Diagnostic yield of exome sequencing-based copy number variation analysis in Mendelian disorders: a clinical application

Abstract

Next-generation sequencing (NGS) coupled with bioinformatic tools has revolutionized the detection of copy number variations (CNVs), which are implicated in the emergence of Mendelian disorders. In this study, we evaluated the diagnostic yield of exome sequencing-based CNV analysis in 449 patients with suspected Mendelian disorders. We aimed to assess the diagnostic yield of this recently utilized method and expand the clinical spectrum of intragenic CNVs. The cohort underwent whole exome sequencing (WES) and clinical exome sequencing (CES). Using GATK-gCNV, we identified 12 pathogenic CNVs that correlated with their clinical findings and resulting in a diagnostic yield of 2.67%. Importantly, the study emphasizes the role of CNVs in the etiology of Mendelian disorders and highlights the value of exome sequencing-based CNV analysis in routine diagnostic processes.

Introduction

Copy number variations (CNVs), including deletions and duplications within the genome, play a pivotal role in genetic diversity and disease development. Detecting CNVs is crucial for diagnosing a variety of disorders [1]. Traditional techniques such as microarrays and Multiplex Ligation-dependent Probe Amplification (MLPA) have commonly been used for CNV detection, but the introduction of next-generation sequencing (NGS) has revolutionized the field due to its efficiency and cost-effectiveness [2,3,4].

The microarray has been a powerful tool for CNV detection in the human genome. However, it has limitations, particularly in resolution. While it can accurately detect CNVs of a certain size depending on the method used, its ability to identify smaller variants, such as exonic deletions and duplications, is limited [5]. Conversely, MLPA is a valuable technique for targeting specific genes and regions of interest. As a PCR-based technology, it uses probes to selectively bind and amplify targeted sequences. However, it requires the design and synthesis of specific probes for each gene or region of interest [6].

CNV detection can be accomplished by analyzing the depth of coverage of short reads generated by NGS. This process involves dividing the genome into specific regions, or ‘bins’ (e.g., 100 base pairs), predetermined by the laboratory. After sequencing, read alignment, data filtering, and normalization, the number of reads in each bin of the patient sample is compared with the corresponding bin in a reference set composed of normal samples. Sophisticated algorithms are employed to detect deletions and duplications by comparing the read depths. The resulting data, which is used to calculate relative copy number information, can be visualized and interpreted through a user interface for further analysis [4, 7].

In this study, we evaluated 12 patients diagnosed using NGS-based CNV analysis. Our aim was to underline the role of intragenic deletions and duplications in the etiology of Mendelian disorders and evaluate the diagnostic yield of NGS-based CNV analysis. By using this advanced genomic approach, we sought to gain insights into the genetic mechanisms underlying these disorders and enhance our understanding of the clinical implications of CNVs.

Materials and methods

Ethical approval

The present study was approved by the Ethics Committee of Ege University Faculty of Medicine (Reference No: 23-7T/27). All procedures involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with Helsinki Declaration. Written informed consent was obtained from the parents or guardians of the participants, ensuring transparency and adherence to privacy regulations.

Patients

A total of 215 whole exome sequencing (WES) analyses and 234 clinical exome sequencing (CES) analyses were performed at the Ege University Faculty of Medicine, Department of Pediatric Genetics, from September 1, 2021, to November 1, 2022. These analyses included a cohort of patients aged 1–18 years, all of whom presented with diverse clinical findings suggestive of monogenic disorders. The exome sequencing-based CNV analysis identified 12 patients with a molecular etiology. The cohort included patients with various genetic indications, such as cardiomyopathy, lactic acidosis, spastic paraplegia, neuronal ceroid lipofuscinosis, epidermolysis bullosa, unidentified vision loss with deafness, and preliminary diagnoses of Kabuki, Bardet-Biedl, Cornelia de Lange, and Ehlers-Danlos syndromes.

Sample collection, clinical exome sequencing, whole exome sequencing, and CNV analysis

Sample Collection: Peripheral venous blood samples (2 ml) were collected from patients and their family members. Genomic DNA was extracted from these samples using the MagNA Pure 96 DNA and Viral NA Small Volume Kit.

Sample Preparation: CES was performed using the Roche HyperCap Designed Share (DS) Inherited Disease Panel kit covering 4125 genes, while WES was executed using the Roche Kapa Hyperexome kit. Samples were prepared following the respective kit protocols, which involved capturing exonic regions of interest using targeted probes.

Sequencing: The prepared CES and WES samples were sequenced on the MGI sequencing platform, DNBSEQ-G400. This process generated raw sequencing data in the form of short reads, representing the DNA fragments.

Variant Classification and Analysis: The raw sequencing data (FASTQ files) were uploaded to the SEQ Platform developed by Genomize Inc. The reads were aligned to the human reference genome GRCh37 using the Burrows-Wheeler Aligner [8]. Variant calling was performed using FreeBayes [9]., followed by additional steps such as PCR deduplication and in-del realignment using Genomize’s proprietary algorithms. Identified variants were annotated using VEP v102 [10] to provide functional annotations. ACMG pathogenicity classification was conducted using Genomize’s proprietary algorithm based on the guideline published by Richards et al. [11].

Copy Number Variation Analysis: CNV analysis was performed using the GATK-gCNV tool, part of the Genome Analysis Toolkit (GATK), version 4.1.8.1. GATK-gCNV utilizes a Bayesian algorithm designed to detect rare CNVs from whole-exome sequencing (WES) data by analyzing sequencing read-depth information. Data Preparation and Preprocessing: Binning was disabled, and all intervals targeted by the whole exome and clinical exome were padded by 250 base pairs (bps). Initially, genomic regions were filtered based on GC content, mappability, and segmental duplication content. Regions with a GC content below 0.1 or above 0.9 were excluded to mitigate GC bias. Additionally, regions with a mappability score below 0.9 or above 1.0 were filtered out to ensure high confidence in read alignment. Genomic intervals with segmental duplication content above 0.5 were excluded to avoid potential false positives. Subsequently, intervals with fewer than 10 read counts in more than 50% of the cohort were filtered out. Extreme read counts falling within the 1st percentile (minimum) and 99th percentile (maximum) across 90% of the cohort were also excluded from further analysis. CNV Filtering and Evaluation: The detected CNVs were first filtered based on their frequency within the same NGS run. CNVs that were observed in multiple patients with different phenotypes were excluded from further analysis. These remaining CNVs were then evaluated using databases such as DGV, ClinVar, and OMIM to assess their pathogenicity. Following this process, pathogenic, likely pathogenic, and variants of uncertain significance that were considered to explain the patient’s clinical findings were reviewed for read depth in IGV (Integrative Genomics Viewer). For clinically relevant CNVs, IGV was used to validate the findings. In cases of homozygous deletions, IGV displayed no reads in the affected region, while in heterozygous deletions, SNPs within the region appeared as homozygous.

Results

The exome sequencing-based CNV analysis yielded an overall diagnostic yield of 2.67% (12/449) in our study. The patients enrolled in the study presented with various indications for genetic testing, including cardiomyopathy, lactic acidosis, spastic paraplegia, neuronal ceroid lipofuscinosis, epidermolysis bullosa, unidentified vision loss with deafness, and preliminary diagnoses of Kabuki, Bardet Biedl, Cornelia de Lange, and Ehlers Danlos syndromes. Clinical exome sequencing (CES) was performed for 5 patients, whole exome sequencing (WES) for 3 patients, and trio-WES for 4 patients. The NGS-based CNV analysis revealed 12 distinct CNVs that correlated with the patients’ clinical presentations. These included monoallelic deletions in the KMT2D, EIF5A, and LAMP2 genes, as well as biallelic deletions in the ZFYVE26, AP4S1, PPT1, PDHX, BBS9, COL7A1, and RPGRIP1 genes. Furthermore, a monoallelic duplication was identified in the HDAC8 gene, and a biallelic duplication was observed in the PLOD1 gene. Biallelic deletions were seen in the Integrative Genomics Viewer (IGV). Figure 1A, B, C and D display coverage images from some the patients with homozygous deletions alongside controls, as visualized in the IGV.

Fig. 1

ABiallelic deletion of the Exon 13–19 in the ZFYVE26 gene in the Patient 3. B Biallelic deletion of exons 13 to 24 in the COL7A1 gene was in Patient 4. C Biallelic deletion of the Exon 4 in the AP4S1 gene in the patient 9 and affected siblings. D Biallelic deletion of the Exon 19 in the RPGRIP1 gene in the Patient 10.

The sizes of the detected CNVs ranged from approximately 485 base pairs (bp) for the smallest deletion to 228,307 bp for the largest deletion.

Table 1 summarizes the clinical findings of the patients and presents the results of the NGS-based CNV analysis., Table 2 shows the average read depth, 20X and 50X coverage, and the total number of CNVs detected per sample. Figure 2 illustrates the dysmorphologic features of patients 1, 2, and 8.

Table 1 Clinical and molecular features of the patients

Table 2 The average read depth, 20X and 50X coverage, and the total number of CNVs detected per sample

Fig. 2

Dysmorphologic features of patients 1, 2, and 8. (Patient 1 has arched eyebrows, synophrys, long eyelashes, depressed nasal bridge, long filtrum and pointed chin. Patient 2 has arched eyebrows, long palpebral fissures, eversion of lateral third of lower eyelids, left iris coloboma, large prominent ears, thin upper lip vermillion. Patient 8 has laterally sparsed eyebrowes, long eyelashes, long and down-slanting palpebral fissures, bulbous nose and cupped ears)

Family segregation studies utilizing NGS-based CNV analysis were conducted in six patients. The results of the segregation analysis in all six patients were consistent with the inheritance models of the diseases. Specifically, trio-WES was performed for four patients (Patients 4, 7, 8, and 11). For the parents of Patient 12, as well as the parents and affected siblings of Patient 9, CES was carried out. It is worth noting that in two patients (Patient 10 and 12), a dual molecular diagnosis was established, with one diagnosis attributed to a single nucleotide variant and the other diagnosis resulting from a CNV. Below, we present detailed clinical findings for some of the patients. Further information about the remaining patients can be accessed through the supplementary material.

Patient 8

Patient 8 was a 4-year-old boy evaluated due to microcephaly, failure to thrive, developmental delay and dysmorphic features including laterally sparsed eyebrowes, long eyelashes, long and down slanting palpebral fissures, bulbous nose and cupped ears suggesting Kabuki syndrome. Trio WES was performed and there was no disease-causing single nucleotide variation. CNV analysis revealed a 228-kb heterozygous deletion on chromosomal region 17p13.1 including EIF5A and DLG4 genes responsible for Faundes-Banka syndrome and Autosomal dominant intellectual developmental disorder-62, respectively. NGS-based CNV analysis in the parents’ WES did not detect this CNV, suggesting its de novo origin.

Patient 10

Patient 10 was a 14-month-old girl born to consanguineous marriage presented with mild developmental delay, congenital sensorineural deafness, decrease in visual acuity, horizontal and rotatuar nistagmus. On physical examination, she had sparse hair and eyebrows, frontal bossing, esotropia in the right eye, depressed nasal bridge, bulbous nasal tip and retromicrognathia. Optic nerves were evaluated as normal and retinal vessels were found thinned in the eye examination. In the CES analysis, there was a homozygous, likely pathogenic variant (c.3679 C > T, p.Arg1227Ter) in the OTOF (NM_194248.3) gene (Fig. 3) responsible for autosomal recessive nonsyndromic hearing loss 9 consistent with sensorineural hearing loss. CNV analysis revealed a homozygous, 620 bp deletion including exon 19 of the RPGRIP1 (NM_020366.3) gene responsible for Cone-rod dystrophy 13 and Leber congenital amaurosis 6.

Fig. 3

Homozygous c.3679 C > T, (p.Arg1227Ter) variant in the OTOF (NM_194248.3) gene

Patient 12

Patient 12, a 9-year-old girl, presented with retinal dystrophy, postaxial polydactyly, axonal polyneuropathy, and myositis episodes during infections. Whole exome sequencing was performed and CNV analysis revealed a homozygous deletion in the BBS9 (NM_198428.2), gene specifically including Exon 18–19, which is responsible for Bardet-Biedl Syndrome type 9. Additionally, Patient 12 had a homozygous missense variant (c.487G > A, p.Gly163Ser) classified as “likely pathogenic” in the HADHB (NM_000183.3) gene which is responsible for Mitochondrial Trifunctional Protein Deficiency 1. This variant explains the patient’s myositis episodes and axonal polyneuropathy. Clinical exome sequencing were performed for segregation analysis and both of the parents were heterozygous for the deletion of exon 18–19 in the BBS9 gene and the single nucleotide variant in the HADHB gene.

Discussion

In this study, we performed exome-based CNV screening via a read-depth analysis of exome sequencing data from a large patient cohort with various phenotypes. While prior studies have demonstrated the efficacy of exome-based CNV analysis in identifying genomic CNVs, it’s noteworthy that the specificity of the technique is contingent on the employed CNV detection algorithms [2,3,4, 17]. For our cohort, the read-depth CNV analysis yielded an overall diagnostic improvement of 2.67%, spanning a diverse array of genetic disorders. This finding is in concordance with the diagnostic yield obtained through NGS-based CNV analysis in heterogeneous patient cohorts including 2418 and 2603 patients, 2% and 1,7% respectively [3, 18].

Disease-specific studies have reported varying diagnostic CNV detection rates. For instance, an 18.92% diagnostic rate was achieved using exome-based CNV analysis in a retrospective study involving 74 families afflicted with neurodevelopmental disorders [19]. In our study, patient 1, under evaluation for neurodevelopmental delay and dysmorphic features, was found to possess a duplication in exons 6–9 of the HDAC8 gene, previously documented by Kaiser et al. [13], leading to a diagnosis of Cornelia de Lange Syndrome Type 5. In patient 2, initially diagnosed with Kabuki syndrome, our NGS-based CNV analysis identified a novel deletion in the KMT2D gene. In a compelling case, Patient 8, who exhibited neurodevelopmental delay and dysmorphic features akin to those seen in Kabuki syndrome, was found to carry a de novo, heterozygous, total deletion of the EIF5A and DLG4 genes. These genes are known to be associated with Faundes-Banka syndrome and Intellectual Developmental Disorder, autosomal dominant 62, respectively. Notably, Faundes-Banka syndrome, first described by Faundes et al. in 2021 [20], is a rare disorder that arises as a result of loss-of-function mutations in the EIF5A gene. Up until now, only seven patients with this syndrome have been reported. Intriguingly, among these patients, three initially presented with clinical features that mimicked Kabuki syndrome, reminiscent of our patient’s case [20]. By identifying this unusual genetic alteration, our study contributes to broadening both the clinical and molecular spectrum associated with Faundes-Banka syndrome.

A comprehensive study spearheaded by Ceyhan-Birsoy et al., meticulously evaluated over 1400 patients diagnosed with cardiomyopathy. Their efforts led to the discovery of clinically significant CNVs in approximately 0.63% of the examined population [21]. Notably, one of the affected genes identified among the detected deletions was the LAMP2 gene responsible for Danon disease, mirroring the genetic findings observed in patient 11 from our cohort. Evidence from previous research has indicated the presence of 12 substantial deletions in the LAMP2 gene [22]. While the work by Ceyhan-Birsoy et al. suggests that the majority of cardiomyopathy cases may not be precipitated by small-scale CNVs, it is essential to not rule out this possibility, especially in instances where the fundamental molecular cause remains unidentified.

To our current knowledge, more than 80 genes have been implicated in the etiology of hereditary spastic paraplegias (HSPs) [23]. A recent comprehensive review conducted by Galatolo et al. [24], which synthesized findings from 19 WES studies, revealed that the rate of molecular etiology detection in HSPs stands at an aggregate of 49.7%. Despite these strides in understanding the genetic basis of HSPs, the role of CNVs in the onset and progression of these disorders remains insufficiently elucidated. Our study further contributes to this evolving field by identifying homozygous deletions in the ZFYVE26 and AP4S1 genes in two cases preliminarily diagnosed with HSP. These genes are known to be causative for Spastic Paraplegia 15 and Spastic Paraplegia 52, respectively. While the ZFYVE26 gene has had two major deletions previously reported in the Human Gene Mutation Database [22], our study has the distinction of reporting a substantial deletion in the AP4S1 gene for the first time.

In a study where 500 patients with inherited retinal diseases were evaluated, the molecular etiology of 8.8% of patients was successfully identified through the application of NGS-based CNV analysis [25]. This method was particularly effective in diagnosing a variety of ocular disorders. In one noteworthy case (Patient 10), a deletion in exon 19 of the RPGRIP1 gene associated with Cone-rod dystrophy 13 and Leber congenital amaurosis 6 was detected. The observed deletion has been previously documented and is one among 14 major deletions reported in the RPGRIP1 gene, according to the Human Gene Mutation Database [2, 22]. In another patient (Patient 12), who was evaluated for retinal dystrophy and postaxial polydactyly, exome-based CNV analysis was instrumental in identifying a homozygous deletion of exons 18 and 19 in the BBS9 gene which is responsible for Bardet-Biedl Syndrome type 9. Although deletions in exons 18 and 19 of the BBS9 gene had not been previously reported, the gene itself is not novel to CNV research, with a total of 11 different large deletions having been documented to date [22].

Epidermolysis bullosa (EB) is a rare hereditary genetic disorder characterized by fragile skin and blister formation, associated with mutations in 16 different genes, highlighting its genetic heterogeneity [26]. One of these genes, COL7A1, is implicated in dystrophic EB, a particularly severe form of the condition. In cases of dystrophic EB, pathogenic variants are detectable through sequence analysis in approximately 95% of cases. However, for the autosomal recessive form of dystrophic EB, less than 2% of cases present pathogenic variants identifiable through deletion/duplication analysis [27]. During the analysis of Patient 4 in this study, NGS-based CNV analysis proved invaluable. This method identified a deletion encompassing exons 13–24 of the COL7A1 gene, a finding previously reported by Taghizadeh et al. [14]. The identification of this molecular etiology is not just an academic triumph; it has immediate practical implications. Not only does it allow for improved genetic counseling, enhancing our understanding of inheritance patterns and risks, but it also paves the way for novel therapeutic interventions. For instance, a local gene therapy called Beremagene Geperpave has recently demonstrated effectiveness in treating dystrophic EB [28] and has received FDA approval [29]. Therefore, understanding the specific genetic makeup of a patient’s disease can significantly impact the treatment strategies available, potentially leading to improved patient outcomes.

Establishing the molecular cause of genetically diverse disease groups typically involves a variety of testing methodologies. These often include targeted NGS panels, clinical exome sequencing or whole exome sequencing. These methods are utilized to explore a broad spectrum of genetic variability, from single nucleotide changes to large-scale chromosomal rearrangements. One of the significant advantages of NGS-based CNV analysis is its ability to concurrently evaluate both single nucleotide variations and CNVs across multiple genes. In the present study, we were able to establish a dual molecular diagnosis for two patients (Patient 10 and 12). Each diagnosis involved distinct molecular events; one was attributed to a single nucleotide variant and the other was a consequence of a CNV. This makes it an especially powerful tool in studying genetically heterogeneous conditions, where pathogenic variations can occur in several different genes.

In the context of this study, NGS-based CNV analysis was used to evaluate patients with genetically diverse conditions such as Ehlers-Danlos syndrome, neuronal ceroid lipofuscinosis, congenital lactic acidosis, hereditary spastic paraplegia, and epidermolysis bullosa. Each of these conditions exhibits significant genetic heterogeneity, underscoring the utility of a testing methodology capable of broad and simultaneous analysis of multiple genes. The use of NGS-based CNV analysis in this setting also served to overcome the limitations of MLPA. It is a commonly used technique for the detection of CNVs, but it is restricted to the evaluation of specific targeted regions using designated probes. Furthermore, it incurs additional costs due to the need for separate probe sets for each gene or region of interest [6]. NGS-based CNV analysis, in contrast, provides a more holistic view of the genomic landscape, enabling the detection of a wider range of potential pathogenic variants and paving the way for comprehensive genetic diagnoses.

One of the limitations of our study is that we did not use a secondary method to confirm the CNVs detected by NGS-based CNV analysis, nor did we employ multiple CNV callers, which could enhance sensitivity and reduce inconsistencies. Despite this, all CNVs identified were consistent with the clinical findings of each patient, reinforcing the validity of our results. For instance, single-exon deletions explaining the clinical presentations were detected in Patients 6, 9, and 10. In Patient 6, the palmitoyl-protein thioesterase activity was measured as “0,” aligning with the clinical diagnosis, and in Patient 9, exome-based family segregation analysis confirmed the expected inheritance pattern for the identified CNV.

Although GATK-gCNV is a reliable tool for detecting CNVs across multiple exons, its sensitivity decreases for single-exon deletions. Despite this, single-exon deletions were successfully identified in our study, as highlighted in Patients 6, 9, and 10. The NGS-based CNV analysis method employed in this study demonstrates a recall rate of 97% for rare coding CNVs (with a site frequency of ≤ 1%) compared to those identified by microarray analysis, and a robust recall rate of 95% for rare coding CNVs identified by genome sequencing, with a resolution exceeding two exons [12].

Our findings underscore the fact that exome-based CNV analysis significantly enhances molecular diagnostic success across a range of disease groups. Consistent with these findings, a recent statement by the American College of Medical Genetics and Genomics (ACMG) emphasizes the critical role structural variants play in understanding the genetic basis of diseases. The statement also highlights the inherent challenges associated with detecting these variants due to their complex nature. While it acknowledges the limitations of traditional cytogenetic methods, it accentuates the superior resolution and comprehensive genome coverage offered by NGS-based approaches in detecting structural variants [30].

In conclusion, our results underscore the effectiveness of NGS-based CNV analysis in determining the molecular etiology of Mendelian diseases. The precise detection of CNVs through this method showcases its potential as a robust tool in genetic diagnostics. Therefore, our study underscores the necessity of integrating NGS-based CNV analysis into regular genetic testing, which could potentially result in improved clinical outcomes and individualized management for people with Mendelian disorders.