Evaluating placental inter-ordinal phylogenies with novel sequences including RAG1, γ-fibrinogen, ND6, and mt-tRNA, plus MCMC-driven nucleotide, amino acid, and codon models

It is essential to test a priori scientific hypotheses with independent data, not least to partly negate factors such as gene-specific base composition biases misleading our models. Seven new gene segments and sequences plus Bayesian likelihood phylogenetic methods were used to compare and test five recent placental phylogenies. These five phylogenies are similar to each other, yet quite different from Fthose of previously proposed trees, and span Waddell et al. [Syst. Biol. 48 (1999) 1] to Murphy et al. [Science 294 (2001b) 2348]. Trees for RAGI, gamma-fibrinogen, ND6, mt-tRNA, mt-RNA, c-MYC, epsilon-globin, and GHR are significantly congruent with the four main groups of mammals common to the five phylogenies, i.e., Afrotheria, Laurasiatheria, Euarchontoglires, Xenarthra plus Boreoeutheria (Laurasiatheria plus Euarchontoglires). Where these five a priori phylogenies differ, remain areas generally hard to resolve with the new sequences. The root remains ambiguous and does not reject a basal Afrotheria (the Exafroplacentalia hypothesis), Afrotheria plus Xenarthra together with basal (Atlantogenata), or Epitheria (Xenarthra basal) convincingly. Good evidence is found that Eulipotyphla is monophyletic and is located at the base of Laurasiatheria. The shrew mole., Uropsilus, is found to cluster consistently with other moles, while Solenodon may be the sister taxa to all other eulipotyphlans. Support is found for a probable sister pairing of just hedgehogs/gymnures and shrews. Relationships within Afrotheria, except the Paenungulata clade, remain hard to resolve, although there is congruent support for Afroinsectiphillia (aardvark, elephant shrews golden moles, and tenrecs). A first-time use is made of MCMC enacted general time-reversible (GTR) amino acid and codon-based models for general tree selection. Even with ND6, a GTR amino acid model provided resolution of fine features, such as the sister group relationship of walrus to Otatriidae, and with BRCA a more reasonable rooting. An extensive analysis of GHR sequences reveals strong congruence with prior phylogenies, including strong support for Eulipotyphla, and good resolution within Rodentia. A codon model gives a worse likelihood than a nucleotide model and sometimes switches support, e.g., with RAG1 + gamma-fibrinogen from a hyrax-sirenian association to support for Tethytheria. An analysis of the concatenated data is in accordance with well-resolved features of the gene trees. Taken all together, this work suggests that we are on the right path finding strong confirmation of prior phylogenies. However, with the use of robust criteria for assessing trees (i.e., not Bayesian posteriors), it is apparent parts of the tree remain hard to resolve. Since our current models are far from fitting the sequence data, we should continue with our exploratory analyses to arrive at a refined set of hypotheses for future testing using more model independent characters (e.g., rare indels. gene rearrangement, and SINE data). (C) 2003 Elsevier Science (USA). All rights reserved.