Genomic imprinting is a phenomenon that has evolved in certain sexually-reproducing clades, including mammals and flowering plants. It is characterized by parent-of-origin-dependent expression of certain loci due to differing chromatin signatures or `imprints' between maternally-inherited and paternally-inherited DNA. Previous work in the endosperm of plants such as maize and Arabidopsis thaliana showed that maternal DNA hypomethylation serves as the primary imprint for imprinted expression at certain genes. Subsequent genome-wide surveys of DNA methylation in Arabidopsis thaliana endosperm uncovered extensive localized DNA demethylation on maternally-inherited but not paternally-inherited chromosomes. Though this demethylation mediates imprinted expression of both maternally- and paternally-biased genes, it is apparently unnecessary for the extensive accumulation of maternally-biased small RNA molecules detected in Arabidopsis seeds. Whether imprinting function and genome-wide patterns of imprinting regulation are conserved across monocots and dicots remains to be tested.
Here, I show that extensive localized hypomethylation of rice endosperm DNA is likewise due to hypomethylation of maternally-inherited but not paternally-inherited chromosomes. Maternal hypomethylation preferentially occurs at regions of high DNA accessibility and is enriched within both imprinted genes and imprinted small RNA-producing loci. Maternally expressed imprinted genes are enriched for hypomethylation at putative promoter regions and transcriptional start and end sites, while paternally expressed genes are enriched for hypomethylation at promoters and gene bodies, mirroring recent results in A. thaliana.
I also show that, unlike in other rice tissues, small RNA populations in rice endosperm are dominated by a limited number of strong small RNA-producing loci. Whereas small RNAs in rice seedling tissues primarily originate from small Class II (cut-and-paste) transposable elements, those in endosperm are more uniformly derived, and include sequences from other transposon classes, as well as genic and intergenic regions. Imprinted 24-nt small RNAs are expressed from either parental genome and correlate with maternal DNA hypomethylation. Overlaps between imprinted small RNA loci and imprinted genes expressed suggest that small RNAs are associated with parent-of-origin-specific silencing.
In order to further investigate imprinting mechanisms and their targets in rice, I analyzed imprinting divergence among four cultivars that span the diversity within the rice species Oryza sativa. While the imprinting of 395 out of 413 genes is conserved among rice cultivars, I estimate that 4 to 11% of imprinted genes show imprinting divergence. For 16 out of 20 genes with diverged imprinting, DNA methylation epialleles were observed in key regulatory regions identified by our genome-scale enrichment analysis. These regions included the promoter and transcription start site for maternally expressed imprinted genes, and the promoter and gene body for paternally expressed imprinted genes. I did not observe an obvious association of imprinted small RNAs with imprinting divergence. In general, DNA methylation and small RNA profiles are conserved among rice cultivars.
To test whether genetic mutations contributed either to the formation of epialleles associated with imprinting divergence or to imprinting divergence that could not be correlated with an epiallele, I assessed genetic variation at seven of the 20 loci with diverged imprinting. At three genes, imprinting divergence was due to the insertion or precise excision of retrotransposons. DNA methylation epialleles in key regulatory regions at the other four loci did not appear to be associated with genetic variation. I also attempted to evaluate whether some apparent variation in parental bias might be due to a resetting of the imprint leading to biallelic expression in later endosperm development, as has been suggested by another group. However, I could not correlate changes in parental bias with changes in DNA methylation and small RNA production, suggesting either that developmental resetting occurs through another mechanism, or that the analysis was confounded by a mixed population of endosperm cells, or that the apparent variation is a result of technical artefacts rather than imprint resetting.
In summary, I identified the key regulatory regions of maternally- and paternally-expressed imprinted genes where maternal DNA hypomethylation is associated with imprinting, and developed a model for imprinted expression where maternal DNA hypomethylation serves as a primary imprint at both maternally- and paternally-expressed loci. Recent data supports the applicability of this model to other plant species such as maize and Arabidopsis. Divergence in imprinting in rice is associated with disruptions to either the genetic sequence or epigenetic state of regulatory regions, with retrotransposons playing a major role in sequence-related imprinting divergence over short evolutionary scales. Further work is required to elucidate the specific biogenesis and function of imprinted small RNAs and to verify potential cases of imprint resetting. The findings I present here make contributions to current understanding and models of mechanisms of epigenetic regulation, seed development, hybrid incompatibilities, and plant evolution. The production of high quality seeds is essential to many agricultural ventures, both as a mechanism for propagating plants and as means of generating valuable food commodities. In combination with other research, my work has the potential to improve breeding strategies for cereal crops that feed much of the world.