Natural methylation epialleles correlate with gene expression in maize

DNA methylation (5-methylcytosine) represses transposon activity and contributes to inaccessible chromatin structure of repetitive DNA in plants. It is depleted from cis regulatory elements in and near genes but is present in some gene bodies, including exons. Methylation in exons solely in the CG context is called gene body methylation (gbM). Methylation in exons in both CG and non-CG contexts is called TE-like methylation (teM). Assigning functions to both forms of methylation in genes has proven to be challenging. Toward that end, we utilized recent genome assemblies, gene annotations, transcription data, and methylome data to quantify common patterns of gene methylation and their relations to gene expression in maize. To compare between genomes, we analyzed each data source relative to its own genome assembly rather than the easier but less accurate method of using one assembly as reference for all. We found that gbM genes exist in a continuum of CG methylation levels without a clear demarcation between unmethylated genes and gbM genes. Analysis of expression levels across diverse maize stocks and tissues revealed a weak but highly significant positive correlation between gbM and gene expression except in endosperm. gbM epialleles were associated with an approximately 3% increase in steady-state expression level relative to unmethylated epialleles. In contrast to gbM genes, which were conserved and were broadly expressed across tissues, we found that teM genes, which make up about 12% of genes, are mainly silent, are limited to specific maize stocks, and exhibit evidence of annotation errors. We used these data to flag all teM genes in the 26 NAM founder genome assemblies. While some teM genes are likely functional, these data suggest that the majority are not, and their inclusion can confound interpretation of whole-genome studies.