Phylogeny of the Liliales (Monocotyledons) with special emphasis on data partition congruence and RNA editing

A phylogenetic analysis of the monocot order Liliales was performed using sequence data from three mitochondrial (atp1, cob, nad5) and two plastid genes (rbcL, ndhF). The complete data matrix includes 46 terminals representing all 10 families currently included in Liliales. The two major partitions, mitochondrial and plastid data, were congruent, and parsimony analysis resulted in 50 equally parsimonious trees and a well resolved consensus tree confirming monophyly of all families. Mitochondrial genes are known to include RNA edited sites, and in some cases unprocessed genes are replaced by retro-processed gene copies, that is processed paralogs. To test the effects on phylogeny reconstruction of predicted edited sites and potentially unintentionally sampled processed paralogs, a number of analyses were performed using subsets of the complete data matrix. In general, predicted edited sites were more homoplasious than the other characters and increased incongruence among most data partitions. The predicted edited sites have a non-random phylogenetic signal in conflict with the signal of the non-edited sites. The potentially misleading signal was caused partially by the apparent presence of processed paralogs in Galanthus (Amaryllidaceae), part of the outgroup, but also by a deviating evolutionary pattern of predicted edited sites in Liliaceae compared with the remainder of the Liliales. Despite the problems that processed paralogs may cause, we argue that they should not a priori be excluded from phylogenetic analysis.