1
Next-Generation Sequencing for Clinical Diagnostics of Cardiomyopathies

Jan Haas, Hugo A. Katus, and Benjamin Meder

1.1 Introduction

The vast progress next-generation sequencing (NGS) has undergone during the past few years [1,2] has opened doors for a more advanced genetic diagnostic for many inherited diseases, such as Miller syndrome or Charcot–Marie–Tooth neuropathy [3,4]. Here, we want to describe the paradigm change in genetic diagnostics using the example of cardiomyopathies.

1.2 Cardiomyopathies and Why Genetic Testing is Needed

Cardiomyopathies are a heterogeneous group of cardiac diseases that can either be acquired through, for example, inflammation (myocarditis), be stress-induced (tako-tsubo), or be due to a genetic cause [5,6]. Examples of genetic forms are hypertrophic cardiomyopathy (HCM), dilated cardiomyopathy (DCM), arrhythmogenic right ventricular cardiomyopathy, and left-ventricular non-compaction cardiomyopathy. Together with the channellopathies, such as long-QT syndrome and Brugada syndrome, they account for the most common heart diseases and belong to the most prevalent causes of premature death in western civilizations [7,8]. A point mutation in exon 13 of the β-myosin heavy chain gene was the first detected mutation diagnosed to be relevant for HCM in 1990 [9]. Driven by this finding, genetic research has progressed tremendously over the past two decades. Mutations in genes coding for a diverse set of proteins (e.g., sarcomeric, cytoskeletal, desmosomal, channel and channel-associated, membrane, and nuclear proteins, but also mitochondrial proteins or proteins relevant for mRNA splicing) have now been found to be implicated in disease onset and progression [10]. With currently more than 90 known disease genes with more than 1000 exons and multiple malign mutations per gene, the disease's heterogeneity is high and poses a challenge for classical Sanger-based sequencing. Although Sanger sequencing is able to detect mutations by testing only the most heavily affected genes, such as the β-myosin heavy chain gene (MYH7), where it is possible to find mutations in up to 30–50% of HCM patients, the mutation frequency in most genes is very low [10–12]. Therefore, new methods were needed to further improve genetic diagnostics in cardiomyopathy patients.

1.3 NGS

In contrast to Sanger sequencing, which is only capable of sequencing a few megabases, NGS is able to sequence hundreds of gigabases per run [13–15]. Currently, the most widely used NGS systems are the “sequencing by synthesis”-based sequencer HiSeq 2000 (Illumina), the “ligation and two-base coding”-based system SOLIDv4 (Life technologies), and the 454 GS FLX (Roche), which relies on “pyrosequencing” technology [7,16]. A detailed comparison of currently used systems including performance benchmarks, such as read lengths and output amounts, has been published recently by Liu et al. [1,2,17]. In addition to the mature NGS systems, so-called benchtop sequencers have emerged. Those instruments, such as the MiSeq (Illumina), the Ion Torrent (PGM), or the GS Junior (Roche), benefit from a significantly shorter run time (hours compared to days) and a lower price, taking into account a reduced amount of sequenced bases [17,18].

Originally, NGS was designed to sequence whole genomes. In order to reduce costs, methods for target enrichment were developed to restrict sequencing to the regions of interest only [19–22]. The sequencing of only selected segments of the genome allowed the sequencing of a larger number of individuals per run [23]. For the enrichment of the target regions, array-based, in-solution based, or polymerase chain reaction (PCR)-based approaches exist [7,24]. Depending on the sequence composition of the target region in terms of, for example, GC content or sequence heterogeneity, the efficiency of the different methods might vary [25]. In recent years, the number of NGS applications has grown considerably. In addition to the target-enrichment methods, which can be either used for custom gene panels or whole-exome sequencing (WES), RNA-Seq methods to study messenger RNA as well as microRNA are now also routinely used. Furthermore, Methyl-Seq or Chip-Seq methods to study DNA methylation are frequently used. It can be assumed that the number of applications will grow further together with the need for more advanced bioinformatics analysis tools [26]. Here, a shift in the cost distribution from sequencing to downstream analysis is expected [27].

1.4 NGS for Cardiomyopathies

As mentioned above, whole-genome sequencing (WGS) is an unbiased approach to determine the exact order of every base in a studied genome. In the case of patient studies, the human genome, consisting of more than 3 billion bases, needs to be analyzed. Although costs have been decreasing dramatically (http://www.genome.gov/sequencingcosts), it is still too expensive to perform WGS with a sufficient coverage for routine diagnostics in cardiomyopathies. Downstream bioinformatics analyses are also much more demanding for WGS compared with WES or partial-exome sequencing (PES), which is preferred to be used instead. Whereas WES is mainly used to discover new disease genes, PES has started to become a standard approach for high-throughput testing of multiple genes and patients. Meder and Haas, for example, applied an array-based enrichment of 47 genes (0.27 Mb) on cardiomyopathy patients, and were able to identify disease-causing mutations in both HCM (80%) and DCM (40%) patients [28]. A similar smaller-scaled array-based approach was used by Mook et al. to study 23 genes in cardiomyopathy patients [29]. PCR-based, filter-based, and in-solution-based methods have also successfully been applied in combination with the different NGS systems mentioned above to study cardiomyopathy-relevant genomic loci by NGS [30–34]. Haas et al. for example were able to for the first time show the mutational landscape for DCM across a large european cohort of 639 patients using in solution based target enrichment (Haas et al. EUR HEAR J. 2014). These studies show the feasibility of using NGS in a clinical environment, but also show the diversity of currently used approaches. Although studies exist which claim NGS to be ready as a stand-alone diagnostic test [35], most centers still rely on Sanger validation of the relevant NGS variants, which are still mainly produced within research projects and are not yet part of the daily routine. Although initial guidelines for clinical NGS testing exist [36,37], it is still difficult to compare results between different centers.

1.5 Sample Preparation

Well-established protocols exist for sample preparation that enable technicians to reproducibly prepare high-quality sequencing libraries, if high-quality DNA is used. Therefore, an initial quality check of the DNA through, for example, Bioanalyzer (Agilent) or Qubit (Invitrogen) measurements is inevitable. Library preparation protocols have been improved tremendously and now only require a few nanograms of input DNA compared with the several micrograms that were needed not so long ago. Also, the time needed for sample preparation has been shortened from a couple of days to a few hours, making it possible to finish preparation in a single working day, achieved, for example, with the Haloplex system (Agilent). Recently, disease-specific enrichment assays were introduced by Haloplex, including an optimized predesigned cardiomyopathy and arrhythmia panel, removing the need for manual target region design. Such panels already exist for cancer, for example, and are expected to be developed for other diseases. Another important aspect in the course of sample preparation is the tracking of the samples to guarantee sample integrity when used in a high-throughput manner. Use of a laboratory information management system (LIMS) is desired. Here, freely as well as commercially available tools exist that help to reduce manual intervention and lower the overall turnaround time [38,39].

1.6 Bioinformatics Analysis Pipeline

Nowadays, the decreasing costs per base enable researchers to sequence larger target regions, exomes or genomes at a higher depth. However, those high-quality gigabase-scale datasets pose an immense challenge for downstream analyses. One major problem for comparison of results is the variety of mainly “homegrown” analysis strategies that have been developed at individual sites. Briefly, they consist of mapping, variant calling, annotation, filtering, and validation of selected variants. Although most approaches rely on similar strategies for filtering (e.g., filtering variants present in databases like dbSNP or the 1000 Genome Project) caution has to be taken. Andreasen et al. showed, for example, that 14% of HCM and 17% of DCM previously disease-causing reported missense and nonsense variants are present within the National Heart, Lung and Blood Institute “Grand Opportunity” Exome Sequencing Project (GO-ESP) cohort, which contains exome data from 6500 individuals [40]. Depending on the chosen filters and also due to differences in pipeline tools, tool combinations, or even versions of the programs, this will lead to a low concordance among the analyses [41]. Despite those drawbacks, it is expected that these hurdles will be overcome by newly developed, improved algorithms. Growing sequencing quality and performance of analysis tools will contribute to provide reliable variant calls, with no need for validation to gain acceptable clinical...