Migration of Isonandreae
According to the BEAST analyses, Sapotaceae began to diversify from 84.5 Ma (HPD: 67.1–105.0 Ma; node A, Figs 1, 2) and the origin of the family is probably Asia as the rest of the family is derived in a grade made up of Sarcosperma and Eberhardtia that are Asian genera. Bell, Soltis & Soltis (2010) estimated the age of major clades of angiosperms using multiple fossils and found that the stem node of Sapotaceae was 74 (62–84) Ma using BEAST with fossils ages calibrated as exponential priors. In contrast Bremer, Friis & Bremer (2004) estimated a Sapotaceae stem node age of c. 102 Ma that is more consistent with the crown node age of the family estimated here. The age of the stem node of the Isonandreae/Mimusopeae clade is Mid Eocene, matching the age of type I and II pollen grains (Harley, 1991) characteristic of Mimusopeae and Isonandreae from the Mid Eocene of England (Gruas-Cavagnetto, 1976). The Isonandreae clade is nested in a grade, the early diverging lineages of which are African, consistent with a migration of this lineage from Africa into Asia from the time of the age of the stem node of the tribe at 40.5 Ma (HPD: 32.6–48.3 Ma; node D, Figs 1, 2). This migration could have occurred via Laurasia during this period as temperatures were high enough to support megathermal taxa that are now predominantly restricted to lower latitudes. The historical presence of Sapotaceae in Laurasia is demonstrated by pollen characteristic of Mimusopeae, Isonandreae and Sideroxyleae in the Eocene of England (Gruas-Cavagnetto, 1976; Harley, 1991).
The results are consistent with the hypothesis that Isonandreae began to diversify first in Laurasia or Sundania. Gondwanan (i.e. Indian or Australasian) Isonandreae are nested in a Sundanian lineage supporting the fact that early diversification in the tribe was in Sundania. Migration from Sundania into New Guinea/Australia, across Wallace's Line, occurred on five occasions between c. 24.2 and 1.3 Ma. The number of migrations may of course be increased by adding to our limited sampling to the east of Wallace's Line. This west to east migration is also evident in studies of Aglaieae (Meliaceae) by Muellner et al. (2008), Pseuduvaria Miquel (Annonaceae) by Su & Saunders (2009) and Begonia L. (Thomas et al., 2012). These results are consistent with previous hypotheses on the origin and patterns of migration of South-East Asian lineages (e.g. Good, 1960; Van Balgooy, 1976; Van Welzen & Slik, 2009), i.e. that migration of angiosperms across Wallace's Line was predominantly from west to east. This has recently been reinforced in a re-assessment of patterns of migration based on a review of phylogenetic studies for the region (Richardson, Costion & Muellner, 2012).
Migration could have occurred at any point between stem and crown nodes with the arrival time only being confirmed at the crown node by diversification. The age of the crown node of Burckella, 12.7 Ma (HPD: 7.7–18.2 Ma; node J, Figs 1, 2), indicated that it had definitely reached New Guinea by this time. New Guinea was believed to be submerged until the Mid Miocene (Hall, 2009), so the age of the onset of diversification of Burckella is consistent with dispersal onto the island following its emergence. The stem node of Burckella is c. 24.2 Ma (node G, Figs 1, 2) and migration from Sundania to New Guinea could have occurred at any time from that point onwards. It is possible that this lineage could have migrated to New Guinea via a now submerged island and/or that the sister lineage of extant Burckella may now be extinct. An extinct sister lineage would have decreased the age of the stem node of Burckella.
Many west-to-east migrations documented in the literature were accompanied by substantial radiation in the east (e.g. Muellner et al., 2008; Thomas et al., 2012). Extensive radiation has also occurred in Palaquium with approximately half of its 110 species occurring to the east of Wallace's Line. Smaller radiations have occurred in Madhuca, a genus of 100 species, which has only 14 species to the east of Wallace's Line and Payena with only one of 20 species to the east of the line. Burckella has about 14 species, all to the east of Wallace's Line in the Western Pacific from the Moluccas and New Guinea to Fiji, Samoa and Tonga. Our results indicate only one poorly supported example of a back migration from east of Wallace's Line to the west, although additional sampling may of course reveal more examples of this. The absence or low frequency of back-dispersal has been noted in Malesian fanged frogs (Limnonectes, Evans et al., 2003) and in Pseuduvaria (Annonaceae; Su & Saunders, 2009) and was also highlighted by Thomas et al. (2012) in his study of Begonia, who suggested that it may be due to niche pre-emption (Silvertown, 2004; Silvertown, Francisco-Ortega & Carine, 2005). Prior filling of niche space may inhibit later arrivals by close relatives. Niche space would have been filled in the more ancient terranes of Sundania in comparison with much of the area to the east of Wallace's Line that only emerged recently.
Despite this relative lack of back migration it is clear that Isonandreae have migrated over water across Wallace's Line on numerous occasions. One criticism of using fossil-calibrated dated trees has been that they only give minimum age estimates that may result in under-estimation of the true ages (Heads, 2005). If this were true, then many phylogenetic splits that have been attributed to trans-oceanic long-distance dispersal on the basis of fossil calibrated dated trees (e.g. Bartish et al., 2011) could in fact have been older and therefore may have been caused by tectonic events such as the break-up of Gondwanaland. If we are underestimating the ages of splits between Sundania and Australasia, then we could also be underestimating the ability of Sapotaceae to disperse across oceans because these events would have occurred earlier over an even wider expanse of ocean when these two regions were more geographically distant from one another. This is also true of any other tectonic convergence event such as that between the North and South American plates (see Cody et al., 2010; J. E. Richardson, G. A. Mondragon, J. Serrano, J. A. Hawkins, I. V. Bartish, U. Swenson, M. Gonzalez, J. Chave, J. Vieu & S. Madriñán, unpubl. data).
Our results show Sri Lankan lineages nested in Sundanian ones, indicating an origin in Sundania and either dispersal or overland migration to Sri Lanka. Migration occurred from Malesia into Sri Lanka (India) three times at c. 22.4 Ma (node H, Figs 1, 2), 19.8 Ma (node I, Figs 1, 2) and 7.7 Ma (node L, Figs 1, 2). More sampling, particularly of Indo-Chinese taxa, will be needed to determine whether Sri Lankan lineages arrived overland or by long-distance dispersal. An ‘Out of India’ scenario was suggested for Crypteroniaceae by Rutschmann et al. (2004) with overland migration from India occurring as the subcontinent collided with Laurasia. Based on the fossil record, Morley (1998) suggested an over water invasion of Sundania from India during the Mid Eocene when India and South-East Asia were positioned at similar latitudes and in the same climatic zone. In the Mid and Late Eocene evidence appeared for the presence in South-East Asia of pollen of taxa characteristic of the Palaeocene and Early Eocene of India, such as numerous palms, bombacoid Malvaceae, Sapindaceae, Thymeleaceae, Proteaceae, Linaceae, Olacaceae, Polygalaceae and Ctenolophonaceae (Morley, 1998, 2000; Lelono, 2000; Collinson & Hooker, 2003), suggesting dispersal of Indian elements to South-East Asia in these periods. Our results provide strong evidence to indicate that migration into sub-continental India and Sri Lanka from South-East Asia has also occurred during a different period of time than those migrations into Sundania indicated by the fossil record.
Pleistocene climatic events resulted in changes in land area (and therefore the potential area of occupation of terrestrial organisms) as a consequence of changing sea levels (Voris, 2000). Sea levels were much lower during glacial periods and consequently many present day islands, particularly in Sundania, would have been connected. Inter-glacial periods would have seen contraction of land areas into island ‘refugia’, and this contraction may have resulted in allopatric speciation. According to the mean ages of splits from a most recent common ancestor, 24 species evolved in Isonandreae during the Pleistocene, indicating that sea-level changes could have had resulted in substantial speciation, although it is, of course, possible that other events might have caused speciation during that epoch. In contrast, only three speciation events occurred in Neotropical Chrysophylloideae (J. E. Richardson, G. A. Mondragon, J. Serrano, J. A. Hawkins, I. V. Bartish, U. Swenson, M. Gonzalez, J. Chave, J. Vieu & S. Madriñán, unpubl. data) during the Pleistocene, although this was based on a sample of only 25% of species compared with the 40% of Isonandreae sampled here. If, on complete sampling, this pattern proves to be the case then it would indicate that for Sapotaceae, South-East Asia has been a more active laboratory of speciation than the Neotropics during the Pleistocene and that, despite Sapotaceae occupying similar biomes in both regions, speciation patterns and processes may differ between them.