�� 2006 Nature Publishing Group Sympatric speciation in palms on an oceanic island Vincent Savolainen1, Marie-Charlotte Anstett2, Christian Lexer1, Ian Hutton3, James J. Clarkson1, Maria V. Norup1,4, Martyn P. Powell1, David Springate1, Nicolas Salamin5 & William J. Baker1 The origin of species diversity has challenged biologists for over two centuries. Allopatric speciation, the divergence of species resulting from geographical isolation, is well documented1. How- ever, sympatric speciation, divergence without geographical iso- lation, is highly controversial2. Claims of sympatric speciation must demonstrate species sympatry, sister relationships, repro- ductive isolation, and that an earlier allopatric phase is highly unlikely1. Here we provide clear support for sympatric speciation in a case study of two species of palm (Arecaceae) on an oceanic island. A large dated phylogenetic tree shows that the two species of Howea, endemic to the remote Lord Howe Island, are sister taxa and diverged from each other well after the island was formed 6.9 million years ago3. During fieldwork, we found a substantial disjunction in flowering time that is correlated with soil prefer- ence. In addition, a genome scan4,5 indicates that few genetic loci are more divergent between the two species than expected under neutrality, a finding consistent with models of sympatric specia- tion involving disruptive/divergent selection2. This case study of sympatric speciation in plants provides an opportunity for refin- ing theoretical models on the origin of species, and new impetus for exploring putative plant and animal examples on oceanic islands. Speciation, the division of populations into evolutionarily inde- pendent units, involves genetic separation and phenotypic differen- tiation. Genetic divergence following geographic isolation gives rise to allopatric speciation: ���the conceptual rationale is simply that, given enough time, speciation is an inevitable consequence of populations evolving in allopatry���6. Numerous empirical examples support this uncontroversial scenario1. In theory, however, popu- lations can become genetically isolated without geographical sepa- ration, resulting in sympatric speciation, a much more contentious model. Sympatric speciation is more strictly defined as the emergence of two species from a population in which mating has been random with respect to the place of birth of the mating partners2. Mathematical models have shown that sympatric speciation is possible2,7���10, but very few examples have been documented in nature11,12. Cichlid fish seem to have radiated sympatrically in African crater lakes. Molecular phylogenetic analyses show that the fish species in each lake share a common ancestor, with sexual selection and ecology possibly driving speciation13. Second, races of apple and hawthorn maggot have shifted to different hosts in sympatry and differ in reproductive behaviour and breeding time14. Third, a genetic study of African indigobirds, which are host-specific brood parasites, showed that they might have recently speciated sympatrically after new hosts were colonized15. These examples are all from animal taxa with relatively large continental geographic distributions. This leaves the door open to controversy, given that truly convincing cases of sympatric speciation must involve biogeographic and phylogenetic histories that make the existence of an allopatric phase highly unlikely1. For this reason, we focused on a plant species-pair confined to a remote oceanic island. Lord Howe Island (Fig. 1a) is a small subtropical island of less than 12 km2, situated 580 km off the eastern coast of Australia. The island was formed by volcanic activity 6.4���6.9 million years (Myr) ago3. Lord Howe Island is the most southerly member of a 1,000-km chain of nine underwater volcanoes along the Lord Howe Rise. The closest link in the Lord Howe Island chain is Elizabeth Reef, 160 km to the north this seamount was an island 10.2 Myr ago3. Lord Howe Island LETTERS Figure 1 | Lord Howe Island and its endemic palms. a, Lord Howe Island is a World Heritage Site, and a permanent park preserve now protects 70% of the island. The waters surrounding Lord Howe Island are protected as a marine park, which also holds the world���s southernmost coral reef. The island is inhabited by approximately 300 residents, and less than 20% of the vegetation has been disturbed. b, The kentia or thatch palm, Howea forsteriana, is characterized by multiple spikes in each inflorescence and has straight leaves with drooping leaflets. c, The curly palm, H. belmoreana, bears a single spike in each inflorescence and has recurved leaves with ascending leaflets. 1Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3DS, UK. 2Centre for Evolutionary and Functional Ecology, UMR 5175, 1919 Route de Mende, 34293 Montpellier cedex 5, France. 3PO Box 157, Lord Howe Island, New South Wales 2898, Australia. 4Department of Systematic Botany, University of Aarhus, DK-8000 Aarhus C, Denmark. 5Department of Ecology and Evolution, University of Lausanne, CH-1015 Lausanne, Switzerland. Vol 441|11 May 2006|doi:10.1038/nature04566 210

�� 2006 Nature Publishing Group itself has eroded rapidly and will be awash within 200,000 years. Apart from Ball���s Pyramid, a precipitous sea stack 23 km southeast of Lord Howe Island that supports only limited plant life, there are no islands nearby, and Australia is the nearest land mass. However, the indigenous vascular flora of Lord Howe Island has greater affinities with New Zealand and New Caledonia than with Australia. Of its 241 plant species, almost half are endemic, and the terrestrial fauna of the island shows similar levels of endemism. Lord Howe Island represents an ideal location to study sympatric speciation because it has long been isolated, is of known age and is so small that geographical isolation on the island cannot realistically occur3. The palm family is represented on Lord Howe Island by four species in three strictly endemic genera. The two species of Howea, H. belmoreana and H. forsteriana, are extremely abundant, occurring in more than 70% of the island���s vegetation16. Howea forsteriana, the kentia palm, is one of the most widely traded houseplants in the world and is worth over e7 million per year in the Dutch horticul- tural industry alone. The two species of Howea display striking morphological differences and their taxonomic status is indisputa- ble17 (Fig. 1b, c). They occur sympatrically in numerous places on Lord Howe Island16, and yet putative hybrids have only rarely been reported18 our thorough fieldwork identified only five specimens with intermediate morphologies. We have also confirmed that both species are diploid (2n �� 32) using conventional cytological techniques, thereby excluding polyploid speciation19. We have produced the most comprehensive DNA-based phylo- genetic tree for the largest subfamily of palms (Arecoideae), com- prising 132 taxa, including all 67 genera of the Indo-Pacific Areceae20. These data strongly support the monophyly of Howea and a sister relationship to the monotypic Australian genus Laccospadix. Cor- recting for molecular rate heterogeneity across lineages, we have dated this tree using four calibration points simultaneously, all independent from Lord Howe Island. Using two different molecular dating methods21,22, we estimated the split between Howea and Laccospadix to be 4.57���5.53 Myr old, and that the two Howea species diverged 1.92 ^ 0.53 Myr ago (nonparametric rate smoothing (NPRS) Table 1) or just less than 1 Myr ago (bayesian Table 1), long after Lord Howe Island was formed. Other dates in the tree are also consistent with geological history, including the root nodes of the other two endemic Lord Howe Island palm genera, Hedyscepe and Lepidorrhachis, and the root node of Carpoxylon, which is younger than the age of Vanuatu23 on which it is endemic (Table 1). During fieldwork, we monitored both Howea species throughout the flowering season. Phenological data indicate that the species are reproductively isolated, with H. forsteriana flowering before H. belmoreana. The peak flowering of each species is separated by approximately six weeks and has limited overlap (Fig. 2). Howea forsteriana is protandrous at the population level, with male flowering peaking two weeks before female receptivity, whereas H. belmoreana is synchronous (Fig. 2). Notably, when H. forsteriana occurs on volcanic rather than calcareous substrates, asynchronous maturation is not observed (Supplementary Information t �� 0.49 (t-test), n �� 12, P �� 0.63). Thus, flowering-time differences seem to be directly influenced by substrate-induced physiological changes. We also found that Howea is wind-pollinated, a rare syndrome in palms (contrary to popular belief), and complete exclusion experi- ments demonstrated the absence of apomixis in both species. The distributions of H. forsteriana and H. belmoreana are also dependent on soil pH (Fig. 3). Howea belmoreana is restricted to neutral and acidic soils, whereas H. forsteriana prefers calcarenite, a recent basic sedimentary formation that dominates low-lying parts of the island3. The same pattern is observed for both adults and juveniles (r 2 �� 0.79, n �� 22 sites, P , 1024 for H. forsteriana r 2 �� 0.69, n �� 43, P , 1024 for H. belmoreana). Despite this pref- erence, both species occur sympatrically in 11 out of the 55 quadrats that contain palms (see Methods). Consistent with sympatric speciation, genetic divergence (F ST ) within the genome, estimated using 274 polymorphic amplified fragment-length polymorphism (AFLP) loci, follows an L-shaped distribution, with most loci showing low F ST and only a small number Figure 2 | Flowering phenology for each Howea species. H. belmoreana is shown in grey (n �� 198), H. forsteriana in black (n �� 177), with male (solid line) and female (dotted line) phases (see Methods). The flowering times of the two species are strongly displaced. In addition, both species are significantly different in their sexual synchrony index (SI F 1,222 �� 27.26, P , 1024). The SI is not significantly different from zero (n �� 132, t �� ���0.30 weeks (t-test), P �� 0.18) and does not differ among sites of H. belmoreana (F 8,131 �� 1.54, P �� 0.15), whereas H. forsteriana is strongly protandrous (mean SI �� 1.3 weeks, n �� 92, t �� 6.75, P , 1024) and there are significant differences among sites (F 8,83 �� 1.96, P �� 0.0619). Table 1 | Ages for root nodes (in Myr) calculated from molecular phylogenetic trees Taxa Node number* Distribution Node age ^ s.d. (NPRS) Node age ^ s.d. (bayesian) Acanthophoenix/Tectiphiala 1 Mascarenes��� 7.61 ^ 1.70 6.52 ^ 1.10 Dictyosperma (hurricane palm) 2 Mascarenes��� 7.75 ^ 1.83 6.60 ^ 1.22 Hyophorbe (bottle palm) 3 Mascarenes��� 7.75 ^ 0.02 7.45 ^ 0.38 Palms 4 Widespread�� 88.93 ^ 9.38 85.73 ^ 2.20 Carpoxylon 5 Vanuatuk 7.07 ^ 2.05 11.98 ^ 5.36 Hedyscepe (big mountain palm) 6 LHI 6.66 ^ 2.45 8.24 ^ 4.30 Lepidorrhachis (little mountain palm) 7 LHI 7.77 ^ 3.07 4.62 ^ 3.28 Howea 8 LHI 4.57 ^ 1.45 5.53 ^ 2.89 Howea belmoreana (curly palm)/Howea forsteriana (kentia palm) split 9 LHI 1.92 ^ 0.53 ,1.00{ LHI, Lord Howe Island. *Nodes 1���4 are calibration points. See also Supplementary Information. ���Acanthophoenix occurs on La Reunion �� (2 Myr) and Mauritius (7.8 Myr) Tectiphiala occurs on Mauritius only. ���Dictyosperma and Hyophorbe occur on La Reunion, �� Mauritius and Rodrigues. ��Santonian fossil, Sabalites magothiensis, 83.5 Myr old (ref. 30). kThe age of Vanuatu is about 10 Myr, although the exact figure is disputed23. Therefore, we did not use this point as a calibration it is given here for comparative purposes only. {Age under the 1-Myr detection level in the bayesian dating calculation of this data set. NATURE|Vol 441|11 May 2006 LETTERS 211