Dispersal, when accompanied by reduced gene flow and natural selection, influences speciation rates among groups of organisms. We used molecular phylogenetics, divergence time estimates, and population genetics to reconstruct the mode, pattern, and tempo of diversification within the wide-ranging purple swamphens (genus Porphyrio), with emphasis on the “supertramp” P. porphyrio. Our results suggest that the Porphyrio clade arose during the Middle Miocene in Africa, with a single colonization in the Americas and several other colonizations in Southeast Asia and the Indo-Pacific around 10 mya. We found that the widespread P. porphyrio is not monophyletic. Indeed, several subspecies and subspecies groups may represent species-level lineages. The P. p. melanotus lineage probably reached Australasia during the Pleistocene (600 kya), although some islands were colonized only in the past few hundred years. New Zealand, and some other islands, had previously been colonized (∼2.5 mya) by flying Porphyrio that evolved into flightless endemic species. Early and recent lineages are now sympatric. Widespread occupation of oceanic islands implies high dispersal and colonization rates, but gene flow probably occurs episodically and follows varying routes at different times. This pattern of colonization enables populations to differentiate and, ultimately, speciate.
The ability to disperse and colonize new habitats provides organisms with ecological opportunities to harvest novel resources and establish new populations. Vagrant species that reach, via long-distance dispersal, new regions or isolated islands that are thousands of kilometers from their traditional breeding range may generate new flocks of dispersers before being totally or partially displaced by more efficient competitors (Diamond 1974, 1975). This “supertramp” strategy can also give rise to numerous, phenotypically distinct variants, which may lead to speciation (Simpson 1953, Diamond 1974, Diamond et al. 1976, Grant 1986, Whittaker 1998, Crisp et al. 2011). However, for a single species to persist, without differentiation, over a large breeding range of fragmented habitat, individuals must move between habitat patches at a rate sufficient to counter the evolutionary effects of isolation. This requirement predicts that such widespread species have a high level of gene flow among populations distributed across several landscapes. Alternatively, relatively small, isolated populations can segregate and accumulate phenotypic differences if range expansion is not constant through time or if the direction of dispersal changes. Outcomes may vary, depending on the number of independent radiations, differences in diversification rates, rate and pattern of gene flow, and rapidity of species radiation following a wave of dispersal (Mayr and Diamond 2001, Moyle et al. 2009, Cibois et al. 2011).
Inferring the geographic origin and temporal diversification of organisms is an essential part of biogeography and depends on an accurate estimate of evolutionary relationships among species (Rosen 1978, Filardi and Moyle 2005). In the case of birds, enabled by flight to disperse long distances, the spatiotemporal patterns of diversification can be challenging to analyze, especially when shallow radiations on islands generate differential morphological traits that obscure evolutionary affinities (Filardi and Moyle 2005, Irestedt et al. 2013). Dispersal and adaptation together are important drivers of insular diversification of many bird groups (Pratt 2005, Grant and Grant 2008) and account for much of the diversity that we find today in archipelagos (Trewick and Gibb 2010, Trewick 2011). Colonization of islands sometimes involves loss of the capability for further long-distance dispersal, when flight is not integral to foraging, social interaction, or predator avoidance (McNab 1994, McNab and Ellis 2006, Steadman 2006). Reduction, and even loss, of flight capacity can be an adaptive response to island life, and some speciation may occur with adaptation of flightlessness, as a result of altered selective environments (Milá et al. 2010, Sly et al. 2011, Alonso et al. 2012, Runemark et al. 2012).
Family Rallidae (Aves: Gruiformes) is diverse and cosmopolitan. It includes common species that are good dispersers, as well as regional and island endemics. Many oceanic islands that were naturally without terrestrial mammal predators appear to have favored reversion to a terrestrial lifestyle after colonization and speciation by flying ancestors (Ripley 1977, Steadman 2006). This combination of high dispersal and high endemicity associated with the loss of flight makes them interesting subjects for evolutionary analysis. In particular, the large, flamboyant purple swamphens (genus Porphyrio) demonstrate extraordinary dispersal capabilities, with evidence of multiple invasions, apparently spaced out in time, that resulted in divergences of size, color, and other traits (Ripley 1977, Remsen and Parker 1990, Trewick 1996). Seven species of purple swamphens are currently recognized, 4 of which are or were present in the Oceania region (Trewick 1996, Taylor 1998). Principal among these is the widespread “supertramp” Purple Swamphen (Porphyrio porphyrio), which occurs from Africa and the Mediterranean east to the Pacific (Ripley 1977). This taxon comprises apparently parapatric morphological variants that have, at times, been classified into ∼13 subspecies or species (Figure 1A; Ripley 1977, Taylor 1998). Although sometimes considered a reluctant flier (Craig 1977, Craig and Jamieson 1990), this taxon has nevertheless established populations on many oceanic islands, throughout the Indian and western Pacific Ocean (Mayr 1949, Ripley 1977, Trewick 1997, 2011). At least 2 colonizations of New Zealand resulted in the presence of the North Island Takahe (P. mantelli) and South Island Takahe (P. hochstetteri), endemic flightless herbivores that were sympatric with flying swamphens (Trewick 1997, Trewick and Worthy 2001). Some island populations and subspecies are known only from fossils that reveal the numerous extinctions that followed colonization of those islands by people (Steadman 1995, 2006, Steadman et al. 1999). Insular endemics have been recognized as distinct species on New Caledonia and New Zealand (Balouet and Olson 1989, Trewick and Worthy 2001) and on other Pacific islands (Steadman 1988, Kirchman and Steadman 2006).
We used multilocus DNA sequence data to generate a dated phylogenetic hypothesis of relationships within Porphyrio and to explore the pattern of gene flow among populations of P. porphyrio. We address the following questions to gain insights into the biogeographic origin and diversification of these birds: (1) What is the phylogenetic structure of the genus? (2) What time of diversification, and pattern of dispersal and colonization, explains current diversity? (3) Is there support for a single or multiple range expansions? (4) Are regional subspecies of P. porphyrio monophyletic, or is there a mismatch between clade structure and taxonomy?
To obtain DNA for analysis, we sampled bones, toe pads, feathers, blood, and muscle tissue from specimens of the 7 known species in the genus Porphyrio (Figure 1B), including representatives of P. porphyrio subspecies from Africa, Europe, Asia, and Pacific islands (Table 1 and Figure 1A). Additionally, we sampled several populations in New Zealand and Australia separated by ∼1,500 km of sea and graded terrestrial landscapes to explore gene flow at different spatial scales.
Taxa, museum voucher numbers, locality, type of tissue, and GenBank accession numbers of data included in our study. An asterisk indicates sequences <200 bp (see Appendix B).
DNA extractions from bones and toe pads were carried out in a dedicated ancient DNA (aDNA) laboratory (Ecology Group, Massey University, Palmerston North, New Zealand; see DNA Toolkit at http://evolves.massey.ac.nz). DNA extractions from toe pad samples were performed using the QiAMP DNA Minikit (Qiagen, Valencia, California, USA), following the manufacturer's instructions and standard procedures for aDNA (Cooper and Poinar 2000, Rohland and Hofreiter 2007). DNA from bones was extracted using decalcification with EDTA and proteinase K digestion in Tris-buffered saline, followed by purification with phenol–chloroform. DNA from fresh tissues was extracted in a laboratory geographically separated from the aDNA laboratory, using either Tissue DNeasy kit (Qiagen; following the manufacturer's instructions) or incubation at 55°C with proteinase K and a CTAB buffer (2% Hexadecyl trimethyl ammonium bromide, 100 mM Tris–HCl, pH 8.0, 1.4 M NaCl, 20 mM EDTA), followed by a combined phenol–chloroform–isoamyl alcohol (25:24:1) cleanup.
Mitochondrial and Nuclear DNA Amplification
We sequenced 2 mitochondrial genes and 1 nuclear gene for population genetic analyses of P. porphyrio in Australia and New Zealand: mitochondrial control region (CR) and cytochrome oxidase b (cyt b), plus a fragment of the nuclear beta-fibrinogen intron 7 (BFG-7). For phylogenetic analysis of Porphyrio, parts of 2 additional mitochondrial genes (ribosomal RNA 12S and 16S) and 1 nuclear gene (recombination activating gene 1 [RAG-1]) were amplified from representative specimens of the currently recognized species and subspecies of P. porphyrio. Additional sequences were downloaded from GenBank (Table 4 in Appendix A). Standard polymerase chain reaction (PCR) methods (using the primers listed in Table 5 in Appendix A) were used for amplification of nuclear and mitochondrial fragments. Amplification products were purified with QIAquick PCR cleanup kit (Qiagen) or ExoI/SAP digest. For each PCR product, both strands were sequenced using Big Dye Terminator version 3.1 reagents and an ABI 3730XL automated DNA sequencer (Applied Biosystems, Foster City, California, USA). Short sequences without GenBank accession numbers are provided in Appendix B.
Phylogeny and Divergence Times
All sequences were edited, assembled, and aligned using Geneious version 6.0.5 (Drummond et al. 2012a) and checked by eye. Alignments of ribosomal genes, CR, and BFG-7 were conducted using Gblocks version 0.91b (Castresana 2000) and evaluated by eye. Cytochrome oxidase b and RAG-1 were checked for the presence of indels and stop codons. Prior to concatenated analyses, we performed individual gene tree analysis to detect spurious branch-length patterns and evidence of significant incongruence. We built a supermatrix with a six-way partition by gene: 12S, 16S, CR, cyt b, BFG-7, and RAG-1 (Wiens 2006, Holland et al. 2007, Wiens and Moen 2008, Johnson et al. 2012). Maximum likelihood (ML) trees were implemented in RAxML version 8.0.24 via the CIPRES portal (Miller et al. 2010). We used a general time-reversible model with gamma distribution (GTR + Γ), which allowed RAxML to halt bootstrap resampling automatically (bootstopping) once split support values converged (Pattengale et al. 2010). We conducted Bayesian phylogenetic analyses using MrBayes version 3.2.2 as implemented in the CIPRES portal under a GTR + Γ+ I model of evolution. The model was estimated in ModelTest version 3.7, using Akaike's Information Criterion (Posada and Crandall 1998). After performing shorter test runs, we conducted 3 parallel runs of the Metropolis Coupled Markov Chain Monte Carlo (MCMCMC) algorithm for 5 million generations each, sampling 1 tree with associated parameter values per 5,000 generations, and employing 3 heated chains and 1 cold chain. Convergence and diagnostics of the Markov process were visualized using Tracer version 1.6 (see Acknowledgments). The first half million generations (10%) were discarded as burn-in. A burn-in of 10% gave optimal results, and we obtained effective sample sizes (ESS) >200 for 95% of the parameters. The ML and Bayesian trees were viewed using FigTree version 1.4.2 (see Acknowledgments) and SplitsTree version 4.12.8 (Huson and Bryant 2006). Amaurornis flavirostra was used as an outgroup to root the tree.
Divergence times among lineages of Porphyrio were estimated using a relaxed Bayesian clock implemented in BEAST version 1.7.5 (Drummond et al. 2012b). For calibration constraints, we used the basal divergence estimate of Porphyrio with a normal distribution of 11–20 Ma (95% range) and the basal split of Amaurornis flavirostra and Porphyrio with a normal distribution of 27–35 Ma (95% range), as previously calculated from an analysis using a widely sampled dataset of mitochondrial and nuclear genes (Garcia-R. et al. 2014). We combined the results of 3 independent runs of 30 million generations to ensure ESS scores >200 for 95% of the parameters in each run. Chains were sampled every 4,000 generations, and a burn-in of 10% (3 million generations) was used. The tree, with times of divergence and highest posterior density (HPD) intervals, was visualized using FigTree.
Population Differentiation and Demographic History
Population-level analyses were carried out with 2 datasets: (1) concatenated mitochondrial loci CR and cyt b; and (2) the autosomal locus BFG-7. Sequence ambiguities at heterozygous sites in BFG-7 that indicated separate alleles were resolved using PHASE implemented in DnaSP version 5.0 (Librado and Rozas 2009) with the default parameters. To test for intralocus recombination in BFG-7, we used the PHI test (Bruen et al. 2006) implemented in SplitsTree. This is a robust test that can reliably detect recombination and report few false positives (Martin et al. 2011). We calculated the following summary statistics for genetic variation of each population in DnaSP: number of haplotypes (h), nucleotide diversity per site (π), number of segregating sites (S), Watterson's estimator of per site population mutation rate (θW), Tajima's D statistic (DT), and Ramos and Rozas's R2-test. For each population, DT was also analyzed, using 103 coalescent simulations conditioned on the sample size and the observed number of segregating sites (Hudson 1990). Demographic expansion was assessed using the R2-test implemented in DnaSP because it is the most powerful test when dealing with limited sample sizes (Rozas et al. 2003).
Lynch and Crease's pairwise FST (Lynch and Crease 1990) was calculated in DnaSP, using 5,000 replicates and a significance level of ≤0.05 to test the null hypothesis of panmixia (Raymond and Rousset 1995) between pairs of P. porphyrio populations in Oceania. The level of population genetic structure was tested using an analysis of molecular variance (AMOVA) implemented in Arlequin version 3.5 (Excoffier et al. 2005). To complement the phylogenetic inferences, haplotype networks were constructed using Network version 126.96.36.199 (Bandelt et al. 1999) with median joining to visualize the relationship between haplotypes and their geographic distribution. A Mantel test was conducted for correlation between uncorrected mitochondrial genetic distances estimated in MEGA version 5.2 (Tamura et al. 2011) and linear geographic distances (Jensen et al. 2005) using the ade4 package (Dray and Dufour 2007) in R (R Development Core Team 2008) with 10,000 permutations.
The present study includes wider taxonomic representation than the first molecular phylogenetic study of Porphyrio, which compared 4 taxa using a single gene, mitochondrial 12S rRNA (Trewick 1997). We have included Porphyrio alleni, P. flavirostris, and P. albus, as well as the P. porphyrio subspecies porphyrio, indicus, poliocephalus and other Australasian subspecies (note that Trewick  included P. p. seistanicus from Turkey). The complete alignment of 6 gene fragments contained 4,304 base pairs [bp], comprising 816 bp of cyt b, 699 bp of CR, 728 bp of 16S, 402 bp of 12S, 868 bp of RAG-1, and 791 bp of BFG-7. No premature stop codons were detected in the 2 protein-coding genes. Individual gene trees did not reveal spurious sequences or significant conflict among individual phylogenies. Phylogenetic analyses that included just 1 representative of each species and subspecies that had the most complete gene sets yielded similar topologies (results not shown). Topologies from ML and Bayesian analyses were congruent for the concatenated dataset. The African species P. alleni is sister to the New World species pair P. martinica and P. flavirostris (Figure 1C). Porphyrio porphyrio did not form a monophyletic group. Instead, it comprised 6 distinct clades (porphyrio, indicus, madagascariensis, pulverulentus, poliocephalus [including seistanicus], and melanotus), and it was paraphyletic with respect to 3 species-level taxa: P. mantelli and P. hochstetteri from New Zealand, and P. albus from Lord Howe Island (Figure 1C, clade A). Porphyrio p. melanotus (Figure 1C, clade B) includes the parapatrically distributed subspecies bellus, caledonicus, samoensis, vitiensis, palliatus, pelewensis, melanopterus, and chathamensis (the latter was previously demonstrated to be invariant at the 12S locus by Trewick ). This phylogenetic and spatial structure corresponds with significant differences in color and size that have been previously described in some detail (Mayr 1949, Ripley 1977, Simmons et al. 1980, Sangster 1998, Taylor 1998). One Indonesian specimen (P. p. indicus) did not group with other specimens from this region but instead was more closely related to the clade composed of P. p. pulverulentus and P. albus (Figure 1C). The close similarity of P. p. pulverulentus and P. albus sequences indicates a complex history of exchange, because their lineage is not recorded in islands between. Sequence data for P. albus came from old and rare museum specimens, and such aDNA sources have to be treated with caution. However, consistent results were obtained from separate samples and replicate PCRs. There is the possibility of mislabeling of museum specimens, but the white plumage characteristic of P. albus is uncommon in other populations.
Divergence time analysis suggests a Middle Miocene origin of diversification within Porphyrio, with the split between the lineages that led to the African species P. alleni and to other species occurring earliest, around 14 (19–9) mya. Splitting among P. porphyrio “subspecies” was estimated to have occurred about 6 (11–2) mya, with a likely colonization of P. p. melanotus in Australasia occurring in the late Pleistocene (600 kya). However, an earlier colonization by a flying P. porphyrio at the start of the Pleistocene (∼2.5 mya) resulted in the flightless takahe endemic to New Zealand.
Population Genetic Structure
The mitochondrial data contained 1 to 4 haplotypes (h) in each of the populations sampled (Figure 2A). Nucleotide diversity (π) at sampling localities with n ≥ 2 was variable, ranging from 0.0 in Northland and Otago, New Zealand, to 0.0059 in Victoria, Australia. Zero to 12 segregating sites (S) were observed in each population, yielding a population mutation rate per site (θW) between 0.0 and 0.0065 (Table 2). For BFG-7, two alternative haplotypes were identified for alleles possessed by heterozygous individuals. No statistically significant evidence for recombination (P = 0.06) was detected using a phi test. Between 4 and 14 inferred BFG-7 sequences were sampled per population, and 1 to 4 unique haplotypes were found (Table 2). The number of segregating sites in each population varied from 0 to 6. Aside from invariant samples from New South Wales and Western Australia (north), nucleotide diversity ranged from 0.00069 in Palmerston North, New Zealand, to 0.00440 in Victoria. This latter population also had the highest inferred population mutation rate per site (0.00414). No population showed a significantly skewed DT for mitochondrial DNA (mtDNA) or nuclear data, although power to reject the null hypothesis of neutrality may have been hampered by small sample sizes. The population-size-change (R2) test did not find evidence of demographic expansion of the populations sampled (Table 2).
Summary of descriptive statistics for mtDNA data (cyt b and CR) and the nuclear locus (BFG-7) used in population genetic analyses.
Panmixia was evident in 2 localities sampled on the same island and separated by ∼500 km: Palmerston North and Northland on North Island, New Zealand. Nevertheless, population genetic structuring was found among other populations more geographically remote from one another, with the exception of some of the comparisons among Western Australia (north) and Victoria, which was probably due to the low sample size (Table 3). Analysis of BFG-7 showed little population genetic structuring (Table 3) among localities at which mtDNA diversity was clearly partitioned. Significantly different values were obtained for most pairwise comparisons among Victoria populations because this location had endemic haplotypes. Consistent with the population pairwise differentiation analysis of the mitochondrial data, most genetic variation was explained by differences among populations (70.5%; P < 0.0001). The AMOVA of the BFG-7 data indicated that a small but significant (14%; P < 0.01) component of variance was attributable among populations. Population geographic structure was evident in the mitochondrial and nuclear haplotype networks (Figure 2B, 2C), and the Mantel test showed a significant correlation between genetic and geographic distances among populations (r = 0.508, P < 0.05), even though only ∼26% of genetic divergence was explained by geographic distance.
Pairwise comparisons between populations using mtDNA data (above diagonal) and the nuclear locus BFG-7 (below diagonal). Negative values represent a program idiosyncrasy due to the small sample size and are effectively zero. Significant values are in bold (*0.01 < P < 0.05; **0.001 < P < 0.01).
Biogeography and Evolution of Swamphens
Our phylogenetic analyses and molecular dating support independent and temporally nonoverlapping colonization events among Porphyrio species. This interpretation is, however, based on surviving or recently extinct lineages only; other colonizations are represented by fossils on Oceanic islands (Steadman 1988, 2006, Steadman et al. 1999) or have left no trace at all. The most likely area of origin of Porphyrio is Africa, with colonization westward into the Americas and several other colonizations northeastward (Europe, Asia, and Oceania) during the Miocene and Pleistocene. The oldest split among the currently recognized P. porphyrio lineage (Figure 1C, clade A) occurred in the Late Miocene (∼6 mya), giving rise to P. p. porphyrio on the Mediterranean coast of Europe and P. p. indicus in Indonesia. Further diversification took place during the Pliocene, giving rise to P. p. madagascariensis in Africa and a radiation into Oceania.
The unique sequence obtained from the extinct P. albus of Lord Howe Island suggests a close affinity to Philippine P. p. pulverulentus, indicating that it was perhaps a white color variant founded from P. p. pulverulentus migrants. The flightless status of P. albus appears to be equivocal, and the population seems to have been polymorphic for plumage, with a high frequency of white individuals (White 1790, Hindwood 1940, Greenway 1967). Aberrations in color have been found in some insular populations, caused perhaps by an allele fixed through a founder effect (Cunningham 1955, Steadman 2006, Uy et al. 2009). White Porphyrio occur intermittently, and recent observations include an individual P. p. melanotus in Otago, New Zealand (Trewick and Morgan-Richards 2014). The Lord Howe population may have been established from a small number of colonizing individuals from the Philippines during the late Pleistocene (∼500 kya), but this would have involved dispersal from the Philippines to Lord Howe Island over other islands. We remain cautious about the short DNA sequence obtained from P. albus.
Despite the appearance that flight is used only infrequently among P. porphyrio subspecies, the lineage has dispersed, colonized, and established populations multiple times across open expanses of water. Haplotypes from Indonesia (a specimen from Java is closely related to a specimen from the Philippines) and New Caledonia (specimens from this locality are closely related to specimens from localities as far away as Sulawesi and Palau) support the inference of exchange (Figure 1C). Although rare misplaced haplotypes of this sort might be evidence of ongoing exchange among island populations, they could also be the product of incomplete lineage sorting or past migration events.
Porphyrio p. melanotus (Figure 1C, clade B) appears to have entered Australasia within the past 600,000 yr, but bone deposits show a more recent arrival on some remote islands (Millener 1981, Taylor 1998, Steadman 2006). This includes New Zealand, where deposits indicate colonization ∼500 yr ago, after Polynesian settlement (Trewick and Worthy 2001). This dating is much more recent than the estimated divergence of P. porphyrio and takahe lineages (P. hochstetteri and P. mantelli) that must represent an earlier, separate colonization. Within the Australia–New Zealand geographic region, the distribution of genetic variability (phylogenetic analysis, FST, AMOVA, haplotype networks, and Mantel test) indicates that the genetic structure of P. p. melanotus populations is not homogeneous. This lineage may have originated in Wallacea, and eustatic sea-level changes could have aided colonization by reducing overwater dispersal distances. Lowered sea level during glacial phases of the Pleistocene reduced the distance between some land areas, including between Papua New Guinea and Australia (Voris 2000, Hall 2009, Jønsson et al. 2010, Wurster et al. 2010, Lohman et al. 2011, Condamine et al. 2013, Irestedt et al. 2013). They did not, however, significantly alter the overwater distance between Australia and New Zealand (Graham 2008). This colonization pattern has created the allopatric distribution currently shown in Oceania. We note that higher genetic diversity in Australia than across the Tasman Sea reflects a recent arrival in New Zealand and indicates an influence of distribution due to persistence of habitat rather than geographic distance correlated with human settlement and clearance of forest.
Pairwise differences among Australasian populations were higher for the mtDNA data than for the nuclear DNA gene. This difference may be explained by the higher mutation rate and lack of recombination in the mitochondrial genome (Neiman and Taylor 2009). However, different population genetic (e.g., background selection), demographic (e.g., effective sex ratio and/or male-biased migration rates), or natural selection also must be considered for those higher mtDNA FST estimates (Palumbi and Baker 1994, Charlesworth 1998, 2009, Stinchcombe and Hoekstra 2008, Muir et al. 2012). Despite the indication that the populations studied have not undergone recent demographic changes, with a marked reduction of genetic variation within populations and increasing genetic differentiation, the swamphens within P. p. melanotus of south Western Australia (Figure 1C, clade B) show exceptional and not subtle differentiation in plumage color pattern (Whittell 1934). The current nominate subspecies P. p. bellus in south Western Australia has a prominently brighter blue breast and throat color than P. p. melanotus (see images in Figure 2A). Differences in color, size, and other traits are evident among other lineages within the melanotus clade (Ripley 1977) and in other clades. For instance, within the poliocephalus clade, the Middle Eastern “seistanicus” population is grayish compared with individuals from India (see color photographs in Figure 1C). The mismatch between plumage patterns and the distribution of neutral population genetic markers suggests that differentiation in color and other traits have arisen rapidly in Porphyrio and are subject to selection in local environments (Mayr 1954, Nosil et al. 2009, Feder et al. 2012) or to stochastic genetic drift (Clegg et al. 2002a, 2002b). The lack of sorting at the BFG-7 locus suggests that fixed plumage-color differences among populations are not the result of drift but are perhaps better explained by “purifying” sexual selection.
Selection that results in character divergence among populations can occur without being detectable by neutral genetic markers (Charlesworth and Charlesworth 2009, Nosil et al. 2009). There may be lineages with genomic region(s) involved in adaptive divergence, and these regions may respond independently to environment pressures via selection (Schneider et al. 1999, Clegg et al. 2002a, 2002b, Schluter 2009, Via 2009, Cooke et al. 2012). Appearance is a trait that is important in assortative mating by individuals within a population (Schluter 2009, Maan and Seehausen 2011) and is strongly implicated in the behavior of communally breeding P. p. melanotus (Jamieson 1988). As such, it may drive monomorphism in local populations. Selection on mate choice and kin fitness likely maintains local population phenotypes in stable frequencies (Andersson et al. 1998, Eaton 2005, Johnsen et al. 2006, Pryke and Griffith 2006, Murphy 2008).
Although P. p. melanotus is not the lineage that gave rise to the flightless insular species of New Zealand (P. mantelli and P. hochstetteri), the recent success of P. p. melanotus in reaching several remote islands in the Pacific is testimony to the high success rate of dispersal and colonization. However, the restriction of gene flow evident in FST values suggests that range expansion is probably episodic, and this enhances the opportunity for speciation and establishment of reproductive barriers. Sexual selection could help drive locus-specific evolution without being evident in genes that are (with respect to those traits) neutral, and reproductive isolation could evolve as a consequence of local adaptation and selection on characters involved in mate choice and inclusive fitness by way of mating behavior.
The sampling of Porphyrio species and the extensive sampling of P. porphyrio in Africa, Europe, Asia, and Oceania were supplemented by contributions from museum collections, researchers, and institutions. We are grateful to the following individuals and institutions who provided tissues necessary for our study: W. Boles and J. Sladek at the Australian Museum (Australia); L. Joseph and R. Palmer at the Australian National Wildlife Collection (Australia); J. Summer at the Museum Victoria (Australia); C. Stevenson at the Western Australian Museum (Australia); C. Miskelly and A. Tennyson at the Te Papa Museum (New Zealand); C. Fisher and T. Parker at the Liverpool Museum (England); R. Prys-Jones at the Tring Museum of Natural History (England); R. T. Brumfield at the Louisiana University Museum of Natural Science (USA); D. Willard at the Field Museum of Natural History (USA); P. Sweet and C. Filardi at the American Museum of Natural History (USA); J. Fjeldså and J. Kristensen at the Natural History Museum of Denmark (Denmark); S. Birks at the Burke Museum (USA); E. Garcia Franquesa at the Barcelona Museum (Spain); M. Marquez at the Banco de Tejidos Animales de Cataluya (Spain); G. Davies at the Ditsong Museum (South Africa); M. Trotter, R. Adams, and S. Pilkington at Fish and Game (New Zealand); T. Burns, M.Sc. student at Massey University (New Zealand); A. Gamauf at the Vienna Museum (Austria); C. M. Milensky at the Smithsonian Institution (USA); R. Urriza at the Philippines National Museum (Philippines); R. Hinlo (Philippines); D. Philippe (France); and S. A. Subrata (Indonesia). For helpful advice on the analysis, we thank B. Holland (University of Tasmania). Images of birds used in figures were generously provided by J. Dolphijn, L. and V. Dunis, A. d'Affonseca, C. Gouraud, J. Sagardia, H. Weerman, J. Babbington, M. Montoro, V. Moreno, S. A. Subrata, N. Ramesh, P. Bourdin, and D. Gambas Izquierdo. We are grateful to I. Castro and F. Sheldon for helpful comments that greatly improved the manuscript. J.C.G.-R thanks the Massey Doctoral Scholarship, New Zealand International Doctoral Scholarship, A. M. Soto, and T. Garcia. Tracer software is available at http://tree.bio.ed.ac.uk/software/tracer/. FigTree software is available at http://tree.bio.ed.ac.uk/software/figtree/.
Taxa, GenBank accession numbers, and original source of data for additional DNA sequences included in our study.
Primers for PCR and DNA sequencing employed in our study. Asterisks denote primers taken from the primer database of the Allan Wilson Centre for Molecular Ecology and Evolution, Massey University.
These sequences are <200 bp, too short to be submitted to GenBank. Specimen name is followed by museum voucher when available. Acronyms for museums are the same as in Table 1.
>Porphyrio porphyrio porphyrio_BM 93-0242-T
>Porphyrio porphyrio pelewensis_LIV T9774
AACTGGGATTAGATACCCCACTATGCTTGGCCCTAAATCCAGATACTCACCACCACT AGAGTATCCGCCTGAGAACTACGAGCACAAACGCTTAAAACTCTAAGGACTTGGCGGTGCC CCAAACCCACCTAGAGGAGCCTGTTCTGTAATCGATAACCCACGATATACCCAACCCCTT CTTGCCCAAAGCAGC
>Porphyrio porphyrio palliatus_LIV T9048
AACTGGGATTAGATACCCCACTATGCTTGGCCCTAAATCCAGATACTCACTACCACTAG AGTATCCGCCTGAGAACTACGAGCACAAACGCTTAAAACTCTAAGGACTTGGCGGTGCCCCA AACCCACCTAGAGGAGCCTGTTCTGTAATCGATAACCCACGATATACCCAACCCCTTCTTGC CCAAAGCAGC
>Porphyrio porphyrio poliocephalus_AMNH DOT17002
CGATATACCCAACCCCTTCTTGCCCAAAGCAGCCTACATACCGCCGTCCCCAGCTCACCTCC CCTGAGAGCCTAAATAGTGAGCACAACAACACCTCGCTAATAAGACAGGTCAAGGTATAGCCCAT GAAGGGGTAGAAATGGGCTACATTTTCTAAAATAGAAA
>Porphyrio porphyrio caledonicus
CGATATACCCAACCCCTTCTTGCCCAAAGCAGCCTACATACCGCCGTCCCCAGCTCACCT CCCCTGAGAGCCTAAATAGTGAGCACAACAACACCTCGCTAATAAGACAGGTCAAGGTATAGC CCATGAAGGGGTAGAAATGGGCTACATTTTCTAAAATAGAAA
>Porphyrio mantelli_NMNZ DM7930
GGATCATACCTCTATAAAGAAACCTGAAACACAGGAATCATCCTACTACTCACCCTAATA GCCACTGCCTTCGTAGGCTATGTCCTACCATGAGGACAAATATCCTTCTGAGGCGCTACAGTC ATTACAAACCTATTCTCAGCCATC