Localized Phylogenetic Discordance Among Nuclear Loci Due to Incomplete Lineage Sorting and Introgression in the Family of Cotton and Cacao (Malvaceae)

The economically important cotton and cacao family (Malvaceae sensu lato) have long been recognized as a monophyletic group. However, the relationships among some subfamilies are still unclear as discordant phylogenetic hypotheses keep arising when different sources of molecular data are analyzed. Phylogenetic discordance has previously been hypothesized to be the result of both introgression and incomplete lineage sorting (ILS), but the extent and source of discordance have not yet been evaluated in the context of loci derived from massive sequencing strategies and for a wide representation of the family. Furthermore, no formal methods have been applied to evaluate if the detected phylogenetic discordance among phylogenomic datasets influences phylogenetic dating estimates of the concordant relationships. The objective of this research was to generate a phylogenetic hypothesis of Malvaceae from nuclear genes, specifically we aimed to (1) investigate the presence of major discordance among hundreds of nuclear gene histories of Malvaceae; (2) evaluate the potential source of discordance; and (3) examine whether discordance and loci heterogeneity influence on time estimates of the origin and diversification of subfamilies. Our study is based on a comprehensive dataset representing 96 genera of the nine subfamilies and 268 nuclear loci. Both concatenated and coalescence-based approaches were followed for phylogenetic inference. Using branch lengths and topology, we located the placement of introgression events to directly evaluate whether discordance is due to introgression rather than ILS. To estimate divergence times, concordance and molecular rate were considered. We filtered loci based on congruence with the species tree and then obtained the molecular rate of each locus to distribute them into three different sets corresponding to shared molecular rate ranges. Bayesian dating was performed for each of the different sets of loci with the same parameters and calibrations. Phylogenomic discordance was detected between methods, as well as gene histories. At deep coalescent times, we found discordance in the position of five subclades probably due to ILS and a relatively small proportion of introgression. Divergence time estimation with each set of loci generated overlapping clade ages, indicating that, even with different molecular rate and gene histories, calibrations generally provide a strong prior.