Fig. 2: Global and fine-scale architectural changes characterize the TSA chromatin state.
a, Aggregate plots of homotypic interactions between A and B compartments in cis. b, Metaplots showing normalized H3K27ac ChIP–seq read density over A and B compartments (bin size = 1 kb). Shading represents s.d. of the mean. c, Distribution of H3K27ac ChIP–seq reads (left) and upregulated and downregulated TSSs (right) by compartment. d, Schematic representation of biophysical modeling where EAA and EBB correspond to the attraction energies and KθA and KθB correspond to the stiffness of the chromatin fiber, in A and B domains, respectively. Each chromosome was modeled by a 20-Mb chain with beads representing 5-kb DNA. A and B domains were set at 1.5 Mb in size to match the mean compartment size derived from the Micro-C data, resulting in six A domains, six B domains and two telomeric regions of 1 Mb at the extremities of the chain. A nucleus was modeled using 20 chains. e, Trans-contact ratio in DMSO and TSA in the Micro-C data (n = 19 chromosomes) and in the model (n = 600 corresponding to 20 simulated chains in 30 replicates). Box plots show median (central line), the 25th and 75th percentiles (box limits) and 1.5× IQR (whiskers). Outliers are not shown. f, Compartment strength in DMSO (left) and TSA (right) in the Micro-C data and in the model (n = 150 corresponding to number of values in average compartment profiles). Box plot elements are as in e. g, Volcano plot of differential loops between DMSO and TSA (significance cutoffs (dashed lines)—adjusted P value (derived from Wald test, corrected for multiple testing with BH approach) <0.05, absolute FC > 1.5). Positive log2(FC) indicates stronger interaction in TSA. h, Micro-C contact maps showing differential looping at the Bcar1, Zfp462 and Tbx3 loci. i, Aggregate plots of Micro-C signal around differential loops at 4-kb resolution. j, GO enrichment among genes closest to differential loop anchors. P values correspond to one-sided Fisher’s exact test corrected for FDR with the BH method. k, Pile-ups showing Micro-C signal around all loops identified in DMSO (top) and TSA (bottom) (resolution = 4 kb). Quantification of aggregate loop signal (n = 9 corresponding to the central 3 × 3 pixels) is shown on the right (paired two-tailed t test; **P < 0.01). Box plot elements are as in e. l, Pile-ups of Micro-C signal around loops stratified by the presence (CTCF loops) or absence (non-CTCF loops) of CTCF ChIP–seq peaks at loop bases (resolution = 4 kb). Non-CTCF loops have been further divided into active and repressive based on the presence of activating (H3K27ac, H3K4me1) or repressive (H3K27me3, H3K9me3) ChIP–seq signal at loop bases. Quantification of aggregate loop strength (n = 9 corresponding to the central 3 × 3 pixels) is shown on the right (paired two-tailed t test; *P < 0.05, ***P < 0.001). Box plot elements are as in e. IQR, interquartile range; Res, resolution.
