High-Resolution Regulatory Maps Connect Vascular Risk Variants to Disease-Related Pathways

Abstract

Background: Genetic variant landscape of coronary artery disease is dominated by noncoding variants among which many occur within putative enhancers regulating the expression levels of relevant genes. It is crucial to assign the genetic variants to their correct genes both to gain insights into perturbed functions and better assess the risk of disease.

Methods: In this study, we generated high-resolution genomic interaction maps (≈750 bases) in aortic endothelial, smooth muscle cells and THP-1 (human leukemia monocytic cell line) macrophages stimulated with lipopolysaccharide using Hi-C coupled with sequence capture targeting 25 429 features, including variants associated with coronary artery disease. We also sequenced their transcriptomes and mapped putative enhancers using chromatin immunoprecipitation with an antibody against H3K27Ac.

Results: The regions interacting with promoters showed strong enrichment for enhancer elements and validated several previously known interactions and enhancers. We detected interactions for 727 risk variants obtained by genome-wide association studies and identified novel, as well as established genes and functions associated with cardiovascular diseases. We were able to assign potential target genes for additional 398 genome-wide association studies variants using haplotype information, thereby identifying additional relevant genes and functions. Importantly, we discovered that a subset of risk variants interact with multiple promoters and their expression levels were strongly correlated.

Conclusions: In summary, we present a catalog of candidate genes regulated by coronary artery disease-related variants and think that it will be an invaluable resource to further the investigation of cardiovascular pathologies and disease.

Keywords: coronary artery disease; gene; genomics; haplotype; inflammation.

Figure 1.

Promoter interacting regions were enriched for regulatory elements. A, Three types of Hi-C with sequence capture-established interactions; promotor (P)-distal (D), P-P and genome-wide association studies (GWAS)-P (G-P), respectively; for the Gene A promoter. In P-D is analyzed probed promoters’ interaction with D element (DEs); the latter separated by restriction sites. In P-P and G-P both ends of the interaction are probed. B, The largest connected subgraph in chr9 in aortic endothelial cell (AEC). There are 22 promoters, 3 GWAS SNPs and 271 DE (only those overlapping with H3K27ac marks are included). The blue track represents gene expression levels, gray boxes represent transcripts, and innermost layers represent H3K27Ac marks (2 replicates). Purple arcs represent G-P or P-P, and orange arcs represent P-D interactions respectively. C, Principal component analysis of P-D interaction of genes not expressed in none of the cell types D, Overlap enrichment relative to a segment-length and distance-from-P controlled random set. The AEC P-D and G-P datasets were overlapped with general cardiovascular, HUVEC, and HAEC transcription factor marker data from ChipAtlas. E, Three hundred and fifty-seven kilobase region containing MTAP-rs944797 interaction and (F) 393 kb region containing BMP6-rs9328448 interaction visualized using Gviz package. Overlap with H3K27Ac marks are shown in the lower panes as well as the signal from the input chromatin. P-D (including P-G) and P-P interactions are colored as green and purple respectively. ASMC indicates aortic smooth muscle cells; BMP6, Bone Morphogenetic Protein 6; HAEC, human aortic endothelial cell line; HUVEC, Human umbilical vein endothelial cells; LPS, lipopolysaccharides; MTAP, Methylthioadenosine Phosphorylase; mTHP, macrophage–THP-1; and SNP, single nucleotide polymorphism.

Figure 2.

HiCap can inform on regulatory potential of variants in LD with risk variants. A, The aortic endothelial cell (AEC) Prom-Dist (PD) set was searched for interactions between the promoter and distal elements close to a variant interacting with the same promoter (green curve). The comparison was made relative to a site at the same distance from the promoter it but located at the other side of it (light blue curve). The latter is nearly a horizontal line as expected, whereas the blue curve strongly deviates from that at distances not too far from the variant site. The bin size used to count interactions is 5 kb. B, Variants in the AEC, aortic smooth muscle cell (ASMC) and macrophage–THP-1 (mTHP-1)-lipopolysaccharide (LPS) merged genome-wide association studies (GWAS)-Prom (GP) set and their interaction preferences with genes at distance 0 (no gene-jumping), 1 (nearest gene is jumped over), etc. Distal regions in the corresponding PD set are shown for reference. The GP interaction set is further split in equal-sized halves depending on the variants’ distance to its nearest gene (GP without and with a nearby gene, respectively). C, The coronary artery disease (CAD)–related variant rs12239436 (red box) interacts with 62 kb distant gene USP24 (gray box) in AEC (dotted line), ASMC (dashed,) and mTHP-1–LPS (solid). Strengths of interactions are represented with the P value recorded and indicated by arrow thickness. Our result set further include strong interactions with the 1.3 Mb distant previously CAD associated gene PLPP3 (also known as PPAP2B). Less significant interactions with nearby FYB2 (ASMC) and PRKAA2 (mTHP-1–LPS) are potentially bystanders. D, Earlier reported interaction between variant rs934937 and the EDN1 gene is, to a varying degree, present in both AEC patients. Not so in less relevant tissues ASMC and mTHP-1–LPS. E, Comparison of overlaps between PD dataset (AEC) vs 100 matched random datasets and CVD_GWAS and matched SNP datasets (see methods) shows that there is an enrichment for variants in linkage disequilibrium (LD) with CVD_GWAS found in PD dataset when the genomic distance between SNP and its LD proxy is ≤80 kb. No such enrichment was seen for random PD datasets vs real or random SNP sets. chr indicates chromosome; CVD, cardiovascular disease; SMC, smooth muscle cells; SNP, single nucleotide polymorphism; and SP, supporting pair.

Figure 3.

Assigned target genes of CVD variants were enriched for pathways relevant for vascular pathologies. A, Expression quantitative trait loci contained in the merged CVD genome-wide association studies (GWAS)-Promotor (prom), Prom- distal (Dist) and Prom-Dist_ linkage disequilibrium (LD) datasets plotted vs a size and distance-corrected random set. Deviation from diagonal is present among approximately 30% of the data. B, GWAS traits that are overrepresented in the interaction datasets. Only traits containing at least 14 variants were taken forward. Fold enrichment is calculated by dividing the actual number of trait variants in the interaction dataset to that of expected (fraction of trait variants in the full trait set). The bar labels denote the fraction of variants found in the interaction datasets. C, Gene ontology (GO) term enrichment analysis of genes interacting with variants or those in LD with CVD_GWAS set using TopGO package. GO terms enriched using only nearest genes to the variants are not reported. Terms with >5 genes and enrichment score >0.05 were not included. ASAP_H indicates The Advanced Study of Aortic Pathology, heart tissue; HiCap, Hi-C with sequence capture; MI, myocardial infarction; PR interval, the period, measured in milliseconds, that extends from the beginning of the P wave (the onset of atrial depolarization) until the beginning of the QRS complex; and Pri-miRNA, primary transcript of micro RNA.

Figure 4.

Expression levels of promoters interacting with the same variant were correlated. Expression correlation between (A) genes MTAP and IFT74 (P value 4.2×10−16) and (B) genes SMARCAD1 and BMP2K (P value 2.1×10−16) using aortic intima-media expression from 131 individuals. The values on axes are RPKM values, and each dot corresponds to each individual where both expression information are obtained from. C and D, Circos plot representation of interactions between rs16998073 and rest of the genome in C aortic endothelial cell (AEC) and D aortic smooth muscle cell for comparison. The plot spans chr4:73000000-103000000. The blue track represents gene expression levels, gray boxes represent transcripts, and innermost layers represent H3K27Ac marks in AEC cells (2 replicates). Purple arcs represent genome-wide association studies-promotor (Prom) or Prom-Prom, and orange arcs represent Prom-Distal interactions. BMP2K indicates BMP2 Inducible Kinase; BMP6, Bone Morphogenetic Protein; IFT74, Intraflagellar Transport 74; 6 MTAP, Methylthioadenosine Phosphorylase; RPKM, read counts per kilobase million; and SMARCAD1, SWI/SNF-Related, Matrix-Associated Actin-Dependent Regulator Of Chromatin, Subfamily A, Containing DEAD/H Box 1.