The diagnostic value of next generation sequencing in familial nonsyndromic congenital heart defects

INTRODUCTION

Isolated CHD are generally thought to have a multifactorial origin, though in a small proportion of cases, estimated at 4%, there is a clear familial recurrence [Oyen et al., 2009; Oyen et al., 2010; van der Bom et al., 2011], suggesting a single genetic cause. The inheritance pattern is typically autosomal dominant with variable expression and reduced penetrance. More rarely, autosomal recessive or X-linked inheritance exists. Traditional cloning approaches have led to the identification of a large number of genes for nonsyndromic CHD [Schott et al., 1998; Garg et al., 2003; Garg et al., 2005; Kirk et al., 2007], but each of these genes is implicated only in a small proportion of familial CHD. Genes associated with CHD can be found in CHDWiki [Barriot et al., 2010] (http://www.esat.kuleuven.be/∼bioiuser/chdwiki).

Traditional genetic diagnosis of a large panel of candidate genes suffers from limitations in sequencing capacity. Nowadays, this is overcome by next-generation sequencing (NGS) technologies, which enable large-scale DNA sequencing. Selective enrichment followed by massive parallel sequencing of the genomic regions of interest has been widely applied in genetic research, and is gradually being implemented in clinical diagnostic settings [Ellard et al., 2013; Mook et al., 2013; Wang et al., 2013].

MATERIALS AND METHODS

Families and Clinic Phenotypes

Thirteen families with multiple affected individuals with CHD were recruited from the genetic and cardiology clinics in three University Hospitals in Belgium (Leuven 11, Ghent 1 and Louvain-la-Neuve 1). Inclusion criteria were families with a nonsyndromic CHD in at least three first or second degree relatives. The type of CHD needed to be confirmed by cardiac ultrasound, and the predominant heart defect in the family is known to be etiologically heterogeneous (thus excluding e.g. families with supravalvular aortic stenosis, known to be exclusively caused by ELN gene mutations). Unaffected individuals included in the study were examined by cardiac ultrasound. Moreover, DNA from at least one affected individual was available. Informed consent was obtained from all participants or their legal representative. The study conforms to the principles outlined in the Declaration of Helsinki, and was approved by the local ethical committee of the UZ Leuven (S52853). The pedigrees of the families are shown in Figure 1 and Supplementary Figure 1. Genomic DNA was extracted from peripheral white blood cells in all individuals.

RESULTS

DISCUSSION

Given the genetic heterogeneity, genetic testing for familial nonsyndromic CHD is challenging. The exceptions are families with defined CHD with a single known causative gene, such as ELN and SVAS/PPS [Urban et al., 2000] or NKX2-5 in families with ASD associated to AV conduction defects [Gutierrez–Roelens et al., 2002]. The present results show that in a significant proportion of well-selected familial CHD (6 out of 13; 46%), a causative variant can be identified using the newest sequencing technology, allowing the massive parallel resequencing of all known cardiac genes in one single assay, which thus holds a value as a diagnostic test. In this study, pathogenic variants were only detected in three genes, TBX5, NOTCH1 and MYH6. Of interest, pathogenic variants in the same genes were detected in a recent report using a similar approach as the present study [Blue et al., 2014]. This suggests that in future, a smaller gene panel might be established, taking advantage of the parallel analysis of several genes commonly causing familial CHD, and thus avoiding the burden of numerous variants of unknown significance.

Given the enormous amount of human genetic variation, we anticipated and encountered interpretational difficulties of the found variants. Restricting the analysis to a well-defined panel of validated genes for CHD facilitated the interpretation. In total, using stringent filtering criteria (as described in Materials and Methods section), only 45 potential functionally significant variants were found in these 13 families. In the families where a functional deleterious variant was found, the interpretational challenges remain in assigning causality to these variants.

In all families where a certain or likely causative variant was identified, the found gene matched the previously reported genotype-cardiac phenotype associations. For instance, in Family 1, we identified a MYH6 missense variant. MYH6 has been described as the predominant sarcomeric disease gene for familial ASD, and particularly, perturbations in the MYH6 head domain is thought to be the major genetic cause for familial ASD [Posch et al., 2011]. However, association of mutations in MYH6 with TGA has previously only been reported once in a mutation analysis study of MYH6 in patients with a wide spectrum of sporadic CHD [Granados–Riveron et al., 2010]. In three families with LVOTO (Families 2, 3, and 4), we identified three functional deleterious variants in the NOTCH1 gene: a frameshift deletion, a splice site variant and a missense variant. Notch signaling plays an important role in cardiac outflow tract development [de la Pompa and Epstein, 2012]. Missense mutations in NOTCH1 have been found in bicuspid aortic valve (BAV) and left ventricular outflow tract malformations including AS, COA, and HLHS [McBride et al., 2008], NOTCH1 haploinsufficiency resulted from nonsense or frameshift mutations is thought to cause familial aortic valve disease [Garg et al., 2005]. However, missense variants with pathogenic in silico predictions sometimes can be found in normal population. For this reason, without further functional studies, it is difficult to confirm the causality of the missense variant found in Family 4. In Families 5 and 6, two missense variants were identified in TBX5. Mutations in TBX5 are known to cause Holt–Oram Syndrome, which is characterized by skeletal malformations in upper limbs and cardiac defects, most commonly ASD and VSD [Mori and Bruneau, 2004]. Clinically the affected members in these two families present as isolated CHD, without observable limb defect. This is of interest, since TBX5 mutations typically are associated with limb defects, which can be variable. For instance, Cross et al. [2000] reported a family with very mild limb defects in only one affected individual. The fact that TBX5 mutations can be associated with apparently isolated CHD is also supported by the recent finding by Blue et al. [2014] who detected a TBX5 mutation in two families with isolated CHD. A recent study by Baban et al. [2014] also reported TBX5 mutations in patients with isolated tetralogy of Fallot (TOF). This indicated that screening of a larger, unbiased panel of genes might lead to novel genotype–phenotype correlations. Previously, targeted gene testing was only performed for families with suggestive features, for instance, heart-limb anomalies in the case of TBX5 testing [McDermott et al., 2005].

Genetic testing aims to improve genetic counseling in familial CHD. At first sight, counseling for CHD appears to be straightforward. However, in most families, even though the inheritance pattern is autosomal dominant, there is reduced penetrance and variable expressivity, which complicates risk determination. For instance, in Family 2, several clinically unaffected individuals (besides the obligate carriers) were shown to be carriers. They have thus an increased risk for affected offspring. Moreover, the high incidence of CHD raises the possibility of phenocopies. Indeed, in two of the families, we encountered affected individuals who did not carry the found variant. In Family 2, Individual I2, who had an AS, did not carry the familial NOTCH1 variant. It is uncertain if the patient had congenital abnormalities of the aortic valve, but given her advanced age, this would most likely fit with the more common age-dependent degenerative aortic valve stenosis. The identified novel frameshift deletion in NOTCH1 is probably associated with the defect in the family. NOTCH1 mutations are known to cause a spectrum of aortic valve anomalies, which fit with the phenotypes in Family 2. The frameshift variant was not found in 1,000 normal controls, and is thought to have a high functional impact by resulting in a truncated protein. Finding the same variant in a replication cohort would provide further validity. In Family 4, two sisters present a LVOTO, but only in the sister with COA, a NOTCH1 variant is implicated. This is unexpected, since it is commonly assumed that AS, COA, and HLHS are part of a spectrum of CHD, with shared pathogenesis and etiology. One explanation is that the found NOTCH1 missense variant may not be causal in this family. Alternatively, the two sisters may carry a different variant. The two affected members with AS carried a same missense variant in EHMT1 (c.752C>G, p.Pro251Arg), which alters a conserved amino acid and is predicted to be pathogenic (SIFT: 0, PolyPhen2:0.995, MutationTaster: disease causing). Mutations in EHMT1 are known to cause Kleefstra Syndrome, an autosomal dominant disorder characterized by severe intellectual disability and characteristic dysmorphic features. However, neither patient showed any manifestation of the syndrome, and this variant was also detected in a control person (1/1000), indicating the results by in silico prediction programs need to be interpreted with caution. This also indicates the value of studying as many affected family members as possible in order to evaluate the significance of found variants.

Familial CHD is characterized by remarkable variability in expression. One hypothesis is that this is related to additional genetic and/or environmental modifiers. Thus, it is possible that additional rare variants in one or more of the other CHD-associated genes cause a more complex or severe CHD in affected individuals. For this reason, we have searched for genetic modifiers in the gene panel we investigated. For instance, the MYH6 missense variant identified in Family 1 explains the prevailing phenotype (ASD II) in the family, but the proband also has TGA, which is a distinct phenotype from other family members, and there is no clear genotype–phenotype correlation between MYH6 and TGA. We hypothesized that besides the MYH6 variant explaining the common defects in the family, a variation in another gene could contribute to the additional phenotype. However, we did not identify in this individual any novel or rare variant in any of the genes associated with TGA, including CFC1, NODAL, ZIC3, MED13L, ACVR2B, GDF1, and LEFTY2 [Goldmuntz et al., 2002; Muncke et al., 2003 ; Chhin et al., 2007; Mohapatra et al., 2009; D'Alessandro et al., 2013].

In seven families, no functional deleterious variant was detected in the analyzed genes. There are a number of possible explanations. First, there are technical issues. Incomplete capture of targets may occur because repetitive regions are removed during the tiling probe design, to ensure the uniqueness of the capturing. Certain regions of our targets were excluded due to the presence of a pseudogene or highly similar sequences in other genomic regions. In addition, during the library preparation, genomic regions with very high GC content undergo difficulties in amplification, resulting in reduced or no sequencing coverage. Although the paired-end deep sequencing strategy sometimes enables sequencing of poorly covered parts of the targets, ∼10% of total targets were not well sequenced, where less than 30 reads passing the QC criteria aligned. However, coding sequences of the canonical transcripts of target genes account for only ∼ 6% of not well sequenced region.

Second, our analysis focused on variations in exons and exon–intron boundaries of the target genes exclusively, known to harbor the majority of monogenic mutations. Variants disrupting regulatory elements located in deep introns, UTR or intergenic regions thus remain undetected. A recent study shows possible pathogenic effect of variations disrupting cardiac enhancers located upstream of SOX9 [Sanchez–Castro et al., 2013].

Third, current NGS data processing methods are limited in detecting genomic structural variants from targeted sequencing data. Structural variants have been implicated in the pathogenesis of CHD, partial gene deletions or duplications disrupting gene function are rare causes of familial CHD. For instance, chromosomal translocations or deletions disrupting the elastin gene can cause familial supravalvular aortic stenosis and supravalvular pulmonary stenosis [Morris, 1998]. A TBX5 intragenic duplication was indicated to cause atypical Holt–Oram syndrome phenotype in a family [Patel et al., 2012]. Our analysis pipeline did not include tools to detect structural variants.

Fourth, in those families where no causative variant was detected, a novel gene may be implicated. Novel genes have been constantly reported in association with CHD. Since the design of this capture panel, mutations in GATA5, BVES, and GJA5 were shown to cause sporadic or familial TOF [Guida et al., 2013; Wei et al., 2013; Wu et al., 2013], MEF2C mutations were identified in patients with outflow tract defects [Kodo et al., 2012], NPHP4 and DICER1 variants were described in association with TGA [Foulkes et al., 2011; French et al., 2012].

Finally, CHD is genetically heterogeneous: a same phenotype can result from alterations in different genes, while different variations interrupting the same gene can lead to different phenotypes. Due to the complexity of the genotype–phenotype correlation of CHD, the monogenic model may not apply to all nonsyndromic CHD families.

In conclusion, although the genetic heterogeneity of CHD and remarkable variability of expression make it challenging to interpret the large number of identified variants, setting up a well-defined analysis pipeline enables effective identification of causative variants. Application of targeted massive parallel sequencing in clinical diagnosis represents a break-through for genetic testing of (familial) nonsyndromic CHD and other complex diseases with high genetic heterogeneity. However, difficulties assigning causality with certainty of the found variants is the main challenge. Nevertheless, in over half of our CHD families with obvious genetic factors, no potential causative variant was identified by screening a panel of known CHD associated genes, revealing a large proportion of unknown genetic causes of familial CHD.