Genome-Wide Analysis Identifies Germ-Line Risk Factors Associated with Canine Mammary Tumours

Abstract

Canine mammary tumours (CMT) are the most common neoplasia in unspayed female dogs. CMTs are suitable naturally occurring models for human breast cancer and share many characteristics, indicating that the genetic causes could also be shared. We have performed a genome-wide association study (GWAS) in English Springer Spaniel dogs and identified a genome-wide significant locus on chromosome 11 (praw = 5.6x10-7, pperm = 0.019). The most associated haplotype spans a 446 kb region overlapping the CDK5RAP2 gene. The CDK5RAP2 protein has a function in cell cycle regulation and could potentially have an impact on response to chemotherapy treatment. Two additional loci, both on chromosome 27, were nominally associated (praw = 1.97x10-5 and praw = 8.30x10-6). The three loci explain 28.1±10.0% of the phenotypic variation seen in the cohort, whereas the top ten associated regions account for 38.2±10.8% of the risk. Furthermore, the ten GWAS loci and regions with reduced genetic variability are significantly enriched for snoRNAs and tumour-associated antigen genes, suggesting a role for these genes in CMT development. We have identified several candidate genes associated with canine mammary tumours, including CDK5RAP2. Our findings enable further comparative studies to investigate the genes and pathways in human breast cancer patients.

Fig 1. Genome-wide association results.

(A) Quantile-quantile plot displaying a lambda of 1.00, indicating no residual inflation. Thin lines indicate 95% CI. SNPs with -log₁₀(p) values > 4 deviates from the expected distribution and are associated with CMT. (B) Manhattan plot displaying the results from the GWAS based on the Swedish ESS Illumina 170K genotypes. Genome-wide significance is reached for one SNP on chromosome 11 (73,290,522 bp) and nominal association is reached for seven SNPs on chromosomes 11 and 27.

Fig 2. Association results for chromosome 11.

(A) Association plot and (B) Minor allele frequency plot for chromosome 11. (C) Candidate region with association results colour-coded according to pair-wise LD (r²) with the top SNP (index). (D) Candidate region showing association results for the merged and imputed GWAS and sequence candidate SNP dataset with colours according to pair-wise LD (r²) with the top SNP. The top haplotype spans a region containing CDK5RAP2, MEGF9 and potentially also MIR147A and LINC01613. Black arrows indicate direction of transcription and red arrow the top SNP position.

Fig 3. Haplotypes in the chromosome 11 candidate region.

(A) Phylogenetic tree displaying haplotype relationship of 15 SNPs in the candidate region on chromosome 11. The 51 haplotypes can be formed into three groups based on the tree clusters (separated by dashed lines). (B) Case/control frequencies in the three haplotype groups in the ESS cohort. There is a lower proportion of cases in haplotype group 1 compared to group 3.

Fig 4. Association results for chromosome 27.

(A) Chromosome 27 with association results colour-coded according to pair-wise LD (r²) with the top two SNPs (index 1 = chr27:745,156 bp and 2 = chr27:7,706,463 bp). (B) Minor allele frequency plot over chromosome 27. (C) Association results for the merged and imputed GWAS and sequence candidate SNPs dataset at the 0.7 Mb peak. The top SNP (chr27:735,281 bp) is located 418 bp upstream of LACRT, SNP position indicated by red arrow. Gene annotations are lifted over from the human genome. (D) Association results for the 7.7 Mb region. The top SNP (chr27:7,706,463 bp) is potentially located in an intron of SLC38A4.