Prevalence of TECTA mutation in patients with mid-frequency sensorineural hearing loss

The criteria proposed by the GENDEAF study group published in Hereditary Hearing Loss Homepage [2] were used to categorize hearing-loss levels as follows: mild, 20–40 dB HL; moderate, 41–70 dB HL; severe, 71–95 dB HL; profound, >95 dB HL (better hearing ear, averaged over 0.5, 1, 2, and 4 kHz). The following criteria were used to define hearing loss progression: progressive, deterioration of >15 dB HL on average over the frequencies 0.5, 1, and 2 kHz within a 10-year period [2]. We judged the progression only in cases we could follow up the audiograms for more than 10 years.

The process for selecting patients for TECTA gene mutation testing is illustrated in Fig. 2. First, patients with hearing loss due to GJB2, or mitochondrial m.1555A > G or m.3243A > G, mutations were excluded based on previous test results. Next, patients with bilateral non-syndromic SNHL were selected, followed by exclusion of patients with inner ear malformations by CT or MRI, if performed. CT or MRI was performed in 1153 patients among 1410 bilateral non-syndromic SNHL patients. Next, patients were selected based on the mid-frequency U-shaped criterion proposed by the GENDEAF study group [2]: >15 dB HL difference between the poorest threshold at mid-frequencies (1, 2 kHz) and those at higher (4, 8 kHz) and lower (0.125, 0.25, 0.5 kHz) frequencies. Since many patients with shallow U-shaped audiograms did not meet this requirement, an additional new criterion was developed to include such patients: the poorest thresholds were identified at 0.5, 1, or 2 kHz, and the criterion was that the threshold at 0.5 kHz was worse than those at 0.125 and 0.25 kHz, and the threshold at 2 kHz was worse than those at 4 and 8 kHz; patients who did not meet the U-shaped criterion, but met this criterion for a shallow U-shaped audiogram, were also selected for TECTA analysis.

Genomic DNA samples were extracted from the peripheral blood of patients using a DNA extraction kit Genomix (Biologica, Japan). The methods of screening for GJB2 and the mitochondrial m.1555A > G and m.3243A > G mutations were shown in Additional file 1. For TECTA analysis, the primers described in Additional file 2 were used to amplify all exons of TECTA by PCR [17]. The PCR conditions were as follows: 5 min denaturation at 95 °C; 35 cycles of 95 °C for 1 min, 58 °C for 1 min, and 72 °C for 2 min; followed by 72 °C for 2 min, and ending with a holding period at 4 °C. PCR products were purified and subjected to Sanger sequencing. Analysis was performed using SeqScape ver3.0 Software (Applied Biosystems), by comparison with the NCBI human reference sequence (GRCh37.p13). Frameshift, splice site (splice site within ±2 nucleotides), and nonsense mutations were judged pathogenic. For changes in splice sites within 10 bp of exon-intron boundaries, NNSPLICE (0.9 version) [18] was used to predict the effect on splicing.

The reported TECTA missense variants were judged as pathogenic when the variants co-segregate with the phenotypes in family members, and judged as variants of uncertain significance (VUS) when the variants did not co-segregate with the phenotypes. Novel missense variants were judged as possibly pathogenic if they met all of the following criteria: 1) non-synonymous; 2) minor allele frequency < 1% in all public databases, including 1000 GENOMES [19], NHLBI Exome Sequencing Project (ESP6500) [20], Exome Aggregation Consortium (ExAC) [21], Human Genetic Variation Database (HGVD) [22], and integrative Japanese Genome Variation Database (iJGVD) [23]; 3) high amino acid conservation (>90% in primates and mammals, >50% in vertebrates) among up to 12 primate, 50 mammal, and 38 vertebrate species (Additional file 3) using the UCSC conservation tool [24]; 4) consistency with phenotypes confirmed by hearing tests in family members. Variants that did not satisfy criteria 1)–3) were judged as non-pathogenic. As patients with TECTA mutations often show mild to moderate hearing loss from childhood, individuals may not be aware of their hearing loss; therefore, variants that met criteria 1)–3), but failed criterion 4), were considered VUS. Furthermore, the effect of missense mutation on the function of the TECTA protein was predicted by computer analysis using PolyPhen-2 [25] and Protein Variation Effect Analyzer (PROVEAN v1.1.5) [26].

To evaluate novel possibly pathogenic variants (p.Y1900C and p.S2017P), changes in the TECTA protein structure were predicted by molecular modeling. The tyrosine at residue 1900 of the TECTA protein is located in the N-terminal part of the ZP domain (ZP-N), while the serine at residue 2017 is located in the C-terminus of the ZP domain (ZP-C). As the structure of the TECTA ZP domain is unknown, the human ZP domain structure was predicted using the crystal structure of chicken ZP3 (PDB ID: 3NK3, chain A), which has a high amino acid homology to human ZP domains, as a template, using SWISS-MODEL [27‐29]. The TECTA protein can form dimers through binding sites in the ZP-N domain, and the amino acid sequence of this domain is important for dimer formation [30, 31]. Therefore, as the p.Y1900C mutation may affect ZP-N dimerization, additional analysis of the effect of this variant on dimerization was performed. Since the dimeric structure of the ZP3 protein is unknown, a predicted structure of the human TECTA dimer was generated, using the predicted human ZP domain structure and the dimeric human uromodulin protein structure (PDB ID: 4WRN, chain A, chain B), which has high homology with the human ZP domain, as a template, using the display software, UCSF Chimera 1.11 [30, 32], and the influence of p.Y1900C mutation in this context was predicted.

As shown in the patient selection and exclusion diagram (Fig. 2), 1215 patients received audiogram selection from 1896 probands. There were 21 patients with U-shaped audiograms meeting the GENDEAF criterion in one or both ears (1.7% of patients who received U-shaped audiogram selection) and 46 with shallow U-shaped audiograms (3.9% of patients who were excluded by U-shaped audiogram selection). Family histories of these 67 patients with U-shaped or shallow U-shaped audiograms were compatible with AD in 23 patients (34.3%), AR in 10 patients (14.9%), and sporadic in 34 patients (50.7%). Gene analysis identified pathogenic and possibly pathogenic variants in 4 patients (6.0%) of the 67 patients. In patients with U-shaped audiograms, none had pathogenic or possibly pathogenic variants, one patient (4.8%) had VUS, and 20 patients (95.2%) had non-pathogenic variants or no variants (Additional file 4). In patients with shallow U-shaped audiograms, two patients (4.3%) had pathogenic variants, the other two patients (4.3%) had possibly pathogenic variants, five patients (10.9%) had VUS, and 37 patients (80.4%) had non-pathogenic variants or no variants. Pathogenic or possibly pathogenic variants were identified in 3 of 23 families (13.0%) which have the family histories compatible with AD and 1 of 44 families (2.3%) which have the family histories compatible with sporadic or AR.

A list of pathogenic variants, possibly pathogenic variants, and VUS found in this study is presented in Table 1, and a list of non-pathogenic variants is provided in Additional file 5. The DNA chromatograms of pathogenic variants, possibly pathogenic variants, and VUS are provided in Additional file 6. Three pathogenic variants (p.R1890C, p.C633*, and c.65-3C > T) have previously been reported. Two novel variants, p.Y1900C and p.S2017P, identified in this study met criteria 1)–4) (See Methods), and were predicted to have high pathogenicity by PolyPhen-2 and PROVEAN. Thus, they were considered possibly pathogenic variants.

Modeling structure of p.Y1900C in TECTA is shown in Fig. 3. The tyrosine at residue 1900 (Y1900) is located in the ZP-N domain, which is involved in dimer formation via hydrogen bonds (arrows in Fig. 3a). Y1900 is predicted to form a hydrogen bond with the lysine at residue 1807 (K1807) (arrowheads in Fig. 3a), to maintain the structure of the dimerization site (Fig. 3a, b). However, in the model of the protein containing the p.Y1900C mutation, this hydrogen bond was absent (Fig. 3c). Hence, the p.Y1900C mutation is likely to disrupt dimerization of the TECTA protein via the ZP-N domain, and result in structural abnormality of the tectorial membrane.

The effect of the p.S2017P mutation, located in the ZP-C domain, was not determined, as no significant alteration in protein structure was predicted after introduction of this mutation in the model; moreover, it is unknown whether this region of the protein is involved in interactions with other molecules.

Figure 4 shows the pedigrees of four families in which pathogenic or possibly pathogenic variants were identified. Genotyping of the proband and family members, as well as phenotypic characteristics, indicated an AD inheritance pattern in Family 1, Family 3, and Family 4 and an AR inheritance pattern in Family 2. In Family 3 and Family 4, novel possibly pathogenic variants with an AD inheritance pattern were identified. In Family 4, the mother of the proband (I-2) had the possibly pathogenic variant and mild hearing loss, with a flat audiogram, in contrast to the moderate hearing loss and U-shaped audiogram of the proband (II-1). Given this discrepancy, the potential pathogenicity of this variant should be considered with caution.

Clinical data of the four families with pathogenic or possibly pathogenic variants are presented in Table 2. The onset of hearing loss was in the first decade of life in all families. The level of hearing loss was mild to moderate in all cases. All of the probands had shallow U-shaped audiograms. No progression was observed in any case during the follow-up period according to the criteria used in the present study. In Family 1, the most recent audiogram of the father of the proband, who had the mutation, was flat, and the hearing of the proband deteriorated at both high and low frequencies with age, resulting in an audiogram similar to that of the father (Fig. 4). The word recognition score (WRS) was 70–90% in all cases (Additional file 7).

Figure 5 and Table 3 shows the pedigrees, audiograms, and clinical data of Family 5–10 in which VUS were identified. In Family 7, Family 8 and Family 10, the reported TECTA mutations did not show typical family history of AD or did not segregate in family members who were not tested for hearing loss, and/or TECTA mutations. It is possible that this may be explained by mild phenotype, de novo mutation, or phenocopy. In Family 7, the parents of the proband (II-3 and II-4) were not known to have hearing loss. Therefore, the mutation may be de novo in the proband. Alternatively, the father (II-3) may carry the mutation and have mild hearing loss that had not been noted. In Family 8, the father (II-1) carried the mutation but did not present with hearing loss. In addition, the allele frequency of this variant in Japanese is high (0.008) for the cause of DFNA8/12. Although the possibility of the low penetrance of this variant cannot be denied, these data suggest that this variant is unlikely to be pathogenic. In Family 10, the mother of the proband (II-2) presented with hearing loss, but did not have the mutation. As the peak frequency of the U-shaped audiogram and the onset of the hearing loss were quite different in the mother from those in the proband (III-2), the cause of hearing loss may be different in the two individuals (suggesting phenocopy), and the mutation in the proband could be de novo or inherited from the father (II-1), who may have the mutation and mild hearing loss that had not been noted. In Family 5, Family 6, and Family 9, novel VUS were identified. In these families, similar to Family 7 and Family 10, individuals reporting normal hearing who had not had hearing tests (Family 5, II-4; Family 6, I-1 and I-2; Family 9, II-1) may have had mild hearing loss. In Family 5, the lack of TECTA mutation in the father (of proband, II-1) with hearing loss, suggests that his hearing loss may be due to other causes, such as presbycusis. In Family 6, the mutation in the proband could be de novo.