Genome-wide high resolution parental-specific DNA and histone methylation maps uncover patterns of imprinting regulation in maize

Abstract

Genetic imprinting is a specific epigenetic phenomenon in which a subset of genes is expressed depending on their parent-of-origin. Two types of chromatin modifications, DNA methylation and histone modification, are generally believed to be involved in the regulation of imprinting. However, the genome-wide correlation between allele-specific chromatin modifications and imprinted gene expression in maize remains elusive. Here we report genome-wide high resolution allele-specific maps of DNA methylation and histone H3 lysine 27 trimethylation (H3K27me3) in maize endosperm. For DNA methylation, thousands of parent-of-origin dependent differentially methylated regions (pDMRs) were identified. All pDMRs were uniformly paternally hypermethylated and maternally hypomethylated. We also identified 1131 allele-specific H3K27me3 peaks that are preferentially present in the maternal alleles. Maternally expressed imprinted genes (MEGs) and paternally expressed imprinted genes (PEGs) had different patterns of allele-specific DNA methylation and H3K27me3. Allele-specific expression of MEGs was not directly related to allele-specific H3K27me3, and only a subset of MEGs was associated with maternal-specific DNA demethylation, which was primarily located in the upstream and 5' portion of gene body regions. In contrast, allele-specific expression of a majority of PEGs was related to maternal-specific H3K27me3, with a subgroup of PEGs also associated with maternal-specific DNA demethylation. Both pDMRs and maternal H3K27me3 peaks associated with PEGs are enriched in gene body regions. Our results indicate highly complex patterns of regulation on genetic imprinting in maize endosperm.

Figure 1.

Differential DNA methylation among tissues and between two parental genomes for maternally and paternally expressed imprinted genes. (A–C) Average DNA methylation levels of MEGs, PEGs, and nonimprinted genes (Non-imp) for shoot, embryo, and endosperm in CG context throughout the gene body and its 2-kb up- and downstream regions. (D–F) Comparison of average DNA methylation levels between two parental genomes of MEGs, PEGs, and nonimprinted genes (Non-imp) in CG context throughout the gene body and its 2-kb up- and downstream regions. (G–I) Average DNA methylation levels of MEGs, PEGs, and nonimprinted genes (Non-imp) for shoot, embryo, and endosperm in CHG context throughout the gene body and its 2-kb up- and downstream regions. (J–L) Comparison of average DNA methylation levels between two parental genomes of MEGs, PEGs, and nonimprinted genes (Non-imp) in CHG context throughout the gene body and its 2-kb up- and downstream regions. (A–L) Gene body regions were separated into 60 bins, and extended 2-kb up- and downstream regions were separated into 20 bins. The average methylation levels were calculated with the same method as in Supplemental Figure S7.

Figure 2.

Analysis of maize H3K27me3 peaks. (A,B) Genomic distribution of 12,125 H3K27me3 peaks (A) and 1131 maternal peaks (B) identified from hybrid endosperm of reciprocal crosses. (C) Overlap between maternal peaks and MEGs/PEGs/nonimprinted genes (Non-imp). “Genes with peaks” indicates the genes that are overlapped with H3K27me3 peaks identified in the endosperm of both BM and MB. “Genes with analyzable peaks” refers to genes that had SNPs within the H3K27me3 enriched region. “Genes with maternal peaks” refers to genes that contained maternal-specific H3K27me3 peaks identified in this study. The number in parentheses indicates the number of genes that were analyzed. (D) The distribution of maternal-specific H3K27me3 peaks located in PEGs and their 2-kb up- and downstream regions. (E,F) The distributions of DNA methylation levels for maize endosperm H3K27me3 peaks (E) and maternal-specific peaks (F) in shoot, embryo, and endosperm of B73. (G) The distribution of parental DNA methylation of maternal peaks.

Figure 3.

Allelic views of H3K27me3 and DNA methylation in a genomic region. (A) Allelic levels of H3K27me3 and CG DNA methylation are shown for an ∼400-kb region in both BM and MB endosperm. The region contained nine genes with unknown or nonimprinting status (black block), one PEG (red block), and one MNC (blue block). (B) A zoomed-in allelic view of RNA-seq, H3K27me3, and CG DNA methylation for a confirmed PEG that overlapped with a MNC. The overall expression level of transcribed regions is shown in light blue for both BM and MB. The relative expression levels, the percentage of allelic reads of H3K27me3 ChIP-seq data, and the DNA methylation level for specific SNP sites are shown for both maternal and paternal alleles, with red lines for the paternal allele (P) and blue lines for the maternal allele (M). Black rectangle, exon; black line, intron. The gray rectangles highlight the maternal H3K27me3 peaks and pDMRs identified in this region. The lines (or dots) correspond to all analyzable sites.

Figure 4.

Models proposed for the regulation of imprinting in maize endosperm. (A) Model for maternally activated MEGs. MEGs expressing specifically in endosperm are probably maternally activated. Most of these MEGs are associated with pDMRs located around the upstream and 5′ portion of gene body regions. In central cells, the methylation of pDMR regions can be removed by DME-like demethylase, but not in sperm cells. After fertilization, pDMRs that were identified in the endosperm exhibit maternal DNA demethylation and paternal DNA hypermethylation. Maternal DNA demethylation of these MEGs results in their maternally preferred expression. (B) Model for paternally repressed MEGs. MEGs expressing in endosperm but also in other tissues are probably paternally repressed. The nonexpression of the paternal alleles of these MEGs may have resulted from paternally specific unknown repressor(s). (C) Model for maternally repressed PEGs. PEGs expressing in endosperm but also in other tissues are probably maternally repressed. Paternal preferred expressions of these PEGs require both DNA methylation and H3K27me3. In central cells, DNA methylation in pDMRs is removed, but not in sperm cells. In the endosperm, H3K27me3 only targeted to the maternal alleles with DNA demethylation of the pDMR regions (demonstrated by maternal-specific H3K27me3). As the paternal alleles remain hypermethylated, they cannot be targeted by H3K27me3. Repression of the maternal alleles by H3K27me3 led to the paternal preferred expression of these PEGs. (D) Model for paternally activated PEGs. PEGs expressing specifically in endosperm are probably paternally activated (de-repressed) PEGs. In other tissues, these PEGs possibly possess biallelic H3K27me3 peaks. As only maternal-specific H3K27me3 peaks were identified in endosperm, it is likely that for these PEGs, their paternal alleles have lost their H3K27me3, which resulted in paternally specific expression. The dashed boxes around the central cell and sperm cell indicate that the gene expression and chromatin modification in these tissues have not been supported by experimental data.