Molecular phylogeny of African Rytidosperma–affiliated danthonioid grasses reveals generic polyphyly and convergent evolution in spikelet morphology

Abstract

AbstractOf the 90–95 species included in the Rytidosperma clade of grasses (Danthonioideae: Poaceae) about 20 are native to southern Africa, most of these belonging to the genera Karroochloa, Schismus and Tribolium. Whereas morphological data identify all three genera as being monophyletic, the robustness of this pattern is compromised by high levels of homoplasy. Here we use molecular data (plastid trnL–trnF and nuclear ITS sequences) to re–evaluate phylogenetic relationships within the Rytidosperma clade, with the specific aim of testing the monophyly of the African genera. Separate parsimony analyses of the two partitions reveal low levels of incongruence, which may be attributable to incomplete lineage sorting or past hybridization. Analyses of the combined data produce well resolved, strongly supported topologies that are virtually identical, regardless of the inference method used (parsimony or Bayesian inference) or whether or not "conflict" taxa are included. Our data support the monophyly of the Rytidosperma clade (synapomorphies include a punctate–ovate hilum, a large embryo, and lodicule microhairs) as well as the monophyly of the Australasian–South American Rytidosperma s.l. However, the monophyly of Karroochloa, Schismus and Tribolium is strongly contradicted, all three genera being polyphyletic as currently defined. Although not implemented here, a minimum of five new combinations would be required to rectify this situation. Character reconstructions reveal extensive convergence in a number of morphological features, especially those previously used to define genera. We suggest that the convergent pattern of evolution displayed by these traits, most of which describe spikelet/diaspore morphology, is linked to their functionality in seed dispersal and burial, and that this has compromised their utility as phylogenetic markers.