Correct and incorrect vertebrate phylogenies obtained by the entire mitochondrial DNA sequences

Concatenated sequences of all protein-coding genes in mitochondria recovered a known phylogeny of 11 vertebrate species correctly with statistical significance. However, when it was rooted by lampreys or sea urchins, the root of the vertebrate tree was placed between the mammal cluster and the chicken-frog-fish cluster or between the mammal-chicken cluster and the frog-fish cluster, depending on the tree-making method used. Although the frog-fish or chicken-frog-fish cluster was biologically incorrect, it was again supported with a significantly high bootstrap value. In this study, we investigated the reasons why this happened. It has been suggested that an incorrect phylogeny may be constructed due to a change of amino acid composition in different lineages or due to homoplasies at sites with hydrophobic amino acids. However, our results indicated that these were not the causes of the incorrect rooting of the vertebrate tree. Rather, it was important to take into account an extensive rate variation across sites and different probabilities of substitution among different amino acids. The substitution rates for mitochondrial sequences vary considerably for different vertebrate lineages. In such a case, it is known to be important to use the model that reflects the actual substitution probability to obtain a correct tree topology. The correct rooting of the vertebrate tree was recovered when rate variation across sites was properly accounted for.