Screening for functional regulatory variants in open chromatin using GenIE-ATAC

Abstract

Understanding the effects of genetic variation in gene regulatory elements is crucial to interpreting genome function. This is particularly pertinent for the hundreds of thousands of disease-associated variants identified by GWAS, which frequently sit within gene regulatory elements but whose functional effects are often unknown. Current methods are limited in their scalability and ability to assay regulatory variants in their endogenous context, independently of other tightly linked variants. Here, we present a new medium-throughput screening system: genome engineering based interrogation of enhancers assay for transposase accessible chromatin (GenIE-ATAC), that measures the effect of individual variants on chromatin accessibility in their endogenous genomic and chromatin context. We employ this assay to screen for the effects of regulatory variants in human induced pluripotent stem cells, validating a subset of causal variants, and extend our software package (rgenie) to analyse these new data. We demonstrate that this methodology can be used to understand the impact of defined deletions and point mutations within transcription factor binding sites. We thus establish GenIE-ATAC as a method to screen for the effect of gene regulatory element variation, allowing identification and prioritisation of causal variants from GWAS for functional follow-up and understanding the mechanisms of regulatory element function.

Graphical Abstract

GenIE-ATAC is a medium throughput screening method that allows identification and validation of the effect of defined nucleotide variants on chromatin accessibility in their native chromatin context. This includes identifying causal variants from GWAS, or understanding transcription factor binding in chromatin. It obviates the need for derivation of clonal lines making it rapid and amenable to primary or differentiated cells.

Figure 1.

Schematic of GenIE-ATAC assay. (A) Pie chart to show the proportion of GWAS-associated variants (20) that can be assayed using GenIE and GenIE-ATAC. DHS: DNase 1 hypersensitive site. (B) Edited pools of cells contain a mixture of WT, edited point mutation (SNP) and a variety of deletion alleles (indel 1, 2 etc.). The chromatin accessibility (and therefore transcription factor occupancy) of each of the genotypes can be quantified by amplicon sequencing of the ATAC material and gDNA extracted from the same population of cells. Values are internally normalised to the WT allele for each library and displayed as a ratio of ATAC to gDNA reads as a measure of relative chromatin accessibility for each allele (Effect size).

Figure 2.

GenIE-ATAC recapitulates genome-wide ATAC allele specific SNP effects on chromatin accessibility. GenIE-ATAC was performed on KOLF_2 iPSCs at six heterozygous SNPs: three positive control SNPs (top) with predicted effects on chromatin accessibility based on chromatin QTL analysis in iPSCs, and three negative control SNPs (bottom) that would not be predicted to have an effect. Graphs show the ratio of chromatin accessibility of the alt allele normalised to the reference genome sequence (effect size). Red bars indicate amplicon sequencing of gDNA (PCR), green bars from amplicon sequencing of ATAC material (Amplicon-based ATAC-seq), blue bars show allele-specific reads from genome-wide ATAC-seq data (Genome-wide ATAC-seq) from KOLF_2 iPSCs. Individual repeats are indicated with dots; boxplot hinges represent the 25th and 75th percentiles, where these are interpolated due to the small number of points (n = 3), and whiskers extend to the most extreme data point not further than 1.5 times the inter-quartile range from the hinge.

Figure 3.

GenIE-ATAC CRISPR screening identifies functional regulatory SNPs. (A) Homozygous SNPs (1–8) were edited using CRISPR to their alternative alleles in KOLF_2 hiPSCs and assayed by GenIE-ATAC. The top panel shows the effect on chromatin accessibility of the HDR-introduced allele relative to the WT allele. The middle panel shows the effect size of deletions within a ±6 bp window of the SNP site relative to the WT allele. The bottom panel shows editing rates (HDR, orange; indels, blue). Data was processed using rgenie. All error bars represent 95% confidence intervals, n = 3. (B) The most frequent deletion types by read count from GenIE-ATAC targeting of SNP 2 (rs12269414). The effect size of ATAC for each allele relative to the WT is indicated on the right hand side. The HDR allele is highlighted in green. Del 1 and Del 2 (highlighted in blue and orange respectively) are small defined deletions around SNP of interest. Deletion alleles which have no effect change are indicated by red boxes, and recapitulate the CTCF motif (left).

Figure 4.

Combined GenIE-ATAC and GenIE shows the importance of an OCT4 binding site within the FGFR super-enhancer for both chromatin accessibility and transcription of the FGFR gene. (A) Schematic of the FGFR super-enhancer showing the ATAC peak in hiPSCs (ATAC-seq data, average of three repeats (19)), the region of DNA that gave luciferase activity in reporter assay (26), and the POU5F1:SOX2 binding site along with CRISPR edits that were introduced to abolish OCT4 binding (either T-G HDR_1 or ATG-TGC HDR_3). Pie chart shows editing rates. (B) Graphs showing effect size of HDR or deletions in GenIE-ATAC (left) or GenIE (right) assay. Data processed using rgenie. All error bars represent 95% confidence intervals, n = 9.

Figure 5.

GenIE-ATAC across a CTCF binding site highlights SNPs important for binding. (A) Schematic of the CTCF binding site around rs12269414 and mutated bases. Position 3 is the N within the NGG PAM sequence of the CRISPR and therefore could not be mutated due to recutting and subsequent indel formation. (B) Top panel shows the effect on chromatin accessibility of the HDR allele normalised to the WT allele with mutated bases indicated at the bottom. Bottom panel shows editing rates. Data was processed using rgenie. All error bars represent 95% confidence intervals, and P values are as indicated. (C) Sequence logo showing the consensus binding site as determined by GenIE-ATAC, with letter size proportional to the square of the relative GenIE-ATAC effect size when mutated to that nucleotide (relative to the WT/consensus, set to 1). (D) Scatter plot showing correlation between GenIE-ATAC motif proportion size and Jasper CTCF motif proportion size. Points represent the GenIE effect size for each nucleotide divided by the sum over all four nucleotides at the same position; error bars are the 95% confidence intervals for a given nucleotide divided by the sum of these over all 4 nucleotides at the same position.