Molecular Phylogenetics and Evolution 83 (2015) 20–32 Contents lists available at ScienceDirect Molecular Phylogenetics and Evolution journal homepage: www.elsevier.com/locate/ympev A long distance dispersal hypothesis for the Pandanaceae and the origins of the Pandanus tectorius complex Timothy Gallaher a,⇑, Martin W. Callmander b,c, Sven Buerki d, Sterling C. Keeley a a University of Hawaii, Botany Department, 3190 Maile Way, Honolulu, HI 96822, USA Missouri Botanical Garden, P.O. Box 299, St. Louis, MO 63166-0299, USA c Conservatoire et Jardin botaniques de la Ville de Genève, ch. de l’Impératrice 1, case postale 60, 1292 Chambésy, Switzerland d Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3DS, UK b a r t i c l e i n f o Article history: Received 30 May 2014 Revised 28 October 2014 Accepted 3 November 2014 Available online 13 November 2014 Keywords: BEAST Dispersal Divergence times Gondwana Pacific biogeography RASP a b s t r a c t Pandanaceae (screwpines) is a monocot family composed of c. 750 species widely distributed in the Paleotropics. It has been proposed that the family may have a Gondwanan origin with an extant Paleotropical distribution resulting from the breakup of that supercontinent. However, fossils supporting that hypothesis have been recently reassigned to other families while new fossil discoveries suggest an alternate hypothesis. In the present study, nuclear and chloroplast sequences were used to resolve relationships among Pandanaceae genera. Two well-supported fossils were used to produce a chronogram to infer whether the age of major intra-familial lineages corresponds with the breakup of Gondwana. The Pandanaceae has a Late Cretaceous origin, and genera on former Gondwanan landmasses began to diverge in the Late Eocene, well after many of the southern hemisphere continents became isolated. The results suggest an extant distribution influenced by long-distance-dispersal. The most widespread group within the family, the Pandanus tectorius species complex, originated in Eastern Queensland within the past six million years and has spread to encompass nearly the entire geographic extent of the family from Africa through Polynesia. The spread of that group is likely due to dispersal via hydrochory as well as a combination of traits such as agamospermy, anemophily, and multi-seeded propagules which can facilitate the establishment of new populations in remote locations. Ó 2014 Elsevier Inc. All rights reserved. 1. Introduction 1.1. Pandanaceae lineages: distribution, dispersal modes, and keycharacter innovations Pandanaceae is a family of c. 750 dioecious species in the order Pandanales composed of trees, shrubs, epiphytes, and lianas. The family consists of five genera: Benstonea Callm. & Buerki (c. 60 spp.), Freycinetia Gaudich. (c. 250 spp.), Martellidendron (Pic. Serm.) Callm. & Chassot (6 spp.), Pandanus Parkinson (c. 450 spp.), and Sararanga Hemsl. (2 spp.) (Callmander et al., 2003, 2012). Molecular phylogenetic work supports the monophyly of each of the recognized genera, (Buerki et al., 2012; Callmander et al., 2003). In the most recent study based on chloroplast sequences, Abbreviations: BPP, Bayesian posterior probability; HPD, highest probability distribution; LDD, long distance dispersal; Ma, mega-annum; s.s., sensu stricto. ⇑ Corresponding author at: Iowa State University Department of Ecology, Evolution, and Organismal Biology, 251 Bessey Hall. Ames, IA 50011, USA. E-mail address: tjgallaher@gmail.com (T. Gallaher). http://dx.doi.org/10.1016/j.ympev.2014.11.002 1055-7903/Ó 2014 Elsevier Inc. All rights reserved. Sararanga was found to have a sister relationship to the remaining genera with Freycinetia as sister to a clade consisting of Martellidendron, Benstonea, and Pandanus (hereafter the MBP clade = AMcP clade of Buerki et al., 2012); however, relationships among these last three genera remain to be shown. Pandanus was previously classified into eight subgenera by Stone (1974). Two of these, Martellidendron and Acrostigma (now Benstonea), were elevated to generic rank (Callmander et al., 2003, 2012). The chloroplast phylogeny estimate of Buerki et al. (2012) recovered two subclades within Pandanus and their results suggest that up to five of the remaining subgenera recognized by Stone may not form monophyletic groups. The Pandanaceae is distributed throughout the Paleotropics from West Africa to the islands of the eastern Pacific (Fig. 1). With the exception of Martellidendron, which is endemic to Madagascar and the Seychelles, the genera have overlapping distributions centered on Malesia, a region which includes Indonesia, Borneo, the Philippines, and New Guinea. This region perhaps represents the most complex geological history on the planet due to the convergence of landforms of Gondwanan, Laurasian, and Pacific plate 21 T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 origin (Hall, 2002, 2009; Hill and Hall, 2002). Sararanga is narrowly distributed in the Philippines, the Solomon Islands, northern mainland New Guinea, and on the nearby Yepen and Manus islands (Stone, 1961). Freycinetia has a center of diversity in New Guinea and Indonesia and its distribution reaches Hawai‘i and Micronesia indicating that part of its range at least has been achieved through LDD (Stone, 1967a, 1968, 1969). Benstonea is distributed from India to Fiji with the majority of species found in Borneo (Callmander et al., 2012). While the first Pandanus subclade, ‘‘subclade 1’’ has a distribution similar to that of Benstonea, the second Pandanus subclade, ‘‘subclade 2’’ has achieved the widest extant distribution among Pandanaceae lineages reaching Africa, Madagascar, much of Southeast Asia, and nearly all of the islands of the tropical IndoPacific (Buerki et al., 2012; Stone, 1974). The biogeography of the family has been previously discussed under an assumption of an East-Gondwana origin with subfamilial lineages resulting from the breakup of the supercontinent during the Cretaceous (Callmander and Laivao, 2002; Callmander et al., 2003; Nadaf and Zanan, 2012; Stone, 1976a). However, there have been no explicit biogeographical analyses of the family and it is unclear what role Gondwana-derived vicariance or potentially recent long-distance-dispersal (LDD) have played in the historical biogeography of the family. Further, recent fossil discoveries, fossil reassignments, and available age estimates for the family cast some doubt on a Gondwana origin. The Gondwana supercontinent formed in the Southern Hemisphere prior to the Cambrian from the fusion of several continental blocks (see McLoughlin (2001) for a chronology of the Gondwana breakup summarized below). In the Early Jurassic, the supercontinent began to break up starting with rifts between West Gondwana and East Gondwana which became fully separated by 130 Ma. West-Gondwana split into South America and Africa with the two becoming isolated by 105 Ma. The breakup of East-Gondwana isolated Madagascar by 65 Ma and the Indian subcontinent was isolated from 65 Ma until it fused with Laurasia at c.43 Ma. Antarctica, and Australia finally separated by 35.5 Ma resulting in the set of southern hemisphere continents that we recognize today. The oldest Pandanaceae fossils are not found on Gondwanan landmasses, instead they are leaf fragments assigned to the genus Pandanites Tuzson from Maastrichtian (66.0–72.1 Ma) strata from North America and from Early Campanian (79–83.6 Ma) strata in Austria and Romania (Kvaček and Herman, 2004; Popa et al., 2011). Pandanites exhibit a suite of characters including lateral Legend Benstonea Freycinetia Martellidendron nerves resulting in an ‘M’-shaped leaf cross-section, apical drip tip, spiral leaf arrangement, marginal prickles, parallel veins, and tetracytic stomata which allies this taxon to the Pandanaceae (Kvaček and Herman, 2004). Several more recent fossils have affinities with Pandanaceae. Pandanus eocenicus Guleria & Lakhanpal from Early Eocene strata (41.2–56.0 Ma) of Gujarat, India may be Pandanaceae although the fossils lack the characters to place them in the family with certainty. Leaf fragments described as Pandanus helicopus Kurz. from the Neogene (2.6–23.0 Ma) of Sumatra are likely assignable to the Pandanaceae; however, like P. eocenicus, the taxon is not definitively assignable to Pandanus due to the lack of reproductive material (Kvaček and Herman, 2004). Two fossil pollen genera, Dryptopollenites Stover in Stover & Partridge and Lateropora Pocknall & Mildenhall are present in the fossil records of Australia and New Zealand beginning in the Paleocene (56–59 Ma) and are thought to have affinities with Freycinetia although this awaits a more detailed comparative analysis (Macphail et al., 1994). The fossil genus Megaporites Krutzsch described from Miocene (5.3– 23.0 Ma) Taiwan is also believed to have affinities with Freycinetia (Huang, 1980). Late Miocene to recent fossil pollen attributed to Pandanus has been found from throughout Southeast Asia and Islands of the Pacific (Jarzen, 1983; Leopold, 1969). Fossil impressions of P. tectorius fruit (>1.2 Ma) were found in a rejuvenation stage lava flow along the north coast of the island of Kauai, Hawaii with subsequent Holocene Pandanus pollen and macro sub-fossil specimens found in deposits along the south coast of that island (Burney et al., 2001; Burney, 2002). Several other fossil taxa previously assigned to the Pandanaceae from Gondwana-derived continents including Pandanaceoxylon Patil & Datar from Cretaceous (66.0–145.0 Ma) India and Viracarpon Sahni and Pandanusocarpon Bonde from the Eocene (33.9– 56.0 Ma) of India are not attributable to the Pandanaceae due to a lack of synapomorphic characters and these have been assigned to other extant families (Bonde, 2005; Cox et al., 1995; Herman and Kvaček, 2010). Age estimates for the family are incongruous. Janssen and Bremer (2004) estimated the origin of the Pandanaceae at 98 Ma using a plastid sequence-based phylogeny estimate of the monocots and a calibration based on the crown age of the monocots as calculated by Bremer (2000). However, in an analysis based on 36 fossil calibration points scattered throughout the angiosperm lineage, Bell et al. (2010) calculated the age of the family at Pandanus subclade 1 Pandanus subclade 2 Sararanga Cyclanthaceae Fig. 1. Natural extant distributions of the genera of the Pandanaceae, including the two Pandanus subclades, and the Cyclanthaceae. Red, Benstonea Callm. & Buerki; Dark green, Freycinetia Gaudich.; Light green, Martellidendron (Martelli) Callm. & Chassot; Blue, Pandanus Parkinson subclade 1; Purple, Pandanus subclade 2; Orange, Sararanga Hemsl.; Black, Cyclanthaceae Poit. ex A. Rich. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 22 T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 47 Ma (31–62 Ma 95% HPD) or 52 Ma (36–71 Ma) depending on whether an exponential or lognormal relaxed clock algorithm was implemented. There have been no published age estimates for the genera or intra-generic taxa of the Pandanaceae. In order to support a Gondwana vicariance diversification pattern for the family, Pandanaceae lineages that occupy the former Gondwanan continents would have to have divergence dates coinciding in time with the breakup of that continent into the modern continental landforms. An alternate explanation for the current distribution of the family would be LDD to Gondwana derived continents following the breakup of the supercontinent. This is particularly plausible given the age and distribution of fossils and the fact that several lineages within the Pandanaceae have evolved characters which facilitate both LDD and the successful establishment of new populations in remote locations. Different fruit types have evolved in the Pandanaceae to take advantage of zoochory, hydrochory (flotation), or a combination of these two dispersal modes (Fig. 2). A variety of animals have been observed to disperse the fruit of several species and others have fruit that can float on seawater and remain viable for months (Nakanishi, 1988; Stone, 1976a). Sararanga and Freycinetia have multi-seeded fruit composed of sweet, fleshy, animal-dispersed berries. While seed dispersal has not been reported for Sararanga, the fruit of Freycinetia is known to be dispersed by birds and bats (Cox, 1990; Ingle, 2003). The fruit of the MBP clade on the other hand are single or multi-seeded drupes. The fruit of Martellidendron each contain two seeds while all species of Benstonea have fruit composed of single-seeded drupes. The fruit of Pandanus are either single-seeded drupes or multi-seeded polydrupes, with the later condition believed to result from fusion of adjacent singleseeded floral units. The fruit of MBP clade species have a fleshy proximal mesocarp that attracts animals as well as, in many species, aerenchyma, often concentrated in the distal mesocarp, which allows the fruit to float on water. The aerenchyma tissue is absent or poorly developed in Benstonea and in some Pandanus (i.e. P. subg. Kurzia) and subsequently the fruit of those species are poor floaters and are presumed to be primarily animal-dispersed (Stone, 1976a). Several characters promote establishment following dispersal. For example, the multi-seeded propagule increases the probability of both a male and female establishing in a single dispersal event (Cox, 1985). Asexual reproduction via agamospermy, which would allow a single founding female to establish a population, has been documented in P. tectorius and anecdotal accounts of isolated trees of several species bearing fertile fruit suggest that this trait may be widespread in Pandanus (Cox, 1985; Stone, 1972). The report of supernumerary nuclei in the embryo sac of Benstonea parva (Ridl.) Callm. & Buerki suggests that the condition may be found in that genus as well (Cheah and Stone, 1975). Agamospermy has not been reported in Sararanga, Freycinetia, or Martellidendron; however, some species of Freycinetia have been observed to occasionally produce both male and female inflorescences or bisexual flowers and self-pollination has been documented in at least one case (Huynh and Cox, 1992; Poppendieck, 1987). Anemophily which is believed to occur in Martellidendron, Benstonea, and Pandanus eliminates dependence on animal pollinators (Cox, 1990). The inflorescence of Freycinetia allows for a potentially diverse array of animal pollinators including birds, bats, and other small mammals while Sararanga may rely on specific flying insects for pollination (Cox, 1990; Lord, 1991). 1.2. The Pandanus tectorius complex The most widespread group in the Pandanaceae, and the most problematic in terms of taxonomy, is Pandanus subg. Pandanus section Pandanus (within Pandanus subclade 2). Over 300 species have been ascribed to this section. Many of these have been synonymized with two widespread coastal species, P. tectorius Parkinson and P. odoratissimus L.f., based on the observation that a great deal of the morphological variation used to distinguish between species within the section can often be observed within a single population (Fosberg, 1977; Stone, 1967b, 1982). As a result of this taxonomic uncertainty, this group has been termed the Pandanus tectorius complex (Stone, 1976b). This lineage is distributed from east Africa to east Polynesia, nearly encompassing the entire geographic range of the family. Many of the included taxa are coastal although several inland species are included in this group (Huynh, 1980; Stone, 1974). Fruit of many of these species can float in water for months and the multiple seeds contained within each propagule remain viable (Nakanishi, 1988). The fruit of species of subg. Pandanus are eaten by animals including crabs (Lee, 1985), tortoises (Hnatiuk, 1978), cassowary (Ramsey, 2010), rats (McConkey et al., 2003), and fruit bats (Ash, 1987; Wiles et al., 1991). This may allow for dispersal inland and away from coastal populations. Further, the subgenus includes species with confirmed agamospermy, vegetative reproduction, and wind pollination (Carlquist, 1967). Since the composition of the P. tectorius complex is debated and its relationship to other sections has not been tested, biogeographic patterns of this most widespread ocean-dispersed lineage remain obscure. In order to infer whether the age of Pandanaceae lineages supports a vicariance hypothesis from a Gondwanan origin, chloroplast and nuclear molecular markers were used along with fossil data to reconstruct an age-calibrated phylogeny estimate of the Pandanaceae. Samples were selected to represent each of the Pandanaceae genera and Pandanus subgenera as recognized by Stone (1974, 1978, 1983) including the later updates by Callmander et al. (2003, 2012). Samples of three species from the Cyclanthaceae, shown in previous studies (Chase et al., 2006; Duvall et al., 1993; Tamura et al., 2004) to be the sister family to the Pandanaceae, were selected as the outgroup. In a second analysis, intensive sampling of taxa attributed to the Pandanus tectorius complex, along with species found to be closely related to the complex in the first analysis, were used to infer the origins and composition of the widespread, ocean-dispersed complex. 2. Materials and methods 2.1. Data generation Genomic DNA was extracted from desiccated leaf tissue using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocols with the lysis phase increased from 5 to 120 min. Three regions of the chloroplast: trnL-trnF, ndhF-rpl32, and the trnQ-rps16 intergenic spacers; and two nuclear regions: the FLORICAULA/LEAFY second intron (LFY) and a coding portion of the phytochrome C gene (PHYC) were amplified. PCR parameters and primer references are given in Supplementary Table 1. Amplifications were purified with EXO-SAP (Affymetrix, Santa Clara, California, USA) and sequenced on a 3730XL DNA Analyzer (Applied Biosystems, Carlsbad, California, USA). In total, sequences from 126 samples representing 75 species were newly generated for this study. Sample information including herbarium vouchers, Genbank accession numbers, and the distributional information for the species represented by the samples are presented in Supplementary Table 2. The nuclear regions for some samples indicated heterozygous sequences indicated by two strong peaks at up to six base-pair positions per sequence for the LFY intron and up to four positions per sequence for PHYC. These positions were coded as missing data for the subsequent analyses. DNA contigs were assembled using T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 a d f 23 c b e g Fig. 2. Reproductive diversity in the family Pandanaceae. (a) Paniculate inflorescence of multi-seeded berries of Sararanga sinuoa Hemsl.; (b) spikes of multi-seeded berries of Freycinetia arborea Gaudich.; (c) multi-headed infructescence of single seeded drupes of Benstonea epiphytica (Martelli) Callm. & Buerki; (d) solitary head of 2-seeded drupes of Martellidendron hornei (Balf. f.) Callm. & Chassot; (e) solitary head of multi-seeded drupes of Pandanus tectorius Parkinson; (f) staminate inflorescence with several spikes of Pandanus tectorius Parkinson; (g) multi-headed infructescence of single seeded drupes of Pandanus polycephalus Lam. Photographs: (a) T. Gallaher; (b) P. Lowry. (c–g) M. Callmander; Vouchers: (a) Gallaher 272; (b) Callmander & al. 846; (c) Callmander & al. 1014; (d) Callmander & Meystre 94; (e) Callmander & al. 957; (f) Callmander & Gallaher 920; (g) Callmander 984. 24 T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 Sequencher 4.9 (Gene Codes Corp., Ann Arbor, Michigan, USA) and sequence alignments were produced using MUSCLE (Edgar, 2004) as implemented in MEGA 5.0 (Tamura et al., 2011). Alignments were manually adjusted to maximize homology and microsatellite loci with variable numbers of repeats were removed from the data matrix. The first and second base pairs of each codon of PHYC were coded as a separate partition from the 3rd base pair to allow greater flexibility in modeling the higher mutation rate in the third codon position. The best-fit model of molecular evolution and model parameters for each partition was determined using the sample size corrected Akaike information criteria calculated for 88 models using the Maximum Likelihood optimized base-tree and ‘‘Best’’ tree-search option in jModelTest 2.1.4 (Darriba et al., 2012; Guindon and Gascuel, 2003) (see Supplementary Table 1). 2.2. Analysis 1: dated phylogeny estimate of the Pandanaceae The first analysis included 56 samples including three Cyclanthaceae taxa and representative taxa from each of the Pandanaceae genera: three species of Benstonea, one species of Martellidendron, five species of Freycinetia, both Sararanga species, and 42 taxa collectively representing all Pandanus subgenera. Maximum likelihood inference was done using RAxML-HPC2 (XSEDE) (Stamatakis, 2006) on the CIPRES Science Gateway (Miller et al., 2010). Species of Cyclanthaceae were set as the outgroup taxa and gene partitions were specified. The analysis used the GTRGAMMA model for tree inference and node support was estimated with 5000 bootstrap replicates and displayed on the best-scoring tree. In addition to the combined dataset, the analysis was repeated separately for the combined chloroplast regions and the combined nuclear regions in order to compare phylogenetic signal between these two genomes. Congruence between the phylogenetic signals from each genome was assessed visually. Mega 5.2.1 (Tamura et al., 2011) was used to test for a strict molecular clock assuming the GTR + G + I model and the maximum likelihood tree. The strict molecular clock was rejected (p = 0) and the lognormal relaxed clock prior was used in subsequent analyses. Bayesian analysis and node age inference were conducted using BEAST 1.8 (Drummond et al., 2012). In order to record and evaluate age estimate statistics, well-supported clades recovered in the maximum likelihood analysis were specified in BEAUti 1.8, but not constrained to monophyly. Substitution and clock models were unlinked for each partition. The Birth–Death tree prior was used given the assumption that significant levels of extinction have likely occurred within lineages (Yang and Rannala, 1997). The XML output from BEAUti was modified to incorporate the substitution models selected using jModelTest2. The BEAST analysis was run for 40,000,000 generations with trees sampled every 1000 generations. For each analysis two independent runs were done, each with random starting seeds. Upon completion of each BEAST run, the resulting log files were inspected with Tracer 1.5 (Rambaut and Drummond, 2011) to ensure that all parameters had reached a stable value, that effective sample sizes were greater than 200, and that separate runs had converged on tree topologies and node age estimates. The first 10,000 trees were discarded, resulting in 30,000 trees. The maximum clade credibility tree was selected and the posterior probability of sampled trees was annotated onto the tree along with median node heights in TreeAnnotator 1.7.5 (Drummond et al., 2012). 2.2.1. Fossil constraints Two independent age calibrations were applied to the BEAST analysis. Due to the sparse fossil records of Pandanaceae and Cyclanthaceae, uniform rather than lognormal priors were used. While the lognormal prior assumes that the actual age of a clade is likely to be slightly older than the fossil ages, the uniform prior makes no assumption about the age of the clade within a predefined upper and lower bound. The first calibration was a uniform prior range of 45–150 Ma placed on the crown node of the Cyclanthaceae. The early age corresponds to the estimated minimum age of well-preserved fossil infructescences and seeds from Germany and England assigned to the genus Cyclanthus Poit. ex A. Rich. (Smith et al., 2008), while the upper value was set to a date older than recent estimates of the Monocot-Eudicot split (Bell et al., 2005; Chaw et al., 2004; Janssen and Bremer, 2004). Previous studies identified Cyclanthus as the sister to all other extant Cyclanthaceae (Eriksson, 1994; Mennes et al., 2013) allowing this constraint to be placed on the crown node of the sampled Cyclanthaceae rather than the stem node. The second age constraint, a uniform prior range of 79–150 Ma, was applied to the root age of the tree. This constraint is based on minimum age estimates of Austrian Pandanaceae leaf fossils assigned to the extinct genus Pandanites (Kvaček and Herman, 2004). Since the relationship between Pandanites and extant genera is not known, the fossil constraint was placed on the common ancestor of the family. The fossil pollen taxon Dryptopollenites was not used as a calibration point due to uncertainty in the literature as to the precise placement of this taxon within Pandanaceae (Macphail et al., 1994). The analysis sequence was run three times; the first with only the Cyclanthaceae fossil calibration, the second with only the Pandanaceae fossil calibration and the third with both calibration points included. 2.3. Analysis 2: the origins and composition of subg. Pandanus s.s A second analysis was performed to investigate the origins and taxonomic composition of the P. tectorius complex. This was done as a separate analysis in order to avoid asymmetric over-sampling of subg. Pandanus within the family-level phylogeny which could lead to improper age estimates. Pandanus whitmeeanus in Pandanus subclade 1 was used as the outgroup to 92 samples representing 43 putative taxa from Pandanus subclade 2. The ingroup taxa included nine species of Pandanus subg. Vinsonia (Warb.) B.C. Stone and three samples of P. dubius Spreng. which had been reconstructed as the sister taxon to the Australian species of subg. Pandanus, including P. tectorius, in analysis 1. Multiple accessions of each of the recognized Australian species of subg. Pandanus including multiple samples assigned to the P. tectorius complex from throughout its geographical range were included in the analysis. In addition to the five DNA partitions, a binary partition coded with the presence/absence of indels was included in this analysis. The indel partition was coded manually with a gap present in any sequence in the alignment given a score of 0 and an insertion given a score of 1. This resulted in a partition of 10 characters. Variable microsatellite repeats were removed prior to analysis and were not considered in the scoring of indels. The ‘‘simple’’ binary substitution model implemented in BEAST was used for the indel partition. Other parameters of the BEAST 1.8 analysis were similar to that in analysis 1 except that no age calibrations were applied and the analysis was run for 60,000,000 generations with trees sampled every 1000 generations. The first 15,000 trees were discarded. The set of trees produced in analysis 2 was used in an ancestral area reconstruction for all nodes with greater than 0.7 BPP using the Bayesian binary MCMC analysis implemented in RASP 2.1 (Yu et al., 2013). All samples were classified into one of five areas based on the collection locality (or the original collection locality for cultivated specimens) (1) Africa/Madagascar, (2) WA/NT Australia, (3) Eastern Queensland Australia, (4) the coastal tropical Indo-Pacific, and (5) the inland tropical Indo-Pacific. The last two areas comprise the coastal and inland portions respectively of the region which stretches from tropical India to the eastern Pacific exclusive of the previously defined areas. The outgroup, representing T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 P. subclade 1 was coded with distribution 4 and 5. Run parameters included 250,000 cycles, and 10 chains sampled every 100 generations. The first 500 samples were discarded. The F81 model for state frequencies was used with a gamma among-site rate variation. Following the analysis, the distance between runs 1 and 2 was assessed and found to be below 0.01 (distance of run 1 and run 2 = 0.0003). Results were annotated onto the phylogeny. 3. Results 3.1. Analysis 1: dated phylogeny estimate of the Pandanaceae The nuclear, chloroplast, and combined analysis strongly support the monophyly of the five described Pandanaceae genera (Fig. 3). Sararanga is sister to all other Pandanaceae and Freycinetia is inferred to be the sister taxon to the MBP clade (both BPP 1.0, 100% bootstrap), consistent with the results of Buerki et al. (2012). Further, Martellidendron is supported as the sister taxon to Benstonea + Pandanus (BPP 1.0, 100% bootstrap in the combined analysis). There was no substantial conflict between the chloroplast and nuclear genomes (congruent topology and BPP > 0.9, >80% bootstrap for major clades, see Supplementary Fig. 1). However, bootstrap support for the sister relationship between Benstonea 25 and Pandanus was low (26%) in the chloroplast analysis and high (98%) in the analysis of the nuclear partitions and combined analysis. The analysis recovered two Pandanus subclades consistent with the results of Buerki et al. (2012). Pandanus subclade 1 includes species from P. subg. Coronata Martelli, P. subg. Lophostigma (Brongn.) H. St. John, P. subg. Kurzia B.C. Stone, and P. subg. Rykia (de Vriese) B.C. Stone. In addition, subclade 1 was also found to include species previously assigned to P. subg. Pandanus. Pandanus subclade 2 includes a monophyletic P. subg. Vinsonia that is sister to a clade (hereafter P. subg. Pandanus sensu stricto) consisting of P. dubius (formerly assigned to P. subg. Rykia) plus the Australian species of P. subg. Pandanus and members of the widespread P. tectorius clade. Node age estimates were similar among analyses relying on either the fossil Cyclanthus, fossil Pandanaceae or both fossils (Table 1). All analyses inferred a root age with the highest posterior density (HPD) centered in the Late Cretaceous, albeit with a wide 95% HPD between 49.0 and 132.8 Ma depending upon the analysis. In all three analyses, the ages of the constrained nodes were close to the minimum age estimates of the two fossils indicating that the lower bounds of the fossil calibrations had a strong effect on age estimation whereas the upper bounds were not as influential. Fig. 3. Time-calibrated Bayesian phylogeny estimate of the Pandanaceae inferred with BEAST 1.8. ML bootstrap support and Bayesian posterior probabilities (BPP) are given along branches. Only probabilities greater than 0.7 are displayed. The blue bars represent the 95% highest probability density (HPD) interval of age estimates. Nodes are situated at the median age values. For the two nodes with a truncated HPD due to a fossil constraint, the age of the node was adjusted to the peak of the HPD rather than the median age. Geologic periods based on the GSA Geologic Time Scale (v. 4.0) (Walker et al., 2013). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 26 T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 Table 1 Median node age and 95% highest posterior density (HPD) estimates in millions of years for all three permutations of the 56-sample family-level analysis (Analysis 1). For constrained nodes (indicated by an asterisks) the peak of the HPD estimate is reported rather than the median value. Node Analysis Median and 95% highest posterior density (Ma) Root age Cyclanthus fossil only Pandanaceae fossil only Both fossils 86.0 81⁄ 85⁄ (49.0–139.3) (79–121.5) (79–132.8) Cyclanthaceae crown Cyclanthus fossil only Pandanaceae fossil only Both fossils 45⁄ 42.3 45⁄ (45.0–72.7) (22.9–66.6) (45–72.1) Pandanaceae crown Cyclanthus fossil only Pandanaceae fossil only Both fossils 58.0 60.0 65.0 (45.5–101.0) (38.3–89.7) (41.9–97.5) Freycinetia/MBP split Cyclanthus fossil only Pandanaceae fossil only Both fossils 34.7 36.1 39.3 (16.1–60.3) (22.2–54.9) (24.6–60.7) Martellidendron/B + P split Cyclanthus fossil only Pandanaceae fossil only Both fossils 21.5 22.3 24.3 (10.0–37.1) (13.4–33.9) (15.1–38.1) Benstonea/Pandanus split Cyclanthus fossil only Pandanaceae fossil only Both fossils 18.7 19.3 21.1 (8.6–32.5) (11.6–29.7) (12.5–33.0) Sararanga crown Cyclanthus fossil only Pandanaceae fossil only Both fossils 6.1 6.4 7.0 (2.3–12.3) (2.7–11.7) (3.1–12.7) Pandanus crown Cyclanthus fossil only Pandanaceae fossil only Both fossils 11.5 11.8 12.9 (5.4–19.9) (6.8–18.0) (7.7–20.2) Pandanus subclade 1 crown age Cyclanthus fossil only Pandanaceae fossil only Both fossils 10.2 10.5 11.5 (5.0–17.9) (6.1–15.9) (6.7–17.8) Pandanus subclade 2 crown age Cyclanthus fossil only Pandanaceae fossil only Both fossils 7.3 7.4 8.1 (3.1–12.8) (4.1– 11.7) (4.5–13.1) Subg. Pandanus s.s. crown age (incl. P. dubius) Cyclanthus fossil only Pandanaceae fossil only Both fossils 4.6 4.7 5.3 (2.0–8.3) (2.6–7.6) (2.8–8.6) Australian species of subg. Pandanus crown age Cyclanthus fossil only Pandanaceae fossil only Both fossils 3.1 3.1 3.5 (1.3–5.6) (1.7–5.1) (1.9–5.7) 3.2. Analysis 2: the Pandanus tectorius complex The 93 sample analysis of Pandanus subclade 2 recovered several well supported clades (Fig. 4). Pandanus sub. Vinsonia is supported as sister to P. subg. Pandanus s.s. (1.0 BPP). The three samples of P. dubius were supported (1.0 BPP) as sister to an unresolved polytomy of four clades: (1) a single sample of P. basedowii C.H. Wright from the Northern Territory Australia; (2) a group of three species, P. spiralis R. Br. (with numerous recognized varieties), P. aquaticus F. Muell., and P. rheophilus B.C. Stone found in Western Australia, and the Northern Territory and extending into western Queensland; (3) a group of three species, P. solms-laubachii F. Muell., P. gemmifer H. St. John, and a yet undescribed species, P. sp. ‘‘Russell River’’ from east Queensland, and (4) a lineage which includes P. cookii Martelli, and P. oblatus H. St. John from eastern Queensland. Nested within this final clade are P. tectorius and its allies. The Bayesian binary analysis inferred the ancestor of Pandanus subclade 2 to have occurred within the coastal Indo-Western Pacific (79.7%) exclusive of Africa/Madagascar, and Australia, or Africa/Madagascar (16.3%). The ancestor of P. subg. Pandanus s.s. (including P. dubius) is also inferred to have originated in the Indo-West Pacific (95.3%). The ancestor of the Australian clade which includes the P. tectorius complex is inferred to have occurred within Australia (78.2%: WA/NT 25.0%, Queensland 53.2%, or within the coastal Indo-Pacific (20.0%). The ancestor of the P. tectorius complex plus P. oblatus and P. sanderi Sander, and that group plus P. cookii was inferred to have evolved in east Queensland (99.9% for both). The ancestor of the complex, exclusive of P. oblatus and P. sanderi, was inferred to have evolved within the coastal Indo-Pacific (88.1% or in Queensland 11.8%). Dispersal back to Australia was inferred to have occurred on up to two occasions leading to P. conicus H. St. John and P. tectorius populations in coastal Queensland. Dispersal to Africa occurred on at least one occasion leading to P. kirkii Rendle and up to three dispersals inland into islands of the Indo-Pacific from the coastal Indo-Pacific were inferred. 4. Discussion 4.1. Origins and diversification of the Pandanaceae The family level analysis supports the generic relationships recovered by Buerki et al. (2012) and the inclusion of nuclear markers resolved relationships among Martellidendron, Benstonea, and Pandanus. There is sufficient resolution in the combined nuclear and chloroplast dataset to confirm that at least three of the six remaining Pandanus subgenera recognized by Stone (1974): P. subg. Pandanus, P. subg. Kurzia, and P. subg. Rykia are not monophyletic and the phylogeny estimate suggests that P. subg. Lophostigma is likewise not monophyletic. Only P. subg. T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 27 Fig. 4. Bayesian Phylogeny estimate of Pandanus subclade 2 focusing on Pandanus subg. Pandanus. Ancestral geographic areas were reconstructed for all nodes with BPP greater than 0.7 using Bayesian Binary analysis in RASP 2.1. Background and branch colors indicate geographic area: Purple, Madagascar, Africa, and Indian Ocean islands; Pink, Western Australia and Northern Territory; Green, eastern Queensland; Blue, coastal Indo-Pacific (India-Hawaii excl. Australia); Grey, inland Pacific islands; Black, sample coded as both coastal Indo-Pacific and inland Pacific islands. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Vinsonia remains a monophyletic group in all analyses completed as of this study (Buerki et al., 2012). The two-fossil analysis inferred that the split between Pandanaceae and Cyclanthaceae occurred in the early to late Cretaceous (79–132.8 Ma). This result could be consistent with a vicariance event between west and east Gondwana, or between the South American and African continents; however, the occurrence of fossils from both families in Laurasia complicates that scenario and allows for an origin for both families in Laurasia. Pandanaceae lineages on the former Gondwanan landmasses of Africa, Madagascar, and Australia are inferred to have split from their sister groups well after the geologic isolation of each of those continents. This suggests that the distributions of major lineages of the Pandanaceae are due to LDD and fails to support a distribution due to Gondwana vicariance. Further, the congruence between the estimated stem age of the Pandanaceae and the age of the oldest fossils assigned to the family in Europe and North America suggests a Laurasian origin; however, a Gondwanan origin cannot be rejected based on this analysis. The sequence of events suggested here for the Pandanaceae is perhaps similar to that of the Sapindaceae which originated in Laurasia and then dispersed to South East Asia which served as a staging area for subsequent dispersals to its extant worldwide distribution (Buerki et al., 2011). The Pandanaceae may therefore join a growing group of angiosperm 28 T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 lineages once thought to have a distribution explained by the breakup of Gondwanan but which were subsequently found to have colonized former Gondwana landmasses via LDD (Birch and Keeley, 2013; Cook and Crisp, 2005; Davis et al., 2002). Sararanga diverged from the remaining Pandanaceae lineages in the Late Cretaceous to Early Eocene 65.0 Ma (41.9–97.5 Ma 95% HPD) and the two extant species of Sararanga diverged from each other in the Late Miocene to Pliocene 7.0 Ma (3.1–12.7 Ma 95% HPD). The extant distribution of Sararanga is restricted to landmasses of Pacific-plate origin which likely emerged in the Eocene as islands at the margins of the Pacific plate and have had limited contact with Laurasian or Gondwana derived areas since that time until the Miocene (Hall, 2002). The restricted distribution and low present day species diversity in Sararanga may result from a lack of the characters (discussed below) which appear to facilitate successful establishment in the remaining Pandanaceae lineages. Alternatively, extinction may account for the distribution of Sararanga with extant species representing relicts of a once more widespread lineage. Freycinetia split from its common ancestor with the MBP clade in the Late Paleocene to Oligocene 39.3 Ma (24.6–60.7 Ma 95% HPD). This 95% confidence interval includes the first occurrences of fossil pollen Dryptopollenites in Australia thought to have affinities with Freycinetia (Macphail et al., 1994; Stoian, 2002). The potential for self-pollination in some species, the copious number of seeds per fruit characteristic of Freycinetia, and the ability to attract an array of vertebrate pollinators may have facilitated establishment in remote locations (Huynh and Cox, 1992; Lord, 1991; Poppendieck, 1987). Greater sampling of Freycinetia and a better understanding of the extent of bisexuality in the genus is required to understand the origins and spread of this group (Cox, 1981). The Martellidendron lineage diverged from its common ancestor with Benstonea and Pandanus in the Late Eocene to Early Miocene, 24.3 Ma (15.1–38.1 Ma 95% HPD). This divergence is well after the isolation of Madagascar from the Seychelles and India which occurred in the Late Cretaceous. In addition, the climate of Madagascar was mostly arid prior to the Late Eocene, after which time much of the continent entered a period of stable wet-tropical to subtropical climate more suitable for Pandanaceae (Yoder and Nowak, 2006). Given the presence of ample aerenchyma positioned between the longitudinal mesocarp fibers in the fruit of extant species and that the fruit are too large to be carried long distances by birds, the lineage may have colonized Madagascar through long-distance ocean-dispersal. In lab conditions, the fruit of Martellidendron hornei (Balf. f.) Callm. & Chassot remain buoyant in fresh water for several months which would allow time for fruit to cross the Indian Ocean (Gallaher, unpublished data). By the middle Oligocene, oceanic currents were established which move ocean surface water from the Indonesian throughflow westward into the Indian Ocean (Gourlan et al., 2008; Kuhnt et al., 2004). The biogeographic history of Martellidendron may represent a close parallel to the Chuniophoeniceae palm tribe which had its origins in Eurasia at the Miocene/Oligocene boundary, subsequently diversified in Malesia, and reached Madagascar through LDD in the Miocene (Baker and Couvreur, 2012). Benstonea diverged from Pandanus in the late Oligocene or early Miocene 21.1 Ma (12.5–33.0 Ma 95% HPD). The extant distribution of Benstonea suggests an origin on Asiatic landmasses of the Sundashelf region; however, there are not enough species sampled here to test that hypothesis. The fusing of floral units to form multi-seeded propagules in Pandanus may promote successful establishment resulting in the wider distribution of that genus relative to the single seeded Benstonea. This character appears to be quite variable in Pandanus with several likely reversals, such as in the single-seeded P. aquaticus which is nested within a group of species with polydrupes (Pandanus subg. Pandanus s.s. section Pandanus). All members of the MBP clade may be wind pollinated as was demonstrated with P. tectorius; however, pollination studies are lacking outside of that species. The two Pandanus subclades diverged from one another in the Miocene, 12.9 Ma (7.7–20.2 Ma 95% BPP). Subclade 1 has a distribution primarily in Southeast Asia and the islands of the Western Pacific. The similarity to Bentonea in the geographic range of this group may reflect their shared dispersal mode. Many of the sampled species in subclade 1, such as species assigned to Pandanus subg. Kurzia, are animal-dispersed or may be dispersed short distances over water (Stone, 1976a). Only one species within this lineage, P. polycephalus Lam., is suspected of ongoing oceandispersal and it has a relatively wide distribution from the Moluccas Islands, Indonesia eastward to the Solomon Islands (Guppy, 1906; Stone, 1976a). Other species including P. whitmeeanus Martelli distributed on islands of the south Pacific, P. forsteri Moore & Muell.on Lord Howe Island and P. kanehirae Martelli on Palau, have large multi-seeded fruit and well-developed aerenchyma. It is unlikely that these species would have been dispersed by birds over long distances and they may have reached their current distribution by flotation. Pandanus subclade 2 has achieved the widest distribution among Pandanaceae lineages. This group accounts for much of the diversity on former Gondwana landmasses, yet the crown age of the group (4.5–13.1 Ma 95% HPD) is much too recent to suggest a vicariance distribution due to the breakup of the supercontinent; rather, dispersal is implicated in the distribution of this clade. Zoochory and ocean-dispersal are important dispersal modes in this group and some species utilize both dispersal modes (Stone, 1976a). The occurrence of ocean-dispersal in several species within each of the subclade 2 sub-lineages (subg. Vinsonia and subg. Pandanus s.s.) suggests that their common ancestor may have been ocean-dispersed. Pandanus subg. Vinsonia, with over 100 species, is distributed in Madagascar, Africa, and the islands of the western Indian Ocean while subg. Pandanus s. s., with a center of diversity in the western Pacific, is broadly distributed throughout the Paleotropics, but is notably absent from Madagascar and is restricted to the coast in Africa. The cladogenesis event separating Vinsonia from subg. Pandanus s.s. suggests isolation between ocean basins. In the case of Vinsonia, the common ancestor likely dispersed to Madagascar and Africa from an origin in the western Pacific through the Indonesian throughflow. Although these prevailing currents have been operating almost continuously since the Oligocene, between 15–17 Ma and 3.4–5.5 Ma, geologic activity may have effectively closed the Indonesian seaway leading to periods of temporary isolation of the Pacific and Indian Oceans (Martin and Scher, 2006; Nishimura and Suparka, 1997). Vinsonia has diversified with the aid of animal dispersal in Madagascar and Africa while some species retain flotation as an important dispersal mechanism (Stone, 1976a). Pandanus subg. Pandanus s.s. includes P. dubius, which is sister to the Australian species of subg. Pandanus including members of the P. tectorius complex (Fig. 5). Species formerly assigned to P. subg. Pandanus in the Philippines, Lord Howe Island, and Palau, once thought to be allied to the P. tectorius complex, were found to be not closely related to the widespread complex; rather, these species were inferred to be members of Pandanus subclade 1. The ocean-dispersed P. dubius, previously assigned to subg. Kurzia from the western Pacific, and its allies P. borneensis Warb. of Borneo and P. elatus Ridl. of Christmas Island (Indian Ocean) (Buerki et al., 2012) have a combined distribution which straddles the divide between the Indian and Pacific Ocean. Although it contains ocean-dispersed members, unlike its sister group, the lineage has not spread widely throughout the Indo-Pacific. 29 T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 a c b d e f g h i j k l m n o Fig. 5. Diversity within Pandanus subgenus Pandanus s.s.: (a) Pandanus dubius Spreng., (b) P. basedowii Wright, (c) P. aquaticus Muell., (d) P. rheophilus Stone, (e) P. spiralis R. Br., (f) P. gemmifer St. John, (g) P. solms-laubachii Muell., (h) P. cookii Martelli, (i) P. oblatus St. John, (j) P. tectorius Parkinson, (k) P. tectorius Parkinson, (l) P. odoratissimus L. f., (m) P. conicus St. John, (n) P. arapepe St. John, (o) Pandanus sp. ‘‘Makatea’’. Photographs: (a, c, d, f–m) T. Gallaher; (b and e) R. Barrett, (n) G. McCormack; (o) J-F. Butaud. 4.2. Adaptive radiations in Australia and the origins of the Pandanus tectorius complex The subg. Pandanus s.s. lineage arrived on the Australian continent in the Late Miocene to Pliocene, 3.5 Ma (1.9–5.7 Ma 95% HPD) and subsequently diversified into two Western Australia/Northern Territory lineages and two lineages in eastern Queensland. The Western Australia/Northern Territory lineages include four species that occupy distinct habitats. The sandstone endemic, P. basedowii is unique among the Australian Pandanus in that it has light fruit 30 T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 which may be wind-dispersed (Wright, 1930). Pandanus spiralis is distributed on the open savannah near ephemeral water sources and forms a clade with two species, P. aquaticus and P. rheophilus, which are adapted to persistent flooded conditions such as would be found in a swamp or billabong (Wilson, 2011). Although these last three species are morphologically and ecologically distinct, there is little genetic differentiation between these taxa indicating a very recent divergence and/or ongoing gene flow. The two eastern Queensland clades likewise include species that are found in distinct habitats indicating ecological speciation. The first clade includes P. solms-laubachii which occupies swamp habitats and two species, P. gemmifer and an undescribed species known as ‘‘Russell River’’, which inhabit the banks of large rivers. The second clade includes P. cookii, found on savannah near ephemeral water sources, P. oblatus, found in the tropical rainforests of coastal northeast Queensland, P. sanderi, a sterile ornamental species first collected from Timor, and a clade consisting of the P. tectorius complex. In Australia, P. tectorius is strictly a coastal species with a distribution along the east coast of Queensland from the tip of the Cape York Peninsula to New South Wales. Pandanus conicus, which was found to be nested within the P. tectorius complex, is found in tropical coastal forests in eastern Queensland. From an eastern Queensland origin and within the past six million years, the P. tectorius complex has dispersed to areas throughout the Indo-Pacific from coastal Africa to Hawaii and the South Pacific. In several cases dispersal from the coast to inland localities has occurred. Several morphologically distinct taxa including P. odoratissimus, P. arapepe H. St. John, P. boninensis Warb., and an undescribed species from Makatea Island in the Tuamotu Archipelago are nested within the complex resulting in a paraphyletic P. tectorius. The pattern of distinct species nested within an otherwise widespread species does not necessarily call into question their distinctiveness. In this case, coastal metapopulations of P. tectorius from throughout the Indo-Pacific may continue to be linked by regular gene flow while inland populations derived from coastal populations and coastal populations on remote islands out of the pathway of major ocean-currents may become isolated and evolve into distinct species. 5. Conclusions The divergence time estimates for intra-familial lineages within the Pandanaceae fails to support a Gondwanan origin and suggests an important role for long-distance-dispersal throughout the evolutionary history of the family. Long-distance-dispersal rather than vicariance is the most likely explanation for the distribution of Pandanaceae on the former Gondwanan landmasses of Madagascar, Africa, and Australia. This analysis suggests that LDD to the western Indian Ocean occurred on up to three independent occasions, including the dispersals to Madagascar by Martellidendron and P. subg. Vinsonia separated by approximately 15 Ma and the colonization of Africa by P. subg. Pandanus s.s. within the past six Ma. The colonization of Australia by P. subg. Pandanus likewise took place within the past six million years. These events occurred well after the isolation of those continents. Ocean dispersal is implicated in each case and subsequent inland radiations may have been facilitated by the retention of fleshy tissue that attracts animal dispersal agents. In P. subg. Pandanus s.s., and potentially in P. subg. Vinsonia, inland radiations have later recolonized coastal areas having retained the aerenchyma tissue and other adaptations necessary for flotation. The P. tectorius complex has achieved the widest distribution encompassing nearly the entire geographic range of the family within the past six million years. This success is likely due to propagules that float for long durations on the ocean while retaining a fleshy mesocarp that can aid in animal dispersal on land as well as a combination of characters which facilitate establishment including multiple seeds per fruit, agamospermy, and wind pollination. The origin of this clade in eastern Queensland also may have contributed to its spread since prevailing ocean currents in this area would potentially allow floating propagules to move into both the North and South Pacific as well as into the Indian Ocean. While this paper did not address the identity of all of the numerous taxa attributed to the Pandanus tectorius complex, results indicate that several distinct species are included and that the complex should not be considered monotypic. While reticulation among coastal populations is probably widespread, there appear to be mechanisms preserving distinct species within the complex. In some cases, (i.e. P. papenooensis or P. conicus), dispersal inland, into new habitats and away from coastal populations may result in isolation and the formation of new species. Ocean currents may also play an important role in maintaining a barrier to gene flow between geographic areas of the Indo-Pacific. Acknowledgments The Authors thank the staff of the Institut de Recherche pour le Développement, Nouméa, New Caledonia (NOU), the University of South Pacific, Suva, Fiji (SUVA) and the Australian Tropical Herbarium (ATH) in Cairns particularly Darryn Crayn, Ashley Field, Mark Harrington, Andrea Lim, and Frank Zich, for facilitating fieldwork and processing sample in Australia. In the Philippines, research assistance was provided by Regielene Gonzales, Lori Tongco (Institute of Biology, University of the Philippines), Tope Ordoñez, and the staff of BacMan Sorsogon and Mindanao Geothermal. Other assistance was provided by Jeff Boutain, Mark Merlin, Art Whistler, and Pei-Luen Lu (University of Hawaii – UH); James Ishimoto; Carlos Cianchini; Craig Costion (ATH); David Lorence and Tim Flynn (National Tropical Botanical Garden, Kaua‘i [PTBG]); David Orr (Waimea Botanical Garden, Honolulu); Jean-Yves Meyer (Ministère d’Education, de l’Enseignement Supérieur et de la Recherche, Papeete, Tahiti); Jean-Francois Butaud and Fred Jacq (French Polynesia); Jacques Florence (Muséum National d’Histoire Naturelle, Paris [P] and Institut de Recherche pour le Développement [IRD]); Karen Shigematsu and Mashuri Waite (Lyon Arboretum, Honolulu), Gerald McCormack (Natural Heritage Trust, Cook Islands); Russell Barrett (Botanic Gardens & Parks Authority, Western Australia); Steve Jackson (Cairns Botanical Garden, Queensland); Julie Roach (Townsville Botanical Garden); Joanne Birch (Royal Botanic Gardens Melbourne); Amanda Harbottle, Barbara Kennedy, Shelley James, and Clyde Imada [BISH], Michael Thomas [HAW]; Ian Hutton (Lord Howe Island); Yahya Abeid, Roy Gereau, and Pete Lowry [MO], Vicki Funk [US]; and Nicola Schoenenberger [LUG]; We are grateful to the Ministry of Education, Heritage, and Arts (Fiji), the authorities of New Caledonia’s Northern and Southern Provinces, the Queensland government, the government of the Philippines, and the USDA for providing collecting, export, and import permits for this research. Reviews of early drafts were provided by Lynn Clark (Iowa State University), Jason Cantley, Marion Chau (UH) and two anonymous reviewers. Financial support to TG was provided by the UH Graduate Student Organization, the UH Arts and Sciences Student Research Award and the Systematics Research Fund; to MWC and SB by a National Geographic Society Exploration Grant (# 904211) and the Idaho Botanical Research Foundation. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ympev.2014.11. 002. T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 References Ash, J., 1987. Demography, dispersal and production of Pandanus tectorius (Pandanaceae) in Fiji. Aust. J. Bot. 35 (3), 313–330. Baker, W.J., Couvreur, T.L.P., 2012. Global biogeography and diversification of palms sheds light on the evolution of tropical lineages. Historical biogeography. J. Biogeogr. 40 (2), 274–285. Bell, C.D., Soltis, D.E., Soltis, P.S., 2005. The age of the angiosperms: a molecular timescale without a clock. Evolution 59 (6), 1245–1258. Bell, C.D., Soltis, D.E., Soltis, P.S., 2010. The age and diversification of the angiosperms re-revisited. Am. J. Bot. 97 (8), 1296–1303. Birch, J.L., Keeley, S.C., 2013. Dispersal pathways across the Pacific: the historical biogeography of Astelia s.l. (Asteliaceae, Asparagales). J. Biogeogr. 40 (10), 1914– 1927. Bonde, S.D., 2005. Indian fossil monocotyledons: current status, recent developments and future directions. Challenges Indian Palaeobiol. 21. Bremer, K., 2000. Early Cretaceous lineages of monocot flowering plants. Proc. Natl. Acad. Sci. USA 97 (9), 4707. Buerki, S., Forest, F., Alvarez, N., Nylander, J.A.A., Arrigo, N., Sanmartin, I., 2011. An evaluation of new parsimony based versus parametric inference methods in biogeography: a case study using the globally distributed plant family Sapindaceae. J. Biogeogr. 38 (3), 531–550. Buerki, S., Callmander, M.W., Devey, D.S., Chappell, L., Gallaher, T., Munzinger, J., Haevermans, T., Forest, F., 2012. Straightening out the screw-pines: a first step in understanding phylogenetic relationships within Pandanaceae. Taxon 61 (5), 1010–1020. Burney, D.A., 2002. Late Quaternary chronology and stratigraphy of twelve sites on Kauai. Radiocarbon 44 (1), 13–44. Burney, D., James, H., Burney, L., Olson, S., Kikuchi, W., Wagner, W., Burney, M., McCloskey, D., Kikuchi, D., Grady, F., 2001. Fossil evidence for a diverse biota from Kaua’i and its transformation since human arrival. Ecol. Monogr. 71 (4), 615–641. Callmander, M., Laivao, M., 2002. Endemisme et biogeographie du genre Pandanus (Pandanaceae) en Afrique et a Madagascar. Bioterre Rev. Inter. Sci. de la Vie et de la Terre, 125–138. Callmander, M.W., Chassot, P., Kupfer, P., Lowry, P.P., 2003. Recognition of Martellidendron, a new genus of Pandanaceae, and its biogeographic implications. Taxon 52 (4), 747–762. Callmander, M.W., Lowry II, P.P., Forest, F., Devey, D.S., Beentje, H., Buerki, S., 2012. Benstonea Callm. & Buerki (Pandanaceae): characterization, circumscription, and distribution of a new genus of screw-pines, with a synopsis of accepted species. Candollea 67 (2), 323–345. Carlquist, S., 1967. The biota of long-distance dispersal. V. Plant dispersal to Pacific Islands. Bull. Torrey Bot. Club 94 (3), 129–162. Chase, M.W., Fay, M.F., Devey, D.S., Maurin, O., Rønsted, N., Davies, T.J., Pillon, Y., Petersen, G., Seberg, O., Tamura, M.N., Asmussen, C.B., Hilu, K., Borsch, T., Davis, J.I., Stevenson, D.W., Pires, J.C., Givnish, T.J., Sytsma, K.J., Mcpherson, M.A., Graham, S.W., Rai, H.S., 2006. Multigene analyses of monocot relationships: a summary. Aliso 22, 63–75. Chaw, S.M., Chang, C.C., Chen, H.L., Li, W.H., 2004. Dating the monocot–dicot divergence and the origin of core eudicots using whole chloroplast genomes. J. Mol. Evol. 58 (4), 424–441. Cheah, C.H., Stone, B.C., 1975. Embryo sac and microsporangium development in Pandanus (Pandanaceae). Phytomorphology 25 (2), 228–238. Cook, L.G., Crisp, M.D., 2005. Not so ancient: the extant crown group of Nothofagus represents a post-Gondwanan radiation. Proc. Royal Soc. B: Biol. Sci. 272 (1580), 2535–2544. Cox, P.A., 1981. Bisexuality in the Pandanaceae: new findings in the genus Freycinetia. Biotropica 13 (3), 195–198. Cox, P.A., 1985. Islands and dioecism: insights from the reproductive ecology of Pandanus tectorius in Polynesia. In: White, J. (Ed.), Studies of plant demography: A festschrift for John L. Harper. London, Academic Press, pp. 359–372. Cox, P.A., 1990. Pollination and the evolution of breeding systems in Pandanaceae. Ann. Mo. Bot. Gard. 77, 816–840. Cox, P.A., Huynh, K.L., Stone, B.C., 1995. Evolution and systematics of Pandanaceae. In: Rudall, P.J., Cribb, P.J., Cutler, D.F., Humphries, C.J. (Eds.), Monocotyledons: Systematics and Evolution, vol. 2. The Royal Botanic Gardens, Kew, pp. 663–684. Darriba, D., Taboada, G.L., Doallo, R., Posada, D., 2012. JModelTest 2: more models, new heuristics and parallel computing. Nat. Methods 9, 772. Davis, C.C., Bell, C.D., Mathews, S., Donoghue, M.J., 2002. Laurasian migration explains Gondwanan disjunctions: evidence from Malpighiaceae. Proc. Natl. Acad. Sci. 99 (10), 6833–6837. Drummond, A.J., Suchard, M.A., Xie, D., Rambaut, A., 2012. Bayesian phylogenetics with BEAUti and the BEAST 1.7. Mol. Biol. Evol. 29 (8), 1969–1973. Duvall, M.R., Clegg, M.T., Chase, M.W., Clark, D.W., Kress, J.W., Hills, H.G., Eguiarte, L.E., Smith, J.F., Gaut, B.S., Zimmer, E.A., 1993. Phylogenetic hypotheses for the Monocotyledons constructed from rbcL sequence data. Ann. Mo. Bot. Gard. 80 (3), 607–619. Edgar, R.C., 2004. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 32 (5), 1792–1797. Eriksson, R., 1994. Phylogeny of the Cyclanthaceae. Plant Syst. Evol. 190 (1–2), 31– 47. Fosberg, F.R., 1977. Miscellaneous Notes on the Flora of Aldabra and Neighbouring Islands: V: Pandanus tectorius Parkinson sensu latissimo (Pandanaceæ). Kew. Bull. 31 (4), 837–840. 31 Gourlan, A.T., Meynadier, L., Allègre, C.J., 2008. Tectonically driven changes in the Indian Ocean circulation over the last 25 Ma: neodymium isotope evidence. Earth Planet. Sci. Lett. 267 (1), 353–364. Guindon, S., Gascuel, O., 2003. A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst. Biol. 52 (5), 696–704. Guppy, H., 1906. Observations of a Naturalist in the Pacific Between 1896 and 1899: Plant-dispersal, vol. 2. Macmillan and Co., Limited, London. Hall, R., 2002. Cenozoic geological and plate tectonic evolution of SE Asia and the SW Pacific: computer-based reconstructions, model and animations. J. Asian Earth Sci. 20 (4), 353–431. Hall, R., 2009. Southeast Asia’s changing palaeogeography. Blumea 54, 148–161. Herman, A.B., Kvaček, J., 2010. Late Cretaceous Grunbach flora of Austria. Osterreich, Naturhistorisches Museum. Hill, K.C., Hall, R., 2002. Mesozoic-Cenozoic evolution of Australia’s New Guinea margin in a west Pacific context. In: Hillis, R., Muller, R.D. (Eds.), Defining Australia: the Australian Plate as part of Planet Earth, vol. 372. Geological Society of America Special Papers, pp. 265–290. Hnatiuk, S.H., 1978. Plant dispersal by the Aldabran giant tortoise, Geochelone gigantea (Schweigger). Oecologia 36 (3), 345–350. Huang, T.C., 1980. Miocene Palynomorphs of Taiwan. V. Angiosperm Grains. Taiwania, 25, pp. 57–103. Huynh, K.L., 1980. Quelques especes de Pandanus (Pandanaceae) peu connues des archipels des Philippines, de Palaos et de Salomon. Candollea 35, 385–419. Huynh, K.-L., Cox, P.A., 1992. Flower structure and potential bisexuality in Freycinetia reineckei (Pandanaceae), a species of the Samoa Islands. Bot. J. Linn. Soc. 110 (3), 235–265. Ingle, N.R., 2003. Seed dispersal by wind, birds, and bats between Philippine montane rainforest and successional vegetation. Oecologia 134 (2), 251–261. Janssen, T., Bremer, K., 2004. The age of major monocot groups inferred from 800+ rbcL sequences. Bot. J. Linn. Soc. 146 (4), 385–398. Jarzen, D.M., 1983. The fossil pollen record of the Pandanaceae. Gard. Bull. Singapore 36 (2), 163–175. Kuhnt, W., Holbourn, A., Hall, R., Zuvela, M., Kase, R., 2004. Neogene history of the Indonesian throughflow. In: Clift, P., Kuhnt, W., Wang, P., Hayes, D. (Eds.), Continent-Ocean Interactions within East Asian Marginal Seas, 149 ed. The American Geophysical Union, Washington, DC, pp. 299–320. Kvaček, J., Herman, A.B., 2004. Monocotyledons from the Early Campanian (Cretaceous) of Grünbach, Lower Austria. Rev. Palaeobot. Palynol. 128 (3–4), 323–353. Lee, M.A.B., 1985. The dispersal of Pandanus tectorius by the land crab Cardisoma carnifex. Oikos 45 (2), 169–173. Leopold, E.B., 1969. Miocene Pollen and Spore Flora of Eniwetok Atoll. US Government Printing Office, Marshall Islands. Lord, J.M., 1991. Pollination and seed dispersal in Freycinetia baueriana, a dioecious liane that has lost its bat pollinator. NZ J. Bot. 29 (1), 83–86. Macphail, M.K., Alley, N.F., Truswell, E.M., Sluiter, I.R.K., 1994. Early Tertiary vegetation: evidence from spores and pollen. In: Hill, R.S. (Ed.), History of the Australian Vegetation: Cretaceous to Recent. Cambridge University Press, Cambridge, pp. 189–261. Martin, E.E., Scher, H., 2006. A Nd isotopic study of southern sourced waters and Indonesian throughflow at intermediate depths in the Cenozoic Indian Ocean. Geochem. Geophys. Geosyst. 7 (9). McConkey, K.R., Drake, D.R., Meehan, H.J., Parsons, N., 2003. Husking stations provide evidence of seed predation by introduced rodents in Tongan rain forests. Biol. Conserv. 109 (2), 221–225. McLoughlin, S., 2001. The breakup history of Gondwana and its impact on preCenozoic floristic provincialism. Aust. J. Bot. 49 (3), 271–300. Mennes, C.B., Smets, E.F., Moses, S.N., Merckx, V.S., 2013. New insights in the longdebated evolutionary history of Triuridaceae (Pandanales). Mol. Phylog. Evol. 69 (3), 994–1004. Miller, M.A., Pfeiffer, W., Schwartz, T., 2010. Creating the CIPRES Science Gateway for inference of large phylogenetic trees. In: Paper presented at Proceedings of the Gateway Computing Environments Workshop (GCE), New Orleans, LA (14 November 2010). Nadaf, A., Zanan, R., 2012. World Pandanaceae: an Overview in Indian Pandanaceaean Overview. Springer, pp. 1–9. Nakanishi, H., 1988. Dispersal ecology of the maritime plants in the Ryukyu Islands, Japan. Ecol. Res. 3 (2), 163–173. Nishimura, S., Suparka, S., 1997. Tectonic approach to the Neogene evolution of Pacific-Indian Ocean seaways. Tectonophysics 281 (1), 1–16. Popa, M.E., Csiki, Z., Vasile, S., Savescu, M.O.B., 2011. The Maastrichtian macroflora of the Hateg and Rusca Montana basins. Paper presented at Eighth Romanian symposium on paleontology, Bucharest (29–30 September 2011). Poppendieck, H.-H., 1987. Monoecy and sex changes in Freycinetia (Pandanaceae). Ann. Mo. Bot. Gard. 74 (2), 314–320. Rambaut, A., Drummond, A., 2011. Tracer 1.5. Ramsey, D., 2010. Washed up: how plants spread along Australia’s tropical seashores. Wildlife Australia 47 (4), 34–39. Smith, S.Y., Collinson, M.E., Rudall, P.J., 2008. Fossil Cyclanthus (Cyclanthaceae, Pandanales) from the Eocene of Germany and England. Am. J. Bot. 95 (6), 688– 699. Stamatakis, A., 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 22 (21), 2688–2690. Stoian, L.M., 2002. Palynological analyses and dating of eight core samples from Caroline 1 well, Gambier-Otway Basins, South Australia: Government of South Australia, Office of Minerals and Energy Resources. 32 T. Gallaher et al. / Molecular Phylogenetics and Evolution 83 (2015) 20–32 Stone, B.C., 1961. The genus Sararanga (Pandanaceae). Brittonia 13 (2), 212–224. Stone, B.C., 1967a. Materials for a Monograph of Freycinetia (Pandanaceae). I. Gardens’ Bulletin Singapore, 22, 129–149. Stone, B.C., 1967b. Studies of Malesian Pandanaceae, I: (Polymorphism in Pandanus odoratissimus L.f. of Asia). Gard. Bull. Singapore 22(2), 231–257. Stone, B.C., 1968. Materials for a monograph of Freycinetia Gaud. IV. Subdivision of the genus, with fifteen new sections. Blumea 16, 361–372. Stone, B.C., 1969. Materials for a monograph of Freycinetia Gaud. (Pandanaceae) VIII. A revised list of Philippine species, with critical notes and some new taxa. Webbia 23 (2), 597–607. Stone, B.C., 1972. A reconsideration of the evolutionary status of the family Pandanaceae and its significance in monocotyledon phylogeny. Quart. Rev. Biol. 47 (1), 34–45. Stone, B.C., 1974. Towards an improved infrageneric classification in Pandanus (Pandanaceae). Botanische Jahrbucher fur Systematik, Pflanzengeschichte und Pflanzengeographie 94 (4), 459–540. Stone, B.C., 1976a. On the biogeography of Pandanus (Pandanaceae). Compte R. Seances Soc. Biogeogr. 452–460, 69–90. Stone, B.C., 1976b. The Pandanaceae of the New Hebrides, with an essay on intraspecific variation in Pandanus tectorius. Kew. Bull. 31 (1), 47–70. Stone, B.C., 1978. Revisio Pandanacearum: Part 1. Pandanus subgenera Coronata Martelli and Acrostigina (Kurz) Stone. Flora Malesiana, Precursores. Federation Museums J. 23, 1–73. Stone, B.C., 1982. Pandanus tectorius Parkins. in Australia: a conservative view. Bot. J. Linn. Soc. 85 (2), 133–146. Stone, B.C., 1983. Revisio Pandanacearum: part 2. Pandanus subgenus Rykia (De Vr.) Kurz (sections Gressittia, Rykiopsis and Solmsia excepted). Federation Museums J. 28, 1–99. Tamura, M.N., Yamashita, J., Fuse, S., Haraguchi, M., 2004. Molecular phylogeny of monocotyledons inferred from combined analysis of plastid matK and rbcL gene sequences. J. Plant. Res. 117 (2), 109–120. Tamura, K., Peterson, D., Peterson, N., Stecher, G., Nei, M., Kumar, S., 2011. MEGA5: molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum parsimony methods. Mol. Biol. Evol. 28 (10), 2731–2739. Walker, J.D., Geissman, J.W., Bowring, S.A., Babcock, L.E., 2013. The geological society of America geologic time scale. Geol. Soc. Am. Bull. 125 (3–4), 259–272. Wiles, G.J., Engbring, J., Falanruw, M.V.C., 1991. Population status and natural history of Pteropus mariannus on Ulithi Atoll, Caroline Islands. Pacific Science 45 (1), 76–84. Wilson, A.J.G., 2011. Pandanaceae. In: Wilson, A. (Ed.), Flora of Australia, vol. 39. ABRS, Canberra/CSIRO Publishing, Melbourne, pp. 211–235. Wright, C.H., 1930. Pandanus (Vinsonia) Basedowii C. H. Wright [Pandanaceae]. Bulletin of Miscellaneous Information (Royal Gardens, Kew), 4, 158–159. Yang, Z., Rannala, B., 1997. Bayesian phylogenetic inference using DNA sequences: a Markov chain Monte Carlo method. Mol. Biol. Evol. 14 (7), 717–724. Yoder, A.D., Nowak, M.D., 2006. Has vicariance or dispersal been the predominant biogeographic force in Madagascar? Only time will tell. Annu. Rev. Ecol. Evol. Syst. 37 (1), 405–431. Yu, Y., Harris, A.J., He, X.J., 2013. RASP (Reconstruct Ancestral State in Phylogenies). Version 2.1.