Fig. 2

Cohesin and CTCF ChIP-seq binding strength and proximity to genes. a Box plots of normalized ChIP-seq signal for the peak sets indicated on the x-axis. Peaks with sex differential binding for cohesin (top graph) and CTCF (bottom graph) are shown. Each pair of boxplots represents the male and female ChIP-seq signal for the same set of peaks, defined by their sex bias and peak type (CAC or CNC, for ΔCohesin peaks; and CAC or Lone CTCF, for ΔCTCF peaks), as indicated below the x-axis. Peak scores were calculated by average intra-peak ChIP signal, normalized by total sequence reads per million in peak (RIPM; see “Materials and methods”). Female-biased peaks were, on average, stronger than male-biased peaks by M–W test: p ≤ 0.001 for female vs male CAC(ΔCoh), CAC(ΔCTCF), and for CNC, but not for Lone CTCF peaks. b Distance from each indicated set of cohesin and CTCF peaks to the nearest enhancer DHS. Cumulative frequency curves indicate the fraction of each group on the y-axis, within the distance in kb to the nearest enhancer DHS indicated on the x-axis. Enhancer DHS were defined based on their high ratio of the enhancer histone mark H3K4me1 over the promoter mark H3K4me3 at DHS [31]. Sex-biased CNC peaks are closer to enhancer DHS (median distance to eDHS of 0.22 kb for male-biased CNCs and 0.12 kb for female-biased CNCs; KS pval < 0.0001 for all comparisons) than the other CTCF and cohesin peak classes (M CAC(ΔCTCF): 14.98 kb; F CAC(ΔCTCF) 13.76 kb; M Lone ΔCTCF: 13.88 kb; F Lone ΔCTCF: 7.17 kb). Female-biased CNC peaks are significantly closer to enhancer DHS than are male-biased CNC peaks (p = 0.0351; KS t-test). Male-biased CAC(ΔCohesin) peaks were closer to enhancers than female-biased CAC(ΔCohesin) peaks (p = 0.002; KS t-test), however, the reverse was found for CAC(ΔCTCF) peaks (p = 0.0052; KS t-test). Distance to nearest enhancer was not significantly different between male-biased and female-biased Lone CTCF peaks (p = 0.1068; KS t-test). P values for comparisons between male-biased and female-biased peaks of the same class are shown for each plot (KS t-test). c Distance from each indicated set of cohesin and CTCF peaks to the nearest TSS. Cumulative frequency curves indicate the fraction of each group on the y-axis within the distance in kb to the nearest TSS indicated on the x-axis. TSS for protein coding (RefSeq) and liver lncRNA genes were considered [80]. Female-biased cohesin and CTCF peaks are closer to TSS than male-biased CTCF and cohesin peaks of the same class (significance by KS t-test is indicated at top left of each plot). Distance to the TSS was not significantly different for male-biased versus female-biased CNC peaks (p = 0.1458; KS t-test). d Proximity of sex-biased cohesin and CTCF binding sites to sex-biased genes. Peak designations were as follows: Proximal, peaks < 20 kb from a sex-biased gene TSS; Intra-TAD, peaks within the same intra-TAD loop as a sex-biased gene; or TAD, peaks in the same TAD as a sex-biased gene. Each of these groups is mutually exclusive. TAD loop [34] and intra-TAD loop [31] coordinates were from the indicated references. A set of 983 sex-biased biased protein-coding genes was used in this analysis (see Additional file 3: Table S1 of [54]). e Cumulative frequency curves show the fraction of each group (y-axis) within the distance in kb to the nearest sex-biased DHS or H3K27ac genomic region (x-axis), based on a merged list of published sex-biased DHS [13] and sex-biased H3K27ac ChIP-seq peaks [55] for male and female mouse liver. For this analysis, CAC peaks with sex-biased binding of CTCF and cohesin were combined and presented as a single group [CAC (Both)]. Male-biased and female-biased CNC peaks are significantly closer to sex-biased DHS/H3K27ac than the four other peak classes (p < 0.001; KS t-test). Female-biased CNC peaks were significantly closer to sex-biased DHS/H3K27ac than male-biased CNC peaks (p = 0.0094; KS)