Functional Testing of SLC26A4 Variants—Clinical and Molecular Analysis of a Cohort with Enlarged Vestibular Aqueduct from Austria

The prevalence and spectrum of sequence alterations in the SLC26A4 gene, which codes for the anion exchanger pendrin, are population-specific and account for at least 50% of cases of non-syndromic hearing loss associated with an enlarged vestibular aqueduct. A cohort of nineteen patients from Austria with hearing loss and a radiological alteration of the vestibular aqueduct underwent Sanger sequencing of SLC26A4 and GJB2, coding for connexin 26. The pathogenicity of sequence alterations detected was assessed by determining ion transport and molecular features of the corresponding SLC26A4 protein variants. In this group, four uncharacterized sequence alterations within the SLC26A4 coding region were found. Three of these lead to protein variants with abnormal functional and molecular features, while one should be considered with no pathogenic potential. Pathogenic SLC26A4 sequence alterations were only found in 12% of patients. SLC26A4 sequence alterations commonly found in other Caucasian populations were not detected. This survey represents the first study on the prevalence and spectrum of SLC26A4 sequence alterations in an Austrian cohort and further suggests that genetic testing should always be integrated with functional characterization and determination of the molecular features of protein variants in order to unequivocally identify or exclude a causal link between genotype and phenotype.


Introduction
To provide the best possible care for patients affected by hearing loss, and their families, the search for a possible genetic etiology of the disease is mandatory when environmental causes are likely to be excluded. Identification of the genetic cause of deafness provides valuable information for accurate diagnosis, reliable prognosis and precise genetic counseling, and can be essential for some forms of genetic deafness that may become amenable for treatment in the future [1]. Molecular diagnostic based on sequencing of candidate genes plays a major role in this context. However, hearing loss is genetically heterogeneous and, with the exception of syndromic cases, the genotype-phenotype

Function of SLC26A4 Protein Variants
The function of the four pendrin protein variants (p.M21V, p.Y115D, p.A434D and p.V577A) was determined based on the ability of transporting the iodide anion following heterologous expression in HEK 293 Phoenix cells (Figure 1).

Function of SLC26A4 Protein Variants
The function of the four pendrin protein variants (p.M21V, p.Y115D, p.A434D and p.V577A) was determined based on the ability of transporting the iodide anion following heterologous expression in HEK 293 Phoenix cells (Figure 1). Iodide influx measured in cells expressing pendrin p.Y115D, p.A434D or p.V577A was significantly reduced compared to that measured in cells expressing wild type pendrin and significantly higher than that measured in control cells, indicating that the ion transport function of these variants, although not completely lost, is compromised. The transport activity of pendrin p.M21V was indistinguishable from wild type. The results of the functional test are summarized in Table 3. Iodide influx measured in cells expressing pendrin p.Y115D, p.A434D or p.V577A was significantly reduced compared to that measured in cells expressing wild type pendrin and significantly higher than that measured in control cells, indicating that the ion transport function of these variants, although not completely lost, is compromised. The transport activity of pendrin p.M21V was indistinguishable from wild type. The results of the functional test are summarized in Table 3.  [26] based on functional and molecular features is given, and refers to a condition where a given variant is found in compound heterozygosity with another pathogenic variant. ER, endoplasmic reticulum.

Subcellular Localization of Pendrin Variants
To define the subcellular localization of the pendrin variants, their possible co-localization with the plasma membrane ( Figure 2) or ER ( Figure 3) was determined by confocal imaging following expression in HeLa cells.

Subcellular Localization of Pendrin Variants
To define the subcellular localization of the pendrin variants, their possible co-localization with the plasma membrane ( Figure 2) or ER ( Figure 3) was determined by confocal imaging following expression in HeLa cells.    Similarly to wild type pendrin, pendrin p.M21V co-localized with the plasma membrane and not with the ER. Conversely, pendrin p.Y115D and p.V577A did not reach the plasma membrane, but did co-localize with the ER. Pendrin p.A434D co-localized with both the plasma membrane and ER. The results of the subcellular localization studies are summarized in Table 3.

Total and Plasma Membrane Expression Levels of Pendrin Variants
Expression levels of pendrin variants were determined in the whole cell and in the region of the plasma membrane by confocal imaging following expression in HeLa cells. The total expression levels of all of the functionally impaired pendrin variants (p.Y115D, p.A434D and p.V577A) were substantially reduced with respect to those of the wild type. In contrast, the total expression levels of the fully functional pendrin variant p.M21V were not significantly different from the wild type ( Figure 4). Similarly to wild type pendrin, pendrin p.M21V co-localized with the plasma membrane and not with the ER. Conversely, pendrin p.Y115D and p.V577A did not reach the plasma membrane, but did co-localize with the ER. Pendrin p.A434D co-localized with both the plasma membrane and ER. The results of the subcellular localization studies are summarized in Table 3.

Total and Plasma Membrane Expression Levels of Pendrin Variants
Expression levels of pendrin variants were determined in the whole cell and in the region of the plasma membrane by confocal imaging following expression in HeLa cells. The total expression levels of all of the functionally impaired pendrin variants (p.Y115D, p.A434D and p.V577A) were substantially reduced with respect to those of the wild type. In contrast, the total expression levels of the fully functional pendrin variant p.M21V were not significantly different from the wild type ( Figure 4).  Consistent with previous observations [27], expression levels of pendrin variants in the plasma membrane region determined on a single-cell level ( Figure 5) mirrored total expression levels. Expression levels of pendrin variants are summarized in Table 3. Consistent with previous observations [27], expression levels of pendrin variants in the plasma membrane region determined on a single-cell level ( Figure 5) mirrored total expression levels. Expression levels of pendrin variants are summarized in Table 3.

Population Frequencies of c.343T>G and c.1301C>A
The presence of the two (c.343T>G and c.1301C>A) exonic SLC26A4 sequence variations was assessed in 125 healthy volunteers with no self-reported hearing disorders using custom designed genotyping assays. The genotyping results for patient 616 (heterozygous for c.343T>G) and patient 271 (heterozygous for c.1301C>A) and all other patients outlined in Table 2 (wild-type at both positions) matched that from the Sanger sequencing results, indicating that both custom genotyping assays worked correctly ( Figure 6). All of the normal hearing control samples were homozygous wild-type for both sequence variations. The frequency of both sequence variants was 0.69% and each fulfilled the Hardy-Weinberg equilibrium expectation.

Population Frequencies of c.343T>G and c.1301C>A
The presence of the two (c.343T>G and c.1301C>A) exonic SLC26A4 sequence variations was assessed in 125 healthy volunteers with no self-reported hearing disorders using custom designed genotyping assays. The genotyping results for patient 616 (heterozygous for c.343T>G) and patient 271 (heterozygous for c.1301C>A) and all other patients outlined in Table 2 (wild-type at both positions) matched that from the Sanger sequencing results, indicating that both custom genotyping assays worked correctly ( Figure 6). All of the normal hearing control samples were homozygous wild-type for both sequence variations. The frequency of both sequence variants was 0.69% and each fulfilled the Hardy-Weinberg equilibrium expectation.

Discussion
A great diversity of clinical manifestations, in particular audiometric characteristics of hearing loss and vestibular symptoms, typically reported among patients with EVA [28][29][30], was confirmed in our cohort of 19 patients from Austria, including one patient with agenesis of the vestibular aqueduct on both sides (Table 1, S1 and S2). An intraoperative gusher phenomenon of cerebro-spinal fluid during cochlea implant surgery was observed in 7 of 23 operated ears (30.4%) (Table S1), an incidence that is relatively high compared to other groups [31,32]. However, the condition was manageable in all cases by accurate sealing during surgery without any postoperative leakage observed. The incidence was independent from SLC26A4 genotype.
A total of 12 of 19 patients (63.2%), including the patient with agenesis of the vestibular aqueduct, complained of dizziness or vertigo to a certain extent (Table S2). This is in agreement with other reports [29,33]. However, Zalewski et al. [29] concluded that, although vestibular symptoms are commonly found in EVA patients, a causative relation cannot be identified. Vestibular disorders, such as Menière's disease or vestibular migraine, need to be considered in cases of reported episodic or progressive dizziness or vertigo. Clinical diagnosis of these vestibular disorders is based on clinical findings defined by Lopez-Escamez et al. [34] for Menière's disease and Lempert et al. [35] for vestibular migraine. None of our patients fulfilled the diagnostic criteria for any of those two diagnoses. None of the patients had a low frequency sensorineural hearing loss. Reported vestibular symptoms did not appear in repeated attacks accompanied by cochlear symptoms, such as tinnitus or aural fullness. Only one patient (patient ID 395) reported headache but not in connection with episodic vertigo and without any sensation of an aura (Table S2). Caloric testing and head impulse testing results are reported in Table S2. Vestibular diagnosis in patients undergoing cochlear implant at our department is done prior to surgery in patients older than 15 years. Therefore, data on vestibular function are not available for all patients. Unfortunately, there is no objective data on vestibular function of patient 271 (bearing biallelic sequence alterations in SLC26A4) due to young age. However, parents did not report difficulties in motor development, giving a clinical hint on adequate vestibular function. Patient 358 reported episodic dizziness and showed symmetric reduction of excitability during caloric testing on both sides. Patient 616 did not show signs of vestibular dysfunction with physiological caloric testing and head impulse test.
Two previously reported GJB2 variants (c.35delG and c.88A>G) were found in our cohort ( Table  2). The GJB2 variant c.35delG, leading to premature protein chain termination, was unequivocally established by numerous studies as pathogenic [36,37] and represents the most common cause of hearing loss worldwide [24]. In our cohort, the variant c.35delG was detected in homozygosity in two

Discussion
A great diversity of clinical manifestations, in particular audiometric characteristics of hearing loss and vestibular symptoms, typically reported among patients with EVA [28][29][30], was confirmed in our cohort of 19 patients from Austria, including one patient with agenesis of the vestibular aqueduct on both sides (Table 1, Tables S1 and S2). An intraoperative gusher phenomenon of cerebro-spinal fluid during cochlea implant surgery was observed in 7 of 23 operated ears (30.4%) (Table S1), an incidence that is relatively high compared to other groups [31,32]. However, the condition was manageable in all cases by accurate sealing during surgery without any postoperative leakage observed. The incidence was independent from SLC26A4 genotype.
A total of 12 of 19 patients (63.2%), including the patient with agenesis of the vestibular aqueduct, complained of dizziness or vertigo to a certain extent (Table S2). This is in agreement with other reports [29,33]. However, Zalewski et al. [29] concluded that, although vestibular symptoms are commonly found in EVA patients, a causative relation cannot be identified. Vestibular disorders, such as Menière's disease or vestibular migraine, need to be considered in cases of reported episodic or progressive dizziness or vertigo. Clinical diagnosis of these vestibular disorders is based on clinical findings defined by Lopez-Escamez et al. [34] for Menière's disease and Lempert et al. [35] for vestibular migraine. None of our patients fulfilled the diagnostic criteria for any of those two diagnoses. None of the patients had a low frequency sensorineural hearing loss. Reported vestibular symptoms did not appear in repeated attacks accompanied by cochlear symptoms, such as tinnitus or aural fullness. Only one patient (patient ID 395) reported headache but not in connection with episodic vertigo and without any sensation of an aura (Table S2). Caloric testing and head impulse testing results are reported in Table S2. Vestibular diagnosis in patients undergoing cochlear implant at our department is done prior to surgery in patients older than 15 years. Therefore, data on vestibular function are not available for all patients. Unfortunately, there is no objective data on vestibular function of patient 271 (bearing biallelic sequence alterations in SLC26A4) due to young age. However, parents did not report difficulties in motor development, giving a clinical hint on adequate vestibular function. Patient 358 reported episodic dizziness and showed symmetric reduction of excitability during caloric testing on both sides. Patient 616 did not show signs of vestibular dysfunction with physiological caloric testing and head impulse test.
Two previously reported GJB2 variants (c.35delG and c.88A>G) were found in our cohort ( Table 2). The GJB2 variant c.35delG, leading to premature protein chain termination, was unequivocally established by numerous studies as pathogenic [36,37] and represents the most common cause of hearing loss worldwide [24]. In our cohort, the variant c.35delG was detected in homozygosity in two unrelated Caucasian patients with bilateral EVA and no SLC26A4 sequence alterations, and can reasonably be considered as the genetic determinant of deafness in these individuals ( Table 2).
The GJB2 variant c.88A>G leads to the amino acid substitution p.I30V, and was first detected as a single allelic mutation in one patient of a cohort of 324 Taiwanese individuals with prelingual deafness [38]. To our knowledge, no functional or molecular tests for the GJB2 variant p.I30V have been performed. However, bioinformatic predictions indicate that this amino acid change may be tolerated, and the corresponding protein variant may have no pathogenic potential [39]. In our cohort, the c.88A>G GJB2 variant was found in two siblings of Arab descent (patient ID 267 and 308, Tables 1 and 2) with bilateral EVA, renal tubular acidosis (RTA) and no sequence alterations on the other GJB2 allele. Whether monoallelic GJB2 sequence alterations can be sufficient to determine deafness is currently unclear [37,40]. Importantly, mutations in ATP6V1B1, which encodes the B1-subunit of H + -ATPase, can cause RTA, early onset hearing loss and bilateral EVA [41] and may reasonably account for the pathological phenotype encountered in these two subjects.
Although the link between GJB2 mutations and temporal bone anomalies is not clearly established [42,43], GJB2 variants can be detected in EVA patients with variable prevalence depending on the specific population and the criterion adopted to define EVA [43][44][45]. In our cohort, both patients with biallelic GJB2 mutations (patient ID 307 and 421) presented bilateral EVA, thus supporting a link between GJB2 dysfunction and inner ear malformations. It is possible, however, that the homozygous variant c.35delG determines non-syndromic hearing loss in these patients, which also manifest EVA as the consequence of undetected pathogenic sequence alterations in SLC26A4 or other genes. Of note, both patients show signs of vestibular dysfunction (Table S2) and patient 307 presents the most severe hearing loss phenotype (bilateral complete absence of hearing function) observed in this cohort (Table S1).
Four SLC26A4 variants were found in this cohort ( Table 2); for two of them (c.61A>G, SNP ID rs375716219, and c.1730T>C, SNP ID rs56017519) an SNP ID was assigned, however, functional characterization is missing (according to: https://www.ncbi.nlm.nih.gov/snp/, the clinical significance was not assessed for the former and uncertain for the latter, respectively). The other two SLC26A4 variants (c.343T>G and c.1301C>A) are not reported in the SNP database. To assess their pathogenic potential, the functional and molecular features of the corresponding protein variants were assessed by in vitro studies, and the protein variants were subsequently classified according to the American College of Medical Genetics and Genomics Standards and Guidelines [26] (Table 3).
Sequence alterations c.1301C>A and c.1730T>C encoding the pendrin protein variants p.A434D and p.V577A were found in compound heterozygosity in a patient with congenital hearing loss, incomplete partition 2 (IP2) and no sequence alterations in GJB2 (patient ID 271, Tables 1 and 2). The former sequence alteration was inherited from the mother and the latter from the father, both of whom had normal hearing. Variant p.A434D showed a 16% reduction of iodide transport activity ( Figure 1 and Table 3). Following heterologous expression, this variant co-localized with both the plasma membrane ( Figure 2) and the ER (Figure 3). The total cellular and plasma membrane expression levels of this variant appear substantially reduced compared to the wild type (Figures 4 and 5), although residual expression (p < 0.001 compared to p.V577A, Figure 4) and partial plasma membrane targeting ( Figure 2) allowed for residual transport activity (Figure 1). The functional and molecular defects of variant p.V577A appeared more severe in terms of loss in ion transport function (76% reduction, Figure 1 and Table 3), complete retention in the ER (Figures 2 and 3) and dramatic reduction of expression (Figures 4 and 5). We conclude that both SLC26A4 variants show functional and molecular features compatible with the pathological phenotype of the patient ( Table 3). Evidence of thyroid malfunction due to pendrin dysfunction may unequivocally be obtained through perchlorate discharge testing [22]. Assessing the thyroid function by perchlorate testing in this patient was not possible; therefore, definite discrimination between non-syndromic EVA and Pendred syndrome could not be ascertained. Ultrasound of the thyroid gland showed no signs of pathologic volume changes. To date, the patient is eight years old, and manifest goiter may appear in the second decade of life [46,47]. A clinical follow-up will be performed including perchlorate testing when possible.
The SLC26A4 variant c.343T>G was found in heterozygosity with the wild type allele in an individual with fluctuating hearing loss, bilateral EVA and no GJB2 mutations (patient ID 616, Tables 1  and 2). The corresponding protein variant p.Y115D showed substantially reduced activity and expression levels (Figures 1, 4 and 5 and Table 3), loss of targeting at the plasma membrane and retention within the ER (Figures 2 and 3). The functional and molecular features of this variant are consistent with a pathogenic potential when another dysfunctional allele is present (Table 3). Whether monoallelic pendrin mutations are sufficient to determine a pathological phenotype is currently unclear. Some studies support the hypothesis of a second, undetected SLC26A4 mutation that accounts for EVA in these cases [48,49]. For example, sequence alterations within the promoter or cis-regulatory (enhancer) regions leading to a reduced transcription may per se be insufficient to determine a pathological phenotype, but can be pathogenic in combination with another dysfunctional allele. These sequence alterations may reside in upstream regions distant from SLC26A4, intergenic regions or even other genes [50]. For this patient, the only other sequence alteration in SLC26A4 with an allele frequency <1%, of which the clinical significance was not assessed, was found in the 3'UTR (g.55910G>A, rs375630290, Table S3). Therefore, the genetic cause of deafness for this patient remains undetermined.
For 7 of 19 patients (6 of 18 families, 33%) a family history of hearing loss was recorded (Table S1). The pedigree of these families is presented in Figure S1. No candidate causative variants were identified for these familial cases (as discussed above, the GJB2 variant c.88A>G found in patients 267 and 308 and the SLC26A4 variant c.61A>G found in patient 358 are most likely with no pathogenic potential).
PS/non-syndromic EVA have also been associated with monoallelic sequence alterations in the SLC26A4 and FOXI1 or KCNJ10 genes, with a double-heterozygosity inheritance model [51,52]. FOXI1 (OMIM #601093) is a transcription factor for SLC26A4 [52,53], and KCNJ10 (OMIM #602208) is an inwardly rectifying K + channel essential for the maintenance of the endocochlear potential [54]. Although later studies showed that sequence alterations in FOXI1 or KCNJ10 are rare in the context of PS/non-syndromic EVA [55][56][57][58], analysis of these genes may be of value in individuals with monoallelic pathogenic SLC26A4 sequence alterations.
SLC26A4 variants with no reduction in ion transport function are found in the general [59,60], deaf with no EVA [59,60] and deaf with EVA [27,[61][62][63][64] populations. Including these variants into genotype-phenotype correlation analyses may lead to false conclusions [65]. Therefore, unequivocal assessment of pathogenic potential of a SLC26A4 sequence alteration requires functional studies of the corresponding protein variant. In our cohort, the SLC26A4 variant c.61A>G was found in heterozygosity with the wild type allele in an individual with profound hearing loss, bilateral IP2 and no GJB2 mutations (patient ID 358, Tables 1 and 2). The corresponding protein variant p.M21V did not show alterations in function (Figure 1), subcellular localization (Figures 2 and 3) or expression levels (Figures 4 and 5) compared to the wild type, and should therefore be considered as having no pathogenic potential (Table 3). No additional sequence alterations with an allele frequency <1% were found in SLC26A4 in this patient (Table S3), therefore pendrin dysfunction most likely does not account for the hearing loss in this patient.
In Caucasian cohorts, approximately 50% of all individuals with unilateral or bilateral hearing loss and EVA carry at least one mutant SLC26A4 allele [22,23,66]. For example, in a cohort of 100 unrelated French individuals with EVA, SLC26A4 mutations (either monoallelic or biallelic) were found in 40% of subjects [67]. In our cohort of 17 unrelated individuals with EVA, only 2 (12%) presented at least one pathogenic SLC26A4 allele (p = 0.02 compared to the above mentioned French cohort). Therefore, SLC26A4 mutations seem to be under-represented in our cohort, thus indicating that additional genes may be involved in determining EVA in the Austrian population.
The spectrum of SLC26A4 variants is strictly related to the ethnicity of patients. The c.1001+1G>A, p.V138F, p.T416P, p.L236P and p.G209V pathogenic variants are prevalent in the Caucasian population, with p.V138F being predominant in Germany and Czech Republic, two countries geographically close to Austria [24]. None of these mutations were found in our cohort. Indeed, a novel pathogenic variant (p.Y115D) was detected. The novel pathogenic variant was not detected in 125 population-matched healthy control individuals ( Figure 6). In addition, the previously reported SLC26A4 variants found in our cohort (p.M21V and p.V577A) are not frequently found in the general population-according to: https://www.ncbi.nlm.nih.gov/snp/, the minor allele frequency (MAF) is 0.0042% for c.61A>G (p.M21V), SNP ID: rs375716219, and 0.001647% for c.1730T>C (p.V577A), SNP ID: rs56017519, respectively (Table S4)-nor in deaf individuals [24]. The MAF of all four exonic SLC26A4 variants (Table S4) is lower than 0.005, a threshold value for pathogenic autosomal-recessive variants suggested by Shearer et al. [68]. These observations underscore that SLC26A4 variants found in our cohort, including the variant characterized as non-pathogenic in this work, are uncommon in the general population. Therefore, in terms of prevalence and spectrum of SLC26A4 mutations, the Austrian population seems to present characteristics distinct from those of other Caucasian-European groups. Expansion of the cohort subject of this study will be essential to establish if the pathogenic SLC26A4 allelic variants more commonly found in Caucasians are either rare or absent in the Austrian population.

Patients
Eighteen Austrian subjects (10 females and 8 males aged between 8 and 51 years; average age 27 years) diagnosed with sensorineural (n = 16), mixed (n = 1) or conductive (n = 1) hearing loss and EVA, as well as one subject with agenesis of the vestibular aqueduct (patient ID 365) were included in the study (Table 1). Of these, two (patient ID 267 and 308) were siblings of Arab origin. All the other patients were unrelated subjects of Caucasian, or mixed Caucasian and African (patient ID 395) ethnicity. The research was prospectively reviewed and approved by a duly constituted ethics committee (415-E/2092/6-2017, 9 May 2017) and has therefore been performed in accordance with the principles embodied in the 1964 Declaration of Helsinki and its later amendments (Available online: https://www.wma.net/policies-post/wma-declaration-of-helsinkiethical-principles-for-medical-research-involving-human-subjects/). Written informed consent was obtained from all subjects or their legal representatives prior to blood sampling and genetic testing. For all patients, imaging studies of the inner ear by computer tomography (CT) of the temporal bones were performed. EVA was defined as an enlargement of the vestibular aqueduct ≥1.5 mm midway between the endolymphatic sac and the vestibule. An abnormal cochlea was considered as Mondini malformation in cases of a normal basal turn and cavity-like appearing distal turns with a missing interscalar ridge between the basal turn and the distal turns on the axial plane of the CT scan [69]. IP2 was defined as the combination of EVA and Mondini malformation [70]. None of the patients presented overt goiter or alterations of clinical or laboratory thyroid functional parameters. Individual characterization of hearing loss (HL) was based on side-specific pure-tone audiometric testing results, as well as the patient's history. The type of HL was defined as conductive, mixed or sensorineural based on air-and bone-conduction thresholds. The severity of HL was classified on decibel hearing level (dB HL) thresholds, either as mild for values between 26 and 40 dB HL, moderate between 41 and 60 dB HL, severe between 61 and 80 dB HL or profound for values above 80 dB HL. Ears with no perception at all were considered as deaf. Frequencies possibly affected by HL were divided into low, middle or high frequency range. Onset of HL was defined through patient's history, since newborn screening tests were either not performed or results were not available. Related to the time of speech development, the onset of HL was defined as congenital, prelingual, perilingual or postlingual. Clinical course of HL with possible hearing drops was obtained through audiometric tests and patient's history. Surgical reports were screened for intraoperative gusher phenomenon, defined as visible efflux of cerebro-spinal fluid after cochleostomy, during cochlear implant surgery.

Genomic DNA Samples
Patient whole blood was collected in plastic tubes with potassium-ethylenediaminetetraacetic acid (S-Monovette ® , Sarstedt, Nümbrecht, Germany) via venipuncture. Total genomic DNA (gDNA) was purified from~350 µL blood with the EZ1 DNA Blood 350 µL kit (Qiagen, Hilden, Germany) using the EZ1 Advanced XL platform (Qiagen) according to the manufacturer's instructions. Quantification was performed with the QIAxpert (Qiagen) spectrophotometer. Only samples with an A260/A280 between 1.7 and 1.9 were used for downstream analysis.  Table S5.

Genomic DNA Analysis
The entire coding region of GJB2 (exon 2) was amplified by PCR as described above using the following primers: forward, 5 CACGTTCAAGAGGGTTTG 3 , and reverse, 5 TGAGCCTTGA CAGCTGAGC 3 . The same primers were used for sequencing.

Plasmid Constructs
The pTARGET (Promega Corporation, Madison, WI, USA) vector, containing the cDNA of wild type or mutated human pendrin [9], was used for functional tests.
The pEYFPN1 vector (Clontech, Mountain View, CA, USA), containing the cDNA of wild type or mutated human pendrin, was used for co-localization and determination of pendrin expression levels via imaging. Following transfection of this construct in cells, pendrin is produced with the enhanced yellow fluorescent protein (EYFP) fused to its C-terminus [27].
Sequence alterations in the cDNA of pendrin were obtained using the QuikChange ® site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA) and the primers listed in the Table S6.
The sequence of all plasmid inserts was verified by Sanger sequencing (Microsynth AG).

Pendrin Functional Test
For testing the function of pendrin variants, the influx of iodide was measured in cells expressing wild type or mutated pendrin and in control cells. The functional test was performed as already described [9,59,72,73], with adaptations allowing for the use of a multiplate reader [27,60,[74][75][76]. Shortly, cells were seeded into black 96-well plates, grown overnight and co-transfected with 0.1 µg of a plasmid encoding for EYFP H148Q;I152L (an EYFP variant with substantially improved sensitivity for iodide [77]) and 0.1 µg of pTARGET plasmid bearing the cDNA of wild type or mutated pendrin. Control cells were co-transfected with EYFP H148Q;I152L and the empty pTARGET vectors. Transfection was done by the calcium phosphate co-precipitation method. Experiments were performed at room temperature 48 h post-transfection. Cells were initially bathed in 70 µL of high chloride solution (in mM: KCl 2, NaCl 135, CaCl 2 1, MgCl 2 1, D-glucose 10, 4-(2-hydroxyethyl)-1-piperazineethane sulfonic acid (HEPES) 20, 308 mOsm/KgH 2 O with mannitol, pH 7.4), and the baseline fluorescence intensity was measured (1 measurement/s for 3 s). Subsequently, 140 µL of high iodide solution (in mM: KCl 2, NaI 135, CaCl 2 1, MgCl 2 1, D-glucose 10, HEPES 20, 308 mOsm/KgH 2 O with mannitol, pH 7.4) were injected into each well and the fluorescence intensity was measured again (1 measurement/s for 16 s). Owing on the chloride/iodide exchanger activity of pendrin [9][10][11], iodide enters the cytosol of pendrin-expressing cells from the extracellular medium and leads to a decrease of EYFP H148Q;I152L fluorescence. Therefore, negative % fluorescence variations (∆F%) indicate a flux of iodide from the extracellular to the intracellular milieu and reflect pendrin transport efficiency.
Subcellular localization of pendrin variants was determined by co-localization between wild type or mutant SLC26A4 with EYFP fused to the C-terminus and markers of the plasma membrane (CellMask TM Deep Red plasma membrane stain, C10046, Invitrogen Molecular Probes, Waltham, MA, USA) or endoplasmic reticulum (ER) (ER-Tracker TM Red, BODIPY TM TR glibenclamide, E34250, Invitrogen Molecular Probes).
To stain the plasma membrane, living cells were washed thrice with ice-cold Hank's balanced salt solution (HBSS, Sigma-Aldrich), incubated on ice for 5 min with 1.25 µg/mL CellMask TM Deep Red plasma membrane stain in HBSS, washed again thrice with ice-cold HBSS and imaged immediately in HBSS.
To stain the ER, living cells were washed thrice with room temperature HBSS, incubated for 20 min at 37 • C and 5% CO 2 with 1 µM ER-Tracker TM Red in Krebs-Henseleit buffer (Sigma-Aldrich), washed again thrice and immediately imaged in HBSS at room temperature.
Co-localization was detected and quantified as previously described [27]. Shortly, imaging was performed by sequential acquisition with a Leica TCS SP5II AOBS confocal microscope (Leica Microsystems, Wetzlar, Germany) equipped with a HCX PL APO 63×/1.20 Lambda blue water immersion objective and controlled by the LAS AF SP5 software (Leica Microsystems). For co-localization with the plasma membrane, EYFP was excited with the 514 nm line of the Argon laser and emission was detected in the 525-600 nm range; the CellMask TM Deep Red stain was excited at 633 nm (Helium Neon laser) and emission was detected in the 655-750 nm range. For co-localization with the ER, EYFP was excited with the 514 nm line of the Argon laser and emission was detected in the 525-555 nm range; ER-Tracker TM Red was excited at 561 nm (diode-pumped solid-state (DPSS) laser) and emission was detected in the 571-650 nm range. Co-localization with plasma membrane or ER markers was quantified and expressed as the Pearson's correlation coefficient [78], overlap coefficient and co-localization rate. The co-localization parameters of wild type pendrin were considered as indicators of a preferential co-localization with the plasma membrane and poor co-localization with the ER.

Determination of Wild Type and Mutant Pendrin Total Expression Levels by Imaging
Determination of total expression levels of pendrin variants was performed as previously described [27]. Shortly, HeLa cells expressing wild type or mutant SLC26A4 with EYFP fused to the C-terminus were fixed with 4% paraformaldehyde for 30 min, counterstained with 0.1 µg/mL 4 ,6-diamidino-2-phenylindole (DAPI) for 10 min, washed and imaged in HBSS. Wild type or mutant SLC26A4 expression levels, given by the fluorescence intensity of EYFP, were normalized for the cell density, given by the fluorescence intensity of DAPI. Imaging was performed by confocal microscopy as described above. EYFP was excited with the 514 nm line of the Argon laser and emission was detected between 525 and 600 nm; DAPI was excited with a diode laser (405 nm) and emission was detected between 430 and 470 nm. Laser power and photomultipliers gain were kept rigorously constant for acquisition of all images.

Determination of Wild Type and Mutant Pendrin Expression Levels in the Plasma Membrane Region
Determination of expression levels of pendrin variants in the region of the plasma membrane was performed as previously described [27]. Shortly, HeLa cells expressing wild type or mutant SLC26A4 with EYFP fused to the C-terminus and the enhanced cyan fluorescent protein (ECFP) (pECFPC1 vector, Clontech) were stained with CellMask TM Deep Red plasma membrane stain, washed and imaged in HBSS. EYFP fluorescence intensity of three regions of interest of the plasma membrane of a single cell was corrected for the background fluorescence, averaged and normalized for the background-subtracted ECFP fluorescence intensity measured in the cytosol of the same cell, to obtain wild type or mutant SLC26A4 expression levels normalized for the transfection efficiency of the single cell. Imaging was performed by confocal microscopy as described above. EYFP was excited with the 514 nm line of the Argon laser and emission was detected between 525 and 580 nm. ECFP was excited with a diode laser (405 nm) and emission was detected between 450 and 490 nm. CellMask TM Deep Red stain was excited at 633 nm (HeNe laser) and emission was detected in the 643-750 nm range. Laser power and photomultipliers gain were kept rigorously constant for acquisition of all images.

Sequence Variation Genotyping Assays
Ten µL end-point genotyping reactions containing 1× TaqMan ® Universal PCR Master Mix no AmpErase ® UNG, 1× hydrolysis probe assay and 13.5 ng gDNA were performed in MicroAmp ® Optical 384-well Reaction Plates on the QuantStudio™ 12K Flex Real-Time PCR System (all from Thermo Fisher Scientific). Cycling parameters were: 60 • C for 30 s, 95 • C for 10 min., followed by 50 cycles of 92 • C for 15 s and 60 • C for 90 s, and a final 30 s at 60 • C. The two hydrolysis probe assays used in the 384-well format targeted SLC26A4 c.343T>G and c.1301C>A and were custom designed and manufactured by Thermo Fisher Scientific.

Salts and Chemicals
All salts and chemicals used were of pro analysis grade.

Statistical Analysis
All data are expressed as arithmetic means ± SEM and analyzed with GraphPad Prism software (version 4.00 for Windows, GraphPad Software, San Diego, CA, USA). Significant differences between data sets were verified by one way Analysis of Variance (ANOVA) with Bonferroni's or Dunnet's post-tests, or by the Fisher exact test, as appropriate. A p value < 0.05 was considered as statistically significant; (n) corresponds to the number of independent measurements. To determine departure from Hardy-Weinberg equilibrium, a chi-squared test with one degree of freedom was used.

Conclusions
To conclude, we presented the first characterization of a cohort of Austrian deaf patients with EVA. Biallelic GJB2 and SLC26A4 sequence variations with functional impact could be identified as the genetic determinant of deafness for two and one patient, respectively. Prevalence and spectrum of SLC26A4 pathogenic variants are distinct from other Caucasian cohorts. Specifically, SLC26A4 pathogenic variants seem to be under-represented in this cohort, thus providing an imperative towards the investigation of additional genes and/or genomic regions that may be involved in determining deafness and EVA. High throughput sequencing technologies, such as whole exome sequencing, would be of great value in this context. Further, this study underscores the importance of integrating genetic analysis, functional testing and identification of molecular features of protein variants to unambiguously identify or exclude the genetic cause of deafness.
Supplementary Materials: Supplementary materials can be found at www.mdpi.com/1422-0067/19/1/209/s1. Acknowledgments: This work was supported in part by the Paracelsus Medical University Research Fund-PMU FFF: #L-16/04/005-ROS to Sebastian Roesch. We would sincerely like to thank the participants in this study and their families. Also, we gratefully acknowledge the expert secretarial assistance of Elisabeth Mooslechner, as well as laboratory assistance of Michaela Zopf and Markus Kwik.