Mouse Epigenome Atlas Decodes Two DNA Modifications Across 29 Tissues and Ages

DNA cytosine modifications are central to gene regulation and are increasingly used as disease biomarkers. The two most abundant modifications—5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC)—have distinct biological roles but are conflated by the widely used bisulfite sequencing method, which cannot distinguish between them. This study addresses that gap by generating the first large-scale, base-resolution ternary-code atlas (unmodified C, 5mC, and 5hmC) across diverse mouse tissues.

The team applied the bisulfite-assisted APOBEC-Coupled Epigenetic (bACE) conversion strategy to the Infinium Mouse Methylation BeadChip, profiling 265 samples from 29 tissue types collected from 32 C57BL/6J mice spanning 8–76 weeks of age and both sexes. The bACE approach works by splitting bisulfite-converted DNA: one aliquot profiles 5mC+5hmC combined (5modC), and the other undergoes additional APOBEC deamination that selectively deaminates 5mC while leaving bisulfite-converted 5hmC intact, enabling direct 5hmC quantification. A reference interpolation curve corrected for array background, improving concordance with orthogonal sequencing methods (ACE-seq, TAB-seq).

Key findings reveal a complex grammar of 5hmC distribution. Globally, 5hmC ranged from 0.5% in immune tissues (blood, thymus) to 14.6% in neuron-rich brain regions, inversely correlated with 5mC levels. Tissues with similar total modification levels can have dramatically different 5mC-to-5hmC ratios—for instance, subcortical brain and blood share similar total 5modC but diverge markedly in 5hmC content. The distribution of 5hmC is jointly shaped by cell mitotic activity (rapidly dividing cells carry less 5hmC, consistent with its dilution without maintenance methylation), chromatin states, and cis-interactions between neighboring CpG sites. Notably, 5hmC at non-CpG sites (CpH context) was elevated specifically in neuronal tissues at CpA sites.

For tissue identity classification, the atlas demonstrates that 5hmC alone—independent of 5mC—enables high-precision machine learning discrimination of tissue types. Moreover, 5hmC significantly complements 5mC-based biomarkers in distinguishing both brain and non-brain cell identities, suggesting that ternary-code profiling provides a richer molecular fingerprint than 5modC alone. Age-related epigenetic changes in both modifications correlate with tissue-specific gene expression variation, implying that the two marks co-regulate transcriptional programs during aging in a tissue-dependent fashion.

The atlas provides a foundational resource for exploring epigenetic dynamics in development, aging, and disease. Because the mouse C57BL/6J model has a well-characterized genetic background and parallels many human epigenetic phenomena, findings are likely translatable to human biology. The work also expands the potential scope of liquid biopsy and epigenetic clock applications by demonstrating that 5hmC carries independent, actionable biological information.