Compositional heterogeneity and outgroup choice influence the internal phylogeny of the ants

Knowledge of the internal phylogeny and evolutionary history of ants (Formicidae), the world's most species-rich clade of eusocial organisms, has dramatically improved since the advent of molecular phylogenetics. A number of relationships at the subfamily level, however, remain uncertain. Key unresolved issues include placement of the root of the ant tree of life and the relationships among the so-called poneroid subfamilies. Here we assemble a new data set to attempt a resolution of these two problems and carry out divergence dating, focusing on the age of the root node of crown Formicidae. For the phylogenetic analyses we included data from 110 ant species, including the key species Martialis heureka. We focused taxon sampling on non-formicoid lineages of ants to gain insight about deep nodes in the ant phylogeny. For divergence dating we retained a subset of 62 extant taxa and 42 fossils in order to approximate diversified sampling in the context of the fossilized birth-death process. We sequenced 11 nuclear gene fragments for a total of ∼7.5 kb and investigated the DNA sequence data for the presence of among-taxon compositional heterogeneity, a property known to mislead phylogenetic inference, and for its potential to affect the rooting of the ant phylogeny. We found sequences of the Leptanillinae and several outgroup taxa to be rich in adenine and thymine (51% average AT content) compared to the remaining ants (45% average). To investigate whether this heterogeneity could bias phylogenetic inference we performed outgroup removal experiments, analysis of compositionally homogeneous sites, and a simulation study. We found that compositional heterogeneity indeed appears to affect the placement of the root of the ant tree but has limited impact on more recent nodes. Our findings have implications for outgroup choice in phylogenetics, which should be made not only on the basis of close relationship to the ingroup, but should also take into account sequence divergence and other properties relative to the ingroup. We put forward a hypothesis regarding the rooting of the ant phylogeny, in which Martialis and the Leptanillinae together constitute a clade that is sister to all other ants. After correcting for compositional heterogeneity this emerges as the best-supported hypothesis of relationships at deep nodes in the ant tree. The results of our divergence dating under the fossilized birth-death process and diversified sampling suggest that the crown Formicidae originated during the Albian or Aptian ages of the Lower Cretaceous (103-124 Ma). In addition, we found support for monophyletic poneroids comprising the subfamilies Agroecomyrmecinae, Amblyoponinae, Apomyrminae, Paraponerinae, Ponerinae, and Proceratiinae, and well-supported relationships among these subfamilies except for the placement of Proceratiinae and (Amblyoponinae + Apomyrminae). Our phylogeny also highlights the non-monophyly of several ant genera, including Protanilla and Leptanilla in the Leptanillinae, Proceratium in the Proceratiinae, and Cryptopone, Euponera, and Mesoponera within the Ponerinae.