The population frequency of human mitochondrial DNA variants is highly dependent upon mutational bias
Genome-wide association studies (GWASs) typically seek common genetic variants that can influence disease likelihood. However, these analyses often fail to convincingly link specific genes and their variants with highly penetrant phenotypic effects. To solve the 'missing heritability problem' that characterizes GWASs, researchers have turned to rare variants revealed by next-generation sequencing when seeking genomic changes that may be pathogenic, as a reduction in variant frequency is an expected outcome of selection. While triage of rare variants has led to some success in illuminating genes linked to heritable disease, the interpretation and utilization of rare genomic changes remains very challenging. Human mitochondrial DNA (mtDNA) encodes proteins and RNAs required for the essential process of oxidative phosphorylation, and a number of metabolic diseases are linked to mitochondrial mutations. Recently, the mtDNAs of nearly 200,000 individuals were sequenced in order to produce the HelixMT database (HelixMTdb), a large catalog of human mtDNA variation. Here, we were surprised to find that many synonymous nucleotide substitutions were never detected within this quite substantial survey of human mtDNA. Subsequent study of more than 1000 mammalian mtDNAs suggested that selection on synonymous sites within mitochondrial protein-coding genes is minimal and unlikely to explain the rarity of most synonymous changes among humans. Rather, the mutational propensities of mtDNA are more likely to determine variant frequency. Our findings have general implications for the interpretation of variant frequencies when studying heritable disease.