Results of next‐generation sequencing gene panel diagnostics including copy‐number variation analysis in 810 patients suspected of heritable thoracic aortic disorders

Abstract Simultaneous analysis of multiple genes using next‐generation sequencing (NGS) technology has become widely available. Copy‐number variations (CNVs) in disease‐associated genes have emerged as a cause for several hereditary disorders. CNVs are, however, not routinely detected using NGS analysis. The aim of this study was to assess the diagnostic yield and the prevalence of CNVs using our panel of Hereditary Thoracic Aortic Disease (H‐TAD)‐associated genes. Eight hundred ten patients suspected of H‐TAD were analyzed by targeted NGS analysis of 21 H‐TAD associated genes. In addition, the eXome hidden Markov model (XHMM; an algorithm to identify CNVs in targeted NGS data) was used to detect CNVs in these genes. A pathogenic or likely pathogenic variant was found in 66 of 810 patients (8.1%). Of these 66 pathogenic or likely pathogenic variants, six (9.1%) were CNVs not detectable by routine NGS analysis. These CNVs were four intragenic (multi‐)exon deletions in MYLK, TGFB2, SMAD3, and PRKG1, respectively. In addition, a large duplication including NOTCH1 and a large deletion encompassing SCARF2 were detected. As confirmed by additional analyses, both CNVs indicated larger chromosomal abnormalities, which could explain the phenotype in both patients. Given the clinical relevance of the identification of a genetic cause, CNV analysis using a method such as XHMM should be incorporated into the clinical diagnostic care for H‐TAD patients.

could explain the phenotype in both patients. Given the clinical relevance of the identification of a genetic cause, CNV analysis using a method such as XHMM should be incorporated into the clinical diagnostic care for H-TAD patients.

K E Y W O R D S
copy-number variations, eXome hidden Markov model, genetics, thoracic aortic aneurysm, thoracic aortic dissection

BACKGROUND
Over the last decade, advances in clinical genetics have led to the identification of disease-associated genes at a rapid pace. Especially when surveillance, early detection, and/or treatment provide health benefits for the index patient and at-risk relatives, identification of an underlying genetic cause is highly relevant. Therefore, recommendations for genetic counseling and DNA testing are increasingly being incorporated into clinical guidelines (Ackerman et al., 2011;Eccles et al., 2016). Thoracic aortic aneurysms and aortic dissections (TAAD) are a significant cause of sudden death at young age and is an example of a disease where screening of at-risk relatives can be lifesaving (Hoyert, Arias, Smith, Murphy, & Kochanek, 2001;Olsson, Thelin, Stahle, Ekbom, & Granath, 2006). Because aortic aneurysms are often asymptomatic and aortic dissections are often fatal and preventable by timely surgical intervention, the identification and clinical screening of at-risk relatives are clinically highly relevant and recommended (Hiratzka et al., 2010). In the majority of cases, TAAD is a sporadic occurrence, associated with, among others, hypertension, bicuspid aortic valve, and older age. However, in approximately 20% of cases TAAD is reported to be familial (FTAAD), often with an autosomal dominant pattern of inheritance with incomplete penetrance (Biddinger, Rocklin, Coselli, & Milewicz, 1997;Coady et al., 1999;Robertson et al., 2016). Of note, variants in genes originally associated with syndromic H-TAD have also been reported in patients presenting with apparently nonsyndromic H-TAD (Gago-Diaz et al., 2014;Regalado et al., 2011Regalado et al., , 2016. Given the incomplete penetrance and the highly variable age of onset within both heritable and sporadic TAAD (Campens et al., 2015;Coady et al., 1999;Khalique et al., 2009;Robertson et al., 2016), follow-up of at-risk relatives with normal aortic diameters at initial cardiologic screening is important. The identification of a pathogenic variant in a TAAD patient allows for targeted screening of relatives and enables prenatal and preimplantation genetic diagnosis. In addition, specific recommendations on imaging, surgical, and pharmacological treatment based on the underlying genetic cause are emerging (den Hartog et al., 2016;Franken et al., 2015;. A causative variant can be identified in approximately 20% of FTAAD families (D. M. Milewicz, Regalado, Shendure, Nickerson, & Guo, 2014).
Next-generation sequencing (NGS) allows for the rapid analysis of multiple genes in a diagnostic setting at relatively low costs. Therefore, DNA testing is increasingly offered to TAAD patients. The majority of the detected variants are single-nucleotide changes. CNVs have emerged as a relevant cause for several genetic disorders including cancer, intellectual disability, and neuropsychiatric disorders (Pollack et al., 2002;Shlien & Malkin, 2010;Thapar & Cooper, 2013). Routine diagnostic variant-calling analysis by (short reads-)NGS technology is not suitable for detecting CNVs. Therefore, CNVs may be missed unless additional testing is performed, for example, by multiplex ligation-dependent probe amplification (MLPA) or targeted array analysis. However, these tests are often not routinely performed and/or do not include all the relevant genes. The detection of CNVs in NGS sequencing data using statistical and computational tools is an alternative approach. The eXome hidden Markov model (XHMM) is one of several algorithms developed for the detection of CNVs through NGS data (Fromer & Purcell, 2014;Fromer et al., 2012). XHMM has identified (potential) causative CNVs in, for example, patients with Parkinson's disease, autism spectrum disorders, and rare diseases like Joubert syndrome and very early onset inflammatory bowel disease (Kelsen et al., 2015;Koyama et al., 2017;Poultney et al., 2013;Spataro et al., 2017). The aim of this study was to assess both the diagnostic yield of our panel of H-TAD-associated genes and the prevalence of CNVs in these genes. Here, we present the results of routine NGS analysis (variant-calling analysis) and XHMM analysis on the NGS sequencing data of the largest series of TAAD patients described so far (n = 810) referred for analyses of the H-TAD panel. In addition, we provide an overview of the clinical data of patients with a pathogenic or likely pathogenic variant, with a special focus on patients with CNVs. The results of this study underline the importance of CNV analysis in routine diagnostic testing in patients with H-TAD.  COL3A1, EFEMP2, ELN, FBN1, FBN2, MYH11, MYLK, NOTCH1, PLOD1, PRKG1, SCARF2, SKI, SLC2A10, SMAD2, SMAD3, SMAD4, TGFB2, TGFB3, TGFBR1, and TGFBR2. Since October 2016, the BGN gene was added to the panel (analyzed in 166 patients), while SCARF2, which was not associated with TAD but had previously been selected in view of a possible differential diagnosis 'Congenital contractural arachnodactyly' and 'Van den Ende-Gupta syndrome,' was excluded from routine analysis. The previously described bioinformatics read-depth-based tool XHMM was used for CNV detection in the NGS sequencing data. CNV confirmation was performed using either a home-made MLPA test, in combination with the P300 or the P200 MLPA kit of MRC Holland, or an SNP array. Detailed information on the analyzed genes and applied methodologies are available in the Supporting Materials and Methods.

Clinical data
Informed consent for NGS gene panel analysis was obtained from all 810 patients after genetic counseling by the referring physician.
The main reasons for analysis of this gene panel include familial or early onset aortic aneurysms or dissections or signs of generalized connective tissue disorders. The majority of patients was referred by a clinical geneticist who frequently participated in a multidisciplinary team specialized in connective tissue disorders. A standardized survey was sent to the referring physicians in order to collect the medical data of patients carrying an identified genetic variant (including ophthalmologic and cardiologic findings, family history, and physical examination).
Written informed consent was obtained from the patients and/or their parents with an aberration detected by XHMM, as more detailed medical data were published. Under Dutch law, assessment of the study protocol by our ethics committee was not indicated because only genetic and clinical data collected during regular patient care were used.

Genetic and clinical data in patients with variants identified by variant-calling analysis
In addition, 90 VUS were identified (patients 9, 52, 67-150; Table 1 and Supporting Information Table S1). In six patients (patients 9 and 52 in Table 1

Genetic and clinical data in patients with a CNV identified by XHMM analysis
The results of the XHMM analysis in the six patients with a CNV (patients 61-66) are depicted in Figure 1 and are summarized in Table 3.
In patient 61, a deletion of two exons in the MYLK gene was identified (NM_053025.3: c.(2390+1_2391-1)_(3448+1_3449-1)del). This deletion is predicted to generate an out-of-frame deletion in the long transcript of the MYLK gene (NM_053025.3) and a loss of the first 682 coding nucleotides, including the alternative translation initiation codon in the smooth-muscle cell-specific transcript encoding isoform 5 (Uniprot Q15746-7). This male patient was diagnosed with a type B dissection at the age of 60 years and developed a type A dissection at the age of 65 years. He was treated surgically (Bentall procedure).
Medical history and physical examination did not reveal any other signs of a connective tissue disorder. Pedigree analysis revealed that his sister suddenly died at the age of 53 years. No medical records, autopsy, or DNA were available. The 35-year-old son of the index patient did not carry the two-exon deletion of MYLK. Until now, no other relatives opted for genetic testing.
In patient 62, a deletion of one exon of PRKG1 was detected (NM_001098512.2: c.(433+1_434-1)_(547+1_548-1)del). This deletion is predicted to lead to an in-frame deletion of 39 amino acids and the insertion of an Alanine residue and encompasses a large part of the high-affinity cGMP-binding domain of the PRKG1 protein including Arginine177. A recurrent substitution of this arginine for glutamine has been reported in patients with H-TAD and shown to have a gainof-function effect (Guo et al., 2013). At the age of 35 years, this male patient was diagnosed with an aortic root dilatation, a type A dissection, aortic valve insufficiency, and dilated cardiomyopathy.

2012;
Stheneur et al.,   No further alignment available. c This family is recently described in literature (Overwater & Houweling, 2017). d A variant of unknown significance was identified in these patients as well (Supporting Information Table S1).   (Overwater & Houweling, 2017). d A variant of unknown significance was identified in these patients as well (Supporting Information Table S1). e Paternity and maternity not confirmed. vious studies that identified mutations in 10.3% to 35.5% (Campens et al., 2015;Lerner-Ellis et al., 2014;Poninska et al., 2016;Proost et al., 2015;Wooderchak-Donahue et al., 2015;Ziganshin et al., 2015). This wide range is likely to be explained by differences in clinical and demo- )exon deletions did not differ notably from the known phenotypic manifestations related to variants in these genes. Moreover, a large duplication including the whole NOTCH1 gene and a large deletion F I G U R E 2 Further characterization of XHMM results by additional (cyto-) genetic testing. BAF, B allele frequency; Chr, chromosome; der, derivate chromosome; LLR, log R ratio; FISH, fluorescence in situ hybridization. A, SNP array profile of chromosomes 7 and 9 are shown on the left. The top plot of each image shows the LRR, which provides an estimation of the copy number for each marker aligned to its chromosomal position. The bottom plot of each image shows the BAF for each SNP aligned to its chromosomal position. SNP array analysis revealed a terminal copy-number loss at 7p22.3 (2Mb; GRCh37; chr7:43360-2067625) indicated with a red arrow and a terminal copy-number gain at 9q33.3-q34.43 (11.8Mb; GRCh37; chr9:129172353-141020389) indicated with a green arrow. Chromosomes 7 and 9 from the index (left) with the unbalanced translocation and the father (right) carrying the balanced translocation are shown on the right. The breakpoints of the reciprocal translocation are indicated with an arrow. The index has the derivative chromosome 7 lacking a short segment from the short arm of chromosome 7 that is replaced by an extra copy of a terminal segment of chromosome 9q. The father has two derivative chromosomes 7 and 9, each carrying a segment of the other chromosome. B, SNP array profile of chromosome 22 is shown on the left. SNP array analysis revealed a copy-number loss at 22q11.2 (3.2Mb; GRCh37; chr22:20779645_20792061) indicated with a red arrow. The results of metaphase FISH on blood from the mother is presented on the right. The 22q11.2 region is recognized by the HIRA probe, producing a red signal. The green signal is from the ARSA probe hybridizing with the ARSA gene on chromosome band 22q13.33. The 22q11.2 deletion is indicated by a blue arrow. Metaphase FISH analysis revealed that the mother is also a carrier of the 22q11.2 deletion (ish del(22)(q11.2q11.2)(HIRA-)) encompassing SCARF2 were detected by XHMM analysis.
The results of this study underline the importance of CNV analysis using a bioinformatics tool such as XHMM in the clinical diagnostic care for TAAD patients. As CNV analysis is often not routinely performed for most genes included in this NGS platform, these CNVs would not have been detected by regular genetic analysis. Four of the six detected CNVs in this study were small intragenic deletions (two single-exon deletions, one 2-exon, and one 4-exon deletion). These are generally not detected by routine CGH or SNP array analysis. This highlights the importance of using a CNV detection tool, which allows detection of CNVs with (small) single-exon resolution. Based on the results of this study, single-exon-sensitive deletion/duplication analysis on a routine basis should be recommended in patients suspected of H-TAD.

CONCLUSION
In 66 of 810 (8.1%) patients suspected of H-TAD, a pathogenic or likely pathogenic variant was identified using our NGS gene panel in combination with XHMM analysis. Six of these 66 pathogenic or likely pathogenic variants (9.1%) were a CNV, not detectable by routine NGS analysis. This study is the first to describe the incremental yield of CNV analysis in patients suspected of H-TAD. Our study underscores the importance of CNV analysis using a bioinformatics tool such as XHMM in the clinical diagnostic care for H-TAD patients.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE
Under Dutch law, assessment of the study protocol by our ethics committee was not indicated because only genetic and clinical data collected during regular patient care were used.

CONSENT FOR PUBLICATION
Written informed consent was obtained from the patients and/or their parents with an aberration detected by XHMM, as more detailed medical data were published. Informed consent for DNA diagnostics was obtained from all 810 patients after genetic counseling by the referring physician.

AVAILABILITY OF DATA AND MATERIAL
All data and protocols used for this study are either included in the article (or in its supporting files) or are available upon request.