Introduction

Hearing loss is a genetically heterogenous condition. An autosomal recessive inheritance pattern is most common and accounts for 70% of genetic deafness [1]. To date, more than 100 genomic loci and at least 70 genes have been implicated in autosomal recessive nonsyndromic deafness (designated as DFNB) (Van Camp G, Smith RJH. Hereditary Hearing Loss Homepage; https://hereditaryhearingloss.org).

Comprehensive genetic testing using high-throughput DNA sequencing technology has enabled etiologic diagnosis in 40–50% of hearing loss patients, but in a large number of cases, a genetic etiology remains inconclusive [2]. In some cases, only a single heterozygous pathogenic or likely pathogenic variant is detected in a gene associated with autosomal recessive hearing loss; in others, the clinical significance of identified variants is uncertain. Relatively common (e.g., allele frequency > 0.3%) silent variants such as synonymous changes that do not alter the amino acid sequence are often removed in the bioinformatic analysis because most are classified as likely benign according to the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) recommendation and the ClinGen Hearing Loss Expert Panel specifications [3, 4].

Autosomal recessive non-syndromic hearing loss (ARNSHL) at the DFNB3 locus on the short arm of chromosome 17 in band p11.2 is caused by biallelic pathogenic variants in MYO15A, a gene of 66 exons that spans ~71 kb and encodes the unconventional myosin XV [5]. Loss of MYO15A function typically leads to congenital profound hearing loss in humans and deafness and vestibular defects in mice [6]. Since DFNB3 was first described [7], more than 200 pathogenic variants in MYO15A have been reported, including some in the Ashkenazi Jewish (AJ) population [8,9,10,11]. Residual hearing have been reported in patients with MYO15A variants, mostly in the N-terminal region; the pathogenicity of variants in other regions found in patients with partial hearing loss has not been rigorously assessed according to the ACMG/ClinGen framework [12,13,14,15].

In this study, we implicate as pathogenic a relatively common synonymous variant c.9861 C > T/p.(Gly3287=) in MYO15A identified in the AJ population, present evidence for its pathogenicity in ARNSHL, and summarize the phenotypic characteristics associated with this variant.

Subjects and methods

Patients

Eighty families with nonsyndromic bilateral sensorineural hearing loss of AJ ancestry were genetically screened at Dor Yeshorim (DY) and referred for follow -up clinical genetic testing when indicated (Supplementary information). Retrospective review of genetic testing results identified other families with relevant variants from those who underwent clinical or research genetic testing at the Children’s Hospital of Philadelphia (CHOP), the Molecular Otolaryngology and Renal Research Laboratories (MORL), the Laboratory for Molecular Medicine (LMM), and Tel Aviv University (TAU). Pedigrees with variants discussed in this study are shown in Fig. 1. Clinical information including newborn hearing screening results, age at onset, and characteristics of hearing loss were retrieved from patients’ medical records, clinical genetic test requisition documents, and direct communications with study participants. Jewish population controls not known to have hearing loss include 9596 anonymous samples (6487 Ashkenazi, 1727 Sephardi, and 1382 both). This study was conducted in compliance with protocols approved by the Institutional Review Boards of DY, Partners HealthCare, the American University of Beirut, the CHOP, and the University of Iowa, and the Helsinki Committees of Tel Aviv University and the Israel Ministry of Health. Written informed consent was obtained from all participants.

Fig. 1: Pedigrees of eight Jewish families who received genetic screening at Dor Yeshorim (DY), and six families who received genetic testing at Tel Aviv University (TAU), the Molecular Otolaryngology and Renal Research Laboratories (MORL), the Laboratory for Molecular Medicine (LMM), and the Children’s Hospital of Philadelphia.
figure 1

Squares denote males and circles denote females. Solid symbols represent individuals affected with bilateral sensorineural hearing loss; clear symbols represent unaffected individuals; dotted symbols represent unaffected heterozygous carriers. Arrowheads point to probands. V, c.9861 C > T/p.(Gly3287=); V1, c.8050 T > C/p.(Tyr2684His); V2, c.8183 G > A/p.(Arg2728His); V3, c.8467 G > A/p.(Asp2823Asn); V4, c.8897_8900dup/p.(Ala2968Profs*33); V5, c.4198 G > A/p.(Val1400Met); V6, c.6292 G > A/p.(Asp2098Asn); V7, c.8090 T > C/p.(Val2697Ala); V8, c.4642 G > A/p.(Ala1548Thr); V9, c.707 A > G/p.(Tyr236Cys); V10, c.10584del/p.(Thr3528Profs*26); +, reference allele; ?, genotype unknown; /, variants in trans; *, phase unknown but assumed in trans because the same variants have not been observed in multiple unrelated families. [V9;V10] indicates that V9 and V10 are in cis. In audiograms, L, left; R, right; yo, years old.

Full size image

Molecular genetic testing

AJ patients were screened by the DY targeted hearing loss variant panel (Supplementary information). Cases in which the genetic cause for hearing loss could not be determined after the DY screening were referred for clinical next-generation sequencing gene panels, including the MORL OtoSCOPE panel [2], the Sema4 hearing loss panel, or clinical exome sequencing at GeneDx or the Hadassah Medical Center (Table 1). Other hearing loss probands were tested by the HEar-Seq panel at TAU [8, 9], by the OtoSCOPE panel at the MORL [2], by the OtoGenome panel at the LMM, or by the Audiome panel at the CHOP. Clinically significant variants were confirmed by Sanger sequencing. Family members were tested by targeted Sanger sequencing of familial variants.

Table 1 Clinical characteristics of patients with MYO15A variants in this study.
Full size table

Variant interpretation

Sequenced variants were described using the Human Genome Variation Society (HGVS) nomenclature (http://varnomen.hgvs.org). NM_016239.4 (NC_000017.10) was used as the cDNA reference sequence. Variants were reviewed and classified according to the ACMG/AMP and ClinGen Hearing Loss Expert Panel guidelines [3, 4]. Variants are listed in ClinVar (Table 2). Population frequencies were estimated based on data from the Genome Aggregation database (http://gnomAD.broadinstitute.org). Population-specific carrier frequencies were determined by screening the variant in the general Jewish population, including 6,487 Ashkenazi, 1,727 Sephardi, and 1,382 mixed Ashkenazi and Sephardi Jewish individuals by DY. Computational predictions were obtained from Varcards (http://varcards.biols.ac.cn/) and Human Splicing Finder (HSF) (http://www.umd.be/HSF3/HSF.shtml). Statistical analysis of the odds ratio, 95% confidence interval, p value, and Z score were calculated using MEDCALC (https://www.medcalc.org/calc/odds_ratio.php).

Table 2 Classification of MYO15A (NM_016239.4; NC_000017.10) variants found in patients with hearing loss in this study.
Full size table

In vitro splicing analysis

In vitro splicing minigene assays were carried out as described [16,17,18]. Briefly, genomic sequence at chr17:18069512-18069966 (hg19) including exon 61 (161 bp) plus 163 and 131 nucleotides from the 5′ and 3′ flanking sequences, respectively, of MYO15A (NM_016239.4) was PCR amplified from a DNA sample heterozygous for the c.9861 C > T variant using gene-specific primers designed with embedded SalI or SacII restriction enzyme recognition sites. After digestion, PCR fragments were ligated into the pre-constructed pET01 Exontrap vector (MoBiTec, Goettingen, Germany). Sequencing of selected colonies confirmed proper orientation of the cloned fragment and identified both wild-type and variant colonies. Next, the wild-type and variant minigenes were transfected in triplicate into HEK293 cells and total RNA was extracted 36 h post transfection using the Quick-RNA MiniPrep Plus kit (ZYMO Research). cDNA synthesis was performed using RNA SuperScript III Reverse Transcriptase (ThermoFisher Scientific, Waltham, Massachusetts) with a primer specific to the 3′ native exon of the pET01 vector. After PCR amplification, products were visualized on a 1.5% agarose gel, extracted and then Sanger sequenced.

Results

Identification of the c.9861 C > T variant in MYO15A

The probands in DY1 and DY2 (Fig. 1 and Table 1) initially had the OtoSCOPE panel test at the MORL. The c.8050 T > C/p.(Tyr2684His) variant in MYO15A was identified in DY1; DY2 carried the c.8183 G > A/p.(Arg2728His) variant. Further investigation by exome sequencing of the DY1 proband at GeneDx reported MYO15A c.8050 T > C/p.(Tyr2684His) as well as the c.9861 C > T/p.(Gly3287=) variant, which was classified as likely benign at the time because of its relatively high allele frequency (0.4% in AJ) and predicted synonymous impact. The c.9861 C > T/p.(Gly3287=) variant had been filtered out bioinformatically and not reported by the MORL because of its predicted synonymous impact. Follow-up segregation studies confirmed that c.9861 C > T was in trans with c.8050 T > C and co-segregated with hearing loss in DY1 (Fig. 1). Targeted screening of c.9861 C > T and reanalysis of sequencing data confirmed its trans configuration with c.8183 G > A and co-segregation with hearing loss in DY2 (Fig. 1 and Table 1).

Because genetic results from families DY1 and DY2 raised the possibility that c.9861 C > T was potentially pathogenic, the variant was added to the DY hearing loss screening panel. Additional genetically undiagnosed AJ families with hearing loss were screened, and c.9861 C > T was identified in DY3 and DY4 with previously unsolved exome sequencing at Hadassah Medical Center (Table 1). DY5 was positive for the same two variants identified in DY2 and DY4. DY6, DY7, and DY8 were found to have c.9861 C > T and second alleles identified by additional tests (Table 1).

Classification of c.9861 C > T variant in MYO15A according to ACMG/AMP criteria

The c.9861 C > T MYO15A variant has been identified in the heterozygous state in 0.5% (63/12,974) of alleles in the general AJ population and in 4.4% (7/160) AJ alleles in hearing loss probands by DY. It has also been identified in the heterozygous state in 0.4% (42/10,362) of AJ alleles in gnomAD (https://gnomad.broadinstitute.org/variant/17-18069748-C-T). The allele frequency in Sephardi and mixed Ashkenazi and Sephardi Jewish populations are 0.06% (2/3,454) and 0.14% (8/2,764), respectively. This allele is significantly enriched in hearing loss patients over the general AJ population (odds ratio 9.4, 95% confidence interval 4.2–20.8, Z = 5.5, p < 0.0001), thus providing strong evidence for pathogenicity (PS4) [3, 4]. The variant segregates with phenotypes in nine affected and 23 unaffected siblings of the probands in eight AJ families, which provided strong segregation evidence (PP1_Strong) [4]. It was found in one homozygous and 11 compound heterozygous probands with hearing loss (Fig. 1). The variant was in trans with seven different alleles including two variants classified as pathogenic and four as likely pathogenic (Tables 1 and 2). However, because the allele frequency of c.9861 C > T is >0.3% in the AJ population, we considered the allelic evidence moderate (PM3) instead of strong.

The c.9861 C > T variant is predicted to alter splicing by HSF by either activating a cryptic donor site, creating a novel exon splicing silencer (ESS) motif and/or abolishing an exonic splicing enhancer (ESE) motif. To characterize the impact of c.9861 C > T on RNA splicing, we cloned the wild-type and variant sequences of MYO15A (NM_016239.4) exon 61 and flanking introns into the pET01 exon trap vector and transfected them into the HEK293 cell line. Visualization of the splicing products showed that cells transfected with the wild-type vector yielded the expected 407-bp band, which contains exon 61 of MYO15A (Fig. 2). In contrast, cells transfected with the variant construct yielded a single 246-bp band, which corresponds to splicing of native 5′ and 3′ exons of the pET01 vector and skipping of the cloned exon 61 of MYO15A (Fig. 2). Sequencing of purified PCR products confirmed breakpoints and splicing events. No differences were detected among replicates. Skipping of the 161 bp exon is predicted to cause a frameshift that would lead to premature protein truncation. The C > T transition at c.9861 position is computationally predicted to create a cryptic donor site (GT) at c.9860_9861from the reference sequence GC. Should the cryptic donor site be used, c.9860_9948 of 89 bp in exon 61 would be deleted and result in a frameshift that would lead to premature protein truncation. We did not detect the activation of the predicted cryptic donor site as a splice event in our experiments. Results of minigene assays provide functional evidence to support pathogenicity, but the in vitro study may not fully recapitulate the impact under physiological conditions in vivo; therefore, we considered the functional evidence supporting (PS3_Supporting).

Fig. 2: Minigene splicing assay.
figure 2

a Electrophoresis of RT-PCR products from total RNA extracted from HEK293 cells transfected with wild type, c.9861 C > T, or empty vectors. b Schematic drawings of minigene constructs. Boxes indicate exons. The blue box in the middle is either wild-type (with C) or variant (with T). Spliced products and sequencing results of the RT-PCR products are shown on the right.

Full size image

In summary, the c.9861 C > T variant in MYO15A is classified as pathogenic based on PP1_Strong, PS4, PM3, and PS3_Supporting according to the ACMG/AMP recommendation for sequence variant interpretation and specifications of the ClinGen Hearing Loss Working Group Expert Panel [3, 4].

Clinical characteristics of patients with c.9861 C > T in MYO15A

We identified 12 probands and nine affected siblings with biallelic MYO15A variants including c.9861 C > T (Table 1 and Fig. 1). Affected individuals all had bilateral nonsyndromic sensorineural hearing loss, but none had congenital profound deafness typically found in patients with variants that impact MYO15A function [19]. Eight of ten individuals with newborn screening results passed the screening and subsequently developed mild-to-moderate hearing loss, which was typically noticed during early childhood. The loss progressed to moderate-to-severe and even profound with advancing age.

Discussion

Comprehensive genetic testing using high-throughput sequencing generates a large number of genetic variants that require interpretation. Synonymous variants are often excluded from this list as they are typically considered likely benign unless there is compelling evidence for the contrary or they have extremely low allele frequencies. It is known that synonymous variants may be disease-causing by several mechanisms including altered translation rate, mRNA secondary structure, or RNA splicing [20,21,22,23,24,25]. In some instances, multiple mechanisms work together in inducing disease phenotype [23]. However, computational tools to predict these effects are not always reliable [26]. Prior to this study, c.9861 C > T was classified as likely benign because it was found in 0.4% of AJ, which is above the threshold for Criterion BS1, and it is not evolutionarily conserved at the nucleotide level [4]. However, its allele frequency is consistent with the carrier frequency for ARNSHL, meaning that BS1 criterion is not applicable in light of conflicting evidence. ARNSHL is a genetically heterogeneous condition with an estimated prevalence of >1 in 10,000 in various populations, including AJ as well. The prevalence of deafness is not particularly higher in AJ than in other populations. An allele frequency of 0.4% would be consistent with carrier frequency for ARNSHL in any population. The difference among different populations are different major contributors that need to be included on the exception list of ClinGen allele frequency cutoff rules. Our data established the pathogenicity of the variant based on strong segregation, statistical, moderate allelic, and supporting functional evidence to indicate that abnormal splicing is a probable cause (PP1_Strong, PS4, PM3, PS3_Supporting) [3, 4].

Establishing the pathogenicity of c.9861 C > T, together with other specific variant criteria, allowed us to classify c.8090 T > C/p.(Val2697Ala) as pathogenic, which led to classification of c.4642 G > A/p.(Ala1548Thr) as pathogenic. Hence, conclusive diagnoses were reached for 13 families: DY1, DY2, DY3, DY4, DY5, DY6, DY7, DY8, TAU9, MORL10, LMM12, LMM13, and CHOP14. Additional data are required to clarify the clinical significance of c.6292 G > A/p.(Asp2098Asn) identified in LMM11.

Patients with c.9861 C > T reported in this study presented with delayed-onset, mild-to-moderate hearing loss, as opposed to the congenital profound deafness typically observed in individuals with MYO15A-related hearing loss, suggesting the splicing defect of c.9861 C > T may be leaky. Notably, the majority of patients with c.9861 C > T passed newborn hearing screening, developed hearing loss during early childhood, and had a progressive loss when long-term follow -up information was available. Because this variant is relatively common in the AJ population, genetic screening of this variant may identify carrier parents and newborns with AJ ancestry at risk of hearing loss.

In conclusion, the c.9861 C > T variant in MYO15A is a relatively common synonymous variant enriched in the AJ population. It leads to abnormal splicing. Individuals homozygous or compound heterozygous for the variant show childhood onset progressive hearing loss.