Whole-Exome Sequencing Targeting a Gene Panel for Sensorineural Hearing Loss: The First Portuguese Cohort Study Free

Over 5% of the world’s population suffer from disabling hearing loss. In the United States, 2–3 out of every 1,000 children are born with a detectable level of hearing loss in one or both ears [Centers for Disease Control and Prevention, 2010], and approximately 15% of the adults aged 18 years and over report some trouble hearing [Blackwell et al., 2014]. Up to 60% of the congenital/early-onset hearing loss is due to genetic causes [Marazita et al., 1993].

Sensorineural hearing loss (SNHL) without other features (nonsyndromic) accounts for 70% of the hereditary cases; the other 30% are syndromic, with more than 400 syndromes described. Also, more than a hundred SNHL-related genes have been identified to date [Alford et al., 2014]. GJB2 and GJB6 genes account for a significant proportion of cases, although about 70–90% remain undiagnosed after GJB2/GJB6 testing [Pandya et al., 2003; Putcha et al., 2007; Burke et al., 2016].

By addressing the challenge of genetic heterogeneity, next-generation sequencing technologies revolutionized the way SNHL molecular diagnosis is approached in clinical practice and are now a standard of care [Richards et al., 2015]. These technologies allow either the sequencing of the entire genome or parts thereof, the latter being a more practical and affordable alternative [Shearer and Smith, 2015]. Whole-exome sequencing (WES) involves sequencing of all exons of all genes, where about 85% of known disease-causing variants occur [Majewski et al., 2011]. More selective alternatives include the clinical exome sequencing, which covers genes that have a known clinical association with human disease, and the targeted gene panels (GP), that cover a subset of genes deemed to be relevant to the phenotype. The GP approach minimizes the secondary findings in genes unrelated to the disease under investigation [Saudi Mendeliome Group, 2015], yet more inclusive or non-targeted approaches allow the identification of novel causative variants.

The aim of this study is to evaluate the diagnostic yield of WES targeting a GP for SNHL in a group of 71 Portuguese index patients with presumed hereditary SNHL, selected on the basis of the absence of causative variants in the GJB2 gene or causative deletions/duplications in the GJB6 gene.

A retrospective cohort study was performed in 71 probands with SNHL followed at the Genetics Clinic of “Hereditary Hearing Loss” of Centro Hospitalar Universitário de São João, who underwent WES and targeted analysis of a GP in the period of June 2011 to December 2019. All individuals had been previously screened by Sanger sequencing for mutations in GJB2 and/or GJB6 genes, with no alterations found. Non-genetic causes of SNHL, including noise exposure, were investigated through medical history and examination. None had suspicious findings, except for 1 patient who had a history of neonatal hypoxia. For each patient, the age of onset of SNHL, personal and familial medical history were ascertained. Patients with either nonsyndromic or syndromic SNHL were included. Audiological evaluation was performed using auditory brainstem response tests or pure tone audiometry. The severity of SNHL was classified by the pure tone average (PTA) of 500, 1,000, 2,000, and 4,000 Hz thresholds, according to the GenDeaf study group recommendations: mild SNHL corresponds to a PTA of 20–40 dB; moderate to 41–70 dB; severe to 71–95 dB; and profound to PTA >95 dB [Mazzoli et al., 2003]. For genetic testing, peripheral blood samples were collected. From all patients, written informed consent for genetic testing was obtained after counselling by a clinical geneticist.

The exomes of 71 individuals with SNHL were sequenced. Then, targeted analysis of WES data was performed for a panel of 158 genes that have been related to nonsyndromic and syndromic SNHL (listed with the expected coverage rates in online suppl. Table S1; see www.karger.com/doi/10.1159/000523840 for all online suppl. material). Genetic tests and reports were performed by IPATIMUP (Porto, Portugal). Exome capture was performed using the Ion AmpliSeq^TM Exome RDY Kit (Thermo Fisher Scientific) and sequenced using the Ion semiconductor sequencing technology (Ion Torrent S5XL or S5 platform; Thermo Fisher Scientific). High quality sequence reads were mapped to the human genome reference (UCSC hg19) and annotated with the IonReporter server. Variants were classified using several databases containing functional (Ensembl), populational (GnomAD, ExAC), and disease-related (ClinVar, Deafness Variation Database [Azaiez et al., 2018]) information, as well as computational methods for prediction of the impact caused by the variant on the structure and function of the protein (PolyPhen-2 [Adzhubei et al., 2010], MutationTaster [Schwarz et al., 2010], SIFT [Ng and Henikoff, 2003], and Human Splicing Finder [Desmet et al., 2009]). Then, sequence variants were categorised according to the current guidelines from the American College of Medical Genetics and Genomics in 5 classes: benign, likely benign, uncertain significance, likely pathogenic, and pathogenic [Richards et al., 2015; Oza et al., 2018]. Every pathogenic or probably pathogenic variant was independently confirmed by Sanger sequencing (the primer sequences used in this process are listed in online suppl. Table S2). Variants were described following the standard nomenclature recommended by the Human Genome Variation Society.

Identified variants were interpreted considering the patient’s clinical evaluation to assess causality. Variant classification and its predicted functional impact, segregation analysis, and matching with the gene-associated phenotypes known from the literature were taken into account. Data resources included ClinVar, OMIM, PubMed, LOVD, and Deafness Variation Database [Azaiez et al., 2018]. A genetic diagnosis was assumed in patients with pathogenic/likely pathogenic variants matching the phenotype known from literature, and these variants were defined as “causative.” We also defined a subset of patients harbouring variants of uncertain significance that, whether for the matching phenotype, the compatible segregation analysis, or the prediction of functional impact, were considered to be “presumptive causative variants.” The remaining patients were considered undiagnosed.

The statistical analysis was performed using GraphPad® software. Confidence intervals (CI) were calculated using the modified Wald method [Agresti and Coull, 1998]. Differences in diagnostic yield according to the age of SNHL onset were tested by the χ² and Fisher’s exact tests, with p values ≤0.05 being considered as statistically significant.

In total, the WES data of 71 probands were analysed for a panel of SNHL-related genes. Sixty-six patients were Portuguese, 1 was Brazilian, 3 were Romani, and 1 had a parent of Guinea-Bissau origin. SNHL was congenital or had pre-lingual onset in 21 individuals (29.6%). Regarding the other cases, 36 (51%) reported onset during the first or second decades, 9 (12.7%) during the third or fourth decades, and 3 (4.2%) after the age of 40. For the remaining 2 patients (2.8%), information was missing. In the present cohort, of the 158 genes included in the present GP (online suppl. Table S1), 56 did not show any variant which was pathogenic, likely pathogenic or of uncertain significance. Altered genes had an average of 2.3 variants.

Of 71 probands, pathogenic/likely pathogenic variants capable of explaining the SNHL phenotype were found in 15 (21.1%, 95% CI 13.1–32.1%). The patients’ phenotypes and genotypes are summarized in Table 1, as well as information on whether the variant segregation with SNHL was observed or not.

Of these 15 probands, 9 presented heterozygous causative variants in genes associated with dominant SNHL: EYA4 (2 patients), MYO3A (2 patients), TECTA, EDNRB, KCNQ4, MYO6, and POU4F3. For 5 (A3, A5, A7, A8, and A9), genetic testing of the available family members was not enough for a conclusive segregation analysis of the variant, but the family history was consistent with a dominant trait; also, for proband A7, incomplete penetrance is suggested (Fig. 1). Within each of the families A2 and A4, 2 affected individuals carried the mutation (mother and daughter in A2; 2 siblings in A4), but information about previous generations was unavailable. In individuals A1, A2, and A8, cardiomyopathy has been excluded.

Four probands had homozygous or compound heterozygous mutations in recessive SNHL-related genes: SLC26A4 (2 patients), OTOGL, and TMPRSS3. Patient A11 had Mondini dysplasia, revealed by computed tomography of the temporal bone, and increased serum levels of free tri-iodothyronine (T3). Proband A12 had undergone thyroidectomy at age 28 because of follicular hyperplasia; also, he had no relevant family history.

Finally, 2 probands harboured a variant in an X-linked gene (COL4A5).

Six probands (8.5%, 95% CI 3.6–17.6%) had presumptive causative variants (Table 2): 1 proband was compound heterozygous for variants in a recessive gene (MYO15A), 2 were homozygous for variants in recessive genes (TMIE and TBC1D24), 1 digenic (PCDH15 and CDH23), 1 hemizygous for an X-linked gene (SMPX), and another one heterozygous for a GJB3 variant. Genograms are shown in Figure 1. Overall, the diagnostic yield was 29.6% (95% CI 20.2–41.1%).

The clinical presentation of both proband B1 and her hearing-impaired brother match the phenotype associated with the MYO15A gene [Friedman et al., 1995]. No other family member had SNHL. Computational predictive methods indicate that both variants might have a functional impact, underscoring their possible pathogenic role.

In respect to the TBC1D24 variant (proband B2), computational predictive methods are not in agreement regarding the functional impact of this genetic change. Besides the proband’s sister, family history of SNHL was negative.

Regarding patient B3, computational methods predicted a relevant functional impact for the c.3178C>T, p.(Arg1060Trp) variant in the CDH23 gene, whereas conflicting predictions of functional impact were obtained for the c.3832C>A, p.(Arg1278Ser) variant in the PCDH15 gene. Family history of SNHL or consanguinity was absent, which is compatible with the digenic recessive trait. However, progenitors DNA samples were unavailable, making it impossible to validate the findings. At the age of 30 (her last follow-up appointment), the patient had never had vestibular or ocular symptoms (electroretinography and vestibular tests were not performed), but these features can manifest later in life [Zheng et al., 2005].

The SMPX variant identified in proband B4 was predicted to have a relevant functional impact, and the phenotype was compatible with previous descriptions for this gene [Del Castillo et al., 1996]. No family member was tested, but family history was compatible with the X-linked trait.

The variant c.362-1G>A in the TMIE gene (proband B5) has not been previously described. Its location in a canonical region of splicing makes it reasonable to assume that disruption of this process will occur. Both the patient and her sister had congenital SNHL, in consistency with previous findings for this gene [Sirmaci et al., 2009]. Their parents were both hearing, in keeping with a possible recessive trait. The validation of the variant’s pathogenicity was not possible since none of the family members was genetically tested.

Proband B6 presented SNHL and multifocal demyelinating motor neuropathy, which is in agreement with previous findings for the GJB3 gene [López-Bigas et al., 2001]. To our knowledge, no other family member had history of SNHL or neuropathy, but 2 sisters had skin problems.

In 50 out of 71 cases (70.4%, 95% CI 58.9–79.8%), a genetic diagnosis was not achieved. Two cases did not display any variant. In the remaining 48, variants of uncertain significance were detected. In 16 individuals, single variants were identified in genes known to underlie autosomal recessive SNHL and, when 2 or more genes were affected, no reference to digenism was found in the literature. Non-matching phenotypes led to the refutation of a molecular diagnosis in 3 other cases. In another one, the phenotype-genotype matching was inconclusive. This was a 2-year-old boy carrying the variant c.190T>C, p.(Phe64Leu) in the COL4A5 gene, which was predicted to have a functional impact by computational methods. However, the absence of typical Alport phenotype was difficult to valorise due to the proband’s age [Kashtan, 1999]. Also, the patient had no family history of SNHL, and his mother, heterozygous for the same variant, did not perform evaluation of kidney function. In another 9 cases, the familial study revealed that the identified variants were not segregating with SNHL. For the remaining 19 probands (26.8%, 95% CI 17.8–38.1%), segregation analysis of the variants could not be performed due to the unavailability of family members’ DNA samples. An adequate segregation analysis could have clarified the significance of the variants found.

In summary, in at least 30 out of 71 patients (42.3%, 95% CI 31.5–53.9%), no potentially causative variants were identified, with the other 20 cases (28.2%, 95% CI 19.0–39.6%) being inconclusive – 19 due to the lack of segregation analysis and 1 due to uncertain phenotype-genotype matching.

Causative variants were identified in 21.1% (95% CI 12.3–33.4%) and 25.0% (95% CI 8.3–53.9%) of probands with SNHL onset before and after 20 years old, respectively, with this difference being not statistically significant (p = 0.72). For a cut-off at the age of 40 years, causative variants were identified in 21.2% (95% CI 13.0–32.6%) and 33.3% (95% CI 5.6–79.8%) of cases with onset before and after that age, respectively, again with no statistically significant difference (p = 0.53). Differences are still not statistically significant when cases with presumptive causative variants are considered for the diagnostic rates (p = 0.49 for a cut-off of 20 years; p = 1.00 for a cut-off of 40 years). In online supplementary Table S3, the results of genetic diagnosis by age of SNHL onset are presented, with 4 categories of results being considered: cases with causative variants, cases with presumptive causative variants, undiagnosed cases, and inconclusive cases (due to the lack of segregation analysis or to uncertain phenotype-genotype matching).

Among the 71 probands, we found 9 heterozygotes (12.7%, 95% CI 6.6–22.6%) for pathogenic/likely pathogenic variants in recessive SNHL-related genes: 5 patients had variants in genes linked to nonsyndromic hearing loss (3 in GJB2 and 2 in TMPRSS3); and each of the other 4 patients had one variant related to Pendred (SLC26A4 gene), Alström (ALMS1 gene), Bartter (BSND gene), and Usher (ADGRV1) syndromes.

In this study, we evaluated the diagnostic yield of WES targeting a GP for SNHL in a Portuguese cohort of patients. To our knowledge, this is the first study on the application of next-generation sequencing technologies in SNHL based on a Portuguese cohort. Our results contribute to characterize the mutational spectrum in the Portuguese population and further corroborate the genetic heterogeneity of SNHL, since causative variants found in 15 patients involved 11 different genes. The genes most frequently affected were EYA4, MYO3A, SLC26A4, and COL4A5, each of them being identified in 2 patients.

An interesting finding of our study was the identification of a MYO3A-related case of autosomal dominant nonsyndromic SNHL. In fact, this gene was exclusively associated with autosomal recessive nonsyndromic SNHL for many years, and only recently it was associated with autosomal dominant SNHL [Grati et al., 2016; Dantas et al., 2018; Doll et al., 2020]. Also, both probands reported here had the c.2090T>G, p.(Leu697Trp) MYO3A variant, which was previously identified in 2 large unrelated Brazilian families with autosomal dominant nonsyndromic SNHL [Dantas et al., 2018]. Interestingly, one of our probands is of Brazilian origin.

Similarly, the association of variants in the EDNRB gene with Waardenburg syndrome type 2 (proband A5) was only recently uncovered, with the suggestion that they may account for 5–6% of the Waardenburg syndrome type 2 cases [Issa et al., 2017], thereby highlighting their relevance.

We also identified a likely pathogenic variant in the TECTA gene, c.2262dup, p.(Asn755Lysfs*104), which was not previously described. It generates a premature stop codon, being a potential loss-of-function variant, a hypothesis supported by MutationTaster scores. Surprisingly, segregation analysis revealed that the proband inherited the TECTA variant from her mother who had normal hearing function, which suggests that incomplete penetrance might account for this finding. In fact, clinical penetrance may vary with the variant type, as is well known, for example, for variants in COL3A1 and BMPR2 genes, associated with Ehlers-Danlos syndrome and heritable pulmonary arterial hypertension, respectively [Cooper et al., 2013]. Notwithstanding, incomplete penetrance phenomenon can be explained by different mechanisms [Cooper et al., 2013]. Obviously, the possibility that the c.2262dup, p.(Asn755Lysfs*104) TECTA variant is not the cause of SNHL in this family cannot be ruled out.

Concerning the TBC1D24 gene, as far as we know, this is the first study describing a family with homozygosity for the variant c.641G>A, p.(Arg214His). In previous studies, this variant was reported in compound heterozygosity with 2 other likely pathogenic variants [Bakhchane et al., 2015; Tona et al., 2020]. Some authors have even hypothesized that it could act as a hypomorphic variant, leading to abnormal hearing when combined with another severe recessive variant [Bakhchane et al., 2015]. Clinical interpretations on ClinVar are conflicting, varying from benign to likely pathogenic (accessed 2021 November 15), which highlights the need for further data to clarify its meaning. Our results may favour the possibility that this variant is pathogenic.

We also report on a patient with a probable CDH23/PCDH15 digenic form of Usher, albeit without vestibular or ocular symptoms at the age of 30. In fact, cases with balance problems emerging only at the age of 29 and nocturnal blindness manifesting only at 40 years have been reported [Zheng et al., 2005]. Furthermore, there have been several reports of Usher syndrome diagnosis in patients with apparent nonsyndromic hearing loss [Wei et al., 2012; Shearer et al., 2013].

Regarding the GJB3-related case, it must be noticed that the combination of SNHL with demyelinating neuropathy as a result of the c.196_198del, p.(Asp66del) variant has been previously described, with a considerable interindividual variability of SNHL severity and symmetry [López-Bigas et al., 2001]. However, hearing loss of fluctuant severity has never been reported. Serial audiological studies would allow us to clarify this subjective presentation, but audiological assessment results were not available. Elucidation of the skin problems presented by the proband’s sisters would also be valuable, since some GJB3 variants are known to underlie erythrokeratodermia variabilis phenotype [Richard et al., 1998, 2000].

In the present cohort, the targeted analysis of WES data for a panel of SNHL-related genes yielded a molecular diagnosis in 21.1% of the probands (or 29.6%, if cases with presumptive causative variants are accounted). Diagnostic rates between 10 and 83% were reported in a previous review. The studies under review varied considerably in terms of the number of hearing loss genes targeted for sequencing (from 15 to 246), the number of individuals per study (from 6 to 125), the predominant inheritance mode of hearing loss, the execution or not of a pre-screening for common deafness mutations, among other variables [Shearer and Smith, 2015]. In our study, it is worth mentioning the great proportion of probands (27%) for whom the unavailability of DNA from family members precluded a proper segregation analysis of variants of uncertain significance. Segregation analysis would likely uncover presumptive causative variants in some cases. Since diagnosis with pathogenic/likely pathogenic variants was exempted from segregation analysis for many cases, exclusion of those 27% of patients would create a bias, possibly overestimating the diagnostic yield. Thus, we decided to keep them in the study, privileging an intention-to-diagnosis approach which overall accurately reflects the realities of clinical practice.

In at least 42.3% of our patients, the targeted analysis of WES data did not allow a genetic diagnosis. A part of these cases might be explained by variants that were not identified due to insufficient enrichment or coverage. Although most genes had an expected coverage rate equal or close to 100%, 12 genes had coverage rates lower than 95%. The STRC gene, whose variants have been identified as a frequent cause of SNHL [Mehta et al., 2016; Sloan-Heggen et al., 2016], had the lowest coverage rate (42.1%). It is also important to consider that the data presented do not include the study of CNVs, which accounted for 1.5–7.3% of the cases in previous studies [Zaco Seco et al., 2017].

Another part of the undiagnosed cases may be attributable to variants in genes not included in the GP used. In fact, panels systematically miss alterations in genes not yet associated with the disorder. In these patients, reanalysis of their WES data may be considered, either for an updated GP for SNHL – including new genes that, in the meantime, have been associated with this disorder – or even in a non-targeted approach. The design of more inclusive GPs is also an option. Two research groups have projected and assessed GPs covering almost 3,000 genes known to cause mendelian disorders, with diagnostic yields of 54 and 63.6% (GJB2 mutations excluded) [Atik et al., 2015; Saudi Mendeliome Group, 2015]. These enlarged GPs, usually referred to as clinical exome, preserve the GP advantage of avoiding secondary findings in genes not known to be linked to human diseases, while broadening the spectrum of analysed genes and thus, improving detection rates.

Finally, in undiagnosed patients, especially those with late-onset SNHL, non-genetic causes must be reconsidered. Accordingly, some studies demonstrate that the diagnostic yield strongly declines with the increase in age of onset [Shearer et al., 2013; Zaco Seco et al., 2017]. However, an age-of-onset-related trend in diagnostic yield was not evident in the present study, which may be at least partly justified by the fact that all probands had been previously screened for GJB2 and GJB6 causative variants. Other studies included non-screened probands, with GJB2 variants accounting for the diagnostic yield in the congenital and first decade onset SNHL [Shearer et al., 2013; Zaco Seco et al., 2017]. The results of the present study suggest that genetic testing of patients with late-onset SNHL, whose clinical and family history hints at an underlying genetic cause, may still be important. Although this may be true, it should be noted that only 3 individuals with SNHL onset after 40 years old were included.

To sum up, in the present study, we evaluated the diagnostic yield of WES targeting a GP for SNHL in a Portuguese sample, unveiling the spectrum of SNHL variants in this population. We also brought new insights into the contribution of MYO3A, TECTA, EDNRB, TBC1D24, CDH23, and PCDH15 genes to SNHL and reported some novel variants, so the present research may contribute to the understanding of the genetic basis of SNHL. However, a large proportion of patients remained without a diagnosis after GP-targeted analysis. Some of these patients may benefit from more inclusive or non-targeted approaches that allow the identification of novel variants. Then, interpreting the huge amount of genetic information provided and managing incidental findings may be the biggest challenge to face.

This study was based on a clinical etiologic investigation. All phases of the study complied with the Ethical Principles for Medical Research Involving Human Subjects expressed in the Declaration of Helsinki (World Medical Association, 2013). The study was approved by the Centro Hospitalar Universitário de São João ethics committee, approval number 127-19, and a signed informed consent according to the Declaration of Helsinki was obtained from all participants.

The authors declare that no competing interests exist.

The present study was not specifically funded.

Cláudia Sousa Reis: data collection, data analysis and interpretation, scientific information collection, manuscript preparation. Susana Fernandes: data collection, data analysis and interpretation, manuscript preparation, critical revision of the manuscript. Sofia Quental and Sérgio Castedo: manuscript preparation, critical revision of the manuscript. Carla Pinto Moura: conceptualization, data analysis and interpretation, manuscript preparation, final approval of the manuscript.

Relevant data are included in the present article and its supplementary material files. The authors consider submitting the exome data to a exome dataset; for that purpose, permission of patients and relatives is under request. Further enquiries can be directed to the corresponding author.