Rumbling Orchids: How To Assess Divergent Evolution Between Chloroplast Endosymbionts and the Nuclear Host

Phylogenetic relationships inferred from multilocus organellar and nuclear DNA data are often difficult to resolve because of evolutionary conflicts among gene trees. However, conflicting or outlier associations (i.e., linked pairs of operational terminal units in two phylogenies) among these data sets often provide valuable information on evolutionary processes such as chloroplast capture following hybridization, incomplete lineage sorting, and horizontal gene transfer. Statistical tools that to date have been used in cophylogenetic studies only also have the potential to test for the degree of topological congruence between organellar and nuclear data sets and reliably detect outlier associations. Two distance-based methods, namely ParaFit and Procrustean Approach to Cophylogeny (PACo), were used in conjunction to detect those outliers contributing to conflicting phylogenies independently derived from chloroplast and nuclear sequence data. We explored their efficiency of retrieving outlier associations, and the impact of input data (unit branch length and additive trees) between data sets, by using several simulation approaches. To test their performance using real data sets, we additionally inferred the phylogenetic relationships within Neotropical Catasetinae (Epidendroideae, Orchidaceae), which is a suitable group to investigate phylogenetic incongruence because of hybridization processes between some of its constituent species. A comparison between trees derived from chloroplast and nuclear sequence data reflected strong, well-supported incongruence within Catasetum, Cycnoches, and Mormodes. As a result, outliers among chloroplast and nuclear data sets, and in experimental simulations, were successfully detected by PACo when using patristic distance matrices obtained from phylograms, but not from unit branch length trees. The performance of ParaFit was overall inferior compared to PACo, using either phylograms or unit branch lengths as input data. Because workflows for applying cophylogenetic analyses are not standardized yet, we provide a pipeline for executing PACo and ParaFit as well as displaying outlier associations in plots and trees by using the software R. The pipeline renders a method to identify outliers with high reliability and to assess the combinability of the independently derived data sets by means of statistical analyses.