Genome-wide Association for Heart Failure: From Discovery to Clinical Use

Dominic E. Fullenkamp; Megan J. Puckelwartz; Elizabeth M. McNally


Eur Heart J. 2021;42(20):2012-2014. 

Graphical Abstract: Genetic contribution to dilated cardiomyopathy.

The work of Garnier et al. reveals a polygenic risk score that contributes to dilated cardiomyopathy (DCM). It is known that monogenic mutations also predispose to dilated cardiomyopathy. These genetic determinants can combine with other risks such as hypertension (HTN), coronary artery disease (CAD), and metabolic syndrome (MetS) to further increase risk for dilated cardiomyopathy and heart failure (HF).

HF affects >26 million people worldwide, and is a highly heritable condition. The genetic contributions responsible for this heritable component of HF arise from a combination of high-effect rare variants and low-effect common variants. The monogenic contributions to DCM have been well characterized, and clinical testing to define these alleles is useful for arrhythmia risk prediction and family member management. Distinct from family-based genetic testing, genome-wide association studies (GWAS) have also been useful to define the genetic contribution to HF at the population level. HF GWAS have been challenged by the need to amass sizable and accurately phenotyped cohorts. Furthermore, the phenotypes used in GWAS may address different components of the HF clinical spectrum. Garnier and colleagues have now carried out a genome-wide association analysis on a large DCM cohort, as reported in this issue of the European Heart Journal.[1] The authors genetically evaluated 2719 DCM cases and 4440 controls, making this the largest GWAS of DCM. This effort confirmed two previously identified loci, BAG3 and HSPB7, and identified additional DCM-associated loci, and, importantly, identified two new loci, on chromosomes 3p25.1 and 22q11.23.

The discovery cohort and two replication cohorts were used by Garnier and colleagues who conducted GWAS using the phenotype of sporadic DCM, defined as reduced left ventricular ejection fraction (LVEF) (<45%) and enlarged left ventricular (LV) end-diastolic volume/diameter in the absence of obvious pathology including CAD or valve disease. This strategy was aimed at uncovering genetic loci important for cardiac dimensions, rather than other aspects of HF. In the discovery cohort, single nucleotide polymorphisms (SNPs) were tested for association, and five loci reached significance; four of these replicated in the additional two cohorts and the authors focused their analysis on two culprit genes, SLC6AC and SMARCB1 in the novel loci. SLC6A6 encodes a taurine transporter, and a homozygous mutation in human SLC6A6 associates with retinal disease and cardiomyopathy for which taurine supplementation may have benefit.[2] The second potential gene implicated in DCM is SMARCB1, which encodes a chromatin regulator. GWAS can effectively associate genomic loci with phenotypes of interest, in this case DCM, but it can be difficult to identify the mechanism of action or even the gene responsible. In this case, the authors used informatic and expression tools to implicate the SLC6A6 and SMARCB1 genes. Future work will ascertain whether and how these genes influence myocardial function.

In addition to the novel loci, the Garnier study reaffirmed signals seen in prior HF GWAS, providing further support for the importance of these pathways. For example, an early GWAS of advanced HF identified a chromosome 1 locus containing the CLCNKA and HSBP7 genes.[3] A DCM GWAS of 1179 sporadic cases and 1108 controls also provided support for the CLCNKA/HSPB7 locus, but also identified the BAG3 gene as a significant locus.[4] The BAG3 gene encodes a chaperone protein, and it has been linked to both HF and DCM in both GWAS and monogenic studies, providing striking genetic support for its essential role in heart function. A large GWAS of 4100 DCM cases and 7600 controls did not replicate previous results, but rather this study identified a region that affects expression for genes encoding class I and II major histocompatibility complex heavy chain receptors.[5] Using an exome array, Esslinger and colleagues identified eight loci associated with sporadic DCM, including BAG3.[6] In each of these DCM GWAS, cases were defined by enlarged LV diameter and reduced LVEF (≤45%) in the absence of other causal factors including CAD. As in the Garnier study, this design enriches for genetic variation that influences myocardial function and dimensions, rather than the clinical syndrome of HF. The Garnier study did not find any correlation of these loci with metabolic or lipid traits, further underscoring the importance of these loci for myocardial function independent of primary vascular disease.

Garnier et al. also used GWAS data to generate a genetic risk score (GRS) for DCM. A 27% increase in DCM risk was attributed to eight risk alleles [odds ratio (OR) 1.27, 95% confidence interval (CI) 1.14–1.42] and a 21% decreased risk for individuals with only one risk allele (OR 0.79, 95% CI 0.66–0.95). There was also a significant association in subjects with an increased LV end-diastolic diameter (n = 2187, OR 1.53, 95% CI 1.05–2.23), indicating that these loci are targeting particular phenotypes within the DCM diagnosis. A limitation of the Garnier study is its focus on European ancestry DCM, a population with narrow genetic architecture. Many HF and DCM GWAS studies, including the Garnier study, exclude familial DCM cases. Yet, when cohorts are examined with deep sequencing of cardiomyopathy genes, pathogenic and probable pathogenic variants are uncovered. This is not unexpected since HF and DCM arise from both rare and common variants (Graphical Abstract).

The question persists of whether GWAS would yield more genetic information with the use of more refined phenotypes. For example, Aragam et al. examined whether more robust HF phenotyping would improve genetic association in the UK Biobank, which has >450 000 mostly European ancestry participants.[7] HF was refined by defining LV dysfunction in the absence of CAD as non-ischaemic cardiomyopathy. GWAS using all-cause HF (n = 7382 cases) identified loci predominantly linked to well-known HF risk factors such as CAD and atrial fibrillation. Refining the phenotype to include only non-ischaemic cardiomyopathy (n = 2038 cases) uncovered loci implicated in DCM. These signals survived replication in independent cohorts and after adjustment for other HF risk factors, and these loci associated with metrics of LV dysfunction in subjects without clinical HF, suggesting genetic determinants of cardiac function itself, rather than intermediaries. Although phenotypic refinement reduced cohort size by a third, it produced more robust signals for DCM. This result reflects that HF is not a single entity, but rather a compendium of phenotypes driven by an assortment of genetic and environmental processes. Distillation of HF into more distinct subphenotypes improved the power of GWAS to identify genetic drivers of key elements underlying HF.

From patient to patient, the clinical syndrome of HF varies in its progression or disease trajectory. Most case–control studies rely on ascertaining phenotype at a single point in time, and consider the binary state of disease or no disease. However, emerging evidence supports that the dimension of disease trajectory adds additional power to phenotypic information, as some individuals progress quickly while others remain relatively stable. By adding disease progression into phenotypic assessment, even smaller cohorts can be suitable for GWAS.[8,9]

LV structural changes are a defining feature of DCM, and several large-scale GWAS have been performed using echocardiographic measures. These studies have identified several loci known to cause monogenic cardiomyopathies.[10,11] Pirruccello et al. used cardiac magnetic resonance imaging (MRI)-derived LV measures in ~36 000 UK Biobank subjects.[12] They found that the MRI phenotypes were influenced by participants' genetics, with SNP-based heritability ranging from 34% to 43% for cardiac dimensional and functional measurements. They identified 45 loci associated with cardiac structure and function, and 17 overlapped with known monogenic cardiomyopathy genes. Using these GWAS data, a GRS was created which associated with incident DCM in the general population after exclusion of subjects with DCM. Even in the presence of TTN truncating genetic variants, which are known to cause ~20% of inherited DCM,[13] this score influenced the size and shape of the heart, implicating common variants in DCM as additive to rare variants. Tadros and colleagues provide further evidence that common genetic variants influence cardiac measures.[14] They conducted GWAS and multitrait analysis in DCM (n = 5521), hypertrophic cardiomyopathy (HCM) (n = 1733), and LV traits (19 260 UK Biobank subjects). In HCM subjects with a rare monogenic variant, common genetic variation explained a significant proportion of the phenotypic variability that is a hallmark of inherited cardiomyopathies. These data suggest that common genetic variants influence pathological changes in cardiac structure and function and that the genetic burden of common variants may significantly contribute to disease pathology in cardiomyopathies. Thus, GWAS of HF, DCM, and cardiac dimensions provide insight into pathways important for pathogenesis. Additionally, there is emerging evidence that a GRS generated from GWAS could prove useful in the clinic, especially with precise phenotyping.

Methods to integrate common and rare HF genetic information will continue to evolve and provide insight on disease progression, potentially providing biomarkers and clues for useful therapeutic pathways to guide drug development. At present, rare cardiomyopathy variants have clinical utility in predicting risk, especially arrhythmia risk. Polygenic risk scores for HF or DCM progression are expected to come to clinical use, especially with the addition of broader GWAS-derived data. For polygenic risk scores to realize better clinical utility, additional GWAS are needed, especially from individuals from diverse genetic ancestries. Combining genetic risk data with clinical and social determinants should help identify those at greatest risk, offering the opportunity for risk reduction.