Deep metazoan phylogeny: When different genes tell different stories

Abstract

Molecular phylogenetic analyses have produced a plethora of controversial hypotheses regarding the patterns of diversification of non-bilaterian animals. To unravel the causes for the patterns of extreme inconsistencies at the base of the metazoan tree of life, we constructed a novel supermatrix containing 122 genes, enriched with non-bilaterian taxa. Comparative analyses of this supermatrix and its two non-overlapping multi-gene partitions (including ribosomal and non-ribosomal genes) revealed conflicting phylogenetic signals. We show that the levels of saturation and long branch attraction artifacts in the two partitions correlate with gene sampling. The ribosomal gene partition exhibits significantly lower saturation levels than the non-ribosomal one. Additional systematic errors derive from significant variations in amino acid substitution patterns among the metazoan lineages that violate the stationarity assumption of evolutionary models frequently used to reconstruct phylogenies. By modifying gene sampling and the taxonomic composition of the outgroup, we were able to construct three different yet well-supported phylogenies. These results show that the accuracy of phylogenetic inference may be substantially improved by selecting genes that evolve slowly across the Metazoa and applying more realistic substitution models. Additional sequence-independent genomic markers are also necessary to assess the validity of the phylogenetic hypotheses.