Variants encoding a restricted carboxy-terminal domain of SLC12A2 cause hereditary hearing loss in humans

Summary

Hereditary hearing loss is challenging to diagnose because of the heterogeneity of the causative genes. Further, some genes involved in hereditary hearing loss have yet to be identified. Using whole-exome analysis of three families with congenital, severe-to-profound hearing loss, we identified a missense variant of SLC12A2 in five affected members of one family showing a dominant inheritance mode, along with de novo splice-site and missense variants of SLC12A2 in two sporadic cases, as promising candidates associated with hearing loss. Furthermore, we detected another de novo missense variant of SLC12A2 in a sporadic case. SLC12A2 encodes Na⁺, K⁺, 2Cl⁻ cotransporter (NKCC) 1 and plays critical roles in the homeostasis of K⁺-enriched endolymph. Slc12a2-deficient mice have congenital, profound deafness; however, no human variant of SLC12A2 has been reported as associated with hearing loss. All identified SLC12A2 variants mapped to exon 21 or its 3’-splice site. In vitro analysis indicated that the splice-site variant generates an exon 21-skipped SLC12A2 mRNA transcript expressed at much lower levels than the exon 21-included transcript in the cochlea, suggesting a tissue-specific role for the exon 21-encoded region in the carboy-terminal domain. In vitro functional analysis demonstrated that Cl⁻ influx was significantly decreased in all SLC12A2 variants studied. Immunohistochemistry revealed that SLC12A2 is located on the plasma membrane of several types of cells in the cochlea, including the strial marginal cells, which are critical for endolymph homeostasis. Overall, this study suggests that variants affecting exon 21 of the SLC12A2 transcript are responsible for hereditary hearing loss in humans.

Keywords:

Cell membranes – Reverse transcriptase-polymerase chain reaction – Yellow fluorescent protein – Audiology – Cochlea – Dideoxy DNA sequencing – Exon mapping

Introduction

Sensorineural hearing loss is one of the most common sensory disorders in humans, and its onset can be influenced by multiple environmental and genetic factors. Approximately 1 in 500 to 1,000 newborns has congenital sensorineural hearing loss [1], and more than 100 genes have been reported as associated with nonsyndromic hearing loss to date [2].

Hair bundles of auditory sensory cells (hair cells) in the mammalian cochlea are bathed in K⁺-rich endolymph at the apical membrane. Proper depolarization of hair cells is achieved via the influx of K⁺ from the endolymph through a mechanotransduction channel located on the hair bundles in response to auditory stimuli. Production and maintenance of endolymph with a high K⁺ concentration (> 150 mM) and endocochlear potential (> +80 mV) is primarily accomplished by the secretion of K⁺ from the marginal cells of the stria vascularis in the cochlear lateral wall [3, 4]. To achieve and maintain the unique composition of the endolymph, multiple energy-neutral ion transporters, such as Na⁺, K⁺, and 2Cl⁻ cotransporters (NKCCs), are active in the lateral wall cells, along with active transport of K⁺ by ion pumps, such as Na⁺, K⁺ ATPases, and voltage-gated K⁺ channels [3]. Dysfunction of the genes contributing to homeostasis of the cochlear endolymph, such as ATP1A3 [5], KCNQ1 [6], and KCNE1 [7], leads to CAPOS (cerebellar ataxia, areflexia, pes cavus, optic atrophy, and sensorineural hearing loss) syndrome (MIM: 601338), Jervell and Lange-Nielsen syndrome (JLNS) 1 (MIM: 220400), and JLNS2 (MIM: 612347), respectively, which all have hearing loss as a major clinical feature. Inner-ear dysfunction, based on the failure to maintain endolymph homeostasis, has been proposed as “endolymphatic deafness” [8]. Although less clearly elucidated, Cl⁻ has also been recognized as important for proper cochlear function, since dysfunction of CLIC5, which mediates Cl⁻ efflux [9], or SLC26A4, a Cl⁻/I⁻ and Cl⁻/HCO₃⁻ exchanger that mediates Cl⁻ influx [10], leads to the autosomal-recessive hearing loss conditions DFNB103 (MIM: 616042) and DFNB4 (600791), respectively.

Genetic analyses of known deafness genes [11–13] can be applied for diagnostic tests, while whole-exome sequencing (WES) is effective for discovering novel disease-associated genes [14, 15]. Using WES, we identified three variants of SLC12A2, which encodes Na⁺, K⁺, 2Cl⁻ cotransporter 1 (NKCC1), in three independent families with segregating hearing loss. We also identified SLC12A2 variant in the fourth family. The importance of SLC12A2 in hearing has previously been reported in Mus musculus [16–19], but not in humans. We conducted splicing assays and functional analysis of the identified variants in vitro, and the results support the pathogenicity of SLC12A2 variants in humans.

Results

Clinical presentation of patients

In family 1, the proband was a 7-month-old female infant (Fig 1A, III-1). The proband and her twin brother (III-2) both presented with congenital, profound hearing loss (Fig 1D and 1E, S1A and S1B Fig). The family exhibited an autosomal-dominant inheritance mode. Hearing was normal in III-3 (S1C Fig), whereas II-3, II-4, I-3, and I-4 also showed congenital profound hearing loss (S1D–S1G Fig).

In family 2, the proband was a 1 year, 5-month-old female toddler (Fig 1B, III-2), who presented as a sporadic case of congenital, severe hearing loss (Fig 1F). The proband could not hold her head up or maintain a sitting position until age 10 months, indicating a minor motor developmental delay.

In family 3, the proband was a 1 year, 10-month-old male toddler (Fig 1C, III-2), who had congenital, profound hearing loss as a sporadic case (Fig 1G, S1H Fig). The proband could not hold his head up until 6 months old and only started walking alone at 1 year, 8 months, indicating borderline motor developmental delay.

CT of the temporal bones did not detect any anomaly of the inner, middle, or outer ears in any of the patients examined (family 1, III-1 and III-2; family 2, III-2; family 3, III-2). Visual inspection of the entire body, palpation of the head and neck, interview of affected individuals or their parents did not detect additional symptoms of multiple organs [20], global developmental delay [21], macrocephaly and epilepsy [22], or schizophrenia [23], all of which have been reported to be associated with SLC12A2 variants. Body weight and height, motor and neuronal development, as well as behavior and cognition were within normal range.

Clinical presentation of family 4 is briefly described in the section of genomic analysis.

Genomic analysis

Details of the three candidate pathogenic variants detected by genomic analysis are presented in Table 1, and a summary of the WES analyses is presented in S1 Table.

For family 1, based on trio WES analysis, including samples from the proband (III-1) and the parents (II-1, II-4), we selected 38 coding and splice-site variants with low minor allele frequencies in population databases that co-segregated with hearing loss (see Materials and methods for details, S4 Table). To narrow down these candidate variants, genes were classified into three categories and prioritized for analysis, as described in the Materials and Methods. No Tier 1 genes (previously identified deafness genes in humans; S2 Table) with autosomal dominant inheritance were among the candidate variants, and SLC12A2 [16–19] was the only Tier 2 gene (genes associated with hearing loss in mouse models; S3 Table) among the variants. Sanger sequencing identified a heterozygous variant of SLC12A2 (c.2941G>T), resulting in a D981Y amino acid change in I-4, II-3, II-4, III-1, and III-2, who had hearing loss, but not in the other subjects tested (Figs 1A and 2A, S2A Fig), indicating co-segregation of the variant with hearing loss, with the exception of I-3, who had hearing loss, but did not carry this variant. To further investigate the possible cause of hearing loss in I-3, we conducted sequencing analysis targeting coding exons of 154 deafness genes and candidate deafness genes [11] (S7 Table) in this individual. The results indicated that he carried a heterozygous variant of MYH19 (NM_002473:c.5765+9C>G, rs201008102) associated with DFNA17 with uncertain pathogenicity; however, neither II-4 nor the proband (III-1) carried this variant, indicating that it did not co-segregate with hearing loss in the family. Hence, the cause of hearing loss in I-3 is unknown at present. Overall, although we cannot exclude other possibilities, such as linkage disequilibrium of the SLC12A2 variant with a possible cryptic pathogenic change, or variants in unknown deafness genes, also present in I-3, which could be responsible for the disease, the SLC12A2 variant was considered the best candidate for association with hearing loss.

For family 2, trio WES analysis identified five candidate variants in four genes (one variant each of SLC12A2, CASC1, and C5orf51, and 2 variants of SZT2, S5 Table). No Tier 1 gene was among the identified variants, and SLC12A2 was the only Tier 2 gene. Sanger sequencing confirmed that the heterozygous c.2930‒2A>G variant of SLC12A2 in the proband reflected a de novo mutation (Fig 2B, S2B Fig). Human Splicing Finder ver3.0 [24] generated a prediction value for the altered acceptor site of −41.22%, well beyond the threshold (−10%), suggesting that inclusion of exon 21 is disrupted by the variant (Table 1). In this family, parenthood of the proband was investigated by counting the numbers of variants she inherited. We found that 98 and 103 of 217 protein-affecting variants with MAF < 0.003 appeared to be derived from the father (II-3) and mother (II-4), respectively, confirming their parenthood (S3A Fig).

For family 3, singleton WES analysis of the proband focusing on Tier 1 and 2 genes detected candidate heterozygous variants of TECTA (NM_005422.2: c.4495G>C, MIM: 601543) [25, 26] among the Tier 1 genes and SLC12A2 and ACAN (NM_001135.3: c.83C>G) among the Tier 2 genes. ACAN has been associated with hearing loss in M. musculus [27]; however, it is associated with other diseases in humans (short stature and advanced bone age, with or without early-onset osteoarthritis and/or osteochondritis dissecans (MIM: 165800) and spondyloepiphyseal dysplasia (MIM: 608361 and 612813)). Sanger sequencing revealed that the TECTA and ACAN variants detected in the proband (III-2) appeared to be derived from the father (II-3) and the mother (II-4), respectively (S4A and S4B Fig, S6 Table). The both parents had normal hearing; thus these variants are unlikely to be associated with hearing loss. The heterozygous c.2962C>A (p.P988T) variant of SLC12A2 was confirmed to be a de novo variant by Sanger sequencing, and considered the best candidate for association with hearing loss (Fig 2C, S2C Fig). In this family, parenthood was studied by comparison of two short tandem repeat (STR) markers among the proband and his parents, since the proband was subjected to singlet WES analysis and it was not possible to count the numbers of variants inherited from the parents. STR markers detected in the proband appeared to be derived from the parents (II-3 and II-4), confirming their parenthood (S3B Fig).

These three SLC12A2 variants were considered the best candidates for causing hearing loss in the families. The wild-type (WT) sequences at the positions of each variant are highly conserved among 90 vertebrate species, and none of the variants were present in public population databases (Table 1).

In parallel to WES analysis, we conducted sequencing analysis focused on 154 deafness genes and candidate deafness genes (S7 Table) to subjects with hearing loss. In the analysis, a sporadic case was found to have de novo c.2935G>A (p.E979K) variant of SLC12A2 (S5A, S5E–S5G Fig). The proband was a 7 months-old female infant (S5A Fig, III-2), who had congenital, severe hearing loss (S5B and S5C Fig). The position of E979 appeared to encode acidic glutamate or aspartate among vertebrate species; thus, polarity of the amino acid residue at this position was conserved. The proband did not show a motor developmental delay or inner ear malformation. STR markers detected in the proband appeared to be derived from the parents (II-1 and II-4), confirming their parenthood (S5H Fig).

Genomic analysis of human population databases revealed that SLC12A2 is intolerant of functional variation (the z score for the constraint metric for missense variants is positive (2.4), and the probability of SLC12A2 being loss of function-intolerant (pLI) is ≥ 0.9 (0.96), according to gnomAD version 2.1 [28]). These scores support the possibility that the missense variants detected in families 1 and 3 alter the function of SLC12A2.

Conservation of exon 21-encoded amino acid residues

Intriguingly, all four SLC12A2 variants mapped to exon 21 or its 3’-splice site (Fig 2D and 2E). Among the 27 exons within the SLC12A2 genomic DNA (Fig 2F), exon 21 consists of 48 base pairs and encodes 17 amino acid residues (including partial codons for two residues at each end) in the long cytoplasmic stretch of the SLC12A2 protein, after the 12 transmembrane regions. There are two known alternatively spliced transcripts of human SLC12A2: an exon 21-included transcript encoding a long isoform (NP_001037.1), and an exon 21-skipped transcript encoding an in-frame, 16 residue truncated isoform (NP_001243390.1) (Fig 2E and 2F). Therefore, the three missense variants were expected to affect the function of the long isoform, and the splicing variant was anticipated to lead to the excision of exon 21, leading to increased production of the truncated isoform.

Although residues E979, D981, and P988 are well conserved among SLC12A2 orthologs in vertebrate species (Table 1, S5D Fig), multiple alignment of the eight SLC12A family proteins (S6 Fig) revealed that the residues encoded by exon 21 are unique to SLC12A2 and are not conserved, even in the most closely related proteins in this family, SLC12A1 (also known as NKCC2) and SLC12A3 (NCC) [29]. This finding suggests that, during the molecular evolution of SLC12A family members, exon 21 encoded a peptide sequence that conferred a specific functional characteristic to the protein.

Analysis of SLC12A2 exon 21 splicing in vitro and in mammalian tissues

To investigate whether the c.2930‒2A>G (family 2) variant affects inclusion of exon 21 of SLC12A2, reporter minigenes were constructed by inserting the genomic region, including exons 21 and 22 of the gene, with or without the splicing variant, into the pSpliceExpress vector [30] (Fig 3A). Sanger sequencing revealed that the WT and candidate splicing variant alleles harbored additional variants, c.2930‒733A>G (rs251217) and c.2930‒734T>C (rs10057122), respectively (Fig 3A). As rs10057122 and rs251217 are common variants (35.2% and 36.4% in 1000Genomes, respectively), they were considered non-pathogenic and unlikely to substantially affect splicing of exon 21.

When HEK293T cells were transfected with the vector harboring the WT allele, RT-PCR detected three PCR products as reporter transcripts (Fig 3B). The longest product corresponded to the reporter transcript including both exon 21 and exon 22 of SLC12A2 (filled arrowhead in Fig 3B), a shorter PCR product corresponded to the reporter transcript without exon 21 (open arrowhead in Fig 3B), and the shortest PCR product corresponded to a transcript containing neither exon 21 nor exon 22 (arrow in Fig 3B). The shortest product was also detected in the cells transfected with empty vector (mock), indicating that it was generated from the vector construct. When cells were transfected with vector harboring the c.2930‒2A>G variant allele, the longest PCR product disappeared, whereas the two shorter PCR products remained detectable. These results strongly suggest that the c.2930‒2A>G variant disrupts the inclusion of exon 21 in the SLC12A2 transcript, resulting in expression of only the exon 21-skipped transcript from the variant allele.

We next explored whether both the exon 21-included and -skipped alternative transcripts of SLC12A2 were expressed in cochlear tissues. Quantitative RT-PCR (qRT-PCR) demonstrated that expression ratios of exon 21-included/skipped transcripts were dramatically different among mammalian tissues. In whole cochlea and lateral wall, exon 21-included transcript was expressed at significantly higher levels than the exon 21-skipped transcript (Fig 3C, S7A Fig), with ratios of exon 21-included/skipped transcripts similar to that in the choroid plexus; in these tissues, levels of the exon 21-included transcript were significantly higher than those in the cortex and cerebellum (Fig 3C, S7A Fig). Further, expression of the exon 21-included transcript was significantly lower than that of the exon 21-skipped transcript in the cerebellum, with no significant difference detected in the cortex (Fig 3C, S7A Fig). Similar to the results using M. musculus tissues, exon 21-included transcript levels in cochleae dissected from a 3-year-old male Macaca fascicularis were significantly higher than those of the exon 21-skipped transcript, while the exon 21-included transcript was expressed at significantly lower levels in Homo sapiens brain (Fig 3D, S7B Fig).

Consistent with the qRT-PCR results, RT-PCR analysis detected only exon 21-included transcript in whole cochlea and lateral wall samples from M. musculus, and predominant expression of the exon 21-included transcript in cochlea from M. fascicularis (S8A and S7B Figs). Predominant to exclusive exon 21-included transcript expression was also observed in the choroid plexus, eye, spleen, liver, intestine, kidney, and ovary of M. musculus. By contrast, both transcripts were detectable in the cortex and cerebellum of M. musculus, as previously reported [31]. Similarly, both transcripts were detectable in H. sapiens brain, as previously reported [32]. Overall, expression patterns of Slc12a2 in mammalian cochlea and central nervous systems determined by RT-PCR resembled those determined by qRT-PCR (summarized in S9 Fig). Junction expression of SLC12A2 in various H. sapiens tissues, determined as part of the Genotype-Tissue Expression Project (GTEx) [33], also demonstrated that the exon 21-skipped transcript (ETS00000343225.4) is detectable at high levels in the central nervous system (S10 Fig). These data indicate that mammalian cochleae almost exclusively express the exon 21-included transcript of SLC12A2, and that exon 21 skipping is a minor event in the mammalian cochlea and tissues other than those of the central nervous system.

Functional consequences of recombinant (rec) SLC12A2 variants

To determine the effects of the SLC12A2 variants on Na⁺, K⁺, 2Cl⁻ cotransport activity, we established HEK293T cell lines that heterologously expressed WT, mutated (D981Y or P988T), or exon 21 truncated (corresponding to p.977_992del) recSLC12A2 proteins in a doxycycline-dependent manner. Cl⁻-sensitive yellow fluorescent protein (YFP) was attached to the amino-terminus of each of these SLC12A2 constructs so that the ion transport activity of WT, mutated, and truncated recSLC12A2 proteins could be assessed by monitoring Cl⁻ influx [34]. mTurquoise2 (mTq2) was constitutively expressed in these cell lines as a reference fluorescent protein. As cell suspensions were used for our Cl⁻ influx assay, simultaneous measurements of YFP fluorescence (F_YFP) and reference mTq2 fluorescence (F_mTq2) allowed correction for fluctuations in F_YFP between measurements that could be attributed to changes in the number of cells in the light path (Fig 4A). Prior to measurement, all cell lines were incubated in a low Cl⁻-containing solution, which is essential for activating SLC12A2 [35, 36]. The Cl⁻ influx assay was initiated by the injection of a high-Cl⁻ buffer. WT recSLC12A2-expressing cells showed biphasic Cl⁻ influx kinetics, which was in stark contrast to the monophasic kinetics of empty vector-derived YFP-expressing cells (mock, Fig 4B). The monophasic Cl⁻ influx kinetics that were apparent for the YFP-expressing cells indicated that the HEK293T cells retained endogenous Cl⁻ influx activity. The time constant for this endogenous Cl⁻ influx resembled the slower time constant (τ_slow) observed for WT recSLC12A2-expressing cells (152 sec vs. 179 sec, Fig 4B). A previous study also reported biphasic Cl⁻ influx kinetics in SLC12A2-expressing cells, likely ascribable to deactivation of SLC12A2 upon Cl⁻ influx [36]. Thus, we assessed the ion transport activity of recSLC12A2 and its variants, based on the Cl⁻ influx rates determined within 15 sec after the injection of the high-Cl⁻ solution (Fig 4C). We found that the Cl⁻ influx rate for each of the variants (D981Y and P988T) and the truncated recSLC12A2-expressing cells was significantly lower than that of WT recSLC12A2-expressing cells, with values statistically indistinguishable from those of YFP-expressing cells (Fig 4D), suggesting that the Na⁺, K⁺, 2Cl⁻ cotransport activity of SLC12A2 was significantly compromised in cells expressing each of these variants.

The exon 21-encoded region of SLC12A2 contains a dileucine sorting motif, which is important for intracellular/membrane trafficking [37]. To gain further insights into the dysfunction of these recSLC12A2 variants, we examined the subcellular localization of the recSLC12A2 proteins in the HEK293T cells used for the Cl⁻ influx assay. As shown in Fig 5, the WT, variant, and truncated recSLC12A2 proteins all appeared to be successfully targeted to the plasma membrane. Therefore, the compromised transport activities of the recSLC12A2 constructs could be attributed to alterations in their function and not to impaired membrane targeting.

SLC12A2 signals in M. fascicularis cochlea

Immunohistochemical study using anti-SLC12A2 antibody performed on M. fascicularis cochleae demonstrated positive SLC12A2 signals in the stria vascularis, the spiral prominence, and parts of the spiral ligament (Fig 6A, S11 Fig). SLC12A2-positive cells in the spiral ligament were also positive for the type II/IV/V fibrocyte marker, ATP1B1 (also referred to as Na⁺, K⁺-ATPase β1; Fig 6A). These results demonstrate that the distribution of SLC12A2 in the M. fascicularis cochlea is essentially the same as reported for guinea pig, gerbil, and rat [38–40]. The spiral limbus and a small area of Reissner’s membrane connected to the spiral limbus were also positive for SLC12A2 (Fig 6A). No SLC12A2 signal was detected in the organ of Corti or epithelial cells of the spiral prominence (Fig 6A and 6B). In the strial marginal cells, SLC12A2 signals colocalized with those of ATP1B1 (Fig 6C) and were visible at the basolateral membrane, but not the apical surface, as shown by the absence of colocalization with KCNQ1, a specific marker for the apical membrane of strial marginal cells (Fig 6D. S11B Fig).

Discussion

Association of SLC12A2 with hearing loss in humans

In this study, we identified four candidate variants of SLC12A2 in four families with sensorineural hearing loss. In vitro studies demonstrated that a splice-site variant at the exon 21 acceptor site (c.2930‒2A>G) inhibited inclusion of exon 21 and increased generation of an exon 21-skipped transcript from the variant allele. Further, the exon 21-skipped transcript was expressed at very low levels in mammalian cochleae. HEK293T cells expressing recSLC12A2 proteins carrying each of the three variants showed significantly decreased Cl⁻ influx activity. Moreover, we observed abundant SLC12A2 signals in the basolateral membrane of strial marginal cells and lateral wall fibrocytes in non-human primate cochlea, a distribution identical to that reported for rodent cochleae [16], suggesting a conserved role for SLC12A2 in homeostasis of the endolymph by recycling K⁺ from the perilymph to the stria vascularis in mammalian cochlea, including in humans [3, 4]. The congenital, severe-to-profound hearing loss in our patients is consistent with the phenotype of Slc12a2-deficient M. musculus [16–19]. In addition to hearing loss, two of our patients (III-2 in family 2 and III-2 in family 3) presented with motor developmental delay. Since the patients with motor developmental delay caught up with expected milestones later, the feature was considered to be comorbid with vestibular impairment that could be compensated, unlike muscle, bone, or brain disorders, which cannot be compensated. Patients with hearing loss and symptoms derived from inner ear dysfunction have been categorized as having nonsyndromic hearing loss; for example, DFNA9 (MIM: 601369), DFNA11 (601317), and DFNB4 (600791) all exhibit vestibular abnormalities as a major clinical feature associated with hearing loss and are categorized as nonsyndromic hearing loss. Therefore, we propose that mutations in SLC12A2 are associated with autosomal-dominant nonsyndromic hearing loss.

SLC12A2 is the entry site for Na⁺, K⁺, and 2Cl⁻ from the intrastrial space to the strial marginal cells in the cochlear lateral wall [4]. Slc12a2-deficient M. musculus show loss of scala media [16–19], while homozygous Slc12a2 deficiency results in loss of endolymph volume, collapse of the otic vesicles with functional sensory hair cells, and a vestibular disorder phenotype in Danio rerio larvae [41]. The disrupted Cl⁻ influx in cells expressing the three variants reported here strongly suggests that the mechanisms underlying hearing loss include, at least in part, impaired production of endolymph secreted from the strial marginal cells, attributable to decreased activity of the Na⁺, K⁺, 2Cl⁻ cotransporter.

Intriguingly, several reports have argued that SLC12A2 is associated with other clinical features; for example, a heterozygous p.V1026Ffs*2 variant or a splice variant, affecting both exon 21-included and -skipped transcripts, might be associated with malformation of multiple organs without hearing loss [20], or global developmental delay [21], respectively. Further, missense variants mapping to exon 1 [22] or 17 [23] may increase the risk of morphological or functional disorders of the central nervous system. In addition, Slc12a2-deficient M. musculus exhibit hearing loss, vestibular defects, and additional symptoms, including small body size, hyper-excitability, and male infertility (reviewed in [29]). These additional phenotypes were not observed in the probands or in the other family members with hearing loss in this study. These differences could be attributable to the fact that all the candidate variants identified in this study are predicted to affect exon 21, without degeneration of the transcript. We speculate that the variants affecting the exon 21-encoded region without protein degeneration not only deteriorate cotransporter activity, but also affect an unrecognized function of SLC12A2, both of which are critical for hearing. This possibility awaits further investigation before firm conclusions can be drawn.

Difference in the inheritance pattern of hearing loss associated with SLC12A2 variants between humans and M. musculus

A discrepancy was noted between the results obtained with animal models and observations in our patients in the inheritance pattern of the hearing loss phenotype; our data indicate that heterozygous variants of SLC12A2 are associated with hearing loss in humans, whereas Slc12a2 is responsible for autosomal-recessive hearing loss in M. musculus [16–19]. One possible explanation for this difference is that the functional SLC12A2 derived from a single allele is not sufficient for normal hearing in humans, whereas it is sufficient in M. musculus; however, based on the assumption that SLC12A2 is responsible for monogenic disease, the idea that haploinsufficiency of SLC12A2 is associated with hearing loss is inconsistent with the fact that multiple loss-of-function variants of SLC12A2 have been detected in subjects with multiorgan dysfunction and normal hearing (p.V1026Ffs*2) [20] or registered in ExAC and gnomAD databases (such as p.G55Vfs*5, Y257Cfs*10, Q435*, R482*, P988Rfs*4, K1007Sfs*29, Q1099Ifs*3, Y1111*, and R1133*) [28, 42], although the allele frequencies of all the loss-of-function variants are very low (the highest MAF among loss-of-function variants in gnomAD is 6.49 × 10⁻⁵ for the p.G55Vfs*5 variant). All those nonsense and indel variants, including p.P988Rfs*4 in exon 21, are predicted to result in degeneration of the transcript by nonsense-mediated mRNA decay [43]. Therefore, although the reason for intolerance to loss-of-function variants of SLC12A2 in population databases remains unclear, haploinsufficiency does not appear to adequately explain the molecular mechanism underlying hearing loss caused by SLC12A2, without taking additional modifier factors into consideration [44].

An alternative explanation for the difference between human patients and Slc12a2-deficient mouse models is that variants affecting the exon 21-encoding region without degeneration of the transcript may alter as yet unidentified roles of SLC12A2, in addition to disrupting ion transport activity, leading to hearing loss as gain-of-function or dominant negative effect. It is intriguing to note that the expression level of the exon 21-skipped SLC12A2 transcript was significantly lower than that of the exon 21-included transcript in mammalian cochlear tissues, unlike previous reports in brain tissues [31, 32]. Our in vitro experiments suggest that the c.2930–2A>G variant leads to expression of the exon 21-skipped transcript and increased overall expression levels of this exon 21-skipped transcript in the cochlea. The exon 21-skipped or altered isoform of SLC12A2 may have roles in addition to ion transport that are beneficial for the central nervous system, but unfavorable for the cochlea. Slc12a2-knockout mice have been generated by elimination of both the exon 21-included and -skipped isoforms, by disruption of exon 9 [16], exon 6 [18], or exons 9–11 [19]. All these knockout strains are reported to result in degeneration of both Slc12a2 transcripts. Another Slc12a2-deficient mouse, with the c.2958dupA variant in exon 21 [17], results in absence of Slc12a2 signals in the cochlea, where the exon 21-included transcript is predominantly expressed. By contrast, the four variants found in our patients are considered to result in expression of the transcript without degeneration of the variant allele, or increased expression of the exon 21-skipped transcript, in the case of the c.2930–2A>G variant. Therefore the consequences of these variants would be expected to differ fundamentally from those reported for previous mouse strains. Hence, to verify the role of the exon 21 variants of SLC12A2 in hearing loss, generation and investigation of animal models carrying the equivalent variants is required.

The discrepancies between this study and a previous report demonstrating that the exon 21-skipped SLC12A2 isoform retains K⁺ influx activity comparable to the exon 21-included SLC12A2 isoform, determined using Rb⁺ in MDCK cells [32], may reflect methodological differences in the measurement of the activity of the exon 21-skipped SLC12A2 isoform.

Role of the exon 21-encoded region of SLC12A2

One possible mechanism of hearing loss by variants affecting exon 21 could be that they exert a dominant negative effect, in addition to resulting in decreased Cl⁻ influx activity. SLC12A2 and other proteins in the SLC12A family are thought to reside on the plasma membrane as homodimers [45–47] that form via association of the two carboxy-terminal domains [36, 47]. Whether the WT and variant and SLC12A2 molecules could generate a heterodimer, and thereby alter ion transport activity relative to the WT homodimer, warrants investigation in the future.

Another possible mechanism that could underlie hearing loss caused by the variants affecting exon 21 is that they may affect targeting of SLC12A2 to the basolateral membrane [37]. If variants affecting exon 21 alter the asymmetric distribution of properly functioning SLC12A2 on the strial marginal cells, this would affect its ability to secrete K⁺ into the endolymph; however, the choroid plexus, which predominantly expresses the exon 21-included variant, exclusively harbors SLC12A2 in the apical microvilli [48], indicating that asymmetric distribution is not fully dependent on the exon 21-encoded region. Whether targeting of SLC12A2 to the apical or basolateral membrane in the strial marginal cells is affected by variants in the exon 21-encoded region of SLC12A2 will be investigated in future experiments.

A third possible explanation for the effects of the exon 21 variants is that they may influence more complex properties of the strial marginal cells, through interaction with other molecules. In vitro studies have shown that SLC12A2 can be positively or negatively regulated through interaction of its carboxy-terminal region with the K⁺ and Cl⁻ cotransporters SLC12A7 (KCC4) [49] and SLC12A9 (CIP1) [50], respectively. Intriguingly, deficiency of Slc12a7 causes progressive hearing loss in M. musculus [51]. Identification of partners that interact with the exon 21-encoded region of SLC12A2 may provide insight into the tissue-specific properties of the ion transporter isoforms, as well as the molecular pathology of hearing loss associated with SLC12A2 variants.

The limitations of this study are as follows: 1) the small number of families investigated; additional cases and families, as well as careful clinical follow-up, are required to prove conclusively that SLC12A2 is associated with nonsyndromic hearing loss; 2) lack of an animal model for investigation of the specific variants identified in this study; a mouse model carrying the Slc12a2 variant equivalent to those identified in our patients should be generated and their phenotypes should be studied; 3) molecular effect of the variants on exon 21 is only partially explained by in vitro analysis; additional studies including transporter activity of the hetero dimers between WT and each variant, plasma membrane distribution in polarized cells, and identification of molecules associated with the exon 21-encoded region should be carried out.

In conclusion, whole-exome analysis or sequencing analysis targeting 154 deafness genes of four families with hearing loss identified SLC12A2 as a novel candidate deafness gene. The molecular mechanism of hearing loss could include dysfunction of the cotransporter activity of SLC12A2 caused by the exon 21 variants, and may reflect other affected roles of the exon 21-encoded region. Future studies of the molecular regulatory mechanism of splicing of exon 21, and molecules associated with the exon 21-encoded region, may lead to a more precise understanding of the function of SLC12A2 in the inner ear, as well as of mechanisms underlying hearing loss, and provide insight for future development of drugs to treat hearing loss related to SLC12A2 dysfunction.

Materials and methods

Ethics statement

This study was approved by the institutional ethics review board at the National Hospital Organization Tokyo Medical Center (approval number: R1-0703009), the ethics committee at the at the National Hospital Organization Mie Hospital (R1-0703009), the ethics committee at the National Center for Child Health and Development (661), the ethics committee at the RIKEN Center for Integrative Medical Sciences (YokohamaH17-16(27), YokohamaH24-15(3)), and the ethics committee at the Tokyo Medical and Dental University (O2015-503). Studies were conducted only after patients or their parents signed informed consent to participate. For in vitro experiments, all procedures were approved by the Institutional Safety Committee for recombinant DNA Experiments at the National Hospital Organization Tokyo Medical Center and the Institutional Biosafety Committee at Northwestern University. Collection of animal tissue samples was approved by the Institutional Animal Care and Use Committee at the National Hospital Organization Tokyo Medical Center.

Subjects

Medical histories and clinical information were obtained from probands and family members, when possible, by experienced otologists with the expertise in pediatric hearing loss (SM, NM, TM). All probands and/or their parents were interviewed to determine family history, onset of hearing loss, possible causes, progression, and comorbidity, and each individual underwent physical and otologic evaluations. Audiological examinations included pure-tone audiometry, auditory steady-state response, or conditioned orientation reflex audiometry, according to the age of the patient. Severity of hearing loss was determined according to the recommendations of the Genetic Deafness study group [52]. Temporal bones and inner ears were inspected by computed tomography. Physical findings and clinical data other than audiological or otological tests were obtained by visual inspection of the entire body, palpation of the head and neck, and interview of the affected individuals or their parents. Body weight and height, motor and neuronal development, as well as behavior and cognition were checked.

Genetic analysis

Genomic DNA samples were extracted from blood collected from the probands and their family members. Before conducting WES, probands were screened for GJB2 or mitochondrial m.1555A>G or m.3243A>G mutations, which are frequent causes of hearing loss in Japanese patients [26].

For families 1 and 2, genomic DNA was subjected to capture of whole-exome regions, using a Nextera Rapid Capture Exome kit (Illumina) [53], and to massively parallel sequencing, using the HiSeq2500 platform (Illumina). Paired-end read sequences were checked for quality and mapped onto the human reference genome with decoy sequence (hs37d5) using BWE-mem (v0.7.5a), and variants and genotypes were called using the Genome Analysis Toolkit 3.4.46 (GATK) [54] and HaplotypeCaller, after removal of duplicated reads using Picard (v1.106). Individual variants were joint-called together with in-house data (WES data, N = 498; and whole-genome sequencing data, N = 1,037) [55] using GenotypeGVCFs. Variants and genotypes were filtered and refined using the Variant Quality Score Recalibration workflow, followed by CalculateGenotypePosteriors and VariantFiltration, using previously obtained in-house population WGS data. In the case of the proband of family 3, whole-exome regions were captured using SureSelect Human All Exon v5 (Agilent Technologies) and sequenced using the HiSeq4000 platform (Illumina). Both procedures provided a mean sequence coverage of ≥ 100×, with > 95% of target bases having ≥ 20× coverage (S1 Table).

To identify candidate pathogenic variants, changes considered to affect the encoded protein were filtered according to minor allele frequency (MAF) (≥ 0.001 for autosomal-dominant inheritance mode, ≥ 0.003 for X-linked and autosomal-recessive inheritance mode) in public databases: Database of Single Nucleotide Polymorphisms (dbSNP142) [56], 1000Genomes [57], NHLBI Exome Variant Server (ESP6500) [58], Exome Aggregation Consortium (ExAC) [42], gnomAD [28], and the Human Genetic Variation Database (HGVD) ver1.42 for Japanese subjects [59]. In addition, variants from families 1 and 2 were filtered for variants from in-house data with MAF ≥ 0.003, while those from family 3 were filtered for variants with MAF ≥ 0.01 in 600 individuals and family members with or without congenital brain anomaly or hearing loss. The threshold MAF for autosomal-dominant variants was set to 0.001, according to the criterion of “automatically benign” variants for autosomal-dominant variants adopted by the CLINGEN hearing loss expert panel [60]. The threshold MAF for autosomal-recessive variants was set to 0.003 in population databases and 0.005 in Japanese population databases, to avoid excluding an established pathogenic variant in CDH23 (NM_022124: c.719C>T) [61, 62], which is present at an exceptionally high MAF in Japanese population databases (0.00248139 in HGVD [59] and 0.00327511 in our in-house data [55]).

The remaining variants were prioritized into three categories prior to analysis of their co-segregation with the disease: Tier 1 genes, 293 genes reported in OMIM as associated with nonsyndromic or syndromic hearing loss (S2 Table); Tier 2 genes, 328 genes associated with hearing loss in animal models by the Mouse Genome Informatics [63] or International Mouse Phenotyping Consortium [27], and not included in Tier 1 (S3 Table); and Tier 3 genes, all other genes. The effect of variants on splicing was predicted using Human Splice Finder 3.0 [24]. Candidate variants were validated by Sanger sequencing, using PrimeSTAR polymerase (TaKaRa BIO). Samples from individual I-3 in family 1 and III-2 in family 4 were subjected to targeted analysis of 154 deafness genes (S7 Table), which was modified from the previous study [11]. The analytical procedures including thresholds of MAF to filter candidate pathogenic variants were identical to those used for WES analysis. For paternity testing in family 3 and 4, the STR markers D1S80 (MCT118) and D17S5 (YNZ22) were amplified using AmpliTaqGold 360 (ThermoFisher Scientific) and analyzed using a BioAnalyzer 2100. Primers used for validation of candidate variants by Sanger sequencing and STR marker analysis are shown in S8 Table.

Splicing and gene expression analyses of SLC12A2 transcripts

A genomic region spanning from within intron 20 of SLC12A2 to within intron 22 (2,507 bp) was amplified from the proband of family 2 and inserted into the vector pSpliceExpress (Addgene: 32485) [30]. Host HEK293T cells (RIKEN Bio Resource Center) were transfected with each vector using Lipofectamine 2000 (ThermoFisher Scientific) and incubated for 2 days (three independent experiments). Total RNA was extracted using the RNeasy mini kit (QIAGEN) and reverse-transcribed with SuperScript III (ThermoFisher Scientific). PCR was carried out at 35 cycles of 98°C for 10 sec, 60°C for 10 sec, and 72°C for 30 sec, followed by 72°C for 5 min, using PrimeSTAR polymerase (TaKaRa BIO).

To quantify the endogenous expression levels of human, M. fascicularis, and M. musculus SLC12A2/Slc12a2 transcripts, specific primers to distinguish the exon 21-included transcript (variant 1 in human (NM_001046.2), variant X1 in M. fascicularis (XM_005557674.2), and variant 1 in M. musculus (NM_009194.3)) from the exon 21-skipped transcript (variant 2 in human (NM_001256461.2), variant X2 in M. fascicularis (XM_005557675.2), and variant X1 in M. musculus (XM_006525732.3)) were designed and analyzed in triplicate experiments, using a Power SybrGreen kit (Agilent Technologies) and the QuantStudio3 system (Agilent Technologies), according to the manufacturer’s protocols. All primer sets used to amplify the genomic region or transcripts of SLC12A2, as well as the glyceraldehyde-3-dehydrogenase (GAPDH) housekeeping gene, are listed in S8 Table. cDNA templates were generated from human brain total RNA (TaKaRaBIO) and total RNA extracted from M. fascicularis cochleae (a male, 3 years old, from the Primate Research Center, Ibaraki, Japan [64]), and various tissues (cerebral cortex, cerebellum, cochlea, choroid plexus, eye, spleen, liver, intestine, kidney, and ovary) from M. musculus strain FVB/NJ (both sexes except ovary, 4–12 weeks old, N = 3). RT-PCR was carried out using the Type-iT Multi Detection PCR kit (QIAGEN). The PCR conditions were as follows: 95°C for 5 min, followed by 40 cycles of 95°C for 30 sec, 54°C for 90 sec, and 72°C for 40 sec, and a final cycle of 68°C for 5 min. Statistical analysis was conducted by one-way ANOVA, combined with the Tukey-Kramer test. P < 0.05 was considered statistically significant.

Cl⁻ influx assay of cells transfected with recSLC12A2 constructs

A plasmid expressing WT SLC12A2, amino-terminally attached to Cl⁻-sensitive YFP (NT13) [34], was purchased from Addgene (49060) and cloned into the vector pSBtet-Pur (Addgene: 60507) [65]. The D981Y and P988T missense mutations were introduced by site-directed mutagenesis. A recSLC12A2 construct without the exon 21-encoded region, representing the expected transcript generated due to the c.2930–2A>G mutation (Δex21), was generated by multistep PCR [66]. An mTurquoise2 (mTq2) construct, cloned into the vector pSBbi-Bla (Addgene: 60526), was also generated, and the mTq2 fluorescence was used as a reference to correct for fluctuations in the observed YFP fluorescence. Stable HEK293T cell lines were established using these constructs, as previously described [65]. Expression of recSLC12A2 and its variants was induced by the application of doxycycline (1 μg/ml) 1 day prior to experiments. mTq2 was constitutively expressed. Cells were cultured in DMEM containing 10% FBS and penicillin/streptomycin (100 U/ml). Prior to the Cl⁻ influx assay, cells were cultured in suspension for 30 min at room temperature in a low-Cl⁻ buffer, containing 135 mM sodium gluconate, 5 mM KCl, 0.5 mM CaCl₂, 0.5 mM MgCl₂, and 15 mM HEPES (pH 7.4). An aliquot (150 μl) of each cell suspension was transferred to a 96-well plate (approximately 1.5 × 10⁵ cells/well). The Cl⁻ influx assay was initiated by an automated injection of 100 μl of a high-Cl⁻ buffer, containing 135 mM NaCl, 5 mM KCl, 0.5 mM CaCl₂, 0.5 mM MgCl₂, and 15 mM HEPES (pH 7.4), in a Synergy Neo2 plate reader (BioTek). The fluorescence intensities of YFP (F_YFP) and mTq2 (F_mTq2) were simultaneously measured over time. The fluorescence ratio, F_YFP/F_mTq2, was plotted against the corresponding measurement time, and the ion transport activity (sec⁻¹) was determined by exponential curve-fit analyses. All band-pass and dichroic filters and mirrors were purchased from Chroma Technologies. One-way ANOVA, combined with the Tukey-Kramer test, was performed for multiple comparisons, and p < 0.05 was considered statistically significant.

Cell imaging

For fluorescence imaging of the recSLC12A2 constructs, cells were fixed in 4% formaldehyde for 5 min at room temperature. Images were captured using an A1R confocal microscope with a Plan Apo 60× oil objective (Nikon).

Immunohistochemistry

Cochleae were dissected from saline-perfused, formalin-fixed temporal bones of young adult female M. fascicularis (3 years old, N = 3) and decalcified with 0.1 M EDTA/phosphate buffer at 4°C for at least 6 weeks, then embedded in paraffin. The 5 μm sections were rehydrated, pretreated with sodium citrate buffer, then blocked in PBS containing 5% FBS, 1% BSA, and 0.05% Tween-20. The primary antibodies used in this study were rabbit antiserum against SLC12A2, generated using the amino-terminal region of the protein as an immunogen, and thus able to detect both the exon 21-included and -skipped isoforms (Abcam: ab59791, 1:100), mouse monoclonal antibody against ATP1B1 (Santa Cruz Biotechnology: sc21713, 1:400), and goat antiserum against KCNQ1 (Santa Cruz Biotechnology: sc10646, 1:200). Sections were incubated with primary antibodies at 4°C overnight, and signals were visualized using appropriate secondary antibodies, conjugated with either Alexa 488 or 568, followed by counterstaining with DAPI. Images were captured by fluorescence (DM2500, Leica Microsystems) and confocal (Axio 700) microscopy.

Supporting information

S1 Fig [c]
Audiograms and estimated audiograms for family members in this study.

S2 Fig [pdf]
Partial electropherograms generated by Sanger sequencing.

S3 Fig [a]
Parenthood testing in family 2 and 3.

S4 Fig [a]
Non-candidate variants in family 3.

S5 Fig [a]
variant found in III-2 of family 4.

S6 Fig [pdf]
Partial results of multiple alignment of eight SLC12A family proteins.

S7 Fig [a]
Expression levels of transcripts determined by qRT-PCR.

S8 Fig [a]
Assessment of transcripts in mammalian tissues by RT-PCR.

S9 Fig [a]
Summary of expression in mammalian tissues.

S10 Fig [pdf]
Junction expression of in human tissues.

S11 Fig [a]
Histology of . cochlear tissues.

S1 Table [pdf]
Summary of whole-exome sequencing in this study.

S2 Table [pdf]
List of genes categorized in Tier 1 in this study.

S3 Table [pdf]
List of genes categorized in Tier 2 in this study.

S4 Table [pdf]
Primary candidate variants co-segregated with hearing loss in family 1.

S5 Table [pdf]
Primary candidate variants found in family 2 with hearing loss.

S6 Table [pdf]
Primary candidate variants found in family 3 with hearing loss.

S7 Table [pdf]
List of deafness genes for targeted sequencing.

S8 Table [pdf]
Primers used in this study.