Extensive variation in chromatin states across humans

Abstract

The majority of disease-associated variants lie outside protein-coding regions, suggesting a link between variation in regulatory regions and disease predisposition. We studied differences in chromatin states using five histone modifications, cohesin, and CTCF in lymphoblastoid lines from 19 individuals of diverse ancestry. We found extensive signal variation in regulatory regions, which often switch between active and repressed states across individuals. Enhancer activity is particularly diverse among individuals, whereas gene expression remains relatively stable. Chromatin variability shows genetic inheritance in trios, correlates with genetic variation and population divergence, and is associated with disruptions of transcription factor binding motifs. Overall, our results provide insights into chromatin variation among humans.

Fig. 1

Variation in chromatin, factors, and expression across individuals. (A) Number and fraction of enriched regions (for chromatin marks and factors) and expressed genes (for RNA) that are variable across individuals. (B) Composition (emission probability) of five chromatin marks and CTCF in 15 chromatin states: TssA (active promoters), TssF (flanking active promoters), Tx (strong transcription), TxW (weak transcription), EnhA (active enhancers with H3K4me3), TxEnhA (active enhancers in transcribed regions), Enh (active enhancers without H3K4me3), TxEnh (active enhancers without H3K4me3 in transcribed regions), EnhW (weak enhancers), TxEnhW (weak enhancers in transcribed regions), TssP (poised promoters), EnhP (poised enhancers), ReprPC (Polycomb repressed), Ctcf (CTCF enriched regions), and Low (low signal). (C) Examples of a nonvariable and a variable region. Coordinates are in build hg19 of the human reference sequence. State colors are as in (B). (D) log10 ratio of the observed probability that a region switches from one state (row) to another (column) in any pair of individuals relative to background switching across pairs of replicates.

Fig. 2

Genetic basis of chromatin variation. (A) Spearman correlation between genotype and signal at variable and nonvariable H3K27ac regions after correcting for differences in length and signal strength. For the null sets, we shuffled the signal. (B) Correlation of allelic biases between the parents and the daughter of the YRI trio at allele-specific SNPs of the daughter that are homozygous in both parents (Pearson correlation coefficients are in the legend; linear fits are shown as lines). Only marks with at least 50 SNPs are shown.

Fig. 3

Correlation between chromatin signal and ancestry. (A) Fraction and number of regions where ancestry has a significant contribution to signal variation. (B) Row-standardized signal at H3K27ac peaks from (A), grouped into four clusters (C1 to C4). (C) Fraction of regions from (B) with SNPs characteristic of individuals in each ancestry group. Each column is divided by its maximum. The maximum genetic divergence for each ancestry group (squares with the value 1) is achieved in the cluster that shows the most divergent signal for that group [from (B)]. (D) Enrichment of regions from (A) for SNPs with high FST. Stars indicate P < 0.01 (binomial test) after accounting for the overall enrichment for SNPs.

Fig. 4

Mechanism and functional consequences of chromatin variation. (A) Correlation coefficients of TF motif disruption scores and H3K27ac signal across individuals. Motifs are sorted based on the number of associated peaks; peaks are sorted based on their associated motifs. (B) Log2 fold-enrichment of motifs in promoter (red) versus enhancer (orange) states. Only significant enrichments (Fisher’s exact test P < 0.05) are shown. (C) eQTLs and GWAS hits in variable regions. Stars indicate P < 0.05.