The MC1R and ASIP Coat Color Loci May Impact Behavior in the Horse

Abstract

Shared signaling pathways utilized by melanocytes and neurons result in pleiotropic traits of coat color and behavior in many mammalian species. For example, in humans polymorphisms at MC1R cause red hair, increased heat sensitivity, and lower pain tolerance. In deer mice, rats, and foxes, ASIP polymorphisms causing black coat color lead to more docile demeanors and reduced activity. Horse (Equus caballus) base coat color is primarily determined by polymorphisms at the Melanocortin-1 Receptor (MC1R) and Agouti Signaling Protein (ASIP) loci, creating a black, bay, or chestnut coat. Our goal was to investigate correlations between genetic loci for coat color and temperament traits in the horse. We genotyped a total of 215 North American Tennessee Walking Horses for the 2 most common alleles at the MC1R (E/e) and ASIP (A/a) loci using previously published PCR and RFLP methods. The horses had a mean age of 10.5 years and comprised 83 geldings, 25 stallions, and 107 mares. To assess behavior, we adapted a previously published survey for handlers to score horses from 1 to 9 on 20 questions related to specific aspects of temperament. We utilized principle component analysis to combine the individual survey scores into 4 factors of variation in temperament phenotype. A factor component detailing self-reliance correlated with genotypes at the ASIP locus; black mares (aa) were more independent than bay mares (A_) (P = 0.0063). These findings illuminate a promising and novel animal model for future study of neuroendocrine mechanisms in complex behavioral phenotypes.

Genes controlling coat color dictate the quantity and distribution of melanin pigments in the skin and hair. In many mammalian species, these same genes often have pleiotropic effects on behavioral phenotypes. For example, alterations of the stress response are noted in colorphase foxes (Keeler et al. 1968; Keeler et al. 1970; Kukekova et al. 2012) and agouti-variant deer mice (Harris et al. 2001). Aggressiveness is associated with red coat color versus black Cocker Spaniel dogs (Pérez-Guisado et al. 2006). Additionally, a fear reaction specific to Icelandic horses with the silver coat color was recently described (Brunberg et al. 2013). Across diverse species, individuals with loss-of-function ASIP variants tend to be calmer and have a lesser stress or fear response. The 2 genes primarily responsible for determination of base coat color in the horse (Equus caballus) are the Melanocortin-1 Receptor (MC1R) and its antagonist, Agouti-Signaling Protein (ASIP) (Rieder et al. 2001; Bailey and Brooks 2013, Table 1; Figure 1). Notated as the “extension” (E/e) locus, MC1R is epistatic to ASIP, or the “agouti” (A/a) locus.

Table 1.

Genotype and corresponding base coat color phenotype in the horse

		MC1R genotype
		EE	Ee	ee
ASIP genotype	AA	Bay	Bay	Chestnut
	Aa	Bay	Bay	Chestnut
	aa	Black	Black	Chestnut

Figure 1.

Photos illustrating the variations of base coat colors in the Tennessee Walking Horse. Examples of a dark chestnut (a), light chestnut (b), black (c), and bay (d) horse.

Called chestnut, red or sorrel in horses, a recessively inherited red coat results from a single base substitution in MC1R, leading to loss-of-function in the receptor and a coat bearing little to no black eumelanin that might obscure the underlying red/brown pheomelanin (Rieder et al. 2001). MC1R loss-of-function mutations in humans also lead to red hair color, as well as increased sensitivity to the sun, thermal discomfort, and a reduced pain tolerance threshold (Mogil et al. 2003; Liem et al. 2005). Mogil et al. (2003) specifically observed that, in comparison to their darker-pigmented counterparts, red-haired women had a greater analgesic response to certain opioid drugs.

The second gene influencing base coat color in the horse is ASIP, a paracrine signaling protein that acts as a ligand antagonist for the MC1R receptor by competing with α-melanocyte-stimulating hormone (α-MSH) for binding (Rieder et al. 2001). In the horse, a loss-of-function mutation in ASIP, a recessive 11 base pair (bp) deletion resulting in a frameshift, causes the uniform production of eumelanin across the body and the appearance of a black coat in the homozygote (Rieder et al. 2001). Bay base coat color results if at least one of the ASIP alleles is the functional wild-type. There is evidence for a relationship between stress and the ASIP gene in deer mice. There is increased reactivity to stressors in black-colored individuals, and the black color in deer mice is due to a loss-of-function ASIP mutation similar to black horses (Harris et al. 2001). There is also a well-documented relationship between ASIP alleles and food intake; those with functional variants of the gene often become obese (Harris et al. 2001), though an obesity phenotype was not examined in the current study.

To date, the only studies assessing the relationship between coat color and behavior phenotypes in the horse have examined just one specific coat color pattern, Silver. However, the Silver locus is also responsible for ocular abnormalities and blindness, confounding measures of startle behavior using visual stimuli (Ramsey et al. 1999; Brunberg et al. 2013). This study aimed to determine the relationship between the presence of alleles of the genes for base coat color and behavioral characteristics in the horse. Results of this research should provide people involved with horses—owners, riders, trainers, etc.—a greater understanding or foresight in choosing and working with horses of a particular color.

Materials and Methods

Horse Sampling

A total of 276 Tennessee Walking Horses (TWH) were sampled in conjunction with another study (Staiger 2015). The TWH breed was chosen for this study due to its popularity for use in trail/pleasure riding and show competition, and the TWH also possesses many coat colors providing a variety of MC1R and ASIP genotypes. Owners and trainers of horses either volunteered for participation in the study and mailed in required materials or were visited in person by lab staff at private farms or at horse shows to collect the necessary data. Sampled horses included in the study resided in the continental United States or in Alberta and Quebec, Canada. Data were collected from September 2011 through August 2014. Approximately 50–100 hair bulbs were pulled from the underside of the tail or from the mane of individuals to extract DNA. Pedigrees, registration information, and photographs were collected from each horse for identification purposes.

A behavior survey (Staiger 2015) was adapted from Momozawa et al. (2003 and 2005), containing 20 questions referring to temperament traits with answer scores rated on a scale from one (1) to nine (9) (Supplementary Table S1). Answers to the survey questions quantified the quality of a given horse’s temperament. Those who filled out the survey relayed how long they had known the horse, their exact relationship with the horse (e.g. owner, trainer, or barn manager) and their amount of exposure and familiarity to the animal’s behavior.

Two candidate genes, those for Melanocortin-1 Receptor (MC1R) and Agouti-Signaling Protein (ASIP), were chosen for their conservation in different species and potential to impact behavior in the horse. A behavior survey (Staiger 2015) was used to subjectively assess the behavior of horses. Such a methodology was previously utilized successfully in the horse (Momozawa et al. 2003; 2005), fox (Kukekova et al. 2012), and dog (Hsu and Serpell 2003). Seaman et al. (2002) validated a behavior survey as mirroring the results of physical handling and novel object tests like those of Wolff et al. (1997) and Visser et al. (2001) in determining temperament in the horse.

In total, 215 horses contributed to the temperament scores after filtering for incomplete behavior surveys and relatedness. The horses sampled had a mean age of 10.5 years, ranging from 1 to 32 years. The samples consisted of an equal gender ratio with 83 geldings and 25 stallions, for a total of 108 males, and 107 mares or females.

MC1R and ASIP Genotyping

DNA was extracted from 5 to 15 collected hair bulbs per horse by 2 methods: a kit-based procedure for alcohol precipitation or a crude lysis technique. The kit-based hair prep protocol was modified from the Gentra PUREGENE^TM DNA Isolation Kit (Cook et al. 2010). The crude lysis hair preparation followed the method outlined by Locke et al. (2002). Samples were genotyped at the coat color gene loci following the restriction fragment length polymorphism (RFLP) methods at MC1R (Marklund et al. 1996; Royo et al. 2008) and polymerase chain reaction (PCR) techniques at ASIP (Royo et al. 2008).

DNA was amplified by PCR using FastStart Taq DNA Polymerase and included all reagents per the manufacturer’s recommended conditions (Roche Diagnostics), with the primers obtained from Marklund et al. (1996) and Rieder et al. (2001), based on the assembly of the equine genome, EquCab2.0 (Table 2). Thermocycling on an Eppendorf Mastercycler® EP Gradient or Thermo Electron Corporation Px2 was also according to the manufacturers’ recommendations with respective annealing temperatures presented in Table 2 and run a total of 40 cycles for both amplicons.

Table 2.

Polymerase chain reaction (PCR) primers and annealing temperatures for MC1R and ASIP

Gene	Annealing temperature	Primer sequences (Forward and Reverse)
ASIP	58°C	Forward: 5′ CTT TTG TCT CTC TTT GAA GCA TTG 3′
		Reverse: 5′ GAG AAG TCC AAG GCC TAC CTT G 3′
MC1R	63°C	Forward: 5′ CCT CGG GCT GAC CAC CAA CCA GAC GGG GCC 3′
		Reverse: 5′ CCA TGG AGC CGC AGA TGA GCA CAT 3′

Genotypes for the MC1R locus were determined from a RFLP test using Taqα1 (1.0U per reaction; New England Biolabs, Inc.) incubated at 65°C overnight. The resulting MC1R products were visualized by electrophoresis following standard conditions on a 1% agarose gel (Omnipur Agarose; EMD Chemicals, Inc.). The presence of 150 and 200 base pair (bp) fragments indicate genotype homozygous recessive (ee), a single 350bp fragment (not digested) indicates a homozygous dominant (EE) horse, while all 3 bands (150, 200, and 350bp) represents the heterozygote (Figure 2). The ASIP products were visualized by electrophoresis following standard conditions on a 4% agarose gel (Omnipur Agarose, EMD Chemicals, Inc.). Genotyping at the ASIP locus was called via the presence of a 91-bp fragment indicating genotype homozygous recessive (aa); a 102-bp fragment indicates a homozygous dominant (AA), and a heterozygote (Aa) had both bands (91 and 102bp) (Figure 2). Genotypes from the MC1R and ASIP loci determined the base coat colors of the horses sampled (Table 1, Bailey and Brooks 2013).

Figure 2.

MC1R (a) and ASIP (b) genotyping gels showing the ladder (L), as well as the non-template control (NTC) in the ASIP gel. Homozygous recessive, homozygous dominant, and heterozygous conditions are genotyped from left to right.

Statistical Analysis

Statistical analyses were performed using the software JMP Pro 10 (SAS Institute, Inc., Cary, NC). To analyze the temperament trait data, principle component analysis (PCA) was conducted on the scores of the 20-question survey. The multivariate method of PCA on covariances was employed and principle components (PCs) were retained for analysis based on Peres-Neto et al. (2005) recommendations. PCs were connected to the original behavioral qualities based on their eigenvalues (Hatcher 1994). Factor analysis using prior communality and Varimax rotation for orthogonal transformation was carried out on the significant PCs to correct for human-derived non-normal distribution of the behavior survey scores, yielding factor components (FCs).

A standard least squares multiple regression was used to model the dependence of the behavior traits on genetic base coat color, gender (mare, gelding, or stallion), and age. This method controlled for the effects of age, gender, genotype, and their interactions with the behavior traits to explain the best set of covariances to use in analysis of variance (ANOVA). Owner relationship and observation frequency were also examined, but found to be not statistically significant and therefore not included in the model. ANOVA was used to detect significant differences between the FCs and genetic base coat color by gender. Statistical significance was set at any P value equal to or below 0.05.

Data Availability

Genotypes, behavior scores, and sampling information for all horses used in these analyses are available in a table submitted to Dryad.

Results

Allele Frequencies

The 215 TWH were successfully genotyped at the MC1R and ASIP loci. Genotypes were more evenly distributed at the MC1R locus than at ASIP. Base coat color phenotypes of the horses and allele frequencies were determined based on the genotypes (Table 3). At MC1R, the allele frequencies were 0.395 for E and 0.605 for e; at ASIP, A had a frequency of 0.116 compared to a with 0.884.

Table 3.

Distribution of MC1R and ASIP genotypes and base coat color phenotypes of the 215 Tennessee Walking Horses

		MC1R genotype
		EE	Ee	Ee
ASIP genotype	AA	0	0	3
	Aa	6	21	17
	aa	31	75	62

Coat Color and Temperament

PCs 1 through 4 were retained for analysis as they accounted for 34.1, 9.27, 7.80, and 5.78% of the data variance, respectively. Four FCs were derived from the significant PCs retained for analysis, accounting for 16.9, 23.5, 8.59, and 7.94% of the data variance. The temperament traits that described each FC were determined by graphing the eigenvectors of the surveyed temperament traits against the FC scores (Supplementary Figure S1). Factor 1 reflected overall trainability, with positive FC1 scores indicating a less trainable horse versus negative scores defining a more trainable horse. Factor 2 described neophobia, with negative scores correlated to a fearful or flighty horse and positive scores a calmer animal. Negative FC3 scores explained a less friendly, more competitive horse, while positive scores a friendlier, less competitive horse among peers. Factor 4 strongly loaded with self-reliance, indicating that positively scoring horses were restless when left alone and negatively scoring individuals more at ease when solitary.

The results of the standard least squares multiple regression showed significant values delineating a dependence of behavior on genotype at the ASIP locus and gender for FC1 and FC4 (P = 0.0449 and P = 0.0383, respectively) and with age for FC2 and FC3 (P = 0.0010 and P = 0.0471, respectively). No significant association was identified with the MC1R locus or combined genotype effects for ASIP and MC1R in this sample population. Analyses of variance indicated black mares (aa) had statistically significantly lower FC4 scores than bay mares (A_) (P = 0.0063), reflecting a lower behavior survey score or more self-reliant horse that is at ease when alone (Figure 3).

Figure 3.

Mean Factor 4 “self-reliance” scores in mares are significantly different between the dominant and recessive ASIP genotypes.

Discussion

There are 2 genes that determine base coat color in the horse that are anecdotally reported to affect behavior: MC1R and ASIP. Our candidate gene approach using a behavior survey found an association between ASIP genotype and temperament that may influence how humans interact with horses. A black mare (aa as defined by ASIP) is more self-reliant and perhaps preferable in a solitary situation than a bay-colored horse (A_). This is reflected in foxes, where recessive alleles at ASIP correlated to smaller adrenal size and thus less stress response-related hormone (adrenaline or epinephrine) production (Keeler et al. 1968). Similar effects were described in deer mice and rats (Harris et al. 2001), and analogous results were obtained in the English Cocker Spaniel dog (Podberscek and Serpell 1997; Pérez-Guisado et al. 2006). In a loss-of-function mutation at ASIP, α-MSH binds to melanocortin receptors which inhibit stress-induced glucocorticoid release, decreasing the stress response (Harris et al. 2001). When α-MSH binds, it also plays a part in dopamine pathways, increasing feeding and grooming behaviors, while the protein product of ASIP binds antagonistically (Roseberry et al. 2015). Implications of this finding include considerations for change in the treatment of horses and potential human applications. Similar human hair color genetics may show the same correlations with behavior and physiology.

Age correlated positively with FC2 and negatively with FC3, indicating that as a horse aged it exhibited a less neophobic behavior phenotype and a more competitive nature, respectively. The effects of genotype and age could not be separated in this study, but the results seen reflected those of Baragli et al. (2014).

Further study in this area is warranted. Ideally, the research sample should be expanded, a more objective measure of behavior developed, and additional coat color genes analyzed. Distributions of gene alleles and behavior scores were skewed in the TWH (Table 3), and this bias may have affected our results. The TWH in this study had a high a allele frequency, likely due to coat color selection following the success of a black-colored World Grand Champion of the breed from 1946 to 1965, though there was selection for chestnut (ee as defined by MC1R) for the same reason at a later time in breed development in the 1980s–1990s (Womack 1994). Line breeding and selection of TWH may have also affected behavioral tendencies, as a calm disposition is desired both innately and requisitely by the breed association (Tennessee Walking Horse Breeders’ and Exhibitors’ Association 2011), limiting the temperament variation we expected. A revised behavior survey or another type of behavioral assessment can be performed for a more objective measure of temperament; these assessments include arena tests, handling tests, and novel object tests (Wolff et al. 1997) or measuring heart rate (Visser et al. 2002). Other gene loci that encode coat color could influence behavior in the horse. One study (Brunberg et al. 2013) examined the gene for silver coat color (PMEL17) and a novel object test, which may have been a more objective phenotype than the present study, yet did not consider the impact of the genes for base coat color. A genome wide association study (GWAS) might be a first step in future studies to find other such significant relationships, by analyzing the allele variation within many genes influencing coat color and behavior.

From the current study, the genes governing base coat color in the horse seem to influence behavior. Historically chestnut horses were considered excitable, but horse breeds, coat color, and their associated behavior may have changed since that proposition was originally made. We found no significant relationship with behavior at the MC1R locus, which codes for chestnut coat color, though detection of such an effect may have been hampered by the low frequency of the e allele in this population. The correlation between the recessive ASIP allele and a more independent temperament will help guide horse choice and care, as well as improve our understanding of pigmentation and behavioral pathways.

Supplementary Material

Supplementary material can be found at http://www.jhered.oxfordjournals.org/.

Funding

This work was supported by the Cornell University College of Agriculture and Life Sciences Dextra Undergraduate Research Endowment Fund.

Data Availability

Data deposited at Dryad: http://dx.doi.org/doi:10.5061/dryad.3q111