Introduction

Canine hip dysplasia (CHD) is a common orthopedic trait in all dog breeds. CHD causes instability and subluxation of the hip with secondary signs of osteoarthritis and clinical signs of lameness. Breed occurrence varies widely from 1 to 75 % (Janutta and Distl 2006). In German shepherd dogs, prevalence of CHD is estimated at 35 % (Janutta et al. 2008). Assessment of CHD is based on radiography both of hip joints. Scoring systems have been developed to achieve precise diagnoses across all dog breeds from different countries in Europe. There is strong evidence in support of a genetic predisposition to CHD in German shepherd dogs and many other dog breeds. Heritability estimates for German shepherd dogs from different European countries were at h 2 = 0.20–0.35 (Hamann et al. 2003; Janutta et al. 2008; Stock et al. 2011). Recently, involvement of a major gene was demonstrated for the German shepherd dog using mixed inheritance models for statistical analyses (Mäki et al. 2004; Janutta et al. 2006). Genome-wide linkage studies in eleven paternal half-sib groups of German shepherd dogs identified 21 quantitative trait loci (QTL) and nine genome-wide QTL for CHD (Marschall and Distl 2007). A QTL study in a Labrador retriever Greyhound crossbred family revealed twelve chromosomes with chromosome-wide significant markers for CHD (Todhunter et al. 2005). In Portuguese water dogs, QTL for signs of CHD were demonstrated on chromosomes 1 and 3 (Chase et al. 2004, 2005). A genome-wide association study for CHD and osteoarthritis (OA) in 721 dogs from several breeds and Labrador Retriever-Greyhound crosses identified four CHD-associated and two OA-associated SNPs. The CHD-associated SNPs were located on dog chromosomes (CFA) 3, 11, and 30 (Zhou et al. 2010).

We conducted a two-stage association study to identify SNPs located within CHD-QTL and with a significant contribution to the phenotypic expression of CHD in German shepherd dogs. For this purpose, we sequenced approximately 200 kb within 17 QTL in each of 24 to 48 German shepherd dogs and searched these sequences for polymorphisms, particularly for single nucleotide polymorphisms (SNPs). We have chosen the German population of German shepherd dogs as this population is well suited for an association study, since this breed represents the largest purebred dog population in Europe with a large phenotypic and genetic variance for CHD and a consistent recording system of CHD including the collection of EDTA-blood samples from dogs when the radiographic examination is performed. Validation of CHD-associated SNPs was performed in a random sample of 843 dogs, representative for the German shepherd dog population.

Materials and Methods

Animals

A total of 843 German shepherd dogs had been included in the present association study for CHD. Cases and controls with equal proportion of sexes and birth cohorts from the years 2000–2005 were selected from the whole German breeding population of German shepherd dogs including more than 12,000 individuals. Radiographs and CHD-scores according to the official guidelines of the Fédération Cynologique Internationale (FCI) as well as EDTA-blood samples were available for all dogs. Blood samples were collected for parentage testing by the German Association for German shepherd dogs (SV, Augsburg, Germany). For all dogs included in the present study, parentage has been confirmed using an approved set of microsatellite for parentage control. The radiographs were made by veterinarians officially approved by the association of radiographic diagnostic of genetic influenced skeletal diseases of small animals (GRSK) and the German Association for Dog Breeding and Husbandry (VDH) strictly following the requirements for radiographs of hip joints as stated by Brass (1993) and the FCI. All X-rays were evaluated by a recognized veterinary expert and a subsample of 200 X-rays was re-evaluated by further two experts. Consistency of diagnoses was >99%. These officially recorded CHD-grades were supplied by the SV and used in all subsequent analyses. CHD-A includes dogs with normal hips, CHD-C dogs with slight signs of CHD, CHD-D dogs with moderate signs of CHD and CHD-E dogs with severe signs of CHD using X-rays from both hip joints. Dogs with mild to severe signs of CHD were defined as cases. Controls were dogs free from any signs of CHD. Cases and controls were matched by sex, birth year, and coancestry (Table 1). The samples were designed in such a way that dogs of both sexes and the birth years 2000–2005 from all over Germany were represented in the data set and coancestry among cases and controls as well as within cases and controls was minimized. Using the available pedigree records, mean relationship coefficients among all dogs, cases, and controls were lower than 0.1 %. Relationship coefficients were calculated for all registered German shepherd dogs considering all available pedigree information, with a maximum of eleven ancestral generations. For this purpose PEDIG software (Boichard 2002) was used. These relationship coefficients were employed to select out of a group of 12,096 dogs those dogs fitting to our study design and with relationship coefficients as low as possible. The final data set included 843 German shepherd dogs with 405 males and 438 females (Table 1). Out of these 834 dogs, 277/834 dogs were CHD-A and 557/834 with CHD-C, CHD-D or CHD-E. Dogs scored as CHD-A were treated as unaffected (n = 277), dogs scored as CHD-C (n = 349), CHD-D (n = 152), and CHD-E (n = 65) as affected. The animals were born between 2000 and 2005 and purebred following the rules of the SV. The study population was a sample stratified by coancestry from all registered and X-rayed dogs of the German population of German shepherd dogs born between 2000 and 2005. In order to distinguish clearly between CHD-affected and CHD-unaffected dogs, dogs with the diagnosis of a borderline for CHD were not used in the present study.

Table 1 Distribution of canine hip dysplasia status (CHD) by sex and birth year in the German shepherd dogs employed for the association study
Full size table

SNP Detection

According to the results of a whole genome scan using 261 microsatellites genotyped for 459 purebred German shepherd dogs from eleven paternal half-sib families (Marschall and Distl 2007), 129 genes on 14 canine chromosomes within previously identified QTL significant after Bonferroni correction were selected from the dog genome assembly 2.1 (http://www.ncbi.nlm.nih.gov/mapview/map_search.cgi?taxid=9615) for SNP detection (Table S1). Particularly, peak regions of these 14 genome-wide significant QTL of the previous study by Marschall and Distl (2007) were selected. In case that in these QTL-regions polymorphic sites for German shepherd dogs were already known from the SNP collection of the Broad Institute (http://www.broadinstitute.org/mammals/dog/snp2), we used these SNPs. At least two amplicons of approximately 500 base pairs in size were re-sequenced per gene (Table S2). We genotyped 312 PCR products located within these QTL and performed sequence analysis in a set of 24 German shepherd dogs with equal proportions of cases and controls from both sexes. Sequencing in further 24 German shepherd dogs was done if a polymorphism was detected. These dogs were a random sample and not related with each other according to the pedigree records up to eight generations. This detection sample was an independently collected data set and not identical with the validation sample. Sequence analyses revealed a total of 177 SNPs. From these 177 SNPs, 73 SNPs were between breed polymorphisms among the boxer sequence of the dog genome assembly 2.1 (http://www.ncbi.nlm.nih.gov/mapview/map_search.cgi?taxid=9615) and the German shepherd dogs investigated here. Further 67 SNPs were polymorphic in German shepherd dogs but not informative for the CHD status due to a low minor allele frequency (<0.05). In 218 PCR products no SNPs were found, and in the 94 other PCR products the number of identified SNPs ranged between one and nine SNPs. SNPs found were located within 65 genes. A set of 37 SNPs was polymorphic in German shepherd dogs as well in CHD-affected and unaffected German shepherd dogs.

DNA was isolated from EDTA-blood samples using the NucleoSpin Kit 96 Blood Quick Pure Kit (Macherey-Nagel, Düren, Germany). Primer pairs were designed using the software RepeatMasker (http://www.repeatmasker.org/cgi-bin/WEBRepeatMasker) and Primer3 (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi). These primer pairs yielded PCR products within a size range of 173 to 682 base pairs. PCR amplification was done using PTC 100TM, PTC 200TM (MJ Research, Watertown, MA, USA) or Biometra TProfessional thermocyclers (Biometra, Göttingen, Germany) with standard PCR programmes. The PCRs were performed in a total volume of 30 μl using 10 ng of genomic DNA as template, 3 μl 10× incubation buffer containing 15 mM MgCl2, 1 μl DMSO, 1 μl of the forward and reverse primer (10 pmol/μl), 1 μl dNTPs (100 μM each) and 0.2 μl Taq (5 U/μl) Polymerase (Qbiogene, Heidelberg, Germany). After cleaning of the PCR products with MinElute® 96 UF Plate (Qiagen, Hilden, Germany), they were sequenced in 7.5 μl volumes using 3.5 μl of the PCR product, 1 μl primer (forward or reverse, 10 pmol/μl) and 1.5 μl DYEnamic-ET-Terminator Cycle Sequencing Kit (GE Healthcare, Freiburg, Germany) with a standard programme. The reaction product was cleaned up using a SephadexTM G-50 Superfine filtration (GE Healthcare). The amplicons were sequenced on a MegaBACE 1000 automated sequencer (GE Healthcare). The forward and reverse sequences were searched for SNPs by visual examination using the Sequencher 4.7 program (GeneCodes, Ann Arbor, MI, USA) and the reference sequence of the dog genome assembly 2.1 (http://www.ncbi.nlm.nih.gov/mapview/map_search.cgi?taxid=9615).

Genotyping

For genotyping, 37 SNPs distributed in 30 different genes and seven intergenic regions within CHD-regions on 13 chromosomes were used for the association study (Table S3). Genotyping was performed via restriction fragment length polymorphisms (RFLPs) or when no RFLP was available using custom TaqMan® SNP genotyping assays (Applied Biosystems, Darmstadt, Germany). Also for the extended validation set of additional 14 SNPs, custom TaqMan® SNP genotyping assays (Applied Biosystems) were employed. For RFLPs, the amplification of the PCR products containing the SNPs were performed as described above for the search of polymorphisms. RFLPs were done in a 20 μl volume using 4 μl buffer, possibly 0.4 μl bovine serum albumin (BSA) dependent on the used endonucleases and 3 U endonuclease with 10 μl of the PCR product. The SNP genotypes were determined by gel electrophoresis using 2 % agarose gels and evaluated by visual examination. The genotyping custom TaqMan assays were analyzed on a 7300 Real Time PCR System (Applied Biosystems) in 12 μl volume using 5.3 μl SensiMix DNA kit (Quantance, London, UK), 0.3 μl custom TaqMan® SNP genotyping assays (Applied Biosystems) and a DNA template of 10 ng. After a 10 min initial denaturation at 95 °C, 40 cycles of 15 s at 92 °C and 60 s at 60 °C were used. In the first step genotyping was performed for 190 German shepherd dogs. If a SNP was found to be associated with the CHD status in this sample at a P value < 0.05 for the genotype and allele test statistic, all 843 dogs from the validation sample were genotyped.

Genotyping of the final SNP set consisting of ten SNPs was performed using the ABI Prism® SNaPshotTM Multiplex System (Life Technologies by Applied Biosystems, Darmstadt, Germany). Single base extension (SBE) primers were designed using online primer design software BatchPrimer3 (http://probes.pw.usda.gov/cgi-bin/batchprimer3/batchprimer3.cgi). According to the manufacturer’s instructions, the primer sequences should differ in length by at least four to six nucleotides and should not contain possible extendable hairpin structures. All primers used here had undergone HPLC purification. The primer design was in this way that a difference of seven nucleotides between the SBE primers was achieved by adding non-homologous polynucleotides (poly (dNTP)) at the 5′ end. In order to configurate the SBE primer set more efficiently we designed two primers with opposed motifs at one length (Table S4). The SBE detection was performed using an ABI Genetic Analyzer 3500 (Life Technologies by Applied Biosystems). Data evaluation was done using GeneMapper, version 4.2 (Life Technologies by Applied Biosystems).

Statistical Analysis

Allele and genotype frequencies of the SNPs were calculated for the different CHD-grades as well as the observed heterozygosity (H 0), the polymorphism information content (PIC) and Hardy-Weinberg equilibrium were estimated using the ALLELE procedure of SAS/Genetics (Statistical Analysis System, version 9.3, SAS Institute, Cary, NC, USA, 2013). Association analyses were performed using odds ratios with their 95% confidence limits and χ 2-tests of the CASECONTROL procedure of SAS/Genetics for genotypic and allelic associations and the allelic trend with the CHD status (CHD-A vs. CHD-C/D/E) and with CHD-A versus CHD-C, CHD-D, or CHD-E. These four definitions of phenotypes were used for all analyses. Testing for the three different CHD grades was done to see whether the association is consistent within each grade or may be caused by a specific CHD grade. A marker was regarded as significantly associated for −log10  P values >2.87 (0.00135 = 0.05/37). We chose this threshold because only 37 SNPs were informative for the panel of dogs genotyped here. We discarded 67/104 SNPs due to their low minor allele frequency. A general linear model including all CHD-associated SNP genotypes was used to estimate the proportion of the phenotypic variance explained by these SNPs for all CHD grades and each of the different CHD grades separately. Calculations were performed using SAS, version 9.3.

Results and Discussion

In the first stage of the association study for German shepherd dogs, we studied 37 SNPs in 95 CHD cases and 95 controls (Table S3). These SNPs were located on 13 dog chromosomes (CFA) including CFA1, 3, 4, 8, 9, 16, 19, 26, 29, 32, 33, 34, and 35. A total of ten SNPs were significantly associated at a nominal P value <0.05 with CHD and these ten SNPs on CFA3, 9, 26, 33, and 34 were evaluated in all 843 German shepherd dogs. Four SNPs were located on CFA34, three SNPs on CFA33 and each one SNP on CFA3, 9 and 26 (Table S4). The minor allele frequencies (MAF) ranged from 0.09 to 0.48 (Table 2).

Table 2 Distribution of the minor allele frequencies of the ten SNPs genotyped in German shepherd dogs including 277 unaffected dogs (CHD-grade A) and 566 affected dogs with 349 dogs classified as CHD-grade C, 152 dogs as CHD-grade D and 65 dogs as CHD-grade E
Full size table

Out of these ten SNPs, eight SNPs were located within the genes PGM2 on CFA3, EPHA3, EPHA6, and PCNP on CFA33, TRIO, SEMA5A, SLC6A3, and FGF12 on CFA34. Each one SNP was intergenic on CFA9 and 26. All SNPs but the intragenic SEMA5A SNP on CFA34 were significantly associated with CHD (Table 3). The CHD-QTL on CFA3, 9, 26, 33, and 34 were confirmed in all analyses when comparing all CHD-affected dogs with CHD-free dogs and dogs with the different grades of CHD or combinations of different grades (Tables 4, 5, 6). A combined analysis for dogs free from CHD and dogs with CHD-D and CHD-E gave slightly higher −log10  P values compared to the analysis with CHD-A versus CHD-D dogs (data not shown).

Table 3 Odds ratios (OR) of the ten SNPs genotyped with their 95 % lower (CL) and upper confidence limits (CU), χ 2- and −log10  P values for genotypic, allelic and trend association considering 277 CHD-A German shepherd dogs as controls and 566 German shepherd dogs classified as CHD-C/D/E as cases
Full size table
Table 4 Odds ratios (OR) of the ten SNPs genotyped with their 95 % lower (CL) and upper confidence limits (CU), χ 2-values and −log10  P values for genotypic, allelic and trend association considering 277 CHD-A German shepherd dogs as controls and 349 German shepherd dogs classified as CHD-C as cases
Full size table
Table 5 Odds ratios (OR) of the ten SNPs genotyped with their 95 % lower (CL) and upper confidence limits (CU), χ 2- and −log10  P values for genotypic, allelic and trend association considering 277 CHD-A German shepherd dogs as controls and 152 German shepherd dogs classified as CHD-D as cases
Full size table
Table 6 Odds ratios (OR) of the ten SNPs genotyped with their 95 % lower (CL) and upper confidence limits (CU), χ 2- and −log10  P values for genotypic, allelic and trend association considering 277 CHD-A German shepherd dogs as controls and 65 German shepherd dogs classified as CHD-E as cases
Full size table

The strongest associations were found for the SNPs on CFA33 and 34. In all analyses considering CHD-free German shepherd dogs versus mildly to severely CHD-affected dogs (C + D + E or C or D or E or D + E), the TRIO-, PCNP-, and SLC6A3-associated SNPs reached −log10  P values at 9–23. Odds ratios (ORs) were highest for the TRIO- and SLC6A3-associated SNPs with values at 0.12–0.17 and 5.8–8.6, respectively.

A further validation was performed for a set of 14 SNPs flanking the highly associated SNPs on CFA26, 33, and 34 (Table S5). The associations for this extended validation set of SNPs were significant at experimental P values and supported the genomic regions identified in the validation sample. Particularly, the associations were strong for the SNPs flanking the genes PCNP on CFA33 and TRIO on CFA34. Thus, the extended set of SNPs strongly supported these two genes as being involved in CHD in German shepherd dogs.

On CFA33, the PCNP-associated SNP had the strongest associations with CHD. The EPHA3- and EPHA6-associated SNPs were in low linkage disequilibrium (r 2 = 0.11 and 0.18) with the PCNP-associated SNP. Thus, several CHD-loci may be located on CFA33. However, a multiple analysis of variance for the three SNPs on CFA33 showed that only the PCNP-SNP was significantly associated. This may indicate that the genotype effects of the SNPs within EPHA3 and EPHA6 are overlapping with the genotype effects of the PCNP-SNP. The strong associations of the TRIO- and SLC6A3-associated SNPs, their distance of approximately eleven Mb on CFA34 and their low linkage disequilibrium (r 2 = 0.23) suggest that these two SNPs may represent two different CHD-loci on this chromosome. In a multiple analysis of variance for the four SNPs on CFA34, the TRIO- and SLC6A3-SNPs remained significantly associated with CHD, whereas the SEMA5A- and FGF12-SNPs were no longer significant. In addition, the SEMA5A- and FGF12-SNPs on CFA34 were in very low linkage disequilibrium (r 2 < 0.04) with the TRIO- and SLC6A3-associated SNPs. Therefore, we assume that at least two different CHD-loci may be located on CFA34.

The proportion of phenotypic variance of CHD explained by the cumulative effects of the nine CHD-SNPs was 22.4 % (all CHD-grades), 24.0 % (CHD-A vs CHD-C), 31.1 % (CHD-A vs CHD-D), and 34.2 % (CHD-A vs. CHD-E).

Several CHD-associated SNPs on CFA3, 11, and 30 were reported from an across-breed association study including a locus on CFA3 at 74.72 Mb (Zhou et al. 2010). This CHD-locus on CFA3 may be different from the locus we identified here. We also assume that the CHD-locus on CFA3 for German shepherd dogs is located more distally on CFA3 because the linkage peak region was at 89–92 Mb. However, we could not find an informative SNP for German shepherd dogs in this region. On CFA11 and 30, we did not look for informative SNPs because on these chromosomes QTL were not detected for German shepherd dogs (Marschall and Distl 2007). Therefore, the loci identified here had not been found in another linkage or association study. If these loci are specific for German shepherd dogs remains unclear as only a few dog breeds had been investigated for CHD-loci.

The triple functional domain (PTPRF interacting) (TRIO)-associated SNP g.414871G>A showed the strongest associations with CHD across and within the different CHD scores. The protein encoded by TRIO contains three functional domains, a serine/threonine kinase domain and two guanine nucleotide exchange factor (GEF) domains for the family of Rho-like GTPases, specific for Rac1 and RhoA (Debant et al. 1996). TRIO is a unique Rho GEF, because it has two separate GEF domains, GEFD1 and GEFD2, controlling the GTPases RhoG/Rac1 and RhoA, respectively (Bellanger et al. 2000). Rac1 and RhoA act as antagonists, both playing an important role in chondrogenic proliferation and differentiation (Wang and Beier 2005). Chondrocyte-specific deletion of Rac1 in mice leads to dwarfism due to reduced chondrocyte proliferation (Wang et al. 2007). Inhibition of Rac1 expression in micromass culture resulted in reduced mRNA levels of the chondrogenic markers collagen II and aggrecan, and decreased accumulation of glycosaminoglycans indicating that Rac1 promotes chondrogenesis (Woods et al. 2007). Rac1-deficient chondrocytes had severely reduced levels of inducible nitric oxide synthase protein (iNOS) and nitric oxide production (Wang et al. 2011). Mice deficient for iNOS had reduced chondrocyte proliferation and resembled the phenotype of Rac1-deficient growth plates. RhoA overexpression in chondrogenic ATDC5 cells resulted in increased proliferation and a marked delay of hypertrophic differentiation, whereas inhibition of Rho/ROCK signaling inhibited chondrocyte proliferation and accelerates hypertrophic differentiation (Wang et al. 2004). Therefore, changing the balance between the GTPases RHOA and RAC1 due to mutations in TRIO will lead to disturbances in cartilage development.

For all candidate genes on CFA3, 9, 26, 33, and 34, we could find a joint network indicating that the loci identified harbour genes contributing to CHD (Table S6).

In conclusion, the results of this study confirmed CHD-loci on five different chromosomes for German shepherd dogs and are a further step towards clarification of the genes underlying CHD. The strongest associated SNP is located in close proximity to a functional candidate gene being involved in regulating bone formation, osteoclast activity, chondrocyte proliferation, and differentiation. A joint network could be shown using the candidate genes of all these loci. In further studies, these CHD-regions should be sequenced in a larger number of cases and controls using next generation sequencing in order to clarify the possible role of genetic variants for CHD.