Identification of Homozygous Mutations for Hearing Loss

Abstract

Background

Hearing loss is the most common sensory disorder worldwide, affecting about 1 out of every 1000 newborns. The disease has major genetic components, and can be inherited as a single gene disorder either in autosomal dominant or recessive fashions. Due to the high rate of consanguineous unions, Iran has one of the highest prevalence of autosomal recessive nonsyndromic deafness (ARNSD) in the world.

Methods

We carried out a genetic screening of ten Iranian kindreds with more than one offspring affected by ARNSD caused by consanguineous unions. Sanger sequencing and whole exome sequencing together with in silico 3D structure modeling and protein stability prediction were used to identify the underlying disease causing genes.

Conclusion

We identified the causes of deafness in all 10 kindred. In six kindreds homozygous mutations were identified in GJB2 gene by Sanger sequencing. By using whole exome sequencing (WES), a homozygous missense mutation was identified in ESRRB gene as the first ever reported disease gene in Iran. Also two novel homozygous frameshift and missense mutations were identified in MYO15A gene and one previously reported mutation in TMC1 gene in three independent kindred.

Our study shows the efficacy of WES for unraveling new pathogenic mutations in ARNSD patients and expands the spectrum of genes contributing to ARNSD in the Iranian population. The findings of our study can facilitate future genetic screening of patients with ARNSD , early screening and optimal design of novel therapeutics.

Background:

Hearing loss (HL) is the most prevalent sensory defect worldwide, affecting 0.1% to 0.3% of newborns (Morton & Nance, 2006).The frequency of deafness in Iran has been reported to be higher, creating a large medical and financial burden for the health care system and an urgent need to understand disease pathogenesis and identify the causal factors. Increased prevalence of consanguinity in the country suggests a major contribution of homozygous alleles and autosomal recessive nonsyndromic inheritance of deafness (ARNSD), accounting for about 50% of cases (Bademci et al., 2016; Mahdieh, Rabbani, Wiley, Akbari, & Zeinali, 2010). About two-thirds of hearing loss are nonsyndromic (Mahdieh, et al., 2010). Based on earlier genetic studies there is a high level of genetic heterogeneity with mutations in GJB2 (OMIM#121011) comprising the most frequent cause of autosomal recessive hearing loss worldwide and about16–20% of cases in Iran (Daneshi et al., 2011; Hashemi, Ashraf, Saboori, Azarpira, & Darai, 2012; Kenneson, Van Naarden Braun, & Boyle, 2002; Najmabadi et al., 2005; Zelante et al., 1997). The high prevalence of the disease, the excessive degree of consanguinity and ethnic the diversity in Iran raises the possibility of unknown genetic causes of HL (Babanejad et al., 2012; Bazazzadegan et al., 2012; Davarnia et al., 2012; Mehrjoo, Babanejad, Kahrizi, & Najmabadi, 2015; Naghavi et al., 2008).

Whole Exome Sequencing (WES) is a high throughput technique which enables researchers and clinicians to obtain nearly complete sequences of all coding regions and exon-intron boundaries as well as copy number variation (S. B. Ng et al., 2009; Warr et al., 2015). Since rare monogenic diseases are most commonly caused by mutation in coding regions, WES offers a more efficient and cost-effective method for analysis of hereditary mutation detection compared to whole genome or targeted sequencing.

Accordingly, growing number of clinical laboratories and researchers use WES as the method of choice for diagnosis of genetic variants (Atik, Bademci, Diaz-Horta, Blanton, & Tekin, 2015; Diaz-Horta et al., 2012).

We recruited 10 kindreds with two or more offsprings of consanguineous unions affected with deafness. All affected subjects had nonsyndromic hearing loss and the transmission of disease was consistent with an autosomal recessive pattern of inheritance. Genetic screening to identify the cause of deafness was carried out by targeted sequencing of GJB2 gene followed by exome sequencing. We identified the genetic causes of deafness in all kindreds. The yield was much greater compared to previous studies, in part due to the use of a potent tool of high throughput compared to targeted sequencing.

Materials and methods:

Patient recruitment

Patients were clinically evaluated by the ENT specialists of Khalili hospital – Shiraz – Iran. Complete physical examinations, otoscopy and electrophysiological testing like otoacoustic emission (OAE) testing, auditory brainstem response (ABR) and auditory steady-state response (ASSR) testing were performed for all patients. The comprehensive evaluation included computed tomography (CT) scan or magnetic resonance imaging (MRI), which were carried out based on symptoms and findings. Subsequently, patients were referred to Department of Medical Genetics for pedigree construction and genetic evaluations. The representative pedigree of the kindreds is shown in Fig. 1. All kindreds had at least 2 affected individuals and parents were first cousin.

Figure 1.

The representative pedigree of the kindreds is shown. All kindreds had 2 affected individuals and parents were first cousin. The index case is denoted by arrow, the deafness phenotype is indicated by filled black symbols; Squares represent males and circles females.

Collectively 10 consanguineous families with two or more affected children were enrolled. Most of the families were from Fars province in the south west of Iran. The probands had prelingual hearing loss with no syndromic features and the inheritance of deafness was consistent with an autosomal recessive pattern. All affected individuals showed a severe to profound hearing loss from early in childhood. The Ethics Committee of Shiraz University of Medical Sciences approved the study. The patients signed a written informed consent to participate in this study.

Targeted Sanger sequencing of GJB2 gene

The peripheral blood samples of patients were collected after obtaining written informed consent. The DNA of individuals was extracted by DNA easy kit (Qiagen, USA) according to manufacturer protocol. The samples were analyzed by Sanger sequencing for c.35delG in GJB2 gene, the most common mutation in autosomal recessive hearing loss.

Whole-exome sequencing

Subjects with a negative result for GJB2 mutations were selected for exome sequencing. Sequence Capture Human Exome 2.1M Array (Roche NimbleGen), was used for enrichment of the coding exons and flanking intronic sequences instructed by the manufacturer. Captured coding DNA samples were sequenced on the HiSeq 4000 instrument (Illumina San Diego, CA). The sequence data were processed by mean of MAQ software(Li, Ruan, & Durbin, 2008). Sequence reads were aligned by human reference genome (hg19 NCBI). GATK software (https://www.broadinstitute.org/gatk/) was used to call variants (McKenna et al., 2010). Single-nucleotide variants were detected by SAMtools software. The raw data was filtered again to discard common variants reported in reference genomes (P. C. Ng et al., 2008). Variants were filtered for allele frequencies greater than 0.001% in the ExAC database. Mutation pathogenicity was assessed using PolyPhen-2 and SIFT predicting software and filtered if considered not damaging by either software.

Variants were annotated according to the conservation, novelty and tissue expression by the mean of a pipeline for genome annotation (Choi et al., 2009).Variants were confirmed by Sanger sequencing in all affected and non affected members of the family.

Bioinformatics analysis

WES data were analyzed for coverage and average read depth of sequenced regions.

Variants were filtered based on allele frequency of <0.001 in dbSNP141 (Shearer et al., 2014) and Yale Center for Genome Analysis exome database (Keramati et al., 2014). After these filtering steps, significant alterations such as nonsense, missense, splice site, in-frame and outframe INDELs were selected. Pathogenicity of non-reported alteration was checked by prediction tools: PolyPhen2 (Adzhubei et al., 2010), SIFT (Kumar, Henikoff, & Ng, 2009), PROVEAN, and Mutation Taster.

The variant filtering strategy is shown in a flow chart (Fig. 2).

Figure 2.

A flow chart demonstrating genomic variant filtering strategy of exome sequencing data. Total of 52785 genomic variants were identified in exome sequencing of 4 patients who were excluded for GJB2 gene mutations. LOF represents: splice-site, frameshift, stop-gained, stop-lost and frameshift indels mutations.

3D structure modeling, validation and refinement of modeled structure

I-TASSER and GalaxyTBM servers were used for generating 3D structure of proteins. I-TASSER server at http://zhanglab.ccmb.med.umich.edu/I-TASSER, an automated prediction tool, utilizes multiple-threading alignments, and iterative template fragment assembly simulations strategies for modeling protein 3D structure (Yang & Zhang, 2015). In Galaxy TBM at http://galaxy.seoklab.org/cgi-bin/submit.cgi?type=TBM, modeling 3D structure is done based on two steps; first, more reliable core structures are generating by selecting multiple templates (up to 20). In the second step, less reliable loop or terminus regions are identified and re‐modeled applying an optimization-based refinement method (Ko, Park, & Seok, 2012).

The 3D structures of modeled proteins were validated by three servers; Ramachandran plot analysis at http://mordred.bioc.cam.ac.uk/Brapper/rampage.php (Lovell et al., 2003), the ProSA-web at https://prosa.services.came.sbg.ac.at/prosa.php (Wiederstein & Sippl, 2007), and the ERRAT server at http://services.mbi.ucla.edu/ERRAT/ (Colovos & Yeates, 1993). The Ramachandran plot verified the residue-by-residue stereochemical qualities of models,and represents the number of residues that are located in favored, allowed, and outer regions. The ProSA-web z-score represents the overall model quality in the plot that includes the Z-scores of experimentally defined structures deposited in PDB. The ERRAT server evaluates the statistics of non-bonded atom–atom interactions in compared to a database of reliable highly refined crystallography structures.

The refining of some modeled protein was performed by Galaxy Refine server at http://galaxy.seoklab.org/cgi-bin/help.cgi?key=METHOD&type=REFINE (Heo, Park, & Seok, 2013).GalaxyRefine first rebuilds all side-chain conformations and then constantly relaxes the structure by short-range molecular dynamics simulations after side-chain repacking perturbations in order to improve the initial model.

Prediction of mutational effect on protein Stability

The stability changes of mutants were predicted by five different web-based prediction tools to analyze and predict protein stability changes upon point mutations: CUPSAT at http://cupsat.tu-bs.de (Parthiban, Gromiha, & Schomburg, 2006), DynaMut at http://biosig.unimelb.edu.au/dynamut/ (Rodrigues, Pires, & Ascher, 2018), SDM at http://marid.bioc.cam.ac.uk/sdm2/ (Rodrigues, et al., 2018), I-Mutant at http://folding.biofold.org/i-mutant/i-mutant2.0.html (Rodrigues, et al., 2018), and MUpro at http://mupro.proteomics.ics.uci.edu (Cheng, Randall, & Baldi, 2006); the first three servers are structure-based, and the last two servers are sequence-based. CUPSAT calculates the effect of mutations on the protein stability applying protein environment specific mean force potentials. The potentials are extracted from statistical evaluation of protein structure data sets. DynaMut applies Normal Mode Analysis (NMA) via two different methods, Bio3D and ENCoM for analyzing protein. Moreover, DynaMut can evaluate the effect on mutations on a protein stability through vibrational entropy changes.SDM is a knowledge-based method that applies conformationally constrained environmentally-dependent amino acid substitution tables to predict the change to the protein stability between wild-type and mutant protein. MUpro and I-Mutant are support vector machine (SVM) based methods to predict stabilizing or destabilizing amino acid substitutions based on free energy change Gibbs free energy (ΔΔG). The results of all servers are designated in terms of ΔΔG value, in which the negative value corresponds to the destabilizing effect of mutant.

Results

The causative mutations were identified in all 10 families. Homozygous mutations were identified in GJB2 (c.35delG) genes in 6 independent families by Sanger sequencing. We analyzed data from whole-exome sequencing in the four remaining index patients. Four homozygous deleterious mutations were identified in four affected families (Table1). Three were missense and one was a frameshift mutation. Two of these mutations were novel and had never been previously reported. The homozygous state in all affecteds and the absence of the mutation or heterozygous state in unaffected parents or sibling were confirmed by Sanger sequencing. In all families affected persons were homozygous; parents were heterozygous and unaffected sibs, if present, were normal or heterozygous.

Table 1.

Mutations detected in this study.

Locus	Gene	Family	Prelingual/ Postlingual	Progressive/ nonprogressive	Severity	Type of mutation	Nucleotide change	Amino acid change
DFNB1	GJB2	1,2,4,6, 7,10	Prelingual	nonprogressive	Severe to profound	Deletion	c.35delG	p.G12Vfs*2
DFNB3	MYO15A	3	Prelingual	nonprogressive	Severe to profound	Insertion	c.414dupA	p.F141Vfs*87
		5	Prelingual	nonprogressive	Severe to profound	missense	c.9467T>C	p.L3156P
DFNB7/11	TMC1	8	Prelingual	nonprogressive	Severe to profound	missense	c.1334G> A	p.R445H
DFNB35	ESRRB	9	Prelingual	nonprogressive	Severe to profound	missense	c.536G>A	p.R179H

The missense c.1334G> A mutation in exon 16 of transmembrane channel-like 1(TMC1) (NM_138691; chr9:75406911; rs760532554) was a homozygous variant, leading to arginine 445 for histidine amino acid substitution. Based on gnomAD database, the allele frequency of this variant is extremely low (0.0000119). This mutation has not been reported previously as disease-causing and is pathogenic in most predicting software (Table2).

Table 2.

ACMG classification, pathogenicity scores and allele frequencies of mutations detected in this study.

	GJB2 c.35delG	ESRRB c.536G>A	MYO15A c.414dupA	MYO15A c.9467T>C	TMC1 c.1334G>A
ACMG	Pathogenic	Likely pathogenic	Pathogenic	VUS	Likely pathogenic
Mutation Taster	Disease causing	Disease causing	Disease causing	Disease causing	Disease causing
Polyphen2	NA	Damaging	NA	Damaging	Damaging
SIFT	NA	Damaging	NA	Damaging	Damaging
PROVEAN	NA	Damaging	NA	Damaging	Damaging
gnomAD frequency	0.00597	0.0000159	0	0	0.0000119
Clinvar	Pathogenic	(Not reported)	(Not reported)	(Not reported)	(Not reported)
Iranome frequency	0.001875	0	0	0	0
Ref.	Zelante, et al., 1997	Richard, et al., 2019	This study	This study	Santos, et al., 2005

A missense mutation was identified in ESRRB gene (NM_004452.4: exon6:c.536G>A; chr 14:76948380; rs752185665), resulting in arginine 179 for histidine substitution in estrogen-related receptor 2 domain. The mutation is predicted as damaging by most prediction software (Table2). Based on gnomAD database, the allele frequency of this variant is extremely low (0.0000159). The disease gene has never been reported as the cause of ARNSD in Iran.

Two homozygous mutations were identified in MYO15A (NM_016239.4) gene in two independent kindred, which are novel in the Iranian population. One was a frame shift (exon 2: c.414dupA; chr 17:18022527; rs750130520) and the other was a missense mutation (exon 56: c.9467T>C; chr 17:18064711) leading to p.L3156P substitution. Both mutations are pathogenic according to PolyPhen and SIFT prediction software and are absent in gnomAD database (Table2).

3D structure modeling, refinement, and validation

In this study, I-TASSER and Galaxy Web servers were employed for 3D structure modeling of studied proteins. After the refinement process, the best model of each protein was selected for mutational analysis. The ProSA-web, ERRAT, and Ramachandran plot validations of the final 3D model ofTMC1, ESRRB, and MYO15A proteins are shown in Table 3. Due to the extreme length of MYO15A protein (3530 amino acids), only the c-terminal segment of protein containing the mutated amino acid, starting from codon 671 (D) was modeled. The validation results presented the 3D models had high-quality structures and were used for mutational analyses.

Table 3.

The results of final 3D model of TMC1, ESRRB, and MYO15A proteins evaluated by ProSA-web, ERRAT, and Ramachandran plot.

Protein Name	ProSA-Web (Z scores)	Ramachandran Plot (Residue Distribution in three regions)	ERRAT Overall Quality Factor
TMC1	−6.4	Favored region: 93.7 Allowed Region: 4.1 Outlier region: 2.2	86.09
ESRRB	−6.29	Favored region: 96.3 Allowed Region: 3.0 Outlier region: 0.7	83.84
MYO15A	−9.1	Favored region: 92.1 Allowed Region: 7.4 Outlier region: 0.5	90.59

Effects of the substitution mutation on protein stability

In order to increase the accuracy of prediction of protein stability four other web-based tools, including DynaMut, SDM, MUpro, and CUPSAT were used. As seen in Table 4, ΔΔG of all mutated proteins by all prediction tools except CUPSAT for Unconventional myosin-XV (L3156P) and Transmembrane channel-like protein 1 (R445H), showed negative values, indicating reduced protein stability caused by the mutations. The prediction for the two aforementioned mutations by CUPSAT was stabilizing, but unfavorable. The 3D structures of three proteins and the related mutated amino acids in each protein are shown in Fig 3. The duplication in nucleotide 414 (A) of MYO15A gene results in frame shift mutation and subsequent translation termination (TGA stop codon in position 679–681), generating a truncated protein of 226 amino acids or no protein at all due to the nonsense-mediated RNA decay (Fig 4).

Table 4.

Change in protein stability and the folding free energy (ΔΔG) upon mutation in Unconventional myosin-XV, Steroid hormone receptor ERR2, and Transmembrane channel-like protein 1, evaluated via three servers.

Servers	Protein name	ΔΔG (kcal/mol)	Stability
CUPSAT a	Unconventional myosin-XV (L3156P)	3.2	Stabilizing (Unfavorable)
	Steroid hormone receptor ERR2 (R179H)	−0.32	Destabilizing (Unfavorable)
	Transmembrane channel-like protein 1 (R445H)	0.73	Stabilizing (Unfavorable)
Dynamut a	Unconventional myosin-XV (L3156P)	−0.883	Destabilizing
	Steroid hormone receptor ERR2 (R179H)	−0.61	Destabilizing
	Transmembrane channel-like protein 1 (R445H)	−1.36	Destabilizing
SDMa	Unconventional myosin-XV (L3156P)	−2.23	Reduced Stability
	Steroid hormone receptor ERR2 (R179H)	−0.13	Reduced Stability
	Transmembrane channel-like protein 1 (R445H)	−0.15	Reduced Stability
I-Mutant b	Unconventional myosin-XV (L3156P)	----------	Reduced Stability
	Steroid hormone receptor ERR2 (R179H)	----------	Reduced Stability
	Transmembrane channel-like protein 1 (R445H)	----------	Reduced Stability
Muprob	Unconventional myosin-XV (R445H)	−1.90	Reduced Stability
	Steroid hormone receptor ERR2 (R179H)	−0.64	Reduced Stability
	Transmembrane channel-like protein 1 (R445H)	−1.58	Reduced Stability

^a3D model of protein is used as an input.

^bPrimary sequence is used as an input.

Figure 3.

3D structures of three proteins and the corresponding substituted amino acids: A) Arg179His amino acid substitution in steroid hormone receptor ERR2 protein. B) Leu297Pro amino acid substitution in unconventional myosin-XV. C) Arg445His amino acid substitution in transmembrane channel-like protein 1.

Wild-type and mutant residues are colored in light-green and are also represented as sticks alongside with the surrounding interacting residues.

Aromatic contacts, water mediated weak hydrogen bonds, hydrophobic contacts, hydrogen bond, and ionic interactions are shown in blue, orange, green, red, and yellow dotted lines, respectively.

Figure 4.

Mutation in MYO15A gene c.414dupA (p.Phe141ValfsTer87). A) Duplication in adenine 414 (in black box) of the MYO15A gene, results in a frame shift, causing a premature termination (TGA) at position 679–681, shown in red. The duplicated adenine is presented in green. B) The truncated protein consists of 226 amino acids. The stop codon is shown in red.

Discussion

The present study aimed to shed light on the genetic etiology of autosomal recessive non-syndromic deafness (ARNSD) in the Iranian population. The genetic diagnosis of ARNSD is critical as it can help with early diagnosis, management and development of strategies for treatment of affected patients such as early speech therapy and cochlear implantation and for the determination of carrier status. In general, the success rate in identifying disease genes for ARNSD is about 70% (Sloan-Heggen, Babanejad et al. 2015). Strikingly, we found causative mutation in all ten patients, which is likely due to the advantage of WES over targeted sequencing of candidate genes. More than half of ARNSD cases are caused by mutations in GJB2, with c.35delG being the most common, making up about 70% of all GJB2 mutations (Duman & Tekin, 2012; Kelley et al., 1998)[18, 19]. Most other ARNSD are caused by rare private mutations (Diaz-Horta, et al., 2012). Najmabadi et al., reported a lower prevalence of GJB2-related deafness in the entire Iran while Bonyadi et al. showed a higher prevalence of GJB2 mutations in Iranian Azeri patients. In our study 60% of the patients had a GJB2 mutation, which is higher than expected compared to the study by Najmabadi. This discrepancy is likely due to the population diversity in Iran (Babanejad, et al., 2012; Bonyadi, Esmaeili, Abhari, & Lotfi, 2009).

Since identification of connexin 26 (GJB2) as a cause of severe deafness in 1997 to date, around 66 genes have been implicated in ARNSHL (Kelsell et al., 1997; Van Camp & Smith, 2017). In the present study mutations in GJB2, MYO15A, ESRRB and TMC1 genes were found in ten Iranian kindred with ARNSD. Six kindred had homozygous mutations in GJB2.

In one kindred a previously reported homozygous mutation in the TMC1 gene (c.1334G<A) was identified. TMC1encodes a 6-pass integral membrane protein and have cytoplasmic orientation of N and C termini which is a component of mechanotransduction channels in inner ear’s hair cells (Kurima et al., 2002; Pan et al., 2013). TMC1 and TMC2 expression are essential for permeation properties and sensory transduction in auditory and vestibular hair cells and mutation in TMC1gene may lead to reduced calcium selectivity and single-channel current. More than 30 TMC1 mutations in Middle Eastern families with ARNSD have been identified (Atik et al., 2015; Kurima, Yang, Sorber, & Griffith, 2003; Nakanishi, Kurima, Kawashima, & Griffith, 2014; Pan, et al., 2013).

In one of the kindreds a homozygous mutation in ESRRB gene (c.536G>A) was identified that is novel in the Iranian population. Affected individuals of this family showed bilateral severe to profound hearing loss and absent of otoacoustic emission (OAE) and no syndromic manifestations (Fig 5). All individuals underwent a brain MRI examination and no structural abnormalities were identified. There was no unique clinical characteristics between the patients in our study and those reported by other (Collin et al., 2008). This gene has been mapped to DFNB35 locus and encodes a protein that is structurally and in sequence similar to the estrogen receptor. Its expression has been shown to be essential for somatic cell reprogramming, pluripotency and inner ear development and function. It is postnatally expressed in the human cochlea (Collin et al., 2008; Doege et al., 2012; Luo et al., 1997). This is the first report of mutation in this gene in patients with ARNSD in Iran and should be included in the diagnostic testing panel for inherited deafness. In our study two novel homozygous mutations in the MYO15A gene (exon 2: c.414dupA, p.F141Vfs*87; and exon 56: c.9467T>C, p.L3156P ) were identified in two independent kindreds. MYO15A gene is composed of 66 exons and consists about 71 kb, encoding myosin XV protein. MYO15A is expressed in the human fetal cochlea and brain and is a new branch of the myosin superfamily, which possesses a conserved N-terminal motor domain, 2 light-chain binding IQ motifs, a tail region containing a MyTH4, a talin-like domain and a PDZ ligand at the C-terminal tip. C-terminal PDZ ligand of MYO15A interacts with the third PDZ domain of whirlin and is required for the targeting of whirlin in to the tips of stereocilia and hair bundle morphogenesis (Belyantseva et al., 2005; Liang et al., 1999; Nal et al., 2007).

Figure 5.

Audiogram of affected person in family 9 with mutation in ESRRB gene. Audiogram demonstrating bilateral severe to profound hearing loss.

In order to obtain precise insight into the effect of mutation on 3D structure of proteins, first the tertiary structures of three proteins (TMC1, ESRRB, and MYO15A) were modeled by GalaxyTBM server. According to validation results, the best 3D structure of each protein was used for the evolution of the effect of substitution mutation on the stability and 3D structure of proteins.

Substitution mutation of Arg179His in ESRRB resulted in formation of one aromatic contact (blue dotted line) (Fig. 3A). Substitution mutation of Leu297Pro resulted in formation of one water mediated weak hydrogen bond (orange dotted line), and one hydrophobic contact (green dotted line); moreover, one hydrogen bond has been broken (red dotted line) (Fig. 3B). Substitution mutation of Arg445His in TMC1 resulted in formation of one water mediated weak hydrogen bond (orange dotted line), and two hydrogen bonds (red dotted line) (Fig. 3C).

Conclusion

Our study identified several novel recessive mutations in Iranian population. Identification of causative mutations for inherited deafness could facilitate diagnosis and management of patients with ARNSD and help with early design of therapeutic plans. Our results broaden the spectrum of genes contributing to ARNSD in the Iranian population. Finally, our study showed the efficacy of WES for unraveling new pathogenic mutations in ARNSD patients.

Highlights.

• Hearing loss is the most common sensory disorder in the world.

• Genetic screening of ten Iranian kindreds with more than one offspring affected by autosomal recessive non-syndromic deafness (ARNSD) was performed.

• Two novel homozygous missense mutations were identified in MYO15A gene and two previously reported mutations were identified in ESSRB and TMC1 genes in Iran.

• The structural computational analyses, also confirmed our finding.

Abbreviation

ARNSD

Autosomal recessive nonsyndromic deafness

HL

Hearing loss

WES

Whole Exome Sequencing

OAE

otoacoustic emission

ABR

auditory brainstem response

CT

Computed tomography

MRI

magnetic resonance imaging

SVM

support vector machine

NMA

Normal Mode Analysis

ΔΔG

change Gibbs free energy

ARNSHL

auotosomal recessive non-syndromic hearing loss