Personal and population genomics of human regulatory variation

Abstract

The characteristics and evolutionary forces acting on regulatory variation in humans remains elusive because of the difficulty in defining functionally important noncoding DNA. Here, we combine genome-scale maps of regulatory DNA marked by DNase I hypersensitive sites (DHSs) from 138 cell and tissue types with whole-genome sequences of 53 geographically diverse individuals in order to better delimit the patterns of regulatory variation in humans. We estimate that individuals likely harbor many more functionally important variants in regulatory DNA compared with protein-coding regions, although they are likely to have, on average, smaller effect sizes. Moreover, we demonstrate that there is significant heterogeneity in the level of functional constraint in regulatory DNA among different cell types. We also find marked variability in functional constraint among transcription factor motifs in regulatory DNA, with sequence motifs for major developmental regulators, such as HOX proteins, exhibiting levels of constraint comparable to protein-coding regions. Finally, we perform a genome-wide scan of recent positive selection and identify hundreds of novel substrates of adaptive regulatory evolution that are enriched for biologically interesting pathways such as melanogenesis and adipocytokine signaling. These data and results provide new insights into patterns of regulatory variation in individuals and populations and demonstrate that a large proportion of functionally important variation lies beyond the exome.

Figure 1.

Overview of data used in the analyses. (A) Schematic of the DNase I data. Binding of regulatory proteins to DNA (blue rectangle) results in nucleosome (open circles) displacement and local chromatin remodeling, and these regions are susceptible to cleavage with the endonuclease DNase I. High-throughput sequencing of libraries made from digested nuclei reveals DNase I hypersensitive sites, detectable by increased depth of coverage. Peaks are defined as 150-bp windows centered on the area of maximum cleavage (The ENCODE Project Consortium 2012). Within hypersensitive sites, footprints of regulatory factor binding are observed as decreased cleavage. (B) Unrooted neighbor-joining tree of the 53 unrelated individuals colored by population. Abbreviations are described in Supplemental Table 2.

Figure 2.

Characteristics of regulatory variation among individuals. (A) Total number of variants in DNase I peaks, footprints, and the exome stratified by GERP score. (B) Distribution of the number of variants per individual in DNase I peaks, footprints, and the exome. (C) Distribution of the number of variants per individual with GERP ≥ 3 in DNase I peaks, footprints, and exomes.

Figure 3.

Significant variation of diversity between 732 cis-regulatory motifs. (A) For each motif, average diversity is plotted as a black circle, and 95% confidence intervals obtained by bootstrapping are shown as gray lines. The light blue and yellow rectangles denote the 95% confidence intervals of diversity in fourfold synonymous sites (FFSs) and the exome, respectively. (Red vertical lines) Motifs that belong to the indicated class of transcription factor. (Black vertical lines) Motifs where at least 50% of all instances of that motif contain a CpG dinucleotide. (B) Normalized diversity in motifs versus non-normalized diversity. Motifs with a CpG (defined as above) are plotted in red. (Dashed line) Best fit for non-CpG motifs (r = 0.70, P < 10⁻¹⁶).

Figure 4.

Heterogeneity of polymorphism across cell types. (A) Distribution of normalized nucleotide diversity (black points) across DNase I peaks in 138 cell types. Vertical bars around peaks indicate 95% confidence intervals obtained by bootstrapping. (Blue rectangle) 95% confidence interval for normalized nucleotide diversity in fourfold degenerate sites. (B) Venn diagram showing the amount of shared and unique sequence for DNase I peaks among normal/primary, malignant, and iPS/ES cell types. The barplot on the left shows average normalized diversity for several categories of peaks in the Venn diagram. Shared all and shared two denote peaks shared among all three categories and between any two categories, respectively. N, M, and SC denotes peaks specific to normal/primary, malignant, and stem cell (iPS/ES) cell types, respectively.

Figure 5.

Malignant cell lines exhibit significantly more singleton DNase I peaks than normal cell lines. (Triangles) Observed proportion of singleton peaks. (Blue and green lines) Distribution (density histograms) of singleton peaks when randomly sampling 29 (blue) or five (green) cell types; this is the distribution of the number of singleton peaks we would expect if malignant or stem cells were similar to normal cells, respectively. Note the malignant category (blue) shows significantly more singleton peaks than expected given its sample size, but the stem cell category (green) falls within the expected range.

Figure 6.

Genome-wide distribution of population structure in regulatory DNA. (A) Genome-wide distribution of locus-specific branch lengths (LSBLs) for Africans, Asians, and Europeans, respectively. Note that the valley of uniform LSBL on chromosome 17 in Europeans corresponds to the MAPT region that is segregating a large chromosomal inversion (Zody et al. 2008). (B) Distribution of the proportion of highly differentiated DNase I peaks found for different categories of cell types. (SC) Stem cells (iPS/ES); (I) immortalized; (M) malignant; (N) normal/primary cell types. (C) Distribution of African LSBL across intron 1 of VDR. (D) Distribution of European LSBL across intron 4 of FTO. In panels C and D, peaks are shown as red rectangles and exons as black rectangles.