This gene-based collapsing analysis assesses the contribution of rare variants to all-cause heart failure with and without reduced left ventricular ejection fraction.
Key Points
Question
What is the contribution of rare genetic variants to all-cause heart failure with and without reduced left ventricular ejection fraction?
Findings
In this gene-based collapsing analysis of 5942 patients with heart failure and 13 156 controls using whole-exome sequencing, a significant enrichment of rare protein-truncating variants in the TTN gene and in general an increased burden of mendelian cardiomyopathy variants was demonstrated in patients with heart failure of mostly presumed ischemic etiology compared with controls.
Meaning
Mendelian genetic conditions may represent an important subset of complex late-onset diseases such as heart failure, irrespective of the clinical presentation.
Abstract
Importance
Sequencing studies have identified causal genetic variants for distinct subtypes of heart failure (HF) such as hypertrophic or dilated cardiomyopathy. However, the role of rare, high-impact variants in HF, for which ischemic heart disease is the leading cause, has not been systematically investigated.
Objective
To assess the contribution of rare variants to all-cause HF with and without reduced left ventricular ejection fraction.
Design, Setting, and Participants
This was a retrospective analysis of clinical trials and a prospective epidemiological resource (UK Biobank). Whole-exome sequencing of patients with HF was conducted from the Candesartan in Heart Failure-Assessment of Reduction in Mortality and Morbidity (CHARM) and Controlled Rosuvastatin Multinational Trial in Heart Failure (CORONA) clinical trials. Data were collected from March 1999 to May 2003 for the CHARM studies and September 2003 to July 2007 for the CORONA study. Using a gene-based collapsing approach, the proportion of patients with HF and controls carrying rare and presumed deleterious variants was compared. The burden of pathogenic variants in known cardiomyopathy genes was also investigated to assess the diagnostic yield. Exome sequencing data were generated between January 2018 and October 2018, and analysis began October 2018 and ended April 2020.
Main Outcomes and Measures
Odds ratios and P values for genes enriched for rare and presumed deleterious variants in either patients with HF or controls and diagnostic yield of pathogenic variants in known cardiomyopathy genes.
Results
This study included 5942 patients with HF and 13 156 controls. The mean (SD) age was 68.9 (9.9) years and 4213 (70.9%) were male. A significant enrichment of protein-truncating variants in the TTN gene (P = 3.35 × 10−13; odds ratio, 2.54; 95% CI, 1.96-3.31) that was further increased after restriction to variants in exons constitutively expressed in the heart (odds ratio, 4.52; 95% CI, 3.10-6.68). Validation using UK Biobank data showed a similar enrichment (odds ratio, 4.97; 95% CI, 3.94-6.19 after restriction). In the clinical trials, 201 of 5916 patients with HF (3.4%) had a pathogenic or likely pathogenic cardiomyopathy variant implicating 21 different genes. Notably, 121 of 201 individuals (60.2%) had ischemic heart disease as the clinically identified etiology for the HF. Individuals with HF and preserved ejection fraction had only a slightly lower yield than individuals with midrange or reduced ejection fraction (20 of 767 [2.6%] vs 15 of 392 [3.8%] vs 166 of 4757 [3.5%]).
Conclusions and Relevance
An increased burden of diagnostic mendelian cardiomyopathy variants in a broad group of patients with HF of mostly ischemic etiology compared with controls was observed. This work provides further evidence that mendelian genetic conditions may represent an important subset of complex late-onset diseases such as HF, irrespective of the clinical presentation.
Introduction
Heart failure (HF) is a complex clinical syndrome affecting approximately 40 million people worldwide.1 Heart failure risk increases with age, leads to a high number of hospitalizations, and creates a major burden for the health care system.2 Although survival after diagnosis has improved, death rates remain high, with a 5-year survival averaging 50%.3
Heart failure is often classified into 3 main subtypes based on the left ventricular ejection fraction (EF). An EF of 50% or more is defined as HF with preserved EF (HFpEF), an EF between 40% and 49% is referred to as HF with midrange EF (HFmrEF), and an EF of 40% or less is called HF with reduced EF (HFrEF).4 All 3 subtypes have overlapping risk factors, but their pathophysiology and response to treatment are distinct.5
The majority of HF cases are attributed to ischemic heart disease, hypertension, or primary/secondary cardiomyopathies.6 Primary cardiomyopathies such as hypertrophic, dilated, or arrhythmogenic cardiomyopathies are often inherited but show variable expressivity with incomplete penetrance.7 They are typically caused by rare variants in genes encoding sarcomere or cytoskeletal proteins, membrane ion channels, or desmosomes. Studies have revealed a significant heritability of HF, estimated at 26%.8 While large-scale genome-wide association studies have highlighted the contribution of common variants to the cause of HF,9,10 the role of rare genetic variants in all-cause HF has not been systematically assessed, to our knowledge. The absence of such studies does not mean that rare, high-impact variants do not also play a role in HF due to ischemic disease, the leading cause of HF, but rather reflects the scarcity of large HF cohorts with sequencing data.
In this study, we performed whole-exome sequencing of patients with HF, representing all broad clinical subtypes, from the Candesartan in Heart Failure-Assessment of Reduction in Mortality and Morbidity (CHARM)11 and Controlled Rosuvastatin Multinational Trial in Heart Failure (CORONA)12 clinical trials. We applied gene-based collapsing analyses to discover genes with an excess of rare, presumably deleterious variants among patients with HF compared with controls. We also performed a diagnostic analysis in known cardiomyopathy genes to investigate the overall burden of pathogenic cardiomyopathy variants. Finally, we used UK Biobank data to validate the main collapsing results.
Methods
Study Population
Sites participating in all studies received approval from local ethics committees for their conduct.11,12 Only patients who gave written informed consent for genetic analysis and for whom a DNA sample was available were included in the present study. This study was performed in accordance with the policy on bioethics and human biologic samples of AstraZeneca. Data were collected from March 1999 to May 2003 for the CHARM study and September 2003 to July 2007 for the CORONA study. Exome sequencing data were generated between January 2018 and October 2018.
CORONA Trial
A total of 5011 patients 60 years and older with chronic HF of investigator-reported ischemic cause and a left ventricular EF of 40% and less were enrolled for the study (NCT00206310) and randomly assigned to receive rosuvastatin or placebo.11
CHARM Trials
Eligible patients, 18 years or older with symptomatic HF, were enrolled into 1 of 3 concurrent studies, according to their left ventricular EF: more than 40% (CHARM Preserved; NCT00634712), 40% or less and treated with angiotensin-converting enzyme inhibitor (CHARM Added; NCT00634309), or 40% or less and not treated with angiotensin-converting enzyme inhibitor due to intolerance (CHARM Alternative; NCT00634400).12 A total of 7601 patients were enrolled in these studies and randomly assigned candesartan or placebo. For the subgroup analysis, CHARM Preserved samples were split according to their EF into HFmrEF (EF, 40%-49%) and HFpEF (EF, ≥50%).13,14
Controls
The control cohort consisted of unrelated individuals from the Institute for Genomic Medicine database collected as part of unrelated studies (eMethods and eTable 2 in the Supplement). Protocols were approved by Columbia University’s institutional review board and participants provided informed consent for the use of DNA in genetic research.
UK Biobank
UK Biobank is a large prospective cohort study with more than 500 000 participants aged 40 to 69 years recruited from the general population in the United Kingdom.15 UK Biobank Field ID 41270 (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes based on hospital in-patient diagnoses) was interrogated for the following codes: I20-I25 ischemic heart diseases, I42 cardiomyopathy, I50 HF, and I70 atherosclerosis. Individuals with a primary or secondary International Statistical Classification of Diseases and Related Health Problems, Tenth Revision code in those categories were considered cases.
Exome Sequencing, Variant Calling, and Quality Control
Whole-exome sequencing of all patients with HF and controls was performed at the Institute for Genomic Medicine. All samples were processed with the same bioinformatic pipeline for variant calling and underwent basic quality control (eMethods in the Supplement).
Clustering
Principal component analysis for dimensionality reduction was performed on a set of predefined variants (previously described by Cameron-Christie et al16) to capture population structure. We applied the Louvain method of community detection17 to the first 6 principal components to identify clusters that reflect ancestry. Louvain and other graph-based clustering methods have advantages compared with other techniques because they do not assume clusters of a particular size, density, or shape.18 To check the quality of the clusters, we performed further dimensionality reduction using the Uniform Manifold Approximation and Projection19 on the first 6 principal components (eFigure 1 in the Supplement). We performed coverage harmonization and gene-based collapsing as detailed in the eMethods in the Supplement and below for each cluster with at least 30 patients with HF and 30 controls.
Gene-Based Collapsing
We performed gene-based collapsing to test for a significant difference in the proportion of patients with HF carrying at least 1 qualifying variant (QV) in a gene compared with controls. A QV is defined as a variant that passes certain filter criteria. While the quality control filters mentioned in the eMethods in the Supplement are the same in all models, other filters such as variant effect, pathogenicity predictors, and internal and external minor allele frequency are specific to a certain model. eTable 3 in the Supplement depicts the core models used for all analyses on the full data set with all patients with HF combined and 6 different subgroups (eTable 4 in the Supplement). The synonymous model was used as a negative control (eFigure 2 in the Supplement). We did not analyze individuals with HFmrEF and HFpEF who were 50 years and younger separately because of small sample sizes. Including models with regional intolerance described in the eMethods in the Supplement, we analyzed 7 synonymous and 119 nonsynonymous models.
Diagnostic Variant Analysis
We performed a diagnostic analysis in known cardiomyopathy genes and classified variants based on the American College of Medical Genetics and Genomics (ACMG) guidelines20 to investigate the overall burden of pathogenic cardiomyopathy variants in our cohort. We selected 41 known cardiomyopathy genes based on the ACMG’s 2016 recommendations for reporting of secondary findings in clinical exome and genome sequencing21 and gene lists obtained from the curated Genomics England PanelApp22 (eMethods and eTable 5 in the Supplement) and extracted all variants with a minor allele frequency 1% or less. While all protein-truncating variants (PTVs) passing the minor allele frequency filter were evaluated, all other variants were only included if they were classified as pathogenic or likely pathogenic in ClinVar23,24 or disease Mutation in the Human Gene Mutation Database.25 All variants that passed above filters were assessed according to ACMG guidelines,20 and variants classified as pathogenic or likely pathogenic were deemed diagnostic variants.
Statistical Analyses
From the collapsing results of the individual clusters, we extracted the number of patients with HF and controls with and without a QV per gene and used the exact 2-sided Cochran-Mantel-Haenszel test26,27 to test for an association between disease status and QV status while controlling for cluster membership (using the stats package in R version 3.6; R Foundation). We created quantile-quantile plots and estimated the genomic inflation factor λ using a permutation-based expected distribution of P values (eMethods in the Supplement). We defined a studywide Bonferroni multiplicity-adjusted significance threshold of P < 2.25 × 10−8 (0.05 / [18 650 genes × 119 nonsynonymous models]). Considering the high degree of correlation among the various models and subgroups, this is a very conservative threshold. Analysis of the sequencing data began in October 2018 and ended April 2020.
Results
In total, we performed whole-exome sequencing of 5942 patients with HF. After removal of related individuals, we performed gene-based collapsing analyses stratified by clusters in 5916 individuals with HF and 13 156 unrelated controls without reported heart disease. In the CORONA trial, 2869 of 2870 unrelated patients (99.97%) had HFrEF due to an ischemic event. The CHARM studies included 3046 patients, of whom 1887 (62.0%) and 767 (25.2%) had HFrEF and HFpEF, respectively, overall comprising 1836 patients (60.3%) with ischemic HF (Table 1; eTable 1 in the Supplement). The mean (SD) age for all patients was 68.9 (9.9) years. Of all patients, 4213 (70.9%) were male. Of unrelated samples, 4199 (71.0%) were male.
Table 1. Characteristics of All Sequenced Cases.
| Characteristic | No. (%)a | ||
|---|---|---|---|
| CHARM | CORONA | All | |
| No. of samples | 3056 | 2886 | 5942 |
| Age, y | |||
| >50 | 2751 (90.0) | 2886 (100.0) | 5637 (94.9) |
| ≤50 | 305 (10.0) | 0 | 305 (5.1) |
| Male | 2000 (65.5) | 2213 (76.7) | 4213 (70.9) |
| Female | 1056 (34.6) | 673 (23.3) | 1729 (29.1) |
| HFrEF | 1890 (61.9) | 2886 (100.0) | 4776 (80.4) |
| HFmrEF | 393 (12.9) | 0 | 393 (6.6) |
| HFpEF | 773 (25.3) | 0 | 773 (13.0) |
| Ischemic | 1842 (60.3) | 2885 (100.0) | 4727 (79.6) |
| Idiopathic | 564 (18.5) | 0 | 564 (9.5) |
| Hypertensive | 418 (13.7) | 0 | 418 (7.0) |
| Other cause | 232 (7.6) | 1 (0.03) | 233 (3.9) |
Abbreviations: CHARM, Candesartan in Heart Failure-Assessment of Reduction in Mortality and Morbidity; CORONA, Controlled Rosuvastatin Multinational Trial in Heart Failure; HFmrEF, heart failure with midrange ejection fraction; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction.
Numbers include a small number of related individuals (see eTables 1 and 4 in the Supplement for unrelated individuals).
Gene-Based Collapsing Analyses
We removed 3 of 11 clusters because they did not contain enough cases, and only 5860 patients with HF and 10 578 controls were included in the final test (eTable 6 in the Supplement reports cluster sizes). In an analysis comprising all HF subtypes, 1 gene, TTN (encoding Titin), reached studywide significance in several models. We detected the strongest association in the dominant PTV model (P = 3.35 × 10−13; odds ratio [OR], 2.54; 95% CI, 1.96-3.31; Figure 1), for which QVs were defined as PTVs with an internal and external minor allele frequency below 0.1%.
Figure 1. Quantile-Quantile Plot of Observed vs Expected P Values for the Full Cohort and Protein-Truncating Variants Model.
Only TTN reached studywide significance with a P value of 3.35 × 10−13. The orange and light blue lines indicate the 2.5th and 97.5th percentile of expected P values, respectively.
PTVs in TTN (TTN–truncating variants; TTNtvs) are a known cause of cardiomyopathies, especially dilated cardiomyopathy (DCM). Not all TTNtvs are considered causal, as the background prevalence of TTNtvs of uncertain significance in the general population is relatively high.28,29 Roberts et al29 showed that TTNtvs in healthy controls are enriched in exons that are usually not expressed in the main cardiac isoforms of TTN, whereas the majority of variants in cardiomyopathy patients cluster in the constitutively expressed distal exons. The proportion spliced-in (PSI) metric is an estimate of the percentage of TTN transcripts that incorporate an exon based on RNA sequencing data and can be used to identify variants with a higher probability of pathogenicity.29,30 Filtering QVs in TTN for a PSI more than 0.9 removes the majority of control variants while retaining most of the case variants increasing the odds ratio from 2.54 to 4.52 and decreasing the P value further to 5.21 × 10−18. The collapsing signal for TTNtvs is detectable in all subgroups (Figure 2A), although it does not reach significance in some of them.
Figure 2. Forest Plot Showing an Enrichment of TTN Truncating Variants in Cases Compared With Controls in CHARM/CORONA and the UK Biobank.
All subgroups show an enrichment that increases after filtering for a proportion spliced in (PSI) more than 0.9 (odds ratios indicated in yellow). Note that subgroups defined by International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes in B are not mutually exclusive (see eTable 16 in the Supplement for more details). CHARM indicates Candesartan in Heart Failure-Assessment of Reduction in Mortality and Morbidity; CORONA, Controlled Rosuvastatin Multinational Trial in Heart Failure; HFmrEF, heart failure with midrange ejection fraction; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction.
Overall, 71 of 127 patients with HF (55.9%) with TTNtvs in exons with a high PSI had ischemic heart disease as the identified primary etiology, 43 (33.9%) had known idiopathic DCM, 7 (5.5%) were classified as having hypertensive HF, and the remaining 6 (4.7%) had various other HF etiologies. While the distribution shows the expected enrichment in known DCM, a surprisingly high number of patients with ischemic etiology harbor potentially diagnostic TTN variants. These results highlight that although TTNtvs are only known to cause nonischemic cardiomyopathies, also patients with a clinical diagnosis of ischemic HF are enriched for this type of variant.
The TET2 gene achieved a significant P value in the dominant PTV model (P = 8.30 × 10−10; OR, 33.73; 95% CI, 7.56-311.77) for the HFrEF CORONA subgroup and suggestive signal in the full cohort (P = 1.22 × 10−6; OR, 5.11; 95% CI, 2.50-10.80). Somatic mutations in TET2 are associated with age-related clonal hematopoiesis.31,32,33 Previous studies have shown an association between age-related clonal hematopoiesis and the risk of coronary heart disease in humans and with accelerated atherosclerosis in mice.31,32,34 The frequency of somatic mutations increases with age and a considerable clonal expansion is needed to increase the alternate allele fraction so that the variant is detected and passes our quality control filters. Hence, a higher age at sample collection increases the chance of detecting an age-related somatic mutation via germline calling and inclusion in the collapsing analysis. Because the patients in this study are notably older than the controls, we cannot disentangle the age-related effect from the true effect that might be responsible for our phenotype of interest. Further analyses with age-matched controls are needed to investigate the role of somatic TET2 mutations in HF.
Restricting the analysis to clusters that reflect European ancestry did not lead to changes in the significant genes because it still included 93.8% of the patients with HF (5494 of 5860). eTables 7 and 8 in the Supplement list the top genes across models and subgroups.
Diagnostic Analysis
In the CHARM and CORONA trials, we performed a diagnostic analysis in 41 known cardiomyopathy genes and classified variants according to ACMG criteria. Overall, we detected 204 diagnostic variants in 201 of 5916 patients with HF (3.4%) affecting 21 different genes (Table 2; eResults in the Supplement). Consistent with the findings from the collapsing analysis, a large proportion of the diagnostic variants (125 of 204 [61.3%]) were TTNtvs. There was a substantial overlap (124 of 127 [97.6%]) between TTNtvs that passed PSI filtering and diagnostic TTN variants showing that the collapsing model with additional PSI filtering truly captures diagnostic variants. Other genes containing multiple diagnostic variants include MYBPC3, known to primarily cause hypertrophic cardiomyopathy; MYH7, a gene in which variants can cause either DCM or hypertrophic cardiomyopathy; DSP, DSC2, DSG2, and SCN5A, genes known to be implicated in dilated and arrhythmogenic cardiomyopathies; and FLNC, in which variants can cause all types of cardiomyopathies.
Table 2. Diagnostic Yield in Different Subgroups.
| EF | Sex | No./total No. (%) | ||||
|---|---|---|---|---|---|---|
| Hypertension | Idiopathic dilated cardiomyopathy | Ischemic heart disease | Other cause | Total | ||
| HFpEF | Female | 3/153 (2.0) | 2/31 (6.5) | 2/151 (1.3) | 2/58 (3.5) | 9/393 (2.3) |
| Male | 4/87 (4.6) | 3/42 (7.1) | 2/194 (1.0) | 2/51 (3.9) | 11/374 (2.9) | |
| HFmrEF | Female | 2/36 (5.6) | 3/23 (13.0) | 1/69 (1.5) | 0/17 | 6/145 (4.1) |
| Male | 2/30 (6.7) | 4/32 (12.5) | 2/174 (1.2) | 1/11 (9.1) | 9/247 (3.6) | |
| HFrEF | Female | 1/38 (2.6) | 9/152 (5.9) | 24/951 (2.5) | 2/38 (5.3) | 36/1179 (3.1) |
| Male | 3/72 (4.2) | 34/283 (12.0) | 90/3166 (2.8) | 3/57 (5.3) | 130/3578 (3.6) | |
| HFpEF | Both | 7/240 (2.9) | 5/73 (6.9) | 4/345 (1.2) | 4/109 (3.7) | 20/767 (2.6) |
| HFmrEF | Both | 4/66 (6.1) | 7/55 (12.7) | 3/243 (1.2) | 1/28 (3.6) | 15/392 (3.8) |
| HFrEF | Both | 4/110 (3.6) | 43/435 (9.9) | 114/4117 (2.8) | 5/95 (5.3) | 166/4757 (3.5) |
| All EF | Female | 6/227 (2.6) | 14/206 (6.8) | 27/1171 (2.3) | 4/113 (3.5) | 51/1717 (3.0) |
| Male | 9/189 (4.8) | 41/357 (11.5) | 94/3534 (2.7) | 6/119 (5.0) | 150/4199 (3.6) | |
| Total | Both | 15/416 (3.6) | 55/563 (9.8) | 121/4705 (2.6) | 10/232 (4.3) | 201/5916 (3.4) |
Abbreviations: EF, ejection fraction; HFmrEF, heart failure with midrange ejection fraction; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction.
While almost all of the diagnostic variants in individuals with idiopathic DCM as the primary etiology affected known DCM genes, 2 individuals had diagnostic variants in MYBPC3, a gene primarily associated with hypertrophic cardiomyopathy. In individuals with ischemic heart disease, diagnostic variants were also primarily found in DCM genes such as TTN, DSG2, and BAG3 or genes known for several types of cardiomyopathy such as FLNC or TNNT2. Diagnostic variants in individuals with a primarily hypertensive etiology were mostly TTNtvs or missense variants in MYBPC3. While there was 1 gene uniquely affected by diagnostic variants in HFmrEF and 11 genes in HFrEF, all genes implicated in HFpEF were a subgroup of HFrEF genes and the small overlap of genes (1 of 23) that were affected in all groups was probably due to the small sample size for HFmrEF and HFpEF, because unique genes also harbored a smaller number of diagnostic variants in general (Figure 3; eTable 9 in the Supplement).
Figure 3. Venn Diagram of Genes With a Diagnostic Variant for the 3 Different Heart Failure Types .
All genes implicated in heart failure with preserved ejection fraction are a subgroup of heart failure with reduced ejection fraction genes, and only 1 heart failure with midrange ejection fraction gene is not implicated in heart failure with reduced ejection fraction.
Patients with HFpEF had a slightly lower diagnostic yield than those with HFmrEF or HFrEF (20 of 767 [2.6%] vs 15 of 392 [3.8%] vs 166 of 4757 [3.5%]). The difference does not seem to be caused by primary etiology or sex as it is present in almost all of the subgroups but could be due to different rates of TTNtvs, since after removal of TTN and individuals with a diagnostic TTNtv, the yields were comparable (10 of 757 [1.3%] for HFpEF, 3 of 380 [0.8%] for HFmrEF, and 66 of 4654 [1.4%] for HFrEF; eTables 10 and 11 in the Supplement).
As expected, individuals with DCM as the primary etiology had a very high diagnostic yield (55 of 563 [9.8%]), especially if they had HFmrEF (7 of 55 [12.7%]) or were male and had HFrEF (34 of 283 [12.0%]). Notably, the diagnostic rate for female individuals with HFrEF and an idiopathic DCM was substantially lower (9 of 152 [5.9%]). Male patients with HFrEF and ischemic heart disease, by far the largest subgroup, had a diagnostic yield of almost 3% (90 of 3166). The diagnostic yield for a hypertensive etiology or other cause was greater than 3% (15 of 416 [3.6%] and 10 of 232 [4.3%]).
Restricting the analysis to patients with HF 50 years or younger increased the diagnostic yield in individuals with DCM to 12.9% (15 of 116) but decreased the yield for all other etiologies, which might be due to the small sample size (eTable 12 in the Supplement). Most of the diagnostic variants found in younger individuals (15 of 18 [83.3%]) were TTNtvs.
Results restricted to only individuals of mostly European ancestry were almost identical to those of the full cohort (eTable 13 in the Supplement) because this group encompasses 5481 of 5916 patients with HF (92.7%). Sample sizes for other ancestries were too low, but we did detect diagnostic TTNtvs in all groups.
To assess the clinical significance of the diagnostic results, we performed the same analysis for all 13 156 controls and observed a 3-fold overall increase in diagnostic yield for patients with HF (201 of 5916 [3.4%] vs 146 of 13 156 [1.1%] for controls) and up to 5-fold increase in TTN alone (eTable 14 in the Supplement).
UK Biobank Validation of TTN Findings
The enrichment observed with PSI filtering in the UK Biobank Europeans for HF (I50) was very similar to the enrichment in the CORONA and CHARM trials (Figure 2). After filtering, the genetic risk was increased from 2.95 (95% CI, 2.42-3.57) to 4.97 (95% CI, 3.94-6.19) compared with 2.54 (95% CI, 1.96-3.31) to 4.52 (95% CI, 3.10-6.68) in the clinical studies. Furthermore, similar enrichment is observed for cardiomyopathy (I42) from 7.32 (95% CI, 5.65-9.50) to 15.15 (95% CI, 11.57-19.84). A smaller enrichment could also be observed in the ischemic heart disease group (I20-25) (1.31 [95% CI, 1.15-1.49] to 1.53 [95% CI, 1.28-1.83]). No association with atherosclerosis (I70) could be observed with or without PSI filtering (eTable 15 in the Supplement). A more detailed analysis of the ischemic heart disease group (I20-25) showed that there is no enrichment in patients with an I20-25 diagnosis that do not have cardiomyopathy (I42) or HF (I50), but there is an enrichment in those with an I20-25 code and an HF diagnosis but no cardiomyopathy (eTable 16 in the Supplement).
Discussion
We performed the first large-scale whole-exome sequencing study of a broad population of patients with HF. Although our criteria for determining diagnostic variants were very stringent, our results demonstrate a surprisingly high diagnostic yield within known cardiomyopathy genes (3.4%) for a cohort clinically classified as mostly ischemic. This is particularly relevant in context of the significantly lower diagnostic yield in controls and highlights the enrichment of deleterious cardiomyopathy variants in patients with HF. Individuals with idiopathic DCM had the highest diagnostic yield (9.8%). However, even within the largest subgroup of ischemic HF, almost 3% of patients with HF had a diagnostic variant. The yield was only slightly lower in HFpEF compared with HFrEF or HFmrEF and the genes implicated largely overlap, suggesting more commonality in the genetic architecture than would be expected from the clinically identified mechanisms involved in these different subgroups. This finding was unexpected because the pathophysiology of the different subtypes is distinct. HFpEF in particular is thought to be more affected by comorbidities and environmental triggers with little genetic contribution.35 Given the complexity of the syndrome of HF, these findings suggest that genetic variants could add to the clinical phenotype and prognosis as this may not completely reflect the underlying causes of the HF.
We also show that after filtering based on expression in the heart, the TTNtvs from our collapsing model were almost all diagnostic variants based on ACMG criteria. Therefore, the almost 5-times increased rate of QVs in TTN found in patients with HF compared with controls reflects the increased diagnostic yield of the patients with HF. The TTN signal is present in HFpEF, HFmrEF, and HFrEF, again underlining the genetic overlap of the subtypes irrespective of the EF and of the underlying clinical diagnosis.
Currently, there is strong interest in linking genetic information to electronic health record data. While many reproducible association results have been reported,36,37 there are also concerns about the quality of the records and suitability for genetic analyses.38 In our analyses, it is reassuring that we observed similar enrichment values of TTNtvs in patients with HF compared with controls between the clinical studies, with stringent inclusion and exclusion criteria, and UK Biobank, a population cohort. Furthermore, the more detailed analysis of the ischemic heart disease group might indicate that TTNtvs are not associated with increased risk of ischemic heart disease in general, but increased risk of developing HF after an ischemic event.
Limitations
Although we sequenced 5942 patients with HF, sample sizes for several subgroups were small, as were the numbers of cases carrying a diagnostic variant. Therefore, especially for small differences in yield, we cannot distinguish true differences from random effects. Furthermore, the stringent criteria used to classify variants as pathogenic may lead to an underdiagnosis of patients with mendelian conditions.
Conclusions
Our results show that rare genetic variants provide complementary information to the clinical phenotypes in patients with HF. For the holistic assessment of patients and understanding of differences in prognosis or treatment response, it is unclear whether grouping patients with HF based on their EF or primary etiology alone is sufficient because a proportion of patients with HF have genetic causes that may influence outcomes and response.
eMethods
eResults
eFigure 1. UMAP plot of the full dataset colored by cluster membership (A) and predicted ancestry (B)
eFigure 2. Quantile-quantile plot of observed vs. expected p-values for the full cohort (“All”) and Synonymous model
eTable 1. Characteristics of unrelated cases
eTable 2. Phenotypes of unrelated control samples
eTable 3. Main collapsing models and their specific filters
eTable 4. Subgroups and number of unrelated cases analyzed
eTable 5. List of genes used in diagnostic analysis and condition from PanelApp
eTable 6. Cluster sizes for the full cohort (“All”)
eTable 7. Genes with enrichment in cases and a p-value < 10-5 across subgroups and models ranked by p-value (includes only top p-value per gene)
eTable 8. Genes with enrichment in cases and a p-value < 10-5 across subgroups and models ranked by p-value (includes only top p-value per gene) from European-only analysis
eTable 9. Number of diagnostic variants per gene
eTable 10. Diagnostic yield TTN only
eTable 11. Diagnostic yield after removal of TTN variants and TTN variant carriers
eTable 12. Diagnostic yield in patients with <= 50 years of age
eTable 13. Diagnostic yield in Europeans
eTable 14. Number of diagnostic variants and yield per gene for cases and controls
eTable 15. TTN results for more detailed phenotypes of I20-I25 (ischemic heart diseases), I42 (cardiomyopathy), I50 (heart failure), and I70 (atherosclerosis) in UK Biobank Europeans
eTable 16. TTN results for different intersections of phenotypes of I20-I25 (ischemic heart diseases) with I42 (cardiomyopathy) and I50 (heart failure) in UK Biobank Europeans
eReferences
References
- 1.Vos T, Allen C, Arora M, et al. ; GBD 2015 Disease and Injury Incidence and Prevalence Collaborators . Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990-2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet. 2016;388(10053):1545-1602. doi: 10.1016/S0140-6736(16)31678-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Cook C, Cole G, Asaria P, Jabbour R, Francis DP. The annual global economic burden of heart failure. Int J Cardiol. 2014;171(3):368-376. doi: 10.1016/j.ijcard.2013.12.028 [DOI] [PubMed] [Google Scholar]
- 3.Savarese G, Lund LH. Global public health burden of heart failure. Card Fail Rev. 2017;3(1):7-11. doi: 10.15420/cfr.2016:25:2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Yancy CW, Jessup M, Bozkurt B, et al. ; American College of Cardiology Foundation; American Heart Association Task Force on Practice Guidelines . 2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol. 2013;62(16):e147-e239. doi: 10.1016/j.jacc.2013.05.019 [DOI] [PubMed] [Google Scholar]
- 5.Tromp J, Westenbrink BD, Ouwerkerk W, et al. Identifying pathophysiological mechanisms in heart failure with reduced versus preserved ejection fraction. J Am Coll Cardiol. 2018;72(10):1081-1090. doi: 10.1016/j.jacc.2018.06.050 [DOI] [PubMed] [Google Scholar]
- 6.Ziaeian B, Fonarow GC. Epidemiology and aetiology of heart failure. Nat Rev Cardiol. 2016;13(6):368-378. doi: 10.1038/nrcardio.2016.25 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Jacoby D, McKenna WJ. Genetics of inherited cardiomyopathy. Eur Heart J. 2012;33(3):296-304. doi: 10.1093/eurheartj/ehr260 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Lindgren MP, PirouziFard M, Smith JG, Sundquist J, Sundquist K, Zöller B. A Swedish nationwide adoption study of the heritability of heart failure. JAMA Cardiol. 2018;3(8):703-710. doi: 10.1001/jamacardio.2018.1919 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Shah S, Henry A, Roselli C, et al. ; Regeneron Genetics Center . Genome-wide association and mendelian randomisation analysis provide insights into the pathogenesis of heart failure. Nat Commun. 2020;11(1):163. doi: 10.1038/s41467-019-13690-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Arvanitis M, Tampakakis E, Zhang Y, et al. ; 23andMe Research Team . Genome-wide association and multi-omic analyses reveal ACTN2 as a gene linked to heart failure. Nat Commun. 2020;11(1):1122. doi: 10.1038/s41467-020-14843-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Pfeffer MA, Swedberg K, Granger CB, et al. ; CHARM Investigators and Committees . Effects of candesartan on mortality and morbidity in patients with chronic heart failure: the CHARM-Overall programme. Lancet. 2003;362(9386):759-766. doi: 10.1016/S0140-6736(03)14282-1 [DOI] [PubMed] [Google Scholar]
- 12.Kjekshus J, Apetrei E, Barrios V, et al. ; CORONA Group . Rosuvastatin in older patients with systolic heart failure. N Engl J Med. 2007;357(22):2248-2261. doi: 10.1056/NEJMoa0706201 [DOI] [PubMed] [Google Scholar]
- 13.Lund LH, Claggett B, Liu J, et al. Heart failure with mid-range ejection fraction in CHARM: characteristics, outcomes and effect of candesartan across the entire ejection fraction spectrum. Eur J Heart Fail. 2018;20(8):1230-1239. doi: 10.1002/ejhf.1149 [DOI] [PubMed] [Google Scholar]
- 14.Solomon SD, Anavekar N, Skali H, et al. ; Candesartan in Heart Failure Reduction in Mortality (CHARM) Investigators . Influence of ejection fraction on cardiovascular outcomes in a broad spectrum of heart failure patients. Circulation. 2005;112(24):3738-3744. doi: 10.1161/CIRCULATIONAHA.105.561423 [DOI] [PubMed] [Google Scholar]
- 15.Bycroft C, Freeman C, Petkova D, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562(7726):203-209. doi: 10.1038/s41586-018-0579-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Cameron-Christie S, Wolock CJ, Groopman E, et al. Exome-based rare-variant analyses in CKD. J Am Soc Nephrol. 2019;30(6):1109-1122. doi: 10.1681/ASN.2018090909 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Blondel VD, Guillaume J-L, Lambiotte R, Lefebvre E. Fast unfolding of communities in large networks. J Stat Mech. 2008;(2008):p10008. doi: 10.1088/1742-5468/2008/10/P10008 [DOI] [Google Scholar]
- 18.Lancichinetti A, Fortunato S. Community detection algorithms: a comparative analysis. Phys Rev E Stat Nonlin Soft Matter Phys. 2009;80(5 pt 2):056117. doi: 10.1103/PhysRevE.80.056117 [DOI] [PubMed] [Google Scholar]
- 19.McInnes L, Healy J, Melville J. UMAP: uniform manifold approximation and projection for dimension reduction. arXiv. Preprint revised September 18, 2020. https://arxiv.org/abs/1802.03426v3
- 20.Richards S, Aziz N, Bale S, et al. ; ACMG Laboratory Quality Assurance Committee . Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med. 2015;17(5):405-424. doi: 10.1038/gim.2015.30 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Kalia SS, Adelman K, Bale SJ, et al. Recommendations for reporting of secondary findings in clinical exome and genome sequencing, 2016 update (ACMG SF v2.0): a policy statement of the American College of Medical Genetics and Genomics. Genet Med. 2017;19(2):249-255. doi: 10.1038/gim.2016.190 [DOI] [PubMed] [Google Scholar]
- 22.Martin AR, Williams E, Foulger RE, et al. PanelApp crowdsources expert knowledge to establish consensus diagnostic gene panels. Nat Genet. 2019;51(11):1560-1565. doi: 10.1038/s41588-019-0528-2 [DOI] [PubMed] [Google Scholar]
- 23.Landrum MJ, Lee JM, Benson M, et al. ClinVar: improving access to variant interpretations and supporting evidence. Nucleic Acids Res. 2018;46(D1):D1062-D1067. doi: 10.1093/nar/gkx1153 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Landrum MJ, Lee JM, Riley GR, et al. ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res. 2014;42(database issue):D980-D985. doi: 10.1093/nar/gkt1113 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Stenson PD, Ball EV, Mort M, et al. Human Gene Mutation Database (HGMD): 2003 update. Hum Mutat. 2003;21(6):577-581. doi: 10.1002/humu.10212 [DOI] [PubMed] [Google Scholar]
- 26.Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst. 1959;22(4):719-748. [PubMed] [Google Scholar]
- 27.Cochran WG. Some methods for strengthening the common χ2 tests. Biometrics. 1954;10(4):417. doi: 10.2307/3001616 [DOI] [Google Scholar]
- 28.Akinrinade O, Koskenvuo JW, Alastalo T-P. Prevalence of titin truncating variants in general population. PLoS One. 2015;10(12):e0145284. doi: 10.1371/journal.pone.0145284 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Roberts AM, Ware JS, Herman DS, et al. Integrated allelic, transcriptional, and phenomic dissection of the cardiac effects of titin truncations in health and disease. Sci Transl Med. 2015;7(270):270ra6. doi: 10.1126/scitranslmed.3010134 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Schafer S, de Marvao A, Adami E, et al. Titin-truncating variants affect heart function in disease cohorts and the general population. Nat Genet. 2017;49(1):46-53. doi: 10.1038/ng.3719 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Jaiswal S, Fontanillas P, Flannick J, et al. Age-related clonal hematopoiesis associated with adverse outcomes. N Engl J Med. 2014;371(26):2488-2498. doi: 10.1056/NEJMoa1408617 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Jaiswal S, Natarajan P, Silver AJ, et al. Clonal hematopoiesis and risk of atherosclerotic cardiovascular disease. N Engl J Med. 2017;377(2):111-121. doi: 10.1056/NEJMoa1701719 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Genovese G, Kähler AK, Handsaker RE, et al. Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence. N Engl J Med. 2014;371(26):2477-2487. doi: 10.1056/NEJMoa1409405 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Fuster JJ, MacLauchlan S, Zuriaga MA, et al. Clonal hematopoiesis associated with TET2 deficiency accelerates atherosclerosis development in mice. Science. 2017;355(6327):842-847. doi: 10.1126/science.aag1381 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Tayal U, Prasad S, Cook SA. Genetics and genomics of dilated cardiomyopathy and systolic heart failure. Genome Med. 2017;9(1):20. doi: 10.1186/s13073-017-0410-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Cirulli ET, White S, Read RW, et al. Genome-wide rare variant analysis for thousands of phenotypes in over 70,000 exomes from two cohorts. Nat Commun. 2020;11(1):542. doi: 10.1038/s41467-020-14288-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Mars N, Koskela JT, Ripatti P, et al. ; FinnGen . Polygenic and clinical risk scores and their impact on age at onset and prediction of cardiometabolic diseases and common cancers. Nat Med. 2020;26(4):549-557. doi: 10.1038/s41591-020-0800-0 [DOI] [PubMed] [Google Scholar]
- 38.Kim E, Rubinstein SM, Nead KT, Wojcieszynski AP, Gabriel PE, Warner JL. The evolving use of electronic health records (EHR) for research. Semin Radiat Oncol. 2019;29(4):354-361. doi: 10.1016/j.semradonc.2019.05.010 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods
eResults
eFigure 1. UMAP plot of the full dataset colored by cluster membership (A) and predicted ancestry (B)
eFigure 2. Quantile-quantile plot of observed vs. expected p-values for the full cohort (“All”) and Synonymous model
eTable 1. Characteristics of unrelated cases
eTable 2. Phenotypes of unrelated control samples
eTable 3. Main collapsing models and their specific filters
eTable 4. Subgroups and number of unrelated cases analyzed
eTable 5. List of genes used in diagnostic analysis and condition from PanelApp
eTable 6. Cluster sizes for the full cohort (“All”)
eTable 7. Genes with enrichment in cases and a p-value < 10-5 across subgroups and models ranked by p-value (includes only top p-value per gene)
eTable 8. Genes with enrichment in cases and a p-value < 10-5 across subgroups and models ranked by p-value (includes only top p-value per gene) from European-only analysis
eTable 9. Number of diagnostic variants per gene
eTable 10. Diagnostic yield TTN only
eTable 11. Diagnostic yield after removal of TTN variants and TTN variant carriers
eTable 12. Diagnostic yield in patients with <= 50 years of age
eTable 13. Diagnostic yield in Europeans
eTable 14. Number of diagnostic variants and yield per gene for cases and controls
eTable 15. TTN results for more detailed phenotypes of I20-I25 (ischemic heart diseases), I42 (cardiomyopathy), I50 (heart failure), and I70 (atherosclerosis) in UK Biobank Europeans
eTable 16. TTN results for different intersections of phenotypes of I20-I25 (ischemic heart diseases) with I42 (cardiomyopathy) and I50 (heart failure) in UK Biobank Europeans
eReferences