RPB2 gene phylogeny in flowering plants, with particular emphasis on asterids.

Two, apparently functional, paralogues of the RPB2 gene, which encodes the second largest subunit of RNA polymerase II, are shown to be present in two major groups of asterid plants. Although all other land plants surveyed so far have been found to have only one of these two copies, the RPB2 gene phylogeny inferred from the 3' half of the gene for 35 angiosperm taxa and six other land plants indicates that the duplication of the RPB2 gene occurred earlier than the time for origin of the asterid group, probably near the origin of "core eudicots." The d copy is present in all plants which are unambiguously assigned to the core eudicots, whereas the I copy is retained only in the lamiid clade, Ericales, and Escallonia, all belonging to the asterid group of plants. Both parsimony and likelihood analyses of sequences from the 3' half of the gene give strong bootstrap support for these conclusions. There is no support for monophyly of the taxa having both copies. Thus, numerous losses of one of the copies must be inferred. Structurally, both paralogues appear functional, and transcription is demonstrated for both copies. In the lamiid group, the d copy has lost introns 18-23. The well supported phylogenetic relationships implied by the RPB2 gene phylogeny are largely congruent with well supported phylogenetic hypotheses based on other sequence data. However, Ilex, usually assigned to the campanuliid clade, is instead supported as being a member of the lamiid clade, both from sequence data and the presence of an I copy as well as the loss of introns 18-23 in the d copy. Escallonia, supported as a member of the campanuliid clade both by RPB2-d-sequences and previously published DNA sequence data, has all the introns 18-23 in its d copy, as do all other members studied from the campanuliid group. We used the Markov Chain Monte Carlo (MCMC) approach of the MrBayes program to implement Maximum Likelihood bootstrapping. Under the same model of evolution, bootstrapping frequencies are significantly lower than the Bayesian posterior probabilities inferred from the MCMC chain.