Tree rooting with outgroups when they differ in their nucleotide composition from the ingroup: the Drosophila saltans and willistoni groups, a case study.

Rooting is frequently the most precarious step in any phylogenetic analysis. Outgroups can become useless for rooting if they are too distantly related to the ingroup. Specifically, little attention has been paid to scenarios where outgroups have evolved different nucleotide frequencies from the ingroup. We investigate one empirical example that arose seeking to determine the phylogenetic relationship between the saltans and the willistoni groups of Drosophila (subgenus Sophophora). We have analyzed 2085 coding nucleotides from the xanthine dehydrogenase (Xdh) gene in 14 species, 6 from the saltans group and 8 from the willistoni group. We adopt a two-step strategy: (1) we investigate the phylogeny without outgroups, rooting the network by the midpoint method; (2) we reinvestigate the rooting of this phylogeny using predefined outgroups in both a parsimony- and a model-based maximum-likelihood framework. A satisfactory description of the substitution process along the Xdh region calls for six substitution types and substitution rate variation among codon positions. When the ingroup sequences are considered alone, the phylogeny obtained using this description corroborates the known relationships derived from anatomical criteria. Inclusion of the outgroups makes the root unstable, apparently because of differences between ingroups and outgroups in the substitution processes; these differences are better accounted for by a simplified model of evolution than by more complex, realistic descriptions of the substitution process.