Identification of CDH23 mutations in Korean families with hearing loss by whole-exome sequencing

Hearing loss is one of the common heterogeneous disorders. Genetic factors account for more than 50% of cases of congenital hearing loss, where the majority of cases exhibit autosomal recessive inheritance [1]. To date, more than 100 mapped loci have been reported, and 55 non-syndromic hearing loss genes have been identified (http://hereditaryhearingloss.org/).

The gene most commonly involved in hearing loss worldwide is GJB2[2], while SLC26A4 is also frequently involved in congenital hearing impairment. The GJB2 and SLC26A4 genes make mutation screening relatively easier, and many studies have focused on only these two genes. However, in many ethnic populations, GJB2 and SLC26A4 are responsible for only a small percentage of deafness cases [3], and screening of mutations in a large number of genes simultaneously is difficult. Also, it is near-impossible to identify pathogenic mutations by traditional linkage analysis when DNA is available from only small families.

Mutations in the CDH23 gene are known to be responsible for both Usher syndrome type ID (USH1D) and non-syndromic hearing loss (DFNB12). To date, more than 50 mutations have been reported in patients with Usher syndrome type I (USH1D) who have congenital hearing loss, retinitis pigmentosa (RP), and vestibular dysfunction. A total of 24 mutations have been reported in patients with non-syndromic hearing loss (DFNB12) [4]. A genotype-phenotype correlation study suggested that USH1D was usually associated with nonsense, whereas DFNB12 with missense mutations [5]. Deafness caused by CDH23 has been found in many populations worldwide, including African–American, Dutch, European, German, Pakistani, Turkish, and Japanese populations [6]. However, clinical application of CDH23 mutation detection has lagged because of the size of the gene.

Recent advances in DNA enrichment and next-generation sequencing (NGS) technology have allowed rapid and cost-effective analysis of the causative mutations of human disorders, especially those that are heterogeneous in nature [7]. The techniques are particularly applicable to analysis of small families [8]. In the present study, we applied whole-exome sequencing (WES) to study small Korean families negative for mutations in GJB2 and SLC26A4, and we identified CDH23 mutations in two families with autosomal recessive non-syndromic hearing loss (ARNSHL).

We identified three CDH23 mutations, p.Pro240Leu, p.Glu1595Lys, and p.Asn342Ser, in 2 (15%) of 13 Korean families with ARNSHL by whole-exome sequencing. The present report is the first to demonstrate that CDH23 is an important causative gene for ARNSHL in Korean patients. Additionally, we screened for eight mutations, including the three (p.Pro240Leu, p.Asn342Ser and p.Glu1595Lys) detected in the present study, and five others (p.Arg301Gln, p.Glu956Lys, p.Arg1417Trp, p.Gln1716Pro, and p.Arg2029Trp) that occur at relatively high frequencies (patient allele frequency > 0.1) in Japanese patients [4], in 93 unrelated Korean hearing loss patients. Two patients and one patient, respectively, were heterozygous for p.Pro240Leu and p.Glu1595Lys, and no other mutation was detected. p.Pro240Leu was reported in Japanese families with ARNSHL as a compound heterozygous or homozygous mutation [5]. Miyagawa et al. reported that mutations of the CDH23 gene are important causes of non-syndromic hearing loss and that p.Pro240Leu accounted for nearly 43.3% (45/105) of all CDH23-mutated families in Japanese [4]. In this study, both families with ARNSHL caused by CDH23 mutations carried the p.Pro240Leu mutation and, additionally, two of 93 hearing-loss patients were heterozygous for the mutation. Thus, p.Pro240Leu is the most common cause of CDH23-associated ARNSHL in Asian populations. p.Glu1595Lys was present in 1 of 93 patients. Astuto et al. reported that ~5% of recessive non-syndromic hearing loss might be caused by mutation of CDH23[6]. Although GJB2 and SLC26A4 mutations were absent in our patients, our results show that CDH23 mutation caused ARNSHL in 2 of 13 (15%) affected Korean families and that 3 of 93 patients with hearing loss carried a heterozygous mutation. The precise frequency will be determined by future mutation analysis of a large patient cohort, but is estimated to be high. Therefore, CDH23, as well as GJB2 and SLC26A4, should be included in screening.

CDH23 (NM_22124) has 69 exons and encodes cadherin 23, a protein of 3,354 amino acids with 27 EC domains, a single transmembrane domain, and a short cytoplasmic domain [12]. It is a putative calcium-dependent adhesion molecule required for proper morphogenesis of hair bundles of inner ear neurosensory cells. Mutations in CDH23 cause the stereocilia of hair cells in the inner ear of Waltzer mice (a model of USH1D) to become disorganized [13, 14]. Many deafness mutations in CDH23 are in the calcium-binding motif of the linker region between EC repeats. All mutations detected in the present study were missense in nature, and in the EC domain. Of the three mutations, two (p.Glu1595Lys and p.Asn342Ser) affected highly conserved EC calcium-binding sites (Figure 2B and D). The sites are LDRE and DXNDN, and are thought to be essential for linearization, rigidification, and dimerization of CDH23 [15, 16]. The p.Glu1595Lys mutation may disrupt a conserved LDRE calcium-binding motif in the fifteenth EC domain. These EC domains are involved in cell-to-cell adhesion via hemophilic calcium-dependent interactions [16]. Molecular modeling of the p.Glu1595Lys mutation shows impairment of calcium binding [17]. Since calcium provides rigidity to the elongated structure of cadherin molecules, thereby enabling hemophilic lateral interaction, this mutation is likely to impair interactions of CDH23 molecules with either CDH23 or other proteins. Likewise, the p.Asn342Ser mutation is also located in a highly conserved EC calcium-binding site (DXNDN). Thus the mutation probably decreases calcium affinity and impairs protein function. Their pathogenic effects on protein structure or function are supported by in silico analysis (Table 2).

Additionally, we identified 26 non-synonymous variants in the CDH23 coding exons from 16 hearing-loss patients and 30 Korean control exomes (Table 3). Among them, four variants have the potential to be pathogenic. pGly266fs in SR-255S showed co-segregation in the family and was absent in 128 Korean control exomes. However, it was heterozygous, so we could not conclude that it is a causative mutation. p.Asp428Asn in SR-68 was also heterozygous. DNA samples from family members were not available. Thus, we could not confirm its pathogenicity by segregation analysis. p.Val2283Ile and p.Phe2801Val were identified as compound heterozygous mutations in SR-1016. Co-segregation was confirmed by Sanger sequencing of family members, but these two variants were detected in three of the 128 Korean control exomes. Thus, we concluded that these two variants are not causative mutations. However, a heterozygous mutation in a recessive gene may be relevant to the hearing loss phenotype if it coexists with another heterozygous mutation [5], and it is a low frequency in the general Korean population might not rule out the possibility that it is pathogenic. Further exome data accumulation and establishment of a database of common mutations in Korean patients with hearing loss will help us to determine whether they are pathogenic or not.

We performed parallel sequencing of the whole exomes of 13 small families and rapidly and successfully identified the CDH23 mutation in two families. In two of the other 11 families, hearing loss was caused by the MYO15A mutation [18], the first report of this mutation in Eastern Asia. Whole-exome sequencing allowed us to screen mutations in a large number of genes at the same time and to detect pathogenic mutations in affected individuals who were not identified by classical genetic studies. However, the causative mutations in nine families have still not been identified. In these families, whole-exome sequencing may have missed some mutations in exons because it did not completely cover the targeted region, which may have had a high GC content or repetitive sequences, or because hearing loss was caused by mutations an as-yet-unknown gene. Our search for deafness genes in the remaining families is ongoing.

We identified CDH23 mutations in two of 13 families with ARNSHL by whole-exome sequencing. The exact frequency will need to be determined in a future larger cohort mutation analysis, but it is estimated to be high. Therefore, as well as GJB2 or SLC26A4 screening, CDH23 should also be considered. Our results show that whole-exome sequencing is effective when used to detect causative mutations in heterogeneous patients with hereditary hearing loss. In the future, we anticipate the rapid discovery of hearing loss-related genes by whole-exome sequencing, which will improve understanding of the condition.

HMW, MHP and SKK designed the study and wrote the article. HMW, MHP and WGY analyzed the data. HJP and JWS collected the samples and examined the family information. HMW and BYK performed the sequencing analyses. All authors read and approved the final manuscript.