Chromatin accessibility profiling uncovers genetic- and T2D disease state-associated changes in cis-regulatory element use in human islets

Abstract

Genetic and environmental factors both contribute to islet dysfunction and failure, resulting in type 2 diabetes (T2D). The islet epigenome integrates these cues and can be remodeled by genetic and environmental variation. However, our knowledge of how genetic variants and T2D disease state alter human islet chromatin landscape and cis-regulatory element (RE) use is lacking. To fill this gap, we profiled and analyzed human islet chromatin accessibility maps from 19 genotyped individuals (5 with T2D) using ATAC-seq technology. Chromatin accessibility quantitative trait locus (caQTL) analyses identified 3001 sequence variants (FDR<10%) altering putative cis-RE use/activity. Islet caQTL were significantly and specifically enriched in islet stretch enhancers and islet-specific transcription factor binding motifs, such as FOXA2, NKX6.1, RFX5/6 and PDX1. Importantly, these analyses identified putative functional single nucleotide variants (SNVs) in 13 T2D-associated GWAS loci, including those previously associated with altered ZMIZ1, MTNR1B, RNF6, and ADCY5 islet expression, and linked the risk alleles to increased (n=8) or decreased (n=5) islet chromatin accessibility. Luciferase reporter assays confirmed allelic differences in cis-RE activity for 5/9 caQTL sequences tested, including a T2D-associated SNV in the IL20RA locus. Comparison of T2D and non-diabetic islets revealed 1882 open chromatin sites exhibiting T2D-associated chromatin accessibility changes (FDR<10%). Together, this study provides new insights into genetic variant and T2D disease state effects on islet cis-RE use and serves as an important resource to identify putative functional variants in T2D-and islet dysfunction-associated GWAS loci and link their risk allele to in vivo loss or gain of chromatin accessibility.

Single cell chromatin accessibility reveals pancreatic islet cell type- and state-specific regulatory programs of diabetes risk

“…In each case, reporter assays showed significant allelic effects on enhancer activity that were directionally consistent with predictions (Figure 4c). We also compared predictions to chromatin accessibility quantitative trait loci (caQTLs) previously identified in ensemble islet samples 72 . We observed highly significant enrichment of caQTLs among variants with predicted effects on alpha or beta cells (obs.=38.8%, exp.=23.6%, two-sided Fisher’s exact P=1.64×10 −66 ) (Figure 4d).…”

“…We obtained raw sequence data of ATAC-seq for 42 bulk islet samples from four prior studies 14,27,28,72 and 4 bulk pancreas samples from ENCODE. We re-processed all samples with a uniform pipeline: we aligned all reads to hg19 with bwa mem, identified and removed duplicate reads with picard MarkDuplicates, and called peaks with MACS2 (v.2.1.2) with the parameters ‘—shift −100 –extsize 200 –keep-dup all’.…”

“…We determined significance using a two-sided binomial test assuming an expected fraction of 50%. We assessed enrichment of predicted effect variants in alpha or beta cell states for islet caQTLs 72 compared to any islet caQTL in alpha or beta cell sites using two-sided Fisher’s exact tests. We then stratified variants with predicted effects by category (hormone-high, hormone-low, alpha, or beta) and assessed enrichment of caQTLs with predicted effects within each category with two-sided Fisher’s exact tests.…”

BiFET: A Bias-free Transcription Factor Footprint Enrichment Test

“…ATAC-seq peaks from five human PBMCs (12) and five human islets (14,16) were called using MACS version 2.1.0 with parameters “-nomodel-f BAMPE”. The peaks from all ten samples were merged to generate one consensus peak set (N = 57,108 peaks) by using R package DiffBind_2.2.5.…”

“…Due to advances in genomewide chromatin accessibility profiling, notably the ATAC-seq (9) technology, increasing numbers of chromatin accessibility maps have been generated in primary human cells to study complex diseases, including cancer (10), systemic lupus erythematosus (11), immunosenescence (12,13), and type 2 diabetes (14–16). Effective detection and analyses of TF footprints from these data will be instrumental to nominate potential regulators associated with a clinical phenotype of interest (e.g., immunosenescence (12) or cancer subtypes (17)).…”