Human non-CpG methylation patterns display both tissue-specific and inter-individual differences suggestive of underlying function

DNA methylation (DNAm) in mammals is mostly examined within the context of CpG dinucleotides. Non-CpG DNAm is also widespread across the human genome, but the functional relevance, tissue-specific disposition, and inter-individual variability has not been widely studied. Our aim was to examine non-CpG DNAm in the wider methylome across multiple tissues from the same individuals to better understand non-CpG DNAm distribution within different tissues and individuals and in relation to known genomic regulatory features. DNA methylation in umbilical cord and cord blood at birth, and peripheral venous blood at age 12–13 y from 20 individuals from the Southampton Women’s Survey cohort was assessed by Agilent SureSelect methyl-seq. Hierarchical cluster analysis (HCA) was performed on CpG and non-CpG sites and stratified by specific cytosine environment. Analysis of tissue and inter-individual variation was then conducted in a second dataset of 12 samples: eight muscle tissues, and four aliquots of cord blood pooled from two individuals. HCA using methylated non-CpG sites showed different clustering patterns specific to the three base-pair triplicate (CNN) sequence. Analysis of CAC sites with non-zero methylation showed that samples clustered first by tissue type, then by individual (as observed for CpG methylation), while analysis using non-zero methylation at CAT sites showed samples grouped predominantly by individual. These clustering patterns were validated in an independent dataset using cord blood and muscle tissue. This research suggests that CAC methylation can have tissue-specific patterns, and that individual effects, either genetic or unmeasured environmental factors, can influence CAT methylation.