Identify 5 examples of coevolution and explain how they do this.

Types of Coevolution

A few different categories of coevolution are often discussed by scientists in ecology and evolutionary biology: pairwise coevolution, diffuse coevolution, and gene-for-gene coevolution. Pairwise coevolution (or ‘specific’ coevolution) describes tight coevolutionary relationships between two species. Diffuse coevolution (or ‘guild’ coevolution) refers to reciprocal evolutionary responses between suites of species. This type of coevolution emphasizes that most species experience a complex suite of selective pressures derived from numerous other species, and their evolutionary responses change the selective environment for other species. Gene-for-gene coevolution (or ‘matching gene’ coevolution) describes the specific case where coevolution involves gene-for-gene correspondence among species, such as when hosts and parasites have complementary genes for resistance and virulence.

Concepts of Coevolution

Coevolution refers to several processes. One possible form of coevolution is cospeciation, the coordinated branching (speciation) of interacting species (such as host and parasite). To the extent that this has occurred, concordant (or matching) phylogenies of host and parasite clades (or evolutionary lines) would be expected. Cospeciation might be caused by the interaction between species, but it could also result from a joint history of geographic isolation, assuming that divergence and reproductive isolation evolve at similar rates in the two groups. Concordance of the two phylogenies implies a longer history of association, and of opportunity for reciprocal adaptation, than, for example, when parasites or symbionts have frequently switched from one host to another. Host switching can be inferred from certain patterns of discordance between host and symbiont phylogenies. Both cospeciation and host switching have been revealed in herbivorous insects, symbiotic bacteria, and parasites. For example, lice associated with gophers and with certain seabirds appear to have cospeciated to a considerable extent, and endosymbiotic, mutualistic bacteria (Buchnera) display almost complete phylogenetic concordance with their aphid hosts, from the family level down through relationships among conspecific populations.

In its most frequent usage, coevolution refers to genetic changes in the characteristics of interacting species resulting from natural selection imposed by each on the other—that is, reciprocal adaptation of lineages to each other. Such changes are referred to as specific or pairwise coevolution if the evolutionary responses of two species to each other have no impact on their interactions with other species. Diffuse or guild coevolution occurs when the genetic change in at least one species affects its interaction with two or more other species. For example, cucumber genotypes with high levels of the chemical cucurbitacin have enhanced resistance to mites but also enhanced attractiveness to cucumber beetles; this is an instance of a negative genetic correlation in resistance. Early season attack by flea beetles makes sumac plants more susceptible to stem-boring cerambycid beetles, and so resistance to the former would also reduce the impact of the latter.

In one of the seminal papers on coevolution, Ehrlich and Raven postulated in 1964 what has since been named “escape and radiate” coevolution—a process in which evolutionary changes temporarily reduce or eliminate the ecological interactions between species. Applying this concept to plants and herbivorous insects, Ehrlich and Raven postulated that in response to selection by herbivores, a plant species may evolve new defenses that enable it to escape herbivory and to flourish so well that it gives rise to a clade of descendant species with similar defenses. At some later time, one or more species of herbivores adapt to the defenses and give rise to an adaptive radiation of species that feed on the plant clade. In this scenario, the evolutionary diversification of both herbivores and plants is enhanced by their interactions.

Despite a common misconception, coevolution need not promote stable coexistence of species, and it certainly need not enhance mutual harmony. For example, parasites may evolve to become more virulent or less, depending on their life history. The Darwinian fitness of a genotype of parasite is measured by the average reproductive success of an individual of that genotype. Extracting more resource from a host, thereby reducing its chance of survival, often enhances the parasite's reproductive success, as long as the parasite individual, or its offspring, can escape to new hosts before the current host dies. Evolution of the parasite, by individual selection, may result in such high virulence that the prey or host population is extinguished. Extinction of prey populations does not alter the relative fitnesses of individual parasite genotypes and so does not select for reduced virulence. However, group selection may favor lower virulence or proficiency. If populations of more virulent parasites suffer higher extinction rates than less virulent populations, the species as a whole might evolve lower virulence. Although individual selection is likely to be stronger than group selection in most species, the population structure of some parasites may provide an opportunity for group selection to affect their evolution.

Sources of Evidence for Coevolution

The study of coevolution comprises the same approaches as studying evolution in general. As in the broader field, the first forms of evidence for coevolution consisted of detailed natural history observations, descriptions of the diversity of adaptive structures that mediate ecological interactions, and comparison among populations and species.

Charles Darwin published the first comprehensive illustration of how intricately species are adapted to one another and how structural complexity can be explained by coevolution in his description The Various Contrivances By Which Orchids Are Fertilized By Insects (1877). By comparing the shape of different orchid flowers and their associated pollinators, Darwin demonstrated that certain moth features are needed to successfully obtain nectar from the flower, features that are indeed exhibited by their specialized pollinators. By restricting nectar collection to a few pollinators, orchids increase the likelihood of cross-fertilization. Such is the correspondence between flower and pollinator shape that on observing the 29-cm-long nectar-bearing spur of the Madagascan orchid Angraecum sesquipedale, Darwin predicted the existence of a pollinating moth with a proboscis of that length. Such a moth, Xanthopan morgani praedicta, was indeed discovered 40 years later.

The description of patterns in plant use by lepidopteran larvae preceded the concept of escape-and-radiate coevolution. Higher taxa of butterflies often feed upon a single group of flowering plants. While some feed on more than one plant family, these tend to be closely related or have similar biochemistries. For instance, larvae of the butterfly subfamily Pierinae, or whites, feed predominantly on the families Capparaceae and Brassicaceae, which are closely related. Some whites also feed on members of the family Tropaeolaceae that share with the other families the production of mustard oil glycosides and a rare fatty acid. These regularities imply an important role for plant secondary metabolites in determining butterfly host use. Given that these compounds affect herbivore behavior, acting often as deterrents, secondary chemistry may have constituted the key feature that allowed plant escape.

Comparisons among conspecific populations have also been suggestive of coevolution. The coloration pattern of the butterfly Heliconius erato, thought to be a signal to predators indicating distastefulness, varies among populations in Central and South America. Strikingly, the wing coloration of H. melomene, an equally distasteful congener with distinct life history and host preference, varies geographically in parallel with H. erato. This pattern is thought to be an example of coevolution of mimicry between prey species that share a predator. Fritz Müller, a contemporary of Darwin, first suggested this particular model of coevolution to explain similarities in wing pattern among unpalatable butterfly species belonging to two distinct genera (Ituna and Thyridia).

Müller also introduced the use of mathematical models to study the coevolutionary process. Modern mathematical and computer simulation models may incorporate population genetics, quantitative genetics, evolutionary game theory, and optimality theory. Mathematical modeling has proven useful in describing the dynamic of the interactions between species and in determining which conditions favor coevolution.

Although ecological interactions usually do not “fossilize,” the analysis of paleontological records has provided some evidence of coevolution. For instance, the appearance in the Ordovician of predaceous cephalopods is associated with the simultaneous appearance of several defensive strategies on the part of their prey (e.g., strong sculpture and coiling in gastropods and shell-bearing cephalopods, spines in echinoderms), suggestive of diffuse coevolution between predators and their prey. The antiquity of certain interactions may also be determined by inspecting fossils of extant species. Several plant families possess structures (domatia) that harbor mites, which attack plant enemies. Domatia similar to the modern form have been discovered in fossilized leaves from the Eocene, 55 million years ago!

The relative age of clades of associated taxa is relevant for demonstrating correlated coevolution or co-speciation. These processes would be necessarily excluded if one group were much older than the other. The age of an association, or of interaction-related adaptations, can often be estimated from phylogenies with time calibration (e.g., using approximate molecular clocks or stem-group fossils). Molecular evidence from deep-sea vesicomyid clams and the sulfur-oxidizing endosymbiotic bacteria on which they depend for nourishment indicates that the interacting clades are both approximately 100 million years old. These two lineages appear to have been in close association since their origin and to have cospeciated, as indicated by the remarkable level of congruence between their estimated phylogenies. Phylogenetic information also becomes relevant in testing whether a character is an adaptation for an ecological interaction or an ancestral feature that exists in the absence of the interaction.

In some instances it has been possible to document the particular genes that affect a species' interaction. H.H. Flor found several genes in flax (Linum usitatissimum) that provide resistance to the rust Melampsora lini. Rust virulence is determined by a set of complementary genes, in a one-to-one relationship. This study inspired the gene-for-gene model (see Special Features in Parasite-Host Coevolution: Gene-for-Gene Systems), which has become a paradigm of phytopathology. Most traits, however, have a complex genetic basis, involving many genes. Such complexity requires a quantitative genetic approach, which partitions the trait variation into genetic and environmental components. This approach has demonstrated that many of the traits relevant to interactions have genetic variability, that is, there is potential for coevolution. For example, the wild parsnip (Pastinaca sativa) and its most important associated herbivore, the parsnip webworm (Depressaria pastinacella), are thought to be engaged in coevolution mediated by the evolution of furanocoumarins and the insect's detoxifying mechanisms. May R. Berenbaum has documented genetic variation both in the production of furanocoumarins and in the webworm's ability to metabolize this group of plant toxins.

Quantitative genetics is also used to measure correlations between traits. The detection of negative genetic correlations is indicative of trade-offs between traits, such that selection for the increase in value of one trait leads to a decrease in value of the correlated trait. Tradeoffs are of particular relevance in explaining evolutionary constraints, and particularly why species are specialized. Pea aphid (Acyrthosiphon pisum) clones collected from two crop plants (alfalfa and red clover) exhibited higher fitness when reared on the plant from which they had been collected, suggesting local adaptation. The negative genetic correlation in fitness across crops may constrain the evolution of generalist clones, as these would be outcompeted on either plant by crop-specialized clones.

Measurement of genetic variation and correlation between traits offers information on the genetic context in which selection can act. Correlations between traits and fitness suggest the form and direction of selection. In a greenhouse study, the wild parsnip exhibited a negative genetic correlation between concentration of several furanocoumarins and seed set, suggesting that the production of the chemicals may impose a cost to reproduction in the absence of the parsnip webworm. These negative correlations were not detected in the field, indicating that presence of furanocoumarins increases fitness in the presence of the herbivore. Ideally such studies are performed in a natural setting, as our ultimate interest is understanding how natural selection works in the wild, but one can use model systems in the laboratory, such as evolving populations of bacteria and bacteriophage.

Finally, studies of interacting species are commonly based on the analysis of single communities. However, most species are composed of many local populations, and increasing importance has been attributed to the geographical structure of species and their interactions. Across the distribution of an interaction one is likely to observe a mosaic of selection pressures as a result of variation in abiotic and biotic factors, and the particular demographic and genetic histories of local populations. Some localities may be coevolutionary hot spots, that is, sites of reciprocal coevolution, whereas in others selection may be unidirectional or act on neither species. The geographic variation in outcomes is further modified by gene flow among populations. Consequently, different degrees of coadaptation are to be expected among populations. Clearly our understanding of the dynamics of a species interaction requires the study of many communities and interpopulation processes.

A few cases that satisfy the requirements of long-term multipopulation studies have emerged recently and have reinforced our need for a geographic mosaic theory of coevolution. For instance, resistance and virulence structures of Linum and Melampsora (referred to earlier), studied in New South Wales, sometimes vary dramatically across populations and time. The frequency of susceptible genotypes of flax will affect the local frequency of a particular strain of flax rust, but additional factors were found to play a role, namely, drift, extinction, and migration from neighboring populations. The geographic structure of flax and rust proved to be an essential factor in explaining the persistence of the interaction.

Methods of Studying Coevolution

The methods of studying coevolution correspond to those of studying evolution generally. Long-term, macroevolutionary patterns of coevolution, for example, are analyzed by paleobiologists and by the use of phylogenetic studies of extant species. Phylogenies can indicate whether related parasites have speciated and diverged in concert with their hosts or have shifted among host species, in a process analogous to colonizing new geographic areas. They can indicate whether or not a repeatedly evolved feature that affects ecological interactions, such as a defense against parasites, has been consistently associated with increased diversification of species. By plotting characters such as defenses on a phylogeny based on other data, patterns in the evolution of the characters can be discerned. Phylogenetic information can be important for demonstrating that a character is an adaptation for an ecological interaction, rather than a widely shared primitive character that happens to confer a benefit in a novel ecological context.

Mathematical and computer models of processes of coevolution within populations and species play an important role in studying coevolution. Such models are based on population genetics, quantitative genetics, or optimality theory. Some models couple genetic dynamics to population dynamics, based on assumptions about how the outcomes of interactions between individuals with specified phenotypes will affect demography. As more such realism is introduced, the dynamics and possible outcomes are often found to become increasingly complex and dependent on initial conditions.

Many empirical studies document changes in features that mediate ecological interactions (e.g., size of bills, teeth, or other trophic structures) by comparing features of related species or conspecific populations, or by characterizing rapid changes in populations that have been moved to new regions by humans and have engaged in new interactions. Some comparisons test predictions from coevolutionary models; others test important assumptions of the models, such as the importance of costs of adaptation. Finally, coevolution can be studied directly in model systems in the laboratory, such as rapidly evolving populations of bacteria and bacteriophage.

Traditionally, research on coevolution has had a strong ecological and phenotypic emphasis, with relatively little analysis of genetic, biochemical, and cellular mechanisms. Contemporary research increasingly includes the study of variation and evolutionary change in interspecific interactions, especially between hosts and their pathogens and parasites, at molecular and genomic levels.

Introduction

Coevolution, or coevolution, is the reciprocal evolutionary change in a set of interacting populations over time resulting from the interactions between those populations. Usually, the interacting populations are different species, like plant–pollinator, predator–prey, or host–parasite. Some of the first scientific papers to use the term coevolution in the middle of the twentieth century studied these systems. In 1964, Erlich and Raven published a now famous study of the coevolution of butterflies and their interactions with flowers, but even earlier, in 1958, Mode published his mathematical model on the coevolution of obligate parasites and their hosts.

Note that the term ‘coevolution’ has a rather broad scope. For instance, at the global scale, one may refer to the coevolution between humankind and the biosphere. However, in evolutionary ecology, we are downscaling to the level of ecosystems, biological communities, and populations. Occasionally, the same coevolutionary terminology is extended to the discussion of reciprocal coevolutionary change between males and females of the same species as a result of sexual selection. In all cases, coevolution implies evolutionary changes resulting from interactions, so one must first understand the major types of ecological interactions, primarily mutualism and antagonism, and their potential coevolutionary results. This article provides an overview of coevolution as the result of ecological interactions. The major evolutionary results of mutualism, antagonism, and mixed ecological interactions are each discussed using archetypal examples and current mathematical theory. Finally, potential confounding factors in the research of coevolution are discussed.

Abstract

Coevolution plays a key role in shaping the biodiversity on Earth. Coevolution is commonly defined as reciprocal evolutionary changes brought about by interactions between species, implying that interacting species impose selection on each other. The covariance between fitness and trait value determines the strength of natural selection. Thus, to understand coevolution, it is necessary to understand fitness both within and between species. A large amount of research has focused on fitness consequences resulting from interacting with other species (e.g., antagonisms, mutualisms). The geographic mosaic theory places coevolutionary fitness outcomes in a broader landscape of both spatial and environmental variability.

Abstract

Co-evolution occurs when one species evolves in response to evolutionary changes in another, the result being an evolutionary feedback involving two or more species.

It is an important and ubiquitous process in nature that can occur any time populations of different species interact through evolutionary time, with foraging behaviour involved whenever interactions include an animal species.Rather than continuing indefinitely, co-evolution is expected to reach an equilibrium, with no further evolutionary change in the interacting species, because of ‘trade-offs’ experienced by individuals of the species involved.Co-evolution can therefore be thought of as an ‘evolutionary game’, with various different species as ‘players’, that reaches an Evolutionarily Stable Strategy (ESS) at equilibrium, such that mutants for each interacting species, deviating slightly from average members of population, are selected against.It is thus possible to develop mathematical models of co-evolution as evolutionary games between species, with ESS’s as the hypothesised end result, and therefore predicted to agree with our observations. But this is unlikely to be easy! Because foraging behaviour is generally involved, but foraging decisions are difficult to determine directly, Optimal Foraging Theory (OFT) will frequently be necessary as part of model development. Developing such models will also likely be complex, given the number of component relationships involved and the need to evaluate them quantitatively, and little development of such models has so far occurred.As an example of how mathematical models of co-evolution can be developed, with foraging behaviour included, I consider co-evolution between plants and their pollinators in terms of floral nectar production (in this case by the plant commonly known as scarlet gilia – Ipomopsis aggregata) and its utilisation by nectar-feeding birds (in this case hummingbirds). This example illustrates how it is possible to deal with the inherent complexity of co-evolution and determine an ESS for floral nectar production and hummingbird foraging behaviour. It also illustrates how, in the absence of appropriate models of co-evolution, explanations for observed patterns are often unsatisfactory. Furthermore, in this example, good agreement was obtained between observations and the predicted ESS!This example therefore illustrates the potential for development of mathematical models of co-evolution, leading to quantitative comparisons of observations and predictions, and to thus better understand both the process of co-evolution and its outcomes, while demonstrating the importance of foraging behaviour. Hopefully, this potential will continue to be realised.

Introduction

Coevolution is the reciprocal adaptation among interacting organisms, and it is one of the major forces that organize biodiversity by linking the genomes of interacting species (Thompson and Cunningham, 2002). It is also one of the major forces creating biodiversity because diversifying coevolutionary selection can lead to speciation (Thompson, 2005, 2009). Measuring coevolution is difficult because it involves determining how two species interact through time, and without a time machine it is difficult to measure how traits of interacting species have changed through time. The difficulties in rigorously testing coevolutionary hypotheses led to questions about its importance as an evolutionary force. To solve the problems in testing the coevolutionary hypothesis, Thompson (1999, 2005) developed the geographic mosaic theory of coevolution that examines geographic variation in species interactions to use space as a proxy for time to measure coevolution. A geographic mosaic of coevolution (GMC) occurs when local variation in species interactions produces a spatially variable pattern of reciprocal adaptation in these species.

Coevolution involves at least three steps: first one species evolves a response to a trait of a second species, and this is followed by a response of the second species to the first (Janzen, 1980). These steps can be measured by comparing sites where two species are exerting reciprocal selection on each other that are termed coevolutionary hotspots, and sites where two species are not exerting reciprocal selection on each other that are termed coevolutionary coldspots (Figure 1). If coevolution is occurring, then it can be measured by comparing traits in coevolutionary hotspots versus coldspots. Interactions between species may also differ in different environments, with those environments changing the selection on coevolved traits, so coevolution can also be measured by comparing hotspots in different environments (Figure 1). The predicted outcome of geographic mosaics of coevolution is that populations of the interacting species should become geographically differentiated and locally adapted due to the presence or absence of diversifying selection, or by different strengths of selection between interacting species in two hotspots. However, if movement of individuals causes gene flow between populations adapted to environments with different coevolutionary forces then this can result in ‘trait remixing’ and ‘maladaptation’ (Figure 1). For example, if there is migration from a coldspot where the population is not adapted to an interacting species to a hotspot where the interacting species is present then non-coevolved individuals with low fitness may be found in the hotspot.

The geographic mixture of coldspots and hotspots produces a geographic selection mosaic, which consists of a matrix of interactions where the genotypes of one species interact with the genotypes of a second species in a specific environment, or a G×G×E interaction. The G×G interaction that determines coevolution varies geographically from hotspots (Figure 1), where species frequently interact resulting in strong reciprocal selection, to coldspots, where interactions are rare or absent and reciprocal selection is weak (Figure 1). Geographic variation in the environment can alter the G×G×E interaction so that selection differs among hotspots (Figure 1). The evolutionary outcomes produced by the divergent geographic selection mosaics can range from transient polymorphisms to speciation and adaptive radiation.

The pervasive patterns of geographic variation in selection and local adaptation that are found among natural populations (see Kawecki and Ebert, 2004; Greischar and Koskella, 2008; Hokesema and Forde, 2008 for reviews) are consistent with the geographic mosaic of coevolution theory. However, to demonstrate that geographic population variation between interacting species is due to their reciprocal selection requires separating selection due to genotype×genotype interaction from those of a myriad of other species and abiotic factors represented by the E in the mosaic of G×G×E interactions. To completely describe a GMC it is necessary to measure the G×G×E coevolution mosaic, gene flow among populations, and the mosaic of local adaptation and maladaptation that these processes produce.

One approach to testing hypotheses generated by the GMC theory is determining whether observed ecological patterns are consistent with the predictions of the GMC. Thompson (2005) has made three such predictions, and these predictions are illustrated by the interactions between a predatory fish and three species of freshwater snail (Chaves-Compos et al., 2011). First, Thompson predicts that geographic populations will differ in traits shaped by the interaction. In the fish–snail example, since the success of a fish that feeds on snails depends on snail shell thickness relative to fish crushing ability, then the geographic variation found in these traits indicates that they are coevolving. Thompson’s second prediction is that if these species are coevolving, then traits of interacting species will be matched in some, but not all, of the populations (Thompson et al., 2002). Chaves-Compos et al. (2011) found snail shell thickness and teeth morphology is matched in some areas as predicted if coevolution produced reciprocal local adaptation, but not matched in others because of gene flow. Finally, Thompson predicts that geographic variation in diversifying selection will result in few coevolutionary traits being fixed in all populations. In the snail–fish example, neither shell thickness nor fish tooth morphology is a fixed trait, but rather varies geographically.

A match with the ecological predictions is consistent with the existence of a GMC, but alternative hypotheses are also consistent with these patterns and additional studies are often needed to distinguish GMC from alternatives. Experimental measurements of the geographic mosaics of selection and local adaptation can provide rigorous tests of the GMC hypotheses. A variety of approaches have been used to measure and partition the effects components of G1×G2×Ex to measure coevolutionary forces. The GMC predictions that the outcome of interactions of species 1, G1, and species 2, G2 from different locations will differ due to coevolutionary selection for local adaptation. To test this hypothesis, ‘common garden experiments’ measure interactions between a range of populations of a species from different areas (i.e., EA, or EB) in a single environment (either EA, or EB or a laboratory or garden (EC)) in order to measure the effects of reciprocal selection by comparing the G1A×G2A, G1B×G2B, G1A×G2B interactions, in the absence of environmental variation (Lively, 1989; Lively and Jokela, 1996; Laine 2005). While these experiments can provide indications of geographic variation in the interaction of two species it cannot accurately measure coevolutionary forces because the crucial G×E interactions are not accounted for.

‘Reciprocal transplant experiments’ measure local adaptation of populations, a crucial prediction of the GMC theory, by comparing the performance of each population in their local environment and foreign environments, for example, the performance of the population G1A is measured in the local G1A×G2A×EA and in the ‘foreign’ G1A×G2B×EB environments. Local adaptation is measured by either comparing whether a local population performs better within its own habitat than a foreign population does or whether a local population performs better in its local habitat than in a foreign habitat (Gandon, 1998; Kawecki and Ebert, 2004; Greischar and Koskella, 2008). However, merely demonstrating that populations are locally adapted does not demonstrate that coevolution has occurred between two focal species, because each species could be locally adapted to the abiotic environment or to other non-focal species. To measure coevolutionary forces, Nuismer and Gandon (2008) have developed an experimental design and mathematical analysis that allows full measurement of the G×G×E coevolutionary forces. It requires a double reciprocal transplant experiment where multiple geographic populations interact in multiple environments.

Concluding Remarks

Co-evolution of certain hosts and pathogens for millions of years has resulted in a fine-tuned equilibrium that enables survival of both. Antigen presentation is one of the critical elements in this balance. While antigen presentation is an essential process for long-term effective host defense, targeting APCs represents a common maneuver of many viruses to avoid host surveillance and establish a chronic or persistent infection. A major challenge in biomedical research is to thwart microbial APC subversion to promote eradication of the pathogen. A better understanding of the mechanisms used by APC to display microbial antigens as well as the virus strategies to subvert APC functions during immune responses will provide new tools for designing novel vaccination approaches and immunotherapeutic treatments for human infectious diseases.

The Importance of Coevolution between Hosts and Pathogens

Coevolution can be defined as evolutionary changes that occur within two or more organisms as a response to interactions between them, and the mutual selective pressures that those interactions cause (Van Blerkom, 2003). Haldane (1949) was one of the first scientists to emphasize the importance of infectious diseases as agents of selection within humans as well as other organisms. Although this work largely consisted of speculations derived from the work of others rather than discussions of his own experimental work, Haldane’s ideas have inspired much thinking on disease and evolution (Lederberg, 1999).

Haldane (1949) recognized that disease can be either a selective advantage or a disadvantage to a host when competing with other species. For example, some wild African ungulates have evolved the ability to tolerate nonlethal trypanosome infections and thus outcompete introduced cattle that die from the infections. More often the situation is one of disadvantages, as in the risk of predation in a species that is highly susceptible to a pathogen and has a high prevalence of infected individuals. Such a species may not survive and reproduce as well when faced with a predator, such as a cat, that chases after its prey. Haldane further pointed out that individuals who possess rare phenotypes may be more likely to be resistant to pathogens than individuals with common phenotypes, and therefore frequency-dependent natural selection in favor of rare types in an environment of parasitism may foster the evolution of new alleles (increased levels of genetic variation) in both hosts and pathogens. This idea is at the core of a prominent hypothesis in studies of the coevolution of host and pathogens, the Red Queen hypothesis.

Van Valen (1973) used Alice’s interactions with the Red Queen in Lewis Carroll’s Through the Looking Glass and What Alice Found There to illustrate the nature of all kinds of species interactions (Figure 1). In that story, Alice moves backward when she wants to go forward, and if she wants to go quickly, she comes to an abrupt stop. She wants to talk to the Red Queen, but in order to do so, she must walk away from her. Finally, at one point she and the Red Queen start running somewhere (although Alice is not quite sure where), and at the end of a long hard run, Alice observes that after all that running, they are still right where they started.

Van Valen used this story as a metaphor for the impact of species interactions on rates of extinction over long spans of time, but others have used the same story to explain coevolutionary interactions between hosts and pathogens (see, for example, Bell, 1982; Lively and Dybdahl, 2000; Lively, 2010). This latter version of the Red Queen hypotheses is also used to illustrate how host–pathogen interactions may help to explain the maintenance of sexual reproduction. Sexual reproduction provides increased opportunities for new (and rare) offspring phenotypes that pathogens would not have evolved to infect, which could allow more opportunity for hosts to escape the negative effects of the parasitism. It is important to note that changes in the genetic structure of hosts, however, often stimulate the pathogen to evolve new strategies to infect the host (Muehlenbein, 2010).

The Red Queen hypothesis for host–pathogen coevolution and the maintenance of sexual reproduction is attractive and logically compelling, but empirical evidence to support it is rare, largely because it is difficult to track oscillations in genotype and phenotype frequencies as well as linkage disequilibrium effects over time (Salathé et al., 2008). In addition, it has been questioned on theoretical grounds, but it is still being studied and remains one of the primary hypotheses related to host–pathogen coevolution.