Madagascar sheds new light on the molecular systematics and biogeography of grammitid ferns: New unexpected lineages and numerous long-distance dispersal events.

Based on a worldwide phylogenetic framework filling the taxonomic gap of Madagascar and surrounding islands of the Western Indian Ocean (WIO), we revisited the systematics of grammitid fern species (Polypodiaceae). We also investigated the biogeographic origin of the extant diversity in Madagascar and estimated the relative influence of vicariance, long-distance dispersals (LDD) and in situ diversification. Phylogenetic inferences were based on five plastid DNA regions (atpB, rbcL, rps4-trnS, trnG-trnR, trnL-trnF) and the most comprehensive taxonomic sampling ever assembled (224 species belonging to 31 out of 33 recognized grammitids genera). 31 species from Madagascar were included representing 87% of the described diversity and 77% of the endemics. Our results confirmed a Paleotropical clade nested within an amphi-Atlantic grade. In addition, we identified three new major clades involving species currently belonging to Grammitis s.l., Ctenopterella and Enterosora. We resolved for the first time Grammitis s.s. as monophyletic, and Ctenopterella (newly tested here) and Enterosora as polyphyletic. The Neotropical genus Moranopteris was shown to also occur in Madagascar through a newly discovered species. Most importantly, we suggest a >30% inflation of the species number in the WIO due to the hidden diversity in >10 cryptic lineages, best explained by high morphological homoplasy. Molecular dating and ancestral areas reconstruction allowed identifying the Neotropics as the predominant source of LDD to the African-WIO region, with at least 12 colonization events within the last 20Ma. Repeated eastward migrations may be explained by transoceanic westerly winds transporting the dust-like spores. Tropical Asia s.l. would also have played a (minor) role through one dispersal event to Madagascar at the end of the Oligocene. Last, within the complex Malagasy region made of a mosaic of continental and oceanic islands located close to the African continent, we showed that contrary to theoretical expectations and empirical evidence in angiosperms, Africa does not act as a dispersal source and Madagascar seems to have a more important influence on the regional dynamics: we observed both in situ species diversification and dispersal out of Madagascar. This influence also extends beyond the region, since one dispersal event probably originated from Madagascar and reached the Subantarctic island of Amsterdam.