Sunda– Sahul floristic exchange and pathways into the Southwest Pacific: New insights from wet tropical forest trees

Aim: Recent investigations on the floristic exchange between Southeast Asia and Australia have shown a clear dispersal directionality bias (West to East) of wet- adapted plant taxa. However, dispersal routes and directions of wet forest taxa into the South Pacific remain insufficiently known. We here aimed to establish the most likely routes and directions of plant dispersal into the Southwest Pacific islands. Location: Southeast Asia, East Asia, Australia, Southwest Pacific. Taxon: Dysoxylum


| INTRODUC TI ON
Southeast Asia covers only 4% of the Earth's land area but harbours 20-25% of the planet's fauna and flora. It is one of the most biodiverse regions in the world (Woodruff et al., 2010). Due to the high number of endemic vascular plant species threatened with extinction, four biodiversity hotspots are recognised in this region: Indo-Burma, Sundaland, Wallacea and the Philippines (Myers et al., 2000;Woodruff et al., 2010). Its complex geological history includes continental landmasses and numerous oceanic islands, such as Indochina, Malesia and the Philippines (Hall, 2009(Hall, , 2012(Hall, , 2017. Due to its position, Southeast Asia plays an important role for the floristic exchange between the Asian and the Australian flora, with Wallacea acting as an important transition zone for both Laurasian and Gondwanan plant lineages in both directions (Crayn et al., 2015;Morley, 2003;Richardson et al., 2012;Takhtajan, 1986;Turner et al., 2010;van Welzen & Slik, 2009).
These predominant patterns for plants, based on molecular phylogenetic trees and modern biogeographic tools, are very well in line with early observations based merely on changes in species numbers and abundance across Southeast Asia for both plants and animals (Dickerson, 1928). Second, it was found that temporal dynamics of the Sunda-Sahul floristic exchange (disjunctions resulting from dispersal/migration events) are reflecting the region's geological history, with most disjunctions occurring after Sahul and Sunda started merging (Sunda-Sahul convergence, ca. 25-12 million years ago, Ma), and zoochorous lineages being overrepresented among successful dispersers (Crayn et al., 2015). Third, biogeographic barriers originally described to account for the discontinuous distribution patterns of animals (e.g. Wallace's line; Wallace, 1863; with Huxley's modification, Huxley, 1868) are less effective for plants (Grudinski et al., 2014;Joyce et al., 2021;Richardson et al., 2012), which has resulted in several attempts to find a more suitable geographic division (e.g. van Welzen et al., 2011) and underlying factors influencing these patterns (Joyce et al., 2021).
In contrast to Malesia, the floristic composition of Southwest Pacific islands and underlying processes forming the floras have so far been examined by only a few studies (Gillespie et al., 2011: dry forests; Keppel et al., 2010: lowland rain forest). Due to wind patterns established during the Miocene (Barker & Burrell, 1982), the circumpolar westerly winds favour dispersal with eastward directionality into the southern Pacific (Winkworth et al., 2002). The Southwest Pacific is neighbouring Southeast Asia, Australia, and Papua and displays a similar complexity of island groups that have varying origins, both volcanic and continental (Collot et al., 2020). Based on wind directionality and geographical proximity, New Guinea, Australia and New Caledonia are considered main source areas for dispersal into the Southwest Pacific (Keppel et al., 2010;Winkworth et al., 2002). New Zealand and New Caledonia, both being biodiversity hotspots (Myers et al., 2000), have received more attention than other islands of the region, and show a high number of elements of Australian origin (Dawson & Sneddon, 1969;Morat et al., 1986;Thorne, 1969). Furthermore, New Caledonia shares around 32% of its genera with the Solomon Islands (Thorne, 1969).
Studies have exhibited dispersal from New Caledonia to New Guinea/ Solomon Islands and Australia (Gossia: McLay et al., 2018;Planchonella: Swenson et al., 2019). Recently, dispersal from New Guinea and the Solomon Islands further east into the Pacific through Fiji has been proposed for several woody plant genera, such as Planchonella, Alphitonia (Hauenschild et al., 2018;Swenson et al., 2019), as well as for woody taxa of Cyrtandra (Atkins et al., 2020;Johnson et al., 2017).
However, routes of colonisation from the islands further east into the southern Pacific are not well investigated. In general, longdistance dispersal (LDD), although in the past considered an infrequent event, has had a major impact on distribution patterns of island plants (Cowie & Holland, 2006;Jordano, 2017;Nathan, 2006).
Indication of LDD events are disjunctions in distribution patterns over larger geographical distances, following the progression rule (Whittaker et al., 2017). According to this rule, the age of taxa in an archipelago follows the emergence of islands, with older taxa occurring on older islands and younger taxa on younger islands (Whittaker et al., 2017). In contrast, metapopulation vicariance would be indicated by older taxa occurring on younger islands, which has not found support from recent empirical data (Swenson et al., 2019). Oceanic and continental islands have a higher level of endemism than the continental mainland (Kier et al., 2009). In island biogeography, the level of endemism is viewed in relation to the level of isolation (MacArthur & Wilson, 1967). More recently, rather than isolation, species richness of islands has been discussed a more suitable indicator for endemism (Emerson & Kolm, 2005;Witt & Maliakal-Witt, 2007). As for the tropical South Pacific, species richness itself has no impact on the level of endemism on highly isolated islands, since species richness is negatively correlated to a high level of isolation (Keppel et al., 2010). Palaeoendemics are relics of extinction, while neoendemics are the result of diversification and speciation after colonisation (Mishler et al., 2014). Palaeoendemics are often associated with long emergent islands, such as continental islands like New Caledonia, the latter which has the highest level of plant endemism worldwide (Kier et al., 2009).
A suitable model group to investigate those biogeographic patterns of wet forest angiosperm lineages in a geographically extensive area, such as the one spanning Indomalesia, Australasia and the Pacific islands, can be found in Meliaceae, a family of high ecological importance. Within Meliaceae, Dysoxylum s.l. has one of the largest distribution ranges in the study region. Meliaceae comprise c. 740 validly described, accepted tree and shrublet species, taxonomically arranged in 58 genera (Muellner-Riehl & Rojas-Andrés, 2021 and Pseudocarapa (six species) (Holzmeyer et al., 2021). Zoochory by vertebrates, commonly birds, is the expected mode of seed dispersal for all of Dysoxylum s.l. and has been observed in multiple cases (Bodare et al., 2017;Ganesh & Davidar, 2001;Mabberley, 1995Mabberley, , 2013Manuyath, 2003;Myers & Court, 2013;Sethi & Howe, 2009;Whittaker & Turner, 1994). Didymocheton spectabile: Sethi & Howe, 2009) and a variety of other species (yellow-vented bulbul, Philippine glossy starling, black-naped orioles, sunbirds & collared kingfisher, Whittaker & Turner, 1994;tui, kakariki & blackbirds, Myers & Court, 2013). As for Dysoxylum s.l. as a group, even though seed characteristics are varying (Cheek, 1989), the bicolorous pattern of the fruits promotes bird dispersal (Willson & Thompson, 1982). The genera have wide, overlapping distribution ranges ( Figure 1a, Supplementary Documentation S1). However, Didymocheton is very distinct, with almost all species occurring east of Wallace's Line, except for a few widely distributed species reaching mainland Asia (Figure 1a, Supplementary Documentation S1). The distribution allows to investigate dispersal routes into the Southwest Pacific and resulting endemism (Holzmeyer et al., 2021).
To test if the biogeographic history of Dysoxylum s.l. follows well-established patterns in the Sunda-Sahul convergence zone (defined here as the area comprising Australia, New Guinea and Southeast Asia) and to obtain a more in-depth, detailed understanding of potential angiosperm dispersal routes and directions into the Southwest Pacific, we used the biggest data set on Dysoxylum s.s. and allied genera to date. Specifically, we test whether (i) the group originated on mainland Asia; (ii) diversification and geographic expansion are in temporal agreement with the convergence of the Asian and Australian tectonic plates since the Miocene, as has been shown for most previous plant studies (Crayn et al., 2015); (iii) Didymocheton uses a dispersal route from New Guinea eastwards into the Southwest Pacific through Fiji, as may be expected based on previously presumed rainforest tree dispersal routes (Keppel et al., 2010); and (iv) LDD is the main force behind the distribution expansion of Dysoxylum s.l., and therefore follows the progression rule.

| Divergence time estimation and calibrations
Divergence time estimates were calculated using BEAST 1.10.4 . Substitution models were chosen based on jModeltest 2 2.1.6 (Darriba et al., 2012) for the partitions: HKY+ G (Hasegawa et al., 1985) for the plastid partitions (trnL-F, rps15-ycf1), GTR+I+G for ITS and GTR+G for ETS (Tavaré, 1986). The birth-death model with incomplete sampling (Stadler, 2009) was chosen because of incomplete species sampling. The uncorrelated relaxed clock model was used and unlinked between nucleotide and plastid sequences, the clock models were checked and affirmed in Tracer 1.7.1  based on the coefficients of variation (>0.1) (Drummond & Bouckaert, 2015). Prior and clock settings were tested by performing marginal likelihood estimation (MLE) using path sampling (PS)/stepping stone sampling (SS) (Baele et al., 2012a;2012b). Two temporal constraints set in the analyses were based on six fossils. Psilastephanocoporites laevigatus    (Graham & Jarzen, 1969). Two independent runs with 100 million generations each, where every 10,000th tree was saved, were performed.
A burn-in of 3% was excluded after checking the log-files in Tracer 1.7.1  and the two tree files were merged with LogCombiner (implemented in the beast package). The tree files were annotated in TreeAnnotator (implemented in the beast package), node heights were annotated using the 'keep target heights' option.

| Ancestral area reconstruction
Ancestral area reconstruction (AAR) was conducted in R using the package 'BioGeoBEARS 1.1.2' (Matzke, 2018). Areas allowed and manual dispersal multipliers were edited for three time slices based on geological data; dispersal was solely restricted in time slices when the areas concerned were not emergent (Hall, 2002(Hall, , 2012Heaney, 1986;Nugraha & Hall, 2018;Seton et al., 2012;Zahirovic et al., 2016). All files are available via Dryad (doi:10.5061/dryad. rfj6q57f8). The three time slices were set before 30, 29-20 and 19 Ma to present, based on geological activity, to address the appearance of the Philippines, as well as Sulawesi and the Lesser Sunda Islands, and slowly increasing connectivity in Malesia (Hall, 2002(Hall, , 2011(Hall, , 2012Heaney, 1986;Nugraha & Hall, 2018;Seton et al., 2012;Zahirovic et al., 2016). The delimitation of areas for AAR was based Every node in the phylogeny was assigned probabilities for the potential ancestral areas. The ancestral area can either contain several areas or constitute a single area. We noted the probabilities for the areas being ancestral. For those nodes for which no area reached a probability of at least 0.5, we listed the next most likely areas, which often were part of a composition of best areas. In some cases, the analyses listed multiple areas with probabilities below 0.5 without overlap in the composition of best areas, these ancestral areas are not further discussed because of the uncertainty. Biogeographical Stochastic Mapping (BSM) was performed subsequent to the AAR in 'BioGeoBEARS' (Dupin et al., 2016;Matzke, 2016) for the bestfitting model. A total of 100 BSMs were simulated and biogeographical events (within-area speciation, vicariance and dispersal events; range expansions and founder events) averaged over it.
An additional Pacific-specific AAR was conducted (Supplementary Documentation S3).

| Age-Area correlation
To assess whether Dysoxylum s.l. is following the progression rule, we matched the lineages to their current distribution areas and identified the ages of the latest emergence of these areas from litera-

| Divergence time estimation and calibrations
The topology of our dated tree was compared to a recently published phylogenetic study (Holzmeyer et al., 2021) to check for consistency with the current classification, which could be con-

| Pacific subset analysis
The AAR for the Pacific subset analysis by 'BioGeoBEARS' resulted in the DEC+J model (Matzke, 2014) as the most favourable model

| Age-area correlation
We compared estimated ages of lineages with the age of the long-

| Origin of Dysoxylum s.l. and floristic exchange between Sunda and Sahul
As expected, all genera, except the predominantly Sahul-Pacific dis-

| Pacific dispersal routes
Based on the distribution and ancestral area from the main analysis (Figures 1b,c, 2), we focus on discussing Didymocheton for insight on Pacific routes. The genus has its ancestral area estimated to be shared between New Guinea, the Solomon Islands, Australia, and the Southwest Pacific (Figure 2), and a current distribution range from Australia to Niue, with some exceptionally widespread species with the Australian flora (Thorne, 1969), which agrees with dispersal from and back to the Sahul area, observed here. Dispersal east into the Pacific from New Caledonia has been reported before in a few studies (Barrabé et al., 2012;Keppel et al., 2009 (Onstein et al., 2017).
A larger fruit size has been shown to be more likely related to dispersal and distribution limitation in Meliaceae species (e.g. Aglaia mackiana, dispersed by the dwarf cassowary; Pannell, 1997) than smaller fruited species (Hopkins et al., 1998). However, D. bijugus is distributed from New Caledonia to Vanuatu. In our analysis, the two samples of D. bijugus used constitute two independent evolutionary lineages rendering the taxon not monophyletic. Both samples were collected from New Caledonia, so we cannot confidently infer dispersal routes of the species to Vanuatu and/or New Caledonia.
The dispersal of Didymocheton is hindered on this route and the barrier might only affect the dispersal vector ('The land and fresh-water birds of the southwest Pacific islands derive mainly from New Guinea' Diamond, 1970, p. 529).

| The impact of LDD on Dysoxylum s.l.
The current distribution patterns of Dysoxylum s.l. are mainly driven by dispersal events, as expected. This is also supported by the high percentage of dispersal returned in the overall BSM (Table 1), while it has to be noted that a portion of additional dispersal is disguised within the within-area speciation due to large geographical areas used in the analysis, for example, 'A' representing Africa and mainland Asia. Similarly, the high percentage of within-area speciation returned for the Pacific subset analysis can be accounted to the enlarged geographical area of the non-Pacific region. However, 34% of biogeographical events were simulated dispersal events ( Table 1). In both cases, vicariance contributes the smallest fraction of biogeographical events simulated ( Table 1). Based on the correlation of the age of lineages and time since the last emergence of areas, Dysoxylum s.l.
follows the progression rule. The ages of lineages match and follow the ages of colonisable landmass. The comparison shows that younger lineages occur in the younger subset of geographic areas (Supplementary Documentation S10). Also, none of the lineages was found to be older than the longest emergent area of their current distribution range. This suggests an overwhelming role of LDD on distribution patterns in this group rather than metapopulation vicariance. Extinction can be construed as dispersal (Keppel et al., 2009), but based on the young age of endemic taxa, we would not assume an elevated level of extinction. This is also corroborated by extreme disjunctions, such as in the distribution of  (Pennington, 1981). In order to get a better insight on the evolution of its current distribution range, more detailed research on Cabralea, possibly including population-level sampling, is needed. This will also aid the delimitation of potential subspecies. We suggest a single successful LDD event from Australasia as explanation for the trans-Pacific disjunction.
As mentioned above, Dysoxylum s.l. dispersal vectors are most likely birds and based on the multiple dispersal events observed (Bodare et al., 2017;Ganesh & Davidar, 2001;Mabberley, 1995Mabberley, , 2013Manjunath, 2003;Myers & Court, 2013;Sethi & Howe, 2009;Whittaker & Turner, 1994), especially into Southwest Pacific, occurred quite regularly. Migratory birds and their routes have often been associated with LDD events in other groups (Baldwin & Wagner, 2010;Gillespie et al., 2012) and might explain the higher frequency of LDD events in Didymocheton. Similar patterns in biogeographical analyses of dispersal vectors themselves can also hint at co-dependent evolutionary processes, where plant distribution follows the biogeographical patterns of potential vectors (Diamond, 1970). Successful dispersal has to be followed by successful establishment to be detectable, and this is often associated with the capacity of uniparental reproduction (Pannell et al., 2015).
Dysoxylum s.l. mostly comprises dioecious species, which is fairly common for island flora (Pannell et al., 2015). Studies on the reproductive systems of Dysoxylum s.l. on population level would provide better insights into possible changes in the reproductive systems that might have occurred after successful establishment.
However, successful establishment of taxa with woody habit and fleshy fruits has been found to be less negatively impacted by dioecy and subsequent mate limitations, through compensation by other traits, such as perenniality and higher propagule pressure, the latter which would enhance the frequency of dispersal events (Pannell et al., 2015;Vamosi et al., 2007). This could explain the suc-  and East (Tonga, Samoa), and therefore plays an important role for the distribution expansion of rainforest plants into the SW Pacific.

| CON CLUS IONS
Dispersal from New Caledonia westwards into the Pacific appears to be impeded. A detailed study on fruit and seed sizes as well as association with dispersal vectors will bring more insight as to the conditions of traits needed for successful dispersal and establishment.
Dysoxylum s.l. follows the progression rule and plant diversity on oceanic islands is driven by LDD as main mechanism of distribution. In order to get a better understanding of the trans-Pacific split between Didymocheton and Cabralea, a phylogeographic analysis of Cabralea could shed light on the ancestral area in South and Central America and subsequent dispersal. The impact of isolation on the likelihood of species to be(come) endemic, in our study, concurs with the theory of island biogeography, and plays a major role for the observed high levels of diversity in SE Asian and Pacific Meliaceae.

ACK N O WLE D G E M ENTS
We thank Claudia Krüger for assistance in laboratory work, Jan Schnitzler for general support and David J. Mabberley

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Sequence data are deposited in GenBank and cited in the