Evolution of mitochondrial-encoded cytochrome oxidase subunits in endothermic fish: The importance of taxon-sampling in codon-based models

While endothermy is ubiquitous in birds and mammals, it is not exclusive to these most recently arisen vertebrate classes. The ability to warm specific organs and/or tissues above ambient temperature (regional endothermy) has evolved at least three times in phylogentically discrete fish lineages: lamnid sharks (Lamnidae), tunas (Scombridae) and billfishes (Istiophoridae and Xiphidae). Given the links between endothermy and metabolic rate, we looked for evidence of convergent molecular evolution in mtDNA-encoded cytochrome c oxidase (COX) subunits in each of these discrete lineages. We found no evidence that the endothermic phenotype in fishes is driven or accompanied by molecular convergence. Though we found little evidence for positively-selected sites in any of the lineages in any subunit, the conclusions were sensitive to the choice of maximum-likelihood model. Several sites identified by Naïve Empirical Bayes (NEB) were not found when Bayes Empirical Bayes (BEB) was employed. As well, conclusions were profoundly influenced by taxon-sampling. Several of the putative sites of positive selection in COX II were no longer apparent as we augmented taxon sampling. The lack of convergent molecular evolution in these remarkable taxa, combined with the profound influence of model choice and taxon sampling provide a cautionary note on the use of rates of non-synonymous to synonymous mutations (dN/dS) to explore questions of the evolution of physiological function.

Figure optionsDownload as PowerPoint slideHighlights
► Convergent molecular evolution as a mechanism for convergent evolution of endothermy in fish.
► Effect of taxon-sampling on codon-based evolutionary models for selection pressures.
► No evidence that convergence at molecular level underlies evolution of endothermy in fish.
► When sampled adequately, no positive selection on cytochrome oxidase in endothermic fish.
► Suggest codon-based models be used with increasing caution in large-scale phylogenies.