Phylogeny of the <i>Centaurea</i> group (Centaurea, Compositae): geography is a better predictor than morphology

The Centaurea group is part of the Circum-Mediterranean Clade (CMC) of genus Centaurea subgenus Centaurea, a mainly Mediterranean plant group with more than 200 described species. The group is traditionally split on morphological basis into three sections: Centaurea, Phalolepis and Willkommia. This division, however, is doubtful, especially in light of molecular approaches. In this study we try to resolve this phylogenetic problem and to consolidate the circumscription and delimitation of the entire group against other closely related groups. We analyzed nuclear (internal transcribed spacer of the ribosomal genes) and chloroplast (rpl32-trnL intergenic spacer) DNA regions for most of the described species of the Centaurea group using phylogenetic and network approaches, and we checked the data for recombination. Phylogeny was used to reconstruct the evolution of the lacerate-membranaceous bract appendages using parsimony. The magnitude of incomplete lineage sorting was tested estimating the effective population sizes. Molecular dating was performed using a Bayesian approach, and the ancestral area reconstruction was conducted using the Dispersal–Extinction–Cladogenesis method. Monophyly of the Centaurea group is confirmed if a few species are removed. Our results do not support the traditional sectional division. There is a high incongruence between the two markers and between genetic data and morphology. However, there is a clear relation between geography and the structure of the molecular data. Diversification in the Centaurea group mainly took place during the Pliocene and Pleistocene. The ancestral area infered for the Circum-Mediterranean Clade of Centaurea is the Eastern Mediterranean, whereas for the Centaurea group it is most likely NW-Africa. The large incongruencies, which hamper phylogenetic reconstruction, are probably the result of introgression, even though the presence of incomplete lineage sorting as an additional factor cannot be ruled out. Convergent evolution of morphological traits may have led to incongruence between morphology-based, traditional systematics and molecular results. Our results also cast major doubts about current species delimitation