Through extensive re-recruitment efforts, an affected individual from the originally published DFNA33 family (Bönsch et al., 2009) was re-ascertained as part of a large diverse-population rare disease study (ethics commission approval number 197/2019BO1). The pedigree was re-evaluated and re-drawn.

The affected individual underwent a complete ear, nose, and throat examination that included binocular ear microscopy and external ear inspection. Pure-tone audiometry was performed according to current standards to determine hearing thresholds at routinely measured frequencies (0.25, 0.5, 1, 2, 4, 6, and 8 kHz). Both air and bone conduction thresholds were measured. Tympanometry was performed to measure tympanic membrane compliance and middle ear pressure.

Whole genome sequencing was performed on genomic DNA extracted from peripheral blood using standard protocols, as previously described (Weisschuh et al., 2020). The sequencing library was prepared with the TruSeq DNA PCR-free protocol (Illumina, San Diego, USA) for subsequent sequencing as 2 × 150 bp paired-end reads on a NovaSeq 6000 platform (Illumina, San Diego, USA). Bioinformatic processing of raw read data, annotation, and variant calling was performed using the megSAP pipeline (https://github.com/imgag/megSAP) developed at the Institute of Medical Genetics and Applied Genomics, University Hospital Tübingen, Germany, and includes an in-house genome database called the NGS Database (NGSD), which consists data on more than 3,000 individuals to ascertain sequencing artifacts and variant frequencies. Visualization was performed using the Integrative Genomics Viewer.

Bioinformatics filters were used to support the analysis of variants, following autosomal dominant and recessive modes of inheritance, as described previously (Falb et al., 2023). This included analysis of sequence variants and copy number variations in 298 known hearing loss-associated genes (Supplementary Table S1). Allele-based filters selecting variants with an allele frequency (including sub-populations) ≤ 1%, variants present in ≤20 other individuals in NGSD, and those previously annotated as pathogenic/likely pathogenic in HGMD and ClinVar yielded a short list of heterozygous variants in these hearing loss-associated genes. Minor allele frequencies were derived from gnomAD, 1000 Genomes Project, ExAC, dbSNP, and NGSD. The Alamut Visual (Interactive Biosoftware, SOPHiA GENETICS, Rouen, France) splice window that includes splice predictions from SpliceSiteFinder-like, MaxEntScan, NNSPLICE, GeneSplicer, and ESEfinder, as well as RESCUE-ESE predictors, was used to assess the effect of variants on splicing.

Variant classification applied the American College of Medical Genetics and Genomics (ACMG) guidelines that are adapted for hereditary hearing loss (Oza et al., 2018) and referenced the Deafness Variation Database (Azaiez et al., 2018). The classification was assisted through the use of the public version of Varsome (Kopanos et al., 2019).

A region spanning ATP11A exons 8 to 10 was PCR-amplified from the genomic DNA of the patient and a healthy control using gene-specific primers with a XhoI restriction site (ATP11A Ex8-10 forward: 5′-aat​tct​cga​gAA​AAT​CAC​CGA​AGC​CAT​GAG-3′) and a BamHI restriction site (ATP11A Ex8-10 reverse: 5′-att​gga​tcc​ATG​GTG​AGA​GGA​GCT​GTT​GG-3′). The 5,014-bp amplicon was ligated into a multiple cloning site between native exons A and B in a linearized pSPL3 exon-trapping vector. The vector was transformed into DH5α competent cells (NEB 5-alpha, New England Biolabs) and plated overnight. The wild-type and mutant-containing vector sequences were confirmed by Sanger sequencing and transfected into HEK 293T cells using the FuGENE Transfection Reagent (Promega). Total RNA was prepared from 24-h post-transfected cells using an miRNeasy Mini Kit (QIAGEN), and reverse transcription was carried out using a High-Capacity RNA-to-cDNA Kit (Applied Biosystems). Amplified fragments were visualized on 1.5% agarose gel and, subsequently, Sanger sequenced.

The affected individual, who was not in the original DFNA33 report, is a direct descendent of a hearing-impaired father who had four affected siblings, three of whom were previously described (Bönsch et al., 2009) (Figure 1A). The individual reported progressive, high-frequency sensorineural hearing impairment without subjective vertigo and tinnitus. The individual's thresholds mirror the average thresholds of individuals previously described in their family (Bönsch et al., 2009) up to 2 kHz, where the individual shows steeply increasing thresholds to severe-to-profound hearing loss at 8 kHz in the left ear and no thresholds at 6 and 8 kHz in the right ear. An audiogram at the age of 66.9 years is shown in Figure 1B. A tympanogram showed normal bilateral results with regular tympanic membrane compliance. Upon comparison of the audiogram with the hearing of men in the same age group (ISO, 2017), the hearing thresholds at all measured frequencies were found to be significantly lower than the expected hearing threshold for this age group.

Whole-genome sequencing resulted in an average 51.7 × coverage of the genome. An analysis of sequence variants and copy number variations in 289 known hearing loss-associated genes (Supplementary Table S1) excluded a putative pathogenic genetic aberration. These variants, shown in Supplementary Table S2, were excluded based on criteria such as their presence in a homozygous state in other presumably normal hearing individuals in NGSD, low in silico pathogenicity prediction scores, no predicted effect on splicing, or an obvious clinical mismatch.

Following the initial hypothesis that ATP11A may be the gene for the DFNA33 locus and the growing clinical and functional evidence asserting ATP11A is an essential protein for a normal functioning auditory system, our analysis then focused on the previously mapped DFNA33 locus coordinates (GRCh37:110,300,001-115,169,878), as well as ATP11A (Supplementary Table S3), and uncovered variants that were either identified in other in-house patients with presumed normal hearing or were likely benign based on bioinformatics assessment. Only one deep intronic variant in ATP11A intron 8 (chr13:113421269C>G (GRCh37), ENST00000375630:c.725 + 737C>T) was predicted via in silico tools to activate exonic splice enhancers. This variant was tested with an in vitro splicing assay using established protocols (Rad et al., 2021; Vona et al., 2021) and yielded a functionally normal result (Figure 2). Taken together, the variant is classified as likely benign (PM2_Supporting, BP4_Supporting, and BS3_Strong; −4 points (likely benign point range: −6 to −1)). On this basis, we conclude that following short-read genome sequencing, a method employed for the identification of the first described ATP11A variant (Pater et al., 2022), we present evidence that does not support ATP11A as being the gene for the DFNA33 locus and acknowledge that we have been extremely fortunate to re-recruit an individual from the DFNA33 family.

Upon close inspection, we noted that the haplotypes that were used to define the DFNA33 boundaries of the disease interval in the original Bönsch et al. paper show several unlikely events (Bönsch et al., 2009, shown in Figure 1A for clarity). These include three recombination events in a small chromosomal interval in individual 503 and two recombination events in a relatively short interval in individuals 412 and 419. The fine mapping and recapitulation of haplotypes seem improbable and may be due to genotyping or other errors. Due to this observation, the DFNA33 disease interval may be inaccurate. Therefore, we performed genome-wide analysis of variants that did not yield an obvious candidate and informative result for follow-up and thus required additional family members in order to significantly narrow variants.