Phylogenomic dating and Bayesian biogeography illuminate an antitropical pattern for eucerine bees

Abstract

Aim An antitropical pattern is characterized by the occurrence of closely related taxa south and north of the tropics but absent or uncommonly represented closer to the equator, in contrast to most taxa, which tend to have their highest diversity in the tropical regions. We investigate the antitropical distribution of eucerine bees with the aim of contributing to the characterization and understanding of this pattern. Location All continents except Antarctica and Australia. Taxon Eucerine bees (Hymenoptera: Apidae: Eucerinae). Methods We carried out phylogenomic dating under two different clock models and used multiple strategies to vary matrix composition, evaluating the overlapping of divergence times estimated across models using Bhattacharyya coefficients. Lastly, we reconstructed the biogeographic history of eucerine bees using a Bayesian implementation of the DEC model. Results Eucerinae is estimated to have started diversifying during the Palaeocene, with all its tribes originating during the Palaeocene/Eocene transition in southern South America. At least two range expansions happened into North America before the full closure of the Isthmus of Panama. We show that divergence between closely related groups with disjunct distributions would have happened in periods when the climate favoured the expansion of open habitats and became isolated when the forests were re-established. Main conclusions We describe the early diversification of eucerine bees, revealing an intimate association with southern South America. Events of range evolution of Eucerinae were likely affected by periods of global cooling and aridification, and palaeoclimatic and vegetational conditions probably have been more relevant to the formation of the antitropical distribution of Eucerinae than the consolidation of the Isthmus of Panama connecting the Americas. We also demonstrate that most uncertainty in divergence time estimation is not due to the amount of molecular data being used, but more likely other factors like fossil calibrations and violations of clock models.