Species delimitation, patterns of diversification and historical biogeography of the Neotropical frog genus Adenomera (Anura, Leptodactylidae)

Aim For many taxa, inaccuracy of species boundaries and distributions hampers inferences about diversity and evolution. This is particularly true in the Neotropics where prevalence of cryptic species has often been demonstrated. The frog genus Adenomera is suspected to harbour many more species than the 16 currently recognized. These small terrestrial species occur in Amazonia, Atlantic Forest (AF), and in the open formations of the Dry Diagonal (DD: Chaco, Cerrado and Caatinga). This widespread and taxonomically complex taxon provides a good opportunity to (1) test species boundaries, and (2) investigate historical connectivity between Amazonia and the AF and associated patterns of diversification. Location Tropical South America east of the Andes. Methods We used molecular data (four loci) to estimate phylogenetic relationships among 320 Adenomera samples. These results were integrated with other lines of evidence to propose a conservative species delineation. We subsequently used an extended dataset (seven loci) and investigated ancestral area distributions, dispersal-vicariance events, and the temporal pattern of diversification within Adenomera. Results Our conservative delineation identified 31 Confirmed Candidate Species (four remaining unconfirmed) representing a 94% increase in species richness. The biogeographical analysis suggested an Amazonian origin of Adenomera with as many as three dispersals to the DD and one to the AF during the Miocene. These dispersals were associated with habitat shifts from forest towards open habitats. Main conclusions The DD played a major role in the history of Adenomera in limiting dispersal and favouring diversification of open-habitat lineages. Moreover, a forest bridge during the Miocene Climatic Optimum may have permitted dispersal from Amazonia towards the AF and subsequent diversification. Uncovering species boundaries and distributions might drastically change inferences based on currently perceived distribution patterns.