Fifteen clues to the early diversification of flowers: first results from the eFLOWER initiative

Flowers are central to the biology and evolution of angiosperms (flowering plants). Here, we focus on the history of floral diversification through deep evolutionary time. A comprehensive understanding of global patterns of floral evolution has been prevented so far by the lack of an adequate morphological data set spanning all major angiosperm lineages. Using a new flexible and innovative approach, centered on a multi-user database (PROTEUS), we have built such a data set. Unlike most earlier studies of character evolution at the scale of angiosperms, we have recorded exemplar species instead of higher taxa such as genera or families. This approach allows for explicit reconstructions without assumptions about ancestral states or monophyly of supraspecific taxa, although we acknowledge that it also entails the risk of undersampling morphological variation. Importantly, the exemplar approach allows a direct match with the species sampled in molecular phylogenetic trees, providing the possibility to take branch lengths into account and to explore character evolution with probabilistic models. We report our first results from a data set of floral traits for 792 species representing all orders and 367 families (87%) of angiosperms, sampled from a recently published timetree calibrated with 136 fossil age constraints. Using this data set and tree, we reconstruct ancestral states for 26 key floral traits using parsimony, maximum likelihood, and Bayesian approaches and infer ancestral flowers for 15 key nodes (incl. Angiospermae, Mesangiospermae, Magnoliidae, Monocotyledonae, Eudicotyledonae, Pentapetalae, Rosidae, and Asteridae) as well as all extant orders of flowering plants. Further, we explore the sensitivity of our results to phylogenetic, dating, and model uncertainty. Importantly, all of our analyses converge on the same results in most cases, except near the root of the angiosperm tree, where parsimony continues to provide equivocal answers for some important floral traits, consistent with previous work. Lastly, the use of an explicit temporal framework for reconstructing ancestral states allows, for the first time, a direct comparison of inferred ancestral flowers with the known fossil record of each geological time interval. These results shed new light on long-standing questions on floral evolution across angiosperms, with important implications for a wide range of disciplines including Evo-Devo, pollination biology, and paleobotany.