Ethics statement
All examinations and animal experiments were carried out after obtaining informed written consent for participation by the owner and in accordance with local laws, regulations, and ethical guidelines. Blood samples or buccal swabs were collected with the approval of the Cantonal Committee for Animal Experiments (Canton of Bern; permits BE 71/19 and BE94/2022) or with ethical approval of Purdue University (IACUC # 1901001840).
Animal selection, and definition of breed and phenotype
This study was performed with DNA samples from a total of 937 dogs. MAS dogs represent a relatively young breed that was developed from (standard) Australian Shepherd dogs. MAS dogs were recognized as an independent breed in 2012 by the American Kennel Club and provisionally recognized by the European FCI in 2019. Furthermore, unofficial breed designations, such as Miniature Australian Shepherd or Toy Australian Shepherd have also been used by some breeders. We provide the owner-reported breed assignments of all dogs in our study in Table S1. The 937 dogs consist of 485 MAS, 321 Australian Shepherds, 127 Miniature Australian Shepherds, and 4 Toy Australian Shepherds. Samples originated either from the College of Veterinary Medicine, Purdue University (n = 151) or the Vetsuisse Biobank at the University of Bern (n = 786).
Each dog was assigned to one of three phenotypic categories. The neurological phenotypes were classified by a board-certified neurologist based on neurological examination, clinical reports, or videos of the dogs submitted by the owners. Twenty-five dogs were classified as ‘neurological, clinical signs compatible with NAD’ by the presence of hind limb weaknesses with or without ataxia, abnormal gait, scuffing of paws/dragging of digits, and kyphosis. All affected dogs exhibited a pacing/ambling gait (atypical ipsilateral movement of limbs), when gait could be adequately visualized. Six neurological dogs with signs that differed from the above mentioned were classified as ‘neurological, other’. The remaining 906 dogs were used as controls. For the majority of the controls, the owners reported their dogs as healthy at the time of sampling. For some of the control dogs, no phenotype information regarding neurological diseases was available. More detailed information on the 937 dogs is compiled in Table S1.
Clinical examinations
A proportion of affected dogs were evaluated clinically by board-certified veterinary neurologists (n = 10 out of 23 cases that ultimately were homozygous del/del at the discovered deletion variant, RNF170:XM_038559916.1:c.367delG, Table S1), with a complete neurological evaluation performed. The remaining 13 cases were variably under the care of general practice veterinarians. Following a call for screening of NAD in the breed, additional affected dogs were submitted to the study but not evaluated in person by a veterinary clinician. In this scenario, owners typically submitted videos of the dogs' gait, and these videos were reviewed by a board-certified neurologist (n = 9 out of 23 del/del cases, Table S1). The remaining four del/del cases were submitted with clinical descriptions only provided by the owners.
Pathological examinations
Necropsy reports were available for two neurologically affected MAS dogs (Case #3/Dog #79 and Case #4/Dog #139, Table S1), which were euthanized at the owner’s discretion under the care of their general practitioner veterinarian, due to severely diminished quality of life.
A complete necropsy with harvesting and evaluation of all the major organs, the brain, and the spinal cord, was performed on both cases. The tissues were fixed for 48–72 hours in 10% neutral buffered formalin and routinely trimmed and processed. Paraffin-embedded tissues were sectioned at 4–5 microns and stained with hematoxylin and eosin.
Additional sections of the central nervous system were mounted on charged slides (ProbeOn™ Thermo Fisher Scientific) and were used for immunohistochemistry staining for glial cells, including glial fibrillary acidic protein (GFAP) for astrocytes, and ionized calcium-binding adapter molecule 1 (Iba1) for microglia/macrophages. The immunostaining was performed using the Leica Bond RXm automated platform combined with the Bond Polymer Refine Detection kit (Leica #DS9800). Briefly, after dewaxing and rehydration, sections were pretreated with the epitope retrieval BOND ER1 low pH buffer (Leica #AR9961) for 20 min at 98°C. Endogenous peroxidase was inactivated with 3% H2O2 for 10 min at room temperature (RT). Nonspecific tissue-antibody interactions were blocked with Leica PowerVision IHC/ISH Super Blocking solution (PV6122) for 30 min at RT. The same blocking solution also served as diluent for the primary antibody. A rabbit polyclonal primary antibody against GFAP (Agilent (Dako), Z0334) and a rabbit monoclonal primary antibody against Iba1 (WAKO, 019-19741) at a concentration of 1/5000 and 1/1500, respectively, were used and incubated on the slides for 45 min at RT. A biotin-free polymeric IHC detection system consisting of HRP conjugated anti-rabbit was then applied for 25 min at RT. Immunoreactivity was revealed with the diaminobenzidine (DAB) chromogen reaction. Slides were finally counterstained in hematoxylin, dehydrated in an ethanol series, cleared in xylene, and permanently mounted with a resinous mounting medium (Thermo Scientific ClearVue™ coverslip). Normal canine brain and spinal cord sections from a young, unaffected mixed-breed dog were used as positive controls. Negative controls were obtained either by omission of the primary antibodies or replacement with an irrelevant isotype-matched rabbit polyclonal or rat monoclonal antibody.
DNA extraction and exclusion of PNPLA8
Genomic DNA was isolated from the EDTA blood samples or buccal swabs with either 1) the Maxwell RSC Whole Blood DNA Kit using a Maxwell RSC instrument (Promega, Dübendorf, Switzerland), 2) the Qiagen Puregene Blood and Tissue kit or the Qiagen DNeasy Blood and Tissue kit (Qiagen, Hilden, Germany) following the manufacturer’s protocol, or 3) with a standard phenol-chloroform extraction.
Due to the close relatedness of MAS to Australian Shepherds and the overlap of clinical signs in affected dogs, the previously-published PNPLA8:XM_005630935.2:c.1169_1170dup frameshift variant, discovered in Australian Shepherds with hereditary ataxia, OMIA variant ID1470 (Abitbol et al. 2022) was first tested in a very small subset (n = 16) of the total 485 MAS dogs. Specifically, four neurologically affected MAS, five unaffected but obligate carrier MAS, and seven old (9 + years of age when tested), distantly related, phenotypically normal MAS were genotyped, and all dogs were homozygous wildtype at PNPLA8I:XM_005630935.2:c.1169_1170dup, suggesting this variant was likely not causing the NAD in this breed.
SNV genotyping
Genomic DNA from a total of 54 dogs were genotyped on the Illumina CanineHD BeadChips containing 220,853 markers (Neogen, Lincoln, NE, USA) (as designated in Table S1). All SNV positions reported herein correspond to the UU_Cfam_GSD_1.0/CanFam4 assembly.
GWAS
GWAS was performed with 54 samples (24 cases and 30 controls). Quality control of the SNV genotype data was performed using PLINK v.1.9 (Chang et al. 2015). SNVs with a minor allele frequency of less than 5% or more than 10% missing genotype data and individuals with a genotyping rate of less than 90% were removed. Additionally, unplaced markers with unknown chromosomal position and mtDNA markers were excluded. After pruning, 54 dogs and 156,213 markers remained in the analysis and were used for the GWAS using the linear mixed model implemented in the GEMMA software (v0.94.1). The genomic inflation factor in the analysis was 0.98, indicating that the population stratification was appropriately controlled for by the mixed model. Bonferroni correction was used to estimate the genome-wide significance threshold at p = 0.05/156,213 = 3.2 x 10− 7. Manhattan- and QQ-plots were created using the qqman package in R(Turner 2014; Team 2019). The raw SNV genotype data are available in Supplementary File S1.
Autozygosity mapping
The genotype data of 22 dogs that were homozygous for the disease-associated allele at the best associated marker were used for autozygosity mapping. The analysis was done using PLINK v.1.9 (Chang et al. 2015). Markers with missing genotypes in one of the cases were excluded. Additionally, a .tped file with all markers from chromosome 16 was created. Visual inspection of this file in an Excel spreadsheet was performed to exactly determine the shared homozygous haplotype in the 22 cases.
Whole-genome sequencing and variant filtering
A PCR-free genomic DNA library was prepared from Case #9/Dog #161 and whole-genome sequencing at 24.8x coverage was performed on an Illumina Novaseq 6000 instrument (Illumina, Zurich, Switzerland). Reads were mapped to the UU_Cfam_GSD_1.0 reference genome assembly and variant calling was performed as described in Jagannathan et al., 2019. SnpEff software (Cingolani et al. 2012) together with NCBI annotation release 106 for the UU_Cfam_GSD_1.0 genome reference assembly was used to predict the functional effects of the called variants. The sequencing data of this single affected dog was compared against 960 control genomes of different breeds to filter for private variants (Table S2).
PCR and Sanger sequencing
The candidate variant RNF170:XM_038559916.1:c.367delG was genotyped by direct Sanger sequencing of PCR amplicons. The amplification of a 368 bp (or 367 bp in case of the mutant allele) PCR product was performed using the primers 5’-TTTTTCAGCATTGGAGCAGTT-3’ (forward) and 5’-TGATGCTTTCTGGATACAAACATT-3’ (reverse) and AmpliTaqGold360MasterMix (Thermo Fisher Scientific, Waltham, MA, USA) together with additional 20% of GC enhancer (Thermo Fisher Scientific). Post amplification, the samples were treated with exonuclease I and alkaline phosphatase and subsequently sequenced with ABI BigDye v3.1. PCR amplicons were sequenced with the PCR primers on an ABI 3730 DNA Analyzer (Thermo Fisher Scientific) and the resulting Sanger sequences were analyzed using the Sequencher 5.1 software (GeneCodes, Ann Arbor, MI, USA).
Post-hoc linkage mapping
Affected dogs from both Europe and North America were all connected via one large pedigree (Figure S1). A post-hoc LOD score was calculated using dogs with pedigree information and genotypes for Chr16:23,653,869:delG in LAMP software (Li et al. 2006). Due to the complexity of this large canine pedigree, it was divided into eight smaller families to allow the program to run. The recessive genetic model option was used, and disease prevalence was set at 2%.
Estimation of age of mutation
Runs of homozygosity encompassing the identified variant in RNF170 were identified in PLINK v1.9 (Chang et al. 2015) with default parameters, and linkage map positions across the region were approximated (Wong et al. 2010). The distance from the presumed causal variant and decay of homozygosity in either direction was calculated. The variant was dated using a previously described methodology (Gandolfo et al. 2014) specifically designed for SNP array data from small datasets.