Interpreting Secondary Cardiac Disease Variants in an Exome Cohort

Abstract

Background

Massively parallel sequencing to identify rare variants is widely practiced in medical research and in the clinic. Genome and exome sequencing can identify the genetic cause of a disease (primary results), but can also identify pathogenic variants underlying diseases that are not being sought (secondary or incidental results). A major controversy has developed surrounding the return of secondary results to research participants. We have piloted a method to analyze exomes to identify participants at-risk for cardiac arrhythmias, cardiomyopathies or sudden death.

Methods and Results

Exome sequencing was performed on 870 participants not selected for arrhythmia, cardiomyopathy, or a family history of sudden death. Exome data from 22 cardiac arrhythmia and 41 cardiomyopathy-associated genes were analyzed using an algorithm that filtered results on genotype quality, frequency, and database information. We identified 1367 variants in the cardiomyopathy genes and 360 variants in the arrhythmia genes. Six participants had pathogenic variants associated with dilated cardiomyopathy (n=1), hypertrophic cardiomyopathy (n=2), left ventricular noncompaction (n=1) or long QT syndrome (n=2). Two of these participants had evidence of cardiomyopathy and one had left ventricular noncompaction on ECHO. Three participants with likely pathogenic variants had prolonged QTc. Family history included unexplained sudden death among relatives.

Conclusions

Approximately 0.5% of participants in this study had pathogenic variants in known cardiomyopathy or arrhythmia genes. This high frequency may be due to self-selection, false positives, or underestimation of the prevalence of these conditions. We conclude that clinically important cardiomyopathy and dysrhythmia secondary variants can be identified in unselected exomes.

Introduction

Massively parallel sequencing (MPS), either whole genome sequencing (WGS) or whole exome sequencing (WES) has been successful in facilitating the discovery of mutations for Mendelian disorders and more common conditions such as diabetes. Clinical MPS is also available, allowing thousands of research subjects and increasing numbers of patients to undergo this test. The highly parallel nature of MPS ensures that many genes other than those being targeted are evaluated. As a result, secondary (or incidental) genetic data on patients are rapidly accumulating. The majority of these secondary variants are benign or variants of uncertain significance (VUS). However, a few are expected to be clinically relevant.¹ There is considerable debate regarding the responsibilities of researchers and clinicians in returning secondary genetic variants.^2–4 The challenges are formidable – while recognizing that clinically important genetic results may be present in MPS datasets, researchers and clinicians must develop approaches to dealing with secondary variants that appropriately balance the sensitivity and the false positive rates of this testing.

We previously evaluated secondary variants in cancer susceptibility genes, showing that the detection of these mutations is practicable albeit challenging.¹ As part of our continuing efforts to address the issue of secondary variants, we selected 22 arrhythmia and 41 cardiomyopathy-associated genes for study. We selected these phenotypes because they have well-delineated diagnostic criteria, effective treatments are available, and causative mutations have been identified.

To study the feasibility and utility of identifying highly penetrant secondary arrhythmia and cardiomyopathy-associated gene variants from WES data, we analyzed 870 participants who were not pre-selected for a personal or family history of arrhythmia, cardiomyopathy or sudden cardiac death (SCD). We analyzed the WES data for variants in these genes, filtered them and evaluated the phenotypes and family histories of those participants with predicted pathogenic variants. We present our approach, findings, and challenges in interpretation using available genetic and phenotypic data coupled with information from databases and literature for the identification of highly penetrant arrhythmia and cardiomyopathy-associated gene variants in an unselected cohort.

Patients and Methods

Study Participants

Selection criteria for ClinSeq^® participants have been described.⁵ Participants between 45 and 65 years were consented for initial phenotyping, which included clinical laboratory tests,⁵ ECG, ECHO, MDCT coronary Ca, WES or WGS, and return of results. Participants were selected with a spectrum of coronary artery disease (CAD) risk (10-year Framingham risk <5%, 5–10%, >10%, and a group with known CAD). These participants were not selected for a history of arrhythmia, cardiomyopathy or family history of SCD. The NHGRI IRB reviewed and approved this study.

Gene List

We developed a list of 41 cardiomyopathy-associated genes (Supplemental Table 1, arrhythmogenic right ventricular cardiomyopathy/dysplasia [ARVC/D], dilated cardiomyopathy [DCM], hypertrophic cardiomyopathy [HCM], left ventricular noncompaction [LVNC]) and 22 heritable cardiac arrhythmia genes (Supplemental Table 2, atrial fibrillation, Brugada syndrome [BS], catecholaminergic polymorphic ventricular tachycardia [CPVT], long-QT syndrome [LQTS], short-QT syndrome [SQTS]). This list comprises conditions inherited in an autosomal dominant pattern with the exception of CASQ2-associated CPVT, which is inherited in an autosomal recessive pattern. Cardiomyopathy-associated genes attributed to metabolic or developmental syndromes were excluded.

Next-Generation sequencing and variant analysis

DNA isolation, library preparation, sequencing, alignment, base calling, and filtering for sequence and base-calling metrics were performed as described.¹ Reads were aligned to hg18 (NCBI build 36) with Casava 1.8 ELAND (Illumina) and cross_match (http://www.phrap.org/phredphrapconsed.html). Genotypes were determined with Most Probable Genotype (MPG).⁶ Our goal was to identify variations highly likely to be pathogenic while minimizing false positives; therefore sensitivity was sacrificed. We analyzed nonsense, frameshift, splice-site, and non-synonymous variants in the above-listed genes in 870 participants. Variants were graded using a modified version of an established scale¹ from class 0 (poor quality) to class 5 (pathogenic). Variants that failed quality filters were designated class 0. Our goal was to identify rare, highly penetrant disease-susceptibility alleles rather than common alleles with modest disease liability. We did not look for gene-gene interactions from more common alleles (minor allele frequency [MAF] ≥0.015) as the size of our cohort (n=870) limited the power to detect additive effects. We reasoned that no single allele with a frequency of >0.01 in ClinSeq^® or dbSNP MAF of >0.015 (dbSNP⁷ build 132, minimum 120 chromosomes) could cause a disorder with a prevalence of 1/500 and designated these class 1.

The most common of the conditions analyzed here were DCM and HCM with a frequency of ~1/500.^{8, 9} We used the allele frequency from ClinSeq^® and the NHLBI Exome Sequencing Project ([ESP], allele frequencies downloaded 11/24/12) to set a threshold such that any single variant that occurred at a frequency greater than the published estimates for the disease prevalence (>1/2000 BS,¹⁰ >1/10,000 CPVT,¹¹ >1/2500 LQTS,¹² >1/1000 ARVC,¹³ >1/500 HCM,¹² >1/500 DCM^{8, 14, 15}), was likely not pathogenic (class 2) irrespective of any prior pathogenic classification (Figure 1). For all other variants, the Human Gene Mutation Database (HGMD¹⁶) and locus-specific databases (LSDB^{17, 18}) were consulted, relevant publications were reviewed and variants were assigned to pathogenicity classes according to the criteria in Table 1. Variants were designated class 2 (likely benign) if they had been reported multiple times as benign or multiple lines of evidence argued against pathogenicity. Evidence against pathogenicity included: comparable frequency in cases and controls, co-occurrence with a known pathogenic mutation, and/or normal functional data. Novel missense or in-frame insertion/deletion variants were assigned to class 3 (variant of uncertain significance). Missense and in-frame insertion/deletion variants reported a single time as pathogenic without supporting evidence or multiple times with evidence against pathogenicity and loss of function ([LOF], i.e. nonsense, frameshift or splice-site) variants reported once with single evidence against pathogenicity or multiple times with multiple evidence against pathogenicity were also assigned to class 3. For novel nonsense, frameshift and splice-site variants, the characteristics of the gene and the variant, and the participant’s family history, were considered. Variants were designated class 4 (likely pathogenic) if there was a single case reported as pathogenic with supporting evidence (i.e., segregation, absent in controls, functional studies), two cases reported as pathogenic without additional evidence for or against pathogenicity, or ≥ three cases without sufficient race-matched control data to exclude a high population frequency. Class 5 was assigned when two cases with additional supporting evidence were presented or ≥ three cases were reported as pathogenic without evidence against pathogenicity and with sufficient race-matched control data to exclude a high population frequency. Two investigators analyzed variants and assigned a consensus pathogenicity score. Variants are described by their predicted protein changes in the text, but both the cDNA and predicted protein changes are described in the tables.

Figure 1.

Framework for Variant Interpretation. Variants were filtered on genotype quality, coverage and allele frequency. Variants occurring at a frequency greater than the disease prevalence were designated class 2. Remaining variants were assigned pathogenicity scores based on data in HGMD and LSDBs (Table 1). dbSNP, Single Nucleotide Polymorphism Database; ESP, NHLBI Exome Sequencing Project; HGMD, Human Gene Mutation Database; LSDB, locus-specific database; MAF, minor allele frequency; MPG, most probable genotype score.

Table 1.

Criteria for gene variant assignment of pathogenicity class 1 to 5

Designation	Novel	Novel	Pathogenic	Pathogenic	VUS	Benign
Mutation Type	Missense In-frame insertion/deletion	Nonsense Frameshift Splice	Missense In-frame insertion/deletion	Nonsense Frameshift Splice	Any	Any
Class 5 (Pathogenic)		ClinSeq®/ESP MAF ≤ disease prevalence AND sufficient race-matched control data AND
		Multiple LOF reported as pathogenic, consistent FHx	≥ 3 reported cases as pathogenic and no evidence against or 2 cases with supporting evidence	Multiple LOF reported as pathogenic and no evidence against
Class 4 (Likely pathogenic)		ClinSeq®/ESP MAF ≤ disease prevalence AND
		Multiple LOF mutations reported as pathogenic, equivocal FHx	Single reported case as pathogenic with supporting evidence OR 2 reported cases as pathogenic and no evidence against OR ≥ 3 reported cases as pathogenic and no evidence against, but with insufficient race-matched control data	No other LOF mutation reported as pathogenic and no evidence against
Class 3 (Uncertain)	CS®/ESP MAF ≤ disease prevalence AND
	Novel missense or in-frame insertions, deletions	No LOF mutations reported as pathogenic OR inconsistent FHx	Single reported case as pathogenic and no supporting evidence OR multiple reported cases as pathogenic with evidence against	Single reported case as pathogenic with single evidence against OR multiple reported cases as pathogenic and multiple evidence against	Single reported case as VUS	Single reported case as benign with no supporting evidence
Class 2 (Likely not pathogenic)	ClinSeq®/ESP MAF > disease prevalence OR
			Single reported case as pathogenic with multiple evidence against	Single reported case as pathogenic with multiple evidence against	Multiple evidence against pathogenic	Multiple cases reported as benign with no supporting evidence OR single report case as benign with supporting evidence
Class 1 (Benign)	MAF > 0.01 in ClinSeq® or ESP or > 0.015 in dbSNP

Variants that passed quality and frequency filters were assigned to pathogenicity classes based on data available in HGMD, locus-specific databases, and family history (Figure 1). FHx, family history; ESP, National Heart Lung Blood Institute Exome Sequencing Project; LOF, loss of function; MAF, minor allele frequency; VUS, variants of unknown significance.

Statistical analyses

Participants with arrhythmia- or cardiomyopathy-associated rare variants (classes 4–5) were grouped, respectively and compared to participants without identifiable class 3–5 arrhythmia or cardiomyopathy-associated variants. Due to the high number of variants in the cardiomyopathy data set, there were no controls with homozygous wildtype genotype across all class 3–5 variant positions. Since the majority of TTN variants are unclassified, class 3 TTN variants (n=366) were not considered in selecting cardiomyopathy controls (cardiomyopathy control group I, n=67). Cardiomyopathy-associated rare variants were also compared against all individuals with class 1–3 variants (cardiomyopathy control group II, n=838). Fisher’s Exact test was used to compare categorical variables and Mann-Whitney U test was used for group comparisons of continuous clinical variables. Calculations were performed on GraphPad Instat v.3.1.a and presented as mean ± SD. Multiple linear regression was used to compute effect estimates of independent variables (age, BMI, history of hypertension, race, sex, use of anti-hypertensive medications) on ECG and ECHO parameters, identify confounding covariates and calculate adjusted means. The leaps and effects packages (R software v.2.15.2) were used for linear model analyses and selection. Results are shown as mean ± SEM. A 2-tailed P-value of <0.05 was considered significant. A Benjamini-Hochberg correction was applied to decrease false positives due to multiple testing. Multi-dimensional scaling in PLINK was used to assess population structure.

Results

Demographics

There were 870 individuals analyzed in this study. Most were Caucasian (89%) and not Hispanic (96.3%). The median age at consent was 57 years (range 45 to 65). There was a slight preponderance of males (50.3%).

Sequence data

For the 870 exomes, 68.2 billion reads were generated resulting in 6.13 trillion bp of sequence and 2,452,318 unique variants. Copy number variants and indels >10 bp were not assessed. A total of 244,541 variants were nonsynonymous, frameshift, nonsense, or splicing. The 41 cardiomyopathy-associated genes comprised 185,997 bp of targeted coding sequence. Average gene coverage ranged from 51.8% for (MYBPC3) to 100% (PLN) with a median of 97.6%. The 22 arrhythmia-associated genes comprised 72,984 bp of targeted coding sequence. Average gene coverage ranged from 60.3% (HCN4) to 99.9% (CACNB2, KCNE2, KCNE3) with a median of 97.4% (Figure 2).

Figure 2.

Whole exome gene coverage. A. Cardiomyopathy-associated genes. B. Arrhythmia-associated genes. Box and whisker plots showing base coverage for each gene across 870 probands. The bottom and top lines of the box represent the upper bounds of the first and third quartiles respectively; the mid-line represents the median. The bottom and top whiskers represent the lowest and highest values within 1.5 times the interquartile range. Outliers have been excluded. The y-axis represents the fraction of total coding bases covered by a high quality genotype call.

Variant classification

There were 1367 variants identified in the 41 cardiomyopathy-associated genes (no variants were identified for ACTC1 or MYL3, Supplemental Table 1) and 360 variants in the 22 arrhythmia-associated genes (Supplemental Table 2). Fifteen variants (cardiomyopathy n=9, arrhythmia n=6) did not meet the quality metrics and were designated score 0 (Figure 3). There were 440 variants designated class 1 (cardiomyopathy n=363, arrhythmia n=77) due to frequency (>0.01 ClinSeq^® or NHLBI ESP and >0.015 dbSNP) (Figure 1). There were 380 variants designated class 2 (cardiomyopathy n=282, arrhythmia n=98) due to allele frequency in ClinSeq^® or NHLBI ESP exceeding the disease prevalence. The remainder of the variants (cardiomyopathy n=713, arrhythmia n=179) were scored based on information in HGMD and/or an LSDB. Variants listed as a “polymorphism” in HGMD or “benign polymorphism” in an LSDB were scored class 2 after literature review (cardiomyopathy n=4, arrhythmia n=5). Of the remaining variants, 832 were scored class 3 ([uncertain], cardiomyopathy n=677, arrhythmia n=155, Figure 3) due to no publications or no HGMD/LSDB entry or the predicted protein change was in a transcript other than the HGMD reference or single case reports without supporting evidence or with conflicting evidence. Forty-five variants were scored class 4 ([likely pathogenic], cardiomyopathy n=28, arrhythmia n=17) based on a single reported case with supporting evidence or two reported cases with no evidence against or have a predicted LOF variant when LOF mutations have been reported as causative, but the proband’s family history was equivocal or if there were insufficient race-matched control data. Six variants were designated class 5 ([pathogenic], cardiomyopathy n=4, arrhythmia n=2, Figure 3) based on three or more reported cases of pathogenicity and no evidence against or two reported cases with additional supporting evidence and sufficient race-matched control data are available. Details of the scoring are available (Supplemental Tables 1–4). All reported class 4–5 variants were confirmed by Sanger sequencing.

Figure 3.

Summary of variants by pathogenicity class. A. Distribution of 1367 cardiomyopathy-associated variants by pathogenicity class. B. Distribution of 360 arrhythmia-associated variants by pathogenicity class.

Individuals with class 5 cardiomyopathy-associated variants

Four cardiomyopathy-associated variants identified in one participant each (MYBPC3 p.Gly490Arg, MYBPC3 p.Arg495Gln, MYH7 IVS8+1G>A, PLN p.Leu39X) were scored class 5 (Table 2).

Table 2.

Summary of ClinSeq^® participants with cardiomyopathy-associated pathogenic class 5 variants

Gene	cDNA reference, predicted protein/splice alteration	Path score	ID Age Sex Race Ethnicity	QTc (mS)	ECHO septum (mm) nl 6–11	ECHO LVEF (%)nl 55–74	ECG/ECHO comments	HTN	Family history
MYBPC3	NM_000256.3 c.1468G>A p.Gly490Arg	5	182262 56 Male Cauc NH	388	12	65	Nl/mild asymmetric basal septal, hypertrophy aortic root mildly dilated	No	Proband palpitations, sister cardiomyopathy, paternal cousin d. 68y CHF, paternal uncle d. 2y
MYBPC3	NM_000256.3 c.1484G>A p.Arg495Gln	5	135629 62 Male Cauc NH	417	11	57	SB57, LAFB/aortic root, AA, RA, RV are mildly dilated, nl RV fxn, no pulmonary HTN	No	Proband A-fib, mild concentric LVH, sister A-fib x1y, mother A-fib
MYH7	NM_000257.2 c.732+1G>A IVS8+1G>A	5	120682 62 Female Cauc NH	431	8	60	1AVB/prominent trabeculations LV apex, nl LV fxn, mild AR, no pulmonary HTN	Yes	Proband LVNC on cardiac MRI, mild LAE, daughter resting heart rate in 90’s
PLN	NM_002667.3 c.116T>G p.Leu39X	5	114451 57 Female Cauc NH	398	6	60	Nl/mild-moderate AR	No	Brother CAD & stent 61y, father MI 61y, defibrillator & pacemaker 70’s

1AVB, first degree AV block; AA, ascending aorta; A-fib, atrial fibrillation; AR, aortic regurgitation; CAD, coronary artery disease; Cauc, Caucasian; CHF, congestive heart failure; d., died; fxn, function; HTN, hypertension; LAFB, left anterior fascicular block; LV, left ventricle; LVH, left ventricular hypertrophy; LVEF, left ventricular ejection fraction; LVNC, left ventricular noncompaction; MI, myocardial infarction; MRI, magnetic resonance imaging; NH, Non-Hispanic; Nl, normal; Path, pathogenicity; RA, right atrium; RV, right ventricle; SB, sinus bradycardia (followed by the heart rate).

Participant 182262 had MYBPC3 p.Gly490Arg, a missense mutation reported in three individuals with three different phenotypes (DCM¹⁹, HCM²⁰ and LVNC²¹). He was a 56-year old Caucasian male with a normal ECG and mild asymmetric basal septal hypertrophy and normal left ventricular ejection fraction (LVEF 65%) on ECHO. He had a history of elevated cholesterol, hyperuricemia, and renal cancer diagnosed at age 55, but no hypertension. His family history was significant for maternal grandparents dying from congestive heart failure (CHF), a sister diagnosed with cardiomyopathy in her 60’s, a paternal cousin who died of CHF at age 68 and a paternal uncle who died at age 2. MYBPC3 p.Gly490Arg was reported in NHLBI ESP with a MAF of 3/8476.

Participant 135629 had MYBPC3 p.Arg495Gln, a missense mutation cosegregating with HCM in one family,²² reported in one sporadic HCM²⁰ and associated with left ventricular hypertrophy (LVH) in asymptomatic relatives.²³ He was a 62-year old Caucasian male who had sinus bradycardia and left anterior fascicular block on ECG and a mildly dilated right atrium and ventricle with normal LVEF (57%) on baseline ECHO. He was subsequently diagnosed with asymptomatic atrial fibrillation (A-fib) on pre-op ECG and mild concentric LVH in 2012. His sister developed intermittent A-fib in her 60’s and his mother had “severe” A-fib. MYBPC3 p.Arg495Gln was not identified in NHLBI ESP.

Participant 120682 has MYH7 IVS8+1G>A, a splice variant that has been reported to co-segregate with LVNC in two families.²⁴ She was a 62-year old Caucasian female with first degree AV-block on ECG and prominent trabeculations at the left ventricular (LV) apex with normal LVEF (60%) on ECHO. Follow-up cardiac MRI (2008, 2012, Figure 4) confirmed LVNC changes with normal LVEF (60%). Her family history was negative for cardiomyopathy or CHF, one daughter was noted to have a persistent high resting pulse (90’s). MYH7 IVS8+1G>A was not identified in NHLBI ESP.

Figure 4.

Cardiac MRI. A. Four-chamber view of a normal heart. B. Four-chamber view of heart from participant 120682 with hypertrabeculation and increased noncompaction.

Participant 114451 had PLN p.Leu39X, a loss of function mutation identified in three unrelated families with DCM.²⁵ She was a 57-year old Caucasian female with a normal ECG, thickened aortic valves, moderate aortic regurgitation and normal LVEF (60%) on ECHO. She had a brother with CAD and stent placement at age 61, a father with a myocardial infarction (MI) and placement of a cardiac defibrillator and dual chamber pacemaker at age 61. PLN p.Leu39X was not identified in the NHLBI ESP.

See Supplemental Table 5 for phenotypic summary of class 4 cardiomyopathy variants.

Class 5 arrhythmia-associated variants

Two variants were scored class 5 in the 22 arrhythmia-associated genes (Table 3).

Table 3.

Summary of ClinSeq^® participants with arrhythmia-associated pathogenic class 5 variants

Gene	cDNA reference, predicted protein/splice alteration	Path score	ID Age Sex Race Ethnicity	QTc (mS)	ECHO septum (mm) nl 6–11	ECHO LVEF (%)nl 55–74	ECG/ECHO comments	HTN	Family history
KCNE1	NM_000219.3 c.292C>T p.Arg98Trp	5	155279 62 Female Cauc NH	438	10	65	Nl/Nl	No	Sister palpitations
KCNH2	NM_000238.3 c.934C>T p.Arg312Cys	5	173996 55 Male Cauc NH	402	10	60	SB56, 1AVB/Impaired LV diastole, moderate MR, mild AR	No	Proband palpitations, 2 paternal aunts & 1 uncle d. 70’s unknown cause

1AVB, first degree AV block; Cauc, Caucasian; d., died; LV, left ventricle; LVEF, left ventricular ejection fraction; NH, Non-Hispanic; Nl, normal; Path, pathogenicity; SB, sinus bradycardia (followed by the heart rate).

Participant 155279 had KCNE1 p.Arg98Trp. She was a 62-year old Caucasian female with a normal ECG and ECHO. Her family history included a father who died of an MI at age 45, a sister who had two MIs and valve surgery before age 70 and another sister with stress associated palpitations. KCNE1 p.Arg98Trp was identified in two individuals diagnosed with LQTS^{26, 27} and not identified in the NHLBI ESP. Functional studies of this variant showed a positive shift in voltage dependent activation of the slow potassium channel (I_Ks) and defects in trafficking KCNQ1.²⁸

Participant 173996 had KCNH2 p.Arg312Cys. He was a 55-year old Caucasian male with sinus bradycardia (not on beta blockers) and first degree AV block, moderate mitral regurgitation and impaired left ventricular relaxation on ECHO. He had a history of an episode of palpitations without syncope and a family history of two paternal aunts and one uncle who died suddenly in their 70’s from unknown causes. KCNH2 p.Arg312Cys was identified in three unrelated individuals, two with LQTS,^{26, 29} one referred for LQTS genetic testing³⁰ and was not identified in the NHLBI ESP.

See Supplemental Table 6 for phenotypic summary of class 4 arrhythmia variants.

Comparison of baseline phenotypic data

Individuals with class 4–5 arrhythmia-associated variants were assessed as a group (n=19) for comparison of age, sex, race, history of hypertension, baseline ECG and ECHO measurements to participants who did not have identifiable class 3–5 variants in the respective arrhythmia-associated genes (n=42). Individuals with class 4–5 cardiomyopathy-associated variants (n=32) were compared to two control groups as described. The groups were similar in regards to age, sex and race. There was a lower percentage of individuals with a history of hypertension in the arrhythmia cases and a higher percentage in the cardiomyopathy cases compared to controls, but the difference was not significant. There were no significant differences in QTc interval, septal thickness, left ventricular mass or ejection fraction in the respective cases and controls both before and after adjustment for covariates (data not shown).

Discussion

We piloted a method to identify rare, high penetrant cardiomyopathy- and arrhythmia-associated alleles. Analysis of WES from 870 participants not selected for a clinical phenotype or family history of arrhythmia or cardiomyopathy identified six individuals with class 5 (pathogenic) variants. Three of these individuals had subclinical evidence of cardiomyopathy at enrollment in ClinSeq^® and would not otherwise have sought medical evaluation. Consensus recommendations for asymptomatic, mutation-positive individuals are available,⁹ but there are still many areas where more research is needed to formulate recommendations.

Individual 182262 (MYBPC3 p.Gly490Arg) had asymptomatic asymmetric basal septal hypertrophy when he enrolled in ClinSeq^® at age 56. On follow-up, he related intermittent caffeine-associated palpitations (x2 years), but no dyspnea on exertion (DOE), chest pain, or syncope. His family history is significant for a sister developing HCM in her late 60’s. Individuals with MYBPC3 mutations have been reported to display age-dependent penetrance for HCM.²² There are no specific recommendations for individuals with incidental HCM susceptibility variants. We applied American Heart Association (AHA) recommendations for genotype-positive/asymptomatic individuals with HCM mutations identified through family studies to our participants. Recommendations for participant 182262 include: 1) periodic assessment with ECG and transthoracic ECHO (TTE) to screen for dysrhythmia and LV dysfunction; 2) his palpitations should be evaluated with a 24-hour Holter to screen for ventricular tachycardia (VT); 3) optimal management of his hyperlipidemia and maintenance of normal blood pressure and weight; 4) cascade genetic testing for his first-degree relatives.⁹ We are applying those recommendations to this participant, although we recognize those recommendations did not consider individuals identified by WES.

Participant 135629 (MYBPC3 p.Arg495Gln) had an asymptomatic dilated right atrium and ventricle at age 62. At follow-up, he reported DOE at age 66. Pre-op ECG showed new onset A-fib and ECHO showed mild concentric LVH and decreased LVEF (50%). AHA recommendations for him include: 1) anticoagulation with warfarin for stroke prevention; 2) annual ECG to monitor for asymptomatic progressive conduction block or rhythm changes; 3) TTE every 1–2 years to reassess the degree of HCM, and monitor for left ventricular outflow tract obstruction; 4) initiation of cascade genetic testing for his first-degree relatives.⁹

Participant 120682 (MYH7 IVS8+1G>A) was asymptomatic when she enrolled in ClinSeq^® at age 62. She has no complaints of DOE, palpitations or fainting on follow-up at age 66. There are no set recommendations for asymptomatic individuals with LVNC. Monitoring is tailored to individual symptoms.³¹ The consulting cardiologist recommended repeat cardiac MRI or ECHO every two years to reassess the severity of the aortic regurgitation and left ventricular function. We recommended genetic testing of her first-degree relatives.

These data show that a focused analysis of MPS data can successfully identify incidental, medically relevant genetic variants that may benefit research participants. This genomics-first approach is a major paradigm shift for medical practice as it changes the focus from diagnosis and treatment of manifest disease to identification, monitoring and early treatment aimed at disease modification for asymptomatic individuals with genetic susceptibility. The potential benefit of early detection will require clinical studies and long-term follow-up to establish evidence based effective monitoring and prophylaxis.

The three remaining individuals (PLN p.Leu39X, KCNE1p.Arg98Trp, KCNH2 p.Arg312Cys) were non-penetrant. Long term follow-up of these individuals and family studies may help to elucidate genetic and environmental factors that influence incomplete penetrance and variable expressivity.

The ClinSeq^® frequency of 2/870 for LQTS associated class 5 variants is above the reported prevalence of 1/2,500 for LQTS.¹² We suggest four possible explanations for the high observed frequency of LQTS variants: First, participants may have self-selected when enrolling in ClinSeq^®. Among the 19 participants with class 4 or 5 arrhythmia-associated variants, eight had mild bradycardia, one developed LBBB without evidence of CAD in middle age, while several of the participants had family histories of sudden death that might have lead them to enroll in the study. Second, it is possible that the true prevalence of LQTS is underestimated. Third, it is possible that the penetrance of these disorders is lower than hitherto appreciated and these individuals are (to date) non-penetrant. Finally, it is possible that some of these variants are actually benign and that the literature supporting their pathogenicity is incorrect. There are no formal standards for classifying a variant as causative and the genetics literature and mutation databases contain substantial numbers of incorrect assignments of pathogenicity.^{32, 33} Data from cardiomyopathy and dysrhythmia disease registries are starting to fill in this void by providing estimates of the frequency of rare variants in healthy controls,³¹ but additional criteria are needed to support or refute pathogenicity such as in vitro functional studies or long-term follow-up of healthy controls.

Deciding which secondary variants to return remains a challenge. We suggest following NHLBI working group guidelines³⁴ to determine which variants to return. NHLBI criteria for return of genetic results include: a. variant has established, important health implications, b. the genetic finding is actionable, c. testing is analytically valid and d. the participant has consented to receive his/her genetic results. Additionally, the American College of Medical Genetics and Genomics (ACMG) has released recommendations for the reporting of incidental findings from clinically indicated (i.e., not research) sequencing (www.acmg.net). Their recommended minimum gene list to be considered for return of incidental findings includes a number of disorders studied here. Even in asymptomatic individuals, we suggest that there are potential clinical benefits for returning class 5 variants that can be derived from the individual’s increased awareness of changes in symptomatology and enhanced medical surveillance. Our approach is similar to proposals for using genetic data to identify asymptomatic family members with LQTS mutations who should be followed and medically managed if indicated³⁵ and ACCF/AHA guidelines recommending periodic clinical assessment for genotype-positive/asymptomatic individuals with pathogenic mutations who do not express the HCM phenotype.⁹

These variants can be considered secondary (or incidental) findings because neither we, nor the patients were seeking genetic test results for these traits at the time the patients were consented. The secondary variants issue is being debated in both the research and clinical testing realms and has spawned efforts to develop policies to handle such results.^{4, 34} These policies support the notion that secondary findings of high medical importance should be sought and returned to research participants by a team of experts supported by a genetic counselor. Others however disagree and suggest that individual clinical results should not be returned to research participants, even when medically compelling, to prevent therapeutic misconception and avoid stressing research resources. Because we are studying the impact of returning results in ClinSeq^®, we have opted to returned class 5 and some class 4 (i.e., those with abnormal findings or suspicious family history) variants to our participants. We recognize that other clinical researchers may reach different conclusions based on their study designs and IRB protocols. In the clinical realm, the ACMG recommendations are controversial and will undoubtedly evolve, but are a starting point for developing an approach to incidental variants. Among the six class 5 variants we detected, four met the current ACMG incidental findings reporting recommendations (MYBPC3 p.Gly490Arg, MYBPC3 p.Arg495Gln, MYH7 IVS8+1G>A, and KCNH2 p.Arg312Cys), but two of our class 5 variants are in genes not on the ACMG list (PLN p.Leu39X and KCNE1 p.Arg98Trp). As in all areas of medicine, clinicians must use their judgment in interpreting and returning incidental findings. The data presented here do not lead to specific conclusions supporting the return of results in the research or clinical setting, but do provide data on the feasibility and yield of such an analysis.

In the clinical setting, identifying apparently pathogenic gene variants in an asymptomatic individual will require a multi-step process to determine clinical significance and appropriate medical management. Research on the clinical utility and implementation of individual genomic information must be supplemented by additional clinical evaluation focused on the disease in question. This iterative evaluation of asymptomatic individuals with potentially pathogenic variants will be important as studies have shown that family members with normal QTc and pathogenic arrhythmia variants have a 4% risk of developing an aborted cardiac arrest or SCD³⁵ and individuals with MYBPC3 mutations and no HCM have evidence of impaired myocardial energetics³⁶ illustrating that an asymptomatic individual does not necessarily have a benign clinical outcome.

Estimating disease risk from an individual’s genetic profile for disease prevention and treatment is the objective of individualized medicine. Although genomic medicine may appear within reach, the development and fine-tuning of the analyses and clinical interpretation of sequencing data remains a major hurdle. One challenge in the interpretation of WES is to minimize false positives. This is not a trivial problem as the probability of false positive findings increases substantially in an asymptomatic cohort. Single gene mutation screening for cardiomyopathy or dysrhythmia is not indicated in the general population due to the large number of people identified with VUS.³¹ Although we support recommendations that proscribe gene screening for LQTS or DCM variants in the general population, this must be considered as distinct from the secondary variant issue. An appropriate analogy is the chest radiograph – while it is inappropriate to screen the general population for lung cancer,³⁷ all radiologists report lung lesions noted incidentally when a chest radiograph is performed, irrespective of the indication, while recognizing that most such masses are not cancerous.³⁸ Our challenge is to improve the positive predictive value and sensitivity of genomic screening so that useful medical predictions can be generated.

Limitations of this study include decreased sequence coverage of MYBPC3 due to the capture kit used on the initial 391 DNA samples not targeting this gene. Hence, there may be additional MYBPC3 disease-associated variants present in our cohort of 870 that were not picked up by WES. ClinSeq^® participants are not representative of the general population, as a group they have above average income, a high level of education and are highly motivated to participate in research.³⁹ Another source of potential bias is that some of our participants may have self-selected for enrollment based on their personal or family histories, hence the frequency of pathogenic variants may not be representative of the general population.

In summary, we analyzed 870 WES to identify pathogenic arrhythmia and cardiomyopathy-associated gene variants using an algorithm that filtered results based on genotype quality, frequencies, published data from databases and identified six participants possessing pathogenic variants for DCM, HCM and LQTS. As with other Mendelian disorders, familial arrhythmias and cardiomyopathies exhibit variable expressivity,⁴⁰ reduced penetrance,⁴¹ environmental⁴² and genetic modifiers⁴³ that pose challenges to the interpretation of gene variants. As such, these disease models serve as a paradigm to highlight the challenges in translating genomic sequence data into evidence-based medicine. Yet, these data suggest that significant numbers of patients, in this case approximately 0.5% of the participants, have a putative pathogenic secondary variant in a gene for these disorders.