40 Years of Evolution: Darwin’s Finches on Daphne Major Island

The ground finches differ in size and shape. The three, coexisting, granivore species on Daphne differ by at least 15% in one beak dimension, which is equivalent to approximately two standard deviations either side of each mean value (Grant 1986). Beak dimensions change across decades as a result of fluctuating selection (chapter 11), but how much selection is required to transform one species into another by changing a mean by 15%? This question can be approached at two different but complementary life history stages: adults or embryos. The adult approach uses genetic variances and covariances for the species on Daphne and the phenotypic (morphological) differences between them (Price et al. 1984a) to estimate the minimum cumulative or net forces of selection required to effect the transformation of adults (Lande 1979, Price et al. 1984a, Schluter 1984). The other approach examines the underlying genetic developmental program leading to different beak sizes and shapes. We will discuss the two approaches in turn, and consider how together they reveal the way changes in size and shape can occur.

These points are illustrated with calculations of residual shape change that remains after size changes have been accomplished in the transformation of one species to another (box 12.1). A low value of residual shape change in standard deviation units implies that the transition from one species to another has been largely accomplished by selection on size alone. A high value implies that selection on shape itself is needed. In this exercise we are not attempting to reproduce the actual transitions in both size and shape, nor are we using a known, statistically well-supported phylogeny, which is still not available (Petren et al. 2005). Instead, we use the calculations to illustrate the fact that some morphological transformations are much more easily accomplished than others, perhaps more likely to have occurred than others (Schluter 1996), and possibly to have occurred in less time. In doing so we assume that the genetic covariance matrix changes relatively little during the transformations, which may not always be correct (Grant and Grant 2000a, Björklund et al. 2013). We also assume a continued and sufficient supply of genetic variation, which is supported by experience of long-term responses to selection in animal and plant breeding (Hill and Kirkpatrick 2010). Strong directional selection, continued for up to 50 or 100 generations, has produced continued evolutionary responses in flies (Weber 1996), mice (Keightley 1998), and corn (Walsh 2004). This has been interpreted as the result of new mutations arising continuously and supplementing standing genetic variation (Hill and Kirkpatrick 2010).

Selection on a size factor alone can easily transform species differing in size but scarcely in shape (fig. 12.1). The transformation from fortis to fuliginosa morphology is a clear example of this. The little variation in shape that is left after selection on size is actually less than the observed change in beak shape from selection in 1985 and 1986 (fig. 11.12). The fortis to magnirostris transition differs in that the correlated effect on shape of selection on size has been too strong. Genetic facilitation rather than genetic resistance has caused an overshoot of the shape target.

Box 12.1. Species Transitions in Morphology

The core species with genetic and phenotypic data are fortis and scandens from Daphne (chapter 3). We lack quantitative genetic data from fuliginosa and magnirostris, and therefore used fortis, the most similar species, as the starting point for reconstructing the transitions to fuliginosa and magnirostris morphologies. We used equal numbers of families for a correlation-based principal components analysis to characterize structural size variation among pairs of these species as the first component. Between 63.0% and 84.9% of the total variation is accounted for in each analysis by PC1, and in all analyses the six morphological traits have approximately equal loadings, which justifies using it as a measure of overall size. The second principal component accounts for most of the remaining variation, 7.8%–25.7%. In each analysis, loadings on PC2 for beak depth and beak width are large and positive, whereas the loading for beak length is negative, and loadings for the other traits are generally small. Therefore, PC2 is mainly a beak-shape factor, varying from pointed to blunt (also Boag 1983, Grant and Grant 1989; chapter 4).

Next we calculated heritabilities for PC1 and PC2 scores for fortis and scandens and genetic correlations between them. Even though the two axes of variation are uncorrelated in the combined data, the individual species do sometimes display positive correlations. Heritable variation for the size and shape factors and the additive genetic correlation between them vary strongly among the species (table 1.10 in Grant and Grant 2000a). Third, we calculated the net forces of selection on the size factor (alone) involved in each of the four possible (reciprocal) species transformations using these genetic parameters and the phenotypic distances in standard deviation units of principal components scores. Shape changes arising from selection solely on size were calculated from these measures of selection and genetic variation and covariation; they are the product of the net forces of selection on size, the square root of the heritability of size, and the genetic correlation between size and shape (Lande 1979, Price et al. 1984b, Schluter 1984). What remains is residual shape differences between species.

Among the four species on Daphne those that differ most in shape, fortis and scandens, are the least interconvertible by selection. Asterisks in the lower part of figure 12.1 show that selection on size of either species results in a correlated shape change in the opposite direction, so that the difference between species in shape becomes magnified as selection decreases the difference in size, and vice versa. We reason from this conflict that the ancestor was possibly smaller than both of the extant species and possessed a pointed beak, for then each transformation would have proceeded with little genetic resistance. This reasoning is consistent with the molecular phylogeny (Petren et al. 1999b, 2005, Sato et al. 1999) that shows pointed-beaked species, and specifically Geospiza difficilis (Sharp-beaked Ground Finch), to be basal to cactus and granivorous ground finches.

In the absence of genetic incompatibilities, the genetics of speciation is the genetics of beak divergence. What are the genetic changes involved in the morphological changes taking place during speciation? The place to look for them is in development (Price and Grant 1985, Björklund 1993, Grant et al. 2006). The four species differ in beak proportions at the time of hatching, which implies differences in embryonic growth, and they also differ in relative growth rates of different dimensions after hatching (Grant 1981c). Hence the genetic divergence that accompanied species formation involved evolution of developmental differences. Recent discoveries in developmental genetics have identified genes affecting embryonic beak development in the finches, and have characterized differences in expression patterns among closely related species (fig. 12.2). These are parallel to discoveries in plants and animals in other systems, in both plants and animals, of key regulatory genes that have undergone change by mutation (Shapiro et al. 2004), introgression (Kim et al. 2008), or polyploidization (Barrier et al. 2001), and been subject to selection.

G. magnirostris, fortis, and fuliginosa differ in the size and robustness of their beaks. This variation in adult beak morphology is paralleled by variation in the expression of a gene responsible for producing a signaling molecule, Bmp4, in the prenasal cartilage (pnc) in all three species at days 5 to 6 during 12 days of embryonic development—earlier and more intensely in the development of magnirostris than in the other species (Abzhanov et al. 2004; also Wu et al. 2004, Campàs et al. 2010). The association is more than just a correlation: experiments have demonstrated a causal role of the molecule (Abzhanov et al. 2004). Thus agents that govern the timing and intensity of expression of the gene, such as other gene products and feedback loops, modulate the development of these beaks differently in the three species. Later in embryonic development the premaxillary bone (pmx) is laid down, and another interacting set of genes—TGFβIIr, βcatenin and Dickkopf-3—have similar differential effects on beak development of the three species (Mallarino et al. 2011). Relatively simple changes in gene regulation and the timing of gene expression must have occurred along the fuliginosa-fortis-magnirostris axis of the Darwin’s finch radiation.

The same conclusion of potentially rapid speciation has been reached in a study of three species of crossbills (Loxia) in Europe. They differ in body size and beak size in a manner parallel to the fuliginosa-fortis-magnirostris sequence, and differ ecologically too (Marquiss and Rae 2002), yet no difference in mitochondrial and microsatellite DNA has been found, a failure that cannot be attributed to introgressive hybridization (Piertney et al. 2001). Reproductively and ecologically they are three species, whereas genetically, at the studied loci, they are one (Piertney et al. 2001). These examples share with each other and with several other closely related species of birds (e.g., Lack 1944, Schoener 1984) the characteristic of feeding on foods of different size. As more examples of rapid speciation come to light (Milá et al. 2007, Moyle et al. 2009, Kirchman 2012), more will be sought and more will be found.

A critical stage in the evolution of beak differences between species is the encounter of two populations derived from a common ancestor. There must have been countless dispersal and colonization events throughout the archipelago in the course of the Darwin’s finch radiation. We witnessed one in 40 years. It gave us insights into both ecological and reproductive interactions between colonists and residents at this critical stage of the speciation cycle when divergence can be enhanced or reversed. Such interactions, occurring frequently, may have contributed to the rapid diversification of the ground finches.

Colonization by magnirostris led to ecological character displacement of fortis (chapter 7). This demonstrates an evolutionary adjustment of competitors to a limited resource. The two species did not interbreed either at initial contact or any time in the following 30 years, and therefore they are not strictly an example of secondary contact of differentiated populations as in figure 1.1. A hypothetical example of this process would be colonization of Daphne by unusually large fortis from another island such as San Cristóbal. They are 20% larger in average beak depth on this island than on Daphne, and with such large beaks they would be, like magnirostris, superior exploiters of Tribulus fruits.

Lack of interbreeding between fortis and magnirostris on Daphne shows that they were sufficiently different in beak morphology and song. Both cues are important in mate recognition and choice, but one piece of evidence suggests morphology can be the more important. A minimum of nine fortis males sang magnirostris song, presumably through learning the song during their sensitive period (figs. 8.6 and 8.9), yet none of them bred with magnirostris females. G. magnirostris males recognized their songs and responded by chasing the fortis. We interpret these observations to signify that the beak and body size differences between the species were large enough to prevent interbreeding, regardless of their song.

Nevertheless interactions involving song differences are also important at this stage of secondary contact. The songs of both fortis and scandens underwent a change after magnirostris numbers had increased, starting in the 1990s (Grant and Grant 2010c). Songs of both species became faster in rate of note repetition or trill rate, and in doing so they diverged from the songs of magnirostris but not from each other (fig. 12.5). Divergence began well before the character displacement shift in fortis, and therefore cannot be simply accounted for as a correlated effect of a change in beak morphology that affects song characteristics (Podos 1997, 2001, Huber and Podos 2006). It happened without a change in the structural habitat more than one meter above ground where birds sing, therefore it cannot be explained as a change in the sound transmission properties of the environment (Wiley 1991, Luther and Wiley 2009). Since songs of two species, fortis and scandens, changed in the same direction and at the same time, divergence is unlikely to have been caused by random changes (cultural drift). The most likely explanation is acoustic interference between species (Nelson and Marler 1990, Luther 2009). G. magnirostris are large and aggressive, frequently interfering with and attempting to take over the nests of fortis and scandens. Their songs are louder and are in the same frequency bandwidth as the songs of fortis and scandens. According to this explanation, divergence reduces acoustic and physical interference (Grether et al. 2009), as well as improving communication through learning (Peters et al. 2012, Ríos-Chelén et al. 2012) and transmission (Luther and Wiley 2009, Nemeth et al. 2013) in a noisy environment.