|Home | About | Journals | Submit | Contact Us | Français|
Hearing loss with enlargement of the vestibular aqueduct (EVA) can be associated with mutations of the SLC26A4 gene encoding pendrin, a transmembrane Cl−/I−/HCO3− exchanger. Pendrin’s critical transport substrates are thought to be I− in the thyroid gland and HCO3− in the inner ear. We previously reported that bi-allelic SLC26A4 mutations are associated with Pendred syndromic EVA whereas one or zero mutant alleles are associated with nonsyndromic EVA. One study proposed a correlation of nonsyndromic EVA with SLC26A4 alleles encoding pendrin with residual transport activity. Here we describe the phenotypes and SLC26A4 genotypes of 47 EVA patients ascertained since our first report of 39 patients. We sought to determine the pathogenic potential of each variant in our full cohort of 86 patients. We evaluated the trafficking of 11 missense pendrin products expressed in COS-7 cells. Products that targeted to the plasma membrane were expressed in Xenopus oocytes for measurement of anion exchange activity. p.F335L, p.C565Y, p.L597S, p.M775T, and p.R776C had Cl−/I− and Cl−/HCO3− exchange rate constants that ranged from 13 to 93% of wild type values. p.F335L, p.L597S, p.M775T and p.R776C are typically found as mono-allelic variants in nonsyndromic EVA. The high normal control carrier rate for p.L597S indicates it is a coincidentally detected nonpathogenic variant in this context. We observed moderate differential effects of hypo-functional variants upon exchange of HCO3− versus I− but their magnitude does not support a causal association with nonsyndromic EVA. However, these alleles could be pathogenic in trans configuration with a mutant allele in Pendred syndrome.
Enlargement of the vestibular aqueduct (EVA; MIM# 600791) is the most common radiologic malformation of the inner ear associated with sensorineural hearing loss (Valvassori and Clemis, 1978). EVA is a fully penetrant feature of Pendred syndrome (PS; MIM# 274600) (Phelps, et al., 1998), an autosomal recessive disorder comprising bilateral hearing loss, EVA with or without other malformations of the inner ear, and an iodine organification defect in the thyroid gland, which may lead to goiter. Goiter is an incompletely penetrant feature of PS and goiter phenocopies due to other etiologies are common. The perchlorate discharge test (PDT) can detect the iodine organification defect in PS, but is considered by some to be insensitive and nonspecific (Borck, et al., 2003). We have found the PDT to be sensitive and specific when it is administered precisely and the results are carefully interpreted (Madeo, et al., 2006; Pryor, et al., 2005b; Pryor, et al., 2003).
Mutations of the SLC26A4 gene (alias PDS; MIM# 605646) cause PS (Everett, et al., 1997) and are a comparatively common cause of childhood deafness among many populations throughout the world (Park, et al., 2003). SLC26A4 encodes pendrin, a polytopic transmembrane protein that can exchange a variety of anions including HCO3−, Cl−, I−, and formate across the plasma membrane (Lopez-Bigas, et al., 2001; Lopez-Bigas, et al., 2002; Scott, et al., 1999; Taylor, et al., 2002; Tsukamoto, et al., 2003). Its in vivo transport mechanisms are thought to include Cl−/I− exchange in the thyroid gland (Royaux, et al., 2000) and Cl−/HCO3− exchange in the inner ear (Wangemann, et al., 2007). SLC26A4 mutations are also detected in patients with nonsyndromic EVA (NSEVA) (Li, et al., 1998; Usami, et al., 1999), leading some to conclude that PS and NSEVA are variable manifestations of the same underlying disorder (Azaiez, et al., 2007; Tsukamoto, et al., 2003).
Scott et al. (Scott, et al., 2000) explored the basis for phenotypic variability by measuring anion influx activities for selected missense pendrin variants expressed in Xenopus oocytes. They concluded that PS variants were functional null alleles whereas NSEVA alleles were hypomorphic alleles. They proposed that normal thyroid function in NSEVA may be the consequence of residual pendrin activity encoded by the NSEVA variants. This hypothesis appeared to be inconsistent with subsequent reports of common EVA variants associated with both PS and NSEVA (Lopez-Bigas, et al., 2001; Lopez-Bigas, et al., 2002; Taylor, et al., 2002; Tsukamoto, et al., 2003).
Later studies indicated that the EVA phenotype is correlated with the number, not type, of variant alleles of SLC26A4. Among a cohort of 39 EVA subjects, we identified a strong correlation of PS with bi-allelic SLC26A4 mutations, whereas NSEVA was associated with one or zero mutations of SLC26A4 (Pryor, et al., 2005b). Azaiez et al. (2007) and Pera et al. (2008) observed the same correlation in their cohorts. We also observed bi-allelic SLC26A4 mutations only in bilateral EVA whereas zero or mono-allelic mutations were associated with either uni- or bilateral EVA (Pryor, et al., 2005b). Similarly, Albert et al. (2006) and Madden et al. (2007) reported correlations of hearing loss severity with the number of mutant alleles of SLC26A4.
In western and European populations, SLC26A4 mutations cannot be detected in about one third of patients with EVA, whereas only one mutant SLC26A4 allele is identified in another third (Albert, et al., 2006; Campbell, et al., 2001; Coyle, et al., 1998; Pryor, et al., 2005b). Discordant segregation of NSEVA with SLC26A4-linked markers in some families indicates the existence of other genetic or environmental factors underlying this disorder (Pryor, et al., 2005b). This etiologic heterogeneity of NSEVA confounds the interpretation of rare SLC26A4 variants with uncertain effects upon expression or function. Some of these variants, such as those reported by Scott et al. (2000), may be detected in NSEVA but not PS. Do these variants represent benign polymorphic variation or are they causally correlated with NSEVA? Perhaps they have differential effects upon transport of critical substrates in the inner ear (HCO3−) versus the thyroid (I−). It has also been suggested the pathogenic potential of hypo-functional SLC26A4 variants may depend upon the allele in trans configuration (Scott, et al., 2000; Taylor, et al., 2002; Tsukamoto, et al., 2003).
Here we describe 47 previously unreported EVA patients and four novel variants of SLC26A4. We sought to define the pathogenic potential of these novel SLC26A4 variants and all other alleles that we detected among our entire cohort of 86 EVA patients. Our data continue to support a strong correlation of phenotype with number of mutant alleles, but a causal correlation of hypo-functional SLC26A4 variants with NSEVA is unlikely.
This study was approved by the Combined Neuroscience (CNS) Institutional Review Board (National Institutes of Health, Bethesda, Maryland). We obtained written informed consent for all subjects. We defined EVA as previously described (Pryor, et al., 2005b). Our subjects comprised 47 individuals with EVA and their unaffected relatives from 41 families. There were 6 multiplex families with 12 subjects, including monozygotic twins 1659 and 1660. We classified self-described subject ethnicity according to our IRB reporting guidelines. Forty (85%) of the subjects from 35 of the families were classified as “white”, one subject was “black”, and six (13%) were “other/unknown”. Thirty-seven subjects had bilateral EVA and 10 had unilateral EVA (Table 1). Pure-tone (0.5/1/2/4 kHz) audiometric threshold averages for the 84 EVA ears were classified as normal (n=4), mild (n=17), moderate (n=28), severe (n=16), profound (n=13), or unknown (n=6) as described previously (King, et al., 2007).
We prepared DNA samples from EVA subjects as described (Pryor, et al., 2005b). We used Caucasian control (HD200CAU) and human diversity control samples from Coriell Cell Repositories (Camden, NJ, USA) (Kitajiri, et al., 2007) and previously reported African-American control samples (Bonilla, et al., 2005).
We evaluated affected subjects at the NIH Clinical Center as described by Pryor et al. (2005b) with a few exceptions (see Table 1). We also performed perchlorate discharge tests and classified thyroid phenotype as described (Pryor, et al., 2005b). We considered a discharge test to be ‘uninterpretable’ (Table 1) if discharge values fluctuated above and below 15%. The thyroid phenotype was classified as indeterminate in subjects with anti-thyroid antibodies (1619, 1693 and 1823) or atypical goiter with a hyper-functional nodule (1847).
We isolated DNA and performed SLC26A4 sequence analysis as described (Pryor, et al., 2005b) with a few modifications (Supplementary Methods). We determined segregation and meiotic phase configuration of variant alleles by sequence analysis of affected exons in parents and siblings, when available. We analyzed 14 unlinked short tandem repeat (STR) markers in subjects 1659 and 1660 to determine zygosity. Mutations were numbered according to NM_000441.1 (cDNA) and NP_000432.1 (protein). cDNA number +1 corresponds to the A of the ATG translation initiation codon. Names of all variants were checked with Mutalyzer (http://www.LOVD.nl/mutalyzer/).
We cloned wild type SLC26A4 cDNA into the KpnI and XhoI sites of pEGFP-N1 (Taylor, et al., 2002) and the XbaI and BamHI sites of pGEM-HE (Scott, et al., 2000) for site-directed mutagenesis (QuikChange, Stratagene, La Jolla, CA) to generate missense expression constructs. The entire cDNA inserts were sequenced to confirm the absence of unintended mutations.
COS-7 cells were cultured, transfected with SLC26A4-EGFP cDNA expression constructs, and stained with Concanavalin A (see Supplementary Methods). Fusion of GFP to the N-terminus resulted in comparatively lower plasma membrane localization and increased perinuclear aggregation (not shown) so we chose to analyze mutant allele products fused at their C-termini to EGFP.
We detected 18 different sequence variants of SLC26A4 among the 47 newly ascertained EVA subjects (Table 1). Genotype analyses of parents confirmed vertical transmission of the variants in all cases. Fourteen variants have been previously reported and four are novel. Novel variants include c.-60A>G and c.-66C>G in the 5’-untranslated region (5’-UTR; GenBank accession NM_000441) in subjects 1726 and 1643, respectively. Neither c.-60A>G nor c.-66C>G was in trans configuration with a pathogenic allele. c.-60A is a conserved nucleotide within a conserved sequence among mammalian orthologs. c.-66C is a non-conserved nucleotide within a non-conserved sequence among mammals. The self-reported ethnicities of the subjects were “black” (1726) and “white” (1643). We did not detect c.-60A>G or c.-66C>G among 146 Caucasian control chromosomes. c.-66C>G is annotated in dbSNP as a single nucleotide polymorphism (SNP; rs17154282) that is polymorphic among African Americans but not Europeans. We identified c.-66C>G in African American (GM14682), Mexican (GM00244), Puerto Rican (GM07546 and GM12928), and Middle Eastern (GM02016) control chromosomes from a panel of 57 samples of diverse ethnic origins (Kitajiri, et al., 2007). These results suggest that the presence of c.-66C>G in subject 1643 may reflect unrecognized ethnic heterogeneity in this individual. We did not detect c.-60A>G among diversity control samples.
We identified novel missense variants p.V402M (c.1204G>A) and p.M775T (c.2324T>C) co-segregating with EVA in affected sibling pairs in families 243 and 147, respectively (Table 1). Both variants are non-conservative substitutions of residues that are conserved among SLC26A4 orthologs but not among SLC26A paralogs (http://genome.ucsc.edu/). We did not detect p.V402M and p.M775T among 172 and 190 ethnically matched (Caucasian) control chromosomes, respectively.
We sought to predict the pathogenic potential of these and the other SLC26A4 variant alleles that we detected in these 47 subjects and our previously reported 39 patients (Table 2).
We considered frameshift variants c.783dupT, c.1340_1343dup, c.1392delG, and c.1536_1537delAG (Table 2) to be pathogenic based upon predicted and demonstrated deleterious effects of other truncation mutations upon SLC26A4 expression and function (Brownstein, et al., 2008; Yoon, et al., 2008).
We considered c.1001+1G>A and c.1264-1G>C to be pathogenic since they affect canonical splice donor and acceptor nucleotide positions, respectively (Pryor, et al., 2005b). We previously identified c.-3-2A>G as a mono-allelic SLC26A4 variant in a subject with NSEVA (Pryor, et al., 2005b). c.-3-2A>G is conserved among mammalian orthologous sequences (http://genome.ucsc.edu/). We did not detect c.-3-2A>G among 378 Caucasian control chromosomes. c.-3-2A>G affects a canonical splice acceptor site, but is located upstream of the translation start codon in exon two. Retention of the entire first intron might thus result in a functional transcript. There are also several potential cryptic splice acceptor sites (http://www.fruitfly.org/) whose utilization would not disrupt the predicted ribosome binding site or open reading frame (not shown).
In addition to c.-60A>G and c.-66C>G, we detected the c.-103T>C variant described by Yang et al. (2007). We identified c.-103T>C in trans configuration with the p.L236P mutation in an eight month-old “white” subject (1766) with bilateral EVA and an indeterminate thyroid phenotype. We did not detect c.-103T>C among 146 ethnically matched (Caucasian) control chromosomes or 106 diversity control chromosomes.
Our cohort includes 17 missense variants, including the two novel variants p.V402M and p.M775T. p.V138F, p.G209V, p.L236P, p.E384G, p.T416P, p.L445W, and p.Y530H are prevalent alleles usually observed in homozygosity or in compound heterozygosity with pathogenic mutations in PS subjects (Table 1 and Supplementary Table S1) (Blons, et al., 2004; Coyle, et al., 1998; Pryor, et al., 2005b; Van Hauwe, et al., 1998). The pathogenicity of p.V138F, p.G209V, p.L236P, p.E384G, and p.T416P has been confirmed by functional studies of their protein products (Table 2) (Scott, et al., 2000; Taylor, et al., 2002). The functional properties of p.L445W and p.Y530H allele products have not been reported, but the affected residues are conserved among all known orthologs of SLC26A4 and all or most of the human paralogs, respectively (Table 2).
We detected p. Q514R in three subjects: one (1616) had PS (Pryor, et al., 2005b) and two (1700 and 1847) had bilateral EVA with indeterminate thyroid phenotypes (Table 1). All three p.Q514R carriers had a second SLC26A4 variant (p.C565Y, c.1343_1344insAGTC, and p.G209V) in trans configuration. We detected p.C565Y only in the subject (1616) with PS and p.Q514R. p.C565Y has been reported only twice previously: in trans configuration with p.L236P in a PS patient (Van Hauwe, et al., 1998) and in trans configuration with the p.H723R mutation in a Japanese EVA patient (Supplementary Table S1) (Tsukamoto, et al., 2003). We did not detect p.Q514R or p.C565Y among 192 ethnically matched (Caucasian) chromosomes.
Seven of the missense variants are rare (one or two families with these alleles in our cohort), detected only as mono-allelic variants, or observed only in subjects with NSEVA or indeterminate phenotypes (Table 1 and Supplementary Table S1). These seven variants are p.M1T (c.2T>C), p.F335L (c.1003T>C), p.V402M (c.1204G>A), p.Y530S (c.1589A>C), p.V609G (c.1826T>G), p.M775T (c.2324T>C), and p.R776C (c.2326C>T). Pfarr et al. (2006) recently concluded that p.R776C is a benign polymorphism. We did not detect these variants among 144 to 192 Caucasian control chromosomes.
We detected p.V609G in two African American EVA subjects: 1590 (Pryor, et al., 2005b) and 1726 (Table 1). p.V609G has been previously reported as a mutation in EVA subjects (Supplementary Table S1). It was detected as a mono-allelic variant in one subject (1590) and in trans configuration with the indeterminate variant c.-60A>G in another (1726). We identified 13 heterozygous carriers and two homozygotes for p.V609G among 182 African American control subjects. p.V609G was later annotated as a SNP (rs17154335) that is polymorphic among African Americans (http://www.ncbi.nlm.nih.gov/projects/SNP/).
We identified eight independent chromosomes with p.L597S among our EVA cohort (Table 1). This substitution affects a leucine that is conserved among pendrin orthologs but not human SLC26A paralogs (http://genome.ucsc.edu/). We detected it as a mono-allelic variant in seven subjects with NSEVA or indeterminate phenotypes. Similar findings have been observed in other studies (Supplementary Table S1) (Yang, et al., 2007). We detected p.L597S in trans configuration with the p.E384G mutation in one subject (1447) with PS. We identified p.L597S in five (1.25%) of 400 Caucasian control chromosomes. Albert et al. (Albert, et al., 2006) detected p.L597S in four of 100 control chromosomes and concluded that it was nonpathogenic.
We sought to characterize functionally the pendrin products encoded by p.M1T, p.F335L, p.V402M, p.L445W, p.Q514R, p.Y530H, p.Y530S, p.C565Y, p.L597S, p.M775T, and p.R776C. We expressed wild type or variant pendrin, fused at its C-terminus to GFP (green fluorescent protein), in COS-7 cells. The p.M1T, p.V402M, and p.L445W products displayed an intracellular reticular pattern with no surface expression (Figure 1). This is consistent with endoplasmic reticulum (ER) retention described for other SLC26A4 missense mutant products (Rotman-Pikielny, et al., 2002; Taylor, et al., 2002). In contrast, we detected wild type, p.F335L, p.C565Y, p.L597S, p.M775T, and p.R776C products colocalized with conconavalin A at the cell periphery, consistent with plasmalemmal localization (Figure 1). p.Q514R, p.Y530H, and p.Y530S showed intermediate patterns with prominent ER and post-ER punctate staining that extended to the periphery of the cell and into filopodial processes. These patterns were reproducible at 14 and 18 hours post-transfection (not shown).
Pathogenic pendrin mutants might, in addition to disrupting normal polypeptide trafficking and stability, also alter anion transport rate, anion affinity, and anion selectivity. To examine these possibilities, we expressed wild type pendrin and pendrin missense mutant polypeptides p.F335L, p.C565Y, p.L597S, p.M775T and p.R776C in Xenopus oocytes. We also expressed p.L236P, p.E384G, and p.T416P as loss-of-function controls (Scott, et al., 2000; Taylor, et al., 2002). The Cl−/I− exchange rate constants for these loss-of-function controls ranged from 3 to 7% of the corresponding wild type rate constant (Figure 2B). Rate constants for p.F335L, p.C565Y, p.L597S, p.M775T, and p.R776C were approximately 10-fold higher, ranging from 37 to 77% of the wild type rate constant (Figure 2B). Cl−/HCO3− exchange rate constants of hypomorphic pendrin variants ranged from 13 to 93% of the wild type rate constant (Figure 3B). p.R776C had the highest rate constants among our tested variants, consistent with the observation of Pfarr et al. (2006) that its activity was indistinguishable from that of wild type pendrin.
Normalized Cl−/I− exchange rate constants for the tested hypomorphic variants (gray bars in Figure 2C) were moderately but significantly higher than the wild type value. Normalized Cl−/HCO3− exchange rate constants for the tested variants were similar to the wild type value (Figure 3C), except for p.F335L and p.L597S which showed mildly but significantly decreased and increased values, respectively.
c.-103T>C has been proposed to exert a pathogenic effect by disrupting a conserved binding site for the FOXI1 transcription factor (Yang, et al., 2007). However, another promoter analysis did not reveal a significant effect of this region (exon one) upon SLC26A4 transcription in vitro (Adler, et al., 2008). The unusual location of this postulated transcription factor binding site within an exon and the detection of c.-103T>C only as a mono-allelic variant (Yang, et al., 2007) (Supplementary Table S1) preclude a definitive classification of its pathogenic potential. Although we detected c.-103T>C in trans configuration with p.L236P, our subject (1766) had an indeterminate thyroid phenotype that is not inconsistent with the presence of one mutant allele of SLC26A4 (Pryor, et al., 2005b).
The discordant segregation of c.-3-2A>G with EVA in an affected fraternal twin pair (Pryor, et al., 2005b) supports in silico predictions that it may not affect SLC26A4 expression or function. An exon trapping analysis could not accurately model in situ splicing because it is important to include adjacent exon(s) but the immediate upstream (first) exon does not have its own acceptor site to splice to the vector donor site (not shown). Although we were unable to perform a direct analysis of SLC26A4 transcripts from our c.-3-2A>G carrier to assess its effects on SLC26A4 expression in situ, we suspect that this variant is not pathogenic.
The p.M1T, p.V402M and p.L445W allele products were clearly retained within the ER, confirming their pathogenicity (Figure 1). The p.Q514R, p.Y530H and p.Y530S allele products were also retained inside COS-7 cells but appeared to traffic to post-ER locations (Figure 1). This different localization is probably due to heterogeneity in processing of mutant pendrin polypeptides (Yoon, et al., 2008). Since we and others have nearly always detected these alleles in trans configuration with other mutations in subjects with PS (Supplementary Table S1) (Blons, et al., 2004; Coyle, et al., 1998; Pryor, et al., 2005b), their aberrant expression patterns likely reflect a pathogenic role in EVA. The pathogenic potential of p.Q514R is further supported by the recent description of a prevalent founder mutation, p.Q514K, that affects the same amino acid in a Spanish EVA population (Pera, et al., 2008).
The p.F335L, p.C565Y, p.L597S, p.M775T, and p.R776C allele products all traffic to the plasma membrane in a manner indistinguishable from wild type pendrin (Figure 1). Although their rate constants were reduced in comparison to wild type pendrin, they still had substantial residual activity that was an order of magnitude greater than for functional null controls (Figures 2 and and3).3). Therefore these alleles may not be pathogenic as mono-allelic variants. This hypothesis is supported by normal control carrier rates (1.25 – 4%) of p.L597S that indicate its detection in NSEVA subjects is coincidental (Albert, et al., 2006). Similarly, the detection of p.F335L as a mono-allelic SLC26A4 variant in 12 (2.0%) of 609 EVA subjects reported by Albert et al. (2006), Madden et al. (2007), and Yang et al. (2007) (Supplementary Table S1) is not significantly different (p > 0.05, Fisher’s exact test) from control carrier rates of 0/50 reported by Albert et al. (2006) and our rate of 0/94. Similar analyses of other, less common, variants may be more difficult to interpret due to insufficient numbers of ethnically matched cases and controls.
The pathogenic potential of p.F335L, p.C565Y, p.L597S, p.M775T, or p.R776C may be different in trans configuration with an SLC26A4 mutation. Although p.C565Y is usually detected in trans configuration with a mutant allele, p.F335L, p.L597S, p.M775T and p.R776C are almost always detected as mono-allelic variants (Supplementary Table S1). We detected p.C565Y and p.L597S in two PS subjects (1616 and 1447, respectively) in trans configuration with other mutant alleles (p.Q514R and p.E384G, respectively). The correlation of two mutant alleles with PS suggests a pathogenic role for p.C565Y and p.L597S in these subjects (Pryor, et al., 2005b). In contrast, we observed p.F335L in trans configuration with a mutant allele (c.1001+1G>A) in a subject (1627) with unilateral EVA and an indeterminate thyroid phenotype. The correlation of one mutant allele with unilateral EVA is consistent with a nonpathogenic role for p.F335L in this subject (Pryor, et al., 2005b). Alternatively, the correlation of phenotype with number of mutant alleles may not apply to these hypofunctional variants. The phenotype associated with two variant alleles may depend upon the degree of residual transport activity encoded by the variant allele(s), as observed for skeletal dysplasia phenotypes and SLC26A2 mutations (Karniski, 2001). The phenotype may also reflect interactions of SLC26A4 variants with environmental or genetic modifiers via a complex or oligogenic mechanism (Pryor, et al., 2005a; Pryor, et al., 2005b; Yang, et al., 2007), as described for Bardet-Biedl syndrome (Badano, et al., 2006).
p.G209V is a hypo-functional variant (Taylor, et al., 2002) whose pathogenic potential in trans configuration with SLC26A4 mutations is strongly supported by its frequent detection in this genotypic context (Supplementary Table S1). Its pathogenic role as a mono-allelic variant in EVA subjects seems likely since the relative frequency of detection in these genetic contexts is similar to those for functional null mutations such as p.L236P (Supplementary Table S1).
Our results are similar to those reported by Scott et al (Scott, et al., 2000). Moreover, we observed a degree of loss-of-function of Cl−/I− exchange that was proportionately less for most variants tested than for Cl−/HCO3− exchange, and for all tested variants with respect to Cl−/Cl− exchange. This could explain, in part, the lower penetrance of thyroid abnormalities than of EVA and hearing loss. However, the magnitude and consistency of the hypomorphic variants’ effects upon exchange of different anions seems insufficient to infer a causal association with nonsyndromic EVA. We thus conclude that rare, heterozygous, hypofunctional SLC26A4 alleles detected in NSEVA subjects are likely to be coincidentally detected variants. Some of these variants may be pathogenic in trans configuration with pathogenic SLC26A4 mutations in patients with PS. A potential correlation of clinical phenotype with mutant pendrin function may be more robust with systems that better model in situ expression (Karniski, 2004). Our results may have been affected by the use of non-polarized COS-7 cells for trafficking assays or non-mammalian cells for functional assays or, in the case of some mutants, by ER stress or protein degradation induced by transient overexpression.
Proposed correlations of SLC26A4 genotype with EVA or thyroid phenotypes remain controversial. The varying opinions may reflect differences and inadequacies in phenotype analyses and mutation detection. Our results highlight another potential source of the discrepancies: misclassification of benign polymorphic variants as pathogenic alleles. The inclusion of benign variants of SLC26A4 in genotype-phenotype correlation analyses could have led to false positive or false negative results. These confounding factors should be accounted for when evaluating SLC26A4 genotypes and EVA phenotypes for evidence of correlations.
We thank the study families for their participation, Richard Trembath and Paul Yen for expression constructs, Yandan Yang for technical contributions, NIDCD clinic staff, Stephanie Moran, Angela Stuber, Nick Patronas, Monica Skarulis, and Clara Chen for contributions to clinical evaluations, Ning Hu for clinical data management, Rachel Fisher, Darcia Dierking and Leisha Eiten for referral of patients, Robert Morell for advice on zygosity analysis, Inna Belyantseva and Shin-ichiro Kitajiri for confocal microscopy technical support and advice, and Tom Friedman, Dennis Drayna and Susan Sullivan for critical review of the manuscript. This study was supported by NIDCD/NIH intramural research funds 1-Z01-000039 (T.F.), 1-Z01-DC-000060 and 1-Z01-DC-000064 (A.J.G.) and NIH grants DK43495 (S.L.A.) and DK34854 (Harvard Digestive Diseases Center support to S.L.A.).