Molecular Phylogenetics and Evolution 116 (2017) 172–181

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

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

The confounding effects of hybridization on phylogenetic estimation in the MARK New Zealand cicada genus Kikihia ⁎ Sarah E. Bankera,b, , Elizabeth J. Wadea,c, Chris Simona a University of Connecticut, Department of Ecology and Evolutionary Biology, 75 North Eagleville Road, Unit 3043, Storrs, CT 06269, USA b University of California, Department of Integrative Biology, 3101 Valley Life Sciences Building, Berkeley, CA 94720, USA c United States Department of Agriculture-Agricultural Research Service, Center for Medical, Agricultural, and Veterinary Entomology, 1600 SW 23rd Dr., Gainesville, FL 32608, USA

ARTICLE INFO ABSTRACT

Keywords: Phylogenetic studies of multiple independently inherited nuclear genes considered in combination with patterns Habitat heterogeneity of inheritance of organelle DNA have provided considerable insight into the history of species evolution. In particular, investigations of cicadas in the New Zealand genus Kikihia have identified interesting cases where Gene trees mitochondrial DNA (mtDNA) crosses species boundaries in some species pairs but not others. Previous phylo- Biogeography genetic studies focusing on mtDNA largely corroborated Kikihia species groups identified by song, morphology Species tree methods and ecology with the exception of a unique South Island mitochondrial haplotype clade—the Westlandica group. Concatenation fi fi ff Nuclear-mitochondrial discordance This newly identi ed group consists of diverse taxa previously classi ed as belonging to three di erent sub- generic clades. We sequenced five nuclear loci from multiple individuals from every species of Kikihia to assess the nuclear gene concordance for this newly-identified mtDNA lineage. Bayes Factor analysis of the constrained phylogeny suggests some support for the mtDNA-based hypotheses, despite the fact that neither concatenation nor multiple species tree methods resolve the Westlandica group as monophyletic. The nuclear analyses suggest a geographic distinction between clearly defined monophyletic North Island clades and unresolved South Island clades. We suggest that more extreme habitat modification on South Island during the Pliocene and Pleistocene resulted in secondary contact and hybridization between species pairs and a series of mitochondrial capture events followed by subsequent lineage evolution.

1. Introduction In the present study, we investigate gene trees from multiple nuclear loci to infer the validity of an unexpected clade of NZ cicada species Discordance between individual gene trees and the underlying reconstructed using mitochondrial DNA (Arensburger et al., 2004b; species tree is a well-documented obstacle to understanding evolu- Marshall et al., 2008, 2011). Kikihia is a large endemic genus of cicadas tionary history. Gene-tree discordance can be caused by a variety of that evolved in concert with a rapidly changing New Zealand (NZ) evolutionary processes, including gene duplication and extinction, landscape. Following the typical pattern of island biodiversity, the an- horizontal gene transfer, hybridization, and incomplete lineage sorting cestral species of Kikihia, is estimated to have diversified into three (Maddison, 1997; Avise et al., 1990; Maddison and Knowles, 2006; genera ∼10 Ma, shortly after colonization of NZ in the mid Miocene Carstens and Knowles, 2007; Sullivan et al., 2014; Warnow, 2015). To ∼14 Ma. (Arensburger et al., 2004a; Buckley et al., 2002; Marshall overcome this hurdle, recent studies have focused both on using a et al., 2016). Increased habitat heterogeneity caused by Pliocene era multi-gene approach and developing methods that explicitly account mountain building events triggered a species radiation ∼3–5Ma for gene-tree-species tree discord (Tonini et al., 2015 for list). However, (Arensburger et al., 2004b; Marshall et al., 2008; Williams et al., 2015; the results of empirical and theoretical studies assessing the perfor- Wood et al., 2017) and the genus experienced relatively constant di- mance of these methods versus traditional supermatrix or concatenated versification since that time (Marshall et al., 2008, 2011; Ellis et al., data analyses have varied depending on the data analyzed (Thompson 2015). Given the wealth of ecological, behavioral and mtDNA genetic et al., 2014; Tonini et al., 2015; Warnow, 2015; Edwards et al., 2016; data, the low dispersal rates of these species, and the dramatic geolo- Gatesy and Springer, 2014, 2016). gical history of New Zealand, the genus Kikihia is an excellent model

Abbreviations: New Zealand, NZ; mitochondrial DNA, mtDNA; Maximum Likelihood, ML; Bayesian Inference, BI; North Island, NI; South Island, SI ⁎ Corresponding author at: University of California, Department of Integrative Biology, 3101 Valley Life Sciences Building, Berkeley, CA 94720, USA. E-mail addresses: [email protected] (S.E. Banker), [email protected] (E.J. Wade), [email protected] (C. Simon). http://dx.doi.org/10.1016/j.ympev.2017.08.009 Received 24 March 2017; Received in revised form 4 August 2017; Accepted 17 August 2017 Available online 19 August 2017 1055-7903/ © 2017 Elsevier Inc. All rights reserved. S.E. Banker et al. Molecular Phylogenetics and Evolution 116 (2017) 172–181

Fig. 1. (A) “Intuitive tree” of Fleming and Dugdale based on ecology and morphology, redrawn from the Dugdale papers of the Archives of the NZ Collection, Landcare Research, NZ. Triangles right of branch tips represent current distribution with upwards facing triangles for North Island, downwards facing for South Island, and “O” for restricted distribution on an offshore island. Blue circles to the right of taxa names = individuals belonging to the mtDNA “Westlandica” clades in Fig. 1C*=K. “westlandica north” and K. “westlandica south” were not separated from K. muta muta by Fleming and Dugdale. (B) Map of sample collection. Taxon and colors correspond to the mtDNA groups shown in Fig. 1C, sample ID labels indicate taxon name, district code (Crosby et al., 1998), collection locality (Simon Lab code), and year collected (C) Partitioned Bayesian phylogeny (post-burnin consensus phylogram) of the genus Kikihia based on 2152 bp of mtDNA sequence redrawn from Marshall et al., 2008. Bayesian posterior probabilities (left) and maximum likelihood bootstraps (right).

173 S.E. Banker et al. Molecular Phylogenetics and Evolution 116 (2017) 172–181 system with which to examine the effects of landscape and climate et al., 2004b) have consistently placed the shade singer taxa, K. scu- change on speciation. The New Zealand entomologist John Dugdale and tellaris and K. cauta as sister to the main Kikihia species radiation, geologist/naturalist Charles Fleming described six of the 13 currently therefore all phylogenies were rooted with K. scutellaris. recognized Kikihia species. Their cladogram (Archives of the NZ Na- Cicadas were collected in the field by hand and with nets. tional Insect Collection, Landcare Research, Auckland) of Kikihia spe- Whenever possible, songs were recorded. Specimens were preserved in cies contains four major clades: shade singers (forest understory), Green 95% ethanol either as whole bodies or leg-only samples, kept re- Foliage (evergreen shrub and forest), Muta (grass and shrub) and Rosea frigerated in the field and stored in ethanol at − 20 to −80°C in the (dry scrub) (Fig. 1A). These groupings were created based on dis- Simon Lab at the University of Connecticut. Identification of all speci- tributions, morphology/color pattern, male calling songs, and ecology mens used for molecular sequencing was based on song, location, and with an emphasis on habitat preference. morphology. Genomic DNA was extracted from 1 to 2 legs using the Molecular analyses (Arensburger et al., 2004b; Marshall et al., 2008; Qiagen DNeasy Blood & Tissue kit (Qiagen, 2006) or Clontech Nu- Ellis et al., 2015) based largely on mtDNA support the major patterns of cleospin Tissue kits (Clontech, Mountain View, CA, USA) according to Kikihia relationships hypothesized by Dugdale and Fleming, including manufacturer’s instructions, except that the digestion was conducted the early divergence of the two shade singer taxa and a later radiation over 8–18 h at 54 °C. of several species groups that seem to be organized by habitat type. One striking difference between the Dugdale and Fleming cladogram and 2.2. Gene selection and sequencing the mtDNA phylogeny is the presence of a fourth mtDNA group in the major species radiation, nicknamed the Westlandica group by Marshall Genes were amplified using previously published PCR methods et al. (2008) (Fig. 1C). While the three other groups (Green Foliage, (Marshall et al., 2011). The mitochondrial cytochrome oxidase I (COI) Muta, and Rosea) seem to be patterned by habitat type, the Westlandica gene was sequenced to ensure that species were identified correctly and group contains grass species (K. “westlandica north” and “K. westlan- that the patterns of mitochondrial evolution (Marshall et al., 2008) dica south” previously considered to be K. muta muta and placed in the were consistent in this set of specimens. A touchdown program with Muta group by Dugdale and Fleming, Fig. 1A), a tussock specialist (K. annealing temperatures of 55–45 °C was used with the primers C1-J- angusta, Muta group) and two shrub-inhabiting taxa (K. “tasmani”–wet 1513 (5′CATTTTTGGTATTTGATCAGG 3′) and TL2-N-3014 (Simon shrub, Green Foliage group; K. “murihikua”–dry scrub, Rosea group). et al., 1994). A portion of the calmodulin intron (Cal) was amplified To differentiate between the intuitive habitat group designations and with Cal60For (Buckley et al., 2006) or Cal_short_2For (AACGAAGTA- the mtDNA clades, the remaining green cicadas (previously Green Fo- GATGCCGATGG, and Cal2Rev (UBC insect primer kit) or Cal_shor- liage group) are renamed the “Cutora” group, using the species name of t_716Rev (CGTGCCATCATTGTAAGGAA) using a touchdown program the majority member of this group. with annealing temperatures of 60–50 °C. An intron of period was The identification of the Westlandica group by molecular data also amplified with Per 393F (CTGTCGATCTGCATATAAATGTGAG), Per raised phylogenetic concern because it contained two species formerly 722R (GCAATGGYTTGAGATGCCTRAGGG), and a touchdown program believed to be synonymous with K. muta: K. “westlandica north” and K. with annealing temperatures from 55 to 45 °C. Multiple exons and in- “westlandica south”. These two species share morphological, ecological trons of elongation factor 1α (EF-1α) were sequenced in two sections and song characters with K. muta. Marshall et al. (2011) demonstrated with the first part amplified with primers PA F650-ambig (Lee and Hill, considerable mtDNA divergence between Westlandica group species (K. 2010), EF1-N-1419 (Sueur et al., 2007), and annealing temperature “westlandica north” and K. “westlandica south”) and the rest of the K. 58 °C; the second part was amplified with EF1-F001-cicada muta complex. Here we use five nuclear loci and multiple gene tree- (Arensburger et al., 2004a), EF1-R752-cicada (Arensburger et al., species tree methods as well as gene sequence concatenation to test the 2004a) and annealing temperature 55 °C. EF-1α was chosen for phy- validity of the Westlandica group and to estimate the phylogeny of this logenetic analysis because it worked well for many other cicada species genus and to examine patterns of mito-nuclear discordance in all 31 phylogenies in the past and does not appear to have multiple paralogs New Zealand Kikihia taxa. in this group (Arensburger et al., 2004a; Buckley et al., 2006; Marshall et al., 2008, 2012; Owen et al., 2015). Exons of the glutaminyl tRNA 2. Materials and methods synthase (QtRNA) gene were amplified with QtRNA 612F (Owen et al., 2015), QtRNA 1714R (Owen et al., 2015), and a touchdown program 2.1. Taxon sampling and DNA extraction with annealing temperatures from 55 to 45 °C. Exons of the acetyl transferase 1 (ARD1) gene were amplified with 1041F (Owen et al., Specimens were chosen to represent pure species populations from 2015), 1733R (Owen et al., 2015) and a touchdown program with collection localities where little to no hybridization has been observed annealing temperatures from 55 to 45 °C. or hypothesized during 20+ years of research by Fleming and Dugdale DNA products were purified using a 20–50% dilution of Exo-SAP-IT and 20+ additional years of research by Simon Lab members. Sixty (Affymetrix). Cycle sequencing was conducted using the Applied total specimens were analyzed for this study representing 30 Kikihia Biosystems Big Dye Terminator v1.1 or v3.1 cycle sequencing kit taxa: 16 described Kikihia species and subspecies, and 14 putative (Perkin Elmer) at a ⅛ to ¼ scale reaction volume using BDX64 Big Dye species (denoted with quotation marks) represented by two specimens enhancing buffer (MCLAB), and the product was cleaned by Sephadex per taxon, each from a different collecting locality and collection year (GE Healthcare). Sequences were obtained using an ABI 3100 se- (Table S1). Several of the specimens used in this study were previously quencer, analyzed using ABI Prism Sequencing software (Applied sequenced for COI and EF-1α by Marshall et al. (2008) and are noted in Biosystems) and edited in Geneious (Biomatters Ltd). Supplemental Table 1. There are no representatives from the Chatham Island species, K. longula (very closely related to K. “aotea east” 2.3. Sequence alignment and model selection [Marshall et al. 2011]), and only a single specimen from the Kermadec Island subspecies K. cutora exulis (very closely related to K. cutora cutora All sequences were aligned with MUSCLE (Edgar, 2004) in Geneious [Marshall et al., 2011]), but there is little conflict over the phylogenetic using default parameters and checked by eye. In some chromatograms, position of these two taxa (Arensburger et al., 2004b; Marshall et al., sequences were shifted due to heterozygous insertions or deletions, 2011). Kikihia “tuta” is represented by three specimens to account for therefore, Indelligent (Dmitriev and Rakitov 2008) was used on both the large geographic variability found in three different mitochondrial the forward and reverse sequences. If both directions showed the same clades within this song-defined putative species (Marshall et al., 2011). insertion, the sequence was edited to include the insertion for further Previous studies using outgroups (Marshall et al., 2008; Arensburger analyses. Exon regions of EF1-α, QtRNA, ARD1 and COI were translated

174 S.E. Banker et al. Molecular Phylogenetics and Evolution 116 (2017) 172–181 into amino acid code in Geneious (Biomatters Ltd.) and were aligned after a burn-in of 25% was checked with Tracer v1.6 and a maximum according to the translated sequence. clade credibility tree was constructed. Alternative partitioning schemes and models were tested in Unlike ∗BEAST, the multi-locus coalescent-based ASTRAL-II v4.10.9 PartitionFinder v.1.1.1 (Lanfear et al., 2012) by giving the program (Mirarab and Warnow, 2015) estimates species trees by minimizing the different potential groups of introns or codon positions for the mtDNA distance between quartets based on individual gene trees and the spe- and nuclear exons. The Bayesian Information Criterion (BIC) was used cies tree. The best maximum likelihood gene tree for each nuclear locus to compare alternative partitioning schemes and models of evolution, as well as 1000 bootstrap replicates were estimated with RAxML which were estimated on a neighbor-joining tree created by Parti- v8.1.17 (Stamatakis, 2014). ASTRAL was used to estimate the species tionFinder (Lanfear et al., 2012). tree based on the five best gene trees and then again on 1000 bootstrap replicates for each gene to determine whether gene tree uncertainty 2.4. Gene tree estimation influenced the species tree estimation. Specimens were assigned a species designation a priori and a heuristic search of the unrooted trees Gene trees for each of the six (5 nuclear, 1 mitochondrial) genes was performed. were constructed using both Maximum Likelihood (ML) and Bayesian As an alternative to multi-locus coalescent based methods, the Inference (BI) using the data partitioning and model of evolution program SVDquartets (Chifman and Kubatko, 2014) estimates patterns schemes for each gene sequence determined by PartitionFinder of shared nucleotides across taxa at each nucleotide site (site patterns) (Supplemental Table 2). ML analysis of each gene tree was conducted to infer the species tree. Unlike the other two species tree methods, with GARLI v2.0 (Zwickl, 2006) with bootstrap support values esti- species designation is not required a priori. The SVDquartets algorithm mated using 1000 non-parametric bootstrap replicates and default was implemented in PAUP∗ v4.0a150 (Swofford, 2016) with exhaustive parameters. BI was performed using Mr. Bayes v3.2.1 (Ronquist and quartet sampling (N = 455,126 quartets), 4 threads, using the multi- Huelsenbeck, 2003), which was run twice, for 50 million generations species coalescent with 100 bootstrap replicates, a seed of 123 and the with four chains (three heated) with data partitions unlinked. A burn-in QFM quartet assembly algorithm. The 59-taxon concatenated nuclear of 25% was sufficient for all analyses, as determined by checking for data set was used and the phylogeny was rooted with the two K. scu- stability of the parameter estimates using Tracer v1.6 (Rambaut et al., tellaris specimens based on information from prior studies (Marshall 2014). et al., 2008).

2.5. Concatenation phylogeny estimation 2.7. Test of monophyly of Westlandica group

The five nuclear genes were concatenated and the best partitioning To assess the monophyly of the Westlandica clade we compared and model schemes were estimated with PartitionFinder before con- unconstrained and constrained tree topologies using the concatenated ducting ML analyses using Garli v2.0 and BI analyses using BEAST nuclear data set. We used the concatenated data set because there was v1.8.2 (Drummond et al., 2012). Bootstrap support for ML analyses was very little support in any of the individual nuclear gene trees. The un- assessed with 1000 non-parametric bootstrap replicates and models constrained concatenated phylogenetic analysis obtained with BEAST were unlinked between partitions. BEAST was run with an uncorrelated (above) was compared to analyses estimated with the same parameters lognormal relaxed clock with models unlinked across partitions using except a clade of taxa was constrained to be monophyletic. Five ana- the models and partitions determined with PartitionFinder lyses were run that constrained each of the major clades to be mono- (Supplemental Table 2). Multiple tree priors were tested: 1 – constant phyletic: (1) Westlandica (conservative): K. “westlandica north”, K. size coalescent, 2 – coalescent exponential growth and 3 – Yule birth- “westlandica south”, K. angusta, K. “murihikua” and K. “tasmani” ; (2) death to test the effects of these priors on topology and tree length. The Westlandica (inclusive): K. “westlandica north”, K. “westlandica chain was run for 50 million iterations sampling trees and logging south”, K. angusta, K. “murihikua” K. “tasmani”, K. subalpina, and K. parameters every 5000 steps. Convergence was assessed with Tracer “flemingi”; (3) Rosea: K. rosea, K. “balaena”, K. “acoustica”, K. “pe- v1.6 (Rambaut et al., 2014), which confirmed adequate mixing. Trees ninsularis” and K. horologium; (4) Cutora: K. cutora cutora, K. cumberi were summarized as maximum clade credibility trees after a discarded cumberi, K. c. “cumberi east”, K. dugdalei, K. ochrina, K. lanorum, and (5) 10% burnin. DensiTree v2.2 (Bouckaert, 2010; Bouckaert and Heled, Muta: K. muta muta, K. “muta east”, K. “aotea east”, K. “aotea west”, K. 2014) was used to visualize conflicts within consensus trees based on paxillulae, K. “tuta”, K. “nelsonensis”, and K. “astragali”. The topologies the topologies sampled in the posterior. were compared by examining the marginal likelihoods via path sam- pling and stepping stone algorithms (Baele et al., 2012, 2013) im- 2.6. Species tree estimation plemented in BEAST with 100 path steps of one million iterations and 10% burnin. We compared the fit of the resulting inferences with the Species trees were inferred using ASTRAL, ∗BEAST, and data by calculating Bayes Factors (BF). The strength of BF comparisons SVDQuartets. Two species trees were constructed using ∗BEAST was interpreted according to the framework of Kass and Raftery (1995) (Drummond 2007): (1) using combined data from all six genes (5 nu- where 0 < lnBF < 1 implies there is no difference between the models, clear + 1 mitochondrial); and (2) using data only from the five nuclear 1 < lnBF < 3 implies there is evidence for one model over another, genes. ∗BEAST uses a multi-locus coalescence approach, which si- 3 < lnBF < 5 implies there is strong support that one model is better multaneously estimates trees for each partition and a combined species than another and lnBF > 5 implies that one model is decisively better. tree. Each specimen was assigned a species designation a priori and missing taxa were either removed from the analysis or combined with 3. Results taxa for which there is high support for a close relationship in Marshall et al. (2008, 2011). Kikihia “acoustica” was included in the K. rosea 3.1. DNA sequences and gene phylogenies species; K. convicta and K. cutora exulis were included within the K. cutora cutora sub-species. Each analysis used the partitioning scheme A total of 5413 bases were sequenced from most Kikihia specimens. and evolutionary model suggested by PartitonFinder. All model para- Sequence length, parsimony informative sites, partitioning schemes, meters except branch lengths and topology were unlinked across par- and best fitting model of evolution for each gene and for the con- titions. Each of the partitions (7 nuc, 10 nuc + mito) used a strict clock catenated data matrix are summarized in Supplemental Table 2. All and the Yule model for the species tree prior, and were run for 100 sequences were submitted to Genbank (Supplemental Table 1). million generations, sampling every 20,000th iteration, Convergence This mtDNA gene segment (COI) was sequenced to confirm the

175 S.E. Banker et al. Molecular Phylogenetics and Evolution 116 (2017) 172–181 identity of the included taxa and provide a comparison to the Kikihia supported (pp = 0.98) in the ∗BEAST species tree, with the exclusion of clades inferred by Marshall et al. (2008). The major species group de- K. “westlandica north”, K. “westlandica south”, and K. angusta, which limitations of this tree (Fig. S1A) are largely identical to the Marshall mtDNA places in the Westlandica group (Fig. 1C and Marshall et al. et al. (2008) mtDNA phylogeny with the one exception, the lack of 2008). These three taxa were identified as Muta group taxa in the in- monophyly of the Rosea group. In addition, some tip level relationships tuitive tree. Within the Muta group, there was strong support for the are unresolved or have conflicting topology with low support most monophyly of the four North Island species (pp = 0.98). This pattern is likely due to the shorter mtDNA sequences that were used in this not consistent with mtDNA tree analyses (Marshall et al. 2008, 2011), analysis compared to the previously published phylogeny (741bp vs which included South Island K. “nelsonensis” and K. “tuta” in a poly- 2152 bp). Overall, the same major patterns were observed, particularly tomous relationship with North Island K. muta muta and K. “aotea” the identification of the Westlandica group. clades. In the five-gene tree, the Westlandica and Rosea groups are The five nuclear gene trees individually show some interesting unresolved. signal but overall are poorly resolved (Fig. S1). With a few exceptions, When the mitochondrial gene COI was included in ∗BEAST analysis, they do not conflict with the mitochondrial tree. The strongest signal the Cutora and Muta groups were further resolved. K. “westlandica present (Fig. S1b, Calmodulin, and Fig. S1d, Period gene) supports the north and south” were excluded from the Muta group as in all other monophyly of the Cutora group as defined by mtDNA. The Cutora group analyses except the calmodulin gene tree,. Neither the Westlandica is unresolved in the remaining three gene trees. The monophyly of the group nor the Rosea group were resolved as monophyletic (Fig. S3). mtDNA-defined Muta group (Marshall et al., 2008, i.e., not including The ASTRAL species tree analysis of the five nuclear genes, based on the Dugdale-and-Fleming grass cicadas K. “westlandica north”, K. 5000 gene trees (Fig. 3), is more similar to the mtDNA gene tree than to “westlandica south”, and K. angusta) is not supported in any of the any of the other analyses. In the ASTRAL tree, the Muta, Cutora, and individual nuclear gene trees. In contrast, we find agreement with the Westlandica groups are all supported by 100% bootstrap scores. The intuitive tree in that Calmodulin strongly supports the inclusion of K. only differences are that K. “flemingi” and K. subalpina are excluded “westlandica north” and K. “westlandica south” in a monophyletic from the Westlandica group, K. horologium is included in the Westlan- group with the rest of the Muta group taxa. Both EF-1α and QtRNA dica group, and the Rosea group is polyphyletic. The closely related provide strong support for a monophyletic group consisting of the four sister taxa K. subalpina and K. “flemingi” form a clade that is part of “the Muta group taxa that are distributed on North Island (K. muta muta, K. main Kikihia radiation”–a polytomy with all Kikihia clades except the “aotea east”, K. “aotea west” and K. “muta east”) but do not resolve shade singers. The normalized quartet score of this species tree is 0.61, relationships among the remaining Muta group taxa. Neither of these which means that 61% of the quartet trees induced by the gene trees are patterns is consistent with the mitochondrial tree (Marshall et al., 2008, in the estimated species tree. The species tree estimated with the five 2011). Multiple trees (EF-1α, Period, QtRNA) support an independent best gene trees had a normalized quartet score of 0.65 and resolved the clade consisting of K. subalpina and K. “flemingi” separate from the same major clades. This is not surprising given the low resolution and Westlandica group. None of the individual gene-trees support the lack of support for most relationships in the individual gene trees. monophyly of the Westlandica group. The species tree produced by SVDquartets (Fig. S4) is similar to the concatenated data set phylogeny in that it shows support for the 3.2. Concatenated phylogenies monophyly of the Cutora group, the exclusion of K. subalpina and K. “flemingi” from the Westlandica group, and the placement of these two The concatenated nuclear data set shows more similarity to the taxa as a clade sister to all Kikihia other than K. scutellaris and K. cauta. mitochondrial tree than do any of the five nuclear gene trees (Fig. 2). This analysis shows low support for all other relationships. The to- There were no significant differences between BI and ML analyses (Fig. pology of the consensus tree is consistent with the other species tree S2). Multiple tree priors had no effect on branch length or topology in analyses, although with much less resolution and nodal support. the BEAST analyses. This concatenated phylogeny shows the strongest support (pp = 1, bs = 99) for the NI Cutora group. It also shows strong 3.4. Tests of monophyly support (pp = 1, bs = 95) for the monophyly of the NI Muta clade (plus the one South Island K. muta muta from mid Canterbury) and moderate Both path sampling and stepping stone algorithms estimated almost support (pp = 1, bs = 61) for the entire Muta group (excluding the K. the exact same marginal likelihood scores for every tree tested, thus “westlandica north & south” taxa). Unlike the published mtDNA tree only stepping stone likelihood scores are reported (Table 1). While all (Marshall et al. 2008), the concatenated phylogeny does not show five constrained trees show higher support than the unconstrained tree support for the monophyly of the entire Westlandica or Rosea groups, with lnBF > 3, the trees with the Cutora and Muta groups constrained both are polyphyletic with very few relationships other than K. rosea + are more highly supported, lnBF > 4, than the other three constrained “balaena” + “acoustica” well supported by either ML or BI analyses. trees. Constraint of the Westlandica group, with (“inclusive”)or DensiTree consensus trees reflect the posterior probabilities and show without (“conservative”) the sister taxa K. subalpina and K. “flemingi”, that within clades there is a high amount of uncertainty. K. subalpina show approximately the same support. and K. “flemingi” form a monophyletic clade, that is part of a large polytomy involving all lineages except the shade singers. The polyphyly 4. Discussion of the shade singers at the base of the phylogeny is expected since the tree is rooted with K. scutellaris. In contrast to the well-supported mitochondrial phylogeny (Marshall et al., 2008), none of the individual nuclear-gene species 3.3. Species tree analyses trees or concatenated phylogenies showed overall high resolution. Only the Cutora group (found only on NI) and the NI Muta group were In the ∗BEAST analysis of the five nuclear genes combined (without consistently supported as monophyletic. None of the combined gene mtDNA; Fig. 3) that uses clock rooting, K. scutellaris was identified as analyses identified the Westlandica or Rosea groups as monophyletic sister to the rest of the genus with strong support (pp = 0.99), followed although the test of constrained trees using Bayes Factors showed some by K. cauta (pp = 0.96); Marshall et al. (2008) found the relationship to support for the Westlandica clade. Calmodulin was the only gene tree be unresolved. The monophyly of the NI Cutora group was strongly that strongly supported inclusion of K. “westlandica north” and K. supported (pp = 0.98) with the exclusion of K. “tasmani”, which the “westlandica south” in the Muta group, as suggested by the intuitive mtDNA tree placed in the Westlandica group, and the intuitive tree tree of Dugdale & Fleming (Fig. 1A). In general, the two NI Kikihia placed in the Cutora group. The monophyly of the Muta group was also species groups were well supported as monophyletic and the SI species

176 S.E. Banker et al. Molecular Phylogenetics and Evolution 116 (2017) 172–181

Fig. 2. Left: Bayesian phylogeny (BEAST) of 5 nuclear genes concatenated. The mtDNA sequence COI is not included in this analysis. Bayesian posterior probabilities (≥0.95)/ML bootstrap percentages (≥60) are shown. Right: DensiTree consensus. Taxon and colors correspond to the mtDNA groups shown in Fig. 1C. Dark gray bars within groups identify sub- groups referenced in the text. Triangles on branch tips represent current distribution with upwards facing triangles for North Island, and downwards facing for South Island, and “O” for restricted distribution on an offshore island. Tip labels indicate taxon name, district code (Crosby et al., 1998), collection locality (Simon Lab code), and year collected. groups were not. Below we discuss the results of our analyses and our species tree. ASTRAL has been shown to be more accurate than con- reasons for concluding that relationships among South Island taxa are catenation methods, especially when incomplete lineage sorting (ILS) is confounded by extensive hybridization. high (Chou et al., 2015). In addition, ASTRAL has also been shown to be more accurate than SVD quartets when moderate to high ILS is occur- ring (Chou et al., 2015). However, coalescent-based species tree 4.1. Analysis method comparisons methods, especially ones like ASTRAL that are summary methods (and do not include input-tree branch length information) are sensitive to Individual gene trees had very little resolution when analyzed se- gene tree estimation error (Roch and Warnow, 2015). The SVD quartets parately, which is not surprising given the lower substitution rates of analysis (Fig. S4) recovered the Cutora clade (bs = 79, Fig. S4) and nuclear versus mitochondrial genomes and the smaller effective popu- strongly supported (bs = 91, Fig. S4) the position of the Subalpina lation size of mitochondrial DNA due to its haploid, uniparental in- group as sister to all Kikihia other than the shade singers, however, heritance (Zink and Barrowclough 2008). Two of the nuclear markers there was low support for all other inferred relationships throughout contained only protein-coding regions which are under functional the tree. constraints that further reduce the substitution rate and phylogenetic The challenge in investigating the evolutionary history of Kikihia is informativeness. Interestingly, the mitochondrial phylogeny estimated the extensive past and present hybridization and introgression (Fleming with COI in this study did not group K. subalpina and K. “flemingi” with and Dugdale, unpubl notes; Marshall et al., 2001). Both species tree and the Westlandica group (Fig. S1). It also failed to group K. “peninsularis” concatenation methods can be misled by introgression and both ap- with the rest of the Rosea group taxa but support for this relationship proaches assume that gene tree discordance is due to ILS (Solís-Lemus did not exist in the mtDNA phylogeny of Marshall et al. (2008) either. et al., 2016). Differences in our mtDNA phylogeny and Marshall et al.’s (2008) tree In this study, we find no significant differences between con- are likely due to our study’s use of a smaller portion of the mitochon- catenated-data analyses and species tree methods. While the topologies drial genome. of the trees are not consistent in inferring within-group relationships The two coalescent species tree methods used, ASTRAL and among Westlandica and Rosea taxa, they are consistent in having little ∗BEAST, were congruent. Both methods recovered the Muta and Cutora support for these groups. Within the monophyletic Muta and Cutora groups with high support and found a lack of support for monophyletic groups, there is no support for relationships among taxa except in the Rosea and Westlandica groups. Interestingly, the ∗BEAST method found ASTRAL analysis. Concatenation can be misleading, especially when ILS little support for any of relationships within the Rosea and Westlandica is high because it assumes all sites evolve identically and groups, while ASTRAL strongly supported almost every node in the

177 S.E. Banker et al. Molecular Phylogenetics and Evolution 116 (2017) 172–181

Fig. 3. Species tree comparisons based the five nuclear gene regions (calmodulin, EF-1α, period, QtRNA, and ARD1). The mtDNA COI data is not included in this analysis [See Fig. S3 for 6-gene *BEAST]. Left: Maximum Likelihood species tree estimated with ASTRAL. Branch labels are local posterior probabilities and branch lengths are given in coalescent units. Right: *BEAST analysis of the same data, Bayesian posterior probabilities > 0.95 are presented, and all non-significant branches were collapsed. Taxon and colors correspond to the mtDNA groups shown in Fig. 1C. Triangles on branch tips represent current distribution with upwards facing triangles for North Island, and downwards facing for South Island, and “O” for restricted distribution on an offshore island.

Table 1 4.2. Kikihia phylogenetic relationships Bayes Factor of multiple tests of monophyly. Likelihood scores are reported from stepping stone algorithm for unconstrained analysis, and with the trees constrained for the four The intuitive tree of Dugdale and Fleming (Fig. 1A) included a ff di erent sub-generic mitochondrial groups (Muta, Cutora, Rosea, and Westlandica). “ ” Westlandica group was constrained both including and excluding the two sister taxa K. group known as Green Foliage Cicadas . Mitochondrial phylogenies subalpina and K. “flemingi” because these two taxa consistently came out paraphyletic to (Arensburger et al., 2004a; Marshall et al., 2008) suggested that some the rest of the group in other analyses. Green Foliage Cicadas–specifically K. subalpina, K. “flemingi”, and K. “tasmani”–were not members of this group but rather part of a newly Clade constrained Path Stepping BF lnBF Log10BF envisioned “Westlandica group.” The remaining green cicadas were sampling stone “ ” likelihood likelihood well-supported as a group and renamed the Cutora group. The Cutora group consists entirely of North Island species that radiated rapidly a Unconstrained −10,148 −10,148 little more than 2 Ma, and continued to differentiate geographically − − − Westlandica (excl. K. 10,129 10,129 38 3.637586 1.579784 over the last two million years to form subspecies and geographically subalpina & K. fi “flemingi”) de ned clades (Marshall et al., 2008; Ellis et al., 2015). Rosea −10,131 −10,131 −34 3.526361 1.531479 Dugdale and Fleming also recognized a group of grass cicadas found Cutora −10,117 −10,117 −62 4.127134 1.792392 on both North and South Islands as the Muta group (Fig. 1A). They − − − Muta 10,114 10,114 68 4.219508 1.832509 identified K. “westlandica north and south” as part of the species K. Westlandica (incl. K. −10,131 −10,130 −34 3.526361 1.531479 subalpina & K. muta (thus they are not listed separately on their intuitive tree). Con- “flemingi”) trary to our initial hypothesis of Muta group monophyly as defined by Dugdale and Fleming based on song and morphology, there is strong support from all nuclear-gene species trees for the exclusion of K. independently, which due to recombination is not necessary true. This “westlandica north” and K. “westlandica south” from the Muta group. does not appear to be the case with this dataset given the strong si- This agrees with the conclusions of the mitochondrial tree (Marshall milarities in results between concatenation (Fig. 2) and the three dif- et al., 2008, Fig. S1A). Within the Muta group there is one subclade that ferent species-tree-estimation methods (Fig. 3, Fig. S4); this result is is consistently recovered and includes all the NI grass species, K. “muta congruent with other empirical studies (Tonini et al., 2015; Pyron et al., east”, K. “aotea east”, K. “aotea west”, and K. muta muta (found on both 2014; Kozak et al., 2015). The mtDNA phylogeny estimated the split islands). This pattern is not supported in the mitochondrial phylogeny and radiation of these sub-generic groups at roughly the same time. If (Marshall et al., 2008) where the NI K. muta muta group weakly with SI ILS was rampant, we would expect similar levels of conflicting re- K. “nelsonensis” before joining a clade containing the two K. “aotea” lationships between taxa within different groups. However, our results species). The nuclear-gene based relationships within the Muta group show nearly all conflict involves the Westlandica and Rosea group taxa, are neither reconstructed with consistency nor supported by the dif- and almost none involves taxa in the Muta or Cutora groups. Therefore, ferent analysis methods. we conclude that ILS likely did not have a strong influence. Nuclear-based species tree estimation of a Cutora-group clade is

178 S.E. Banker et al. Molecular Phylogenetics and Evolution 116 (2017) 172–181 congruent with the mtDNA analyses and well supported in all cases. All multi-gene analysis methods of nuclear markers (Coalescent- The nuclear tree differs from the intuitive tree in that the two closely based and site-pattern-based species tree analyses, and concatenation related species K. subalpina, and K. “flemingi” are excluded from the analyses) fail to support the monophyly of the mtDNA Westlandica Cutora group. Within the Cutora clade no analysis of nuclear genes group. Instead the relationships between the taxa are unresolved and show well resolved relationships. The only exception is the grouping of have varying support for relationships with Rosea group taxa de- K. ochrina and K. dugdalei as sister taxa with high support in every pending on the analysis. When BEAST was run using the concatenated analysis. nuclear dataset, there was moderate Bayes Factor support for the Unlike the mtDNA phylogeny of Marshall et al. (2008), analyses of monophyly of the Westlandica clade over a non-constrained model the nuclear markers consistently place K. subalpina and K. “flemingi”, (Table 1). However, tests of monophyly using Bayes Factors are biased as a monophyletic group that is part of a deep polytomy for all Kikihia towards accepting the monophyletic hypothesis versus an un- species groups except the shade singers. These sister species were ori- constrained one (Bergsten et al., 2013) so this result should be inter- ginally hypothesized to belong to the Cutora group based on song and preted with caution. morphology, and to the Westlandica group based on mitochondrial Given the lack of resolution in our analyses and a well-supported DNA data. K. “flemingi” occurs only on South Island and is among the estimation of the evolution of taxa within this genus, additional testing few tree-dwelling species in this genus (K. laneorum of the Cutora is needed to fully accept or reject our hypotheses. The strongly sup- group, and the shade singers, K. scutellaris, and K. cauta, are the only ported mtDNA Westlandica group, we propose, resulted from an an- other tree-dwelling species). K. subalpina, occurs only on North Island cient mitochondrial capture of a K. subalpina-like mtDNA three to four and is a shrub-dwelling species found in hills and subalpine mountain million years ago when the Kikihia species began to radiate (Marshall environments. Molecular dating of these species estimated the diver- et al., 2008). This invading mitochondrion was then sequentially cap- gence between K. subalpina and K. “flemingi” to be 1.2–1.35 Ma and tured by species from multiple ecological groups (grass, dry scrub, and there are differences in the segments of their calling songs that cue the wet shrub). Nuclear genes would have hybridized as part of these female response (Marshall et al., 2009). capture events and this gene mixing would have confused nuclear-gene Every analysis of the nuclear gene trees shows that the Rosea group species-tree methods that did not allow reticulation. The Calmodulin is either unresolved or paraphyletic due to the exclusion of K. “pe- gene was the only nuclear gene we sequenced that retained the signal of ninsularis” and K. horologium. In the mitochondrial tree of Marshall et al. K. “westlandica north” and K. “westlandica south” former membership (2008), K. horologium is not included and the ML support for the re- in the Muta group. Contemporary hybridization may also be con- maining Rosea-group taxa is only 69%. It is only their combined mtDNA founding nuclear gene trees; Marshall et al. (2011) reported 20 pairs of and EF-1a phylogeny that places K. horologium in the Rosea group. In Kikihia species from which they documented hybrid song at contact our nuclear gene analyses, the three taxa, K. rosea, K. “balaena” and K. zones. They also provided DNA evidence for hybridization and mi- “acoustica”, are recovered with strong support in the ASTRAL and tochondrial capture involving among three species (K. muta, K. tuta, ∗BEAST species trees and in the concatenated Bayesian analysis. Note and K. paxillulae). that in some analyses K. “acoustica” and K. rosea are combined because they do not differ in nuclear sequence. These three taxa are allopa- 4.4. North Island (NI) vs. South Island (SI) patterns of support trically distributed on the SI, their habitats and male calling songs are very similar, and they form a strongly supported clade on the mtDNA We suggest that extensive hybridization among multiple species on tree (Marshall et al., 2008; Fig. 1C; Fig. S1A). It is uncertain, without South Island but not on North Island explains why our nuclear gene further evidence, whether these should be considered separate species trees show little resolution of SI clades and strong support for mono- or subspecies. phyletic NI clades. We suggest that this pattern could be related to differences in habitat instability and hybridization opportunity on the 4.3. The Westlandica clade and nuclear-mitochondrial discordance two islands throughout the Pleistocene (2.6–0.01 Ma). Steady, uni- directional environmental changes (e.g., mountain building events) Conflict between the history of mitochondrial and nuclear genomes could have led to inter-species isolation and well defined species-group is not uncommon (Marshall et al., 2011; Toews and Brelsford, 2012; clades illustrated by North Island species, while on South Island the Sullivan et al., 2014; Good et al., 2015; Folk et al., 2017). In addition to more dramatic oscillatory environmental changes associated with gla- ILS and hybridization, which can cause discordance between any gene ciation (Alloway et al., 2007; Williams et al., 2015; Wood et al., 2017) trees, nuclear and mitochondrial genomes can estimate different evo- may have increased hybridization, resulting in the lack of monophyletic lutionary histories due to a number of additional genomic features such species group clades and discordance between mitochondrial and nu- as ploidy, uni- vs biparental inheritance, and recombination and sub- clear markers. Cycles of glaciation and retreat impacted the amount and stitution rates (Toews and Brelsford, 2012). Given the disparities in type of habitat available more severely on South Island and this in turn nuclear-gene-encoded characters of morphology and male courtship influenced the population structure of many organisms (Craw et al., song between members of the Westlandica clade, it is a reasonable in- 2017; Wallis et al., 2016; Buckley et al., 2009, 2015; Wallis and ference that mitochondrial sequence similarity is due to mtDNA capture Trewick, 2009; Marske et al., 2009) including cicadas (Marshall et al., during an ancient hybridization event rather than multi-trait con- 2009; Marshall et al., 2011; Marshall et al., 2012). On South Island vergences. At this point, it is unclear when or from which taxon this approximately 30% glacial ice cover, and dense forest habitat limited mtDNA capture event originated. suitable habitat to a small number of isolated glacial refugia (Williams We expected that the phylogeny estimated with multiple nuclear et al., 2015; Wood et al., 2017; Alloway et al., 2007). Marshall et al. markers would support the relationships hypothesized in the intuitive (2009) identified at least three regions of probable forest refugia for the cladogram of Dugdale and Fleming (based on ecology, song, mor- species K. “flemingi” in the northern region of South Island: Northwest phology, and distribution). We hypothesized that the relationships with Nelson, Marlborough Sounds and Kaikoura. The remaining portions of the Westlandica group would not be supported, but instead Westlandica the island were covered in a mosaic of sparse scrub, grasslands, and rare group taxa would nest with taxa more similar in morphology and song forests which oscillated in their ranges along with the changing climate (e.g., K. “westlandica north” and K. “westlandica south” would be sister (Fleming, 1977; Alloway et al., 2007; Wood et al., 2017). Globally these to K. muta muta). Our results support the opposite conclusion; severe range changes were the cause of extinctions and population Westlandica group taxa are not nested within the Muta and Cutora bottleneck events in other organisms (Hewitt, 2000, 2004; Wallis et al., clades whose members they resemble, instead these taxa fall outside of 2016). During melting periods, forest and shrub habitat on South Island these well-supported monophyletic clades. expanded and habitat heterogeneity increased. During this period,

179 S.E. Banker et al. Molecular Phylogenetics and Evolution 116 (2017) 172–181 species restricted to forest and shrub habitats must have expanded their Data accessibility ranges rapidly to take advantage of the newly available resources, a pattern consistent with other NZ invertebrates (Marske et al., 2009; DNA sequences: COI-GenBank accessions MF796031–MF796059; Wallis and Trewick, 2009; Buckley et al., 2015). It is likely that as re- EF-1α MF796172–MF796214; CAL MF795980–MF796030; PER treating glaciers allowed some populations to expand their ranges MF795920–MF795979; ARD1 MF796114–MF796171; QtRNA southward and towards the interior, contact zones allowed hybridiza- MF796060–MF796113. tion among the endemic grass species and between grass species and shrub and/or scrub species. In contrast to the extreme habitat cycles Acknowledgments experienced on South Island, North Island experienced little to no glacial cover during glacial maxima. The northern third was densely A subset of the CAL, COI, and EF-1α sequences were generated by forested and the lower two thirds retained forest patches or shrubs or David C. Marshall and Megan Ribak. We thank members of the Simon was submerged to various degrees between 5 and 2 Mya (Alloway et al., lab for help in field collecting, especially David Marshall and Kathy Hill. 2007; Marske et al., 2009; Ellis et al., 2015; Wood et al., 2017). This We thank David Marshall, Kathy Hill & Geert Goemans for useful dis- habitat continuity in both time and space on North Island may have cussions and the NZ Department of Conservation for permits and as- discouraged hybridization between populations specialized to other sistance in the field. We thank Francesco Frati, Adrian Patterson, and an habitat types. anonymous reviewer for helpful suggestions in the review stage. Most Multiple lines of evidence give strong support that the extensive of these analyses were run on the University of Oslo Bioportal (Kumar hybridization of NZ cicadas currently observed on South Island et al., 2009) or the University of Connecticut Bioinformatics Facility (Marshall et al., 2011) has been occurring throughout the diversifica- computer cluster. This project was supported by National Science tion of the genus, driven largely by changing environments. The ex- Foundation grant DEB-09-55849 (to C.S.) and the University of Con- tensive grasslands found in modern day New Zealand were absent until necticut Summer Undergraduate Research Fund (to S.E.B) & UConn the Quaternary, long after the origination of the Muta group 3–4 Ma. Honors Program Life Science Research Award (to S.E.B.). Grass specialist species such as K. muta muta must have originated in a different environment, likely low-lying wetland habitat dominated by Author Contributions monocotyledonous vegetation upon which K. muta muta has been shown to grow larger and faster compared to grasses (Fleming and SEB, EW, and CS conceived of ideas and designed experiments. SEB Scott, 1970). Adaptation of K. muta muta to grasslands allowed the collected data, SEB and EW conducted data analysis. SEB led manu- range of the species to expand from North Island onto South Island and script writing, with contributions from EW and CS. down the east coast, where it came into contact with other grassland species in the Muta group, a pattern supported by molecular evidence Appendix A. Supplementary material (Marshall et al., 2008). There is abundant evidence for hybridization between the invading North Island K. muta muta and the resident South Supplementary data associated with this article can be found, in the Island saltmarsh species K. “tuta” (Marshall et al., 2011). Additionally, online version, at http://dx.doi.org/10.1016/j.ympev.2017.08.009. there is evidence for hybridization between these hybrid “muta-tuta” individuals and South Island lowland wet-scrub resident K. paxillulae References (Marshall et al., 2011). Our results add an additional layer of complexity to the under- Alloway, B.V., Lowe, D.J., Barrell, D.J.A., 2007. Towards a climate event stratigraphy for standing of South Island biogeography. Wallis et al. (2016) reported New Zealand over the past 30,000 years. J. Quat. Sci. 22, 9–35. Arensburger, P., Buckley, T.R., Simon, C., Moulds, M., Holsinger, K.E., 2004a. patterns of phylogenetic splits between taxa distributed northwest of Biogeography and phylogeny of the New Zealand cicada genera (Homoptera: New Zealand’s Southern Alps, and taxa distributed to the southeast. Cicadidae) based on nuclear and mitochondrial DNA data. J. Biogeogr. 31, 557–569. Within the genus Kikihia,itisdifficult to test these biogeographic hy- Arensburger, P., Simon, C., Holsinger, K., 2004b. Evolution and phylogeny of the New Zealand cicada genus Kikihia Dugdale (Homoptera: Auchenorrhycha: Cicadidae) with potheses without well supported relationships between South Island special reference to the origin of the Kermadec and Norfolk Islands’ species. J. taxa. If the mitochondrial clades are supported by nuclear genes as well, Biogeogr. 31, 1769–1783. [e.g., ((K. “westlandica north & south”, “tasmani”), (K. “murihikua”, K. Avise, J.C., Ankney, C.D., Nelson, W.S., 1990. Mitochondrial gene trees and evolutionary – angusta))] then the patterns described by Wallis et al. (2016) are sup- relationships of mallard and black ducks. Evolution 44, 1109 1119. Baele, G., Lemey, P., Bedford, T., Rambaut, A., Suchard, M.A., Alekseyenko, A.V., 2012. ported. However, our analyses of nuclear genes confound this general Improving the accuracy of demographic and molecular clock model comparison pattern [e.g., 5-gene ASTRAL analysis ((K. “westlandica north & south, while accommodating phylogenetic uncertainty. Mol. Biol. Evol. 29 (9), 2157–2167. K.”murihikua”), (K. “tasmani”, K. angusta)). The strong gene-tree in- Baele, G., Li, W.L.S., Drummond, A.J., Suchard, M.A., Lemey, P., 2013. Accurate model selection of relaxed molecular clocks in Bayesian phylogenetics. Mol. Biol. Evol. 30 congruence suggests that hybridization was ongoing between South (2), 239–243. Island Kikihia taxa throughout the Pleistocene, even between species Bergsten, J., Nilsson, A.N., Ronquist, F., 2013. Bayesian tests of topology hypotheses with separated by the Southern Alps. an example from diving beetles. Syst. Biol. 62 (5), 660–673. Bouckaert, R.R., 2010. DensiTree: making sense of sets of phylogenetic trees. As the number of markers used in phylogenetic datasets has in- Bioinformatics 26 (10), 1372–1373. creased in the past decade so has the detection of gene-flow between Bouckaert, R., Heled, J., 2014. DensiTree 2: Seeing Trees Through the Forest. bioRxiv, species. Understanding the interplay of factors that affect the pre- 012401. fl Buckley, T.R., Arensburger, P., Simon, C., Chambers, G.K., 2002. Combined data, valence of inter-species gene ow are important not only for biogeo- Bayesian phylogenetics, and the origin of the New Zealand cicada genera. Syst. Biol. graphy, but also for fundamental understanding of speciation. Our 51 (1), 4–18. finding of little gene-tree discordance between North Island taxa and Buckley, T.R., Cordeiro, M., Marshall, D.C., Simon, C., 2006. Differentiating between hypotheses of lineage sorting and introgression in New Zealand alpine cicadas strong gene-tree discordance between South Island taxa suggests that (Maoricicada Dugdale). Syst. Biol. 55 (3), 411–425. fluctuating environments may increase the likelihood of gene flow Buckley, T.R., Marske, K.A., Attanayake, D., 2009. Identifying glacial refugia in a geo- throughout the speciation process. Further investigations of how cli- graphic parthenogen using palaeoclimate modelling and phylogeography: the New – matic fluctuations of varying period and intensity affect gene flow may Zealand stick insect Argosarchus horridus (White). Mol. Ecol. 18 (22), 4650 4663. Buckley, T.R., Krosch, M., Leschen, 2015. Evolution of New Zealand : summary and provide additional insights to the relationship of gene-flow and diver- prospectus for future research. Austral Entomol. 54, 1–27. gence. Carstens, B.C., Knowles, L.L., 2007. Estimating species phylogeny from gene-tree prob- abilities despite incomplete lineage sorting: an example from Melanoplus grass- hoppers. Syst. Biol. 56 (3), 400–411. Chifman, J., Kubatko, L., 2014. Quartet inference from SNP data under the coalescent model. Bioinformatics 30 (23), 3317–3334.

180 S.E. Banker et al. Molecular Phylogenetics and Evolution 116 (2017) 172–181

Chou, J., Gupta, A., Yaduvanshi, S., Davidson, R., Nute, M., Mirarab, S., Warnow, T., Marshall, D.C., Hill, K.B., Moulds, M., Vanderpool, D., Cooley, J.R., Mohagan, A.B., 2015. A comparative study of SVDquartets and other coalescent-based species tree Simon, C., 2016. Inflation of molecular clock rates and dates: molecular phyloge- estimation methods. BMC Genom. 16 (10), 1. netics, biogeography, and diversification of a global cicada radiation from Australasia Craw, D., Upton, P., Waters, J., Wallis, G., 2017. Biological memory of the first (: Cicadidae: ). Syst. Biol. 65, 16–34. Pleistocene glaciation in New Zealand. Geology 45 (7), 595–598. Marske, K.A., Leschen, R.A., Barker, G.M., Buckley, T.R., 2009. Phylogeography and Crosby, T.K., Dugdale, J.S., Watt, J.C., 1998. Area codes for recording specimen localities ecological niche modelling implicate coastal refugia and trans-alpine dispersal of a in the New Zealand subregion. New Zealand J. Zool. 25, 175–183. New Zealand fungus beetle. Mol. Ecol. 18 (24), 5126–5142. Dmitriev, D.A., Rakitov, R.A., 2008. Decoding of superimposed traces produced by direct Mirarab, S., Warnow, T., 2015. ASTRAL-II: coalescent-based species tree estimation with sequencing of heterozygous indels. PLoS Comput. Biol. 4 (7), e1000113. many hundreds of taxa and thousands of genes. Bioinformatics 31 (12), i44–i52. Edgar, R.C., 2004. MUSCLE: multiple sequence alignment with high accuracy and high Owen, C.L., Marshall, D.C., Hill, K.B.R., Simon, C., 2015. The phylogenetic utility of throughput. Nucleic Acids Res. 32 (5), 1792–1797. acetyltransferase (ARD1) and glutaminyl tRNA synthetase (QtRNA) for re- Edwards, S.V., Xi, Z., Janke, A., Faircloth, B.C., McCormack, J.E., Glenn, T.C., Zhong, B., constructing Cenozoic relationships as exemplified by the large Australian cicada Wu, S., Lemmon, E.M., Lemmon, A.R., 2016. Implementing and testing the multi- Pauropsalta generic complex. Mol. Phylogenet. Evol. 83, 258–277. species coalescent model: a valuable paradigm for phylogenomics. Mol. Phylogenet. Pyron, R.A., Hendry, C.R., Chou, V.M., Lemmon, E.M., Lemmon, A.R., Burbrink, F.T., Evol. 94, 447–462. 2014. Effectiveness of phylogenomic data and coalescent species-tree methods for Ellis, E.A., Marshall, D.C., Hill, K.B., Owen, C.L., Kamp, P.J., Simon, C., 2015. resolving difficult nodes in the phylogeny of advanced snakes (Serpentes: Phylogeography of six codistributed New Zealand cicadas and their relationship to Caenophidia). Mol. Phylogenet. Evol. 81, 221–231. multiple biogeographical boundaries suggest a re-evaluation of the Taupo Line. J. Qiagen, 2006. DNeasy Blood and Tissue Handbook, Valencia, CA, USA. Biogeogr. 42 (9), 1761–1775. Rambaut, A., Suchard, M., Xie, W., Drummond, A., 2014. Tracer v. 1.6. Institute of Fleming, C.A., 1977. The history of life in New Zealand forests. NZ J. Forest. 22, 249–262. Evolutionary Biology. University of Edinburgh. Fleming, C.A., Scott, G.H., 1970. Size differences in cicadas from different plant com- Roch, S., Warnow, T., 2015. On the robustness to gene tree estimation error (or lack munities. NZ Entomol. 4 (4), 38–42. thereof) of coalescent-based species tree methods. Syst. Biol. syv016. Folk, R.A., Mandel, J.R., Freudenstein, J.V., 2017. Ancestral gene flow and parallel or- Ronquist, F., Huelsenbeck, J.P., 2003. MrBayes 3: Bayesian phylogenetic inference under ganellar genome capture result in extreme phylogenomic discord in a lineage of mixed models. Bioinformatics 19 (12), 1572–1574. angiosperms. Syst. Biol. 66, 320–337. Simon, C., Frati, F., Beckenbach, A., Crespi, B., Liu, H., Flook, P., 1994. Evolution, Gatesy, J., Springer, M., 2014. Phylogenetic analysis at deep timescales: unreliable gene weighting, and phylogenetic utility of mitochondrial gene sequences. Entomol. Soc. trees, bypassed hidden support, and the coalescence/concatalescence conundrum. Am. 87, 651–701. Mol. Phylogenet. Evol. 80, 231–266. Solís-Lemus, C., Yang, M., Ané, C., 2016. Inconsistency of species tree methods under Good, J.M., Vanderpool, D., Keeble, S., Bi, K., 2015. Negligible nuclear introgression gene flow. Syst. Biol. 65 (5), 843–851. despite complete mitochondrial capture between two species of chipmunks. Stamatakis, A., 2014. RAxML version 8: a tool for phylogenetic analysis and post-analysis Evolution 69, 1961–1972. of large phylogenies. Bioinformatics 30 (9), 1312–1313. Hewitt, G., 2000. The genetic legacy of the Quaternary ice ages. Nature 405, 907–913. Sueur, J., Vanderpool, D., Simon, C., Ouvrard, D., Bourgoin, T., 2007. Molecular phylo- Hewitt, G., 2004. Genetic consequences of climatic oscillations in the Quaternary. Philos. geny of the genus Tibicina (Hemiptera, Cicadidae): rapid radiation and acoustic be- Trans. R. Soc. Lond. B Biol. Sci. 359, 183–195. havior. Biol. J. Linn. Soc. Lond 91 (2007), 611–626. Kass, R.E., Raftery, A.E., 1995. Bayes factors. J. Am. Stat. Assoc. 90 (430), 773–795. Sullivan, J., Demboski, J., Bell, K., Hird, S., Sarver, B., Reid, N., Good, J., 2014. Kozak, K.M., Wahlberg, N., Neild, A.F., Dasmahapatra, K.K., Mallet, J., Jiggins, C.D., Divergence with gene flow within the recent chipmunk radiation (Tamias). Heredity 2015. Multilocus species trees show the recent adaptive radiation of the mimetic 113, 185–194. Heliconius butterflies. Syst. Biol. syv007. Swofford, D., 2016. < https://people.sc.fsu.edu/~dswofford/paup_test/ >. Kumar, S., Skjæveland, A., Orr, R., Enger, P., Ruden, T., Mevik, B., Burki, F., Botnen, A., Thompson, A.W., Betancur-R, R., López-Fernández, H., Ortí, G., 2014. A time-calibrated, Shalchian-Tabrizi, K., 2009. AIR: a batch-oriented web program package for con- multi-locus phylogeny of piranhas and pacus (Characiformes: Serrasalmidae) and a struction of supermatrices ready for phylogenomic analyses. BMC Bioinformatics 10, comparison of species tree methods. Molcular Phylogenetics Evol. 81, 242–257. 357. Toews, D.P., Brelsford, A., 2012. The biogeography of mitochondrial and nuclear dis- Lanfear, R., Calcott, B., Ho, S.Y., Guindon, Y., 2012. PartitionFinder: combined selection cordance in . Mol. Ecol. 21 (16), 3907–3930. of partitioning schemes and substitution models for phylogenetic analyses. Mol. Biol. Tonini, J., Moore, A., Stern, D., Shcheglovitova, M., Ortí, G., 2015. Concatenation and Evol. 29, 1695–1701. species tree methods exhibit statistically indistinguishable accuracy under a range of Lee, Y.J., Hill, K.B., 2010. Systematic revision of the genus Psithyristria Stål (Hemiptera: simulated conditions. PLOS Currents Tree of Life. Cicadidae) with seven new species and a molecular phylogeny of the genus and Wallis, G.P., Trewick, S.A., 2009. New Zealand phylogeography: evolution on a small higher taxa. Syst. Entomol. 35 (2), 277–305. continent. Mol. Ecol. 18 (17), 3548–3580. Maddison, W.P., 1997. Gene trees in species trees. Syst. Biol. 46 (3), 523–536. Wallis, G.P., Waters, J.M., Upton, P., Craw, D., 2016. Transverse alpine speciation driven Maddison, W.P., Knowles, L.L., 2006. Inferring phylogeny despite incomplete lineage by glaciation. Trends Ecol. Evol. 12, 916–926. sorting. Syst. Biol. 55, 21–30. Warnow, T., 2015. Concatenation analyses in the presence of incomplete lineage sorting. Marshall, D.C., Slon, K., Cooley, J.R., Hill, K.B.R., Simon, C., 2008. Steady Plio- PLOS Curr. Tree Life. Pleistocene diversification and a 2-million-year sympatry threshold in a New Zealand Williams, P.W., McGlone, M., Neil, H., Zhao, J.-X., 2015. A review of New Zealand pa- cicada radiation. Mol. Phylogenet. Evol. 48, 1054–1066. leoclimate from the Last Interglacial to the global Last Glacial Maximum. Quatern. Marshall, D.C., Hill, K.B.R., Frontaine, K.M., Buckley, T.R., Simon, C., 2009. Glacial re- Sci. Rev. 110, 92–106. fugia in a maritime temperate climate: cicada (Kikihia subalpina) mtDNA phylo- Wood, J, Wilmhurst, J., Newnham, R., McGlone, M., 2017. Evolution and ecological geography in New Zealand. Mol. Ecol. 18, 1995–2009. change during the New Zealand Quaternary. Pp 235–291. In: Landscape and Marshall, D.C., Hill, K.B.R., Cooley, J.R., Simon, C., 2011. Hybridization, mitochondria Quaternary Environmental Change in New Zealand. (J. Shulmeister, Ed.) Advances in DNA phylogeography, and prediction of the early stages of reproductive isolation: Quaternary Science 3. Atlantis Press, Paris. lessons from New Zealand Cicadas (Genus Kikihia). Syst. Biol. 60, 482–502. Zink, R.M., Barrowclough, G.F., 2008. Mitochondrial DNA under siege in avian phylo- Marshall, D.C., Hill, K.B.R., Marske, K.A., Chambers, C., Buckley, T.R., Simon, C., 2012. geography. Mol. Ecol. 17 (9), 2107–2121. Limited, episodic diversification and contrasting phylogeography in a New Zealand Zwickl, D.J., 2006. GARLI—genetic algorithm for rapid likelihood inference. cicada radiation. BMC Evol. Biol. 12 (1), 177. See < http://www.bio.utexas.edu/faculty/antisense/garli/Garli.html >.

181