DNA/DNA hybridization studies of the carnivorous marsupials. I: The intergeneric relationships of bandicoots (Marsupialia: Perameloidea)

Publisher Website
Google Scholar
Abstract
A complete suite of comparisons among six bandicoot species and one outgroup marsupial was generated using the hydroxyapatite chromatography method of DNA/DNA hybridization; heterologous comparisons were also made with three other bandicoot taxa. Matrices of ΔTm's, Δmodes, and ΔT50Hs were generated and corrected for nonreciprocity, homoplasy, and, in the case of ΔTm's, normalized percent hybridization; these matrices were analyzed using the FITCH algorithm in Felsenstein's PHYLIP (version 3.1). Uncorrected and nonreciprocity-corrected matrices were also jackknifed and analyzed with FITCH to test for consistency. Finally, sample scores for ΔTm, Δmode, and ΔT50H matrices were bootstrapped and then subjected to phylogenetic analysis. These manipulations were carried out, in part, to address criticisms of the statistics used to summarize DNA/DNA hybridization (especially T50H) and the method itself. However, with the exception of an unresolved trichotomy among the twoEchymipera species andPeroryctes longicauda, all trees showed the same branchpoints. Except in the case of the tree generated from reciprocal-corrected ΔTm data, nodes were stable under jackknifing; and, again excepting the above-mentioned trichotomy, all nodes were supported by 95% or more of the bootstrapped trees. These results suggest that, despite arguments to the contrary, all three summary statistics can be valid for DNA/DNA hybridization data. Of taxonomic interest is the placement ofEchymipera spp. andPeroryctes longicauda together and separate from the more distantPeroryctes raffrayanus; the genusPeroryctes is thus at least paraphyletic. The trees further groupedEchymipera-plus-Peroryctes as the sister group ofIsoodon-plus-Perameles. Limited hybridizations withMacrotis lagotis suggest that its current position as representative of an entirely distinct family of perameloids is correct.