Abstract
Physical interaction of regulatory elements in three-dimensional space poses a challenge for studies of disease because non-coding risk variants may be great distances from the genes they regulate. Experimental methods to capture these interactions, such as chromosome conformation capture, usually cannot assign causal direction of effect between regulatory elements, an important component of fine-mapping studies. We developed a Bayesian hierarchical approach that uses two-stage least squares and applied it to an ATAC-seq (assay for transposase-accessible chromatin using sequencing) data set from 100 individuals, to identify over 15,000 high-confidence causal interactions. Most (60%) interactions occurred over <20 kb, where chromosome conformation capture-based methods perform poorly. For a fraction of loci, we identified a single variant that alters accessibility across multiple regions, and experimentally validated the BLK locus, which is associated with multiple autoimmune diseases, using CRISPR genome editing. Our study highlights how association genetics of chromatin state is a powerful approach for identifying interactions between regulatory elements.
This is a preview of subscription content, access via your institution
Access options
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$29.99 / 30 days
cancel any time
Subscribe to this journal
Receive 12 print issues and online access
$209.00 per year
only $17.42 per issue
Buy this article
- Purchase on Springer Link
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Data availability
ATAC-seq data for the 100 LCLs used in this study are available from the European Nucleotide Archive (ERP110508). All summary statistics are available from Zenodo (https://doi.org/10.5281/zenodo.1405945).
References
Pombo, A. & Dillon, N. Three-dimensional genome architecture: players and mechanisms. Nat. Rev. Mol. Cell Biol. 16, 245–257 (2015).
Haarhuis, J. H. I. et al. The cohesin release factor WAPL restricts chromatin loop extension. Cell 169, 693–707.e14 (2017).
Fudenberg, G. et al. Formation of chromosomal domains by loop extrusion. Cell Rep. 15, 2038–2049 (2016).
Claussnitzer, M. et al. FTO obesity variant circuitry and adipocyte browning in humans. N. Engl. J. Med. 373, 895–907 (2015).
Smemo, S. et al. Obesity-associated variants within FTO form long-range functional connections with IRX3. Nature 507, 371–375 (2014).
Denker, A. & de Laat, W. The second decade of 3C technologies: detailed insights into nuclear organization. Genes Dev. 30, 1357–1382 (2016).
de Wit, E. & de Laat, W. A decade of 3C technologies: insights into nuclear organization. Genes Dev. 26, 11–24 (2012).
Bonev, B. & Cavalli, G. Organization and function of the 3D genome. Nat. Rev. Genet. 17, 661–678 (2016).
Lieberman-Aiden, E. et al. Comprehensive mapping of long-range interactions reveals folding principles of the human genome. Science 326, 289–293 (2009).
Rao, S. S. et al. A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping. Cell 159, 1665–1680 (2014).
Mifsud, B. et al. Mapping long-range promoter contacts in human cells with high-resolution capture Hi-C. Nat. Genet. 47, 598–606 (2015).
Mumbach, M. R. et al. Enhancer connectome in primary human cells identifies target genes of disease-associated DNA elements. Nat. Genet. 49, 1602–1612 (2017).
Cairns, J. et al. CHiCAGO: robust detection of DNA looping interactions in Capture Hi-C data. Genome Biol. 17, 127 (2016).
Grubert, F. et al. Genetic control of chromatin states in humans involves local and distal chromosomal interactions. Cell 162, 1051–1065 (2015).
Waszak, S. M. et al. Population variation and genetic control of modular chromatin architecture in humans. Cell 162, 1039–1050 (2015).
Kumasaka, N., Knights, A. J. & Gaffney, D. J. Fine-mapping cellular QTLs with RASQUAL and ATAC-seq. Nat. Genet. 48, 206–213 (2016).
Delaneau, O. et al. Intra- and inter-chromosomal chromatin interactions mediate genetic effects on regulatory networks. Preprint at bioRxiv (2017).
Giambartolomei, C. et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet. 10, e1004383 (2014).
Voight, B. F. et al. Plasma HDL cholesterol and risk of myocardial infarction: a mendelian randomisation study. Lancet 380, 572–580 (2012).
Do, R. et al. Common variants associated with plasma triglycerides and risk for coronary artery disease. Nat. Genet. 45, 1345–1352 (2013).
Day, F. R. et al. Genomic analyses identify hundreds of variants associated with age at menarche and support a role for puberty timing in cancer risk. Nat. Genet. 49, 834–841 (2017).
Burgess, S. & Thompson, S. G. Mendelian Randomization: Methods for Using Genetic Variants in Causal Estimation. (CRC Press, Taylor & Francis Group, Boca Raton, 2015).
Veyrieras, J. B. et al. High-resolution mapping of expression-QTLs yields insight into human gene regulation. PLoS Genet. 4, e1000214 (2008).
Ignatiadis, N., Klaus, B., Zaugg, J. B. & Huber, W. Data-driven hypothesis weighting increases detection power in genome-scale multiple testing. Nat. Methods 13, 577–580 (2016).
Wakefield, J. A Bayesian measure of the probability of false discovery in genetic epidemiology studies. Am. J. Hum. Genet. 81, 208–227 (2007).
Hemani, G., Tilling, K. & Davey Smith, G. Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet. 13, e1007081 (2017).
The ENCODE Project Consortium. An integrated encyclopedia of DNA elements in the human genome. Nature 489, 57–74 (2012).
Hoffman, M. M. et al. Integrative annotation of chromatin elements from ENCODE data. Nucleic Acids Res. 41, 827–841 (2013).
Tewhey, R. et al. Direct identification of hundreds of expression-modulating variants using a multiplexed reporter assay. Cell 165, 1519–1529 (2016).
Okada, Y. et al. Genetics of rheumatoid arthritis contributes to biology and drug discovery. Nature 506, 376–381 (2014).
Bentham, J. et al. Genetic association analyses implicate aberrant regulation of innate and adaptive immunity genes in the pathogenesis of systemic lupus erythematosus. Nat. Genet. 47, 1457–1464 (2015).
Liu, J. Z. et al. Association analyses identify 38 susceptibility loci for inflammatory bowel disease and highlight shared genetic risk across populations. Nat. Genet. 47, 979–986 (2015).
Lambert, J. C. et al. Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer’s disease. Nat. Genet. 45, 1452–1458 (2013).
Paternoster, L. et al. Multi-ancestry genome-wide association study of 21,000 cases and 95,000 controls identifies new risk loci for atopic dermatitis. Nat. Genet. 47, 1449–1456 (2015).
Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014).
Scott, R. A. et al. An expanded genome-wide association study of type 2 diabetes in europeans. Diabetes 66, 2888–2902 (2017).
Nikpay, M. et al. A comprehensive 1,000 Genomes-based genome-wide association meta-analysis of coronary artery disease. Nat. Genet. 47, 1121–1130 (2015).
Guthridge, J. M. et al. Two functional lupus-associated BLK promoter variants control cell-type- and developmental-stage-specific transcription. Am. J. Hum. Genet. 94, 586–598 (2014).
GTEx Consortium. Genetic effects on gene expression across human tissues. Nature 550, 204–213 (2017).
Battle, A. et al. Characterizing the genetic basis of transcriptome diversity through RNA-sequencing of 922 individuals. Genome Res. 24, 14–24 (2014).
Lappalainen, T. et al. Transcriptome and genome sequencing uncovers functional variation in humans. Nature 501, 506–511 (2013).
Shin, H. Y. et al. Hierarchy within the mammary STAT5-driven Wap super-enhancer. Nat. Genet. 48, 904–911 (2016).
Chen, L. et al. Genetic drivers of epigenetic and transcriptional variation in human immune cells. Cell 167, 1398–1414.e24 (2016).
Jiang, H., Lei, R., Ding, S. W. & Zhu, S. Skewer: a fast and accurate adapter trimmer for next-generation sequencing paired-end reads. BMC Bioinformatics 15, 182 (2014).
Ding, Z. et al. Quantitative genetics of CTCF binding reveal local sequence effects and different modes of X-chromosome association. PLoS Genet. 10, e1004798 (2014).
Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformatics 25, 1754–1760 (2009).
Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357–359 (2012).
Kim, D. et al. TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome Biol. 14, R36 (2013).
Browning, B. L. & Browning, S. R. Genotype imputation with millions of reference samples. Am. J. Hum. Genet. 98, 116–126 (2016).
Durand, N. C. et al. Juicer provides a one-click system for analyzing loop-resolution Hi-C experiments. Cell Syst. 3, 95–98 (2016).
Hoffman, M. M. et al. Unsupervised pattern discovery in human chromatin structure through genomic segmentation. Nat. Methods 9, 473–476 (2012).
Ernst, J. & Kellis, M. ChromHMM: automating chromatin-state discovery and characterization. Nat. Methods 9, 215–216 (2012).
Wood, S. N. Generalized Additive Models: An Introduction with R. (Chapman & Hall/CRC, Boca Raton, Florida, USA, 2006).
Weirauch, M. T. et al. Determination and inference of eukaryotic transcription factor sequence specificity. Cell 158, 1431–1443 (2014).
Anders, S. & Huber, W. Differential expression analysis for sequence count data. Genome Biol. 11, R106 (2010).
Acknowledgements
N.K., A.J.K., and D.J.G. were funded by Wellcome Trust grant no. WT206194.
Author information
Authors and Affiliations
Contributions
D.J.G. and N.K. conceived and designed the experiments. A.J.K. performed the experiments. N.K. performed statistical analysis and analyzed the data. N.K. and A.J.K. contributed reagents, materials, and analysis tools. D.J.G., N.K., and A.J.K. wrote the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Integrated supplementary information
Supplementary Figure 1 Mapped causal interactions using the pairwise hierarchical model.
(a), An example of a peak pair in a causal or pleiotropic relationship. The inset heatmaps show posterior probabilities for the interaction categories among the three peaks. The left heatmap shows the posterior probability of peak pairs in a causal relationship PPCjk, where the row indicates upstream peak j and the column indicates downstream peak k, suggesting the leftmost peak A is likely to be the upstream peak for the others (B and C). The upper triangle element of the right heatmap indicates the posterior probability of pleiotropy and the lower triangle element indicates the posterior probability of the linkage category. Our model predicts a pleiotropic relationship at B and C, driven by the variant in peak A. (b), Numbers of different types of directed acyclic graphs (DAGs) found in a high-confidence set of causally interacting peaks (see Section 4.6 in the Supplementary Note for details). (c), The MB21D2 locus with causal interactions predicted across 60 peaks; the highest peak in the middle was inferred to be the master regulatory peak with the putative causal variant rs2886870 inside the peak. (d), The numbers of caQTLs mapped by the standard hierarchical model (blue) and causal peak pairs mapped by PHM (red) detected in downsampled data with sample sizes ranging from 20 to 80.
Supplementary Figure 2 Model performance assessment by simulation and real data.
(a), Confusion matrix of mapped interactions for seven different simulated hypotheses. The row corresponds to the true hypothesis and the column corresponds to the inferred hypothesis by PHM. The first four hypotheses are with two causal variants for each peak and the bottom three hypotheses are hybrid hypotheses with a combination of linkage, pleiotropy and causality. See section 4.14 in the Supplementary Note for details. (b), Scatterplot of the posterior probability of causality from peak k to peak j (PPCkj) against the strength of genetic association from the causal variant to peak j (BFj) under the simulation scenario of causality (j → k). The red line shows the misclassification rate of causality (k → j) from the true causality (j → k). The rate is around 40% when caQTL signal for peak j is very weak. It decreases below 1% for BFj > 100. (c), Result of correlation analysis across 53 different cell types from the Roadmap Epigenomics Project. The top panel shows a heatmap of Spearman’s correlation between 32 chromatin accessibility peaks in the TTC34 gene body. The second panel shows normalized chromatin accessibility (see section 2.8 of the Supplementary Note) at 12 peaks across 53 different cell types. Those 12 peaks were selected according to the highly correlated block in the top panel. The third panel shows our ATAC-seq coverage plot stratified by the putative causal variant (rs4648682[G > A]) within the inferred master regulatory peak. PPCjk from the master peak is shown by arrows. (d), Inferred high-confidence (PPCjk = 0.99) causal interaction around the 3ʹ end of the RAP1GAP2 gene. ATAC-seq coverage is stratified by genotype at the putative causal SNP (rs6502671[C > T]) in the left peak. This example also showed causal interaction spanning two adjacent TADs. (e), Box plot of allelic imbalance for three genotypes at rs6502671 (n = 27, 12 and 2 samples for CC, CT and TT, respectively) estimated by aggregated allele-specific counts at heterozygous SNPs in the downstream peak (right peak) in d. Allelic imbalance is observed only in individuals whose genotype at rs6502671 is heterozygous (light blue dots). The downstream peak (right peak) is a strong caQTL given the regulatory SNP (rs6502671) with allelic imbalance π = 0.383 (P = 1.9 × 10–35) estimated by RASQUAL (Nat Genet. 48, 206–213, 2016). This suggests the accessibility linked to the T allele at rs6502671 is significantly lower than that linked to the C allele at rs6502671. In the box plots, the box represents the interquartile range (IQR), the black line is the median, and the whiskers are 1.5 times the IQR above or below the first and third quartiles, with data points outside the whiskers shown by open circles. (f), Scatterplot showing the ratio of putative TF binding affinities between the reference and alternative alleles at each lead SNP (predicted by simple hierarchical model) against the ratio of ATAC-seq allele-specific (AS) chromatin accessibility counts (n = 26,213 SNPs). AS counts for each lead SNP were generated by aggregating AS counts for heterozygous individuals at each lead SNP. The red line shows the linear regression line (beta = 0.36 and P = 2.0 × 10–25).
Supplementary Figure 3 Comparison with 3C-based assays.
(a), Classification of Hi-C annotation. Causal interactions were classified into (1) outside TAD: two peaks located in an unannotated TAD region, (2) one-side TAD: one of two peaks located in an annotated TAD, (3) across TADs: two peaks span two different annotated TADs, (4) within TAD: two peaks located inside the same TAD annotation, (5) across inner TADs: two peaks spanning one or more TAD boundaries within a TAD, and (6) loop anchor; two peaks, both located in a Hi-C loop anchor region. (b), The number of causal interactions overlapping with the six different annotations shown in a. (c), Enrichment of causal interactions overlapping with the six different annotations shown in a. To compute the odds ratio (OR), n = 15,884,515 peak pairs (peak distance > 35 kb) were used. (d), QQ plots of RASQUAL P values for allele-specific chromatin accessibility analysis stratified by the Hi-C annotation in a. (e), Enrichment OR with 95% confidence interval of the same effect direction in different 3C-based annotations. The number of causal peak pairs to compute the OR for each category was n = 3,060, 3,060, 3,060, 3,060, 3,060, 3,060, 11,534, 11,534, 11,534, 3,060, 11,534 and 11,534 from top to bottom. The bottom two annotations show the effect direction is less concordant when two peaks span TAD boundaries or insulators (CTCF peaks). (f), Estimated peak-pair-level prior probability excluding peak pairs separated by short or intermediate distances (1, 3 or 5 kb). (g), Number of mapped causal interactions with estimated prior probabilities in f. The left four bars show the total sum of posterior probabilities and the right four bars are based on the high-confidence set with PPCjk > 0.5.
Supplementary Figure 4 Comparison with genome segmentation and the gene body effect on chromatin accessibility.
(a), Percentage of concordant effect directions between peaks overlapping seven different segmentation annotations. The white dots show the effect direction is significantly higher or lower than the background after multiple-testing correction (q value < 0.05 for the 7 × 7 = 49 tests). (b), 2 by 2 table illustrating the testing for enrichment of peaks within a gene body, downstream of a promoter peak colocalized with the gene. The background is the downstream peak found in a mirrored gene body 5ʹ of the promoter. A promoter peak was defined by colocalizing with the target gene expression QTL (posterior probability > 0.5) and overlapping with the ENCODE active promoter. (c), Scatterplot of the expression level of a gene against average chromatin accessibility in the gene body (n = 20,046 protein-coding genes with one or more chromatin accessibility peaks found within the gene body). The red line shows the regression line (beta = 0.18, P = 2.8 × 10–1,300).
Supplementary Figure 5 Fine-mapping eQTLs of the GPATCH2L gene using mapped causal interactions as an annotation.
(a), ATAC-seq coverage plot stratified by the genotype of rs74067641[T > C]. There are ten causally interacting peaks in the regulatory region of the GPATCH2L gene. Peak 6 is inferred as the master regulatory peak. The strength of PPCjk is shown by the arrows with color coding. (b), Heatmap showing the posterior probability of causality from peak j (row) to peak k (column) (PPCjk) between the ten peaks in a. A row corresponds to an upstream peak j and a column corresponds to a downstream peak k. The heatmap illustrates that peak 6 regulates multiple peaks (on row 6), while there is no upstream peak that regulates peak 6 (on column 6). (c), Probability of master regulator (PMR) for each peak in a.
Supplementary Figure 6 CRISPR validation of the BLK locus.
(a,b), Scatterplots of log10 Bayes factors (BF) for SLE (Nat. Genet. 47, 1457–1464, 2015) and RA (Nature 506, 376–381, 2014) against the log10 Bayes factor of the eQTL for the BLK gene. The color of each point represents the LD index/r2 between a variant in the cis region and the index SNP (indicated by a purple point). (c,d), Regional plots of disease associations (based on log10 Bayes factor) for SLE and RA. The color of each point is the r2 value between a variant in the cis region and the index SNP (indicated by a purple point). (e–g), ATAC-seq coverage plots around the putative causal variants. Coverage is stratified by the three genotypes of the causal variants indicated in each panel. The sequence logo is drawn based on the known transcription factor binding motif in CISBP (Cell 158, 1431–1443, 2014). We selected the motif that gives the highest binding affinity difference between the reference and alternative alleles and the orientation of the affinity was also the same as the chromatin openness. (h), Aggregated ATAC-seq coverage around the BLK locus for the two deletion lines (green, D1; orange, D2) superimposed on the parental line (R; navy). (i), Average chromatin accessibility in terms of FPKM at each peak (square dots) segregated by the three different cell lines (navy, parental line; green, D1 heterozygous line; orange, D2 heterozygous line). The peaks surrounded by red rectangles showed a consistent pattern to the caQTL results. (j), Expression levels of the FAM167A gene for the three different cell lines with two replications (n = 6 replicates in total; see section 5.7 in the Supplementary Note for details of statistical hypothesis testing).
Supplementary Figure 7 Hypothetical features of causal interactions mapped by PHM and CHi-C.
(a), Scatterplot of CHi-C contact distance and read count that supports the contact (CHiCAGO score greater than 1). The color of each point is proportional to the CHiCAGO score. Contacts with CHiCAGO score > 10 are colored the same as a score of 10. (b), An example of causal interactions between strong enhancers found in an intergenic region. These peaks are not colocalised with any eQTLs within the 1-Mb cis window. (c), An example of mapped causal interaction starting from a strong enhancer (SE) to an active promoter (AP). For comparison, the coverage plot of H3K27ac ChIP-seq (Cell 162, 1039–1050, 2015) is shown in parallel with our ATAC-seq. Coverages are stratified by the three genotypes of the putative causal caQTL SNP (rs6454679[G > C]). The putative causal SNP exists around the summit of an ATAC peak that corresponds to the nucleosomal-depleted region between two histone modification ChIP-seq peaks. (d), Distributions of absolute effect sizes (|beta|) at peaks that are upstream caQTL peak (U; n = 3,469 peaks), downstream caQTL peak (D; n = 4,321 peaks), downstream caQTL peak labeled as an active promoter (AP; n = 309 peaks), poised promoter (PP; n = 75 peaks), strong enhancer (SE; n = 599 peaks), weak enhancer (WE; n = 483 peaks), repressed region (R; n = 1,202 peaks), transcribed region (T; n = 293 peaks) or insulator (I; n = 370 peaks). The upstream peaks were defined as peak j with PPCjk > 0.5 (or peak k with PPCkj > 0.5), and the downstream peaks were defined as peak k with PPCjk > 0.5 (or peak j with PPCkj > 0.5). Two-sided Kolmogorov–Smirnov test showed the average effect size for AP is significantly lower than that of SE (P = 3.0 × 10–12).
Supplementary information
Supplementary Text and Figures
Supplementary Figures 1–7 and Supplementary Note
Rights and permissions
About this article
Cite this article
Kumasaka, N., Knights, A.J. & Gaffney, D.J. High-resolution genetic mapping of putative causal interactions between regions of open chromatin. Nat Genet 51, 128–137 (2019). https://doi.org/10.1038/s41588-018-0278-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41588-018-0278-6
This article is cited by
-
Integrative analysis of transcriptomic and epigenomic data reveals distinct patterns for developmental and housekeeping gene regulation
BMC Biology (2024)
-
Fundamentals for predicting transcriptional regulations from DNA sequence patterns
Journal of Human Genetics (2024)
-
Single-cell chromatin accessibility and transcriptomic characterization of Behcet’s disease
Communications Biology (2023)
-
Multimodal single cell analysis infers widespread enhancer co-activity in a lymphoblastoid cell line
Communications Biology (2023)
-
Functional genomics identify causal variant underlying the protective CTSH locus for Alzheimer’s disease
Neuropsychopharmacology (2023)
