|Home | About | Journals | Submit | Contact Us | Français|
Melanins are a ubiquitous component of plumage colouration in birds and serve a wide variety of functions. Although the genetic control of melanism has been studied in chickens and other domestic species, little was known about the molecular genetics of melanin distribution in wild birds until recently. Studies have now revealed that a single locus, the melanocortin-1 receptor (MC1R) locus, is responsible for melanic polymorphisms in at least three unrelated species: the bananaquit, the snow goose and the arctic skua. Results show that melanism was a derived trait and allow other evolutionary inferences about the history of melanism to be made. The role of MC1R in plumage patterning is surprisingly diverse among different species. The conserved molecular basis for the evolution of melanism in birds and several other vertebrates is probably related to low pleiotropic effects at the MC1R.
The dazzling array of plumage colour variation in birds has long attracted the attention of ecologists and evolutionary biologists. Plumage colouration has figured prominently in many fields, including sexual selection (Darwin 1871; Andersson 1994), geographical differentiation and speciation (Mayr 1963), evolution of sexual dimorphism (Dunn et al. 2001) and evolution of polymorphisms (Roulin 2004). Much plumage colour variation both within and between species has a strong genetic component (Buckley 1987; Merilä & Sheldon 2001; Price 2002; Mundy 2005) but there has been surprisingly little progress until recently in defining the genetic changes involved. More generally, there is growing interest in the genetic basis of adaptation and phenotypic evolution, but still very few examples in which the chain of causation from genetic change through to adaptive change in phenotype is fully understood. The knowledge of vertebrate pigmentation genetics gained from studies of domesticated vertebrates, particularly mice and chickens, provides a sound basis for investigating the genetic basis of colour variation in the wild.
Melanins have a wide range of functions in birds, including physical protection (Barrowclough & Sibley 1980; Burtt 1986), protection from parasites (Goldstein et al. 2004), camouflage and a variety of signalling functions (Bókony et al. 2003; Jawor & Breitwisch 2003; Roulin 2004). Here I review recent studies on the molecular genetic basis of melanism in birds that provide a framework for future studies involving more complex phenotypic changes both within and between species. The results reveal strong conservation in evolutionary mechanisms of melanism in birds that extends to other groups of vertebrates.
The bananaquit (Coereba flaveola), lesser snow goose (Anser c. caerulescens) and arctic skua (Stercorarius parasiticus) are distantly related, have widely differing ecologies and show different roles of melanism. Bananaquits are nectarivorous passerines with a wide range in the neotropics. Melanic morph bananaquits are completely black and occur on a few Caribbean islands, particularly Grenada and St. Vincent. On these islands, melanic morphs are restricted to moist forest at low and high altitudes whereas yellow morph birds occupy disturbed dry lowland habitat (Wunderle 1981a,b). Melanic morphs show a stronger shade preference than yellow morphs (Wunderle 1981a). In the best-studied region in south-western Grenada, there is a hybrid zone between melanic and yellow morphs that has not moved in 20 years, suggesting the presence of local adaptation, although the adaptive advantage is poorly understood (MacColl & Stevenson 2003). There is no evidence for assortative mating by morph type (Wunderle 1981a).
Arctic skuas have a holarctic distribution, and a north–south cline in morph frequency with melanic morph birds commoner in the southern part of the range. They were the subject of a long-term study on Foula, Shetland Islands, UK, which found that melanic morph males breed earlier in the season than paler morph males in their first breeding year but that paler morph birds start breeding at an earlier age (O'Donald 1983). Patterns of assortative mating vary somewhat from year to year and are also variable at other colonies, with both positive assortative and disassortative mating being found (Phillips & Furness 1998).
Lesser snow geese show an approximate east–west cline in morph frequency in their breeding distribution across the nearctic into the eastern tip of Russia, with blue morph birds commonest in the east. Extensive studies of fitness components in a colony at La Pérouse bay, Canada, failed to find any adaptive advantages related to plumage morphs (Cooke et al. 1995). However, geese show strong mating preferences based on parental colour, which leads to positive assortative mating (Cooke et al. 1976).
The melanocortin-1 receptor gene (MC1R) was first isolated in laboratory mice, in which it was found to be the gene encoding the classical extension locus that affects coat colour (Mountjoy et al. 1992). The gene encodes the MC1R protein that is a seven-transmembrane domain G-protein coupled receptor expressed primarily in melanocytes. Activation of MC1R leads to increased synthesis of black or brown eumelanin, whereas low MC1R activity generally leads to increased synthesis of red or yellow phaeomelanin (Robbins et al. 1993). In birds, MC1R was first cloned from chickens (Takeuchi et al. 1996) and a point substitution in the gene was subsequently found to be associated with melanism (Takeuchi et al. 1998).
Remarkably, MC1R variation is associated with melanism in bananaquits, snow geese and arctic skuas (Theron et al. 2001; Mundy et al. 2004). A single non-synonymous change is perfectly associated with the presence of melanism in all three species, but different substitutions are involved (figure 1). Three pieces of evidence indicate that the MC1R genotype–phenotype association is specific to the MC1R locus in these three species and is not simply due to the shared demographic history of populations containing the different morphs. First, the association is present in populations segregating for the different morphs. Second, no association is found between the phenotype and neutral markers (mitochondrial DNA; Seutin et al. 1994; Mundy et al. 2004). Third, considering sympatric melanic and non-melanic morphs as separate populations, population differentiation as measured by Fst is far greater at MC1R than neutral markers (mtDNA, microsatellites; Mundy et al. 2004).
In bananaquits, there are two discrete morphs. In snow geese and skuas, there is a discrete difference between non-melanic (white geese, pale skuas) and melanic morphs in the degree of melanization but substantial variation in degree of melanization among melanic morphs (blue geese, intermediate to dark skuas; figure 1). The degree of melanization among the melanic morphs is strongly correlated with the number of variant MC1R alleles (figure 1; Mundy et al. 2004). MC1R can therefore be considered to be the major quantitative trait locus (QTL) for melanism in geese and skuas. The other genetic and environmental factors involved are unknown. Patterns of inheritance of variant MC1R alleles are in accordance with field observations of transmission genetics in the three species (O'Donald & Davis 1959; Wunderle 1981a; Rattray & Cooke 1984).
Two other studies have investigated a relationship between MC1R and avian melanism. An island subspecies of the white-winged fairy-wren (Malurus leucopterus) in which males in nuptial plumage are melanic shows strong differentiation at MC1R from a mainland subspecies in which males are blue (Doucet et al. 2004). An interesting aspect of this example is that there is sexual dimorphism in melanism, and females are not melanic. Ultrastructural studies indicate that the melanic subspecies arose from an ancestor with blue males. However, comparison with an out-group shows that, unexpectedly, many MC1R substitutions have occurred in the lineage leading to the blue subspecies (Doucet et al. 2004), so it will be important in this case to demonstrate that the differences at MC1R do not merely reflect population divergence.
Species of Old World leaf-warblers (Phylloscopus) differ in patterns of small unmelanized patches on the head, wings and rump that are known to have important signalling functions. MC1R variation across species does not correlate with any of these traits (MacDougall-Shackleton et al. 2003), but it would be surprising if MC1R coding differences could have such a fine-scaled patterning effect.
Non-synonymous substitutions have been identified that are strongly associated with melanism, but do they cause melanism? Comparison of substitutions with the known structure–function relationships of the MC1R protein suggest that the substitutions in bananaquits, snow geese and arctic skuas could increase MC1R activity and hence eumelanin deposition (figure 2). The best evidence comes from bananaquits, in which the Glu92Lys substitution, which also occurs in domestic mice and chickens, locks the MC1R in an active state (constitutive activation; Robbins et al. 1993; Ling et al. 2003). In snow geese, the Val85Met substitution occurs in the second transmembrane domain where many mutations affect MC1R activity (Majerus & Mundy 2003), while the Arg233His substitution in arctic skuas occurs at the same site as one of the substitutions in melanic populations of the rock pocket mouse (Chaetodipus intermedius; Nachman et al. 2003; figure 2).
A more direct approach is to determine biochemical function of different MC1R alleles. MC1R can be expressed in vitro and assayed for ligand binding and cAMP activation, and this has been widely performed for MC1R variants in domestic species and humans (Robbins et al. 1993; Schiöth et al. 1999). Such studies on wild MC1R variants are currently underway.
Although all cases of MC1R association described so far may be due to changes in MC1R coding sequence, it is likely that substitutions in MC1R regulatory regions are also involved in plumage colour evolution. Some of the important MC1R cis-regulatory sequences and transcription factors that interact with the MC1R promoter have been identified in mice and humans (Moro et al. 1999). Detection and identification of substitutions affecting MC1R regulation is challenging, but developing in vitro assays to assess the effects of specific substitutions on MC1R expression is tractable.
Networks reconstructing evolutionary relationships among MC1R haplotypes show that alleles containing substitutions associated with melanism are present in derived positions in bananaquits, snow geese and arctic skuas (Theron et al. 2001; Mundy et al. 2004). Thus, there is genetic evidence that melanism was a derived trait in these species, as had been predicted from other information. It is important to note that this inference does not depend on the sites involved in the association being the functional sites, since if they are non-functional they must be in tight linkage disequilibrium to functional sites.
In the case of bananaquits, the haplotype network provides strong evidence for two waves of migration to Grenada and St. Vincent (Theron et al. 2001). An older colonization event involved populations that evolved the melanic MC1R alleles, whereas a younger migration brought yellow MC1R alleles that are only distantly related to the melanic alleles. These inferences are supported by historical data that suggests that yellow birds arrived on Grenada and St. Vincent around 100 years ago (Wunderle 1981a,b).
Although the inference of the evolution of melanism using MC1R is straightforward in these examples, it does provide an example of a broader approach that is an alternative method to the reconstruction of phenotypic evolution. Inferences of trait evolution over a phylogeny are generally performed using the phenotypes of extant species (e.g. Christidis et al. 1988; Omland & Lanyon 2000). The new method involves first reconstructing the genotype and then using a known phenotype–genotype association to infer phenotypic evolution. This method has the advantage that models of DNA sequence evolution are far more sophisticated than models of phenotypic evolution and so reconstruction of the genotype is relatively robust. It is obviously dependent, however, on accurate prediction of phenotype from genotype.
Another class of historical inferences involves patterns of natural selection on MC1R, which provides an important route to documenting adaptive evolution at MC1R in relation to melanism or other phenotypic changes. So far, the MC1R has been found to be under positive selection between the bananaquit and tanager lineages (Theron et al. 2001), although the phenotypic change involved in this case is unclear. Further intraspecific and interspecific tests for selection will be important in defining the adaptive role of MC1R in plumage evolution.
A surprising aspect of the involvement of MC1R in melanic polymorphisms in a range of birds is the different effects of the locus on patterning (table 1). Melanic morph bananaquits have an extreme melanic phenotype, with melanin deposited throughout all feathers. Variation among arctic skuas morphs involves gradual darkening of the neck, breast and belly. Snow geese show the most interesting variation. In increasingly melanic morphs, there is an increase in the number of feathers that are melanized, with gradual spread over the back, then the neck, then the breast and finally the belly. There is relatively little variation in the amount of melanin deposited in each feather, except at the boundary between melanized and non-melanized regions. This patterning is strongly suggestive of a gradient of a factor that interacts with the melanogenic pathway containing MC1R. Work on the developmental biology of feather tracts has shown that variation in patterning along tracts is primarily determined by the epidermis rather than by melanocytes (Rawles 1959; Richardson et al. 1991). Hence, one possibility is that there is a gradient of MC1R agonist secreted by the epidermis.
Melanism is derived within snow geese and the common ancestor of snow and the closely related Ross' geese was presumably largely white. Why would such a complex pattern of melanism develop? The answer appears to lie further back in evolutionary history. The closest relative of snow and Ross' geese is the emperor goose (Anser canagicus). This species has melanin-containing feathers on most of the body apart from the upper neck and head that is in a strikingly similar pattern to that found in blue morph snow geese (there is additional fine barring of the feathers in emperor geese that is absent in snow geese). It therefore appears that regional melanic patterning across the body was already present in the common ancestor of emperor and snow geese and became lost in the snow goose lineage, due to a loss-of-function mutation at MC1R or another locus. Its reappearance in geese with the variant MC1R is thus an atavism.
The species of birds in which MC1R is associated with melanism are distantly related to each other and, in particular, the Anseriforms (including geese) are generally thought to form a basal out-group (together with the Galliforms) to the other groups in the Neoaves, which include Charadriiforms (skuas) and passerines (bananaquits and fairy-wrens; Groth & Barrowclough 1999; Tuinen et al. 2000; Johansson et al. 2001; Poe & Chubb 2004). The repeated involvement of MC1R in avian melanism therefore represents a striking example of conservation of the genetic basis of evolution. In fact, this conservation extends to mammals and probably also to reptiles. MC1R variation is associated with melanism in a population of rock pocket mice (Chaetodipus intermedius; Nachman et al. 2003; Hoekstra et al. 2004), in two species of cat (the jaguar, Panthera onca, and the jaguarundi, Herpailurus yaguarondi; Eizirik et al. 2003) and in the little striped whiptail lizard (Aspidoscelis inornata; Rosenblum et al. 2004). MC1R variation is also associated with pale coat colour in a population of black bears (Ursus arctos; Ritland et al. 2001) and with red hairs and pale skin in some human populations (Valverde et al. 1995).
There is clearly some ascertainment bias in the examples presented above, for example, bananaquits were chosen as a good candidate for an MC1R effect. Also, although there are no counterexamples yet in birds, there are several examples of mammals with large changes in melanin distribution in which an MC1R coding sequence variation is not involved, for example, some populations of rock pocket mouse (Hoekstra & Nachman 2003) and several primate species (Mundy & Kelly 2003). Nevertheless, there is enough data to show that MC1R, and probably MC1R coding variation, has been repeatedly involved in the evolution of plumage (and coat) colour. There are several potential reasons for this, which are not mutually exclusive:
Theoretically the position of MC1R on the pathway to melanogenesis might give it such a special role that the same phenotypic effect could not be achieved by substitutions in other loci. A good example of such a critical role in another system is wavelength sensitivity in vertebrate vision, where changes in the opsin components of photopigments are the major way of achieving a change in phenotype (Yokoyama & Yokoyama 1996). However, this is not the case for MC1R—for example, substitutions at many loci can cause melanism in mice (Jackson 1997).
The issue of pleiotropy at MC1R is important for the potential of MC1R-mediated effects to be involved in evolutionary trade-offs, since melanin production is probably more costly than lack of production. The possibility that melanized secondary sexual traits are used as honest indicators of quality is currently receiving more attention (Jawor & Breitwisch 2003) but see Badyaev & Hill (2000). On the other hand, if there are low negative pleiotropic effects of MC1R variants, this would increase the frequency of their occurrence in plumage evolution. Correlated phenotypic changes of MC1R-associated effects on pigmentation have not been described in wild populations, but potential pleiotropic effects of MC1R have been explored in laboratory studies of MC1R function.
In mammals, there have been several studies investigating a possible role of MC1R in the immune system. MC1R is expressed in many leucocytes (Catania et al. 1996; Neumann et al. 2001) and the melanocortin peptide α-MSH, which is an MC1R agonist, has well-established anti-inflammatory effects (Star et al. 1995). However, no evidence of immune dysfunction has been reported in mice or humans which lack functional MC1R. Furthermore, two recent studies have found that anti-inflammatory effects in mice are due to MC3R and not MC1R (Ichiyama et al. 1999; Getting et al. 2003). Overall, therefore, despite some speculation to the contrary, the evidence of a function of MC1R in immunity in mammals is poor. There is little information on the relationship between MC1R and immunity in birds and it would be interesting to investigate immune function in chickens with MC1R variants.
The only strong evidence for a pleiotropic action of MC1R is in the nervous system. In mice and humans, MC1R mediates a female-specific pathway of analgesia (Mogil et al. 2003). Although this effect is pharmacologically well-defined, its functional role is unclear.
The real issue is whether MC1R changes have lower pleiotropic effects than other genes affecting pigmentation. Over 40 loci controlling plumage colour have been identified in chickens (Smyth 1990). The total is probably much greater, however, since more than 120 loci are involved in coat colour control in mice (Bennett & Lamoreux 2003) and the higher complexity of within feather patterning compared with within hair patterning suggests that more loci are involved in birds than mammals. The loci affecting coat colour can be grouped into four major categories: melanocyte development and differentiation, melanosome development, melanin synthesis and regulation of melanin synthesis (Jackson 1997). It is notable that most loci affecting the first two processes have strong deleterious pleiotropic effects since common developmental pathways are shared among different systems: mutations affecting melanocyte development usually affect other neural crest-derived cells and mutations affecting melanosomes also affect lysosomes.
In summary, therefore, there is little evidence for negative pleiotropic effects of MC1R. It may be that MC1R is one of a relatively small number of genes that could be used to change pigmentation without having other deleterious phenotypic consequences.
It is possible that the overall mutation rate at the MC1R locus is high. This would lead both to a relatively high rate of loss-of-function variants that would be mostly subject to purifying selection, and a high rate of beneficial variants. The synonymous substitution rate of MC1R gives an appropriate measure of the mutation rate of the locus, and comparison of this rate among avian and mammalian MC1R genes and other nuclear genes reveals no evidence that this rate is elevated (N. Mundy, unpublished results).
In bananaquits, snow geese and skuas, melanism is derived and the novel melanic alleles would have had an immediate phenotypic effect and been available for selection. As the adaptive value of melanism in these species is uncertain, it is difficult to gauge the importance of this effect in these examples. The situation is far clearer in rock pocket mice, in which coat colour is related to substrate colour (Hoekstra & Nachman 2003) and almost certainly provides camouflage against predators. It is worth noting that derived recessive MC1R alleles have arisen to appreciable frequency at least twice in evolution, in humans in Europe (Valverde et al. 1995; Harding et al. 2000) and in a population of black bears on Queen Charlotte Island, Canada (Ritland et al. 2001). As both of these cases plausibly involved isolated populations, it is possible that inbreeding played a role.
In conclusion, low pleiotropic effects can be identified as a major reason why MC1R is an evolutionary favourite, with dominance also a likely contributor. It will be interesting to compare the spectrum of MC1R effects on the phenotype with those at other pigmentation loci when more information on the evolutionary role of these becomes available. These results also raise the issue of how the findings on the evolutionary genetics of pigmentation will compare with other phenotypic changes. Are particular loci repeatedly involved in, say, changes in body size or are genetic effects more dispersed? Variation in the underlying genetic architecture of different traits could strongly influence their evolution.
The MC1R examples discussed here demonstrate the utility of a candidate gene approach to adaptive phenotypic evolution (see also Fitzpatrick et al. 2005; Nachman 2005), which can be contrasted with methods based on genomic scans using neutral markers. Association studies using MC1R and other pigmentation loci (e.g. the recently documented PMEL17 in chickens (Kerje et al. 2004)) may be useful when studying both Mendelian and quantitative plumage traits, either on their own or in conjunction with neutral markers. The disadvantages of a candidate gene strategy are apparent when a negative result is obtained, that is, there is no association between genotype at the candidate locus and phenotype. The negative conclusions in such a case only relate to the part of the gene studied—for example, in the Phylloscopus study mentioned above (MacDougall-Shackleton et al. 2003), it is still possible that there is an association with the small portion of MC1R coding region that was not sequenced or with the MC1R promoter region. In general, the success of the approach depends on such factors as the availability of candidates, the practicability of assaying genetic variation in them and the number of loci potentially involved in evolution of a particular trait (see previous section). Careful selection of candidates according to the specificity of their phenotypic effects seems to be important and deleterious effects at loci identified in inbred captive lines may not always be expressed in captivity.
The examples discussed here clearly only represent the tip of the iceberg as far as avian plumage colouration genetics in the wild is concerned. Over 300 avian species exhibit plumage polymorphisms (Galeotti et al. 2003). Although fitness effects of the melanic polymorphisms studied here are incompletely understood, in general these polymorphisms are expected to have important effects on fitness, as found, for example, in buzzard (Buteo buteo; Lank et al. 1995; Krüger et al. 2001), ruff (Philomachus pugnax; Lank et al. 1995) and feral pigeons (Columba livia; Murton et al. 1974). None of the examples discussed above involve conspicuous changes in phaeomelanin distribution and there is very little information on molecular genetic mechanisms affecting carotenoid distribution, structural colouration, sexually dimorphic traits and within feather patterning (Prum & Williamson 2002).
I thank Trevor Price and an anonymous reviewer for helpful comments on the manuscript and the Leverhulme Trust, BBSRC and NERC for financial support.