New Genetic Loci Implicated in Cardiac Morphology and Function Using Three-Dimensional Population Phenotyping

Abstract

Background: Cardiac remodeling occurs in the mature heart and is a cascade of adaptations in response to stress, which are primed in early life. A key question remains as to the processes that regulate the geometry and motion of the heart and how it adapts to stress.

Methods: We performed spatially resolved phenotyping using machine learning-based analysis of cardiac magnetic resonance imaging in 47 549 UK Biobank participants. We analyzed 16 left ventricular spatial phenotypes, including regional myocardial wall thickness and systolic strain in both circumferential and radial directions. In up to 40 058 participants, genetic associations across the allele frequency spectrum were assessed using genome-wide association studies with imputed genotype participants, and exome-wide association studies and gene-based burden tests using whole-exome sequencing data. We integrated transcriptomic data from the GTEx project and used pathway enrichment analyses to further interpret the biological relevance of identified loci. To investigate causal relationships, we conducted Mendelian randomization analyses to evaluate the effects of blood pressure on regional cardiac traits and the effects of these traits on cardiomyopathy risk.

Results: We found 42 loci associated with cardiac structure and contractility, many of which reveal patterns of spatial organization in the heart. Whole-exome sequencing revealed 3 additional variants not captured by the genome-wide association study, including a missense variant in CSRP3 (minor allele frequency 0.5%). The majority of newly discovered loci are found in cardiomyopathy-associated genes, suggesting that they regulate spatially distinct patterns of remodeling in the left ventricle in an adult population. Our causal analysis also found regional modulation of blood pressure on cardiac wall thickness and strain.

Conclusions: These findings provide a comprehensive description of the pathways that orchestrate heart development and cardiac remodeling. These data highlight the role that cardiomyopathy-associated genes have on the regulation of spatial adaptations in those without known disease.

Keywords: cardiomyopathies; genome-wide association study; heart ventricles; hypertrophy; ventricular dysfunction.

Figure 1.

Cardiac image analysis. Cardiac magnetic resonance (CMR) cine imaging was performed in short and long axis planes. A fully convolutional neural network was used to segment the left ventricular myocardium and determine regional wall thickness. Motion tracking was performed using image registration to map myocardial deformation between frames and calculate regional Eulerian strain. The left ventricle was divided into 17 segments using the American Heart Association model. Short-axis (A), 4-chamber long-axis (B), and 2-chamber long-axis (C) CMR imaging with corresponding segments. Segments are grouped by anatomic location and numbered sequentially allowing characteristics to be represented on bullseye (D) and 3-dimensional (E) models of the left ventricle (right ventricle shown in outline).

Figure 2.

Correlations between spatial left ventricle (LV) traits and association with known predictors. A through C, Distribution of regional traits in up to 47 549 participants by sex is shown with violin plots. The Scott rule was used to determine the smoothing bandwidth for the kernel estimation. Inner dotted lines show the data quartiles. D through F, Pearson correlation among the 16 regional wall thickness (WT), strain^circ, and strain^rad before (lower diagonal) and after (upper diagonal) adjustment by the indicated confounding factors. The correlation coefficients are shown as heatmaps. The color indicates a correlation efficiency from −1 to 1 (blue to red). The basal, mid-cavity, and apical segments are separated by light blue lines. G, Distribution of regional trait Pearson correlations before and after adjustment by the indicated confounding factors. The boxes show the quartiles of the data set, and all values are overlaid as dots.

Figure 3.

Genetic correlation (r_g) between spatial left ventricular traits, blood pressure (BP), and cardiomyopathies. The r_g between spatial traits and both systolic BP (SBP; A) and diastolic BP (DBP; B) were calculated before (blue) and after (orange) adjustment for BP at the imaging visit. A genome-wide association study (GWAS) of spatial left ventricular traits was performed on up to 40 058 UK Biobank participants of white British ancestry. Summary statistics of SBP and DBP came from a GWAS on up to 801 644 individuals. Two-sided P values were estimated using linkage disequilibrium score regression. Significant correlations (P<0.05) are shown in darker colors. Center values are the estimated r_g, and error bars indicate SE. C, r_g between BP-adjusted spatial traits and hypertrophic cardiomyopathy (HCM; blue) and dilated cardiomyopathy (DCM; red). Summary statistics of HCM came from GWAS of 1733 cases and 6628 controls, and DCM from GWAS of 5521 cases and 397 323 controls. AHA indicates American Heart Association; and WT, wall thickness.

Figure 4.

Genome-wide association study (GWAS) of spatially resolved wall thickness and contractility. A, Manhattan plots for spatially resolved analysis on 3 left ventricular traits, namely, wall thickness, strain^circ, and strain^rad. The plots show variant-based 2-sided minimum P values from 16 GWAS of left ventricular American Heart Association (AHA) segments for the indicated traits for up to 40 186 Europeans in the UK Biobank. In the merged spatial analysis, each locus was defined at above multiple hypothesis–adjusted genome-wide significance threshold of 3.125×10⁻⁹ (green dotted line) and labeled sequentially by chromosome location across all loci. The conventional genome-wide significance at 5×10⁻⁸ is shown as a gray dotted line. Spatial-only loci are labeled in bold green, and those that were found significant in global traits by conventional genome-wide significance were labeled orange. Loci 8, 28, 29, and 39 contain >1 independent lead SNP (LD r² < 0.1). B, Locus look up in cardiomyopathy genes. The loci not significant in the corresponding globally averaged trait are highlighted with a gray background. Significance is defined as having at least 1 SNP in the corresponding locus window that reached conventional genome-wide significance (dark pink) or 5% FDR (light pink) for the indicated cardiac disease. Locus naming was performed primarily by gene prioritization considering FUMA and prior gene association with Mendelian hypertrophic cardiomyopathy (HCM) or dilated cardiomyopathy (DCM). See Table S3 for a list of locus genes. Figures S7 through S9 show β values of lead SNPs in the spatial GWAS loci by individual segments. FDR indicates false discovery rate; FUMA, functional mapping and annotation of genome-wide association studies; LD, linkage disequilibrium; and SNP, single nucleotide polymorphism.

Figure 5.

Predicted regulatory effects of genome-wide association study (GWAS) variants on expression. Regulatory effects were calculated using GWAS summary statistics and the GTEx, version 8, expression quantitative trait loci (eQTL) MASH-R model for the heart left ventricle; 10 498 genes were tested. The Manhattan plot shows the minimum P value for segmental wall thickness, with chromosomes colored by alternate dark and light blue. Bullseye plots were colored with red (positive effect of predicted gene expression on wall thickness) and blue (negative effect of predicted gene expression on wall thickness). Similarly, we performed eQTL based transcriptiome wide association studies for strain^circ and strain^rad and the results are shown in Figure S10.

Figure 6.

Exome-wide variant-level association study (ExWAS) on spatially resolved wall thickness and contractility. A, Manhattan plot of ExWAS for spatial traits. The smallest P values across 16 segments were plotted for wall thickness, strain^circ, and strain^rad. Locus definitions were taken from the genome-wide association study (GWAS) analysis. Three exome variants that reached significance but were not mapped to the 42 GWAS loci are labeled with the gene names. The loci are colored by the most severe consequence of the variants. Spatial-only loci from the GWAS analysis are circled in green. B-D, Forest plots to show the association β and P values for selected variants on spatial wall thickness, global mean wall thickness (WT G), maximum wall thickness (WT Max), and left ventricular mass (LVM). Forest plots are colored by log-scaled P values, and significant ones are labeled with asterisks. The insets are bullseye plots of β values. The direction of β values is specific to the effect alleles. The variants plotted are (in human genome guild 38 [GRCh38]): chr6:118566140:T:C (CEP85L), chr14:23392602:A:G (MYH6), and chr17:39657827:G:A (STARD3), from left to right.

Figure 7.

Mendelian randomization (MR) analysis of blood pressure on spatial traits and spatial traits on the risk of hypertrophic cardiomyopathy (HCM) and dilated cardiomyopathy (DCM). Odds ratios (ORs) represented are those inferred from the inverse variance–weighted (IVW) 2-sample MR per SD increase. The error bars represent the 95% CI of the OR. OR for circumferential strain reflects those of increased contractility. Asterisk indicates that the MR IVW test P value is below multiple hypothesis–adjusted threshold (P<0.05/16). A, MR results on increased systolic blood pressure (SBP)/diastolic blood pressure (DBP) on risk of increased spatial traits, including the global mean (G) and regional mean on American Heart Association (AHA) segments (1–16) for wall thickness, strain^circ, and strain^rad. Spatial traits used a genome-wide association study (GWAS) on rank-based inverse transformed (rIVT) adjusted with sex, age, body mass index (BMI), and body surface area (BSA) at magnetic resonance imaging (MRI), in 40 058 participants of the UK Biobank without cardiomyopathy and with available cardiac magnetic resonance (CMR) imaging. Genetic instruments for SBP and DBP were selected from a published GWAS including up to 801 644 individuals. B, MR results on increased global and spatial left ventricular (LV) wall thickness, and strain in circumferential and radial directions on risk of HCM and DCM. Genetic instruments for spatial traits were selected from the present GWAS, and the number of single nucleotide polymorphisms involved in performing MR to HCM is listed on the right side of each figure. The outcome HCM GWAS included 5927 cases vs 68 359 controls; DCM included 14 255 cases vs 1 199 156 controls.

Figure 8.

Cluster groups and gene interactions for left ventricular wall thickness (WT). A, Genes in the global mean wall thickness genome-wide association study (GWAS) loci were clustered by the STRING-DB. Enrichment analysis on the largest cluster is shown in C. Edges show protein-protein relationships where a protein works together to perform a common task. B, Genes in the spatial-only loci of regional wall thickness were clustered by the STRING-DB (Search Tool for the Retrieval of Interacting Genes/Proteins). Enrichment analysis on the largest cluster was shown in D. Light blue line: known interactions from curated data sets; purple line: experimentally determined known interactions; dark blue line: predicted interactions from gene neighborhood; red line: predicted interactions from gene fusions; dark blue line: predicted interactions from gene co-occurrences; light green line: predicted interaction from text mining; black line: co-expression; and light purple: gene homology.