THE GENETIC BASIS OF SEXUAL DIMORPHISM IN BIRDS

As implied by Fisher in the quotation given above, there are two ways that genes can produce the striking sexual dimorphisms seen in many species of animals. The first is that the alleles for dimorphic traits could be male-limited from the outset, with their expression perhaps conditioned on physiological or hormonal differences between the sexes (we assume for convenience that such traits are seen in males but not females). Alternatively, these alleles could be expressed initially in both sexes, but then become limited to males by the accumulation via selection of modifier genes (the presumption is that male-specific ornaments are sexually selected traits that are deleterious in females). This sex limitation could itself evolve via two routes: the accumulation of modifiers (alleles or regulatory elements) that (1) suppress expression only in females, or (2) promote expression only in males. At the end of all of these processes, only males would show sexually dimorphic traits. How, then, do we determine which genetic pathway has occurred?

Unable to see how selection could limit to a single sex a trait originally expressed in both sexes, Darwin (1871, p. 182) believed that dimorphic traits were male-limited from their first appearance:

Fisher (1930), however, with characteristic enthusiasm for the evolution of modifier alleles (a famous aspect of his theory about the evolution of dominance), ascribed sexual dimorphisms almost entirely to the appearance of the trait in both sexes followed by the accumulation of suppressors in females. Yet there was no empirical evidence for this hypothesis at the time of Fisher's writing, nor is there any today.

Here we suggest one approach to this question. This involves studying hybrids between species showing different forms or degrees of sexual dimorphism. Because the different evolutionary ways that alleles could become male-specific in expression make different predictions about what one might see in female hybrids, an analysis of these hybrids might help us discriminate between the hypotheses.

We envision three scenarios. In the first, we assume that in two closely related species, males differ in the type or degree of their sexually dimorphic traits, and that the traits were originally expressed in both sexes but later suppressed in females by the accumulation of modifiers. (These modifiers could be either trans-acting genes or regulatory elements, or closely linked cis-regulatory elements). If these two species were interfertile, their F₁ female hybrids would carry some genes for sexual dimorphisms from each parental species (except for W-linked genes, because female birds are heterogametic), but only a haploid set of suppressors from each species. Unless the suppressors are completely dominant, hybrid females should show some expression of the male traits.

In the second scenario, the alleles now expressed only in males could have originally been expressed in both sexes, but the evolution of “modifiers” (again, trans-acting genes or trans- or cis-acting regulatory elements), could have led to those alleles being expressed only in males, perhaps by bringing their transcription solely under hormonal control. This leads to the same prediction as that for female suppressors: unless the male-limited modifiers or regulatory elements are completely dominant, hybrid females between two species differing in male-specific traits should show some expression of those traits. Although the first and second scenarios are mechanistically different, they could presumably occur together, with the simultaneous evolution of female suppressors and male promoters.

It is important to stress here that although alleles may be presently activated or suppressed by evolved sex-specific modifiers (perhaps regulated by hormones), this does not mean that male traits could not appear in female hybrids. This is because these females resemble in genetic constitution a partial ancestral state in which (if the genes were not sex-limited from the outset) “dimorphism” alleles were expressed in both sexes.

The third possibility is that alleles could be male-limited from the outset, perhaps because they arose at genes that were already male-limited, or arose fortuitously in regions under control of cis-dominant elements activated only in males. In such cases, female hybrids would show no expression of male traits and would presumably resemble intermediates between the nonornamented females of the parental species.

Hybridization between species thus offers a way to study the genetic basis of sexual dimorphisms. Species hybrids have been used in other studies to reconstruct ancestral genetic conditions, for example to look for meiotic drive alleles that have been fixed in species but whose presence is undetectable unless they are made heterozygous in species hybrids (Coyne 1989; Coyne and Orr 1993).

Here we examine hybrid birds to determine whether female interspecific hybrids show reexpression of “dimorphism” alleles that may have been originally expressed in both sexes but later suppressed in females. Birds are an obvious group in which to examine this question because many species show extreme sexual dimorphisms, are easily crossed, and hybrids have been documented extensively (e.g., Gray 1958; McCarthy 2006). Mundy (2006, p. 495) even suggested that these hybrids could be used to study the genetics of sexual dichromatism:

We will show that, based on the hybrids and hybrid specimens available, we find virtually no evidence for the expression of male-limited traits in female species hybrids, suggesting that the genes for such traits are either male-limited from the outset or controlled by completely dominant sex-limited modifiers.

Methods and Materials

We began our search for hybrid females with the recently published book Handbook of Avian Hybrids of the World (McCarthy 2006), which describes approximately 4000 interspecific and intersubspecific hybrids. We also looked for other female hybrids by communicating with ornithologists, bird curators, and other researchers who study avian hybrid zones. All of these workers had personally examined hybrids.

We narrowed our search to female hybrids between sexually dimorphic species whose males possess different types of ornamental traits (e.g., distribution of color patches; presence/absence of headwires, elongated feathers, or wattles). This maximized our chances of seeing expression of dimorphic traits in female hybrids. Related species whose males share similar types of dimorphic traits (i.e., species having the same area of the body pigmented, but in different colors) may be less useful for investigating the origin of sexual dimorphisms: such species might have inherited both the dimorphic trait and any female suppressors (or male promoters) from a common ancestor, and the lack of expression in female hybrids might simply reflect their homozygosity for identical suppressors or promoters from that ancestor. We thus excluded these cases (which were not numerous) from our study. For similar reasons, we excluded hybrids between subspecies.

We further restricted our search to female hybrids who were adults at the time of collection or observation. Age was determined from either information on the specimen tag or the literature. In nearly all sexually dimorphic birds, males do not acquire dimorphic traits until they reach or approach breeding age. By eliminating female hybrids collected as juveniles or immatures, we avoided the possibility of their not showing male traits simply because they were too young.

Some bird species also show seasonal variation in dimorphic traits, with males of most species molting into a plumage less bright or less exaggerated during the nonbreeding season. Were this the case for some of our hybrids, it is possible that they would not display male traits simply because they were collected during the winter months. However, in nearly every case included in our analysis, the winter plumage of male parental species differs substantially from that of females. Thus, our results should not be markedly affected by the season at which the hybrid individuals were collected, as some evidence of male traits in mature individuals would be detectable year round. The only two cases in our dataset in which male traits disappear during molting involve the two duck hybrids. In most duck species, males drop their body feathers rapidly twice each year during a period of “eclipse plumage,” and resemble females for a few weeks. It is possible that if the female duck hybrids given in Table 1 were collected during this time, we may have missed some male-like traits expressed during most of the year. Nevertheless, we consider this fairly unlikely because the interval of eclipse plumage occupies only a small proportion of the year.

The number of known hybrid females that meet all of the above criteria was surprisingly small (see below), although we did not have the resources to personally examine all known hybrid specimens in museums. We acknowledge that our list is incomplete. Nevertheless, it is a representative list because it includes hybrids in a wide range of taxa collected in many places.

It is also likely that many female hybrids have been collected or observed but have not been correctly identified. The reason for this has to do with how hybrids are recognized. Excluding cases of hybridization in captivity, where the parentage of a hybrid is known precisely, all hybrid specimens that we found were identified as hybrids either through DNA sequencing (one case: Lazuli × Indigo Bunting), because they were reared in captivity and parentage was known (six cases), or because the individual showed morphological or plumage traits intermediate between females of parental species who themselves had nonoverlapping distributions of these traits (six cases). This latter method has been used in the ornithological literature for over 100 years as the standard way to identify hybrids. Table 1 shows the methods by which each hybrid was identified.

Because in many cases males of the parental species differ in plumage but not size, it is often quite easy to identify male hybrids but difficult or impossible to identify female hybrids. There are, for example, no fewer than 30 known specimens of hybrid males between the King Bird of Paradise (Cicinnurus regius) and the Magnificent Bird of Paradise (C. magnificus), yet not a single description of a female hybrid. The explanation is almost certainly that females of the two parental species are nearly identical in plumage and morphology, making it impossible without genetic evidence to identify female hybrids.

Results and Discussion

Table 1 summarizes the crosses, the types of traits differing between male parent species, and a description of female hybrids. Of the more than 4000 hybrids we considered, only 13 pairs met our criteria. However, every one of these cases gives the same result: F₁ female hybrids do not show any sexually dimorphic traits seen in males of the parental species. We note that this lack of expression is not because alleles for male-specific traits are simply recessive in hybrids, for these traits are seen in male hybrids, which invariably show dimorphic traits seen in males of both parental species (see, for example, plate 15 of Frith and Beehler 1998, showing the appearance of male interspecific hybrids in birds of paradise).

A striking illustration of the absence of male traits in female hybrids is given in plate 14 of Frith and Beehler (1998), depicting hybrids in two pair of birds of paradise, each pair representing an intergeneric cross (Table 1, see cover). Despite the fact that males in each cross differ by diverse and elaborate characters such as head wires, tail streamers, and iridescent colors, the female hybrids are drab brown, lacking any trace of male ornaments, and resembling the females of the parental species.

Because only 0.4% of all known bird hybrids are represented in Table 1, it is possible that there is some ascertainment bias in our table, and that our conclusions might not be general. But this possibility seems unlikely because our sampling represents four orders and eight families of birds.

We conclude, then, that in our sample of species (and perhaps in birds in general), male-limited dimorphic traits evolved in either (or both) of two ways (1) The alleles responsible for such traits were from their initial appearance expressed only in males. (2) The alleles responsible for such traits were initially expressed in both sexes, but then were either suppressed in females or became limited to males by alleles or regulatory regions that are completely dominant in hybrids.

We cannot from our data alone discriminate between these two hypotheses. If the second hypothesis is true, the most likely explanation for dominance would be cis-dominance, perhaps because “dimorphism” alleles have been brought under control by a promoter region either suppressed or activated by hormones, or by some product of genes in a pathway that is itself activated by hormones. (One example of a cis-dominant promoter region controlling a male-specific allele is the promoter element of the yellow gene in Drosophila biarmipes[Gompel et al. 2005], which is involved in creating a male-specific wing spot used in mating.) Moreover, if the second hypothesis is true, then the absence of male traits in any female hybrids implies that the “dimorphism” alleles come under the control of regulatory genes fairly quickly, before speciation is completed.

In principle, repressors or activators that are not dominant could be identified by backcrossing hybrids to either parental species. Some backcross females should be homozygous for suppressors or activators from one species but carrying “dimorphism” alleles for the other, and should thus show male traits. Alternatively, one could look for male-like females in hybrid zones (which may contain “natural” advanced intercross or backcross individuals) between two species having different dimorphisms. These techniques will not work, however, if the dimorphism genes are controlled by cis-dominant elements, as the genes and their regulators will not be separable in backcrosses.

We should emphasize that modifier alleles or regulatory elements need not inevitably act in a dominant fashion. Even cis-acting regulators need not be completely dominant (Duncan 2002). Regulators (and proteins involved in regulation, whether linked or not) can also be trans-acting, and thus not invariably dominant. Transcription factors, for example, are structural proteins that need not be linked to the genes they regulate, and other proteins can act as cofactors during gene regulation; these possibilities show that gene regulation can evolve that is not cis-dominant (Hoekstra and Coyne 2007). For example, in the study of Gompel et al. (2005) described above, expression of the full wing spot requires trans-acting in addition to cis-acting factors.

What traits could any male-limited genes “cue on” to limit their expression to only one sex? The most obvious—and most likely—is hormones. It is well known that in birds the expression of sexually dimorphic traits is strongly affected by hormones (Kimball 2006). In most groups of birds, plumage traits (e.g., feather color or elongation, appearance of color patches) appear to be estrogen dependent, with male-specific traits expressed only in the absence of that hormone. In the Charadriiformes and Passeriformes, however, male plumage appears to be testosterone dependent. In most groups, testosterone is also involved in inducing nonplumage traits such as bill and leg color, presence of wattles, spurs, and combs, as well as behavioral traits such as aggression and male-specific sex displays (Owens and Short 1995; Kimball 2006). (In a few species, male-specific plumage traits are produced directly by the presence of homogametic sex chromosomes.)

The molecular-genetic basis of sexually dimorphic morphological traits in birds is unknown, but much is known about the molecular basis of sex-specific gene expression in other taxa. In some cases, hormones such as androgen or estrogen bind to transcription factors, and these complexes then interact with target sequences such as enhancers, causing sex limitation of gene expression—and of the traits affected (e.g., Ning and Robins 1999; Claessens 2001). This raises the possibility that evolution of the transcription factors or cofactors, rather than of regulatory sequences themselves, may be involved in the sex-limitation of avian dimorphic traits.

It seems likely, then, that any male-specific genes somehow cue their expression on the absence of estrogen or the presence of testosterone. This is supported by two anecdotal reports that female hybrids in pheasants and ducks develop partial male-like plumage when they are old or diseased, conditions that can reduce the amount of estrogen (White 1900; Thomas 1914), a hormone whose absence is known to induce male traits in Galliformes and Anseriformes (Kimball 2006).

A supplemental approach to studying the genetics of sexual dimorphism could involve manipulations, such as ovarectomies or testosterone injections, which change the hormone titer in female hybrids of reciprocal crosses. The consequent expression of male plumage in the reciprocal-hybrid heterogametic females (assuming that such plumage is hormone dependent) could indicate whether interspecific differences in such traits reside largely on the X chromosome. Such manipulations in backcross females could further elucidate the genetics of dimorphic traits, including whether any modifiers are dominant and always cosegregate with the male trait (implying cis-dominant regulation).

We should note that our approach was limited to plumage colors and ornaments in birds, but in principle could be extended to behavioral traits. For example, the behavior of hybrid females might also be examined in crosses between species whose males show different mating displays. Our approach could, of course, also be used in species other than birds.

Finally, there is evidence that some alleles involved in some sexual dimorphisms have not been male-limited from the outset. These are cases in which females display some exaggerated male-like traits, but traits less extreme than seen in males. In the birds of paradise, for example, male astrapias have very long tails (Frith and Beehler 1998, plate 6); females have tails shorter than those of conspecific males, but clearly longer than those of females from related species. It is not clear whether in such cases the exaggerated female trait may actually have an adaptive function (e.g., social signaling or species recognition). Rudimentary female versions of male traits might also function in both attracting mates and in female–female contest competitions (Amundsen 2000). Nevertheless, the presence of similar traits in both sexes, with the males showing a more exaggerated form, suggests (but does not prove) that such traits are based on the same alleles in males and females.

In sum, we reinforce the suggestion of Mundy (2006) that the study of hybrids can illuminate, in part, the genetic basis of important morphological evolution. Because sexually dimorphic traits are involved in processes as critical as mate choice and reproductive isolation, this approach may ultimately help us understand sexual selection in general, and speciation in particular. If our results are general, they suggest that the origin of sexually dimorphic traits in birds has followed a consistent pathway in all taxa.

Associate Editor: J. True