Cytosine DNA methylation is an ancient form of transcriptional control that is conserved across all kingdoms of eukaryotes. DNA methylation plays a major role in silencing of selfish genetic elements, such as transposons. Additionally, in some instances, DNA methylation is required for genomic imprinting and regulation of endogenous genes. In the model plant Arabidopsis thaliana, at least three pathways, each with its own methyltransferase, maintain DNA methylation. MET1 targets CG dinucleotide sequences; due to the inherent symmetry across DNA strands, MET1 is able to recognize hemimethylated sites after DNA replication, thus maintains faithful methylation patterns. CMT3 typically has preference for CHG sites (where H is A, T, or C), and is targeted to chromatin via its chromodomain, which has specificity for histone 3 lysine 9 dimethylation--another epigenetic mark associated with heterochromatin. Finally, DRM2 maintains CHH, or asymmetric, methylation through targeting by a dual siRNA/long non- coding RNA pathway termed RNA-directed DNA methylation (RdDM). It should be noted that CMT3 and DRM2 are both capable of methylating non-CG sites. While all three methyltransferases maintain existing DNA methylation patterns, only the RdDM pathway
establishes the mark in all sequence contexts, in a process known as de novo methylation.
In this dissertation I will describe both forward and reverse genetic techniques I have used to uncover factors required for de novo methylation. For both techniques, I made use of the FWA transgene. In wild-type plants, the RdDM pathway is able to target, methylate, and silence the transgene at the repeats in its 5' UTR. However, in RdDM mutants, the transgene remains unmethylated and expresses, leading to a late-flowering phenotype. From a mutagenesis screen, I discovered novel mutations in 11 genes required for DNA methylation establishment. I will describe the methodologies of cloning and characterizing those mutants. Additionally, from the same study, I was able to show a de novo methylation phenotype from previously described RdDM mutant alleles.
In a reverse genetic screen, utilizing a collection of insertional mutations in known or putative RNA binding proteins, I helped characterize a known RNA splicing factor, the first such RNA processing protein shown to be required for RdDM. I also showed that two partially redundant paralogs of the IDN2 RNA-binding protein are required for RdDM and de novo methylation. Further biochemical analysis revealed that the paralogs form a complex with IDN2. In collaboration with a structural biology group, we solved the structure of the RNA binding motif of IDN2.
Finally, I will discuss the data explicating the relationship between histone 3 lysine 4 (H3K4) demethylases and the RdDM pathway. We made the surprising discovery that active demethylation is required for RdDM maintenance, but not establishment. In sum, the work in this dissertation contributes to our knowledge of the components and mechanism of RdDM, and how the RNA polymerase-dependent pathway is affected by perturbations in local chromatin.