In this study we investigated 28 genetically unexplained Dutch index cases with HL and a unilateral or bilateral EVA. To elucidate the missing heritability in monoallelic SLC26A4 cases, who represent 14–31% of subjects with HL and EVA (Azaiez et al. 2007; Mey et al. 2019), extensive genomic analyses as well as phenotyping were performed. Important findings in this study were (1) the enrichment of a shared (V1-)CEVA haplotype in M1 SLC26A4 cases, (2) two novel SLC26A4 splice variants and (3) the identification of a FOXI1 variant in three subjects suggesting a contribution of this variant to the etiology of HL and EVA. Furthermore, the genotype-phenotype analyses revealed that the severity of the HL associated with biallelic variants (M2) in SLC26A4 is comparable to the HL associated with a monoallelic variant in SLC26A4 with or without the CEVA haplotype (M1 and M1/CEVA).
The V1-CEVA haplotype was also previously reported in a single M1 case and suggested to refine the CEVA haplotype or to harbor a different genetic defect (Chattaraj et al. 2017). The shared VNTR marker alleles of the V1-CEVA and the CEVA haplotype suggest that V1-CEVA refines the boundaries of the shared genomic region to 0.57 Mb.
We anticipated that a pathogenic variant co-segregates with the CEVA haplotype. Therefore, we subjected the shared genomic region to extensive genomic analyses that included WES, short- and long-read WGS, and optical genome mapping, to reveal any potential variants missed or misinterpreted in earlier studies. None of the applied sequencing or imaging techniques revealed rare SVs that overlap or are present within the CEVA haplotype. In the light of the proven accuracy and efficacy of especially optical genome mapping and long-read sequencing in SV detection (Chaisson et al. 2019), we deem it unlikely that any SVs within the CEVA region escaped detection. Additionally, we evaluated all SNVs with an AF ≤ 5% (gnomAD) present within the region for predicted regulatory or splice altering effects but for none of the 20 SNVs a potential effect was predicted by SpliceAI. Two SNVs overlap with a potential regulatory element of SLC26A4 (GeneHancer, EnhancerAtlas), and one variant is present within the intronic regions of this gene. However, all three variants are located within a highly repetitive element (LINE). Although little is known about the effects of genetic variation within LINE elements, a potential effect on the methylation landscape and consequently gene expression levels has been suggested (Xie et al. 2009) and such an effect can therefore not be excluded for the three indicated variants. For the remaining SNVs, no potential effects on transcript splicing or gene regulation were predicted. Nevertheless, we cannot rule out combinatory effects of the SNVs, since they are all located in cis. A thorough experimental (multi-omic) analysis is required to optimally assess the effects of the identified variants. RNA studies can be performed to detect quantitative or qualitative changes affecting the SLC26A4 transcripts. A defect observed on the RNA level could provide valuable insights that may point towards the true pathogenic defect, and prioritize one, or a combination, of the variants on the CEVA allele. However, SLC26A4 is not or at extremely low levels expressed in readily accessible patient cell types (e.g., fibroblasts and blood cells). The same holds true for induced pluripotent stem cells or otic progenitor cells (Hosoya et al. 2017). However, Hosoya and co-workers have successfully developed a protocol that allows the differentiation of otic progenitor cells into outer sulcus-like cells that express SLC26A4 at high levels. This protocol could potentially be a powerful tool to evaluate the consequences of CEVA haplotype at the RNA level.
SLC26A4 is not the only gene present within the CEVA haplotype, which also spans BCAP29, COG5, DUS4L, HBP1, PIK3CG, and PRKAR2B. For none of these genes, pathogenic variants associated with (syndromic) HL have been reported, nor has a function in the inner ear been described. The majority of the CEVA-associated SNVs (16/20) are located within an intronic region of these genes, however, for none of these variants a splice altering effect is predicted by SpliceAI.
Since the genetic defect on the CEVA haplotype could not be pinpointed by the genetic analyses, we could not determine whether the AF of the defect is lower than that of the CEVA haplotype and more in line with the expected frequency based on the prevalence of HL (1:1,000 newborns (Morton and Nance 2006)) and the genetic heterogeneity of the condition. Alternatively, the CEVA haplotype could be considered a hypomorphic allele, of which the penetrance depends on the contribution of other co-existing (common) variants.
Not all M0 or M1 SLC26A4 cases could be genetically explained by the presence of the CEVA haplotype. Therefore, digenic inheritance with variants in EPHA2, FOXI1, and KCN10 was also explored as a potential explanation for the missing heritability. Digenic inheritance of defects in SLC26A4 and EPHA2 has recently been reported in two Japanese Pendred syndrome cases (Li et al. 2020). A c.1063G > A (p.(Gly355Arg)) and a c.1532C > T (p.(Thr511Met)) variant in EPHA2 were each found ‘in trans’ with a reported pathogenic variant in SLC26A4 (Deafness Variation Database (Azaiez et al. 2018)). EPHA2 was identified as a binding partner of pendrin, with a crucial role in regulating pendrin localization (Li et al. 2020). The identified variants in EPHA2 were predicted to be pathogenic by several in silico predictions tools. However, the c.1532C > T variant has a relatively high allele frequency of 3.03% in the East Asian population, including 11 homozygotes (gnomAD). Yet, in the present study, we did not obtain indications for digenic inheritance of variants in SLC26A4 and EPHA2 in subjects with HL and EVA. Besides for EPHA2, a digenic mechanism has also been reported and debated for variants in SLC26A4 and KCNJ10 or FOXI1, with currently no consensus (Jonard et al. 2010; Landa et al. 2013; Pique et al. 2014; Yang et al. 2009; Yang et al. 2007). FOXI1 is a transcriptional regulator of SLC26A4 (Yang et al. 2007). We identified a c.677C > T (p.(Thr226Ile)) FOXI1 variant in three subjects (2 M0/FOXI1 and 1 M0/CEVA/FOXI1). This variant was previously detected in an individual diagnosed with Pendred syndrome and a monoallelic pathogenic SLC26A4 variant (Pique et al. 2014). The variant has an allele frequency of 0.71% in non-Finnish Europeans (gnomAD) and affects an amino acid residue located outside the DNA-binding domain but close to the nuclear localization signal (NucPred (Brameier et al. 2007)). Previously reported pathogenic FOXI1 variants have been shown to affect the DNA-binding properties of the protein (Enerbäck et al. 2018). We speculate that a variant affecting the localization motif of the protein could potentially have a loss of function effect as well. Although the variant is classified as benign in the Deafness Variation Database, we identified the variant three times in our cohort of genetically unexplained SLC26A4 cases and combined with the fact that it has been reported in a previous study (Pique et al. 2014), this suggests that the variant might actually contribute to the etiology of HL and EVA. This is in line with the expansion of the endolymphatic compartment and an audio-vestibular phenotype observed in Foxi1−/− mice (Hulander et al. 2003). In situ hybridization of the endolymphatic duct and sac of these mice revealed complete absence of Slc26a4 mRNA expression. Functional studies, among which cellular localizations assays, are warranted to evaluate the effect of the c.677C > T FOXI1 variant. We did not identify likely pathogenic variants in KCNJ10 (AF ≤ 5%) in our cohort
WGS did not reveal strong candidate regulatory variants based on data derived from enhancer databases and transcription factor binding site predictions. Nevertheless, interpretation of regulatory variants is still considered complex and is limited by the lack of available epigenetic datasets for the inner ear. Also, no SVs overlapping with SLC26A4 were detected using WGS, suggesting a limited contribution of SVs to the mutational landscape of SLC26A4. This is in line with earlier observations described in literature (Liu et al. 2021; Pique et al. 2014). For the monoallelic cases (M1, M0/CEVA), no long-read sequencing or optical genome mapping was performed. As it is generally accepted that most SVs could not be accurately detected using short-read sequencing approaches only (Chaisson et al. 2019), it cannot be excluded that causative SVs are present but missed due to technical limitations.
The present study did not confirm that the CEVA allele is associated with a milder HL compared to SLC26A4 variants affecting the protein-coding sequences, as indicated by Chao et al. (Chao et al. 2019). They discerned a significantly milder HL in their cohort of M1/CEVA subjects (n = 20 ears, median 47.5 dB HL) than we have seen in our cohort of M1/CEVA subjects (n = 16 ears, median 84 dB HL). A possible explanation for this discrepancy could be the progression of HL combined with a ~ 5-year difference in average age between the cohorts (7.5 and 12.8 years, respectively). Progression of HL is seen in up to 39.6% of EVA-ears (Alemi and Chan 2015), with progression rates of ~ 3.5 – ~5.5 dB/y (Govaerts et al. 1999; Jackler and de la Cruz 1989). On the other hand, the older subjects in our M1/CEVA cohort show less severe HL than the younger subjects, which is questioning the relationship with age. Furthermore, there is also an average age difference of 5 years between the M2 groups in both studies (13.2 years and 18.4 years, respectively), while the severity of HL is comparable (85 and 86.3 dB HL, respectively).
The reported variability of the auditory phenotype associated with EVAs (Arjmand and Webber 2004; Gopen et al. 2011; Griffith and Wangemann 2011) may be another explanation for the observed differences in severity of HL in both studies. In literature, many prognostic factors such as genotype, EVA size and morphology, age, head trauma, and gender are reported as underlying explanations for this variability, although some of these studies draw contradicting conclusions (Alemi and Chan 2015; Archibald et al. 2019; Ascha et al. 2017; Gopen et al. 2011; Miyagawa et al. 2014; Rah et al. 2015; Saeed et al. 2021). In the same line, Song et al reported intrafamilial differences in the severity of hearing loss in siblings with the same biallelic variants in SLC26A4 (Song et al. 2014). Larger sample sizes are needed to confirm or reject the hypothesis that the CEVA haplotype is associated with a milder HL phenotype.
The significant difference in HL severity between the M2 and M1/CEVA groups versus the M0 group suggests that SLC26A4 defects have a prognostic value which can be strengthened in the future by the identification of the underlying genetic defects in subjects of the M0 group.
In conclusion, the HL and EVA in 12 of the 28 studied subjects could be associated with SLC26A4. In addition, we have identified genetic factors that might (partially) explain the phenotype in four additional subjects. However, we could not pinpoint the genetic defect that is present on the CEVA haplotype. The arrival of third-generation sequencing techniques, the expansion of epigenetic and transcriptomic datasets and the increasing understanding of non-coding, structural, and regulatory variants will aid in solving the missing heritability in SLC26A4 in the coming years. This is of great importance for counseling patients about the underlying cause and expected prognosis of their HL. Furthermore, as variants in SLC26A4 are a frequent cause of HL (Sloan-Heggen et al. 2016), it is an interesting target for the development of a genetic therapy (Kim et al. 2019). Although the involved molecular defect of the CEVA haplotype is still not resolved, the high prevalence of the CEVA haplotype suggests that a significant portion of monoallelic SLC26A4 cases can be associated with SLC26A4 defects by testing for the presence of this haplotype.