Importance of synonymous substitutions under dense taxon sampling and appropriate modeling in reconstructingthemitogenomic tree of Eutheria

Abstract

Although molecular phylogenetics is a strong tool for reconstructing the tree of life, many problems persist due to systematic errors caused by model mis-specifications. Resolving misconstructed trees should lead us to better understand the processes of molecular evolution. Mammalian mitogenomes provide us with a good opportunity in this respect, because the mammalian tree is well established on the basis of multiple nuclear genes, and mitogenome trees are sometimes in conflict with it, for example concerning the positions of tarsiers and colugos.The utility of mitogenomes as a phylogenetic marker is therefore sometimes questioned, and an important problem is whether any method can overcome the misleading phylogenetic signals of mitogenomes. Here we show that the maximum likelihood tree of 463 eutherianmitogenomes reconstructed from nucleotide sequences of protein-encoding genes gives positions of tarsiers and colugos that are consistent with the well-established nuclear tree; this is the first study to obtain a consistent tree with respect to the positions of tarsiers and colugos using mitogenomes. Furthermore, our mitogenome tree of the 463 eutherians is mostly consistent with the nuclear gene tree. Previous mitogenomic studies have been hampered by sparse taxon sampling, and our analysis demonstrates the importance of dense taxon sampling to relieve the misleading phylogenetic signals of mitogenomes. However, because there are many convergent and parallel substitutions in the amino acid sequences, the effect of dense taxon sampling on the accuracy of tree reconstruction seems to be very limited. We further show the importance of using synonymous substitutions with dense taxon sampling as well as with appropriate modeling in recovering the well-established tree from lower to even higher levels of eutherian phylogeny.