Introduction

Inherited retinal diseases (IRDs) are a group of conditions affecting humans and other mammals, which cause loss of sight due to faulty development, metabolic dysfunction, or degeneration of the retinal photoreceptors1,2,3,4. The common theme between IRDs is that visual impairment results from dysfunction or degeneration of the highly specialized photoreceptors that convert light energy into neuronal signals transmitted through retinal interneurons to higher visual centers. Inherited retinal disorders are very variable in their clinical manifestation and in their genetic etiology (which contributes to their common occurrence)5,6.

In humans, more than 300 genes (https://web.sph.uth.edu/RetNet/- accessed 12/8/2024) are associated with IRDs. These can occur in isolation but can also be a part of larger multi-systemic disorders7; in fact, manifestations equivalent to retinitis pigmentosa, cone-rod dystrophy, or maculopathy may belong to the overall presenting symptom of a syndromic disease5. In certain cases, patients affected may, in fact, have other systemic manifestations at the time retinal disease is diagnosed, even with significant morbidity7. More than 80 forms of syndromic retinal diseases have been described in humans so far, including ciliopathies8. syndromic metabolic disorders of glycosylation, neuronal ceroid lipofuscinoses9, mucopolysaccharidoses (MPSs)10, and peroxisomal diseases11, most often recessively inherited5. Phenotyping can also be non-obvious as similar phenotypes can be associated with variants occurring in different genes12,13, or systemic manifestations may appear later after the appearance of isolated visual symptoms7,14.

Nonetheless, an overall significant portion of the variable phenotypes associated with IRDs can be associated with the specific gene carrying the disease-causing variation6. For example, individuals with pathogenic dominant variants in BEST1have a build-up of lipofuscin in the retinal pigment epithelium (RPE) along with subretinal elevations with auto-fluorescent material, which results in visual impairment; this is due to the role and function, and therefore expression, of the gene in specific retinal tissues15. In other cases, there is a reported allelic hierarchy of specific variants associated with a phenotype. There, genetic variants can be classified as ‘mild’, ‘moderate’, ‘severe’ as an example for ABCA416, or ‘syndromic’ versus ‘non-syndromic’ for USH2A6,17.

Comparable diseases are also mapped in mammalian species like dogs, cats, sheep, horses, and non-human primates3,18,19,20. Often, common dog breeding practices lead to a high likelihood of purebred dog breeds being naturally-occurring models for human inherited diseases (https://omia.org/home/, accessed 12/8/2024), IRDs included. Indeed, more than 40 genes have been associated with different forms of IRDs in dogs, most of which classified as progressive retinal atrophy (PRA) in veterinary medicine2.

As in man, retinal diseases in dog can be isolated disorders, usually affecting the retinal pigment epithelium (RPE), photoreceptors - either rods and/or cones - or the bipolar cell layer2,21,22,23,24,25,26. Nonetheless, other cases, also in dogs, IRDs can be syndromic, and associated with oculo-skeletal defects27, neurological abnormalities28, an ensemble of obesity with renal, sperm and olfactory manifestations29, or craniofacial alterations30.

Research into the molecular basis of blinding retinal diseases in dogs is vital for the welfare and health of numerous canine breeds. Additionally, dogs are invaluable animal models for the development of retinal therapies transferrable to a clinical setting31,32,33, providing essential insights into the molecular players and mechanisms driving IRDs in humans3,32,34.

The progress made in identifying many different inherited retinal diseases and the causative genes/mutations in dogs has created confusion about how these diseases are named and classified. To address this issue, a recent publication has proposed consensus guidelines for nomenclature for the retinal diseases affecting dogs and other companion animals2.

In this study, we report on a litter of six dogs where three siblings presented signs of retinal degeneration, and no other specific disease-associated manifestations were detected. The other three littermates and parents were non-affected.

Results

Presentation of cases

Retinal defects were reported in a litter of Labrador retriever dogs bred as guide dogs for the blind/service dogs for veterans or first responders (Fig. 1). Firstly, a 7-months Labrador retriever dog (Fig. 2a, LR5) showed signs of visual impairment, and eye examination indicated that retinal degeneration, estimated as mid-stage disease (according to staging guidelines previously reported35), was present. A littermate (Fig. 2a, LR3) examined at 7.5 months showed no retinal abnormalities. By 10.5 months of age, LR3 had distinct retinal degeneration lesions on ophthalmoscopy and visual impairment; he was not following balls thrown and was also manifestly using his nose to test for the cage door or walking into a cage door. A review of the videos indicates that the dog has lost most of its peripheral visual fields as he cannot track a moving ball that is rolled from his side. However, his central vision, although compromised, can locate the ball when it stops moving (Video S1). At 11.5 months of age, ophthalmic examination confirmed a progressive retinal atrophy (PRA) diagnosis. A third dog from the same litter (LR8, Fig. 2a) with the same clinical signs was later identified (1.5 y). The remaining 3 siblings and parents had ophthalmic examinations and were normal. Compared to an unaffected 15-months Labrador retriever, fundus photographs of L3 taken at 25 months of age showed late-stage retinal degeneration (Fig. 1a, b). Re-examination of L3 at 3.3 yrs indicated complete mature cataracts which precluded retinal examination, and an ERG was non-recordable. Complete mature cataracts also developed in dogs LR3 and LR8. In LR8 the 2ry cataract developed at a comparable age but were accelerated in L3 because diabetes mellitus.

Fig. 1
figure 1

Phenotype. Fundus pictures of unaffected (a) and affected (b) dogs. a - Picture of the left retina of a 15-months old unaffected and unrelated Labrador from the same kennel. b– Fundus the right eye of a PRA-affected case at 25 months of age demonstrating late-stage disease. Note the marked thinning of the retinal vessels and pale atrophic optic nerve head. Both normal and affected dogs have a normal variation of the tapetum color which appears brown to dull gray.

Full size image
Fig. 2
figure 2

Mapping. (a) Family tree of the affected dogs. Males in squares, females in circles, affected in blue. Obligate carriers (and unaffected sibling later identified as carriers) as half-filled shapes. Autosomal recessive inheritance mechanism was suspected. The whole family was genotyped on canine SNP chip, whole-genome sequenced dogs are marked with red asterisk. (b) Homozygosity mapping of the three cases, compared against the three available unaffected siblings and two parents was carried out. The homozygous regions shared by all the cases and exclusive to them compared to the controls are marked in red (37 Mb region in CFA12 marked with an asterisk).

Full size image

Family tree and inheritance mechanism

A total of three cases (two males, one female) and three unaffected siblings (two females, one male) and the unaffected parents were initially examined and collected; the family tree is shown in Fig. 2a. Other than the three cases, no other dog belonging to the breeding program was reported to present comparable vision deficits orretinal abnormalities (Figure 2a). Based on pedigree analysis we positedautosomal recessive mode of inheritance.

Mapping of the candidate regions

We used homozygosity mapping to map the critical interval containing the candidate variant. To this end, we genotyped the entire eight-dog family. With the assumption of a monogenic recessive inheritance, searched for extended regions of homozygosity, and detected regions > 1 Mb in CFA8, CFA9, CFA12, CFA17, CFA22, CFA35. Once the critical homozygous candidate regions were confirmed, the coordinates of the candidate region were converted from Canfam 3.1 (NCBI Assembly ID: 317138) to the UU_Cfam_GSD_1.0 (Canfam4, NCBI Assembly ID: 6119491) to serve as a reference for whole genome sequencing. The whole list of intervals is reported in Table 1. No other region was shared by all the cases not shared by the controls. The results for homozygosity mapping are shown in Fig. 2b, in which all the shared exclusive homozygous regions are marked in red.

Table 1 Candidate intervals obtained through homozygosity mapping of three affected siblings and five unaffected individuals (three siblings and two parents). We detected six different candidate regions shared by the cases and not by the controls. The large CFA12 interval contains the putative causative variant. All coordinates are reported on the Canfam4 reference.
Full size table

Variant detection and genotyping

Four dogs were sequenced, two affected and two unaffected parents, and mapped against the Canfam4 reference, and the variants called as described in methods. The Canfam4 reference was selected because the called variant could then be easily filtered against the large Dog10k database36which could then be used as control. As an additional filtering measure, SNPs and small indels positions were converted to canfam3.1 and filtered against two other databases: Dog Biomedical Variant Database Consortium (DBVDC)37, and European Variation Archive (EVA - https://www.ebi.ac.uk/eva/, accessed 12/8/2024). All the small indels and large structural homozygous variants falling within the candidate intervals were filtered as follows: (I) excluded if not heterozygous in the sequenced parents; (II) excluded if present in the DBVDC or Dog10k databases. A single 3-bp deletion in GTPBP2 (NC_049233.1:12,264,348_12,264,350del – Exon 11) was the only exclusive variant detected (Fig. 3a). Reported genetic variants in human for GTPBP2 are associated with Jaberi-Elahi syndrome (OMIM #607434).

Fig. 3
figure 3

DNA sequencing. (a) Detail of the of the 3-bp in-frame GTPBP2 deletion variant exclusive for the cases. Screenshots of the interval visualized with IGV. On the top, a sequenced case is shown, homozygous of the haplotype and variant, and the deleted 3-bp variant. Further down, a carriers visible. Bottom panel of (a) normal retinal RNA-seq shows the variant falling within the coding region of the GTPBP2 gene (position based on the c-terminal part of isoform 1: c.1606_1608del, p.Ala536del). (b) Sanger results for the variant showing the electropherograms for wild-type, carrier and affected. In the wild-type (top), the red rectangle marks the triplet deleted in the affected, in which the deletion position is marked with an arrow (bottom).

Full size image

Lastly, we genotyped the 3-bp deletion (Fig. 3b) in all the available Labrador retrievers of the same kennel (91 plus the affected) and in all the 569 Labrador retrievers available in our database (all of these of North American origin) This confirmed the variant’s extreme rarity (Table 2).

Table 2 Distribution of the 3-bp deletion candidate. Variants in our Labrador sample pool and in a database of canine variants (Dog 10k multiple breeds) are shown, indicating a concordant segregation within and outside the breed.
Full size table

Variant predicted impact

Due to the nature of the variant, and the fact it was private to the small PRA-affected family, GTPBP2detection in our in-house RNA-seq retinal libraries (average TPM: 44.46, st. dev. ±5.77)38, the concordant segregation with the phenotype, and the fact that variants in GTPBP2 result in retinal manifestation in human, we concluded that we had a very plausible candidate. The variant is predicted to affect the c-terminal part of isoform 1: c.1606_1608del, p.Ala536del (File S2). The amino acid loss was submitted to PolyPhen-2 (HumDiv and HumVar), PROVEAN and MutPred2 Indel prediction programs. The first two reported a “Probably Damaging” predicted, while Provean and MutPred reported “Deleterious”, with a MutPred score of −9.771.

Software like Interproscan (see methods) and NCBI protein repositories were used to detect patterns and predict domains. The information obtained, alongside the alignment of the canine protein with other mammalian orthologs (Fig. 4a), showed that a highly conserved alanine is deleted, and the mutated portion of the protein is well outside the GTPase domain of GTPBP2 (Interpro ID: IPR000795). In fact, the mutant deletion falls within a predicted domain found in the translation elongation factor EF1A (IPR009001), Fig. 4b and File S2, which recognizes and transports aminoacyl-RNA.

Fig. 4
figure 4

Variant position and GTPBP2 sequence. (a) Alignment of the c-terminal Isoform 1 (NCBI: XP_038539566.1, entry under UU_Cfam_GSD_1.0) of canine GTPBP2 with the same interval in other five different mammalian species. Observe the high degree of conservation of the protein. The deleted Alanine is highlighted. (b) 3D structure of human GTPBP2 with the known human syndromic diseases associated variants marked in red for frameshifts and premature stop codons, and in yellow for amino acid changes. alongside the canine variant transposed, in green. In purple, D1 and D2 show the boundaries of the GTP domain, with variants numbered 15 and above outside of it. Bottom right corner, whole structure without the cutoff of the N-terminal loop. 1-Ser3X; 2-Leu93P; 3-Arg121X; 4-Lys125R; 5-Arg131X; 6-Arg144X; 7-Arg219X; 8-Asp319N; 9-Gln332fs; 10-Glu352fs; 11-Gln407X; 12-Val413fs; 13-Arg423X; 14-Arg470X; 15-Glu509fs; 16-Arg520X; 17-Arg521fs; 18-Ala536del.

Full size image

Expression of WT and Ala536del mutant GTPBP2 protein in HeLa cell in vitro model

To investigate the effect of deletion Ala536 (A536del) on GTPBP2 protein, the mutation in the dog full-length GTPBP2 cDNA was cloned in pcDNA3.1 expression vector by site-directed mutagenesis. Heterologous cellular systems were then transfected with cDNA encoding WT or A536del mutant GTPBP2 protein. Immunoblot analysis of total cell lysate with antibodies to GTPBP2 showed that compared to the WT form, the expression level of the mutant GTPBP2 protein is slightly but not significantly affected (Fig. 5a). The expression of endogenous GTPBP2 protein in Hela cells is negligible. To examine the possible role of ubiquitin-proteasome system (UPS) in the disease, after transfection the cells were incubated with the specific UPS inhibitor MG132. Immunoblot data showed that the treatment with MG132 does not increase the expression level of the A536del GTPBP2 (Fig. 5a, b).

Fig. 5
figure 5

Expression level analysis of GTPBP2 protein in HeLa cells transfected with WT and mutant A536del GTPBP2 constructs. Cells were transfected with WT GTPBP2 or with A536del GTPBP2 cDNAs. Cells transfected with A536del GTPBP2 were then treated with the proteasome inhibitor MG132 (+). Untransfected cells were used as control. a) Total protein lysates from untransfected and transfected cells were obtained by solubilization. An equal quantity of protein was separated by SDS-PAGE and blotted onto nitrocellulose paper. The blots were incubated with polyclonal antibodies to GTPBP2. A cropped representative Western blot is shown. The lower panel is a Ponceau Red staining of the same cell lysates, used as loading control. b) Quantification of GTPBP2 protein bands was performed by densitometric analysis on western blots. Data (mean values from at least three independent experiments + S.D.) are reported as the percentage of values from WT GTPBP2 protein. Statistical analysis was performed by One-way ANOVA test, followed by multiple comparisons Dunnett’s test. * p < 0.05.

Full size image

The cellular distribution of GTPBP2 protein was investigated using confocal microscopy analysis of cells transfected with WT or A536del GTPBP2 proteins. Immunolabeling showed that the WT form is present diffusely throughout the cytoplasm (Fig. 6a, b). In contrast, expression of mutant form resulted in the formation of cytosolic aggregates in ~ 70–80% of cells (Fig. 6c, e), while in the remaining ~ 20–30% of cells, the distribution is similar to that of the WT protein (Fig. 6c-f). Cells were also incubated with antibodies to Lamp2 (lysosome-associated membrane protein 2, a marker of the membrane of lysosomes, the organelles involved in degradation and recycling processes in cells), and co-localization of A536del GTPBP2 protein aggregates with Lamp2 protein marker was not found (Fig. 6d, f).

Fig. 6
figure 6

Cellular localization of WT and A536del mutant GTPBP2 proteins. HeLa cells were transfected with WT (a,b) or with A536del mutated GTPBP2 cDNAs (c,d,e,f). Untransfected cells were used as control (g,h). Cells were immunolabelled with polyclonal antibodies to GTPBP2 (red fluorescence) and subsequently with monoclonal antibodies to Lamp2 (green fluorescence), were then incubated with the appropriate secondary antibodies. Nuclear morphology was demonstrated by staining with Hoechst. Images were recorded at the same setting conditions and magnification (scale bar 10 um).

Full size image

Assessment of clinical disease phenotype in affected dogs and comparison with patients with JES

Once the causative GTPBP2 variant was found associated with the retinal disease in dogs, we reassessed the clinical findings in the 3 affected dogs considering the disease in human patients. Most Jaberi-Elahi syndrome human patients have neurologic abnormalities. These include motor alterations, seizures and intellectual disabilities (summarized in Table S3). As well, there are general morphologic abnormalities. In contrast, the affected dogs had no neurologic or external morphologic abnormalities (Table 3). They had normal facial features, appendicular skeleton and haircoat (see Video S1 for details of general physical characteristics). Two of the dogs developed Addison’s disease which was readily managed with appropriate medical treatment; one of these later developed diabetes mellitus and died from a hypoglycemic event (Table 3). Overall, there were no common findings in the three affected dogs that could be interpreted as representing a syndrome associated with the retinal degeneration and GTPBP2 mutation.

Table 3 Clinical findings in affected dogs.
Full size table

Discussion

Through careful phenotyping and sample collection, we used the power of combined SNP-based homozygosity mapping and whole genome sequencing to detect an extremely rare in-frame 3-bp deletion in the canine GTPBP2associated with a non-syndromic retinal degeneration. The gene encodes a G-protein expressed in several tissues, including canine retina38. While we had access to a large control pool within and outside the breed, we are aware that the low number of cases, all within the same population, suggests that this is a ‘private’ mutation with a low frequency in the overall Labrador retriever population.

Following the canine IRDs classification guidelines detailed in Mowat et al.2, since there sufficient evidence that the condition we report is (I) heritable (II) progressive in nature (III) broadly, a retinal atrophy, we can classify it as PRA-GTPBP2.

The deletion we report results in the loss of a conserved Alanine. Interestingly, while the deleted amino acid falls within a region highly conserved in mammals, but not belonging to the major GTPase domain of the protein. The nature of the in-frame deletion precluded, a priori, the use of some prediction tool, if these assumed an amino acid substitution as the only possible input, but the ones that accepted a deletion as input gave an overall consistent output of a confident prediction of a damaging effect.

With this starting information, we hypothesized that the amino acid Ala 536 could have a role in the proper folding of the GTPBP2 protein. It is well known that misfolded proteins can have an enhanced susceptibility to the protein degradation via ubiquitin-proteasome and/or autophagic-lysosomal pathways39– therefore leading to a less amount of detectable protein in the cell lines transformed with the mutant. Our preliminary results show that the amino acid deletion does not reduce the expression of GTPBP2 protein in heterologous cellular model, and that the inhibition of proteasome by MG132, does not change the content of the mutant protein. Furthermore, using the same cellular model, we found that mutant GTPBP2 accumulates in the cytoplasm as round or elongated clusters not associated with lysosome markers even though we cannot exclude that, by increasing the incubation time, this co-localization may be detected. Overall our results suggest that the GTPBP2 variant has no apparent significant impact on the proteolysis of the mutant, but dramatically changes its cellular localization and probably creates aggregates that can be harmful to the cellular environment which includes altering proteostasis, lysosome function or inducing DNA damage40.

The accumulation of misfolded and protein aggregates is a typical feature of many neurodegenerative diseases, including Alzheimer’s, Parkinson’s, and Huntington’s diseases41. GTPBP2 mutations have been linked to neurodegeneration in humans as well42. When retinal degeneration was detected in these cases, it was always associated with other manifestations and was therefore syndromic. We could speculate that the Ala 536 amino acid deletion, despite not falling within the functional GTPase domain, results in a damaged GTPBP2 protein that causes its aggregation, which is just sufficient to impact the retinal environment with no major, detectable effects elsewhere.

GTPBP2 belongs to the GTPase superfamily, also called G-proteins43. GTPBP2 has been associated with several cellular processes including positively regulating the invasion, migration and proliferation of lung cancer non-small cells44; BMP signaling and mesoderm patterning in Xenopusembryos45, and is also a positive regulator of Wnt signaling while maintaining low levels of the negative regulator Axin46. Ishimura and colleagues described a splice variant in Gtpbpin mice shown to cause neurodegeneration and retinal damage and highlighted the role of GTPBP2 as a ribosome rescue factor47.

The first reported significant genetic variants in human for GTPBP2are associated with Jaberi-Elahi syndrome (JES)42,48. Among the findings, apoptosis of neurons and of mutant granule cells in the dentate gyrus as well as CA2 pyramidal neurons, and layer IV cortical neurons, are reported. Importantly, neurons in the retina, including photoreceptors and amacrine, horizontal, and ganglion cells, also degenerate42,48Significant genetic variants associated with JES (all premature stop codons or frameshifts) are p.Val413fs42,48, p.Arg144X, p.Gln407 ×48, p.Arg470X, p.Arg520X, p.Arg219 ×49, ; p.Arg470X and compound heterozygous p.Arg432X/p.Arg131 ×49.

Recently, Salpietro and colleagues50 described homozygous variants in GTPBP2occurring in 16 individuals presenting variable symptoms, including neurodevelopmental impairment, pathognomonic craniofacial features, microcephaly and brain atrophy. Four of the individuals reported had retinal dysfunctions. Novel frameshifts and premature stop codons included p.Ser3X, p.Arg121X, p.Gln332fs, p.Glu352fs, p.Glu509fs, p.Arg521fs. Interestingly, the authors also report three amino acid substitutions as pathological (p.Leu93P, p.Lys125R, p.Asp319N50). Salpietro et al. also observe that only part of the cases they describe can fall within JAS50. Of note, other variable manifestations are occasionally reported in human51. This shows that even within GTPBP2-related human syndromes the phenotype variability is significant.

The variant in dogs is characterized by being a single-amino acid deletion in the C-terminal region, while many JAS-associated variants truncate the protein, sometimes very early, before or within the GTP-binding domain. Variants closer to the C-terminus are also reported in human, like p.Arg520X or p.Arg521fs49,50). which is outside the GTP domain like the one described in the dog. While three reported pathological variants found in human are amino acid changes, no single amino acid deletion was reported so far, and the variant described here is closer to the C-terminal of the protein than any of those.

Wallen et al. suggested GTPBP2-associated diseases may be functionally related to certain peripheral neuropathies caused by variants in aminoacyl-tRNA synthetase genes52. Interestingly, in a study focusing on GTPBP2 and on its likely (66% identity) paralog GTPBP1, Zinoviev and colleagues reported binding of GTPBP2 by Phe-tRNAPhe (which they attributed to the possibility of interaction with aa-tRNA)53; additionally, Zuko and colleagues demonstrated how inactivating GTPBP2 exacerbated peripheral neuropathy in mice with mutated tRNA synthetase54. The variant we describe falls within a predicted aa-tRNA interacting domain (and highly conserved sequence). For this reason, while the phenotype could be indirectly related to GTPase activity (i.e., lack of activity due to mislocalization), we cannot exclude that the variant could have a damaging impact not directly related to the GTPas activity of GTPBP2. Intriguingly, using ribosome profiling, Salpietro et al. did not detect any ribosome stalling or other translational defects in fibroblasts derived from the affected individuals described in their study, albeit they observed that cell-type specificity or specific cellular stresses not captured by their assay could have played a role in their results50.

In this study, we show a retinal-specific phenotype occurring in dogs with a deletion in a conserved Alanine, in a conserved domain, in a canine protein with a 99.50% identity with human GTPBP2. Albeit JAS is reported as having a retinal component, GTPBP2 as a retinal gene is underrecognized, missing from RetNet (https://web.sph.uth.edu/RetNet/ - accessed 12/8/2024). The retina-specific phenotypic manifestation associated with an in-frame 3 nucleotide deletion of GTPBP2support the lesser impact of the variant described compared to JAS (and other human syndromes) associated ones, with signs apparently circumscribed to the metabolically more demanding and vulnerable retinal environment55.

In fact, we can speculate that the mutant protein is not degraded by the ubiquitin proteasomal pathway, because the 24 h of monitoring should be a sufficiently long time for us to detect its activity. Similarly, lack of lysosomal co-localization in the cultured cells argues for degradation of the mutant protein by lysosomes. More probably, the protein could instead be destroyed by lysosomal activity (the 24 h of monitoring time could be in this case an explanation for the lack of localization). A likely possibility is that the alteration of the conserved RNA-synthetase putative domain predicted by InterPro could have a role in this malfunction. A future study focusing on polysome localization of mutant and wild-type proteins along with interactome of the putative C-terminal domain and an extended half-life analysis will further inform on the impact of the variant.

Our study reports apparently non-syndromic retinal degeneration associated with a variant occurring in a gene usually causative of syndromic conditions including retinal degeneration. The variant is present in a conserved domain of GTPBP2 outside of the GTP-domain; it is so far the closest to the C-terminus of the GTPBP2-associated pathological variants reported and seems to have an impact on the cellular localization of the protein. Our study adds a novel variant to the spectrum of canine retinopathies and large animal models and contributes to the identification of non-syndromic retinal degeneration associated with genes usually causing syndromic diseases.

Materials and methods

Case identification and sample collection

Ethical statement

The research and the eye examinations were conducted in full compliance and strict accordance with the Association for Research in Vision and Ophthalmology (ARVO) “Resolution on the Use of Animals in Ophthalmic and Vision Research”, and the study protocol was approved by the “Institutional Animal Care and Use Committee” (IACUCs), University of Pennsylvania (code 806301), and were performed in accordance with relevant guidelines and regulations. Eye examinations were carried out by ACVO board certificated veterinary ophthalmologists (GDA, and diplomates who referred the cases). All methods are reported in accordance with ARRIVE guidelines.

Phenotype assessment

Ophthalmic testing

After diagnosis of a presumed retinal degeneration in a 7-months Labrador retriever dog (LR5, Figs. 1 and 2a) during the ACVO/Epicur National Service Animal Eye Exam Event, clinical records and pedigree information were collected, and examinations were carried out on the first affected dog and its five littermates (of these, there were two additional affected and three unaffected), as well of the parents. (Fig. 2a). All dogs received a dilated fundus examination and slit-lamp biomicroscopy, and fundus photography (Genesis) was done in selected cases. Objective vision assessment and video recordings were provided by the owners to demonstrate the degree of vision impairment in selected cases.

General clinical assessment

The dogs are raised at the breeding facility until 8–10 weeks of age, or slightly later, where they are under veterinary supervision and receive physical examinations aimed at detecting structural or medical issues that would prevent their use as a guide dog for the blind/service dog for veterans or first responders. They are then placed with puppy raisers until 16 months of age and returned to the facility for formal training. Local veterinarians care for puppies’ needs where they are being raised. If medical or other issues preclude their use as a guide or service dog, they are adopted by volunteers who manage their medical care with their local veterinarian. Special examinations/testing are done on an as needed basis.

Labrador retriever samples and single nucleotide variant (SNV) genotyping

Blood- and cheek-swab-derived genomic DNA samples from a total of 663 Labrador retrievers were used. Of these, three were cases, five were unaffected family members (both parents and three unaffected siblings). In addition, 86 dogs belonging to the same service dog breeding program, and 569 unaffected, unrelated Labrador retrievers from USA belonging to different kennels were used. Genomic DNA was extracted with the Illustra DNA extraction kit BACC2 (GE Healthcare), following the manufacturer’s instructions. DNA samples from three cases, three unaffected siblings, and two parents were genotyped on the Illumina Canine 220k HD SNP chip. Genotyping data were pre-processed, and a homozygosity analysis was carried out with Plink v1.9 using the default parameters.

Detection of the candidate variant

Homozygosity mapping

SNP genotyping was followed by homozygosity mapping, based on the assumed recessive mode of inheritance. The mapping was carried out with PLINK v.1.956 to detect extended intervals of homozygosity with shared alleles. The fine-mapping intervals were set to a minimum of 1 M base pairs; we also considered any instance in which cases and controls would be clustered into the same overlapping homozygous hits by PLINK, but a small fraction exclusive of the cases existed in the flanking intervals. Homozygosity analysis was carried out using the commands “--dog” (to account for the species-specific chromosome quantities), “--homozyg” and “--homozyg group”, and the standard parameters. Coordinates of the output were then converted to Canfam4 using the NCBI remapping service (accessed 08/03/2023, now unavailable) and the UCSC liftover service (https://genome.ucsc.edu/cgi-bin/hgLiftOver, accessed 12/8/2024).

Whole-genome sequencing

Two of the cases, assumed homozygous for the variant within the candidate intervals, and the two parents, assumed obligate heterozygous carriers, were selected for whole-genome sequencing (Fig. 2a, dogs LR1, LR2, LR3, LR8). Libraries of 300 bp insert size were prepared and Illumina HiSeq2500 paired-end reads (2 × 100 bp) were collected (one lane per sample). Fastq files were generated using Casava 1.8. A total of 561.77 millions reads (100 bp paired-end reads) for the four libraries were collected for the sequenced dogs from a shotgun fragment library (corresponding to an average coverage of the genome of 28.2x and 27.6x for the cases, and 26.7x and 28.4x for the parents). The paired-end reads were mapped against the dog reference genome CanFam4. The alignment was done using Burrows-Wheeler Aligner (BWA) version 0.5.9-r1657(default settings). The SAM file obtained by BWA was converted to a BAM file and sorted using samtools58. Sorted BAM files were visualized using Integrative Genomics Viewer (IGV)59.

SNV and short in-del discovery

Variant calling was carried out using GATK (version 4.2.9)60in the “HaplotypeCaller” mode with the output set to variant call format (vcf, version 4.0); the raw calls for all samples and sites were flagged using the standard variant filtration module of GATK, following the best practices. The prediction of the impact of the detected variants was done using SnpEff61, comparing the data with the Canfam4 reference. Data were then compared with the heterozygous parent, selecting variants homozygous in the cases and heterozygous in the obligate carriers. Remaining variants were then filtered again the Dog10k database36.

Filtered candidate positions were converted to the Canfam 3.1 canine reference using the NCBI remapping service (https://www.ncbi.nlm.nih.gov/genome/tools/remap, accessed 08/03/2023 now discontinued) and the UCSC liftover service (https://genome.ucsc.edu/cgi-bin/hgLiftOver, accessed 12/8/2024) to compare them against databases using this reference. Candidate variants in the homozygous interval present in the cases were also filtered against the Dog Biomedical Variant Database Consortium (DBVDC)37, using the software BCFtools57,62, plus additional searching was done in the and the European Variation Archive variant browser (https://www.ebi.ac.uk/eva/?Variant-Browser, accessed 12/8/2024).

Structural variants and mobile elements discovery

Delly263 was used to detect five type of structural variants in the BAM files (Duplications, Inversions, Insertions, Deletions, Translocations). BAM files from unrelated dogs belonging to other studies carried out by our group, and from the Dog10k database were also called and used to filter out potential variants. The commands for deletions, insertions, inversion, translocations and duplications were all executed separately. Each of these analyses was carried out focusing on the candidate regions found by mapping.

Data availability

Fastq for four whole genome sequenced data and the SNP-genotyped dogs are deposited in DRYAD (DOI: https://doi.org/10.5061/dryad.xwdbrv1kd, accessed 12/8/2024).

Genotyping

The 3-bp deletion proposed candidate variant was verified in the 3 cases and the available 660 controls by re-sequencing targeted PCR products and sequenced by Sanger sequencing after enzymatically cleaning with ExoSap-IT diluted in water 1:10 in an AB Genetic Analyzer. PCR primers were designed using PRIMER364; the PCR products were run on 1.5% agarose gel with a 0.5 μg/ml ethidium bromide added. PCR products were amplified using flanking primers for the CFA12 GTPBP2 3-bp deletion, F (5’-GGCCTCTTTCCTTTCTGGAG-3’, chr12: 12,264,251-12,264,270), R (5’-TTCGAGGCAGAGATTGTCCT-3’, chr12: 12,264,422-12,264,441). Amplification was carried out with AmpliTaqGold360Mastermix (Life Technologies). All the coordinates are indicated for Canfam4. The wild-type amplicon spans 171 bp. Sequence data were visualized using 4Peaks.

Protein sequence analysis

Protein sequence alignment was carried out using Clustal Omega. AA deletion impact was predicted with PolyPhen-2, PROVEAN, and MutPred-indel. Domain detection with InterPro. The degree of identity between orthologs and isoforms was calculated by BLAST. 3D models of GTPBP2 was created by Alphaphold for the human GTPBP2 isoform, uniport ID Q9BX10 and modified (https://alphafold.ebi.ac.uk/entry/Q9BX10, accessed 07/31/2024).

Functional studies

GTPBP2 construct and site-directed mutagenesis

The full-length adult Canis lupus GTP binding protein 2 (GTPBP2 transcript variant) cDNA was synthetically generated by Genscript Biotech (Piscataway, New Jersey, USA) based on the published database sequence (NCBI Reference Sequence: XM_538939.7). The ORF sequence was cloned into pcDNA3.1(+) via HindIII/NotI restriction sites, added upstream of the ATG and downstream of the STOP codon respectively. GTPBP2 mutation was generated by using the QuikChange site-directed mutagenesis kit (Stratagene, San Diego, CA, USA), according to the manufacturer’s instructions. The mutated GTPBP2 was performed with the following mutagenic primers: F (5’-GCAACGTACGTCAGACGGTGGTGGAAAAGATCCATG-3’) and R (5’-CATGGATCTTTTCCACCACCGTCTGACGTACGTTGC-3’). All constructs were verified by Sanger sequencing as described above.

TPM calculation

To search in our internally generated canine RNA-seq datasets for GTPBP2 transcripts to verify the expression of the gene, we used a previously published dataset consisting of nine dogs38, three normal and two groups of three affected with a non-allelic retinal disease. We used the RNA-seq analysis software Kallisto to calculate the average TPM (transcript per million) for GTPBP2 canine transcripts within all three groups.

Cell culture, transfection and treatment with proteasome inhibitor

HeLa cells (ATCC, Manassas, VA, USA) were counted, seeded at 100 000 cells/cm2 and grown in DMEM (high glucose medium), supplemented with 10% FBS. Cells were transiently transfected with wild-type (WT) or mutant GTPBP2 cDNAs using jetOPTIMUS® DNA (Polyplus Transfection, New York, NY, USA) transfection reagent, according to the manufacturer’s instructions. Sixteen hours after transfection, the inhibitor of the ubiquitin proteasomal system (UPS) MG132 (Sigma-Aldrich, St. Louis, MO, USA) was added (10 µM final concentration dissolved in DMSO 0.1%), and cells were incubated for 8 h. At the end of treatment, cells were gently washed twice with phosphate-buffered saline (PBS). The cells, which were grown on 13 mm glass coverslips, were fixed with 4% paraformaldehyde for 15 min at room temperature or lysed in a buffer containing Tris-HCl 50 mM pH 7.5, NaCl 150 mM and NP40 1% (v/v) supplemented with protease inhibitors (Sigma-Aldrich, St. Louis, MO, USA). Protein concentrations were determined by the Bicinchoninic Protein Assay Kit (Quantum Protein Assay Kit, EuroClone Pero, MI, Italy).

Gel electrophoresis and immunoblotting

Proteins were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto a nitrocellulose membrane; the blots were exposed to 3% bovine serum albumin in Tris-buffered saline, Tween 20 (TBS-T) (TBS + 0.05% Tween-20) for an hour and then probed with anti-GTPBP2 rabbit-monoclonal antibody (PA5-56738, 0.1µg/ml, ThermoFisher Scientific, Waltham, MA, USA) The membranes were then incubated with anti-rabbit- HRP secondary antibody (dilution 1:30,000, Sigma-Aldrich, St. Louis, MO, USA) and developed with DAB. The blots were imaged with iBright 1500 (ThermoFischer Scientific). Quantification of protein bands was performed by densitometric analysis on Western blots from at least three independent experiments. Uncropped gels and ponceau are shown in Figure S4.

Immunofluorescence analysis

Cells fixed with 4% paraformaldehyde were washed with PBS and then permeabilized in 0.5% Triton X-100 in PBS. As primary antibodies, rabbit anti-GTPBP2 PA5-56738 (1µg/ml, ThermoFisher Scientific) and mouse monoclonal Lamp2 antibody (Abcam ab25631) were used in 1% BSA/PBS solution. After the incubation period, cells were gently washed with PBS and incubated with the Alexa Fluor 568 red and Alexa Fluor 488 green secondary antibodies (dilution 2 μg/mL, ThermoFisher). Nuclear morphology was characterized by staining with Hoechst 33342 (dilution 1 μg/mL, ThermoFisher Scientific). Glass coverslips were sealed with mowiol (Sigma-Aldrich, St. Louis, MO, USA). Confocal microscopy was performed using a TCS-SP5 II confocal laser scanning microscope (Wetzlar, Germany).