Phylogeny of the higher Libelluloidea (Anisoptera: Odonata): an exploration of the most speciose superfamily of dragonflies.

Although libelluloid dragonflies are diverse, numerous, and commonly observed and studied, their phylogenetic history is uncertain. Over 150 years of taxonomic study of Libelluloidea Rambur, 1842, beginning with Hagen (1840), [Rambur, M.P., 1842. Neuropteres. Histoire naturelle des Insectes, Paris, pp. 534; Hagen, H., 1840. Synonymia Libellularum Europaearum. Dissertation inaugularis quam consensu et auctoritate gratiosi medicorum ordinis in academia albertina ad summos in medicina et chirurgia honores.] and Selys (1850), [de Selys Longchamps, E., 1850. Revue des Odonates ou Libellules d'Europe [avec la collaboration de H.A. Hagen]. Muquardt, Bruxelles; Leipzig, 1-408.], has failed to produce a consensus about family and subfamily relationships. The present study provides a well-substantiated phylogeny of the Libelluloidea generated from gene fragments of two independent genes, the 16S and 28S ribosomal RNA (rRNA), and using models that take into account non-independence of correlated rRNA sites. Ninety-three ingroup taxa and six outgroup taxa were amplified for the 28S fragment; 78 ingroup taxa and five outgroup taxa were amplified for the 16S fragment. Bayesian, likelihood and parsimony analyses of the combined data produce well-resolved phylogenetic hypotheses and several previously suggested monophyletic groups were supported by each analysis. Macromiinae, Corduliidae s. s., and Libellulidae are each monophyletic. The corduliid (s.l.) subfamilies Synthemistinae, Gomphomacromiinae, and Idionychinae form a monophyletic group, separate from the Corduliinae. Libellulidae comprises three previously accepted subfamilies (Urothemistinae, a very restricted Tetrathemistinae, and a modified Libellulinae) and five additional consistently recovered groups. None of the other previously proposed subfamilies are supported. Bayesian analyses run with an additional 71 sequences obtained from GenBank did not alter our conclusions. The evolution of adult and larval morphological characters is discussed here to suggest areas for future focus. This study shows the inherent problems in using poorly defined and sometimes inaccurately scored characters, basing groups on symplesiomorphies, and failure to recognize the widespread effects of character correlation and convergence, especially in aspects of wing venation.