Abstract
<jats:p>Mitochondrial protein-coding genes are compact, functionally constrained, and generally under strong purifying selection, leading to an expectation of few amino acid-altering variants within populations. We characterized variation in the 13 protein-coding genes of 141 Fundulus heteroclitus from three proximate admixed populations each with two ancestral mitochondrial clades. Across 11,417 bp of coding sequence we identified 927 variants (8% of sites): 763 synonymous (82%) and 164 nonsynonymous (18%). Purifying selection varied widely among genes (dN/dS 0.003 in COX1 to 0.159 in ATP8), with Complexes I and V showing higher dN/dS than Complexes III and IV. Interestingly, when comparing the two mitochondrial clades, the number of synonymous and nonsynonymous variant and their ratio (N/S) is a function of the reference genome. This occurs because the long branch length between the two clades dominated by synonymous variants alters the accounting of variants within clades. This discrepancy between clades in N/S disappears when a random selection of variants serves as a reference. Comparison across taxa indicates that nucleotide diversity (π) in F. heteroclitus was nearly 20-fold higher than a single Drosophila population and comparable to or greater than a worldwide Caenorhabditis elegans collection, 1,176 sub-Saharan African humans, and 5,718 humans from admixed populations. This π reflects the admixture of two divergent mitochondrial clades. Yet the ratio of nonsynonymous to synonymous polymorphism (N/S, πN/πS, θN/θS) was lower than in these species, evidence of strong purifying selection. Heteroplasmy was common (53 of 141 individuals, 37.6%; mean 20 sites, 1.7 nonsynonymous); no heteroplasmic variant was unique, indicating segregation of existing polymorphisms rather than de novo mutation. Overall, these admixed populations harbor exceptionally high mitochondrial diversity while maintaining strong purifying selection.</jats:p>