Coat colour variation has fascinated humans for centuries and consequently it has become one of the most extensively studied traits, mainly in mice [1] but also in domestic animals [2], [3]. There is an extensive diversity of white spotting patterns within and between dog breeds caused by genetic factors and stochastic effects during melanocyte development. White spotting indicates an absence of melanocytes in the hair follicles and/or skin due to a failure of melanoblast migration, proliferation or survival during development. In dogs, there is one major white spotting (S) locus that was defined by C. Little during the 1950s [4]. He described four different alleles at this locus with phenotypic effects ranging from solid (S, Figure 1A), to a completely white coat, caused by homozygosity for the Extreme white allele (s^w, Figure 1B). The two intermediate phenotypes were named Irish spotting (sⁱ, Figure 1D) and piebald (s^p, Figure 1E). Irish spotting is characterized by modest white spotting, often present as a white collar and a white belly, as demonstrated by breeds such as the Bernese Mountain dog and Basenji. Piebald-coloured dogs display limited to extensive white spotting and the phenotype is observed in several breeds, including the Beagle and Fox Terrier. Two S alleles are segregating in Boxers: Solid (S) and Extreme white (s^w). These give rise to three different phenotypes: solid (S/S), flash (S/s^w, Figure 1C) and extreme white (s^w/s^w). The flash (S/s^w) phenotype is similar to Irish spotting (sⁱ/sⁱ) and is therefore often called pseudo-Irish.

In 2006 and 2007, three groups independently mapped some or all of the white spotting phenotypes to the MITF (microphtalmia-associated transcription factor) locus [5]–[7]. The highest resolution was obtained by Karlsson et al. [6] who mapped the Spotting locus to a 1 Mb region on chromosome 20 based on a genome-wide association study comparing 10 homozygous white (s^w/s^w) and 10 homozygous solid (S/S) Boxers. Fine-mapping to a 100 kb region which included MITF was achieved by including a second breed, bull terrier, segregating for the same two alleles [6].

MITF was an obvious candidate gene because it encodes a transcription factor controlling neural crest-derived melanocyte development and migration [8], [9]. MITF has nine alternative promoters that produce multiple isoforms expressed in different tissues. All isoforms share the sequence encoded by exons 2–9, but have unique amino acids in their N-termini, determined by distinct first exons. There is clear conservation across mammals at this gene, as several of the MITF isoforms expressed in human and mouse have also been identified in dog [10], including the melanocyte-specific MITF-M isoform.

To date, 29 mutations in murine Mitf have been identified which affect the development and function of melanocytes in the skin, eye and inner ear [1]. Some variants affect vision by reducing eye size (microphtalmia) or cause early onset hearing disorders [11]. In humans, deleterious MITF mutations cause disorders of vision and hearing, including the Waardenburg and Tietz syndromes [9], [12], [13]. Deafness has also been recorded in white dogs, where approximately 2% of white dogs (s^w/s^w) present with bilateral deafness and 18% are unilaterally deaf [14]. The majority of mutations reported in mice and humans that cause severe pleiotropic effects are generally loss-of-function mutations affecting the coding regions [15]. This is not the case with dog MITF alleles. A comparison of S and s^w haplotypes, using BACs from an S/s^w heterozygote, across the 100 kb canine white spotting candidate region revealed 124 sequence polymorphisms, all of which were located in non-coding regions [6]. This demonstrated that the extreme white coat colour phenotype is controlled by one or several regulatory mutations. This hypothesis is strongly supported by the fact that coloured patches on white spotted dogs display normal pigmentation. Thus, this suggests that the canine MITF variants primarily affects migration and survival of melanocytes during development, but have no or only minor effects in mature melanocytes in the hair follicle; pigmentation of the hair requires MITF protein expression [16].

Further analysis of these 124 polymorphisms resulted in a short list of three candidate mutations within or in the vicinity of the MITF-M promoter, and one in the MITF-1B exon located upstream of MITF-M. The first of these is a canine-specific short interspersed nucleotide element (SINEC-Cf element), located about 3 kb upstream of the MITF-M transcription start site (TSS). The SINE insertion was only found in dogs presenting the extreme white (s^w/s^w) or piebald (s^p/s^p) phenotypes, and was absent in Irish-spotted (sⁱ/sⁱ) and solid (S/S) dogs [6]. This very strong association between the SINE insertion and the extreme white and piebald phenotypes was confirmed in a subsequent study based on 324 dogs from 45 breeds although a few exceptions from this rule were noted [17]. The second candidate (SNP#21), a SNP located approximately 1.2 kb upstream of MITF-M TSS, occurs in a highly conserved region and the A allele at this locus is associated with white spotting alleles [6]. The third polymorphism is a variable length polymorphism (Lp) approximately 100 bp upstream of the MITF-M TSS. Long variants of the Lp (LpWhite) are associated with all three white-spotting alleles (s^w, s^p and sⁱ), whereas all solid dogs examined carried short Lp variants [6], from here on named LpSolid. The fourth candidate mutation, a 12 bp deletion in exon 1B (Exon1B_del) also showed a very strong association with white spotting. It was found on all Extreme white and Piebald chromosomes tested, but it was also found in the heterozygous state in 4 out of 76 solid dogs. This previous work showed that the Extreme white and Piebald alleles share the same sequence variants for the SINE, SNP#21, Lp and Exon1B_del polymorphisms, implying that other sequence variants must explain the phenotypic differences between the two alleles.

The aims of the current study were two-fold: (i) to screen for the presence of the candidate mutations in wolves that are expected to have the S wild-type genotype and (ii) to evaluate the functional significance of the candidate mutations associated with the MITF-M promoter, alone and in combination. Given the high level of conservation across mammals at this region, any mechanistic insights gained in the canine model could also be used to further understand the role of the master melanocyte regulator, MITF.

Coat colour was probably one of the first traits altered during dog domestication. White spotting in dogs can be traced back as long as there are written records or dog portraits, and have been present for thousands of years. Changes in coat colour occur during the domestication due to selection against wild-type colour; as a measure to facilitate animal husbandry (less camouflage), to facilitate the identification of undesired domestic/wild crossbred animals or for fashion and because of relaxed purifying selection [19]. In fact, Columella, the Roman authority on agriculture, wrote already in the first century AD that shepherds prefer white sheep dogs “because it is unlike a wild beast, and sometimes a plain means of distinction is required in the dogs when one is driving off wolves in the obscurity of early morning or even at dusk” [20]. Our previous study demonstrated that the Extreme white (s^w) allele in Boxers is not associated with coding changes in MITF [6]. This conclusion was achieved by taking advantage of the fact that the reference individual (Tasha) for the dog genome assembly showed the flash phenotype (S/s^w). This allowed us to sequence BACs representing each of the two alleles. A careful examination of the entire non-coding region showing the strongest association to white spotting, combined with a screen across a diverse panel of dogs, revealed four candidate mutations that showed the strongest association to white spotting. The present study demonstrated that none of these four candidates could be the sole causal mutation for Extreme white. This is because the four candidate mutations either occur at a relatively high frequency in the wild ancestor of dogs or do not perfectly co-segregate with phenotype among dog breeds. It is possible that these candidate mutations have an effect on pigmentation in wolves and they may even cause some subtle forms of white spotting, as this could be difficult to notice in a wolf with its long fur, basic grey colour and greying with age effect. However, we are convinced that the Extreme white allele is not common in wolves, as to the best of our knowledge Extreme white has never been reported in wolves. This colour morph would also be strongly selected against in a natural population due to both the lack of camouflage (except in arctic environments) and the well established association with deafness [14].

Our results combined with the previous characterization of the canine MITF locus strongly suggest that the different white spotting alleles (s^w, s^p and sⁱ) are not caused by three independent mutation events, but rather reflect haplotype effects due to different combinations of multiple causative regulatory mutations selected during dog domestication and breed formation. This interpretation is supported by the observation that there is more extensive sequence similarity among the s^w, s^p and sⁱ haplotypes than between any of these and haplotypes associated with solid colour [6]. For instance, Extreme white and Piebald alleles may share all four candidate mutations examined in this study. Thus, MITF in dogs is another example of ‘evolution of alleles’ by consecutive accumulation of multiple causal mutations that occurs in domestic animals at loci under strong selection [21].

This study has provided strong genetic and functional evidence that the length polymorphism (Lp) in the MITF-M promoter is one of the causal polymorphisms. It is probably the most important regulatory mutation causing decreased MITF-M transcription, and consequently affecting white spotting in dogs. Firstly, we have not yet found any overlap between the Lp alleles associated with white-spotting or solid colour in dogs (Table 2). The white-spotting alleles had a longer C mononucleotide repeat and all, except the one found in Dalmatians, had a longer total repeat region. We detected an extensive genetic diversity of the Lp among wolves. We found as many as 15 different alleles when only examining 17 wolves. A very striking and unexpected observation was the considerable Lp allele-sharing between white-spotting alleles in dogs and wolves, but no allele-sharing between solid alleles and wolf alleles. This is unexpected because Solid is assumed to be the wild-type allele at this locus. We propose that the explanation for this pattern is that there has been selection both for and against white spotting (thus for solid colour in the latter case) during dog domestication and breed formation. In the early history of domestication there was selection against wild-type colours, while during breed formation there was selection for breed-specific colour such as white spotting in Boxers and solid black colour in Labradors. A repeat polymorphism like this is expected to have a very dynamic evolution, where new alleles are generated by slippage during DNA replication, consistent with the rich Lp diversity among wolves. Thus, the very tight Lp distribution observed in solid dogs and the lack of overlap with alleles found in wolves is hardly in agreement with selective neutrality.

The extensive Lp diversity may have no obvious phenotypic effects in wolves, because such effects may depend on epistatic interaction with other sequence polymorphisms not present in wolves at MITF or at other loci in the genome. However, it is also possible that the MITF polymorphisms contribute to coat colour diversity in wolves, ranging from dark grey to almost white colour. This is worth exploring using samples from phenotypically well-characterized wolves.

This study also provided functional data supporting the conclusion that the Lp directly affects white spotting. Luciferase reporter constructs containing the longer Lp variants, associated with Extreme white (s^w) in Boxers and Dalmatians and with the Irish spotting (sⁱ) allele in Bernese Mountain Dogs, had all significantly lower activities compared with constructs containing the short Lp Solid variant, indicating that the Lp affects MITF-M transcription (Figures 3B and 4). The Extreme white (s^w) and Solid Lp variants in Boxers differ in size by four nucleotides, which corresponds to approximately one half turn of the DNA helix. This may have important consequences for the functional interaction between transcription factors binding to either flanking side of the Lp (Figure 3C). Transcription factors that face the same side of the DNA helix in the context of the short Lp Solid variant will face opposite sides of the DNA helix in the Lp White variant, which is expected to result in reduced capacity to establish functional interaction. Previous studies have shown that transcription is markedly affected when crucial promoter elements are separated by half a helical turn [22]–[23]. In favour of such a scenario is the cluster of LEF1, SOX10 and PAX3 binding sites which are located approximately 200 bp upstream of the Lp. Importantly, the PAX3 and SOX10 transcription factors have been shown to physically interact and to be critical for MITF-M transcription [24]. Furthermore, a 10 bp insertion that disrupts a PAX3 binding site in the MITF-M promoter is causing the “splashed white” phenotype in horses [25]. The corresponding PAX3 site in the dog promoter is the one located between the Lp and the TATA-box (Figure 3C). LEF1 is known to physically interact with MITF and to be involved in transcriptional self-activation of the MITF-M promoter [26]. Our data support previous findings from other species that the cluster of LEF1, SOX10 and PAX3 binding sites are crucial for MITF-M expression (Figure 3C). Our data also demonstrate that it is unlikely that the Lp region has a direct effect on transcription factor binding, as we observed no effect of the Lp variants unless the upstream 200 bp region containing the cluster of LEF1, SOX10 and PAX3 sites were included in the construct.

The SINEC-Cf insertion located about three kb upstream of the MITF-M transcription start site had a widespread distribution among wolf populations, implying that this must be an ancestral polymorphism rather than a recent derived mutation. However, the SINE insertion shows a very strong association to the Extreme white and Piebald haplotypes but is rare or absent among Solid and Irish spotting haplotypes (those associated with no spotting or the least white spotting). Our reporter assay data suggested that the SINE-region acts as a weak silencer element and that it acts in an additive mode together with the long Lp variants to reduce MITF-M promoter activity associated with s^w and s^p. The mechanism underlying the repressive function of the region containing the SINE insertion remains to be defined, but for example, SINE elements are known to be targets for DNA methylation [27]. It is possible that differential methylation within this region is involved in controlling MITF-M transcription. Unfortunately, we have not been able to analyse the MITF-M expression or methylation status during different stages of canine melanocyte development due to the implicit difficulties in sampling dog embryos.

SNP#21 occurs in the vicinity of an evolutionary conserved element but this actual nucleotide position is not strongly conserved. It could be a derived mutation since it is rare in wolves. The association between SNP#21*A and white spotting across breeds is far from perfect and the reporter assay did not reveal any effect on transcriptional regulation. However, this does not exclude the possibility that this substitution could have a minor effect on white spotting, as a reporter assay will never perfectly replicate transcriptional regulation in vivo.

We also provided support for the possibility that the 12 bp deletion in exon 1B is a derived allele that promotes white spotting. This variant was only found in the highly inbred Scandinavian wolves, and was thought unlikely to be causative in the previous study because it was found at a low frequency in some solid dogs [6]. However, the present study has shown that no MITF polymorphism shows complete agreement with phenotype and in fact, the insertion in exon 1B, together with the Lp, shows the best concordance with white spotting. The deletion appears to be fixed in dogs with the extreme white and piebald phenotypes, but rare or absent in wolves, solid dogs and dogs with the Irish spotting phenotype. No functional assays were attempted for this polymorphism since it occurs in an exon and in our previous study we concluded that it was unclear if this exon is functional in carnivores [6]. However, recently released RNAseq data (January 2104; Broad CanFam3.1/canFam3; http://genome.ucsc.edu) show that MITF exon 1B is transcribed in dogs and was found in one transcript from blood (CUFF.25976.1) and one from lung (CUFF.27047.1) Both transcripts originated from the derived allele associated with Extreme white and Piebald. This finding suggests that the 12 bp deletion in exon 1B may very well affect MITF function during the development of melanocytes. Thus, the extensive white spotting in Extreme white and Piebald dogs may be due to the combined effect of mutations affecting MITF-M transcription and a coding mutation in exon 1B.

The luciferase assay indicated that the Lp variant associated with Irish spotting had an even lower MITF-M promoter activity than the variants associated with Extreme white and Piebald, although the two latter alleles cause more extensive white spotting (Figure 4). However, this is consistent with our interpretation that canine MITF alleles are not due to single mutations but the combined effect of multiple mutations in the MITF region. An important difference between Irish spotting and Extreme white/Piebald is that only the two latter alleles carry the SINE insertion upstream of MITF-1M and the 12 bp deletion in MITF-1B. Another interesting difference between these two groups of alleles is that sⁱ/sⁱ homozygotes show a high degree of symmetric white spotting whereas s^p/s^p and s^w/s^w homozygotes show asymmetric white spotting (Figure 1). This is probably not caused by the nature of the underlying mutations but rather a dosage effect, i.e. to which extent MITF function is affected during melanocyte development, because s^w/S heterozygotes also show symmetric white spotting.

MITF must be one of the loci in the dog genome that has been under strongest positive selection during domestication and breed formation. It is likely that Little's [4] original classification of three white spotting alleles, Extreme white, Piebald and Irish spotting, is an underestimate and that a more extensive allelic diversity is created by combining the variability at the Lp with other sequence variants in the MITF region. For instance, Beagles are considered to be homozygous Piebald but as Little described, they vary extensively from almost solid coloured to a phenotype mimicking extreme white. It is still an open question whether this variability is caused by genetic heterogeneity at MITF or genetic variation at other loci affecting melanocyte development. In order to advance our knowledge about MITF it will be essential to make further genetic studies of dogs that are very well characterized for the white spotting phenotype. We recommend that coat colour is carefully registered when performing genome-wide associations studies of other traits or disorders in breeds with a variable white spotting phenotype, like the Beagle. Such studies will reveal the extent to which the variability in white spotting is caused by genetic heterogeneity at MITF or if the influence of yet unidentified modifying loci is at play. Furthermore, it is essential to resequence the entire MITF region from Piebald and Irish spotting chromosomes since they may harbour sequence variants in addition to the ones evaluated in the current study, which were detected in a sequence comparison of the Extreme white and Solid alleles from a single heterozygous dog [6].