether a single ancient event has led to diverse asexual lin-eages. Furthermore, we determined the number of independentsexual lineages (potential cryptic species) within the morphospe-cies because srcin of asexuals from distinct sexual ancestorsmay contribute to clonal diversity. 2. Materials and methods  2.1. Specimen collection and DNA extraction We conducted a large-scale sampling campaign covering mostof Europe and the Mediterranean including North Africa (Fig. 1).Between 50 and 100 specimens of   E. virens  were collected with ahand net from each of 135 temporary pools during winter andearly spring of 2005/2006 and 2006/2007 and were kept alive inmineral water overnight to eliminate stomach contents. The ani-mals were slowly killed in diluted ethanol, in order that their car-apace valves stayed open, and specimens were subsequentlytransferred into 100% ethanol and kept at 4  C. Details of the sam-pling sites utilised in this study, selected from a database of fieldsampling data compiled by all members of the SexAsex project(http://evirens.group.shef.ac.uk/), are reported in Table 1S and have been visualised in Fig. 1 using  DIVA-GIS v.5.4  software (Hij-mans et al., 2001).Specimens were chosen for analysis at random except that wepreferentially used individuals whose valves were open and intact,as those were good indications that specimens were alive andhealthy when they were killed. Genomic DNA was extracted fromthe ostracod softparts using the DNeasy Blood and Tissue Kit (Qia-gen) according to the manufacturer’s protocol, after washing eachspecimen with ultrapure water and PBS (1  concentration) buffer.Valves were retained for morphometric analysis (to be reportedelsewhere). A subset of ostracods was prepared for both allozymeanalysis and DNA extraction with subsequent sequencing of COI(Adolfsson et al., submitted for publication).  2.2. DNA sequencing  Our choice of the first marker, the mitochondrial cytochromeoxidase I (COI), was based on the availability of universal primers,its high rate of evolution (Lunt et al., 1996) and its widespread usefor DNA barcoding which make it especially suited to examine in-tra-specific relationships in a taxon where there is limited geneticinformation. A 657bp fragment of the COI gene was amplifiedusing the following universal invertebrate primers: HCO2198 (for-ward) 5 0 -TAAACTTCAGGGTGACCAAAAAATCA-3 0 and LCO1490 (re-verse) 5 0 -GGTCAACAAATCATAAAGATATTGG-3 0 (Folmer et al.,1994). A few populations from Corfu and Italy could not be ampli-fiedwiththisCOI primerpairandthereforeanotherpairofspecificprimers (FMCO 5 0 -TAGGACAGCCRGGATCWCT-3 0 and RMCO 5 0 - S.N.S. Bode et al./Molecular Phylogenetics and Evolution 54 (2010) 542–552  543  CGGTCTGTTAAWAGCATWGTGA-3 0 ), resulting in a 476bp frag-ment, was designed using  Primer3 v.0.4.0  (Rozen and Skaletsky,2000).PCR amplification was performed using a Westburg Biometra TPersonal thermal cycler starting with 3min of denaturation at95  C followed by 35 cycles of 30 s at 95  C, 54.5  C for 30 s of annealing time, 1.5min at 72  C of extension and a final extensionstage of 10min at 72  C. The 20 l l volume reactions contained0.5Urecombinant Taq  DNApolymerase(Invitrogen),1  PCRBuffer(200mM Tris–HCl [pH 8.4], 500 mM KCl), 2mM dNTP, 10 l M of each primer, 1.5mM MgCl 2 , and   8ng template DNA, althoughDNA yield was variable. Difficult template DNA was amplifiedusing the HotStar Taq Master Mix Plus Kit (Qiagen) applying themanufacturer’s protocol with the same conditions as above, but atotal of 37 cycles were used with an annealing temperature of 48  C, or 45  C for the primer pair FMCO and RMCO.The second marker, 16S, was chosen to complement the COIinformation with a more conserved marker (DeSalle et al., 1987).A480bpfragmentwasamplifiedfromthemitochondrial16SrRNAgene usingprimer 16SL (forward) 5 0 -CGCCTGTTTAACAAAAACAT-3 0 and as reverse 16SH 5 0 -CCGGTCTGAACTCAGATCACGT-3 0 or 16SBr5 0 -CCGGTCTGAACTCAGATCACGT-3 0 (Palumbi, 1996). Amplificationwas performed with the HotStar Taq Master Mix Plus Kit and 35cycles of 50 s denaturation at 94  C, 50 s annealing at 50  C, and1min 20 s extension at 72  C. Conditions were otherwise asdescribed above, except that less DNA was used. PCR productswere run on 1.4% agarose gels, stained with SYBR   Safe (Invitro-gen) and photographed. The remaining products were cleanedusing the GFX TM PCR DNA and Gel Band purificationkit (GE Health-care) following the manufacturer’s protocol and eluting accordingto concentration estimates from the gel images. The cleaned PCR products were directly sequenced in both directions with the PCR primers and the ABI BigDye Terminator v.1.1. kit and then furtherpurified with an ethanol/EDTA precipitation method (ABI cyclesequencing kit manual) to be run on an ABI 3130xl GeneticAnalyser.Chromatograms were edited in  CodonCode Aligner v.1.6.3 . andunambiguously aligned through the ClustalW algorithm in  MEGAv.4.0  (Tamura et al., 2007). Sequences were checked for identitybyBLASTsearchforhighlysimilarsequences(megablast)(Altschulet al., 1990). Putative COI fragments scored highly with other COIsequences of   E. virens  in Genbank (Schön et al., 2000). So far, therehave been no submissions of 16S from  E. virens  to Genbank. How-ever, the 16S sequences generated had high similarity with other16S ostracod sequences. COI was also translated in  MEGA  usingthe invertebrate mitochondrial genetic code to test for the pres-ence of amplifiedpseudogenes (Bensassonet al., 2001) identifiablethroughstopcodonswithinthereadingframe(Songetal., 2008).Atotal of 449  E. virens  sequences was obtained for COI (Table 1S). Inorder to reduce the computational effort, we used 374 sequencesin phylogenetic analyses, excluding 75 that were identical to se-quences already present in the data set. For 16S, 219 sequenceswere obtained. Three outgroups from the same subfamily (Eucyp-ridinae) (Meisch, 2000),  Eucypris pigra ,  Tonnacypris lutaria  (COI and16S) and  Eucypris crassa  (16S only) were also sequenced using thesame conditions as for  E. virens . All edited sequences are availablefrom Genbank, Accession Nos. GQ914057 – GQ914280 (16S) andGQ914281 – GQ914731 (COI).Some PCR amplifications were only successful after several at-tempts, probably because of low quantity or quality of DNA or, insome cases, because of sequence divergence in the primer regions(especially for COI). This increased the risk of amplifying contami-nating DNA. Therefore, we checked for compatibility of COI and16Ssequenceswherebothwereavailableandremoved18individ-uals where there was evidence of contamination (identical COI se-quences for individuals with divergent 16S sequences, or viceversa).EightoftheexcludedindividualswerefromoneTurkishsite(LAD) whichhad divergent 16S sequences and inconsistent COI se-quences. Some individuals from this site were also divergent mor-phologically and probably do not belong to  E. virens .  2.3. Phylogenetic analyses Bayesian phylogenetic analyses were performed using  Beast v.1.4.8 . (Drummond and Rambaut, 2007) and maximum likelihood(ML)analysesusing PhyML v.2.4.4. (GuindonandGascuel,2003),on Fig. 1.  Sampling sites of   Eucypris virens , corresponding to  Table  1S. The map was produced using  DIVA-GIS v.5.4. 544  S.N.S. Bode et al./Molecular Phylogenetics and Evolution 54 (2010) 542–552  the COI and 16S datasets separately, and on the concatenation of the two as a partitioned supermatrix.  Modeltest v.3.7   (Posada andCrandall, 1998) was run on both separate datasets to test for themost adequate model of evolution. The programs  Dnaml  and Dnamlk  (Felsenstein and Churchill, 1996) from the  Phylip packagev.3.68  (http://Evolution.genetics.washington.edu/phylip.html)were run in order to test whether the data conformto a molecularclock. In addition, the following three models implemented in Beast  , were applied to a preliminary COI dataset and their likeli-hoodswerecompared:General timereversiblewithgammadistri-bution and invariable sites (GTR+ C +I); SRD06 (Hasegawa–Kishino–Yano [HKY] with the first and second position partitionedfrom the third position for each codon); and HKY with three parti-tions for the three bases per codon (HKY+ C +3). Xml files for the Beast   runs were prepared using  Beauti v.1.4.8.  The searches wererun assuming a strict molecular clock but no clock rate was speci-fied. For the remaining parameters, the default settings of   Beauti were used, except that Jeffrey’s prior distribution parameter wassetto5.0.WeappliedaYulealgorithmdescribingapure-birthspe-ciation process rather than a population-based coalescent process(this decision was based on preliminary runs which showed thatdeep branches were present in the dataset).For each dataset, three independent searches were run for10,000,000 generations each.  Tracer v.1.4.  (Rambaut and Drum-mond, 2007) was used for visualising the output of the Bayesiantreesearchesandthetreesfromallrunswerecombinedusing Log-Combiner v.1.4.8,  re-sampling every 10,000th tree and therebyobtaining 3,000 trees.  TreeAnnotator v.1.4.8.  was used to discard aburn-in of the first 1,000 trees and to choose one tree out of 2,000 remaining trees based on the likelihoods of their topologies.Trees were manipulated and rotated to facilitate comparisons of the topologies between the various datasets (and methods) using FigTree v.1.2  (http://tree.bio.ed.ac.uk/software/figtree/).A partition homogeneity test (Farris et al., 1994), also calledincongruence length difference test (ILD), was performed with1,000 replicates on the concatenated COI+16S dataset using  PAUP v.4.0 b10 . (Swofford, 2003). In addition, the two separate datasets,with congruent individuals only, were run in separate Beastsearches and their combined likelihoods were compared with thelikelihood from the concatenated dataset to confirm that the twosets of sequences are compatible with a common tree, as expectedfor two sections of the mitochondrial genome. The concatenateddatasetconsistedof 156individualsof  E. virens  andtwooutgroups,for which both regions were sequenced (with ends trimmed: COIbp67-bp605; 16S bp82-bp411). The Bayesian search parametersand models were as above with the single gene datasets, but man-ually combined in Xml format to allow partitioned evolutionarymodels but a common tree search.TwoMLsearchesin PhyML  withneighbour-joiningstartingtree,substitution model parameters estimated during the search and500 non-parametric bootstrap replicates each were performed forthe COI andtheconcatenateddatasets. Theirtopologieswerecom-pared and bootstrap values were combined, to provide 1,000 boot-strap replicates.  2.4. Quantitative cluster analysis In order to test whether our set of samples can be divided intodistinctclustersofrelatedhaplotypes,weappliedamaximumlike-lihood method developed by Pons et al. (2006, provided as an R script by T. Barraclough and M.T. Monaghan). This method teststhe proposition that the gene tree can be described by two pro-cesses, a pure-birth process generating a set of populations and acoalescent process within populations, and determines the mostlikely transition time between these phases of evolution. The tran-sition point defines a set of clusters or singleton sequences each of whichcanbeconsideredasasamplefromanindependentlyevolv-ingpopulation(seeSection4).ThemethodwasappliedtothestrictclocktreeoftheCOIdatasetobtainedfrom Beast, usingthepackage R v.2.8.0.  (R Development Core Team, 2008) withthe extension apev.2.2–2  (Paradis et al., 2004) for phylogenetic applications. Thenumbers of lineages were plotted against time and a threshold be-tween speciation and coalescence was fitted using the generalmixed Yule coalescent (GMYC) model. A  v 2 test was performed(M.T.Monaghan,pers.comm.)totestforsignificanceofapplicationof this clustering model against the null model of a uniformbranching pattern. Support limits for the number of clusters wereestimated fromthe likelihoodsurface for the time of transitionbe-tween phases. Clusters of sequences, or singletons, were labellednumerically (with prefix ‘p’) and classified as either sexual, in thepresence of at least one male, asexual in the absence of males, ormixed in the case of presence of males in the same cluster as fe-males from asexual populations (sites where males have not beenobserved). Assignments of reproductive modes to clusters wereaided by allozyme and flow cytometry analyses of a subset of thesequencedindividuals(Adolfssonet al., submittedfor publication).The clusters defined by this analysis of the COI data set were fittedonto the topologies of the 16S and ML trees. Fromhere on, we willrefer to the sequence clusters inferred by this analysis simply as‘clusters’.A limitation of the Pons et al. (2006) method is that it fits a sin-gle transition time between the pure-birth and coalescent phases.This may make the outcome sensitive to over-sampling of individ-ual genotypes. Therefore, we applied the method not only to theCOI dataset used in the phylogenetic analysis (374 sequences)but also to thefull set of 449sequences andto aset withtwolargegroups of identical haplotypes reduced to a single sequence persite (406 sequences, excluding 26 sequences from cluster p34 col-lectedatsitesCOAandCOBand17sequencesfromclusterp25col-lected at sites DR1, BER, BRC, CC3, CC4, CC5 and MF2).  2.5. Other statistical analyses Meansequencediversities( p ) withinandbetweenclusters, andTajima’s  D  statistics within clusters, were calculated in MEGA v.4.1. (Tamura et al., 2007) using the complete deletion option. Diversityvalueswereusedtotestwhethersister-pairsofclustersconformtothe 4X rule (Birky et al., 2005), which aims to delineate indepen-dent populations in a simpler, but conceptually similar way tothe method of  Pons et al. (2006). AT content in COI was also calcu-lated with  MEGA. Although we sampled widely across Europe, it is almost certainthat our dataset does not include all extant lineages of   E. virens .This is especially true for the extant sexual lineages, since sexualpopulationshaverestricteddistributions(seebelow)andnosexualpopulations to the east of the Mediterranean basin were found.Therefore, we conducted rarefaction analyses, based on the COIdata set. We simulated rarefaction for all asexuals (249 individu-als; 96 localities), all sexuals (127 individuals; 28 localities), andthe two reproductive modes together (376 individuals; 117 locali-ties of which 6 contain both sexual and asexual individuals) usingthe program  EstimateS v.8.0.0.  (Colwell, 2005) in order to estimatehow many more clusters we would expect to find if the samplingeffort were increased, given the existing data and sampling infor-mation. Samples were randomized 1,000 times without replace-ment. The following estimators for the total number of clustersto be expected, and their respective confidence intervals (whereapplicable), were extracted: Chao 1 and Chao 2 with bias-correc-tion (Chao, 1987), Jackknife 2 (Smith and van Belle, 1984) and Michaelis–Menten (Colwell and Coddington, 1994), using the rec-ommended ‘MMMeans’ option whereby the estimates of eachsample pooling level are computed just once. S.N.S. Bode et al./Molecular Phylogenetics and Evolution 54 (2010) 542–552  545  In order to summarise the spatial distribution of each cluster,we calculated the centroid of all sites where the cluster was pres-entandthemeandistanceofindividualsamplesitesfromthatcen-tral position, using the Haversine formula (see http://www.movable-type.co.uk/scripts/gis-faq-5.1.html). 3. Results  3.1. Phylogenetic analysis of COI data Increasing evidence for frequent nuclear insertions of mito-chondrial DNA, known as numts, has lead to the conclusion thatthe high diversity of COI lineages detected in DNA barcoding pro- jects may be an overestimate of the true diversity (Song et al.,2008). This phenomenon is particularly frequent in crustaceans(Bensasson et al., 2000, 2001). However, the sequences obtainedhere contain neither stop codons nor indels and can be alignedunambiguously, despite a large number of base pair differences.It has also been suggested (Song et al., 2008) that a lower AT con-tent can indicate the presence of numts due to a different compo-sitional bias between mitochondrial and nuclear DNA. In our data,the base composition of the COI sequences was biased towardsA+T (A=27.4; T=33.0; G=17.3; C=22.3), which is consistentwith expectations for the mitochondrial genome.Modeltest v.3.7 found TIM+ C +I to be the preferred model of substitution for the total COI dataset according to the AIC andTVM+ C +I according to the hierarchical likelihood ratio tests(  lnLk=9226.437 and   lnLk=9229.16 for models chosen byAIC and hLRTs, respectively). The TIM+ C +I model requires lessparameters to be estimated ( K   =8) than TVM+ C +I ( K   =9). Bothmodels are nested within the HKY model, which is implementedin the packages used and so was our chosen model. In ML treesearches,weusedHKY+ C +Iwith6ratecategoriesforthegammadistribution. In the independent model comparison implementedin  Beast  , the HKY model with three codon positions partitionedwas preferred over the SRD06 and GTR+ C +I models (Bayes fac-tors calculated in  Tracer   (Suchard et al., 2001) 300.6nats and52.5nats, respectively, where ‘nats’ are Bayes factor units on a nat-ural logarithm scale) and was therefore chosen for subsequentanalysesexceptwhereotherwiseindicated.Weused6ratecatego-ries. Because this model tends to fit a lowsubstitution rate for 2ndpositionsofcodons,thereisnorequirementforaseparatecategoryof invariant sites. Three combined independent  Beast   runs weresufficient to achieve effective sample size (ESS) parameter valuesabove the recommended threshold of 200 (Drummond and Ram-baut, 2007) for confident parameter estimation and a smooth uni-modal trace curve. The likelihoods of each run were similar and onthis basis it was concluded that the three independent runs con-verged to a similar optimum. The test of the molecular clock with Dnaml/Dnamlk  showed a higher likelihood for the non-clocklikemode of evolution, but the improvement was only marginally sig-nificant (  2 D LL=420.13, df=372,  p  =0.043). Therefore we haveassumed clock-like evolution in subsequent analyses.TheCOIdataarecharacterisedbygroupsofcloselyrelatedhaplo-types which are separated by surprisingly large distances (  p -dis-tance up to 14.5% within  E. virens ). This is reflected in thephylogenetic analysis (Fig. 2) which shows multiple clades of closely related sequences, well supported by both Bayesian andML analyses, which are connected by long branches. Support fordeeper nodes was relatively weak because of saturation of COIsequence divergence, leading to some inconsistencies between the Fig. 2.  Bayesian phylogenetic strict clock tree of COI dataset. Triangles at the tips represent clusters determined with the Pons et al. (2006) method. Lines without trianglesare singletons. Clustersare labelled numerically, p1top40, pstandingfor ‘provisional species’, andlabels are next tothe triangles onthe right. Widths of triangles reflect thenumbers of individuals in the sample belonging to the clusters (part a only), whereas depth indicates maximum divergence within clusters (parts a–c). Full lines and blackfills represent purely sexual groups. White triangles refer to asexual clusters with dashed lines, which indicate that the transition from sexual to asexual reproductive modemust have happened somewhere along this branch. A dotted line with a grey triangle indicates a mixed cluster. Outgroups are indicated as grey lines. Values of posteriorprobabilities (above nodes) and bootstrap percentages (below nodes) are represented by symbols (posterior probability; bootstrap): star (>0.95; 95%), black dot (>0.7; 70%)white dot (>0.5; 50%). Values for the basal nodes of clusters are placed beside the cluster names (posterior probability; bootstrap). The bootstrap values obtained by  PhyMLv.2.4.4.  with 1000 bootstrap replicates, were superimposed from congruent ML topologies. (a) Full Bayesian COI tree .  The bar indicates units of substitutions per site. (b)Expansionof thetoppart of thephylogeny, indicatedas(b) in(a). Thesameappliesto(c) whichisthe bottompart ofthe phylogeny, indicatedas(c) in(a). This treeandbothML trees with bootstrap supports are available in Newick format as Supplementary data.546  S.N.S. Bode et al./Molecular Phylogenetics and Evolution 54 (2010) 542–552
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks