Epigenetic Natural Variation in Arabidopsis thaliana

Abstract
Cytosine methylation of repetitive sequences is widespread in plant genomes, occurring in both symmetric (CpG and CpNpG) as well as asymmetric sequence contexts. We used the methylation-dependent restriction enzyme McrBC to profile methylated DNA using tiling microarrays of Arabidopsis Chromosome 4 in two distinct ecotypes, Columbia and Landsberg erecta. We also used comparative genome hybridization to profile copy number polymorphisms. Repeated sequences and transposable elements (TEs), especially long terminal repeat retrotransposons, are densely methylated, but one third of genes also have low but detectable methylation in their transcribed regions. While TEs are almost always methylated, genic methylation is highly polymorphic, with half of all methylated genes being methylated in only one of the two ecotypes. A survey of loci in 96 Arabidopsis accessions revealed a similar degree of methylation polymorphism. Within-gene methylation is heritable, but is lost at a high frequency in segregating F2 families. Promoter methylation is rare, and gene expression is not generally affected by differences in DNA methylation. Small interfering RNA are preferentially associated with methylated TEs, but not with methylated genes, indicating that most genic methylation is not guided by small interfering RNA. This may account for the instability of gene methylation, if occasional failure of maintenance methylation cannot be restored by other means. In plants and animals, many DNA sequences are modified by the addition of methyl groups, but the principles governing methylation patterns are not well understood. In Arabidopsis, we show that repetitive sequences, derived from mobile (transposable) elements, are densely methylated throughout their length, while about one third of all protein-coding genes are internally methylated. Methylated transposons are silent, homologous to small interfering RNA, and coated with histone H3 dimethylated on lysine-9. In contrast, methylated coding-sequence genes are highly expressed, do not have corresponding small RNAs, and are coated with histone H3 dimethylated on lysine-4. Comparing two different ecotypes of Arabidopsis, we find that transposons are twice as likely as genes to have suffered insertion and deletion, although gene deletion is surprisingly prevalent. While the pattern of transposon methylation is conserved between ecotypes, protein-coding gene methylation is polymorphic so that only half of all gene methylation on any one chromosome is shared between natural accessions collected from around the world.