Fig. 1

The chromatin insulator protein Beaf32 protects genes from H3K27me3 spreading at borders. a Genomic view of our ChIP-seq data for H3K27me3 reads (y-axis) in S2 cells depleted for the insulator protein Beaf32 (“Beaf-KD”) or in control, siRNA-treated wild-type cells (“WT” control). The orange and blue triangles represent Beaf32 and GAF binding sites, respectively, as detected by ChIP-seq (see “Methods”). The blue bar represents a large H3K27me3 heterochromatin domain as detected by hidden Markov model (HMM; see “Methods”) often bordered by binding sites of insulator proteins such as Beaf32. Note that Beaf32 binding sites are enriched at the borders of H3K27me3 domains (930 sites; Fisher exact test, p value < 1e−151) (see panel c). The red bars represent “micro-domains” of H3K27me3 (see text for details). The dashed rectangles in black highlight the corresponding regions where such H3K27me3 levels may decrease in Beaf-KD compared to WT control, contrasting with the apparent increase in H3K27me3 levels near borders flanked by a Beaf32 site (see dashed rectangle in orange; see also panels b and c). b, c Averaged H3K27me3 levels centered surrounding the H3K27me3 domain borders (see “Methods” for details) in absence or in presence of a Beaf32 site (panels b and c, respectively) or a Beaf32 site flanking heterochromatin domains. The asterisks indicate a significant p value (Wilcoxon pairwise test, p value < 1.e−4) for the statistical difference in H3K27me3 levels (in the 0–4 kb euchromatin segments next to heterochromatin), between Beaf-KD compared to WT control cells (c) compared to borders without Beaf32 sites (“NS”; not significant). d Distribution of sites (bins) with increasing H3K27me3 levels relative to heterochromatin borders (x-axis, position 0) with Beaf32 (blue bars) or without (red bars). Note that Beaf32 specifically deregulates H3K27me3 at sites flanking repressive H3K27me3 heterochromatin flanked by a Beaf32 binding site. The error bars represent the variability of the signal from independent replicates with all bins in the indicated intervals (see “Methods”). e Genome-wide analysis representing the relative enrichments of genes associated with H3K27me3 variations scored in their TSSs (± 1 kb), depending on binding of insulator proteins (Beaf32, Cohesin, GAF, dCTCF, DREF) in the same windows (TSSs ± 1 kb). All TSSs were systematically ranked according to variations between Beaf-KD and control cells of H3K27me3 levels (TSSs ± 1 kb) (see “Methods”). Log odds ratios were then calculated for all genes for ranking them in quintiles. Enrichment tests were performed by intersecting such quintiles (groups of genes ranked depending on H3K27me3 variations ± 1 kb surrounding TSS) with TSSs with or without insulator protein sites in the same interval (± 1 kb TSS) (see also Additional file 1: Fig. S1F). The indicated p values (asterisks) were calculated using Fisher’s exact test in presence as compared to absence of sites