Comprehensive molecular analysis of 61 Egyptian families with hereditary nonsyndromic hearing loss

Nonsyndromic hearing loss is an extremely heterogeneous disorder. Thus, clinical diagnostics is challenging, in particular due to differences in the etiology of hearing loss between populations. With this study, we wanted to elucidate the genetic basis of hearing loss in 61 consanguineous Egyptian families. In 25 families, linkage analysis was used as a prescreening to identify regions for targeted sequencing of candidate genes. Initially, the coding regions of 12 and later of 94 genes associated with hearing loss were enriched and subjected to massively parallel sequencing (MPS) with diagnostic yields of 36% and 75%, respectively. Causative variants were identified in 48 families (79%). They were found in 23 different genes with the majority being located in MYO15A (15.3%), SLC26A4 (9.7%), GJB2 (8.3%), and MYO7A (6.4%). As many as 32 variants were novel ones at the time of detection. Five variants were shared by two, three, or even four families. Our study provides a first survey of the mutational spectrum of deaf patients in Egypt revealing less GJB2 variants than in many European populations. It underlines the value of targeted enrichment of well‐selected deafness genes in combination with MPS in the diagnostics of this frequent and genetically heterogeneous disorder.

30%). Eighty percent of nonsyndromic hearing loss (NSHL) follow an autosomal recessive (AR) mode of inheritance, <20% an autosomal dominant one (AD), and 1% are either X-linked or mitochondrial forms. 2 Inherited nonsyndromic hearing loss is characterized by a remarkable genetic heterogeneity. According to recently published data, more than 1000 causative variants have been identified in more than 90 genes. 3 Despite this enormous heterogeneity, there is one very common cause of severe-to-profound autosomal recessive nonsyndromic hearing loss (ARNSHL) in populations of Europe, the Middle East, and North America, namely variants of the GJB2 gene, which codes for Connexin 26. In particular, the variant NM_004004. 5: c.35delG has been found to be the most common cause of ARNSHL in the mentioned populations, being responsible for up to 63% of cases. 4 Previous studies on the cause of deafness in African populations or countries, including the Egyptian population, were mainly focused on the analysis of GJB2. Snoeckx et al. analyzed the coding sequence of GJB2 in 111 Egyptian families (159 individuals) with nonsyndromic hearing loss. They found that in only 14.4% of them ARNSHL was due to mutations in GJB2, yet similar to Caucasian populations NM_004004.5:c.35delG was the most frequently found mutation in those families. 5 Another study, performed on 97 families (155 individuals) from Southern Egypt, revealed an allele frequency of 8.7% for c.35delG and one other GJB2 mutation in single family-the novel missense variant NM_004004.5:c.T212A (p.I71N). 6 A further study reported the mitochondrial variant in MTRNR1 NC_012920.1: m.1555A>G to rarely cause SNHL in Egyptian patients. 7 This is the first study aimed at a comprehensive detection of gene variants causing hearing loss in Egyptian families by taking advantage of massively parallel sequencing (MPS) of beforehand-enriched DNA targets. In total, we analyzed 61 consanguineous families from Southern Egypt. In a first step, 25 families without causal variants in GJB2 were genotyped on a 250 K SNP array for homozygosity mapping resulting in 12 candidate genes analyzed by MPS. In a second step, the remaining 36 families plus 16 families that could not be solved in the first step were analyzed using targeted enrichment of 94 genes and MPS. With this partially two-tiered approach, we identified candidate variants in 48 out of the 61 families under investigation, which we confirmed by segregation analysis. This corresponds to a diagnostic yield of 78.7%. The largest number of causal variants per gene was found in MYO15A.

| Subjects
The study group included 61 families with hereditary nonsyndromic hearing loss recruited from Governorates of South Egypt (for pedigrees see Figure S3). The families with A-ID numbers represent a subset of an earlier study on the frequency of GJB2 variants in 97 families from Southern Egypt. 6 We selected 25 families from the negatively tested ones for further analysis in this study. The 36 families with B-ID numbers had not been screened for GJB2 variants prior to inclusion into this study. Informed consent was obtained from all participants in the study.
The study was approved by the ethical committee, Faculty of Medicine, Assiut University, Egypt and performed in compliance with national legislation and international standards (Declaration of Helsinki).

| Clinical evaluation
All participants had no history or manifestation suggestive of environmental causes of hearing loss, and no active or recent history of an otological problem. Clinical evaluation of all individuals included history taking, otoscopic examination and basic audiological evaluation using a dual channel clinical audiometer (Obiter 922) and immittancemetry measurement with Impedance Audiometer Interacoustic AZ 26. When reliable responses via conventional audiometry were not possible an Auditory Brainstem Response (ABR) was used for threshold estimation (Nicolet Spirit OS/2).

| DNA extraction and prescreening
DNA was extracted from peripheral blood using standard methods. In the first set of families (A-ID numbers) a prescreening for GJB2 variants by restriction assays, single strand conformational polymorphism analysis and Sanger sequencing had been performed as part of a previous project. 6 A subset of 25 families without HL associated variants in GJB2 was transferred to this study. The second set of 36 families (B-ID numbers) was not part of the previous project and had not been subjected to a prescreening for GJB2 variants.

| Linkage analysis
Genome-wide linkage analysis of 51 individuals from the prescreened 25 families was carried out using GeneChip Human Mapping 250 K SNP Array data (Thermo Fisher Scientific). Genotypes were called by the GeneChip DNA Analysis Software (GDAS v3.0, Thermo Fisher Scientific). Subsequent data handling was performed using the graphical user interface ALOHOMORA. 8 Relationship errors were identified using the program Graphical Relationship Representation. 9 The program PedCheck was applied to find Mendelian errors 10 and data for SNPs with such errors were removed from the data set. Non-Mendelian errors were identified using the program MERLIN 11 and unlikely genotypes for related samples were deleted. Linkage analysis was performed assuming autosomal recessive inheritance, full penetrance, consanguinity and a disease allele frequency of 0.0001. Multipoint LOD scores were calculated using the programs ALLEGRO 12 or MERLIN. 11 Also haplotypes were reconstructed with either of these programs and presented graphically with HaploPainter. 13 2.5 | Estimation of variant age by IBD DNA samples of 14 individuals belonging to five families were genotyped on the Axiom Precision Medicine Research Array (Thermo Fisher Scientific). Genotypes were called by the Axiom Analysis Suite v4.0 (Thermo Fisher Scientific). Genotype data were checked for relationship errors, Mendelian errors and unlikely genotypes, as described before. Haplotypes in linkage regions covering the causing variant were reconstructed in order to determine the length of the shared haplotype between two or more families. The minimum variant age was estimated from the shared haplotype length as described elsewhere. 14 2.6 | Targeted enrichment of 12 genes (gene panel 1, GP1) and MPS The RainDance enrichment assay was designed to enrich all exons including the exon-intron boundaries of 12 genes known to be associated with hearing loss or located in linkage regions (for gene list see Table S1). The assay comprised 373 exons to capture 377 targets with   Table S1). GP2 was designed with the SureDesign online tool covering 1504 targets with a total probe size of 496 kb and finally ordered from Agilent. In total, 62 patient DNA samples of 52 families were analyzed together with four patient DNA control samples. Library preparation and subsequent enrichment were performed using the SureSelect custom XT enrichment protocol (Agilent Technologies). For this, 3 μg of genomic DNA underwent fragmentation using the Bioruptor (Diagenode) sonication method.
Fragments were end-repaired, A-tailed and adapter-ligated and individually enriched according to the standard protocol. A pool of 16 libraries was loaded on a MiSeq, HiSeq2000 or HiSeq4000 sequencer (Illumina). Runs of 2 × 100 bp or 2 × 75 bp, respectively, generated on average 152 Mb of raw data per sample.

| Controls
We analyzed eight patient DNAs with already known mutations as positive controls in the GP1 enrichment experiments. 15

| Linkage Analysis
After exclusion of GJB2 variants, 6  F I G U R E 2 Genome-wide HLOD plot of 25 families. MERLIN with LD modeling was used for Linkage analysis assuming autosomal recessive inheritance, full penetrance, consanguinity, and a disease allele frequency of 0.0001. 11 Chromosome numbers are given on the top of the plot; genetic distance and HLOD are displayed on the xand y-axis, respectively POU4F3, TCOF1). We added six further genes, which are all frequently mutated deafness genes (see Table S1). We enriched the coding  Table S2.
Since we could not uncover the genetic cause of ARNSHL in 16 out of the 25 analyzed families, a more comprehensive analysis was required. Therefore, we designed a panel of 94 hearing loss genes (GP2, see Table S1) and analyzed at least one member of each of the 16 remaining families with unknown mutations and of 36 additional, mainly consanguineous families, which were not prescreened for Most variants were homozygous; only three compound heterozygous states were observed, concerning variants in ADGRV1 (2x) and LOXHD1 (1x), and one heterozygous or hemizygous state each in families with autosomal dominant or X-linked inheritance, respectively (for summary of variants see Table 1, for variants of each family see Table S2).
Taking the results from both gene panel sequencing analyses together, we noticed that some variants of the tested candidate genes are present in more than one family (see Table 1 increases from Western Sahara to Egypt. 46 The low frequencies of  Figure 3A and Figure S2). In total, we F I G U R E 3 Frequency and distribution of gene variants co-segregating with hearing loss in 61 Egyptian families. A, Percentages of gene variants associated with SNHL. The frequency of families with GJB2 variants refers to a total of 133 families merging the results of the present study for the 36 families without prescreening for GJB2 variants with data of our previous study on GJB2 variants in 97 families. 6 Genes with variants in one family only are lumped together as "other genes." In total, variants of deafness genes were detected and confirmed by segregation analysis in 48 of 61 families (79%). B, Distribution of putative consequences at the protein level within subgroups of patients showing the same phenotype. Abbreviations: ntSNHL, near total SNHL; pSNHL, profound SNHL; spSNHL, severe to profound SNHL; sSNHL, severe SNHL; msSNHL, moderately severe SNHL; mSNHL, moderate SNHL found 44 different variants, 32 of which were novel ones at the time of detection. Types of variants in terms of different consequences at the protein level were widely spread over patients grouped by graduated severity of hearing loss, except that milder phenotypes seemed to be mainly due to missense variants ( Figure 3B).
Splice site variants were predicted for five different genes in seven families (see Table 2 The missing goiter in our family is probably due to its later manifestation, as all affected children were younger than 12 years at the time of examination. The median age of goiter appearance in a previous study was 14.9 years for patients with Pendred syndrome. 49 So a follow-up of this family will be important. (Thr36Asn) that is located in the second transmembrane domain (TD) of the encoded protein barttin. 52 The variant is predicted to be probably damaging by PolyPhen-2 and damaging by SIFT. Since all affected individuals of both families were not diagnosed with renal dysfunction (normal blood urea and serum creatinine levels), the variant could be solely responsible for ARNSHL as described in other studies. [52][53][54] One of the described BSND mutations associated with hearing loss only was also found in the second TD. 54