Apparent methylation of mitochondrial DNA in WGBS is dominated by technical artifacts

DNA methylation is a critical epigenetic modification that occurs on cytosine residues and plays essential roles in regulating gene expression, cellular differentiation, and genomic imprinting. However, the existence of cytosine methylation in mitochondrial DNA (mtDNA) remains highly controversial. Whole-genome bisulfite sequencing (WGBS) has been widely used to assess mtDNA methylation, but signals detected in mitochondrial DNA are frequently influenced by technical artifacts.

We systematically evaluated technical factors that influence mtDNA methylation measurements using zebrafish WGBS datasets derived from a wide range of tissues and experimental conditions, complemented by analyses of human and mouse datasets. mtDNA methylation measurements varied widely among samples within CpG, CHG, and CHH contexts, and mtDNA CpG methylation estimates above 2% were often accompanied by comparable CHG and CHH methylation estimates, even when lambda DNA CpG methylation remained below 2%. Notably, samples lacking documented DNA fragmentation tended to show substantially elevated mtDNA methylation compared with those subjected to fragmentation. Although strand-specific sequencing biases were common, they were not a sufficient cause of elevated false-positive methylation signals. The degree of strand coverage imbalance did not predict high methylation in datasets with documented fragmentation, and high methylation in undocumented samples was preferentially observed where strand coverage was balanced rather than imbalanced. Kit annotations were associated with dataset-level variation in mtDNA methylation, particularly for library preparation kits, but elevated signals were not confined to any single bisulfite conversion or library preparation kit. Cross-species analyses showed that human and mouse datasets displayed comparable patterns of elevated mtDNA methylation.

Our findings indicate that high mtDNA methylation signals detected in WGBS data are dominated by technical artifacts, particularly DNA fragmentation status and bisulfite conversion efficiency. These results highlight the need for careful interpretation of mitochondrial methylation signals and provide a framework for distinguishing technical noise from potential biological signals in WGBS-based analyses.