Cheating Viruses and Game Theory

The theory of games can explain how viruses evolve when they compete against one another in a test of evolutionary fitness

Paul E. Turner

he 19th-century circus showman reptitious copulations. This strategy them easy subjects for manipulation TP. T. Barnum is reputed to have is very successful for maintaining a and study. Although the experiments coined the phrase “There’s a sucker subpopulation of sneakers, but it’s un- are conducted in the laboratory, evo- born every minute”—although Bar- likely that the population will evolve lution proceeds by natural selection num denied the saying was his, and to contain only cheaters because ter- because the laboratory habitat dictates it has been variously attributed by ritorial males are most attractive to fe- which genetic variants are favored to biographers. In any event, whoever male mates. contribute their genes to the next gen- did voice this cynical view of human In general, cheaters are highly suc- eration. This is very different from ar- gullibility could not have predicted cessful when they are rare because tificial selection, such as dog breeding, that the terms “cheaters” and “suck- they frequently encounter suckers. The where the experimenter determines ers” would describe individuals in benefits of cheating wane as more indi- the variants that will reproduce. Per- the world of microorganisms as well. viduals in the population opt to cheat. haps most important, microorganisms However, my colleagues and I have In the parlance of evolutionary biology, can be stored in a freezer indefinitely, been studying interactions between vi- the success of cheaters should be gov- creating a “fossil record” that permits ruses, and it seems that strategies for erned by frequency-dependent selection. direct comparisons between the genet- taking advantage of the other fellow That is, some cost should be associated ic makeup of an ancestral population are just another way for the viruses, with a cheating strategy so that selfish and that of its evolved descendants. too, to make a “living.” individuals are at an advantage when Our experiments suggest that yes, The temptation to cheat appears they are rare, but disadvantaged when perhaps at this moment, there may be to be a universal fact of life. In the they are common. cheaters among the viruses vying for struggle to survive and reproduce that Game theory is a useful approach survival within and near your own drives evolution, selfish individuals for mathematically predicting which cells. But in the long run, such crimes may be favored over cooperators be- strategy, if any, will dominate such don’t always pay. cause they are more energy efficient. a contest. Social scientists use game By definition, cheaters expend rela- theory to predict which behaviors will Viruses in Conflict tively little energy in a task because spread through a population, especial- We experience viruses mainly through they specialize in taking advantage of ly in contests involving classic strat- the symptoms, such as fever and fa- others—“suckers”—whose efforts they egies such as “hawk” versus “dove,” tigue, that signal that our body is de- co-opt to their own advantage. In cer- and “cooperator” versus “cheater.” One fending against an invader’s attempt tain animal species some males exert of the most intriguing results of this to commandeer the normal activities tremendous energy maintaining and approach is a mathematical proof dem- of the body’s cells. Viruses are para- defending territories to attract females. onstrating that cheating can take over sites that rely on the genetic machinery Meanwhile, the population may con- a population, even though deceit can of a host organism to make copies of tain subordinate “sneaker” males that be considered an irrational behavior themselves. At any particular moment are uninterested in territory but linger because it is punishable. an infected individual can harbor sev- at the boundaries and specialize in sur- My colleagues and I have applied eral species of viruses or even genetic game theory to the experimental evo- variants (genotypes) of the same spe- lution of viruses in the laboratory. This cies. So the host serves as an ecosystem Paul E. Turner is an assistant professor of ecology is a field that is relatively new, but is for potential interactions between the and evolutionary biology at Yale University. He re- proving to be powerful for testing fun- viruses. These interactions may be in- ceived his Ph.D. from Michigan State University in damental questions in evolutionary direct—for example, when the host’s 1995. He is interested in the ecology and evolution of infectious diseases and the use of laboratory popu- biology. It’s an approach with many immune system takes action against lations of microbes as models to address these topics. advantages: Viruses are easy to cul- one species of virus, while simultane- Address: Department of Ecology and Evolutionary ture. They have rapid generation times ously affecting other viruses. A host’s Biology, Yale University, P. O. Box 208106, New and large population sizes. In addi- fever may be a generalized response to Haven, CT 06520. Internet: [email protected] tion, an array of modern tools makes a specific infection, but the high tem-

© 2005 Sigma Xi, The Scientific Research Society. Reproduction 428 American Scientist, Volume 93 with permission only. Contact [email protected]. The Bridgeman Art Library Figure 1. Cheaters often win in the simulations used in game theory, a branch of mathematics that analyzes competitive interactions between individuals. Regrettably too common among human beings, cheating has also been found among many other animal species and, perhaps surprisingly, among viruses. The cheating viruses were discovered in laboratory experiments by the author. In Le Tricheur (“The Cheat”), the 17th-century French painter Georges de La Tour depicts a card game with at least one deceitful player. perature can impair the growth of all extremely difficult to detect. More of- plants, but they are unable to hook up viruses within the body. ten, viruses seem to experience a con- with the aphid vector. This situation When viruses interact directly, their flict of interest over the resource pool. changes if luteoviruses and umbra- effects on each other are more difficult When this happens one virus can self- viruses happen to co-infect the same to detect because they take place within ishly usurp resources to the detriment plant. The umbraviruses steal some of an individual cell. When a virus enters a of other virus species or genotypes. the capsid proteins from the luteovi- cell, it hijacks the host’s metabolism, in- In a form of complementation known rus resource pool, attach themselves structing it to make the bits and pieces as phenotypic mixing, a virus acquires to the aphids and so move on to new needed to assemble other viral particles. certain observable (phenotypic) traits host plants. Meanwhile, the hapless lu- When more than one virus infects a cell, from another virus. This phenomenon teoviruses experience a net loss in the the metabolic products are freely acces- frequently involves a conflict over pro- capsid proteins they need to assemble sible to any of the co-infecting viruses. teins used to create the viral capsid, a their progeny. In a process called complementation, one shell that protects the genetic material Complementation can also be a fac- virus provides a useful product that of the virus. Phenotypic mixing allows a tor in conflicts when one virus usurps cannot be made by another virus. If the virus to acquire capsid proteins from the an enzyme needed for replication from viruses provide each other with useful resource pool of a different virus. This another virus. The best known exam- resources, the interaction is of mutual is critical because some proteins on the ples involve an ordinary virus and a benefit. Consider the co-infection of a capsid dictate whether a virus can at- defective form, typically a virus with cell by two mutant viruses, which differ tach to a particular host cell and thrive. a “shortened” genome, one that lacks by having inactivated genes at different An interaction between two plant one or more essential genes. This phe- locations in the genome. The common viruses illustrates how this strategy nomenon was first described in labo- resource pool allows the viruses to use can be important in the transmission ratory experiments involving poliovi- each other’s protein products. The co- of a virus. Luteoviruses infect nearly ruses. When these viruses are grown infection rescues the mutants, allowing all the crops that people grow for food at high densities there is strong selec- them to reproduce when they couldn’t or fiber. These viruses can easily travel tion pressure for them to lose genes otherwise do so. between plants by hitching a ride with (become defective) because shortened Such mutually beneficial interac- the tiny plant-sucking insects called viruses replicate much faster than vi- tions between viruses are either rare or aphids. Umbraviruses also infect crop ruses with genomes of normal length.

© 2005 Sigma Xi, The Scientific Research Society. Reproduction www.americanscientist.org 2005 September–October 429 with permission only. Contact [email protected]. If the replication advantage of the defective-interfering viruses drives the helpers to extinction, both strains will die. Most virologists view defective- interfering viruses as an unfortunate nui- sance, one that may compromise their research goals—as when, for example, such a virus contaminates the purity of a commercial vaccine. To a microbial ecol- ogist, however, the defective-interfering viruses are especially intriguing because they are parasites on parasites, or hyper-parasites, something rarely seen elsewhere in biology. This naturally raises the question of whether defective-interfering viruses are merely laboratory artifacts. Some recent evidence suggests that a related phenomenon may not be uncommon outside the laboratory. Natural hyper- parasitism is seen among viruses that infect farm animals and crops. The majority of the defective viruses identified Figure 2. Cheating as an evolutionary strategy can be studied by measuring the reproductive from these systems are satellite viruses; success of animals that engage in covert copulations or fertilizations, such as these bluegill they are usually unrelated to their sunfish (Lepomis macrochirus). Here a male bluegill (right) gains access to a nesting female helpers. By contrast, defective-interfer- (center) by mimicking her appearance, and so cuckolds a so-called “parental” male (left) who courts the female and cares for the offspring. Both breeding tactics used by the males are suc- ing viruses have recognizable genetic cessful, and so neither can displace the other in the population. Cheating viruses use tactics similarities to the helpers from which very different from the mimicking male fish, but it is possible to conduct experiments that they evolved by losing certain genes. show how cheating affects their reproductive success and thus the viruses’ evolutionary fit- Defective-interfering viruses may be ness. (Image courtesy of Brian D. Neff, University of Western Ontario.) rare in nature because the helper viruses appear to evolve a resistance to The defective viruses interfere with the vantage in replication rate can result in being parasitized by closely related vi- reproductive success of the ordinary drastic differences in the relative suc- ruses. For some reason satellite viruses viruses by using their gene products, cess of the viruses. The problem faced are more easily able to circumvent any so that the ordinary viruses take on the by the “defective-interfering” viruses resistance put up by the helpers. role of helpers. Because viruses repro- is that they are entirely dependent on Defective viruses are not known duce exponentially, even a slight ad- helper viruses to provide key proteins. to play a widespread role in human disease. A notable exception is the surface hepatitis delta virus, a satellite virus proteins associated with its helper, the hepatitis-B virus. Together these two viruses cause unusually severe liver damage in cases of chronic active hepatitis. It is not clear why defective viruses are replication not commonly implicated in human disease, but it is conceivable that other examples have yet to be discovered by the medical community. Although hyperparasitic viruses common resource may evolve readily, they can easily virus A pool wind up as evolutionary dead ends be- virus B virus B cause of their strict reliance on helper with surface proteins from virus A viruses. It would be a daunting task to study the relative success of parasitic Figure 3. Viruses compete for resources when they co-infect the same cell. In some instances, viruses in nature because there are so one virus may provide a useful product for another virus, a phenomenon called complemen- many uncontrolled variables in field tation. In this hypothetical example, virus A carries a gene that codes for a valuable surface studies. However, virus interactions protein—for example, one that allows it to infect other cell types. Although virus B lacks the gene for this protein, it can steal the protein from the common resource pool inside the host can be examined under controlled lab- cell. Virus B gains the use of the protein, whereas the offspring of virus A now experience a oratory conditions in experiments that shortage of the protein. A form of complementation called phenotypic mixing takes place measure relative growth rates. This ap- when one virus acquires observable (phenotypic) traits from another virus, as shown here. proach can be augmented with math- Complementation may be involved when viruses cheat. ematical models that explore how eas-

© 2005 Sigma Xi, The Scientific Research Society. Reproduction 430 American Scientist, Volume 93 with permission only. Contact [email protected]. ily parasitic viruses arise and whether uncertain, reward can drive individuals they can persist. As it happens, game to behave in a way that is collectively cooperator cheater theorists have had a long-standing in- irrational: When both prisoners follow terest in the success of parasites and their individual interests, both lose. other cheaters, so there are many math- When the differing strategies are sucker’s reward ematical models to choose from. associated with different underlying payoff genetics, game theory can be applied cooperator Cheaters Sometimes Prosper to the study of evolution. Popularized To understand how game theory might by the late British biologist John May- be applied to virus interactions, consid- nard Smith, evolutionary game theory er a game involving cooperators and comes into play when an individu- cheaters, known as the “prisoner’s di- al’s reproductive success, or fitness, temptation punishment to cheat lemma.” This scenario, which has been is frequency dependent. Consider a cheater used to explore philosophical, politi- predator that preferentially feeds on cal, economic and biological questions the most common type of organism in for half a century, involves two felons a prey population because these indi- Figure 4. Payoff matrix for a contest between who are separately interrogated about viduals provide an easy “search im- a “cheater” and a “cooperator” shows the a crime they committed. In one ver- age.” Prey types with a rare appear- outcome of each pairwise interaction for one sion of the game, an understanding be- ance, perhaps sporting an uncommon individual (left side of matrix) who encoun- tween the prisoners is assumed: They fur color, will have a higher fitness ters another (top of matrix). A cooperator who can cooperate in denying the crime because they escape the predator’s meets another cooperator is rewarded, whereas a cooperator who meets a cheater receives the in the hope that they’ll both be let off notice. This advantage will wane as “sucker’s payoff,” typically a loss of some use- the hook. The interrogator, however, their type becomes more common in ful resource. A cheater gains this valuable re- offers two choices to each prisoner. If the population and therefore more ob- source when interacting with a cooperator and both stay silent (cooperate), each will vious to the predator. so is tempted to cheat. When a cheater meets receive a light one-year jail sentence. If Evolutionary game theory weighs another cheater, nothing is gained, and the two both confess, each goes to prison for 10 the costs and benefits in terms of fit- are usually punished in some way. The rela- years. However, if one confesses—in ness associated with different strate- tive success of cheaters and cooperators can be other words, cheats—the cooperator gies, and so predicts the evolutionary determined if the respective costs and benefits will receive a 20-year sentence while fate of the different types. This is done can be quantified for each of the interactions. the cheater goes free. So what’s a pris- by creating a 2×2 matrix that contains When these strategies are associated with different underlying genetics, the payoff matrix oner to do? all pairwise interactions between two can predict whether one tactic will displace Game theory holds that in such a di- different strategies. Each entry within another over the course of evolution. lemma, it always pays the individual the matrix consists of the fitness pay- to cheat because cheating, even though off to one strategist when it interacts a population evolves to contain only it risks a long sentence, offers the only with the other. The matrix reveals the individuals with a single strategy, it possibility of obtaining the best pay- relative success of the strategies in the is defined as an “evolutionary stable off—freedom. The result is intriguing contest so long as the mathematical strategy.” If two strategies are unable because it explains how a potential, but fitness values can be calculated. When to displace one another, as hinted at

RNA

capsid

surface proteins

Figure 5. Viral particles of the bacteriophage phi-6 (left, small circles), which grow on Pseudomonas phaseolicola bacteria (larger lozenge shapes) in the laboratory, are the subjects of experiments in evolutionary game theory. Each phi-6 particle (above) consists of the genetic material RNA, which is housed in a protein shell (or capsid) and a lipid mem- brane. The phi-6 bacteriophages may compete against one another for the capsid proteins in the common resource pool inside the bacterial host. (Electron micrograph courtesy of Dennis Bamford, University of Helsinki.)

© 2005 Sigma Xi, The Scientific Research Society. Reproduction www.americanscientist.org 2005 September–October 431 with permission only. Contact [email protected]. in the predator-prey example above, phaseolicola bacteria, which are easily both strategies will coexist indefinite- cultured on agar plates. By combining ly; this is defined as a “mixed evolu- the phage and bacterial populations in phi-6 tionary stable strategy.” different ratios it is simple to control bacteriophage In the prisoner’s dilemma, evolution- whether a virus will infect a bacterial ary game theory suggests that cheat- cell on its own or co-infect the same ers should take over the population— cell with other viruses. selfishness turns out to be the evolution- Chao and I created six laboratory pop- ary stable strategy. The result is strik- ulations of phi-6 growing on the P. phaseo- ing because it is somewhat counter to licola bacteria. Three populations were Darwin’s theory of evolution by natu- allowed to evolve at phage-host ratios ral selection. Darwinism holds that dif- that resulted in strictly single infections. mix viruses ferential performance allows the fittest The other three populations were grown with bacteria individuals to produce more offspring, at ratios that allowed co-infection with which steers the population to become an average of about two or three virus- better adapted to its environment over es entering each bacterium. We let the time. The prisoner’s dilemma indicates viruses grow for 50 days, which cor- high low that cheaters can successfully displace responds to about 250 generations of co-infection co-infection cooperators, while simultaneously low- phage evolution. By comparison, a simi- treatment treatment ering the average fitness of the popula- lar experiment using a human population. The prisoner’s dilemma is very tion would take 5,000 years (assuming easy to prove mathematically, but it about 20 years per generation). took laboratory experiments on viruses After 250 viral generations had to demonstrate that the strategy may elapsed, the evolved populations were allow take place in a biological population. each placed into an agar-plate “arena” 40 minutes for virus to compete against the revived ancestor attachment Experiments in Evolution that had been stored in the freezer. This My colleague Lin Chao, of the Uni- allowed us to gauge changes in viral versity of California, San Diego, and fitness, which we defined by measur- I designed a series of experiments to ing the growth rate of the virus on the explore the evolution of behavioral bacteria. If both viruses grew equally interactions between viruses. In this well, the fitness of an evolved virus case, the players in our experimental relative to its ancestor was said to equal game were bacteriophages, or “phag- one. However, if evolution had either plaques form es,” viruses that infect bacteria. Phages improved or worsened the virus’s abil- overnight on bacterial “lawn” are not typically thought to exhibit ity to grow, then the fitness was respec- behavior, but they have proved to be tively either greater or less than one. very useful to test models of conflict- A conspicuous result of the study harvest ing behavioral strategies in evolution- was that the viruses cultured in the co- plaques ary game theory. Such tests would be infecting populations had much high- difficult if not impossible to do with er fitnesses during co-infection, than higher organisms. during single infections. This result We employed the services of a is consistent with the possibility that phage called phi-6, an RNA virus (one evolution under co-infection had se- remove bacteria with filter that has RNA instead of DNA as its lected for cheater viruses—genotypes genetic material) in the family Cystovi- that could efficiently use the products ridae, which attacks legume-infecting of other viruses in the resource pool, bacteria. In the laboratory, the virus but that were less efficient on their is typically grown on Pseudomonas own. The evolved viruses also had the ability to infect the bacteria and Figure 6. Different strains of phi-6 bacterio- replicate on their own, indicating that phages (dots) can be created in the laboratory the cheaters were not simply defec- by controlling the number of bacteriophage tive-interfering viruses that had lost particles that can co-infect a single bacterial mix viruses again key genes. Because the ancestral virus cell (lozenges). Three populations were grown with fresh bacteria did not show any fitness advantage on a bacterial “lawn” at phage-bacterium ra- during co-infection with other virus tios that allowed at least two or three bacterio- genotypes, we defined the ancestral phages to enter each host cell (left protocol), strategy as cooperation. whereas three other populations were allowed to evolve in ratios where no more than one phage entered a bacterial cell (right protocol). Game-Theory Solutions After 50 days, or 250 phage generations, each The evolution of cheater viruses con- repeat process for 50 days population was allowed to compete against an taining a full set of genes provided a three populations per treatment ancestral strain (see Figure 7). unique opportunity to examine wheth-

© 2005 Sigma Xi, The Scientific Research Society. Reproduction 432 American Scientist, Volume 93 with permission only. Contact [email protected]. er the phage was caught in the prison- Interactions between selfish defec- co-infection is common. If a mutant er’s dilemma. We needed to confirm tive-interfering viruses and cooperative defective-interfering virus enters the two key predictions. First, the fitness helper viruses can also be explained us- population, it has a very large fitness of the cheaters relative to the ancestral ing game theory. Imagine a population advantage because it is surrounded cooperator had to be frequency depen- composed entirely of cooperative help- by cooperators that provide essential dent—sensitive to the ratio between ers growing in an environment where gene products. So defective-interfering cheaters and cooperators—because the model predicts that cheaters will show their greatest fitness advantage when they are rare relative to the cooperators. Second, the cheaters had to displace the ancestral cooperators com- filter out pletely and take over the population. bacteria If these two criteria were met, and the takeover by cheaters resulted in a de- ancestral evolved cline in the average fitness of the popu- phi-6 phi-6 bacteriophage lation, then the results would be consis- bacteriophage tent with the prisoner’s dilemma. We set up a series of competitions be- 1:1 mix tween an evolved cheater and the ancestral cooperator. To test for frequency- dependent fitness, the two strains were represented at different initial frequen- cies (ranging from 0.1 to 0.9) for each competition and the numbers were suf- ficiently high to allow co-infection. After allowing the strains to compete for five generations, we found that, indeed, the fitness of the cheater decreased sharply as its initial frequency increased. In oth- add add er words, when cheaters were rare, they viruses to viruses to generally were involved in co-infections bacterial bacterial with cooperators (rather than with other “lawn” “lawn” cheaters), and so they gained a large fitness advantage through their ability to incubate incubate 24 hours 24 hours usurp the components in the resource pool. However, when the cheaters were common, they tended to co-infect cells with other cheaters and so could not profit from their selfish behavior. The same experiment also support- determine ed the second prediction: The cheaters confirm change 1:1 ratio must takeover the population. The fit- in ratio ness of the evolved cheaters relative to number of evolved phi-6 phages the ancestral cooperator was always number of evolved phi-6 phages R = R = 1 number of ancestral phi-6 phages greater than one at all of the initial ra- 0 number of ancestral phi-6 phages tios. This global advantage predicts that the evolved cheater will always R displace the ancestral cooperator. The fitness = 1 strong competitive advantage allows R0 the cheaters to increase their numbers rapidly when they are initially rare. Figure 7. The fitness of an evolved strain of phi-6 bacteriophages relative to the ancestral phi-6 Even though the cost of cheating in- bacteriophages can be determined by having the two strains compete for bacterial hosts. The creases as the number of cheaters in- ability of the two populations to establish themselves and reproduce on a bacterial “lawn” is creases, the cooperators that interacted first established by confirming that they maintain a 1:1 ratio after a 24-hour incubation period with the cheaters always had the low- (left column). When the phages are again counted after a second 24-hour incubation period, est fitness in the system. For this rea- the change in the ratios of the populations (R1 vs. R0) is a measure of their ability to compete against each other. In the author’s experiments, bacteriophages that evolved under conditions son, nothing could prevent the cheat- of high co-infection had a higher fitness during co-infection than during single infections. ers from taking over the population. This result is consistent with the possibility that evolution had selected for “cheating” bacte- Our study was the first to demonstrate riophages—strains that could usurp the products of other bacteriophages (“cooperators”) from the evolution of irrational, selfish be- the resource pool, but were less efficient on their own. The evolutionary costs and benefits of havior in a biological system (see “Esti- the interactions between cheaters and cooperators can be calculated with further experiments mating the Payoffs,” next page). (see “Estimating the Payoffs,” next page).

Experiments between cheating and co- for cheating is defined as P = 1 – c. A operating viruses allow scientists to es- cooperator also loses fitness when in- cooperator cheater timate the fitness payoffs for each of the teracting with a cheater and receives strategies in a 2×2 matrix (right). In the the “sucker’s payoff,” S = 1 – s1. Here prisoner’s dilemma, when two coop- we set P/S = (1 – c)/( 1 – s1) equal to sucker’s erators interact the reward is defined as the right y-intercept of the regres- reward cooperator payoff R = 1. When a cheater meets a coop- sion line, which represents the case erator, the temptation to cheat, T, must when there is the greatest frequen- R = 1 S = 1 – s1 exceed the reward for cooperating by cy of cheaters. Because our results some value, say s2, so that T = 1 + s2. So found that the right y-intercept was the fitness of the cheater relative to the a number greater than one, P had to temptation cooperator is T/R, when the cheaters are be a value greater than S. However, cheater punishment to cheat rare. The value of s2 can be determined the ratio cannot be solved because of P = 1 – c from experimental data. My colleagues two unknown variables, c and s1. We T = 1 + s2 and I set the equation, T/R = (1 + s2)/1, needed to devise an additional ex- equal to the left y-intercept of the re- periment to directly estimate P or S. gression line for the data (see graph We did this by measuring the growth substitute the value c = 0.17 in the below, left). The y-intercept represents rates of the cheaters and cooperators equation for P/S. It was then trivial the case where a cheater is very rare when co-infecting cells on their own. to estimate the parameter s1, so we and therefore guaranteed of interact- This experiment mimics the situa- could fill in the payoff matrix. Be- ing with cooperators. So an individual tion when the cheater viruses take cause the cheater ultimately replaces cheater receives the maximum benefit: over the virus population. We found the cooperator, while lowering the T/R = (1 + s2)/1 = 1.99, and s2 = 0.99. that the growth rate of the cheaters fitness of the population (see graph When two cheaters meet, there is was 83 percent of the cooperators’ below, right), the results are consistent a loss of fitness, and the punishment growth rate. So P = 1 – c = 0.83. We with the prisoner’s dilemma.

2.5 2.0 e

2.0

1.5

1.0 1.0 to ancestral cooperators fitness of evolved cheaters relativ fitness relative to ancestral cooperator .5 0.0 0.2 0.4 0.6 0.8 1.0 0.5 evolved cheaters and evolved cheaters initial frequency of cheater ancestral cooperators viruses become increasingly common detriment of their cooperative ancestor. idea assumes that complementation is in the population. However, as the Evidence from other experiments with not always entirely passive and so can be selfish individuals increase in relative phi-6 suggest that complementation may improved through selection. One possi- frequency, their fitness will decline be- be involved. When the ancestral phi-6 bility is that the cheaters may be poor at cause there are fewer cooperators pres- strain is allowed to co-infect the same producing capsid proteins, which could ent. If the defective-interfering viruses cell with various less fit mutants of the explain their low productivity when take over, their fitness falls to zero be- virus, a greater-than-expected number they infect cells on their own. However, cause they cannot reproduce on their of mutants appear among the offspring. they may be very efficient at stealing own. In this case, the strategy of the This suggests that complementation can entry into capsids produced by coopera- defective-interfering viruses can only take place passively whenever multiple tors during co-infection. It may be that persist through a mixed evolutionary phi-6 genotypes co-infect the same cell. cheaters have evolved mechanisms that stable strategy involving a helper virus. The evolution of cheating viruses may recognize attachment and entry into vi- Evolutionary game theorists call this occur because their prolonged exposure ral capsids. the “chicken game.” to co-infection results in a strong selec- It is still not clear how the selfish phi- tion for the virus to become more ef- Outside the Laboratory 6 genotypes can so efficiently sequester ficient at complementation—a trait that Laboratory studies of cheating viruses products from the resource pool to the was already present in the ancestor. This and bacteria may seem esoteric, but

© 2005 Sigma Xi, The Scientific Research Society. Reproduction 434 American Scientist, Volume 93 with permission only. Contact [email protected]. they have much to offer for under- ogy and those working in basic and of one virus by another. One differ- standing the ecology and evolution applied aspects of microbiology. ence may be that scroungers broadly of microorganisms in nature and in excel at indirectly competing for rare medical and commercial settings. Very A Dilemma? resources, whereas cheaters narrowly little is known about the interactions Following our study, other scientists specialize in directly competing with between microorganisms in the wild. suggested that certain populations of their particular helper virus. So, exam- In fact, the vast majority of microbial yeast and bacteria may also evolve ining frequency-dependent fitness rela- species in nature have yet to be de- according to the prisoner’s dilemma. tive to a variety of viral genotypes might scribed. Cheating has been observed in Some yeast cells forgo the production resolve whether one conflict is a better laboratory experiments among viruses, of a sugar-digesting enzyme, opting to descriptor than the other. In either case, bacteria and slime molds, and it seems steal sugar that was digested by coop- there’s always a cheater and a sucker in- likely that we will discover cheaters in erators. And certain bacterial mutants volved. A showman aware of these virus- natural communities of these microor- cheat by ignoring a chemical signal to es and their prodigious rate of replication ganisms as well. stop growing, while others in the pop- might have quipped instead, “There’s a Human beings have a long history ulation enter a stationary phase. But sucker born every microsecond.” of using bacteria and yeasts for the pro- not everyone agrees that the prisoner’s Bibliography duction and flavoring of foods and bev- dilemma is the best way to describe Denehy, J. J., and P. E. Turner. 2004. Reduced erages, including cheese, bread, wine interactions between microorganisms. fecundity is the cost of cheating in RNA and beer. More recently, we have pur- Microbes obviously lack the complex virus Φ6. Proceedings of the Royal Society: posefully cultured microorganisms to behavior of “higher” life forms, and so Biological Sciences 271:2275–2282. create vaccines, which are often weak- it’s been suggested that mathematical Froissart, R., C. Wilke, R. Montville, S. Remold, ened or inactivated microbes that are models that are not based on animal L. Chao and P. E. Turner. 2004. Co-infection weakens selection against epistatic muta- administered to elicit an immune re- behavior may be more accurate to de- tions in RNA viruses. Genetics 168:9–19. sponse. Vaccines are widely available scribe these phenomena. Giraldeau, L.-A., and T. Caraco. 2000. An in- for combating infectious diseases such An alternative interpretation involves troduction to producer-scrounger games. In as polio, measles and mumps, and there “producers” and “scroungers.” A pro- Social Foraging Theory, ed. L.-A. Giraldeau are now efforts to develop vaccines for ducer expends energy generating op- and T. Caraco. Princeton, N.J.: University other diseases such as AIDS and malar- portunities to exploit resources that are Press, pp. 151–173. Lopez-Ferber, M., O. Simon, T. Williams and ia. Similar strategies are used in agricul- essential to survival and reproduction, P. Caballero. 2003. Defective or effective? ture, where vaccines are administered whereas a scrounger takes advantage Mutualistic interactions between virus to prevent diseases in livestock, and of these opportunities, usurping the re- genotypes. Proceedings of the Royal Society: crops are sprayed with viruses or bacte- sources that producers extract from the Biological Sciences 270:2249–2257. ria to combat plant diseases or to target environment. In this view, the ancestral Maynard Smith, J. 1982. Evolution and the The- insects that destroy crops. The research phi-6 phage is the producer, whereas ory of Games. Cambridge, U.K.: Cambridge University Press. described in this article suggests that the descendant viruses that evolved Nowak, M. A., and K. Sigmund. 1999. Phage- industrial producers of microorganisms under co-infection are scroungers. The lift for game theory. Nature 398:367–368. should be wary of contamination by limited resources could be replication Roux, L., A. E. Simon and J. J. Holland. 1991. Ef- cheaters, which may compromise the enzymes or other proteins essential fects of defective interfering particles on viral desired tastes of foods and beverages, for the production of progeny. When replication and pathogenesis in vitro and in or the effectiveness of vaccines and bio- scroungers are rare, they frequently en- vivo. Advances in Virus Research 40:181–211. logical pesticides. counter producers, so they have many Travisano, M., and G. J. Velicer. 2004. Strate- gies for microbial cheater control. Trends in On the other hand, cheating viruses opportunities to grab the resources. Microbiology 12:72–78. and bacteria may provide desirable and The scroungers have an advantage and Turner, P. E., and L. Chao. 1998. Sex and the exciting new avenues for the applica- should increase when they are rare. evolution of intrahost competition in RNA tion of microorganisms. For example, The producer/scrounger analogy virus Φ6. Genetics 150:523–532. scientists are now trying to determine assumes there is a cost associated Turner, P. E., and L. Chao. 1999. Prisoner’s di- whether defective HIV strains can in- with scrounging. This may be simply lemma in an RNA virus. Nature 398:441–443. terfere with the ability of ordinary HIV the increased competition between Turner, P. E., and L. Chao. 2003. Escape from to replicate and spread within the body scroungers when they become com- prisoner’s dilemma in RNA phage Φ6. American Naturalist 161:497–505. and so prevent or delay the onset of mon. If the cost of scrounging is not Vogt, P. K., and A. O. Jackson. 1999. Satellites AIDS in HIV-infected individuals. too great, the scrounging strategy will and defective viral RNAs. Current Topics Biologists still need to determine the replace the producer strategy in the in Microbiology and Immunology 239. New conditions that promote or hinder the population. However, if each producer York: Springer-Verlag. growth of microbial cheaters. Math- retains a sizable fraction of the resourc- Wilke, C. O., and I. S. Novella. 2003. Phenotypic ematical models such as evolution- es it creates, despite a high frequency mixing and hiding may contribute to memory ary game theory will be valuable for of scroungers, the producers will in- in viral quasispecies. BMC Microbiology 3:11. predicting their long-term success. As crease when they are rare and drive we continue to discover the intriguing the two strategies into a mixed equilib- For relevant Web links, consult this interactions between microorganisms rium in the population. issue of American Scientist Online: that foster the evolution of cheating The cooperator/cheater and the pro- strategies, we will in turn provide op- ducer/scrounger conflicts obviously http://www.americanscientist.org/ portunities for exchanges between sci- have many similarities—most notably, IssueTOC/issue/761 entists interested in evolutionary biol- both analogies deal with the parasitism