Phylogeography of the prehensile-tailed skink Corucia zebrata on the Solomon Archipelago

The biogeography of islands is often strongly influenced by prior geological events. Corucia zebrata (Squamata: Scincidae) is endemic to the geologically complex Solomon Archipelago in Northern Melanesia. We examined the level of divergence for different island populations of C. zebrata and discussed these patterns in light of Pleistocene land bridges, island isolation, and island age. Corucia zebrata was sampled from 14 locations across the Solomon Archipelago and sequenced at two mitochondrial genes (ND2 and ND4; 1697 bp in total) and four nuclear loci (rhodopsin, an unknown intron, AKAP9, and PTPN12). Measures of genetic divergence, analyses of genetic variation, and Bayesian phylogenetic inference were used and the data assessed in light of geological information. Populations of C. zebrata on separate islands were found to be genetically different from each other, with reciprocal monophyly on mitochondrial DNA. Populations on islands previously connected by Pleistocene land bridges were marginally less divergent from each other than from populations on other nearby but isolated islands. There are indications that C. zebrata has radiated across the eastern islands of the archipelago within the last 1–4 million years. Nuclear loci were not sufficiently informative to yield further information about the phylogeography of C. zebrata on the Solomon Archipelago. Analyses of the mitochondrial data suggest that dispersal between islands has been very limited and that there are barriers to gene flow within the major islands. Islands that have been isolated during the Pleistocene glacial cycles are somewhat divergent in their mitochondrial genotypes, however, isolation by distance (IBD) and recent colonization of isolated but geologically younger islands appear to have had stronger effects on the phylogeography of C. zebrata than the Pleistocene glacial cycles. This contrasts with patterns reported for avian taxa, and highlights the fact that biogeographic regions for island species cannot be directly extrapolated among taxa of differing dispersal ability.


Introduction
Islands represent natural laboratories that allow for simpler examination of evolution and ecology than is possible on large continental landmasses. Long periods of geographic isolation, with subsequent genetic drift and differing selection pressures lead to genetic divergence (Balloux and Lugon-Moulin 2002;Vanderwerf et al. 2010) that may result in allopatric speciation on separate islands (Orr and Smith 1998;Glor et al. 2004). The geological histories of archipela-gos play an important role in shaping the biogeography of islands. Island colonization often moves in the direction from older to younger islands, as seen on Hawaii (Hormiga et al. 2003;Rubinoff 2008) and the Galapagos (Parent et al. 2008;Benavides et al. 2009). The last glacial maximum (LGM) 18,000 years ago, when the sea level was 120 m or more lower than today (Fairbanks 1989) has also affected the biogeography of many islands. The lowered sea level exposed land bridges and eliminated or decreased the over-water distance between landmasses, which facilitated dispersal between islands. Consequently, current populations of species on islands that were connected by Pleistocene land bridges often show decreased genetic distance compared to those on islands that have remained isolated, and this pattern can be detected for both volant and nonvolant species (Heaney et al. 2005).
The Melanesian Solomon Archipelago is situated in the South West Pacific and spans two countries, Papua New Guinea (PNG), which has political claim on the island of Bougainville, and the Solomon Islands, to which the remaining islands belong (Fig. 1). The islands have complex geological histories that create excellent conditions for assessment of biogeography, ecology, and evolution. The islands of the Solomon Archipelago have never been connected to neighboring continents, but emerged from the ocean as a consequence of a collision between the Indian and Pacific Plates (Hall 2002). All of the endemic biota on the archipelago has therefore originated from speciation through founder events rather than vicariance. The various islands of the archipelago have different geological compositions, origins, and emergence times. The Western Province is the youngest part of the archipelago and is a result of volcanic activity during the Pliocene and Pleistocene (Petterson et al. 1999;Cowley et al. 2004). The islands of the Western Province are therefore likely to have emerged less than 5 million years ago (MYA) (Petterson et al. 1999). The earlier emergence times of the remaining islands are largely uncertain, but are estimated to range from 30 to 90 MYA (Petterson et al. 1999;Hall 2002). Several islands of the archipelago were connected to each other during the LGM (Fig. 1). Bougainville, Choiseul, Isabel, and Ngela were joined in the Great Bukida landmass and Guadalcanal would have been either connected to the same landmass or separated from it by a narrow channel. At the same time, the Western Province was represented probably by two major islands Diamond 1976, 2001). Makira and Malaita are separated from each other and from the rest of the archipelago by deeper channels and probably have been isolated for a substantial geological time (Mayr and Diamond 2001; Fig. 1). The historical colonization patterns of the archipelago are important in order to understand the origins of the Pacific biota and are now being explored with molecular genetic methods on a number of taxonomic groups, such as birds (Filardi and Smith 2005;Smith and Filardi 2007), bats (Pulvers and Colgan 2007), and reptiles (Austin et al. 2010). Pulvers and Colgan (2007) described a close association between genetic lineages of bats on the Figure 1. The Solomon Archipelago. The broken line represents the subspecies boundary according to McCoy (2006). Sample locations are indicated by black triangles with letters for reference in the text. Yellow ovals indicate the Pleistocene land bridges of the Great Bukida landmass, which included Bougainville, Choiseul, Isabel, and Ngela. The Western Province was separated into two larger islands. Malaita and Makira have been isolated throughout the glacial cycles, while Guadalcanal may have been separated from Ngela by a narrow channel.  Solomon Archipelago and phylogenetic relationships predicted from presumed Pleistocene landmass conformations. A similar pattern was found for birds (Smith and Filardi 2007;Uy et al. 2009). Nonvolant animals such as reptiles and amphibians generally have slower dispersal rates than avian taxa, which could result in smaller scale genetic structure and higher local population differentiation (Hughes et al. 1997). Currently, no extensive biogeographic study that is based on molecular phylogenetics and that includes samples of a single species from all major islands has been conducted on the reptiles (or amphibians) of the Solomon Archipelago.

The genus Corucia
The prehensile tailed skink, Corucia zebrata (Gray 1855) is an ecologically and evolutionary unique reptile endemic to the archipelago (Fig. 2). Belonging to the Egernia group (Gardner et al. 2008), a largely Australian lineage of skinks, C. zebrata is the sister lineage to the remainder of the group and molecular clock estimates suggest that it has been separated from the other seven genera in the group for about 26 MYA (Skinner et al. 2011); a divergence that postdates the period when the Solomon Archipelago is believed to have emerged (Hall 2002). Corucia has traditionally been considered monotypic, with its single species represented by two geographically separated subspecies, C. z. zebrata to the east and C. z. alfredschmidti to the west (Kóhler 1997). The exact geographic location of the subspecies boundary is uncertain, but is likely to lie between the Shortlands/Bougainville and Choiseul (Kóhler 1997;McCoy 2006) (Fig. 1).
Corucia gives birth to live young and may reach a weight of over 1000 g. The combination of large size, a prehensile tail and a nocturnal, herbivorous, and arboreal nature makes it ecologically unique among the Scincidae (McCoy 2006;Ha-gen and Bull 2011). The species has been subject to poaching for the international pet market, a process that is believed to have put severe pressure on some populations (McCoy 2006). Although the conservation status for C. zebrata is unknown, the species' forest habitat is under threat, due to forest fragmentation and destruction as a result of intense logging (Dauvergne 1998).
We used molecular genetic tools to elucidate the phylogeography of Corucia under the hypothesis that the geographic distance between islands and the sea level changes during the LGM have affected the phylogeography of the species. Specifically, the following predictions were made: (1) The genetic divergences will be lower between populations on islands that were connected by Pleistocene land bridges than between those on islands that have remained isolated. (2) The genetic distance between populations will increase with geographic distance between islands (Isolation by distance [IBD] [Wright 1943]), reflecting that gene flow decreases with increasing distance between populations and that selection and genetic drift due to environmental conditions differ increasingly with increasing geographic distance between populations.
Additionally, we aimed to make inferences on the geographical origin and dispersal pattern of Corucia within the Solomon Archipelago and to use molecular clock calibrations and knowledge of the geological history to estimate the dates of the colonization events for the different islands across the archipelago.

Materials and Methods
We collected samples from 46 C. zebrata caught in the Solomon Islands between July 2007 and May 2008. The sample locations are indicated on Figure 1. The lizards were captured by hand and their GPS locations were recorded. Scale clips were collected and preserved in 70% ethanol (no Corucia was killed and no vouchers were taken for conservation and permitting reasons). Tissue samples from another 17 individuals from the frozen tissue collections of the Australian and South Australian Museums were also included. The sample locations for all 63 samples are listed in Appendix 1. The 63 samples included representatives from all major island groups with the exception of Bougainville, as permits to enter Bougainville for sampling purposes were not obtained during the study period. The sample sizes from the respective locations are as follows, with the letters in brackets referring to the locations on Figure

DNA extraction, PCR amplification, and sequencing
DNA was extracted from scales or liver using a Puregene TM DNA Isolation Tissue Kit, D-7000A (Gentra Systems, Minneapolis, MN), following the manufacturer's instructions. We amplified two sections of the mitochondrial genome (ND2 and ND4) and four nuclear loci (rhodopsin intron 4, an unknown intron, AKAP9 coding for amino acid positions 482-688 in the human gene, and PTPN12 coding for amino acid positions 298-497 in the human gene). See Table 1 for primer sequences and PCR conditions for each locus. The two mitochondrial markers have been informative for previous phylogeographic studies in reptiles (e.g., Gardner et al. 2008). The noncoding nuclear locus rhodopsin (Bossuyt and Milinkovitch 2000;Page 2000;Austin et al. 2010) is known to show intraspecific variation in a number of taxa. The two nuclear protein encoding genes were reported by Townsend (2008) to be among the most variable nuclear genes screened across a range of squamate reptile species, and therefore likely to be informative in phylogeographic studies.
The PCR reaction mixtures were prepared using Ampli-Taq Gold (Applied Biosystems, Foster City, CA) following the manufacturer's recommendations, with approximately 100 ng template DNA, 10mM total of each dNTP, 0.2 μM of each primer, 0.5 μg/μL BSA, and 4 mM MgCl, in a final volume of 25 μl. The following PCR cycle was used: 94 • C, 5 min; (94 • C, 45 sec; T a , 45 sec; 72 • C, 1 min) × 38 cycles; 72 • C, 10 min. The PCR reactions were purified using a Multi-Screen vacuum manifold (Millipore, Billerica, MA), following the manufacturer's protocol. The sequencing reactions were prepared with BigDye Terminator v3.1 Cycle sequencing kit (Applied Biosystems), following the manufacturer's protocol. The samples were sequenced in both directions. Electrophoresis was carried out on an ABI PRISM 3730 Analyser (Applied Biosystems). The sequences were edited using the ContigExpress function in Vector NTI v10 (Invitrogen, Carlsbad, CA). For the nuclear genes, we used parsimony to manually resolve alleles in heterozygous individuals. In individuals where phase could not be resolved, PCR products were cloned and several clones sequenced for each product. We used a StrataClone PCR Cloning Kit (Agilent Technologies, Santa Clara, CA), following the manufacturer's recommendations. The clones were sequenced as described above.

Data analysis
The sequences were aligned in Geneious v4.5.5 ). The coding sequences were translated to protein sequences in ContigExpress to verify that the reading frame was not disrupted by premature stop codons or deletions, as a further check of sequence quality and locus identity. We used DNAsp v5 (Librado and Rozas 2009) to characterize withinpopulation genetic diversity, estimate haplotype diversity, and characterize statistical properties of the sequences. Haplotype networks were generated in TCS v1.21 (Clement Table 1. Sequences, PCR conditions, and references for the six loci used.

Locus
Reference . We used the corrected AIC score (Akaike 1979) in Mr.ModelTest v2.3 (Nylander 2004) to select the appropriate models of nucleotide substitution for each locus and data partition. We investigated the degree of substitution saturation between the outgroup taxa and the ingroup using DAMBE (Xia 2001).
Bayesian analysis was carried out in MrBayes v3.1.2 (Huelsenbeck and Ronquist 2001), using samples with sequence for both ND2 and ND4. Due to the low number of variable sites and hence limited information, we chose to make the lowest possible number of partitions; thus concatenating the two datasets and combining the first and second codon positions. The following partitions and models were used: first and second (GTR+I+G), third (GTR+G), and tRNA (HKY+I). The analysis was run four times with 10,000,000 MCMC iterations, and sampled every 1000 generation, thus generating 10,000 trees per run. The analyses were checked for convergence in Tracer v1.5  and AWTY (Nylander et al. 2008). The runs were combined and the first 4000 of the 40,000 trees were discarded as burn-in. We used TreeAnnotator v1.5.4 (Rambaut and Drummond 2008) to generate a maximum clade credibility consensus tree. Fig Tree v1.2.2 (Rambaut 2008) was used to edit the consensus tree.
We quantified the divergence between different islands for the concatenated mitochondrial dataset using the Tamura and Nei distance method with 10,000 permutations in AR-LEQUIN v3.1 (Excoffier et al. 2005). Due to small sample sizes, the populations from Ngela and Malaita were omitted from this analysis. The larger sample sizes and multiple collecting locations for Choiseul and Isabel allowed for an additional pairwise comparison between the eastern and western sampling areas on each of these islands.
A Mantel test for IBD using the mitochondrial data was carried out using the IBDWeb Service v3.16 (Jensen et al. 2005). We performed two separate analyses wherein the geographic distances were defined as: (1) the distance (km) separating the actual sample locations on the respective islands, and (2) the minimum span of open water between the islands, without taking the specific locations on the respective islands into consideration.
We attempted to date the splits between different Corucia clades using BEAST v1.5.4 . The partitions and models of evolution were identical to the above-described MrBayes analysis. We selected a UPGMA (Unweighted-Pair Group Method with Arithmetic means) generated starting tree with a coalescent constant size tree prior and a relaxed clock with uncorrelated log-normal site model. The priors were specified with the following calibrations: Emergence of the Western Province provides an upper limit for colonization of this part of the archipelago and was set to maximum 3 MYA with a uniform distribution. The split between the ingroup and outgroups has been estimated to be 26 MYA (±5.1) (Skinner et al. 2011). The MCMC chain was run four times for 10,000,000 generations, sampling every 1000 trees. After examining the posterior probabilities in Tracer, we discarded the first 10% of trees as burn-in. A maximum clade credibility consensus tree was generated from the remaining trees using TreeAnnotator. The consensus tree was edited in Fig Tree.

Results
There were large differences among loci in the number of variable sites ( Table 2). As expected, the mitochondrial sequences were the most informative, followed by nuclear noncoding and nuclear coding loci. Different numbers of haplotypes were identified using TCS and DNAsp, this is due to how the two programs treat missing data. Analysis in DAMBE suggests the mitochondrial data have not been subject to substitution saturation.

Phylogenetic relationships between island populations
A maximum clade credibility tree from the 36,000 trees using the four combined runs of Bayesian inference is presented in Figure 3. There was weak support in the deeper nodes in the tree, while the support toward the tips was very strong for some clades. The deepest split within the ingroup was between populations from Makira/Malaita and those from the remainder of the sampled islands. Apart from populations from Choiseul and the Western Province, all islands were monophyletic with strong support. The Western Province samples were nested within the Choiseul clade with strong support. Additionally, there was monophyly for the eastern (location I) and western (location H) sampling locations on Isabel, with a support of 0.7 (support not shown on tree). As for the two sampling locations on Choiseul, the western (B) sampling location was monophyletic but was nested among the east Choiseul (C) haplotypes, and was therefore paraphyletic with both west Choiseul and the Western Province. The sample from Ngela was placed as a sister group to the samples from Isabel, while Makira and Malaita were sister clades. The samples from Shortland Islands were somewhat divergent from the remainder but nested close to the samples from Isabel. The trace files for the runs indicated that these analyses had converged (data not shown). However, due to the divergence between Corucia and the outgroup (26 MYA), the placement of the root in the ingroup is probably imprecise (Piller and Bart 2009).

Interisland genetic divergence
We calculated genetic divergence based on mitochondrial data between the different island populations in ARLEQUIN (Fig. 4)  between the eastern and western sampling site of Choiseul and 0.62 (P ≪ 0.001) between the eastern and western sampling site of Isabel.

Isolation by distance
A Mantel test did not detect a significant IBD from the combined data (r 2 = 0.04; P = 0.12) when analyzed with the geographic distance between sample locations. However, when geographic distance was set as the minimum span of open water between sampled islands, there was significant IBDM (r 2 = 0.19; P = 0.045; Fig. 5).

Haplotype networks
For the mitochondrial loci (Fig. 6), the separate islands constituted separate networks with the exception of Malaita that was included in a network with Makira, and the Western Province that was included in a network with Choiseul. The mitochondrial networks suggested that the haplotypes in the Western Province originated from Choiseul and that the samples from Malaita were genetically close to those from Makira. For the nuclear loci (Fig. 7) on the other hand, there was no concordance between haplotype and island origin as the most frequent allele in rhodopsin and PTPN12 was present on all islands. There were few haplotypes present in AKAP9. Here, one haplotype was restricted to the Shortland Islands and another haplotype was restricted to Isabel, but there was no further sorting. The multiple linkages in the unknown intron indicate recombination, and involve alleles from all the sampled islands except the Shortland Islands.

Dating of colonization events within the Solomon Islands
Our attempt to derive a date for the colonization events of C. zebrata using priors with specified divergence times produced a tree topology (Fig. 8) that was different from the tree presented in Figure 3. The samples were placed in the same clades in both trees, but the placement of the different clades in relation to each other was not concordant in the two trees. This was reflected in the support for the respective nodes, which in both trees was low in the deeper nodes and high toward the tips. The analysis suggested that C. zebrata colonized the Solomon Archipelago within the last 1-4 million years, and that some islands may have been colonized within the last 100,000-500,000 years, although no firm inferences about the site and timing of initial colonization and subsequent spread could be made. The large range for the 95% highest posterior densities for the outgroup and the root to the ingroup indicated the large degree of uncertainty in relation to the dating of the nodes.

Genetic distance and monophyly in relation to Pleistocene land bridges
Analyses of F st values did not support the Pleistocene land bridge hypothesis that predicted reduced genetic distance between islands connected during the LGM and higher genetic divergence between islands that remained isolated during the LGM. This hypothesis predicted that populations on Makira and Malaita should have the highest pairwise F st values, both to each other and to populations on other islands, while the F st value between populations on Choiseul and Isabel should be low. Our results only partly supported those predictions in that the samples from the isolated island of Makira were highly divergent. Although samples from the previously connected islands of Choiseul and Isabel showed somewhat lower genetic divergence (Figs. 3 and 4), they were at the same time highly divergent from each other, with a high F st value and contained different strongly supported clades (see Figs. 3, 4, and 6). Most islands, including those that were previously connected during the LGM, were monophyletic, which indicated that episodes of gene flow between the islands have been rare and that the genetic lineages on separate islands are probably the result of single colonization events (or single sources of colonization) on the respective islands. Bougainville, Choiseul, Isabel, and Ngela were joined by land bridges during the LGM (see Fig. 1). The Shortland Islands are situated within less than 8 km of Bougainville and were likely to have been included in the Great Bukida landmass. Indeed, the lowest pairwise F st values for samples from the Shortland Islands were between the islands of Choiseul and Isabel, thus suggesting that the animals on the Shortland Islands are genetically closer to Choiseul and Isabel than to animals from the rest of the archipelago. This is congruent with expectations based on geological data. According to McCoy (2006), Corucia on the Shortland Islands belong to C. z. alfredschmidti; however, our results were somewhat ambiguous in this respect: the samples from the Shortland  Islands were divergent, but still nested within the remainder of the samples (Fig. 3) and with only slightly higher F st values than those for Makira (Fig. 4). Hybridization between C. z. alfredschmidti and C. c. zebrata on the Shortland Islands is a possibility. Ultimately, inclusion of samples from Bougainville is required to provide a robust molecular genetic test of the systematic status of C. z. alfredschmidti and C. z. zebrata.

Nuclear markers and incomplete lineage sorting
Nuclear haplotypes shared between different islands (Fig. 7) indicated incomplete lineage sorting, which may produce gene trees that are either not congruent with species trees (Moran and Kornfield 1993), or incongruent with the true phylogeographic pattern of C. zebrata, in the case of the current study. The more structured data from mitochondrial loci are likely a result of the smaller effective population size and thus shorter coalescence time for the mitochondrial genome. Monophyly is expected to require considerably more time to develop for nuclear loci due to the larger effective population size for the nuclear genome (Nei 1987), sometimes up to several million years (Hudson and Turelli 2003). The smaller effective population size for the mitochondrial genome will lead to stronger effects of demographic stochasticity. These effects can be enhanced during founding events by small population sizes and by multiple impregnations of founding females. The latter will be an important factor among agestructured species with overlapping generations such as C. zebrata. In light of the relatively recent divergence of C. zebrata (Fig. 8), incomplete lineage sorting on nuclear loci was not surprising, and suggested that gene flow between islands has happened too recently to allow for complete sorting of nuclear loci, yet rarely enough to allow for reciprocal sorting of mitochondrial haplotypes (with the exception of Choiseul and the Western Province). Theoretically, this pat-tern could also be explained by male-biased dispersal, however given that inter-island gene flow in C. zebrata is likely to be the result of stochastic events (see below), this explanation is improbable. Nuclear loci were included in the analysis as evolutionary relationships inferred from one gene tree may be insufficient for estimations of the true species tree (Moore 1995). However, in this case the divergences between the different populations were too shallow for nuclear loci to yield useful phylogenetic information.

Intraspecific variation in C. zebrata versus general patterns in pacific taxa
Based on the data presented in this paper, we rejected the hypothesis that Pleistocene land bridges can explain much of the intraspecific variation found in C. zebrata. This differed from the pattern reported for the Solomon Archipelago Melonycteris fruit bats by Pulvers and Colgan (2007), where the current biogeographic regions were best explained by geographic conditions during the LGM, and where no genetic c 2012 The Authors. Published by Blackwell Publishing Ltd.
differences were detected between populations inhabiting islands previously connected by land bridges. A similar pattern has been found for a number of bird species (Smith and Filardi 2007;Uy et al. 2009), again suggesting that Pleistocene land bridges have affected the current patterns of genetic diversity in avian taxa on the Solomon Archipelago. It is plausible that the different pattern for C. zebrata is a consequence of the much greater dispersal ability of avian taxa. It is not known if the Pleistocene habitat between the islands of Bougainville, Choiseul, Isabel, and Ngela was suitable for C. zebrata or whether the land bridge habitats supported a vegetation matrix that would encourage dispersal to more appropriate habitat during the last LGM. Habitat unsuitable for C. zebrata may have allowed easier dispersal for volant animals than for large lizards. The fact that the extreme ends of Choiseul and Isabel represent distinct populations of C. zebrata (Figs. 3 and 6) suggests that the species infrequently disperses over distances such as the length of Choiseul or Isabel. Siler et al. (2010), found that genetic diversity within geckos on the Philippine Archipelago were often greater within islands than between islands, a result that is comparable to the large genetic divergence found for C. zebrata within the islands of Choiseul and Isabel. Moreover, according to Esselstyn and Brown (2009), the distribution of genetic lineages of shrews on the Philippine Archipelago can also be explained by current islands and to a certain degree by IBD. Again, this is concordant with results on C. zebrata, where genetic lineages are largely structured into clades that represent the modern islands (Figs. 3 and 6) and where the size of the dispersal barrier (IBD) appeared to explain approximately 20% of the intraspecific variation (Fig. 5). It is possible that the pattern described in Heaney (2005) where the current allozyme distribution in rodents is largely explained by Pleistocene shore lines is an oversimplification, and that ecological factors such as habitat preference and IBD (Roberts 2006) are equally important. Conservation managers should be aware that biogeographic regions cannot be directly extrapolated between taxa with differing dispersal abilities.

Colonization history and dating of colonization events within the Solomon Islands
The Western Province has been isolated since it was formed but is situated in geographical proximity to Choiseul, and separated from it by approximately 60 km of open water. The low F st value between the Western Province and Choiseul was in concordance both with the tree topologies in Figure 4, where the Western Province was nested within Choiseul and with the haplotype network in Figure 6. This suggested that the Western Province was colonized by animals from the ge-ologically much older Choiseul. Given the young age of the Western Province, a direction of colonization events from the nearby Choiseul to the Western Province is concordant with expectations based on geological data. With the exception of this Western Province example, it was difficult to infer a dispersal pattern among any other of the islands from our phylogenetic reconstruction (Figs. 4 and 8). Pulvers and Colgan (2007) argued that fruit bats (Melonycteris) most likely colonized the Solomon Archipelago from east to west, starting at Makira, as bats on Makira appear most basal in the Melonycteris phylogeny. Pulvers and Colgan (2007) did not acquire samples from Bougainville. When a mitochondrial phylogeny of C. zebrata is reconstructed without samples from the Shortland Islands (data not shown), the tree topology resembles that of Melonycteris in Pulvers and Colgan (2007). A more complete sampling regime may thus alter the topology of the tree and it is plausible that a similar effect may occur if Bougainville samples were included in the Melonycteris phylogeny. A rapid and recent radiation across the archipelago with subsequent isolation could explain the results of the current study. This hypothesis is supported by results from dating using BEAST, which suggested that C. zebrata colonized the archipelago during the last 1-4 million years (Fig. 8). This is surprising given that C. zebrata diverged from the other genera of the Egernia group approximately 26 MYA, and one would therefore expect a deeper divergence for some clades. It is possible that Bougainville is the site of initial colonization and that the species was restricted to this island for several million years prior to dispersal to other islands. Another scincid taxon Tribolonotus radiated primarily from Bougainville and it is plausible that Bougainville (which is close to the larger land mass of New Guinea) is the source of a number taxa that have spread to the eastern islands of the Solomon Archipelago (Austin et al. 2010). Bougainville must be included among the C. zebrata sample locations in order to test this hypothesis. Corucia zebrata has probably colonized across the Solomon Archipelago by over-water dispersal on flotsam, which is a kind of colonization mechanism that has been observed for Iguana iguana in the Caribbean archipelago (Censky et al. 1998) and that is assumed to the mechanism behind the Caribbean Anolis radiation (Schoener and Schoener 1984;Glor et al. 2005), as well as radiation of reptiles of the Mexican Revillagige Islands (Brattstrom 1990). Corucia zebrata is a canopy dwelling species, and one tree may be inhabited by several lizards (Hagen and Bull 2011). Rafts comprising large trees may therefore contain a number of C. zebrata, thus facilitating successful establishment upon arrival. The herbivorous diet of the species would prolong survival of individuals on rafts that contained tree foliage. Phylogenetic comparisons with other species, such as the native frog Litoria lutea, with a lower tolerance for exposure to sea water, fewer on-raft feeding opportunities and thus lower dispersal ability than C. zebrata, would give a more complete understanding of the historical biogeography of the Solomon Archipelago. Table A1. ABTC numbers, date of collection, Island, sample location, latitude, and longitude for Corucia zebrata sampled across the Solomon Archipelago. Sample locations refer to letters in Figure 1.
The following samples were used as outgroup samples (ABCT and Genbank accession numbers for ND2 and ND4 in brackets):