Grass-evolution and diversification : a phylogenetic approach

Abstract

The growth in size of phylogenetic trees, over the last 20 years, has allowed evolutionary biologists to better test hypotheses about the evolutionary history of organisms, and especially those of species rich taxa such as the grasses. Grasses are one of the most diverse families in the angiosperms, consisting of approximately 10,000 species and 600-700 genera and it is essential to investigate evolution and diversification in this group to advance the understanding of the processes shaping the diversity of its life forms. Therefore, this thesis aimed to provide comprehensive phylogenetic trees of the grass family in order to establish macro-evolutionary hypotheses and investigate patterns and processes of grass diversification. One aspect of this thesis was to infer the most comprehensive phylogenetic tree of the grasses in order to establish robust phylogenetic relationships among grass lineages. In Chapter 2, a much larger representation of grass diversity (82% of tribes and 42% of genera) was included than any previous study. Phylogenetic inferences using DNA sequences of three plastid regions: rbcL, matK and trnL-F were performed using maximum parsimony and Bayesian inferences. The resulting trees resolved most of the subfamily relationships within the BEP(Bambusoideae, Ehrhartoideae and Pooideae) and PACCAD (Panicoideae. Aristidoideae. Centothecoideae, Chloridoideae, Arundinoideae and Danthonioideae) clades, which had previously been unclear, such as, among others: (i) the composition of the BEP clade and the sister-relationship of Ehrhartoideae and Bambusoideae + Pooideae, (ii)the paraphyly of tribe Bambuseae, (iii) the position of Gynerium as sister to Panicoideae, and (iv) the monophyly of Eriachne + Micraira. The thesis also highlights how phylogenetic accuracy has been largely neglected in phylogenetic studies of grasses and other organisms with respect to missing data. It is shown that accuracy can be maintained even with the presence of a relatively large amount of missing data in combined analyses (i.e. 33 % of the taxa lacking one or more genes in the combined analysis). However, bootstrap support values, and to a lesser extent Bayesian inference posterior probabilities, are generally louver in combined gene analyses involving missing data than those not including them. We propose a fully resolved tree for the grass family at subfamily level and indicate the most likely interrelationships of all included tribes in our analysis (i.e. 82% of total grass tribes) …