Phylogenetic Uncertainty and Taxonomic Re-revisions: An Example from the Australian Short-necked Turtles (Testudines: Chelidae)

Abstract

Molecular data have greatly influenced our concepts of species and their relationships in the last few decades, and as a consequence the taxonomy of most vertebrate clades has been repeatedly revised to reflect phylogeny. However, as larger and more complete molecular data sets become available, the sometimes striking disparities between taxonomic revisions based on individual gene trees (particularly those based on mitochondrial DNA) and species trees has become increasingly apparent. Here, we present data from 13 nuclear and one mitochondrial gene. Our results demonstrate that the recent taxonomic proposal erecting the new Australian chelid genus Flaviemys (Testudines: Chelidae) was an unnecessary action, and that recognition of Flaviemys confuses, rather than clarifies, a phylogeny-based taxonomy of the group. Taxonomic actions have many broad repercussions, and we recommend that taxonomic changes should be proposed cautiously and only when they are based on the strongest possible data and analyses. D URING the last decade, turtles (Testudines), perhaps more than any other vertebrate group, have been the focus of extensive taxonomic revisions. Essentially every major clade of turtle has been revised to some degree based on the results of one or more molecular phylogenetic analysis. Over the same time period, molecular systematics has grown as a field, both in terms of the amount of molecular data that are often available and the sophistication of downstream analyses. Consequently, one should expect that some of these recent taxonomic decisions may end up being partly or fully revised, reflecting progress in the phylogenetic knowledge on which they were based. Most recently, a number of studies (including some of our own) have analyzed one or more mitochondrial genes, perhaps augmented with a few nuclear loci, recovered trees that are mostly or fully resolved and well supported by the data in hand, discovered apparently paraphyletic taxa, and revised taxonomy under the guiding principle that named lineages should be monophyletic. However, as we now recognize, while the recovered trees might be well supported, the phylogenies may not accurately reflect the actual species phylogeny, particularly if one or a few gene trees dominate the analysis. This has led to a great deal of taxonomic confusion as names are changed, changed again, and sometimes changed back to the original configuration based on trees from different analyses that are each well supported but incongruent with one another. Most of these previous analyses of testudines are based on analyses of relatively few independent loci (i.e., 1–6 mitochondrial DNA genes plus 1–6 nuclear loci). For some groups, this level of gene sampling appears adequate, while for others, phylogenies based on sparse gene sampling are inadequate. For example, Naro-Maciel et al. (2008) generated phylogenies for the sea turtles (Chelonioidea) from mtDNA (two genes) and nuDNA (five loci). Analyses of both data sets independently recovered the same tree topology, suggesting that the phylogeny for the sea turtles appears to be well resolved, and as a consequence the resulting taxonomy is probably stable. Revisions of other groups now appear to have been premature as subsequent phylogenetic analyses, based on expanded taxon or data sampling, produced phylogenies that are incongruent with those upon which the earlier taxonomic revisions were based. As one recent example, Iverson et al. (2013) generated phylogenies for the mud and musk turtles (family Kinosternidae) based on three mitochondrial DNA (mtDNA) and three nuclear DNA (nuDNA) loci, recovered the longrecognized turtle genus Kinosternon as paraphyletic with respect to Sternotherus, and reassigned six species of Kinosternon to the new genus Cryptochelys. However, in a follow up analysis, Spinks et al. (2014) generated phylogenies for the Kinosternidae based on 14 nuclear loci and recovered Kinosternon as monophyletic with respect to Sternotherus with strong support, but Cryptochelys as nonmonophyletic with respect to the more restricted Kinosternon. In this case, the tree topology of Iverson et al. (2013) was driven by mitochondrial sequence variation, and not subsequently well supported by more extensive nuclear data. Based on the non-monophyly of Cryptochelys, Spinks et al. (2014) suggested that the recognition of Cryptochelys was premature because the phylogeny of the Kinosternidae is not stable (see also the Guidelines for Taxonomic Changes in Turtle Taxonomy Working Group [TTWG], 2014). Although taxonomy should always be based on the best available data, and data and analyses are always subject to change and revision, we also recognize that it is now much easier to bring larger, more robust data sets to critically important taxonomic revisions. Such data sets should help settle the sometimes unsettled taxonomy surrounding many groups.