1 INTRODUCTION

Exome and genome sequencing is increasingly used in the molecular diagnostics of Mendelian disorders. In addition to disease-causing alterations, clinical germline sequencing may reveal pathogenic variants in other genes, unrelated to the primary diagnosis (secondary findings), but of medical value to the patients or their family members. To improve patient care, the American College of Medical Genetics and Genomics (ACMG) recommended reporting of known/expected pathogenic variants in a set of medically actionable genes.^{1, 2} The current ACMG “secondary findings” (SF) list (ACMG SF v2.0)² contains 59 genes that were selected by a team of experts on the basis of their association with common, highly penetrant disorders; severity of a phenotype with substantial health implications; established evidence for effective preventive care and/or treatment options; indefinite length of an asymptomatic phase in which preventive actions may reduce disease risk; and availability of molecular testing that could be used to diagnose family members at-risk. Recognizing the limitations in technology, the ACMG recommendations have focused on the reporting of exome and genome sequencing variants, avoided interpretations of disorders caused by structural variants, repeat expansions, or copy number variations (CNVs).¹ The main reasons for reporting secondary findings to ordering clinicians are to improve clinical care and to reduce morbidity and mortality in patients undergoing clinical genome-wide sequencing for concomitant conditions.

Since the first publication of the ACMG recommendations, clinical laboratories offering clinical genomic sequencing have begun to disclose medically actionable secondary findings to patients opting to receive such results. Multiple studies indicate that, reportable pathogenic variants for the ACMG-recommended genes are present in at least 1.2%–2.5% of tested individuals^3-6 and can be found more frequent in specific ethnic populations.⁷ It is anticipated that a significant number of people—including subjects undergoing next-generation sequencing (NGS), such as exome or genome sequencing, and their family members—may benefit from prevention or medical intervention for these life-threatening conditions. CNVs, a highly prevalent type of genomic alterations, are well-recognized causes for many Mendelian disorders and contiguous gene syndromes making it likely that pathogenic CNVs would be present in the 59-ACMG genes. In 2010, the ACMG recommended that chromosomal microarray analysis (CMA) be used as the first-tier test in patients with neurodevelopmental disorders and/or congenital anomalies as whole-genome arrays yield pathogenic CNVs in 15%–22% of affected individuals.^{8, 9} Importantly, these patients may harbor medically actionable dosage-sensitive genes within the deleted or duplicated segments and are unlikely to be tested by exome/genome sequencing.^10-14 Whole gene duplication or deletion CNVs can be responsible for up 80% of the molecular diagnoses involving dosage-sensitive genes.¹⁵ Relevantly, intragenic CNVs account for about 10% of disease-causing variants^16-18 and represent another type of finding that can be detected by CMA or clinical exome/genome sequencing in laboratories that have adopted copy number calling pipelines.^{13, 14, 19, 20} Although the reporting of pathogenic single-nucleotide variants (SNVs) in sequencing has been recommended, no guidelines exist regarding the returning of equally significant CNVs involving medically actionable genes. To facilitate such discussion, we evaluated—from among 8865 patients tested by clinical CMA in our laboratory—the prevalence, nature, and consequences of CNVs involving the 59 genes currently recommended by the ACMG for the reporting of secondary findings.

2 PATIENTS AND METHODS

This study was approved by the IRB at the University of Pittsburgh (PRO13060436; PRO16060611; STUDY20010147). We retrospectively reviewed the findings of CMA performed by the Pittsburgh Cytogenetic Laboratory within an 8-year period (January 2011–December 2018) on 8865 pediatric and adult patients referred for genetic testing. DNA was obtained from peripheral blood samples and evaluated for CNVs by array comparative genomic hybridization (aCGH) using either a 135 K CGH (Roche NimbleGen, Madison, WI) or a 180 K CGH + SNP oligonucleotide array (ISCA design, Agilent, Santa Clara, CA).²¹ Parental blood samples were tested by karyotype, fluorescence in situ hybridization (FISH), or CMA to establish the inheritance mode in cases with pathogenic alterations as well as variants of uncertain clinical significance. To determine the relevance of CNVs to a phenotype for each gene in the list, we reviewed whether a clinically actionable phenotype is instigated by either a haploinsufficiency, extra gene dosage, or a dominant-negative effect, which can be caused by an entire gene deletion, entire gene duplication, or intragenic deletion/duplications, respectively. To establish evidence for dosage pathogenicity for a medically actionable phenotype, information regarding a mechanism of action, type, and position of reported variants, haploinsufficiency and triplosensitivity, gross genomic abnormalities, and intragenic alterations was collected from ClinGen (https://dosage.clinicalgenome.org/acmg.shtml), ClinVar Tools (www.ncbi.nlm.nih.gov/clinvar/docs/acmg/), the Human Gene Mutation Database, and published literature on 59 genes. The laboratory database search was conducted to identify patients with relevant CNVs comprising the 59 actionable genes recommended by the ACMG to be reported as the secondary findings in patients undergoing clinical exome sequencing.² Patients' clinical presentations were obtained from a retrospective medical chart review.

3 RESULTS

CNVs (involving either an entire gene or only a part) have not been described in the literature for 8 of the 59 genes. For 14 genes, the clinical phenotype has been observed in patients who had intragenic deletions or duplications that result in the synthesis of abnormal protein molecules, while the gain or loss in copy number of the entire gene does not result in disease. For 36 of the remaining 37 genes, haploinsufficiency and loss-of-function variants are considered to be disease-causing or disease-predisposing; therefore, entire gene deletions as well as intragenic CNVs are known or expected to be pathogenic (Table 1). For one gene (PCSK9), the whole gene duplication has been reported as a pathogenic CNV leading to severe hypercholesterolemia.²² For two genes (TGFBR1 and MYH11) implicated in autosomal dominant aortic aneurysm syndromes, we identified publications proposing a gain in copy number (duplication) of an entire gene as a mechanism for thoracic aortic aneurysm and dissection.^{23, 24} Gross genomic duplications have not been reported in patients with associated phenotypes for 56/59 actionable genes. In the ClinGen database, none of the 59 actionable genes is listed as a triplesensitive. Collective evidence from the published literature, the ClinGen haploinsufficiency/triplosensitivity database, and a gene's mode of action is summarized in Table 1 to outline CNV types of a likely pathogenic and pathogenic significance.

Based on this information we conducted a search through our database of microarray results and identified 23/8865 (0.26%) relevant patients (Table 2) with either deletions comprising the entire gene or intragenic alterations involving medically actionable gene(s). In addition, we identified 48 unrelated patients with a CNV in 16p13.11 comprising the MYH11 gene, including 14 probands with microdeletions and 34/8865 (0.38%) with microduplications (27 with isolated 16p13.11 gains and seven patients with additional CNVs on other chromosomes).

Among the 23 patients, two (PM7 and PM11) were the affected parents who had microarray testing due to the finding of CNVs in their children (PM6 and PM10, respectively). Parental testing has been completed on 11 patients. Pathogenic CNVs were found to be de novo in six individuals, inherited from a parent in four patients, and resulted from a maternal balanced chromosome rearrangement in one proband.

Medically actionable genes were divided into four categories based on the risk for a cancer predisposition (25 genes), sudden cardiac arrest (SCA, 20 genes), thoracic aortic disease (TAD, 7 genes), and metabolic disorder (Met, 7 genes). A total of 17/59 (29%) genes were affected by CNVs in our study (Figure 1(A)). The affected cohort included seven children (PM2-4, PM17, PM20-22) with deletions of four cancer-predisposing genes, 11 patients (PM1, PM6-9, PM12-14, PM16, PM19) with CNVs in eight SCA genes, including one patient (PM15) with a deletion of both SCA and cancer-predisposing genes, four individuals (PM5, PM10, PM11, PM23) with three genes implicated in TAD, and one woman (PM18) with a metabolic disorder (Table 2). Overall, 15/23 (∼65%) of patients in our group had pathogenic CNVs affecting genes associated with cardiomyopathy, channelopathy, thoracic aortic dissection, or an increased risk for sudden death. Overall, the SCA- and TAD-risk genes show the highest probability of pathogenic CNVs per gene in our cohort of patients tested by CMA (Figure 1(B)).

3.1 Hot spots for genomic rearrangement comprising actionable SF genes

In our cohort, CNVs in four genomic regions reoccurred among unrelated individuals. We observed three patients (PM2-PM4, Table 2, Figure 2(A)) carrying almost identical 10q22.3q23.2 deletions encompassing the BMPR1A gene; three patients (PM6, PM8, PM9) with interstitial deletions of variable size in the subtelomeric region of the short arm of chromosome 6 (6p24.3) comprising the DSP gene; two children (PM12, PM13) with terminal 7q deletions containing two actionable genes, KCNH2 and PRKAG2, in each case; and interstitial 18q21q22 deletions comprising the SMAD4 gene in two patients (PM21, PM22). The ACTC1 gene represents another hot spot region, recurrent deletions in which might be mediated by low-copy repeats (LCRs) as seen in our patient PM1 (Figure 2(B)) and those reported previously²⁵ with the 15q14 deletion syndrome (OMIM#616898).

4 DISCUSSION

Implementation of advanced genome-wide tools, such as CMA, genome and exome sequencing, into clinical diagnosis raises important questions about the standardization of reportable results, including terminology, criteria for classification, approaches for evaluation of supporting evidence to establish pathogenicity and the clinical significance of copy number and sequence variants in clinical genetic laboratories. The ACMG recommendations for reporting of secondary findings are a step toward personalized medicine, enabling clinicians to obtain health-related genetic information for patients and choose the most appropriate course of clinical care through preventive monitoring, prophylactic procedures, and targeted treatment. Notably, the minimum 59-gene list recommended for reporting in clinical sequencing includes 36 haploinsufficient genes meaning that deletion of a gene, in part or entirety, is expected to cause the same effect as a loss-of-function variant. The current guidelines for interpretation and reporting of postnatal constitutional CNVs³² have general recommendations for the reporting of chromosomal imbalances associated with the risk of neoplasia and for presymptomatic conditions; however, the list of such conditions is not defined. Moreover, laboratories are not obligated to report these findings and, due to an increased number of genes associated with cancer predisposition and presymptomatic conditions, may adopt nondisclosure policies. There are currently no professional guidelines in regards to interpretation and reporting of CNVs for the 59 secondary findings genes detected via CMA.

At the time of the original CMA guideline, answers to the questions regarding clinical actionability and potential medical benefit for a set of specific genes were imperative, as the scale and clinical utility of such findings were unknown. Over the last years, numerous research and clinical studies have significantly advanced our knowledge, revealing the implications of genome sequencing findings and options for treatment or prevention. Questions have surfaced, however, concerning the clinical significance of CNV detection involving secondary finding genes, standards and availability of supporting tools for CNV interpretation and reporting, specific consideration for assessment of CNV pathogenicity, and applicability of preventive measures to patients with CNVs. Should CMA interpretation guidelines be revised, including recommendation on CNVs affecting SF genes? CNVs in what medically actionable set of genes are appropriate for reporting? In this study, we try to address some of the questions raised above and provide ground for further discussion.

We reviewed the results of microarray testing in our laboratory and determined that 0.26% of tested patients have pathogenic CNVs affecting one or more of the 59 secondary findings genes indicated by the ACMG. In addition, 0.38% of patients carry a duplication of the MYH11 gene, which appears to be a risk factor for aortic aneurysm and sudden death.^{24, 28} In at least two families, CMA results of the proband prompted parental testing and accurate diagnosis in the affected parent. For example, patients PM6 and PM10 were referred to genetic testing due to neurodevelopmental problems. In each patient, CMA revealed a deletion comprising multiple genes, including the DSP gene and FBN1 gene in PM6 and PM10, respectively. The affected parents (PM7 and PM11) were tested and found to carry the same deletion as their children, asymptomatic for cardiac phenotype. These findings were consistent with manifestations of arrhythmogenic right ventricular dysplasia in PM7 and Marfan syndrome in PM11. Our study suggests that the reporting of CNVs for secondary finding genes is equally significant as the reporting of pathogenic SNVs in clinical exome /genome sequencing. It has become necessary to develop a guideline for returning secondary results to the interested patients/families when CMA is performed, comparable to the NGS approach. Interpretation of CNVs for secondary finding genes requires professional assessment and a clear statement regarding pathogenicity in the report. In our cohort,19/23 patients have deletions encompassing at least one haploinsufficient gene from the secondary findings list. This includes eight patients with loss of tumor suppressor genes and 11 patients with other disorders. Since loss in copy of a tumor suppressor gene is commonly associated with a risk of neoplasia, interpretation of whole gene deletion is straightforward. Assessment of clinical significance for CNVs involving other genes in an asymptomatic patient is more challenging and requires detailed analysis of gene structure and mode of action, knowledge of haploinsufficiency score, penetrance, published supporting evidence, and available clinical management. Recent ACMG technical standards for interpretation of constitutional CNVs and consensus recommendations for CNV reporting of cancer susceptibility genes are important steps toward standardized and harmonized guidelines for returning of SNV and CNV results.^{33, 34}

Ordering clinicians and genetic counselors often view the results of microarray analysis as a set of the genes that are deleted or duplicated in their patients and examine each gene for its association with a known disease in the OMIM database. Unless a clear interpretation of the CNV for a specific gene is included in the laboratory report, this approach may lead to a misinterpretation of clinical significance. This is particularly relevant to genes for which entire gene deletion or duplication is not associated with disease-risk phenotype (Table 1).

In our study, the search for patients with the relevant CNVs revealed two patients (PM12 and PM13) with deletion in the distal 7q36.1q36.3 comprising the PRKAG2 and KCNH2 genes. Since autosomal dominant hypertrophic cardiomyopathy is caused by PRKAG2 missense variants via a gain-of-function mechanism, a loss of the entire gene should not be considered pathogenic, diagnostic, or actionable.³⁵ In contrast, loss in copy number for KCNH2 is pathogenic.³⁶ The presence of intragenic deletions and duplications and the possible formation of a truncated protein should be emphasized in the reports and considered by clinicians if such a statement is absent from interpretation. Patients with pathogenic findings often have multiple genes affected by CNVs.

As recommended, all genes affected by reportable CNVs are included in the laboratory CMA reports; however, it is impractical to provide detailed assessment and clinical significance, as discussed above, for each gene. A predefined list of medically actionable genes and reporting criteria for each condition are essential to eliminate inconsistency and misinterpretation of the clinical significance of CNVs, particularly those associated with noncancerous presymptomatic conditions.

Another challenge is interpretation of CNVs involving a nonhaploinsufficient gene. In PM15 (Table 2), a deletion involving the 3p chromosomal region encompasses the MLH1 gene (pathogenic finding) and the 3′ end of the SCN5A gene, while the retained gene segment, including the promoter, 5'UTR, and exons 1–13 is expected to produce a truncated protein. There is some limited evidence that SCN5A is a dosage-sensitive gene; therefore, whole gene deletions are not likely to be associated with Long QT or Brugada syndrome. Patients with partial gene deletions in SCN5A, resulting in a prematurely truncated protein, were reported to have a clinical phenotype similar to those harboring missense mutations, while clinical significance of deletions embracing the whole SCN5A gene or the 5′ end of the gene are likely to be benign. The CNV pathogenicity can be ascertained through multiple resources of the ClinGen CNV database, such as the ClinGen Dosage Sensitivity Map and ClinGen CNV Interpretation Calculator, and should be further curated and readily available for all medically actionable genes, particularly those recommended by the ACMG for reporting. A standard approach to reporting both CNVs and other variants identified by sequencing for secondary finding genes would minimize inconsistency in reporting and reduce the burden on physicians and other providers.

Our study determined a few chromosomal regions with hot spots for pathogenic CNVs containing medically actionable genes. These include the MYH11 and BMPR1A genes (Figure 2), duplications and deletions of which are mediated by flanking repetitive sequences; DSP and KCNH2 genes, which are located in the subtelomeric segments and have a higher risk for CNVs than genes mapped in other chromosomal regions. Other known recurrent rearrangements, mediated by Alu-Alu recombination or founder mutations such as deletions of exon 3 of the RYR2 gene³⁷ and multi-exon deletion in the MYBP3 gene,³⁸ are more likely to be detected in clinical exome/genome sequencing.

The reporting of pathogenic CNVs as a secondary finding can also benefit at-risk relatives who may have inherited the same pathogenic variant and require surveillance or preventive management to reduce disease complications. The OTC gene was added to version 2 of the secondary findings list due to a possibility of OTC deficiency in 20%–30% of women carrying pathogenic gene alterations.³⁹ Among our patients, a 17-year-old female (PM18) was admitted to hospital with altered mental status and hyperammonemia. Her family history included an older brother who died a few days after birth due to hyperammonemia, and multiple maternal male relatives who died in infancy. However, she had not been tested for OTC gene alterations prior to hospitalization. The OTC gene sequencing results were negative; she was tested by CMA to rule out a carrier status of OTC deficiency. CMA detected a 9.5 Mb deletion in Xp21 comprising 34 genes, including nine disease-causing genes: IL1RAPL1, NR0B1, GK, DMD, XK, CYBB, RPGR, OTC, and TSPAN7. Deletions account for at least 10%–15% of pathogenic findings in patients with OTC deficiency and, in heterozygous female carriers, can lead to a chronic brain and liver damage and a rapid onset of hyperammonemia in the absence of available prophylactic interventions, as observed in our patient. Female carriers of OTC pathogenic variants are also at risk of hyperammonemic complications during pregnancy.

Another important fact is that, besides OTC, the CYBB gene is also deleted in this patient. Pathogenic variants and deletions of CYBB are the cause of chronic granulomatous disease (OMIM#306400), an immunodeficiency disorder characterized by severe recurrent bacterial and fungal infections. Chronic granulomatous disease is associated with high morbidity and mortality in affected males and some heterozygous female carriers. Early diagnosis of patients and relatives at-risk allows the use of antimicrobial prophylaxis, early detection, and treatment of minimally symptomatic infections, minimizing noninfectious complications such as liver involvement, colitis, and pulmonary fibrosis. Likewise, other X-linked immunodeficiency conditions, such as Wiskott–Aldrich syndrome (WAS, OMIM#301000), immuno-dysregulation, polyendocrinopathy, enteropathy (FOXP3, OMIM#304790), severe combined immunodeficiency (IL2RG, OMIM#300400), agammaglobulinemia (BTK, OMIM#300755), lymphoproliferative syndrome (XIAP and SH2D1A, OMIM#300635), hyper IgM syndrome (CD40LG, OMIM#308230), and properdin deficiency (PFC, OMIM#312060), are associated with severe life-threatening manifestations and may be prevented or treated if diagnosed early. Whole gene deletions and intragenic CNVs are seen in 5%–15% of affected children with these diseases.⁴⁰ Patients with immunodeficiency disorders commonly present with failure to thrive, anemia, diarrhea, and neurologic complications, and may undergo CMA and genome/exome sequencing for clinical diagnosis. These genes meet the criteria defined for the secondary finding genes; therefore, we propose to include these genes into the list for reporting in both clinical CMA and genome/exome sequencing.

Pathogenic variants for the 59-ACMG genes are present in at least 1.2%–2.5% of individuals tested by exome/genome sequencing. The prevalence of secondary findings in our cohort (0.26%) is consistent with the knowledge that at least 10% of affected patients have pathogenic CNVs.^16-18 However, the impact of pathogenic CNVs might be much higher if the 16p13.11 duplications containing the MYH11 gene are included in the calculation. Also, there is a limitation in detection of intragenic CNVs by CMA in genes that are small in size or do not have sufficient probe coverage. Contiguous gene deletion/duplication CNVs with a breakpoint located within the gene were observed in five of our patients with partial deletion of SCN5A (PM15), partial KCNQ1 gene duplication, including promoter and the 5′ end of the gene (PM14), exon 1 deletion of the DSP gene (PM6, 7), and partial MYH11 gene duplication (Supplementary Table S1, patient PR27). In the microarray design we use for CMA, 30/59 genes do not have the coverage necessary to detect intragenic CNVs, including 11/59 genes for which only intragenic CNVs are expected to be pathogenic.

Possible CNV detection by cytogenetic and molecular methods has become a natural bridge between the molecular genetics and cytogenetics disciplines. Many groups are evaluating the accuracy of exome sequencing and virtual gene panels for detecting whole gene and exonic CNVs and developing new algorithms to enhance CNV- calling capabilities in a single clinical test. With the advance of NGS and incorporation of CNV-calling algorithms into the exome/genome sequencing pipelines, the rate of detected pathogenic CNVs is anticipated to grow. Developing guidelines for CNV assessment, consolidation of SNV and CNV detection methods, and standardization of approaches in CMA and genome-wide, sequence-based clinical testing is an essential and timely step that will enable the most accurate and effective reporting for medically actionable genes.

CONFLICT OF INTEREST

The authors have no conflict of interest to report.