Pigeonetics takes flight: evolution, development, and genetics of intraspecific variation

Abstract

Intensive artificial selection over thousands of years has produced hundreds of varieties of domestic pigeon. As Charles Darwin observed, the morphological differences among breeds can rise to the magnitude of variation typically observed among different species. Nevertheless, different pigeon varieties are interfertile, thereby enabling forward genetic and genomic approaches to identify genes that underlie derived traits. Building on classical genetic studies of pigeon variation, recent molecular investigations find a spectrum of coding and regulatory alleles controlling derived traits, including plumage color, feather growth polarity, and limb identity. Developmental and genetic analyses of pigeons are revealing the molecular basis of variation in a classic example of extreme intraspecific diversity, and have the potential to nominate genes that control variation among other birds and vertebrates in general.

Introduction

“In the late summer or early autumn of 1859, Whitwell Elwin, editor of the respected British journal the Quarterly Review, was sent an advance copy of a new book by the naturalist Charles Darwin. Elwin… agreed that it had merit, but feared that the subject matter was too narrow to attract a wide audience. He urged Darwin to write a book about pigeons instead. ‘Everyone is interested in pigeons,’ he observed helpfully.” (Bryson, 2003)

Fortunately, Darwin did not follow Elwin’s advice, and On the Origin of Species by Means of Natural Selection remains one of the most important volumes in the history of human thought. Yet Darwin was indeed a pigeon aficionado and immersed himself in the pigeon societies of London. His observations of the myriad breeds, as well as associations with “eminent fanciers” (Darwin, 1859, p. 20), had a profound influence on how he chose to communicate his ideas about how selection works (Baptista et al., 2009; Montgomerie, 2009; Secord, 1981; Theunissen, 2012). “Believing that it is always best to study some special group, I have, after deliberation, taken up domestic pigeons… The diversity of the breeds is something astonishing,” he wrote in the Origin (Darwin, 1859). In Variation of Animals and Plants under Domestication, he dedicated two chapters to this species, and singled out pigeons in his Introduction to the two-volume set: “I have selected this case, because, as we shall hereafter see, the materials are better than in any other; and one case fully described will in fact illustrate all others” (Darwin, 1868, p. 1). Darwin noted that based on morphology alone, a taxonomist might classify unique pigeon breeds as completely different genera (see Fig. 1 for representative breeds, many of which were familiar to Darwin), yet he also observed that all breeds, no matter how morphologically or behaviorally divergent, could be interbred to produce viable offspring (Darwin, 1859; Darwin, 1868). Therefore, he concluded, different breeds did not represent different species, but were instead diverse descendants of Columba livia, the wild rock pigeon.

Fig. 1.

Variation among domestic pigeon breeds. A, frillback; B, Old German owl; C, African owl; D, Scandaroon; E, English trumpeter; F, Indian fantail; G, archangel; H, fairy swallow; I, Old Dutch Capuchin; J, figurita; K, English pouter. Images are not to scale. Photo credit: Sydney Stringham.

The rock pigeon was likely domesticated at least 5000 years ago in the Middle East and/or Mediterranean region, making it one of the oldest avian domesticates, and also the most phenotypically diverse (Hansell, 1998; Levi, 1986; Price, 2002). Their feral descendants are often disdained in urban environments (Jerolmack, 2008) by the very humans that create ideal cliff-like habitats for them, provide them with plentiful food year-round, and are largely responsible for their nearly worldwide distribution. Yet pigeons have also assumed the roles of sacred religious symbols and war heroes (Hansell, 1998), and they are important model organisms for research in parasitology, behavior, physiology, psychology, neuroscience, microbiology, and other disciplines (Abs, 1983; Blaisdell and Cook, 2005; Fagot and Cook, 2006; Harbison et al., 2009; Klopfleisch et al., 2006; Lauwers et al., 2013; Mehlhorn and Rehkamper, 2009; Ros et al., 2011; Wagner et al., 2007). Here, we review the motivation and recent progress towards understanding the genetic and developmental mechanisms controlling pigeon diversity, including some of the same traits that struck “Darwin’s fancy” (Montgomerie, 2009).

Pigeon genetics: then and now

Shortly after the re-discovery of Mendel’s work, scientists began investigating the genetic basis of trait variation in a large variety of organisms, of which today’s canonical model organisms (e.g., mouse, chicken, fruit fly) represent only a subset. Indeed, the genetic basis of trait variation in pigeon was studied by a number of researchers including W. Christie and C. Wriedt (Norway), L. Cole (University of Wisconsin), W.M. Levi (Palmetto Pigeon Plant, South Carolina), W.F. Hollander (Iowa State University), and even the renowned Drosophila geneticist T.H. Morgan (Columbia University) (Christie and Wriedt, 1927; Cole, 1912; Hollander, 1937, 1983; Levi, 1965, 1986; Morgan, 1911). These studies provided case examples of such genetic phenomena as dominance, sex-linkage, sex differences in recombination frequency, and genetic mosaicism (Cole, 1912; Cole and Kelley, 1919; Hollander and Cole, 1940). Among the hobbyist community, an interest in pigeon genetics has continued to the present day, and results of experimental crosses are reported on citizen scientist websites and in newsletters. In academic circles, however, pigeon genetics grew out of favor in the mid-20^th Century, perhaps in part due to the development of molecular genetic tools for a limited set of model genetic organisms. A notable exception to this trend was W.F. Hollander, who continued to study pigeon genetics professionally and as a hobbyist until his passing in 2004 (Hollander, 1983; Hollander and Mangile, 1994). Sell (1994, 2012) provides outstanding and current summaries of most known genetic variants.

In more recent years, technological advances and precipitous price drops in next-generation sequencing have allowed biologists to revisit many of the early, classical pigeon genetic experiments, this time employing modern genomic, genetic, and developmental biology techniques. The rock pigeon now has a high-quality reference genome, additional re-sequenced genomes representing a wide variety of breeds (NCBI Bioproject PRJNA284526), a catalog of over 25 million SNPs and genomic structural variants across breeds, and a draft genome-wide linkage map (Domyan et al., 2014; Domyan et al., 2016; Shapiro et al., 2013). These extensive resources, combined with standard experimental techniques developed for canonical model systems, make the pigeon a viable model to understand the genetic basis of phenotypic variation, and to test the developmental consequences of these genetic variants in developing embryos.

Why study molecular genetics and development in pigeons?

Pigeons offer several advantages as a model to understand the molecular basis of avian and vertebrate diversity. First, pigeon breeders have selected for dramatic variation over the past several thousand years in numerous plumage, anatomical, and behavioral traits that are represented in over 350 breeds (Baptista et al., 2009; Darwin, 1868; Levi, 1965; Price, 2002; Shapiro and Domyan, 2013). As a result, the extent of variation between breeds can approach the number and extent of differences typically observed between different species (Darwin, 1859; Price, 2002).

Second, pigeons are easy to breed, and virtually all varieties can be successfully interbred to produce fertile offspring. This is a critical strength because it permits traditional genetic mapping experiments using birds with highly divergent phenotypes. In contrast, hybrid incompatibility typically precludes genetic mapping studies between wild species. Levi (1965, 1986) and many others summarize reproduction and husbandry techniques, with key aspects mentioned here. Healthy females can potentially lay a clutch of two eggs every 10 days, if the eggs are removed after laying. Pigeon hobbyists commonly remove eggs from their parents’ nest and place them with foster parents, thereby allowing the first set of parents to produce additional offspring. Gestation is shorter in pigeon (17–19 days at 37.5°C) than in chicken (21 days at 37.5°C), and pigeon hatchlings are altricial. Both male and female parents share nesting duties and produce a substance known as “crop milk” to feed to newly-hatched squabs. Crop milk meals gradually transition to regurgitated seeds until the young can feed on their own. Typically, juvenile pigeons are weaned after 3–4 weeks, and reach sexual maturity at approximately 6 months of age. The small body size of pigeons also allows large crosses to be kept in a manageable space.

Third, classical studies show that the genetic architecture of many traits in pigeons is probably oligogenic (Hollander, 1937, 1983; Levi, 1986; Morgan, 1911; Sell, 1994, 2012), thereby increasing the chances of mapping and cloning the genes responsible for differences among breeds. Furthermore, relatively complex traits (feather color, for instance) can sometimes be genetically teased apart into discrete components (e.g., by treating the type and intensity of melanin as separate traits), thereby facilitating mapping of each component (Domyan et al., 2014).

Fourth, many variable traits in pigeons are also variable among birds in general (Baptista et al., 1997; Darwin, 1868; Price, 2002), providing a critical entry point to finding the genes that control variation in the wild. For example, Baptista et al. (2009; 1997) noted the striking similarities between many breeds of domestic pigeon and wild columbids, and proposed that the same genes might control derived phenotypes in naturally and artificially selected birds.

Fifth, many derived traits in domestic pigeons are constructive rather than regressive; that is, they represent phenotypes that are absent in the wild ancestor (Stringham et al., 2012). Although regressive traits have attracted considerable attention in vertebrate evolutionary genetics, the genetics of constructive traits are less understood (Miller et al., 2014). Pigeons present a prime opportunity to address this gap.

Finally, phenotypic convergence is rampant among breeds, with similar traits under selection in breeds that are not necessarily genetically similar (National_Pigeon_Association, 2010; Stringham et al., 2012). This situation offers a rich set of animal resources to test whether the same or different genes – and the same or different mutations in those genes – underlie this phenotypic similarity. Understanding how these similarities arise is an area of intense interest in evolutionary developmental biology and genetics (Christin et al., 2010; Hovav et al., 2008; Martin and Orgogozo, 2013; Paterson et al., 1995; Rosenblum et al., 2014; Stern, 2013; Vickrey et al., 2015).

In short, pigeons are amenable to a combination of classical breeding strategies and genomics to dissect different components of complex phenotypes and define functional interactions among genes. Notably, equivalent insights about the genetic architecture of phenotypic divergence between wild vertebrate species are often very difficult to acquire.

Practical pigeon breeding – advantages for genetic association studies

As with show dogs, fancy mice, orchids, and a multitude of other domestic species, fancy pigeons are typically bred to compete at shows. Pigeon breeds have strict breed standards, but no pedigree requirements (e.g., parents and grandparents need not be registered members of a breed) (National_Pigeon_Association, 2010). Due to these strict standards, different individuals within a breed are – not surprisingly – genetically very similar (Shapiro et al., 2013; Stringham et al., 2012). However, breeds that are phenotypically highly divergent also show genetic signatures of shared ancestry, which is also expected given the recent divergence time among breeds (Shapiro et al., 2013; Stringham et al., 2012). Some of this genetic similarity also likely results from introgression of traits from one breed to another.

Introgression and hybridization among breeds have two fortunate consequences for the molecular geneticist. First, introgression raises the possibility that the same mutations were repeatedly passed from breed to breed. If this is the case, then we should expect to detect the same haplotype blocks in multiple breeds with the same trait. Second, hybridization of two (or more) breeds followed by backcrossing to remove undesirable donor traits from recipient breeds should break long haplotype blocks in the genome. Therefore, instead of having near-isogenic inbred lines, pigeons with similar traits might share identical – but small – haplotypes in different genetic backgrounds. Both of these factors contribute to the power of genome-wide association mapping across breeds. Below, we highlight recent progress using these strategies to identify the genes and molecular mechanisms that underlie morphological diversity among breeds.

The molecular basis of phenotypic variation in pigeons

Head crests

Price (2002) likened the process of artificial selection on birds under domestication (breeder choice) to sexual selection in the wild (typically female choice). The types of traits selected in both cases – color, head crests, and other ornamental, structural, and behavioral variants – also show remarkable parallels. Head crests of various kinds are found in numerous wild and domesticated avian species and can serve important functions in mate choice (Amundsen, 2000; Price, 2008; Price, 2002). By analyzing the genomes of a variety of pigeon breeds, we identified a single-nucleotide coding mutation in the highly conserved intracellular kinase domain of EphB2 that is perfectly associated with crests of reversed feathers on the head (Shapiro et al., 2013; Fig. 2). All crested pigeons appear to share the same mutant haplotype as a result of introgression among breeds. EphB2 is involved in numerous aspects of vertebrate development, including feather development, palate morphogenesis, melanocyte migration, tissue polarity, and axon pathfinding (Henkemeyer et al., 1996; Kelsh et al., 2009; Risley et al., 2009; Suksaweang et al., 2012; Xu et al., 2000). However, because only the kinase domain is affected in crested pigeons (Shapiro et al., 2013), forward signaling should be disrupted, but reverse signaling (through ephrin ligands) should be intact (Mendes et al., 2006). The EphB2 mutation is found in pigeons with different crest phenotypes – both the area comprising the crest and the length of its feathers can vary – so other modifier loci likely contribute to the extent of crest formation. Developmentally, head crests in domestic pigeons originate as a reversal of feather placodes on the head and neck (Shapiro et al., 2013). Head crest development in domestic chickens, in contrast, comprises elongation of head feathers but not a change in the direction of growth, and is not associated with EPHB2, but instead with ectopic expression of HOXC8 (Wang et al., 2012).

Fig. 2.

Head crests in domestic pigeons and ringneck doves. A. The simple “peak” crests of both species are associated with different amino acid substitutions in the kinase domain of EphB2. B. Pigeons also have elaborate crests, such as the mane of the Old Dutch Capuchin, that require contributions from modifier loci. The molecular identities of these additional loci are unknown. C. Occipital region of an English trumpeter embryo shows reversed growth trajectory of crest feathers (arrowheads). Feathers have a rostral instead of caudal orientation. The top of the head is at the upper right of the frame, beak is to the lower right (out of frame). e, eye. Photo credits for A, B: Sydney Stringham.

Baptista (2009) predicted that derived traits in pigeons and other columbids might have a similar genetic basis. In another domesticated columbid species, the crested ringneck dove (Streptopelia risoria), this appears to be the case (Fig. 2). As in domestic pigeons, crested ringneck doves have an amino acid substitution at a highly conserved residue in the catalytic site of the EPHB2 kinase domain (Vickrey et al., 2015). Both mutations produce kinase-dead versions of EPHB2, yet the amino acid substitutions found in pigeon and ringneck dove are separated by more than 100 residues and are predicted to have different effects on protein function. The pigeon mutation is predicted to destabilize the active site by breaking a critical salt bridge, while the dove mutation likely hinders entrance of ATP into the catalytic domain (Carrera et al., 1993; Hemmer et al., 1997; Vickrey et al., 2015). Thus, different mutations in the same gene result in similar phenotypes in two columbid species that diverged over 20 million years ago (Goodwin, 1983; Pereira et al., 2007).

Plumage color

Whether for aesthetic or practical reasons, or a combination of the two, the genetic basis of pigmentation has been one of the most-intensively investigated fields in pigeons and other vertebrates (Domyan et al., 2014; Hubbard et al., 2010; Manceau et al., 2010; Rosenblum et al., 2014; Sulem et al., 2007). Pigeons display a tremendous amount of color variation, and rely on essentially the same pigmentation system as mammals; by varying the absolute and relative amounts of black eumelanin and red pheomelanin, a wide variety of colors can be generated (Haase et al., 1992). In addition to being a visually pleasing phenotype to breeders, color variation also has relevance in free-living populations of pigeons, as color variants are correlated with differences in parasite load, juvenile survivorship, and predator avoidance (Jacquin et al., 2011; Jacquin et al., 2012; Jacquin et al., 2013; Palleroni et al., 2005; Recapet et al., 2013).

For more than a century, transmission genetic experiments have provided clues about the number of color genes and alleles, their dominance and epistatic relationships, and their cumulative effects on melanin phenotype (Hollander, 1937, 1983; Levi, 1986; Sell, 1994, 2012). For example, classical studies show that the basic or “ground” color locus B is sex-linked and has three alleles: a wild-type allele (B⁺) that causes primarily black eumelanin to be deposited on feathers, a dominant ash-red allele (B^A) that results in red pheomelanin, and a recessive brown allele (b) that primarily produces eumelanin (Fig. 3; Cole, 1912; Cole and Kelley, 1919; Haase et al., 1992; Hawkins, 1931; Steele, 1931). Using a combination of genomics and breeding experiments, tyrosinase-related protein 1 (Tyrp1), which encodes a melanin biosynthesis enzyme, was identified as the B locus (Domyan et al., 2014). All ash-red birds tested to date harbor the same Tyrp1 haplotype, suggesting that this trait was transferred to different breeds via hybridization and introgression. The B^A allele has a single missense mutation in Tyrp1 that is predicted to cause an alanine-to-proline substitution in the TYRP1 signal peptide, which greatly diminishes cleavage of the mature protein (Domyan et al., 2014). While the cellular consequences of this processing defect are unknown, a similar cleavage site mutation in the corn gene fl2 causes the encoded protein to be retained in the ER (Coleman et al., 1995; Gillikin et al., 1997), suggesting a similar fate for the ash-red version of TYRP1. How misprocessing of TYRP1 causes such a dramatic shift in pigment type, even in the presence of the wild-type allele, is an area of active investigation. Given the dominant inheritance pattern of the ash-red phenotype, either dominant-negative or neomorphic functions are likely.

Fig. 3.

Interactions among 3 loci produce 8 different melanin phenotypes. Allelic variation at Tyrp1 determines blue (wild-type), ash-red, or brown base color. The d allele of Slc45a2 reduces the intensity of the base color. Recessive red pheomelanism results from regulatory mutations at the Sox10 locus. Consequently, its downstream target Tyrp1 – and probably other pigmentation genes (PGs) – is also downregulated. Modified from Domyan et al., 2014.

Sequencing of Tyrp1 in brown birds, the other derived phenotype at the classic B locus, revealed a surprising amount of allelic heterogeneity: three distinct mutant alleles were identified among breeds (Domyan et al., 2014). All three mutations (two frameshift and one nonsense) are predicted to result in null alleles, consistent with Tyrp1 inactivation phenotypes in other vertebrates (Hubbard et al., 2010; Jackson, 1988). Collectively, variation at the B color locus in pigeons shows a wide range of effects on color. This includes the dominant B^A allele that causes a shift from eumelanism to pheomelanism, and recessive b alleles that arose at least 3 times within the rock pigeon.

Like the dominant ash-red (B^A) allele of Tyrp1, an autosomal mutation known as recessive red (e) also causes red feathers, and is epistatic to the B locus (Cole and Kelley, 1919; Haase et al., 1992). Consistent with these classical genetic results, recessive red birds have reduced expression of the gene Sox10 (Domyan et al., 2014), which encodes a transcriptional activator of Tyrp1 (Murisier et al., 2006). This downregulation is associated with two distinct, partially overlapping deletions of a known melanocyte enhancer of Sox10 in recessive red birds species-wide. It remains unclear why recessive red pigeons are red, however, given that the Tyrp1 loss-of-function phenotype is brown rather than red. This disparity strongly suggests that additional targets of SOX10 are also misregulated in the melanocytes of recessive red birds.

A third mutation, at the sex-linked dilute (d) locus (Cole, 1912; Cole and Kelley, 1919; Haase et al., 1992), is in a gene encoding a melanosomal solute carrier, Slc45a2, and is associated with lightened feather color (Domyan et al., 2014). This mutation is predicted to cause a histidine-to-arginine substitution in one of the protein’s transmembrane domains, presumably altering its ability to transport substrate across the melanosomal membrane.

One of the most-intensively studied pigmentation genes in vertebrates is melanocortin-1-receptor (Mc1r), which has been implicated in color variation in many organisms, including humans (Andersson, 2009; Gross et al., 2009; Hoekstra, 2006; Manceau et al., 2010; Mundy, 2005; Rosenblum et al., 2014; Sulem et al., 2007). Analysis of Mc1r sequences in feral pigeons with different color phenotypes did not identify an association between Mc1r genotype and pigment variation (Derelle et al., 2013). However, a second study found that a Val85Met substitution in MC1R was weakly associated with pheomelanism in domestic populations (Guernsey et al., 2013). Functional assays in cell culture demonstrate that the Val85Met allele is a hypomorph, and is correlated with reduced MC1R protein localization at the plasma membrane. Paradoxically, the same amino acid substitution is strongly associated with eumelanism in two other avian species (Baiao et al., 2007; Mundy et al., 2004). However, the contradicting associations are difficult to reconcile without functional tests of Val85Met alleles from these other species.

Together, variation in Tyrp1, Sox10, and Slc45a2 accounts for some of the most common melanin phenotypes in pigeons. The various combinations of genotypes at these three loci generate a broad spectrum of color diversity through both coding and regulatory mutations. A large number of modifier pigment loci also contribute to color variation (Levi, 1986; Sell, 1994, 2012), and an important line of future research is to determine their molecular identities and interplay with the already-identified loci.

Feathered feet

In domestic pigeons, the epidermis of the ankle and foot can be completely scaled or covered with feathers (Fig. 4). Classical breeding experiments demonstrate that diversity in epidermal appendage type is governed by a small number of loci (Doncaster, 1912; Hollander, 1937, 1983; Wexelsen, 1934). This epidermal variability is reminiscent of interspecific variation among wild birds, as some raptor and boreal species have feathered feet (e.g., some ptarmigans and owls). Previous studies of chicken development and genetics suggest changes in dermal-epidermal interactions as a mechanism governing the decision between scaled and feathered epidermis (Chang et al., 2004; Crowe et al., 1998; Dorshorst et al., 2010; Harris et al., 2002; Harris et al., 2004; Somes, 1992; Zou and Niswander, 1996). Recent findings about the molecular basis of epidermal variation in pigeons, however, suggest a more fundamental developmental basis (Domyan et al., 2016).

Fig. 4.

Developmental and genetic basis of feathered feet in pigeons. A–D. Ontogeny of the feathered-foot phenotype. A. Hamburger-Hamilton (Hamburger and Hamilton, 1951) stage 25 English trumpeter pigeon embryo showing ectopic expression of Tbx5 (arrows), a gene that normally is expressed in the forelimb (f) but not in the hindlimb (h). B. Lateral view of right hindlimb of late-stage English trumpeter embryo, showing elongated feather buds on the toes (arrows). C. Juvenile English trumpeter with large foot feathers (arrows). D. Comparison between scaled foot of a racing homer (left) and feathered foot of an English trumpeter (right). Feathers have been removed, but the large insertion points of lateral feathers on digit 3 are visible (arrowheads) and hide digit 4. Digit 1 is rotated anteriorly, unlike the posterior-facing digit 1 of the scaled breed (arrows). E. Two genomic regions are associated with foot feathering among pigeon breeds. Scaffold 79 contains Pitx1 and scaffold 70 contains Tbx5. F. Gene-expression model of epidermal appendage variation among pigeons. Grouse is associated with slight downregulation of Pitx1, which is transformed into small muff by synergy with ectopic Tbx5 expression. Large muff results from more-severe misregulation of both genes. f, forelimb. Modified from Domyan et al., 2016.

Darwin’s observations of heavily feathered pigeon feet (“muffs” in the parlance of hobbyists) provide some of the earliest anatomical clues about the potential molecular mechanisms that govern the transition from scaled to feathered epidermis. For example, he noted the feet of the English trumpeter breed were “so heavily feathered, they almost appear like little wings,” and, “pigeons with feathered feet very generally have the bases of their outer toes connected by skin” (Darwin, 1868). Furthermore, in the heavily muffed English trumpeters, the first digit of the foot (homologous to the big toe of humans) is sometimes rotated forward, rather than pointing backward as in other pigeons (Domyan et al., 2016). Together, these lines of anatomical evidence suggest forelimb-like traits in the hindlimbs of muffed birds, and support a model that the underlying mutations causing feathered feet are not simply altering the fate of individual epidermal placodes, but instead are altering hindlimb identity itself.

Consistent with a model of limb identity change, and with the small number of large-effect loci predicted by classical experiments, whole-genome scans of allele frequency differentiation showed that two loci are associated with the conversion of scaled to feathered epidermis across multiple pigeon breeds (Domyan et al., 2016). The first region contains the transcription factor Pitx1, a critical determinant of hindlimb identity and morphology (Duboc and Logan, 2011; Logan and Tabin, 1999; Szeto et al., 1999). Variation at the Pitx1 locus also underlies the repeated evolution of pelvic reduction in two species of stickleback fish (Chan et al., 2010; Coyle et al., 2007; Cresko et al., 2004; Shapiro et al., 2006; Shapiro et al., 2004; Shikano et al., 2013). The second genomic region contains the transcription factor Tbx5, which plays key roles in forelimb bud outgrowth and identity (Duboc and Logan, 2011; Hasson et al., 2007; Hasson et al., 2010; Minguillon et al., 2005; Rallis et al., 2003; Rodriguez-Esteban et al., 1999; Takeuchi et al., 1999; but see Minguillon et al., 2005). In an F₂ laboratory cross, genetic markers for Pitx1 and Tbx5 co-segregate with the extent of feather coverage on the foot and the length of foot feathers, respectively, providing independent confirmation of the association between phenotype and genotype.

Studies of pigeon diversity can move beyond genetic associations to measure the effects of allelic variants during embryonic development. In embryos of feather-footed birds, Pitx1 hindlimb expression levels are reduced relative to embryos of scaled birds. In contrast, Tbx5, which is normally not expressed in the vertebrate hindlimb, is expressed ectopically in the hindlimbs of feathered-foot embryos. Allele-specific expression tests in hybrid embryos of scale- and feather-footed parents show that these differences in Pitx1 and Tbx5 expression result from cis-regulatory changes. Surprisingly, ectopic expression of Tbx5 is not diffuse throughout the hindlimb, as is the endogenous pattern in the forelimb. Instead, it is expressed robustly at the posterior aspect of the limb bud, which is the site of outgrowth of the largest foot feathers (Fig. 4). Remarkably, feather-footed chicken embryos also manifest ectopic expression of TBX5 in the posterior hindlimb, due (at least in part) to a cis-regulatory change (Domyan et al., 2016). Thus, feathered feet in two domestic bird species are associated with tissue-specific regulatory changes in key limb development genes.

Taken together, these anatomical, genetic, and developmental data suggest that the feathered legs and feet of muffed pigeons result from the partial transformation of hindlimbs into forelimbs, and that this is due to cis-regulatory changes in two crucial determinants of limb identity and outgrowth (Fig. 4). Notably, some dinosaurs and basal birds had nearly completely feathered hindlimbs, including large, muff-like feathers on the posterior aspect of the leg and foot (Foth et al., 2014; Godefroit et al., 2014; Hu et al., 2009; Turner et al., 2012; Xu et al., 2003; Zheng et al., 2013). Although scales replaced these distal feathers in modern birds, feather-footed pigeons, chickens, and other species demonstrate that bird feet are still competent to form feathers when given the appropriate molecular signals. Thus, by understanding the molecular changes that can lead to the gain of feathers in modern pigeons, we might also begin to understand the genetic and developmental changes that led to the loss of foot feathers in proto-avian dinosaurs and basal birds.

Discussion

A novel entry point to general mechanisms of variation

Studies of pigeons illustrate how comparative genomics can inform developmental analyses to discover the molecular basis of variation. Notably, many of the traits that vary among pigeons also vary among other avian species (Baptista et al., 2009; Price, 2002). Complementary genetic, genomic, and developmental approaches are enabling the rapid discovery of the molecular basis of astonishing variation among pigeons, thereby opening new avenues to understand the potential roles of specific genes in variation among vertebrates in general (Baptista et al., 2009; Domyan et al., 2016; Vickrey et al., 2015). The limited fecundity of females and dearth of commercial hatcheries makes obtaining material for embryonic manipluation more difficult in pigeon than in chicken, but embryos of both species develop in ovo, and we expect that gene perturbation technologies developed for chicken will be readily-transferrable to the pigeon model, as well.

Future work holds promise to exploit the remarkable diversity of pigeons to define the genetic and developmental basis of additional phenotypic variation. For example, pigeon breeds differ in body mass by more than an order of magnitude (Levi, 1965). In addition, craniofacial variation is extreme within this species, ranging from the inconspicuous beak of the pigeon breed dubbed the African owl to the curved, buzzard-like face of the Scandaroon (Fig. 1C,D) (Levi, 1965). While we now have a better understanding of how different pigment types are specified, we know considerably less about how these pigments are deposited to generate color patterns within and among feathers (Lin et al., 2013; Willier and Rawles, 1940). Together, color pattern and pigment traits in myriad vertebrate species have a profound influence on mate choice, species recognition, and crypsis – traits that are probably under selection in free-living pigeon populations as well (Jacquin et al., 2012). Pigeons also offer numerous opportunities to understand the genetic basis of intriguing behaviors (Darwin, 1868; Hollander and Mangile, 1994; Sell, 1994). Of particular interest, rolling is a heritable behavior in which pigeons uncontrollably perform backward somersaults while in flight or when trying to initiate flight; it was described by Darwin (1868) as “one of the most remarkable inherited habits or instincts ever recorded” (p. 150). This behavior might be rooted in changes to serotonin signaling (Smith et al., 1987), providing a potential model for serotonergic control of skeletal muscle. In summary, selection in pigeons has produced an unusually broad exploration of morphospace for a single species, thereby providing ample opportunities to understand the genetic and developmental basis of variation.

Importance of domesticated animals in understanding evolutionary diversity

Domesticated species are important models for rapid evolutionary change over short time scales (Akey et al., 2010; Andersson, 2009; Boyko et al., 2010; Driscoll et al., 2009; Gaut, 2015; Price, 2002; Ross-Ibarra et al., 2007). Some of the same evolutionary processes that operate during selective sweeps and sexual selection in natural populations also shape domestic populations (Andersson, 2009; Careau et al., 2010; Gregory, 2009; Price, 2008; Price, 2002). Like their wild ancestors, domesticated species have undergone genetic adaptation to various climates and living conditions, including human preferences, making them excellent models for variation under natural selection and divergence by prezygotic isolation (Andersson, 2009; Driscoll et al., 2009; Hartl and Clark, 2007).

Furthermore, studies of domesticated species have led to the surprising realization that the same genes are sometimes involved in cases of natural and artificial selection. For example, derived alleles of Mc1r are associated with color variants not only in domestic cows, pigs, pigeons, and sheep, but also in wild populations of mammals, reptiles, birds, and fish (Andersson, 2009; Gross et al., 2009; Guernsey et al., 2013; Hoekstra, 2006; Manceau et al., 2010; Mundy, 2005; Rosenblum et al., 2014). In domestic dogs, variation at the IGF1 locus is associated with variation in body mass and age of death (Jones et al., 2008; Sutter et al., 2007), and this same gene plays critical and conserved roles in variation in body size and longevity in other vertebrates as well (Suh et al., 2008). Further illuminating the genetic basis of derived traits in domesticated and natural populations will help resolve whether selection is biased toward repeated changes in the same genes, or whether multiple developmental pathways are utilized in the repeated evolution of derived phenotypes.

Table 1.

Summary of known molecular bases of trait variation in pigeons

Trait	Symbol	Gene	Coding or regulatory	Molecular & developmental correlates	Notes (Reference)
Head crest	cr	EphB2	coding	Salt bridge disrupted in kinase domain, reversed polarity of feather placodes	Coding change in EphB2 associated with crest in ringneck dove (Shapiro et al., 2013)
Ash-red	BA	Tyrp1	coding	Signal peptide cleavage site disrupted, mislocalization of pigment sythesis	Dominant to B+ WT allele (Domyan et al., 2014)
Brown	b	Tyrp1	coding	Nonsense/missense mutations, decay of mRNA	b alleles evolved at least 3 times (Domyan et al., 2014)
Recessive red	e	Sox10	regulatory	Deletions of melanocyte enhancer, downregulation of Tyrp1	e alleles evolved at least 2 times; epistatic to B (Domyan et al., 2014)
Dilute	d	Slc45a2	coding	Substitution in transmembrane domain, functional effect unknown	Same gene associated with pigment dilution in several vertebrates (Domyan et al., 2014)
Pheomelanism	NA	Mc1r	coding	Substitution in intracellular loop, reduced plasma membrane-integrated protein	Same Val85Met substitution associated with eumalanism in other birds (Guernsey et al., 2013)
Grouse	gr	Pitx1	regulatory	Usually associated with 44-kb upstream deletion, controls extent of foot feathering	Synergizes with Sl to form muffs, second allele likely (Domyan et al., 2016)
Slipper	Sl	Tbx5	regulatory	Ectopic enhancer activity and hindlimb expression, controls extent and size of foot feathers	Synergizes with gr to form muffs (Domyan et al., 2016)

Table 2.

Selected derived and variable traits in domestic pigeons and their evolutionary and/or biomedical relevance; see Baptista et al. (2009) and Price (2002) for further discussion of avian traits

Trait in pigeon	Identified loci (where applicable)	Evolutionary and/or biomedical relevance	Selected References
Head crest	Ephb2	Avian head crests, sexual selection, directionality of epidermal appendage outgrowth, cell migration, palate development, axon guidance, prostate and colorectal cancer	(Baptista et al., 1997; Fu et al., 2009; Henkemeyer et al., 1996; Huusko et al., 2004; Kelsh et al., 2009; Risley et al., 2009; Vickrey et al., 2015)
Foot feathers	Pitx1, Tbx5	Epidermal appendage and limb identity, feathered feet of dinosaurs and birds, pelvic reduction in sticklebacks, clubfoot, Liebenberg and Holt- Oram syndromes	(Adachi et al., 2016; Al-Qattan et al., 2013; Chan et al., 2010; Gurnett et al., 2008; Postma et al., 2008; Shapiro et al., 2004)
Feather color (melanin type)	Tyrp1, Sox10, Slc45a2, Mc1r	Plumage, hair, and skin color diversity; crypsis and mate choice; melanoma risk	(Lao et al., 2007; Pho and Leachman, 2010; Rosenblum et al., 2014; Sturm, 2009; Tully, 2007)
Color pattern (melanin distribution)		Plumage, hair, and skin color diversity; crypsis and mate choice; melanocyte migration and survival, piebaldism; vision defects associated with at least 2 pattern phenotypes in pigeons	(Fleischman, 1992; Fleischman et al., 1991; Hollander, 1983; Levi, 1986; Lin et al., 2013; Linnen et al., 2013; Singh and Nusslein-Volhard, 2015)
Beak morphology		Adaptive diversity in craniofacial structures (including beaks), craniofacial developmental disorders	(Abzhanov and Tabin, 2004; Helms and Brugmann, 2007; Hu et al., 2015; Schneider, 2007; Smith et al., 2015)
Rolling behavior		Neurotransmitter and stereotypic behavior patterns	(D’Souza and Craig, 2008; Darwin, 1868; Mowrer, 1940; Smith et al., 1987)
Body size		Body size	(Durand and Rappold, 2013; Lettre, 2013; Wood et al., 2014)
Eye and nasal ceres		Ornamentation and mate choice, epidermal growth	(Lahaye et al., 2014)
Supernumerary tail feathers		Epidermal placode specification	(Houghton et al., 2005; Houghton et al., 2007; Morgan, 1911)
Homing behavior		Navigation and migration	(Beason and Wiltschko, 2015; Edelman et al., 2015; Guilford and Biro, 2014; Treiber et al., 2012)

• Domestic rock pigeons display dramatic variation within a single species.

• Pigeons are amenable to genetic, genomic, and developmental studies of diversity.

• The molecular basis for several classic traits is now known.

• Variation in pigeons is relevant to other evolutionary and biomedical phenotypes.

• Pigeons can reveal general mechanisms of avian and vertebrate diversity.