Eukaryotic DNA cytosine methylation can be used to transcriptionally silence repetitive sequences, including transposons and retroviruses. This silencing is stable between cell generations as cytosine methylation is maintained epigenetically through DNA replication. The Arabidopsis thaliana Dnmt3 cytosine methyltransferase ortholog DOMAINS REARRANGED METHYLTRANSFERASE2 (DRM2) is required for establishment of small interfering RNA (siRNA) directed DNA methylation. In mammals PIWI proteins and piRNA act in a convergently evolved RNA–directed DNA methylation system that is required to repress transposon expression in the germ line. De novo methylation may also be independent of RNA interference and small RNAs, as in Neurospora crassa. Here we identify a clade of catalytically mutated DRM2 paralogs in flowering plant genomes, which in A.thaliana we term DOMAINS REARRANGED METHYLTRANSFERASE3 (DRM3). Despite being catalytically mutated, DRM3 is required for normal maintenance of non-CG DNA methylation, establishment of RNA–directed DNA methylation triggered by repeat sequences and accumulation of repeat-associated small RNAs. Although the mammalian catalytically inactive Dnmt3L paralogs act in an analogous manner, phylogenetic analysis indicates that the DRM and Dnmt3 protein families diverged independently in plants and animals. We also show by site-directed mutagenesis that both the DRM2 N-terminal UBA domains and C-terminal methyltransferase domain are required for normal RNA–directed DNA methylation, supporting an essential targeting function for the UBA domains. These results suggest that plant and mammalian RNA–directed DNA methylation systems consist of a combination of ancestral and convergent features.
Nuclear DNA quantity varies widely between species and is poorly correlated with gene number. Variation in genome size can be explained by differing amounts of repetitive DNA. Repetitive DNA may be mobile, meaning it can increase its copy number within genomes. To prevent this, plants and animals suppress expression of repeats, often by marking the repeated sequence with a methyl group on cytosine bases. DRM2 is an enzyme capable of establishing this methylation, which can be guided to target sequences by short complementary RNA guides. As repeated sequences are prone to generate short RNAs they are efficiently recognized and silenced. We show that DRM2 requires a related inactive DRM3 protein to normally silence repeated sequences. A similar situation exists in mammals, where active and inactive DNA methyltransferases act together to silence repeats. We also demonstrate that non-catalytic regions of the DRM2 enzyme are functionally important, which we speculate to be involved in targeting the enzyme to the genome. Although plant and mammal RNA–directed DNA methylation systems share key similarities, there are important mechanistic differences, meaning that they are likely to have arisen convergently.