Epigenetics has emerged as an important research field for crop improvement under the on-going climatic changes. Heritable epigenetic changes can arise independently of DNA sequence alterations and have been associated with altered gene expression and transmitted phenotypic variation. By modulating plant development and physiological responses to environmental conditions, epigenetic diversity—naturally, genetically, chemically, or environmentally induced—can help optimise crop traits in an era challenged by global climate change. Beyond DNA sequence variation, the epigenetic modifications may contribute to breeding by providing useful markers and allowing the use of epigenome diversity to predict plant performance and increase final crop production. Given the difficulties in transferring the knowledge of the epigenetic mechanisms from model plants to crops, various strategies have emerged. Among those strategies are modelling frameworks dedicated to predicting epigenetically controlled-adaptive traits, the use of epigenetics for in vitro regeneration to accelerate crop breeding, and changes of specific epigenetic marks that modulate gene expression of traits of interest. The key challenge that agriculture faces in the 21st century is to increase crop production by speeding up the breeding of resilient crop species. Therefore, epigenetics provides fundamental molecular information with potential direct applications in crop enhancement, tolerance, and adaptation within the context of climate change.

Epialleles are meiotically heritable variations in expression states that are independent from changes in DNA sequence. Although they are common in plant genomes, their molecular origins are unknown. Here we show, using mutant and experimental populations, that epialleles in Arabidopsis thaliana that result from ectopic hypermethylation are due to feedback regulation of pathways that primarily function to maintain DNA methylation at heterochromatin. Perturbations to maintenance of heterochromatin methylation leads to feedback regulation of DNA methylation in genes. Using single base resolution methylomes from epigenetic recombinant inbred lines (epiRIL), we show that epiallelic variation is abundant in euchromatin, yet, associates with QTL primarily in heterochromatin regions. Mapping three-dimensional chromatin contacts shows that genes that are hotspots for ectopic hypermethylation have increases in contact frequencies with regions possessing H3K9me2. Altogether, these data show that feedback regulation of pathways that have evolved to maintain heterochromatin silencing leads to the origins of spontaneous hypermethylated epialleles.

Heterosis is the superior phenotypic performance of F1 hybrids relative to their parents. Although this phenomenon is extensively exploited commercially, its molecular causes remain elusive. A central challenge is to understand how specific features of parental (epi)genomes contribute to the widespread functional remodelling that occurs in hybrids. Using Arabidopsis, we show that differentially methylated regions (DMRs) in parental pericentromeres act as major re-organizers of hybrid methylomes and transcriptomes, even in the absence of genetic variation. We demonstrate that these parental DMRs facilitate methylation changes in the hybrids not only in cis, but also in trans at thousands of target regions throughout the genome. Many of these trans-induced changes facilitate the expression of nearby genes, and are significantly associated with phenotypic heterosis. Our study establishes the epigenetic status of parental pericentromeres as an important predictor of heterosis and elucidates its pleiotropic potential in the functional remodelling of hybrid genomes.

Whole-genome bisulfite sequencing (WGBS) is the standard method for profiling DNA methylation at single-nucleotide resolution. Many WGBS-based studies aim to identify biologically relevant loci that display differential methylation between genotypes, treatment groups, tissues, or developmental stages. Over the years, different tools have been developed to extract differentially methylated regions (DMRs) from whole-genome data. Often, such tools are built upon assumptions from mammalian data and do not consider the substantially more complex and variable nature of plant DNA methylation. Here, we present MethylScore, a pipeline to analyze WGBS data and to account for plant-specific DNA methylation properties. MethylScore processes data from genomic alignments to DMR output and is designed to be usable by novice and expert users alike. It uses an unsupervised machine learning approach to segment the genome by classification into states of high and low methylation, substantially reducing the number of necessary statistical tests while increasing the signal-to-noise ratio and the statistical power. We show how MethylScore can identify DMRs from hundreds of samples and how its data-driven approach can stratify associated samples without prior information. We identify DMRs in the A. thaliana 1001 Genomes dataset to unveil known and unknown genotype-epigenotype associations. MethylScore is an accessible pipeline for plant WGBS data, with unprecedented features for DMR calling in small- and large-scale datasets; it is built as a Nextflow pipeline and its source code is available at https://github.com/Computomics/MethylScore.

SUMMARYF1 hybrids derived from a cross between two inbred parental lines often display widespread changes in DNA methylation patterns relative to their parents. To which extent these changes drive non‐additive gene expression levels and phenotypic heterosis in F1 individuals is not fully resolved. Current mechanistic models propose that DNA methylation remodeling in hybrids is the result of epigenetic interactions between parental alleles via small interfering RNA (sRNA). These models have strong empirical support but are limited to genomic regions where the two parental lines differ in DNA methylation status. However, most remodeling events occur in parental regions with similar methylation patterns, and seem to be strongly conditioned by distally acting factors, even in isogenic hybrid systems. The molecular basis of these distal interactions is currently unknown, and will likely emerge as an active area of research in the future. Despite these gaps in our molecular understanding, parental DNA methylation states are statistically associated with heterosis, independent of genetic information, and may serve as biomarkers in crop breeding.