Morphological QTLs were identified by association across breeds using breed stereotypes (fixed phenotypes) based on data described in our previous publication (Jones et al. 2008
). Unlike within-breed mapping of segregant phenotypes, across-breed mapping identifies loci within a much smaller linkage disequilibrium (LD) distance 400–500 kb (Jones et al. 2008
). Thus, relevant candidate genes are identified from a much smaller genome region. In all, 30 morphological QTLs were identified. A subset of these, listed in , contained relevant candidate genes. These represent immediate targets for fine mapping and/or validation in specific breeds (e.g., hair length and FGF5 in Dachshunds and Corgis [Housley and Venta 2006
Details of QTLs for size-related traits and other aspects of morphology
To test the potential of the technique, we have analyzed several behavioral phenotypes. Using an ethological approach, we asked a dog trainer with more than 30 years experience training and judging various dog breeds to score all the breeds for boldness, herding, pointing, and trainability. In all, 10 loci were identified 5 of which contained the suggestive candidate genes presented in . Four of these candidate genes relate to various loci affecting the nervous system and/or have been implicated in behavior: MC2R
on CFA 1 (27381939 bp) is a melanocortin receptor, and C18orf1
(27572327 bp) has been implicated in schizophrenia. DRD1
, on CFA 4 (40743436 bp), encodes a dopamine subtype receptor. CNIH
, on CFA 8 (33396000), has been implicated in cranial nerve development. Finally, PCDH9
, on CFA 22 (24273482 bp), encodes a protein localized to synaptic junctions and believed to be involved in specific neural connections and signal transduction. Although the behaviors involved are poorly defined, the presence of major candidate genes appropriate to behavior is encouraging. The apparent exception is IGF1
that might be expected to affect boldness on the basis of size (large dogs bold, small timid). However, the locus on CFA 15 does not appear to be related to size, as approximately equal numbers of large and small breeds were found to be bold (Jones et al. 2008
), and boldness and size were not correlated (r
QTLs associated with behavior
and present previous across-breed mapping results (Jones et al. 2008
) for morphology and behavior. Across-breed mapping successfully overlapped previously identified loci for the traits analyzed and suggested a number of new candidate loci appropriate to the stereotypic phenotype in question. In all cases, we were able to rule out spurious effects due to a simple measure of between-breed relatedness (i.e., average breed genome similarity [Jones et al. 2008
]). However, the genotyping database that we used only interrogated about 25% of the genome. Thus, many potential loci will not have been identified. More importantly, false positives due to nonsystenic LD and more complex breed relationships cannot be ruled out without complete genome coverage.
Many of the limitations of the current data set will be addressed by an effort, termed CanMap, (http://www.sciencemag.org/cgi/content/full/sci;317/5845/1668
) currently being completed. The goal of the CanMap project is to produce dense SNP profiles of a dozen dogs from each of nearly a hundred breeds. Unrelated dogs from each breed (deemed unrelated if they shared no common grandparents) will have been genotyped to provide much more complete genome coverage.
Despite the power of across-breed mapping to identify genomic regions of interest, the potential for false positives, whereby causative regions of the genome cannot be distinguished from noncausative, will always necessitate validation using within-breed segregation analysis. Most often, across-breed mapping identifies markers that tend to be near or at fixation (homozygous) in breeds with the associated phenotype. Breeds in which the phenotype is still segregating will not contribute to the power of QTL identification. However, they will provide a resource in which the association can be validated using within-breed segregation analysis. Such breeds are readily identified from the across-breed SNP genotyping database (e.g., alelle frequencies near 0.5). It should be possible now to validate the most significant (P
≤0.001) loci in and using breeds in which the implicated SNPs are segregating (e.g., the locus on CFA 32 for short coat () was identified by segregation analysis using Dachshunds or Corgis (Housley and Venta 2006
Another facet of across-breed mapping is the ability to investigate phenotypes that vary between breeds but do not appear to segregate within any breed. One such trait is longevity. It has long been known that, on average, dogs from small breeds live longer than those from large breeds (Egenvall et al. 2005). However, a detailed study of longevity and size within breeds has failed to provide any evidence for a similar relationship of size and longevity. Indeed, in that study, there was a trend suggesting that within-breed larger dogs may live longer (Galis et al. 2007
Using mean breed longevity (Jones et al. 2008
), across-breed mapping identified the 2 major size loci on CFA 7 and CFA 15 as highly significant loci regulating longevity (). Given the inverse correlation of size and longevity, this may seem like a trivial result. However, the difference in significance between size and age of death in the IGF1 region of CFA 15 suggests that more than just an effect on size is involved. The fact that within breeds increasing size fails to decrease longevity also suggests that size per se is not a causal factor in altering longevity. Moreover, the association of IGF1 with breed longevity is not unexpected considering that it has been implicated in regulating longevity in a number of organisms (Kenyon 2001
; Bartke 2005
QTLs associated with age of death (AOD) and the probability that size is also associated with that SNP. Trait, chromosome (CFA), CanFam2 nucleotide position on the chromosome, log P, and genome-wide significance threshold (GWT) are as in Table 1
One explanation may be that introducing loci that drastically change growth rate may detrimentally perturb physiological balance, exposing the system to various senescent pathologies without the compensating mutations that normally would arise in the course of a prolonged evolutionary process. More generally, strong selection for morphological or behavioral traits has generated a variety of genome configurations. Some of these, though functional through reproductive maturity, may increase risk of complex disease (Egenvall 2005). Across-breed mapping can be used to determine the impact of the breed-fixed genome configurations on disease risk, provided that accurate estimates of breed disease frequencies are available.
As a first test of this hypothesis, we have analyzed the association of IGF1 with variation in the frequency of certain diseases. It is known that breed size is correlated with the frequency of certain orthopedic diseases found in many dog breeds. Recently, pancreatitis has been correlated with the concentration of IGF1 in rodent serum, and it has been possible to decrease the incidence of pancreatitis by increasing serum levels of IGF1 (Warzecha et al. 2003
; Dembinski et al. 2006
). We therefore expected that across-breed mapping might associate IGF1
haplotypes with such diseases. We obtained disease frequencies () for hip dysplasia, patella luxation, and pancreatitis from the VMDB (see Methods). presents the frequency of these diseases in small and large dog breeds. Here, small and large breeds are defined by the frequency of an SNP allele on CFA 15 (44228468 bp) that is part of the nucleotide sequence common to all small dogs (Sutter et al. 2007
). For all 3 diseases, the association with IGF1
is very significant (genome-wide significance P
< 0.01). These associations were not the result of genotypic relatedness between breeds as they became more significant after correcting for breed relationships. This implies that selection for size has influenced the frequency of these diseases.
Figure 1 Disease frequencies are graphed as cumulative distributions (x axis, log scale) for hip dysplasia (open circles), patella luxation (closed circles), and pancreatitis (closed triangles). Records from 22 veterinary hospitals were obtained from the VMDB (more ...)
Figure 2 Frequencies of hip dysplasia, patella luxation, and pancreatitis in large and small breeds of dogs. Frequencies of these diseases () were divided between the 2 groups of breeds. Cumulative distributions are shown for each group. We defined small (more ...)
Across-breed mapping combined with within-breed mapping is a powerful approach for exploration of the wide range of behavior, morphology, and disease frequencies observed across the spectrum of dog breeds. The CanMap project promises the characterization of the allele frequencies for many breeds, thus creating an invaluable enduring fixed resource. As we obtain a more detailed picture of the complex relationships between breeds, across-breed mapping will become more sensitive and less prone to false positives. Characterization of breed phenotypes can be applied to the CanMap resource as an easy starting point from which to carry out genetic analysis of many complex phenotypes. Results from across-breed mapping will provide a narrower range of focus for validation in specific breeds. Different breeds may be fixed for different haplotypes affecting a complex phenotype. These provide snapshots of genotypes, each of which display a simplified aspect of the interactive network. Integration of these should become a powerful tool for understanding the network as a whole.