In order to determine the degree to which dogs could be assigned correctly to their breed group, my lab utilized data from 96 microsatellite markers spanning all the dog's 38 autosomes in a set of 414 dogs representing 85 breeds. We found, first, that nearly all individual dogs were assigned correctly into their breed group when we used a set of statistical tools called clustering algorithms, which look for similarities in the frequency and distribution of alleles between individuals. The exceptions largely included six sets of closely related breed pairs (for example whippet-greyhound and mastiff-bullmastiff) that could only be assigned to their respective breeds when considered in isolation from other breeds.

We also showed that the genetic variation between dog breeds is much greater than the variation within breeds. Between-breed variation is estimated at 27.5 percent. By comparison, genetic variation between human populations is only 5.4 percent. Thus the concept of a dog breed is very real and can be defined not only by the dog's appearance but genetically as well.

A second part of the study used an assignment test to determine whether we could correctly identify each dog's breed by its genetic profile alone. In a blinded study, where the computer program did not know what data set came from which breed, 99 percent of dogs were correctly assigned to their breed based on their DNA profile alone.

To determine the ancestral relationship between breeds, Heidi Parker from my lab used data from the same set of dogs and sought to determine, ideally, which dog breeds were most closely related to one another. To do this we utilized a computer program called structure, which was developed by Jonathan Pritchard at the University of Chicago and his colleagues. The program identifies genetically distinct subpopulations within a group based on patterns of allele frequencies, presumably from a shared ancestral pool.

The structure analysis initially ordered the 85 breeds into four clusters, generating a new canine classification system. Cluster 1 comprised dogs of Asian and African origin—thought to be older lineages—as well as gray wolves. Cluster 2 included largely mastiff-type dogs with big, boxy heads and large, sturdy bodies. The third and fourth clusters split a group of herding dogs and sight hounds away from the general population of modern hunting dogs, the latter of which includes terriers, hounds and gun dogs. As more dog breeds have been added to the study, additional groupings have emerged.

These data are extremely useful for disease-gene mapping studies. In some cases, dogs from breeds that are members of the same cluster can be analyzed simultaneously to increase the statistical power of the study. This will not only aid in the identification of genomic regions in which the disease gene lies, but will also assist in "fine mapping" studies which aim to reduce the region of DNA linkage to a manageable size of about 1 million bases. Once a region is well defined, we can begin to select candidate genes for mutation testing.

The Shape of Things

Our research group, along with others, has been interested for several years in identifying genes that define the differences in body size, shape and appearance between breeds. Dog breeds vary not only in overall body size, but also in leg length, head shape and many other body features, all of which are controlled at least in part at the genetic level. The amount of morphologic variation observed in the dog is reported to surpass that of all living land mammals. The first important molecular study aimed at understanding the genetics of canine morphology was done at the University of Utah and led by Gordon Lark and Kevin Chase. The project, termed the Georgie Project in memory of a favored dog, focused on the Portuguese water dog, which is ideal for this type of study because it derives from a small number of founders, largely from two kennels, that came to the United States in the early 1950s. The breed standard permits a significant amount of variation in body size compared with other breeds. The community supporting the project is composed of highly motivated owners and breeders who have sought to improve the health of the breed through collaboration with scientists. To date, the project has collected DNA from more than 1,000 dogs and has completed a genome-wide scan using more than 500 microsatellite markers on nearly 500 dogs. In addition to family history and medical data, more than 90 measurements have been collected for nearly 500 animals. These were derived from a set of five x-rays taken at the time of initial sample collection. Analysis of these metrics led to the development of four primary principal components (PCs), sets of correlated traits that define Portuguese water dog morphology. It is important to keep in mind that PCs are not genes but traits, and as such, they are susceptible to genetic analysis. Analysis of the genome scan data and four PCs initially highlighted 44 putative quantitative trait loci (QTLs) on 22 chromosomes that are important for heritable skeletal phenotypes in the Portuguese water dog. QTLs derive from complicated statistical analysis and indicate locations in the genome that contribute coordinately to a particular trait. Of particular interest to us was a locus on canine chromosome 15 (CFA15) that showed a strong association with overall body size. Although this was only one of seven loci hypothesized to play a role in body size in the dog, we chose it as an initial focus because of the strength of the effect and the proximity to a compelling candidate gene.

Figure 4. Based on patterns in the frequency of alleles (variations in the genome), the author’s group used a computer program to identify genetically distinct subpopulations within a group of 85 dog breeds. The breeds were divided into four groups based on the dominance of a particular cluster of alleles (vertical bar). The first group comprises dogs of Asian and African origin, thought to be older lineages (yellow). The second cluster contains herding dogs and sight hounds (green), whereas the third includes modern hunting dogs such as terriers, hounds and gun dogs (orange). The fourth cluster includes largely mastiff-type dogs with big, boxy heads and large, sturdy bodies (blue). Dots next to breed names correspond, from top to bottom, with breeds shown at right. Adapted by Barbara Aulicino and Linda Huff from Parker et al. 2004.

To find the gene on CFA15, we searched for SNPs in a 15 million-base-pair region and then genotyped the resulting set of markers on all the Portuguese water dogs for which size information was available. The distribution of these markers displayed a single peak close to the insulin-like growth factor-1 gene (IGF1), which is known to influence body size in humans and mice. We investigated IGF1 in detail and showed that 96 percent of Portuguese water dog chromosomes carry one of just two patterns of alleles, which are termed haplotypes. The haplotype associated with small dogs was termed "B" and the one associated with large dogs "I." Portuguese water dogs homozygous for haplotype B—that is, dogs that have the B pattern on both chromosomes—have the smallest median skeletal size, whereas dogs homozygous for I are largest. Dogs that are heterozygous—that is, those with a different pattern on each chromosome—fall between.

To study the presumably more general role of IGF1 in size differentiation among breeds, we surveyed genetic variation associated with 122 SNPs, spanning the relevant 34 million- to 49 million-base-pair interval of chromosome 15 in 353 dogs representing 14 small breeds and 9 giant breeds. Several lines of evidence pointed to IGF1 as the gene likely to account for small body size in the dogs.

Figure 5. When characterizing a dog breed by its physical traits, one metric often used is the average height of the dog at the shoulder compared to the average weight of males of the breed. This ratio represents a trade-off between speed and strength: Dogs that have a higher height-to-weight ratio (blue) tend to have long, thin legs, whereas heavier dogs have thicker bones (green). This is one of the principal components, or sets of correlated traits, helpful in understanding the genetics controlling morphologic variation between dog breeds. Adapted by Barbara Aulicino and Linda Huff from Parker and Ostrander 2005.

Most notably, we observed a dramatic reduction in heterozygosity in small breeds over the IGF1 gene. These results demonstrate the presence of a selective sweep in this region, showing that IGF1 has been under tight selection by breeders seeking to create ever smaller dogs. In addition, the dominance of a single unique haplotype in our panel of many unrelated small dog breeds, together with its near absence in giant breeds, suggests that the mutation is ancient and likely evolved early in the history of domestic dogs.

Sexual Dimorphism The Georgie Project is remarkable for the number of putative loci that have been discovered by the initial analysis. In addition to loci for head shape, body size, leg length and a host of other traits, loci have also been described that reportedly control differences in size between the sexes, so-called sexual dimorphism. Sexual dimorphism is observed in almost all mammals including, of course, dogs. The mechanisms for maintaining sexualdimorphism are not well understood. It has been shown that the Sry locus on the Y chromosome plays an important role in sex determination and dimorphism, but this is clearly only a small part of the story. The study of the Portuguese water dog has filled in some additional pieces of this interesting puzzle. This vignette has its roots in the original observation that a locus on chromosome 15, which may or may not be IGF1, interacts with other genes to make males larger and females smaller. Figure 6. Portuguese water dogs were x-rayed to collect more than 90 physical measurements, which led to the development of four principal components that define the breed’s morphology. The areas measured included the hind limbs (left), pelvis (middle) and skull (right). Measurements included the lengths of the femur (a), tibia (b), foot (c) and skullbase (h), as well as the widths of the ilium (d), hip (e), trochanter (f) and skull (g). These measurements were used in a genetic analysis of size differences between male and female dogs (Figure 7). Photographs courtesy of Kevin Chase.   On average, female Portuguese water dogs are 15 percent smaller then males. Chase, Lark and their colleagues observed that in females, a particular haplotype is dominant for small body size. In males, a different set of variants (another unique haplotype) associated with large overall body size is dominant. The locus on CFA15 interacts with another locus on the X chromosome that is known to escape inactivation, meaning that both copies of the genes in this region are turned on (in most locations on the X, only one copy is active). Females who are homozygous at the X-chromosome locus and who are also homozygous for the large-size CFA15 haplotype are, on average, as large as large males. However, all females that are heterozygous at the X-chromosome marker are small, regardless of their CFA15 genotype. This result suggests several scenarios for how genes interact to affect major complex traits, such as body size, and suggests a mechanism for the evolution of sexual dimorphism. Two observations from the study must be accounted for in the development of any model to explain canine sexual dimorphism. The explanation must include a discussion of the reversal of dominant haplotypes between males and females associated with CFA15 locus as well as an explanation for the interaction between the CFA15 and X-chromosome loci. To address the first question, Chase and his colleagues propose the existence of another sex-specific factor. For example, the CFA15 locus might contain two distinct genes associated with two haplotypes; the so-called Ahaplotype acts in both males and females to upregulate size, while the B haplotype and its associated allelesdo not upregulate size but rather contain another gene that suppresses the up-regulator. Figure 7. Like almost all mammals, Portuguese water dogs show sexual dimorphism—size differences between males and females. Analysis indicates that a locus on chromosome 15 interacts with another locus on the X chromosome to make males larger and females smaller. One haplotype or pattern of alleles (A) is dominant for males and is associated with large overall body size. A different haplotype (B) is dominant in females and associated with small body size. Following the rules of classical inheritance, homozygous (AA) males, or males with the A pattern on both sets of their chromosomes, are largest and BB females smallest. These data support an evolutionary hypothesis in which females become smaller as the result of natural selection for optimal size, through the inhibition of major genes that enhance growth. Data courtesy of Kevin Chase. Adapted by Barbara Aulicino from Chase et al. 2005. The second phenomenon, heterozygote-specific interaction, could be explained by arguing that the activation of haplotype A's critical upregulator gene requires interaction with a protein produced by the X chromosome. The data of Chase, Lark and their colleagues are consistent with predictions made in the early 1980s that sexual dimorphism evolves because females secondarily become smaller than males as a result of naturalselection for optimal size. Reduction of female size relative to that of males takes place, according to this hypothesis, through an inhibition of major genes that enhance growth, such as the locus on CFA15.

Remaining Selective

The advances of the past three years in canine genetics have been enormous. The dog genome has been mapped and sequenced. A host of disease loci have been mapped, and in many cases the underlying mutations identified. Our understanding of how dog breeds relate to one another is beginning to develop, and we have a fundamental understanding of the organization of the canine genome. The issue of complex traits is no longer off-limits. We have begun to understand the genetic portfolio that leads to variation in body size and shape, and even some performance-associated behaviors.

Certainly the next few years will bring an explosion of disease-gene mapping. The genetics of canine cancer, heart disease, hip dysplasia, vision and hearing anomalies have all been areas of intense study, and investigators working on these problems are poised to take advantage of the recent advances described here. Whole-genome association studies are likely to replace family-based linkage studies as a way of finding genes associated with not only disease susceptibility and progression, but morphology and behavior as well.

What will the companion-animal and scientific communities do with this new information? It is certainly hoped that the disease-gene mapping will lead to the production of genetic tests and more thoughtful breeding programs associated with healthier, more long-lived dogs. It will be easier to select for particular physical traits such as body size or coat color, not only because we understand the underlying genetic pathways, but because genetic tests are likely to be made available as quickly as results are published. Finally, canine geneticists will finally have a chance to develop an understanding of the genes that cause both breed-specific behaviors (why do pointers point and herders herd?).

What is far less clear is whether we will come to understand what makes the domestic dog unique to us among all the animals in the mammalian world. We have domesticated dogs to the point that they display loyalty, friendship and companionship. We seek their company and approval and bring them into our homes, often as equal members of our family. We rejoice in their victories and mourn their deaths, often as we celebrate or mourn our own children. Is the genetics that defines this relationship within the dog, within ourselves, or both? None of the studies proposed are likely to answer that question, and perhaps that is okay. The comparative-genome projects of humans and dogs were designed to bring about an understanding of our similarities and differences. Perhaps scientists will have to be satisfied to understand that much, and leave as a mystery the genetic basis of approval, adoration and loyalty. At least for me and my dog, it's enough.