Figure 1.
P53 genomic binding relative to chromatin state and histone modifications.
Upper Panel: UCSC (University of California Santa Cruz, California, United States) Genome Browser view of experimental data integrated with ENCODE data. Tracks A–B (p53 genomic binding), F (gene expression fold change), H, and I are experimental data for LCLs generated in this study. Track G: ENCODE/Broad chromatin state segmentation (ChromHMM model, unstressed GM12878 cells downloaded from ENCODE). Tracks A–E display p53 ChIP-seq peaks in the GDF15 region (e.g. orange box). Additional genome-wide information on p53 occupancy was derived from published data as described in methods. Track C: Nutlin–treated MCF7 (breast cancer cell line), Track D: 5-fluorouracil-treated MCF7, Track E: 5-fluorouracil treated IMR90 (embryonic fibroblast). Histone H3 lysine 4 tri-methylation (H3K4me3, tracks H, I) is present at an actively transcribed promoter (PGPEP1, red box) but absent (GDF15, Track I purple box) at the GDF15 promoter displaying insulating CTCF marks (CTCF, track J, blue arrow, blue box. Red box highlights the promoter region of the p53 inducible gene, PGPEP1. Numerous histone modifications align at this position and the ChromHMM track displays red indicative of an active promoter (State 1). In the region within the orange box where p53 peaks were detected, the ChromHMM track displays orange, a high degree of DNase I hypersensitivity (track K, DHS), and multiple histone modifications. Following treatment of GM12878 with DXR, tracks H and I (purple box), change in H3K4me3 marks. Track L shows placental mammalian conservation score (PhyloP). Lower Panel: Illustration of the use of the chromatin state model classification at the p53 ChIPseq maximum (p53RE) or at gene TSS for data analysis in this study.
Table 1.
Chromatin state characteristics of some known p53 regulated genes.
Figure 2.
Dynamic change in histone H3K4me3 is correlated with stress-induced gene expression change.
(A) Black bars (NT) display linear trend for increasing mean gene expression values with increasing H3K4me3 decile. Grey and open bars show gene expression change is greatest among genes that lack H3K4me3 at baseline. (B) Positive correlation (r2 = 0.41, p <0.00001) of DXR-induced change in H3K4me3 modification at the TSS with the DXR-induced change in gene expression. (C) Overrepresentation of CTCF, Polycomb, heterochromatic marks among up-regulated genes relative to repressed genes. p53-bound genes were grouped into quintiles based on levels of H3K4me3 at TSS. The frequency of CTCF, polycomb and heterochromatic marks in up-regulated genes relative to their frequencies in down-regulated genes was calculated as enrichment ratio and the enrichment ratio for each quintile is graphed (Y-axis).
Figure 3.
Distribution of p53 peaks among ENCODE combined chromatin states.
(A) Distribution of p53 ChIP-seq peaks among chromatin states (ChromHMM7) and the median distance (kb) of the p53-binding locations to TSS of nearby genes. (B) Distribution of chromatin states at TSS of genes nearby p53 peaks.
Figure 4.
Dynamic changes in chromatin accessibility accompany gene expression change.
(A, B) Genes (RRAD and SULF2) with poised promoters display H3K27me3 and H3K4me2 marks (purple boxes), CTCF (blue arrows), no H3K4me3 marks present at the TSS, and low gene expression levels. DNase I-seq experiments were carried out in duplicate and DHS data for each replicate sample is displayed (lower tracks, orange boxes). Following DXR treatment, H3K4me3 marks appear (red box), mRNA expression increases, and increases in DHS occur at 4 and 18 hrs (orange boxes). (C) TGFA with heterochromatin marks at the location of its p53 binding peak (green box). TGFA has similar H3K4me2, H3K27me3, and CTCF marks indicative of a poised promoter at baseline and also displays a large p53 peak, however, little change was observed for DHS, H3K4me3 or gene expression.
Figure 5.
Chromatin state organizes p53 occupancy and gene transactivation.
(A) P53-upregulated genes are grouped by ChromHMM at TSS. Genes with CTCF insulation at the TSS prior to treatment were associated with the largest change in gene expression (fold change +/− SEM). Genes displayed in Fig. 4 are examples. (B) Balloon plot depicts the correlation between peak density and change in gene expression after stratifying by ChromHMM state at the TSS. Linear regression was performed on means of the log2 values for each chromatin state and the trend line shown. The size of the circles indicates the relative number of genes in each group.
Figure 6.
p53 binding occurs at a highly conserved motif and sequence content affect p53 transactivation.
(A) Display of the most enriched sequence motif among 2932 peaks identified by de novo motif discovery in this study. (B) 2932 p53 peaks were grouped by the chromatin states at the RE. Average PWM scores between chromatin state groups were all significantly different. (C) Distribution of down-regulated and up-regulated genes among ChromHMM states (upper panel). The up-regulated and down-regulated genes were compared for p53 binding, RE Motif and PWM scores (lower panel). Among the genes with open promoters, the p53REs for down-regulated genes have lower average levels of p53 occupancy and PWMs than the up-regulated genes, and display nucleotide differences at the junction of two half-sites in p53RE motifs (red box).
Figure 7.
Best PWMs are correlated with chromatin accessibility and p53 occupancy.
(A) For each p53 occupied region, the best PWM is the one closest to the position with maximal p53 binding signal. The best p53RE PWM scores are ordered from low to high, and grouped into deciles. The mean value of p53RE PWM scores for each decile is displayed as diamond symbols on the line graph (left X axis). Stacked bars represent frequency of spacers in each p53RE decile (right Y axis). P53REs with weak, small PWM scores more frequently contain spacers between half-sites. (B) P53 peaks containing the strongest PWM scores are distal to the TSS (decile D10). (C) Regression of average peak density versus average best PWM of REs grouped by PWM deciles. (D) Regression of average p53 peak density against average DHS signal at REs under NT conditions grouped by PWM deciles. Peaks containing the weakest PWMs have lower occupancy and are associated with the highest DHS scores.
Figure 8.
p53REs show diverse levels of conservation, occupancy, strength, and transactivation across chromatin states and among families of transposable elements.
(A) p53REs were grouped by ChromHMM and an average PhyloP score (placental mammals) across the 20nts of the p53RE was calculated. Values are plotted as averages for each ChromHMM group. (B) p53REs were ordered by their PWM value into deciles, and p53 peak density was regressed against human-mouse conservation score, displaying an inverse correlation between peak density and human-mouse conservation. (C) Mean PhyloP conservation scores differ among p53REs embedded in different TE families. (D) Mean PWM scores for p53REs embedded in different TE families were significantly higher than nonTE p53REs. (E) p53 occupancy was significantly higher for p53REs embedded in MLT1H, LTR10B and MER61 families. (F) Mean H3K4me3 fold change at nearby gene TSSs were significantly less among most TEs but higher for L2 and MIR families. * p <5×10−2; ** p <1×10−2; ∧p <1×10−4; # p <1×10−12, + p <1×10−27.