Gene Panel Screening Across Canine Eye Disorders Highlights Genetic Heterogeneity and the Need for Molecular Discoveries


 Blinding inherited eye diseases affect millions of people worldwide. Despite a large number of gene discoveries, many patients lack the molecular description of their condition. The domestic dog has become a widely used model to study inherited eye disease genetics and therapeutics during the past 15 years, and nearly 50 genes have been implicated across breeds. Despite a continuous discovery of new causative variants across canine eye disease groups, the genetic cause in most cases is expected to remain unknown. We tested this hypothesis by screening 49 known variants in 194 dogs from 71 breeds affected with progressive retinal atrophy (PRA), glaucoma, or lens luxation and validated the results in additional 1180 dogs. We found that eleven variants in ten genes explained 28% of the studied cases. We also observed new PRA-affected breeds for the RPGRIP1 c.340_341insA29GGAAGCAACAGGATG variant, and clinical support for the pathogenicity of the PDE6A c.1940delA and ADAMTS10 c.2231G>A variants for PRA and glaucoma in two new breeds, respectively. Our findings indicate extensive genetic heterogeneity and the lack of molecular descriptions in more than two-thirds of the LL, glaucoma, and PRA cases by the current gene tests. There is an urgent need for discoveries that would be critical not only for veterinary molecular diagnostics and breeding programs but also for establishing models to characterize pathophysiology and new therapeutic options for the corresponding human eye disorders.

During the past 15 years, due to their similar eye physiologies and retinal metabolism, the dog has become a favorite spontaneous large animal model to study the genetics of inherited eye diseases. This has been facilitated by their distinctive population structure, which results from multiple evolutionary bottlenecks (Lindblad-Toh et al., 2005). Inbreeding and popular sires have diminished genetic variation within breeds and have contributed to the number of eye diseases observable today by enriching the disease-causing variants (Aguirre-Hernández & Sargan, 2005). Interestingly, due to their relatively short breedspeci c histories, large-scale studies have revealed a widespread presence of recessive pathogenic variants across the general dog population, but notably higher disease case prevalence in pure breeds (Donner et al., 2016(Donner et al., , 2018. For example, over 100 breeds to date are known to be affected by progressive retinal atrophy (PRA), the canine equivalent of human RP, which affects millions of individuals worldwide (Haim, 2002;Petersen-Jones & Komáromy, 2014).
The bene ts of using the dog as a spontaneous model to study many human eye diseases have been of growing emphasis from the 1980s onwards when genes of the phosphodiesterase-family were established as harboring causative variants to both the canine PRA and human RP40 (Aguirre et al., 2007;Bunel et al., 2019;Dekomien & Epplen, 2003;Dvir et al., 2010;Miyadera et al., 2012;Suber et al., 1993). Since then, at least 15 genes have been observed to be altered in both PRA and RP (Online Mendelian Database, OMIA, https://omia.org/; RetNet, https://sph.uth.edu/RetNet/). A similar pattern can be further recognized for glaucoma and ectopia lentis (EL), a canine equivalent of lens luxation (LL): ADAM metallopeptidase with thrombospondin type 1 motif-proteins -ADAMTS10 and ADAMTS17 -genes have been observed to harbor variants causative for primary glaucoma and LL in dogs and syndromic diseases, with symptoms including glaucoma and EL, in humans (Weill-Marchessani syndrome) (Farias et al., 2010;Forman et al., 2015;Jeanes et al., 2019;Morales et al., 2009). Overall, this pattern of shared genes harboring eye disease-causing variants indicates a need for further research in both species. Furthermore, the comparable phenotypes and genotypes have allowed meticulous studying of the pathophysiology and successful testing of therapies -especially gene therapies -in dogs (Aguirre et al., 2007;Petersen-Jones & Komáromy, 2014).
To date, genetic testing for inherited canine eye diseases, both syndromic and non-syndromic, is available for more than 60 variants in nearly 50 genes (OMIA, https://omia.org/). These variants represent occurring phenotypes detected in over 130 breeds, yet many eye diseases across multiple breeds have unknown genetic risk factors.
Here, we demonstrate that gene panel screening of a cohort of 194 dogs from 71 breeds identi es the causative mutation in only 28% of the LL, glaucoma, and PRA cases. The blinding conditions in most breeds are not explained by any of the tested 49 known eye disease variants. Future research is urgently needed to improve veterinary diagnostics, breeding plans and establish disease models to the corresponding human conditions.

Study cohorts
The gene panel cohort study cohort was established from the dog biobank at the University of Helsinki. The cohort included 194 privately owned purebred pet dogs from 71 breeds with a diagnosis of LL affected (n=22) and LL suspected (n=14), glaucoma affected (n=28) and glaucoma suspected (n=2), PRA affected (n=115) and PRA suspected (n=3), other retinopathy affected (n=6), or other retinopathy suspected (n=2), and dogs affected with overlapping PRA and glaucoma (n=2) ( Table 1). The gene panel cohort included 110 males, 82 females, and two dogs of unknown gender. All dogs had been eye examined by veterinary ophthalmologists board-certi ed by the European College of Veterinary Ophthalmologists (ECVO). For clarity, we refer to cases in each phenotype with either affected or suspected diagnosis.
To validate the gene panel cohort's preliminary ndings, a validation cohort with additional 1180 samples was selected from the dog biobank at the University of Helsinki. The validation cohort included dogs from six breeds (Table 1) not included in the gene panel cohort. These breeds are known to carry both the RPGRIP1 c.340_341insA 29 GGAAGCAACAGGATG insertion and MAP9 deletion variants based on the commercial genetic testing results at Wisdom Health. We also included samples from two other additional breeds, East Siberian Laikas (ESL, n=71) and Labrador Retrievers (LR, n=190), to test the presence of the ADAMTS10 c.2231G>A glaucoma associated variant and the TTC8 c.669delA PRA associated variant, respectively. We had been contacted by an owner of an ESL that had been tested heterozygous for the ADAMTS10 variant in a commercial gene test and screened for the frequency of the variant in our samples. Also, we wanted to know the possible presence of the TTC8 variant in the Finnish LR population, as reported in LR populations outside Finland (Downs et al., 2014).

DNA isolation
Genomic DNA was extracted from the EDTA blood samples using a semi-automated Chemagen extraction robot (PerkinElmer Chemagen Technologie GmbH, Baeswieler, Germany) according to the manufacturer's instructions. DNA concentration and purity were assessed by using the Qubit uorometer (Thermo Fisher Scienti c, Waltham, MA, USA) and Nanodrop ND-1000 UV/Vis Spectrophotometer (Nanodrop technologies, Wilmington, DE, USA) and samples were stored at +4°C (dilutions) and -20°C (stocks).

Gene panel screening
Genotyping of 49 eye disease variants in 36 genes, including 12 variants associated with syndromic diseases with eye phenotypes (Online resource 1), was carried out according to manufacturer-recommended standard protocols on a customdesigned Illumina In nium XT genotyping bead chip (Illumina, San Diego, CA, USA) commercially available as the MyDogDNA™ or Optimal Selection™ test (Wisdom Health, Vancouver, WA, USA). The design, validation, and use of the custom-designed genotyping array to explore canine diseases and traits have been described in detail earlier (Donner et al., 2016(Donner et al., , 2018Dreger et al., 2019).

Validation screening
Genotyping for validation cohorts (  (44 bp), the PCR products were run on 1.5% agarose gel for 1.5h. The MAP9 deletion PCR-products (10k bp) were run on 1% agarose gels for 40 min. The genotypes were analyzed from the gels imaged with the ChemiDoc MP imaging system (Hercules, California, United States).

Optical coherence tomography imaging
To examine the retinal phenotype associated with the RPGRIP1 insertion variant in a higher resolution, spectral-domain optical coherence tomography (SD-OCT) imaging was performed using the Heidelberg Spectralis HRA+OCT instrument (Heidelberg Engineering, Heidelberg, Germany) in the Department of Ophthalmology, Helsinki University Hospital, Helsinki, Finland. We invited ve dogs homozygous for the RPGRIP1 insertion, wild-type for the MAP9 deletion, and three age-and breed-matched controls (RPGRIP1 insertion wild-type, MAP9 deletion wild-type) for SD-OCT. Three of the RPGRIP1 insertion homozygotes were beagles, one was American Staffordshire Terrier, and one was Miniature Shorthaired Dachshund. Their ages ranged between 1.3-7.8 years.

Gene panel screening reveals genetic heterogeneity in eye disorders
To understand the extent of the genetic heterogeneity and how well the current eye disease gene panel detects genetic causes in affected dogs, we selected 194 eye disease-affected dogs from 71 breeds to be screened for 49 known eye disease variants in 36 genes (Online resource 1). We found causative variants only in 53 affected dogs, representing only 28% of the gene panel cohort (Table 2). The identi ed eleven causative variants in ten genes are speci ed in Table 3.  Table 3). All of the homozygous cases were affected. Two heterozygotes were LL affected and six LL suspected. Heterozygosity increases the risk of developing the LL phenotype (Farias et al., 2010;Gould et al., 2011).

ADAMTS10variant likely to cause glaucoma in East Siberian Laikas
The ADAMTS10 c.2231G>A variant was originally described to cause primary open-angle glaucoma in a laboratory colony of beagles (Kuchtey et al., 2011). An ESL owner contacted us, reporting that her dog carries the ADAMTS10 c.2231G>A variant observed in a gene panel screening. To follow up on this nding, we screened the variant in 71 ESLs available in our biobank.
We found 10 heterozygotes and one homozygote. The homozygote ESL was reported to have buphthalmos and was euthanized due to blindness at ve years of age, both typical signs of complicated glaucoma. One of the heterozygous dogs was reported to have a full sibling with complicated glaucoma. These results suggest that the ADAMTS10 variant causes glaucoma also in ESLs.

Molecular descriptions missing for PRA in many breeds
PRA is known to be genetically heterogeneous. Our gene panel cohort included 118 PRA cases from 52 breeds (Table 1). We found that 32 out of the 118 dogs (27%) were homozygous for one of eight retinal phenotype-causing variants ( Table 2). The most common variant amongst the PRA cases was the PRCD c.5G>A variant, homozygous in 28 PRA-affected dogs ( Table 3).
The PDE6A c.1940delA variant causing rcd3 was found in two breeds, Pomeranian and Chinese Crested Dog. The variant explained one of the two affected Pomeranians' phenotype, suggesting the presence of additional PRA variants in the breed. In Chinese Crested Dogs, one LL suspected dog was heterozygous, but the two PRA affected were wild-type, indicating yet another genetic cause of PRA.
In the gene panel cohort, we observed RPGRIP1 insertion homozygotes or heterozygotes in ve new breeds: Chihuahuas (n=3), English Toy Terriers (n=1), Lowchens (n=1), Swedish Elkhounds (n=1), and Chinese Crested Dogs (n=1) ( Table 3). We further screened 1013 dogs to evaluate carrier frequencies in the new affected breeds and ten additional breeds known to harbor the variant (Table 4). Altogether, 23 RPGRIP1 insertion homozygotes were identi ed in both screenings. Three homozygous Chihuahuas were identi ed, of which two were PRA-affected. The owner of the third dog reported that it had been eye examined healthy at three years of age and had no problems with visual acuity by 10 years of age. Three homozygous English Toy Terriers were found, of which two were PRA-affected, and one was not eye examined. Of the 95 Lowchens tested, seven were homozygous for the RPGRIP1 insertion. One of the homozygotes was PRA-affected, and another was reported to have decreased visual acuity at 10 years of age. The other Lowchens had been examined at a young age (1-3 years old) or not at all. Five short-haired Miniature Dachshunds were also found to be RPGRIP1 insertion homozygotes, of which one was PRA-affected. Homozygotes were also observed in three other breeds, with no reported signs of PRA: American Staffordshire Terrier (n=1), Beagle (n=3), and long-haired Dachshund (n=1) ( Table 4). We found heterozygous carriers in nine of the 15 screened breeds (Table 4).
We next screened for the MAP9 deletion, a known early-onset modi er of the RPGRIP1 insertion associated PRA. In the screening, we included the above-mentioned 23 RPGRIP1 insertion homozygous dogs and 757 additional dogs in MAP9 deletion positive breeds found in commercial testing. All RPGRIP1 insertion homozygotes were wild-type for the MAP9 deletion. Only four MAP9 deletion heterozygotes were found, all of which were wild-type for the RPGRIP1 insertion (Table 4). To assess the retinal phenotype in a higher resolution, we chose ve RPGRIP1 insertion homozygotes (wild-type for the MAP9 deletion, aged 1.3-7.8 years) and three age-and breed-matched controls for spectral-domain optical coherence tomography (SD-OCT) imaging. We did not nd observable changes in the retinal thickness between the cases and controls (Online resource 3). Despite the extensive population screening of the MAP9 deletion variant, we did not nd homozygotes for imaging. The PRA-affected RPGRIP1 insertion homozygotes were not available for SD-OCT.
As the last example of PRA, we screened the TTC8 c.669delA variant associated with gr2-PRA (Downs et al., 2014) in Finnish Labrador Retrievers (LR) as it has been found in other LR populations. Three heterozygotes were identi ed in the screening of 190 LRs, indicating a 1.6 % (3/190) carrier frequency.
We did not nd any causal variant within the 49 tested variants in other retinopathy cases or the dogs affected with PRA and glaucoma (Table 2).

Discussion
Dogs and humans share signi cant similarities in eye physiology, and dogs have become a valuable spontaneous animal model to study the genetics and therapeutics of blinding eye diseases. Despite an increasing speed of discovery of new causative variants across canine eye disease groups, the genetic cause in most cases is still expected to remain unknown. We wanted to test this hypothesis by screening 49 known variants in nearly 200 dogs affected with PRA, glaucoma, or lens luxation and validated the results in additional samples from almost 1200 dogs. The panel screening identi ed causal variants in ten different genes across breeds, including new affected breeds, but explained only 28% of the studied cases. These results reveal remarkable genetic heterogeneity and highlight the extensive lack of molecular descriptions in canine eye disorders. Our study underlines an urgent need for new molecular discoveries that would be critical not only to improve veterinary molecular diagnostics and the understanding of pathophysiology but also to explore new therapeutic options for the corresponding human eye disorders.
The availability of gene panel screening has made it feasible to quickly and affordably test for the most known variants associated with a given phenotype with important implications. For example, it is assumed that up to half of the diagnosed human eye diseases could be treatable before advancing to deterioration of visual acuity (Assi et al., 2021;Solebo et al., 2017), highlighting the relevance of molecular diagnostic tools for early disease risk prediction. The same could also be true for dogs. We applied a gene panel screening approach with the most up-to-date list of known canine syndromic or nonsyndromic variants, covering over 90% of OMIA-listed eye disease variants. Yet, over 70% of the cases remained unexplained.
It is possible that some of the phenotypes observed could have been explained by other known variants not included in our gene panel. For example, we missed the variant in the C2ORF71 gene, which causes a late-onset PRA (rcd4) in the Irish Setter breeds (Downs et al., 2013). We had two PRA-affected Irish Setters in the screening but failed to nd an associated variant in both. Similarly, seven Miniature Schnauzers affected with PRA and one with overlapping PRA and glaucoma were screened without detected variants. Genetic causes for these cases could be found if the two recent PRA variants had been present in the panel: an intronic variant in the HIVEP3-gene and a structural variant in PPT1-gene (Kaukonen et al., 2020;Murgiano et al., 2018). However, the lack of some of the recent variants in the panel does not change the critical conclusion of the results, highlighting the need for further research in all eye disease groups for progress in the eld.
Lens luxation has been associated with a variant in the ADAMTS17 gene (Farias et al., 2010;Gould et al., 2011). Most of our cases were not explained by that variant as 20 out of 36 cases were wild-type. These LL cases are likely to harbor another genetic variant, or their phenotype is secondary to an underlying cause due to trauma or other conditions. In humans, the FBN1 gene, which is best known for its association with Marfan-and Weill-Marchesani syndromes with EL and skeletal changes, is also known to cause an isolated EL phenotype, thus making it a possible candidate gene (Comeglio et al., 2002).
However, none of the affected dogs in our cohort were reported to have skeletal abnormalities.
Panel screening sometimes reveals unexpected results: in this study, we validated a surprising discovery of the ADAMTS10 glaucoma-associated variant in East Siberian Laikas. The POAG-associated ADAMTS10 variant was originally identi ed in an isolated and inbred laboratory beagle population (Kuchtey et al., 2011). The two breeds are not closely related (Donner et al., 2018;Parker et al., 2017), suggesting that the variant is old or just recreated by chance. The ESL originates from early Russian and Siberian dogs, which come from the areas of the earliest archeological evidence of canids (Shannon et al., 2015). Gene panel screening can result in surprising ndings that may not have an immediate explanation. However, the variant appears to cause glaucoma in the ESL breed with a notable carrier frequency (14%), and gene testing is recommended to eradicate the condition from future populations.
PRA, a retinal disease closely resembling human retinitis pigmentosa, is the most studied eye disorder in dogs with most gene discoveries. We screened over 20 known PRA variants but those explained only 27% of the cases. Unsurprisingly, many of these cases were homozygous for the PRCD variant, which affects over 40 breeds (Miyadera, Acland, et al., 2012). Many breeds still lack molecular descriptions. The TTC8 variant associated with gr-PRA2 in Labrador Retrievers had a very low frequency (2%) in the Finnish LRs, correlating with the simultaneous presence of PRA causing PRCD variant and the low number of PRA diagnoses in the breed according to the public database of the Finnish Kennel Club. In addition, we found that the PDE6A variant, which causes rod photoreceptor degeneration through cGMP accumulation (Barbehenn, E et al., 1988;Petersen-Jones et al., 1999), causes PRA in Pomeranians, but is not the only causative variant in the breed as one of the two cases was wild-type for it.
The RPGRIP1 insertion and MAP9 deletion variants have been associated with PRA or its age of onset (Mellersh et al., 2006;Forman et al., 2016;Miyadera et al., 2012). However, the clinical signi cance of these variants remains unclear with a suggested incomplete penetrance (Das et al., 2017 (Forman et al., 2016;Miyadera et al., 2012). The MAP9 discovery in Coton de Tulears was new; however, we did not nd the RPGRIP1 variant in the breed. None of the PRA-affected dogs carried either variant, suggesting other disease-causing variants in the breed. Of the 23 identi ed RPGRIP1 insertion homozygotes, six were PRA affected. The average age at diagnosis for the affected dogs was 8.1 years, whereas the other 17 homozygotes had no reported signs of PRA. Ten of the 17 had been ye examined on average at four years old, and the youngest dog was only 1.9 years old at the time of the study. The remaining seven dogs had not been eye examined and were either young or not available for eye examinations. As none of the homozygotes had the early onset modi er, it might be assumed that the symptoms might appear much later in life, or in the case of the oldest homozygotes, have possibly gone unnoticed as eye examinations had not been performed.
The developments in gene therapy highlight the importance of identifying novel risk variants in canine eye diseases. For example, the RPE65 gene, which harbors RP and PRA-associated variants, was the rst target for gene therapy development to treat blindness. The therapy was rst developed in spontaneous canine models with RPE65 associated PRA and successfully restored the target cells' function (Aguirre et al., 2007;Le Meur et al., 2007;Narfström et al., 2003). Following the success in animal models, the gene-therapy was developed for human use in Leber's congenital amaurosis (Bainbridge et al., 2009;Cideciyan et al., 2008). Since then, gene therapy development has been of constant interest in dog models: at least four genes causing retinal diseases in both species are being studied for gene therapy development. In summary, we have performed the most comprehensive gene panel screening for three common canine eye disorders and report extensive genetic heterogeneity, new affected breeds, and a striking lack of molecular description in most cases. Gene panel tests are already powerful tools for veterinary diagnostics and breeding programs, especially for eye disorders with many variants and many forms of the disease in the breeds, and panels will only get better with improving content over time.
Simultaneously, the affected breeds and dogs may serve as spontaneous models to human eye disorders to investigate the disease pathophysiology and explore possibilities for new therapeutics powered by the emergence of the latest gene-editing technologies. Declarations