The natural history and genotype-phenotype correlations of TMPRSS3 hearing loss: an international, multi-center, cohort analysis

Abstract

TMPRSS3-related hearing loss presents challenges in correlating genotypic variants with clinical phenotypes due to the small sample sizes of previous studies. We conducted a cross-sectional genomics study coupled with retrospective clinical phenotype analysis on 127 individuals. These individuals were from 16 academic medical centers across 6 countries. Key findings revealed 47 unique TMPRSS3 variants with significant differences in hearing thresholds between those with missense variants versus those with loss-of-function genotypes. The hearing loss progression rate for the DFNB8 subtype was 0.3 dB/year. Post-cochlear implantation, an average word recognition score of 76% was observed. Of the 51 individuals with two missense variants, 10 had DFNB10 with profound hearing loss. These 10 all had at least one of 4 TMPRSS3 variants predicted by computational modeling to be damaging to TMPRSS3 structure and function. To our knowledge, this is the largest study of TMPRSS3 genotype-phenotype correlations. We find significant differences in hearing thresholds, hearing loss progression, and age of presentation, by TMPRSS3 genotype and protein domain affected. Most individuals with TMPRSS3 variants perform well on speech recognition tests after cochlear implant, however increased age at implant is associated with worse outcomes. These findings provide insight for genetic counseling and the on-going design of novel therapeutic approaches.

Fig. 1

A Participants included in each analysis. We received data about 148 individuals in 135 families from 16 centers in 6 countries. 21 had variants in other known HL-related genes, had no TMPRSS3 genotype reported, or were heterozygous for TMPRSS3 mutations and were excluded from analysis. 78 participants had audiometric testing reported. 74 participants had cochlear implants and 39 of these reported the results of speech perception testing after cochlear implantation. B Locations of the TMPRSS3 variants in the data set. There were 48 unique TMPRSS3 variants in the data set. Variants included nonsense, missense, splice site, and indel frameshift variants. Variants in red were associated with DFNB10 and green with DFNB8. Variants in blue were associated with both DFNB8 and DFNB10. The black ‘X’ indicates loss of function variants and the black circle indicates missense

Fig. 2

Audiometric thresholds by DFNB8/10 and by genotypic categories. Audiograms were collected for 78 individuals. Grey lines are a family (when a family had more than one audiogram, one was chosen at random for A, B and D–F; C and G are representative of all audiograms [see methods]), blue line is the mean at each frequency, and the blue shading is the 95% CI for DFNB8 (A), DFNB10 (B), and the 3 genotypic categories, M/M (D), M/LoF (E), LoF/LoF (F). C The mean thresholds of DFNB8 and DFNB10 differ at low (250 + 500 Hz) and middle (1 + 2 kHz) frequencies (student’s t test; error, 95% CI). G The mean thresholds of the 3 genotypes at low, middle, and high frequencies (one-way ANOVA with Tukey’s post-hoc comparison). ***p < 0.0005; **p < 0.005; *p < 0.05, ns = not significant

Fig. 3

Hearing loss progression for individuals with DFNB8 genotype groups and protein domains. Pure tone averages (PTA) and hearing thresholds were plotted by age at test for individuals with DFNB8. A Changes in PTA by age were not significant (0.2db/year; p = 0.07 linear regression; dashed lines are 95% CI). B Hearing thresholds by frequency for each genotype group involved in DFNB8. Significant progression was seen for M/M genotypes at 1000 Hz, 2000 Hz, 4000 Hz, and 8000 Hz. C Progression was seen for individuals with two variants in the SRCR domain at 750–3000 Hz

Fig. 4

Speech perception score after cochlear implantation by DFNB8/10 and by genotypic categories. Word recognitions scores (WRS) were reported for 36 individuals. A The mean is 76% (95% CI 70–82%). Analysis by genotype (B) and DFNB8/10 (C) did not reveal any associations with WRS. D Worse WRS were associated with increased age at implantation (Dashed lines represent the 95% CI; Slope − 0.3190, R² 0.1733, p = 0.0068, linear regression)

Fig. 5

Protein modeling shows deleterious affects of 4 missense mutations. A Human TMPRSS3 model predicted by AlphaFold2, positioned in a lipid bilayer generated by Charmm-gui. The 4 domains of the protein are highlighted and labeled. B Overlap of missense variants that lead to severe hearing loss or M/WT with clinical hearing loss. The consistent overlap suggests these 4 variants are more severe in their effects. C Zoomed-in structures showing the differences in interactions due to these 4 missense mutations i) R106C mutation showing the potential disulfide bond formation in yellow dotted lines between C106 and C92. ii) V116M showing the clashing of the mutant Methionine with N114. iii) A138E shows the insertion of the large negatively charged Aspartate residue, inducing steric clashes with nearby amino acids W133, K134, and S153. iv) A306T showing two extra backbone hydrogen bonds formed by the mutant methionine residue with A255 and A256