biogeography of circum-antarctic springtails

Molecular Phylogenetics and Evolution 57 (2010) 48–58

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

journal homepage: www.elsevier .com/locate /ympev

Biogeography of circum-Antarctic springtails

Angela McGaughran a,*, Mark I. Stevens b,c, Barbara R. Holland a,d

a Allan Wilson Centre for Molecular Ecology & Evolution, Massey University, Private Bag 11-222, Palmerston North, New Zealandb South Australian Museum, Adelaide, SA 5000, Australiac School of Earth and Environmental Sciences, University of Adelaide, Adelaide, SA 5000, Australiad Institute of Fundamental Sciences, Massey University, Private Bag 11-222, Palmerston North, New Zealand

a r t i c l e i n f o

Article history:Received 7 September 2009Revised 20 May 2010Accepted 5 June 2010Available online 15 June 2010

Keywords:AntarcticaCryptopygusDispersalPhylogenySpringtailVicariance

1055-7903/$ - see front matter � 2010 Elsevier Inc. Adoi:10.1016/j.ympev.2010.06.003

* Corresponding author. Present address: Max PlancBiology, Department for Evolutionary Biology, SpemaD-72076, Germany. Fax: +49 (0)7071 601 498.

E-mail address: [email protected] (A. M

a b s t r a c t

We examine the effects of isolation over both ancient and contemporary timescales on evolutionarydiversification and speciation patterns of springtail species in circum-Antarctica, with special focus onmembers of the genus Cryptopygus (Collembola, Isotomidae).

We employ phylogenetic analysis of mitochondrial DNA (cox1), and ribosomal DNA (18S and 28S)genes in the programmes MrBayes and RAxML. Our aims are twofold: (1) we evaluate existing taxonomyin light of previous work which found dubious taxonomic classification in several taxa based on cox1analysis; (2) we evaluate the biogeographic origin of our chosen suite of springtail species based on dis-persal/vicariance scenarios, the magnitude of genetic divergence among lineages and the age and acces-sibility of potential habitat.

The dubious taxonomic characterisation of Cryptopygus species highlighted previously is confirmed byour multi-gene phylogenetic analyses. Specifically, according to the current taxonomy, Cryptopygus ant-arcticus subspecies are not completely monophyletic and neither are Cryptopygus species in general. Weshow that distribution patterns among species/lineages are both dispersal- and vicariance-driven. Epi-sodes of colonisation appear to have occurred frequently, the routes of which may have followed currentsin the Southern Ocean. In several cases, the estimated divergence dates among species correspond wellwith the timing of terrestrial habitat availability.

We conclude that these isotomid springtails have a varied and diverse evolutionary history in the cir-cum-Antarctic that consists of both ancient and recent elements and is reflected in a dynamic contempo-rary fauna.

� 2010 Elsevier Inc. All rights reserved.

1. Introduction

Antarctic terrestrial ecosystems are structured by many of thesame forces that influence evolution elsewhere; however, theimportance of isolation and its subsequent effects on populationand species differentiation is perhaps easier to appreciate on thecontinent locked in ice. In Antarctica, previous glacial cycling andcurrent dispersal barriers (e.g. glaciers) play a role in defining thelimits of distribution of different species, but other less obviousforces are also important (Rogers, 2007). For example, ecologicalproperties such as availability of liquid water and ice-free soilare variable, and patchiness in available habitat is likely to influ-ence population structure. Intrinsic characteristics of Antarctic ter-restrial biota such as microarthropods include an absence of wings,limited desiccation tolerance and reduced body size. Since these

ll rights reserved.

k Institute for Developmentalnnstrasse 37-39/IV, Tübingen

cGaughran).

characteristics are likely to have marked effects on dispersal capa-bilities of such taxa, they are also likely to strongly influence pop-ulation structure over time. Collectively, glaciological/ecologicalforces and species life history traits are therefore likely to havedominated and defined evolutionary processes in Antarctic terres-trial taxa through their isolating effects (e.g. Frati et al., 2001;McGaughran et al., 2008). In this context, the patchily distributedspringtails of Antarctic terrestrial ecosystems are good candidatesfor studies of the mechanisms of evolutionary processes such asspeciation.

To date 25 springtail species have been described in Antarctica(Greenslade, 1995), with most genera belonging to the Isotomidae,members of which are also widely distributed globally (Frati andCarapelli, 1999). Antarctic springtail distribution and taxonomyreceived considerable attention in the 1960s and early 1970s (e.g.Gressitt et al., 1963; Wise and Gressitt, 1965; Gressitt, 1967;Strandtmann, 1967; Wise, 1967, 1971; Wise and Spain, 1967; Wiseand Shoup, 1971). However, phylogenetic classification amongglobally distributed springtails has been a subject of disagreementamong various authors (e.g. D’Haese, 2002, 2003; Xiong et al., 2008)

http://dx.doi.org/10.1016/j.ympev.2010.06.003

mailto:[email protected]


http://www.sciencedirect.com/science/journal/10557903

http://www.elsevier.com/locate/ympev

A. McGaughran et al. / Molecular Phylogenetics and Evolution 57 (2010) 48–58 49

and Antarctic species have received little attention in this context(Frati et al., 2000). In fact, only one paper (based on mitochondrialDNA (mtDNA) cox1 variation) has examined phylogenetic relation-ships within Antarctic Isotomidae (Stevens et al., 2006).

This work examined several southern hemisphere isotomidspringtails, paying particular attention to the genetic relationshipsamong Cryptopygus species (Stevens et al., 2006). These authorsfound that several isotomid species were dubiously classified, col-lectively forming a non-monophyletic group. In fact, several mor-phologically recognisable but currently undescribed Cryptopygusspecies clustered into paraphyletic sister relationships with de-scribed Antarctic species (Stevens et al., 2006). An additional keyfinding of Stevens et al. (2006) was that some biogeographicallyseparate locations harboured species with shallow divergence(<2 MA), while other species showed deeper genetic divergences(>10 MA). This suggests that springtail taxa in Antarctica may havevariable evolutionary origins.

Antarctica is essentially a conglomerate of distinct regionsencompassing the sub-, maritime- and continental-Antarctic. How-ever, even regions within the continent are now recognised as dis-crete biogeographical elements (Chown and Convey, 2007). Forexample, the eastern and western parts of the Antarctic continentare very different (Chown and Convey, 2007) and at finer scales,no springtail species are shared across Antarctica’s ‘‘Gressitt Line”– a divide between continental-Antarctica and the Antarctic Penin-sula (Chown and Convey, 2007; Torricelli et al., 2010) (see Fig. 1).

The springtail fauna of Antarctica includes a high proportion ofendemic genera. For example, of the 10 species in easternAntarctica (six of which are from the family Isotomidae; Sinclair

Fig. 1. Distribution map of Antarctica and surrounding islands to show locations referredthe base of the Antarctic Peninsula. Colour codes inside location circles correspond to ouonly recently become available for habitation and we predict to find recently diverged taxthat are more deeply diverged (see text for further details).

and Stevens, 2006), 60% of genera and all species are endemic(Wise, 1967, 1971; Greenslade, 1995; Stevens and Hogg, 2006;Pugh and Convey, 2008; Torricelli et al., 2010). The high occurrenceof Antarctic endemics is often taken as evidence of divergence inisolation, suggesting survival of these groups through ancienttimes (Stevens et al., 2006; Convey and Stevens, 2007; Rogers,2007; Convey et al., 2008, 2009). In the context of Antarctic phylo-geography, recent mtDNA work has confirmed that some speciesshow divergence/speciation over multi-million year timescales(e.g. Stevens et al., 2007; McGaughran et al., 2008, 2009). Alterna-tively, non-endemic species with wider cosmopolitan distributionsare generally presumed to represent more recent introductions.

These concepts – ‘divergence in isolation’ and ‘recent introduc-tion’ – largely correspond to two evolutionary processes that areoften invoked to explain distribution patterns and the genetic rela-tionships among species. These are fragmentation of ancestral pop-ulations by vicariant events or colonisation of new areas bydispersal across a pre-existing barrier (Sanmartín and Ronquist,2004). In general terms, taxa are expected to have either: (1) sur-vived environmental conditions in Antarctica since the break-upof Gondwanaland and subsequent settlement of the major south-ern landmasses (Lawver et al., 1992; Crame, 1999; Convey et al.,2008, 2009) over 30 MA (i.e. vicariance origin) or (2) dispersed toAntarctica in more recent times (<5 MA; dispersal origin) followingcurrents in the Southern Ocean (e.g. the Antarctic CircumpolarCurrent (ACC), the West Wind Drift (WWD); Williams et al.,2003, and references therein). Of course, taxa may have dispersedto Antarctica in ancient times and then undergone vicariant frag-mentation of populations. Additionally, palaeoclimatic changes

to in the text, and from which samples were collected. Note the ‘‘Gressitt Line” nearr biogeographic hypotheses – ‘‘light grey” indicates habitats that are young or havea there; ‘‘black” indicates older habitats, which we predict will be inhabited by taxa

50 A. McGaughran et al. / Molecular Phylogenetics and Evolution 57 (2010) 48–58

(including very recent Pleistocene (<2 MA) climatic cycles) mayhave caused vicariance patterns among populations in more recenttimes (Clarke and Crame, 1992).

Variable habitat has been available across a variety of time-scales in the southern hemisphere for vicariant (fragmentation)and/or dispersal (colonisation) processes. In addition to the ancient‘Gondwanan’ landmasses (continental-Antarctica, Australia, NewZealand, South America), sub-Antarctic Îles Crozet (�8.7 Myr)and Îles Kerguelen (�100 Myr) have had available habitat overthe long-term. Elsewhere in the Southern Ocean, islands such asHeard, Macquarie and Marion Island, and coastal regions of theAntarctic Peninsula have all become available within the last mil-lion years. Thus, contemporary species distributions most likely re-flect a varied origin for Antarctic springtails (Frati and Carapelli,1999) that includes both relic and more recent immigrant species(e.g. Wallwork, 1973; Greenslade, 1995; Marshall and Pugh, 1996;Marshall and Coetzee, 2000; Pugh and Convey, 2000) and the ageof available habitat is likely to have been a key component in theoverall structuring of contemporary lineages.

To examine this more closely, we significantly extend the workof Stevens et al. (2006) outlined above. Our first aim is to evaluateexisting taxonomy of a suite of species of the springtail family Iso-tomidae from Antarctica and circum-Antarctic locations in light ofthe dubious taxonomic classifications identified in several taxa byStevens et al. (2006). We do this by adding several new individuals/locations to the existing cox1 dataset, and combining analyses ofmaternally inherited mtDNA (cox1) with bi-parentally inheritednuclear (18S and 28S rDNA) genes. We are particularly interestedto see if our expanded dataset confirms/refutes the paraphyly ofCryptopygus species identified by Stevens et al. (2006). Our secondaim is to evaluate the biogeographic origin of the genus Cryptopy-gus by using dating estimates and dispersal–vicariance (DIVA)analysis. We hypothesise that the Cryptopygus taxa will show a dy-namic evolutionary origin that includes both dispersal and vicari-ant-based episodes. In addition, we hypothesise that the blend ofboth shallow and deeply diverged Cryptopygus species identified

Table 1Species and locations.

Species Record L

Cryptopygus antarcticus antarcticus Willem (1901) AAS

‘‘Cryptopygus antarcticus” complex ACHMNT

Cryptopygus antarcticus maximus Deharveng (1981) ÎM

Cryptopygus antarcticus reagens Deharveng (1981) ÎCryptopygus antarcticus travei Deharveng (1981) MCryptopygus caecus Wahlgren (1906) M

SCryptopygus cisantarcticus Wise (1967) CCryptopygus dubius Deharveng (1981) MCryptopygus sverdrupi Lawrence (1978) SCryptopygus tricuspis Enderlein (1909) MAntarctophorus subpolaris Salmon (1962) BGressittacantha terranova Wise (1967) TNeocryptopygus nivicolus Salmon (1965) G

OutgroupDesoria klovstadi Carpenter (1902) CFolsomotoma marionensis Deharveng (1981) MParisotoma notabilis Schäffer (1896) MIsotomurus cf. palustris Müller (1876) M

List of species used in this study and relevant information, including species record, conumber of individuals (n).

by Stevens et al. (2006) will correspond to certain geographicalconstraints (i.e. age of available habitat). In particular, we predictthat species found in younger habitats (or habitats that have be-come available more recently) will correspondingly be home to lin-eages that have diverged/speciated more recently, and that olderless penetrable regions will harbour more deeply diverged lineagesthat have survived a long evolutionary period in isolation (see loca-tion colour codes in Fig. 1).

2. Methods

2.1. Species and locations

The nominate species Cryptopygus antarcticus was described in1901 (see Table 1). Subsequent work has shown there to be a vari-ety of undescribed species and subspecies of this genus (e.g. Dehar-veng, 1981; Potapov, 2001; Rusek, 2002; Deharveng et al., 2005;Stevens et al., 2006). Based on this, specimens of the genus Crypt-opygus, including several undescribed Cryptopygus antarcticus ‘sub-species’ (hereafter referred to as Cryptopygus a. ‘complex’) wereextensively sampled from a variety of locations across their entirerange throughout the continental-, maritime- and sub-Antarctic(Fig. 1). Twelve other related species (including four geographicallyrelevant isotomid springtails selected as outgroup taxa) with morerestricted geographic distributional ranges were heterogeneouslysampled based on specimen availability and successful sequencegeneration. In addition, several cox1 sequences from Stevenset al. (2006) were downloaded from GenBank. All sampling infor-mation including location and species are given in Table 1 and rel-evant GenBank accession numbers are given in Appendix A.

2.2. Sequence generation

Total DNA was extracted from specimens following a ‘saltingout’ protocol (Sunnucks and Hales, 1996). Upon extraction, frag-ments of the mitochondrial cytochrome c oxidase subunit I

ocation DIVA distribution n

ntarctic Peninsula 1 (PEN1) K 3ntarctic Peninsula 2 (PEN2) K 4outh Shetland Is. (SSI) J 7ustralia (AUS) E 1hile (CHI) B 5eard Is (HEA) M 1acquarie Island (MAQ) D 7ew Zealand (NZ) H 1asmania (TAS) E 2les Kerguelen (KER) I 8

acquarie Island (MAQ) D 1les Crozet (CRO) G 8

arion Island (MAR) A 1arion Island (MAR) A 1

outh Shetland Is. (SSI) J 2ape Hallett (CHA) C 3arion Island (MAR) A 5

or Rondane Mountains (SVE) O 1arion Island (MAR) A 1

eardmore Glacier (BEA) F 1erra Nova Bay (TER) N 3ranite Harbour (GRA) L 2

ape Hallett (CHA) n/a 1arion Island (MAR) n/a 1arion Island (MAR) n/a 1arion Island (MAR) n/a 1

llection location, distribution assigned to dispersal–vicariance (DIVA) analyses and

Table 2Primer list.

Primer Sequence (50–30)

TY-J-1460 tacaatttatcgcctaaacttcagccLCO1490 gttcaacaaatcataaagatattggaHCO2198 taaacttcagggtgaccaaaaaatca18S1F tacctggttgatcctgccagtag18S5R cttggcaaatgctttcgc18S3F gttcgattccggagaggga18SBI gagtctcgttcgttatcgga18SA2.0 atggttgcaaagctgaaac18S9R gatccttccgcaggttcacctac28SB tcggaaggaaccagctac28Srd1.2A cccssgtaatttaagcatatta28Sbout cccacagcgccagttctgcttacc

List of primers used in this study, and their sequences(50–30).

Table 3Datasets.

Datasets No. of base pairs

Individualcox1 (52 taxa) 36718S (39 taxa) 42728S (40 taxa) 502

CombinedND: 18S + 28S (27 taxa) 929AG: cox1 + 18S + 28S (27 taxa) 1296AG-short: cox1* + 18S + 28S (27 taxa) 1052AI: cox1 + 18S + 28S (71 taxa) 1296

List of the main DNA datasets (including outgroup taxa) used in this study and theirsequence length (No. of base pairs).* Second-codon positions only.


(cox1), 18S rDNA and 28S rDNA genes were amplified using theuniversal primers TY-J-1460 or LCO1490 and HCO2198 (Folmeret al., 1994) for cox1; 18S1F, 18S3R, 18S3F, 18S5R, 18SA2.0,18S9R (Giribet et al., 1996; Whiting et al., 1997) for 18S; and28Srd1.2A, 28SB (Whiting, 2002) or 28Sbout (Giribet et al., 2001;Prendini et al., 2003) for 28S (see Table 2).

Amplifications for each specimen used a 10 ll reaction volumecontaining 1 ll of extracted DNA (unquantified), 1� PCR buffer(Roche, Penzberg, Germany) 2.2 mM MgCl2, 0.2 mM of each dNTP(Boehringer-Mannheim, Mannheim, Germany), 1.0 lM of eachprimer and 0.5 U of Red Hot DNA polymerase (Thermo Scientific,United Kingdom). The thermal cycling conditions for cox1 were94 �C for 1 min followed by five cycles of denaturation and poly-merase amplification (94 �C for 1 min, 45 �C for 1.5 min, 1 min at68 �C) followed by 35 cycles of 94 �C for 1 min, 51 �C for 1.5 minand 1 min at 68 �C, followed by 5 min at 72 �C; and for 18S wereinitial denaturation at 94 �C for 3 min followed by 40 cycles ofdenaturation and polymerase amplification (94 �C for 30 s, 52 �Cfor 45 s and 1.30 min at 68 �C), followed by 4 min at 72 �C. Cyclingconditions for 28S were the same as those used for 18S except for a50 �C annealing temperature and a total cycle number of 35. Allreaction products were purified using SAP/EXO (USB Corp., Cleve-land, United States). Sequencing used forward and/or reverse prim-ers and was performed directly on a capillary ABI3730 geneticanalyser (Applied Biosystems Inc., California, United States) atthe Allan Wilson Centre Genome Service, Massey University.

All sequences were checked for consistency with springtail DNAusing the GenBank BlastN search. Nucleotide sequences from a to-tal of 16 described species and several members of the undescribedCryptopygus a. ‘complex’ (71 individuals in total) were edited inContigExpress (Vector NTI Advance ver. 10.3.0, Invitrogen Corpora-tion, United States) and aligned using default settings in ClustalWas implemented in MEGA ver. 4.1 (Tamura et al., 2007).

2.3. Data exploration

Initially, we focused on our separate ‘cox1’, ‘18S’ and ‘28S’ data-sets (Table 3). The cox1 dataset had no insertions or deletions,while any gaps generated in the alignment of 18S and 28S se-quences were excluded from all analyses (as were nucleotides oneither side of the gap until the next invariant site was encountered– this resulted in removal of 5 and 273 nucleotides from the 18Sand 28S datasets, respectively).

Exploratory data analyses were performed on these datasets inthe programme MrBayes ver. 3.1.2 (Ronquist and Huelsenbeck,2003) using the GTR+I+C model as chosen under the Akaike Infor-mation Criterion (AIC) in ModelTest ver. 3.7 (Posada and Crandall,1998). The Monte Carlo Markov Chain (MCMC) was run from 10

to 30 million iterations (to result in a final standard deviation of splitfrequencies of <0.01; all runs were further evaluated for chain con-vergence using Tracer ver. 1.4.1; Rambaut and Drummond, 2007)with a sampling frequency of 10,000. All other parameters in MrBa-yes were set to default. Three runs were performed for each dataset– one with no constraints enforced (‘unconstrained’), one with theoutgroup (corresponding to: Desoria klovstadi, Isotomurus palustris,Folsomotoma marionensis and Parisotoma notabilis; Table 1) set asa constraint (‘constrained’), and one with the outgroup excluded(‘no outgroup’). These analyses were repeated using default settings(i.e. GTR+C) and estimating the proportion of invariable sites fol-lowing a Maximum Likelihood (ML) approach in the programmeRAxML ver. 7.0.4 on its web-server (Stamatakis et al., 2008).

The 18S and 28S datasets behaved as expected during theseanalyses, with the outgroup consistently falling outside the in-group in consensus networks (Holland et al., 2005) generated inthe programme SplitsTree ver. 4.10 (Huson and Bryant, 2006) usinga 10% burn-in (as determined via the programme Tracer) and asplits threshold of 0.1. However, this was not the case for analyseswith the cox1 dataset, where the outgroup did not come togetherunless constrained to do so. Hence, we re-coded third-codon posi-tions as ‘‘RY” in this dataset and repeated the analyses in bothMrBayes and RAxML. Unfortunately, this did not correct the out-group problem and (as expected) caused a loss in resolution.

Comparison of trees generated for the single-gene datasets alsorevealed several areas of apparent gene discordance (see Section 3).Thus, we decided to generate several concatenated-gene datasets tocontinue our evaluation of Cryptopygus phylogeny, in the hope thatthe combined data would clarify this discordance. Four concate-nated datasets were generated: (1) an 18S + 28S dataset containingthe 27 individuals for which a sequence of both genes existed (‘nu-clear dataset’; hereafter referred to as ‘ND’); (2) a cox1 + 18S + 28Sdataset for those same 27 individuals (‘allgenes dataset’; or ‘AG’dataset); (3) the dataset from (2) but using only second-codon posi-tions for the cox1 gene (to clarify whether removal of the more rap-idly evolving first and third nucleotide positions would revealdeeper-level relationships that could be obscured by homoplasyfrom those sites) (‘all genes-short dataset’; or ‘AG-short’ dataset);and (4) a dataset for all 71 individuals in our study, some of whichwere missing a sequence for one or two genes (‘allindividuals’; or‘AI’ dataset) (Table 3). A partition homogeneity test performed inPAUP* ver. 4.0b10 (Swofford, 2002) showed that 18S and 28S data-set concatenation was inappropriate (P = 0.01), hence, all of ouranalyses on combined datasets treated partitions separately.

2.4. Phylogenetic analysis

As a result of our exploratory data analysis (above), we con-cluded that the best approach to generating final tree hypotheseswas to first determine the ingroup relationships using data with


the outgroup excluded (RAxML/MrBayes) and to then run RAxML/MrBayes on the complete dataset (including the outgroup) withthe ingroup relationships constrained to see where the outgroupwas positioned in relation to the (constrained) ingroup (see Hol-land et al. (2003) and Harrison et al. (2004) for discussion as tohow an outgroup can disrupt a correct ingroup).

We followed this approach to generate trees for the concate-nated datasets. Specifically, for each dataset, we used ingroup-onlyanalyses to generate an ingroup hypothesis, with bootstrap values(RAxML) and posterior probabilities (MrBayes), and then ran eachanalysis with the ingroup constrained (MrBayes) or set as a back-bone (RAxML) in order to get the best outgroup position. For eachanalysis, we assigned partitions according to genes, such that eachpartition was able to evolve independently under the GTR+I+Cmodel and all other parameter settings followed those employedduring data exploration (see above). For MrBayes, this involvedthe settings: lset applyto=(all) nst=6 rates=invgamma; unlinkstatefreq=(all) revmat=(all) shape=(all) pinvar=(all); prset app-lyto=(all) ratepr=variable. This means that the relative branchlengths are fixed/linked across partitions (but their scale variesacross partitions according to the rate multiplier parameters),and each partition is allowed its own GTR+I+C model parameters.All other priors in MrBayes were left at default settings. The RAxMLoutput includes the heuristically best ML-tree based on un-boot-strapped data. For each analysis, we used this RAxML ML-tree(no outgroup) as the final tree hypothesis, and rooted this accord-ing to the constrained analysis (above) following concordancechecks of the ML-tree ingroup relationships against majority-ruleconsensus trees of the RAxML (bootstrap) and MrBayes results.

2.5. Dating estimates

To obtain date estimates in order to help distinguish the relativeage of lineages, we used the programme PAUP* to estimateGTR+I+C (as chosen under the AIC in ModelTest; see above) dis-tances between sequences in the cox1 dataset. These distanceswere then averaged as appropriate to include only single represen-tative individuals of each species and/or location (i.e. lineage). Onlythis dataset was used because a commonly employed molecularclock exists for this gene in arthropods (see below).

Before proceeding with inference of relative dating estimates,we tested whether our cox1 sequences were evolving in a clock-like manner using the relative rates test of Tajima (1993) and ana-lysed the significance of these tests using a Bonferroni correction.This test requires a single outgroup taxon, so we repeated it foreach of the four outgroup species in the cox1 data; all tests indi-cated that the molecular clock assumption could not be rejected(see Section 3).

Because no fossil or geological calibrations are available forthese taxa, dating estimates were based on the molecular clockrate of 1.5–2.3% divergence per million years commonly appliedto arthropods (see Brower, 1994; Juan et al., 1996; Quek et al.,2004). The divergence estimate for the lower bound (1.5%) is de-rived from uncorrected p-distances (see Quek et al., 2004), whichdo not correct for multiple changes at nucleotide sites, so will be-come increasingly biased towards underestimating dates as diver-gences become older. Thus, we caution that our branch lengthsestimates will be longer and our older age estimates will likelybe larger compared to dating estimates quoted in the literaturefor uncorrected p-distances.

2.6. Dispersal–vicariance analysis

To reconstruct the distribution history of our springtail taxa, weused the dispersal–vicariance approach implemented in the pro-gram DIVA ver. 1.1 (Ronquist, 1996). In DIVA, a fully bifurcating

phylogeny is used to parsimoniously optimise the distribution ofancestral species. The method is based on optimisation of athree-dimensional cost matrix derived from a simple biogeo-graphic model. Distributions are described in terms of a set of unitareas, and speciation is assumed to divide ancestral distributionsallopatrically into mutually exclusive sets of unit areas. DIVA findsthe optimal distributions of ancestral species by minimising thenumber of dispersal and extinction events (Ronquist, 1996).

We performed our DIVA analysis based on the rooted ML-tree(see above) generated in RAxML for the AI dataset (no outgroup).We chose this dataset because it adds the widest taxon sampleand thus is less likely to be affected by taxon sampling bias (e.g.Zwickl and Hillis, 2002). In addition, the AI tree hypothesis hadthe highest bootstrap support/posterior distribution values of allof our analyses (see Section 3).

We pruned the AI tree to obtain a final tree hypothesis whereeach ‘lineage’ within each clade was represented (i.e. monophyleticgroups of the same species from the same location were pruned tohave only one representative). This resulted in a tree containing 29lineages.

The distribution of each species was classified as present/absentin 15 different areas corresponding to those given in Table 1. Theselection of these areas was based on the geographic distributionof species and, where possible, matched springtail species ende-mism. In the optimisations, we also ran analyses with the outgroupincluded, and both without constraint on the maximum number ofancestral areas allowed, and with ‘maxareas’ set to limit the max-imum allowable number of geographical areas of ancestral speciesto 2, 3, 4 and 5. This latter ‘maxareas’ evaluation was performed todeal with the known bias of DIVA toward ever increasing ancestralarea sizes as it approaches the tree root (Ronquist, 1996).

3. Results

3.1. Phylogenetic analysis

We present the rooted ML-trees from our RAxML exploratoryanalyses of the single-gene datasets in Appendix B. Appendix B(i)presents the cox1 results (for 48 taxa without the outgroup) andthis can be compared directly with Fig. 2 of Stevens et al. (2006),which presents RY-coded data. Our results are largely consistentwith those of Stevens et al. (2006), in particular confirming theplacement of their groups II and III (Appendix B(i)). All discrepan-cies between the two trees result from the addition of new speciesto our study. For example, our tree places Antarctophorus subpolarisdifferently – this species paired with C. dubius in Stevens et al.(2006), and here, falls in with C. sverdrupi (absent in Stevenset al., 2006) (Appendix B(i)). In addition, lineages of C. a. ‘complex’from Macquarie and Heard Islands fell within group I in Stevenset al. (2006), and here fall outside of that equivalent group witheither C. tricuspis or C. caecus (both absent in Stevens et al., 2006)(see Appendix B(i)). Our larger cox1 dataset confirms the mainfinding of Stevens et al. (2006) for potential non-monophyly withinCryptopygus – this is indicated in Appendix B(i) by the presence ofan asterisk everywhere a non-Cryptopygus species nests with aCryptopygus species. Appendix B(i) also confirms the dubious nat-ure of subspecies level classifications identified for C. a. ‘complex’species in Stevens et al. (2006). For example, C. a. ‘complex’ speciesfrom New Zealand, Chile, Heard and Macquarie Islands group withC. cisantarcticus, C. sverdrupi, A. subpolaris, Gressittacantha terranova,C. tricuspis, Neocryptopygus nivicolus and C. caecus. Thus, the addi-tion of new individuals/locations has not rectified the dubious tax-onomy currently available for Antarctic Cryptopygus species.

Comparison of the single-gene trees (Appendix B) revealsapparent evidence of gene discordance which is best evaluated


in the context of our biogeographic predictions (see location col-our codes in Fig. 1). When evaluated in this light, it becomesclear that the majority of predicted recently diverged lineages(i.e. inhabitants of younger habitats) fall out as expected in eachanalysis. For the cox1 dataset, the exceptions to this include thepredicted recently diverged C. a. travei lineage from Marion Is-land, which groups with C. a. maximus (Îles Kerguelen) and C.a. reagens (Îles Crozet) – these are both predicted as deeply di-verged lineages due to their distribution on older landmasses(Appendix B(i)). In addition, several predicted recently divergedlineages of C. a. ‘complex’ from Marion, Macquarie and Heard Is-lands group with predicted deeply diverged lineages from Chile,New Zealand and continental-Antarctic locations (Appendix B(i)).In the 18S dataset, predicted deeply diverged lineages from ÎlesKerguelen group with predicted recently diverged C. antarcticussubspecies (from Macquarie and Heard Islands, and the AntarcticPeninsula) (Appendix B(ii)) and C. a. travei again groups with apredicted deeply diverged lineage (C. a. ‘complex’ from New Zea-land). Similar examples can be cited from the 28S dataset(Appendix B(iii)).

In Fig. 2, we present the results for the ND (Fig. 2a) and ND-short (Fig. 2b) datasets. These trees are largely in agreement withboth each other and the trees shown in Appendix B. For example,both trees show that C. tricuspis and C. caecus (both from MarionIsland) are deeply diverged lineages that fall outside the mainCryptopygus group. In addition, a group of more recently divergedC. a. antarcticus species (from Antarctic Peninsula and South Shet-land Islands) and C. a. maximus species (from Macquarie Island andÎles Kerguelen) is present in both trees; two divergent lineages of C.a. ‘complex’ from Macquarie Island are apparent; and non-Crypto-pygus species fall with Cryptopygus species in several potentiallyparaphyletic relationships.

Fig. 2. ML-tree for of Cryptopygus and its close relatives, including the outgroup for:18S + 28S + cox1 dataset (second-codon positions only for the cox1 dataset) (27 taxa; 10posterior probabilities from the MrBayes Bayesian analyses (>50%) below nodes. In theunder the GTR+I+C model (see Section 2 for further information). Instances identifying polocation codes.

Because employment of wider taxon sampling generally resultsin tree hypotheses that are less biased by systematic error (seeZwickl and Hillis, 2002), we suspect that the tree topology gener-ated from the AI dataset is more accurate than that resulting fromthe AG analyses. Hence, we present the results of the AI and AGanalyses in Fig. 3, and Appendix C, respectively. Fig. 3 (the AI data-set results) largely replicates the findings described above forFig. 2. As for the single-gene datasets (Appendix B), few patternsin Fig. 3 and Appendix C go against our biogeographic predictions(see location colour codes in Fig. 1). These patterns again relate topredicted deeply diverged lineages of C. a. maximus from Îles Ker-guelen grouping with predicted recently diverged lineages of C. a.maximus from Macquarie Island and C. a. antarcticus (AntarcticPeninsula, South Shetland Islands) and C. a. ‘complex’ (MacquarieIsland). The subspecies C. a. travei (from Marion Island, and pre-dicted to be recently diverged) also groups with the predicted dee-ply diverged C. a. reagens (Îles Crozet) lineage in Fig. 3. Finally, C.tricuspis, C. dubius and C. caecus (all predicted recently diverged lin-eages from Marion Island) show deep divergence from the otherCryptopygus species (Fig. 3).

The relationships from the AI analyses (Fig. 3) show two maindifferences from those of the AG dataset (Appendix C). First, C.sverdrupi and A. subpolaris (new to Fig. 3) come out as sister taxain the AI dataset and no longer group with C. tricuspis and C. caecusas they do for the AG dataset (Appendix C). Second, C. a. travei andC. a reagens (absent from Appendix C) form a sister relationship inFig. 3, while the position of C. a. travei is unresolved in Appendix C.

In summary, our concatenated datasets present results that arelargely in agreement with the single-gene analyses, and we find ageneral improvement in bootstrap support values and posteriorprobabilities as we move from the single-gene to the concatenateddatasets.

(a) the concatenated 18S + 28S dataset (27 taxa; 929 bp); (b) the concatenated52 bp). ML bootstrap values (100 replicates) >50% are indicated above nodes withanalysis, partitions according to each gene were allowed to evolve independentlytential non-monophyly of Cryptopygus are indicated with an asterisk. See Table 1 for

Fig. 3. ML-tree for the allgenes (‘AI’) dataset of Cryptopygus and its close relatives, including the outgroup (total 71 individuals). ML bootstrap values (100 replicates) >50% areindicated above nodes with posterior probabilities from the MrBayes Bayesian analyses (>50%) below nodes. In each analysis, partitions according to the cox1, 18S and 28Sgenes were allowed to evolve independently under the GTR+I+C model (see Section 2 for further information). See Table 1 for location codes. Instances identifying potentialnon-monophyly of Cryptopygus are indicated with an asterisk.


3.2. Dating estimates

The Tajima (1993) test of relative rates did not indicate anypairs of taxa with significantly different rates in cox1 once the Bon-ferroni correction was applied. If the Bonferroni correction was ig-nored then, depending on the outgroup chosen, the test rejected 80(D. klovstadi), 121 (I. marionensis), 217 (I. palustris), or 28 (I. notabi-lis) pairs of taxa out of 1128 pairs tested at the 5% significance le-vel. This indicates that, although the clock assumption is notperfect for this data, it is not badly violated.

In the cox1 dataset (containing 27 representative ‘lineages’; seeSection 2 and Appendix D) GTR+I+C-based distances ranged from0.01 to 0.51 substitutions/site (mean: 0.30) within the ingroup[note that these GTR+I+C-based distances were generated inPAUP* while the branch lengths in Appendix B(i) were generatedin RAxML – the scale of these two distances should not becompared because ML values in RAxML cannot yet be directly com-

pared to likelihood values of other ML programmes (Stamatakis,2006)]. The average distance from the ingroup to the outgroupwas 0.36 substitutions/site, and among ‘subgroups’ of the ingroupranged from 0.25 to 0.35 substitutions/site. The only examples ofshallow divergence among lineages corresponded to three casesof Cryptopygus subspecies. This includes C. a. antarcticus from thetwo Antarctic Peninsula and South Shetland Island locations(�0.03 substitutions/site), C. a. ‘complex’ from Macquarie andHeard Islands (0.06) and C. a. maximus from Îles Kerguelen andMacquarie Island (�0.03). All other subspecies and species showeddeep divergence – an average distance of 0.51 substitutions/siteseparated the two Macquarie Island C. a. ‘complex’ lineages in thisdataset and divergence among the remaining C. a. ‘complex’individuals averaged 0.30 substitutions/site (Appendix D).

Using the generalised invertebrate molecular clock rate of 1.5–2.3% divergence per Myr suggests that the deepest splits withinCryptopygus occurred 22.2–34.0 MA (i.e. 0.51 substitutions/site),

Fig. 4. Re-worked version of Fig. 1 to show species distributions according to the results of the AI-DIVA analysis (see Section 2) for the ancient split of lineages correspondingto: ‘‘FLNO” (black); ‘‘EGHIJK” (light grey); ‘‘B” ‘‘C” (white). ‘‘A” (half white/half grey) and ‘‘D” (half black/half light grey) are shaded so as to indicate their presence across cladesin the DIVA output (see also Appendix E). Dispersal events as indicated in the DIVA analysis are shown by arrows connecting locations. Distributions correspond to those usedin Table 1 and Appendix E. See text and associated figures for further details.


while the shallowest splits have occurred in the last million yearsand up to �2.6–4 MA (for C. a. ‘complex’ from Macquarie and HeardIslands).

3.3. Dispersal–vicariance analysis

The best dispersal–vicariance analysis constructions (i.e. withthe fewest number of dispersals) in all cases were obtained usingno outgroup and no ‘maxareas’ restrictions. In addition, the ances-tral distributions towards the root were the same both with andwithout outgroups included.

As is the usual tendency in DIVA, the root node distributions arelarge and include many (or all) of the areas occupied by the termi-nal species (see Ronquist, 1996). However, the distributions atslightly shallower nodes of the tree were shown to be quite dis-junct. In particular, ancient splits were identified amongst lineagesfrom: ‘‘FLNO(M)”, ‘‘EGHIJK”, ‘‘B” and ‘‘C” and two special cases (‘‘A”and ‘‘D”) showed distributions which extended across >1 clade(Fig. 4, Appendix E; see also Table 1).

This differentiation relates closely to our biogeographic predic-tions (see location colour codes in Fig. 1). In fact the only true over-lap in distribution between older/younger habitats was for lineagesof C. a. reagens from older Îles Crozet and younger Marion Islandand C. a. maximus from older Îles Kerguelen and younger Marion

Island (see Appendix E). This pattern corresponds closely to thephylogenetic analyses.

The optimisation without outgroup and with no restriction on‘maxareas’ inferred a biogeographic scenario requiring 10 dispersalevents for the AI analysis. Exact dispersal events depend on selec-tion of specific ancestral areas when >1 is considered parsimoni-ous, however, the dispersal events present in all scenarios foreach analysis (indicated by ‘*’ in Appendix E) were between:Australia and New Zealand; Antarctic Peninsula and SouthShetland Islands; Macquarie and Heard Islands; Macquarie Islandand Îles Kerguelen; and Marion Island and Îles Crozet.

With a total of 27 individuals, there are up to 26 possible eventsin a fully resolved tree without polytomies for the AI analysis. Since10 of these are dispersal events and extinction events are consid-ered a cost in DIVA (Ronquist, 1996), many vicariant and/or speci-ation episodes are also predicted (i.e. an upper limit of 16). Thusthe biogeographic scenario inferred by DIVA suggests that thecurrent distribution is a result of a combination of vicariance anddispersal events from a disjunct ancestral distribution (see Fig. 4).

4. Discussion

The addition of more taxa/locations and two nuclear genes tothe existing cox1 dataset presented in Stevens et al. (2006) allowsus to confirm the taxonomic conclusions set forth in that paper. In


particular, we can confirm the potential non-monophyly of Antarc-tic Cryptopygus at higher levels. For example, sister relationshipswere confirmed or identified between various Cryptopygus species(e.g. C. caecus, C. sverdrupi, C. tricuspis) and several non-Cryptopygusspecies (G. terranova, N. nivicolus and A. subpolaris). It may be thatthese latter ‘non-Cryptopygus’ are actually Cryptopygus speciesrather than paraphyletic components of the Antarctic Isotomidae.The genus is in need of revision and careful attention should be gi-ven to this issue in future phylogenetic evaluations.

Species that were taxonomically described as derived subspe-cies of C. antarcticus were also confirmed as non-monophyletic inour analyses. These largely related to C. a. ‘complex’ individualsfrom ancient continental landmasses (Chile, Australia, TasmaniaNew Zealand, Antarctic continent), suggesting that these lineagesare not derived subspecies, but actually species as suggested byStevens et al. (2006). The high levels of genetic distance separatingthese lineages from other members of the ingroup (average of 0.30substitutions/site) support this claim.

Our phylogenetic analyses were affected by problems of genediscordance, low bootstrap support/posterior probabilities andinconsistent outgroup separation. This latter problem affectedthe cox1 dataset and begs the question as to whether our outgroupspecies were appropriate outgroups in this case. However, we feelthat the cox1 result was an artefact because the nuclear datasetsresolved the outgroup correctly in all analyses. Regardless, it isclear that more work is required to further resolve relationshipsfor both Antarctic Cryptopygus and southern hemisphere Isotomi-dae in general. Such work should also consider the possibility thatthere may be biological reasons for the conflict detected amongstour various datasets. For example, mitochondrial introgressionthrough hybridisation has been observed in other insects andnon-monophyly of morphologically defined species could be anadditional sign of potential introgression (e.g. Linnen and Farrell,2009). Of course, in the current case, our findings of non-monophyly may be due to taxonomic mis-specifications ratherthan non-monophyly within morphologically defined species (seeabove).

We predicted that our analyses would support a combinedvicariant/dispersal scenario explaining the evolutionary history ofour Antarctic Cryptopygus lineages. The DIVA analyses supportedthis conclusion, assigning 10 dispersal events and an upper boundof 16 vicariant events to optimally explain current distribution pat-terns for the AI dataset.

Hypothesised vicariant events are also supported by the veryhigh levels of genetic divergence found between several species(up to 0.51 substitutions/site). Our cox1 dating estimates suggestthat the deepest evolutionary splits among lineages occurred>20 MA (during the Miocene) and this requires the persistence ofthose lineages in Antarctica through multiple glacial cycles andother potential vicariant events. In practice, the vicariant scenariomight have involved ice sheet oscillations through time. These orother climatic factors may have induced repeated range contrac-tions, leading to fragmentation and isolation (allopatric speciation)of species in refugia during glacial maxima (Convey and Stevens,2007; Rogers, 2007; Convey et al., 2008, 2009). Several relevantstudies have invoked such vicariant explanations for contemporarydistributions of Antarctic taxa (e.g. Marshall and Coetzee, 2000;Allegrucci et al., 2006; Stevens et al., 2006).

On the other hand, dispersal events are indicated in the datasetby the presence of paraphyletic relationships among species fromdiffering geographic locations (assuming the current taxonomy iscorrect). Such potential paraphyly was found in the AI dataset be-tween C. a. antarcticus populations from the Antarctic Peninsulaand South Shetland Islands, C. a. reagens from Îles Crozet and Mar-ion Island, C. a. ‘complex’ from Australia and New Zealand and fromMacquarie and Heard Islands; and among C. a. maximus lineages

from Îles Kerguelen and Macquarie Island. In all five cases, theDIVA analysis predicted a dispersal event between these locations.Depending on the dataset, additional sister relationships werefound between C. a. travei (Marion Island) and both C. a. reagens(Îles Crozet; Appendix B(i)) and C. a. ‘complex’ (New Zealand;Appendix B(ii)), and between C. tricuspis (Marion Island) and C. cae-cus (Marion Island and South Shetland Islands; Appendix B(ii)).Among sub-Antarctic islands, dispersal events may have beenaided by oceanic currents between latitudes 40� and 60�. Certainly,the WWD has favoured such dispersal in circum-Antarctica, wherethe islands are proposed to act as ‘stepping stones’ to aid this pro-cess (McDowall, 1970; Fleming, 1979; Winkworth et al., 2002; San-martín and Ronquist, 2004). Other means of dispersal includewind- and bird-mediated transport (particularly via atmosphericcurrents in the case of wind), and contemporary introductionand/or spread by humans over recent timescales (e.g. Stevensand Hogg, 2002), especially in the sub-Antarctic (e.g. Frenotet al., 2005; Hughes et al., 2006).

In addition to predicting an evolutionary history involving dis-persal and vicariant events combined, we made biogeographicpredictions at the outset of this paper. Generally, we predictedthat we would identify both deep and shallow divergence amongdifferent lineages in our dataset (as shown by Stevens et al., 2006),and that this would correspond to a distinction between accessibleand non-accessible habitats (home to recently diverged andancient lineages, respectively). Our analyses largely confirmed thisprediction. For example, recently diverged lineages of C. a. antarc-ticus were found on the Antarctic Peninsula and South ShetlandIslands (both recently available). In addition, small genetic dis-tances separated C. a. ‘complex’ individuals from Heard (�1 Myr)and Macquarie (�0.3 Myr) Islands. Further, we did not find anyexamples of very closely related lineages anywhere on theAntarctic continent. This suggests that deglaciation on mainlandAntarctica has not exposed substantial contiguous land areaanywhere that has recently been colonised, and matches well withwhat we know about the comparative ease of species invasions inthe maritime- and sub-Antarctic (e.g. Frenot et al., 2005; Hugheset al., 2006).

There were some exceptions to our prediction, however. Thesmall genetic distance separating C. a. maximus from Îles Kerguelenand Macquarie Island is likely explained by a dispersal event fromthe older Îles Kerguelen (�100 Myr). A small genetic distance alsoseparated the predicted recently diverged C. a. travei lineagefrom Marion Island (�0.3 Myr) from the predicted deeply divergedlineage of C. a. reagens (Îles Crozet) and three other predictedrecently diverged lineages from Marion Island (C. tricuspis, C. caecusand C. dubius) also fell out as deeply diverged lineages in our concat-enated analyses. Since Marion Island lies just �19 km from the old-er Prince Edward Island (�8–18 Myr), the presence of deeplydiverged lineages on Marion Island may indicate the maintenanceof close faunal ties between these two locations. Analysis of sam-ples from Prince Edward Island could strengthen this suggestion.The final exception to our prediction was the presence of two Mac-quarie Island (�0.3 Myr) lineages that were separated by large ge-netic distances (0.51 substitutions/site). These clearly ancientlydiverged lineages must surely reflect separate colonisations oftwo distinct evolutionary lineages to this recently available habitat.

In summary, the evolutionary history of this collection ofAntarctic springtails reflects a diverse origin of both vicariant(Miocene) isolation and dispersal through time. We suggest thatthe former must imply the continuous presence of refugia onancient continental landmasses with appropriate environmentalconditions, and that the latter may have been aided in some partby the WWD and other oceanic currents as new island habitatshave become available. Finally, our results emphasise the stand-alone nature of deeper continental-Antarctica, which we attribute


in part to the difficulties of accessibility of this less penetrableregion.

Acknowledgments

We thank two anonymous reviewers, D. Penny and P. Conveyfor helpful comments on the manuscript, and D. Penny, J. Wangand M. Phillips for assistance and/or helpful suggestions regardinganalyses for data exploration. We are grateful to the many whohave contributed to sample collection over the years (includingC. D’Haese, I. Hogg, R. Seppelt, T.G.A. Green, C. Beard, B.J. Sinclair,M.R. Worland, P. Convey, S.L. Chown, C. Scheepers, E.A. Hugo, B.Rocko-Meyer, L. Sancho, K. Green, S. Thiele, D. Bergstrom, K. Kiefer,W. Vincent, R. Edwards and P. Greenslade). We are also grateful toD. Penny, P. Sunnucks, P. Greenslade and S.L. Chown for their sup-port. This work was funded by the New Zealand Tertiary EducationCommission (Top Achievers Doctoral Scholarship to A.M.), the Al-lan Wilson Centre for Molecular Ecology and Evolution, AustralianAntarctic Division (ASAC 2355 to M.S.), National Geographic (CREGrant 7790-05 to M.S.), Museum National d’Histoire Naturelle,Centre National de la Recherche Scientifique, Belgium 2009 BELDI-VA Expedition and Antarctica New Zealand (Programmes K024/K028). This paper contributes to the SCAR EBA and AntarcticaNew Zealand LGP research programmes.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.ympev.2010.06.003.

References

Allegrucci, G., Carchine, G., Todisco, V., Convey, P., Sbordoni, V., 2006. A molecularphylogeny of Antarctic Chironomidae and its implications for biogeographicalhistory. Polar Biol. 29, 320–326.

Brower, A.V.Z., 1994. Rapid morphological radiation and convergence among racesof the butterfly Heliconius erato inferred from patterns of mitochondrial DNAevolution. Proc. Natl. Acad. Sci. USA 91, 6491–6495.

Chown, S.L., Convey, P., 2007. Spatial and temporal variability across life’shierarchies in the terrestrial Antarctic. Philos. Trans. R. Soc. B 362, 2307–2331.

Clarke, A., Crame, J.A., 1992. The Southern Ocean benthic fauna and climate change:a historical perspective. Philos. Trans. R. Soc. B 338, 299–309.

Convey, P., Stevens, M.I., 2007. Antarctic biodiversity. Science 317, 1877–1878.Convey, P., Gibson, J.A.E., Hillenbrand, C.-D., Hodgson, D.A., Pugh, P.J.A., Smellie, J.L.,

Stevens, M.I., 2008. Antarctic terrestrial life – challenging the history of thefrozen continent? Biol. Rev. 83, 103–117.

Convey, P., Stevens, M.I., Hodgson, D.A., Hillenbrand, C.-D., Clarke, A., Pugh, P.J.A.,Smellie, J.L., Cary, S.C., 2009. Antarctic terrestrial life – ancient evolutionarypersistence or recent colonisation? Q. Sci. Rev. 28, 3035–3048.

Crame, J.A., 1999. An evolutionary perspective on marine faunal connectionsbetween southernmost South America and Antarctica. Sci. Mar. 63 (S1), 1–14.

Deharveng, L., 1981. Collemboles des Îles subantarctiques de l’Ocean Indien. ComitéNatl. French Res. Antarct. 48, 1–169.

Deharveng, L., Potapov, M., Bedos, A., 2005. Cylindropygus ferox gen. n., sp. n.: a newmember of the Cryptopygus complex (Collembola, Isotomidae) from centralFrance. J. Nat. Hist. 39, 2179–2185.

D’Haese, C.A., 2002. Were the first springtails semi-aquatic? A phylogeneticapproach by means of 28S rDNA and optimization alignment. Philos. Trans. R.Soc. B 269, 1143–1151.

D’Haese, C.A., 2003. Morphological appraisal of Collembola phylogeny with specialemphasis on Poduromorpha and a test of the aquatic origin hypothesis. Zool. Scr.32, 563–586.

Fleming, C.A., 1979. The Geological History of New Zealand and its Life. AucklandUniversity Press, Auckland.

Folmer, O., Black, M., Hoeh, W., Lutz, R., Vrijenhoek, R., 1994. DNA primers foramplification of mitochondrial cytochrome c oxidase subunit I from diversemetazoan invertebrates. Mol. Mar. Biol. Biotechnol. 3, 294–299.

Frati, F., Carapelli, A., 1999. An assessment of the value of nuclear and mitochondrialgenes in elucidating the origin and evolution of Isotoma klovstadi Carpenter(Insecta, Collembola). Antarct. Sci. 11, 157–171.

Frati, F., Fanciulli, P.P., Carapelli, A., Dell’ampio, E., Nardi, F., Spinsanti, G., Dallai, R.,2000. DNA sequence analysis to study the evolution of Antarctic Collembola.Ital. J. Zool. S1, 133–139.

Frati, F., Spinsanti, G., Dallai, R., 2001. Genetic variation of mtCOII gene sequences inthe collembolan Isotoma klovstadi from Victoria Land, Antarctica: evidence ofpopulation differentiation. Polar Biol. 12, 934–940.

Frenot, Y., Chown, S.L., Whinam, J., Selkirk, P.M., Convey, P., Skotnicki, M.,Bergstrom, D.M., 2005. Biological invasions in the Antarctic: extent, impactsand implications. Biol. Rev. 80, 45–72.

Giribet, G., Carranze, S., Baguñà, J., Riutort, M., Ribera, C., 1996. First molecularevidence for the existence of a Tardigrada + Arthropoda clade. Mol. Biol. Evol.13, 76–84.

Giribet, G., Edgecombe, G.D., Wheeler, W.C., 2001. Arthropod phylogeny based oneight molecular loci and morphology. Nature 413, 157–161.

Greenslade, P., 1995. Collembola from the Scotia Arc and Antarctic Peninsulaincluding descriptions of two new species and notes on biogeography. PolarPismo Entomol. 64, 305–319.

Gressitt, J.L., Leech, T.S., Wise, K.A.J., 1963. Entomological investigations inAntarctica. Pac. Insects 5, 287–304.

Gressitt, J.L. (Ed.), 1967. Entomology of Antarctica Amer. Geophys. Union. Antarct.Res. Ser., vol. 10, pp. 1–395.

Harrison, G.L., McLenachan, P.A., Phillips, M.J., Slack, K.E., Cooper, A., Penny, D., 2004.Four new avian mitochondrial genomes help get to basic evolutionary questionsin the late Cretaceous. Mol. Biol. Evol. 21, 974–983.

Holland, B.R., Penny, D., Hendy, M.D., 2003. Outgroup misplacement andphylogenetic inaccuracy under a molecular clock: a simulation study. Syst.Biol. 52, 229–238.

Holland, B.R., Delsuc, F., Moulton, V., 2005. Visualizing conflicting evolutionaryhypotheses in large collections of trees: using consensus networks to study theorigins of Placentals and Hexapods. Syst. Biol. 54, 66–76.

Hughes, K.A., Ott, S., Bölter, M., Convey, P., 2006. Colonisation processes. In:Bergstrom, D.M., Convey, P., Huiskes, A.H.L. (Eds.), Trends in AntarcticTerrestrial and Limnetic Ecosystems: Antarctica as a Global Indicator.Springer, Netherlands, pp. 35–54.

Huson, D.H., Bryant, D., 2006. Application of phylogenetic networks in evolutionarystudies. Mol. Biol. Evol. 23, 254–267.

Juan, C., Oromi, P., Hewitt, G.M., 1996. Phylogeny of the genus Hegeter(Tenebrionidae, Coleoptera) and its colonisation of the Canary Islandsdeduced from cytochrome oxidase I mitochondrial DNA sequences. Heredity76, 392–403.

Lawver, L.A., Gahagan, L.M., Coffin, M.F., 1992. The development of paleoseawaysaround Antarctica. In: Kennett, J.P., Warnke, D.A. (Eds.), The AntarcticPaleoenvironment: a perspective on global change. Antarct. Res. Ser., vol. 56.AGU, Washington, DC, pp. 7–30.

Linnen, C.R., Farrell, B.D., 2009. Phylogenetic analysis of nuclear and mitochondrialgenes reveals evolutionary relationships and mitochondrial introgression in thesertifer species group of the genus Neodiprion (Hymenoptera: Diprionidae). Mol.Phylogenet. Evol. 48, 240–257.

Marshall, D.M., Pugh, P.J.A., 1996. Origin of the inland Acari of ContinentalAntarctica, with particular reference to Dronning Maud Land. Zool. J. Linn.Soc. 118, 101–118.

Marshall, D.M., Coetzee, L., 2000. Historical biogeography and ecology of aContinental Antarctic mite genus, Maudheimia (Acari, Oribatida): evidence fora Gondwanan origin and Pliocene–Pleistocene speciation. Zool. J. Linn. Soc. 129,111–128.

McDowall, R.M., 1970. The galaxiid fishes of New Zealand. Bull. Mus. Comp. Zool.139, 341–432.

McGaughran, A., Hogg, I.D., Stevens, M.I., 2008. Patterns of population geneticstructure for springtails and mites in southern Victoria Land, Antarctica. Mol.Phylogenet. Evol. 46, 606–618.

McGaughran, A., Torricelli, G., Carapelli, A., Frati, F., Stevens, M.I., Convey, P., Hogg,I.D., 2009. Contrasting phylogeographic patterns for springtails reflect differentevolutionary histories between the Antarctic Peninsula and continentalAntarctica. J. Biogeogr. 37, 103–119.

Posada, D., Crandall, K.A., 1998. Modeltest: testing the model of DNA substitution.Bioinformatics 14, 817–818.

Potapov, M., 2001. Isotomidae. In: Dunger, W. (Ed.), Synopses on PalaearcticCollembola. Staatliche Museum für Naturkunde, Görlitz, Germany, pp. 1–603.

Prendini, L., Crowe, T.M., Wheeler, W.C., 2003. Systematics and biogeography of thefamily Scorpionidae Latreille, with a discussion on phylogenetic methods.Invert. Syst. 17, 185–259.

Pugh, P.J.A., Convey, P., 2000. Scotia Arc Acari: antiquity and origin. Zool. J. Linn. Soc.130, 309–328.

Pugh, P.J.A., Convey, P., 2008. Surviving out in the cold: Antarctic endemicinvertebrates and their refugia. J. Biogeogr. 35, 2176–2186.

Quek, S.-P., Davies, S.J., Itino, T., Pierce, N.E., 2004. Codiversification in an ant-plantmutualism: stem texture and the evolution of host use in Crematogaster(Formicidae: Myrmicinae) inhabitants of Macaranga (Euphorbiaceae). Evolution58, 554–570.

Rambaut, A., Drummond, A.J., 2007. Tracer ver. 1.4. Available from: <http://beast.bio.ed.ac.uk/Tracer/>.

Rogers, A.D., 2007. Evolution and biodiversity of Antarctic organisms: a molecularperspective. Philos. Trans. R. Soc. B 362, 2191–2214.

Ronquist, F., 1996. DIVA version 1.1. Computer Program and Manual Available byAnonymous FTP from Uppsala University (ftp.uu.se or ftp.systbot.uu.se).

Ronquist, F., Huelsenbeck, J.P., 2003. MRBAYES 3: Bayesian phylogenetic inferenceunder mixed models. Bioinformatics 19, 1572–1574.

Rusek, J., 2002. Do we have Cryptopygus-representatives (Collembola: Isotomidae)in Europe? Pedobiologia 46, 302–310.

Sanmartín, I., Ronquist, F., 2004. Southern hemisphere biogeography inferredby event-based models: plant versus animal patterns. Syst. Biol. 53, 216–243.


http://beast.bio.ed.ac.uk/Tracer

http://beast.bio.ed.ac.uk/Tracer


Sinclair, B.J., Stevens, M.I., 2006. Terrestrial microarthropods of Victoria Land andQueen Maud Mountains, Antarctica: implications of climate change. Soil Biol.Biochem. 38, 3158–3170.

Stamatakis, A., 2006. RAxML-VI-HPC: maximum likelihood-based phylogeneticanalyses with thousands of taxa and mixed models. Bioinformatics 22, 2688–2690.

Stamatakis, A., Hoover, P., Rougemont, J., 2008. A rapid bootstrap algorithm for theRAxML web-servers. Syst. Biol. 75, 758–771.

Stevens, M.I., Hogg, I.D., 2002. Expanded distributional records of Collembola andAcari in southern Victoria Land, Antarctica. Pedobiologia 46, 485–496.

Stevens, M.I., Greenslade, P., Hogg, I.D., Sunnucks, P., 2006. Examining southernhemisphere springtails: could any have survived glaciation of Antarctica? Mol.Biol. Evol. 23, 874–882.

Stevens, M.I., Hogg, I.D., 2006. Contrasting levels of mitochondrial DNA variabilitybetween mites (Penthalodidae) and springtails (Hypogastruridae) from theTrans-Antarctic Mountains suggest long-term effects of glaciation and lifehistory on substitution rates, and speciation processes. Soil Biol. Biochem. 38,3171–3180.

Stevens, M.I., Frati, F., McGaughran, A., Spinsanti, G., Hogg, I.D., 2007.Phylogeographic structure suggests multiple glacial refugia in northernVictoria Land for the endemic Antarctic springtail Desoria klovstadi(Collembola, Isotomidae). Zool. Scr. 36, 201–212.

Strandtmann, R.W., 1967. Terrestrial prostigmata (Trombidiform mites). Antarct.Res. Ser. 10, 51–95.

Sunnucks, P., Hales, D.F., 1996. Numerous transposed sequences of mitochondrialcytochrome oxidase I–II in aphids of the genus Sitobion (Hemiptera: Aphididae).Mol. Biol. Evol. 13, 510–524.

Swofford, D.L., 2002. PAUP*: Phylogenetic Analysis using Parsimony (* and OtherMethods), Version 4.0b10. Sinauer, Sunderland.

Tajima, F., 1993. Simple methods for testing the molecular evolutionary clockhypothesis. Genetics 135, 599–607.

Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular evolutionarygenetics analysis (MEGA) software version 4.0. Mol. Biol. Evol. 24, 1596–1599.

Torricelli, G., Carapelli, A., Convey, P., Nardi, F., Boore, J.L., Frati, F., 2010. Highdivergence across the whole mitochondrial genome in the ‘‘pan-Antarctic”springtail Friesea grisea: evidence for cryptic species? Gene 449, 30–40.

Wallwork, J.A., 1973. Zoogeography of some terrestrial micro-arthropoda inAntarctica. Biol. Rev. 48, 233–259.

Whiting, M.F., Carpenter, J.C., Wheeler, Q.D., Wheeler, W.C., 1997. The Strepsiptera-problem: phylogeny of the holometabolous insect orders inferred from 18S and28S ribosomal DNA sequences and morphology. Syst. Biol. 46, 1–68.

Whiting, M.F., 2002. Mecoptera is paraphyletic: multiple genes and phylogeny ofMecoptera and Siphonaptera. Zool. Scr. 31, 93–104.

Williams, S.T., Reid, D.G., Littlewood, D.T.J., 2003. A molecular phylogeny of theLittorininae (Gastropoda: Littorinidae): unequal evolutionary rates,morphological parallelism, and biogeography of the Southern Ocean. Mol.Phylogenet. Evol. 20, 60–86.

Winkworth, R.C., Wagstaff, S.J., Glenny, D., Lockhart, P.J., 2002. The origins andevolution of the genus Myosotis L. (Boraginaceae). Mol. Phylogenet. Evol. 24,180–193.

Wise, K.A.J., Gressitt, J.L., 1965. Far southern animals and plants. Nature 207, 101–102.

Wise, K.A.J., 1967. Collembola (springtails). Antarct. Res. Ser. 10, 23–148.Wise, K.A.J., Spain, A.V., 1967. Entomological investigations in Antarctica, 1963–64

season. Pac. Insects 9, 271–293.Wise, K.A.J., 1971. The Collembola of Antarctica. Pac. Insects Monogr. 25, 57–74.Wise, K.A.J., Shoup, J., 1971. Entomological investigations in Antarctica, 1964–65

season. Pac. Insects Monogr. 25, 27–56.Xiong, Y., Gao, Y., Yin, W., Luan, Y., 2008. Molecular phylogeny of Collembola

inferred from ribosomal RNA genes. Mol. Phylogenet. Evol. 49, 728–735.Zwickl, D.J., Hillis, D.M., 2002. Increased taxon sampling greatly reduces

phylogenetic error. Syst. Biol. 52, 588–598.

biogeography of circum-antarctic springtails

Documents