A total of 90 unrelated Brazilian individuals were selected for having hearing loss of presumptive autosomal recessive inheritance. They were referred by the following institutions: DERDIC (Divisão de Educação e Reabilitação de Distúrbios Auditivos da Comunicação da Pontifícia Universidade Católica), Hospital das Clínicas da Faculdade de Medicina da USP e UNIFESP (Universidade Federal de São Paulo), all in the state of São Paulo, and other institutions. They were ascertained between 2000 and 2019 at the Human Genome and Stem Cell Research Center, Instituto de Biociências, Universidade de São Paulo, São Paulo, SP, Brazil. They were selected based on pedigree criteria, being either born from consanguineous couples or belonging to families with two or more affected individuals in the sibship and most cases were of normal-hearing parents. Although most probands presented with non-syndromic hearing loss, individuals with suspected syndromic deafness were not excluded, as long as they followed at least one of the previous criteria. The study was approved by the Ethics Committee, Institute of Biosciences, University of São Paulo (1.133.416–23/06/2015). Written informed consent was obtained from all hearing-impaired individuals, and relatives, or their legal guardians. Information about secondary actionable findings was not offered to patients because samples were collected in research focused on hearing loss, and written informed consent authorized investigation of genetic factors related to hearing loss, exclusively.

In all probands, the molecular cause of hearing loss had been previously investigated by screening of the c.35delG at GJB2 (NM_004004.5, GRC37) (Scott et al., 1998), screening of two GJB6 deletions, del(GJB6-D13S1830) – chr13: 20,797,177–21,105,967:GRCh37 ∼309Kb, and del(GJB6-D13S1854) – chr13: 20,802,726–21,034,755:GRCh37 ∼232 Kb (Del Castillo et al., 2005), and screening of m.1555A>G at MT-RNR1 (NC_012920.1) (Estivil et al., 1998). Most probands also had the coding region of GJB2 previously analyzed through Sanger sequencing. Only individuals without an identified molecular cause for hearing loss after this screening were included. However, four cases identified with hypomorphic variants in GJB2 were included in order to verify if there was another plausible genetic cause for their hearing loss.

Blood samples were collected and DNA was extracted using different techniques available at the time of collection of samples: conventional phenol/chloroform extraction, commercial kits, or through the equipment Autopure LS (Gentra Systems, Minneapolis, United States) or QiaSimphony (Qiagen, Hilden, Germany).

The custom panel was designed using the SureDesign tool (https://earray.chem.agilent.com/suredesign/) and the DNA libraries were prepared according to the SureSelect QXT Target Enrichment System Kit for Illumina Multiplexed Sequencing (Agilent, Santa Clara, CA) protocol. Sequencing was performed on the HiSeq 2500 platform (Illumina, San Diego, California, United States). Sequences were aligned with the Burrows-Wheeler Aligner (BWA) (http://bio-bwa.sourceforge.net/). The Picard tool (http://broadinstitute.github.io/picard/) was used to eliminate PCR duplicates, identify indels, and realign reads. Recalibration of base qualities and variant calling were performed using Genome Analysis ToolKit (GATK) (https://github.com/broadinstitute/gatk) and the annotation of variants was performed using Annovar (http://annovar.openbioinformatics.org/en/latest/). The list of 99 genes is presented in Supplementary Material, in Supplementary Table S1.

Libraries were captured and constructed through one of the following kits: Nextera Rapid Capture (Illumina, San Diego, California, United States), TruSeq Exome Library Prep Kit (Illumina, San Diego, California, United States), SureSelectQXT Target Enrichment (Agilent Technologies, Santa Clara, California, United States) and xGen Exome Research Panel v1 (Integrated DNA Technologies, Coralville, Iowa, United States). Sequencing was performed on the HiSeq 2500 platform (Illumina, San Diego, California, United States). Bioinformatics procedures were the same as previously described for the next-generation targeted sequencing of 99 genes.

Exonic and splicing variants were filtered considering their minor allele frequency under 1% in the following databases: Exome Aggregation Consortium (ExAC, exac.broadinstitute.org), NHLBI Exome Sequencing Project (ESP, http://evs.gs.washington.edu/EVS/), 1000 Genomes Project (www.internationalgenome.org), and the Brazilian ABraOM (http://abraom.ib.usp.br). The pathogenicity of the candidate variants was predicted through REVEL, CADD Phred, and for splicing variants through MAX EntScan, ADA Score, RF Score, and SpliceAI. Pathogenicity information was checked at ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) and Deafness Variation Database (http://deafnessvariationdatabase.org/). The GRC37 assembly was used as a reference.

The whole exome sequencing analysis focused on variants present in genes previously associated with hearing loss included at “The Hereditary Hearing Loss Homepage” (https://hereditaryhearingloss.org/), Online Mendelian Inheritance in Man (OMIM, https://www.omim.org/phenotypicSeriesTitles/all) or “Deafness Variation Database”.

Candidate variants were classified according to the American College of Medical Genetics and Genomics (ACMG) and Association for Molecular Pathology (AMP) guidelines for genetic hearing loss (Richards et al., 2015; Oza et al., 2018).

Sanger sequencing was used to confirm the presence of candidate variants and, whenever DNA samples of family members were available, to determine segregation and trans configuration. This information was used in the ACMG classification. Primers to amplify exons containing candidate variants were designed using Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). PCR products were purified using Exonuclease I (Affymetrix, Santa Clara, California, United States) and Shrimp Alkaline Phosphatase (Affymetrix, Santa Clara, California, United States). The purified products were prepared using BigDye terminator v3.1 (Thermo Fisher Scientific, Waltham, Massachusetts, United States) and analyzed with the 48-capillary ABI 3730 DNA Analyser (Thermo Fisher Scientific, Waltham, Massachusetts, United States).

Among the 90 probands, 38% (34/90) were from families with two or more affected individuals in the sibship, 44% (40/90) were born from consanguineous couples, and 18% (16/90) fulfilled both criteria. Nine of the 90 probands had been previously clinically diagnosed with Usher syndrome (Figure 1A).

Forty-six probands had their genetic material analyzed through next-generation targeted sequencing of 99 hearing loss-related genes (Supplementary Material). Due to the discovery of new genes associated with hearing loss that resulted in the need to update repeatedly the customized panel, 63 probands had their genetic material analyzed through whole exome sequencing: 19 probands without identified causative variants identified after the next-generation targeted sequencing of 99 genes panel and 44 additional probands, that had not been previously studied through NGS technologies.

In 32 of the 90 probands (36%) the causative variants were identified: 18 probands after next-generation targeted sequencing of 99 genes, 11 probands after whole exome sequencing and three cases whose causative variants were found after both (Tables 1, 2; Figure 1A). Pedigrees and a summary of the investigation of segregation in families are shown in Supplementary Figure S1. Tables 1, 2 and Figure 1A summarize the identified variants through all the strategies.

Among the 46 subjects analyzed through the customized panel, 18 cases were solved (39%), and four had previously been clinically diagnosed with Usher syndrome.

The genetic cause of hearing loss in one proband was found to be a variant associated with autosomal dominant syndromic hearing loss (SIX1). In total, 24% (6/25) of the disease-causing variants were novel, being the variant c.436C>T in CEACAM16 first described in Dias et al. (2019).

In 14 of the 63 probands (22%) in this group, the causative variants were identified. In three of these probands, no causative variants had been identified previously with the next-generation targeted sequencing of 99 genes: in one of them, the gene harboring the causative variant was not included in the customized panel (FGFR3). In the other two cases (P2RX2 and PDZD7), the variants were not detected after targeted sequencing probably because of technical issues. Causative variants related to syndromic hearing loss were found in two probands (BSND and FGFR3). A causative variant leading to nonsyndromic autosomal dominant hearing loss was found in one proband (P2RX2). In total, 29% (4/14) of the disease-causing variants were novel.

Causative variants in genes described as leading to syndromic hearing loss were found in three probands (RC-26, RC-23, and RC-07) born from consanguineous parents.

Proband RC-26 had a likely pathogenic variant in homozygosis in FGFR3 (NM_000142.4:.c.1618A>G:p.(Asn540Asp). The variant had been described with controversial status, but FGFR3 was already associated with camptodactyly, tall stature, and hearing loss syndrome (CATSHLS). In the first evaluation, at 5 months old, the proband had only camptodactyly in one digit. At the age of 11 years and 2 months, microcephaly was observed. Height was average and flexion restriction of the third digit was observed in both hands. Information about phenotypic variability is limited in CATSHLS syndrome since only four families have been reported (Toydemir et al., 2006; Makrythanasis et al., 2014; Escobar et al., 2016; Cannova et al., 2024). However, the presence of camptodactyly associated with hearing loss and a likely pathogenic variant in FGFR3 pointed to a phenotype related to the CATSHL syndrome spectrum. In three reported cases so far, heterozygous missense variants c.1861C>T:p.(Arg621Cys), and c.1637C>A:p.(Arg621His) were detected and dominant transmission was observed. In a fourth case, c.1637C>A:p.(Thr546Lys) in homozygosis was found in two brothers (Makrythanasis et al., 2014). The catalytic loop of the tyrosine kinase domain of the FGFR3 gene is affected in all cases. The aminoacid residue affected by the variant here described, 540, is very close to the p.(Thr546Lys) variant and interestingly, autosomal recessive inheritance was observed (Makrythanasis et al., 2014), as in the present case. Substitutions of residue 540 were previously associated with hypochondroplasia, but the resulting aminoacid is lysin, p.(Asn540Lys), and not Asp, as in the present case (MIM * 134934). According to UniProt, all variants described above are related to residues located within the protein kinase domain (472–761).

Proband RC-23 presented a pathogenic variant in homozygosis, c.139G>A: p.(Gly47Arg) in BSND, previously associated with Bartter syndrome, type 4A. She was first evaluated at the Genetics Unit at the age of 2 years and 8 months, the second child of consanguineous and healthy parents, presenting failure to thrive and speech delay. She was born at term (37 weeks), with a birth weight of 2,240 g and length of 45 cm, after a pregnancy with no history of polyhydramnios. The mother denies vomiting and the child had an appropriate food intake. No facial dysmorphisms were retrieved. Audiologic evaluation revealed profound hearing loss. In the investigation for her failure to thrive, biochemical analysis revealed metabolic alkalosis (venous blood pH: 7.46, plasma bicarbonate: 30 mmol/L and hypokalemia (2.4mEq/L - NV:3.7–5.0). Other electrolyte abnormalities included hypochloremia (94mEq/L - NV: 97–107), hypocalcemia (1.08 mmol/L - NV:1.12–1.32), as well as increased urine choride/creatinine ratio (390.9mEq/gcreatinine - NV:39–348). She has been referred to the nephrology unit and has been followed with a diagnosis of Bartter syndrome. At the last evaluation, at the age of 8 years and 10 months, she was taking oral potassium supplementation and attending a special school, learning sign language. There was a catch-up on weight gain and growth: W: 22.7 Kg (p.10^th-25^th centile) and H: 120 cm (p.5^th-10^th centile). This syndrome exhibits great phenotypic variability (Bartter et al., 1962; Landau et al., 1995). When first ascertained by us, there was no report of kidney problems. The proband, after a recent reassessment, received a clinical diagnosis of Bartter syndrome due to defects in renal reabsorption associated with hearing loss, which is common in Bartter syndrome type 4A.

Proband RC-07 had a heterozygous pathogenic variant c.386A>G: p.(Tyr129Cys) in SIX1 previously associated with Branchiootic syndrome type 3, which exhibits autosomal dominant pattern of inheritance. This syndrome also exhibits great phenotypic variability and hearing loss is the most frequent signal, found in more than 90% of the cases (SMITH, 2014). Clinical revaluation is needed, but it is possible that hearing loss is the only clinical finding in the proband.

This corroborates ACMG recommendations of periodic clinical revaluation since some phenotypic manifestations in syndromic forms appear later in life, after the onset of hearing loss (Alford et al., 2014), and, in some cases, clinical presentation is mild and early recognition of a syndrome may be difficult. These findings favor Whole Exome Sequencing as a first-tier exam in hearing loss, instead of customized panels, and depending on costs or feasibility, exclusion of DFNB1-related hearing loss could be done first, as we have done in our casuistic.

Despite the genetic heterogeneity, some genes showed greater contribution to cases of autosomal recessive deafness in our study: GJB2, CDH23 (10.34%), MYO15A (7%), OTOF (7%), and USH2A (7%) (Figure 1B). It is important to note that the frequency of pathogenic variants in GJB2 cannot be precisely calculated because it would be clearly underestimated, since many families with presumptive ARHL had been screened for variants in GJB2 before this study. In fact, nearly 40 families with DFNB1-related pathogenic variants were known in our diagnostic settings before the sample selection for this study.

We confirmed with our data that the variants classified as “hypomorphic” in GJB2, such as c.101T>C: p.(Met34Thr) and c.109G>A:p.(Val37Ile) were in fact causative of the hearing loss, since no other relevant variants were detected, and the degree of hearing impairment is compatible with the occurrence of these variants.

When comparing data from different studies that investigated variants in known hearing loss genes in large cohorts, especially in probands from pedigrees with presumptive autosomal recessive inheritance, it is interesting to notice that, although there is some variation between different populations, some genes appear as the most frequently altered in more than one study. These genes include MYO15A, MYO7A, SLC26A4, STRC, and USH2A (Bademci et al., 2016; Sloan-Heggen et al., 2016; Yan et al., 2016; Zazo Seco et al., 2017; Morgan et al., 2018). It is important to notice that studies that considered STRC as frequently related to autosomal recessive hearing loss also searched for copy number variations. This analysis did not occur in a systematic way in our sample.

In an investigation performed in the city of Campinas, São Paulo, Brazil, 180 probands had 86 specific variants genotyped in known hearing loss genes through the MassARRAYiPLEX^® platform in 17 different genes. This study indicated that GJB2, SLC26A4, MYO15A, OTOF, and CDH23 showed greater contribution in cases of autosomal recessive deafness (Svidnicki et al., 2015) However, this study screened only previously known variants and NGS, as performed in this study, allowed us to identify ten novel variants and solved additional cases.

In the study of Manzoli et al. (2016), conducted in the city of Monte Santo, Bahia, Brazil, 19 probands were selected for having hearing loss of presumptive autosomal recessive inheritance. They had their genetic material analyzed through next-generation targeted sequencing of 180 hearing loss-related genes. The hearing loss genes with causative variants detected in this group were MYO15A (x10) and CLDN14 (x1). The variant c.4198G>A, p.(Val1400Met) in MYO15A was found in eight probands and was probably inherited from a common founder. This study focused on the importance of the p.(Val1400Met) variant in Northeastern Brazil. Other studies also indicated that, apart from GJB2, pathogenic variants in SLC26A4, MYO15A, OTOF, and CDH23 are a frequent cause of hearing loss in European populations (Hilgert, Smith, Van Camp, 2009).

We did not detect variants in the SLC26A4, probably due to a previous study of this cohort (Nonose et al., 2018) that focused on linkage analysis and Sanger sequencing of this gene among candidate pedigrees.

In the sample here presented, 32 of the 90 probands (36%) had the causative variants identified after NGS.

Next-generation sequencing clearly facilitated genetic testing for hearing loss. Targeted sequencing panels allow the identification of causative variants in known hearing loss genes and have been proven effective (Sloan-Heggen et al., 2016). It usually has better coverage and lower costs when compared to whole exome sequencing. However, custom panels need to be periodically updated due to the discovery of new genes, especially in heterogeneous conditions like hearing loss (Atik et al., 2015). In our study, for example, FGFR3 was not included in the custom panel. Therefore, proband RC-26 had a negative result after next-generation targeted sequencing of 99 genes and the causative variant was detected only after whole exome sequencing. It is possible that patients with syndromic hearing loss who present mild clinical features of a syndrome may have been misclassified as “non-syndromic” in some studies.

A limitation of our study was the fact that not all probands were fully screened for copy-number-variants, which is an interesting aspect to investigate, for instance, with arrays or low-pass-whole genome sequencing (Mazzonetto et al., 2024).

Whole exome sequencing also provides the potential for the discovery of novel hearing loss genes, which means that we can focus on unsolved cases to search for candidate variants in genes that were not previously associated with hearing loss.