Identification of a novel CNV at the EYA4 gene in a Chinese family with autosomal dominant nonsyndromic hearing loss

Hereditary hearing loss is a heterogeneous class of disorders that exhibits various patterns of inheritance and involves many genes. Variants in the EYA4 gene in DFNA10 are known to lead to postlingual, progressive, autosomal dominant nonsyndromic hereditary hearing loss. We collected a four-generation Chinese family with autosomal-dominant nonsyndromic hearing loss (ADNSHL). We applied targeted next-generation sequencing (TNGS) in three patients of this pedigree and whole-genome sequencing (WGS) in the proband. The intrafamilial cosegregation of the variant and the deafness phenotype were confirmed by PCR, gap-PCR and Sanger sequencing. A novel CNV deletion at 6q23 in exons 8–11 of the EYA4 gene with a 10 bp insertion was identified by TNGS and WGS and segregated with the ADNSHL phenotypes. Our results expanded the variant spectrum and genotype‒phenotype correlation of the EYA4 gene and autosomal dominant nonsyndromic hereditary hearing loss in Chinese Han individuals. WGS is an accurate and effective method for verifying the genomic features of CNVs.


Introduction
Hearing loss is the most common sensory deficit in modern society and affects approximately 466 million people worldwide, 34 million of whom are children [1,2]. Hereditary hearing loss patterns vary among autosomal dominant, autosomal recessive, X-linked, and mitochondrial patterns of inheritance. Unlike syndromic hearing losses, non-syndromic hearing loss are not associated with other clinical abnormalities [3]. In 75-80% of cases, NSHL is inherited autosomally recessively, while 20% are inherited autosomally dominantly. X-linked (2-5%) or mitochondrial patterns of inheritance (1%) are rare in NSHL [3,4]. The main characteristics of autosomal dominant hearing loss are high genetic and clinical heterogeneity and a delayed onset (postlingual), which are easily overlooked during hearing screening of newborns. At present, 51 different genes and 67 different loci have been linked to autosomal-dominant NSHL [5].
The EYA4 gene, which is located on chromosome 6q22.3-q23.2, was first identified as a causal gene of DFNA10 in a large American family in 1996 [6]. The EYA4 gene encodes eye absent 4 protein and is considered necessary for the proper development of multiple human organs, including the eye, inner ear, and heart. The EYA4 protein comprises two functional domains that contain a 271-amino-acid highly conserved C-terminal EYA domain (eyaHR) and an N-terminal variable region with a proline-serine-threonine (PST)-rich transactivation domain (eyaVR) [7]. By mediating interactions with the sine oculis family of proteins (Six1-Six6), mammalian EYA proteins function as transcriptional coactivators [8].
Researchers from different countries have found novel variants of the EYA4 gene and deletions of the EYA4 allele in different families with hearing loss based on sequencing analysis [9][10][11][12][13][14]. Late-onset, progressive, sensorineural hearing loss and age of onset from 6 to 50 years are the common characteristics among the tested families. Mid-frequency hearing is affected first, and all frequencies gradually become affected with increasing age. The degree of hearing loss also ranges from mild to moderate or severe with spontaneous evolution [15].
In recent years, next-generation sequencing (NGS) technology, including both targeted and whole-genome sequencing (WGS), has been considered an efficient and swift method to detect potential variants [16,17]. This method provides a guiding role in the diagnosis and treatment of hereditary hearing loss [18]. Copy number variations (CNVs) are genomic variants within species that reflect differences in copy numbers, including deletions, duplications, and amplifications of DNA sequences. By using cytogenomics techniques such as comparative genomic hybridization (CGH), SNP arrays, WES and WGS, many novel CNVs associated with NSHL phenotypes have been identified [19][20][21][22][23]. With the development of diagnostic WGS, the accessibility, robustness and accuracy of CNVs throughout the genome have dramatically improved. Compared to genomic microarrays, the investigation of genomic features such as copy number, content and positional information has become more precise through the WGS method and algorithm [24,25].
In the present study, we present a four-generation Chinese pedigree with autosomal dominant nonsyndromic hearing loss. A novel CNV deletion at 6q23 was identified in the affected individuals by targeted next-generation sequencing (TNGS) and WGS, and this information sheds new light on the pathogenic mechanism of EYA4 variants.

Family and clinical evaluation.
A Chinese family (FY-140) classified as of Han origin presented with late-onset, progressive hearing loss. Approval for the study was obtained from the Ethics Committee of the Eye & ENT Hospital, Fudan University for Human Studies. Written informed consent was obtained from the participants or the parents of minors. The assessment of all the individuals was based on audiological methods, including pure-tone audiometry, acoustic impedance, auditory brainstem response (ABR), distortion product of otoacoustic emission (DPOAE) and otological examination. Clinical information, such as age of onset, degree of hearing loss, progression of hearing loss, noise exposure and history of using aminoglycosides, was collected from family members if available. Information on deceased family members was obtained from relatives. The proband (II-2) was subjected to a high-resolution CT scan of the temporal bone. All the individuals accepted electrocardiography for the reason that some references indicated that EYA4 gene variants may cause heart disease in patients.

Targeted genomic capturing and next-generation sequencing
Genomic DNA was isolated using the TIANamp Blood DNA Midi Kit (TIANGEN Biotech, Beijing China) and fragmented to 150 bp using an ultrasonoscope (Covaris S220, Massachusetts, USA). End repair, adenylation and adapter ligation were performed for library preparation using a standard library construction kit (MyGenostics Inc., Beijing, China). Targeted DNA fragments were captured by a sequence capture array (MyGenostics Inc., Beijing, China). High-throughput sequencing and processing and bioinformatic data analysis were performed using the DNBSEQ-T7 sequencing platform (MGI Tech Co, Shenzhen China). The raw sequence reads were filtered using the BWA MultiVision software package and then aligned to GRCh38/hg38 (University of California Santa Cruz version). SNPs and indels were identified using the GATK Indel Genotyper and ANNOVA software. A CNV analysis was performed using the log2 ratio of the read depth on each exon.

Whole-genome sequencing
The genomic DNA samples were fragmented by sonication to a size of 300-500 bp. Sequence analysis was performed using the TruSeq Nano DNAHT Sample Prep Kit (Illumina Inc., Massachusetts, USA) following the manufacturer's instructions. The total effective data yield of the sample was approximately 430 million reads, and the data showed a coverage of > 99.62% at 20X. After the raw sequence reads were mapped to the human genome reference sequence (USSC) (GRCh38/hg38), SpeedSeq software was used. SNPs, indels, CNVs and SVs were captured using GATK HaplotypeCaller, ANNOVAR and SpeedSeq software.

Clinical evaluation
The pedigree of the family includes 15 members (eight men and seven women) over four generations (Fig. 1A). Five individuals were diagnosed with NSHL based on their medical history and audiological function examination results. The self-reported age of onset of hearing impairment ranged from 26 to 43 years. Assessments of the four affected living members showed mild to severe bilaterally symmetric NSHL across all frequencies, and the disease affected both sexes (Table 1). In addition to hearing loss, the patients had no other clinical symptoms or signs. No patients complained of vestibular symptoms. The temporal bone scans and cardiac examinations of the proband yielded normal results. The initial hearing loss showed an audiogram pattern called a "cookie bite", which was usually mild and only affected mid-frequencies. Progressive hearing loss expanded to other frequencies at later stages (Fig. 1B). Other audiometric tests' results in affected individuals including ABR, DPOAE, acoustic impedance and otological examination were also consistent with the diagnosis of NSHL.

Identification of novel CNV by TNGS
Patients II-2, III-1, and III-5 were subjected to targeted NGS of 147 deafness-related genes, and 8 SNPs and 1 CNV were detected. A novel CNV in exons 8-11 of the EYA4 gene and two previously identified variants (c.7630T > G, p. Leu2544Val and c.8257G > A p. Ala-2753Thr) in the CDH23 gene were found in all three patients ( Fig. 2A, B). The intrafamilial cosegregation of the variants and the hearing loss phenotype were confirmed by long-range PCR and Sanger sequencing in all family members.
We found that both variants of the CDH23 gene (NM_022124), c.7630T > G (p. Leu2544Val) in exon 52 and c.8257G > A (p. Ala2753Thr) in exon 54, were been carried by four patients (II-2, II-3, II-6 and III-1) and five healthy family members (I-2, III-2, III-3, III-5, IV-1 and IV-2) by Sanger sequencing. These four patients and five healthy family members are all heterozygous for both variants of the CDH23 gene. According to the pedigree diagram, the inheritance pattern of the hearing loss in this family was dominant. Both variants of the CDH23 gene did not co-segregate with the phenotype (BS4-ACMG). The genotype for each variant in the pedigree and the Sanger sequencing results of these family members are described in the supplemental material (Additional file 1: Fig. S1, Additional file 2: Fig. S2, Additional file 3: Fig. S3). We believe they are in cis-mutation and not in compound heterozygosity in each subject. Although they have been reported in a previous paper that detected both variants in a deafness patient from China, these variants did not appear in the general populations databases [26]. The minor allele frequency (MAF) of p. Leu2544Val variant is 0.000051 (East Asian) in GnomAD, while the p. Ala2753Thr variant is novel (PM2-ACMG). These variants were predicted as benign by silico pathogenicity prediction tools REVEL, MutationTaster, SIFT and Polyphen 2 (BP4-ACMG). According to ACMG standards and guidelines, both variants in the CDH23 gene are classified as PM2, BP4 and BS4, and the criteria for both alleles will be of uncertain significance (Table 2). In general, CDH23 gene mutations cause autosomal recessive non-syndromic hearing loss [4,27]. Based on the ACMG classification and intrafamilial cosegregation analysis of these variants, we inferred that the c.7630T > G and c.8257G > A variants in the CDH23 gene may not be the cause of the disease in this family. In contrast, a CNV in the EYA4 gene was found only in the patients, and none of the healthy family members were carriers of this deletion (Fig. 2C). Therefore, this variant in the EYA4 gene could be considered the cause of the disease. Because patient I-1 died, we were unable to collect a sample of his DNA. We inferred that patients II-2, II-3 and II-6 inherited the heterozygotic variant from their father and that patient III-1 inherited the variant from patient II-2.

Verification of CNV by WGS
We performed WGS using genomic DNA from patient III-1 to further identify potential variants. We identified on average 3,585,580 SNPs and 732,397 indels in coding regions or introns. Then, Single-nucleotide variants (SNVs) and InDel from WGS data were filtered as follows: (1) variants with MAF below 0.01 in 1000Genomes, ExAC03 Asian population and gnomAD Asian population (2) coding/splicing variants (3) variants that are predicted to be likely pathogenic/pathogenic by any of the following software such as SIFT, POLYPhen V2, Muta-tionTaster, Cadd, Dann, and dbscSNV, were considered as likely causal variants. CNV analysis was performed using the log2-ratio of read depth on each exon. Priority was given to variants found in deafness genes (annotated as deafness genes in one or more of the following databases: OMIM, HPO, HGMD, InterVar, HPO, MGI, ClinVar, ISCA and MalaCards). There were 64 variants predicted as candidates, including 47 SNVs, 3 indels, and 2 CNVs. With WGS, TNGS, Sanger sequencing, and  cosegregation analysis, we identified a novel CNV that might cause disease. The complex genomic rearrangement in the dataset encompassed a 17.4-kb deletion spanning exons 8 to 11 and a 10-base insertion (chr6: 133461414 insTTT GAA TTTT). As shown in Fig. 2, the distal breakpoint mapped to chr6: 133461414 (GRCh38/hg38), and the proximal breakpoint mapped to chr6: 133478832 (GRCh38/hg38) (Fig. 2C). To validate the identified CNV, gap-PCR of the region spanning the breakpoint was performed. The primers Forward1 and Reverse amplify a gap-PCR product specific for the variant allele, whereas the primers Forward2 and Reverse amplify a product from the normal EYA4 gene. The 500-bp PCR product was obtained from all 4 affected individuals (II-2, II-3, II-6 and III-1) in the family but not from the unaffected family members (I-2, III-2, III-3, III-5, IV-1 and IV-2). The normal EYA4 gene product was found in all family members (Fig. 2D). The generated PCR products were subjected to Sanger sequencing, and a sequence analysis of the resultant amplicons confirmed the junction of the two breakpoints through the 10-bp insertion identified by WGS (Fig. 2E). This deletion has been submitted to LOVD under accession ID 00361730.

Discussion
CNVs are a common cause of hereditary hearing loss and are thought to play a role in nearly 20% of non-syndromic HL diagnoses [19]. In this study, we performed a comprehensive genetic analysis that included TNGS, WGS, gap-PCR and Sanger sequencing in a four-generation Chinese Han family with autosomal dominant NSHL. All the affected individuals in the family exhibited sensorineural deafness, which primarily affected low and mid-frequencies and had onset ages in the range of 26 to 43 years.
We identified a novel CNV deletion in exons 8-11 of the EYA4 gene with a 10 bp insertion. First, this variant cosegregated with NSHL symptoms in patients and was not detected in normal family members. Then, this CNV is predicted to affect the eyaHR domain. By interacting with members of the SIX and DACH protein families in a conserved network, the highly conserved C-terminal region of EYA4 (eyaHR) regulates embryonic development and follow-up functions after development of the mature organ of Corti. It regulates Na+/K+-ATPases and the development of mechanosensory cells of the inner ear [12]. Finally, we attempted to construct a three-dimensional structure of the CNV using the SWISS-MODEL software but failed because the structure was severely affected. These findings may also suggest that this novel CNV deletion is pathogenic in auditory function.
The hearing loss phenotype in the present family is similar to that reported for patients carrying EYA4 variants, i.e., late-onset, postlingual, progressive, and bilateral HL. In previous reports, flat-type hearing loss was observed in patients with truncating EYA4 variants. At onset, hearing impairment was usually mild and detected at mid-frequencies, resulting in an audiometric profile commonly referred to as a "cookie-bite" pattern. During its progression, hearing loss began to involve other frequencies. The progression rate of hearing loss caused by EYA4 was approximately 5.75 dB/year (95% CI 4.50-7.00 dB/year), which is relatively severe compared to POU4F3 and MYO6 gene mutation in ADNSHL patients [28,29].
To date, variants in the EYA4 gene have been associated with HL in more than 50 ethnic groups worldwide. It was believed that EYA4 variants led to syndromic and non-syndromic NSHL, but EYA4 is not a frequently mutated gene in ADNSHL compared with other reported genes. The characteristics of all known EYA4 variants are summarized in Table 3. According to these variants, the severity of hearing loss was not significantly related to the types or locations of variants.
Several CNVs in EYA4 have been linked to deafness (Fig. 3). One CNV disrupts the EYA4 gene and spares only exons 1-3 from the deletion [30]. The deleted sequence of the promoter and the first two exons was previously identified in a Japanese boy [31]. A deletion of four exons and a deletion spanning exons 4-20 have been reported to cause severe ADNSHL in Japanese individuals [9]. In addition, the EYA4 variant reportedly causes dilated cardiomyopathy accompanying NSHL in a single large family. In this family, a 4846-bp genomic deletion    [34] that resulted in loss of the EYA domain (eyaHR) and part of the variable region (eyaVR) was detected [32]. A heterozygous deletion of 2747 bp represented a copy variant loss encompassing exon 15 to exon 17 [13]. Additionally, in Japan, a novel hemizygous indel in the EYA4 gene was predicted to be p. (Val124_Pro323del) [14]. We now add a genomic rearrangement consisting of a deletion and a 10-bp insertion to this list (Fig. 3, Table 3). The EYA4 gene is widely distributed in the inner ear, which includes otic vesicles, the Reissner membrane and the sensory epithelia of the vestibular system and Corti [12,33]. Although the pathogenic mechanism of ADNSHL associated with EYA4 variants remains to be further investigated, haploinsufficiency is generally thought of as the major mechanism. Many reports indicate that EYA4 participates in important pathways in cardiac tissue. The physiological level of Eya4 phosphatase activity is thought to participate in normal cardiac gene regulation. Instead of most DCM genes that encode structural proteins, EYA4 is a transcriptional coactivator. Large deletions comprising the variable domain are most likely to affect cardiac functions. A patient carrying a de novo 9 MB interstitial deletion that disrupts the gene EYA4 presented a patent ductus arteriosus and aortic insufficiency. A family with a 4846-bp deletion was associated with DCM as well as NSHL. In the other 2 unrelated families, polymorphic loci on chromosome 6q23 to 24 were associated with DCM and NSHL. [30,34]. For that reason, we performed electrocardiograms for all family members. None of the members exhibited a cardiac phenotype, and electrocardiograms showed no abnormalities. Including our work, many studies indicated that the genotype-phenotype correlation of large deletions in EYA4 and dilated cardiomyopathy was not very obvious [14]. One reason may be that previously reported cardiopathy was caused by other variants in the large deleted regions. Another reason is that the defect in a contiguous gene could account for the cardiac defects [34]. More experiments, including detecting EYA4 levels in nuclear and cytoplasmic components, may provide evidence for this theory. Increasing the yield of genetic testing among patients with both NSHL and DCM may allow for better detection of the EYA4 gene and cardiac pathology [35].  . 3 Overview of the CNVs identified in this study and those previously identified in EYA4