Introduction
Mendelian genetic defects constitute a leading cause of non-syndromic hearing loss, one of the most common sensorineural defects in humans (Nance, 2003). Although dominant and recessive phenotypes linked to GJB2/GJB6 are a frequent etiology (Yokota et al., 2019), more than 150 genes have been linked with syndromic or non-syndromic deafness (Sloan-Heggen et al., 2016; Azaiez et al., 2018). Consequently, genetic diagnosis has moved from screening of the connexin locus to interrogating simultaneously thousands of exons of the human genome, to reach an overall diagnostic rate of around 40% (Shearer and Smith, 2015; Azaiez et al., 2018).
STRC and OTOA bi-allelic loss-of-function are well-known contributors of mild-to-moderate autosomal recessive hearing loss in humans (Vona et al., 2015; Mehta et al., 2016; Zazo Seco et al., 2017). Both STRC and OTOA reside in repetitive regions of the genome, leading to recurrent deletions or duplications (copy number variants, CNVs) in the general population, through non-allelic homologous recombination (Zhang et al., 2009). These regions also encompass paralogous pseudogenes for both genes: CKMT1A-STRCP1-CATSPER2P1 lie directly downstream of CKTM1B-STRC-CASTPER2 on chromosome 15, whereas OTOAP1 lies approximately 800kbp downstream of OTOA on chromosome 16, with 11 interspersed genes or pseudogenes.
Pathogenic and likely pathogenic frameshift, nonsense, missense and splicing variants have been deposited to ClinVar (Landrum et al., 2016) for OTOA and STRC. Using array Comparative Genomic Hybridization (aCGH), Hoppman et al. (2013) found an allele frequency of 1.09% for the STRC-CATSPER2 deletion, over a cohort of more than 5000 patients. The gnomAD structural variant (SV) dataset v2.1 (Collins et al., 2020), calling CNVs using read depth from more than 10000 whole genome sequencing data, reports an allele frequency of 0.90% for the deletion of STRC specific exons. However, deletions of OTOA are apparently much rarer, with estimates ranging from less than 0.1% in gnomAD SV v2.1 to less than 0.2% in the Database of Genomic Variants (MacDonald et al., 2014).
Gene conversion, defined as the replacement of a locus in the genome by a paralogous sequence, is a mechanism known to generate pathogenic alleles (Chen et al., 2007; Casola et al., 2012). For instance, a pathogenic converted allele has been described for the TMEM231 gene, causing autosomal recessive Joubert and Meckel-Gruber syndromes (Maglic et al., 2016; Watson et al., 2020). Other examples include a benign conversion of 88 to 351 nucleotides of BRCA1 intron 2 (Tessereau et al., 2015) mediated by the same breakpoint sites leading to a 37kbp long deletion of exon 1 and 2 of BRCA1.
In their cohort of 686 patients with hearing-loss, using high throughput sequencing data, Shearer et al. (2014) reported that CNVs in STRC were the most common of all CNVs found in deafness-related genes, followed by CNVs in OTOA. Interestingly, they highlighted the case of a causative homozygous gene conversion of STRC. The authors also mentioned 3 conversions of OTOA, which length was estimated between one to two unspecified exons. No pathogenic CNVs nor single nucleotide variants (SNV) were detected in trans in any of the 3 patients, and the pathogenicity of the conversion itself was not assessed. The conversions were detected by comparative analysis of sequencing depth in the paralogous regions of the genes and pseudogenes, but without confirmation with another molecular approach. Moteki et al. (2016) also reported the case of a homozygous gene conversion of STRC in a cohort of 40 patients, identified by high throughput sequencing and confirmed by aCGH. More recently, Rajagopalan et al. (2020) identified two single exon gene conversions of STRC, initially identified as single exon deletions in exome sequencing CNVs calls, confirmed with long range PCR.
Methods
DNA was extracted from whole blood samples using a Qiagen Symphony instrument (Hilden, Germany). Library preparation and exome capture were performed using the Twist Library Preparation Enzymatic Fragmentation Kit and the Twist Human Core Exome Kit (Twist Bioscience, San Francisco, USA) with the RefSeq spike-in, following the manufacturer protocol. High throughput sequencing was performed on an Illumina (San Diego, USA) NextSeq 500 sequencer.
SNVs were called using an in-house pipeline based on bwa-mem (Li and Durbin, 2009) for alignment against GRCh37 and GATK HaplotypeCaller (Poplin et al., 2018) for genotyping. CNVs were identified with an algorithm based on Conifer (Krumm et al., 2012) and XHMM (Fromer et al., 2012). Briefly, it detects variation of the z-scores of mean normalized sequencing depth values, after filtering low quality and indeterminate mapping, of each exon that are compatible with duplications or deletions. All samples previously sequenced at our facility after exome capture using the same kit are used as reference. Visual inspection of sequencing alignments and z-scores was performed with a local instance of JBrowse (Buels et al., 2016). A panel of 187 genes involved in hearing loss and ear malformations was used to filter SNV and CNV, as well as automatic frequency and functional filters, before manual curation by three different biologists.
MLPA was performed using the Salsa Probemix for Deafness Infertility Syndrome (DIS P461, MRC Holland, Amsterdam, The Netherlands) which includes probes for the OTOA and STRC/CATSPER2 regions. PCR amplification was carried out with AccuPrime Taq DNA Polymerase (ThermoFisher, Waltham, USA) and LA Taq DNA polymerase (Takara, Kusatsu, Japan) for amplicons longer than 10kbp. Sanger sequencing of purified PCR products was performed on an ABI 3500 instrument (ThermoFisher).
Nanopore sequencing of long amplicons was performed using a MinION Mk1B instrument (Oxford Nanopore Technologies, ONT, Oxford, United-Kingdom) and R.9.4.1 flowcells with MinKNOW high accuracy base calling. Libraries were prepared according to ONT’s instructions using NEBNext FFPE repair kit, NEBNext Ultra II end repair/dA-tailing module (New England Biolabs, NEB, Ipswich, USA) and SQK-LSK109 ligation sequencing kit (ONT). Reads were mapped using minimap2 against chromosome 16 of the GRCh38 reference genome (NC_000016.10) and visualized using the Integrative Genomics Viewer (Robinson et al., 2017).
Results
Blood samples of a 2 years old female child (case 1) with non-syndromic hearing loss and her parents were referred to our facility for molecular investigations, after genetic counseling and informed consent was obtained. Exome sequencing did not reveal any pathogenic or likely pathogenic SNV in our panel of 187 genes. However, the algorithm for CNV detection showed z-scores compatible with bi-allelic OTOA variants (Fig. 1A and 1B), consisting of a heterozygous deletion spanning at least 110kb from exon 2 of METTL9 (NM_016025.5) to exon 22 of OTOA (NM_144672.4) and a smaller variant centered around exon 22 (z-score compatible with a homozygous deletion of exon 22). The MLPA analysis confirmed the large 110kb deletion and showed it was inherited from the mother.