Large next-generation sequencing gene panels in genetic heart disease: yield of pathogenic variants and variants of unknown significance

Background Genetic heterogeneity is common in inherited cardiac diseases. Next-generation sequencing gene panels are therefore suitable for genetic diagnosis. We describe the results of implementation of cardiomyopathy and arrhythmia gene panels in clinical care. Methods We present detection rates for variants with unknown (class 3), likely (class 4), and certain (class 5) pathogenicity in cardiogenetic gene panels since their introduction into diagnostics. Results In 936 patients tested on the arrhythmia panel, likely pathogenic and pathogenic variants were detected in 8.8% (4.6% class 5; 4.2% class 4), and one or multiple class 3 variants in 34.8%. In 1970 patients tested on the cardiomyopathy panel, likely pathogenic and pathogenic variants were detected in 19.8% (12.0% class 5; 7.9% class 4), and one or multiple class 3 variants in 40.8%. Detection rates of all different classes of variants increased with the increasing number of genes on the cardiomyopathy gene panel. Multiple variants were detected in 11.7% and 28.5% of patients on the arrhythmia and cardiomyopathy panels respectively. In more recent larger versions of the cardiomyopathy gene panel the detection rate of likely pathogenic and pathogenic variants only slightly increased, but was associated with a large increase of class 3 variants. Conclusion Overall detection rates (class 3, 4, and 5 variants) in a diagnostic setting are 44% and 61% for the arrhythmia and cardiomyopathy gene panel respectively, with only a small minority of likely pathogenic and pathogenic variants (8.8% and 19.8% respectively). Larger gene panels can increase the detection rate of likely pathogenic and pathogenic variants, but mainly increase the frequency of variants of unknown pathogenicity. Electronic supplementary material The online version of this article (10.1007/s12471-019-1250-5) contains supplementary material, which is available to authorized users.


Introduction
Next-generation sequencing (NGS) techniques with sequencing and analysis of multiple genes in a single experiment are being used more frequently in a diagnostic setting in monogenic diseases. Examples are whole exome sequencing (WES) and disease-specific gene panels which can be complemented by Sanger sequencing to cover the entire coding region of the specified genes. Gene panels are particularly well suited in genetic heart diseases, because of their phenotypic and genetic heterogeneity; a single disease associated with different genes, and a single gene associated with different diseases [1].
Testing more genes using a gene panel not only increases the detection rate of disease-causing variants, but also of so-called variants of uncertain/unknown significance (VUS). VUS are variants whose implication with disease cannot be concluded based on current data. With time and further knowledge, VUS can then be classified as either benign without clinical significance or disease-causing.

Next-generation sequencing (NGS) gene panels
Detailed information on NGS, the platforms used and their analytical performance has been published before and is described in the online Supplementary File 1 [2,3]. Tab. 1 shows the genes in the different versions of the arrhythmia and cardiomyopathy panels.
Variants are reported using Human Genome Variation Society nomenclature guidelines (http://www. hgvs.org/mutnomen) and classified into one of five categories (class 1: certainly not pathogenic, class 2: unlikely pathogenic, class 3: unknown pathogenicity, class 4: likely pathogenic; class 5: (certainly) pathogenic) using the classification criteria as indicated in the online Supplementary File 1. Identified class 1 and 2 variants are not reported. Classification was carried out at the moment the sequence data were analysed using the data available at that time. Reinterpretation and reclassification was only done when the variant was detected in another patient or more information became available from the family or other cases in other clinical genetics centres in the Netherlands.

Statistical analysis
Variables are reported as frequency (%). Categorical and dichotomous variables are compared between groups using the chi-squared test. A two-sided p-

Dutch founder variants
In the Netherlands, several variants have been detected more frequently. For many of these variants haplotype analysis and genealogy suggest the variant originates from an ancient founder. In 100 patients a previously described founder variant was detected and in 81 another recurrent variant (online Supplementary File 3). Founder variants and recurrent variants made up 34.9% of class 5 variants in the arrhythmia panel and 70.3% of class 5 variants in the cardiomyopathy panel.

Variant reclassification
We looked at reclassification of variants that occurred between January 1, 2015 and January 1, 2018, based on reinterpretation. Most variants were not reclassified at reinterpretation. For the arrhythmia panel seven variants were reclassified and for the cardiomyopathy panel 13 (upgrade of pathogenicity class in 9, downgrade in 11). Reasons for reclassification were new information on frequency of the variant in control databases (n = 6), results of RNA studies (n = 5), co-segregation in the family (n = 5) and additional patients/families with the same variant (n = 4) (online Supplementary File 2).

Discussion
Overall detection rate (class 3, 4, and 5 variants) is 44% and 61% for the arrhythmia and cardiomyopathy panels respectively, with only a minority of likely pathogenic and pathogenic variants (8.8% and 19.8% respectively). These yields of likely pathogenic and pathogenic variants are lower than published in pre-NGS literature [5][6][7][8][9] and NGS literature (data≥ 2015: HCM: 21.4% class 4/5 variants in our cohort vs 32% in literature [10]; DCM: 16.3% class 4/5 variants in our cohort vs 37% in literature [11]; Brugada syndrome: 9.5% class 4/5 variants in our cohort vs~20% in literature [7,12]). Even the substantial contribution of Dutch founder variants does not raise detection rates. There are several explanations for this. First, indications for diagnostic DNA testing broadened in the past years. It is no longer only patients meeting diagnostic criteria for disease, such as the well-phenotyped research cohorts from literature, who are being tested. A study of our centre also showed decreasing yields in time explained by patients with less severe phenotypes being tested in more recent years [13], and studies have been published confirming that yield is higher in more severe patients [14]. Increased awareness of genetic heart disease also gives rise to more referrals of patients with a merely suspected personal or family history of disease. An unclear phenotype is a likely explanation for the lower yield of the arrhythmia panel, as, for example, sudden cardiac arrest or sudden cardiac death could also have nonmonogenic and/or non-cardiac causes (Tab. 2). Patients with a clear diagnosis of LQTS or CPVT, for which a small set of genes can be tested with a high pathogenic yield, are underrepresented in the arrhythmia panel cohort. Second, standards for calling a vari-ant pathogenic have become more stringent. The difference between likely pathogenic and pathogenic was almost absent in early years. Any rare evolutionary conserved variant not found in~100 healthy controls was considered pathogenic. Third, inherited cardiac diseases are relatively common and show incomplete and age-dependent penetrance, meaning that also pathogenic variants or modifiers can be detected in control databases and can have a relatively high frequency in the general population. Presence of a variant in control databases therefor does not exclude pathogenicity. Whereas for highly penetrant, rare diseases low-frequency variants are labelled likely benign or benign (class 1 or 2 variant), for cardiac diseases they will be classified as unknown pathogenicity (class 3 variant). Fourth, knowledge on many of the genes on the panels and variant classification in these genes is sparse, because clear disease associations have been made in a few patients. Variants are therefore easily classified as a type 3 variant.
The increasing number of genes on different versions of our cardiomyopathy panel did result in a higher detection rate of likely pathogenic and pathogenic variants, but mainly of class 3 variants. This has been shown previously in literature [10,11,15]. A lower number of tested genes makes interpretation of test results easier, with less VUS, but also increases the risk of missing pathogenic variants. The addition of the TTN gene to the cardiomyopathy panel in 2015 only resulted in an increase of class 4 and 5 variants in the DCM patients, but not in other cardiomyopathy patients. In all patients, however, the addition of TTN resulted in an increase of class 3 variants. This can, however, also be the result of less severe patients being tested in more recent years. Developments in diagnostic DNA testing and increasing knowledge on tolerated DNA variants ask for continuous re-evaluation of which test (single genes or large panel) or which filter used for a genome-wide test is best suited for which patient. Initiatives like ClinGen (www.clinicalgenome.org), which evaluates clinical validity of supposed disease genes, can be of help with this, and possibly patients can make an informed choice in the extent of the DNA test and the given results. Although reclassification was uncommon on reinterpretation in our cohort, updates of classification status of class 3-5 variants should be made regularly, including reporting important reclassifications back to the physician who requested the test. Initiatives that allow sharing of variant data could help in solving the meaning of observed VUS.
Multiple variants were detected in 11.8% and 28.5% of patients on the arrhythmia and cardiomyopathy panels respectively. Some variants will eventually be classified as benign variants, but others will be associated with disease. Although inherited cardiac diseases are considered to follow an autosomal dominant inheritance pattern, the presence of multiple variants in our study, and previous descriptions of multiple rare variants, digenic inheritance, homozygous and compound heterozygous variants, all suggest more complex inheritance patterns and modifier effects [16][17][18][19]. This might also explain the variable disease expression of inherited cardiac diseases.
The high frequency of VUS in cardiogenetic gene panels necessitates pre-test counselling on VUS of tested patients. However, incidental findings can still give rise to challenges in counselling/clinical decision making. In a separate article we describe these challenges in more detail. We recommend pre-test and post-test counselling to be performed by a physician/ counsellor with sufficient knowledge on VUS and variant classification. Challenging cases should be discussed in a multidisciplinary cardiogenetics team.

Conclusions
Overall detection rates for cardiogenetic NGS gene panels in a diagnostic setting are 44% and 61% for the arrhythmia and cardiomyopathy gene panels respectively, with only a small minority of likely pathogenic and pathogenic variants (8.8% and 19.8% respectively). Larger gene panels can increase detection rates of likely pathogenic and pathogenic variants, but mainly increase the frequency of VUS. Test results, especially VUS and incidental findings, can be challenging for genetic counselling and ask for proper pre-test and post-test counselling and evaluation by a multidisciplinary cardiogenetics team.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.