Breed-Predispositions to Cancer in Pedigree Dogs

Cancer is a common problem in dogs and although all breeds of dog and crossbred dogs may be affected, it is notable that some breeds of pedigree dogs appear to be at increased risk of certain types of cancer suggesting underlying genetic predisposition to cancer susceptibility. Although the aetiology of most cancers is likely to be multifactorial, the limited genetic diversity seen in purebred dogs facilitates genetic linkage or association studies on relatively small populations as compared to humans, and by using newly developed resources, genome-wide association studies in dog breeds are proving to be a powerful tool for unravelling complex disorders. This paper will review the literature on canine breed susceptibility to histiocytic sarcoma, osteosarcoma, haemangiosarcoma, mast cell tumours, lymphoma, melanoma, and mammary tumours including the recent advances in knowledge through molecular genetic, cytogenetic, and genome wide association studies.


Introduction
Cancer is an important disease in dogs and represents one of the major causes of canine death accounting for 27% of all deaths in purebred dogs in the UK in a recent mortality study [1]. is is slightly higher than what previously reported in a Danish Kennel club study (14.5% by Proschowesky et al.) and an earlier UK study (15.7% by Michell) but similar to a postmortem series of 2000 dogs, in which 23% of all dogs and 45% of dogs over 10 years of age died of cancer [2]. In the absence of reliable historical tumour registries, it is difficult to know whether the prevalence of cancer in dogs is actually increasing; however a number of factors may contribute to an increase in the diagnosis of cancer in dogs; as a result of improvements in health and welfare animals are living longer and cancer is generally a disease of older age [3]. Advances in veterinary medicine, particularly diagnostics and higher expectations of the pet owning public, are likely to result in an increased rate of diagnosis.
As is the case in the human population, many different types of naturally occurring cancer may affect dogs and canine malignancies have been established as strong comparative models for the human disease due to their spontaneous development and frequency; dogs live in our environment and eat similar food and are thus exposed to similar risk factors, so the aetiology and pathogenesis of canine tumours is likely to be similar to that of human tumours [4][5][6][7][8]. A general comparison of the incidence of canine cancer with that of human cancer highlights some striking similarities and differences [9]. Breast cancer is the most common malignancy in women and the mammary gland is a common site for tumour development in bitches, although the risk is reduced in bitches spayed at a young age [10], demonstrating the importance of endogenous hormones in the development of this disease. In contrast, carcinomas of the prostate, a very common condition in men and also associated with hormonal stimulation, is relatively uncommon in dogs and occurs more frequently in neutered dogs [11]. Carcinomas of the lung and large bowel, the most common human tumours excluding breast and prostate, do not feature highly in the canine population, whereas so tissue sarcomas, which are rare in humans, are relatively common in dogs.
However, whilst general trends for the occurrence of canine cancers are well established, accurate �gures for the frequency of different types of tumours in the canine population are limited. One study of insured dogs from the United Kingdom showed the skin and so tissues to be the most common sites for tumour-related claims (both benign and malignant) with a standardised incidence rate of 1437 per 100,000 dogs/year, followed by mammary, urogenital, lymphoid, endocrine, alimentary, and oropharyngeal sites. Canine cutaneous histiocytoma was the most common tumour type overall with a standardised incidence rate of 337 per 100,000 dog/year, followed by lipoma, adenoma, so tissue sarcoma, mast cell tumour, and lymphoma ( Figure  1, [12]). Other epidemiological studies, based on hospital populations or on surveys of intakes into pathology laboratories, are largely supportive of these estimates for the canine population as a whole [13][14][15][16][17], as are recent �gures from a number of European Tumour Registries [18][19][20]. It is well recognised that differences exist between breeds of dog and their risk of developing certain types of cancer but there are few large scale epidemiological studies on the incidence of different types of cancer in the canine population which document the variation between breeds. e breeds with the highest proportional mortality for cancer in the Kennel Club/BSAVA study included the following, in descending order: Irish water spaniel, �at-coated retriever, Hungarian wirehaired vizsla, Bernese mountain dog, rottweiler, Italian spinone, leonberger, Staffordshire bull terrier, Welsh terrier, and giant Schnauzer ( [1], Table 1). In a study of rates and causes of death in insured dogs in Sweden, Bonnett et al. [16] found that the Bernese mountain dog, Irish wolound, �at-coated retriever, boxer, and Saint Bernard were the �ve breeds of dog with the highest mortality from tumour-related death. e Bernese, Irish wolound, and leonberger were the top three in a subsequent examination of the same data base [21]. Bernese mountain dogs, �at-coated retrievers, golden retriever, and rottweilers were in the top 5 breeds with over 20% of deaths due to cancer in Denmark [22]. An ownerbased questionnaire conducted in the UK segregated breeds into "overrepresented, " "average" and "underrepresented" with respect to dying of cancer and showed the same trends ( Table 2) [23]. ese population-based studies provide useful indicators of breeds at risk of cancer, but should not be regarded as completely de�nitive because the outcome oen depends on the breed prevalence within the population at risk, which may explain the differences found in studies from different countries. e existence of other inherited diseases or breed-associated problems is a major confounding factor: for example the reason the bulldog has an apparent low risk of cancer in Table 1 may be due to its short lifeexpectancy due to other health issues that affect the breed [24]. Other limitations of such studies include owner compliance, bias through nonrandom sampling of the pedigree dog population, and accuracy of owner-reported cause of death. Furthermore a small number of dogs can seriously bias the results for numerically small breeds, as demonstrated by the red and white setter in Table 2 and probably the Hungarian wireheriad vizsla, and Welsh terrier in Table 1. However, the fact that all these studies consistently show similar overall breed-related predispositions to development of cancer has important implications in understanding the aetiology of cancer as it infers a genetic and heritable component. e dog has a unique population structure with each breed arising from a limited number of founders. e dog is believed to have evolved from grey wolves possibly from Europe or the Middle East although interbreeding with local wolf populations clearly occurred elsewhere in the early history of dog lineages [25][26][27]. Although the wolf dog became domesticated 20-30,000 years ago, and dog types were gradually established for guarding, herding, and hunting purposes, it is only in the past 200 years or so that selective breeding practices have divided the dog population into over 300 discrete breeds worldwide. e English Kennel Club was established in the Victorian era to meet the demands of the dog breeding Populus, to regulate the registration of dog breeds, and to establish breed standards. Kennel clubs in Europe and America still impose strict standards on registration of pedigree dogs requiring that the ancestors of each dog must be registered as well. is combined with the frequent use of popular sires and inbreeding practices means that each breed is a closed, isolated population with virtually no gene �ow between breeds [28]. Over the past 200 years this practice has resulted in reduced genetic diversity within breeds and greater genetic divergence between breeds. e average nucleotide heterozygosity when considered across dog breeds is comparable to the human population [29], but the level of genetic diversity within any single breed is considerably less than the species as a whole [25] indeed it has been estimated that whilst domestication of wild canid populations resulted in a 5% loss of nucleotide diversity, breed formation caused a 35% loss [30]. In many breeds the effective population size is very small even in normal times but in some breeds, for example, the Bernese mountain dog and the leonberger, genetic variation has been further reduced by serious population declines during war or hard economic times [31]. Mutations in a small number of genes of large effect are responsible for many breed characteristics; such selective breeding for exaggerated traits further reduces genetic diversity, and perhaps risks selection of mutations that predispose to disease [25,32,33]. Over 350 inherited disorders have been described in purebred dogs many of which are equivalent to human diseases [34,35]. e limited genetic diversity seen in purebred dogs, facilitates genetic linkage or association studies on relatively small populations as compared to humans [36,37] and by using newly developed resources, genome wide association studies in dog breeds are proving to be a powerful tool for unravelling complex disorders [38]. Although most forms of cancer are likely multifactorial in aetiology, the fact that different breeds of dog are predisposed to developing certain types of cancer (and conversely some breeds are at lesser risk) offers a unique opportunity to study and understand the genetic mechanisms underpinning cancer susceptibility [39,40]. e purpose of this article is to review the current literature on predispositions to cancer in pedigree dogs, including the recent advances in knowledge through molecular genetic, cytogenetic, and genome wide association studies and to consider how the application of this new knowledge will inform our approach to the diagnosis, treatment, and possibly prevention of cancer in the future. Table 1, taken from the results of a health survey of purebred dogs in the UK [1], lists the proportional mortality for different breeds. e mean mortality rate to cancer for all breeds was 27%, thus those breeds recording greater than 30% deaths from cancer might be considered to be at greater risk or predisposed and those less than 25% cancer deaths, at reduced risk, although that is not to say that these latter breeds do not suffer particular types of cancer; for example, the Belgian shepherd has been documented to be at higher risk of gastric carcinoma [41] and the Scottish deerhound has been shown to have a heritable risk of osteosarcoma [42]. Considering the genetic structure of the purebred dog, phylogenetic analysis has shown separation of several breeds with ancient origin from a large group of breeds with presumed European origins. e former include the spitz type breeds: Shar-Pei, Shiba Inu, Chow Chow, Akita, Siberian Husky, and Alaskan Malamute. Further studies to characterise genetic variation within and among breeds established at least four distinct breed groupings: a subset of breeds with ancient Asian and African origins (as above but also including Basenji from Africa, Saluki, Afghan hound from the Middle East, and tibetan Terrier, Lhasa Apso from china), a group of Mastiff-like breeds, a group re�ecting shared ancestral herding behaviour, and a group of huntingtype dogs ( Figure 2) [43]. Most recently a neighbor-joining tree of domestic dogs showing the relationship among the various dog breeds has been constructed by genotyping 10 to 12 dogs for each of 80 breeds (Figure 3) [26]. Breeds that share either common behaviors or morphologic traits are grouped together on the basis of DNA analysis, indicating that they probably share common ancestors. In this context T 2: Adapted from [23]  it is notable how some breed groupings "rank" in Table  1, with the Mastiff-type breeds dominating the top of the table and most of the ancient spitz type breeds below the mean. However, it is not possible to know whether this observation re�ects genetic predisposition to cancer or body size or perhaps that the two are interrelated. Body size has previously been shown to correlate with longevity, with smaller breeds having a longer life expectancy [1,23]. Of the breeds listed as over-represented in Tables 1 and 2, some have been associated with speci�c types of tumour, for example, Bernese mountain dog-systemic and malignant histiocytosis, Irish wolound-osteosarcoma, and others with a higher risk of more than one tumour type for example Boxer-mast cell tumour and brain tumour, Golden retriever-mast cell tumour, lymphoma, hemangiosarcoma. is observation also has important genetic implications, suggesting that some breeds may be like the rare human Li-Fraumeni families where a germ line mutation in a tumour suppressor gene (TP53) results in a hereditary predisposition to several types of cancer [44] or they may resemble the situation in families with mutations in BRCA1 where the risk of developing breast and ovarian cancer is greater in relatives of affected people, rather than the cancer being transmitted as an autosomal dominant condition [45,46]. In contrast other breeds affected by a speci�c type of tumour may have a more speci�c genetic abnormality leading to that particular type of tumour. For example, renal cystadenocarcinoma and nodular dermato�brosis (RCND) is a rare inherited cancer syndrome in German shepherd dogs. Affected dogs develop bilateral multifocal tumours in the kidneys and multiple dense collagenous skin nodules [47]. In the dog the disease gene was mapped to a region on canine chromosome 5q12, corresponding to a gene encoding tumour-suppressor protein folliculin [48]. It was subsequently shown that the same gene is mutated in Birt-Hogg-Dube syndrome, a similar disease in humans [49].
For the purpose of this paper, breed associations will be considered by tumour type rather than by breed.

Histiocytic Sarcoma
Histiocytic sarcoma is the current term used to describe a spectrum of poorly differentiated, pleomorphic tumours shown to have an immunophenotype consistent with a myeloid dendritic antigen presenting cell origin: CD1+, CD4−, CD11c+, CD11d−, MHC II+, ICAM−1+, and y−1± [50]. As many of these reagents cannot be used on formalin �xed para�n embedded sections, vimentin, MHC II, and the cell surface marker CD18 have become the standard panel for identi�cation of histiocytic sarcoma [51,52]. Localised histiocytic sarcoma describes those lesions which present as solitary masses, previously referred to as malignant �brous histiocytoma [53][54][55][56], and disseminated histiocytic sarcoma, those that present with multifocal lesions, previously referred to as malignant histiocytosis [50]. Whilst histiocytic sarcoma is an uncommon tumour in the dog population as a whole, certain pedigree breeds appear predisposed to this disease, notably the Bernese mountain dog where the disseminated form of histiocytic sarcoma accounts for up to 25% of deaths in the breed [57] and the �at-coated retriever where the localised form is more common and accounts for approximately 25% of all tumours in the breed, and up to 50% of all malignancies [58,59]. Disseminated histiocytic sarcoma in the Bernese typically presents with vague clinical signs including lethargy, anorexia, and weight loss. Investigations reveal multifocal disease primarily affecting the lungs, spleen, liver, bone marrow, and lymph nodes. Haematological abnormalities including anemia and thrombocytopenia are common [60][61][62][63]. e disease is rapidly progressive and fatal; many dogs are euthanized upon diagnosis; survival time from diagnosis has been reported to be 49 days [64]. In �at-coated retrievers localised lesions most commonly develop in the deep musculature of the limbs or in periarticular sites; the elbow is the most common site [65]. Even these localised lesions are highly malignant with rapid dissemination to lymph nodes preceding haematogenous spread to parenchymal organs and the skin in over 70% of cases [66]. us both the localised and disseminated forms of histiocytic sarcoma are highly malignant conditions that are largely refractory to conventional treatments and carry a very poor prognosis. Treatment with Lomustine (CCNU) has been reported to result in some short term responses [56,[66][67][68]. e archetype of Bernese mountain dogs, being affected by the disseminated form of the disease, and �atcoated retrievers by the localised form, is not entirely accurate as up to 30% of histiocytic sarcomas in �at-coated retrievers are actually visceral, arising most commonly in the spleen with other visceral sites including liver, lung, mediastinum, and lymph nodes also affected [65]. Included in the visceral form affecting this breed is the hemophagocytic variant of histiocytic sarcoma, which arises from splenic or bone marrow derived macrophages and expresses CD11d [69,70]. Furthermore localised periarticular histiocytic sarcoma is F 3: Summary of the neighbour-joining tree of domestic dogs [26] showing relationships between different breeds of dog on the basis of DNA analyses. Breeds that share either common behaviours or morphologic traits are grouped together on the basis of DNA analysis, indicating that they probably share common ancestors. e colours indicate breeds that probably share common founders.
also recognised in the BMD, where it is possibly predisposed by joint disease [71]. Histiocytic sarcoma is not exclusive to the Bernese and �at-coated retriever� other breeds of dog where histiocytic sarcoma has been reported with some frequency include rottweilers and golden retrievers [50,[72][73][74][75]. e striking high incidence of histiocytic sarcoma in these breeds of dog suggests a heritable predisposition. A recent study of Danish Bernese mountain dogs described 13 dogs diagnosed with malignant histiocytosis, of which 11 were genealogically related [76]. In 1995 Padgett analysed the inheritance of "histiocytosis" in 127 affected Bernese mountain dogs and suggested a polygenic mode of inheritance and a calculated heritability of 0.298 [57]. More recently a pedigree of 327 Bernese mountain dogs (144 males, 183 females) was developed from 800 French and European dogs. A total of 121 dogs had a clinical diagnosis of histiocytic sarcoma. Detailed analysis of this pedigree showed that the segregation of the disease observed in these families could not be explained by a fully recessive model and that an oligogenic model was likely to be a better description of the genetic model underlying the disease [64]. Histiocytic sarcoma has recently been investigated by molecular cytogenetic pro�ling [77] and genome wide association studies [78]. Using genome wide array comparative genomic hybridization, supplemented with �uorescence in situ hybridization and loss of heterozygosity analysis, copy number aberrations (CNAs) were assessed in 146 histiocytic sarcomas, 101 from Bernese mountain dogs (68 from USA and 33 from France) and 45 from �at-coated retrievers (all from USA) [77]. Numerous CNAs were found, both gains and losses, throughout the genome. irty-one regions of the canine genome presented with recurrent CNAs of which 6 were highly recurrent, all of which were deletions located on dog chromosomes 2, 11, 16, 22, and 31. Almost all these recurrent CNAs were shared between the two breeds, suggesting that they are more associated with the cancer phenotype than with breed and a subset suggested involvement of known cancer-associated genes including deletions of the tumour suppressor genes CDKN2A/B, RB1, and PTEN. A small number of abberations were unique to each breed, and the authors speculated that these may contribute to the major differences in tumour location evident in the two breeds [77]. Interestingly dysregulation of CDKN2 has also been associated with susceptibility to histiocytic sarcoma in Bernese mountain dogs by genome wide association study (GWAS) [78]. DNA was isolated from 474 blood samples from Bernese mountain dogs, 242 cases and 232 controls, 114 cases and 120 controls from North America and 128 cases and 112 controls from Europe. Both independent and combined GWAS were used to identify cancer-associated loci, �ne mapping and sequencing narrowed the primary locus to a single region. Both populations shared the same primary locus which featured a single haplotype spanning MTAP and part of CDKN2A which was present on at least one chromosome in 96% of affected dogs, with 65% of cases being homozygous [78]. is haplotype is within a region homologous to human chromosome 9p21 which is an important tumour suppressor locus and is implicated in many human cancers [79,80]. It is likely that the MTAP-CDKN2 locus is associated with more than one tumour type; Bernese mountain dogs are also susceptible to lymphoma, mast cell tumours, and osteosarcoma which seem to show familial clustering. However, the presence of the risk haplotype among control dogs could be due to the fact that the risk associated with this locus is modest and that at least some of the control dogs lack additional risk alleles at other loci. Aer all, it is likely that more than a single locus will be involved in the predisposition [64,78].
To date there are no published data on whether this haplotype or tumour suppressor locus is important in histiocytic sarcoma in other dog breeds especially rottweilers or golden retrievers. However, there is some evidence that it may be important in other canine sarcomas. Disruption of chromosome 11 involving the loss of the CDKN2b-CDKN2a tumour suppressor gene cluster region has been reported in two �brosarcomas in Labrador retrievers [81]. e same study sequenced exon 1 of CDKN2B using DNA from blood of 141 dogs of 18 different breeds and showed widespread polymorphism of this �rst exon. Seven alleles were recorded and sixteen of the eighteen breeds showed heterozygosity. Further investigations into the role of this tumour suppressor region in other canine so tissue sarcomas may yield some interesting �ndings that may start to explain the relatively high incidence of so tissue sarcomas in dogs in comparison to humans, where such tumours are rare.

Osteosarcoma
Osteosarcoma of the long bones is the most common malignant tumour of bone in dogs accounting for 85 −90% of primary bone tumours and almost exclusively affects the large and giant breeds such as rottweiler, great Dane, Irish wolound, greyhound, Saint Bernard [82]. e aetiology of osteosarcoma is probably multifactorial; the predilection for the tumour to develop at metaphyseal region of long bones, especially the distal radius and proximal humerus, correlates with weight bearing, and rapid bone growth during early development along with bone stress due to weight bearing (possibly resulting in microfractures) has been implicated in the aetiology of these tumours [83]. Increasing weight and height appear to be important predictive factors for the disease in the dog [84]. Growth hormone has been shown to be present in canine osteosarcoma samples [85] and studies evaluating the role of insulin-like growth factor-1 (IGF-1) and its receptor IGF-1R and hepatocyte growth factor (HGF) and its receptor c-Met in osteosarcoma cell lines and tissues have shown that these factors may contribute to the malignant phenotype [86,87]. Sex hormones may also contribute to osteosarcoma risk with intact males and females being reported to be at increased risk [84]. However in the rottweiler neutering before 1 year of age appeared to increase risk of bone sarcoma in both male and female dogs [88]. As is the case in many human and canine cancers, alterations in the function of the tumour suppressor genes RB and TP53 have been implicated in the pathogenesis of canine osteosarcoma [89,90]. e fact that speci�c breeds appear predisposed suggests that more speci�c genetic factors may be implicated in the aetiology of canine osteosarcoma. Breeds reported to be at increased risk of developing osteosarcoma include the doberman, German shepherd, golden retriever, great Dane, Irish setter, rottweiler and Saint Bernard [88,91,92], large sight-hounds such as Irish wolound, Scottish deerhound and Borzoi [84], greyhound, rottweiler and great Dane [83] and Irish wolound, Saint Bernard, and Leonberger [93]. It may be argued that these predispositions may be related to size rather than breed, and this is supported by the fact that greyhounds and whippets consistently clustered together in an analysis of molecular variance in microsatellite loci [26,43], yet whilst appendicular osteosarcoma affects greyhounds [82], the disease is rare in whippets, or indeed in dogs under 25 kg body weight [84,94]. Interestingly IGF1 and its associated regions are a major contributing locus in size diversity in dogs, accounting for about 50% of the genetic variation in size [95]. Osteosarcoma is particularly prevalent in retired racing Greyhounds, possibly implicating stress or trauma in the aetiology of the disease in this breed [83,96]. A familial incidence has been observed in Saint Bernards [91] and more recently a study modelling the transmission of osteosarcoma in a population of over 1000 Scottish deerhounds estimated heritability at 0.69 and modelling suggested that a major gene with dominant effect would explain the pattern of transmission [42].
It is well documented that canine osteosarcoma is a good clinical model for the human disease [97][98][99] and there is a growing body of evidence to show that canine and human osteosarcoma have a similar molecular pathogenesis [100,101]. Gene expression pro�ling of canine osteosarcoma has revealed genes associated with progression, survival, and metastasis that are relevant to human osteosarcoma [100,[102][103][104][105]. As is the case in human osteosarcoma, the canine disease is characterised by an extremely complex karyotype indicating extensive genomic instability. Using two breeds of dog with different relative risk of osteosarcoma-rottweiler (12.5% incidence rate) and golden retriever (5% incident rate)-a recent study has shown that the individual genetic background, as de�ned by breed, in�uences the tumour karyotype in osteosarcoma [106]. Eleven loci (from 8 different chromosomes) showed a signi�cant difference in the distribution of DNA copy number imbalances between tumours from golden retrievers compared with those from rottweilers; the most signi�cant of these was the deletion of the WT1 gene which occurred in 48% of the rottweiler cases (14/29) but which was not observed in any of the 9 golden retrievers. Genomic loss of TP53 and CDKN2A suppressor genes were also restricted to rottweilers (7/29 (24%) and 5.29 (17%), resp.). Overall 15/29 rottweilers in this study showed genomic deletion of at least one of the WT1, TP53, CDKN2A, PTEN, or RB1 tumour suppressor genes. ese breed-associated imbalances may contribute to or result from heritable risk factors. A larger study which pro�led 123 cases of canine osteosarcoma by 1 Mb aCGH also demonstrated a high occurrence of genetic imbalances similar to human osteosarcoma and identi�ed several new candidate genes in regions of the canine genome that had highly recurrent copy number abnormalities [101]. Although this dog population was represented by 4 main breeds, rottweiler ( ), greyhound ( ), great Pyrenees ( ), and golden retriever ( ), no signi�cant differences were found between aCGH de�ned regional abberations and breed groups. It is likely that larger cohorts with less variation in tumour histology will be needed to provide the power to detect signi�cance.

Hemangiosarcoma
Hemangiosarcoma is a highly malignant tumour arising from blood vessels, probably less common than some of the other mesenchymal malignancies, for example, �brosarcoma; it has been estimated to represent 7% of canine malignant tumours [5] and accounted for 24/100,000 dogs/year in one study [12], which is still considerably higher than angiosarcoma in humans [107]. e most common primary sites for hemangiosarcoma in dogs are visceral organs, notably the spleen and liver; it may also arise in the right atrial appendage. A dermal form of haemangiosarcoma is also seen in dogs, which has a predilection for light haired or nonpigmented skin, particularly on the ventral abdomen and UV light has been implicated in the aetiology of this form of the disease [108]. Visceral hemangiosarcoma has a predilection for certain breeds; the German shepherd dog has been reported to have an increased risk with an odds ratio of 4.7 (95% CI 2.7-7.8) compared to other purebred dogs [109], but boxers and golden retrievers have also been identi�ed as being at increased risk [110,111]. More recently hemangiosarcoma appears to have become a signi�cant problem in golden retrievers in North America with an estimated life-time risk of 1 in 5 reported by the Golden Retriever Club of America [112,113].
A small number of studies have examined the molecular genetic aspects of canine hemangiosarcoma with respect to growth regulation genes; mutations in the tumour suppressor gene TP53 have been reported [114,115] and a further study suggested that alteration of the p16-cyclin D1-Rb pathway may be associated with the pathogenesis of canine hemangiosarcoma. In this study only 18% of samples showed alterations in both TP53 and p16 [116]. e Rb (p16) pathway is also commonly deregulated in human cancers [117]. As a naturally occurring tumour of endothelial cells, the role of vascular endothelial growth factor (VEGF) in the pathogenesis and progression of hemangiosarcoma has been examined. VEGF has potent angiogenic, mitogenic, and vascular permeability enhancing properties and plays a major role in tumour growth in human cancers where plasma concentrations of VEGF have been shown to correlate with tumour burden and prognosis [118]. In one study, dogs with hemangiosarcoma ( ) were signi�cantly more likely to have detectable concentrations of plasma VEGF compared to healthy dogs ( ) [119], although the same group were not able to demonstrate a marked difference in VEGF concentration between body cavity effusions associated with malignant versus nonmalignant diseases [120]. VEGF is a target for some of the newly licensed tyrosine kinase inhibitors including masitinib mesylate (Masivet-AB Science) and it has recently been shown that masitinib causes a dose-dependent-cell death in canine haemangiosarcoma cell lines [121]-further implicating a role for VEGF in canine haemangiosarcoma.
Phosphatase and Tensin Homolog Deleted from Chromosome 10 (PTEN) is a tumour suppressor gene which is inactivated in many human cancers [122]. PTEN also inhibits angiogenesis, possibly by regulating VEGF gene expression via the P13K pathway [123,124]. To determine the role of PTEN in the origin or progression of canine hemangiosarcoma, Dickerson et al. [125] �rstly examined by immunohistochemistry the expression of CD31 (endothelial marker), PTEN, VEGF, p27 AKt, and p-AKT in sections from 12 haemangiosarcomas and 5 benign splenic haematomas and observed some abnormalities in PTEN expression in the tumours. Using cell lines established from some of these tumours the same group identi�ed mutations of PTEN in the C-terminal domain that may affect the subcellular localisation and stability of the protein [125]. ey speculated that constitutive activation of P13 K or loss of PTEN function might establish autocrine growth loops that promote autonomous growth and transformation of endothelial cells [122,126]. However the PTEN mutation alone does not fully explain the increased levels of VEGF and other angiogenesis promoting growth factors (platelet derived growth factor and basic �broblast growth factor) thought to be elaborated by haemangiosarcoma cells, or the role of the in�ammatory cells frequently associated with these tumours [125]. Gene expression pro�ling has shown canine hemangiosarcoma to have a gene signature suggesting that in�ammatory and angiogenic pathways play a signi�cant role in its pathogenesis [127]. Genes expressed at signi�cantly higher levels in haemangiosarcomas than osteosarcoma, non-Hodgkin's lymphoma, and leukaemia included VEGFA, TIMP-1, FN-1, ADAM9, PDGFC, MMP14, TNF , and acid ceramidase.
Gene expression pro�ling of canine hemangiosarcoma has also shown signi�cant breed differences segregating hemangiosarcomas derived from golden retrievers from hemangiosarcoma in other breeds, with contributions from transcription factors, survival factors, and from pro-in�ammatory and angiogenic genes. VEGF Receptor 1 was preferentially enriched in tumours from golden retrievers versus other breeds [113], suggesting that heritable factors mould gene expression phenotypes and consequently biological behavior. e high incidence of hemangiosarcoma in golden retrievers in the USA appears to be a relatively recent phenomenon, as the golden retriever was not reported to be over-represented in a study from the University of Pennsylvania in 1988 [109]. Furthermore hemangiosarcoma does not appear to be particularly prevalent in the breed in the UK, indeed the data from the insured dog study [12] shows golden retrievers to be less at risk of hemangiosarcoma than all other breeds pooled (unpublished data). is is particularly interesting because in a study investigating genetic diversity among four common breeds sampled in the US and Europe, the golden retriever showed a high level of genetic difference between European and American subpopulations allowing them to separate into two distinct populations in clustering analyses corresponding to their geographical origins [128]. Whereas less differentiation was seen in the Bernese mountain dog and very little in rottweilers or �at-coated retrievers from the two continents. is result is partly due to breed popularity and population size. e golden retriever is a very popular breed with more than 42,000 American Kennel Club registrations and 8,000 UK Kennel Club registrations, such that there is a large gene pool on both continents and mixing between the two populations is rare. In contrast the �at-coated retriever popularity in the UK declined following the �rst world war leaving a small gene pool from which modern dogs derive; the breed still has a relatively small population size worldwide which limits genetic variation within the breed regardless of geographical origin. Not only does this geographic variation in the genetics of certain breeds indicate the importance of taking account of population substructure in genetic studies using worldwide sampling of pure-bred dogs, the genetic differences between the European and North American golden retriever could also be key to understanding the aetiology of hemangiosarcoma in the breed.

Mast Cell Tumour
Mast cell tumours (MCTs) are common tumours of the canine skin, estimated to represent 7-21 percent of all skin tumours in this species [15,129,130]. In the UK, MCT is the second most common canine malignancy, aer so tissue sarcoma with an incidence of 129 per 100,000 insured dogs per year [12]. Cutaneous MCTs are typically solitary lesions but their clinical appearance can be variable and dogs can develop more than one unrelated MCT [131].
e boxer and bull dog breeds including bullmastiffs, Boston terriers, and Staffordshire bull terriers are reported to show an increased risk of developing MCT and it has been postulated that this might be linked to a common ancestry [132]. ese breeds have now been shown to cluster closely phylogenetically [26]. Rhodesian ridgebacks, pugs, weimaraners, Labrador retrievers, beagles, and golden T 3: Breeds over-and underrepresented in a recent American survey of breed association to canine mast cell tumours [130]. retrievers have also been reported to be at increased risk [13]. A phenomenon of occurrence of mast cell tumours in young Shar-Pei dogs was reported from one US veterinary pathology laboratory in 1995. From a total number of 802 submissions diagnosed as MCT, 18 were from Shar-Peis and 5 of these were in dogs less than 2 years of age. Poorly differentiated (Grade III) tumours were more common in younger dogs [133]. Another recently published North American study considered the breed distribution of various canine cutaneous tumours, including MCT, and documented not only those breed over-represented for the disease but also those at reduced risk [130] (Table 3). In our clinic (Cambridge, UK) Labrador retrievers, golden retrievers, and Staffordshire bull terriers all appear over-represented (in comparison to the general Hospital population) whilst German shepherd dogs, cocker spaniels, border collies, cavalier King Charles spaniels, and west Highland white terriers are under-represented [134]. Mast cell tumours show very variable clinical behavior and interestingly this may be in�uenced by breed. Although boxer dogs and bulldog breeds are at higher risk of developing MCTs, these breeds of dog tend to have low grade, less aggressive tumours, as is also the case for pug dogs [135]. Labrador retrievers tend to have more aggressive tumours and golden retrievers are at risk of developing multiple tumours [131].
Despite a wealth of clinical studies on treatment and prognostic indicators in canine mast cell tumours (review [136]) there is very little published on the molecular genetics of these tumours. Alterations in the p53 tumour suppressor pathway have been identi�ed in some canine MCTs but TP53 sequencing in a small number of cases did not reveal any mutations [137].
Recent work has implicated the stem cell factor receptor (KIT) as having a role in the aetiology of canine MCTs [138]. KIT is normally expressed on haematopoteiic cells and mast cells. It consists of an extracellular ligand-binding domain, a transmembrane region, and a cytoplasmic tail with tyrosine kinase activity. Activation of the KIT signal transduction pathway plays a role in the growth and development of normal mast cells. KIT is encoded by the protooncogene c-KIT, dysregulation of which occurs in many human cancers. In 1996, London and others demonstrated expression of KIT on malignant mast cells derived from 4 spontaneous canine MCTs and subsequently reported internal tandem duplications in exon 11 of c-KIT in approximately 30 percent of canine MCTs [139]. Other studies have shown mutations in the juxtamembrane domain of c-KIT in dogs with MCTs, mainly within exon 11, with duplications, deletions and substitutions being described [140][141][142][143][144][145][146]. Several studies reported a signi�cant association between mutation and higher grade of tumour [142,143,145]. To date no mutations have been identi�ed within the canine c-KIT gene away from the juxtamembrane domain, and in particular no mutations have ever been identi�ed in the c-KIT kinase domain in canine MCT [146]. is is in contrast to mast cell disease in people, where the most common mutation is a single nucleotide substitution resulting in a single amino acid substitution in the kinase domain. Internal tandem duplications have not been reported in the c-KIT gene of people.
e presence of c-KIT mutations in only a proportion of mast cell tumours suggests that although mutations in this gene may be responsible for the development of some mast cell tumours, it is likely that mutational events in other genes are involved in the carcinogenesis of many mast cell tumours which are yet to be identi�ed. Tumour suppressor in lung cancer-1 (TSLC1) is a tumour-suppressor gene coding for an adhesion molecule that is involved in normal mast cell to mast cell, and mast cell to �broblast interactions [147] Loss of TSLC1 expression is associated with a poor prognosis in various human tumours, including non-small cell lung cancer, breast, prostate, oesophageal and gastric tumours [148][149][150] and has been shown to correlate with grade in canine MCTs [151]. In humans, mutations in the mismatch repair (MMR) genes give rise to Lynch syndrome, an inherited predisposition to early onset cancer, especially intestinal adenocarcinoma and skin cancer [152]. MMR expression was investigated immunohistochemically in mast cell tumours from young dogs of predisposed breeds versus old dogs of "non-MCT" predisposed breeds, but no signi�cant differences were observed, suggesting that MMR gene defects are not involved in the pathogenesis of canine mast cell tumours [153].

Lymphoma/Leukaemia
Lymphoma is the most common haematopoietic malignancy in the dog. Dog breed has been shown to play a role in the epidemiology of lymphoma with several studies showing a signi�cantly higher relative risk for boxers, bullmastiff and bulldog breeds compared to other breeds [9,14,154,155]. Other breeds identi�ed with increased relative risk include basset hound, St Bernard, Scottish terrier, Airedale terrier, Bouvier des Flandres, Labrador retriever, and rottweiler [14,156]. Familial associations with lymphoma have also been reported; nine of 59 bullmastiffs from 3 households died due to lymphoma over a 3-year period with most of the dogs that died also having a close ancestor that had also developed lymphoma [154]. Clustering of lymphoma has also been reported in related rottweilers and three directly related Otterhounds [157].
Interrogation of the data on a population of insured dogs in the UK showed a signi�cant breed effect with the boxer, bulldog, and bull mastiff breeds all having a high incidence of lymphoma in certain age ranges, with the English springer spaniel, golden retriever, and rottweiler breeds showing some indication of excess lymphoma incidence over expectation at certain age ranges- Table 4 [158]. In a study of just over 600 dogs diagnosed with lymphoma in France over 1 year, boxers, setters, and cocker spaniels were signi�cantly over-represented, and a possible predisposition in rottweilers and Beauce shepherd dogs was also reported [159]. ese authors also examined distribution of immunophenotype of lymphoma by breed and demonstrated that boxer dogs showed a signi�cant predilection for T-cell lymphoma but were underrepresented among centroblastic polymorphic lymphomas. Other breeds also appeared predisposed to speci�c immunophenotypes, with B-cell lymphoma predominant in German shepherd dogs and probably in the rottweiler. Whilst this study also demonstrated an association of canine lymphoma with waste incinerators, polluted sites, and radioactive waste, the strong breed associations with immunophenotype and histological subtypes also support a genetic aetiology.
It has recently been shown that the prevalence of subtypes of lymphoma also varies with breed or breed group. A distinct B-cell and T-cell prevalence of lymphoproliferative disease by breed was reported in North America; 1263 dogs representing 87 breeds, whose samples had been submitted for PARR analysis-clonal rearrangement of immunoglobulin heavy chain or T-cell receptor y chain-showed breed-speci�c susceptibility to develop B-cell or T-cell tumours. Boxers showed increased risk of developing T-cell tumours as did the "Spitz" breeds and Asian "lap" dogs, whereas border collies, basset hounds, cocker spaniels, and dobermans were affected by predominantly B-cell tumours [160]. Because it is retained in related breed groups (e.g., Spitztype dogs and Asian "lap" dogs) the elevated risk for Tcell lymphoproliferative disease may have arisen ancestrally, whereas increased risk of B cell disease may stem from different risk factors or combinations which arose during the process of breed selection. e boxer T-cell lymphoma has been further classi�ed as being predominantly of TCRalpha/beta+, CD4+ (helper) T-cells with lymphoblastic (high grade) morphology [161]. is strong breed association with different tumour subtypes may go some way to explaining differences in response to chemotherapy and survival time by breed which have been observed in some clinical studies [162].
Using CGH analysis on a subset of these tumours, the same study [160] identi�ed unique patterns of chromosomal gains and losses that segregated speci�cally with B-cell tumours and T-cell tumours (as previously identi�ed by [163]) indicating that consistent genetic abnormalities are associated with different tumour types and thus lending weight to the hypothesis that there is a heritable risk for development of canine lymphoma. A deletion of chromosome 14 12 ISRN Veterinary Science T 4: Patterns of excess lymphoma by breed and age, comparing the observed numbers of lymphoma cases per breed, over four (quartile) age ranges, with expected numbers computed from age incidence. was exclusively seen in diffuse B-cell lymphoma and occurred in 7 out of 7 golden retrievers but only in 13% (4 of 31) of dogs from other breeds. omas et al. [164] have recently reported the results of a genome-wide survey of tumourassociated CNAs through array-based comparative genomic hybridisation analysis in 150 cases of canine non-Hodgkin lymphoma in predominantly three breeds of dog: boxers, Labrador retrievers, and golden retrievers. Interrogation of the recurrent CNAs identi�ed revealed an extensive catalog of chromosomal regions and genes presenting with recurrent DNA copy number imbalance, within which are key genes previously associated with a range of human malignancies; copy number loss of CDNK2A/B occurred in 20/36 T-cell lymphoma cases and was more frequent in high grade than low grade cases. e CNA associated most signi�cantly with B cell lymphoma was a highly recurrent deletion of a discrete region on chromosome 26 (74/106 cases) which encompasses the canine immunoglobulin lambda locus. With regard to breed, no signi�cant correlations were found within canine B cell lymphoma suggesting that B cell lymphoma in different breeds shows a highly conserved genomic copy number status, whereas the cytogenetic pro�les of T-cell lymphoma were more strongly in�uenced by the genetic background of the patient. Seven individual loci (on chromosomes 6, 12, 20, and 31) showed highly signi�cant association with breed, each demonstrating an elevated incidence of copy number gain in boxers with T-cell lymphoma [164].

Melanoma
Melanocytic tumours are relatively common in dogs; they account for 4% of cutaneous tumours [13] and represent one of the most common oral malignancies in the dog [165][166][167]. Ocular and subungual variants are also described [168]. Canine melanocytic neoplasms vary widely in behavior; oral/mucosal forms are usually malignant and provide a good model for the human mucosal melanoma [4], although a small proportion of tumours at this site are well differentiated and follow a more benign course [169]. Cutaneous and ocular tumours are usually benign, but tumours of the eyelid and nail bed (subungual) are usually malignant. Despite these generalisms, it should be acknowledged that the biological behavior of canine melanocytic neoplasms can vary widely and although many studies have evaluated various prognostic markers, an accurate prognostic classi�cation for these tumours has yet to be established [170,171]. Cutaneous melanoma occurs more commonly in dogs with heavily pigmented skin, with Schnauzers (both miniature and standard) and Scottish terriers at increased risk [13]. Small breeds especially cocker spaniels and poodles and dogs with heavily pigmented oral mucosa are reported to be at greater risk of oral melanoma [165,172]. A more recent study of canine oral melanomas showed the Chow Chow, golden retriever, and Pekingese/Poodle mix breeds to be overrepresented, whereas the boxer and German shepherd breeds were under represented [173]. It is not clear whether these breed predilections re�ect an underlying genetic risk or merely re�ect heavy pigmentation in some breeds, or a combination of the two. However, breed and familial clustering does support underlying genetic risk factors for melanoma [168]. Furthermore, one study has reported that breed may have some prognostic signi�cance with more than 75% of melanocytic neoplasms exhibiting benign behavior in the doberman and miniature Schnauzer, in contrast to more than 85% of melanocytic neoplasms being malignant in the miniature poodle [174], although tumour site may be a confounding factor in these results. In humans sunlight exposure is an established environmental factor involved in the pathogenesis of cutaneous malignant melanoma [175]; however 6 to 14% of melanoma patients have a family history of melanoma, and it appears there are complex environmental-genetic interactions in such cases. Familial cases tend to be younger, to have a higher number of moles and to develop multiple primary tumours [176]. It has been shown that there are at least two genes involved in familial melanoma in humans, the tumour suppressors CDKN2A(p16) and CDK4. Families with germline mutation in CDKN2A are also prone to pancreatic cancer [177,178]. A number of pathways and genetic mutations have been identi�ed in nonfamilial cutaneous melanoma; these include activating BRAF or NRAS mutations resulting in hyperactivation of the mitogen-activated protein kinase (MAPK) pathway [179] and loss or mutation of PTEN, a negative regulator of the P13 K pathway [180]. Recent genome wide association studies have identi�ed a number of loci associated with nonfamilial melanoma and many mutations have also been identi�ed through genome sequencing [181][182][183], but the functional role of these mutations and variants within these loci is not known and it is likely that many other genes and environmental interactions, involved in the pathogenesis of melanoma.
e role of tumour suppressors has been evaluated in canine melanoma. In one study using canine melanoma cell lines and tumour tissue samples, the loss or signi�cant reduction in p16 expression was the most common abnormality found in 6/7 cell lines and 21/26 tumour samples [184]. Loss or signi�cant reduction of PTEN expression was also seen in 4/7 cell lines and in 13/27 tumour samples. Changes in other tumour suppressors TP53, Rb, and p21 were also detected, suggesting that loss of tumour suppressor function is a common occurrence in canine melanoma. is study included both dermal and oral melanomas; 14 tumours were benign and 11 malignant but the abnormalities in p16 occurred with equal frequency in both benign and malignant tumours, suggesting that inactivation of this pathway is a critical step in the pathogenesis of melanoma. Loss of the tumour suppressor gene products, p21/Waf1 and p53, has previously been demonstrated in a benign multicentric melanoma from a male Gordon setter [185]. More recently dysregulation of the Wnt/beta catenin signal pathway has been reported in 18 canine cutaneous melanomas, demonstrated by abnormal intracellular accumulation and increased expression of beta catenin [186]. MicroRNA pro�les have been examined in canine melanoma tissues and human and canine melanoma cell lines and microRNAs 145 and 205 have been identi�ed as tumour suppressors in both canine and human melanoma cell lines [187,188]. Other studies have examined KIT expression in cutaneous melanocytic tumours [189] and oral malignant melanomas [190], Cyclooxygenase-2 expression [191,192], expression of matrix metalloproteinases [193], and Ki67 expression [194] in various canine melanomas but more as prognostic indicators than as clues to the pathogenesis of these tumours. It is clear that canine melanoma offers a relevant model for the human disease and work is ongoing to elucidate further genetic abnormalities that contribute to the pathogenesis of the disease.

Mammary Tumours
Tumours of the mammary glands are the most common tumour to affect entire bitches representing between 50−70% of all tumour types [18,195]. A standardised incidence rate of 205/100,000 dogs/year has been reported in a population of insured dogs in the UK [12] and in a Swedish study of insured dogs that the overall rate of mammary tumour development was 111 dogs per 10,000 dog years at risk (DAYR) [196]. Mean age of onset is approximately 8 years. It is well established that ovarian hormone stimulation increases the risk of mammary tumour development in dogs as in other species (including humans) and in the bitch, ovariohysterectomy prior to 2 years of age greatly reduces the risk of mammary tumours in later life [10,197]. e incidence of canine mammary tumours does vary by breed but breeds reported to be at risk vary between different studies and different geographical locations. Poodles (toy and miniature), spaniels (English springer, cocker, and Brittany), Puli, English setter, pointers, German shepherd, Maltese terrier, Yorkshire terrier, and dachshund have all been reported to be predisposed [198]. In the Swedish study where the overall mammary tumour rate was 111 dogs per 10,000 DAYR the English springer spaniel, doberman pinscher, and boxer showed signi�cantly increased incidence rates of 319, 297, and 256 per 10,000 DAYR, respectively, whereas the rough haired collie showed signi�cantly reduced risk of 5 per 10,000 DAYR [196]. It should be noted that very few bitches are routinely neutered in Sweden, so mammary tumours are common [3]. A population based study of mammary tumours in Norwegian dogs showed boxers, cocker spaniels, English springer spaniels, and dachshunds to have the highest relative risk of mammary tumour [195]. A study in Japan reported a lower incidence of malignancy in mammary tumours from small breed dogs [199]. is variation in incidence of mammary tumour risk between breeds suggests a signi�cant heritable genetic component to the disease in dogs. A proportion of human breast cancer is familial; women who have inherited mutations in the BRCA1 or BRCA2 (BRCA 1/2) genes have substantially increased risk of breast cancer [45,46] but it is recognised that mutations in these genes only account for a small part, approximately 10%, of the total inherited effect [200]. Furthermore BRCA1/2 mutation is rare in cases of sporadic breast cancer [201]. Four further genes, FGFR2, LSP1, MAP3K1, and TOX3, were associated with a mild increase in risk of breast cancer in a GWAS [202]; however over 50% of breast cancers occur in women who do not carry these higher risk genotypes. Breast cancer risk is currently believed to be polygenic with liability conferred by a large number of loci, each contributing a small effect [203,204]. Oncogenes reported to play an early role in sporadic breast cancer include MYC, CCND1 (Cyclin D), and ERBB2 (HER2/neu) [201]. To date the increasingly powerful molecular techniques available to sequence breast cancer have not been able to elucidate speci�c genetic solutions but rather have highlighted the substantial genetic diversity underlying this common disease [205].
In contrast to the vast number of gene expression studies in human breast cancer relatively few gene expression studies have been published on canine mammary tumours, most studies having focused on speci�c genes or receptors. A variable proportion of canine mammary tumours have been reported to contain mutations in TP53 [206,207] and studies have shown an association between the level of COX-2 expression, malignant phenotype, and prognosis [208]. A recent comparative study of gene expression in human breast and canine mammary tumours and normal mammary tissue observed a signi�cant overlap of genes deregulated in the tumour samples, as compared to their normal counterparts [209], and pathway analysis of the gene expression data revealed many cancer related pathways, to be similarly perturbed including the P13/AKT, KRAS, PTEN WNT-beta catenin pathways and the MAPK cascade (Table 5). ese �ndings con�rm and support the value of canine mammary cancer as a model for human breast carcinogenesis [7,209]. Although mammary tumours are one of the targets being studies by a European consortium of canine geneticists and clinicians (LUPA), no publications to date have reported differential gene expression underlying susceptibility to mammary cancer by breed [38]. However BRCA1 and BRCA2 have been associated with mammary tumours in English springer spaniels in Sweden. Ten human breast cancer genes were evaluated for association with mammary tumours in 212 mammary tumour cases and 143 controls by genotyping SNPs. BRCA1 and BRCA2 were signi�cantly associated with mammary tumours and the association was stronger to BRCA1 in malignant tumours. A borderline association was seen for FGFR2 [210]. e same group also investigated the role of the dog leucocyte antigen (DLA) system as a genetic risk factor in the aetiology of canine mammary tumours in English springer spaniels, and by genotyping the polymorphic exon 2 of DLA class II loci, identi�ed a signi�cant association between a rare protective haplotype of MCH class II and the incidence of mammary tumours in this population of 363 Spaniels, 218 cases and 145 healthy controls [211]. Not only do these �ndings support the concept that MHC class II molecules play a critical role in tumour surveillance but that immune response to cancer may be in�uenced by genotype. A high interbreed and relatively low intrabreed variation in MHC alleles and haplotypes has

Other Epithelial Malignancies, Carcinomas
In contrast to the high prevalence of lung and large bowel cancer in the human population, particularly in the Western World, carcinomas of the lung and large bowel are relatively uncommon in the canine population. A number of tumour registries and clinical case studies have highlighted breed predispositions for carcinomas arising at other sites, as listed in Table 6. Many of these are really just anecdotal observations, the underlying genetic basis of which has rarely been investigated; however a few are worthy of note. Canine anal sac gland carcinoma (ASGC) is a relatively uncommon malignancy arising from the apocrine glands in the walls of the anal sacs. is tumour is invasive and metastatic in nature and is oen associated with a paraneoplastic hypercalcaemia. Although ASGC may arise in any breed of dog the English cocker spaniel and to a lesser degree other spaniel types (English springer and cavalier King Charles) have been reported to be predisposed to development of this tumour [214]. Predisposing genetic factors have yet to be elucidated but an association between ASGC and dog leucocyte antigen DQB1 has been demonstrated in the English cocker spaniel [215]. e allele distribution in DLA loci DAL-DRB1, -DQA1, and DQB1 was compared between 42 cases and 75 controls; there was no difference in allele distribution in DLA-DRB1 while a signi�cant difference was obtained for DLA-DQA1 and DQB1 alleles, with the DLA-DQB1-00701 allele having a higher frequency in cases than controls [215]. Interestingly a similar DLA-DQB1 allele association has been shown in the English cocker spaniel and immune-mediated hemolytic anaemia [216]. It is not known at this time whether the allele itself has a causative effect on development of disease, perhaps through altered immune function, or whether it is an indirect association as a marker locus with a causative locus located in its vicinity.
Other breeds or types of dog have been associated with less common carcinomas, for example, squamous cell carcinoma of the digit appears to have a predilection for large black dogs including giant Schnauzers and standard poodles [217,218], who interestingly cluster quite closely on the phylogenetic tree [26]. A familial association has been reported in giant schnauzers [219]. Scottish Terriers have been shown to be at 19-fold increased risk of transitional cell carcinoma of the bladder compared with mixed breeds, the cause of which is not known but perhaps represents genetic predisposition through differences in metabolic and detoxi�cation pathways [220]. In addition to the Scottish terrier, Airedale terriers and beagles have also been identi�ed as being predisposed to lower urinary tract tumours, whereas German shepherd dogs appear to be under-represented [221]. In contrast to human prostatic carcinoma, which is androgen dependent, prostate cancer has been reported to occur more commonly in neutered than intact male dogs, possibly because a high proportion of canine prostatic carcinomas are transitional cell origin as opposed to adenocarcinoma [222]. Although neutering status is a strong risk factor for canine prostatic cancer, breed predisposition has been demonstrated with dobermann pinschers, Shetland sheepdogs, Scottish terriers, beagles, German shorthaired pointers, Airedale terriers, and Norwegian elkhounds all having a statistically signi�cantly increased odds of having prostatic cancer of any histology independent of neutering status [11].

Brain Tumours
Intracranial neoplasia is quite well described in the dog where the most common primary CNS tumour is meningioma, followed by glial tumours (astrocytoma and oligodendroglioma). Choroids plexus tumours, medulloblastoma, neuroblastoma, and ependymomas occur less frequently. In a postmortem study of 173 dogs, golden retrievers and boxers were at increased risk to develop primary intracranial tumours relative to their frequency in the general Hospital population [223]. Brachycephalic breeds have previously been reported to be prone to development of glioma, but in this study only boxers and Boston terriers were more likely to have an astrocytoma, oligodenroglioma, or undifferentiated glioma than another type of primary intracranial neoplasm. Boxers account for nearly 50% of dogs with brain tumours presented for radiotherapy at the Cancer erapy Unit, Cambridge, and appear equally affected by glioma and pituitary macroadenoma (unpublished data).

Multiple Primary Tumours
is paper has highlighted strong breed predispositions to certain types of cancer and breeds that are prone to more than one tumour type. In reviewing the more recent gene expression literature it is apparent that tumour suppressor genes, particularly CDKN2A, CDNK2B, and PTEN, are implicated in the pathogenesis of many canine tumours, but to date there is little evidence to demonstrate that defects in these pathways are inherited in the susceptible breeds. Recently a germline mutation in the mesenchymal-epithelial transition factor (MET) protooncogene was found in approximately 70% of rottweiler dogs, a breed predisposed to several types of cancer [224]. is supports the concept that particular dog breeds may carry germline mutations that contribute to high rates of cancer in a manner similar to heritable, cancer-associated mutations in humans. Inherited defects in tumour suppressor genes have been associated with increased risk of early onset cancer and development of multiple primary tumours in humans, notably TP53 and Li-Fraumeni syndrome [44]. e literature contains a number of case reports of multiple tumours in individual dogs, for example, simultaneous aortic body tumour and pulmonary histiocytic sarcoma in a �atcoated retriever [225], but only recently has a detailed analysis of dogs presenting with multiple distinct types of neoplasia been published [226]. ese represented just 3%, 53 of 1722 dogs presented to the oncology service at Colorado State University Veterinary Medical Centre; although no breed or sex predisposition was apparent, dogs with mast cell tumour, malignant melanoma, and thyroid carcinoma were signi�cantly over-represented.

Conclusions
Most of this review has focused discussion on the breeds of dog associated with an increased risk of developing cancer and breed associations recognised in common tumours. It should be acknowledged that for most forms of cancer the aetiology is likely to be multifactorial and although genetics are important environmental factors such as chemical exposure [227] and hormonal/metabolic factors have been shown to increase the risk of development of certain tumours.
e domestic dog has been bred selectively for many years to accentuate traits that are desirable in the eyes of the breeder. Each dog breed with a speci�c pool of alleles represents a genetic isolate, facilitating the identi�cation of susceptibility alleles in dogs breeds as compared to humans. Pure-bred dogs allow the identi�cation of rare variants in the whole canine population because they have been accidentally selected in a given dog breed, and for multifactorial diseases such as cancer, the impact of environmental exposure can be analysed against a reasonably homogenous genetic background. us the fact that different breeds of dog have different predilections to different forms of cancer is not only interesting but could provide a very important insight into the genetic aetiology of many forms of the disease. is is particularly important in a disease like cancer, where the complex disease phenotypes are likely to have developed from a combination of multiple genetic risk factors, each with relatively weak penetrance. It could be that the genetic architecture of cancer in predisposed breeds such as the Bernese, boxer, and golden retriever is basically the same as in other breeds with other tumours. e difference is that in the former, the risk conferred by one or more predisposing alleles could be higher. Whether the same alleles predispose to various types of cancer in Bernese, boxer, and golden retriever, or whether different loci, segregating in each one of those breeds, are responsible for the different tumour types, is an interesting question which remains to be answered.
Much has been made of the many problems associated with pedigree dogs in the popular and veterinary press [228,229]; it is therefore somewhat ironic that as a result of these problems the pedigree dog provides an ideal model to identify phenotype/genotype relationships relevant to human disease. e sequencing of the canine genome [29] along with the new genomic tools and resources now available for the study of the dog has allowed workers to start to analyse complex diseases such as cancer. As this review has shown, the dog is already proving to be a valuable model for this purpose and further well designed and conducted population based studies into breed-related canine cancers would provide an important platform to take forward future genetic research.

Abbreviations
GWAS: Genome wide association study/ies CNA: Copy number abnormality TCR: T-cell receptor PARR: PCR for antigen receptor rearrangements PCR: Polymerase chain reaction VEGF: Vascular endothelial growth factor CGH: Comparative genomic hybridization SNP: Single nucleotide polymorphism DLA: Dog leucocyte antigen MHC: Major histocompatibility complex ASGC: Anal sac gland carcinoma GRCA: Golden retriever club of america DYAR: Dog years at risk.