Sizing up dogs

Why are dogs of interest to human geneticists? Dog breeds display an extraordinary amount of variation in skeletal structure, including overall size, leg length and width, and variants of skull shape. Breeds thus present geneticists with great opportunities for studying the genetics underlying common traits that have become fixed as part of various breed standards. Many traits were selected because they offered particular advantages to their human owners: dogs have been bred to hunt (sight hounds and scent hounds); herd (border collies and Australian shepherds); race (greyhounds); draft (huskies); and search (bloodhounds and German Shepherds). Many breeds have a morphology that benefits their tasks; for instance, short-legged dogs are ideal for fox and rabbit hunting as their short stature allows them to more easily catch small prey.How much size variation is there in dogs? No species of land mammal displays a greater degree of variation than does the dog, and size is certainly one of their most variable traits. The very smallest breeds, such as the Chihuahua and Pekingese, weigh no more than a couple of pounds, while the very largest, including the Newfoundland and the St. Bernard, can weigh over 200 lbs. Some breeds are short while others are of striking height, such as the Scottish deerhound, Irish Wolfhound and borzoi, who will quietly draw attention by placing his nose on his owner's shoulder.Each of the approximately 165 dog breeds that are recognized by the American Kennel Club, and the multitude of breeds that exist worldwide, shares three features. First, each breed is defined by a carefully chosen set of traits that breeders have selected. Second, each breed is a closed population; in order to be an American Kennel Club member, both parents and grandparents, in turn, had to be registered members of the same breed. Finally, as a result of the overbreeding of popular sires, and the fact that many breeds have gone through population bottlenecks, breeds often exhibit a non-random distribution of alleles. We, and others, have therefore hypothesized that dogs have fewer alleles supporting major traits than in species not subject to artificial breeding such as humans. This makes dogs a unique and advantageous system in which to study complex traits. Issues such as lack of samples or locus complexity are at least partially overcome by comparing the DNA profiles from dog breeds that share versus lack the trait in question.How were body size genes found in dogs? The first studies to map body size loci in the dog focused exclusively on the Portuguese water dog, a breed for which a two-fold difference in body size is allowed by the American Kennel Club. Comparing large and small Portuguese water dogs, the first canine body size locus was mapped using linkage to canine chromosome 15 (CFA15). Interestingly, when 20 small dog breeds (weighing less than 15 lbs) were studied, a selective sweep was observed that corresponded precisely with the insulin like growth factor 1 (IGF1) gene on CFA15. A selective sweep, or region where there is an unusually low level of genetic heterogeneity, indicates a gene under strong selection. The fact that nearly all small breeds studied shared the same haplotype, or pattern of single nucleotide polymorphisms (SNPs), around IGF1 told us that the production of small dogs by exploitation of particular IGF1 variants had likely occurred early in domestication. Interestingly, a role for IGF1 as a human height gene was recently suggested by genome-wide association (GWAS) data from Japanese populations, underscoring the likely similarity between genetic mechanisms controlling body size in dogs and humans.How many genes control body size in dogs? While data on the Portuguese water dog implicated at least a dozen loci as being involved in controlling canine body size, it was not until 2008, when hundreds of dogs from dozens of breeds were genotyped, that we found out for sure that there are about 6–10 major loci controlling body size in dogs. Because the group who did this work did not have measurements for each dog they genotyped, they relied instead on the breed standard as the ‘phenotype’ for their genetic studies, thus coining the term ‘breed stereotype’. This turns out to be a powerful concept as it shows that, for breed-fixed traits such as leg length, skull shape, coat length, leg length, and so on, specific measurements are not needed as long as pure-bred dogs are genotyped.Quantitative trait loci (QTLs) that contribute to canine height were identified on CFA7, 10, 15 and 34. The exact same loci, plus one on CFA9, were identified when weight was used as a surrogate for size (Figure 1Figure 1). Several of the identified loci were close to, or within, striking candidate genes, such as HMGA2, encoding a high-mobility group protein, on CFA10; SMAD2 and NPR2 on CFA7, SOCS2 and IGF1 on CFA15; and IGF2BP2 on CFA34. HMGA2 is particularly interesting, as it has been shown to play a role in human height as well as associated phenotypes, including bone mineral density and small gestational size. These findings have been validated in subsequent studies, and additional loci in the proximity of candidate genes such as growth hormone receptor (GHR) have been identified. Mutations in GHR in humans can have profound effects on height; however, functional investigations of GHR are complicated as many of the effects of GHR are mediated by IGF1.Figure 1Size variation among dog breeds.Average height and weight of breeds included in the CanMap project. Points corresponding to pictured breeds are indicated with a diamond shape and labeled with the breed name. The solid line is a linear fit of weight to height; the dashed line shows a locally-weighted fit with a span of 0.7. Photos courtesy of Mary Bloom, American Kennel Club.View Large Image | View Hi-Res Image | Download PowerPoint SlideWhat other morphologic traits have been mapped in dogs? The work on body size became the impetus for us to create ‘CanMap’, a publically available data set that can be used to map a multitude of breed-associated fixed traits. The data set comprises over 60,000 genotypes on each of 915 dogs from 80 breeds and nearly 100 wild canids. Using genome-wide analysis, we observed 51 loci that segregate with 57 traits. Even the genetics of ear position (up or floppy) has been mapped using this data set. Morphologic traits for which genes or mutations have now been found include asymmetrical dwarfism, a breed-defining trait for 20 breeds, including the Corgi, Basset hound and Dachshund: fur length; fur curl; white spotting, wrinkled skin of the Shar-pei; the ridge of the Rhodesian ridgeback; facial fur; and so on. In contrast to human studies, we found that a small number of QTLs (≤3) explain most of phenotypic variation for most of the traits studied, likely reflecting the genetic isolation of breeds.Are human and dog body size controlled by the same genes? The genetics of human height has been studied extensively, with over 180 single SNPS thought to contribute to height variation in European ancestry populations. Still other loci have been defined by studies of Koreans, Japanese, Africans and African Americans. Cumulatively, however, these loci are believed to explain no more than 15% of the variance in human height.Is the genetics of height and body size therefore intrinsically intractable? No, we don't believe it is. Remember that dog breeds exhibit nearly 20-fold variation in skeletal size and thus offer a chance to determine not only the vocabulary of genes controlling body size, but also the hierarchy of these genes. We believe that the GWAS of dogs have already identified the major genes that determine significant size differences in and across mammals. If each of these large-effect genes causes a substantial increase in size, a number of additional loci must exist to exert fine control within each size range. To find those genes, GWAS could be done using small, medium, and large breeds for which significant intrabreed size variation is tolerated. We hypothesize this will highlight additional body size genes and, ultimately, the functional hierarchy of the variants.Support for this idea comes from the recent work of Vaysse et al. (2011), who mapped a similar set of breed stereotypes as was done in CanMap using 46 breeds. Not only did they validate the HMGA and IGF1 findings, but they also demonstrated a steady decline in retention of one IGF1 allele in relation to size of the breed. Even more striking were the data on HMGA2, in which one allele appears at low frequencies in all breeds apart from a number of very small breeds, where it appeared to be close to fixation. Thus, combinations of alleles at a small number of loci appear to control body size in dogs and functional studies of those same loci are now needed to understand their role in humans.Didn't Darwin have something to say on all this? In his musings on genetic variation, Charles Darwin wrote “…whatever you yield in regard to the dog, you will have to concede to every variable species of plant or animal (wild or cultivated), man included”. He thus boldly put forth the hypothesis that underpins our studies of body size. In the future, correlating the results from breed-specific studies of humans with those from GWAS of humans will validate or falsify Darwin's postulate. If, as seems likely, the data continue to support his view, it will open the door to molecular genetic study of the many other biological traits that are shared between domestic dogs and their human companions.