Greyhound Osteosarcoma Genes Mapped!

Mapping disease genes using genome-wide association study (GWAS) in dog breeds, each effectively a genetic isolate only a few hundred years old, requires approximately 10× fewer markers and samples than in human populations. However, population structure, cryptic relatedness, and extensive regions of near fixation in breeds complicate GWAS analysis, and to date, few studies have successfully mapped risk factors for complex, multigenic diseases. In this study, we used new methods for analyzing breed populations to identify genomic loci associated with OSA in the first multi-breed association study of a highly polygenic canine disease.


We evaluated 267 racing Greyhounds (153 affected and 114 unaffected with OSA) and 19 AKC Greyhounds; we also evaluated 135 Rottweilers and 141 Irish Wolfhounds. The 3 breeds are visibly discrete genetic populations; the AKC Greyhounds clustered near but distinct from their racing brethren (Figure 1). The racing Greyhound population was the least inbred, likely reflecting a large effective population size.

The risk allele tagging the top Greyhound locus (A=0.84, U=0.65) is on chromosome 11. This haplotype occurs at lower frequency in the AKC Greyhounds (0.61) and in a panel of 28 other breeds from a previously published data-set (0.51% +/− 0.24). We narrowed the region of association to a 15 kb non-coding regionnear CDKN2A (en- codes p16INK4a and p14ARF), CDKN2B (p15INK4b), and the antisense non-coding gene CDKN2B-AS1/ANRIL. Both CDKN2A and CDKN2B function as cell cycle regulators and tumor suppressors. The risk haplotype at 11q16 in Greyhounds is syntenic (i.e. equivalent) to a non-coding regulatory region on human chromosome 9p21. The human 9p21 locus is deleted in 5% to 21% of human OSA, the absence of p16INK4a expression is correlated with decreased survival in pediatric OSA patients, and increased p16 expression is predictive of better response to chemotherapy. Thus, we hypothesized that the variant(s) carried on the risk haplotype disrupts enhancer elements upstream of the CDKN2A/B locus, thereby altering expression of one or more genes in the region. Our analysis of the breeds at high risk for OSA demonstrates that inherited variants are major factors for determining whether a dog develops OSA.

Through a parallel multi-breed canine association study, we found 33 genomic regions associated with OSA, and identified genes and pathways potentially causing this complex, polygenic, and poorly understood disease. Altogether, the 33 loci identified by GWAS account for 50% to 80% of the disease risk within each of these three breeds, demonstrating that inherited factors are the pre- dominant cause.

None of the OSA GWAS loci overlapped between breeds, a strikingly different genetic architecture from the shared variants previously found by mapping Mendelian traits in multiple dog breeds.

We hypothesize that the top canine OSA risk variant at chr11:44405676 alters regulation of CDKN2A/ARF. The OSA associated variant at dog chr11:44405676 disrupts a highly constrained position in a genomic locus that has strong enhancer activity in a human OSA cell line. CDKN2A/ARF encodes the INK4 family of cyclin-dependent kinase inhibitor proteins (including p16INK4a, p15INK4b, and p14ARF). These proteins control G1-progression by in- activation of D-cyclins, inducing senescence via the RB and p53 pathways. Altered levels of CDKN2A, a master regulator of tissue development, are linked to hematopoietic stem cell senescence and development, key feature of malignancies including OSA. SNPs that disrupt enhancer element binding can change transcriptional activity across the human 9p21 locus, including at CDKN2A. Germline variants affecting regulation of CDKN2A may alter the balance between proliferation and senescence in specific tissues, thereby leading to an increased risk of developing OSA and potentially also other cancers in adolescence and adulthood.

The association of glutamate receptors with OSA in both dogs and humans suggests glutamate signaling as a potential therapeutic target, although the diverse physiological functions of this pathway could make this difficult.