Effects of Coronary Artery Disease-Associated Variants on Vascular Smooth Muscle Cells

Abstract

Background: Genome-wide association studies have identified many genetic loci that are robustly associated with coronary artery disease (CAD). However, the underlying biological mechanisms are still unknown for most of these loci, hindering the progress to medical translation. Evidence suggests that the genetic influence on CAD susceptibility may act partly through vascular smooth muscle cells (VSMCs).

Methods: We undertook genotyping, RNA sequencing, and cell behavior assays on a large bank of VSMCs (n>1499). Expression quantitative trait locus and splicing quantitative trait locus analyses were performed to identify genes with an expression that was influenced by CAD-associated variants. To identify candidate causal genes for CAD, we ascertained colocalizations of VSMC expression quantitative trait locus signals with CAD association signals by performing causal variants identification in associated regions analysis and the summary data-based mendelian randomization test. Druggability analysis was then performed on the candidate causal genes. CAD risk variants were tested for associations with VSMC proliferation, migration, and apoptosis. Collective effects of multiple CAD-associated variants on VSMC behavior were estimated by polygenic scores.

Results: Approximately 60% of the known CAD-associated variants showed statistically significant expression quantitative trait locus or splicing quantitative trait locus effects in VSMCs. Colocalization analyses identified 84 genes with expression quantitative trait locus signals that significantly colocalized with CAD association signals, identifying them as candidate causal genes. Druggability analysis indicated that 38 of the candidate causal genes were druggable, and 13 had evidence of drug-gene interactions. Of the CAD-associated variants tested, 139 showed suggestive associations with VSMC proliferation, migration, or apoptosis. A polygenic score model explained up to 5.94% of variation in several VSMC behavior parameters, consistent with polygenic influences on VSMC behavior.

Conclusions: This comprehensive analysis shows that a large percentage of CAD loci can modulate gene expression in VSMCs and influence VSMC behavior. Several candidate causal genes identified are likely to be druggable and thus represent potential therapeutic targets.

Keywords: coronary artery disease; genetics; muscle, smooth, vascular; transcriptomes.

Figure 1.

Flowchart of the study design. CAD indicates coronary artery disease; eCAVIAR, expression quantitative trait locus and genome-wide association study causal variant identification in associated regions; eQTL, expression quantitative trait locus; SMR, summary-data based mendelian randomization; sQTL, splicing quantitative trait locus; and VSMC, vascular smooth muscle cell.

Figure 2.

Gene expression profiles of VSMCs in this study compared with those of human coronary artery SMCs and other types of cell/tissue. Multidimensional scaling (MDS) analysis comparing the transcriptomes of vascular smooth muscle cell (VSMC) samples of this study with reported transcriptomic data from human coronary artery (CA) smooth muscle cells (SMCs) and transcriptomic data of other types of cell/tissue (data sources described in Table S2). B indicates data from bulk RNA sequencing (RNA-Seq); EC, endothelial cells; FB, fibroblasts; Mono, monocytes; MΦ, macrophages; and S, data from single cell RNA-Seq.

Figure 3.

Colocalization between CAD GWAS and VSMC eQTL signals. A, Results of expression quantitative trait locus (eQTL) and genome-wide association study (GWAS) causal variant identification in associated regions (eCAVIAR) analysis of colocalization between coronary artery disease (CAD) GWAS and vascular smooth muscle cell (VSMC) eQTL signals. Blue line indicates the colocalization posterior probability (CLPP) of 0.05. Each dot represents a genetic variant, and those above the blue line (CLPP>0.05) represent variants with significant colocalization. B, Results of summary data–based mendelian randomization/heterogeneity in dependent instruments (SMR/HEIDI) analysis. Blue line indicates P_HEIDI=0.05. Each dot represents a genetic variant, and those above the blue line represent variants with significant colocalization (P_SMR<0.05 and P_HEIDI>0.05). C, Colocalization of TCF21 eQTL signal in VSMCs with CAD GWAS signal at the TCF21 locus. D, Colocalization of SMAD3 eQTL signal in VSMCs with CAD GWAS signal at the SMAD3 locus. E, Colocalization of FES eQTL signal in VSMCs with CAD GWAS signal at the FES locus. F, Colocalization of MIA3 eQTL signal in VSMCs with CAD GWAS signal at the MIA3 locus. G, Colocalization of TGFB1 eQTL signal in VSMCs with CAD GWAS signal at the TGFB1 locus. H, Colocalization of REST eQTL signal in VSMCs with CAD GWAS signal at the REST locus.

Figure 4.

Chromosomal locations (A) and functional pathways (B) of candidate causal genes. A, Boxes designate different chromosomes that are numbered from 1 to 22, with cytogenetic bands shown inside each box and chromosomal positions labeled by tick marks placed every 25 million bp. Lines on the outer aspect of the boxes symbolize coronary artery disease–associated variants that had expression quantitative trait locus (eQTL) or splicing quantitative trait locus (sQTL) effects in vascular smooth muscle cells. Each locus with both cis-eQTL and cis-sQTL effects is indicated by a red line, and each locus with only cis-eQTL is indicated by a green line. B, Dot plot showing identified functional pathways with enrichment of some of the candidate causal genes. Data shown are from a pathway analysis performed with ToppGene. FDR indicates false discovery rate. *Originally identified in cancer. **Originally identified in osteogenesis imperfecta.

Figure 5.

Druggability of genes with eQTL signals in VSMCs that significantly colocalized with CAD GWAS signals. Shown is a summary of results of an interrogation in the Drug Gene Interaction Database of candidate causal genes with expression quantitative trait locus (eQTL) signals in vascular smooth muscle cells (VSMCs) that showed significant colocalization with coronary artery disease (CAD) genome-wide association study (GWAS) signals in eQTL and GWAS causal variant identification in associated regions (eCAVIAR) or summary data–based mendelian randomization (SMR) analyses. Genes with evidence of drug-gene interactions are highlighted in bold. Details of druggability and drug-gene interactions are described in Tables S10 and S11, respectively.

Figure 6.

Coexpressed gene modules in relation to VSMC proliferation and functional pathways. A, Heat map representation of correlations of co-expression gene modules with parameters of vascular smooth muscle cell (VSMC) proliferation. Values shown are correlation coefficients and P values (in brackets). The prefix ME in each module name stands for module eigengene. B, Scatterplot of MEgreen module membership (x axis) vs coefficient of correlation between individual gene expression level and percentage of VSMCs positive for EdU staining at low intensity (between 2000 and 10 000 arbitrary units; EdU-low; y axis). Each gene is indicated by an open dot in green. C, Effect of the MEgreen module hub gene YIPF6 on VSMC proliferation. Data shown are mean (±SEM) values of EdU-low in VSMCs transfected with either YIPF6 siRNA or negative control siRNA relative to the average value of nontransfected VSMCs, control siRNA-transfected VSMCs, and YIPF6 siRNA-transfected VSMCs (n=8; VSMCs from 8 different individuals; P value is from a 2-tailed Mann-Whitney test). YIPF6 knockdown in VSMCs transfected cells is shown in Figure S5. D, Biological pathways enriched in the MEgreen module. EdU-high indicates percentage of VSMCs positive for EdU staining at high intensity (>10 000 arbitrary units); EdU-total, percentage of VSMCs positive for EdU staining; and FDR, false discovery rate.

Figure 7.

Coexpressed gene modules in relation to VSMC apoptosis and functional pathways. A, Heat map representation of correlations of coexpression gene modules with parameters of vascular smooth muscle cell (VSMC) apoptosis. Values shown are correlation coefficients and P values (in brackets). The prefix ME in each module name stands for module eigengene. B, Scatterplot of MEturquoise module membership (x axis) vs coefficient of correction between individual gene expression level and %D@4h (y axis). Each gene is indicated by an open dot in turquoise. C, Effect of the MEturquoise module hub gene SLC25A36 on VSMC apoptosis. Data shown are mean (±SEM) values of the percentage of dead cells (propidium iodide positive) at 4 hours after staurosporine treatment (%D@4h) of VSMCs transfected with either SLC25A36 siRNA or negative control siRNA relative to nontransfected VSMCs (n=5; VSMCs from 5 different individuals; P value is from a 2-tailed Mann-Whitney test). SLC25A36 knockdown in VSMCs transfected cells is shown in Figure S6. D, Biological pathways enriched in the MEturquoise module. ECM indicates extracellular matrix; EGF, epidermal growth factor; FDR, false discovery rate; FSGS, focal segmental glomerulosclerosis; NA@30m and NA@60m, change in nuclear area at 30 and 60 minutes after staurosporine treatment; NF@30m and NF@60m, change in nuclear fragmentation index at 30 and 60 minutes after treatment with the apoptosis inducer staurosporine; %D@8h, percentage of dead cells (propidium iodide positive) at 8 hours after staurosporine treatment; and TT50D, time in minutes for 50% of cells to become propidium iodide positive after staurosporine treatment.