Phylogenomic evidence for multiple losses of flight in ratite birds

Abstract

Ratites (ostriches, emus, rheas, cassowaries, and kiwis) are large, flightless birds that have long fascinated biologists. Their current distribution on isolated southern land masses is believed to reflect the breakup of the paleocontinent of Gondwana. The prevailing view is that ratites are monophyletic, with the flighted tinamous as their sister group, suggesting a single loss of flight in the common ancestry of ratites. However, phylogenetic analyses of 20 unlinked nuclear genes reveal a genome-wide signal that unequivocally places tinamous within ratites, making ratites polyphyletic and suggesting multiple losses of flight. Phenomena that can mislead phylogenetic analyses, including long branch attraction, base compositional bias, discordance between gene trees and species trees, and sequence alignment errors, have been eliminated as explanations for this result. The most plausible hypothesis requires at least three losses of flight and explains the many morphological and behavioral similarities among ratites by parallel or convergent evolution. Finally, this phylogeny demands fundamental reconsideration of proposals that relate ratite evolution to continental drift.

Fig. 1.

Phylogenetic analyses of a 20-gene, 24-kb nuclear DNA dataset strongly supporting ratite polyphyly. All analyses used Anas, Gallus, Buteo, and Ciconia as outgroups. Branches for which all support measures were 100% or 1.0 are indicated with an asterisk; support for ratite polyphyly is highlighted. Topology obtained by using both partitioned (by locus) and unpartitioned ML and Bayesian analyses. Branch lengths reflect the unpartitioned ML analysis. Support measures are partitioned RAxML bootstrap (Upper Left), unpartitioned ML bootstrap (Upper Right), unpartitioned Bayesian posterior probability (Lower Left), and partitioned Bayesian posterior probability (Lower Right).

Fig. 2.

Phylogenetic analyses including crocodilian outgroups and two passerine birds (Corvus and Smithornis) strongly support the conventional position of the avian root and ratite polyphyly. Analyses were conducted by using all sequences that could be aligned between crocodilians and birds (4,668 bp). Support measures are unpartitioned ML bootstrap (Upper Left), MP bootstrap (Upper Right), unpartitioned Bayesian posterior probability (Lower Left), and partitioned Bayesian posterior probability (Lower Right). Branch lengths shown reflect the unpartitioned ML analysis. Branches for which all support measures were 100% or 1.0 are indicated with an asterisk; the branch with no support values had <50% bootstrap support and <0.5 Bayesian posterior probability in all analyses. MP and ML analyses conducted after Y coding produced similar results (not shown).

Fig. 3.

An 8-bp deletion in ALDOB supports ratite polyphyly. (A) Alignment of the region around the informative deletion in ALDOB (positions 3213–3220). The ostrich shares its character state (+8 bp) with neognaths, whereas tinamous share the character state of all other ratites (−8 bp). (B) The distribution of character states can be mapped as a single deletion on the optimal topology found in this study. (C) The distribution requires at least two steps on the traditional topology (one possible reconstruction shown).