The 9p21.3 Coronary Artery Disease Risk Locus Modulates Vascular Cell-State Transitions via Enhancer-Driven Regulation of MTAP

Abstract

The 9p21.3 locus is the strongest genetic association with coronary artery disease (CAD), yet its causal mechanisms remain unresolved. We map the regulatory architecture of 9p21.3 in disease-relevant vascular cells, identifying 12 enhancers within the CAD risk haplotype that respond dynamically to inflammatory and metabolic stress in fibroblasts and smooth muscle cells. These activated states are enriched for CAD heritability, implicating stress-responsive vascular wall cells in disease pathophysiology. Dense CRISPRi tiling integrated with fine-mapping and genomic constraint across >500,000 individuals nominates MTAP as the effector gene, with rs1537371 as a likely causal variant. Perturbation and multi-modal analyses show that MTAP loss induces pro-fibrotic and angiogenic programs and sensitizes vascular cells to TGF-β-driven pathological transitions. Our findings reveal a vascular-specific enhancer network through which noncoding variation at 9p21.3 modulates CAD risk via MTAP-a previously unrecognized regulator of vascular remodeling located 269 kb from the risk haplotype.

Figure 1.. The 9p21.3 coronary artery disease (CAD) locus maps to an enhancer dense region in CAD-relevant cell types.

(A) Manhattan plot showing Phenome-wide association (PheWAS) of 9p21.3 locus (tagged by rs4977574) to multiple disease traits in FinnGen data freeze 13 (DF13). Each triangle represents a phenotype association at 9p21.3 locus and different triangle colors represent various classification of phenotypes. Direction of the triangle indicates increased directionality of association. Increased risk is indicated by an upward triangle and reduced risk is indicated by a downward facing triangle. (B) Schematic depicting an extensive vascular cell wall profiling at baseline and in response to CAD-relevant stimulatory condition. (C) Chromatin state and epigenetic landscape of 9p21.3 CAD haplotype spanning ~ 50-kb DNA sequence within the 3’ region of long non-coding RNA CDKN2B-AS1 (ANRIL). Also depicted are all 50 CAD associated SNPs (r² ≥ 0.8), tagged by rs4977573 and credible set variants from FinnGen and UKBB finemapping efforts. (D-E) Dot plots showing enrichment for CAD heritability at baseline and in response to CAD-relevant stimulatory conditions, including Inflammatory (IL-1α and IL-6 & TNF-α) vascular fibroblast and smooth muscle cells. Solid dark lines around each dot represent significance at a threshold of −log₁₀(adjusted P-value) ≥ 3.35. (F-G) Dot plots showing the impact of CAD-relevant stimulatory conditions on enhancer activities (proxy of H3K27ac) at 9p21.3 locus in vascular fibroblast and smooth muscle cells. Fold changes are indicated as Red (increased fold change) or Blue (Reduced fold change) and solid dark lines around each dot represent significance at a threshold of −log₁₀(FDR) ≥ 1.3.

Figure 2.. The 9p21.3 locus contains multiple functional vascular cell enhancers regulating cis-expressing genes.

(A) Contact matrix (at 5000 bp resolution) from Micro-C¹ analysis of the H1 human embryonic stem cell (h1-ESC) spanning 1 Mb around 9p21.3 CAD haplotype (indicated as a blue rectangle). (B) Schematic showing experimental workflow to identify all functional non-coding 9p21.3 elements by CRISPR interference (CRISPRi) paired with Multiplex Analysis of Cells (MAC-seq) sequencing readout collectively referred to as CRISPRi-MAC-seq. (C) Donut plot showing the proportion of control, TSS-targeting, enhancer-targeting, and SNP-targeting guides (n=417 total guides) in our densely-tiled 9p21.3 library. (D) Box plots showing TSS repression efficiency via assessing normalized MTAP expression in vascular fibroblasts and smooth muscle cells treated with lentivirus containing transcriptional start site (TSS) guides for MTAP (n=4) compared to non-targeting control (n=13). Significance was determined using a two-tailed t-test, with a significance threshold of P-value<0.05. (E-F) Plot shows a sliding window approach that aggregates sgRNA transcriptional effects across a subset of the 9p21.3 annotated enhancers, including E4, E5, E8 and E9 in vascular fibroblast and smooth muscle cells. To assign significance to observed data relative to the permutation background, a permutation z-score was calculated. A significance level of z-score<−1.96 was used to identify significantly repressed regions – illustrated as red points on the sliding window plots.

Figure 3.. Convergent genomics evidence points to MTAP as the primary 9p21.3 CAD effector gene in vascular fibroblast and smooth muscle cells.

(A) Density plot of Gnocchi scores across the genome. Windows overlapping coding regions (red color) compared to windows overlapping 9p21.3 non-coding regions (blue color). (B) Distribution of CAD/T2D variants in constraint regions spanning ~120-kb region on the P-arm of chromosome 9. Overlaid on this plot is enhancer annotation for 9p21.3 locus along with constraint blocks and constraint variants flanking or overlapping enhancer region. (C) Dot plots showing fine-mapped variants associated with a combined trait for coronary revascularization, including coronary artery bypass grafting (CABG) and percutaneous coronary intervention (ANGIO) using data from large-scale population efforts from FinnGen and UKBB. Credible set variants are shaded grey and have the highest magnitude of association to CAD. (D-E) Using a MAGMA gene prioritization approach, we identified CAD-specific genes by calculating genome-wide ZSTAT scores for genes in comparison to a GWAS. Genes with ZSTAT scores exceeding the 0.95 percentile were considered significantly related to the GWAS trait. (D) First, we prioritized 9p21.3 genes against a CAD GWAS. The results indicated that MTAP, CDKN2A, and CDKN2B surpassed the significance cutoff. (E) To assess disease specificity, we then compared these results against a T2D GWAS. This analysis highlighted that MTAP’s association with CAD is disease-specific, as it did not meet the significance cutoff for T2D. (F) Converging evidence for CAD effector gene prioritization. The matrix denotes evidence to provide support (colored blue) for CAD driver genes across different modalities. Top scoring genes have the highest count across all evidence.

Figure 4.. MTAP regulates transcriptional programs that orchestrate vascular development and remodeling

(A) Volcano plots showing differentially expressed genes in vascular smooth muscle cells as a result of MTAP knockdown via siRNA. Genes with a P< 0.05 are considered significant. (B) Gene ontology (GO) analysis showing biological processes associated with DEGs (p-adjusted<0.05) as a result of MTAP suppression in vascular smooth muscle cells. (C) Network analysis of gene enrichment results to find relevant functional modules. Resulting clusters represent cellular processes overrepresented in statistically significant differentially expressed genes as a result of MTAP knockdown in vascular smooth muscle cells. Within each pathway in a cluster, we can visualize genes as upregulated (red) or downregulated (blue) based on their annotation to that pathway. The size of each pathway denotes its fold enrichment, indicating how well that pathway is enriched in our differentially expressed genes compared to the background. (D) Volcano plots showing differentially expressed genes in vascular fibroblasts as a result of MTAP knockdown via siRNA. Genes with a P< 0.05 are considered significant. (E) Gene ontology (GO) analysis showing biological processes associated with DEGs (p-adjusted<0.05) as a result of MTAP suppression in vascular fibroblasts. (F) Network analysis of gene enrichment results to find relevant functional modules. Resulting clusters represent cellular processes overrepresented in statistically significant differentially expressed genes as a result of MTAP knockdown in vascular fibroblasts. Within each pathway in a cluster, we can visualize genes as upregulated (red) or downregulated (blue) based on their annotation to that pathway. The size of each pathway denotes its fold enrichment, indicating how well that pathway is enriched in our differentially expressed genes compared to the background. (G) Alluvial plot showing total number of extracted LipocyteProfiler features from vascular fibroblast and smooth muscle cells by cellular compartments, stained organelle/structure, and feature attribute (object size, feature intersection, texture, and others). (H) Pearson correlation matrix of full extracted feature profiles from basal groups (siNTCtr and siMTAP) as well as in TGF-β treated groups (siNTCtr and siMTAP) in vascular fibroblast. (I) Pearson correlation matrix of full extracted feature profiles from basal groups (siNTCtr and siMTAP) as well as in TGF-β treated groups (siNTCtr and siMTAP) in vascular smooth muscle cells (J) Stacked bar graph showing counts of differentially changed LipocyteProfiler features at baseline (siNTCtr vs siMTAP) and in response to TGF-β (siNTCtr vs siMTAP) in vascular fibroblasts. All features with q-values ≤ 0.2 were considered significant. (K) Stacked bar graph showing counts of differentially changed LipocyteProfiler features at baseline (siNTCtr vs siMTAP) and in response to (TGF-β siNTCtr vs siMTAP) in vascular smooth muscle cells. No differentially changed features were observed at baseline in vascular smooth muscle cells. All features with q-values ≤ 0.2 were considered significant.

Figure 5.. MTAP functions as a critical regulator of pathological vascular cell transitions

(A) Representative images and box plots showing impact of siRNA mediated MTAP knockdown on alpha smooth muscle actin fiber areas (αSMA) at baseline and in response to TGF-β treatment in vascular fibroblasts. One-way ANOVA was to assess statistical significance followed by pairwise t-tests with Bonferroni correction. Scale bar (bottom right) = 50 μM. (B) Representative images and box plots showing impact of siRNA mediated MTAP knockdown on alpha smooth muscle actin fiber areas (αSMA) at baseline and in response to TGF-β treatment in vascular smooth muscle cells. One-way ANOVA was to assess statistical significance followed by pairwise t-tests with Bonferroni correction. Scale bar (bottom right) = 50 μM. (C) Representative images and box plots showing impact of siRNA mediated MTAP knockdown on Col1a1 levels (Col1a1) at baseline and in response to TGF-β treatment in vascular fibroblasts. One-way ANOVA was to assess statistical significance followed by pairwise t-tests with Bonferroni correction. Scale bar (bottom right) = 50 μM. (D) Representative images and box plots showing impact of siRNA mediated MTAP knockdown on Col1a1 levels (Col1a1) at baseline and in response to TGF-β treatment in vascular smooth muscle cells. One-way ANOVA was to assess statistical significance followed by pairwise t-tests with Bonferroni correction. Scale bar (bottom right) = 50 μM.