Following the characterization of the converted allele in our proband, we identified one ClinVar submission (accession SCV000966854.1) reporting the stop codon mutation contributed by OTOAP1, p.Glu787*, in trans with a deletion, identified in a patient with hearing-loss (case 2). After obtaining DNA from this patient, we confirmed the deletion of OTOA on one chromosome using MLPA and performed the long range amplification and nanopore sequencing as described above. In contrast with the first proband, only 11 positions around exon 22 showed nucleotides specific to OTOAP1, including the three in the coding sequence (p.Thr785Pro, p.Glu787* and p.Tyr806Ser). This allowed us to estimate that the size of the conversion on this second allele was less than 900bp and to code the conversion with the following HGVS nomenclature: NC_000016.10:g.(21735999_21736311)_(21736714_21736886)con(22552122_225552435)_(22552837_22553008).
Discussion
Gene conversion leading to pathogenic variants has already been described in a variety of human Mendelian diseases (Chen et al., 2007, 2010). They have been hypothesized to arise after a double-strand break where the damaged chromosome is repaired, after resection, using a non-allelic homologous region of either the sister chromatid or the homologous chromosome. However, it has been assumed that, in humans, conversion events longer than 3kbp are more likely occurring via a non-allelic homologous double recombination mechanism, because most well characterized gene conversion events typically replace regions of a few hundreds of nucleotides only (Chen et al., 2007). For instance, this mechanism has been proposed to be at the origin of the 12f2 deleted lineages of the human Y chromosome (Blanco et al., 2000). A formal distinction between the two mechanisms would require characterization of the by-product chromosome, which is not possible when investigating human alleles.
Given the lengths of the two conversions we described (more than 9kbp and less than 900 bp), we assume that the mechanisms at the origin of these two alleles are respectively non-allelic homologous double recombination and double strand break repair-mediated gene conversion events between OTOA and OTOAP1. Comparison between OTOA and OTOAP1 reference sequences from GRCh38 shows identity higher than 99% over more than 42 kbp of aligned nucleotides, which is sufficient to promote non-allelic homologous recombination between the two loci (Gu et al., 2008).
Heterozygous conversion of one to two exons of OTOA was reported in 3 patients among a cohort of 686 patients with hearing loss (Shearer et al., 2014), without direct molecular confirmation. Estimating the allelic frequencies of the conversion in the population from exome sequencing or short-reads whole genome sequencing is non-trivial. It requires coverage depth comparison between regions of OTOA and OTOAP1. Using exome sequencing, only the exons showing sufficient divergence, allowing for unique mapping to either loci, can reliably be called as converted or not. The alignments of OTOA and OTOAP1 exonic sequences from GRCh38 show differences only for exons 21, 22 and 29 and their 50 flanking nucleotides, with respectively 1, 7 and 1 changes. For the evaluation of pathogenicity, only the conversion of exon 22 appears relevant, as the change in exon 21 is synonymous and the one in exon 29 lies in the 3’UTR.
It is likely that polymorphisms in regions of segmental duplication in OTOA and OTOAP1 are under evaluated due to the difficulty of genotyping non-unique sequences. Interestingly, the p.(Thr785Pro), p.(Glu787*) and p.(Tyr806Ser) mutations of OTOA exon 22 are present in gnomAD v2.1.1 (Karczewski et al., 2020) as low frequency alleles, as well as low quality alleles for p.(Thr785Pro) and p.(Glu787*). The corresponding opposite mutations on OTOAP1 exon 2 are also present in this database. These observed substitutions could result from errors in the mapping algorithm, leading to sequencing reads being incorrectly assigned to their paralogous regions in the genome. The alternative hypothesis would be that recurrent conversions between the two loci create alleles that appear as a mixture of OTOA and OTOAP1 wild-type specific changes. This hypothesis is consistent with the observation that gene conversions contribute to a significant part of the shared polymorphism between paralogous regions in the human genome (Dumont and Eichler, 2013; Dumont, 2015).
Data Availability Statement
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
Acknowledgments
We thank Alexandre Reymond for critical reading of the manuscript. This work is supported by the Fondation Privée des HUG. None of the authors has a conflict of interest to declare.