Autosomal Dominant Non-Syndromic Hearing Loss Maps to DFNA33 (13q34) and Co-Segregates with Splice Site Variants in ATP11A, A Phospholipid Flippase Gene

Whole genome approaches are superior for identifying recessive genes, however discovery of dominant genes including deafness genes (DFNA) remains challenging. Herein we report a new DFNA gene, ATP11A, in a Newfoundland family with a variable form of bilateral sensorineural hearing loss (SNHL). Targeted screening of DFNA genes based on audioproles was unsuccessful. Genome-wide SNP genotyping linked SNHL to DFNA33 (Lod = 4.77), a locus on 13q34 previously mapped in a German family with variable SNHL in 2009. WGS identied 51 unremarkable positional variants on 13q34. Continuous clinical ascertainment identied several key recombination events and reduced the disease interval to 769 Kb, excluding all but one variant. ATP11A (NC_000013.11: g.190616G > A) is a novel point mutation predicted to be a cryptic donor splice site. RNA studies in patient-derived tissues veried in silico predictions, revealing the retention of 153bp of intron in the 3’ UTR of several ATP11A isoforms. A second, unresolved family from Israel with a similar, variable form of SNHL and a novel duplication in exon 28 of ATP11A that occurs within the splice donor sequence (intron 28). ATP11A is a type of P4-ATPase that transports (ip) phospholipids from the outer to inner leaet of cell membranes to maintain asymmetry. Haploinsuciency of ATP11A, the phospholipid ippase that specially transports phosphatidylserine (PS) and phosphatidylethanolamine (PE), could leave cells with PS/PE at the extracellular side vulnerable to phagocytic degradation. Given that surface PS can be pharmaceutically targeted, hearing loss due to ATP11A could potentially be treated. It is also likely that ATP11A is the gene underlying DFNA33.


Introduction
Hereditary hearing loss is a common sensory disorder exhibiting extensive genetic and clinical heterogeneity (Morton and Nance 2006). Over 200 hearing loss genes have been identi ed (Van Camp G 2015); however, approximately one-third of the 60 mapped autosomal dominant loci (DFNA) evade discovery. Like other autosomal dominant traits, hearing loss is typically characterized by variable expressivity and reduced penetrance (Richard JH Smith 1999). Whole-genome sequencing (WGS) in patients with dominant conditions generates many heterozygous variants, which is particularly problematic when the underlying gene is novel. Well ascertained multiplex families and more traditional methods (e.g., linkage and haplotype analysis) can be critical to reduce the number of candidate variants segregating. In a similar way, well ascertained families provide truly unaffected members that can be used to lter out benign variants. Genetic isolates can also help where apparently unrelated families are members of a clan (Sherry et al. 2001), as well as the fact that sibships are usually large and genealogical connections with distant relatives are known. With the increased power to detect rare variants using WGS compared with whole-exome sequencing (WES), WGS analysis should include all known isoforms of candidate genes, and not be limited to validated isoforms (i.e., RefSeq) (Belkadi et al. 2015;Zhao and Zhang 2015).
A key feature of eukaryotic cell membranes is the non-random (asymmetrical) distribution of phospholipids, an essential feature to maintaining cell integrity (Segawa et al. 2014). Asymmetry is most evident at the plasma membrane (Zachowski 1993) and is maintained by the action of three classes of proteins: scamblases, oppases and ippases. Phospholipid ippases (or P4-ATPases) speci cally transport or " ip" phospholipids from the outer to the inner lea et (Paulusma and Elferink 2010). Mouse studies have demonstrated the importance of the P4-ATPase class of ippases in phospholipid metabolism and maintaining normal auditory function (Coleman et al. 2014;Stapelbroek et al. 2009). For example, pathogenic variants in ATP8B1 (MIM: 605868) cause intrahepatic cholestasis type 1 (MIM: 211600), where patients sporadically develop hearing loss (Stapelbroek et al. 2009). Phospholipids have also been implicated in autoimmune conditions, such as antiphospholipid syndrome characterized by the presence of antiphospholipid antibodies that invoke an autoimmune response causing thrombosis, and complications during pregnancy and hearing loss (Mouadeb and Ruckenstein 2005;Wiles et al. 2006). We map SNHL in a family of Northern European descent to the DFNA33 locus (OMIM 614211) and identify two novel splicing mutations (one likely pathogenic and one unknown) in the 3' region of the ATP11A gene (OMIM: 605868) in two unrelated families with variable SNHL.

Materials And Methods
For the discovery phase, a six-generation family of Northern European descent was ascertained from the province of Newfoundland and Labrador (NL), Canada (Fig. 1). The proband (PID IV-7) presented at age 13 years with a progressive, sloping, bilateral sensorineural hearing loss and a family history consistent with autosomal dominant inheritance. Validation in a second pedigree, representing a four-generation family from Israel with roots in Afghanistan was studied at the University of Tel Aviv (Fig. 2). The proband (PID IV-1) is a 29 year old female with hearing loss who was seeking genetic counseling to genetically diagnose the cause of familial hearing loss. At age 17 years she had normal low-mid frequency hearing and a high frequency bilateral sensorineural hearing loss, sloping above 1 kHz. In both families, hearing loss was evaluated using standard pure tone measurements of air and bone conduction thresholds (Fig. 3a- For the discovery phase, genomic DNA was extracted from peripheral blood of the NL family using a modi ed salting out protocol (Miller et al. 1988) and screening was performed by bi-directional Sanger sequencing and analyzed using Mutation Surveyor Software (v5.0, SoftGenetics LLC State College, PA, USA). The NL proband was Sanger sequenced screened for all pathogenic variants causing hearing loss in this genetic isolated population, followed by select autosomal dominant genes matching the proband's (PID IV-7) audiometric data with that of reference audiopro les of 34 autosomal dominant deafness loci

Hearing Loss Phenotype
In the discovery family, sixteen members exhibit bilateral SNHL. Longitudinal audiograms on PIDs IV-4, IV-6, IV-7, IV-9, IV-12, IV-14, V-4, V-5, V-6, V-9, and V-11 reveal a progressive loss with variable onset and con guration, with auditory pro les ranging from high frequency sloping loss to low-mid frequency and at con gurations (Fig. 3). For example, the proband (PID IV-7) has a sloping high frequency hearing loss whereas his son (PID V-6) has a relatively at con guration. Longitudinal audiograms for a cousin (PID V-4) reveal a low-mid frequency rising con guration. The proband (PIDs IV-7) and his father (PID III-5) were identi ed with hearing loss in their rst decade. In contrast, two members (PIDs V-6, VI-1) failed newborn screening. Other members reported hearing loss in the 2nd or 3rd decade (age 28 for PID V-5, Fig. 3d). The course of hearing deterioration is also variable. Hearing sensitivity declined into the severe to profound range by the 2nd decade for some members (PIDs V-4, V-6) and in the 3rd -6th decade for others (PIDs IV-12, IV-7, IV-9, IV-14, IV-6).
Consecutive audiograms for PID IV-2 show all frequency loss and progression over six years (Fig. 3e). A single audiogram for PID IV-3 at age 40 shows moderate low and mid frequency loss, and a peak at 2kHz with severe high frequency loss (Fig. 3e). Hearing loss was present in the proband (PID IV-1) by the early 2nd decade; onset is reported during the rst decade for a sibling (PID IV-2) and two other members (PID IV-3 and IV-12).
Linkage Analysis Links Hearing loss to DFNA33 (13q34) The proband (PID IV-7) screened negative for all genetic variants we previously identi ed in this genetically isolated population (Table S1). Furthermore, although audiometric data analysis yielded positive gene matches, bidirectional Sanger sequencing of targeted genes based on audiopro les yielded wild-type sequence (data not shown). Genome-wide SNP genotyping and two-point linkage analysis yielded statistically signi cant linkage (LOD = 4.77; Table 1) to a 3.6 Mb region on chromosome 13q34 overlapping DFNA33 (13q34-13qter) (Bonsch et al. 2009).  Fig. 4b) were both absent in Newfoundland population controls. In silico tools predict that COL4A1 c.3326-7dupT does not disrupt splicing, which we con rmed by RNA analysis (Fig. 4c). Furthermore, during the transcriptional analysis, we recruited two unaffected family members (PID III-4 and V-2) that harboured key cross over events and reduced the disease interval to 769 Kb, excluding the COL4A1 c.3326-7dupT (Fig. 1). Subsequently, 202 Newfoundland probands with hearing loss were screened wild-type for ATP11A g.190616G > A. 8,768 bp). In silico algorithms predict this novel ATP11A variant has the potential to disrupt a canonical donor splice site (Medium Impact; Table S4). According to NNSPLICE analysis, the ATP11A variant is predicted to activate a cryptic donor splice site 153 bp downstream of the canonical donor splice site. RT-PCR analysis of the 3' region of ATP11A ( anking the g.190616G > A variant) in unaffecteds and ATP11A g.190616G > A carriers (PID III-1, III-5 and III-7) revealed multiple products (Fig. 4d). PCR ampli cation and size fractionation of cloned ATP11A RT-PCR products revealed three products in unaffecteds and these three products plus three extra higher molecular weight products in ATP11A g.190616G > A carriers (Fig. 4e).

RNA Validation of Aberrant Splicing in Patient-Derived Tissues
NNSPLICE analysis predicted that the ATP11A g.190616G > A variant may activate a cryptic splice site 153 bp downstream of the canonical donor splice site. Sequencing of cloned RT-PCR products in ATP11A g.190616G > A carriers show they retain 153 bp of intronic sequence at the 3'UTR of ATP11A and is consistent with the three aberrant higher-molecular weight products (Figs. 4e and 5). This intronic sequence was not observed in unaffected family members tested (Figs. 4e and 5). Whether or not the 153 bp extends exon 29 in the putative ATP11A-201 isoform (Fig. 5a) and exon 30 in the putative ATP11A-202/212 isoforms (Fig. 5b), or represents a novel transcript is unknown (Fig. 5). As the RT-PCR analysis and sequencing was restricted to the 3' region of ATP11A ( anking the g.190616G > A variant), full-length sequencing is required to con rm the variant effect on ATP11A isoforms.

Unrelated Family from Israel with Novel Splicing Variant in exon 28 of ATP11A
Genomic DNA from the proband (PID IV-1) and two cousins (PIDs IV-3, IV-12; Fig. 2) underwent clinical exomes (CEGAT Laboratory) and screened negative for DFNA genes but were heterozygous for a novel splice site duplication in exon 28 (ATP11A: NM_032189.3:c.3322_3327 + 2dupGTCCAGGT) in all three samples (data not shown). This variant of uncertain clinical signi cance has not been seen in 5,000 previous exome analyses at the CEGAT Laboratory or previously described in the gnomAD database.

Discussion
We report the rst family with autosomal dominant SNHL to be linked to DFNA33 since the locus was mapped in 2009 (Bonsch et al. 2009) and identify two novel splice site variants in the 3' region of ATP11A. Splice mutations in the 3' region of ATP11A cause a bilateral, progressive SNHL with variable onset and con guration. The hearing loss in the German DFNA33 family is quite similar to ATP11A mutation carriers in both the NL and the Israel families. Although we cannot be certain that the ATP11A gene is DFNA33, it was noted to be a functional candidate based on mouse studies (Bonsch et al. 2009). Although a decade has passed since DFNA33 was mapped to chromosome 13q34-qter, no other families have been reportedly mapped to this locus. Although other phospholipid ippases (P4-ATPases) are associated with syndromic forms of hearing loss, this study documents the rst association of ATP11A with a highly penetrant Mendelian phenotype. Of the 17 annotated ATP11A transcripts, the putative ATP11A g.190616G > A variant mapped to a short Ensembl isoform containing only three exons. RT-PCR and cloning analysis of the 3' region of ATP11A ( anking the g.190616G > A variant) in unaffecteds and ATP11A g.190616G > A carriers revealed multiple products, three products in unaffecteds and these three products plus three extra higher molecular weight products in ATP11A g.190616G > A carriers. The predicted variant effect of the insertion of a 153 bp intronic sequence, which activates a cryptic donor splice site 153 bp downstream, extending the 3'UTR was con rmed with RNA analysis.
However, the exact disease mechanism is unclear.
It is known that the 3' end of the mammalian P4-ATPases are important for protein folding and regulation of its activity (Chalat et al. 2017). Given that the splicing variant is located in the 3'UTR, it is unlikely that it affects protein structure; however, it might affect protein function through modulating ATP11A gene expression. Furthermore, it is unknown if the likely pathogenic splicing variant acts in a dominant-negative effect or due to haploinsu ciency.
Evidence suggests that the genomic complexity of ATP11A expression is similar in mouse and humans. Several principal ATP11A transcripts are expressed in mouse auditory system, including sensory hair cells of the inner ear, otic progenitor cells and auditory and vestibular ganglion neurons (Shen et al. 2015). By combining single-cell and long-read RNA-seq technologies, Ranum et al (2019) has recently revealed clusters of genes that de ne inner hair cells, outer hair cells and deiter cells, and identi ed many heretofore unrecognized exons, alternative splicing diversity and isoform abundance in hearing loss genes (Ranum et al. 2019). According to the Molecular Otolaryngology and Renal Research Laboratories database, ATP11A-203 (exon 2) murine homolog, Atp11a (exon 29) was expressed in inner hair cells, outer hair cells and deiter cells. Most importantly, Atp11a was identi ed as a deiter cell de ning gene. ATP11a exhibits moderate expression during mouse embryonic development, followed by a marked increase in expression after birth (Shen et al. 2015). That the mouse ATP11A-201 transcript contains an exon 29 that is similar to the new 104 bp exon that we identi ed between exons 28 and 29 in one RT-PCR product and that Atp11a (exon 29) was expressed in inner hair cells, outer hair cells and deiter cells, suggests that of 3' end of the ATP11A product is likely ATP11A-201 in humans. A mouse model for ATP11A-induced hearing loss could help elucidate the molecular mechanisms.
Over the ten-years of study on the NL family, we have bene ted from ongoing recruitment and comprehensive phenotyping which eventually identi ed key recombinations to the disease-associated haplotype. The use of a mixed approach including linkage and haplotype analysis reduced the number of candidate variants, which made the problem tractable. Given the complexity in genomes, a comprehensive bioinformatics pipeline targeting all known transcripts is essential as is the need to experimentally validate in silico predictions in patient-derived tissues.
Limitations of this study include a lack of insight as to protein structure and/or function, speci cally with respect to hearing loss. In addition, given that the complete ATP11A 201/202/212 transcripts were not sequenced, it is uncertain if the penultimate exon is present in these ATP11A transcripts or represents the 3' end of a novel transcript.

Con icts of Interests
The authors declare they have no con ict of interest.

Ethics Approval
This study was approved by the institutional review board at Memorial University (#1.186).

Consent to participate
Informed consent was obtained from all individual participants included in this study.

Consent for Publication
Participants signed informed consent regarding publishing their data.

Figure 1
Six generation Newfoundland pedigree with autosomal dominant bilateral SNHL segregating a DFNA33linked disease haplotype on 13q34. Haplotypes (coloured vertical bars) represent a 3.6Mb region on chromosome 13q34 with polymorphic markers and SNPs and their relative genomic positions (left). The disease haplotype (yellow) is diminished by several key crossover events (red x). Black squares represent males and circles, females. Dashes above symbols indicate pure tone audiometry hearing testing was performed. Asterisks are the members that were selected for whole-genome sequencing.