Incongruent phylogenies and its implications for the study of diversification, taxonomy and genome size evolution of Rhododendron (Ericaceae)

PREMISE Rhododendron L. (Ericaceae Juss.), is the most species-rich genus of woody flowering plants with > 1000 species. Despite the interest in the genus and numerous previous phylogenetic analysis, the infrageneric classification for the genus is still debated, partly due to its huge diversity, partly due to homoplasy in key characters and partly due to incongruence between phylogenetic markers. Here, we provide a broad coverage of representative species of all Rhododendron subgenera, sections, and most subsections to resolve its infrageneric phylogeny or highlight areas of incongruence, support previous analyses of diversification patterns and establish a relationship between genome size evolution and its diversification. METHODS We generated sequences of two plastid ( trn K and trn L-F) and two nuclear (ITS and rpb 2-i) markers for a total of 259 Rhododendron species, and used likelihood and Bayesian statistics to analyze the data. We analyzed the markers separately to discuss and understand incongruence among the data sets and among previous studies. RESULTS We found that the larger a subgenus, the more strongly it is supported as monophyletic. However, the smaller subgenera pose several problems, e.g., R . subgen. Azaleastrum consists of two sections inferred to be polyphyletic. The main shift to higher diversification in the genus occurred in the Himalayan/SE Asian clade of R . subgen. Hymenanthes . We found that polyploidy occurs in almost all subgenera but most polyploid species are within R . subgen. Rhododendron sections Rhododendron and Schistanthe . CONCLUSION Whereas previous reports stated that genome sizes of tropical plants are lower than those of colder and temperate regions in angiosperms in general, our study provides evidence for such a shift to small genome-tropical species within a genus. Taken together, we see the merit in the recognition of the five major clades at the sub generic level but given the amount of incongruence a large amount of species cannot be confidently assigned to one of these five clades. Further, genome-wide data will be necessary to assess whether these currently unassignable taxa are independent taxa, assignable to one of the five major clades or whether they are inter-subgeneric hybrids.