Missense variant in LOXHD1 is associated with canine nonsyndromic hearing loss

Hearing loss is a common sensory deficit in both humans and dogs. In canines, the genetic basis is largely unknown, as genetic variants have only been identified for a syndromic form of hearing impairment. We observed a congenital or early-onset sensorineural hearing loss in a Rottweiler litter. Assuming an autosomal recessive inheritance, we used a combined approach of homozygosity mapping and genome sequencing to dissect the genetic background of the disorder. We identified a fully segregating missense variant in LOXHD1, a gene that is known to be essential for cochlear hair cell function and associated with nonsyndromic hearing loss in humans and mice. The canine LOXHD1 variant was specific to the Rottweiler breed in our study cohorts of pure-bred dogs. However, it also was present in some mixed-breed dogs, of which the majority showed Rottweiler ancestry. Low allele frequencies in these populations, 2.6% and 0.04%, indicate a rare variant. To summarize, our study describes the first genetic variant for canine nonsyndromic hearing loss, which is clinically and genetically similar to human LOXHD1-related hearing disorder, and therefore, provides a new large animal model for hearing loss. Equally important, the affected breed will benefit from a genetic test to eradicate this LOXHD1-related hearing disorder from the population. Supplementary Information The online version contains supplementary material available at 10.1007/s00439-021-02286-z.


Introduction
Hearing loss (HL) is the most common sensory impairment and consequently constitutes a major medical issue, affecting 1-3/1000 newborns and increasing in prevalence by observation age (Morton and Nance. 2006). It is a both clinically and genetically heterogeneous disorder: the severity of the hearing impairment varies and the age of onset ranges from congenital and early-onset to late-onset.
Other symptoms may accompany hearing impairment, but nonsyndromic hearing loss (NSHL) is the predominant type. Genetic factors contribute to more than half of the congenital and early childhood nonsyndromic forms, and variants have been reported in over 100 genes to date (https://hereditaryhearingloss.org/). Despite the accumulating knowledge, a large number of patients remain without a molecular diagnosis. Thus, there is a need to understand better the molecular underpinnings of hearing loss and the associated genes and variants in cochlear biology and disease pathophysiology.
The molecular genetics of hearing loss can be addressed by studying spontaneous hearing defects in purebred dogs. Over the past century, selective breeding of dogs has led to hundreds of breeds with speci c genomic architectures and the formation of genetic isolates, which can signi cantly facilitate gene discovery in small study cohorts (Lindblad-Toh et al. 2005). Both syndromic and nonsyndromic congenital as well as adult-onset hearing loss have been reported in dogs across breeds (Strain. 2015), yet their genetic background remains mostly uncharacterized. The most common form of hearing loss in dogs is pigment-associated congenital sensorineural deafness which occurs in breeds with a lack of pigmentation or piebaldism (De Risio et al. 2016; Comito et al. 2012; Platt et al. 2006;Strain. 2004). The trait is associated with regulatory variants in melanocyte inducing transcription factor (MITF) (Karlsson et al. 2007). Another syndromic type of HL is a recessive congenital deafness and vestibular syndrome observed in Doberman Pinschers, which is linked to variants in myosin VIIA (MYOA7) and tyrosine phosphatase, receptor type Q (PTPRQ) (Webb et al. 2019;Guevar et al. 2018). In addition to the congenital HL forms, an adult-onset hearing defect has been reported in Border Collies and mapped to canine chromosome 6 (Yokoyama et al. 2012). In our study, we observed sensorineural hearing loss in Rottweilers and utilized genome-wide array genotyping and genome sequencing to identify a fully segregating novel missense variant in LOXHD1, a known hearing loss gene in human. Detection of runs of homozygosity (ROH) was also performed with PLINK 1.9 (Chang et al. 2015). Two rounds of analyses were conducted: rst, the optimization of population-dependent parameters using simulated data; and second, detection of ROH shared by the affected dogs. In both analyses, the minimum marker size for ROH (--homozyg-snp) was set to 70 based on the formula described by Pur eld et al. (2012) with α = 0.05, N s = 149084, N i = 6 and mean SNP heterozygosity = 0.21 as evaluated using an in-house Python script. Furthermore, the parameters --homozyg-window-snp, --homozyg-windowmissing, --homozyg-window-het, --homozyg-window-threshold and --homozyg-kb were set to a xed value depending on the analysis (Table 1).

Materials And Methods
Finally, to establish suitable values for ROH density and maximum gap, genotypes for a fully homozygous individual were simulated based on the population map le and analyzed for maximal genome coverage as described by Meyersman et al. (2020). To calculate maximum genome coverage for the simulated genome, --homozyg-gap was set to 2000 kb and --homosyg-density to 200 kb/snp; genome coverage was determined as the total length of the resulting ROH. The simulated genome was then analyzed by varying --homozyg-density from 10 to 125 kb/snp in increments of 5 kb and --homozyg-gap from 20 to 1000 kb in increments of 20 kb (Table 1).

Results
Sensorineural bilateral deafness was diagnosed in four Rottweiler siblings (one female and three males) in a litter of ten puppies using brainstem auditory evoked response (BAER) testing. BAER testing was performed either at 4 (n=2), 5, or 19 months of age, and no auditory response was detected in any of them. However, owners' observations suggested that the puppies had already been affected by hearing impairment at a few weeks of age. No other clinical signs were observed.
We carried out a genome-wide analysis to identify candidate loci using three affected and three unaffected dogs. Homozygosity mapping resulted in 22 regions of case-speci c, allelically matching runs of homozygosity ( Fig. 1  Subsequently, we performed whole exome sequencing on one affected dog and, later, whole genome sequencing on another case. As a result of variant ltering, 32 homozygous SNVs and indels shared by the two affected dogs were discovered; of these, six were located in case-speci c ROH (   Of the seven case-speci c variants that resided in ROH, two exonic variants were considered for further analyses. First, a G>C missense variant at chr7:44,806,821 in lipoxygenase homology domains 1 (LOXHD1), a gene that is known to cause hearing loss in humans and mice (Grillet et al. 2009), was predicted to result in a glycine-to-alanine substitution (Fig. 2). Similarly, a C>T missense variant at chr24:25,785,932 in maestro heat like repeat family member 8 (MROH8) was predicted to cause a glycine-to-serine substitution. We assessed the pathogenicity of these variants in silico using two protein prediction tools, PROVEAN and PolyPhen-2. First, PROVEAN predicted both the LOXHD1 and MROH8 substitutions as "deleterious" with a score of -4.517 and -3.336, respectively. Similarly, PolyPhen-2 predicted the LOXHD1 variant as "probably damaging" with a HumVar score of 0.992 and the MROH8 variant as "possibly damaging" with a score of 0.550. As MROH8 has been associated with red blood cell volume, body mass index, telomere length and hippocampal atrophy in humans (Buniello et al. 2019) and its predicted impact was less pathogenic, making it an unlikely candidate variant, we focused on LOXHD1. The chr7:44,806,821G>C variant is predicted to cause a glycine-to-alanine substitution p. (G1914A) in the fourteenth PLAT domain of the canine LOXHD1 protein (Fig. 2c).
Additionally, we assessed the conservation of the G1914 residue with multiple alignment of 99 Eutherian LOXHD1 protein orthologs, including the dog (Online Resource 6). In these species, the glycine residue and several anking amino acids are fully conserved.
To validate the LOXHD1 variant, we genotyped it in a cohort of 585 Rottweilers, including the four affected siblings and 581 unaffected dogs. We observed complete segregation of the variant with hearing loss, as all four affected dogs were homozygous for the variant. The unaffected dogs were either heterozygous (n=33) or wild-type (n=548). The six unaffected littermates of the probands were either wild-type or heterozygous for the variant. The allele frequency in the population, excluding the affected family, was 2.6 % and carrier frequency 5.3 %.
An additional sample of dogs submitted for commercial genetic testing was screened for the LOXHD1 variant to explore its distribution across breeds. All 28,116 tested dogs representing 374 breeds, breed varieties or designer dog mixes were found as homozygous for the wild-type allele (Online Resource 7). Finally, the variant was also screened in a larger study sample of 771,864 dogs submitted to genetic testing, including breed detection assessment. A variant carrier frequency of 0.08 % and allele frequency of 0.04 % were observed in this dataset. Interestingly, six dogs were found homozygous for the LOXHD1 variant. We were able to contact the owners of 4/6 of the homozygous dogs and the owners reported profound hearing loss or deafness in all of them. One of the deaf dogs did not show any immediate Rottweiler ancestry, while one was a purebred Rottweiler and two were mixed-breed with Rottweiler ancestry. Altogether, of the dogs carrying at least one copy of the deafness candidate variant, 63.4% showed evidence of Rottweiler ancestry in their immediate three-generation pedigree going back to greatgrandparents, providing further support for a link between this speci c breed background and the presence of the variant.
According to a study by Riazuddin et al. (2012), heterozygous LOXHD1 variants have been suggested to contribute to autosomal dominant late-onset Fuchs endothelial corneal dystrophy (FECD). For this reason, we assessed the eye examinations performed as a part of the breeding programs and regular health check-ups of the Rottweilers genotyped either homozygous or heterozygous for the LOXHD1 variant. Two of the homozygous deaf dogs had been eye examined healthy, one at 2 years and the other at 8 years old. In addition, eye examination reports were available for 22/33 heterozygous dogs, examined between 1 year to 7 years and 4 months old. Three heterozygous dogs were diagnosed with different forms of cataract, which is a relatively common eye disease in Rottweilers. No signs of corneal dystrophy were reported in any of the dogs.

Discussion
We describe here a missense variant in LOXHD1 associated with an autosomal recessive congenital nonsyndromic hearing loss in Rottweilers. The variant is rare, yet we con rmed it to be fully segregating with the disease in the breed. This is the rst genetic defect identi ed for NSHL in dogs.
The dogs with identi ed LOXHD1 variant had either congenital or early-onset hearing loss. The owners' reports suggest that the puppies had at least some hearing impairment already at a few weeks of age. However, the con rmed diagnosis by BAER was acquired later and by that time, the hearing loss was total. Therefore, it is likely that the hearing impairment was congenital and progressed to deafness in a few months, although this remains uncon rmed. The hearing loss in Rottweilers can be de ned as nonsyndromic as the affected dogs showed no other consistent clinical features.
Congenital deafness has been reported in Rottweilers previously as sporadic cases (Coppens et al. 2001;Strain. 1996). Histopathological examination of one 4.5-month-old bilaterally deaf Rottweiler puppy demonstrated severe degeneration of hair cells and spiral ganglion while the vestibular organ was unaffected (Coppens et al. 2001). Strikingly similar changes are seen in the samba mouse line generated in an ethylnitrosourea (ENU) mutagenesis screen, where a Loxhd1 missense variant leads to hearing loss (Grillet et al. 2009). Stereociliary development in samba mice is normal, but hair cell function is altered by postnatal day 21 and hair cells eventually undergo degeneration followed by possibly secondary loss of spiral ganglion neurons. Based on the similarities of the canine and murine models, it is probable that the histopathological changes described in the previous Rottweiler cases represent the canine LOXHD1 c.1914G>A variant identi ed in this study. It can be concluded that LOXHD1 has an essential role in maintaining normal cochlear hair cell function. Interestingly, murine Loxhd1 is mainly expressed in the membrane of mature mechanosensory hair cells in the cochlea and high levels in testis, an organ also enriched with stereocilia. Its numerous PLAT (polycystin/lipoxygenase/alpha-toxin) domains are likely involved in targeting the protein to the plasma membrane (Grillet et al. 2009;Bateman and Sandford. 1999), but the speci c function of LOXHD1 in ciliary structures remains unclear.