Introduction

Sensorineural hearing impairment is the most frequent sensory defect of childhood: one child in 1000 is born deaf in developed countries1 and deafness is mainly prelingual. Recent studies suggested that more than 75% of childhood deafness is of genetic origin. About one quarter of genetic forms are syndromic (deafness is a part of genetic syndrome). The remaining forms are classified as nonsyndromic, but in some cases classification is not obvious due to late onset of some symptoms. To date, over 80 loci for nonsyndromic hearing impairment (NSHI) have been identified, but despite this wide heterogeneity, GJB2 genetic alterations are causative in up to 50% of prelingual NSHI.2 Over 400 syndromes associated with deafness have been reported. Several genes are implicated in syndromic or nonsyndromic forms. For example, GJB2, the most common NSHI gene, has also rarely been found in several syndromic forms of deafness.2

Similar findings apply to the SLC26A4 gene (PDS gene), encoding pendrin. Indeed, SLC26A4 mutations are causative in two autosomal recessive disorders, Pendred's syndrome and a NSHI with inner ear malformation.3, 4 Pendred's syndrome is characterized by congenital sensorineural hearing loss and goitre that usually develops in the second decade of life (the median age of goitre appearance was 14.9 years in a study performed in a cohort of French patients with Pendred's syndrome5) and is associated with hypothyroidism in 50% of cases.5, 6 Pendred's syndrome and DFNB4 patients have in common similar deafness features: bilateral, typically prelingual or early postlingual, frequently severe or profound with variable evolution, and associated with inner ear anomalies, such as enlarged vestibular aqueduct (EVA) or Mondini's dysplasia.7, 8 Many studies have focused on Pendred's syndrome, in which the role of SLC26A4 is well defined, but the prevalence of SLC26A4 gene alterations in NSHI remains unclear and is certainly underestimated.

The SLC26A4 gene, located on the long arm of chromosome 7,9 encodes pendrin, a protein expressed in various tissues, including principally the inner ear, thyroid and kidney. Pendrin is a transmembrane anion exchanger, with 12 predicted transmembrane domains, that belongs to the solute carrier 26 family. It was shown to exchange chloride, iodide,10 bicarbonate11 and formate.12 Recently, Wangemann et al, showed that the absence of pendrin led to altered inner ear tissue pH through impaired HCO3-transport, enhanced oxidative stress, absence of KCNJ10 expression and absence of endolymphatic potential.13

In Pendred's syndrome, the SLC26A4 gene has been extensively studied and over 100 different mutations, spanning the entire coding region, have been described. The prevalence of SLC26A4 mutation in Pendred's syndrome is very high in different studies, up to 90%.14 Screening of this gene in large cohorts of hearing impaired patients has been previously performed on the basis of the presence or not of an enlarged vestibular aqueduct (EVA) detected on CT-scan, but without precise informations concerning the thyroid involvement and the ages of the patients.15, 16, 17 Indeed, in the Pendred syndrome, the goitre usually appears in the second decade of life and the probability to develop or not goitre depends of the age of patients.

The prevalence of SLC26A4 mutations in patients with NSHI and inner ear anomalies has never been studied in large and standardized series of Caucasian patients.

From 1995 to 2002, a national research program on deafness allowed a prospective collection of clinical data and samples of 109 children (100 unrelated families) with NSHI and enlarged vestibular aqueduct. Molecular SLC26A4 screening was undertaken in this series of patients in order to determine the prevalence and the spectrum of SLC26A4 gene mutations in this pathology and to identify genotype/phenotype correlations.

Methods

We enrolled, in a French national collaborative study, 109 deaf children from 100 unrelated families, who presented with a bilateral prelingual (less than 3 years of age) or an early postlingual (during the first decade) nonsyndromic sensorineural hearing impairment associated with an EVA on CT-scan (some patients had other inner ear anomalies like Mondini dysplasia). In all patients, the mode of inheritance was compatible with an autosomal recessive inheritance. None of patients had goitre or other thyroid anomaly. For each patient, a complete medical history was obtained to determine the age of onset of the deafness and to exclude the possibility of environmental causes. The deaf subjects underwent an otoscopic examination of the ear, nose and throat and a general examination, with systematic assessment for signs suggestive of a syndromic form of deafness (in particular, goitre and signs of hypothyroidism) by a clinical geneticist specializing in the management of deaf children. They also had an ophthalmological evaluation, investigation for haematuria and proteinuria and an electrocardiogram. None of patients had GJB2 mutation. Deaf children underwent pure-tone audiometry with a diagnostic audiometer in a sound-proof room, with recording of pure-tone air- and bone-conduction thresholds. Air-conduction pure-tone average (ACPTA) thresholds in the conversational frequencies (0.5, 1, 2 and 4 kHz) were calculated for each deaf ear and were used to define the severity of deafness: mild (20 db⩽39 db), moderate (40 db⩽69 db), severe (70 db⩽89 db) and profound (⩾90 db). The severity of deafness in each child was defined by the degree of hearing loss of the best ear. In accordance with the European Working Group on Genetics of Hearing Impairment criteria, hearing loss was considered as progressive when the patient lost more than 15 dB in the ACPTA thresholds in the conversational frequencies by comparison of two reliable audiometric tests carried out at least 10 years apart, to which we added the further criterion of more than 8 dB loss in tests at least 5 years apart. We considered deafness as fluctuant when the mean hearing level in conversational frequencies had risen by more than 10 dB between two successive audiograms.

Genomic DNA was isolated from whole blood. The protocol was accepted by the Committee for the Protection of Individuals in Biochemical Research as required by French legislation and informed consent was obtained from all patients or their parents.

SLC26A4 mutation screening was performed by Denaturing High-Performance Liquid Chromatography (DHPLC), on a Wave™ DNA Fragment Analysis System (Transgenomic™, USA). All 20 protein-encoding exons (2–21) were previously amplified by Polymerase Chain Reaction (PCR).

Variant profiles were sequenced on a new amplification, using an ABI 310 Genetic Sequencer (Applied Biosystems). Sequences were analysed by Sequence Analysis 3.0 (Applied Biosystems). Exon 20 was directly sequenced for all patients. Wherever possible, mutation segregation was studied by direct sequencing of the variant exon in parents. New variants were tested in 50 unrelated healthy controls.

To analyse correlations between genotype and phenotype, patients were classified according to the presence of biallelic SLC26A4 mutations or the absence of mutations. Groups were compared using χ2 and Student's tests.

Results

The 109 patients from 100 unrelated families were aged 1–32 years (mean age: 10.2 years and median age: 10 years) at the time of the genetic assessment.

Allowing one allele in each consanguineous family, we detected 91/198 (46%) allelic variants, 53 of which were different. For the study of clinical phenotype and the prevalence of SLC26A4 mutations, we considered as possibly deleterious frameshift mutations, missense mutations of the coding sequence and splice site mutations. Intronic variants with unknown pathogenicity were not taken in account in the prevalence determination. The L597S missense mutation was found in four alleles from 50 healthy controls and was then considered as nonpathogenic.

Prevalence of SLC26A4 mutation in this series was 40% (40/100 families), detailed in Table 1. Prevalence of biallelic mutation was 24% (24/100 families, including six families with homozygous mutations, two of them with consanguinity), prevalence of monoallelic mutation was 16% (16/100 families). Among 33 patients tested with caloric vestibular tests because of clinical suspicion of vestibular disorders, 10 had alterations and among them three patients were in the group with monallelic or biallelic mutation.

Table 1 Details of Phenotype and genotype of the 45 patients with SLC26A4 mutation
Full size table

A total of 19 variants have never been described (Tables 2 and 3), consisting of nine missenses (T99R; M147T; I199T; M283I; T307M; Q421L; F355S; C565R; V690A: these mutations were not detected in 50 unrelated healthy controls), two insertions (129InsC; IVS12−3InsCAGT), one deletion (IVS7−12DelTTATT), one splice site (IVS11+1 G>C) and six intronic variants (IVS3+14 A>C; IVS4−7 A>G; IVS12−14 T>C; IVS13+9 C>T; IVS13−5 T>G; IVS20−22 T>C). Some of them were not conserved when compared to proteins with significant sequence homology to human pendrin (Table 3).

Table 2 Details of the 19 new allelic variants
Full size table
Table 3 Characteristics of mutated amino acid of the nine new missense mutations compared to proteins with significant sequence homology to human pendrin
Full size table

Three mutations were frameshift: one deletion (2127delT), and two insertions (IVS12−3insCAGT, 129insC). These mutations led to a stop codon at X719 for 2127delT, at X467 for IVS12−3insCAGT and at X85 for 129insC. One particular mutation was IVS7−12DelTTATT: using two splice site prediction software,18, 19 the acceptor splice site prediction of exon 7 decreased from 0.95 in wild type to 0.19 in mutated type. The mutation was then considered as pathogenic.

The clinical features of patients with biallelic mutations compared to patients without mutation are detailed Table 4. Difference between groups was statistically significant for the following items: the age of deafness discovery, the walking age, the fluctuating evolution and the hearing loss severity.

Table 4 Clinical phenotypes of the two distinct groups with biallelic SLC26A4 mutations and without SLC26A4 mutation
Full size table

Discussion

In 198 alleles tested from unrelated families, 91 allelic variants were found and 53 were different. The analysis of the segregation of mutations between families always gave coherent results. The nine new missense mutations identified in this study showed a variable degree of conservation in their amino acid sequence when compared to proteins with high-sequence homology, and therefore may or not be disease causing. However, because of the low incidence of SLC26A4 mutations in DFNB4 hearing loss, only a very large study in people with normal hearing, combined with a study in patients DFNB4 deafness, could establish the pathogenicity of SLC26A4 mutations.

Mutations found in our series were compared to a study of patients with Pendred's phenotype5 realized in our centre, in patients with a similar origins. Many of the mutations found were common: E29Q, V138F, G209V, L236P, IVS8+1 G>A, R409H, T410M, T416P, Y78C, T193I, F355S, L445W, Y530H, S694P, D724N, 2127delT. Moreover, among the four frequent mutations (G209V, L445W, Y530H, IVS8+1) in the present study, all were also found in Pendred families.5 Molecular studies of Pendred syndrome were performed in other populations. The frequent mutations were: T416P, L236P, E384G.16, 20, 21 These mutations have been investigated by in vitro functional studies and compared to frequent mutations implicated in DFNB4 patients (V480D, V653A and I490D/G497S):3, 11, 22 Significant differences were shown in terms of in vitro activity: the V480D, V653A and I490D/G497S mutations allowed continued pendrin activity, whereas the L236P, T416P and E384G mutations did not. These in vitro results are in contradiction with the SLC26A4 molecular results in DFNB4 and Pendred's syndrome populations (similar spectrum of mutations). In the functional studies, the results reflect the consequences of a single mutation. Thus, the type of mutation in trans or the intervention of a regulatory element, either genetic or external, could be key in determining the degree of activity of pendrin and hence the clinical phenotype (for both DFNB4 deafness, as well as Pendred's syndrome).

The prevalence of SLC26A4 biallelic mutations in our series of 100 families was 24% (24/100 families) and prevalence of monoallelic mutation was 16% (16/100). Comparison with other series is difficult, the criteria of inclusion of the patients for SLC26A4 screening being usually the presence of an EVA, without detailed data about the thyroid status and the age of the patients. Tsukamoto reported a prevalence of 78.5% in a Japanese study of 32 deaf patients with EVA,23 however, it should be noted that that the age of subjects is unclear and that the H723R mutation represents 53% of the mutated alleles in this population.23 In a Caucasian population, Scott11 found that 15% of the 20 patients with a DFNB4 type hearing impairment had a SLC26A4 mutation. Finally, the prevalence of SLC26A4 mutations in our series is lower compared to that seen in Pendred's syndrome, estimated to be 90% (40 patients with Pendred's syndrome from 30 unrelated families recruited in the same population).5

Comparing the prevalence of bi and monoallelic mutations to the data reported in subjects with Pendred's syndrome recruited in the same population5 (biallelic mutations: 24% in SLC26A4 versus 77% in Pendred's syndrome; monoallelic: 16 versus 13%), it raises the question of the pathogenicity of SLC26A4 in the cases of simple heterozygosity. Tsukamoto in a Japanese series showed similar datas: the prevalence of heterozygotes was 0% in Pendred's syndrome patients, 31% in presumed DFNB4 patients.23 Pryor17 studied 18 subjects with nonsyndromic EVA, and only detected monoallelic mutations in these patients.

Nevertheless, some arguments support the role of this gene in the clinical symptomatology of patients with monoallelic mutations: the incidence is higher than that estimated for heterozygotes in the general population, indicating that the second mutation may not have been found. Indeed, the existence of abnormalities in the regions regulating transcription, the promotor site and in exon 1 (which is not translated), as well as deletions involving one or several exons which cannot be amplified by standard PCR, have not been studied. A frequent mutation or deletion in one of those regions could explain the high rates of heterozygotes. In addition, another explanation could be the presence of molecular defects within the regulator genes for SLC26A4: indeed, a murine study of Foxi1 (Fkh10), a positive gene regulator of pendrin, found that in Foxi1(−/−) mutants, there was no expression of pendrin in the inner ear.24 However, there is no evidence in the literature to suggest a digenic aetiology.

With respect to the individuals without mutations, there could be another gene for NSHI, although no other gene responsible for deafness with EVA has yet been identified. In addition, EVA and Mondini malformations of the cochlea are not specific for Pendred's syndrome or DFNB4 type deafness, but can also be found in BOR (branchio-oto-renal) and Waardenburg's syndromes. Thus, it can be postulated that some patients who do not have SLC26A4 mutations may represent minor sporadic, or incomplete forms of BOR or Waardenburg's syndromes. Finally, several studies have implicated a number of ionic channels in the homeostasis of the endolymph. Alterations in these proteins could be responsible for some cases of deafness with dilatation of the inner ear.13

The detailed phenotypic study revealed significant differences between the group with bialleclic mutation and the group without mutation. Biallelic mutations of the SLC26A4 gene were associated with a more severe hearing loss at the time of the first consultation, an earlier age of discovery of the hearing impairment (that may be due to a more severe deafness) and a fluctuating course for the disease. This leads to the definition of a phenotype that is largely associated with DFNB4 type hearing loss and thereby directs the choice of molecular tests.

It is of interest that the median age for learning to walk was lower in the biallelic group than the group without mutation (12.7 versus 15.4 months, respectively), although both are normal. As vestibular tests were not systematically performed, it is difficult to establish if any real vestibular defects occurred more frequently in one or the other group.

Differentiating an isolated deafness with EVA from Pendred's syndrome in children, and in particular in very young children, can be difficult. This is especially the case if the thyroid signs (goitre and hypothyroidism) of Pendred's syndrome, which usually appear in the second decade,6 only manifest themselves late and hence the diagnosis can only be definitively established as the disease progresses. The median age of our cohort being 10 years, we can expect that some of the younger patients may develop goitre.

The perchlorate discharge test has been used to try to differentiate between the two entities. However, there are both false negatives and false positives. In our study, five out 23 perchlorate tests were positive, four of which in group with biallelic mutations. Both the clinical examination of the neck and the thyroid function tests were normal in the patients with positive perchlorate discharge test, whose mean age was younger compared to the mean age of goitre onset in Pendred's syndrome (6.5 years).5, 6 Thyroid signs can present at a later date and it is essential that children with a mutation of the SLC26A4 gene are followed up with thyroid function tests (LT4 and TSH levels have to be normal) and clinical search for a goitre.

An ongoing study in our department has shown that out of 888 cases with a nonsyndromic prelingual or early postlingual deafness, 91 had an associated EVA that is, 10.2%. In the literature, 6–15% of the nonsyndromic deafness carry EVA.7, 25 From our series, in the same population, it can be estimated, considering biallelic or bi and monoallelic mutations, that up to 4% of nonsyndromic deafness in children is attributable to mutations of SLC26A4. Thus, the SLC26A4 gene is the second most frequent cause of nonsyndromic deafness, the commonest being GJB2.26

The identification of this form of deafness can now allow reliable genetic counseling, advice on prognosis and implementation of measures to prevent fluctuations in the deafness (avoidance of trauma). Management requires regular monitoring of the deafness and thyroid function.