Fig. 3: Discovery of candidate silencer CREs using rice ACR atlas.

a, The classification of ACRs based on their proximity to H3K27me3 peaks. We first subdivided ACRs into two groups: H3K27me3-associated ACRs (found within or surrounding H3K27me3 peaks) and H3K27me3-absent ACRs. The H3K27me3-associated ACRs were further divided into broad ACRs, characterized by chromatin accessibility in at least five cell types, and cell-type-specific ACRs, accessible in less than three out of six examined cell types across all the species. b, Leaf H3K27me3 ChIP-seq reads near summits of distinct ACR groups. Zm, Z. mays; Sb, S. bicolor. c, A comparative analysis of expression levels and chromatin accessibility of genes surrounding broad ACRs under and outside of H3K27me3 peaks. **P < 0.01 (ranging from 2.8 × 10⁻³⁴ to 7.4 × 10⁻¹⁹), which was performed using one-tailed Wilcoxon signed rank test (alternative = ‘greater’). The broad ACRs where the H3K27me3 region overlapped >50% of the gene body were positioned within 500 to 5,000 bp upstream of the transcriptional start site of their nearest gene. ‘n’ represents the number of genes analysed. The centre line indicates the median; the box limits indicate the upper and lower quartiles; the whiskers indicate 1.5 times the interquartile range (IQR); the dots represent the outliers. NS indicates not significant. d, A screenshot illustrating an H3K27me3-broad ACR harbouring a reported silencer within an H3K27me3 peak in the panicle organ, located approximately 5.3 kb upstream of FZP. e, Percentage of H3K27me3-broad ACRs in O. sativa, Z. mays and S. bicolor capturing six known motifs enriched in PREs in A. thaliana. **P < 0.01 (ranging from 1.2 × 10⁻¹² to 5.2 × 10⁻⁵), which was performed using one-tailed binomial test (alternative = ‘greater’). Grey bars represent the control which was set by simulating sequences with the same length as ACRs 100 times (see ‘Construction of control sets for enrichment tests’ in the Methods). The error bars indicate the mean ± s.d. f, Alignment of H3K27me3 and EMF2b ChIP-seq reads near summits of distinct ACR groups in O. sativa. g, Percentage of EMF2b ChIP-seq peaks in O. sativa capturing six known motifs enriched in PREs in A. thaliana. **P < 0.01 (ranging from 7.1 × 10⁻¹⁷ to 1.5 × 10⁻³), which was performed using one-tailed binomial test (alternative = ‘greater’). The error bars indicate the mean ± s.d. h, Deletion of a rice H3K27me3-broad ACR significantly increased the expression of the nearby gene LOC_Os08g06320. **P < 0.01. Significance testing was performed using one-tailed t-test (alternative = ‘greater’). WT, wild-type plant. i, Comparison of gene expression linked to H3K27me3-broad ACRs that contain PRE motifs with SNPs in the ‘Zhenshan 97’ (ZS97) genotype using ‘Nipponbare’ (NIP) as the reference. Significance tests were performed using one-tailed Wilcoxon signed rank test (alternative = ‘greater’). The error bars indicate the mean ± s.d. Both the mutants and WT samples include three biological replicates. j, Alignment of leaf H3K27me3 Chip-seq reads at summits of TSS of genes derived from i. The control genes include genes overlapping with H3K27me3, which are shared between both genotypes. k, A screenshot illustrating an H3K27me3-broad ACR containing a PRE-associated motif with an SNP situated at 1.2 kb upstream of LOC_Os11g08020 gene, which associates with lower H3K27me3 signal and higher expression of the LOC_Os11g08020 in the ‘Zhenshan 97’ genotype. l, Four TF families, highlighted in red, were significantly enriched in H3K27me3-broad ACRs. The motif data were collected from 568 TFs from A. thaliana belonging to 24 families within the JASPAR database (ref. ¹²⁵). The P value was computed using a hypergeometric test (alternative = ‘greater’). ZnF, C2H2 zinc-finger; SBP, SQUAMOSA PROMOTER BINDING PROTEIN.