Biallelic mutation of CLRN2 causes non-syndromic hearing loss in humans

Deafness, the most frequent sensory deficit in humans, is extremely heterogenous with hundreds of genes probably involved. Clinical and genetic analyses of an extended consanguineous family with pre-lingual, moderate-to-profound autosomal recessive sensorineural hearing loss, allowed us to identify CLRN2, encoding a tetraspan protein as a new deafness gene. Homozygosity mapping followed by exome sequencing identified a 15.2 Mb locus on chromosome 4p15.32p15.1 containing a missense pathogenic variant in CLRN2 (c.494C>A, NM_001079827.2) segregating with the disease. Using in vitro RNA splicing analysis, we show that the CLRN2 c.494C>A mutation leads to two events: 1) the substitution of a highly conserved threonine (uncharged amino acid) to lysine (charged amino acid) at position 165, p.(Thr165Lys), and 2) aberrant splicing, with the retention of intron 2 resulting in a stop codon after 26 additional amino acids, p.(Gly146Lysfs*26). Expression studies and phenotyping of newly produced zebrafish and mouse models deficient for clarin 2 further confirm that clarin 2, expressed in the inner ear hair cells, is essential for normal organization and maintenance of the auditory hair bundles, and for hearing function. Together, our findings identify CLRN2 as a new deafness gene, which will impact future diagnosis and treatment for deaf patients.


Disclosure
The authors declare no conflict of interest.

Introduction
The mammalian inner ear is an exquisite and highly complex organ, made up of the vestibule, the organ responsible for balance, and the cochlea, the sensory organ for hearing. The auditory sensory cells of the inner ear are called the inner and outer hair cells that are responsible for transduction of sound wave-induced mechanical energy into neuronal signals (1, 2). The functional mechanoelectrical transduction machinery involves intact formation and maintenance of a highly specialized and organized structure, the hair bundle. The hair bundle contains a few dozen F-actin-filled stereocilia, arranged in a highly interconnected and highly organized staircase-like pattern, which is critical for function (3). Knowledge of the mechanisms of formation, maintenance, and function of the transduction complex is limited (4). In this regard, identification of novel genes that encode protein products essential for hearing is likely to improve our understanding of the physical, morphological and molecular properties of hair cells and associated mechanistic processes.
Hereditary hearing loss is one of the most common and genetically heterogeneous disorders in humans (5). Sensorineural hearing loss has an incidence of one to two per 1000 at birth (6). It displays extraordinary phenotypic, genetic and allelic heterogeneity, with up to 1,000 different genes potentially involved (7). So far, about 120 genes and more than 6,000 disease causing variants (8)  more are yet to be discovered. Genetic factors predominate the etiological spectrum and most of hereditary hearing loss appears to follow an autosomal recessive inheritance pattern (9). Approximately 80% of the currently known autosomal recessive genes have been identified by studying extended consanguineous families (10). There are many forms of hearing loss that are clinically indistinguishable but caused by distinct genetic entities that are presently unknown. Identification of additional genes essential for auditory function, through the study of families exhibiting hereditary hearing loss, will not only help increase our understanding of the biology of hearing, but will also identify new molecular targets for therapeutic intervention.
Through the study of an extended consanguineous Iranian family, we have identified a CLRN2 coding lesion as the cause of hearing loss in family members that are homozygous for the allele. We have established that clarin 2 likely plays a critical role in mechanotransducing stereocilia of the hair bundle in zebrafish and mouse.
CLRN2 belongs to the clarin (CLRN) family of proteins that are comprised of three orthologues named clarin 1, 2, and 3 that encode four-transmembrane domain proteins.

Genotyping, homozygosity mapping, copy number variation and exome sequencing data analyses
Due to parental consanguinity and suspected autosomal recessive mode of inheritance, we assumed that the causative variant would be homozygous and identical by descent in affected individuals in the fourth generation of the family. Blood samples from 14 family members were obtained and genomic DNA was isolated from whole blood using standard procedures. DNA from affected (IV-1, IV-6, and IV-8) and unaffected (IV-2, IV-3, IV-4, and IV-5) individuals were genotyped using the Infinium Visualization was performed using the Integrative Genomics Viewer. Analysis of homozygous and compound heterozygous variants between the two sequenced affected individuals (IV-6 and IV-1) followed. We analyzed missense variants by using a combination of criteria that scored conservation using GERP++ and PhyloP, and deleterious or pathogenic scores in Combined Annotation Dependent Depletion (CADD) (18), LRT (19), MutationTaster (20), PolyPhen-2 (21), and SIFT (22). Missense variants were excluded when three out of five in silico pathogenicity prediction tools yielded a benign score. Manual MAF analysis used gnomAD (23), GME (24), and Iranome (25).

Segregation, sequence and in vitro
(UniProt: A0PK11) were aligned and visualized in Jalview (28) with an overview of the pathogenic and likely pathogenic missense and nonsense CLRN1 variants retrieved from the Deafness Variation Database v 8.2 (8).
In addition, secondary protein structure prediction of human CLRN2 (NP_001073296.1) that included the wild-type (WT) and mutated amino acid residues was performed using I-TASSER (29).
To assess the splicing effect of the c.494C>A variant, in vitro splicing assays, also called mini-genes, were carried out as described (30,31). Concurrently, the mini-gene splice assay experiment was conducted in a doubleblind manner as previously described (32). Genomic DNAs of an affected homozygous (IV-6) and WT individual (IV-5) were amplified using a forward primer with a XhoI . All zebrafish handling, embryo care, and microinjections were performed as previously described (33). WT zebrafish strain NHGRI-1 was used for all experiments (34). The zebrafish embryonic staging was determined by morphological features according to Kimmel et al (35).
To produce zebrafish clrn2 crispants, the sgRNA target sequences were selected from the UCSC genome browser tracks generated by the Burgess lab. Five independent targets were chosen and sgRNAs were synthesized by in vitro transcription as described earlier (36). sgRNAs and Cas9 protein complex were used to generate indels. A 6 µL mixture containing 2 µL of 40 µM Spy Cas9 NLS protein (New England Biolabs, MA, USA), 200 ng each of five sgRNAs (in 2 µL) and 2 µL of 1 M potassium chloride was injected into one-cell-stage WT embryos. Injection volumes were calibrated to 1.4 nL per injection. Insertion/deletion (indel) variants were detected by amplifying the target region by PCR and Sanger sequencing as described earlier (36).
The sequencing data were analyzed by Inference of CRISPR Edits (ICE) v2 CRISPR analysis tool. The sgRNA target sequences and PCR primer sequences are listed in Table S1.

Zebrafish RNA extraction and real-time quantitative PCR (RT-qPCR)
Total RNA at different developmental stages, adult tissues, and CRISPR/Cas9 All RT-qPCR primer pairs were designed across exon-exon junctions using NCBI Primer-BLAST program. The sequences are listed in Table S1. The PCR cycling conditions were used as per the manufacturer instructions, and the amplification specificity was assessed by dissociation curve analysis at the end of the PCR cycles.
The cycle threshold values (Ct) data were imported into Microsoft Excel for the relative gene expression analysis. Quantification was based on 2^(-ΔΔCT) method (37), and using 18 hours post fertilization (hpf) for clarin 2 temporal expression, muscle for clarin 2 in different tissue expression and the corresponding age-matched control for clarin 2 CRISPR injected F 0 larvae as normalization control.

Distribution of clrn2 and phalloidin staining in zebrafish
To determine clrn2 expression, we used in situ hybridization on larvae and inner ear-containing cryosections. The full-length coding sequence of zebrafish clarin 2 (NM_001114690.1) was PCR amplified from WT zebrafish cDNA using primer pairs with restriction enzymes BamHI and XhoI restriction sites cloned into pCS2+ vector (a kind gift from Dr. Dave Turner, University of Michigan). After restriction digestion, the resulting clones were sequenced and used as templates for riboprobe synthesis. The digoxigenin-UTP-labeled riboprobes were synthesized according to the manufacturer's instructions (Millipore Sigma, MO, USA). Briefly, the clarin 2 and the pvalb9 plasmids (Horizon Discovery) were linearized by BamHI and NotI restriction enzymes, respectively. The linearized plasmid was purified and used as template for in vitro transcription using T7 RNA polymerase to synthesize anti-sense probes. The sense probe was made using XbaI linearized clarin 2 plasmid and SP6 RNA polymerase.
The whole-mount in situ hybridization (WISH) on 3 and 5 dpf zebrafish embryos/larvae was performed following the procedures as described by Thisse et al.
with minor modifications (38). Age-matched zebrafish embryos were randomly collected by breeding WT zebrafish pairs. The embryos were treated with 0.003% phenylthiourea (PTU) (Millipore Sigma, MO, USA) in embryo medium at 1-day post-fertilization (dpf) until the desired stages reached to reduce the pigment formation that will facilitate color visualization during in situ hybridization. Embryos/larvae were then fixed with 4% (V/V) paraformaldehyde in phosphate-buffered saline (PBS) at 3 and 5 dpf. An additional bleaching step was carried out after fixation by incubating the embryos at room temperature in a 3% hydrogen peroxide and 0.5% potassium hydroxide solution. The permeabilization of 3 dpf embryos and 5 dpf larvae were using 2 µg/mL proteinase K for 12 and 18 minutes respectively. Color development was conducted using the BM-Purple alkaline phosphatase substrate (Millipore Sigma, MO, USA).
For preparation of cryo-section samples after WISH, the 5 dpf larvae were soaked in 25%, 30% (V/V) sucrose/PBS and optimum cutting temperature (OCT) each for at least two days, and embedded in OCT then Cryotome sectioned at a 10micrometer thickness.

Production and phenotyping of clarin 2 deficient mutant in mice
The

Identification of CLRN2 as a novel deafness gene in a consanguineous Iranian family exhibiting autosomal recessive non-syndromic sensorineural hearing loss
A three generation Iranian family of Lurs ethnicity was ascertained as part of a large ethnically diverse Iranian population rare disease study (Fig. 1A). Three individuals that included the proband (IV-6), his sibling (IV-1), and a cousin (IV-8), born form consanguineous marriages, have reported moderate-to-profound bilateral nonsyndromic sensorineural hearing loss (Fig. 1B). The age of onset for these three individuals was between 2 and 3 years of age. Pure-tone air-and bone-conduction audiometry thresholds (Fig. 1B) (Fig. 1C, Fig. S1A, Table S2). This locus contains 30 genes, none of which are presently associated with deafness in humans (Table S3). This approach also revealed four much smaller homozygous intervals on chromosomes 2p21 (137.3 kb), 3p22.2 (262.5 kb), 13q13.1 (90.7 kb), and 17q21.31 (292.6 kb) (Fig. S1A, Table S2) that do not contain known deafness-associated genes (Table S3). Pathogenic copy number variations were excluded. Next, we undertook exome sequencing of affected individual IV-6 (arrow, Fig. 1A). This generated 56,387,543 mappable reads, with 75.5% on-target reads. The mean depth was 57.3-fold, with 97.3% of regions with a 10-fold read depth.
Analysis of the exome data of individual IV-6 excluded any candidate pathogenic variants in known deafness-associated genes (42) prompting an exome-wide analysis followed by filtering and re-analysis of variants in homozygous intervals (Table S4).
Further, close inspection of the exome sequencing data revealed complete sequencing coverage of genes in the homozygous intervals (Table S5) conserved across species and the corresponding amino acid in clarin 1 is a serine residue ( Fig. 2A,B), a scenario often associated with conserved phosphorylation site residue, here by serine/threonine protein kinases (43). here. Since exon 3 is the last exon of CLRN2, we designed our PCR primers to exclude the human poly-A signal and used the poly-A signal native to the pET01 vector. As expected for WT CLRN2 (c.494C), we detected the splicing of the 5' native pET01 exon only to exon 3 of CLRN2 (Fig. 3A, B). The same normal splicing was obtained in all cell types transfected with CLRN2 containing the control (rs117875715) variant (Fig. 3B).
However, the c.494C>A variant yielded two bands; one ~650 bp band matching the expected normally spliced exon, and a second abnormal band that was approximately ~1,360 bp (Fig. 3B). Sequencing of these amplicons validated normal splicing including the c.494A variant and also revealed a retained intron 2 in the aberrantly spliced transcript (Fig. S3C). The retention of intron 2 results in a new reading frame that introduces a stop codon 26 amino acids after the native exon 2 splice site (p.(Gly146Lysfs*26)) (Fig. 3C). These results were replicated using the pSPL3b vector and HEK 293T cells (Fig. S3A-C), confirming the c.494C>A induced normal and aberrant splicing, independent of the cell type context. Following TA-cloning of cDNA amplicons from the homozygous individual (from Fig. S3B), 23 of 26 amplicons (88.5%) showed normal splicing, and 3 of 26 amplicons (11.5%) showed a retained intron. The mini-gene splicing assays and sequence analyses clearly show that the c.494C>A affects a highly conserved and key residue in clarin 2 sequence, while also creating aberrant mRNA splicing in vivo likely leading to a truncated protein. Altogether, this further confirms that variants in CLRN2 can lead to sensorineural hearing loss.

Clrn2, a hair cell expressed gene key to hearing also in zebrafish and mice
To further study the role of clarin 2 in the inner ear, we investigated its expression and analyzed potential impact of Clrn2 loss-of-function in two other species, zebrafish and mice.

clrn2 in zebrafish
Taking advantage of larva transparency, we used zebrafish as a model to investigate the clarin 2 expression during early embryonic development. The RT-qPCR at different developmental stages revealed that clrn2 mRNA was first detected at 18 hpf (Fig. 4A), a stage when the otic placode begins to form the otic vesicle in zebrafish (this stage is similar to mouse embryonic day 9 (E9), a stage of otic placode formation) (48,49). clrn2 mRNA expression increased (2-fold at 72 and 96 hpf compared to 18 hpf) and was maintained at later stages, up to 120 hpf (Fig. 4A). Comparative analyses of clrn2 mRNA expression in different adult tissues of zebrafish revealed a significant enrichment in utricle, saccule and lagena of the inner ear (Fig. 4B). Our data are in agreement with RNA expression data from the Genotype-Tissue Expression (GTEx) project, wherein CLRN2 mRNA in humans is enriched in the nervous system, testis, kidney, salivary gland, and lung. CLRN1 has a similar expression profile in humans.
To determine clrn2 cellular expression, we used WISH in the inner ear of 3-and 5-dpf embryos (Fig. 4C, D). Unlike the clrn2 sense probe, the anti-sense clrn2 revealed strong expression in the otic vesicle, similar to the expression of anti-sense pvalb9, used as a marker of hair cells (Fig. 4C). Histological examination of 5 dpf embryos further confirmed that clrn2 expression is more specifically, restricted to hair cells, and is not expressed in the supporting cells of the inner ear (Fig. 4D).
To elucidate the function of clrn2 in zebrafish, we used CRISPR/Cas9 to generate loss-of-function alleles. To maximize the knockout efficiency, we used five sgRNAs targeting the first and second exon of clrn2 gene (Fig. S4). Injected embryos (crispants) were sequenced and, as expected, a mix of alleles in the form of deletions ranging from 4 bp to 73 bp, as well as insertions spanning +1 to +11 bp were observed.
The majority of the variants were frameshift that would most likely create a premature stop codon in the protein (Fig. S4). The RT-qPCR analyses on injected embryos showed that clrn2 crispants have a significantly reduced amount of clrn2 mRNA (Fig.   4E), suggesting nonsense mediated decay, leading to disrupted clarin 2 protein function.
Considering the expression in hair cells (Fig. 4D), we investigated the mechanosensory structures of the hair cell bundle, which are important for hearing and balance function in zebrafish. Interestingly, fluorescent phalloidin staining of the hair bundles of the inner ear in clrn2 crispants (n=10) showed disrupted hair bundle structure compared to the WT controls; the hair bundles are splayed, thin and split in clrn2 crispants (arrowheads in Fig. 4E). This defective phenotype, suggesting a critical role in hair bundle structures, is similar to the hair bundles in zebrafish clrn1 knockouts (50), the orbiter mutants (defective in protocadherin 15 (pcdh15), a gene associated with human Usher syndrome 1F) (51) and ush1c morphants and ush1c mutants (52).

Clrn2 in mice
To further assess the requirement of clarin 2 for auditory function in mammals, and assess further its role in auditory hair bundles, we extended our analyses to mouse.
To investigate stereocilia bundle morphology in Clrn2 del629/del629 mice, we used scanning electron microscopy to examine the cochlear sensory epithelia. At P28 (± 1 day), the inner and outer hair cell stereocilia bundles of Clrn2 mutant mice display the expected U-and V-shape, respectively, which contrasts with the grossly misshapen OHC bundles found in Clrn1 mutant mice (Fig. 5C). However, while the patterning of the bundles appears normal in Clrn2 del629/del629 mice the heights of their middle and short row stereocilia are visibly more variable compared with those of Clrn2 +/+ littermates, and many of the short row 'mechanotransducing' stereocilia are missing (Fig. 5C).  Together, our findings establish that clarin 2 is key to hearing function in zebrafish and mouse, and likely has a key role in the mechanotransducing stereocilia of the hair bundle.

Discussion
We identify CLRN2 as a novel deafness gene in human and zebrafish and describe a new deafness-causing allele in mice. Genetic study using homozygosity from moderate-severe (IV-6) to profound (IV-1) deafness (Fig. 1A, B).
The c.494C>A variant affects an amino acid that is highly conserved among PMP-22/EMP/EP20/Claudin superfamily proteins ( Fig. 2A-C). In addition, the c.494 cytosine is highly conserved and the exchange to adenine is predicted to create an ESE site that likely impacts splicing efficiency in humans (Fig. S2A, B) but not zebrafish (Fig.  S2C). We confirmed the effect on splicing using mini-gene assays. We showed that the c.494C>A variant acts in two ways: 1) as a missense variant (p.Thr165Lys) producing a mutant full length protein and 2) as a splice variant leading to intron retention (Fig. 3B, and Fig. S3B, C) expected to cause a premature stop codon 26 amino acids into intron 2 (p.Gly146Lysfs*26) (Fig. 3C).
Variants that disrupt splicing machinery signals can impact accurate recognition and removal of intronic sequences from pre-mRNA (27) and are recognized as significant contributors to human genetic diseases (54). ESE sequences are cis-acting elements primarily recognized by the SR family proteins that function by recruiting core splicing machinery components to splice sites or by acting antagonistically against nearby silencing elements (27,55,56). ESEs are often associated with introns that contain weak splicing signals, but they can also reside in exons and impact the splicing process.
Two potential mechanisms could synergistically contribute to the disruptive effect of the missense variant. First, the replacement of threonine with lysine, an amino acid with a positively charged 'bulky' side chain (lysine), may affect protein folding (43) and transport to the plasma membrane. Membrane proteins sort to the plasma membrane via the conventional secretory pathway associated with ER-to-Golgi complex (57).
Misfolded membrane proteins are typically retained in the endoplasmic reticulum (ER) and degraded by the ER-associated degradation pathway (58,59). It is possible that a small fraction of the misfolded clarin 2 p.(Thr165Lys) could reach the plasma membrane via the unconventional secretory pathway, similar to that reported for clarin 1 p.(Asn48Lys) (p.(N48K)) (60). The unconventional secretory pathway is induced by the ER-associated misfolded or unfolded protein response (61,62). However, the mutant clarin 2 reaching the surface may be functionally inactive. Second, evolutionarily conserved threonine residues are also conserved protein phosphorylation sites.
Phosphorylation adds a negative charge to the side chain of the amino acid and it serves as an important post-translational mechanism for regulation of protein function This, together with the severe-to-profound hearing loss already exhibited at P21 in Clrn2 mutant mice points to gene defects likely affecting both inner hair cells (IHCs) and OHCs. This is further supported by scanning electron microscopy data showing loss of shortest row stereocilia in both the cochlear IHCs and OHCs (Fig. 5C). Phalloidin staining of clrn2 crispants also confirms hair bundle abnormalities in zebrafish.
In regard to the observed progressivity of the hearing impairment in clarinet mice (39), the earliest reported clinical diagnosis of hearing loss of the CLRN2 affected individuals in the family we present is between 2 and 3 years of age. Newborn hearing screening was not routinely performed when the affected individuals were born, so we cannot confirm hearing was normal at birth. In light of absent serial audiograms, we cannot report if the hearing loss experienced in these patients is progressive, as is observed in the mouse model (39).
In conclusion, we demonstrate the c.494C>A variant affects exon 3 splicing efficiency. We showed, for the first time, that CLRN2 is a deafness-causing gene in humans. A variant causes hearing loss in humans, replicated by animal studies.
Additional reports of families segregating CLRN2 biallelic variants will be crucial to refine and dissect the clinical course and characteristics of hearing loss due to this gene. Together, our studies in zebrafish and mice establish that hearing loss is probably due to defective protein in the hair cells, where the presence of clarin 2 is essential for normal organization and maintenance of the mechanosensitive hair bundles.