Genetic Variants Affecting Skeletal Morphology in Domestic Dogs

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Trends in Genetics

Review

Genetic Variants Affecting Skeletal Morphology


in Domestic Dogs
Danika L. Bannasch,1,* Christine F. Baes,2,3 and Tosso Leeb3

Purebred dog breeds provide a powerful resource for the discovery of genetic Highlights
variants affecting skeletal morphology. Domesticated and subsequently pure- Extreme differentiation in skeletal mor-
bred dogs have undergone strong artificial selection for a broad range of skeletal phology characterizes purebred dog
breeds. This has allowed the utilization
variation, which include both the size and shapes of their bones. While the phe-
of breed-average measurements or
notypic variation between breeds is high, within-breed morphological variation is breed characteristics as phenotypes in
typically low. Approaches for defining genetic variants associated with canine subsequent genetic analyses.
morphology include quantitative within-breed analyses, as well as across-
Across-breed association analysis has
breed analyses, using breed standards as proxies for individual measurements. positively identified a relatively small num-
The ability to identify variants across the genomes of individual dogs can now ber of loci (b20) for stereotypical dog
be paired with precise measures of morphological variation to define the genetic breed body size.
interactions and the phenotypic effect of variants on skeletal morphology.
In outbred populations and within breeds
there appear to be additional body size
Dogs as Models of Skeletal Morphological Variation variants.
No other mammal, domestic or wild, displays the extreme skeletal diversity that can be observed
In the face of high levels of inbreeding
in the dog. Dogs can vary in weight from 2 to 100 kg and in height from 15–80 cm at the shoulder.
and very specific breed criteria for body
In addition to overall size, they can also vary in long-bone length and shape and vertebral and size, less than 50% of the body size loci
cranial morphology. The vast assortment of dog breed shapes has been created by strong and are fixed within breeds.
continuous artificial selection, leading to the fixation of extreme phenotypes within hundreds of
Whole-genome sequence-based asso-
distinct closed breeding populations. The resulting population structure has become a powerful ciation approaches with millions of vari-
tool for defining chromosomal regions underlying skeletal variation. The decrease in sequencing ants and modest numbers of animals
costs and the ability to identify causal variants from large-scale whole-genome sequencing facilitate the identification of new loci
(WGS) (see Glossary) data provide new opportunities to define the variants responsible for and the causative variants underlying
canine skeletal morphology.
the morphological diversity of dogs. Quantitative phenotyping within individuals combined with
segregation analysis will allow the identification of specific functional variants and their precise
genetic contribution to diverse skeletal morphologies. Here, we review the unique characteristics
of dog breeds and their population structure, classification based on skeletal phenotypes, and
how genetic approaches can be used to link specific genes to their morphological variation.

Population Structure of Purebred Dogs


Dogs were domesticated from wolves 15 000–30 000 years ago and are recognized as the first
1
Department of Population Health and
domesticated animal [1,2]. Clear distinctions in morphological phenotypes are evident from burial
Reproduction, School of Veterinary
sites and art 5000–10 000 years before present [3]. Current domestic dogs originated from Medicine, University of California Davis,
random breeding or purpose-bred populations in which selection of their parents is under Davis, CA 95616, USA
2
Department of Animal Biosciences,
human control [4]. A subset of dogs, belonging to closed breeding populations or breeds, are
University of Guelph, Guelph, ON N1G
referred to as purebred dogs. Around the world, purebred dog registries record the pedigree 2W1, Canada
3
of each dog and allow registration of puppies produced from these dogs. Most of these breeds Institute of Genetics, Vetsuisse Faculty,
University of Bern, 3001 Bern,
were developed from small numbers of animals and have been maintained as closed breeding
Switzerland
populations for the past 100–150 years [5,6]. Each breed has a standard that often includes
size, weight, coat type, coat color, and other unique morphologies that are ideal or disqualifying.
Dogs are subjectively judged based on these standards to continue selection toward the ideal *Correspondence:
dog described in the standard. dlbannasch@ucdavis.edu (D.L. Bannasch).

598 Trends in Genetics, August 2020, Vol. 36, No. 8 https://doi.org/10.1016/j.tig.2020.05.005


© 2020 Elsevier Ltd. All rights reserved.
Trends in Genetics

The population history of purebred dogs and strong selection practices have resulted in relatively Glossary
high levels of inbreeding [7,8]. However, the individual population history of each breed is unique Chiari-like malformation:
and the numbers of founders, the effective population size (Ne), and the number of genera- developmental malformation of the skull
tions since closing of the stud books differ among breeds [9–11]. Based on allele sharing, mod- and craniocervical vertebrae
characterized by a conformational
ern dog breeds can be divided into 23 clades of more closely related breeds. The complexity in
change, overcrowding of the brain, and
how different breeds were developed and have admixed is reflected in genome-wide haplotype subsequent partial herniation of the
sharing among them [11]. Studies to evaluate the population dynamics of purebred dogs using a cerebellum through the foramen
combination of pedigree analysis, single-nucleotide variant (SNV) genotyping data and WGS magnum.
Computed tomography (CT) scan:
demonstrate that the average percentage of the genome shared by individuals within a breed makes use of computer-processed
varies between 2% (Chihuahua) and 34% (bull terrier). Effective population sizes of breeds in combinations of many X-ray
the present day are between 50 and 250 while genomic equivalents vary from 2.2 to 16.1 measurements taken from different
[10]. The unique population history of purebred dogs has resulted in a complex genetic relation- angles to produce cross-sectional
images of specific areas. This allows an
ship between the breeds. exact 3D reconstruction of skeletal
morphology.
Morphological Phenotypes Effective population size (Ne): the
number of breeding individuals in an
Purpose-bred dogs were developed to perform specific functions [12]. Some dogs were used to pull
idealized population that would show
or carry large loads and therefore were selected for larger bones and overall size (Figure 1A–C) [13]. the same amount of dispersion of allele
frequencies under random genetic drift
or the same amount of inbreeding as the
population under consideration.
(A) (B) Founders: the animals that were used
to start a breed.
Genome-wide association study
(GWAS): an approach used in genetic
research to associate specific alleles or
genotypes with particular phenotypes.
The allele or genotype frequency
differences for many genetic variants are
tested for significant differences
between groups.
Genomic equivalent: a founder
genomic equivalent is the number of
founders that would produce a
population with the same diversity of
founder alleles as the pedigree

(C) (D) population assuming all founders


contributed equally to each generation of
descendants.
Growth hormone axis: the hormone
system in the body that controls growth.
Haplotype: a group of alleles of genes
that reside on the same chromosome
and are inherited together.
Inbreeding: the mating of individuals
that are closely related through common
ancestry.
Linkage disequilibrium: refers to the
non-random association of alleles at two
or more loci in a general population.
When alleles are in linkage
disequilibrium, haplotypes do not occur
Trends in Genetics at the expected frequencies.
Magnetic resonance imaging (MRI):
Figure 1. Morphological Diversity within Dogs. (A) The Saint Bernard and the Chihuahua represent the largest and uses a strong magnetic field and radio
smallest of the dog breeds. (B) From left to right, the humerus from a wolf, Saint Bernard, German Shepherd Dog, Border
waves to create detailed 3D images of
Terrier, Dachshund, and Chihuahua. Bar, 10 cm. (C) The articulated skeletons from a Saint Bernard and a Chihuahua.
the organs and tissues.
(D) Skulls from various dog breeds. Top row: Saint Bernard, wolf, Borzoi. Bottom row: Border Terrier, Cavalier King Charles
Principal component (PC): PC
Spaniel, Pug, Scottish Terrier, Chihuahua. Bar, 10 cm. Photographs from skeletons and bones courtesy of Naturhistorisches
analysis is a statistical procedure that
Museum Bern/Lisa Schaeublin.

Trends in Genetics, August 2020, Vol. 36, No. 8 599


Trends in Genetics

Some dogs were selected for speed to hunt game and thus have longer legs and leaner builds uses an orthogonal transformation to
convert a set of observations of possibly
[13,14]. In addition to these functional purposes, some breeds were developed solely as
correlated variables into a set of values of
companion animals [13]. In such cases, morphology was under less strict constraints, allowing linearly uncorrelated variables called
further extremes to be selected as desired. From toy breeds to giant breeds, body size differ- PCs.
ences within the domestic dog can vary up to 50-fold (Figures 1A,C and 2). In addition to having Purebred dogs: a subset of dogs
belonging to closed breeding
proportions reminiscent of a wolf, from which dogs were domesticated, many breeds have populations that share morphological
shorter legs and thicker bodies while others have more slender builds (Figures 1B and 2). and behavioral traits that breed true.
The tremendous diversity in body size across breeds is further illustrated by the development Quantitative trait locus (QTL): a
of growth curves for puppies used in veterinary practice. At least seven different body weight region of DNA controlling phenotypic
variation of a quantitative trait.
categories and therefore unique growth curves were necessary to capture the vast differences Quantitative traits are typically controlled
in size among common dog breeds [15]. by many QTLs and the environment.
Radiograph: radiologists’ preferred
term for the 2D static image generated
During breed development, disproportionately short limbs were desirable for a number of
following the passage of X-rays through
reasons. Shorter-legged dogs were slower and people on foot found them easier to use for an animal.
hunting, leading to short-legged scent hounds [6,13,14]. Dogs with short legs were also able to Retrogene: an RNA transcript reverse
follow animals into small holes and burrows, hence many terriers have shorter legs (Figure 1B) transcribed and inserted somewhere
else in the gene resulting in a functional
[6,13]. Unsurprisingly, skeletal morphology and leg length have a strong effect on speed.
copy of the original gene.
The American Kennel Club (AKC) ranks the top dogs by breed over a 100-yard distance Segregation: the statistical analysis of
(https://www.apps.akc.org/apps/fastcat_ranking). Sight hounds like whippets and greyhounds Mendelian ratios in families.
Selective sweep: the reduction or
elimination of variation among the
nucleotides flanking a functional variant
in DNA. It results from a beneficial allele
having recently reached fixation due to
strong positive selection.
Stud book: registry containing the
pedigree information of purebred
animals.
Wicket: a device to measure the
shoulder height of dogs.
Whole-genome sequencing (WGS):
the process of determining the complete
DNA sequence of an organism’s
genome at a single time.

Trends in Genetics

Figure 2. Height × Weight Comparisons across Dog Breeds. American Kennel Club breed-standard maximal weight
for male dogs (x-axis) plotted against maximal height (y-axis). Breeds with longer, more slender limbs and bodies than the
wolf (gray triangle) or the proto-dog represented by the dingo (gray circle) are rare. Breeds with greater proportional mass
are much more common. The Chihuahua and Saint Bernard are colored red for comparison.

600 Trends in Genetics, August 2020, Vol. 36, No. 8


Trends in Genetics

can run at 56 km/h, while mastiff-type dogs run at 35 km/h and disproportionately short-legged
dogs like the dachshund run at only 27 km/h.

Head shape varies across dog breeds and across the size range of breeds (Figure 1D). Accurate
measurement of canine skulls has demonstrated that variation within dog breeds is greater than
that occurring in the family Canidae or within the order Carnivora [16]. Different skull types are
predisposed to different diseases and clinical presentations [17,18]. Dog skulls are classified
into three major groups based on skull proportions; brachycephalic breeds have short,
wide skulls, dolichocephalic have long, narrow heads, and mesaticephalic are in between. The
morphological differences in skull shape form during development, as new-born puppies can
already be classified into the major groups [19]. Length-to-width comparisons have been used
to place dog breeds into these categories, but in reality there are many breeds whose skull
shapes bridge the categories [20–24]. Changes to skull shape have been documented over
time, demonstrating the profound artificial selective pressure that can be applied within breeds
by dog breeders [25,26].

Variations in craniofacial morphology were originally selected for functional purposes. Bite force
was shown to be a function of skull size and shape, with larger and brachycephalic (broad
headed) skulls having greater bite force [27]. Dolichocephaly (narrow head) exists within breeds
that were selected as sight hounds; retinal ganglion cell distribution is highly correlated with
nose length in dogs [28]. Weight-pull competition results also indicate that brachycephalic
shaped dogs are significantly stronger than more dolichocephalic shaped dogs, even when con-
trolling for body mass [29]. Since companion dogs were not bred for a specific function, there is
more flexibility in their skull morphology and pure aesthetics can be involved. Subsequently, the
greatest diversity in skull shape was observed in companion breeds [16].

While the reasons for morphological variation in modern-day dog breeds are numerous, what
remains is a valuable group of animals with profound differences in inherited morphology. This
repository of genetic variation has been maintained in the form of purebred dog breeds for the
past 150 years and can facilitate research on skeletal development.

Phenotype Measurements
Of the different methods to measure skeletal phenotypes, body weight has proved to be one
of the most quantitative and convenient, as it is obtained during veterinary visits. Height measure-
ments are more challenging since the heights obtained using a wicket may vary and height
cannot be assessed well using CT scans or radiographs since dogs are quadrupeds and can-
not be weight bearing using these techniques. Jeanes et al. found a correlation of 0.58 between
different height measuring methods and 0.79 between repeated measures of the same dog [30].
However, direct measurements from disarticulated bones from skeletal preparations are highly
accurate in comparison [31–33]. Differences in the cortical density or cortical/medullary ratio of
specific bones and bone mineral density have not been utilized in genetic studies in dogs to
date but can be obtained from CT scans, and breed differences have been observed [34,35].
Genes identified to cause alterations in bone mineral density are relevant to such analyses in
human medicine since osteoporosis is a significant health concern [36]. When multiple measure-
ments are taken from an individual, principal component (PC) analysis can be performed to
evaluate correlations across measurements. These PC values can subsequently be used as
phenotypes for genetic analysis [37,38].

Since dog breed standards specifically describe the skeletal attributes of each breed, they may
be used as a proxy for individual measurements in an across-breed study design. Measurements

Trends in Genetics, August 2020, Vol. 36, No. 8 601


Trends in Genetics

of dogs gathered at dog shows verified consistency within dog breeds. However, in 38 of
42 breeds measured, 50% of the dogs fell outside the breed-standard guideline, demonstrating
continued variability within breeds or errors in measuring [38]. For the Portuguese water dog
(PWD) breed, size varied as much as threefold based on measurements from radiographs,
illustrating that there is continued variability within this breed [37]. Comparing results based on
individual measurements with those obtained using breed-standard measurements, the same
loci were identified; however, the significance was reduced using individual measurements [39].
Breed-standard measurements have been successfully used to define underlying genetic varia-
tion among dog breeds; however, not all possible loci affecting skeletal morphologies can be
identified, since within-breed variation is not scored.

Genetic Approaches to Understanding Skeletal Variation


Long linkage disequilibrium is a common phenomenon in the purebred dog and has facili-
tated successful genome-wide association studies (GWASs) using relatively few individ-
uals and markers within breeds (Figure 3A, Key Figure) [40,41]. Many examples of successful
causal variant identification utilizing a modest number of animals exist, highlighting the advan-
tage of the unique population structure of purebred dogs [42–45]. Traits shared across breeds
can also be used to refine critical intervals obtained from association within breeds (Figure 3B)
[41,46–49].

Understanding of the loci and alleles that control morphological variation between breeds has
posed unique challenges, since much of the phenotypic diversity in size and shape is not segre-
gating within closed breeding populations. Initially, crosses between breeds were performed to
understand the inheritance of diverse skeletal phenotypes [21]. In the era of modern molecular
genetics, crosses were also performed and proposed, along with selective sweep mapping
to identify breed-specific traits [50–53]. Across-breed association using breed-based stereotyp-
ical phenotypes rather than individual measurements has been suggested to identify breed
characteristics without the need for crosses [54,55]. The premise is that variants affecting
morphologies would be the same across different breeds and that they would be fixed for
the alleles that define their unique morphology. Association analysis to identify allele frequency
differences would be performed in few dogs from many breeds that differ in some trait of
interest (Figure 3C).

Across-breed association analysis has some predictable caveats that can lead to spurious
associations. When modeling of such approaches was performed for mouse strains, dramatic
increases in power were obtained when multiple animals were measured and when individual
measurements were used in the analysis. Variants that explain N35% of phenotypic variance
were identifiable using these methods but variants with smaller effect were not [56]. Even with
60 strains from the collaborative cross of the mouse community [57], repeated individual
measures (five) were necessary to identify quantitative trait loci (QTLs) with 20% effect size
[58]. Another source of error can occur when there is unequal relatedness between strains or
breeds [59]. Based on the unique population history of different dog breeds, not all breeds are
equally related to one another [11].

The era of WGS facilitates the direct investigation of variants underlying phenotypic effects across
dog breeds. Combining breed-stereotypical measurements or traits with association across all
variants identified by WGS bypasses the array genotyping step [60,61]. Bypassing the SNV
genotyping step can improve causative variant detection for QTL mapping since the most-
associated variant can be detected. Structural variants can be more easily queried at the same
time, which is critical since so many of the causative variants are structural.

602 Trends in Genetics, August 2020, Vol. 36, No. 8


Trends in Genetics

Key Figure
Association Mapping Schemes in Dogs

Trends in Genetics

Figure 3. Different canine mapping schemes for red color as an example. (A) Within-breed genome-wide association studies
(GWASs) where the allele frequency differences between red cases and black controls are used to identify large regions of
associated single-nucleotide variants (SNVs). (B) GWAS using two breeds each segregating red and black comparing the
associated regions to narrow the critical interval. (C) Across-breed association where small numbers of individuals from
groups of genetically isolated breeds are used. In this type of mapping scheme, segregation of the phenotype of interest
within breeding populations does not occur; rather, the phenotype is evaluated across breeds. The squares represent
genetically isolated breeds that differ in the phenotype of interest (red color). Allele frequency differences between groups
of breeds can occur due to the biologically relevant locus (red) or population structure leading to false-positive
associations (orange and purple).

Variants Affecting Morphology


Variants Affecting Height and Body Size
While there is tremendous diversity across domestic dog breeds, the skeletal diversity segregat-
ing within breeds can allow the dissection of variant effects on specific bones and overall skeletal
size. Variants in IGF1, COL11A2, ITGA10, and ADAMTS17 and an FGF4 retrogene insertion on
CFA 12 all have an effect on height that segregates within specific dog breeds. COL11A2 and
ITGA10 cause undesirable short stature based on breed standards. By contrast, ADAMTS17
and the FGF4 retrogene appear to be under positive selection in their breeds, presumably due
to their additive influence on height [30,37,47,62,63]. Interestingly, only IGF1 and the FGF4
retrogene have been implicated in body size in across-breed association studies [39,61,64].
The IGF1 locus has remained the most significant locus for height in all studies of size [49].

Trends in Genetics, August 2020, Vol. 36, No. 8 603


Trends in Genetics

The first across-breed association (148 breeds) used breed-average measurements for height
combined with only 1536 SNV markers identified loci on Canis familiaris (CFA) chromosome 7
(SMAD2), CFA10 (HMGA2), CFA15 (IGF1), and CFA34 (IGF2BP2) [55], all of which have been
replicated [39,61,64–69]. Subsequent across-breed analysis for body size has expanded on
these findings, replicating the association to these loci and identifying new ones. The number of
individual dogs, breeds, and markers, as well as methods of categorizing morphology, vary among
these studies (Table 1) [39,47,61,64–71]. The majority of the genes near the associated SNVs are
involved in the growth hormone axis and are therefore excellent candidates for body size loci.

Alterations in limb length change the look and purpose for which dogs are used and are generally
invariant within dog breeds. Arguably, the most successful use of across-breed association on

Table 1. Replication of Positive Associations for Body Size in Dogsa


Gene Location Phenotype
IGF1 CFA15:41.2 BL, BW W, H W W S W, H P W, H W, H
HMGA2 CFA10:8.4 W, H W W W, H W, H W, H
IGSF1 CFAX:102.3 W W, H B T W, H W
LCORL CFA3:91.2 H W, H T W, H W, H
FGF4–Ins CFA18:20.4 H H W, H W, H W, H
FGF4–Ins CFA12:33.7 H, W W, H H H
STC2 CFA4:39.2 W W, H W, H H
GHR CFA4:67.0 W W, H W, H W, H
TBX19 CFA7:30.2 W W, H W, H W
SMAD2 CFA7:43.8 W, H W, H W, H W, H
IGF1R CFA3:41.8 S W, H W, H H
IRS4 CFAX:82.2 W B T B, W
IGF2BP2 CFA34:18.4 W, H W, H W, H
SMOC2b CFA1:55.9 W, H W, H W
Unknown CFA3:62 H H
ZNF608 CFA11:14 W W
Unknown CFA11:26 W, H W, H
Unknown CFA20:21.5 W, H W, H
Unknown CFA26:7.7 W, H W
Unknown CFA26:12 W, H W, H
GH1 CFA9:11.8 T W, H
ESR1 CFA1:42.3 L
c
DVL CFA5:32 W, H
Unknown CFA6:38.2 B
R3HDM1 CHR19:38.3 W
ADAMTSL9-AS CFA20:26.7 W
HNF4G CFA29:23.8 W
BMPb CHR32:5 W, H
Refs [29] [54] [69] [68] [46] [63] [64] [65] [31] [78] [67] [79] [61] [62]

a
Abbreviations: BL, long-bone length; BW, long-bone width; W, weight; H, height; B, bulky; L, long legs; S, small size; T, tibial measurements; P, pelvic measurements.
b
Loci identified for their effect on skull morphology.
c
Locus identified for vertebral malformations.

604 Trends in Genetics, August 2020, Vol. 36, No. 8


Trends in Genetics

breed stereotypes was for severe disproportionate dwarfism or chondrodysplasia as exemplified


for the dachshund breed [72]. Chondrodysplasia is caused by an FGF4 retrogene insertion on
CFA 18 [47,71] and has also been identified in subsequent across-breed GWASs for size due
to the strong effect of leg length on the overall size of the dog [39,61].

Allele frequency estimates within breeds combined with breed-average measurements have
been used to validate the percentage of variance explained by body size loci across breeds.
Depending on the type of modeling used, the estimates vary, with 46–95% of the variance in
size among breeds explained by fewer than 20 loci [39,61,65,68]. Surprisingly, the majority of
breeds segregate many of their size loci. In one study evaluating seven body size loci in 93 breeds,
only 14 breeds were homozygous for their size alleles (Figure 4) [68].

Validation of size loci was performed using a village dog dataset, which is a randomly bred
population allowing the evaluation of segregating loci. In that population, 30–40% of the variance
was explained using an additive linear model [39]. However, reanalysis of these data correcting
for population structure and sex found that only 13.5% of the variance was explained by those
loci [73]. Ideally, validation of size loci would occur within segregating populations where the
percentage of variance attributed to each locus could be evaluated; however, this has been
done only for IGF1 in the PWD [37,49].

Variants Affecting Skull Shape


Extreme variation in skull shape does not segregate within any breed so this trait must be evalu-
ated across breeds or potentially in crosses. Brachycephalic head phenotype was evaluated

Siberian husky German shepherd dog Saint Bernard Chihuahua

1
0.9
0.8
Allele frequency

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

Body size variants

Trends in Genetics

Figure 4. Allele Distribution within Breeds for Loci That Contribute to Body Size in Dogs. The genes and their allele
frequencies are plotted for four representative breeds: Chihuahua, Siberian husky, German shepherd dog, and Saint
Bernard. Allele frequencies were used from the following publications: FGF4* retrogene on CFA12 and CFA18 from [87],
CFA26, GHI from [64], and the remaining variants from [61].

Trends in Genetics, August 2020, Vol. 36, No. 8 605


Trends in Genetics

using across-breed mapping and breed-standard phenotyping. The strongest association signal
was with CFA1; however, additional loci also appeared to be associated [65,70].

A more thorough investigation into skull morphometry using digitized measurements of


51 landmarks from skulls from different breeds was conducted to quantify the stereotypical
skull morphologies prior to across-breed GWASs for skull shape. This approach identified
six loci, and a missense variant in BMP3 was revealed as the causative variant of one asso-
ciated locus on CFA32 [24]. Ultimately, the causal variant underlying the association on
CFA1 was identified in the SMOC2 gene using skull morphometry from CT scans across
breeds and mixed-breed dogs, which provided the quantitative phenotypes as well as the
segregation evidence to determine a critical interval. Neurocranium size (backskull) was
also associated with overall body size loci (LCORL, SMAD2, HMGA2, IGF1, FGF4 retrogene
CFA18) [23].

The DVL2 locus initially identified as associated with vertebral malformation and a tail phe-
notype has also been identified in multiple across-breed GWASs for skull morphology
[24,60,65]. However, it is unclear whether this is a spurious association due to breed relat-
edness or a real effect of the allele on skull morphology Similar potentially spurious associ-
ations for body size have been identified for loci that affect skull shape (Table 1) [39,61,64].
This may be resolved by the evaluation of quantitative measurements in segregating
populations.

Variants Affecting Health


Perhaps unsurprisingly, morphological traits in dogs have been associated with various health
outcomes. Overall body size is strongly associated with lifespan in dogs, with small dogs living
twice as long as giant breeds [74,75]. In two investigations using different dog breeds to de-
termine loci underlying differences in average lifespan, only a single locus, HMGA2, was
found in common [54,61,76]. One study found loci for coat length significantly associated
with longevity, but this may reflect the subset of breeds used in the analysis [76]. Nonetheless,
this hypothesis is testable since both small and large-sized breeds segregate for different coat
types.

Orthopedic disease is an important manifestation of alterations in bone morphology within the


pet dog population [77]. Quantitative measurements taken from radiographs of dogs with hip
dysplasia and unaffected dogs were used in GWASs across breeds and mixes. Interestingly,
the IGF1 locus was significantly associated with overall pelvis size, which is consistent with
an increased tendency toward hip dysplasia in large-breed dogs [78]. In a study of cranial
cruciate ligament disease, radiographic measurements were obtained for a large set of dogs
across breeds. Significant associations were identified with LCORL, GH1 and IRS4, and
ACSL4, body size loci previously identified in across-breed associations. Novel loci on CFA8
and CFA10 were also identified, which underscores the importance of analyses that use
individual measurements [79].

Subtle differences in skull shape may predispose some brachycephalic dogs to serious
medical conditions like Chiari-like malformation. This syndrome was investigated using
MRI of individual dogs within affected breeds to obtain quantitative measurements used
for GWASs. Novel loci not previously associated with size or skull shape across breeds
were identified [80,81]. Since brachycephaly predisposes dogs to a number of health con-
ditions, additional studies of variants segregating within these breeds should be performed
[82–84].

606 Trends in Genetics, August 2020, Vol. 36, No. 8


Trends in Genetics

Concluding Remarks Outstanding Questions


The investigations of genes and alleles that influence differences in overall size and shape among Across-breed analysis provides a
breeds illustrate some of the successes as well as the limitations of the across-breed mapping means of identifying breed-defining
morphologies without performing seg-
approach using stereotypical measurements. In many cases, variants within strong candidate
regating crosses. How can the results
genes underlie the associations, providing compelling evidence for their validity. However, the be validated? Will replication with
number of loci identified is surprisingly small considering the extreme morphological variation mixed-breed dogs or analysis within
within dogs. Notably, many loci governing size are not fixed within breeds, even under strong breeds be able to demonstrate the
same genetic contribution of variants
artificial selection for specific size and weight by breeders (see Outstanding Questions)
to phenotypes?
(Figure 4). Exaggerated phenotypes have a profound influence on the health of dogs [18,85],
which might lead to the maintenance of diversity at loci controlling skeletal development. A The specific effects of variants on the
reoccurring theme in canine morphological genetics is that limited loci explain the variance across skeleton are largely unknown. Can
quantitative methods of measuring
breeds; however, additional loci can be identified within breeds or with quantitative measure- bone size, shape, mass, and density
ments. The genetics of canine body size may be relatively simple across breeds but may resem- be used to define the percentage of
ble other species, with many loci of smaller effect within breeds and in outbred populations phenotypic variation contributed by
each locus and to disentangle additive
[73,86].
and epistatic effects on phenotypes?

Quantitative measures of variation in bone size, shape, and density within and between dog Multiple genetic variants within each
breeds combined with the ability to define the genetic contributions of the variants underlying breed have accumulated to obtain the
vast diversity of morphologies across
these phenotypes can greatly enhance our understanding of mammalian skeletal morphology.
dog breeds. With such strong selection
Large datasets with quantitative phenotypic measurements are challenging to collect compared for canine morphology within breeds,
with using breed-stereotypical measurements and have been used only a few times. Validation of why is there continued variation in allele
interbreed stereotypical skeletal morphology by quantitative measures within segregating popu- frequency for some of these loci?

lations would allow the determination of the additive, dominant, and epistatic effects of the loci on
the skeleton. An improved understanding of the genes and variants that affect bones by making
them larger or smaller, or more or less dense, can be obtained from studies started by Charles
Stockard in the late 19th century [21] and continued today by modern canine geneticists.

References
1. Frantz, L.A. et al. (2016) Genomic and archaeological evidence 13. American Kennel Club (1998) The Complete Dog Book (19th edn),
suggest a dual origin of domestic dogs. Science 352, Howell Book House
1228–1231 14. Young, A. and Bannasch, D. (2006) Morphological variation in
2. Freedman, A.H. et al. (2014) Genome sequencing highlights the the dog. In The Dog and Its Genome (Ostrander, E.A. et al.,
dynamic early history of dogs. PLoS Genet. 10, e1004016 eds), pp. 47–65, Cold Spring Harbor Laboratory Press
3. Sykes, N. et al. (2020) Humanity’s best friend: a dog-centric 15. Salt, C. et al. (2017) Growth standard charts for monitoring
approach to addressing global challenges. Animals (Basel) 10, 502 bodyweight in dogs of different sizes. PLoS One 12, e0182064
4. Pilot, M. et al. (2015) On the origin of mongrels: evolutionary 16. Drake, A.G. and Klingenberg, C.P. (2010) Large-scale diversifi-
history of free-breeding dogs in Eurasia. Proc. Biol. Sci. 282, cation of skull shape in domestic dogs: disparity and modularity.
20152189 Am. Nat. 175, 289–301
5. Clutton-Brock, J. (1995) Origins of the dog: domestication 17. Bell, J. et al. (2012) Veterinary Medical Guide to Dog and Cat
and early history. In The Domestic Dog: Its Evolution, Behav- Breeds, Teton NewMedia
iour and Interactions with People (Serpell, J., ed.), pp. 7–20, 18. Asher, L. et al. (2009) Inherited defects in pedigree dogs. Part 1:
Cambridge University Press disorders related to breed standards. Vet. J. 182, 402–411
6. Morris, D. (2002) Dogs: The Ultimate Dictionary for Over 1,000 19. Andreis, M.E. et al. (2018) Novel contributions in canine craniom-
Dog Breeds, Trafalgar Square etry: anatomic and radiographic measurements in newborn
7. Leroy, G. (2011) Genetic diversity, inbreeding and breeding puppies. PLoS One 13, e0196959
practices in dogs: results from pedigree analyses. Vet. J. 189, 20. Evans, H.E. and De Lahunta, A. (2013) Miller’s Anatomy of the
177–182 Dog, Saunders
8. Wade, C.M. (2011) Inbreeding and genetic diversity in dogs: 21. Stockard, C.R. (1941) The Genetic and Endocrinic Basis for
results from DNA analysis. Vet. J. 189, 183–188 Differences in Form and Behavior, Wistar Institute of Anatomy
9. Calboli, F.C. et al. (2008) Population structure and inbreeding and Biology
from pedigree analysis of purebred dogs. Genetics 179, 22. Koch, D.A. et al. (2012) Proposal for a new radiological index to
593–601 determine skull conformation in the dog. Schweiz. Arch.
10. Dreger, D.L. et al. (2016) Whole-genome sequence, SNP chips Tierheilkd. 154, 217–220
and pedigree structure: building demographic profiles in domes- 23. Marchant, T.W. et al. (2017) Canine brachycephaly is associated
tic dog breeds to optimize genetic-trait mapping. Dis. Model. with a retrotransposon-mediated missplicing of SMOC2. Curr.
Mech. 9, 1445–1460 Biol. 27, 1573–1584.e6
11. Parker, H.G. et al. (2017) Genomic analyses reveal the influence 24. Schoenebeck, J.J. et al. (2012) Variation of BMP3 contributes to
of geographic origin, migration, and hybridization on modern dog breed skull diversity. PLoS Genet. 8, e1002849
dog breed development. Cell Rep. 19, 697–708 25. Drake, A.G. and Klingenberg, C.P. (2008) The pace of morpho-
12. Brewer, D. et al. (2001) Dogs in Antiquity. Anubis to Cerberus: logical change: historical transformation of skull shape in
The Origins of the Domestic Dog, Aris & Phillips St Bernard dogs. Proc. Biol. Sci. 275, 71–76

Trends in Genetics, August 2020, Vol. 36, No. 8 607


Trends in Genetics

26. Fondon III, J.W. and Garner, H.R. (2004) Molecular origins of 51. Neff, M.W. et al. (1999) A second-generation genetic linkage
rapid and continuous morphological evolution. Proc. Natl. map of the domestic dog, Canis familiaris. Genetics 151,
Acad. Sci. U. S. A. 101, 18058–18063 803–820
27. Ellis, J.L. et al. (2009) Cranial dimensions and forces of biting in 52. Metallinos, D. and Rine, J. (2000) Exclusion of EDNRB and KIT
the domestic dog. J. Anat. 214, 362–373 as the basis for white spotting in border collies. Genome Biol.
28. McGreevy, P. et al. (2004) A strong correlation exists between 1, RESEARCH0004
the distribution of retinal ganglion cells and nose length in the 53. Pollinger, J.P. et al. (2005) Selective sweep mapping of genes
dog. Brain Behav. Evol. 63, 13–22 with large phenotypic effects. Genome Res. 15, 1809–1819
29. Helton, W.S. (2011) Performance constraints in strength events 54. Neff, M.W. and Rine, J. (2006) A fetching model organism. Cell
in dogs (Canis lupus familiaris). Behav. Process. 86, 149–151 124, 229–231
30. Jeanes, E.C. et al. (2019) Glaucoma-causing ADAMTS17 55. Jones, P. et al. (2008) Single-nucleotide-polymorphism-based
mutations are also reproducibly associated with height in two association mapping of dog stereotypes. Genetics 179,
domestic dog breeds: selection for short stature may have con- 1033–1044
tributed to increased prevalence of glaucoma. Canine Genet. 56. Kang, H.M. et al. (2008) Efficient control of population structure
Epidemiol. 6, 5 in model organism association mapping. Genetics 178,
31. Stull, K.E. et al. (2014) Accuracy and reliability of measurements 1709–1723
obtained from computed tomography 3D volume rendered 57. Consortium, C.C (2012) The genome architecture of the Collab-
images. Forensic Sci. Int. 238, 133–140 orative Cross mouse genetic reference population. Genetics
32. Ismail, N.A. et al. (2019) Accuracy and reliability of virtual femur 190, 389–3401
measurement from CT scan. J. Forensic Legal Med. 63, 11–17 58. Keele, G.R. et al. (2019) Determinants of QTL mapping power in
33. Smith, E.J. et al. (2017) Three-dimensional assessment of curvature, the Realized Collaborative Cross. G3 (Bethesda) 9, 1707–1727
torsion, and canal flare index of the humerus of skeletally mature 59. Payseur, B.A. and Place, M. (2007) Prospects for association
nonchondrodystrophic dogs. Am. J. Vet. Res. 78, 1140–1149 mapping in classical inbred mouse strains. Genetics 175,
34. Lorinson, K. et al. (2008) Signalment differences in bone mineral 1999–2008
content and bone mineral density in canine appendicular bones. 60. Mansour, T.A. et al. (2018) Whole genome variant association
A cadaveric study. Vet. Comp. Orthop. Traumatol. 21, 147–151 across 100 dogs identifies a frame shift mutation in
35. Mejia, S. et al. (2019) Comparison of cross-sectional geometrical DISHEVELLED 2 which contributes to Robinow-like syndrome
properties and bone density of the proximal radius between in bulldogs and related screw tail dog breeds. PLoS Genet. 14,
Saint Bernard and other giant breed dogs. Vet. Surg. 48, e1007850
947–955 61. Plassais, J. et al. (2019) Whole genome sequencing of canids re-
36. Wright, N.C. et al. (2014) The recent prevalence of osteoporosis veals genomic regions under selection and variants influencing
and low bone mass in the United States based on bone mineral morphology. Nat. Commun. 10, 1489
density at the femoral neck or lumbar spine. J. Bone Miner. Res. 62. Frischknecht, M. et al. (2013) A COL11A2 mutation in Labrador
29, 2520–2526 retrievers with mild disproportionate dwarfism. PLoS One 8,
37. Chase, K. et al. (2002) Genetic basis for systems of skeletal e60149
quantitative traits: principal component analysis of the canid 63. Kyostila, K. et al. (2013) Canine chondrodysplasia caused by a
skeleton. Proc. Natl. Acad. Sci. U. S. A. 99, 9930–9935 truncating mutation in collagen-binding integrin alpha subunit
38. Sutter, N.B. et al. (2008) Morphometrics within dog breeds are 10. PLoS One 8, e75621
highly reproducible and dispute Rensch’s rule. Mamm. Genome 64. Hayward, J.J. et al. (2019) Imputation of canine genotype
19, 713–723 array data using 365 whole-genome sequences improves
39. Hayward, J.J. et al. (2016) Complex disease and phenotype power of genome-wide association studies. PLoS Genet.
mapping in the domestic dog. Nat. Commun. 7, 10460 15, e1008003
40. Lindblad-Toh, K. et al. (2005) Genome sequence, comparative 65. Boyko, A.R. et al. (2010) A simple genetic architecture underlies
analysis and haplotype structure of the domestic dog. Nature morphological variation in dogs. PLoS Biol. 8, e1000451
438, 803–819 66. Vaysse, A. et al. (2011) Identification of genomic regions asso-
41. Karlsson, E.K. et al. (2007) Efficient mapping of mendelian traits ciated with phenotypic variation between dog breeds using
in dogs through genome-wide association. Nat. Genet. 39, selection mapping. PLoS Genet. 7, e1002316
1321–1328 67. Hoopes, B.C. et al. (2012) The insulin-like growth factor 1 receptor
42. Das, R.G. et al. (2019) Genome-wide association study and (IGF1R) contributes to reduced size in dogs. Mamm. Genome 23,
whole-genome sequencing identify a deletion in LRIT3 associ- 780–790
ated with canine congenital stationary night blindness. Sci. 68. Rimbault, M. et al. (2013) Derived variants at six genes explain
Rep. 9, 14166 nearly half of size reduction in dog breeds. Genome Res. 23,
43. Lucot, K.L. et al. (2018) A missense mutation in the vacuolar pro- 1985–1995
tein sorting 11 (VPS11) gene is associated with neuroaxonal 69. Plassais, J. et al. (2017) Analysis of large versus small dogs re-
dystrophy in Rottweiler dogs. G3 (Bethesda) 8, 2773–2780 veals three genes on the canine X chromosome associated
44. Pedersen, N.C. et al. (2017) An autosomal recessive mutation in with body weight, muscling and back fat thickness. PLoS
SCL24A4 causing enamel hypoplasia in Samoyed and its Genet. 13, e1006661
relationship to breed-wide genetic diversity. Canine Genet. 70. Bannasch, D. et al. (2010) Localization of canine brachycephaly
Epidemiol. 4, 11 using an across breed mapping approach. PLoS One 5, e9632
45. Forman, O.P. et al. (2016) An inversion disrupting FAM134B is 71. Parker, H.G. et al. (2009) An expressed fgf4 retrogene is associ-
associated with sensory neuropathy in the border collie dog ated with breed-defining chondrodysplasia in domestic dogs.
breed. G3 (Bethesda) 6, 2687–2692 Science 325, 995–998
46. Salmon Hillbertz, N.H. et al. (2007) Duplication of FGF3, FGF4, 72. Boyko, A.R. et al. (2009) Complex population structure in African
FGF19 and ORAOV1 causes hair ridge and predisposition to village dogs and its implications for inferring dog domestication
dermoid sinus in ridgeback dogs. Nat. Genet. 39, 1318–1320 history. Proc. Natl. Acad. Sci. U. S. A. 106, 13903–13908
47. Brown, E.A. et al. (2017) FGF4 retrogene on CFA12 is responsi- 73. Bouwman, A.C. et al. (2018) Meta-analysis of genome-wide
ble for chondrodystrophy and intervertebral disc disease in association studies for cattle stature identifies common genes
dogs. Proc. Natl. Acad. Sci. U. S. A. 114, 11476–11481 that regulate body size in mammals. Nat. Genet. 50, 362–367
48. Cadieu, E. et al. (2009) Coat variation in the domestic dog is 74. Kirkwood, J.K. (1985) The influence of size on the biology of the
governed by variants in three genes. Science 326, 150–153 dog. J. Small Anim. Pract. 26, 97–110
49. Sutter, N.B. et al. (2007) A single IGF1 allele is a major determi- 75. Greer, K.A. et al. (2007) Statistical analysis regarding the effects
nant of small size in dogs. Science 316, 112–115 of height and weight on life span of the domestic dog. Res. Vet.
50. Acland, G.M. et al. (1989) Non-allelism of three genes (rcd1, Sci. 82, 208–214
rcd2 and erd) for early-onset hereditary retinal degeneration. 76. Nicholatos, J.W. et al. (2019) Cellular energetics and mitochon-
Exp. Eye Res. 49, 983–998 drial uncoupling in canine aging. Geroscience 41, 229–242

608 Trends in Genetics, August 2020, Vol. 36, No. 8


Trends in Genetics

77. Boge, G.S. et al. (2019) Breed susceptibility for common surgi- 82. Mills, G. (2018) Working together to tackle brachy issues. Vet.
cally treated orthopaedic diseases in 12 dog breeds. Acta Vet. Rec. 182, 5
Scand. 61, 19 83. Evans, M. (2018) Continuing the campaign on brachycephalic
78. Fealey, M.J. et al. (2017) Genetic mapping of principal components dogs. Vet. Rec. 182, 114
of canine pelvic morphology. Canine Genet. Epidemiol. 4, 4 84. Ladlow, J. et al. (2018) Brachycephalic obstructive airway
79. Healey, E. et al. (2019) Genetic mapping of distal femoral, stifle, syndrome. Vet. Rec. 182, 375–378
and tibial radiographic morphology in dogs with cranial cruciate 85. Summers, J.F. et al. (2010) Inherited defects in pedigree dogs.
ligament disease. PLoS One 14, e0223094 Part 2: disorders that are not related to breed standards. Vet.
80. Ancot, F. et al. (2018) A genome-wide association study iden- J. 183, 39–45
tifies candidate loci associated to syringomyelia secondary to 86. Wood, A.R. et al. (2014) Defining the role of common variation in
Chiari-like malformation in cavalier King Charles spaniels. BMC the genomic and biological architecture of adult human height.
Genet. 19, 16 Nat. Genet. 46, 1173–1186
81. Lemay, P. et al. (2014) Quantitative trait loci (QTL) study identifies 87. Batcher, K. et al. (2019) Phenotypic effects of FGF4
novel genomic regions associated to Chiari-like malformation in retrogenes on intervertebral disc disease in dogs. Genes (Basel)
griffon Bruxellois dogs. PLoS One 9, e89816 10, 435

Trends in Genetics, August 2020, Vol. 36, No. 8 609

You might also like