Genotype-Phenotype Correlations in TMPRSS3 (DFNB10/DFNB8) with Emphasis on Natural History

Abstract

Background

Mutations in TMPRSS3 are an important cause of autosomal recessive non-syndromic hearing loss. The hearing loss associated with mutations in TMPRSS3 is characterized by phenotypic heterogeneity, ranging from mild to profound hearing loss, and is generally progressive. Clinical presentation and natural history of TMPRSS3 mutations vary significantly based on the location and type of mutation in the gene. Understanding these genotype-phenotype relationships and associated natural disease histories is necessary for the successful development and application of gene-based therapies and precision medicine approaches to DFNB8/10. The heterogeneous presentation of TMPRSS3-associated disease makes it difficult to identify patients clinically. As the body of literature on TMPRSS3-associated deafness grows, there is need for better categorization of the hearing phenotypes associated with specific mutations in the gene.

Summary

In this review, we summarize TMPRSS3 genotype-phenotype relationships including a thorough description of the natural history of patients with TMPRSS3-associated hearing loss to lay the groundwork for the future of TMPRSS3 treatment using molecular therapy.

Key Messages

TMPRSS3 mutation is a significant cause of genetic hearing loss. All patients with TMPRSS3 mutation display severe-to-profound prelingual (DFNB10) or a postlingual (DFNB8) progressive sensorineural hearing loss. Importantly, TMPRSS3 mutations have not been associated with middle ear or vestibular deficits. The c.916G>A (p.Ala306Thr) missense mutation is the most frequently reported mutation across populations and should be further explored as a target for molecular therapy.

Introduction

Hearing loss (HL) profoundly impacts cognitive, psychosocial, and educational development. Approximately one in 1,000 newborns are deaf, one in 300 children are affected with congenital HL of a lesser degree, and an additional one in 1,000 become profoundly hearing impaired before adulthood [Mason and Herrmann, 1998; Parving, 1999]. It is estimated that two-thirds of prelingual-onset sensorineural hearing loss cases in developed countries have a genetic etiology, of which non-syndromic hearing loss (NSHL) accounts for 70%. Autosomal recessive non-syndromic hearing loss (ARNSHL) is the most common form of NSHL – comprising approximately 80% of cases – with around 20% of cases having autosomal dominant inheritance and a small remaining fraction having either X-linked or mitochondrial inheritance [Van Camp et al., 1997; Smith et al., 2005; Shearer et al., 2010].

The genetic underpinnings of ARNSHL are highly complex. To date, 152 genes and over 8,000 different deafness-causing mutations have been identified [Azaiez et al., 2018]. Genetic deafness is characterized by genetic and phenotypic heterogeneity: different mutations in a single gene can cause both recessive and dominant and syndromic and non-syndromic HL [Prezant et al., 1993]. In addition, due to incomplete penetrance and variable expressivity, identical mutations within a single gene can result in multiple, often markedly different phenotypes. Further, there is a large degree of pleiotropy among deafness genes; mutations in broad subsets of genes confer similar HL phenotypes that may be clinically indistinguishable despite arising from distinct genetic mutations [Magrinelli et al., 2021]. Mechanisms underlying the heterogeneity of HL phenotypes are not fully understood. However, they are thought to involve complex genetic factors (varying mutation types, dynamic mutations, somatic mosaicism, intragenic intra- and inter-allelic interactions, modifiers, epistatic genes, and mitochondrial heteroplasmy), epigenetic factors, and gene-environmental interplay [Magrinelli et al., 2021].

Defining the molecular underpinnings of the complex and delicate human auditory system requires a thorough description of the pathogenic phenotypes associated with the participating genes. One such gene, transmembrane protease serine 3 (TMPRSS3, OMIM 605511) – located at chromosome 21q22 – provides an excellent example of both the challenges and potential benefits associated with untangling the molecular basis of ARNSHL.

The genetic contribution of TMPRSS3 to ARNSHL varies with ethnicity, but mutations in TMPRSS3 are the eighth most common cause of genetic deafness [Shearer et al., 2019]. Another study identified TMPRSS3 as the most common cause of genetic HL in adult cochlear implant (CI) users [Shearer et al., 2017]. Further investigation in other populations and broader implementation of genetic screening is required to evaluate the full extent of TMPRSS3-induced HL.

Mutations in TMPRSS3 have been associated with both prelingual-onset ARNSHL (DFNB10, OMIM 605316 [Bonne-Tamir et al., 1996]) and postlingual-onset NSHL (DFNB8, OMIM 601072) [Veske et al., 1996]. There is a significant difference between the clinical presentation of these two phenotypes, and with nearly 100 deafness-causing mutations in TMPRSS3 identified to date, an understanding of genotype-phenotype correlations is beginning to emerge (Table 1). Given the prevalence of DFNB8/10 and the characterization of associated mutations, TMPRSS3 is an excellent potential target for gene therapy. However, developing these approaches requires a comprehensive understanding of the molecular mechanisms underlying the genotype-phenotype relationship.

Table 1.

Reported phenotypes of TMPRSS3 variants

Variant categories	Domain	DNA change (NM_024022.2)	Amino acid change (NP_076927.1)	Origin	DFNB8/DFNB10*	Phenotype	Published reference (PMID)
Missense variants	–	c.1343T>C	p.Met448Thr	Polish		–	28566687
	Serine protease	c.1306C>G	p.Arg436Gly	Polish		–	28566687
		c.1303C>T	p.Arg435Cys	NA	DFNB10	Severe-to-profound childhood onset HL	26969326
		c.1291C>T	p.Pro431Ser	Italian	DFNB10	Severe-to-profound prelingual HL with further reduction in thresholds at higher frequency	24657061
		c.1283A>G	p.Asn428Ser	NA	DFNB10	Congenital severe-to-profound HL	26969326
		c.1276G>A	p.Ala426Thr	Dutch, Polish	DFNB8	Dutch: mainly high-frequency HL with onset at 13 years of age with deterioration of low frequency after 30 years of age	12920079, 21786053, 28566687
		c.1273T>C	p.Ala425Arg	Pakistani		Childhood onset severe-to-profound HL	21534946
		c.1253C>T	p.Ala418Val	Taiwanese		–	32235586
		c.1250G>A	p.Gly417Glu	Chinese	DFNB10	Severe HL with onset at 2 years of age and flat audiogram; progression to thresholds of 90dB by 8 years of age	28695016
		c.1244T>C	p.Leu415Ser	Chinese	DFNB10	Severe HL with onset at 3 years of age and flat audiogram; progression to downsloping audiogram at 35 years of age	28695016
		c.1219T>C	p.Cys407Arg	Pakistani	DFNB10	Congenital profound HL	11424922, 12920079, 15447792
		c.1211C>T	p.Pro404Leu	Tunisian, Turkish	DFNB8	Severe-to-profound HL (Tunisian – congenital; Turkish – onset at 6–7 years of age)	11462234, 12920079, 16021470
		c.1204G>A	p.Gly402Arg	Chinese	DFNB10	Severe HL with onset at 2 years of age and flat audiogram; progression noted	28695016
		c.1159G>A	p.Ala387Thr	Japanese		See nonsense variant c.607C>T	24130743, 25770132
		c.1156T>C	p.Cys386Arg	Indian		–	24416283
		c.1151T>G	p.Met384Arg	Chinese	DFNB10	Severe HL with onset at 3 years of age and flat audiogram with progression to downsloping audiogram by 35 years of age	28695016
		c.1126G>A	p.Gly376Ser	Turkish		–	26226137
		c.1025G>A	p.Gly342Glu	Turkish		–	21117948
		c.1019C>G	p.Thr340Arg	Italian	DFNB10	Severe-to-profound prelingual HL with reduction in thresholds at higher frequency	24657061
		c.974T>A	p.Leu325Gln	Polish		–	28566687
		c.916G>A	p.Ala306Thr	German, Dutch, Korean, Chinese;	DFNB10/DFNB8	Variable prelingual versus postlingual progressive HL, depending on other mutation present (potential founder)	17551081, 21786053, 24526180, 28246597, 28695016
		c.913A>T	p.Ile305Phe	Turkish		–	26226137
		c.809T>A	p.Ile270Asn	Chinese	DFNB10	Progressive severe HL with onset at 3 years of age with downsloping audiogram	28695016
		c.778G>A	p.Ala260T	Japanese	DFNB8	Progressive HL with onset in childhood/ early adolescence; ski-slope audiogram by 33 years of age	25770132
		c.767C>T	p.Ala256Val	Pakistani		Severe-to-profound HL	21534946
		c.763G>T	p.Ala255Ser	Chinese	DFNB10	Childhood onset severe-to-profound HL	27610647
		c.753G>C	p.Trp251Cys	Tunisian;	DFNB10	Congenital severe-to-profound HL	11462234, 12920079
		c.743C>T	p.Thr248Met	Korean		Moderate HL, preservation of low frequency	24526180
		c.727G>A	p.Gly243Arg	Indian	DFNB10	Severe-to-profound HL	24416283
		c.726C>G	p.Cys242Trp	Pakistani	DFNB8	Postlingual-onset progressive HL	24949729
	–	c.647G>A	p.Arg216His	Iranian	DFNB8	Postlingual severe-to-profound HL	26445815
	–	c.647G>T	p.Arg216Leu	Turkish, Japanese	DFNB10	Severe-to-profound HL, either congenital or onset by 1.5 years of age	16021470, 25770132
	–	c.646C>T	p.Arg216Cys	German, Caucasian	DFNB8	Severe HL by 6 years of age with progression to profound HL by 20 years of age	17551081, 22975204
	-	c.616G>T	p.Ala206Ser	Turkish		–	26226137
	SRCR	c.595G>A	p.Val199Met	Dutch	DFNB8	Prelingual low-frequency HL (onset at 5 years of age)	21786053
		c.581G>T	p.Cys194Phe	Pakistani	DFNB10	Congenital profound HL	11424922, 12920079, 15447792
		c.554A>T	p.His185Leu	Japanese	DFNB8	Postlingual severe HL	23967202
		c.551T>C	p.Leu184Ser	Chinese, Taiwanese	DFNB10	Either congenital or prelingual-onset profound HL	31016883, 32235586
		c.535G>A	p.Ala179Thr	Tibetan	DFNB8	Profound postlingual HL with threshold >85 dB	25474651
		c.436G>A	p.Gly146Ser	Turkish		-	26226137
		c.413C>A	p.Ala138Glu	British, Dutch, Caucasian, Polish	DFNB8	Progressive moderate-to-severe sloping HL at 5 years of age	16283880, 21786053, 2297520, 28566687
		c.399G>C	p.Trp133Cys	Turkish		-	26226137
		c.390C>G	p.His130Arg	Japanese	DFNB8	Progressive HL with ski-slope audiograms	25770132
		c.371C>T	p.Ser124Leu	Polish		–	28566687
		c.346G>A	p.Val116Met	Indian	DFNB10	Severe-to-profound HL	24416283
	LDLRA	c.326G>A	p.Arg109Gln	Chinese	DFNB8	Moderate HL with onset at 7 years of age	24853665
		c.325C>T	p.Arg109Trp	Pakistani, Iranian, Polish	DFNB10	Congenital severe-to-profound HL; some asymmetry in HL reported	11424922, 12920079, 2644581, 28566687
		c.316C>T	p.Arg106Cys	Japanese, Chinese;	DFNB8	Postlingual severe deafness with ski-slope audiogram	23967202, 28246597
		c.310G>A	p.Glu104Lys	Pakistani		Childhood-onset severe-to-profound HL	21534946
		c.308A>G	p.Asp103Gly	Greek	DFNB10	Prelingual severe-to-profound HL	11907649, 12920079
		c.280G>A	p.Gly94Arg	Japanese	DFNB8	Progressive HL with ski-slope audiograms	23967202, 25770132
		c.268G>A	p.Ala90Thr	UK, Moroccan		Progressive HL with ski-slope audiograms	22382023, 25770132
		c.239G>A	p.Arg80His	Taiwanese		–	32235586
		c.218G>A	p.Cys73Tyr	Polish		–	28566687
		c.212T>C	p.Phe71Ser	Japanese	DFNB8	Progressive HL with ski-slope audiograms	25770132
	TM	c.188T>G	p.Leu63Arg	NA		Mild-to-moderate childhood onset HL	26969326
Nonsense variants		c.1192C>T	p.Gln398Ter	Turkish	DFNB10	Congenital HL with minimum threshold 87 dB	16021470
		c.296C>A	p.Ser99Ter	Chinese	DFNB8	Moderate HL with onset at age of 7 years	24853665
		c.607C>T	p.Gln203Ter	Japanese	DFNB8	Slow progressive HL detected in primary school, ski-slope-type HL Postlingual HL from age of 10 years, profound HL at age of 46 years	24130743, 25770132
		c.783T>A	p.Cys194Ter	Palestinian	DFNB10	Prelingual bilateral HL	19888295
		c.310G>T	p.Glu104Ter	Pakistani		Bilateral severe-to-profound HL	21534946
		c.271C>T	p.Arg91Ter	Turkish		–	26226137
		c.115C>T	p.Gln39Ter	NA		–	26969326
		c.46C>T	p.Arg16Ter	Iranian		–	27344577
Frameshift variants		c.999delC	p.Asp334MetfsTer24	Polish		–	28566687
		c.988delA	p.Glu330GlyfsTer28	Palestinian	DFNB10	Prelingual, bilateral, severe-to-profound HL with thresholds <85 dB at all frequencies	16460646
		c.579dupA	p.Cys194MetfsTer17	Slovenian, Polish		-	26036852, 28566687
		c.207delC	p.His70ThrfsTer19	Spanish, Greek, Dutch, Slovenian, Polish	Dutch: DFNB10	Dutch: prelingual, downsloping audiogram configuration with impairment of the low frequencies at a very young age Mainly high-frequency hearing impairment with deterioration of low frequencies after 30 years of age	11907649, 21786053, 26036852, 28566687
		c.36dupC	p.Phe13LeufsTer10	Turkish		–	23226338, 26226137
		c.36delC	p.Phe13SerfsTer12	Chinese	DFNB10	Bilateral hearing impairment at 3 years of age with ski-slope audiogram; mild progression: 3–4 dBHL at 250 Hz and 7–10 dBHL at 500 Hz from 3–6 years of age	28246597
Splice site variants		c.1195-1G>C		Saudi Arabian	DFNB10	Prelingual profound HL	21726435
		c.953-5A>G		Polish		–	28566687
		c.783-1G>A		Korean	DFNB10	Bilateral, symmetrical severe-to-profound HL in the first decade of life	29072634
		c.782+8insT		Newfoundlander	DFNB10	Prelingual hearing impairment	15447792
		c.782+2T>A		Polish		–	28566687
		c.323-6G>A		Pakistani, Dutch, Indian, Chinese	DFNB8	Dutch: age of onset 4–5 years of age with downsloping audiogram and progressive HL with later deterioration of low frequencies Chinese: 9 years of age of onset with downsloping audiogram; normal threshold at 125 and 256 Hz at 14 years (postlingual, milder hearing impairment)	11137999, 21786053, 24416283, 28695016
CNVs		5 exons deletion		NA		–	24963352
		Exon 7–10 duplication		Iranian		–	26445815
		Exon 6–10 deletion		NA		–	26969326
		Complex genomic rearrangement		Chinese	DFNB10	Prelingual profound HL	31016883
Complex small structural variants		8-bp deletion and insertion of 18 monomeric β-satellite repeat units		Palestinian	DFNB10	Profound HL without any hearing remnants at a level of 75 to 80 dB	11137999

*Prelingual (DFNB10) is defined as hearing loss onset at 0–4 years; postlingual (DFNB8) is defined as hearing loss onset ages 5 years and above.

A better understanding of the natural histories of specific DFNB8/10 genotypes will also be necessary for gene therapy clinical trials. Therefore, in this study, we reviewed the English language literature pertaining to the genes and causative mutations associated with DFNB8/10 from the years 1996–2021, focusing on the current state of knowledge regarding genotype-phenotype relationships and natural disease histories of DFNB8/10 variants.

Gene and Function

The TMPRSS3 gene has 13 exons and encodes the TMPRSS3 protein, belonging to a transmembrane serine proteases family. The protein consists of 453 amino acids. It contains a serine protease domain, a transmembrane domain (TM) (49–69 aa), a low-density lipoprotein receptor A (LDLRA) domain (74–107 aa), a scavenger receptor cysteine-rich domain (SRCR) (112–211 aa), and a trypsin-like serine protease domain (216–446 aa) [NCBI, 2016] (Fig. 1). RNA in situ hybridization on rat and mouse cochlea revealed that the TMPRSS3 protein is expressed in the cell bodies of the spiral ganglion neurons, inner hair cells, supporting cells, and stria vascularis of the cochlea [Guipponi et al., 2002]. The observed expression in both the spiral ganglion and the cochlear sensory organ suggests a potential role in both these tissues; however, to date, in vitro experiments have failed to clarify the pathogenic mechanisms underlying the TMPRSS3-associated HL phenotypes. The TMPRSS3 protein plays a vital role in activating epithelial sodium channels (EnaC), which are regulated by serine protease activity and have been hypothesized to maintain low sodium concentration in the endolymph of the inner ear [Vallet et al., 1997; Couloigner et al., 2001; Guipponi et al., 2002; Lee et al., 2003].

Fig. 1.

Schematic representation depicting the predicted functional domains and sequence motifs. The positions of the described TMPRSS3 mutations are shown at the top of the figure.

Studies using the Xenopus oocyte expression system conducted by Guipponi et al. [2002] revealed that TMPRSS3 cleaves ENaC though was also associated with increased ENaC-mediated currents [Guipponi et al., 2002]. The authors also showed co-expression of ENaC and TMPRSS3 in the organ of Corti, stria vascularis, and spiral ganglion neurons, suggesting that TMPRSS3 may have a function in signal transduction between spiral ganglion neurons and hair cells. The implication of this interaction is unknown, given that EnaC is not expressed in cochlear hair cells [Couloigner et al., 2001]. More recently, single-cell RNA sequencing studies have subsequently shown that in the mouse inner ear, TMPRSS3 is not expressed in type I SGNs but only in type II SGNs which constitute 5% of all SGNs and do not appear to be involved in the processing of auditory signals [Shrestha et al., 2018; Sun et al., 2018]. Furthermore, Molina et al. [2013] demonstrated that TMPRSS3 expression in mice also maintains calcium-activated potassium channel subfamily M alpha 1 (KCNMA1) and hence normal outward K+ currents in inner hair cells [Molina et al., 2013].

Molecular genetic studies have shown that mutations in different domains of TMPRSS3 result in distinct HL phenotypes, likely due to the unique functional effects of different mutations on protease activity [Ben-Yosef et al., 2001; Masmoudi et al., 2001; Scott et al., 2001; Guipponi et al., 2002; Wattenhofer et al., 2002, 2005; Lee et al., 2003, 2012; Ahmed et al., 2004; Hutchin et al., 2005; Elbracht et al., 2007; Fasquelle et al., 2011; Charif et al., 2012; Diaz-Horta et al., 2012; Chung et al., 2014; Fan et al., 2014; Ganapathy et al., 2014]. In addition, a tight correlation has been reported between disruption of the protease activity and pathogenesis of TMPRSS3-associated HL, with mutations in the SRCR and LDLRA domains predicted to alter the folding or assembly of the catalytic domain or substrate recognition and binding by the protease [Lee et al., 2003].

TMPRSS3 has not been associated with vestibular dysfunction. Hair cell apoptosis in the vestibular sensory epithelia has been observed in TMPRSS3-mutant mice, though there was relatively greater hair cell loss in the cochlea [Tang et al., 2019]. However, mild hyperreflexia of the velocity step responses in vestibulo-ocular testing was reported in three individuals, suggesting the effect of TMPRSS3 mutations on vestibular function may be subclinical [Gao et al., 2017a].

Prevalence

Nearly 100 different mutations occurring across all functional domains of TMPRSS3 have been described in DFNB8/10 patients (Table 1). In a recent study, TMPRSS3 mutations accounted for 2.3% of genetic diagnoses in 587 diverse patients with HL [Shearer et al., 2019]. However, the prevalence of TMPRSS3 mutations in HL populations varies significantly by ethnicity. Pathogenic TMPRSS3 mutations have been reported in more than 14 ethnic groups worldwide, including Asian, Mediterranean, and Caucasian populations [Guipponi et al., 2007; Gao et al., 2017a, 2017b]. The frequency of TMPRSS3 mutations in subjects with HL negative for GJB2 or variants in other common deafness genes (GJB6 and mitochondrial A1555G mutations) is 0.45% (2/448) in a European population with childhood deafness 21; 12% (3/25) in Turkish families 22; 13.1% (5/38) in Slovenian ARNSHL patients [Battelino et al., 2016]; 1.8% (8/449) in the Pakistani population [Ahmed et al., 2004]; 5% (2/39) in Tunisian families affected by profound ARNSHL [Masmoudi et al., 2001]; and 5.9% (3/51) in a Korean ARNSHL population [Chung et al., 2014]. The TMPRSS3 mutation rate in Japanese patients is 0.36% (4/1120) [Miyagawa et al., 2015]. In the Chinese population, TMPRSS3 mutations account for about 4.6% (7/151) of ARNSHL patients negative for GJB2 and SLC26A4 mutations [Gao et al., 2017a]. TMPRSS3 is the cause of genetic HL in 4.5% of Palestinians with varying severity of deafness [Abu Rayyan et al., 2020]. TMPRSS3 is an important cause of HL worldwide, with varying rates of prevalence based on ethnicity.

Hearing Phenotypes and Relationship to Genotype

The HL phenotype varies among families with TMPRSS3 mutations by the age of onset and severity. A reduction in high frequencies characterizes prelingual TMPRSS3-associated hearing impairment, though eventually, hearing deterioration at low frequencies occurs, leading to a flatter audiogram configuration (i.e., residual hearing) [Weegerink et al., 2011]. A ski-slope-type audiogram configuration is associated with TMPRSS3, though it is not specific, as the auditory phenotype and inheritance are very similar to genes that cause high-frequency HL.

All families reported with the TMPRSS3 mutations displayed either severe-to-profound prelingual (DFNB10) or postlingual (DFNB8), progressive bilateral sensorineural hearing impairment, with no described middle ear or vestibular deficits in either group. As described below, a clear genotype-phenotype pattern is emerging that may account for the clinical variability in HL due to mutations in TMPRSS3.

Missense Mutations

Over 50 pathogenic missense mutations have been identified to date across various ethnic groups [Ben-Yosef et al., 2001; Masmoudi et al., 2001; Wattenhofer et al., 2005; Tang et al., 2019; Wong et al., 2020] (Table 1). The c.916G>A (p.Ala306Thr) mutation is the most common of these mutations. Patients with heterozygous missense mutations have been identified in East Asian, European (non-Finnish), African/African American, South Asian, Latino/Admixed American populations in the Genome Aggregation Database (gnomAD, Broad Institute), appearing in 12/19,954, 21/126,792, 4/24,972, 1/30,616, and 1/35,440 alleles, respectively, with a global minor allele frequency ∼0.00015. So far, patients with compound heterozygous c.916G>A (p.Ala306Thr) TMPRSS3 mutations have been identified in German, Dutch, Korean, and Chinese populations [Elbracht et al., 2007; Weegerink et al., 2011; Lee et al., 2013; Chung et al., 2014].

Analysis of Chinese patients with the missense mutation c.916G>A (p.Ala306Thr) revealed a majority with evidence of progressive HL by early childhood (between 3 and 6 years old) without vestibular symptoms; however, the exact phenotype of their HL depended on the other mutation present (i.e., truncating or non-truncating) and its position [Gao et al., 2017a, 2017b]. Given the significant contribution of c.916G>A (p.Ala306Thr) to the DFNB8/DFNB10 phenotype in multiple populations, it may be helpful in future screening for TMPRSS3 mutations and should be explored as a target for gene therapy.

Nonsense Mutations

Truncating nonsense mutations in TMPRSS3 have been identified in Turkish, Chinese, Japanese, Palestinian, Pakistani, and Iranian populations [Wattenhofer et al., 2005; Hashem et al., 2009; Lee et al., 2012; Miyagawa et al., 2013, 2015; Gu et al., 2015; Bademci et al., 2016; Sloan-Heggen et al., 2016; Yan et al., 2016]. For example, a Turkish patient homozygous for c.1192C>T causing a p.Gln398Ter nonsense codon had prelingual deafness at birth with an audiometric ISO value of 98 dB [Wattenhofer et al., 2005]. Conversely, in a described Chinese patient, the novel nonsense mutation c.296C>A, p.Ser99Ter resulted in a moderate postlingual HL with onset at the age of 7 years [Gu et al., 2015].

Frameshift Mutations

Six distinct frameshift mutations in TMPRSS3 have been identified, with reported phenotype in Palestinian (c.988delA, p.Glu330GlyfsTer28), Dutch (c.207delC, p.His70ThrfsTer19), and Chinese (c.36delC, p.Phe13SerfsTer12) patients [Walsh et al., 2006; Weegerink et al., 2011; Gao et al., 2017b]. In 11 members of a Palestinian family homozygous for the (c.988delA, p.Glu330fs) mutation, all experienced prelingual severe-to-profound HL with <85 dB thresholds for all frequencies. This frameshift mutation results in a truncated TMPRSS3 protein unable to activate EnaC [Walsh et al., 2006]. A Chinese patient with a q c.36delC, which leads to the premature stop p.Phe13SerfsTer12, was heterozygous for a p.Ala306Thr mutation and presented at 3 years of age with bilateral hearing impairment and a ski-slope audiogram. The patient’s HL had a mild progression until cochlear implantation at 6 years of age [Gao et al., 2017b]. Four Dutch patients were described with the c.207delC as well as a c.1276G>A (p.Ala426Thr) mutation. The onset of HL ranged from age 7 to 17 years, with mainly high-frequency deterioration and loss of low frequencies after the age of 30 years [Weegerink et al., 2011].

Splice Site and Copy Number Variants

Splice site variants in TMPRSS3 have been implicated in six cases of HL with varying phenotypes. Three children born to consanguineous Saudi Arabian parents with a c.1195-1G>C variant had profound, prelingual HL. This mutation was found to result in activation of a cryptic splice site and a subsequent frameshift [Imtiaz et al., 2011]. Compound heterozygotes for the c.782+8insT mutation in a Newfoundland family predicted to cause skipping of exon nine also exhibited prelingual hearing impairment [Ahmed et al., 2004].

To date, three copy number variants (CNVs) in TMPRSS3 have been described that are associated with deafness. In addition, a large deletion, spanning five exons, a homozygous duplication of exons 7–10, and a deletion of exons 6 to 10 have been described to cause HL [Sloan-Heggen et al., 2015; Sloan-Heggen et al., 2016; Azaiez et al., 2018]. Finally, in a Chinese family, Li et al. [2019] reported a genomic rearrangement consisting of a deletion of exon 11, part of exon 10, and an inversion of exon 3 to exon 9 [Li et al., 2019]. Additionally, two complex small structural variants have been reported including include an 8-bp deletion and an insertion of 18 monomeric (approximately 68-bp) β-satellite repeat units in exon 11 reported in a Palestinian family [Scott et al., 2001]. Patients with CNVs and small structural variants have been reported to exhibit to a DFNB10 phenotype.

Phenotypic Effects of Heterozygosity for TMPRSS3 Mutations

For heterozygous mutations, the nonrandom distribution of different mutation types suggests a correlation between the type of mutation and the onset age and degree of hearing impairment [Weegerink et al., 2011]. Severe hearing phenotypes have been associated with TMPRSS3-truncating mutations (frameshift, stop codon, CNVs, and splice site mutations) and missense variants located within the catalytic domain of the serine protease or near the active site [Weegerink et al., 2011]. The variability in the clinical features associated with TMPRSS3 deafness might be explained by the disabling effects of pathogenic mutations in TMPRSS3 on the protease activity [Lee et al., 2003].

Less severe mutations are generally missense mutations outside of active sites. The TMPRSS3 phenotype is therefore dependent on the two TMPRSS3-mutant alleles. Compound heterozygosity for either two mild mutations or a combination of a mild and severe mutation leads to postlingual HL (DFNB8), whereas the combination of two severe mutations results in profound hearing impairment with prelingual onset (DFNB10) [Weegerink et al., 2011]. For example, the combinations of p.Phe13Serfs∗12 and p.Ala306Thr, p.Gly417Glu and p.Ala306Thr, p.Gly402Arg and p.Ala306Thr, p.Ile270A3sn and p.Ala306Thr, p.Met384Arg and p.Leu415Ser result in prelingual, profound hearing impairment (DFNB10), whereas the combination of c.323-6G>A and c.916G>A (p.Ala306Thr) results in postlingual, milder hearing impairment (DFNB8).

However, this classification schema does not fully capture the complexity of the relationships. For example, genetic analysis of Dutch populations showed that either the missense mutation p.Ala306Thr or p.Val199Met, in combination with each other or the frameshift variant p.Thr70fs, results in profound prelingual HL (DFNB10); however, when combined with either of the missense mutations p.Ala426Thr or p.Ala138Glu, these mutations result in a milder phenotype with postlingual HL and later progression (DFNB8) [Weegerink et al., 2011]. In addition, analysis of the missense mutation c.1219T>C (p.Cys407Arg), the most common TMPRSS3 variation seen in Pakistani populations, also reveals a mild and a severe hearing phenotype, depending on the second allele [Lee et al., 2012].

This pattern is not followed with the c.323-6G>A (p.Cys107fs) mutation. According to the classification in the study by Weegerink et al. [2011], the c.323-6G>A mutation is relatively severe, implying that a c.323-6G>A mutation in the homozygous state would be expected to result in prelingual (DFNB10) deafness. However, the c.323-6G>A mutation in a homozygous state has been reported to cause postlingual (DFNB8) HL. A possible explanation for this may be variations in splicing. The Berkeley Drosophila Genome Project Splice Site Prediction Program predicts that the c.323-6G>A mutation may influence splicing by introducing a novel splice acceptor site in addition to the normal splice acceptor site [Reese et al., 1997]. Therefore, both normal and abnormal splicing may arise in relatively variable amounts between individuals, which may then result in phenotypic variation.

Epigenetic Regulation of Phenotype

Environmental and genetic modifying factors may also affect the expressivity of TMPRSS3-related HL. Although the Turkish and the Tunisian families harbor identical p.Pro404Leu mutations, each has a distinct hearing phenotype; in the Tunisian family, deafness was congenital [Masmoudi et al., 2001], whereas, in the Turkish family, the onset of deafness was at age 6–7 years [Wattenhofer et al., 2005]. Evaluation of genetic modifying factors associated with TMPRSS3 may help better account for this difference.

The phenotype of the TMPRSS3 p.Pro404Leu mutation has not been well defined [Masmoudi et al., 2001; Wattenhofer et al., 2005]. For the family with a homozygous p.Pro404Leu mutation reported by Wattenhofer et al., the average threshold levels (0.5–4 kH) were already 85–99 dB at the age of HL detection around 6–7 years old. However, the speech development of the affected family members was not reported, and it is unclear whether the course of the HL was prelingual or a rapidly progressing postlingual pattern [Wattenhofer et al., 2005].

For the second family with a homozygous p.Pro404Leu mutation reported by Masmoudi et al. [2001], hearing impairment was congenital and severe-to-profound consistent with prelingual onset. Based on these findings, the p.Pro404Leu is likely to be a severe mutation [Masmoudi et al., 2001]. Modifying genetic factors, or perhaps a milder congenital and rapidly progressive HL that was not detected in the original family, may explain the difference in age of onset of hearing impairment between the two families with the same mutation.

CIs and TMPRSS3

Patients with TMPRSS3 mutations have variable CI outcomes. While a number of studies have reported good speech perception among adult TMPRSS3 CI recipients [Elbracht et al., 2007; Weegerink et al., 2011; Miyagawa et al., 2013], three other studies have shown poorer than expected postimplantation speech perception [Eppsteiner et al., 2012; Shearer et al., 2017; Tropitzsch et al., 2018]. The TMPRSS3 protein is expressed on spiral ganglion cell neurons [Fasquelle et al., 2011]. Eppsteiner et al. hypothesized that poor CI outcomes in some patients with TMPRSS3 variants might be due to negative effects on the spiral ganglion, which is not bypassed by the CI [Eppsteiner et al., 2012]. Recently, Holder et al. reported positive CI outcomes in three children with TMPRSS3 mutations and did not find decline in performance, indicating spiral ganglion cell degeneration at up to 2.5 years postimplantation. They argue that CI is appropriate treatment for children with sloping HL due to TMPRSS3 mutation [Holder et al., 2021]. Characterizing the genotype-phenotype correlations in TMPRSS3 and increasing the identification of patients with pathogenic TMPRSS3 mutations will contribute to a better understanding of CI effectiveness in these patients.

Conclusion

Reported mutations in the TMPRSS3 gene display marked heterogeneity in the HL phenotype. The complex clinical picture of TMPRSS3 mutations is likely due to the multiple roles of the TMPRSS3 protein in the molecular processes contributing to hearing. As one of the most common causes of genetic HL, TMPRSS3 is an excellent target for future molecular therapies. However, the phenotypic variability observed within and among TMPRSS3 mutations makes it challenging to identify patients clinically. Additionally, efforts to develop therapy targeting specific mutations in TMPRSS3 are limited by the heterozygosity of the reported variants, with more than 75 to date, and the diversity of TMPRSS3 mutations may be better addressed with gene replacement techniques.

Identification of patients for the development of gene therapies targeting specific TMPRSS3 mutations will require a comprehensive understanding of the molecular physiology underlying the individual features of TMPRSS3-associated phenotypes. As the most common pathogenic mutation reported across multiple ethnicities, the c.916G>A (p.Ala306Thr) missense mutation should be further explored as a target for molecular therapy. Classification of TMPRSS3 mutation phenotypes, particularly prelingual versus postlingual, will be an essential step in developing effective gene therapy.

Statement of Ethics

IRB approval was not required for this study.

Conflict of Interest Statement

All authors declare no relevant conflicts of interest.

Funding Sources

Dr. Liu’s laboratory is supported by NIH grants of R01DC005575, R01DC012115. Dr. Eric Nisenbaum is supported by T32 DC015995. The funders had no role in study design; collection, analysis, and interpretation of data; and writing of the report and placed no restrictions regarding the submission of the report for publication.

Author Contributions

Eric Nisebaum, Denise Yan, and A. Eliot Shearer contributed to methodology, acquisition of data, and writing of original draft; Evan de Joya, Torin Thielhelm, and Nicole Russell performed data acquisition and writing of original draft; Eric Nisenbaum, Denise Yan, A. Eliot Shearer, and Evan de Joya contributed to content revision; Hinrich Staecker, Zhengyi Chen, and Jeffrey Holt contributed to conceptualization and content revision; and Xuezhong Liu contributed to conceptualization, design, and content revision. All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Funding Statement

Dr. Liu’s laboratory is supported by NIH grants of R01DC005575, R01DC012115. Dr. Eric Nisenbaum is supported by T32 DC015995. The funders had no role in study design; collection, analysis, and interpretation of data; and writing of the report and placed no restrictions regarding the submission of the report for publication.

Data Availability Statement

All data generated or analyzed during this study are included in this article or are publicly available in the cited literature. Further inquiries can be directed to the corresponding author.