Genomic Regions Associated with Variation in Pigmentation Loss in Saddle Tan Beagles

Loss of pigmentation is a hallmark of domestication, and dogs offer a unique model for understanding the genetics of fur coloration. The aim of this study was to use dense genetic mapping to map loci underlying variations in color and whiteness in a population of laboratory beagles. A total of 190 beagles with well-defined pedigrees were phenotyped for the amount of white color in six different body parts, including the saddle. All individuals were genotyped on 85,172 informative and valid SNP-markers and the genome-wide associations for the amount of white in each body part were determined. There was a large variation in the amount of white on different parts of the body, and the whiteness was highly correlated within individuals, except for saddle color which was only moderately correlated with overall whiteness. The GWAS showed significant associations with two loci, one on chromosome 5, containing the MC1R gene, and one on chromosome 20, containing the MITF gene. Our results suggest that the variation in loss of pigmentation is largely a function of regulatory variation related to these genes.


Introduction
Loss of pigmentation is one of the defining phenotypic traits differing domesticated animals from their wild ancestors, and central to what Darwin [1] referred to as the domestication syndrome. The famous Siberian farm fox experiment, where foxes were selected on one behavioral trait only, i.e., friendliness towards humans, found that loss of pigment and piebald white spotting started to appear in the selected population after only 8-10 generations [2]. Hence, increased understanding of the genetics underlying variation in pigmentation of domestic animals may pave the way for closer insights into the mechanisms of domestication. For example, in chickens, a mutation in the gene PMEL17 causes a total loss of black pigmentation and also pleiotropically affects domesticationrelated behavior [3].
Due to the unique domestication and population history of the dog, it now serves as an ideal subject to investigate genetic associations for behavioral as well as morphological traits [4]. Through the process of domestication and selection by humans, the domestic dog (Canis familiaris) has become an impressively diverse species with its almost 400 different breeds with different combinations of fur coloration, color patterns, length, and textures. Some breeds are, more or less, fixed for a fur coloration phenotype e.g., the golden retriever, while other breeds come in a wide variety of colors and patterns such as the boxer. In wolves, the ancestors of all dogs, there is a limited color variation, where most individuals are camouflaged in brown and grey, but alleles for melanism as well as light hair color appear to have been present in the wolf population more than 10,000 years ago [5]. Some of the present-day variation in whiteness could therefore be due to selection of standing variation in the ancestral wolf population, but new mutations in pigmentation-related genes have probably been selected as well.
Coat coloration and pattern in mammals is caused by the production of just two types of pigments: black-brown eumelanin and red-yellow pheomelanin [6]. These pigments are Genes 2021, 12,316 2 of 17 produced by melanocytes in organelles called melanosomes that are then transferred out to hair and skin cells. Precursor cells of melanocytes migrate from the neural crest to hair follicles and epidermis during embryogenesis. White spotting and loss of pigmentation is usually due to an absence of melanocytes caused by e.g., failure in differentiation during development or migration. Although many details of the genetic mechanisms underlying coat color are still insufficiently known, some aspects are more well-described. For example, Melanocortin 1 receptor (MC1R) controls which type of pigment is synthesised by the melanocytes. High MC1R signaling leads to increased eumelanin synthesis. In turn, Agouti signaling protein (ASIP) can inhibit MC1R, causing an increase in pheomelanin synthesis. Additionally, CBD103 can bind to MC1R and block the binding site of ASIP and in that way increase the synthesis of eumelanin. Other genes are involved as well, for example, TYRP1 that is associated with brown color, and MLPH, involved in diluted colors [7]. A recent review summarises the state of art of the complex genetic underpinnings of dog fur color genetics [8].
Based on inheritance, Little [9] defined the genetic nomenclature of dog coat color alleles still used today (summarised by Kaelin and Barsh [6]). For example, Little described the S-locus (Spotting), subsequently mapped to the MITF-gene, E-locus (Extension), later mapped to the MC1R-gene, the A-locus (Agouti), later mapped to the ASIP-gene and the RALY-gene that Little originally believed was an allele of the A-locus. Today, we are able to use powerful analyses such as genome-wide association studies (GWAS) to further disentangle the different loci proposed by Little and identify molecular pathways associated with fur coloration. For example, recent studies have identified regulatory modules for ventral as well as hair cycle ASIP expression, including their evolutionary origins, and found no association with RALY [10].
Beagles are considered fixed for the saddle tan and piebald fur coloration genotype [11][12][13]. This means that all individuals have at least some white markings on the paws, tip of the tail, and on the ventral side of the body including the neck caused by the s p /s p -genotype of the MITF-gene. Additionally, all individuals that are able to express eumelanin at the MC1R-gene will have a saddle on top of a tan background. The saddle is the black fur covering the dorsal area of the body, in beagles, it typically includes the neck and reaches all the way to the base of the tail. According to the data presented by Dreger and Schmutz [14] and Dreger, Hooser [12], beagles genotype a t /a t at the ASIP gene which is required for the saddle tan and black-and-tan phenotypes.
In the beagle laboratory population studied by us, there is a vast range of variation in the amount of white spotting and the amount of black coloration of the saddle. This suggests that other modifier loci than those previously identified are involved in affecting the expression of fur coloration, and in consequence affecting the degree of loss of pigmentation. The aim of this study was therefore to identify genomic region associated with variation in loss of pigmentation on different body parts in a population of laboratory beagles.

Animals
The 190 laboratory beagles phenotyped and genotyped in this study were originally used for another GWAS on human-directed social behaviors [15]. All dogs belonged to the same research population and were on average 1.8 years old (min 0.8 years, max 6.2 years) at the time of phenotyping. The population was intentionally outbred; however, some degree of inbreeding has to be accounted for in the analysis. For a more detailed description on housing, handling, and behavioural phenotyping see Persson and Roth [16].

Phenotyping and Genotyping
Screen shots of all dogs were collected from video recordings. At least three pictures (left side, right side, and frontal view of the face) from each individual were saved in order to phenotype their white patches of the face, neck, legs, body, and tail as well as the black saddle. Each individual was then subjectively scored on a scale from 1 to 3 based upon the coverage of their white spotting and the black saddle ( Table 1). The reliability of the phenotype scoring was assessed by an independent observer recoding 10% of the individuals (N = 19) and then calculating the Spearman correlation coefficients to the two independent classifiers. Except for three individuals, all dogs were tri-colored black and tan with white spotting. However, three individuals were only expressing tan and white spotting and were therefore excluded from the saddle analysis as they may still carry the genetic variants for the saddle without being able to express the phenotype. Table 1. Fur coloration scoring. A description of the scoring criteria for the different fur pigmentation phenotypes. Overall white is a combined score of all white patching phenotypes.

Phenotype
Score Name Description No or residual white Solid colored face with no, or only, residual white around the nose.
2 White muzzle White markings around the nose, partly or completely extending over the muzzle, no blaze but a thin streak may occur.

Blaze
As described above but also including a blaze reaching to or above eye level. DNA was obtained from either blood or buccal cells and subsequently genotyped using Illumina's Infinium assay and the CanineHD BeadChip at the SNP&SEQ Technology Platform, Uppsala University in Sweden, based on the CanFam3 assembly. To confirm sample and marker quality, SNP&SEQ also performed initial quality control using the software GenomeStudio 2011.1 from Illumina Inc. [17]. No samples had to be excluded since all had a call rate > 98% and therefore passed the 90% threshold.
Additional quality control was performed on 172,115 markers using the PLINK 1.07 software (http://pngu.mgh.harvard.edu/~purcell/plink/ (accessed on 20 January 2021)) [18]. Markers failing the criteria of minor allele frequency (MAF) > 5%, Hardy-Weinberg p > 0.05 and genotyping rate >90% were removed. Finally, 85,172 markers passed the criteria and were used in the GWAS. To visualize population structure, multidimen-sional scaling (MDS) analysis was performed in PLINK and the first and second dimensions were plotted.

GWAS
A minor level of population stratification (genomic inflation factor λ = 1.0-1.4) is present within the studied population [15]. To adjust for relatedness as well as population stratification the GWAS analysis was performed using the GEMMA software (http://www. xzlab.org/software.html (accessed on 20 January 2021)) which implements the Genomewide Efficient Mixed Model Association algorithm [19]. Initially, a centered relatedness matrix was estimated for each color phenotype based on genotype. The association tests were then performed with the relatedness matrix, phenotype and genotype fit into a univariate linear mixed model. To reach genome-wide significance, Wald p-values from GEMMA had to pass a Bonferroni threshold at p < 0.05 calculated for 85,172 markers (genome-wide significance threshold: p = 5.9 × 10 −7 ). Additionally, GEMMA reports PVE ("chip heritability") which is an estimate of the proportion of phenotypic variance explained. The regression method in the R-package GenABEL was used to make estimates of λ [20]. Q-Q and Manhattan plots were generated through the qqman-package in R version 3.2.3. We checked whether there was any effect of age on overall "whiteness" score, by splitting the data into three age-classes and using a generalized linear model to analyze any possible differences but there was no such effect (Wald Chi-Square: 2.64; p = 0.27), so age was not included in the model. Subsequent linkage disequilibrium (LD) analysis and visualization of haplotypes was done in the Haploview v4.2 software [21]. LD windows were estimated based upon solid spine of LD and SNPs that were considered to be linked had a D' of 1 and LODscore of >0.15. Visualization of the genomic position of significant SNPs in relation to CanFam3 mapped genes were done using the software Integrative Genomics Viewer (IGV) version 2.8.9 [22]. The association between genotypes of the most significant SNP for each phenotype was investigated using the non-parametric Kruskal-Wallis test in IBM SPSS Statistics 25. Pairwise comparisons were made, and Bonferroni was used to adjust p-values for multiple testing. A Spearman correlation was also carried out to investigate how the different phenotypic measures correlate with each other.

Results
The amount of white fur coverage of the face, neck, legs, body, and tail as well as black saddle coverage for each of the 190 beagles is listed in Supplementary Table S1. The inter-rater reliability analysis showed high agreement for all phenotypes (N = 19, Spearman Rho > 0.92, p < 0.001). Significant SNP-phenotype associations were found for the face, neck, legs, and overall white on a region on chromosome 20. The face and saddle coloration phenotypes were associated with a region on chromosome 5. A detailed description of our results is given below.

Color Phenotype
Frequencies of each phenotype are presented in Figure 1. A set of representative photographs of each of the phenotypes can be found in the supplementary material. There is a significant positive correlation between all of the white phenotypic measurements (Figure 2). This means that an individual with a high coverage of white patching on one body part typically have more white coverage on other areas as well. The saddle however, only correlated with the amount of white in the face and the combined measurement called overall white. The estimated proportion of variance explained (PVE) or "chip heritability" of the different fur coloration phenotypes ranged between 0.40 ± 0.13 (SE) for the coverage of white spotting in the face, to 0.79 ± 0.11 (SE) for the amount of white spotting covering the legs. All PVE estimates are presented in Table 2. Table 2. Proportion of variance explaine (PVE) or "chip heritability" and λ (genomic inflation factor) from the GWAS analysis in GEMMA.

GWAS
After pruning the genotype data from uninformative and low-call rate markers, 85,172 SNPs remained for the GWAS analysis in GEMMA. The GEMMA software successfully accounted for population stratification as shown by the genomic inflation factor λ in Table 2 as well as the Q-Q plots in Supplementary Figure S1. The MDS plot is presented in Supplementary Figure S2.
Significant genotype associations were found for the amount of white coverage of the face, neck, legs, and overall white as well as the black pigmentation coverage of the saddle (Figure 3). No markers were found to be significantly associated with the amount of white fur covering the body and tail. The amount of white in the face is associated with a significant peak of 5 SNPs on chromosome 5 and 5 SNPs on chromosome 20 (  Table S5). The 1st haplotype block is overlapping the region of the MC1R-gene. The gene GALNS is positioned between the two blocks while EOGT is positioned after the 2nd block.

Discussion
This genome-wide association study used a high-density canine SNP chip to identify two genomic regions associated with the variation in loss of pigmentation in a population of piebald saddle-tan beagles. The dogs were scored based on the amount of white patching covering the face, neck, body, tail, legs, overall white, and the amount of solid black fur coverage of the saddle. White patching of the face, neck, legs, and overall white was associated with a region on chromosome 20, overlapping the MITF-gene locus positioned at 21612927-21870578 (CanFam3.1). Additionally, the variation in face and saddle coloration was associated with a region on chromosome 5 overlapping the locus of the MC1R gene. Both these genes have previously been associated with fur coloration phenotypes,

Saddle Coloration
The amount of black coloration of the saddle is significantly associated with 19 SNPs on chromosome 5, where the strongest association was found with three SNPs that share the same Bonferroni-adjusted p-value of 9.8 × 10 −18 . These are the same three SNP markers on chromosome 5 that also had the strongest association with face coloration. Figure 9f presents the frequency of saddle coloration for the different genotypes of the first of these three significant SNPs, BICF2P831934 positioned at chr5: 63728735 (N = 187, χ 2 = 64.129, p < 0.001, df = 2). No pairwise comparisons were made as there are only two genotypes, AG and GG in the model. Note that 187 individuals were included in the GWAS of the saddle phenotype as 3 individuals did not seem to display any black pigments. However, these three individuals turned out to be the only carriers of the AA-genotype at BICF2P831934. To acknowledge this, they were included in Figure 9f but not in the statistical analysis.

Discussion
This genome-wide association study used a high-density canine SNP chip to identify two genomic regions associated with the variation in loss of pigmentation in a population of piebald saddle-tan beagles. The dogs were scored based on the amount of white patching covering the face, neck, body, tail, legs, overall white, and the amount of solid black fur coverage of the saddle. White patching of the face, neck, legs, and overall white was associated with a region on chromosome 20, overlapping the MITF-gene locus positioned at 21612927-21870578 (CanFam3.1). Additionally, the variation in face and saddle coloration was associated with a region on chromosome 5 overlapping the locus of the MC1R gene. Both these genes have previously been associated with fur coloration phenotypes, although the mechanisms are still insufficiently known. The associations demonstrated in the present paper suggest that regulatory mutations may affect loss of pigmentation of different parts of the body in beagles, and therefore adds significant knowledge to the quantitative variation in fur coloration in dogs. Future research should be able to determine the causative mutations by genotyping known variants or DNA-sequencing of individuals with contrasting phenotypes.
The temporal and spatial production of pigments is controlled by the interaction of multiple genes. The current state of knowledge with respect to the genetics of fur coloration in dogs has recently been extensively reviewed [8]. In this paper, we focused on one breed, the beagle, that is typically 3-colored as it has the saddle-tan phenotype with piebald white spotting. Piebald white patching has previously been mapped to the MITFgene [23,24]. To express piebald spotting, the dog needs to carry two copies of the s p -allele of the gene microphthalmia-associated transcription factor (MITF) [13,25]. Beagles are fixed for the piebald white spotting s p -allele at the MITF-gene [12] but yet we see a significant amount of variation in the amount of white fur coverage. The chip heritability of this variation ranged between 0.40 and 0.79 indicating that there is a significant heritable genetic basis for this variation in piebald white patching in the studied beagle population. Additionally, all of the variation in white patching phenotypes in face, neck, body, tail, and legs were significantly positively correlated while there was a considerably smaller correlation between saddle and the rest of the body. This raises the possibility that the same genetic mechanisms underlie most of the white patching except for the saddle, where other or additional mechanisms may play a role.
Since MITF is regulated in a highly complex manner, including several different start codons and with different functions in different tissues, it is most likely that the phenotypic variations are a result of regulatory differences between individuals. Consistent with this, it has been found that variations in the regulation of MITF by means of a simple repeat polymorphism in the MITF-M promoter may explain the degree of white spotting in a variety of dog breeds [13]. In the present study, we found that loss of pigmentation on leg, body, neck, and face was associated with SNP markers closely associated with the MITF-gene, but also with adjacent genes in the same haplotype blocks such as EOGT and SETMAR. EOGT is an N-acetylglucosamine transferase and SETMAR encodes a histone-lysine Nmethyltransferase that may be involved in the methylation of H3 histones. Because of the linkage blocks in this population, we cannot assess which of the associations that are putatively causative, but the results suggest that the variations in phenotype may possibly be related to regulatory variants associated with several genes in this haplotype block.
With respect to the saddle, we found a strong association with a region on chromosome 5, covering MC1R, with GALNS and SLC7A5 in the same haplotype block. Three genes have previously been suggested to be involved in the expression of the saddle tan phenotype, i.e., ASIP, RALY, and MC1R but recent research has convincingly demonstrated that RALY is most likely not involved in regulating pigmentation [10]. Saddle tan requires that the dog has the a t (tan points) allele of the ASIP (Agouti Signaling Protein) that is also associated with the tan points or black-and-tan phenotype [14]. In this study, we did not find any associations with ASIP, which could possibly indicate that the beagles studied here were fixed on this allele.
The expression of the saddle tan phenotype is also relying on the ability of the dog to produce black pigment. Hence, if the dog is homozygous for the e-allele (recessive red) of the MC1R-gene it will only be able to express a pheomelaninistic phenotype (no black pigments) [26,27]. This could possibly cause the lack of black pigment in our three beagles excluded from the analysis of the saddle coloration due to the lack of a saddle. It is possible that the A-allele at the BICF2P831934 SNP locus could be linked to the recessive red variant of the MC1R gene. Although speculative, this is supported by the fact that beagles heterozygous at this locus have a significantly more faded saddle than those carrying the GG-genotype and express a more solid black saddle.
The melanocortin 1 receptor (MC1R) controls which type of melanin that can be produced by melanocytes, and when activated stimulates the cells to produce eumelanin. At least three alleles have been described in dogs, where loss-of-function mutations cause an increased production of phaeomelanin, leading to red coat color [7]. It is not clear in what way the association we have found can be explained, i.e., by which mechanisms MC1R could affect the degree of whiteness. However, similar associations (based on hereditary patterns) have been reported in horses, where genotype on the E-locus (MC1R) has been linked to the quantitative expression of white markings in the face region [28]. It is therefore possible that whiteness is associated with regulatory mutations affecting expression levels of MC1R and also that complex's interactions with other genes may play a role. For example, the KIT gene (dominant white locus) is related to white spotting in many species, [29]. It encodes the tyrosine kinase receptor and interacts reciprocally with MITF but has not been implicated to play a role in relation to MC1R, and in fact, its effect on whiteness in dogs is unclear [29].
It is of course important to remember that we have studied only one specific population of beagles, and the alleles present in this population may not be representative for all other beagles. Hence, several other alleles may exist in the breed, and some of them may have subtle functional differences and may therefore also have been overlooked in the classical studies of Little from 1957 [9].
In conclusion, we find that the complex variation in loss of pigmentation in beagles, including the quantitative variation in expression of the saddle, can be attributed to the genetic variation on mainly two loci. These loci contain two genes, previously found to be associated with pigmentation, namely MC1R and MITF, although other genes are also linked to significant markers associated with the phenotypes. Our results suggest that the large variation in pigmentation patterning may possibly be a function of the regulatory control of MC1R and MITF genes.