Mammalian Genome

, Volume 15, Issue 10, pp 798–808

Characterization of the dog Agouti gene and a nonagoutimutation in German Shepherd Dogs


  • Julie A. Kerns
    • Departments of Genetics and PediatricsStanford University
    • Fred Hutchinson Cancer Research Center
  • J. Newton
    • Departments of Genetics and PediatricsStanford University
  • Tom G. Berryere
    • Department of Animal and Poultry SciencesUniversity of Saskatchewan
  • Edward M. Rubin
    • Genome Sciences DepartmentLawrence Berkeley National Laboratory
  • Jan-Fang Cheng
    • Genome Sciences DepartmentLawrence Berkeley National Laboratory
  • Sheila M. Schmutz
    • Department of Animal and Poultry SciencesUniversity of Saskatchewan
    • Departments of Genetics and PediatricsStanford University
    • Beckman Center B271AStanford University School of Medicine
Original Contributions

DOI: 10.1007/s00335-004-2377-1

Cite this article as:
Kerns, J.A., Newton, J., Berryere, T.G. et al. Mamm Genome (2004) 15: 798. doi:10.1007/s00335-004-2377-1


The interaction between two genes, Agouti and Melanocortin-1receptor (Mc1r), produces diverse pigment patterns in mammals by regulating the type, amount, and distribution pattern of the two pigment types found in mammalian hair: eumelanin (brown/black) and pheomelanin (yellow/red). In domestic dogs (Canis familiaris), there is a tremendous variation in coat color patterns between and within breeds; however, previous studies suggest that the molecular genetics of pigment-type switching in dogs may differ from that of other mammals. Here we report the identification and characterization of the Agouti gene from domestic dogs, predicted to encode a 131-amino-acid secreted protein 98% identical to the fox homolog, and which maps to chromosome CFA24 in a region of conserved linkage. Comparative analysis of the Doberman Pinscher Agouti cDNA, the fox cDNA, and 180 kb of Doberman Pinscher genomic DNA suggests that, as with laboratory mice, different pigment-type-switching patterns in the canine family are controlled by alternative usage of different promoters and untranslated first exons. A small survey of Labrador Retrievers, Greyhounds, Australian Shepherds, and German Shepherd Dogs did not uncover any polymorphisms, but we identified a single nucleotide variant in black German Shepherd Dogs predicted to cause an Arg-to-Cys substitution at codon 96, which is likely to account for recessive inheritance of a uniform black coat.

Diversity in mammalian skin and hair color is achieved by differential expression and regional distribution of two pigment types: brown/black eumelanin and red/yellow pheomelanin. Switching between the synthesis of these two pigment types is regulated by a paracrine signaling molecule, Agouti protein, which acts as an antagonistic ligand for the melanocortin-1 receptor (Mc1r), a seven-transmembrane protein expressed on the surface of melanocytes (Bultman et al. 1992; Miller et al. 1993; Robbins et al. 1993; reviewed in Jackson 1994). Mc1r activation due to constitutive activity or an exogenous agonist such as α-melanocyte stimulating hormone (α-MSH) promotes eumelanin synthesis; by contrast, Agouti protein inhibits Mc1r activation, causing a switch from eumelanin to pheomelanin synthesis (Lu et al. 1994; reviewed in Barsh 1996; Jordan and Jackson 1998; Ollmann et al. 1998). Biochemical and pharmacologic aspects of this pathway have become apparent over the last decade but were predicated by classical genetic studies in laboratory mice (Searle 1968; Silvers 1979). Gain-of-function Agouti alleles and loss-of-function Mc1r alleles each cause a pheomelanic phenotype, loss-of-function Agouti alleles and gain-of-function Mc1r alleles each cause a eumelanic phenotype, and animals carrying combinations of Agouti and Mc1r alleles with potentially opposite effects exhibit a coat color phenotype predicted by Mc1r rather than the Agouti genotype; in other words, Mc1r is epistatic to Agouti.

Comparative genetic and zoological studies suggest that the antagonistic ligand–receptor relationship between Agouti protein and the Mc1r applies to many vertebrate species (reviewed in Mundy et al. 2003). Mc1r missense mutations that promote eumelanin synthesis have been identified in sheep (Vage et al. 1999), cattle (Klungland et al. 1995), pigs (Kijas et al. 1998), jaguars (Eizirik et al. 2003), bananaquits (Theron et al. 2001), and chickens (Ling et al. 2003); in each of these cases, the mutations are thought to cause constitutive receptor activation, i.e., gain-of-function, and exhibit dominant transmission. Conversely, Mc1r mutations that promote pheomelanin synthesis have been identified in cattle (Klungland et al. 1995; Joerg et al. 1996), pigs (Kijas et al. 1998), chickens (Takeuchi et al. 1996), horses (Marklund et al. 1996), humans (Ha et al. 2003; Sturm et al. 2003), bears (Ritland et al. 2001), and dogs (Everts et al. 2000; Newton et al. 2000); in each of these cases, the mutations are thought to cause a loss-of-function and exhibit a recessive pattern of inheritance.

The domestic dog, Canis familiaris, is especially well-suited to studies of coat color genetics. Its evolutionary ancestors in the canid family, whose representatives today include foxes, jackals, coyotes, and wolves, span a range of hair color types and patterns (Vila et al. 1999; reviewed in Wayne and Ostrander 1999; Savolainen et al. 2002). Domestication and selective breeding have established a large number of different breeds with a diversity of pigmentation phenotypes including uniform pheomelanin (the Irish Setter, Vizsla, or Golden Retriever), uniform eumelanin (the Newfoundland, Portuguese Water Dog), and a mixture of pheomelanin and eumelanin in a distinctive “black-and-tan” pattern (the Doberman Pinscher or Rottweiler). Two- and three-generation kindreds with different coat color phenotypes are common and often readily available from an active community of dog breeders and owners with a deep interest in genetics. Most important, certain aspects of pigment-type switching in dogs are either not represented in other mammals or are thought to proceed by different mechanisms (Little 1957; Burns and Fraser 1966; Searle 1968). In particular, the uniform eumelanic appearance referred to above for breeds such as the Newfoundland or Portuguese Water Dog has been attributed to an allele of the Agouti locus, AS, that is dominant to ay, an Agouti allele thought to be responsible for the pheomelanic appearance characteristic of fawn-colored Boxers or Basenjis. However, dominance of a “black” Agouti allele to a “yellow” Agouti allele is not consistent with the biochemical action of Agouti protein and the Mc1r described above, and would represent a departure from genetic studies of pigment-type switching in other mammals. Furthermore, we have recently demonstrated that dominant transmission of a black coat color in a large Labrador × Greyhound cross established by Lust and colleagues at Cornell University is unlinked to Agouti and to Mc1r (Kerns et al. 2003).

In most breeds, eumelanic coat color (sometimes described as “self”-coloration) is thought to be dominantly inherited, as in the Labrador × Greyhound cross. In a few breeds, however, a uniform black appearance can be inherited as a recessive trait. This has been most clearly documented for German Shepherd Dogs, in which black progeny may be produced from two sable, saddle-colored, or black-and-tan parents, with a distribution ratio of 1:3 for black vs. nonblack offspring (Carver 1984).

Some of the uncertainty about the biochemistry and genetics of different pigment types in dogs stems from the paucity of molecular studies. We have previously described the isolation and characterization of the dog Mc1r gene and demonstrated complete association between homozygosity for a nonsense mutation, R306ter, and a uniform pheomelanic coat color in the Irish Setter, Golden Retriever, yellow Labrador Retriever, and Samoyed (Newton et al. 2000). We have also described an association between an Mc1r missense variant, M264V, and the presence of a melanistic mask (Schmutz et al. 2003), although a causal relationship has not yet been confirmed by pedigree and/or functional studies. Here we report isolation and characterization of the dog Agouti gene. Linkage studies reveal that recessive inheritance of a uniform black coat in a German Shepherd pedigree cosegregates with Agouti, and molecular genetic studies identify the probable causative mutation.

Materials and methods

Molecular cloning

A Doberman Agouti cDNA clone was isolated in 3 stages by RT-PCR with internal degenerate oligonucleotide primers based on the human, mouse, and bovine sequences, followed by Rapid Amplification of cDNA Ends (RACE) to obtain the 5′ and 3′ ends of the cDNA. Doberman Pinscher puppy tail fragments were collected during routine tail-docking procedures by a community veterinarian. Total RNA was extracted from the yellow-haired skin on the ventral side of the tail, and a partial cDNA was reverse transcribed and amplified with a reverse primer, 5′-A(A/G)NACNC(G/T)(G/A)CANGT(G/A)CA-3′, and a forward primer, 5′-G(G/T)ITTC(C/T)TITG(C/T)TT(C/T)TT(C/T)AC IG-3′. A second round of PCR was carried out with the same forward primer and a nested reverse primer, 5′-AA(G/A)AANC(T/G)(G/A)CA(T/C)TG(G/A)CA-3′. The resulting cDNA, which contains codons 13–119, was cloned and sequenced.

The 5′ and 3′ ends of the cDNA were isolated separately with a RACE kit (Gibco-BRL) containing an “anchor primer” (for 5′ RACE) and an “adapter/amplification primer” (for 3′ RACE). For 5′ RACE, a primer corresponding to nucleotides 421–440 for reverse transcription, a primer corresponding to nucleotides 366–385 for a first round of amplification, and a primer corresponding to nucleotides 324–343 for a second round of amplification were used in conjunction with the anchor primer. The resultant RACE product, covering nucleotides 1–343, was gel-purified, cloned, and sequenced. For 3′ RACE, a primer corresponding to nucleotides 227–247 for a first round of amplification and a primer corresponding to nucleotides 269–288 for a second round of amplification were used in conjunction with the adapter/amplification primer. The resultant RACE product, covering nucleotides 269–788, was gel-purified, cloned, and sequenced.

Genomic fragments that contain exons 1, 2, and 3 were isolated from a size-selected bacteriophage lambda library constructed from Doberman Pinscher genomic DNA using the cloned cDNA as a hybridization probe. Fragments corresponding to exons 2 and 3 were subsequently used to screen a Doberman Pinscher BAC genomic library generously provided by Emmanuel Mignot and colleagues (Li et al. 1999). Among several BAC clones that were isolated, a 180-kb clone, 69A22, was sequenced in its entirety and used to infer the genomic structure depicted in Fig. 1.
Fig. 1

Molecular characterization of the dog Agouti gene. (A) Location and size of STRs and Agouti exons in BAC69A22 (exon sizes are indicated relative to each other, not the entire BAC), together with translational initation (ATG) and termination (TGA) codons. BAC69A22 is 175633 bp in length; exons 2–4 of the Doberman cDNA are located at positions 165124–170040. Exon 1 of the Doberman cDNA is located at positions 129136–129266. The corresponding region of the red fox cDNA (GenBank Accession Y09877) is 98% identical to Doberman genomic DNA at positions 113361–113417 and 127828–127902. (B) Exon size and sequence boundaries. (C) Comparison of predicted Agouti protein sequences from several mammals. Dots represent amino acids identical to dog Agouti and dashes represent the absence of an amino acid. The position of the R96C nonagouti mutation is indicated with an arrow. Functional domains in the protein are indicated above with gray boxes. (D) Comparison of Agouti cDNA sequences between the red fox (GenBank Accession No. Y09877) and Doberman Pinscher. The protein-coding (residues 149–544 in the fox and residues 142–537 in the dog, both in uppercase) and 3′ untranslated regions are very similar (gray shading), but the 5′ untranslated regions are not (solid lines). The division between regions of low and high sequence similarity occurs very close to the boundary between exons 1 and 2. Comparing the 778-nt fox and 771-nt dog cDNA sequences to each other reveals a 64-nt region in which there are 633 identities and 4 gaps; this 647-nt region corresponds to positions 134–778 in the fox and 127–771 in the dog.

Genetic mapping and analysis of the nonagouti German Shepherd Dog mutation

We identified 3 short tandem repeat (STR) sequences within the 69A22 BAC as follows: DBar1 is a 231-bp (CTTTT)n STR amplified with 5′-CATGATCCTGGTCCCAC GTA-3′ and 5′-TGTTTCAATGTGCTATGCACTG-3′. DBar2 is a 150 bp (CA)n STR amplified with 5′-TCCGGATACTGTTATCTCCCAT-3′ and 5′-CCA ATTGGCTGTGTACCTCA-3′. DBar3 is a 300-bp (CTT)n(CA)n STR amplified with 5′-GCATCCCC TTCCTCATCTAA-3′ and 5′-TTCCCCGTCCCTT TAATAAT-3′. The PCR conditions for each marker were 94°/30 sec, 55°/30 sec, 55°/30 sec, for 35 cycles.

To place BAC 69A22 on the canine genetic map, 5 full-sib pedigrees (Australian Shepherd, German Shepherd, and Great Dane) from our assembly of dog families and 1 German Shepherd Dog half-sib family from the DogMap group (Lingaas et al. 1997) were genotyped for DBar1, DBar2, and 3 markers on CFA24, AHT118, AHT125, and ZuBeCA14 (primer information available through ). To investigate linkage of Agoutito a uniform black coat color, we made use of a German Shepherd Dog litter in which black-and-tan parents had produced 3 black progeny and 6 black-and-tan pups as depicted in Fig. 2. Once linkage between Agouti and recessive black was established, we searched for the causative mutation by PCR-amplifying and sequencing fragments that contained each of the 4 exons as follows: exon 1 (261 bp), 5′-CACCCAACACACTTCTGCG-3′ and 5′-TACCATACCAAAACATCTGC-3′; exon 2, (265 bp) 5′-AGCCTGACTGCTTCTCTGTTCC-3′ and 5′-TGTTTGACATCCTGTAGCCCTT-3′; exon 3 (176 bp), 5′-AGCTCGTGGCCTCCAAAAACA TA-3′ and 5′-GTTCCTAGCTAGAATTCCT-3′; exon 4 (620 bp), 5′-GATGTCTGGTCTGGAGCC TC-3′ and 5′-CCTTCTCAAAGTCCCATCTC-3′.
Fig. 2

Mapping and identification of the nonagouti (a) mutation. (A) Selected individuals from a German Shepherd pedigree segregating at and a were genotyped for DBar1, DBar2, and AHT125 as indicated. Linkage phase and presumptive Agouti genotypes were inferred as described in the text. A single recombinant between AHT125 and the other loci on CFA24 was observed in III-6. (B) Sequence chromatogram traces from individuals of the indicated Agouti genotypes; the C427T transition predicts an Arg-to-Cys substitution at codon 96 as described in the text. The at/a and a/a samples are from individuals in the pedigree shown in (A); the at/at sample is from a red-with-tan-points Australian Shepherd.

We also examined genomic DNA collected with cheek brushes from several other privately owned dogs; DNA isolation from buccal swabs was performed according to the QIAamp DNA swab extraction protocol (


Molecular isolation and characterization of the dog Agouti gene

To isolate a dog Agouti cDNA, we hypothesized that the black-and-tan phenotype characteristic of the Doberman Pinscher breed represented localized expression of Agouti protein in areas that give rise to yellow hair. We obtained fragments of tail tissue removed by a community veterinarian during routine docking of Doberman Pinscher puppy tails between 3 and 7 days of age and isolated total RNA from regions of the ventral tail skin where yellow hair was actively growing.

Degenerate oligonucleotide primers based on the protein-coding regions of the mouse, human, and bovine Agouti sequences were used to RT-PCR a protein-coding fragment of the dog cDNA; subsequent RACE steps were used to recover a 788-bp full-length cDNA that predicts a 131-codon open reading frame with a characteristic signature for mammalian Agouti protein (Fig. 1C). A 22-residue signal sequence is followed by a 63-residue highly basic amino-terminal domain (19 lysines or arginines, predicted pI of 10.7), 4 proline residues, and a cysteine-rich carboxy-terminal domain whose pattern of spacing is highly conserved with Agouti-related protein and, therefore, likely to fold into the same characteristic cysteine knot structure.

A large genomic fragment containing the entire dog Agouti gene was isolated from a Doberman Pinscher BAC library and sequenced. Comparison of the cDNA and genomic sequences reveals an exon–intron structure similar to that of other mammals, with an untranslated first exon, a relatively large first intron (∼36 kb), and protein-coding regions on exons 2, 3, and 4 (Fig. 1). The Doberman Pinscher Agouti sequence was most similar to that of the fox; 129 of 131 codons are identical, and exons 2, 3, and 4 of the dog cDNA are 97% identical to the corresponding region of the red fox (Vulpes vulpes) cDNA (Y09877), which includes the 3′ untranslated region (Fig. 1D). By contrast, the 130-bp Doberman Pinscher exon 1 sequence exhibited no similarity to the corresponding region of the fox cDNA (Fig. 1D). However, comparison of the fox cDNA sequence to the Doberman Pinscher BAC genomic sequence revealed two regions of dog genomic DNA that, overall, are 98% identical to the 5′ end of the fox cDNA and that lie approximately 1.3 kb and 15 kb upstream of the dog exon 1 sequence (Fig. 1A, D). Thus, exons 2, 3, and 4 for the Doberman Pinscher and fox Agouti cDNAs are homologous, but upstream exons of the two species are nonhomologous, and the fox cDNA may contain multiple untranslated exons. This observation is reminiscent of the mouse Agouti gene in which alternative untranslated first exons with different patterns of expression account for distinct coat color patterns, with deposition of pheomelanin that is either hair cycle-specific (the Agouti banded phenotype) or ventral-specific (the black-and-tan phenotype).

The human Agouti gene, also known as ASIP, maps to Chromosome 20q11.2. Despite extensive karyotypic rearrangement during recent evolution of the domestic dog, previous mapping studies (Breen et al. 1999) suggest that sequences from human Chromosome 20 are found on a single dog chromosome, CFA24. To confirm the location of the dog Agouti gene, we developed closely linked markers useful for linkage analysis. Sequencing the protein-coding region of the dog Agouti gene in 14 animals from the Cornell Labrador Retriever × Greyhound families did not identify any variants, but analysis of BAC69A22 identified 3 STRs (DBar1, DBar2, and DBar3), of which 2 (DBar1 and DBar2) were highly polymorphic (Fig. 1A). Using 6 families in which at least one parent was heterozygous for DBar1, we tested for linkage between DBar1 and AHT118, AHT125, and ZuBeCA14, 3 STRs previously mapped to CFA24 whose positions were predicted to lie close to genes from human 20q11 (Fig. 3). DBar1 was completely concordant with AHT118 and exhibited one recombinant with AHT125, leading to two-point LOD scores of 6.32 (at θ = 0) and 5.33 (at θ = 0.5), respectively (Fig. 3). The marker AHTH9BREN lies approximately 10 Mb distal to AHT118 and has been cytogenetically mapped to the distal region of CFA24 (band 24) whose total length is estimated to be 73 Mb; thus, the dog Agouti gene is probably located in the mid-distal region of CFA24 (Fig. 3).

Mapping and molecular analysis of a German Shepherd Dog Agouti mutation

As described above, a uniform eumelanic coat color in most breeds is thought to represent the action of “dominant black,” which is consistent with our previous studies of the Cornell Labrador Retriever × Greyhound families in which this phenotype was unlinked to the Agouti or Mc1r gene. In German Shepherd Dogs, however, a uniform black coat is most simply explained by the action of “recessive black,” where, for example, heterozygous black-and-tan parents give rise to black vs. black-and-tan progeny in a 1:3 ratio. To investigate whether variation at the Agouti locus was linked to recessive inheritance of a uniform black coat in German Shepherd Dogs, we made use of such a pedigree, in which a sibship of 6 black-and-tan and 3 black pups was produced by black-and-tan parents, who themselves were the offspring of a black × black-and-tan cross (Fig. 2A). Consistent with nomenclature in other mammals and the terminology originally suggested by Little (1957), we use at to refer to the presumptive black-and-tan allele and a for the nonagouti (recessive black) allele.

Genotypes for DBar1, DBar2, and AHT125 were determined, and information from the granddam genotypes were used to establish phase. In each of the 9 pups, there was complete cosegregation of the DBar21, and DBar12 alleles with the presumptive a allele, and complete cosegregation of the DBar22 and DBar11 alleles with the presumptive at-allele (Fig. 2). We also examined a second German Shepherd Dog pedigree (not shown) in which 5 black pups were produced from sable and black parents who were heterozygous and homozygous for DBar1, respectively. Each of the black pups carried the same allele from the sable parent, leading to a combined LOD score of 5.5 for linkage of Agouti variation to recessive inheritance of a black coat.

To determine if these results could be explained by an alteration in Agouti coding sequence, we sequenced Agouti exons and flanking regions from genomic DNA and identified a missense alteration in exon 4, C427T, predicted to cause an Arg-to-Cys substitution at codon 96 (R96C). In addition to animals from the pedigree depicted in Fig. 2 (in which the at/a parents were heterozygous and the a/a progeny homozygous for R96C), we examined genomic DNA from several herding breeds. We identified 3 black animals homozygous for R96C (a German Shepherd Dog, a Shetland Sheepdog, and a Groenendael), as well as an Australian Shepherd who was heterozygous (Table 1).
Table 1

Molecular genetic test results for a dog nonagouti mutation



Coat colora



German Shepherd Dog




German Shepherd Dog




German Shepherd Dog




German Shepherd Dog




Shetland Sheep Dog




Shetland Sheep Dog








Australian Shepherd

Blue merle



Australian Shepherd



Pup #3

Australian Shepherd

Blue merle


aExamples of the black-and-tan and black German Shepherd phenotypes are given in Fig. 2. Bi-black animals are black with white markings; tricolor animals have a characteristic distribution of black, yellow, and white hair; blue merle animals are modified by additional dilution and white spots such that the black areas appear bluish-gray.

bResults are given as amino acid predicted at codon 96, either Arg (CGC) or Cys (TGC).

cWicka, 7M, and 5M are the mother and two pups from the pedigree depicted in Fig. 3; other animals are unrelated.
Fig. 3

Genetic and physical map location of the dog Agouti gene. Selected markers from the CFA24 radiation hybrid map are shown together with the cytogenetic position of their human homologs; distances and positions are based on Breen et al. (1999). LOD scores for linkage of DBar1 to AHT118 and AHT125 are indicated on the left, as described in the text.


The diversity of coat colors and patterns represented among different dog breeds provides many opportunities for using the genetics of mammalian pigmentation to study basic questions about gene action and interaction. Ironically, the molecular genetics of coat color in dogs has lagged behind that of many other mammals, particularly with regard to our understanding of Agouti–melanocortin signaling. Our results demonstrate that molecular variation in Agouti underlies phenotypic variation in at least four herding breeds of the domestic dog and provide molecular tools to extend our understanding of canine coat color genetics.

Unlike the German Shepherd Dog, in which all black animals are thought to represent “recessive black,” both dominant and recessive black genes have been suggested to occur in the Australian Shepherd, the Shetland Sheepdog, the Belgian Shepherd (Groenendael), and the Border Collie; genomic DNA of animals from these breeds can now be tested directly for the presence of the R96C mutation, and families in which there are one or more black animals can be tested for linkage with the STR markers described above. These results will help to establish the extent of locus heterogeneity—eumelanin caused by dominant inheritance of a gene unlinked to Agouti vs. eumelanin caused by recessive inheritance of the a allele—within different breeds and, in the case of the latter, will determine if presumptive a alleles in different breeds have the same molecular cause. Although recessive black is easily recognized in German Shepherds, some dogs homozygous for the R96C allele will have black areas of the coat together with spotting mutations (e.g., bi-black Shetland Sheepdogs), while others may have eumelanin diluted or modified by additional mutations that act on melanogenesis or pigment granule biogenesis (e.g., blue- or liver-colored German Shepherds).

Studies of the R96C variant may also provide insight into breed phylogeny and history. One of the first descriptions of recessive black in canines was based on breeding experiments between wolves and domestic dogs carried out in the 1920s by Iljin (1941), who described 1:1 segregation of black vs. nonblack progeny by a black sheepdog and a “zonar grey” wolf, the latter presumably heterozygous for recessive black. Given the close relationship between wolves and the German Shepherd Dog, and the distribution of recessive black across different breeds, it is possible that a single ancestral a allele originally derived from Canis lupus explains all occurrences of recessive black in Canis familiaris. By analogy to the Mc1r locus, recessive inheritance of a uniform yellow coat has the same molecular cause, R306ter, in pale yellow Labrador Retrievers, Irish Setters, and Golden Retrievers (Everts et al. 2000; Newton et al. 2000), but it occurs in combination with several different Mc1r haplotypes, so recurrent mutation cannot be easily distinguished from gene conversion. However, like the R96C Agouti allele described here, the R306ter Mc1r allele responsible for recessive yellow is a C-to-T transition at a CpG dinucleotide, and therefore could represent a hotspot for recurrent mutation.

The R96C mutation is the only sequence variant we identified in DNA from black German Shepherd Dogs; however, we examined only exon and flanking intron sequence, so it is possible that additional variants specific for black German Shepherd Dogs are present in regulatory regions. Nonetheless, several considerations suggest the R96C mutation is the biochemical cause for recessive black. The R96 residue is conserved among all species from which Agouti has been identified, and the cysteine-rich carboxy-terminal domain (residues 92–131) is the portion of the molecule that interacts directly with the Mc1r (Kiefer et al. 1997; Ollmann et al. 1998). Based on studies of Agouti-related protein (Agrp), whose carboxy-terminal domain shares a conserved pattern of cysteine spacing with Agouti protein and whose three-dimensional structure has been determined (Ollmann et al. 1997; Bolin et al. 1999; McNulty et al. 2001), the carboxy-terminal domain of Agouti protein is likely to fold into an inhibitor cysteine knot (ICK) motif in which four of the five disulfide pairs provide a critical role in folding and/or stabilization. The residue corresponding to Agouti Arg96 in Agrp is His 90; however, the deleterious effects of R96C in Agouti are probably due to the presence of the Cys rather than the loss of the Arg. An “extra” cysteine in the R96C black German Shepherd Dog allele would probably prevent the ICK motif from folding correctly; indeed, previous studies by Perry et al. (1995) demonstrated that Agouti protein variants with a single “missing” cysteine in any of the four critical disulfide pairs exhibited little or no biological activity in vivo. In an extensive characterization of loss-of-function Agouti mutations in the laboratory mouse, Miltenberger et al. (2002) found that one “unpaired cysteine” allele, C115S, retained a very small amount of biological activity as manifested by small amounts of pheomelanin synthesis in the perimammary and perineal areas. Black German Shepherds, however, produce absolutely no pheomelanin, which suggests that an “extra” cysteine at position 96 is more disruptive than a “missing” cysteine at position 115.

In agreement with Little’s terminology (and conventional usage in the dog community), we refer to the R96C mutation in black German Shepherd Dogs as nonagouti (a). However, it should be noted that in laboratory mice, the allele of the same name is a regulatory defect with significant levels of residual activity, as manifested by pheomelanic hairs in the pinna, perimammary, and perineal regions (Bultman et al. 1994; Vrieling et al. 1994). In this respect, the nonagouti allele in German Shepherd Dogs is more analogous to the amorphic mutations extreme nonagouti or jet in the mouse (Hustad et al. 1995; Miltenberger et al. 2002). Amorphic Agouti mutations, referred to as nonagouti, have also been identified in the laboratory rat [19-bp deletion (Kuramoto et al. 2001)], the fox [intragenic deletion of exon 2 (Vage et al. 1997)], black horses [11-bp deletion (Rieder et al. 2001)], and black domestic cats [2-bp deletion (Eizirik et al. 2003)].

Additional molecular studies based on our results may also help to resolve the nomenclature for other dog Agouti alleles. The terms wolf-grey (ag), saddle-markings (as), tan-sable (ay), gray-sable (aw), and black-and-tan (at) have been used to designate presumptive Agouti locus genotypes responsible for different German Shepherd phenotypes (Carver 1984; Willis 1989), but a spectrum of intermediate phenotypes may exist, and the terminology does not extend well across different breeds. For example, the black-and-tan pattern characteristic of the Doberman Pinscher, the Rottweiler, and some Dachshunds is considerably different from black-and-tan in German Shepherd Dogs. The Doberman Pinscher pattern exhibits small well-demarcated areas of pheomelanic hairs in the perineum, distal limbs, chest, ventral neck, and supraorbital areas (“eye spots”); by contrast, black-and-tan in German Shepherd Dogs is characterized by a large contiguous area of pheomelanic hair that covers the ventrum, flank, and part of the head and fades gradually into adjacent areas of eumelanic hair covering the saddle, dorsal tail, periauricular region, and nose (Fig. 2A). Furthermore, the prototypic Agouti phenotype in most mammals—a subapical band of pheomelanin on a eumelanic background—is relatively rare among domestic dog breeds, whereas a uniform pheomelanic coat, e.g. fawn-colored Boxers, Basenjiis, or Dachshunds, is relatively common.

In laboratory mice, distinct ventral-specific and hair cycle-specific promoters give rise to different pigment patterns (Vrieling et al. 1994), but the existence of fawn- or tan-colored dogs that carry a functional Mc1r (Newton et al. 2000) suggests that canids may have an additional Agouti promoter expressed throughout the hair cycle over most of the body surface. Such a promoter could account for the observation that the 5′ untranslated region of the red fox Agouti cDNA lacks similarity to other mammalian Agouti cDNA sequences but appears to lie several kilobases upstream of the Doberman Pinscher first exon. Thus, the black-and-tan Doberman Pinscher could represent activity of a “ventral-specific promoter” (by analogy to laboratory mice), while the black-and-tan German Shepherd Dog could represent a combination of different Agouti promoters. Alternatively, Agouti mRNA may be expressed similarly in both types of animal, with the activity of the ventral-specific promoter in German Shepherd Dogs modified by a second locus. A precedent for the latter idea comes from studies of the mouse droopy ear mutation (Candille et al. 2004), which causes ventral areas of pheomelanic hair to extend across the dorsal flank, much like the German Shepherd Dog black-and-tan phenotype. These explanations are not mutually exclusive and can now be tested individually with molecular genetic and linkage studies.

Taken together with our previous studies of the Cornell Labrador Retriever × Greyhound families, there are now at least two loci in which sequence variation may cause exclusive production of eumelanic hairs in dogs, Agouti and Black (K). The latter is likely to account for many if not most instances previously described as As or “dominant black,” and, because it is not allelic with Mc1r or Agouti, it could represent variation in a gene not previously recognized as a mouse coat color mutation. Thus, further genetic and molecular characterization of the K gene may help to explain not only a spectrum of phenotypic variation in domestic dogs, but it may also provide insight into fundamental aspects of pigment-type switching.


We thank J. Faraco and E. Mignot for providing the canine BAC library, and J. Longmire for his support of JAK. We are grateful to the dog breeders who generously submitted DNA samples from their litters and to the DogMap and the FHCRC dog genome project for providing public access to the canine map and marker data at index.html and dog_genome/ .

Copyright information

© Springer-Verlag 2004