Background

In dogs, spinocerebellar ataxia (hereditary ataxia) is characterized by progressive incoordination of gait, loss of balance, hypermetric and spastic movements. This condition may be associated with changes of brainstem auditory evoked potentials, myokymia, neuromyotonia and muscle fasciculation or seizures [1–7].

Hereditary ataxia is classified according to different aspects like age of onset (neonatal, juvenile and adult) and location of the pathomorphological lesions (spinal cord and/or cerebellar cortex) [8]. In Parson (PRT) and Jack Russell Terriers (JRT) lesions were found in the spinal cord and in the brain [2]. In Fox Terriers, lesions were only present in the spinal cord [9], whereas a more recent study reported histopathological degenerative changes in the brainstem [10]. Histopathological examinations of the central nervous system in PRT and JRT demonstrated a bilateral symmetrical myelopathy, predominantly a neuronal axonopathy in the dorsal and ventral or ventromedial funiculi [1, 2]. Loss of axons and myelin and astrogliosis were primarily seen in the spinocerebellar tracts of the cervical cord but were also present in all parts of the brain [1, 2, 7].

A complex segregation analysis using regressive logistic models was performed to elucidate the mode of inheritance in PRT and JRT [2]. The data included 3 pedigrees with a total of 115 dogs, either JRT or PRT, and 27 affected with the typical clinical signs and 9/27 with a histopathological examination showing a bilateral symmetric myelopathy in the nervous system. A monogenic mode of inheritance could be clearly ruled out. Mixed models with a major gene effect and a polygenic model were most likely for these pedigrees. The analysis could not differentiate how many loci may segregate in these two dog breeds.

In JRT, PRT and related breeds, the missense mutation c.627C > G in the KCNJ10 (potassium inwardly-rectifying channel, subfamily J, member 10) gene (Transcript XM_545752.4) was found to be strongly associated with spinocerebellar ataxia (SCA) [7, 11]. This KCNJ10 variant was identified screening a whole-genome sequence of a single JRT with spinocerebellar ataxia and myokymia under the assumptions of an autosomal recessive inheritance and a mutation altering the protein structure. Data filtering included other 81 canids that had not shown SCA. Affected animals showed ataxia, myokymia and/or seizures beginning from 2–12 months of age. The spinal cord sections revealed a bilateral myelopathy. Brain magnetic resonance imaging showed no abnormalities and hearing loss was not evident [7].

A missense mutation in the CAPN1 gene (c.344G > A) was associated with a late-onset spinocerebellar ataxia phenotype (LOA) in a cohort of PRT [12]. Signs of symmetric spinocerebellar ataxia had been noticed in affected dogs between 6 and 12 months of age. The CAPN1 missense mutation c.344G > A was identified using a genome-wide association study (GWAS) followed by the massively parallel sequencing of affected and control PRT for the LOA-associated chromosomal interval.

The objectives of this study were to perform a genome-wide association study (GWAS) in PRT for hereditary ataxia using the canine Illumina high density bead chip and then to show whether the previously reported SCA- and LOA-associated KCNJ10 and CAPN1 genes in PRT and JRT are in windows of significantly associated genomic regions. In addition, we sequenced KCNJ10 and validated the previously reported KCNJ10 and CAPN1 mutations in PRT and JRT and further control samples of other breeds of dog.

Methods

We collected EDTA-blood samples from 77 PRT and 9 JRT whereof 16 PRT and 3 JRT were clinically affected by SCA (Additional file 1). All 19 cases were presented at the clinic for Small Animals, University of Veterinary Medicine Hannover. Controls did not show any signs of ataxia, myokymia or seizures based on medical records of practitioners and clinical examinations at the clinic for Small Animals, Hannover. In addition, controls had to be >4 years of age. For 17 PRT a pedigree could be constructed. A subsample of the cases for the present study had been previously described elsewhere [2]. In brief, all clinically affected dogs showed symmetric generalized ataxia with hypermetric and spastic movements in all 4 legs at 2–9 months of age. Generalized seizures were observed in some cases. In all affected dogs clinical signs worsened and had to be euthanized at an age <3.5 years. All dogs had their origin in Germany. Out of the 16 clinically SCA-affected PRT cases, 14 were confirmed through a histopathological examination. In all examined 14 PRT, a bilateral symmetrical myelopathy was evident. Predominantly an axonopathy combined with myelin loss, in the dorsal and ventral or ventromedial funiculi was seen. Swelling of axons and dilatation of myelin sheaths with loss of myelin adjacent to a mild astrogliosis were observed primarily in the spinocerebellar tracts of the cervical cord but were also observed in all parts of the brain.

Genomic DNA was isolated using standard methods with RBC (Red Blood Cell) lysis buffer and SE (sodium EDTA) buffer. The DNA concentration of the samples was adjusted to 50 ng/μl using the Nanodrop ND-1000 (Peqlab Biotechnology, Erlangen, Germany) and quality control was performed by gel electrophoresis using 1 % agarose gels (peqGold Universal Agarose, Peqlab Biotechnologie).

We performed a GWAS in 44 PRT using the canine Illumina high density beadchip (Illumina, San Diego, CA, USA) containing 173,662 single nucleotide polymorphisms (SNPs). The data set included 15 cases and 29 controls. Out of these 15 cases, 12 had a histopathological examination confirming the clinical diagnosis. Controls had to be unrelated with the cases at the parent level and >4 years of age. Among controls, 11 were male and 18 female. Cases included 7 males and 8 females. After quality control (genotyping rate per SNP and animal >0.90), filtering for minor allele frequency >0.05 and Hardy-Weinberg equilibrium (HWE) (p < 0.000001) using SAS/Genetics, version 9.4 (Statistical Analysis System, SAS-Institute, Cary, NE, USA), 128,863 SNPs were left for the analysis.

The GWAS was performed with a general linear model. The analysis was run using TASSEL, version 3.0.164 [13]. The GLM explained for sex effects and the SNP genotypes. In addition, we tested several extended models employing up to five principal components (PCAs) to show the robustness of the outcome of the GWAS. All these models yielded the same genome-wide associated SNPs as the GLM model. Thus, adding principal components for cryptic data structure did not change the results of the GLM model. A plot of the first two PCAs for cases and controls employed for the GWAS is shown in Additional file 2. We applied the Bonferroni correction to determine the threshold for genome-wide significance in order to avoid false-positive associations. The Bonferroni correction set the threshold for significance at 0.05/128,863 = 3.88e−7 for a P-value of 0.05. A sex-stratified case-control analysis was performed using SAS/Genetics, version 9.4 (SAS). Homozygosity mapping was done in sliding windows with a minimum of 4 consecutive homozygous SNPs within a range of 50 kb using PLINK, version 1.07 [14].

A polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) for a sample of 77 PRT, 9 JRT from Germany and 179 controls from 20 different dog breeds was employed for genotyping the CAPN1:c.344G > A and KCNJ10:c.627C > G mutations. The design of primers was carried out using Primer 3 (http://bioinfo.ut.ee/primer3-0.4.0/) and the dog genome assembly CanFam3.1 (Additional file 3).

The gene model we used for KCNJ10 was based on the transcript XM_005640901.1 of the dog genome assembly CanFam3.1 (Additional file 4). This transcript includes three exons whereof only two exons (exon 2 and 3) are translated into a protein with 414 amino acids (aa). A previous study [7] used the transcript XM_545752.4 as gene model (denoted XM_545752.3 in previous reports [7, 11]) with two exons encoding 379 aa. Exon 2 and 3 of KCNJ10 and their adjacent genomic regions were sequenced in 10 PRT (cases PRT I-III and PRT IV-VIII and controls PRT IX and PRT X) and 3 JRT (cases JRT I-III) using an ABI 3500 Genetic Analyzer (Applied Biosystems by Life Technologies, Darmstadt, Germany) (Additional file 3). In total, 45 genetic variants were detected in these 13 dogs. Exon 1 could not be amplified because several prime pairs designed for this canine genomic region or primer pairs designed from other species including human, mouse and rat did not result in amplicons with the expected size nor the amplicons could be sequenced. The exon 1 region did not seem to be correctly annotated in the dog reference genome and thus, may prevent retrieval of possible genetic variants for KCNJ10 exon 1 in whole genome sequence data.

In addition, we tried to amplify the gene KCNJ9 (gene ID 100855823) annotated at 22,100,627–22,104,165 base pairs (bp) on dog chromosome (CFA) 38 (CanFam3.1). The record of this gene contained large gaps. We searched for canine expressed sequence tags (ESTs) using the canine coding sequence of KCNJ9 and the basic local alignment tool (blast) of NCBI (https://blast.ncbi.nlm.nih.gov/Blast.cgi). The best hit with an E-value of 1e−77 and a total score of 293 (DN374672.1) of this search did not fit to the genomic region where KCNJ9 was annotated in the dog genome reference assembly. The same was found for hits with larger E-values. Therefore, we were not able to build a gene model for KCNJ9 or sequence this gene.

All KCNJ10 genetic variants detected in the 10 PRT and 3 JRT were further evaluated by sequencing KCNJ10 from one dog of 9 other breeds. These dogs had a normal phenotype in regard to SCA and included the following breeds: Akita, Bernese mountain dog, Dalmatian, German Drahthaar, German shepherd, Irish wolfhound, Shar Pei, Tibetan terrier and Briard. Analysis of sequencing results had been performed using Sequencher software, version 4.8 (Gene Codes, Ann Arbor, MI, USA). The evaluation study resulted in 16 private variants for PRT and JRT. We then filtered the 16 genetic variants for homozygous mutant genotypes in affected PRT and JRT. The single nucleotide variant (SNV) reported in a previous study [7] and only one other variant (KCNJ10:g.22141027insC) remained. Validation of this result was performed in further 17 PRT. We developed a test on a LI-COR 4300 DNA Analyzer (LI-COR Biosciences, Lincoln, NE, USA) using a 167-bp amplicon (Additional file 3). The forward primer was labelled with IRD700 at the 5′end. All samples of the 77 PRT and 9 JRT as well as 84 dogs from 40 other different breeds were used to generate the KCNJ10:g.22141027insC containing amplicon. The amplicons were size-fractionated using gel electrophoresis on an automated sequencer LI-COR 4300 DNA Analyzer on 6 % polyacrylamide denaturing gels. Allele sizes were scored against IRD700-labeled DNA ladders (Additional file 5). For bioinformatic analysis of the effects of the KCNJ10:g.22141027insC variant we used the complete fasta-sequence of KCNJ10 and employed the Regulatory RNA Motifs and Elements Finder (RegRNA) (http://regrna.mbc.nctu.edu.tw/html/about.html) [15] and the variant effect predictor (http://www.ensembl.org/Homo_sapiens/Tools/VEP?db=core).

Results

Genome-wide association study

The GWAS using 15 cases and 29 controls identified three genome-wide significantly associated SNPs on dog chromosome 38 (CFA38) (Additional file 6) and a quantile-quantile plot (Q-Q plot) for expected versus observed –log10 P-values was calculated to control for population stratification (Additional file 7). All SNPs with –log10P-values >5 are located on CFA38 and were in HWE. The genomic inflation factor was equal to 1, indicative that data stratification had no effect on the outcomes of the GWAS. The significantly associated SNPs are located at 18.53, 25.21 and 25.31 Mb (CanFam2) and at 15.53, 22.19 and 22.30 Mb (CanFam3.1) on CFA38 with the strongest signal at 22.30 Mb (Praw = 5.14×10−9). The P-values for the SNPs with the highest associations after Bonferroni-correction were at 0.031 (BICF2G63072533), 0.036 (BICF2P378938) and 6.6×10−4 (BICF2P478530) (Additional file 8). In the associated interval, 9/15 cases and one control shared a homozygous region over 147 consecutive SNPs spanning 1893 kb at 23,387,952–25,280,713 bp (CanFam2) and at 20,388,358–22,267,778 (CanFam3.1). Inclusion of further two cases resulted in an even shorter homozygous interval with a size of 124 kb. However, for all other cases, an extended homozygosity region could not be determined (Additional file 9). This may indicate that hereditary ataxia had its origin in several founders or recombinations broke down the homozygous identical-by-descent region caused by a common founder to a very small region not detectable via the beadchip data.

A potential candidate gene in the associated and homozygous interval is KCNJ10 at 25,148,668–25,149,816 bp (CanFam2) and at 22,114,718–22,143,618 bp (CanFam3.1) [7, 11]. The two closest and significantly associated SNPs are 58,212 (BICF2P378938) and 149,816 (BICF2P478530) bp downstream to KCNJ10. An association was not evident for SNPs on CFA18 where CAPN1 is located. The gene KCNJ9 could not be verified using expressed sequence tags (ESTs) for searching this region. Other candidate genes could not be detected in the associated interval. Further members of the KCNJ-gene family were not considered because the GWAS did not indicate significant signals on locations outside CFA38.

Genotyping CAPN1 and KCNJ10 variants

CAPN1 genotyping results showed that all 77 PRT and 9 JRT were homozygous wild-type. Genotyping the KCNJ10:c.627C > G mutation revealed 38/77 PRT (49.4 %) with the wild-type, 22/77 (28.6 %) as heterozygous and 17/77 (22.1 %) as homozygous mutated. In 1/38 of the homozygous wild-type tested PRT, hereditary ataxia was clinically and histopathologically confirmed. Two other homozygous wild-type tested PRT were diagnosed to be affected after clinical examination and/or according to the report of the owner. Among JRT, only 1/9 dogs was homozygously affected and 7/9 heterozygous. Each one of the heterozygous and homozygous wild-type JRT were clinically affected (Table 1). In addition, 179 controls from 20 different dog breeds were genotyped for both mutations. All these dogs were homozygous wild-type for both mutations (Additional file 10).

Table 1 Distribution of ataxia phenotypes by the KCNJ10:c.627C > G genotypes for Parson Russell (PRT) und Jack Russell Terriers (JRT)
Full size table

Sequencing KCNJ10

In order to identify further mutations associated with hereditary ataxia explaining the cases with the homozygous KCNJ10 C/C genotype, exon 2 and 3 of KCNJ10 with their exon/intron or exon/UTR boundaries were sequenced in PRT and JRT. We included 3 affected PRT with the homozygous KCNJ10 C/C genotype and all 3 affected JRT with the different KCNJ10 genotypes. In addition, we sequenced 7 PRT (5 affected PRT and 2 unaffected) with the concordant KCNJ10 genotype. A total of 45 genetic variants could be identified including 41 single nucleotide variants (SNVs) and 4 indels (3 insertions and 1 deletion) (Additional file 11) whereof most of them were located in the 3′UTR (CanFam3.1) (Additional file 4). Sequencing further 9 dogs of other breeds (Additional file 12) and filtering showed 16 variants private for PRT and JRT (Additional file 13). A further filtering for homozygous mutant genotypes in affected PRT and JRT resulted in a 1-bp-insertion (KCNJ10:g.22141027insC) prevalent in these animals. This 1-bp-insertion was located within a stretch of 7 consecutive C bases (Additional file 5). Bioinformatic analysis using RegRNA indicated the KCNJ10:g.22141027insC affecting regulation of gene expression via a regulatory RNA motif or miRNA target site. A regulatory RNA motif (5′-GGGAGGGG-3′) was identified 6 bp downstream to the KCNJ10:g.22141027insC. The motif of the miRNA (hsa-miR-1180) target site contains the KCNJ10:g.22141027insC (5′-acACAACCCCCCCGACGCCGGGAg-3′). Genotyping a pedigree containing 17 PRT for both mutations, KCNJ10:g.22141027insC and KCNJ10:c.627C > G, showed segregation of both mutations with the affection status with exception of one animal (Additional file 14). Genotyping results of the KCNJ10:g.22141027insC are summarized in Table 2. Out of the 77 PRT, 34 (44.2 %) showed the wild-type, 26 (33.8 %) the heterozygous and 17 (22.1 %) the homozygous mutated genotype. In 2/26 of the heterozygous PRT, hereditary ataxia was clinically and histopathologically confirmed and 2/34 homozygous wild-type tested PRT were affected according to a clinical examination and/or according to the report of the owner. The genotyping results for the KCNJ10:g.22141027insC mutation of JRT showed two homozygous mutated genotypes among the affected individuals. A further affected individual had the wildtype genotype. Genotyping further 84 dogs of 40 other breed dogs revealed 59 (70.3 %) wildtype homozygous, 21 (25.0 %) heterozygous and 4 (4.8 %) homozygous mutated genotypes (Additional file 15).

Table 2 Distribution of ataxia phenotypes by the KCNJ10:g.22141027insC genotypes for Parson Russell (PRT) und Jack Russell Terriers (JRT)
Full size table

Discussion

The results of the GWAS indicate that the associated genomic region on CFA38 may harbour the most likely candidate genes for hereditary ataxia in the sample under study. Thus, our study corroborated the previous candidate gene search using whole genome sequence data [7]. We could not detect samples with the CAPN1:c.344G > A mutation. This may indicate that this mutation may be rare in the German PRT and JRT populations. In agreement with a previous study [7], one histopathologically confirmed case of hereditary ataxia and two clinically affected cases carried neither the CAPN1:c.344G > A allele nor the homozygous KCNJ10:c.627C > G genotype. This may indicate that at least one further mutation may be involved in hereditary ataxia of PRT and JRT. As the largest proportion of affected dogs were concordant with the segregation of the KCNJ10:c.627C > G alleles the remaining unexplained cases may harbour a rare mutation. However, breeders have to be aware that selective breeding based on genetic tests for the KCNJ10:c.627C > G and CAPN1:c.344G > A mutations cannot completely preclude affected dogs of the breeds PRT or JRT despite of homozygous wild-type genotypes of the parents.

The present sequence analysis tagged a 1-bp-insertion associated with hereditary ataxia in PRT and JRT. This mutation may have an effect on regulation of KCNJ10 expression as indicated through bioinformatic results from RegRNA. However, the presence of this 1-bp-insertion in other breeds may require further research in phenotyping spinocerebellar ataxia associated with myokymia, seizures, or both. In conclusion, a few but not all cases of hereditary ataxia being heterozygous or homozygous for the wild-type allele of KCNJ10 could be explained by the KCNJ10:g.22141027insC mutation.

Conclusions

Our data confirmed the presence of the KCNJ10:c.627C > G mutation in PRT and JRT affected with hereditary ataxia. However, some cases of hereditary ataxia could not be explained through the KCNJ10:c.627C > G nor the CAPN1:c.344G > A mutation. Genetic testing using the KCNJ10:c.627C > G and the CAPN1:c.344G > A mutations deems not to be sufficient to capture all cases of hereditary ataxia in PRT and JRT. This lets us assume that at least one additional form of hereditary ataxia may segregate in the breeds PRT and JRT. The KCNJ10:g.22141027insC mutation was evident in a few cases but could not explain all not yet resolved cases. Further efforts are necessary to collect cases of hereditary ataxia not explained by the presently known mutations to unravel the responsible mutations in these cases of hereditary ataxia.