1 Taxon sampling and inferences about diatom phylogeny

Abstract

Proper taxon sampling is one of the greatest challenges to understanding phylogenetic relationships, perhaps as important as choice of optimality criterion or data type. This has been demonstrated in diatoms where centric diatoms may either be strongly supported as monophyletic or paraphyletic when analyzing SSU rDNA sequences using the same optimality criterion. The effect of ingroup and outgroup taxon sampling on relationships of diatoms is explored for diatoms as a whole and for the order Thalassiosirales. In the latter case, SSU rDNA and rbcL sequence data result in phylogenetic relationships that appear to be strongly incongruent with morphology and broadly incongruent with the fossil record. For example, Cyclotella stelligera Cleve & Grunow behaves like a rogue taxon, jumping from place to place throughout the tree. Morphological data place C. stelligera near the base of the freshwater group as sister to the extinct genus Mesodictyon Theriot and Bradbury, suggesting that it is an old, long branch that might be expected to “misbehave” in poorly sampled trees. Cyclotella stelligera and C. bodanica Grunow delimit the diameter of morphological diversity in Cyclotella, so increased sampling of intermediate taxa will be critical to resolving this part of the tree. Morphology is sampled for a much greater number of taxa and many transitional states of putative synapomorphies seem to suggest a robust morphological hypothesis. The Thalassiosirales are unstable with regards to taxon sampling in the genetic data, suggesting that perhaps the morphological hypothesis is (for now) preferable.