Functional characterization of T2D-associated SNP effects on baseline and ER stress-responsive β cell transcriptional activation

Abstract

Genome-wide association studies (GWAS) have linked single nucleotide polymorphisms (SNPs) at >250 loci in the human genome to type 2 diabetes (T2D) risk. For each locus, identifying the functional variant(s) among multiple SNPs in high linkage disequilibrium is critical to understand molecular mechanisms underlying T2D genetic risk. Using massively parallel reporter assays (MPRA), we test the cis-regulatory effects of SNPs associated with T2D and altered in vivo islet chromatin accessibility in MIN6 β cells under steady state and pathophysiologic endoplasmic reticulum (ER) stress conditions. We identify 1,982/6,621 (29.9%) SNP-containing elements that activate transcription in MIN6 and 879 SNP alleles that modulate MPRA activity. Multiple T2D-associated SNPs alter the activity of short interspersed nuclear element (SINE)-containing elements that are strongly induced by ER stress. We identify 220 functional variants at 104 T2D association signals, narrowing 54 signals to a single candidate SNP. Together, this study identifies elements driving β cell steady state and ER stress-responsive transcriptional activation, nominates causal T2D SNPs, and uncovers potential roles for repetitive elements in β cell transcriptional stress response and T2D genetics.

Modeling islet enhancers using deep learning identifies candidate causal variants at loci associated with T2D and glycemic traits

“…S7), as would be expected if these SNPs perturbed the binding of islet regulatory factors. Although no individual SNP exhibited allelic imbalance after multiple hypothesis correction (likely due to the small number of ATAC-seq samples), several of these SNPs have been shown to exhibit allelic activity in reporter assays conducted in MIN6 beta cells: rs7732130 at the 5:76435004 ( ZBED3/PDE8B ; GRCh37 coordinates) T2D signal (34), rs7933438 at the 11:128040810 ( ETS1 ) T2D signal (35), and rs4237150 at the 9:4290085 ( GLIS3 ) T2D signal (36). We calculated the overlap of the 94 candidate causal SNPs with predicted TFBSs (based on motifs; Table S3; Methods) and found these SNPs were enriched (FDR<5%) in RFX family binding sites, consistent with previous studies that report RFX6 as an important islet transcription factor for T2D genetic risk (21, 37).…”

“…For 101 signals (Table S2), we were able to nominate a single candidate causal variant in the 99% credible set from statistical fine-mapping (94 unique SNPs). These predictions include multiple SNPs for which functional allelic effects have been detected previously in vitro in islet beta cells (34)(35)(36) and one for which we provide biochemical validation. We anticipate that further validation experiments of candidate causal variants nominated in this study will lead to additional insights into the molecular genetic basis of T2D and T2D-related traits.…”

CoRE-ATAC: A deep learning model for the functional classification of regulatory elements from single cell and bulk ATAC-seq data

“…Regulatory activities of certain open chromatin regions are genetically modulated, which can be detected via chromatin accessibility quantitative trait loci (caQTL) analyses. Previously, we identified caQTLs from human islet samples (n = 19) [ 7 ] for which we generated an MPRA library to test and compare the regulatory activity of reference and alternative alleles forcaQTLs and other variants (n = 4293 SNPs) [ 42 ]. These data gave us the opportunity to test whether CoRE-ATAC predictions can detect genetically driven differences in the regulatory activity.…”

“…Massively Parallel Reporter Assay (MPRA) data were generated in MIN6 pancreatic beta cell line to study the regulatory activity of variants associated with Type-2-Diabetes [ 42 ]. Briefly, 4293 variants within islet ATAC-seq peaks were profiled for regulatory element activity using MPRA.…”

“…We noted that a significant portion of STARR-seq enhancers (~33%)were predicted as promoters in our models and the majority of these regions were close to TSS (S14C and S14D Fig) . In contrast, CoRE-ATAC enhancers overlapping STARR-seq active regions were more Regulatory activities of certain open chromatin regions are genetically modulated, which can be detected viachromatin accessibility quantitative trait loci (caQTL) analyses. Previously, weidentifiedcaQTLsfrom human islet samples (n = 19) [7] for which we generated an MPRA library to test and compare the regulatory activity of reference and alternative alleles for-caQTLs and other variants (n = 4293 SNPs) [42]. These data gave us the opportunity to test whether CoRE-ATAC predictions can detect genetically driven differences in the regulatory activity.…”