Genome-wide association study reveals novel genetic loci: a new polygenic risk score for mitral valve prolapse

Abstract

Aims: Mitral valve prolapse (MVP) is a common valvular heart disease with a prevalence of >2% in the general adult population. Despite this high incidence, there is a limited understanding of the molecular mechanism of this disease, and no medical therapy is available for this disease. We aimed to elucidate the genetic basis of MVP in order to better understand this complex disorder.

Methods and results: We performed a meta-analysis of six genome-wide association studies that included 4884 cases and 434 649 controls. We identified 14 loci associated with MVP in our primary analysis and 2 additional loci associated with a subset of the samples that additionally underwent mitral valve surgery. Integration of epigenetic, transcriptional, and proteomic data identified candidate MVP genes including LMCD1, SPTBN1, LTBP2, TGFB2, NMB, and ALPK3. We created a polygenic risk score (PRS) for MVP and showed an improved MVP risk prediction beyond age, sex, and clinical risk factors.

Conclusion: We identified 14 genetic loci that are associated with MVP. Multiple analyses identified candidate genes including two transforming growth factor-β signalling molecules and spectrin β. We present the first PRS for MVP that could eventually aid risk stratification of patients for MVP screening in a clinical setting. These findings advance our understanding of this common valvular heart disease and may reveal novel therapeutic targets for intervention.

Keywords: Genetic correlation; Genome-wide association study; Mitral valve prolapse; Polygenic risk score; Proteomics; RNA-sequencing.

Structured Graphical Abstract

This study meta-analysed 4884 mitral valve prolapse (MVP) cases versus 434 649 controls, and discovered 16 genetic loci associated to MVP. Downstream analyses implicated candidate genes involved in TGF-beta signalling, cardiomyopathy and the cytoskeleton. The results from the meta-analysis were used to calculate a polygenic risk score (PRS) to aid prediction of MVP. Adding the PRS to a model with age, sex and clinical risk factors improved MVP risk prediction. Abbreviations, MVP, mitral valve prolapse, p, p-value, PRS, polygenic risk score, RF, risk factors.

Figure 1

Overview of the study sample and performed analyses. The upper part of the figure shows the six included biobanks, registries, and studies that provided summary level results for primary and secondary meta-analyses and their respective case counts for mitral valve prolapse. The middle section shows the two main analyses of this work with total number of cases and controls that were included, and the total number of loci that were found in each. The primary meta-analysis was based on the summary statistics for the mitral valve prolapse phenotype. The secondary sensitivity analysis was based on the subset of patients who underwent mitral valve surgery in addition to their mitral valve prolapse diagnosis. Additionally, a meta-analysis was calculated without the UK Biobank for the purpose of creating a polygenic risk score. The lower part of the figure summarizes the downstream work that was performed. Analyses to evaluate the candidate genes at the genome-wide association study loci included eQTL lookups in heart tissue from the genotype-tissue expression, assessment of the variant effects regarding missense variation, generation, and interrogation of a protein expression data set of mitral valve tissue in humans and interrogation of RNA-Seq data set from mitral valve and heart tissue from mice, literature evaluation of candidate genes, and gene-based test based on the summary level results and including gene-set enrichment analyses as a follow-up. Furthermore, we assessed variants in a conditional and joint analysis to establish independent signals and their overlap with open chromatin regions from mitral valve and heart tissue, primarily based on the ATAC-Seq data. We also evaluated the overlap with functional elements in the genome in a partitioned heritability analysis. We checked whether mitral valve prolapse had a shared genetic architecture with related traits via genetic correlation analyses. We assessed the association of the tophits to other genome-wide association study traits with the Phenoscanner tool. Finally, we derived a polygenic risk score with the tool PRS-CS and tested the score in the UK Biobank. For each downstream analyses, the circles indicate whether the analysis was performed for the primary results (green circle), for the mitral valve surgery results (grey circle), or for the meta-analysis without the UK Biobank (white circle with black outline). ATAC-Seq, assay for transposase-accessible chromatin using sequencing; CV, cardiovascular; eQTL, expression quantitative trait locus; GWAS, genome-wide association study; MAGMA, Multi-marker Analysis of GenoMic Annotation; MV, mitral valve; MVP, mitral valve prolapse; PRS-CS, polygenic risk score-continuous shrinkage; RNA-Seq, RNA sequencing; TWAS, transcriptome-wide association study.

Figure 2

The Manhattan plot and evaluation of genes at each locus of mitral valve prolapse meta-analysis. The Manhattan plot of the meta-analysis results for mitral valve prolapse. Highlighted in blue are the 14 loci with a genome-wide significance (P-value <5 × 10⁻⁸). The grey dotted line indicates the genome-wide significance cut-off while the blue solid line indicates the subthreshold cut-off (P-value <5 × 10⁻⁶). Below the Manhattan plot, the candidate genes for each locus are listed and the lines of evidence for each gene are summarized. A gene at the GWAS locus was chosen as a candidate gene if the gene (i) overlapped with a linkage disequilibrium window (R  ² ≥ 0.6) around the sentinel variant, (ii) had an eQTL to a sentinel variant, or (iii) had a genome-wide significant missense mutation. A red quadrant indicates that the evidence is present. eGene, the sentinel variant has an eQTL for the gene in cardiac atrial or ventricular tissue; TWAS, significant in the transcriptome-wide association study in either atrial or ventricular tissue; NG, nearest gene(s) to the sentinel variant; Proteomics, protein levels in human mitral valve were detected; RNA-Seq, detectable expression levels in mitral valve or heart tissue in RNA-Seq data from mice; MAGMA, significantly associated in the multi-marker analysis; Literature, gene has been implicated for mitral valve prolapse in previously published work; Intronic, the sentinel variant fell within the intron of the gene; Missense, a significantly associated variant was missense for the gene. eQTL, expression quantitative trait locus; eGene, expression quantitative trait locus gene; GWAS, genome-wide association study; LD, linkage disequilibrium; MAGMA, Multi-marker Analysis of GenoMic Annotation; MVP, mitral valve prolapse; NG, nearest gene(s); RNA-Seq, RNA sequencing; TWAS, transcriptome-wide association study.

Figure 3

Genetic correlation of mitral valve prolapse with other cardiac phenotypes. Genetic correlation between mitral valve prolapse and cardiovascular phenotypes, including diseases, electrocardiographic traits, and cardiac magnetic resonance imaging traits. The genetic correlation was assessed with the linkage disequilibrium score regression method. ECG, electrocardiogram; LVEF, left ventricular ejection fraction; LVEDVi, left ventricular end-diastolic volume indexed to body surface area; LVESVi, left ventricular end-systolic volume indexed to body surface area; SVi, stroke volume indexed to body surface area; MRI, magnetic resonance imaging; MVP, mitral valve prolapse.

Figure 4

Locus summary for 3p25.3 and the candidate gene LMCD1 and 2p16.2 and the candidate gene SPTBN1. (A) Summary of locus at 3p25.3. Left: regional plot showing the linkage disequilibrium pattern for the sentinel variant rs165177 based on the European ancestry linkage disequilibrium observed from 1000 genomes. The T-allele is associated with an increased risk for mitral valve prolapse. Middle: Expression quantitative trait locus for rs165177 in atrial tissue from the genotype-tissue expression. The T-allele is associated with decreased expression of the LMCD1 gene. Right: Protein level of the candidate gene LMCD1 from human is shown in blue and RNA-Seq expression of Lmcd1 from mouse tissue, cardiac and mitral valve shown in red. (B) Summary of evidence for locus at 2p16.2. Left: regional plot showing linkage disequilibrium pattern for the sentinel variant rs12713274 based on the European ancestry linkage disequilibrium from 1000 genomes. The A-allele is associated with increased risk for mitral valve prolapse. Middle: expression quantitative trait locus for rs12713274 in atrial tissue from the genotype-tissue expression. The A-allele is associated with a decreased expression of the SPTBN1 gene. Right: Protein levels of the candidate gene SPTBN1 from human are shown in blue and RNA-Seq expression from mouse tissue for Sptbn1, mitral valve shown in red. GTEx, genotype-tissue expression; iBAQ, intensity-based absolute quantification; LD, linkage disequilibrium; MV, mitral valve; MVP, mitral valve prolapse; RNA-Seq, RNA sequencing; SNP, single nucleotide polymorphism.

Figure 5

Performance of polygenic risk prediction of mitral valve prolapse in the UK Biobank. (A) Receiver operating characteristic curves of prediction models including age, sex, and polygenic risk score. (B) Receiver operating characteristic curves of prediction models including clinical risk factors [hypertension, all-cause heart failure, myocardial infarction, and diabetes (Type 1 and Type 2)] for mitral valve prolapse and polygenic risk score. (C) Quintiles of polygenic risk are based on a polygenic risk score derived on summary statistics without the UK Biobank, and using the European ancestry samples from the 1000 genomes reference set for linkage disequilibrium. The score was applied to unrelated European ancestry individuals from the UK Biobank (n = 393 229). The individuals were grouped into quintiles based on their polygenic risk score and for each group, the prevalence of mitral valve prolapse was calculated. (D) Distribution of the polygenic risk score in the UK Biobank, marked with a red line is the top 20% cut-off for the score. The top 20% presented with an odds ratio of 1.79 compared with the bottom 80%. LD, linkage disequilibrium; MVP, mitral valve prolapse; OR, odds ratio; PRS, polygenic risk score; RF, risk factors; ROC, receiver operating characteristic.