Testing for misleading effects in the phylogenetic reconstruction of ancient lineages of hexapods: influence of character dependence and character choice in analyses of 28S rRNA sequences

The present analyses employ the almost complete sequence of the 28S rRNA gene to investigate phylogenetic relationships among Pancrustacea, placing special emphasis on the position of basal hexapod lineages. This study utilizes a greater number of characters and taxa of Protura, Collembola and Diplura than previous analyses to focus on conflicts in the reconstruction of the early steps in hexapod evolution. Phylogenetic trees are mainly based on Bayesian approaches, but likewise include analyses with Maximum Likelihood and Maximum Parsimony. Different analyses, including the application of a mixed DNA/RNA substitution model, were performed to narrow possible misleading effects of non-stationarity of nucleotide frequencies, saturation and character independence down to a minimum. This is the first time that a mixed DNA/RNA model is applied to analyse 28S rRNA sequences of basal hexapods. All methods yielded strong support for the monophyly of Collembola, Diplura, Dicondylia and Insecta s.str., as well as for a cluster composed of Diplura and Protura (‘Nonoculata-hypothesis’). However, the last cluster may be an artifact caused by a shared GC bias of the 28S sequences between these orders, in combination with a long branch effect. The instability of the position of the ‘Nonoculata’ within Pancrustacea further bears out the misleading effect of non-stationarity of nucleotide frequencies. Protura and Diplura either form the sister-group to Collembola (Entognatha) or cluster with branchiopod crustaceans. Overall, the phylogenetic signal of the complete sequences of the 28S rRNA gene favours monophyly of Hexapoda over paraphyly. However, further corroboration from independent data is needed to rule out the competing hypothesis of mutually paraphyletic Crustacea and Hexapoda.