Welcome back New Zealand: regional biogeography and Gondwanan origin of three endemic genera of mite harvestmen (Arachnida, Opiliones, Cyphophthalmi)

Authors

  • Sarah L. Boyer,

    Corresponding author
      *Sarah L. Boyer, Department of Biology, Macalester College, 1600 Grand Avenue, Saint Paul, MN 55105, USA.
      E-mail: boyer@macalester.edu
    Search for more papers by this author
  • Gonzalo Giribet

    1. Department of Organismic and Evolutionary Biology and Museum of Comparative Zoology, Harvard University, Cambridge, MA, USA
    Search for more papers by this author

*Sarah L. Boyer, Department of Biology, Macalester College, 1600 Grand Avenue, Saint Paul, MN 55105, USA.
E-mail: boyer@macalester.edu

Abstract

Aim  Biogeographers have long been intrigued by New Zealand’s biota due to its unique combination of typical ‘continental’ and ‘island’ characteristics. The New Zealand plateau rifted from the former supercontinent Gondwana c. 80 Ma, and has been isolated from other land masses ever since. Therefore, the flora and fauna of New Zealand include lineages that are Gondwanan in origin, but also include a very large number of endemics. In this study, we analyse the evolutionary relationships of three genera of mite harvestmen (Arachnida, Opiliones, Cyphophthalmi) endemic to New Zealand, both to each other and to their temperate Gondwanan relatives found in Australia, Chile, Sri Lanka and South Africa.

Location  New Zealand (North Island, South Island and Stewart Island).

Methods  A total of 94 specimens of the family Pettalidae in the suborder Cyphophthalmi were studied, representing 31 species and subspecies belonging to three endemic genera from New Zealand (Aoraki, Neopurcellia and Rakaia) plus six other members of the family from Chile, South Africa, Sri Lanka and Australia. The phylogeny of these taxa was constructed using morphological and molecular data from five nuclear and mitochondrial genes (18S rRNA, 28S rRNA, 16S rRNA, cytochrome c oxidase subunit I and histone H3, totalling c. 5 kb), which were analysed using dynamic as well as static homology under a variety of optimality criteria.

Results  The results showed that each of the three New Zealand cyphophthalmid genera is monophyletic, and occupies a distinct geographical region within the archipelago, grossly corresponding to palaeogeographical regions. All three genera of New Zealand mite harvestmen fall within the family Pettalidae with a classic temperate Gondwanan distribution, but they do not render any other genera paraphyletic.

Main conclusions  Our study shows that New Zealand’s three genera of mite harvestmen are unequivocally related to other members of the temperate Gondwanan family Pettalidae. Monophyly of each genus contradicts the idea of recent dispersal to New Zealand. Within New Zealand, striking biogeographical patterns are apparent in this group of short-range endemics, particularly in the South Island. These patterns are interpreted in the light of New Zealand’s turbulent geological history and present-day patterns of forest cover.

‘With regard to general problems of biogeography, the biota of New Zealand has been, perhaps, the most important of any in the world. It has figured prominently in all discussions of austral biogeography, and all notable authorities have felt obliged to explain its history: explain New Zealand and the world falls into place around it.’

(Nelson, 1975, p. 494)

Introduction

New Zealand has intrigued students of biogeography since the days of Ernst Dieffenbach (1811–55), the German geologist and physician who became the first European naturalist to live and work in New Zealand (1839–41). He corresponded extensively with Charles Darwin about a wide range of topics, including the relationship of the avifauna of the Chatham Islands to that of the New Zealand archipelago. To this day, the biogeography of New Zealand’s biota remains a subject of intense interest and debate.

New Zealand’s unique biogeography is a consequence of the land’s unusual geological history (Gibbs, 2006). Because New Zealand originated as a fragment of the supercontinent Gondwana, with rifting occurring some 80 Ma, but has been isolated from major land masses by some 2000 km for the past 60 Myr, its flora and fauna have characteristics typical of both continents and islands (McDowall, 2008). Therefore, the New Zealand archipelago has been viewed as a unique system for studying the evolution of a continental fauna in a state of isolation usually seen only in volcanic islands formed in the mid-ocean (Daugherty et al., 1993; Cowie & Holland, 2006). But is there truly a Gondwanan character to the New Zealand flora and fauna? This topic has been the subject of fierce debate over the past few years, with some maintaining that Gondwanan vicariance has played a major role in establishing the biota of New Zealand (Harvey, 1996; Cooper et al., 2001; Ericson et al., 2002; Stockler et al., 2002) and others insisting that submersion of New Zealand during the Oligocene and/or long-distance dispersal of organisms to the archipelago have erased any Gondwanan signal in extant communities (McGlone, 2005; Waters & Craw, 2006; Knapp et al., 2007; Trewick et al., 2007; Landis et al., 2008).

Like many isolated islands, New Zealand is characterized by a biota that is depauperate in certain higher taxa, with rates of endemism at or near 100% in many groups (Daugherty et al., 1993; Myers et al., 2000; Gibbs, 2006). Radiations may have been driven in part by fluctuations in global sea level during the Cenozoic, as well as the dramatic Oligocene ‘drowning’ or ‘bottleneck’ period when New Zealand’s land area shrank to a tiny fraction of its present size. An often-cited estimate of the smallest land area is 18% (Cooper & Millener, 1993), although recently some have argued that there is no evidence that New Zealand was not entirely submerged during this time (Landis et al., 2008). As land re-emerged during the Miocene and the climate warmed, the diversity of plants and animals increased (Cooper & Millener, 1993), with radiations in such notable groups as ratite birds and giant wetas occurring during this time (Cooper & Cooper, 1995; Trewick & Morgan-Richards, 2005). The Pliocene saw the uplift of most of the major mountains seen in New Zealand today, including the Southern Alps, and the Pleistocene was a period of cycling glacial and interglacial stages. Together, these events drove diversification in many terrestrial invertebrate groups (Trewick et al., 2000; Morgan-Richards et al., 2001; Trewick, 2001; Trewick & Wallis, 2001; Buckley & Simon, 2007). Such diversification should render a clear phylogenetic pattern of reciprocal monophyly of both the New Zealand clades and their sister taxa.

The evolutionary history of most plant and animal groups has been studied with a focus on a single biogeographical process such as vicariance, dispersal or radiation. One group with the potential to provide a window into both Gondwanan vicariance and more recent patterns of diversification is the family Pettalidae, a group of tiny mite harvestmen (daddy long-legs in the suborder Cyphophthalmi) (Fig. 1) with a typical Gondwanan distribution and a remarkably high diversity in New Zealand. The New Zealand archipelago is home to 57% of the described pettalid mite harvestman species, most of which were first documented by local arachnologist Ray Forster (Forster, 1948, 1952; Boyer & Giribet, 2003, 2007). Three genera of these animals are endemic to New Zealand: NeopurcelliaForster, 1948, including only the widespread species Neopurcellia salmoni, found in the western South Island between Lake Kaniere and Lake Te Anau; Rakaia Hirst, 1925, found on Stewart Island, along the south and east coasts of the South Island and throughout Nelson and in the North Island as far north as Lake Waikaremoana; and AorakiBoyer & Giribet, 2007, with a distribution ranging throughout the North Island and in the South Island as far south as Aoraki/Mount Cook in the South Island (Boyer & Giribet, 2007) (Fig. 2).

Figure 1.

 Live photographs of selected South Island specimens belonging to the three New Zealand genera of Cyphophthalmi. The animals are c. 3–4 mm long. (a) Aoraki healyi from Mt Stokes (S 41°05′, E 174°08′). (b) Aoraki longitarsa from Governor’s Bush, Aoraki/Mt Cook N.P. (S 43°44′, E 170°05′). (c) Neopurcellia salmoni from Roaring Billy Falls, Haast River (S 43°56′, E 169°17′). (d) Rakaia pauli from Kelcey’s Bush, Waimate (S 44°42′, E 170°58′). (e) Rakaia stewartiensis from Fern Gully, Oban, Stewart Island (S 46°53′, E 168°06′). All photographs by G. Giribet.

Figure 2.

 Distributions of the genera Aoraki (white with thin borders), Neopurcellia (white with thick borders), and Rakaia (black) in New Zealand. The map shows the collecting localities of specimens from the current study (circles) as well as museum specimens (squares) assigned to genera using morphological diagnoses established by Boyer & Giribet (2007).

Most species of Pettalidae are known from only a handful of localities, despite extensive collecting by Forster and others, with ranges typically of the order of 100 km or less in diameter. This makes these animals short-range endemics as defined by Harvey (2002). In addition, no species in the entire suborder Cyphophthalmi are known from oceanic islands formed de novo by volcanoes, such as Hawaii or the Society Islands (Giribet, 2000). These slow-moving animals spend their entire life cycle in forest leaf litter and retreat underground during periods of drought or cold. In New Zealand, they have been collected from localities close to sea level through elevations above 1100 m. Although they are not found in grasslands and shrublands, the authors have collected them from tiny patches of forest habitat surrounded by agricultural fields. Therefore, these animals are excellent models for studying historical biogeography in general, and vicariance in particular.

In this paper, we present a phylogeny of the family Pettalidae including 94 specimens, of which 54 are individuals representing 25 of the 29 known species and subspecies of New Zealand’s mite harvestmen, plus six undescribed species – totalling 31 New Zealand taxa. The questions we addressed include: (1) How are the New Zealand genera related to other members of the family Pettalidae from across Temperate Gondwana, and what can be inferred about dispersal vs. vicariance within the family? (2) Are the New Zealand Cyphophthalmi a monophyletic group, or do they constitute multiple lineages? (3) How can the regional biogeographical patterns seen within the New Zealand Cyphophthalmi be related to the geological history of New Zealand?

Materials and methods

Taxon sampling

A total of 94 specimens from the family Pettalidae (suborder Cyphophthalmi) were included in this study. These included 54 individuals representing 31 species and subspecies of three endemic genera (Aoraki, Neopurcellia and Rakaia) from New Zealand, plus other members of the family from Chile (genus Chileogovea), South Africa (genera Purcellia and Parapurcellia), Sri Lanka (Pettalus), and Australia (Austropurcellia and Karripurcellia). Members of the families Sironidae, Stylocellidae, Neogoveidae and Troglosironidae are included as outgroups (see Appendix S1 in Supporting Information). The majority of the material was collected alive by the authors and fixed in 96% EtOH, with a few specimens collected by the New Zealand Department of Conservation (NZ DOC) using pitfall traps. All specimens are deposited as vouchers at the Harvard Museum of Comparative Zoology, Department of Invertebrate Zoology DNA collection, with the exception of the four specimens collected by the NZ DOC, which are deposited at the Museum of New Zealand Te Papa Tongarewa (for voucher numbers see Appendix S1).

Morphological data set

A total of 45 morphological characters were scored for each species used in the molecular data set. These characters were defined in our 2007 study of the phylogeny of Pettalidae and references therein (Giribet & Boyer, 2002; Giribet, 2003a; de Bivort & Giribet, 2004; Boyer & Giribet, 2007) (Appendix S2). When appropriate specimens were available, they were dissected and mounted for examination using an FEI Quanta 200 scanning electron microscope (SEM). Micrographs are available in the online public database MorphoBank at http://morphobank.geongrid.org. Characters were also coded from specimens examined using light microscopy, including holotype material where available. The Harvard Museum of Comparative Zoology online database of photographs of Cyphophthalmi types is available at http://giribet.oeb.harvard.edu/Cyphophthalmi.

DNA extraction

Total DNA was extracted from whole animals using Qiagen’s DNeasy tissue kit (Qiagen, Valencia, CA, USA), either by crushing the individual or one appendage in the lysis buffer, or by incubating an intact animal or appendage in lysis buffer overnight, then removing the specimen before proceeding with the rest of the manufacturer’s extraction protocol, as described by Boyer et al. (2005).

PCR and sequencing

Purified genomic DNA was used as a template for polymerase chain reaction (PCR) amplification of the genes for 18S rRNA, 28S rRNA, 16S rRNA, cytochrome c oxidase subunit I (COI) and histone H3. The complete 18S rRNA (c. 1.8 kb) was amplified in three overlapping fragments of c. 900 bp each, using primer pairs 1F–5R, 3F–18Sbi and 18Sa2.0–9R (Giribet et al., 1996; Whiting et al., 1997). An additional primer internal to 1F–5R was used for sequencing, 4R (Giribet et al., 1996). The first c. 2200 bp of 28S rRNA were amplified using the primer sets 28SD1F/28Srd1a-28Sb (Whiting et al., 1997; Park & Ó Foighil, 2000; Edgecombe & Giribet, 2006), 28Sa–28Srd5b (Whiting et al., 1997; Schwendinger & Giribet, 2005) and 28S4.8a–28S7bi (Schwendinger & Giribet, 2005). Sequencing of the 28S rRNA gene was performed with those primers and some additional internal primers: 28Sa (Whiting et al., 1997) and 28Srd4b (Edgecombe & Giribet, 2006). 16S rRNA was amplified and sequenced using the primer pair 16Sa–16Sb (Xiong & Kocher, 1991; Edgecombe et al., 2002). COI was amplified and sequenced using the primer pair LCO1490–HCO2198 (Folmer et al., 1994). The complete coding region of histone H3 was amplified and sequenced using primer pair H3aF–H3aR (Colgan et al., 1998). A complete table of primer sequences can be found in Boyer & Giribet (2007).

PCR reactions (50 μL) included 4 μL of template DNA, 1 μm of each primer, 200 μm of dNTP (Invitrogen, Carlsbad, CA, USA), 1× PCR buffer containing 1.5 mm MgCl2 (Perkin-Elmer, Waltham, MA, USA) and 1.25 units of AmpliTaq DNA polymerase (Perkin-Elmer). The PCR reactions were carried out using a GeneAmp PCR System 9700 thermal cycler (Applied Biosystems, Carlsbad, CA, USA), and involved an initial denaturation step (5 min at 95°C) following by 35 cycles including denaturation at 95°C for 30 s, annealing (ranging from 42 to 49°C) for 30 s, and extension at 72°C for 1 min, with a final extension step at 72°C for 10 min.

The double-stranded PCR products were visualized by agarose gel electrophoresis (1% agarose), and purified using Qiagen QIAQuick spin columns. The purified PCR products were sequenced directly; each sequence reaction contained a total volume of 10 μL including 2 μL of the PCR product, irrespective of PCR yield, 1 μm of one of the PCR primer pairs, 1 μL of ABI BigDye 5× sequencing buffer and 0.5 μL of Applied Biosystems big dye terminator ver. 3.0. The sequence reactions, performed using the thermal cycler described above, involved an initial denaturation step for 3 min at 95°C, and 25 cycles (95°C for 10 s, 50°C for 5 s, 60°C for 4 min). The BigDye-labelled PCR products were cleaned with AGTC Gel Filtration Cartridges or Plates (Edge Biosystem, Gaithersburg, MD, USA). The sequence reaction products were then analysed using an ABI Prism 3100 or 3730 Genetic Analyzer (Applied Biosystems).

Sequence editing

Chromatograms obtained from the automatic sequencer were read, and both strands for each overlapping fragment were assembled using the sequence-editing software sequencher 4.0 (Gene Codes Corporation, Ann Arbor, MI, USA). Sequence data were edited in macgde 2.2 (Linton, 2005). All new sequences were deposited in GenBank (Appendix S1).

Phylogenetic analysis

Direct optimization (dynamic homology)

Sequences from ribosomal genes were compared against secondary structure models, and then split into accordant fragments using internal primers and some visualized secondary structure features (Giribet, 2002). COI, which showed length variation, was split into three fragments using areas where the amino acid sequences allowed us to do so unambiguously. The protein-coding gene histone H3 showed no length variation and was treated as pre-aligned.

Data from each gene were analysed in poy ver. 3.0 (Wheeler et al., 2004) using the direct optimization method with parsimony as the optimality criterion (Wheeler, 1996). The direct optimization method allows analysis of sequences of unequal length without a predetermined static alignment. poy uses a dynamic optimization process that generates phylogenetic trees by searching for topologies that minimize total transformation cost under specified parameters (e.g. indels, transversions and transitions). Rather than performing sequence alignment and tree construction under different criteria and/or models, the same criterion and model are used consistently throughout the entire phylogenetic analysis.

The data for all genes were analysed simultaneously. In addition, the data for each gene were analysed independently, and 18S and 28S rRNA were combined in an analysis of nuclear ribosomal genes. Tree searches were conducted on a cluster of 30 dual-processor nodes assembled at Harvard University (darwin.oeb.harvard.edu). Commands for load balancing of spawned jobs were in effect to optimize parallelization procedures (–parallel –dpm –dpmacceptratio 1.5 –jobspernode 2). Trees were built through a random addition sequence procedure followed by a combination of branch-swapping steps (subtree pruning and regrafting, and tree bisection and reconnection).

poy facilitates efficient sensitivity analysis (sensuWheeler, 1995). All data sets (individual genes and combinations) were analysed under nine parameter sets for a range of indel-to-transversion ratios and transversion-to-transition ratios (111, 121, 141, 211, 221, 241, 411, 421, 441). A table describing these parameter sets can be found in Boyer & Giribet (2007). Implied alignments (Wheeler, 2003) can easily be generated for each tree (Giribet, 2005).

To identify the parameter set or ‘model’ that maximized congruence among all the data (what we call the ‘optimal’ parameter set), we employed a character-congruence technique which is a modification of the incongruence length difference (ILD) metric developed by Mickevich & Farris (1981). The value is calculated for each parameter set by subtracting the sum of the scores of all partitions from the score of the combined analysis of all partitions, and normalizing it for the score of the combined length. This has been interpreted as a meta-optimality criterion for choosing the parameter set that best explains all partitions in combination, maximizing overall congruence and minimizing character conflict among all the data. This parameter set was given special consideration in the analysis of data from each individual gene, and is referred to throughout the paper as the ‘optimal parameter set’. Although Wheeler (1995) proposed the use of such a measure as an option for choosing between multiple hypotheses, such a measure has been modified subsequently (e.g. Wheeler & Hayashi, 1998). The ILD has been compared with other measures by Aagesen et al. (2005), and criticized for being partition-dependent (Wheeler, 1996). However, we use the ILD test as a rough way to choose a tree, yet still consider the remaining parameter sets by evaluating the alternative trees obtained under the different parameter sets and by presenting the strict consensus of all trees obtained under the different analytical parameter sets (interpretive functions of Wheeler, 2006), as discussed by Giribet (2003b).

We performed an analysis of the morphological data set alone, as well as a combined analysis of molecular and morphological data, in which all trees from across parameter sets used in the analysis of molecular data were used as the basis for tree fusing under each parameter set, with morphological character transitions weighted identically with the maximum weight allowed for any molecular transition. As in the combined molecular analysis, the ILD was calculated and used to identify an optimal parameter set. Nodal support for all topologies was measured using parsimony jackknifing (Farris et al., 1996) with 100 replicates.

Multiple sequence alignments (static homology)

Even though direct optimization approaches such as the one employed here are compelling philosophically (De Laet, 2005; Fleissner et al., 2005; Redelings & Suchard, 2005; Wheeler, 2006; Wheeler & Giribet, 2009), they are not wanting for detractors (e.g. Kjer, 2004; Simmons, 2004; Kjer et al., 2007; Ogden & Rosenberg, 2007). In order to evaluate further the hypotheses derived under direct optimization, we performed a more traditional two-step phylogenetic analysis of the data by submitting the sequence loci with length variation to a static alignment procedure. We generated manual alignments for 18S rRNA, 28S rRNA and COI, excluding areas that were difficult to align visually. For ribosomal data, initial alignments were generated by splitting the sequences into fragments as described for direct optimization, and then modifying the alignments by eye. Alignment of COI was based on translation of the DNA sequences. For the static alignment of each data set, as well as for the entire concatenated data set, we determined the best-fit model using Modeltest 3.7 (Posada & Crandall, 1998) under the Akaike information criterion (AIC), as recommended by Posada & Buckley (2004). The model chosen was GTR + I + G in each case, except for the 18S rRNA data partition, for which TrN + I + G was favoured. We then analysed the data under the recommended model(s) using mrbayes 3.1.1 (Huelsenbeck & Ronquist, 2001), running sufficient generations such that the average deviation of split frequencies reached < 0.01. In each case, four chains were employed, and the number of generations was either 5,000,000 (18S and 28S rRNA data sets) or 2,000,000 (COI and H3 data sets). The combined (concatenated) data set was run for 15,000,000 generations, and failed to achieve an average deviation of split frequencies below 0.154. For the combined data set, each partition was allowed a separate model. In each case, burn-in prior to levelling of –log L was discarded. We also analysed the combined data for 18S rRNA, 28S rRNA, COI and histone H3 using paup* portable version 4.0b10 for Unix on the Harvard University Center for Genomic Research supercomputing cluster (portal.cgr.harvard.edu) under the single model chosen, using the AIC criterion as implemented in Modeltest 3.7, GTR + I + G. We calculated bootstrap support with 100 replicates.

Results

In our analyses we tested the monophyly of each New Zealand genus of Cyphophthalmi: Aoraki, Neopurcellia and Rakaia (Fig. 3). In addition, several subgeneric clades were retrieved in all or most analyses. They are described here to facilitate tree description in this section.

Figure 3.

 Phylogeny of the family Pettalidae. This tree represents the strict consensus of 50 most parsimonious trees of 14,701 weighted steps obtained in direct optimization analysis of all molecular data under the optimal parameter set (411). Numbers on branches represent jackknife frequencies, with stars indicating 100% support. The three lineages from New Zealand are highlighted. Maps correspond to the distribution of each New Zealand genus.

Within Aoraki, two clades were retrieved consistently and will be referred to hereafter as Aoraki clade a and Aoraki clade b. ‘Aoraki clade a’ includes animals from Nelson and North Island that all have a longitudinal carina on the anal plate: Aoraki calcarobtusa westlandica, A. granulosa and A. tumidata, as well as animals from throughout the north-west South Island: Aoraki denticulata denticulata, A. denticulata major and A. longitarsa (Figs 3 & 4). ‘Aoraki clade b’ includes species from the Marlborough Sounds and North Island: Aoraki crypta, A. healyi, A. inerma and Aoraki sp. Mount Stokes.

Figure 4.

 Geographical distributions of clades within the genus Aoraki. The tree corresponds to the Aoraki clade from Fig. 3. Black circles indicate localities represented by specimens included in the present study; white circles represent additional known localities for the genus based on museum specimens. The letter beside each clade corresponds to the map illustrating the distribution of that clade.

Within Rakaia, four clades were retrieved consistently and will be referred to hereafter as Rakaia clades a, b, c and d. ‘Rakaia clade a’ is made up of animals with male tarsus IV bisegmented or bilobed: R. florensis, R. lindsayi, R. minutissima and R. stewartiensis. ‘Rakaia clade b’ is found throughout Otago and Southland and includes Rakaia sorenseni sorenseni, R. sorenseni digitata and Rakaia sp. Beaumont Forest. ‘Rakaia clade c’ includes animals from Otago and Canterbury, mostly found along the east coast: Rakaia antipodiana, R. macra, R. pauli and Rakaia sp. Hinewai. ‘Rakaia clade d’ includes animals from the Wellington area: Rakaia dorothea, R. magna australis, R. media, R. cf. uniloca, R. solitaria, Rakaia sp. Akatarawa, Rakaia sp. Kapiti Island and Rakaia sp. Wi Toko (Figs 3 & 5).

Figure 5.

 Geographical distributions of clades within the genus Rakaia. The tree corresponds to the Rakaia clade from Fig. 3. Black circles indicate localities represented by specimens included in the present study; white circles represent additional known localities for the genus based on museum specimens. The letter beside each clade corresponds to the map illustrating the distribution of that clade.

Molecular data

When all molecular data were combined, a well resolved tree was retrieved under the optimal parameter set (411) (Table 1; Fig. 3). Aoraki, Neopurcellia and Rakaia were all found to be monophyletic, and there was not support for sister-group relationships among pairs of these three genera. Monophyly of each genus was stable to parameter variation, appearing in the strict consensus of all shortest trees from analysis under every parameter set. Each genus also formed a clade in the Bayesian analyses, with posterior probability > 95% and was well supported in the likelihood analyses (Fig. 6).

Table 1.   Tree length and calculated incongruence length difference (ILD) values for the partitions, including total combined molecular data (mol), 18S rRNA (18S), 28S rRNA (28S), 16S rRNA (16S), cytochrome c oxidase subunit I (COI), histone H3 (H3), combined molecular + morphological data (comb) and morphology alone (morph), at different parameter set values corresponding to those described by Boyer & Giribet (2007). Optimal parameter sets are indicated in bold.
 mol18S28S16SCOIH3combmorphILD molILD comb
  1. ILD-mol, ILD when molecular data are analysed without morphology; ILD-comb, ILD for combined molecular and morphological data.

111221984484352606897281194223171000.01820.0191
12117655323359251867402799180362000.02070.0296
1413031451363189374123121187308064000.01980.0228
21112436230301134864893598126852000.01740.021
22120211331520660277451799206992000.01950.03331
24135371529951811024124091187358624000.01970.0222
41114701238465040804919597151924000.01460.0203
42124659347827172007510799251574000.02140.025
441442345611561313322125301187447314000.02290.025
Figure 6.

 Phylogeny of the family Pettalidae. This tree is based on maximum likelihood analysis of combined data from 18S rRNA, 28S rRNA, histone H3 and COI data sets. The overall likelihood score for this tree was 48816.98077. Numbers on branches represent bootstrap support.

Within Aoraki, two lineages were found –Aoraki clade a and Aoraki clade b – within which the Marlborough Sounds species form a paraphyletic grade with respect to the North Island species (Fig. 4). Within Rakaia, a nested pattern was seen: (Rakaia clade a (Rakaia clade b (Rakaia clade c + Rakaia clade d))) (Fig. 5).

Results from analyses of individual molecular data partitions are not discussed here, and can be found in Boyer (2007).

Morphological data

Morphological data alone resolved the family Pettalidae as monophyletic and retrieved the monotypic genus Neopurcellia but not Rakaia, Aoraki, or any of the subgeneric clades of New Zealand species found in analyses of molecular data. Outside of the New Zealand taxa, the pettalid genera Pettalus and Chileogovea were both retrieved –Purcellia was represented by a single species. Parapurcellia appears as a paraphyletic grade sister to all other Pettalidae. Aoraki clade b appears as a paraphyletic grade sister to Pettalus (Sri Lanka) + Chileogovea (Chile).

Combined molecular and morphological data

The optimal parameter set for combined molecular and morphological data was 111 (Table 1). The strict consensus under this parameter set was almost fully resolved, with a monophyletic Pettalidae, Aoraki, Neopurcellia and Rakaia (Fig. 7). Within both Aoraki and Rakaia, the subgeneric clades described in the ‘Molecular data’ section were retrieved once again. Monophyly of each pettalid genus and of most New Zealand subgeneric clades received high jackknife support, and relationships among genera received < 50% jackknife support in every case.

Figure 7.

 Phylogeny of the family Pettalidae. This tree represents the strict consensus of 30 most parsimonious trees of 22,317 weighted steps obtained in direct optimization analysis of molecular plus morphological data under the optimal parameter set (111). Numbers on branches represent jackknife frequencies; stars indicate 100% support. The three lineages from New Zealand are highlighted.

Discussion

Geological context

New Zealand separated from Gondwana c. 80 Ma. Throughout the existence of the supercontinent, the land that today forms New Zealand was adjacent to land that today forms Australia and Antarctica (Sanmartín, 2002). During the Oligocene drowning event c. 27 Ma, the land area of the North and South Islands was reduced and then expanded again due to changing sea levels, dwindling to a tiny fraction of its present-day size (Stevens, 1980; Cooper & Cooper, 1995; Trewick & Morgan-Richards, 2005), although some maintain that New Zealand could have been submerged entirely (Waters & Craw, 2006; Landis et al., 2008; McDowall, 2008). Over the past 20–25 Myr, the Australian and Pacific tectonic plates have slipped c. 500 km relative to one another, with the Australian plate moving north. The Rangitata Orogeny during the Cretaceous resulted in large mountains that eroded and subsided through the Palaeogene. New Zealand featured only limited topographic relief until the Pliocene, when mountain-building began in the South Island (5–2 Ma), during which time all of the major axial mountain ranges were formed.

The development of the Southern Alps provided conditions for the formation of extensive glaciers, which reached their maximum during the Pleistocene, c. 18,000 years ago (Stevens, 1980; Suggate, 1990; Trewick, 2001; Winkworth et al., 2005). During the Last Glacial Maximum (LGM), ice sheets covered much of the Southern Alps and extended to the lowlands of the central west coast of the South Island. New Zealand’s small size and lack of large ice sheets at the glacial maximum allowed vegetation to react quickly to climatic change, with a minimum of periglacial effects from decaying ice sheets. Although Pleistocene pollen records suggest that forest was uncommon in the South Island during the LGM, new data from fossil beetles challenge this view. The presence of certain beetle species indicates that forest habitats have been found in LGM strata at localities previously thought to have been dominated by grasslands during the LGM (Marra, 2008). A major increase in arboreal pollen at localities across the South Island occurred starting 10,500 years ago, and over the next 1000 years a large proportion of the land mass became covered with tall podocarp-hardwood forest (McGlone et al., 1993). Today, New Zealand’s forests are composed of four broad floristic components: podocarps, hardwoods, Nothofagus (the southern beech) and Agathis australis (kauri). New Zealand’s Cyphophthalmi are known from all of these forest types, although a survey of museum collection records and the authors’ own field work experiences indicate that where these animals are found, they are most abundant in Nothofagus forests and least abundant in the kauri forests of the northernmost peninsula of the North Island.

Marine inundation, orogeny and other tectonic activity, and glaciation are all phenomena that would have contributed towards the fragmentation of the forests of New Zealand. The biogeography of New Zealand’s Cyphophthalmi fauna would certainly have felt significant impacts from all of these processes.

Overview: biogeographical relationships of New Zealand Cyphophthalmi

The three New Zealand genera of Cyphophthalmi are unequivocally members of the temperate Gondwanan family Pettalidae, confirming previous work (Boyer & Giribet, 2007; Boyer et al., 2007b). Aoraki, Neopurcellia and Rakaia are not each other’s closest relatives (Figs 2, 5–7). Therefore, we conclude that the present-day diversity of Cyphophthalmi found in New Zealand does not represent a single radiation, but in fact three separate evolutionary lineages.

The lack of resolution among pettalid genera from different land masses prevents the use of a molecular clock in assigning dates to nodes on this particular phylogeny. In the absence of divergence estimates, what can these results tell us about dispersal vs. vicariance within the Gondwanan family Pettalidae? In a previous phylogenetic study of Cyphophthalmi, the age of the family Pettalidae was estimated to be between 178 and 215 Ma, with its origin in the Jurassic/Triassic (Boyer et al., 2007b). As New Zealand was the last major land mass to separate from temperate Gondwana, it is unlikely that the New Zealand genera have originated in the past 25 Ma, while all the other genera have been in place for more than 100 Ma – and that none are paraphyletic – as would be the case under a post-Oligocene drowning dispersal scenario.

If New Zealand has indeed been submerged, and the terrestrial biota of the archipelago has arrived solely through transoceanic dispersal, we would expect to see close relationships between the pettalid fauna of New Zealand and species from other nearby land masses. In many groups of plants, New Zealand representatives have clearly dispersed from Australia (Gibbs, 2006; Sanmartín, 2007). In some of our analyses, the monophyletic genus Austropurcellia from Queensland appears as the sister taxon of Rakaia, but Rakaia never appears as a lineage nested within a paraphyletic Austropurcellia, as one would expect under a scenario of dispersal from eastern Australia. Some elements of the New Zealand biota, such as parakeets, have been postulated to have arrived from New Caledonia (Gibbs, 2006). In the case of mite harvestmen, dispersal to and from New Caledonia is strongly rejected, because the mite harvestmen of New Caledonia belong to a different family endemic to that island, Troglosironidae, which is in turn closely related to the Tropical Gondwanan family Neogoveidae from tropical West Africa and tropical South America, and not to Pettalidae (Boyer et al., 2007b). Although dispersal has certainly played a role in establishing some elements of the New Zealand biota, and although it is not possible to unequivocally reject dispersal in the case of mite harvestmen, our data do not support the evolutionary relationships expected if we assume dispersal as an explanation for the presence of the family Pettalidae in the archipelago.

Within New Zealand, the geographical ranges of the three genera of Cyphophthalmi can be described as follows. The genus Neopurcellia (sensuBoyer & Giribet, 2007) comprises a single widespread species, although we do not discard the possibility of this constituting a series of cryptic species (S.L.B., unpublished data) – found throughout the west coast of the South Island as far north as Lake Kaniere, and as far south as Te Anau (Fig. 2). The genus Rakaia (sensuBoyer & Giribet, 2007) is found on Stewart Island, along the south and east coasts of the South Island and throughout Nelson, and in the North Island as far north as Lake Waikaremoana (Fig. 2). Rakaia does not overlap in distribution with Neopurcellia, but certain lineages of Rakaia do overlap with Aoraki in the Nelson/Marlborough area and in the North Island. Finally, the genus Aoraki is found throughout the North Island as well as in the South Island as far south as Aoraki/Mt Cook. It is not found on Stewart Island, or anywhere along the south and east coasts of the South Island, where Rakaia is highly diverse (Fig. 2).

Taken together, the distributions of these genera indicate two major trends. Firstly, the Southern Alps form a barrier not readily crossed by mite harvestmen. Only Neopurcellia inhabits the very wet forests to the west of this mountain range, except in the most northern reaches of the Southern Alps, where certain members of both Aoraki and Rakaia are found. Secondly, Aoraki and certain lineages of Rakaia have diversified within an overlapping range, including the North Island plus the northern part of the South Island. Close relationships between North and South Island clades have also been observed in other groups of animals, such as brown kiwis and cicadas (Baker et al., 1995; Buckley et al., 2001). These two land masses were connected by a large region of grassland and shrubland during the height of the Pleistocene glaciation only 20,000 years ago (McGlone et al., 1993).

The South Island: west coast, Southland, Otago and Canterbury

The two most widespread species of Cyphophthalmi in New Zealand are Aoraki denticulata (Boyer et al., 2007a) and N. salmoni. Each of these species occurs in the extensive Nothofagus and podocarp forests of the western South Island, where rainfall is extremely high. These are the largest ranges of continuous forest in the South Island, and such large ranges of these two species no doubt reflect the extensive suitable habitat in which they live.

The Southern Alps mountain range has created a barrier that divides the high-rainfall western part of the South Island from the dry eastern part. Forests to the east of the Southern Alps are much patchier and drier than those to the west, possibly driving allopatric speciation in this area. There are three closely related species in the Rakaia clade b with adjacent but non-overlapping distributional ranges in the south coast of the South Island, and four closely related species from Rakaia clade c with adjacent but (again) non-overlapping ranges in the central eastern part of the South Island (Fig. 5). In contrast to the large, continuous forests in the north and west parts of the South Island, which are home to widespread species, the patchy, discontinuous forests along the east coast of the South Island are home to species with very restricted ranges.

Nelson, Marlborough and the North Island

There is a close relationship between Cyphophthalmi from Nelson/Marlborough and the North Island in both Aoraki and Rakaia – with several localities home to pairs of species from the two genera. A. granulosa and R. media, A. inerma and R. media, A. denticulata and R. magna australis, A. denticulata and R. florensis, and A. denticulata and R. minutissima are all pairs of taxa which we have found living sympatrically at single collecting sites. Although the Cook Strait today constitutes a formidable barrier for mite harvestmen to cross, only 20,000 years ago, at the height of the glacial maximum, the Nelson/Marlborough area was connected to the North Island by an area of grassland and shrubland. During this time, scattered forests were present throughout the North Island as well as in the northernmost parts of Nelson and Marlborough. It is therefore plausible that Cyphophthalmi persisted across this area during the Pleistocene.

Stewart Island

Within Rakaia, the two species from Stewart Island and two species from Nelson + North Island form a clade (Rakaia clade a), which is sister to all other species in the genus. This relationship among the Rakaia clade a species is one of the most stable and well supported clades within the New Zealand Cyphophthalmi; it is found in the strict consensus of all the shortest trees under all parameter sets for the combined molecular data set, as well as in the strict consensus of all shortest trees under all parameter sets for the 18S rRNA, 28S rRNA and histone H3 data partitions. Bayesian analyses of every data partition retrieved this clade with > 95% posterior probability, and this relationship was also found in the optimal trees of every likelihood analysis and in our previous study of the family Pettalidae incorporating molecular data (Boyer & Giribet, 2007).

From a morphological point of view, this relationship is not surprising. Two of these species (R. florensis and R. minutissima) have a fully bisegmented male tarsus IV, a condition which is probably plesiomorphic in Pettalidae, as Parapurcellia and Neopurcellia share this character. The two species sister to R. florensis and R. minutissima have male tarsi IV, which display partial bisegmentation apparent in both light microscope and SEM images (see fig. 7d–i in Boyer & Giribet, 2007). Rakaia minutissima and R. stewartiensis also share an anal plate with a highly concave posterior margin not seen in any other New Zealand Cyphophthalmi (see figs 8 & 9 in Boyer & Giribet, 2007).

Although a close relationship between animals from the southernmost South Island and the northern South Island + southern North Island may seem counterintuitive, similar patterns of disjunction are common across many groups of New Zealand plants and animals (Heads, 1998; Trewick & Wallis, 2001; Heads & Craw, 2004; Haase et al., 2007), and two alternative hypotheses have been proposed to explain this pattern. One possibility is that this disjunction reflects the relicts of groups that were once more widespread. During the LGM, ice covered much of the south-west coast of the South Island, comprising the area that would have linked Stewart Island to the regions currently inhabited by R. florensis and R. minutissima. This could have extirpated Cyphophthalmi inhabiting the area, leaving intact only the Stewart Island and Nelson + North Island members of what was once a much more widespread clade. Another possible explanation for South Island disjunctions is displacement along the Alpine Fault, a hypothesis that is supported in a study of hydrobiid gastropods (Haase et al., 2007) but not across larger and more mobile terrestrial arthropods (Trewick, 2001).

Conclusions

The mite harvestmen (Cyphophthalmi) fauna of New Zealand includes three evolutionary lineages, which clearly are members of the temperate Gondwanan family Pettalidae, and which are not closely related to each other. The relationships of the New Zealand genera to relatives in Australia and other adjacent land masses, such as New Caledonia, provide no evidence for recent dispersal as the explanation for the presence of pettalids in New Zealand. In the southern regions of the South Island, Neopurcellia occurs to the west of the Southern Alps, while Rakaia occurs to the east of these mountains. In the North Island, the genus Aoraki predominates. The latter two genera overlap in the southern part of the North Island and in the northern region of the South Island, reflecting the recent historical connection of these two major islands. Despite the lack of precise dating, the history of Cyphophthalmi in New Zealand has clearly been influenced by ancient Gondwanan vicariance, as well as more recent alterations to the New Zealand landscape during the Pliocene uplift of the Southern Alps and Pleistocene glaciation across the South Island. Future comparative studies of phylogeographical relationships within species from disparate geographical regions of New Zealand, such as the South Island’s west coast and east coast, will increase our understanding of the role that habitat fragmentation has played in the diversification of this remarkable group of short-range endemics.

Acknowledgements

For help with field work, the authors acknowledge Greg Edgecombe, Cyrille D’Haese, Jessica Baker, Phil Sirvid, Ricardo Palma and Cor Vink. All permits were arranged with the help of the sublime New Zealand Department of Conservation; special thanks go to Ian Millar at the Nelson office. Phil Sirvid and Ricardo Palma at Te Papa Tongarewa/Museum of New Zealand (Wellington) loaned holotype materials as well as other specimens that provided valuable data regarding distributions of species; some of these materials included the Department of Conservation’s pitfall trap specimens for Aoraki inerma, A. tumidata and Rakaia antipodiana. Additional specimens were loaned by Simon Pollard from the Canterbury Museum (Christchurch), Abigail Blair from the Otago Museum (Dunedin), and Grace Hall at the New Zealand Arthropod Collection of Manaaki Whenua Landcare Research (Auckland). The authors would like to thank Maureen O’Leary for the invitation to participate in the MorpoBank project; Jennifer Lee assisted with the uploading of images to the database. Emily Sabo assisted with the production of Fig. 2. This material is based on work supported by the National Science Foundation under Grants Nos 0236871 and DEB-0508789. Field work for this project was also generously supported by a Museum of Comparative Zoology Putnam Expedition Grant.

Biosketches

Sarah Boyer earned her PhD in Biology from Harvard University and is Assistant Professor of Biology at Macalester College. She is interested in the phylogeny, biogeography and phylogeography of terrestrial and freshwater invertebrates. A particular focus of her research is the fauna of New Zealand.

Gonzalo Giribet is Professor of Organismic and Evolutionary Biology and Curator of Invertebrates at the Museum of Comparative Zoology, Harvard University. He is interested in the origins and maintenance of invertebrate diversity, in both marine and terrestrial environments, and in theoretical aspects of systematics and biogeography.

Editor: Malte Ebach

Ancillary