Molecular epidemiology of DFNB1 deafness in France

Background Mutations in the GJB2 gene have been established as a major cause of inherited non syndromic deafness in different populations. A high number of sequence variations have been described in the GJB2 gene and the associated pathogenic effects are not always clearly established. The prevalence of a number of mutations is known to be population specific, and therefore population specific testing should be a prerequisite step when molecular diagnosis is offered. Moreover, population studies are needed to determine the contribution of GJB2 variants to deafness. We present our findings from the molecular diagnostic screening of the GJB2 and GJB6 genes over a three year period, together with a population-based study of GJB2 variants. Methods and results Molecular studies were performed using denaturing High Performance Liquid Chromatograghy (DHPLC) and sequencing of the GJB2 gene. Over the last 3 years we have studied 159 families presenting sensorineural hearing loss, including 84 with non syndromic, stable, bilateral deafness. Thirty families were genotyped with causative mutations. In parallel, we have performed a molecular epidemiology study on more than 3000 dried blood spots and established the frequency of the GJB2 variants in our population. Finally, we have compared the prevalence of the variants in the hearing impaired population with the general population. Conclusion Although a high heterogeneity of sequence variation was observed in patients and controls, the 35delG mutation remains the most common pathogenic mutation in our population. Genetic counseling is dependent on the knowledge of the pathogenicity of the mutations and remains difficult in a number of cases. By comparing the sequence variations observed in hearing impaired patients with those sequence variants observed in general population, from the same ethnic background, we show that the M34T, V37I and R127H variants can not be responsible for profound or severe deafness.


Background
The genetic origin of deafness is suspected in more than half of the congenital hearing loss cases. More than 400 syndromes can include hearing loss or deficient hearing functions as a component. However, non syndromic expression of deafness is observed in more than 70 % of cases. In the non syndromic forms of hearing loss (NSHL), familial or sporadic cases are observed and the transmission is predominantly autosomal recessive. Genetic heterogeneity has been established by linkage studies: more that 50 loci associated with NSHL (including dominant and recessive autosomal, and X-linked types of transmission) have been localized, making possible the identification of a number of causative deafness genes http://dnalab-www.uia.ac.be/dnalab/hhh/. Despite the extreme genetic heterogeneity, the recessive DFNB1 locus, mapping to chromosome 13q12, is by far the most prevalent. This locus contains the two Gap junction genes GJB2 and GJB6, encoding, respectively connexin 26 (CX26) and connexin 30 (CX30). These proteins associate in hexamers to form homo-and hetero-connexons [1,2]. Two connexons from adjacent cells dock to form a functional channel that will allow, among other small molecules, the diffusion of potassium ions critical for the normal sensory hair cell excitation [3].
The contribution of the GJB2 gene in NSHL varies from 0 to 40 % in diverse populations [4] and this genetic heterogeneity is also emphasized by the variation in frequency of specific mutations among different populations. More than 70 mutations in the GJB2 gene have been reported [5], and although the majority are rare or private, the prevalence of four mutations define specific ethnic origins. The 35delG mutation accounts for approximately 70 % of GJB2 mutant alleles in Northern and Southern European, as well as American Caucasian populations, with a carrier frequency of 2.3 % to 4 % [6][7][8][9]. The three other mutations, 167delT, 235delC or R143W represent the most common pathogenic alleles in Ashkenazi Jews [10], Asian [11][12][13][14] and Ghanian populations [15,16], respectively.
Recently, we and other groups have identified a large 309 kb deletion that includes the 5' region of the GJB6 gene and most of its coding region [17]. It is unclear whether this deletion removes regulatory elements common to GJB6 and GJB2 resulting, in addition to the deletion of GJB6, in reduced expression of the wild type GJB2 gene [17][18][19][20]. This deletion also appears to have an ethnic specific origin as it is absent from the Siberian (manuscript in preparation), Chinese [21], Austrian and Italian populations [17,22]. In this report, when we refer to GJB6 mutations we will consider only this particular mutation ∆(GJB6-D13S1830).
Molecular diagnostic testing of non syndromic deafness was initiated in Montpellier in early 2000. This testing was carried out in parallel with a molecular epidemiology study of GJB2 variants in the Languedoc Roussillon region. Although many reports have estimated the carrier frequency in French and Mediterranean populations (mostly of the 35delG mutation), differences of frequency between samples are observed (0.0 to 2.7 % in France). This is essentially due to the size and composition of control samples [23]; for review see [4]. Assessment of GJB2 variant sequence distribution in Languedoc Roussillon was necessary, as we had observed significant differences in the distribution of CFTR mutations between several French regions [24].
In this study, we present our results from three years of molecular diagnostic testing of GJB2/GJB6 including the clinical and associated audiologic findings and also determine the prevalence and spectrum of DFNB1 mutations in the southern France population. In addition, we report the first screening of the most frequent GJB2 variants, on several thousand dried bloodspots (Guthrie cards) from newborns and thus re-evaluate the pathogenic status of some GJB2 variants.

Patients
A total of 159 unrelated families, comprised of 184 patients with sensorineural hearing loss, were referred from the Genetic Counseling Department and/or the Ear, Nose and Throat specialized clinics (Centre Hospitalier Universitaire of Montpellier and Lyon). All patients had permanent hearing loss not caused by infections, exposure to drugs or other prenatal or perinatal etiology of deafness. Informed consent was obtained for each individual. The hearing loss could be moderate to profound, bilateral or unilateral, symmetrical or asymmetrical, stable or progressive, pre or post-lingual, syndromic or non syndromic. From family histories, 79 patients were classified as sporadic cases, 54 defined as familial cases, including 33 with autosomal recessive, 19 with autosomal dominant, one with X-linked and one ambiguous modes of inheritance. Seven families were consanguineous. For 26 patients, family histories were unknown.
Among these 159 families, 84 were affected with a non syndromic, stable, bilateral, congenital, mild to profound deafness: 55 families had a single deaf child and the others, considered as familial cases, showed an autosomal recessive (24 cases), dominant (4 cases) or unclear (1) mode of inheritance. Hearing loss was moderate in 4 families, mild in 17, severe to profound in 60 and observed with variable expression in 3 families.

Audiological assessment
Pure-tone audiometry (PTA) was performed on every affected family member. PTA was performed for air conduction on each ear using an Interacoustics audiometer. Air-conduction thresholds were obtained at 0.5, 1, 2, 3, 4, 6 and 8 kHz. Severity of deafness was defined as mild (20-40 dB HL), moderate (41-70 dB HL), severe (71-90 dB HL) or profound hearing loss (above 90 dB HL). In young children, behavioral audiometry was used to determine the auditory thresholds in free field conditions. All subjects had an auditory brainstem response (ABR) assessment to determine the hearing loss level for high frequencies.

DNA extraction from patient WBC and mutation analysis
Blood samples were obtained from deaf patients, and their parents and sibs (when possible) and DNA was extracted with the Nucleon BACC3 DNA extraction kit (Amersham Pharmacia Biotech, Piscataway, NJ).

DNA extraction from dried bloodspots and mutation analysis of GJB2
Guthrie cards were obtained from the GREPAM (Center for neonatal screening) of Montpellier. All samples were anonymized and no phenotypic data could therefore be available. Spots of 3 mm diameter, punched from the cards, were distributed in 96-well plates and DNA was extracted using methanol extraction [26]. All experiments were then set up using the robot BioMEK 2000 (Beckman). A first set of 2,777 spots were screened using the PSDM assay, specific for the 35delG mutation. When DHPLC technology became available in the laboratory, a further 3,516 spots were analyzed for the R1' fragment (covering from the ATG codon to position 230 of the coding sequence (CDS)). 528 of these 3 516 spots were also analyzed in the R2 fragment (position 190 to 500 of the CDS). Pools of 2 DNAs were systematically used for DHPLC screening, thus eliminating the risk of missing the detection of a homozygote sample. Any abnormal DHPLC profile was re-evaluated on individual DNA samples followed by sequencing analysis.

Statistical analysis
Statistical analyses were performed with the Rv.1.3.1 software (The free software Fundation, Inc). Proportion, chisquare and Fisher exact test were used to test differences between groups. All p-values were taken to be significant at <0.05. When observed or expected values were below 5, a Fisher exact test was performed.

Patients
The analysis of the coding and non coding GJB2 exons plus the ∆(GJB6-D13S1830) screening allowed us to genotype 30 unrelated individuals with biallelic GJB2 and/or GJB6 (∆(GJB6-D13S1830)) mutations (Table 1). Clinical and audiological evaluation showed that 27 of these patients with biallelic DFNB1 mutations had bilateral congenital severe or profound NSHL with no evidence of progressive phenotype. No mutation was identified among families with an autosomal dominant mode of inheritance.
Sixteen additional patients carried 2 GJB2 mutations/variants with a controversial pathogenic effect or a single GJB2 sequence variation ( Table 2).
is the second most frequent mutation (9/60) and accounts for 15 % of the mutated alleles: in addition to the 2 homozygotes patients, 5 patients were identified as compound heterozygotes ∆(GJB6-D13S1830) /GJB2 mutation.

Parents of deaf individuals
No de novo mutations were detected in patients, as every parent of patients carrying bi-allelic GJB2/GJB6 mutations was heterozygous for one of the mutations. In two unrelated families, 2 normal-hearing parents of children genotyped R127H/M34T or R127H/W24X, were found to be homozygous R127H. Interestingly, in each of these fami-lies, a normal-hearing sib also carried the genotype R127H/M34T or R127H/W24X. Thus, R127H may not be associated with deafness.

Analysis in the general population
Twenty-two different GJB2 sequence variations (lying in R1' and R2) were identified in the general population and are listed in table , together with the calculated carrier frequencies. Ten of these sequence variations were detected more than once (35delG, M34T, V37I, V27I, W24X, E47X, Y68C, R127H, V153I, F83L), with relative frequencies    Controls: patients with no GJB2 mutation. Chi-square test was used to test the independence between the genotype 765T/T and the control group. We also present the allelic frequencies of some GJB2 mutations in individuals referred for NSHL with respect to the frequency in the general population (Table 5). We could not detect any significant difference between the two groups for the V37I, R127H and M34T sequence variations.  [32]. The epidemiologic study presented here is based on the largest number of random samples describing an unbiased general population screen and once more supports the heterogeneous composition of the Languedoc Roussillon population. This situation has direct implications for genetic counseling as well as on the development of potential diagnostic kits (as it was in the case for the design of the CF neonatal screening kit [24]). The explanation for such a different carrier rate lies in the heterogeneity of the migrations "landing" in Languedoc Roussillon, historically and still today. A recent study, carried out on the prevalence of Hemochromatosis gene (HFE) mutations in the Languedoc Roussillon population has annotated the origins of the four grand-parents for each newborn. It was observed that the population originated from various regions in France and also from other European or African countries [33].

The 35delG mutation in patients and in general population
Other sequence variations in the GJB2 gene 35delG remains by far the most frequent mutation in the hearing impaired population, although its frequency is lower than was expected in comparison with other Mediterranean areas. The second most common mutation in our region is the ∆(GJB6-D13S1830) with 15 % of the mutated alleles, and the E47X mutation represents 5 %. The other mutations were identified twice or only once.

Founder effects and sequence variations
A founder effect for the 35delG mutation has recently been described [27], which explains the variable frequency of this mutation in different populations rather than resulting from a mutational hot spot. Similarly, the 235delC mutation, frequent in the Japanese population, is derived from a common ancestor [34]. Founder effects may account for a number of other GJB2 sequence variations, whose frequencies depend on ethnic background (such as M34T [4]). The polymorphism 765C>T (referred to as SNP1 [27]) has been systematically included in our series. The 765T allele showed complete association with the 35delG as well as the ∆(GJB6-D13S1830). The genotype comparison between GJB2 and/or ∆(GJB6-D13S1830) patients with GJB2 negative subjects is significant with P-values < 10 -6 ( Table 3). These results are in accordance with the fact that ∆(GJB6-D13S1830) is absent in some populations and strongly suggest the existence of a founder effect as confirmed by a recent multicenter study [17].

Controversial effects of sequence variations
Among the sequence variations that have controversial consequences, a few variants have been extensively discussed based on families with deafness and on studies using in vitro expression systems. Since we have performed a population based study, in parallel with the analysis of a number of hearing impaired patients, we show that M34T, V37I and R127H represent common variants that are not responsible for severe or profound deafness.
The M34T variation was first described as a dominant mutation [35] and the dominant negative effect was supported by in vitro functional studies [36,37]. However, the description of normal hearing carriers abolished this hypothesis and furthermore, normal hearing patients were found to be compound heterozygotes, M34T/ 167delT or M34T/35delG [29,38]. Since the two alleles in trans of the M34T corresponded to a null allele, Griffith et al. (2000) suggested that the M34T was functional in vivo and therefore, the phenotypic consequences of the M34T allele would depend on the opposing CX26 allele variant [39]. The possibility of considering the M34T as a non pathogenic variant was also raised [40,41].
M34T represents the highest carrier rate in our population. We do not deal with a carrier rate of 1/116 as estimated from a small sample in Paris [29,41] but with 1/43 (2.3 %). We identified one M34T homozygote in the general population, as expected from the carrier frequency (1/ 4,444). The M34T is more frequent than the 35delG and, in contrast to the 35delG mutation, no M34T homozygote or in compound heterozygosity with a deleterious mutation was observed in our cohort of patients. This carrier rate is similar to the one observed by Green et al. [9]. The allele frequency of 1.15 %, based on 3,516 individuals, shows no significant discrepancy with the M34T allele frequency in the deaf population (1.6 %). Although a study based on the general population does not rule out the possibility of hearing deficiency in few individuals, these data eliminate the possibility of considering the M34T as a dominant or recessive mutation associated with severe or profound deafness.
The R127H, first described by Estivill [7] is also contentious and functional studies of this variant are inconsistent [42,43]. The frequency of carrier rate of in our region is 1/75 (1.33 %), not significantly different from that of the deaf population. Moreover, two normal-hearing parents were genotyped R127H/R127H.
Finally two normal-hearing sibs were compound heterozygotes R127H/M34T or R127H/W24X emphasizing the non pathogenic nature of this sequence variation. However, the genotype R127H/M34T was identified twice in our patient cohort (2/159 1.25 %) but never in the general population (odds to be associated randomly of 1.5 × 10 -4 ). The observed frequency in the patients is significant (p = 1.3 %), and therefore we still cannot rule out that the combined genotype R127H/M34T can act as a variant that, under certain circumstances (associated with other modifiers such as alterations in other deafness genes), would contribute to the phenotype.
Similarly, the V37I variation does not show any significant difference in frequency between the general and deaf populations. This is consistent with the fact that it was originally identified as a non pathogenic polymorphism because of its occurrence in a heterozygous state in the general population [40,44]. However, homozygosity for V37I and compound heterozygosity of V37I were often described in patients with NSHL suggesting that V37I acts as a recessive mutation. Recent studies [11,28,45,46] clearly indicated its pathogenicity when associated with another mutated GJB2 allele. Functional analyses [47] also showed that V37I is devoid of functional activity and thus may be pathogically significant.
The likelihood of M34T to be associated with V37I is 1 × 10 -4 , as calculated from the allele frequency, and is significantly different from the observed frequency in patients (p = 0.63 %), and once more can not rule out that the genotype M34T/V37I is not associated with mild or moderate deafness.
These data demonstrate the challenge of interpretating the association of two sequence variations. Moreover, combined genotypes with variants such as M34T, V37I, or R127H could have a phenotypic expression modulated by environmental factors or modifier genes. Conclusions from in vitro transfection assays can not be taken as complete because one sequence variant supposedly acting as a recessive mutation should be at least co-transfected with a second recessive mutation; but above all, one wonders whether the different cell types used for these in vitro experiments really reflect the molecular context existing in vivo. Complementary assays considering co-transfection of two different mutations need to be performed. These, together with compilation of observed combined genotypes correlated with phenotypes, should help in the interpretation of the molecular tests.
Phenotypes associated with GJB2/GJB6 mutations Although a recent report recommended GJB2 screening in cases of progressive and recurrent sudden HL [31]), no GJB2 and/or GJB6 biallelic mutations were identified in patients with progressive, postlingual, asymmetrical hearing loss in this study. The contribution of the GJB2/GJB6 genes, in our cohort is exclusively found in non syndromic, prelingual, bilateral, stable deafness and is about 33 % (28/84). This is similar to the proportion found in Spain (31.6 %) [28], in France (39.8 %) [48] and in Greece (33.3 %) [30]. However the degree of implication of GJB2/GJB6 genes in deafness depends on the composition in degree of severity of the patient group. In this study, the rate of DFNB1-associated deafness would be 41.6 % (25/60) if only congenital profound NSHL was considered.

Conclusions
The data presented here demonstrate that genetic counseling has been greatly improved with the identification of the ∆(GJB6-D13S1830) mutation and a total of 30 families could benefit from a GJB2 and/or GJB6 unambiguous molecular diagnosis. The genetic counseling was more difficult for 16 families (1/3) because of the compound heterozygosity of poorly-defined variations or the presence of a single GJB2 alteration. Additional data certainly need to be collected to evaluate if two sequence variations consid-ered as non-disease causing stay neutral when associated in trans.
The number of families (129) for which no genetic counseling could be provided, based on the mutation screening of the DFNB1 genes, remains very high and demonstrates that criteria for genetic testing must be very well defined and set according to the provided test (CX26). Genetic heterogeneity is still observed when prelingual, non syndromic, stable deafness is present, and the possibility for screening mutations in other genes should be offered.
Finally, neonatal screening programs bring a great improvement in the management of deafness; however, the option to offer molecular based screening remains very uncertain and if offered should be based on specific population studies because of the genetic heterogeneity of deafness.

Competing interests
There are no competing interests

Authors' contributions
FA, PB, GL and MD referred patients; NPR, AV, VF, CT and DL carried out the molecular studies; NM performed the statistical analyses; AFR supervised the whole study in the laboratory of MC.