Summary Histone variants play crucial roles in gene expression, genome integrity and chromosome segregation. However, to what extent histone variants control chromatin architecture remains largely unknown. We report genome-wide profiles of all four types of H2A variants in Arabidopsis and identify that the previously uncharacterized histone variant H2A.W specifically associates with heterochromatin. Genetic analyses show that H2A.W acts in synergy with the heterochromatic marks H3K9me2 and DNA methylation to maintain genome integrity. In vitro, H2A.W enhances chromatin condensation through a higher propensity to promote fiber-to-fiber interactions via its conserved C-terminal motif. In vivo, elimination of H2A.W causes decondensation of heterochromatin and conversely, ectopic expression of H2A.W promotes heterochromatin condensation. These results demonstrate that H2A.W plays critical roles in heterochromatin by promoting higher order chromatin condensation. Since motifs similar to the H2A.W C-terminal motif are present in other histone variants in other organisms, our findings impact our understanding of heterochromatin condensation in eukaryotes.

Histone variants are non-allelic protein isoforms that play key roles in diversifying chromatin structure. The known number of such variants has greatly increased in recent years, but the lack of naming conventions for them has led to a variety of naming styles, multiple synonyms and misleading homographs that obscure variant relationships and complicate database searches. We propose here a unified nomenclature for variants of all five classes of histones that uses consistent but flexible naming conventions to produce names that are informative and readily searchable. The nomenclature builds on historical usage and incorporates phylogenetic relationships, which are strong predictors of structure and function. A key feature is the consistent use of punctuation to represent phylogenetic divergence, making explicit the relationships among variant subtypes that have previously been implicit or unclear. We recommend that by default new histone variants be named with organism-specific paralog-number suffixes that lack phylogenetic implication, while letter suffixes be reserved for structurally distinct clades of variants. For clarity and searchability, we encourage the use of descriptors that are separate from the phylogeny-based variant name to indicate developmental and other properties of variants that may be independent of structure.

DNA methylation maintains genome stability and regulates gene expression [1]. In mammals, DNA methylation is reprogrammed in the germline from one generation to the next [2]. In plants, it was considered that patterns of DNA methylation are stably maintained through sexual reproduction [3-6]. However, a recent report showed discrete variations of DNA methylation profiles from mother to daughter plants [7]. The mechanisms that explain these variations have remained unknown. Here, we report that maintenance DNA methyltransferases are barely expressed during Arabidopsis female gametogenesis. In contrast, after fertilization both maintenance and de novo DNA methyltransferases are expressed strongly in the embryo. Embryogenesis is marked by increased de novo DNA methylation, reaching levels that are further maintained in the adult plant. The accumulation of these epigenetic marks after fertilization silences a methylation-sensitive fluorescent reporter. De novo DNA methylation in the embryo provides a mechanism that could account for the gradual remethylation of experimentally demethylated genomes [8, 9]. In conclusion, we uncover that DNA methylation activity fluctuates during sexual reproduction. This cycle likely explains variations of genome-wide patterns of DNA methylation across generations in Arabidopsis [7, 10] and enables a limited degree of reprogramming of the epigenome.