Molecular Data and Analyses

We selected and adapted the data set of nuclear ribosomal DNA (ETS, ITS), the nuclear gene RPB2, and gap information coded as binary characters used by Swenson et al. (2013). We chose to focus on nuclear DNA because chloroplast markers lack sufficient phylogenetic information at this taxonomic level in Sapotaceae (Swenson et al. 2013). We modified the data set by excluding distant outgroups, while expanding the taxon sampling of interest in the ingroup. Extraction, amplification, and primers followed the protocols established by Bartish et al. (2005) for ITS, Swenson et al. (2008) for ETS, and Swenson et al. (2013) for RPB2. Molecular dating and phylogeographic diffusion analysis were performed in an integrated analysis using BEAST v1.8.4 (Drummond et al. 2012). The data (ETS, ITS, RPB2, gaps) was partitioned with a unique clock and substitution model for each aligned locus, while linking the tree model across all partitions. Each partition was tested for available substitution models in BEAST using jModeltest (Darriba et al. 2012), which assigned an HKY+G substitution model, and a simple model without gamma for the binary data, all with estimated base frequencies. Clock models for all partitions were set to uncorrelated lognormal distributions to allow different substitution rates and patterns of among-lineage rate variation across loci, with rate priors set to weakly informative exponential distributions (offset 0, mean 0.5), and with the default priors for the standard deviation of branch rates. We used a birth–death tree prior with default uninformative uniform priors for birth and death rates. Posterior probability (PP) of 0.95 or more is recognized as strong node support, 0.85–0.94 as weak, and values below 0.85 are not reported. Age estimates are reported with 95% highest posterior density (HPD) intervals. Taxa, vouchers, and GenBank accession numbers, together with a complete, aligned, and partitioned data matrix, are available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.p2gs002 (Supplementary Appendices S1 and S2).

Node Calibration

Two nodes were temporally constrained, one using fossils and the other by applying a secondary calibration. Fossils of Planchonella are scarce and, apart from subfossils from the Marquesas Islands (Huebert and Allen 2016), are known only from New Zealand. These include a leaf impression from Oligocene–Miocene transition strata at Landslip Hill (Campbell 2002) and unpublished microfossils plus a flower bud containing pollen from Foulden Maar near Middlemarch, Otago (Swenson et al. 2014). Foulden Maar once hosted a rich mesothermal Lauraceae-dominated evergreen forest (Bannister et al. 2012) dated to 23.2 0.2 Ma (Lindqvist and Lee 2009; Lee et al. 2016). Fossils from these localities are allied to Planchonella costata, the only extant representative in New Zealand, which is sister to the Australian species P. eerwah (Swenson et al. 2013). Based on these being sister species, having rather long evolutionary histories, being descendants from a most recent common ancestor positioned on a short molecular branch, and a fossil that is very precisely dated, we argue that the fossil does not significantly pre-date the most recent common ancestor of P. costata and P. eerwah (node 37, Fig. 4). Hence, we have applied a fairly informative constraint on node 38 with an offset of 23.2 Ma with variable exponential decays of 0.055, 0.55, and 2.25, respectively, varying the upper 97.5% limit of the probability density. Hence, three different prior distributions were used, one of 23.2–23.4 Ma, one of 23.2–25.2 Ma, and one of 23.2–31.5 Ma. This approach was designed to compensate for the risk of erroneous dating of the fossil and a possible earlier arrival of Planchonella to New Zealand. Secondary calibration information can be used if fossils are scarce or absent (Ho and Philipps 2009; Ho et al. 2015). We used this to suggest a root age of Pleioluma approximated from the corresponding node of an earlier study (Swenson et al. 2014), modeled by a normal distribution (mean 48.6 Ma, SD 5.1) with 97.5% of the probability density between 38.6 and 58.6 Ma. A normal distribution is preferred over a uniform prior, because it does not impose unjustified hard boundaries on the potential node age.

Open in a separate window

Figure 4.

Continuation of the Planchonella (plant family Sapotaceae) chronogram as indicated in Figure 3, showing subclades D3 and D4. Details of the figure follow those of Figure 3, except that accessions from the Seychelles are color-coded in brown. The star points to the 23.2 myr fossil used for calibration. Accessions of the Hawaiian species P. sandwicensis are shown as two triangles, developed in Figure 6.

The Seychelles

North of Madagascar in the Indian Ocean lie the Seychelles islands, which, in part, consist of Gondwanan continental fragments that reached their current position some 63–65 Ma (Frey 2009). Despite providing terrestrial habitats during much of the Cenozoic, the Seychelles have been colonized by dispersal, for example, by Rubiaceae from Africa, Asia, and Madagascar in several waves (Kainulainen et al. 2017). The Seychelles are the westernmost outpost of Planchonella, where the widespread species P. obovata occupies the two main islands Mahé and Praslin. Sister to these accessions is the population from Palau, located in a straight line 8900 km to the east, with a divergence time of roughly 2.2 Ma (0.4–4.8 Ma; node 51, Figs. 4, ​,5c).5c). Disjunctions of this kind can be explained only by LDD, which has also been inferred in Amaracarpus (Rubiaceae; Razafimandimbison et al. 2017).

Planchonella in Fiji and Vanuatu (H2, H3)

The Solomons–Vanuatu–Fiji–Tonga are all part of the Melanesian arc system that developed over the past 40–45 myr (Hall 2002; Crawford et al. 2003; Schellart et al. 2006). By the end of the Eocene (35 Ma), this arc system began to split owing to back-arc spreading in the South Fiji Basin associated with eastward rollback of the Pacific oceanic crustal slab, which eventually resulted in three separate trench systems: Vanuatu, Vitiaz, and Tonga. Around 8–12 Ma, back-arc spreading began to form the North Fiji Basin, causing progressive isolation and clockwise rotation of the Vanuatu segment, counter-clockwise rotation of the Fiji segment, and cessation of activity in the Vitiaz Trench. In short, during the past 45 myr, Vanuatu experienced tectonism and at least three major phases of arc-related volcanism that probably generated some terrestrial environments in the region since the Oligocene. Heads (2006, 2018) suggested that the Vanuatu–Fiji island arc probably inherited most of its biota from the old (Melanesian) arc, demonstrated by the endemic species being much older than any of the individual islands.

Our analyses include eight accessions from Fiji and three from Vanuatu, representing eight species (Planchonella aneityensis P. brevipes, P. linggensis, P. membranacea, . “Munzinger6490”, P. smithii, P. umbonata, and P. vitiensis). These species are placed in seven clades scattered across the phylogeny, two in Clade D2 (Fig. 3) and five in Clade D4 (Fig. 4). The oldest subaerial rocks in Fiji and Vanuatu are dated to around 36 and 23 Ma, respectively (Crawford et al. 2003). Planchonella brevipes from Fiji has an ancestral origin at 23.3 Ma (18.3–29.0 Ma) and P. linggensis from Vanuatu represents the most recent arrival in the area at only 0.4 Ma. Sister relationships of these accessions include New Caledonia/Vanuatu, New Guinea/Fiji (Viti Levu), Solomon Islands/Vanuatu, Micronesia/Fiji (Vanua Levu), and Hawaii/Fiji. The phylogeographic diffusion analysis suggests that this area was colonized twice from New Guinea, with subsequent splits at 23.3 Ma (18.3–29.0 Ma) and 21.6 Ma (17.1–27.0 Ma). The first event resulted in radiation in Vanuatu and Fiji at 6 Ma with an onward colonization of the Cook Islands and French Polynesia, all in line with the progression rule. The second lineage split in an area west of Vanuatu and formed the lineages now present in Fiji and the Hawaiian Islands. An additional lineage, possibly with an origin in New Caledonia, split in the middle Miocene and progressed into New Guinea and Fiji at 6.9 Ma (3.4–11.2 Ma; node 26). Hence, Planchonella has had representatives in Fiji and Vanuatu for the past 20 myr, derived twice from New Guinea, and possibly twice from New Caledonia. The area probably formed an important genetic pool in the Miocene for onward colonization of islands progressively developing in the Pacific, especially supporting the patterns reported for the Society Islands and Hawaii (Hembry and Balukjian 2016; Claridge et al. 2017), but these areas have also received recent immigrants. Our data and analyses reject hypotheses H and H that the flora is older than Vanuatu and Fiji, and that it evolved on the ancient Melanesian arc, but is reconcilable with H in that the lineages have colonized oceanic islands subsequent to their development (Table 1).

Similarity Between the Loyalty Islands and Vanuatu (H4)

The presumed higher biological similarity between the Loyalty Islands and Vanuatu compared with the more closely located Grande Terre (New Caledonia) could be explained by metapopulation vicariance and westward rollback of the Loyalty Ridge from 50 to 35 Ma (Heads 2018). The Loyalty Islands are positioned on the Loyalty Ridge and derived from Oligocene volcanic seamounts capped by thick limestones (Bogdanov et al. 2007). Four of these seamounts have been uplifted during the past 2 myr to about 100 m (Lifou) and 130 m (Maré) above sea level. The islands are undoubtedly young based on their marine caprocks, but old lineages could have become established on the islands.

Planchonella lifuana is the only congener endemic to the Loyalty Islands where it grows on calcareous substrates. This species is embedded in a clade of 13 species endemic to New Caledonia and all except one are restricted to ultramafic substrates. Planchonella lifuana is sister to P. leptostylidifolia from which it split at 5.6 Ma (2.9–8.7 Ma; node 42, Fig. 4). The estimate is slightly older than the age of the Loyalties, but much closer to the island age than the 35–50 myr required in order to concur with ridge rollback and metapopulation vicariance. Another example from Sapotaceae is a yet undescribed species of Pycnandra (. Butaud 3343), endemic to the Loyalties and embedded among species from Grande Terre, with an age of about 2.5 Ma (Swenson et al. 2014, 2015). Other examples of dated disjunctions between these islands are the geckos Bavayia crassicollis and B. cyclura from, respectively, the Loyalties and Grande Terre, dated to some 2 Ma (Skipwith et al. 2016). There are also young age disjunctions between the Loyalties and Vanuatu, exemplified by the sister crickets Lebinthus lifouensis and L. santoensis, dated to 2.1 Ma, and the widespread species Cardiodactylus novaeguineae dated to 0.2 Ma (Nattier et al. 2011). It suffices to say that molecular evidence indicates that sisters and widespread taxa exist in any combination of Grande Terre, Loyalty Islands, and Vanuatu, and that many split during the Pleistocene, rejecting both H and H of metapopulation vicariance dependent on westward ridge rollback (Table 1).

Hawaiian Islands (H2 and H5)

The Hawaiian–Emperor seamount chain comprises eight main islands and numerous atolls and seamounts extending from the young island Hawaii (0.5 Ma) to Kure Atoll (29.8 Ma) in the northwest (Price and Clague 2002). The islands are of hotspot origin and Kauai (4.7 Ma) is currently the oldest high island in the chain. Northwest of Kauai are the small islands of Nihoa (7.3 Ma) and Necker (11.0 Ma). Six islands (Kauai, Oahu, Maui, Molokai, Lanai, and Hawaii) are inhabited by the morphologically confusing species Planchonella sandwicensis (Wagner et al. 2002). The crown age for . sandwicensis is estimated at 11.1 myr (7.3–15.5 myr), which splits into two well-supported lineages with crown ages of 3.6 myr (1.7–6.2 myr) and 7.5 myr (4.1–11.5 myr; Fig. 6). Subsequent splits group the accessions according to island of origin, but some with very weak support. Accessions of the two lineages have, respectively, deep purple and yellow fruits and the split at 11.1 Ma corresponds to a speciation event, but it is unclear if additional species or subspecies are recognizable. Taxonomic revision of these lineages will be explored in a subsequent paper.

The first split in the entire Hawaiian lineage corresponds to the age of Necker Island (11.0 myr), whereas the crown age for the yellow-fruited clade corresponds to the age of Nihoa (7.3 myr). This lineage later split at 2.1 Ma (1.0–3.8 Ma) and colonized Hawaii at 0.7 Ma (0.1–1.7 Ma), which coincides with the development of that island. Maui Nui and Oahu were colonized around 1.5 Ma, concurring with the islands’ ages, but support is weak and the direction of dispersal is not established. The purple-fruited clade began to radiate at 3.6 Ma (1.7–6.2 Ma; Fig. 6). The divergence time between populations in Molokai and Maui is estimated at 0.4 Ma (0.0–1.4 Ma) and coincides with subsidence and isolation of Maui Nui into separate islands (Price 2004). The split between populations on Oahu and Kauai occurred at 1.4 Ma (0.5–2.8 Ma) and coincides with the islands’ ages, but the population on Kauai seems to be derived. The youngest island, Hawaii, has never been colonized by this clade. The yellow-fruited clade concurs in part with the progression rule, having a potential origin in Kauai and a progressive colonization of younger islands, whereas the purple-fruited clade does not, owing to inferred back-dispersal to older islands.

Splits in Hawaiian Planchonella are apparently older than the islands that presently hold forests (nodes 80, 82; Fig. 6) and the gap to the stem node spans myr. Whether an extinct ancestor has ever been present on yet older islands, such as Gardner (15.8 Ma), will remain unknown. During the formation of Kauai, it is possible that no other islands in the chain had peaks over 1000 m (Price and Clague 2002), suggesting that extant Hawaiian taxa in montane habitats may have had an origin after the development of Kauai or may have experienced a severe bottleneck if derived from a pre-Kauai island. Price and Clague (2002) explained that this bottleneck may not have been as severe for coastal and lowland taxa. Planchonella sandwicensis is distributed in various habitats, including relatively dry and humid windward forests. The yellow-fruited clade appears to be distributed predominantly in the drier, leeward sides of islands—regions that have many lowland and coastal ecological attributes. Therefore, it is probable that the ancestor of P. sandwicensis was established from a lowland population on a pre-Kauai island. Other Hawaiian angiosperms, including the Hawaiian lobeliads (Campanulaceae) (Givnish et al. 2009) and Hillebrandia (Begoniaceae) (Clement et al. 2004), have origins much older than the current main islands.

Our results show that the Hawaiian lineage of Planchonella is older than the existing high islands, that the first split is consistent with the age of the two eroded Necker and Gardner islands, and that the species is not of Cretaceous age. Nevertheless, the lineage must have thrived somewhere, since it split from the most recent common ancestor at 20 Ma. It may have survived by island-hopping along the Hawaiian–Emperor seamount chain. Hence, the history of Planchonella in Hawaii concurs with island-hopping, but not with metapopulation vicariance sensu Heads (2018). We thereby reject H but, in part, support H in that islands are progressively colonized by dispersals subsequent to their origin and becoming habitable (Table 1).

Two anonymous reviewers and Associate Editor, Simon Ho provided many useful and constructive comments on this article. Curators at the herbaria BISH, BRI, L, PTBG, and SING kindly gave access to their collections. Charles Morel and Steen Hansen helped us with collections from the Seychelles. New Caledonian environmental services provided us with collecting permits across the territory. Raphaël Colombo and Patrick Taton installed the video system for filming pigeons. Collections made in Hawaii became possible through permits granted through the Hawaii Department of Land and Natural Resources, Division of Forestry and Wildlife (KPI-2015-48, MDF-052215A, ODF-041715R1, 13-209-5), Division of State Parks (K2015-320cc), and Parker Ranch (permit 150121), as well as permissions from Kualoa Ranch, The Department of Hawaiian Home Lands, The National Tropical Botanical Garden, The Nature Conservancy of Molokai, and Waimea Valley Garden. In Hawaii, field support was provided by Maggie Sporck-Koehler, Laurent Pool, Lucas Luehrs, David Orr, Susan Ching-Harbin, Kobey Togikawa, Richard Pender, Russell Kallstrom, William Haase, David Lorence, Art Medeiros, Joseph Camara, Nicolai Barca, Keahi Bustamente, Patti Welton, Elliott Parsons and Hank Oppenheimer, and elsewhere in the Pacific, we extend out thanks to all collectors appearing in Supplementary Appendix S1 available on Dryad.