Dnmt1-independent CG methylation contributes to nucleosome positioning in diverse eukaryotes.

Dnmt1 epigenetically propagates symmetrical CG methylation in many eukaryotes. Their genomes are typically depleted of CG dinucleotides because of imperfect repair of deaminated methylcytosines. Here, we extensively survey diverse species lacking Dnmt1 and show that, surprisingly, symmetrical CG methylation is nonetheless frequently present and catalyzed by a different DNA methyltransferase family, Dnmt5. Numerous Dnmt5-containing organisms that diverged more than a billion years ago exhibit clustered methylation, specifically in nucleosome linkers. Clustered methylation occurs at unprecedented densities and directly disfavors nucleosomes, contributing to nucleosome positioning between clusters. Dense methylation is enabled by a regime of genomic sequence evolution that enriches CG dinucleotides and drives the highest CG frequencies known. Species with linker methylation have small, transcriptionally active nuclei that approach the physical limits of chromatin compaction. These features constitute a previously unappreciated genome architecture, in which dense methylation influences nucleosome positions, likely facilitating nuclear processes under extreme spatial constraints.

The Arabidopsis nucleosome remodeler DDM1 allows DNA methyltransferases to access H1-containing heterochromatin.

Nucleosome remodelers of the DDM1/Lsh family are required for DNA methylation of transposable elements, but the reason for this is unknown. How DDM1 interacts with other methylation pathways, such as small-RNA-directed DNA methylation (RdDM), which is thought to mediate plant asymmetric methylation through DRM enzymes, is also unclear. Here, we show that most asymmetric methylation is facilitated by DDM1 and mediated by the methyltransferase CMT2 separately from RdDM. We find that heterochromatic sequences preferentially require DDM1 for DNA methylation and that this preference depends on linker histone H1. RdDM is instead inhibited by heterochromatin and absolutely requires the nucleosome remodeler DRD1. Together, DDM1 and RdDM mediate nearly all transposon methylation and collaborate to repress transposition and regulate the methylation and expression of genes. Our results indicate that DDM1 provides DNA methyltransferases access to H1-containing heterochromatin to allow stable silencing of transposable elements in cooperation with the RdDM pathway.

DNA Methylation and Histone H1 Jointly Repress Transposable Elements and Aberrant Intragenic Transcripts.

DNA methylation and histone H1 mediate transcriptional silencing of genes and transposable elements, but how they interact is unclear. In plants and animals with mosaic genomic methylation, functionally mysterious methylation is also common within constitutively active housekeeping genes. Here, we show that H1 is enriched in methylated sequences, including genes, of Arabidopsis thaliana, yet this enrichment is independent of DNA methylation. Loss of H1 disperses heterochromatin, globally alters nucleosome organization, and activates H1-bound genes, but only weakly de-represses transposable elements. However, H1 loss strongly activates transposable elements hypomethylated through mutation of DNA methyltransferase MET1. Hypomethylation of genes also activates antisense transcription, which is modestly enhanced by H1 loss. Our results demonstrate that H1 and DNA methylation jointly maintain transcriptional homeostasis by silencing transposable elements and aberrant intragenic transcripts. Such functionality plausibly explains why DNA methylation, a well-known mutagen, has been maintained within coding sequences of crucial plant and animal genes.

Divergence among rice cultivars reveals roles for transposition and epimutation in ongoing evolution of genomic imprinting.

Parent-of-origin-dependent gene expression in mammals and flowering plants results from differing chromatin imprints (genomic imprinting) between maternally and paternally inherited alleles. Imprinted gene expression in the endosperm of seeds is associated with localized hypomethylation of maternally but not paternally inherited DNA, with certain small RNAs also displaying parent-of-origin-specific expression. To understand the evolution of imprinting mechanisms in Oryza sativa (rice), we analyzed imprinting divergence among four cultivars that span both japonica and indica subspecies: Nipponbare, Kitaake, 93-11, and IR64. Most imprinted genes are imprinted across cultivars and enriched for functions in chromatin and transcriptional regulation, development, and signaling. However, 4 to 11% of imprinted genes display divergent imprinting. Analyses of DNA methylation and small RNAs revealed that endosperm-specific 24-nt small RNA-producing loci show weak RNA-directed DNA methylation, frequently overlap genes, and are imprinted four times more often than genes. However, imprinting divergence most often correlated with local DNA methylation epimutations (9 of 17 assessable loci), which were largely stable within subspecies. Small insertion/deletion events and transposable element insertions accompanied 4 of the 9 locally epimutated loci and associated with imprinting divergence at another 4 of the remaining 8 loci. Correlating epigenetic and genetic variation occurred at key regulatory regions-the promoter and transcription start site of maternally biased genes, and the promoter and gene body of paternally biased genes. Our results reinforce models for the role of maternal-specific DNA hypomethylation in imprinting of both maternally and paternally biased genes, and highlight the role of transposition and epimutation in rice imprinting evolution.

Evolution and function of genomic imprinting in plants.

Genomic imprinting, an inherently epigenetic phenomenon defined by parent of origin-dependent gene expression, is observed in mammals and flowering plants. Genome-scale surveys of imprinted expression and the underlying differential epigenetic marks have led to the discovery of hundreds of imprinted plant genes and confirmed DNA and histone methylation as key regulators of plant imprinting. However, the biological roles of the vast majority of imprinted plant genes are unknown, and the evolutionary forces shaping plant imprinting remain rather opaque. Here, we review the mechanisms of plant genomic imprinting and discuss theories of imprinting evolution and biological significance in light of recent findings.

Mechanism for DNA transposons to generate introns on genomic scales.

The discovery of introns four decades ago was one of the most unexpected findings in molecular biology. Introns are sequences interrupting genes that must be removed as part of messenger RNA production. Genome sequencing projects have shown that most eukaryotic genes contain at least one intron, and frequently many. Comparison of these genomes reveals a history of long evolutionary periods during which few introns were gained, punctuated by episodes of rapid, extensive gain. However, although several detailed mechanisms for such episodic intron generation have been proposed, none has been empirically supported on a genomic scale. Here we show how short, non-autonomous DNA transposons independently generated hundreds to thousands of introns in the prasinophyte Micromonas pusilla and the pelagophyte Aureococcus anophagefferens. Each transposon carries one splice site. The other splice site is co-opted from the gene sequence that is duplicated upon transposon insertion, allowing perfect splicing out of the RNA. The distributions of sequences that can be co-opted are biased with respect to codons, and phasing of transposon-generated introns is similarly biased. These transposons insert between pre-existing nucleosomes, so that multiple nearby insertions generate nucleosome-sized intervening segments. Thus, transposon insertion and sequence co-option may explain the intron phase biases and prevalence of nucleosome-sized exons observed in eukaryotes. Overall, the two independent examples of proliferating elements illustrate a general DNA transposon mechanism that can plausibly account for episodes of rapid, extensive intron gain during eukaryotic evolution.

DNA demethylation by ROS1a in rice vegetative cells promotes methylation in sperm.

Epigenetic reprogramming is required for proper regulation of gene expression in eukaryotic organisms. In Arabidopsis, active DNA demethylation is crucial for seed viability, pollen function, and successful reproduction. The DEMETER (DME) DNA glycosylase initiates localized DNA demethylation in vegetative and central cells, so-called companion cells that are adjacent to sperm and egg gametes, respectively. In rice, the central cell genome displays local DNA hypomethylation, suggesting that active DNA demethylation also occurs in rice; however, the enzyme responsible for this process is unknown. One candidate is the rice REPRESSOR OF SILENCING1a (ROS1a) gene, which is related to DME and is essential for rice seed viability and pollen function. Here, we report genome-wide analyses of DNA methylation in wild-type and ros1a mutant sperm and vegetative cells. We find that the rice vegetative cell genome is locally hypomethylated compared with sperm by a process that requires ROS1a activity. We show that many ROS1a target sequences in the vegetative cell are hypomethylated in the rice central cell, suggesting that ROS1a also demethylates the central cell genome. Similar to Arabidopsis, we show that sperm non-CG methylation is indirectly promoted by DNA demethylation in the vegetative cell. These results reveal that DNA glycosylase-mediated DNA demethylation processes are conserved in Arabidopsis and rice, plant species that diverged 150 million years ago. Finally, although global non-CG methylation levels of sperm and egg differ, the maternal and paternal embryo genomes show similar non-CG methylation levels, suggesting that rice gamete genomes undergo dynamic DNA methylation reprogramming after cell fusion.

FACT complex is required for DNA demethylation at heterochromatin during reproduction in Arabidopsis.

The DEMETER (DME) DNA glycosylase catalyzes genome-wide DNA demethylation and is required for endosperm genomic imprinting and embryo viability. Targets of DME-mediated DNA demethylation reside in small, euchromatic, AT-rich transposons and at the boundaries of large transposons, but how DME interacts with these diverse chromatin states is unknown. The STRUCTURE SPECIFIC RECOGNITION PROTEIN 1 (SSRP1) subunit of the chromatin remodeler FACT (facilitates chromatin transactions), was previously shown to be involved in the DME-dependent regulation of genomic imprinting in Arabidopsis endosperm. Therefore, to investigate the interaction between DME and chromatin, we focused on the activity of the two FACT subunits, SSRP1 and SUPPRESSOR of TY16 (SPT16), during reproduction in Arabidopsis We found that FACT colocalizes with nuclear DME in vivo, and that DME has two classes of target sites, the first being euchromatic and accessible to DME, but the second, representing over half of DME targets, requiring the action of FACT for DME-mediated DNA demethylation genome-wide. Our results show that the FACT-dependent DME targets are GC-rich heterochromatin domains with high nucleosome occupancy enriched with H3K9me2 and H3K27me1. Further, we demonstrate that heterochromatin-associated linker histone H1 specifically mediates the requirement for FACT at a subset of DME-target loci. Overall, our results demonstrate that FACT is required for DME targeting by facilitating its access to heterochromatin.

Arabidopsis male sexual lineage exhibits more robust maintenance of CG methylation than somatic tissues.

Cytosine DNA methylation regulates the expression of eukaryotic genes and transposons. Methylation is copied by methyltransferases after DNA replication, which results in faithful transmission of methylation patterns during cell division and, at least in flowering plants, across generations. Transgenerational inheritance is mediated by a small group of cells that includes gametes and their progenitors. However, methylation is usually analyzed in somatic tissues that do not contribute to the next generation, and the mechanisms of transgenerational inheritance are inferred from such studies. To gain a better understanding of how DNA methylation is inherited, we analyzed purified Arabidopsis thaliana sperm and vegetative cells-the cell types that comprise pollen-with mutations in the DRM, CMT2, and CMT3 methyltransferases. We find that DNA methylation dependency on these enzymes is similar in sperm, vegetative cells, and somatic tissues, although DRM activity extends into heterochromatin in vegetative cells, likely reflecting transcription of heterochromatic transposons in this cell type. We also show that lack of histone H1, which elevates heterochromatic DNA methylation in somatic tissues, does not have this effect in pollen. Instead, levels of CG methylation in wild-type sperm and vegetative cells, as well as in wild-type microspores from which both pollen cell types originate, are substantially higher than in wild-type somatic tissues and similar to those of H1-depleted roots. Our results demonstrate that the mechanisms of methylation maintenance are similar between pollen and somatic cells, but the efficiency of CG methylation is higher in pollen, allowing methylation patterns to be accurately inherited across generations.

DNA demethylation is initiated in the central cells of Arabidopsis and rice.

Cytosine methylation is a DNA modification with important regulatory functions in eukaryotes. In flowering plants, sexual reproduction is accompanied by extensive DNA demethylation, which is required for proper gene expression in the endosperm, a nutritive extraembryonic seed tissue. Endosperm arises from a fusion of a sperm cell carried in the pollen and a female central cell. Endosperm DNA demethylation is observed specifically on the chromosomes inherited from the central cell in Arabidopsis thaliana, rice, and maize, and requires the DEMETER DNA demethylase in Arabidopsis DEMETER is expressed in the central cell before fertilization, suggesting that endosperm demethylation patterns are inherited from the central cell. Down-regulation of the MET1 DNA methyltransferase has also been proposed to contribute to central cell demethylation. However, with the exception of three maize genes, central cell DNA methylation has not been directly measured, leaving the origin and mechanism of endosperm demethylation uncertain. Here, we report genome-wide analysis of DNA methylation in the central cells of Arabidopsis and rice-species that diverged 150 million years ago-as well as in rice egg cells. We find that DNA demethylation in both species is initiated in central cells, which requires DEMETER in Arabidopsis However, we do not observe a global reduction of CG methylation that would be indicative of lowered MET1 activity; on the contrary, CG methylation efficiency is elevated in female gametes compared with nonsexual tissues. Our results demonstrate that locus-specific, active DNA demethylation in the central cell is the origin of maternal chromosome hypomethylation in the endosperm.

The AAA-ATPase molecular chaperone Cdc48/p97 disassembles sumoylated centromeres, decondenses heterochromatin, and activates ribosomal RNA genes.

Centromeres mediate chromosome segregation and are defined by the centromere-specific histone H3 variant (CenH3)/centromere protein A (CENP-A). Removal of CenH3 from centromeres is a general property of terminally differentiated cells, and the persistence of CenH3 increases the risk of diseases such as cancer. However, active mechanisms of centromere disassembly are unknown. Nondividing Arabidopsis pollen vegetative cells, which transport engulfed sperm by extended tip growth, undergo loss of CenH3; centromeric heterochromatin decondensation; and bulk activation of silent rRNA genes, accompanied by their translocation into the nucleolus. Here, we show that these processes are blocked by mutations in the evolutionarily conserved AAA-ATPase molecular chaperone, CDC48A, homologous to yeast Cdc48 and human p97 proteins, both of which are implicated in ubiquitin/small ubiquitin-like modifier (SUMO)-targeted protein degradation. We demonstrate that CDC48A physically associates with its heterodimeric cofactor UFD1-NPL4, known to bind ubiquitin and SUMO, as well as with SUMO1-modified CenH3 and mutations in NPL4 phenocopy cdc48a mutations. In WT vegetative cell nuclei, genetically unlinked ribosomal DNA (rDNA) loci are uniquely clustered together within the nucleolus and all major rRNA gene variants, including those rDNA variants silenced in leaves, are transcribed. In cdc48a mutant vegetative cell nuclei, however, these rDNA loci frequently colocalized with condensed centromeric heterochromatin at the external periphery of the nucleolus. Our results indicate that the CDC48A(NPL4) complex actively removes sumoylated CenH3 from centromeres and disrupts centromeric heterochromatin to release bulk rRNA genes into the nucleolus for ribosome production, which fuels single nucleus-driven pollen tube growth and is essential for plant reproduction.

Imprinted expression of genes and small RNA is associated with localized hypomethylation of the maternal genome in rice endosperm.

Arabidopsis thaliana endosperm, a transient tissue that nourishes the embryo, exhibits extensive localized DNA demethylation on maternally inherited chromosomes. Demethylation mediates parent-of-origin-specific (imprinted) gene expression but is apparently unnecessary for the extensive accumulation of maternally biased small RNA (sRNA) molecules detected in seeds. Endosperm DNA in the distantly related monocots rice and maize is likewise locally hypomethylated, but whether this hypomethylation is generally parent-of-origin specific is unknown. Imprinted expression of sRNA also remains uninvestigated in monocot seeds. Here, we report high-coverage sequencing of the Kitaake rice cultivar that enabled us to show that localized hypomethylation in rice endosperm occurs solely on the maternal genome, preferring regions of high DNA accessibility. Maternally expressed imprinted genes are enriched for hypomethylation at putative promoter regions and transcriptional termini and paternally expressed genes at promoters and gene bodies, mirroring our recent results in A. thaliana. However, unlike in A. thaliana, rice endosperm sRNA populations are dominated by specific strong sRNA-producing loci, and imprinted 24-nt sRNAs are expressed from both parental genomes and correlate with hypomethylation. Overlaps between imprinted sRNA loci and imprinted genes expressed from opposite alleles suggest that sRNAs may regulate genomic imprinting. Whereas sRNAs in seedling tissues primarily originate from small class II (cut-and-paste) transposable elements, those in endosperm are more uniformly derived, including sequences from other transposon classes, as well as genic and intergenic regions. Our data indicate that the endosperm exhibits a unique pattern of sRNA expression and suggest that localized hypomethylation of maternal endosperm DNA is conserved in flowering plants.

Genome-wide analysis of Arabidopsis thaliana DNA methylation uncovers an interdependence between methylation and transcription.

Cytosine methylation, a common form of DNA modification that antagonizes transcription, is found at transposons and repeats in vertebrates, plants and fungi. Here we have mapped DNA methylation in the entire Arabidopsis thaliana genome at high resolution. DNA methylation covers transposons and is present within a large fraction of A. thaliana genes. Methylation within genes is conspicuously biased away from gene ends, suggesting a dependence on RNA polymerase transit. Genic methylation is strongly influenced by transcription: moderately transcribed genes are most likely to be methylated, whereas genes at either extreme are least likely. In turn, transcription is influenced by methylation: short methylated genes are poorly expressed, and loss of methylation in the body of a gene leads to enhanced transcription. Our results indicate that genic transcription and DNA methylation are closely interwoven processes.

Deposition of histone variant H2A.Z within gene bodies regulates responsive genes.

The regulation of eukaryotic chromatin relies on interactions between many epigenetic factors, including histone modifications, DNA methylation, and the incorporation of histone variants. H2A.Z, one of the most conserved but enigmatic histone variants that is enriched at the transcriptional start sites of genes, has been implicated in a variety of chromosomal processes. Recently, we reported a genome-wide anticorrelation between H2A.Z and DNA methylation, an epigenetic hallmark of heterochromatin that has also been found in the bodies of active genes in plants and animals. Here, we investigate the basis of this anticorrelation using a novel h2a.z loss-of-function line in Arabidopsis thaliana. Through genome-wide bisulfite sequencing, we demonstrate that loss of H2A.Z in Arabidopsis has only a minor effect on the level or profile of DNA methylation in genes, and we propose that the global anticorrelation between DNA methylation and H2A.Z is primarily caused by the exclusion of H2A.Z from methylated DNA. RNA sequencing and genomic mapping of H2A.Z show that H2A.Z enrichment across gene bodies, rather than at the TSS, is correlated with lower transcription levels and higher measures of gene responsiveness. Loss of H2A.Z causes misregulation of many genes that are disproportionately associated with response to environmental and developmental stimuli. We propose that H2A.Z deposition in gene bodies promotes variability in levels and patterns of gene expression, and that a major function of genic DNA methylation is to exclude H2A.Z from constitutively expressed genes.

EMF1 and PRC2 cooperate to repress key regulators of Arabidopsis development.

EMBRYONIC FLOWER1 (EMF1) is a plant-specific gene crucial to Arabidopsis vegetative development. Loss of function mutants in the EMF1 gene mimic the phenotype caused by mutations in Polycomb Group protein (PcG) genes, which encode epigenetic repressors that regulate many aspects of eukaryotic development. In Arabidopsis, Polycomb Repressor Complex 2 (PRC2), made of PcG proteins, catalyzes trimethylation of lysine 27 on histone H3 (H3K27me3) and PRC1-like proteins catalyze H2AK119 ubiquitination. Despite functional similarity to PcG proteins, EMF1 lacks sequence homology with known PcG proteins; thus, its role in the PcG mechanism is unclear. To study the EMF1 functions and its mechanism of action, we performed genome-wide mapping of EMF1 binding and H3K27me3 modification sites in Arabidopsis seedlings. The EMF1 binding pattern is similar to that of H3K27me3 modification on the chromosomal and genic level. ChIPOTLe peak finding and clustering analyses both show that the highly trimethylated genes also have high enrichment levels of EMF1 binding, termed EMF1_K27 genes. EMF1 interacts with regulatory genes, which are silenced to allow vegetative growth, and with genes specifying cell fates during growth and differentiation. H3K27me3 marks not only these genes but also some genes that are involved in endosperm development and maternal effects. Transcriptome analysis, coupled with the H3K27me3 pattern, of EMF1_K27 genes in emf1 and PRC2 mutants showed that EMF1 represses gene activities via diverse mechanisms and plays a novel role in the PcG mechanism.

Histone H2A.Z and DNA methylation are mutually antagonistic chromatin marks.

Eukaryotic chromatin is separated into functional domains differentiated by post-translational histone modifications, histone variants and DNA methylation. Methylation is associated with repression of transcriptional initiation in plants and animals, and is frequently found in transposable elements. Proper methylation patterns are crucial for eukaryotic development, and aberrant methylation-induced silencing of tumour suppressor genes is a common feature of human cancer. In contrast to methylation, the histone variant H2A.Z is preferentially deposited by the Swr1 ATPase complex near 5' ends of genes where it promotes transcriptional competence. How DNA methylation and H2A.Z influence transcription remains largely unknown. Here we show that in the plant Arabidopsis thaliana regions of DNA methylation are quantitatively deficient in H2A.Z. Exclusion of H2A.Z is seen at sites of DNA methylation in the bodies of actively transcribed genes and in methylated transposons. Mutation of the MET1 DNA methyltransferase, which causes both losses and gains of DNA methylation, engenders opposite changes (gains and losses) in H2A.Z deposition, whereas mutation of the PIE1 subunit of the Swr1 complex that deposits H2A.Z leads to genome-wide hypermethylation. Our findings indicate that DNA methylation can influence chromatin structure and effect gene silencing by excluding H2A.Z, and that H2A.Z protects genes from DNA methylation.

Regulation of imprinted gene expression in Arabidopsis endosperm.

Imprinted genes are expressed primarily or exclusively from either the maternal or paternal allele, a phenomenon that occurs in flowering plants and mammals. Flowering plant imprinted gene expression has been described primarily in endosperm, a terminal nutritive tissue consumed by the embryo during seed development or after germination. Imprinted expression in Arabidopsis thaliana endosperm is orchestrated by differences in cytosine DNA methylation between the paternal and maternal genomes as well as by Polycomb group proteins. Currently, only 11 imprinted A. thaliana genes are known. Here, we use extensive sequencing of cDNA libraries to identify 9 paternally expressed and 34 maternally expressed imprinted genes in A. thaliana endosperm that are regulated by the DNA-demethylating glycosylase DEMETER, the DNA methyltransferase MET1, and/or the core Polycomb group protein FIE. These genes encode transcription factors, proteins involved in hormone signaling, components of the ubiquitin protein degradation pathway, regulators of histone and DNA methylation, and small RNA pathway proteins. We also identify maternally expressed genes that may be regulated by unknown mechanisms or deposited from maternal tissues. We did not detect any imprinted genes in the embryo. Our results show that imprinted gene expression is an extensive mechanistically complex phenomenon that likely affects multiple aspects of seed development.

Local DNA hypomethylation activates genes in rice endosperm.

Cytosine methylation silences transposable elements in plants, vertebrates, and fungi but also regulates gene expression. Plant methylation is catalyzed by three families of enzymes, each with a preferred sequence context: CG, CHG (H = A, C, or T), and CHH, with CHH methylation targeted by the RNAi pathway. Arabidopsis thaliana endosperm, a placenta-like tissue that nourishes the embryo, is globally hypomethylated in the CG context while retaining high non-CG methylation. Global methylation dynamics in seeds of cereal crops that provide the bulk of human nutrition remain unknown. Here, we show that rice endosperm DNA is hypomethylated in all sequence contexts. Non-CG methylation is reduced evenly across the genome, whereas CG hypomethylation is localized. CHH methylation of small transposable elements is increased in embryos, suggesting that endosperm demethylation enhances transposon silencing. Genes preferentially expressed in endosperm, including those coding for major storage proteins and starch synthesizing enzymes, are frequently hypomethylated in endosperm, indicating that DNA methylation is a crucial regulator of rice endosperm biogenesis. Our data show that genome-wide reshaping of seed DNA methylation is conserved among angiosperms and has a profound effect on gene expression in cereal crops.

Concepts, mechanisms and implications of long-term epigenetic inheritance.

Many modes and mechanisms of epigenetic inheritance have been elucidated in eukaryotes. Most of them are relatively short-term, generally not exceeding one or a few organismal generations. However, emerging evidence indicates that one mechanism, cytosine DNA methylation, can mediate epigenetic inheritance over much longer timescales, which are mostly or completely inaccessible in the laboratory. Here we discuss the evidence for, and mechanisms and implications of, such long-term epigenetic inheritance. We argue that compelling evidence supports the long-term epigenetic inheritance of gene body methylation, at least in the model angiosperm Arabidopsis thaliana, and that variation in such methylation can therefore serve as an epigenetic basis for phenotypic variation in natural populations.

Millennia-long epigenetic fluctuations generate intragenic DNA methylation variance in Arabidopsis populations.

Methylation of CG dinucleotides (mCGs), which regulates eukaryotic genome functions, is epigenetically propagated by Dnmt1/MET1 methyltransferases. How mCG is established and transmitted across generations despite imperfect enzyme fidelity is unclear. Whether mCG variation in natural populations is governed by genetic or epigenetic inheritance also remains mysterious. Here, we show that MET1 de novo activity, which is enhanced by existing proximate methylation, seeds and stabilizes mCG in Arabidopsis thaliana genes. MET1 activity is restricted by active demethylation and suppressed by histone variant H2A.Z, producing localized mCG patterns. Based on these observations, we develop a stochastic mathematical model that precisely recapitulates mCG inheritance dynamics and predicts intragenic mCG patterns and their population-scale variation given only CG site spacing. Our results demonstrate that intragenic mCG establishment, inheritance, and variance constitute a unified epigenetic process, revealing that intragenic mCG undergoes large, millennia-long epigenetic fluctuations and can therefore mediate evolution on this timescale.