Inference of phylogenetic relationships among the subfamilies of grasses (Poaceae: Poales) using meso-scale exemplar-based sampling of the plastid genome

Abstract

To clarify phylogenetic relationships among grass (Poaceae) subfamilies and to better resolve the placement of Poaceae in Poales, we surveyed 17 plastid genes and associated noncoding regions (∼15.5 kb per taxon, about 1/10 of the plastid genome) for exemplar representatives from 10 grass subfamilies and a broad sample of related monocots. We found general concordance in relationships and support levels among gene regions and data partitions across analyses, with some exceptions in Bayesian analyses. Different phylogenetic criteria generally agreed on backbone relationships, and the support values we inferred were generally as good as or better than those in other studies that employed more taxa for fewer genes to estimate the same backbone. Within grasses, we found robust support for the monophyly of subfamily Anomochlooideae and for the Bambusoideae–Ehrhartoideae–Pooideae (BEP) clade, and moderate support for a sister-group relationship between Bambusoideae and Pooideae. Most relationships in the strongly supported Panicoideae–Aristidoideae–Chloridoideae–Micrairoideae–Arundinoideae–Danthonioideae (PACMAD) clade were not resolved consistently, probably because the current intensive sampling of genes is still insufficient to resolve short internal branches. Our data infer a well-supported clade that includes the grass family and two small families of grass-like plants, Ecdeiocoleaceae and Joinvilleaceae, but we did not satisfactorily resolve the relationships among these three families. A generally accelerated substitution rate in Poaceae plastid genomes is shared with some, but not all, lineages that are closely related to grasses in the larger clade of commelinid monocots, which may complicate inference of the sister group of the grasses in all current studies.