Parallel diversification of Australian gall-thrips on Acacia

M.J. McLeish a,*, B.J. Crespi b, T.W. Chapman c, M.P. Schwarz d

a South African National Biodiversity Institute, Kirstenbosch Research Centre, Private Bag X7, Claremont, Cape Town 7735, South Africab Behavioural Ecology Research Group, Department of Biological Sciences, Simon Fraser University, Burnaby, BC, Canada V5A 1S6

c Memorial University, St. John’s, Newfoundland, Canada A1B 3X9d Flinders University, Sturt Road, Bedford Park, Adelaide, South Australia 5042, Australia

Received 3 January 2006; revised 12 March 2007; accepted 14 March 2007Available online 27 March 2007

Abstract

The diversification of gall-inducing Australian Kladothrips (Insecta: Thysanoptera) on Acacia has produced a pair of sister-clades,each of which includes a suite of lineages that utilize virtually the same set of 15 closely related host plant species. This pattern of parallelinsect-host plant radiation may be driven by cospeciation, host-shifting to the same set of host plants, or some combination of theseprocesses. We used molecular-phylogenetic data on the two gall-thrips clades to analyze the degree of concordance between their phy-logenies, which is indicative of parallel divergence. Analyses of phylogenetic concordance indicate statistically-significant similaritybetween the two clades. Their topologies also fit with a hypothesis of some degree of host–plant tracking. Based on phylogenetic andtaxonomic information regarding the phylogeny of the Acacia host plants in each clade, one or more species has apparently shiftedto more-divergent Acacia host–plant species, and in each case these shifts have resulted in notable divergence in aspects of the phenotypeincluding morphology, life history and behaviour. Our analyses indicate that gall-thrips on Australian Acacia have undergone paralleldiversification as a result of some combination of cospeciation, highly restricted host–plant shifting, or both processes, but that theevolution of novel phenotypic diversity in this group is a function of relatively few shifts to divergent host plants. This combinationof ecologically restricted and divergent radiation may represent a microcosm for the macroevolution of host plant relationships andphenotypic diversity among other phytophagous insects.! 2007 Elsevier Inc. All rights reserved.

Keywords: Parallel divergence; Cospeciation; Host-switch; Phylogenetics; Gall-thrips; Acacia

1. Introduction

Evolutionary conservation of associations between plantand phytophagous insect groups is a central theme in biol-ogy and provides a platform for testing hypotheses rich inscope (Futuyma and Moreno, 1988; Jermy, 1993; Kelleyet al., 2000; Craig et al., 2001; Johnson et al., 2002; Nyman,2002; Ward et al., 2003; Zerega et al., 2005; Jousselin et al.,2006; McLeish et al., 2007). Coevolution theory (Ehrlichand Raven, 1964) was the historic impetus driving workendeavouring to penetrate factors explaining radiations

of both phytophagous insects and their host–plants viaselective responses to one another over a relatively longperiod. ‘Coevolution’ has also been used to demonstratejoint speciation of interacting lineages, or cospeciation(Herre et al., 1996; Clayton et al., 1999; Page, 2003). How-ever, the extent to which phytophagous insects and theplants with which they interact exert selection on oneanother is complex, highly varied among lineages, andunclear (Jermy, 1984, 1993; Ballabeni et al., 2003). In thisstudy, we infer a phylogeny of gall-inducing thrips on Aus-tralian Acacia and test hypotheses concerning how thisplant–insect assemblage has evolved.

Gall-inducing insects are tightly constrained to mecha-nisms by which speciation might proceed. Australian gall-inducing thrips are phytophagous insects that have evolved

1055-7903/$ - see front matter ! 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.ympev.2007.03.007

* Corresponding author. Fax: +27 0 21 797 6903.E-mail address: mcleish@sanbi.org (M.J. McLeish).

www.elsevier.com/locate/ympev

Molecular Phylogenetics and Evolution 43 (2007) 714–725

strategies permitting specific utilisation of desert Acaciaspecies for food and shelter and for these reasons gall-induction imposes a level of phylogenetic constraint (Cor-nell, 1983; Jermy, 1993; Farrell and Mitter, 1998a; Craiget al., 2001; Ward et al., 2003). Host race formation isapparent in gall-thrips on Acacia (Crespi et al., 2004). Aswell as preadaptation to closely related host plants, cospe-ciation and host switching across related plants has beenshown to result in life history shifts, host specialisation,and the macroevolutionary conservatism in resource use(Ehrlich and Raven, 1964; Berlocher, 2002; Crespi et al.,1998, 2004; Despres and Jaeger, 1999; Cronin and Abra-hamson, 2001; Dres and Mallet, 2002; Machado et al.,2005; Rønsted et al., 2005).

Gall-inducing thrips are a monophyletic group thatinhabit species of Plurinerves, Juliflorae, and PhyllodinaeAcacia subgenera, or sections. Putative gall-thrips specieson closely related hosts are of particular interest. Presum-ably, these taxa have recently diverged and are expectedto include taxa near or below species-level and provide amore transparent interpretation of cladogenesis in gall-thrips with fewer extinction events obscuring thrips-Acaciaassociations. Kladothrips rugosus Froggatt and Kladothripswaterhousei Mound and Crespi induce galls on the same 14Plurinerves host species showing a high degree of distribu-tion overlap. The phylogenetic relationships among thesecryptic taxa are yet to be resolved. Cryptic species are dif-ferent species that cannot be easily distinguished on thebasis of morphology and is indicative of recently divergedspecies (Jaenike, 1981; Parsons and Shaw, 2001). Theapparent cryptic species K. rugosus and K. waterhouseicomplexes appear overwhelmingly host-specific, theyinduce disparate taxon-specific gall morphologies, and pre-liminary molecular work using COI sequence data andmicrosatellites have provided strong evidence for species-level divergence among them (Mound, 1971; Moundet al., 1996; Crespi et al., 1997, 1998; McLeish et al.,2006). However, some of these putative species show littlegenetic divergence, which is suggestive of host race popula-tion’s status.

The scope of this work does not include discussion ofspecies definitions, but contends that levels of polymor-phism, below that of species, exist in our dataset andrequire elaboration. Genetic distances among K. rugosusand among K. waterhousei populations show that a largemajority of the K. rugosus and K. waterhousei complexmembers are apparently di!erent species, though addi-tional diagnoses would be useful. Measures of gene flowhave to be determined to show reproductive isolation. Inaddition to genetic distance, it is also crucial to use behav-ioural and ecological criteria to identify species (Ferguson,2002). Both the phylogenetic inferences indicate gall struc-ture is highly conserved amongst all newly sampled popu-lations. It is commonly accepted that gall morphology islargely under the control of the insect genome and repre-sents an extended phenotype (Stern, 1995; Crespi and Wor-obey, 1998; Morris et al., 2002; Stone and Schonrogge,

2003). Fidelity of gall structure over di!erent host speciesis consistent with gall phenotype being largely determinedby the thrips genotype and therefore a potentially usefuldiagnostic character in species identification. Species-spec-ificity of gall morphology is evident in other insect orders(sawflies: Nyman et al., 2000: wasps: Cook et al., 2002).Recent molecular work (McLeish et al., 2006) has showntwo K. rugosus populations, each of which induces a dis-crete gall type, once believed to be the same species, aredi!erent.

TheK. rugosus andK. waterhousei complexes thus appearto represent ecological replicates (Johnson and Clayton,2003) sharing the same set of host species, each expectedto cluster into a separate clade and respond in parallel tohost speciation via cospeciation, host switching and/or hostrace formation. These clades thus represent an excellentopportunity to test for parallel diversification and evaluatethe roles of historical contingency and selection in evolu-tionary change (Ricklefs and Schluter, 1993).

1.1. Modes of speciation

Speciation in gall-thrips might proceed by the formationof host-related races where there is reduced gene flowamong populations of a single species parasitising two ormore localised host species leading to reproductive isola-tion (Jaenike, 1981, 1990; Emelianov et al., 1995; Parsonsand Shaw, 2001; Dres and Mallet, 2002). Speciation via ahost-shift can be thought of as a transition from polymor-phism (e.g. for host preference) to host race preceding atransition from host race to reproductively isolated species.Host races are maintained by reduced gene flow predomi-nantly via di!erential host preference. By contrast,host-related sibling species are reproductively isolated forreasons in addition to di!erential host preference (Jaenike,1981). Genetic divergence data suggests that host-relatedraces of gall-thrips are actually a series of host specificsibling species, which is consistent with the strong host–plant specificity shown in virtually all other gall-inducinginsects (Crespi et al., 2004; Rohfritsch and Shorthouse,1982).

Cospeciation between phytophagous insects and theirhosts, parasites, or mutualists has been clearly demon-strated in a number of cases, most of which involve stronghost–plant specificity and intimate insect–plant relation-ships such as gall-induction or complex physiological andlife history adaptation (Ronquist and Nylin, 1990; Baker,1996; Herre et al., 1996; Machado et al., 1996; Roderick,1997; Roderick and Metz, 1997; Farrell and Mitter,1998b; Burckhardt and Basset, 2000; Clark et al., 2000; Iti-no et al., 2001; Weiblen and Bush, 2002; Weiblen, 2004).The majority of studies, however, indicate that congruencebetween insect and host plant phylogenies is partial or non-existent, and thus host-shifting appears to be the more pre-valent mechanism in determining the associations of insectsand their hosts (Humphries et al., 1986; Weintraub et al.,1995; Janz and Nylin, 1998; Dobler and Farrell, 1999; Janz

M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725 715

et al., 2001; Jones, 2001; Lopez-Vaamonde et al., 2001;Ronquist and Liljeblad, 2001). Consequent to host-shift-ing, fitness tradeo!s between hosts, or ecological diver-gence of derived, host-shifted populations, may spur theevolution of reproductive isolation (Joshi and Thompson,1997; Hawthorne and Via, 2001; Nosil et al., 2002), andcolonisation of new host–plant lineages may provideopportunities to diversify rapidly (Ehrlich and Raven,1964; Mitter et al., 1988; Farrell and Mitter, 1998b).

Cospeciation between gall-inducing thrips and host Aca-cia lineages has been suggested at a macroevolutionaryscale in an explicitly phylogenetic context (Crespi et al.,2004). Two thrips lineages, each producing a morphologi-cally discrete elongate or pouched gall type on Acacia sec-tions Plurinerves and Juliflorae can be traced from twoancestral gall-inducing species on a single ancestral Acacialineage. Both derived thrips lineages have retained theancestral elongate-pouched gall type combination. Ahypothesis of cospeciation makes three predictions (Crespiet al., 2004): (1) phylogenies of parallel thrips lineagesshould be identical or very similar to each other, (2) phy-logenies of thrips lineages should be identical or very sim-ilar to that of the host plants, and (3) speciation eventsamong parallel thrips lineages and host lineages shouldbe contemporaneous. Deviations from an identical matchwould be indicative of processes other than cospeciationoperating. Alternatively, the K. rugosus and K. waterhouseigroups might have independently converged onto the sameset of closely related Acacia conducive to Kladothrips gall-induction, as host-shifts are reported (Craig et al., 1994) tomore freely occur among taxonomically and phylogeneti-cally similar plants.

Contradicting the model involving complete cospecia-tion is evidence that indicates the apparent coincidencein several species of major morphological and life historychanges accompanying host switches between more dis-tantly related host lineages that are not inhabited by clo-sely related thrips sister-species (Crespi et al., 2004). Theseswitches are also evidenced by the absence of elongate-pouched gall type combinations and by the presence ofonly a single gall type on the novel host species, as inKladothrips intermedius, K. rodwayi and K. morrisi (Crespiet al., 2004). Convergence of thrips lineages amongrelated hosts would predict their phylogenies to be inde-pendent of one another. In this paper we test for parallelspeciation in the K. rugosus and K. waterhousei speciescomplexes, and evaluate hypotheses for their joint diversi-fication on Australian Acacia. To do so, we first extendand revise the current gall-thrips phylogeny (Morriset al., 2001) with addition of K. rugosus, K. waterhousei,K. habrus, and K. intermedius ‘races’ from di!erent Acaciaspecies; and second, use the phylogeny to test for parallelpatterns of diversification between the K. rugsosus and theK. waterhousei groups, which would be indicative of cospe-ciation of each of these groups with their Acacia hosts,or the parallel evolution of the same set of host–plantshifts.

2. Materials and methods

2.1. Collections, DNA extractions, PCR, and sequencing

Taxa were collected from widely distributed Acacia pop-ulations across Australia (Table 1). Voucher specimens ofthese taxa have been deposited in the Australian NationalInsect Collection (ANIC) at CSIRO Entomology in Can-berra. Gall morphology is highly conserved within eachgall-thrips taxon with structural diversity exhibited amongstthem (Crespi andWorobey, 1998), and was used in conjunc-tion with host species identification (Maslin, 2001) to dis-criminate amongst gall-thrips races. To test whether gallstructure can be used as an indicator of association betweenlike-types, we mapped three distinct gall structure categoriesonto focal taxa in our phylogenies: (1) ‘spiky’ galls have veryobvious pointed protrusions: (2) ‘elongate’ galls are thosethat are elongate or tubular, some of which have subtlesurface textural qualities such as fine striations: and(3) ‘pouched’ galls that form from a ‘ballooning’ of petioletissue from one surface of the phyllode with a noticeablymore narrow ostiole than is formed in elongate galls.

Fragments of cytochrome oxidase I (COI), elongation fac-tor — one alpha (EF-1a), wingless, and 16S (ribosomalRNA) gene regions were sequenced. The DNA extractionsused for the sequencing data were from fresh tissue frozento !80". To maximise DNA yield each tissue extractioncomprised of all individuals in a single gall, the brood ofone female (Chapman et al., 2000), using a GENTRA SYS-TEMS DNA Extraction Kit. Amplifications of DNA wasundertaken using the following protocol: 94 "C, 45 s dena-turation; 48 "C, 1 min annealing; 72 "C, 1 min extensionfor 34 cycles; with a final cycle of 72 "C, 6 min extension.The polymerase enzyme used wasAmplitaqGold (ROCHE)that required a 90 "C, 9 min incubation period for the firstcycle only. The PCR mixture was a 25 ll reaction including:1· bu!er (ROCHE), 1 UofAmplitaqGold polymerase. Fourmillimeter of MgCl2, 0.8 mM of dNTPs, 5 pmol of each pri-mer, and unknown concentrations of template DNA.

The following primer pairs were used to amplify the var-ious gene fragments. The COI gene fragment was amplifiedusing two primer pair sets: LCO1490: 50-GGT CAA CAAATC ATA AAG ATA TTG G-30 with HCO2198 50-TAAACT TCA GGG TGA CCA AAA AAT CA-30 (Folmeret al., 1994) and C1-J-2183 50-CAA CAT TTA TTTTGA TTT TTT GG-30 (Simon et al., 1994) with A273550-AAA AAT GTT GAG GGA AAA ATG TTA-30 (Cres-pi B). The EF-1a gene fragment was amplified using primerpair sets M51.9 50-CAR GAC GTA TAC AAA ATC GG-30 (Cho et al., 1995) with 50-AGA CTC AAC ACA CATAGG TTT GGA C-30 (Morris D) and G730 50-ACCTTC GCT CCT GCC AAC TT-30 with G731 50-AAGGGT GAT AAT AGC AGC-30 (McLeish M). The winglessgene fragment was amplified using primer pair 50-TAGACG TAT CGT TAC ACT GC-30 and 50-CGT CAAGAC CTG CTG GAT GC-30 (McLeish M). The 16S (ribo-somal RNA) gene fragment was amplified using CI-J-2195

716 M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725

50-CCG GTC TGA ACT CAG ATC ACG T-30 (Simonet al., 1994) with A2735 50-CGC CTG TTT AAC AAAAAC AT-30 (Crespi B). We amplified up to 1245 bp of theCOI mitochondrial gene, 444 bp of the EF-1a gene,472 bp of the 16S, and 549 of the wingless gene. SeqEdv1.0.3 (http://helix.nih.gov/docs/gcg/seqed.html) was usedto edit sequences. Sequences were aligned usingCustalX 1.81.1a software (Thompson et al., 1997; ftp://ftp-igbmc.u-strasbg.fr/pub/ClustalX/ accessed 24 June2005). All nucleic acid sequence data has been lodgedwith GenBank under the Accession Nos. AY827474–AY827481, AY920988–AY921000, AY921058–AY921069,and DQ246453–DQ246516.

Taxon sampling is known to e!ect the estimation of sub-stitution parameters. Our data set of 57 taxa was in excess ofthe suggested 20, assumed to be appropriate in accountingfor uncertainty caused by too small a sample (Sullivanet al., 1999). In five instances, the dataset includes replicatepopulations of the same ‘host race’ from samples taken ateither di!erent sites or di!erent years. These taxa includepopulations ofK. sterni,K. waterhousei (onA. papyrocarpa),K. intermedius, and K. rugosus (on Acacia cana and x2 racesonA. papyrocarpa). We included these populations to verifyexpectations of phylogenetic coherence within a taxon.

2.2. Phylogenetic analysis

Phylogenies were inferred to validate the independenceof the K. rugosus and K. waterhousei groups and extend

and revise the current gall-inducing thrips phylogeny. Thecurrent robust well resolved and well-supported gall-thripsphylogeny (Morris et al., 2001) comprises 21 described spe-cies and was generated using maximum parsimony (MP)and maximum likelihood approaches. Maximum parsi-mony is used to test the robustness of model-based trees.We infer MP phylogenies with the addition of 32 new taxausing the search parameters consistent with Morris et al.(2001). To accommodate di!erences in substitution rateparameters and base compositional bias in our multiplegene dataset we implemented a model-based maximumlikelihood Bayesian approach. A recent study (Lin andDanforth, 2004) advocates a Bayesian approach to accom-modate substitution and rate dynamics evident in the gall-thrips sequence dataset of Morris et al. (2001). We alsoconducted Bayesian analyses using a combined dataset(i.e. no partitioning of the sequence data) in addition toanalyses using separated data. In all cases, nodal supportfor poorly and well-supported relationships were invariablyreproduced by the combined Bayesian analyses.

Maximum parsimony and Bayesian inferences wereimplemented in PAUP*b4.10 (Swo!ord, 2002) and MrBa-yes (MrBayes 3.0b4, Huelsenbeck and Ronquist, 2001),respectively. Maximum parsimony analysis was imple-mented using a heuristic search, with TBR (tree bisec-tion-reconstruction), branch swapping on all best trees,with 100 random sequence additions holding 10 trees heldat each step. We used 500 heuristic search pseudoreplicatesto calculate bootstrap support values (Felsenstein, 1985),

Table 1Description and locations of collection sites for the Kladothrips rugosus and K. waterhousei species complexes used for the sequence data

Galler Host Site description Date

K. sterni A. aneura 20 km N of Quilpi QLD August 04K. habrus A. pendula 15 km S of Yenda NSW May 04K. intermedius A. oswaldii 15 km W of Mt. Hopeless SA March 04K. rugosus A. papyrocarpa 25 km W of Whyalla SA February 02K. rugosus A. ammophila Lk. Bindegolly QLD April 97K. rugosus A. ancistrophylla 20 km E of Dalwallinu WA April 97K. rugosus A. enervia 3 km W of Merridin WA January 99K. rugosus A. cana 25 km NW of Tibooburra NSW April 98K. rugosus A. loderi 83 km SW Broken Hill NSW April 05K. rugosus A. melvillei 9 km S of Yenda NSW May 04K. rugosus A. microcephala 10 km W of Prairie QLD April 98K. rugosus A. microsperma 40 km E of Quilpie QLD August 04K. rugosus A. omalophylla 19 km S of Miles QLD April 98K. rugosus A. papyrocarpa 25 km W of Whyalla SA February 02K. rugosus A. pendula 95 km S of Gri"th NSW April 04K. rugosus A. sibilans 9 km S of Wooramel WA April 97K. rugosus A. tephrina 8 km N of Barcaldine QLD February 04K. waterhousei A. maranoensis 57 km E of Morven QLD April 97K. waterhousei A. loderi 79 km W of Wilcannia NSW March 96K. waterhousei A. ancistrophylla 20 km E of Dalwallinu WA April 97K. waterhousei A. enervia 3 km W of Merridin WA January 99K. waterhousei A. cana 25 km NW of Tibooburra NSW April 98K. waterhousei A. microsperma 45 km N of Adavale QLD April 97K. waterhousei A. omalophylla 121 km E Quilpie QLD April 98K. waterhousei A. ammophila Lk. Bindegolly QLD April 97K. waterhousei A. microcephala 37 km N of Aramac QLD March 98

For collections details of other species see Morris et al. (2001). Host populations of K. habrus, K. intermedius, and K. waterhousei were also collected.

M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725 717

using the same search parameters as above for thepseudoreplicates. Heuristic search starting trees forbranch-swapping were generated using stepwise addition,swapping on the best trees only.

To accommodate di!erences in substitution rate param-eters and heterogeneity of base composition in our multiplegene fragment dataset we fitted separate models to di!erentgene partitions in the Bayesian analysis (Lin and Danforth,2004). The sequence data used in the MrBayes analysis wasdivided into six partitions comprising 1st, 2nd, and 3rdcodon positions of the cytochrome oxidase one (COI) mito-chondrial data, with single partitions for each of elongationfactor one alpha (EF-1a), wingless, and the 16S gene frag-ments. We used a general time reversible (GTR) DNA sub-stitution model with gamma distributed rates with aproportion of invariant sites. Posterior probabilities andmean branch lengths are derived from 3000 trees takenfrom generations 3.5 to 5.0 million, sampling every 500thgeneration. The sampled trees were derived from post bur-nin generations after the chains had reached apparent sta-tionarity. We ran the Bayesian analysis 3 times to verify therepeatability of the phylogenetic outcome. We summarisedthe repeatability between independent Bayesian runs bypercent variation of the log likelihood arithmetic meansgenerated for all generations sampled and for the post bur-nin sample trees. Log likelihood values reached apparentstationarity rapidly in each of the three Bayesian analyses.The arithmetic mean of the log likelihood values for allgenerations sampled and for post burnin samples were cal-culated for each Bayesian analysis and the percent varia-tion between them determined. Percent variation betweenthe arithmetic means between successive Bayesian analyseswas only 0.11% .

The outgroup, Rhopalothripoides (Bagnall) and Dactyl-othrips (Bagnall), are the most closely related sister-generato Kladothrips Froggatt (Morris et al., 2002). The K. rugo-sus and Kladothrips waterhousei species complexes werechosen as ingroup taxa as these groups are expected tohave diversified recently and in parallel on the same setof host Acacia species.

2.3. Codivergence analysis

Conservative associations between two gall-inducingthrips groups inhabiting the same Acacia host species sug-gest that the groups have diversified in parallel (Crespiet al., 2004). A comprehensive host–plant phylogeny isnot available to compare with a gall-thrips phylogeny.Under a hypothesis of parallel diversification, both groupsare expected to respond to host speciation in tandem. Toaddress the hypothesis of parallel divergence, congruencebetween two gall-thrips phylogenies inhabiting the samehost species was tested. We inferred separate phylogeniesof the K. rugosus, K. acaciae, and K. ellobus group andthe K. waterhousei, K. habrus, and K. hamiltoni group usinga Bayesian approach sampling every 500th of 3 milliongenerations. Stationarity was reached almost instanta-

neously in our phylogenetic inferences and to reduce com-putational time, we sampled every 3 million generationswhen generating the trees used to test codivergence hypoth-eses. The data was analysed using a combined approach inaddition to analysing partitioned data using the same pri-ors as described above. Nodal support di!erences betweenthe separate and combined approaches were negligible, andhave therefore elected to show inferences generated usingthe partitioned analyses. Nearest ancestral sister-specieswere used as outgroups and pruned for the cospeciationanalyses. We assumed that both of the inferences were ‘truephylogenies.’ As codivergence analyses assume that thephylogenies used represent the true relationships amongtaxa, the trees are not collapsed and support values aregiven for all bifurcations to demonstrate regions ofuncertainty.

The extent to which host and parasite phylogenies arecongruent can be used to detect ‘coevolution’ or cospecia-tion. To test how closely the diversification of gall-thripshas been subject to host speciation we tested for concor-dance between the phylogenies of: (I) the K. acaciae,K. ellobus, and the K. rugosus complex; and (ii) theK. harpophyllae, K. hamiltoni, K. habrus, K. intermedius,and K. waterhousei complex. Three approaches using com-puter programs were used, each treating the data di!erentways. First, ParaFit (Legendre et al., 2002 http://www.bio.umontreal.ca/casgrain/en/labo/parafit.html) wasused to test a global null hypothesis that the associationof two trees has been independent. This approach permitsthe treatment of the phylogenies ‘symmetrically’ and isnot directed at reconstructing a putative history of theassociation. The associations between the phylogenies arerandomised and tested. Phylogenies of the K. rugosus andK. waterhousei races are transformed into matrices of prin-ciple coordinates and then combined with another matrixdescribing the associations between the phylogenies. Thesignificance of a global fit is tested without direct inclusionof either one of the phylogenies rather focusing on manip-ulating the associations. To test the global fit between theK. rugosus and K. waterhousei groups, we implementedParaFit using phylogenies of equal length branches; patris-tic distances generated in PAUP, and likelihood valuesfrom our Bayesian consensus phylogram. By using equalbranch lengths (of 1) we were able to test the fit of thetopologies only. Cospeciation predicts that codivergencesmust occur in a contemporaneous manner. The matricesapproach also allows phylogenies to be represented by like-lihood or patristic values that account for concordancesubject to branch length variation.

Finally, as the statistical power a!orded by matrix-dri-ven approaches, such as ParaFit, was considered less thanoptimal as a result of information loss, concordancebetween K. rugosus and K. waterhousei phylogenies wasalso assessed using randomisation tests implemented inthe event-driven TreeMap 1.0 (http://www.evolve.zoo.ox.ac.uk/rod/treemap.html; Page, 1994) and TreeFitter (http://www.ebc.uu.se/systzoo/research/treefitter/treefitter.html;

718 M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725

Ronquist, 1997) approaches. Tree fitting and tree mappingmethods are ‘asymmetrical’ in their treatment of the twotrees to be tested for congruence. TreeMap (Page, 1994)and TreeFitter (Ronquist, 1997) are event-based compari-sons that enable the maximisation of codivergence events(equal to 1) explaining the association between the twophylogenies by down-weighting duplication, sorting, andswitching events (equal to 0 in each case). These eventsare of course invalidated when comparing parasite phylog-enies that presumably do not have historic host–parasiteassociations. The di!erence between these two approachesis that TreeMap requires event costs to be treated in a moreinflexible manner where the codivergence cost is strictly lessthan that of a duplication. TreeFitter allows a range of costevents to be explored where both duplication and codiver-gence can be set to zero. The ability to consider a greaterrange of event cost combinations in a single analysis canbe important for understanding the optimal set of eventsmore likely to represent the mechanisms operating in anygiven phylogenetic association. Here, we find bothapproaches less than ideal for our ‘parasite–parasite’ asso-ciation. However, we considered this type of approach auseful alternative to exploring the data and used each ofthe gall-thrips phylogenies as a pseudo-host tree in separateanalyses.

To easily visualise the host associations among theK. rugosus and K. waterhousei species complex a tangle-gram was generated by inferring Bayesian majority ruleconsensus phylogenies for each group and connecting taxaassociated with the same host Acacia species. These treeswere used to generate a set of possible codivergence eventsas inferred by TreeMap 1.0. TreeMap provides a graphicutility that generates a tanglegram displaying the associa-tions and putative codivergence events between two phy-logenies. The significance of the association betweenphylogenies is determined between observed and random-ised trees generated from either of the phylogenies beingcompared. An approach that randomises a host phylogenyis intuitive for host-parasite associations but not optimalwhen neither tree is a ‘host’ as such. Here, we assume thateither of the phylogenies is likely to closely match that ofthe true host phylogeny under cospeciation criteria, whereeither of the thrips phylogenies can be used to simulate thehost phylogeny.

3. Results

We extended and revised the phylogeny of Morris et al.(2001) with the inclusion of 16 putative races (on di!erentAcacia species) of the K. rugosus complex and 10 putativeraces of the K. waterhousei complex that specialise on thesame 14 host Acacia species. Maximum parsimony andBayesian inferences are in general agreement and show ahigh level of support for each of the clades containing theK. rugosus and K. waterhousei complexes (Figs. 1 and 2).Phylogenetic concordance tests between the K. rugosusand K. waterhousei species complexes showed a significant

level of non-independence. The ParaFit global test of treetopology only, using equal branch lengths, was significant(P = 0.01). TreeMap estimates of the maximum numberof observed codivergence events was significantly more(0.0025 < P(t = 2.93) < 0.001) than 10,000 random treeassociations (t 6 t0.05(1),1, reject Ho of no di!erence) andTreeFitter results indicated a maximum frequency of sevencodivergence events between the observed trees was alwayssignificantly more than those of the randomised trees(where host tree is K. waterhousei group; P = 0.019 andwhere host tree is K. rugosus group; P = 0.016).

3.1. Phylogenetic analysis

The addition of 32 new taxa in our phylogenetic infer-ences yielded results consistent with the general structuring

Fig. 1. A Maximum parsimony phylogeny using COI, EF-1a, wingless,and 16S genes implementing a heuristic search with tree bisection-reconstruction (TBR) branchswapping, random addition of taxa 100replicates per search and 10 trees held at each step, and 500 bootstrapreplicates. K. rugosus and K. waterhousei species complexes are highlightedwith grey boxes. Host tree species are abbreviated in brackets as follows:amm, A. ammophila; anc, A. ancistrophylla; ane, A. aneura; can, A. cana;ene, A. enervia; lod, A. loderi; mar, A. maranoensis; mel, A. melvillei; mcp,A. microcephala; msp, A. microsperma; oma, A. omalophyla; ori, A. orites;pen, A. pendula; pap, A. papyrocarpa; sib, A. sibilans; and tep, A. tephrina.Taxon codes are as follows: QLD, Queensland and WA, WesternAustralia populations; a, population A; b, population B; E, elongate;P, pouched; and S, spiky gall structures.

M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725 719

of the most recent published phylogeny (Morris et al.,2001). Both maximum parsimony and Bayesian inferencesare in general agreement and indicate that the K. rugosusand K. waterhousei clades form well-supported and wellresolved groups of 100% for each of these nodes in theBayesian inference and 99% in the MP inference (Figs. 1and 2). A majority of the putative K. rugosus and K. water-housei host races appear to be at various stages of di!eren-tiation (Fig. 3). In particular, taxa exhibiting apparent lessthan species-level genetic distances with similar gall struc-tures group as clades (unpublished work). Uncorrected‘‘p’’ distances for COI between K. rugosus population bear-ing disparate gall structures were in the order of 6–10%contrasting those among like types with distances notexceeding 0.8%. Uncorrected ‘‘p’’ distances between theoutgroup and Kladothrips species ranged from 7 to 12%.

Both MP and Bayesian inferences grouped ‘pouched,’‘elongate,’ and ‘spiky’ gall structures into clades of liketypes. Replicate taxon samples of populations with thesame host and gall structure collected from di!erent sitesor seasons grouped as sister-taxa. These grouping con-firmed the expectation that such populations, particularlyat relatively recent stages of di!erentiation, maintainedphylogenetic coherence. For example, although currentlyconsidered host races, populations of K. waterhousei inhab-iting A. cana and A. papyrocarpa sampled across di!erentyears and multiple sites, group together.

The K. rugosus complex was paraphyletic with respect toK. maslini and the K. waterhousei complex was paraphyleticwith regard to the putative host races of K. habrus andK. intermedius. The maximum parsimony inference placesKladothrips rodwayi into a paraphyletic relationship withthe K. waterhousei complex. We suspect low bootstrap

Fig. 3. Consensus phylogram using six separately modelled partitionscomprising 1st, 2nd, and 3rd COI codons and separate EF-1a,wingless, and16S sites. TheK. rugosus andK. waterhousei species complexes are indicatedby the grey boxes. Host tree species are abbreviated in brackets as follows:amm, A. ammophila; anc, A. ancistrophylla; ane, A. aneura; can, A. cana;ene, A. enervia; lod, A. loderi; mar, A. maranoensis; mel, A. melvillei; mcp,A. microcephala; msp, A. microsperma; oma, A. omalophyla; ori, A. orites;pen, A. pendula; pap, A. papyrocarpa; sib, A. sibilans; and tep, A. tephrina.Taxon codes are as follows: QLD, Queensland andWA,Western Australiapopulations; a, populationA; b, population B; E, elongate; P, pouched; andS, spiky gall structures.

Fig. 2. Bayesian consensus tree analysis of six separately modelledpartitions comprising 1st, 2nd, and 3rd COI codons and separate EF-1a, wingless, and 16S sites. Posterior probabilities and mean branchlengths are derived from 3000 trees taken from a sample of 5 milliongenerations, sampling every 500th generation. The K. rugosus andK. waterhousei species complexes are indicated by the grey boxes. Hosttree species are abbreviated in brackets as follows: amm, A. ammophila;anc, A. ancistrophylla; ane, A. aneura; can, A. cana; ene, A. enervia; lod,A. loderi; mar, A. maranoensis; mel, A. melvillei; mcp, A. microcephala;msp, A. microsperma; oma, A. omalophyla; ori, A. orites; pen, A. pendula;pap, A. papyrocarpa; sib, A. sibilans; and tep, A. tephrina. Taxon codes areas follows: QLD, Queensland and WA, Western Australia populations; a,population A; b, population B; E, elongate; P, pouched; and S, spiky gallstructures.

720 M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725

support and/or long-branch attraction, due to incompletesequence data for the K. waterhousei complex at EF-1aand wingless gene regions (Wiens, 2005), might have con-tributed to this outcome. The clade comprising K. acaciae,K. ellobus, and the K. rugosus complex, matches closelywith host a"liation by the sister-clade comprisingK. harpophyllae, K. hamiltoni, and the K. waterhousei com-plex. Sister-species from each of these clades specialise onthe same host species, Acacia cambadgei (Baker) andAcacia harpophylla (Muell, Benth) with the K. rugosusand K. waterhousei complexes sharing the same set of hostspecies. Monophyly of the K. rugosus complex becomesinvalid by the presence the K. maslini lineage, though a hostshift is strongly suspected along this lineage (Crespi et al.,2004).

3.2. Codivergence analysis

ParaFit tests a global null hypothesis that the associa-tion between the phylogenies has been independent(P = 0.01). Tests on individual associations indicate thatthis inferred cospeciation was partial rather than complete.Portions of the two trees are apparently independent as 3of the 12 links were significant in their contribution tothe global test statistic. Global tests using patristic dis-tances and likelihood values were non-significant at the lev-els of P = 0.08 and P = 0.18, respectively, suggesting thatnon-contemporaneous associations invalidated the signifi-cant codivergence events or that molecular-evolutionaryrates di!er between lineages in the two clades.

Under the assumption that one of the thrips phylogenieswas equal to the true host phylogeny, both thrips phyloge-nies were used to simulate the host phylogeny in separate

analyses in TreeMap 1.0 and TreeFitter. An exact searchalgorithm implemented in TreeMap 1.0 produced 24 solu-tions explaining the historical relationship between the phy-logenies. All required seven codivergence events, when theK. rugosus group was assumed to represent the true hostphylogeny. One such solution is summarised in Fig. 4 andit also shows gall-thrips taxa that share the same host spe-cies. An exact search generated 19 solutions all incorporat-ing seven codivergence events when the K. waterhouseigroup was assumed to represent the true host phylogeny.Both trees were randomised using a Markovian model togenerate a distribution (n = 10,000) of codivergence events.The results indicated a significant (t 6 t0.05(1),1, rejectHo ofno di!erence) association between the trees. This outcomeindicates that cospeciation apparently occurred more oftenthan by chance. A randomisation approach implemented inTreeFitter was used to test the significance of codivergencesin the observed tree associations in two separate analyses,where alternate phylogenies were assumed to match the truehost phylogeny. Tests of congruence against a randomisedset of trees indicated a significant association (where hosttree is K. waterhousei group; P = 0.019 and where host treeis K. rugosus group; P = 0.016).

4. Discussion

We inferred phylogenies including numerous unde-scribed putative species of gall-inducing thrips to testhypotheses of diversification. The addition of these taxato a previous gall-thrips phylogeny (Morris et al., 2001),results in a tree that includes virtually all known taxonomicentities for this group. A clear pattern has emerged fromthis expanded tree, suggesting that diversity for this group

Fig. 4. Bayesian Consensus phylogenies of the K. rugosus and K. waterhousei groups plus associated species. Posterior probabilities are indicated onbranches. Host species are indicated under thrips taxon names. Lines connected to two phylogenies show host associations. Circles on nodes showcodivergence events as inferred by TreeMap 1.0.

M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725 721

of specialist insects is generated in close association withhost speciation. Maximum parsimony and Bayesian infer-ences group the K. rugosus and K. waterhousei species com-plexes as sister-taxa with a high level of support. Significantcongruence between the phylogenies of the K. rugosus andK. waterhousei groups, including sister-species within each(Fig. 4), indicated both cospeciation and convergence ofthrips across host–plant species of Acacia in the sectionPlurinerves. Codivergence between these thrips groups indi-cates they were subject to the same isolating eventsimposed by host speciation. Poor resolution amongrecently diverged taxa within the two complexes renderscospeciation inferences ambiguous for some associations.Host switching and the formation of host races among clo-sely related species of this Acacia section does not appearto be accompanied by the relatively large shifts in adaptivechange when switching between host sections (Crespi et al.,2004) as implied for K. maslini and K. rodwayi. Highly sim-ilar morphologies and marginal adaptive shifts amongthrips belonging to the K. rugosus complex belie substantialgenetic di!erentiation evident in the COI gene and thestructures of the galls induced by each. This work suggeststhat the potential of specialist phytophagous insects todiversify phenotypically increases with their ability colonisemore distantly related hosts.

4.1. Phylogenetic analyses

Phylogenetic inferences reveal patterns consistent withsome taxa belonging to the K. rugosus and K. waterhouseicomplexes being genetically di!erentiated below the levelevident among the described gall-thrips taxa (Morriset al., 2001; Crespi et al., 2004; McLeish et al., 2006). Theserecently diverged groups o!er an opportunity to test codi-vergence hypotheses where evolutionary processes such asextinction are unlikely to obscure interpretation, as mightbe expected amongst relatively older lineages (Brownet al., 1995).

The morphological di!erence between the K. intermediussamples with ‘pouched’ and ‘elongate’ gall types, was con-sidered negligible (personal communication, Mound LA)though genetic di!erentiation of 5% (uncorrected ‘‘p’’ dis-tance for COI) suggests considerable di!erentiation. TheK. sterni populations from Western Australia (WA) andQueensland (QLD) grouped as sister-taxa with a high levelof support (Figs. 1and 2) although branch length estimates(Fig. 4) indicate genetic divergences consistent with allo-patric factors. The grouping of K. habrus with putativeK. intermedius populations requires further attention toestablish taxonomic boundaries and nomenclature. Indeed,lack of phylogenetic signal evident for taxa within thesegroups compounds tests of phylogenetic concordance dueto considerable levels of uncertainty (Fig. 4). However, thisambiguity does not invalidate a signal that was strongenough to produce a significant level of non-independencein tests of congruence between these parallel clades a"li-ated with the same host species.

4.2. Diversification

It has been hypothesized that cospeciation and hostswitching processes are responsible for the genetic, pheno-typic, and ecological di!erentiation in gall-thrips (Crespiet al., 2004). Strong agreement among phylogenetic infer-ences indicates that the K. rugosus and K. waterhousei com-plexes are both paraphyletic and inhabit Plurinerves hostspecies (Figs. 1and 2). These two thrips clades might havediversified in parallel via each shifting to related, nearbyhosts without cospeciation per se, but broad patterns ofgall-thrips lineages tracking host diversification and evi-dence of parallel diversification between thrips and hostspecies at lower scales appears to be partially an outcomeof cospeciation.

The best available Acacia host phylogeny (Crespi et al.,2004) provides some support for the relationships indicatedby the K. rugosus and K. waterhousei groups and associatedsister-species (Fig. 4). Acacia harpophyllae and A. cambageiare sister-taxa and are hosts to basal sister-species of boththrips clades. Although the phylogenetic relationshipsamong the other Plurinerves species are weakly supported,this group forms a sister clade to A. harpophyllae andA. cambagei. Switching between host plants more distantlyrelated than hosts of gall-thrips sister-taxa, is accompaniedby noticeable life history shifts, is very likely to occurbetween hosts that are taxonomically and phylogeneticallyclose, and have overlapping or adjacent ranges (Maslin,2001; Crespi et al., 2004). Several examples stand out.Within the clade also comprising the K. waterhousei group,losses in sociality (Morris et al., 2001) accompanied by anapparent switch to more distantly related host have beeninferred (Crespi et al., 2004). Kladothrips xiphius is believedto represent a loss in sociality and is found on a speciesbelonging to the Juliflorae section, not Plurinerves as do sis-ter-species. Similarly, K. rodwayi is recognised as a loss insociality and it too is found on a more distantly relatedhost, A. melanoxylon (Fig. 4), a species distributed in tem-perate and not arid climates. Kladothrips intermediusinhabits a host species that appears not to be as closelyrelated to the hosts of thrips sister-species. Like K. rodwayi,K. intermedius ecloses within the gall, contrasting othermembers of this thrips clade that instead disperse as pupa.Unlike its K. rugosus sister-members on a Plurinerves hostand disperse as pupa, K. maslini inhabits a Juliflorae host(Figs. 1 and 2) and ecloses as an adult within the gall.Uncorrected ‘‘p’’ distance (unpublished work) variationshown by these additional taxa implies both similar andbelow those between described gall-thrips species.

The frequency and causes of those host switches travers-ing host sections accompanying relatively large life historychanges is unknown. A period of host radiation may con-tribute to gall-thrips ability to switch among host–plantsunder speciation (Craig et al., 1994). It appears that theopportunity for gall-thrips to diversify might have beenassisted by speciation in Plurinerves hosts presumably dur-ing a rapid radiation as widespread aridity developed in the

722 M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725

early Quaternary (Maslin and Hopper, 1982; Clapperton,1990; Lovejoy and Hannah, 2004). It is expected thatgall-inducing insects should radiate into arid habitats aspressures exerted by parasitoids, predators, and pathogensare not as acute in xeric environments (Fernandes et al.,1994; Blanche and Westoby, 1995; Price et al., 1998; Veldt-man and McGeoch, 2003). Radiations into novel niches torelieve pressures exerted by natural enemies is consistentwith Ehrlich and Raven’s (1964) ‘escape and radiate’hypothesis predicting the generation of diversity in phy-tophagous insects as a consequence of strong selection aftercolonising a novel host.

4.3. Bimodality in divergence

The tanglegram in Fig. 4 shows phylogenies of both theK. rugosus and K. waterhousei complexes connected bylines that join terminal taxa that share the same host spe-cies. One of many possible inferred solutions explainingcodivergence events between the trees is indicated at partic-ular bifurcations. Tests for concordance indicate that thesephylogenies are significantly non-independent, but a nota-ble degree of incongruence is also evident. However, phylo-genetic uncertainty in the trees might contribute to acomponent of ambiguity in the interpretation of codiver-gence events possibly caused by inclusion of taxa in thephylogenies that are marginally divergent. Synchronousgenetic isolation between host and parasite can occur atthe level of individual, population, species, or higher(Rannala and Michalakis, 2003). Codivergence indicatesa frequent incidence of isolating events a!ecting both thripsclades in tandem and that cospeciation between gall-thripsand Acacia operates in a broad sense between lineages andalso between species. A lack of identical fit between thephylogenies shows independent isolating events has possi-bly acted on either clade at some stage. Non-significantcontemporaneous branching episodes between the K. rugo-sus and K. waterhousei groups shown by the ParaFit out-comes for patristic and likelihood branch length valuessuggest bimodality in divergence processes. That is, eachthrips clade appears to have responded somewhat di!eren-tially to isolating events of the host. Given cospeciation,one might expect that the more closely related host specieswould reflect proportionally similar divergences paralleledin both thrips lineages that inhabit these hosts, but this pat-tern does not appear to be consistent with our data.

Bimodality in divergences might reflect di!erential host-utilisation, dispersal ability, or community driven di!er-ences such as escaping natural enemies and interspecificcompetition (Jaenike, 1990). For example, demographicand life history di!erences between the K. rugosus andK. waterhousei complexes, such as social organization(Crespi, 1992) and comparative brood sizes (Wills et al.,2001) could be invoked to develop hypotheses explainingthis bimodality. There is general agreement betweengall-thrips and best available host phylogenies (Crespiet al., 2004) but interpretations are also conditional on

the presence of a degree of phylogenetic uncertainty. Theability to test concordance between gall-thrips lineageshas provided insight into mechanisms driving diversifica-tion in this group where comparisons of insect and plantphylogenies were not feasible. This work shows that diver-sification in phytophagous insects can proceed via a combi-nation of synchronous divergence episodes between insectand host lineages in addition to independent modes of spe-ciation among them. Speciation by host switching and hostrace formation, concurrent with cospeciation, can play arole in generating diversity. The potential for a group ofspecialist phytophages to diversify appears to be closelylinked to their ability to traverse genetic, phenological,chemical, and morphological obstacles among plant speciesvarying in susceptibility to colonisation. These patternsimply that the ability to overcome di"cult barriers to col-onising novel host plants might determine the degree ofdiversity attained in phytophagous insects. Switching tomore distantly related hosts should be accompanied greaterpotentials to diversify.

Acknowledgments

We thank the Evolutionary Biology Unit, South Austra-lia Museum, for their sequencing facilities and technicalsupport. This project was made possible by part fundingfrom the Nature Foundation SA Inc. (Project #7324), theSir Mark Mitchell Research Foundation (CXS10423800),the Australian Research Council (ARC) to SchwarzM.P., Cooper J.B., Crespi B.J., Chapman T.W.,(DP0346322), an ARC Postdoctoral Fellowship to Chap-man T.W., and an NSERC grant to Crespi.

References

Baker, S.C., 1996. Lice, cospeciation and parasitism. Int. J. Parasit. 26,219–222.

Ballabeni, P., Gotthard, K., Kayumba, A., Rahier, M., 2003. Localadaptation and ecological genetics of host–plant specialisation in a leafbeetle. Oikos 101, 70–78.

Blanche, K.R., Westoby, M., 1995. Gall-forming insect diversity is linkedto soil fertility via host plant taxon. Ecology 76, 2334–2339.

Brown, J.M., Abrahamson, W.G., Packer, R.A., Way, P.A., 1995. Therole of natural enemy escape in a gallmaker host–plant shift. Oecologia104, 52–60.

Burckhardt, D., Basset, Y., 2000. The jumping plant-lice (Hemiptera:Psylloidae) associated with Schinus (Anacardiiaceae): systematics,biogeography, and host plant relationships. J. Nat. Hist. 34, 57–155.

Chapman, T.W., Crespi, B.J., Kranz, B.D., Schwarz, M.P., 2000. Highrelatedness and inbreeding at the origin of eusociality in gall-inducingthrips. Proc. Natl. Acad. Sci. USA 97, 1648–1650.

Cho, S., Mitchell, A., Regier, J.C., Mitter, C., Poole, R.W., Friedlander,T.P., Zhao, S., 1995. A highly conserved nuclear gene for low-levelphylogenetics: Elongation Factor-1 recovers morphology-based tree forheliothine moths. Mol. Biol. Evol. 12, 650–656.

Clapperton, C.M., 1990. Quaternary glaciations in the Southern Hemi-sphere: an overview. Quat. Sci. Rev. 9, 299–304.

Clark, M.A., Moran, N.A., Baumann, P., Wernegreen, J.J., 2000.Cospeciation between endosymbiomts (Buchnera) and a recent radi-ation of aphids (Uroleucon) and pitfalls of testing for phylogeneticcongruence. Evolution 54, 517–525.

M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725 723

Clayton, D.H., Lee, P.L.M., Tompkins, D.M., Brodie, D.E., 1999.Reciprocal natural selection on host–parasite phenotypes. Am. Nat.154, 261–270.

Cook, J.M., Rokas, A., Pagel, M., Stone, G.N., 2002. Evolutionary shiftsbetween host oak sections and host–plant organs in Andricusgallwasps. Evolution 56, 1821–1830.

Cornell, H.V., 1983. The secondary chemistry and complex morphology ofgalls formed by Cynipinae (Hymenoptera): why and how? Am. Midl.Nat. 110, 220–234.

Craig, T.P., Itami, J.K., Horner, J.D., Abrahamson, W.G. 1994. Hostshifts and speciation in gall-forming insects. In: Price, P.W., Mattson,W.J., Baranchikov, Y.N. (Eds.), The Ecology and Evolution of Gall-Forming Insects. United States Department of Agriculture, pp. 194–207.

Craig, T.P., Horner, J.D., Itami, J.K., 2001. Genetics, experience, andhost–plant preference in Eurosta solidaginis: implications for hostshifts and speciation. Evolution 55, 773–782.

Crespi, B.J., 1992. Eusociality in Australian gall thrips. Nature 359, 724–726.

Crespi, B.J., Carmean, D.A., Chapman, T.W., 1997. Ecology andevolution of gall thrips and their allies. Annu. Rev. Entomol. 42, 51–71.

Crespi, B.J., Carmean, D.A., Mound, L.A., Worobey, M., Morris, D.C.,1998. Phylogenetics of social behaviour in Australian gall-formingthrips: evidence from mitochondrial DNA sequence, adult morphologyand behaviour, and gall morphology. Mol. Phylogenet. Evol. 9, 163–180.

Crespi, B.J., Morris, D.C., Mound, L.A. 2004. Evolution of Ecologicaland Behavioural Diversity: Australian Acacia Thrips as ModelOrganisms. Australian Biological Resources Study and AustralianNational Insect Collection, CSIRO, Canberra, Australia.

Crespi, B.J., Worobey, M., 1998. Comparative analysis of gall morphol-ogy in Australian gall thrips: the evolution of extended phenotypes.Evolution 52, 1686–1696.

Cronin, J.T., Abrahamson, W.G., 2001. Do parasitoids diversify inresponse to host–plant shifts by herbivorous insects? Ecol. Entomol.26, 347–355.

Despres, L., Jaeger, N., 1999. Evolution of oviposition strategies andspeciation in the globe flower flies Chiastocheta spp. (Anthomyiidae). J.Evol. Biol. 12, 822–831.

Dobler, S., Farrell, B.D., 1999. Host use evolution in Chrysochusmilkweed beetles: evidence from behaviour, population genetics, andphylogeny. Mol. Ecol. 8, 1297–1307.

Dres, M., Mallet, J., 2002. Host races in plant-feeding insects and theirimportance in sympatric speciation. Philos. Trans. R. Soc. Lond., Ser.B: Biol. Sci. 357, 471–492.

Ehrlich, P.R., Raven, P.H., 1964. Butterflies and plants: a study incoevolution. Evolution 18, 586–608.

Emelianov, I., Mallet, J., Baltensweiler, W., 1995. Genetic di!erentiationin Zeiraphera diniana (Lepidoptera: Tortricidae, the larch budmoth):polymorphism, host races or sibling species. Heredity 75, 416–424.

Farrell, B.D., Mitter, C., 1998a. Phylogeny of host a"liation: havePhyllobrotica (Coleoptera: Chrysomelidae) and the Lamiales diversi-fied in parallel? Evolution 44, 1389–1403.

Farrell, B.D., Mitter, C., 1998b. The timing of insect/plant diversification:might Tetraopes (Coleoptera: Cerambycidae) and Asclepias (Asclepi-adaceae) have coevolved? Biol. J. Linn. Soc. 63, 553–577.

Felsenstein, J., 1985. Confidence limits on phylogenies: an approach tousing the bootstrap. Evolution 39, 783–791.

Ferguson, J.W.H., 2002. On the use of genetic divergence for identifyingspecies. Biol. J. Linn. Soc. 75, 509–516.

Fernandes, G.W., Lara, A.C.F., Price, P.W., 1994. The geography ofgalling insects and the mechanisms that result in patterns. In: Mattson,W.J., Baranchikov, Y., Price, P.W. (Eds.), The Ecology and Evolutionof Gall-Forming Insects. United States Department of Agriculture,Minnesota, pp. 42–48.

Folmer, O., Black, M., Hoeh, W., Lutz, R., Vrijenhoek, R., 1994. DNAprimers for amplification of mitochondrial cytochrome c oxidase

subunit I from diverse metazoan invertebrates. Mol. Mar. Biol.Biotechnol. 3, 294–299.

Futuyma, D.J., Moreno, G., 1988. The evolution of ecological speciali-sation. Annu. Rev. Ecol. Syst. 19, 207–233.

Hawthorne, D.J., Via, S., 2001. Genetic linkage of ecological specializa-tion and reproductive isolation in pea aphids. Nature 412, 904–907.

Herre, E.A., Machado, C.A., Bermingham, E., Nason, J.D., Windsor,D.M., McCa!erty, S.S., Bachmann, K., 1996. Molecular phylogeniesof figs and their pollinator wasps. J. Biogeogr. 23, 521–530.

Huelsenbeck, J.P., Ronquist, F., 2001. MRBAYES: Bayesian inference ofphylogeny. Bioinformatics 17, 754–755.

Humphries, C.J., Cox, J.M., Nielsen, E.S., 1986. Nothophagus and itsparasites: a cladistic analysis approach to coevolution. In: Stone, A.R.,Hawksworth, D.L. (Eds.), Coevolution and Systematics. ClarendonPress, Oxford, pp. 55–76.

Itino, T., Davies, S.J., Tada, H., Heida, O., Inoguchi, M., Itioka, T.,Yamane, S., Inoue, T., 2001. Cospeciation of ant and plants. Ecol.Res. 16, 787–793.

Jaenike, J., 1981. Criteria for ascertaining the existence of host races. Am.Nat. 117, 830–834.

Jaenike, J., 1990. Host specialization in phytophagous insects. Annu. Rev.Ecol. Syst. 21, 243–273.

Janz, N., Nyblom, K., Nylin, S., 2001. Evolutionary dynamics of hostplant specialization: a case study of the tribe Nymphalini. Evolution55, 783–796.

Janz, N., Nylin, S., 1998. Butterflies and plants: a phylogenetic study.Evolution 52, 486–502.

Jermy, T., 1993. Evolution of insect–plant relationships—a devil’sadvocate approach. Entomol. Exp. Appl. 66, 3–12.

Jermy, T., 1984. Amt. Nat.. Evolution of insect/host plant relationships124, 609–630.

Johnson, K.P., Clayton, D.H., 2003. Coevolutionary history of ecologicalreplicates: comparing phylogenies of wing and body lice to Columb-iform hosts. In: Page, R.D.M. (Ed.), Tangled Trees. University ofChicago Press, Chicago, Illinois, pp. 262–286.

Johnson, K.P., Williams, B.L., Drown, D.M., Adams, R.J., Clayton, D.H.,2002. The population genetics of host specificity: genetic di!erentiationin dove lice (Insecta: Phthiraptera). Mol. Ecol. 11, 25–38.

Jones, R.W., 2001. Evolution of the host plant associations of theAnthonomous grandis species group (Coleoptera: Curculionidae):Phylogenetic tests of various hypotheses. Ann. Entomol. Soc. Am.94, 51–58.

Joshi, A., Thompson, J.N., 1997. Adaptation and specialization in a two-resource environment in Drosophila species. Evolution 51, 846–855.

Jousselin, E., van Noort, S., Rasplus, J.-Y., Gree!, J.M., 2006. Patterns ofdiversification of Afrotropical Otiteselline fig wasps: phylogeneticstudy reveals a double radiation across host figs and conservatism ofhost association. J. Evol. Biol. 19, 253–266.

Kelley, S.T., Farrell, B.D., Mitton, J.B., 2000. E!ects of specialisation ongenetic di!erentiation in sister species of bark beetles. Heredity 84,218–227.

Legendre, P., Desdevises, Y., Bazin, E., 2002. A statistical test for host–parasite coevolution. Syst. Biol. 51, 217–234.

Lin, C.P., Danforth, B.N., 2004. How do insect nuclear and mitochondrialgene substitution patterns di!er? Insights from Bayesian analyses ofcombined data sets. Mol. Phylogen. Evol. 30, 686–702.

Lopez-Vaamonde, C., Rasplus, J.Y., Weiblen, G.D., Cook, J.M., 2001.Molecular phylogenies of fig wasps: partial cocladogenesis of pollin-ators and parasites. Mol. Phylogen. Evol. 21, 55–71.

Lovejoy, T.E., Hannah, L. (Eds.), 2004. Climate Change and Biodiversity.Yale University Press.

Machado, C.A., Herre, E.A., McCa!erty, S., Bermingham, E., 1996.Molecular phylogenies and fig pollinating and non-pollinating waspsand the implications for the origin and evolution of the fig–fig waspmutualism. J. Biogeogr. 23, 531–542.

Machado, C.A., Robbins, N., Gilbert, M.T.P., Herre, E.A., 2005. reviewof host specifity and its coevolutionary implications in the fig/fig-waspmutualism. Proc. Nat. Acad. Sci. USA 102, 6558–6565.

724 M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725

Maslin, B.R. 2001, CD-ROM - Wattle: Acacias of Australia. CSIROPublishing/Australian Biological Resources Study (ABRS) / Depart-ment of Conservation and Land Management (CALM) WesternAustralia.

Maslin, B.R., Hopper, S.D., 1982. Phytogeography of Acacia (Legumi-nosae: Mimosoideae) in central Australia. In: Barker, W.R., Greens-lade, P.J.M. (Eds.), Evolution of the Flora and Fauna of AridAustralia. Peacock Publications, Hong Kong, pp. 301–316.

McLeish, M.J., Chapman, T.W., Mound, L.A., 2006. Gall morpho-typecorresponds to separate species of gall-inducing thrips (Thysanoptera:Phlaeothripidae). Biol. J. Linn. Soc. 88, 555–563.

McLeish, M.J., Chapman, T.W., Schwarz, M.P., 2007. Host-drivendiversification of gall-inducing Acacia thrips and the aridification ofAustralia. BMC Biol. 5 (3). Available from: <http://www.biomedcen-tral.com/1741-7007/5/3>.

Mitter, C., Farrell, B.D., Wiegmann, B.M., 1988. The phylogenetic studyof adaptive zones: has phytophagy promoted insect diversification?Am. Nat. 132, 107–128.

Morris, D.C., Schwarz, M.P., Crespi, B.J., Cooper, J.B., 2001. Phylog-enetics of gall-inducing thrips on Australian Acacia. Biol. J. Linn. Soc.74, 73–86.

Morris, D.C., Schwarz, M.P., Cooper, S.J.B., Mound, L.A., 2002.Phylogenetics of Australian Acacia thrips: the evolution of behaviourand ecology. Mol. Phylogen. Evol. 25, 278–292.

Mound, L.A., 1971. Gall-forming thrips and allied species (Thysanoptera:Phlaeothripinae) from Acacia trees in Australia. Bulletin of the BritishMuseum (Natural History) Entomology 25, 387–466.

Mound, L.A., Crespi, B.J., Kranz, B., 1996. Gall-inducing Thysanop-tera (Phlaeothripidae) on Acacia phyllodes in Australia: host–plantrelations and keys to genera and species. Invertebr. Taxon. 10,1171–1198.

Nosil, P., Crespi, B.J., Sandoval, C.P., 2002. Host–plant adaptation drivesthe parallel evolution of reproductive isolation. Nature 417, 440–443.

Nyman, T., 2002. The willow bud galler Euura mucronata Hartig(Hymenoptera: Tenthredinidae): one polyphage or many monophages?Heredity 28, 288–295.

Nyman, T., Widmer, A., Roininen, H., 2000. Evolution of gall morphol-ogy and host–plant relationships in willow-feeding sawflies (Hyme-noptera: Tenthredinidae). Evolution 54, 526–533.

Page, R.D.M., 1994. Parallel phylogenies: reconstructing the history ofhost parasite assemblages. Cladistics 10, 155–173.

Page, R.D.M. (Ed.), 2003. Tangled Trees. University of Chicago Press,Chicargo, Illinois.

Parsons, Y.M., Shaw, K.L., 2001. Species boundaries and genetic diversityamong Hawaiian crickets of the genus Laupala indentified usingamplified fragment length polymorphisms. Mol. Ecol. 10, 1765–1772.

Price, P.W., Fernandes, G.W., Lara, A.C.F., Brawn, J., Barrios, H.,Wright, M.G., Ribeiro, S.P., Rothcli!, N., 1998. Global patterns inlocal number of insect galling species. J. Biogeogr. 25, 581–591.

Rannala, B., Michalakis, Y., 2003. Population genetics and cospeciation:from process to pattern. In: Page, R.D.M. (Ed.), Tangled Trees.University of Chicago Press, Chicago, Illinois, pp. 120–143.

Ricklefs, R.E., Schluter, D., 1993. Species Diversity in EcologicalCommunities: Historical and Geographical Perspectives. Universityof Chicago Press, Chicago, Illinois.

Roderick, G.K., 1997. Herbivorous insects and the Hawaiian silverswordalliance: coevolution or cospeciation? Pac. Sci. 51, 440–449.

Roderick, G.K., Metz, E.C., 1997. Biodiversity of Planthoppers (Hemip-tera: Delphacidae) on the Hawaiin silversword alliance: e!ects on hostplant phylogeny and hybridisation. Memoirs of the Museum ofVictoria 56, 393–399.

Rohfritsch, O., Shorthouse, J.D., 1982. Insect galls. In: Kahl, G., Schell,J.S. (Eds.), Molecular Biology of Plant Tumors. Academic, New York,pp. 131–152.

Ronquist, F., 1997. Phylogenetic approaches in coevolution and biogeog-raphy. Zool. Scr. 26, 313–322.

Ronquist, F., Liljeblad, J., 2001. Evolution of the gall wasp—host plantassociation. Evolution 55, 2503–2522.

Ronquist, F., Nylin, S., 1990. Process and pattern in the evolution. Ofspecies associations. Syst. Zool. 39, 323–344.

Rønsted, N., Weiblen, G.D., Cook, J.M., Salamin, N., Machado, C.A.,Savolainen, V., 2005. Sixty million years of co-divergence in the fig-wasp symbiosis. Proc. R. Soc. B. 272, 2593–2599.

Simon, C., Frati, F., Beckenbach, A., Crespi, B., Liu, H., Flook, P., 1994.Evolution, weighting, and phylogenetic utility of mitochondrial genesequences and a compilation of conserved polymerase chain reactionprimers. Ann. Entomol. Soc. Am. 87, 651–701.

Stern, D.L., 1995. Phylogenetic evidence that aphids, rather than plants,determine gall morphology. Proc. R. Soc. Lond., Ser. B: Biol. Sci. 260,85–89.

Stone, G.N., Schonrogge, K., 2003. The adaptive significance of insect gallmorphology. Trends Ecol. Evol. 18, 512–522.

Sullivan, J., Swo!ord, D.L., Naylor, G.J.P., 1999. The e!ects on taxonsampling on estimating rate heterogeniety parameters of maximum-likelihood models. Mol. Biol. Evol. 16, 1347–1356.

Swo!ord, D.L. 2002. PAUP*. Phylogenetic Analysis Using Parsimony(*and Other Methods). Version 4.0b10. Sinauer, Sunderland,Massachusetts.

Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins,D.G., 1997. The ClustalX windows interface: flexible strategies formultiple sequence alignment aided by quality analysis tools. NucleicAcids Res. 24, 4876–4882.

Veldtman, R., McGeoch, M.A., 2003. Gall-forming insect species richnessalong a non-scleromorphic vegetation rainfall gradient in SouthAfrica: the importance of plant community composition. Austral.Ecol. 28, 1–13.

Ward, L.K., Hackshaw, A., Clarke, R.T., 2003. Do food-plant preferencesof modern families of phytophagous insects and mites reflect pastevolution with plants? Biol. J. Linn. Soc. 78, 51–83.

Weiblen, G.D., 2004. Correlated evolution in fig pollination. Syst. Biol. 53(1), 128–139.

Weiblen, G.D., Bush, G.L., 2002. Speciation in fig pollinators andparasites. Mol. Ecol. 11, 1573–1578.

Weintraub, J.D., Lawton, J.H., Scoble, M.J., 1995. Lithinine moths onferns: a phylogenetic study of insect–plant interactions. Biol. J. Linn.Soc. 55, 239–250.

Wiens, J.J., 2005. Can incomplete taxa rescue phylogenetic analyses fromlong-branch attraction? Syst. Biol. 54, 731–742.

Wills, T.E., Chapman, T.W., Kranz, B.D., Schwarz, M.P., 2001. Theevolution of reproductive division of labour and gall size in Australiangall-thrips with soldiers. Naturwissenschaften 88, 526–529.

Zerega, N.J.C., Clement, W.L., Datwyler, S.L., Weiblen, G.D., 2005.Biogeography and divergence times in the mulberry family (Mora-ceae). Mol. Phylogenet. Evol. 37, 402–416.

M.J. McLeish et al. / Molecular Phylogenetics and Evolution 43 (2007) 714–725 725