Mutational biases and selection in mitochondrial genomes: insights from a comparative analysis of natural and laboratory populations of Caenorhabditis elegans.

Spontaneous mutations display biases in their relative frequencies with important consequences for genome structure and composition. While laboratory studies have provided important insights into the spontaneous mutation spectrum, laboratory environments for optimal growth may engender biases that are not representative of natural populations. We analyzed the mitochondrial genomes of 1,524 Caenorhabditis elegans natural isolates comprising 531 unique isotypes to investigate mtDNA polymorphism in the wild. Ancestral reconstruction was used to polarize 2,464 variants (88 indels, 2,376 SNPs) and the results were compared with mutations identified in experimental lines under relaxed selection. MtDNA variant distribution in natural isolates is strongly dependent on site-degeneracy in a manner consistent with purifying selection. There is significant variation in the proportion of synonymous and nonsynonymous polymorphism between genes. Specifically, electron transport chain complex I genes are enriched for nonsynonymous polymorphism. The probability of synonymous mutation is higher at sites with neighboring G/C nucleotides and the per gene synonymous polymorphism is negatively correlated with A+T-content at the first and second codon positions. Furthermore, the 5' and 3' ends of genes have both higher A+T-content and less synonymous polymorphism than central regions. There is evidence of natural selection for preferred codons. We identify the first cases of large heteroplasmic mtDNA structural variants in C. elegans natural isolates, comprising deletions and duplications. Although some patterns of mtDNA mutational bias are similar between laboratory and natural populations, there exist significant differences. In particular, G/C → T/A transversions typically associated with oxidative damage and 8-oxoguanine are strikingly rarer at four-fold degenerate sites in natural populations relative to laboratory populations suggesting that the latter are more prone to mtDNA oxidative damage. Hence, mutational spectra observed in laboratory strains may differ in important aspects from those of natural populations.