2005 Balzan Prize for Population Biology
Population Biology and Speciation – Bern, 11.11.2005
A fundamental problem in biology is to understand the steps involved in the process of speciation, because the question of how one species splits into two addresses the foundation of the biodiversity we see around us in the world today.
Beginning in 1973 we chose to address this problem with a study of Darwin’s finches in the Galápagos archipelago. This young adaptive radiation in an isolated archipelago is a particularly suitable system for asking questions about adaptation and the multiplication of species: how these processes happen, and how to interpret them. The unique advantages are that all 14 species of Darwin’s finches are closely related, having been derived from a common ancestor in the last 2-3 million years. None of the species has become extinct as a result of human activity, thereby creating missing links. They live in the largely undisturbed environment in which they evolved, they are distributed in various combinations across islands, and they exhibit various degrees of morphological distinctness. Thus, considering populations across the entire archipelago, it is as though the whole process of speciation in all of its stages from the initial colonization through divergence to the formation of a reproductive barrier between species has been frozen in space, allowing us to investigate each step in turn.
Three inter-related problems preoccupied us from the outset: how is speciation to be explained; how, and how often, does adaptation to the ecological environment occur; and how is quantitative genetic variation in small populations maintained in the face of depleting forces of natural selection and random genetic drift. To address these problems we developed the strategy of combining archipelago-wide comparative studies of different combinations of finch species with intensive long-term studies of population processes on the islands of Genovesa (11 years) and Daphne Major (33 years). By simultaneously taking account of ecological, behavioral and genetical aspects of evolution, and being alert to their interconnectedness, we hoped to reach a greater understanding of finch evolution than would be possible from a narrower focus on just one of them.
Speciation begins when a new population is established, continues with the divergence of that population from its parent population, and is completed when members of two diverged populations coexist in the same place without interbreeding. The question is how do these processes occur? We stand a virtually negligible chance of observing the whole process under natural conditions. Nevertheless it is possible to make relevant observations in nature of all steps in the process.
The first step involves ecological divergence. One way this happens is through the fundamental evolutionary process of adaptation to the environment through natural selection of heritable phenotypic variation. Usually adaptation is inferred, rarely is it studied directly. Our studies have been able to accomplish both.
Adaptation of finches on different islands had previously been inferred from the association between beak size and shape of several species on the one hand and characteristics of the diet on the other (Lack, D., Darwin’s Finches, Cambridge University Press, Cambridge, 1947). Our initial studies strengthened the inference through sampling, measurement and quantification of the physical characteristics of foods available in the environment on several islands, the foods actually eaten, and beak dimensions.
Earlier, David Lack had two important insights into the adaptive evolutionary history of the finches. These were a) competitive interactions between species explain some mutually exclusive distribution patterns and enhanced beak size differences between sympatric finches (character displacement), and b) competitive interactions are expected to be most severe in the dry season when food is likely to be in short supply. Both were supported by our fieldwork. A modelling exercise produced a better fit between observed and predicted beak sizes when competitive interactions were incorporated than when they were not. Dry season food supplies are indeed lower than wet season supplies, with few exceptions, and as food supply declines so do population sizes. Thus species adapt to the environment, and an important part of the environment is the presence of competitor species. Both of Lack’s insights were valuable and correct.
Natural Selection Observed
Inferences about adaptive evolution in the past are more plausible if supported by a demonstration of evolution in the present. We gained insights into adaptation through long-term studies of ground finch (Geospiza) populations on both Genovesa and Daphne Major. By capturing and measuring a large number of finches and marking them with unique combinations of colored leg bands we were able to follow their fates. This simple and well-used technique of population study revealed the tempo of ecological and evolutionary change. It showed that ground finches can live as long as 16 years, which was a surprise given the occasionally harsh climatic environment of lowland Galápagos habitats, and shorter maximum lives of north temperate relatives. More remarkable was the discovery that significant adaptive change takes place in less than a year, and is therefore, potentially at least, a rapid process and not an immeasurably slow one as was believed at that time.
Natural selection was clearly demonstrated during the drought on Daphne Major Island in 1977. Medium ground finches (Geospiza fortis) consumed a large fraction of the supply of small seeds, and then turned increasingly towards the large and hard seeds, now in relatively high abundance. Only large finches with large beaks could crack the seeds and extract the kernels, hence they survived at a relatively high frequency. Natural selection had occurred. Over the 33 years we discovered that natural selection is not restricted to one trait, one species or one island; it also occurred in the population of cactus finches (Geospiza scandens) on Daphne, and in G. conirostris (large cactus finch) and G. magnirostris (large ground finch) on Genovesa. Natural selection occurs frequently, and varies in direction and strength according to the particular set of environmental conditions.
For natural selection to cause evolution the trait in question must be heritable. We were able to show that the morphological traits subject to natural selection (beak and body size) were highly heritable, and that genetic correlations among them were generally very strong and uniformly positive. These results were used to predict evolutionary responses to repeated natural selection, and predictions were then compared to actual changes from generation to generation. Matches between observations and predictions were close, both after the 1977 drought and following selection events in l ater years, thus demonstrating that natural selection drives frequent evolutionary changes in nature. This provides an important insight into the evolutionary lability of populations when subjected to environmental change. For example it helps to make sense of large changes in several organisms occurring in an apparently short period of time in the paleontological record.
An additional insight was the realization that continuous monitoring of the environment is as important for an understanding of long-term evolution as is the continuous monitoring of the finch populations themselves. This insight emerged from a study of the effects of the El Niño event of 1983. Unprecedented in the century in length and severity, it was followed two years later by a drought. The contrast could scarcely have been more extreme. A simple expectation would have been that the selection event of 1977 would be repeated in 1985. That expectation would have been entirely wrong, because the vegetation and hence food environment had changed profoundly during 1983, from dominance of large-seed producing plants to dominance of small-seed producing ones. This change lasted for a long time. Small seeds continued to be abundant in the seed bank until 1999. In the most recent six, relatively dry, years the vegetation has changed both in composition and quantity back towards the pre-1983 condition, but has not reached it yet.
The important point is that no two droughts have the same effects; the particulars depend on initial conditions. Droughts have selective or non-selective effects on the finches depending on whether large seeds or small ones (and which ones) are plentiful at the start of a drought, as well as on the preceding evolutionary history of the finches. The drought of 1977 had strong selective effects on the beak traits of two species of finches, the 1985-86 drought had weaker effects in an opposite direction on beak traits of one of the species, and the drought of 1988-89 had no selective effects on either.
The study provides an empirical foundation for general theories concerning adaptation to food resources with or without the diversification of species. Additionally it has yielded the surprising fact that as a result of natural selection the finches on Daphne Major Island are not morphologically or genetically the same as their ancestors were 30 years ago. Their environment has changed, and so have they. In collaboration with colleagues at Harvard University, Arkhat Abzhanov and Cliff Tabin, we are currently exploring evolutionary change of beak dimensions at the molecular genetic level.
Colonization and Founder Effects
The classical founder effect model of speciation of Ernst Mayr proposes that the crucial step in speciation occurs right at the beginning of the founding of a new population by a few individuals. Rapid evolution could occur as a result of inbreeding, random genetic drift, the loss of genetic variation, and selection of sets of genes whose interactions change as a result of substantially altered allele frequencies. If the changes are profound enough they may yield a new species, reproductively isolated from the population that gave rise to it when they later come into contact. Although much has been written about possible evolutionary consequences of founder events, there have been very few opportunities to examine the genetic changes that actually occur in nature. We were fortunate to witness the founding of a new population on Daphne Major Island, and were thus able to make a direct assessment of this model. Two female and three male large ground finches Geospiza magnirostris established a breeding population at the end of 1982 when a major El Niño event had just got underway. We followed their fates and the fates of their descendants for the next 20 years.
As expected from the founder effects model inbreeding depression occurred in the first 10 years, and was actually more severe than ever recorded in the two resident populations of G. fortis and G. scandens. However survival of the first inbred individuals was unexpectedly high. Often more can be learned from exceptions like this than from observations that agree with expectations. The insight gained from this surprising observation was that genetic disadvantages from inbreeding can be over-ridden by the ecological advantages of abundant food and the absence of density dependent effects on survival. Another unexpected observation was recurrent immigration, from more than one island source and not just from the island closest to Daphne. Recurrent immigration contributed to the elimination of any long-lasting effects of inbreeding, and to the fact that no substantial change in either phenotypic traits or in microsatellite allele frequencies occurred. Insofar as the observations can be generalized they provide no support for the model of founder effect speciation, and showed how it can fail. Yet even if this specific model is not applicable, speciation may nonetheless be more likely to proceed in small insular populations than in large ones, as discussed below.
Byproduct of Adaptation
An alternative to speciation through founder effects is Theodosius Dobzhansky’s adaptive byproduct hypothesis of speciation. If the same morphological traits that undergo adaptive change also function in a reproductive context when individuals choose mates then speciation might proceed as a simple consequence of the adaptive change. No special genetic factors need be invoked to explain the splitting of ancestral populations into separate species, since ecological factors are the primary driving force of differentiation according to this explanation, and the avoidance of interbreeding (premating isolation) when they later come into contact with each other is a passive, secondary, consequence of the adaptive change.
Our early studies on Daphne, Genovesa and several other islands showed the importance of ecological factors, mainly the composition of the food supply, which varies from island to island. With this in mind it is easy to see how finches that dispersed from one island and established a new breeding population on another would encounter a different constellation of foods, and undergo evolutionary adaptation in beak traits to the new food supply. By itself this would raise barriers to gene exchange between the original and the derived populations if beak size and shape were also the key traits used by finches in choosing a mate, and if adaptive divergence was large. Discrimination experiments with stuffed museum specimens of birds supported the hypothesis. Responding birds of different ground finch species tested on Daphne, Genovesa and other islands discriminated between local members of their own species and either a resident, closely-related, species or different conspecifics from another island. They did so on the basis of beak size and shape differences. Two other possibilities can be ruled out; body sizes were the same in some of the tested pairs and plumages were identical in all of them.
One awkward fact does not fit the byproduct hypothesis: on Daphne and elsewhere the largest members of a small species (e.g. G. fuliginosa) are more similar in beak size to the smallest members of a larger species (G. fortis) than they are to most members of their own population. If beak traits are the only factor in mate choice the species should hybridize, and those individuals that do so should be similar in beak size. Long-term studies were needed to establish the fact that some species of Darwin’s finches do indeed interbreed, though rarely, all previous evidence of hybridization being ambiguous. Gene exchange between species was confirmed by identifying the parents with microsatellite DNA. The study on Daphne yielded the unexpected result that when G. fortis and G. fuliginosa hybridize they do so without apparent regard to their respective beak sizes. This does not contradict the byproduct hypothesis but shows it is insufficient. Something else is involved in mate choice that usually leads to conspecific pairing but occasionally gives rise to hybridization.
That something else is song. The initial argument for the importance of song was developed by Robert Bowman, with rich support from analyses of song and sound transmission profiles. He showed that some features of song (pitch) varied among species as a function of their body size, and others varied as a function of the sound transmitting properties of the habitats they occupied. We later demonstrated with song playback experiments, in the absence of any morphological cues, that individuals could discriminate between their own species song and the song of related species. Thus two sets of experiments show that song and beak morphology are used as cues in mate choice. Why, then, do species sometimes hybridize? The answer lies partly in how finches acquire their songs and other information that they use later in choosing a mate.
A small number of laboratory experiments, conducted by Bowman, with finches exposed to tape-recorded song early in life showed that songs are learned between the ages of 10 and approximately 30 days after hatching by an imprintinglike process. Only one song is sung, by males and not females, and once learned it remains constant for life. These results combined with pedigree analyses on both Daphne and Genovesa demonstrated that song is typically transmitted from father to son. Nevertheless, the normal cultural transmission process can be perturbed by the misimprinting of young birds on the song of another species. Insights into the causes of perturbations have been gleaned over a very long time on Daphne, with difficulty because it is rare and because the circumstances are idiosyncratic. For example in one case an aggressive pair of G. scandens displaced the G. fortis owners of a nest and raised the offspring that hatched from one G. fortis egg together with their own offspring. The fostered G. fortis male later sang a G. scandens song.
A key insight into how song and beak traits are used in mate choice emerged from the pedigrees: paternal song is the primary cue for both males and females in hybridizing and backcrossing pairs. Misimprinted birds mate according to song type rather than morphology. The hybrid offspring breed with members of the species that sing the same song type as their fathers. Beak traits become the primary cue only when there is a large difference in beak size between the hybrid or backcross individual and the potential mates of the species that sing its song. For example, G. fortis individuals that have misimprinted on the song of the much larger G. magnirostris on Daphne have not hybridized. A parsimonious explanation for the development of mate preferences is that song is learned from parents and other conspecific individuals in association with morphological traits, primarily beak size and shape.
A Model of Adaptation and Chance
Speciation begins in allopatry as a result of local adaptation in genetically transmitted beak traits that also function in mate choice. Culturally transmitted song traits also diverge. Some song features such as frequency vary in relation to body size, and others, such as note repetition rate, vary in relation to beak size and jaw musculature possibly independent of body size. Thus several facts about Darwin’s finches are consistent with the adaptive byproduct hypothesis of Dobzhansky. Yet there are two reasons for thinking that a more comprehensive model of speciation is needed.
First, there is a need to make explicit the role of learning in the development of mate preferences; mate preferences are not inherent or “genetically fixed”. One consequence of learning is that species can sing each other’s songs, within limits, despite beak shape differences, and base their choice of mates largely upon learned song. Thus a crucial question is just what features of song are used in mate choice, and how do these diverge? Whatever causes songs to diverge in allopatry promotes speciation.
Second, there is a need to explicitly incorporate the role of chance because it is an important factor in song divergence. Change in the composition of individual songs from one generation to the next probably occurs through copying errors, as we have seen manifested on Daphne in minor ways. The most egregious example was an unusual song variant in the G. scandens population apparently caused by a cactus spine in its throat! Its sons learned and sang the same strange song, and at least two generations later there were five males on the island singing this song. Frequencies of song variants are subject to random change across generations due to some males leaving more sons than others, much as the frequencies of selectively neutral alleles drift at random. We have documented this on Daphne, in both G. fortis and G. scandens populations, as well as in the G. magnirostris population in the first few years following colonization of the island.
The importance of chance in speciation is underlined by the fact that populations of the same species on adjacent islands with nearly identical habitats may sing different songs. The strongest example is the sharp-beaked ground finch, Geospiza difficilis, on the northern islands of Wolf and Darwin. There are no habitat or morphological differences between the populations that would help to explain the differences in song, and we are left with randomness in the colonization of the two islands, or random cultural drift in song characteristics akin to random genetic drift, as explanations. Playback experiments on one or both islands are needed to test if the birds on one island would respond to songs from the other. The differences are so large, it seems highly likely they would ignore each other, because G. difficilis individuals on a third island, Genovesa, do distinguish between songs from Wolf and their own, similar, songs. In this case subtle, not profound, differences between Genovesa and Wolf songs, probably in temporal patterning, set up a potential reproductive barrier between the two populations.
Time-dependent Molecular Change
Another possibility is that populations stay isolated on different islands long enough for reproductive barriers to arise simply as a result of the accumulation of different alleles through mutation. Then, at the extreme, when members of the two populations encounter each other they are unable to produce viable and fertile offspring, even if they choose to mate with each other, because the species are genetically incompatible: during development of the offspring adverse interactions occur between the maternal and paternal genomes.
Genetic incompatibilities have not been detected between any of the four hybridizing pairs of ground finch species on Daphne and Genovesa. This statement applies to both survival and reproductive aspects of hybrid and backcross fitness. In following the fates of hybrids and backcrosses to determine where they might be at an expected disadvantage in relation to the parental species we gained the important insight that their fitness is not fixed but is environment-dependent: just as environmental variation alters the fitness of inbred birds so does it influence the fitness of hybrids and backcrosses.
From 1973 to 1982 the few hybrids that were produced by G. fortis breeding with either G. fuliginosa or G. scandens on Daphne did not survive to breed (nor did many of the non-hybrids). At this time we thought they survived poorly because of genetic incompatibilities. The successful breeding of hybrids and backcrosses from 1983 onwards showed this was wrong. Instead their fates were governed by the food supply. Before 1983 the large, hard Opuntia echios seeds and woody Tribulus cistoides fruits dominated the food supply. Hybrids were unable to crack open Tribulus fruits to reach the seeds and took twice as long to open Opuntia seeds as G. scandens. The switch to an abundant and continued production of small soft seeds after the 1983 El Niño produced a food supply appropriate for hybrids and backcrosses with their intermediate beak sizes. Since 1983 the survival of hybrids and backcrosses has been equal to or even greater than survival of the pure species.
These findings show that in the course of speciation learned behavior is of paramount importance in the establishment of a barrier to interbreeding. Only later do the species accumulate so much genetic change that they are incapable of forming viable and fertile offspring through interbreeding should they attempt to do so. This later point marks the unquestionable completion of speciation. Many if not all coexisting populations of Darwin’s finches have not yet reached this point, although they function as species by remaining distinct even in the face of occasional gene exchange.
Maintaining the Potential for Evolutionary Change
In the last million years or more oscillations between glacial and inter-glacial conditions at temperate latitudes have caused sea level to rise and fall. Small islands have been repeatedly created and submerged in the Galápagos archipelago. Their populations of finches could have been the starting point for species formation through selection and hybridization. Contemporary study of finch populations on a small island such as Daphne can throw light on these dynamics, and in particular on the question of how quantitative genetic variation is maintained.
Genetic variation is lost through drift and selection and regenerated by mutation. In principle rates of losses and gains could be equal, yielding an equilibrial level of genetic variation maintained for a long time. At the outset this simple scheme seemed insufficient to explain the high levels of phenotypic and presumed genetic variation in quantitative traits in many Darwin’s finch populations. Movement of birds from one island to another, resulting in a high rate of gene exchange, is a possible resolution of this problem, but if this is correct it raises another one: why, in the face of extensive gene flow, do the means of traits of different populations differ as much as they do?
An alternative to intraspecific gene flow is gene flow between species. Results of the long-term studies on Daphne and Genovesa showed that rare hybridization events (<1%) and ensuing gene exchange through backcrossing does increase genetic variation, thereby linking the evolutionary fate of one species to another.
The interdependence fluctuates as the direction of backcrossing changes. On Daphne it has changed from a predominant flow of genes from G. scandens to G. fortis to a predominant flow in the opposite direction. Introgression has not been balanced by loss of genetic variation through oscillating directional selection, instead in the last two decades genetic input into the G. scandens population has increased, and most individuals now (2005) have the genetic signature of backcrosses. Thus G. fortis and G. scandens are currently converging phenotypically and genetically, but are kept apart by song. This remarkable state of affairs has occurred through high hybrid and backcross survival rather than through an increase in hybridization.
Implications of Hybridization
Exchange of genes through hybridization has interesting implications in the context of speciation. For example, it implies a dynamic tension between ecologically differentiated species that are derived from a common ancestor. This is quite different from the normal view of speciation as a steady increase in differences between populations leading eventually to the complete cessation of interbreeding.
Events on Daphne illustrate how the process of divergence can be put into reverse under certain environmental conditions. The outcome, if the process continues, is difficult to predict. We can think of four alternatives. The species could remain as they currently are, maintained as separate species by the distinctness of their songs. They could eventually fuse into a single panmictic and bilingual population as a result of an increase in frequency of hybridization each generation due to their increasing morphological similarity. Then again the environment could revert to pre-1983 conditions, in which case gene exchange could diminish, the fitness of hybrids might decline under the changed ecological conditions, the two species would diverge, returning to their respective morphological states in 1973 when we began our study.
If the environment changes, a fourth possibility is a new evolutionary trajectory guided by natural selection and facilitated by hybridization. Hybridization and backcrossing play two facilitatory roles. As well as increasing variation, on which agents of selection can act, hybridization of species with different allometries such as G. fortis and G. scandens weakens genetic correlations between traits. This has the effect of reducing genetic constraints on novel directions of evolutionary change.
We believe these effects of hybridization could be general. Small, transitory, islands like Daphne with unique ecologies might have been arenas for introgression that facilitated novel directions of evolution and contributed to the adaptive radiation. We know that hybridization is not restricted to Daphne and Genovesa. Throughout the archipelago species are often more similar to each other genetically on the same island than on different islands, which is indicative of widespread, low-level, hybridization since species on the same island can hybridize whereas those on different islands cannot. Hybridization may have been going on episodically ever since the adaptive diversification of the ancestral finch species gave rise to two populations living in the same environment. Indeed, hybridization might have been important in the early stages of adaptive radiations in other organisms in other environments around the globe.
The Value of Long-term Study of a Single System
Continuous, annually repeated, study can potentially yield more insight into the causes of change as well as more precise measures of change itself. We have gained benefits that would have eluded us if the study had been intermittent as opposed to annually continuous. Perhaps the most important finding is that long-term effects of rare but strong events can have long-lasting consequences for the environment and for the organisms that exploit it. Short-term or intermittent studies are likely to miss such events and not be able to interpret the consequences.
The best example of this phenomenon is the critical El Niño event of 1982-83 on Daphne. If we had stopped before that year we could have concluded, incorrectly, that droughts invariably select for large body and beak size, hybrids do not breed, and the island cannot support another species (e.g. G. magnirostris). If we had started after 1983 we would have been mystified about the origins of the current interbreeding and convergence of G. fortis and G. scandens, we would have missed the colonization by G. magnirostris and would have no reason to believe that it had not always been a member of the community of breeding finches. Our appreciation of the ecological and evolutionary importance of annual environmental variation would have been much diminished.
We conclude by emphasizing that our study, like other long-term field studies, did not progress with a steady accumulation of knowledge and understanding, on an undeviating linear trajectory. Expectations at the outset were conditioned by our understanding from reading the literature, and almost inevitably surprises were encountered on the way. El Niños and droughts were far more intense than we expected, and evolutionary change was far more rapid. We gradually became aware of the importance of song and the interaction of learned cultural and genetical evolution, a change in the fate of hybrids (and the implications for speciation), the fact that changes in vegetation as a result of changes in climate could have evolutionary consequences as well as the more obvious ecological ones, and the fact that a reversal of weather did not result in a reversal of vegetation characteristics because there is an inertia to the system. The importance of temporal scale shifted from annual to decadal. Some observations forced us to re-think theory, for example the success of inbred G. magnirostris. Finally, we had no basis for expecting that the populations would not be the same phenotypically and genetically at the end as they were at the beginning of the study. Ramifications of this last result are still being explored.
All of these findings have been made possible by protection and conservation of the unique biota of the Galápagos by the Galápagos National Parks Service. These efforts, occasionally opposed strenuously by commercial interests, have been aided by the Charles Darwin Foundation and the Charles Darwin Research Station on Santa Cruz Island, Galápagos, and further supported by fund-raising organizations including the Charles Darwin Foundation Inc. and Friends of Galápagos in Switzerland, Britain, and other European countries. Further understanding of evolutionary processes in the Galápagos archipelago, in marine as well as terrestrial environments, depends crucially on the continuation of such conservation efforts. By recognizing the value of research into the population biology and evolution of Darwin’s finches the Balzan Prize Committee will give a boost to these efforts and, we hope, to similar efforts to conserve natural systems more broadly around the globe. For this we are deeply grateful.