A Screening Approach to Identify Clinically Actionable Variants Causing Congenital Heart Disease in Exome Data

Introduction

See Editorial by Paige et al

Congenital heart disease (CHD) is the most common developmental defect in humans and also one of the most severe causes of infant morbidity, affecting ≈8 per 1000 live births or 1.35 million infants each year worldwide.¹ Moreover, with advancements in prenatal diagnosis and surgical corrections, the number of newborn deaths has declined, and more children with CHD survive into adulthood, including those with severe abnormalities of the heart. The recurrence risk is 1% to 10%, indicative of a genetic basis of disease, depending on the type of CHD and whether the father or mother has CHD.² Therefore, it is necessary to reach a genetic diagnosis in affected families.

Generally, the pathogenesis of CHD is still poorly understood, and in only about one third of cases can an inherited or sporadic genetic abnormality be uncovered as the cause of the defect.^3,4 Single gene mutations can cause some cases of both syndromic and isolated CHD, but many other cases are believed to be multifactorial, implicating multigenic or environmental factors in disease causation.^4,5 Nonsyndromic CHD characterized by Mendelian inheritance is frequently caused by a monogenic mutation, but most nonsyndromic CHD occurs sporadically.^3,5

Massively parallel sequencing technology and variant screening have become widely adopted tools for genetic analysis in research and diagnostics, and a growing number of susceptibility genes causing nonsyndromic types of CHD have been identified.^4,6 Unlike past genetic analyses that only focused on single gene loci or a panel of candidate genes, whole-exome sequencing (WES) can provide gene variant frequencies in populations, which in turn give important insights about the true effects of variants, and facilitate the clinical diagnosis of sporadic and Mendelian diseases, respectively.^7,8 WES offers advantages over whole-genome sequencing in terms of costs and speed, as well as data volume, analysis, and interpretation.^7,9,10 WES has identified many human genes that cause CHD when mutated.^11,12 Considering that targeted deletion of >500 mouse genes leads to cardiac structural defects (http://www.informatics.jax.org/), it is expected that exome and genome sequencing will implicate many more human genes in CHD in the future.¹³ The challenge, however, is to recognize disease-causal variants and distinguish them from numerous benign variants present in every genome. Nevertheless, to quickly obtain diagnostic results for patients or families with CHD, focusing on known disease-causal genes has been reported as a useful approach.^14,15

In this study, we exome-sequenced a cohort of 30 Australian families with various forms of sporadic or Mendelian CHD. First, we curated a contemporary high-confidence list of genes known to cause human CHD, and we report how many cases can be explained by damaging variants in these genes. We use the updated standards and guidelines for the clinical interpretation of sequence variants with respect to human diseases developed by the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) guidelines¹⁶ to rate pathogenicity and thus identify clinically actionable findings. Second, we comprehensively analyzed whole exomes within families to identify potential disease-causative variants in genes not currently considered causal of human CHD but plausible based on known gene function and segregation with disease in families. Our study provides an approach that rapidly yields clinically actionable results and also identifies potentially causal variants that require further information for them to be clinically actionable.

Methods

Because this analysis uses patient samples, the data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure.

Study Participants

Ethical approval for this study was obtained from the Sydney Children’s Hospital Network Human Research Ethics Committee (approval number HREC/16/SCHN/73). The cohort consisted of 30 unrelated families of which 18 were recruited at Princess Margaret Hospital for Children, Perth, Australia, and 12 were selected from the Kids Heart Research DNA Bank based at Children’s Hospital at Westmead, Sydney, Australia. Families originating from Perth were recruited at presentation to a preadmission clinic before cardiac surgery. They included individuals with family histories of CHD and sporadic cases, whereas only families with >1 affected family member were selected from the Kids Heart Research DNA Bank in Sydney. Heart defects in all affected individuals were confirmed using echocardiography. In all families, at the minimum, a trio (parents and proband) was analyzed. Principal component analysis confirmed the self-declared ethnicities of all individuals in the cohort (Figure I in the Data Supplement).

Sample Preparation and WES Library Construction

Blood samples were processed at the Kids Heart Research DNA Bank, Sydney. Whole blood samples originating in Perth were transported to Sydney on dry ice. Genomic DNA was extracted as described previously.¹⁴ Exome libraries were prepared using the Nextera Rapid Capture Exome kit (Illumina, Inc, San Diego, CA) and sequenced on Illumina HiSeq2000 and HiSeq2500. Sequencing coverage analysis was performed using Bedtools (http://bedtools.readthedocs.io/) in conjunction with custom scripts.

Analysis of Sequencing Data

After exome sequencing, DNA sequence reads were aligned to the human reference genome build hg19 using the Burrows-Wheeler Aligner.¹⁷ Single Nucleotide Variant calling was performed using Genome Analysis Toolkit software (Broad Institute, Cambridge, MA) following the germline variation best practices and using the option of HaplotypeCaller. The generated variant call format files from multiple samples were then processed together via a joint genotyping call with Genome Analysis Toolkit to produce 1 multisample variant call format file for all DNA variants. Variants were annotated with various metrics using ANNOVAR (http://annovar.openbioinformatics.org/en/latest/). A count for the number of homozygous variants in the Exome Aggregation Consortium (ExAC) database (http://exac.broadinstitute.org/)¹⁸ was incorporated with an in-house custom script. The resulting ANNOVAR-annotation file was split into individual families using an in-house custom script.

A control data set of sequenced exomes was used to help exclude artificial sequencing-derived variants in our samples. The control data consisted of Binary Alignment/Map files from 200 reference exome sequences of similar ethnicity, aligned to hg19 using Novoalign (http://www.novocraft.com), generated in a previous project using the same sequencing machinery. Sequence variants were identified, and a variant frequency table was prepared. We referred to this annotation as the control cohort number of alleles.

Filtering, Prioritization, and Validation of Variants

A list of 90 high-confidence CHD genes known to cause human CHD when mutated was curated based on stringent criteria and termed high-confidence CHD gene list (hcCHD list; see Methods in the Data Supplement; Table I in the Data Supplement for details). The VarSifter tool¹⁹ was used to filter variants based on genomic location, allele frequency, inheritance pattern, and in silico predicted damaging consequence. In the first analysis approach (hcCHD Gene Screen), exomes of affected probands were screened for rare variants in genes on our manually curated hcCHD list. Variants were defined as rare if the minor allele frequency was <0.01 (<1%) in the ExAC and 1000 Genomes Project databases.²⁰

In the second analysis approach (Comprehensive Screen), inheritance filters were applied to prioritize rare and damaging variants segregating with disease within any gene, and for each family, all possible Mendelian inheritance models were tested. Minor allele frequency thresholds in ExAC and 1000 Genomes were set to <0.001 (<0.1%) for de novo and autosomal dominant inheritance models and to <0.02 (2%) for autosomal recessive and compound heterozygous models. In addition, thresholds for the homozygous variant count in ExAC and heterozygous count in the control cohort, respectively, were set to <1 for de novo and autosomal dominant, <2 (ExAC) and <4 (control cohort) for compound heterozygous, and <3 (ExAC) and <4 (control cohort) for autosomal recessive inheritance models.

Variants passing population frequency thresholds were checked with the Integrative Genomics Viewer²¹ to confirm the variant call. Variants in highly polymorphic genes, which have notoriously high false-positive variant calls (eg, mucins), were removed irrespective of read quality. Subsequent consideration of protein-damaging variants first removed synonymous calls. Remaining variants were prioritized in VarSifter by their in silico predicted impact on protein function using the Combined Annotation-Dependent Depletion (CADD)²² and PolyPhen-2²³ tools. Variants predicted as possibly or probably damaging by PolyPhen-2 (HumVar score, ≥0.446) and having a scaled CADD score of >15 were presumed damaging. Variants lacking PolyPhen-2 and CADD annotations, for example, indels, were checked with MutationTaster2²⁴ and excluded if the variant call was incorrect. Candidacy of top-scoring variants in the Comprehensive Screen was further interrogated with respect to known gene function, available animal models, published in vitro evidence for similar gene variants causing protein dysfunction, and tentative associations of previous human gene variants with CHD. Finally, potential disease-causal variants were evaluated according to the ACMG-AMP guidelines.

Results

Clinical Characteristics of the Cohort

In total, 120 individuals from 30 families underwent exome sequencing. In 14 families, no history of CHD was reported, in which case the trio, consisting of the unaffected parents and affected proband, was sequenced. For the remaining 16 families with apparent Mendelian CHD, trios plus individuals with available DNA were sequenced. Extracardiac anomalies were found in a subset of patients, but the majority (44 of 57 affected individuals) had isolated CHD (Table 1). Pedigrees and individual clinical findings are provided (Figure II in the Data Supplement). In all cases, the sequencing was of sufficient coverage for analysis (Table II in the Data Supplement).

hcCHD Gene Screen Identified Clinically Actionable Variants and Other Predicted Damaging Variants in Known CHD-Causing Genes

Genetic variants passing the initial automated quality control threshold of the Genome Analysis Toolkit software (≈50 000 variants per individual) were subjected to a series of filtering steps. Filtering for variants in genes implicated in CHD has been applied previously,^25,26 but these gene lists tend to be too inclusive and often include genes only linked to CHD by knock-out mouse models. By contrast, we curated a contemporary, high-confidence list of genes known to cause CHD in humans (hcCHD list) by reviewing publicly available CHD gene panels, gene and phenotype databases, and published literature for genes that, when mutated, have been repeatedly reported to cause CHD in humans. Genes were only included in the hcCHD list if variants in the respective gene were reported as the monogenic cause of CHD (isolated or in the context of a syndrome) in ≥3 independent familial or sporadic cases in ≥2 separate publications. In a few cases, single publications reporting multiple individual de novo or autosomal dominant cases with monogenic causative variants were deemed sufficient for inclusion of the respective gene. Importantly, rare predicted damaging variants in genes on the hcCHD list will attract a pathogenic rating by the ACMG-AMP guidelines if the variant segregates with disease. The hcCHD list includes 90 genes that met our stringent criteria defining a hcCHD gene (Table I in the Data Supplement). Curation of the hcCHD list will be ongoing with an up-to-date list accessible online (https://www.victorchang.edu.au/heart-research/embryology), making this list a valuable resource for diagnostic and research use.

Across the 30 families, 99 rare variants were identified in genes on the hcCHD list. Rare variants, predicted to be damaging by CADD,²² PolyPhen-2,²³ and MutationTaster2,²⁴ were identified for further scrutiny (Table III in the Data Supplement). Evidence from the literature supported the candidature of 7 variants that segregated with disease in the respective families (Table 2; Figure II in the Data Supplement). Variants in 3 genes (PTPN11, TBX5, and TFAP2B) are rated pathogenic by the ACMG-AMP guidelines and are thus considered clinically actionable. A de novo missense variant in PTPN11 (NM_002834.4:c.184T>G, p.Y62D), identified in the proband of family 13455 with atrial septal defect, patent ductus arteriosus, and pulmonary valve stenosis (Figure II in the Data Supplement), had been previously documented as pathogenic in ClinVar (SCV000057372.3). A dominantly inherited novel TBX5 stopgain variant (NM_000192.3:c.1221C>G, p.Y407*) was identified in family 00860 with atrial and ventricular septal defects. This variant segregates with disease in all 6 affected family members, and a frameshift variant 38 amino acids further C-terminal has been reported to cause Holt-Oram syndrome, including CHD.²⁷ It is possible that subtle upper limb abnormalities, typical for Holt-Oram syndrome, but only identifiable using radiological imaging, were missed during the clinical examination of family 00860 (Figure II in the Data Supplement).

We also identified a previously described pathogenic TFAP2B variant in family 02946 (NM_003221.3:c.854G>A, p.R285Q, ClinVar: SCV000028717.2), reported to have a dominant-negative effect on the DNA-binding capability of TFAP2B. This variant was present in an obligate parental carrier, potentially indicative of incomplete penetrance, commonly observed in inherited CHD.¹⁴ Like the previous report, the TFAP2B variant was found in individuals with patent ductus arteriosus and distinctive facial features of Char syndrome (Figure II in the Data Supplement).²⁸

Variants were identified in NOTCH1, TBX20, ELN, and GATA4 that did not pass PolyPhen-2 and CADD thresholds (Table 2) but were present in all respective affected family members and exhibited phenotypic concordance with respect to the literature (Figure II in the Data Supplement). These variants, rated as variants of uncertain significance according to ACMG-AMP guidelines, may contribute to the overall phenotype as part of digenic or oligogenic scenarios. The remainder of the variants are currently rated variants of uncertain significance or benign because they neither segregate with disease nor pass CADD and PolyPhen-2 thresholds (Table III in the Data Supplement).

Comprehensive Screen Identified Clinically Actionable Variants and Other Predicted Damaging Variants in Families With Sporadic CHD

In a second approach (Comprehensive Screen), each family was analyzed for potential disease-causal variants across all genes in an unbiased way. In 14 families with no history of CHD (proband/parent trios), de novo variants are most likely to cause disease; however, for completeness, all inheritance models were applied, identifying additional variants (Table IV in the Data Supplement). We identified 18 rare de novo variants in 14 probands. Of these variants, 11 were predicted to be damaging (Table IV in the Data Supplement), but only 4 had links to heart development (Table 3; Figure II in the Data Supplement).

Among them is the de novo variant in PTPN11 also identified in the hcCHD Gene Screen. A missense variant in PBX1 convincingly explains the cardiac defect, Tetralogy of Fallot with absent pulmonary valve, in proband 13467. PBX1 (Pre-B-cell leukemia transcription factor 1) is an evolutionarily highly conserved homeodomain-containing transcription factor, which when deleted in mouse, results in similar cardiac phenotypes (Figure II in the Data Supplement).^29–31 In 4 other families, we identified predicted damaging de novo variants in genes that have not yet been linked to human CHD (Table IV in the Data Supplement). Of these, 2 missense variants in CNOT1 and ZFP36L2 have supportive evidence in the literature (Figure II in the Data Supplement, families 13478 and 13481). Accordingly, variants in PTPN11 and PBX1 were rated pathogenic, according to ACMG-AMP guidelines, and are thus clinically actionable, whereas CNOT1 and ZFP36L2 were rated likely pathogenic (Table 3).

We also applied a filter for recessive inheritance and identified 4 homozygous variants predicted to be damaging. None of the affected genes have a known link to CHD. For a compound heterozygous inheritance model, we identified predicted damaging candidate variants in 7 genes (Table IV in the Data Supplement), also none of which have a known link to CHD. Filtering variants for an autosomal dominant inheritance model that assumes incomplete penetrance in the parental generation returned large variant numbers. In these cases, variants in the hcCHD genes were given priority, but none were compelling candidates (Table IV in the Data Supplement).

Comprehensive Screen Identified Clinically Actionable Variants and Other Predicted Damaging Variants in Families With a History of CHD

The Comprehensive Screen was also applied to 16 families with a history of structural heart defects. As above, for completeness, all inheritance models were applied, identifying additional variants (Table V in the Data Supplement). We identified 1545 rare autosomal dominant variants in 16 probands. Of these, 428 variants were predicted to be damaging, and top candidates present in all affected family members, with supportive evidence in the literature, are highlighted in Table 3.

In 5 families, we identified predicted damaging variants that segregate with disease. In 2 families, variants explain the cardiac lesions: the stopgain in TBX5 as identified with the hcCHD Gene Screen and a known pathogenic missense variant in TEK.

TEK (NM_000459.4:c.2744G>A, p.R915H; SCV000044247.1; Table 3), previously described to cause ventricular septal defects and cutaneomucosal venous malformations,³² was identified in a trio within which the mother and proband had ventricular septal defects. This and the cardiovascular phenotype of Tek-null mice³³ are strong evidence that the TEK variant is disease-causal in this trio.

We found compelling candidates that segregate with disease in 3 families: a missense variant in USP34, UPF2, and a digenic combination of 2 missense variants in KDM5A and KMT2C (Table 3). The novel USP34 variant (NM_014709.3:c.863C>T, p.A288V) segregates with disease, and a similar missense variant in this gene was previously reported in a patient with diverse CHDs.¹² Cardiac-specific silencing of this gene in Drosophila resulted in significant embryonic lethality, highlighting a possible role of this gene in human heart development.³⁴

The novel UPF2 variant (NM_080599.2:c.2606T>G, p.I869S) is nearby a previously identified variant E858R that perturbs the UPF2 (Regulator of nonsense transcripts 2)-UPF3b interaction, necessary for nonsense-mediated decay (Figure II in the Data Supplement).³⁵ UPF2 functions as part of the nonsense-mediated mRNA decay surveillance machinery, and, although the function of UPF2 has not yet been assessed in the heart, loss of this gene in mice results in embryonic lethality by midgestation,³⁶ making this variant a compelling candidate.

KDM5A and KMT2C, and their respective variants (NM_001042603.2:c.4399C>T, p.R1467W and NM_170606.2:c.14258A>T, p.Q4753L), may perturb heart development through altered histone methylation (Figure II in the Data Supplement, family 01179).

Accordingly, variants in TBX5 and TEK were rated pathogenic according to ACMG-AMP guidelines and are thus clinically actionable, whereas USP34 and UPF2 were rated likely pathogenic.

The Comprehensive Screen revealed 4 additional likely disease-causal variants in the remaining families with apparent incomplete penetrance of CHD: the TFAP2B missense variant identified with the hcCHD Gene Screen, a single-base deletion in TIE1 leading to a frameshift and truncated protein, a stopgain variant in TEAD2, and an indel in FLT4 (Table 3).

TIE1 (Tyrosine-protein kinase receptor Tie-1) functionally dimerizes with TEK (Angiopoietin-1 receptor) to mediate development and maintenance of the vascular system. Loss of TIE1 in mice results in abnormal heart morphologies, including a thinning of the atrial wall.^37,38 In this light, the TIE1 variant (NM_005424.4:c.1341del p.V448Cfs*9) is a likely cause of the septal defects seen in family 02529 (Figure II in the Data Supplement). TEAD2 (Transcriptional enhancer factor TEF-4) regulates cardiomyocyte proliferation and survival during development through the Hippo signaling pathway and regulation of Pax3 and Foxa2 transcription factors.^39,40 The novel TEAD2 stopgain (NM_001256659.1:c.1195C>T, p.R399*) is, therefore, a plausible explanation for the CHDs seen in family 13430 (Figure II in the Data Supplement).

Accordingly, the variant in TFAP2B was rated pathogenic according to ACMG-AMP guidelines, and is thus clinically actionable, whereas variants in TIE1, FLT4, and TEAD2 were rated likely pathogenic.

In summary, the hcCHD Gene Screen identified disease-causal variants in 3 cases (1 sporadic and 2 familial), defined as pathogenic by ACMG-AMP guidelines. The Comprehensive Screen also identified these 3 variants and rated them as top candidates in the respective families. Unsurprisingly, the Comprehensive Screen did not identify additional disease-causal variants in hcCHD genes but revealed candidate variants in other genes in a further 10 cases (3 sporadic and 7 familial). In total, we were able to uncover credible disease-causal variants in 3 families with variants in the hcCHD gene list (3 of 30; 10%) and in 10 families with variants in genes not yet considered causal of CHD (10 of 30; 33%; Tables 2 and 3; Figure). Furthermore, the unbiased Comprehensive Screen identified numerous additional rare and predicted damaging variants with no previous association to CHD (Tables IV and V in the Data Supplement).

• Download figure
• Download PowerPoint

Discussion

Here, we applied 2 approaches to identify disease-causal genetic mutations in 30 Australian families affected with CHD using exome sequencing. The hcCHD Gene Screen focusing on genes known to cause CHD in humans efficiently explained the disease pathogenesis in 10% of cases, and the Comprehensive Screen identifying predicted damaging variants in the whole coding region of the genome detected credible gene candidates potentially causing CHD in an additional 33% of cases.

Currently, clinical genetic testing in this field is restricted to cytogenetic screening of patients with multiple congenital abnormalities, copy number variant (CNV) analysis for severe CHDs associated with known syndromes, and single gene testing in a subset of genes with monogenic syndromic association.⁴¹ However, given the complex pathogenesis of CHD, it can be assumed that more genes are involved in heart development and CHD than those currently listed in diagnostic gene panels. To address this unmet need for an up-to-date list defining human CHD genes, we curated the hcCHD list of genes, variants in which will be classified as pathogenic according to ACMG-AMP guidelines if the variant is rare, predicted damaging, and segregates with disease (Table 2).

Performance of Family-Based WES Analysis

The hcCHD Gene Screen identified a gene variant as the definitive cause of CHD in 3 of 30 families (Figure, left). Through subsequent application of the Comprehensive Screen, we identified likely causative variants in 10 additional families (Figure, middle). Although the latter 10 candidate genes have not yet been unequivocally linked to CHD, there is substantial protein function or animal model evidence to indicate a role in human CHD.

Moreover, the Comprehensive Screen returned several additional predicted damaging variants, which we do not currently consider disease-causal because of a lack of data relating them to heart development or the general lack of knowledge about their function (Tables IV and V in the Data Supplement; Figure II in the Data Supplement). Future functional studies, such as analysis of animal models, or assessment of protein function in vitro (Figure), is warranted to determine the actual consequence of these predicted damaging variants.

To accurately evaluate our yield in comparison with previous reports, it is necessary to consider sporadic and familial cases in our cohort separately. Among the 14 families with sporadic CHD, we identified one disease-causal variant within a known CHD gene (PTPN11). With the Comprehensive Screen, we could identify a plausible variant in 3 additional families, yielding an overall diagnosis rate of 29% (4 of 14). This diagnosis rate for sporadic cases exceeds the maximum expected ≈10% contribution by de novo variants, as suggested by recent large-scale studies of sporadic CHD.^12,25

Our diagnosis rate was higher for the 16 families with a history of CHD. The hcCHD Gene Screen identified disease-causal variants in 2 known CHD genes (TBX5 and TFAP2B), and the Comprehensive Screen uncovered candidate variants in 7 additional families, yielding an overall diagnosis rate of 56% (9 of 16).

Comparable analyses of familial CHD have achieved successes between 33% and 46%. Because these studies, however, only interrogated variants in known CHD genes (57–69 genes),^14,15,42 this statistic is appropriately below that of our unbiased Comprehensive Screen. The apparent outperformance of these studies compared with our hcCHD yield can be somewhat accounted for, similar to our de novo yield, by relative differences in sample sizes, variant filtering, and interpretation, between studies.

In familial cases, monogenic mutations are more likely to be the cause of CHD, especially if the disease occurs across ≥2 generations in a Mendelian pattern. By contrast, sporadic cases are more likely to be multifactorial, involving multiple genetic and nongenetic factors.^4,5 Indeed, cumulative data from published reports show that more cases of sporadic CHD remain unresolved by genetic screening than cases of familial CHD.⁴ The absence of convincing damaging recessive or compound heterozygous genotypes in our study is in line with ≈1% incidence previously reported for nonsyndromic CHD cases with no parental consanguinity.^25,26

In contrast to other cohort-based CHD studies, we reported variant pathogenicity in the context of ACMG-AMP standards. We have shown that predicted damaging variants identified in genes on the hcCHD list receive a pathogenic classification and, therefore, are considered causative for diagnostic purposes. As the hcCHD list inevitably grows because of efforts such as our Comprehensive Screen, an increasing number of pathogenic disease-causal variants will be identified. Periodic reinterpretation of clinical exomes in light of new supportive evidence will significantly improve diagnostic success.⁴³ Unlike target panel sequencing, WES data can be reanalyzed when warranted using up-to-date lists, such as the hcCHD list, to achieve higher diagnostic rates.

Limitations

A general limitation of family-based WES variant filtering and analysis is that it assumes a monogenic cause of disease. In cases where no plausible gene variant can be found, it is possible that the disease is caused by ≥2 mutations, which individually need not be particularly rare nor damaging. This has been highlighted in recent studies, utilizing unbiased statistical comparisons from large-scale CHD cohorts,^25,26 showing that patients with nonsyndromic CHD carry an exome-wide burden of rare inherited protein-truncating variants. These findings recognize a potential oligogenic contribution to nonsyndromic CHD and may partially account for observed incomplete penetrance. Another limitation of using WES data concerns cases in which CNVs are the underlying genetic factor. CNVs are estimated to explain ≤10% of sporadic CHD, and even ≤25% of syndromic CHD, but their identification requires high-quality whole-genome sequence data or other specific assays.^4,6 CNV analysis was attempted from whole-exome sequence data of this cohort, but variant calls were of insufficient quality to be practically interpretable, as is generally known for exome data.⁴⁴ Finally, pathogenic mutations in noncoding genome regions will not be identified with WES. However, the functional relevance of noncoding variants is largely unknown, and because they are difficult to test, it is likely that the identified causes of CHD will be largely limited to the exome for some time.⁹ A future approach to overcome these limitations is to sequence the genome, limit initial analysis to the exome, integrate CNV analysis, and analyze noncoding regions when in silico tools to interpret these regions become available.

Conclusions

Filtering WES data using the hcCHD list of genes known to cause CHD quickly identifies variants that are pathogenic according to the ACMG-AMP guidelines and thus are clinically actionable. The hcCHD list, as it is up-to-date, will be invaluable in a diagnostic setting. Therefore, in probands with CHD requiring surgery, with no identified pathogenic CNVs, and their parents, we propose exome sequence analysis of variants in hcCHD genes as a first-line diagnostic approach. The Comprehensive Screen, however, significantly increases the rate of identifying CHD-causing variants, and also implicates additional candidate genes that are likely to be included in the hcCHD list in the future, ultimately improving the rate of patient diagnosis.

Acknowledgments

We are grateful to the families who participated in this research, to Dr David Andrews (Princess Margaret Hospital, Perth), and the individuals and agencies that supported this research.

Footnotes