S C I E N C E ’ S C O M P A S S ● R E V I E W

R E V I E W : E C O L O G Y in the Interactions and of Anurag A. Agrawal

species represents an interaction norm where When individuals of two species interact, they can adjust their in response the response of one species to the other cre- to their respective partner, be they antagonists or mutualists. The reciprocal pheno- ates the environment to which the other spe- typic change between individuals of interacting species can reflect an evolutionary cies may then respond (Fig. 3). The current response to spatial and temporal variation in species interactions and ecologically sign, strength, and variability in the species result in the structuring of food chains. The evolution of adaptive phenotypic interaction then depends on the past recipro- plasticity has led to the success of in novel habitats, and potentially cal responses between the individuals. In this contributes to genetic differentiation and . Taken together, phenotypic simplified view, spatial aspects of the biotic responses in species interactions represent modifications that can lead to reciprocal and abiotic environment are assumed to be change in ecological time, altered community patterns, and expanded evolutionary constant. The decomposition of the environ- potential of species. mental component of the interaction norm into temporal (Fig. 3) and spatial aspects allows for a more detailed analysis of varia- henotypic plasticity is the ability of an Reciprocal phenotypic change in spe- tion in species interactions. to express different pheno- cies interactions. The intersection between Reciprocal phenotypic change in ecologi- Ptypes depending on the biotic or abiotic species interactions and phenotypic plasticity cal time may be (i) a primary determinant of environment (Fig. 1). Single genotypes can has generated considerable interest among an organism’s in nature; (ii) the change their chemistry, , develop- evolutionary ecologists (Table 1). The study result of long-term evolution where the envi- ment, , or behavior or in response of phenotypic responses of one organism to ronment (i.e., the species interaction) has to environmental cues. R. A. Fisher and other another is by definition an investigation of a been variable; and (iii) a stabilizing factor in 20th century evolutionary biologists lacked species interaction. Yet, biologists have al- mutualistic interactions. A signature of recip- explanations for phenotypic plasticity (1). most entirely focused on plasticity in species rocal phenotypic change is the escalation of Evolutionary biologists have been interested interactions as one-sided events: What is the phenotypes between individuals of two spe- in studying the genetic basis of phenotypes, effect of variation in species X on the pheno- cies over an extended bout of interactions. and early work was focused on traits pre- type of species Y? In nature, however, it is This can be a directional change in the phe- sumed to be unaffected by the environment. quite likely that interacting individuals are notype of partners, where exposure to certain Environmentally affected phenotypes were continually responding to their interaction cues activates in a dose-dependent considered of lesser importance because of partners in a reciprocal fashion over ecolog- manner (Fig. 3). For example, in a mutualis- their apparent lack of a genetic basis. The ical time (Fig. 2). Reciprocal interaction sim- tic interaction, individuals may increase re- modern view of phenotypic plasticity rejects ply implies a back-and-forth response in wards in response to increased services from this notion because phenotypic plasticity of- terms of phenotypic change between individ- a partner, and this back-and-forth changing of ten has a genetic basis. Now, many ecologists uals, and does not imply symmetry in the phenotypes can be a continuous or iterative and evolutionary biologists have embraced strength of responses or effects of one partner process. However, reciprocal phenotypic the idea that under many circumstances such on the other. Many antagonistic or mutualis- change does not have to be directional (5–7). phenotypic plasticity can be adaptive (2). tic interactions, including those that are not Here again, there is an analogy to coevolu- This hypothesis parallels Lamarck’s First behavioral, may involve reciprocal phenotyp- tionary dynamics, where two possible out- Law proposed in Philosophie zoologique in ic changes in ecological time. As an analog, comes are escalating (directional) arms races 1809 (3), which stated that organisms accli- studies of have long attempted to or polymorphisms that are stable or fluctuat- mate to their environment to improve perfor- study the reciprocal evolutionary change in ing in space or time. The latter case predicts mance. Of course, Lamarck is largely dis- interacting species. high levels of genetic and phenotypic varia- credited by his Second Law, which suggested Thompson (4) proposed a similar exami- tion between populations of the interacting that these adjustments to the environment nation of phenotypic plasticity in species in- species because of variation in the cost- were heritable. The modern view of plasticity teractions termed an interaction norm. An benefit ratio of a particular to can be generalized to the statement that phe- interaction norm is expressed as a genotype- the partner (8). Nondirectional phenotypic notypic plasticity evolves to maximize by-genotype-by-environment interaction. In changes in ecological time may similarly in variable environments (the adaptive plas- other words, the phenotype of an individual result in variable phenotypes between dif- ticity hypothesis) (2). Here, I take the adap- or the sign and strength of an interaction ferent pairs of interacting species. Recent tive plasticity hypothesis as a starting point between species is determined by the geno- advances in the understanding of transpos- for evaluating ongoing and future research types of interacting individuals and the envi- able elements suggest that stress-induced directions in the ecology and evolution of ronmental conditions in which they occur. I retrotransposons may be a mechanism for species interactions. propose a refinement of the interaction norm nondirectional change. Defensive respons- concept that distinguishes environmental ef- es in and result in increased fects that are generated by spatial versus tem- transpositional activity that may result in Department of Botany, University of Toronto, 25 Willcocks Street, Toronto, ON M5S 3B2, Canada. E- poral variation. Reciprocal phenotypic immunity to parasites (6, 7). mail: [email protected] change between individuals of interacting Although studies have not been conducted

www.sciencemag.org SCIENCE VOL 294 12 OCTOBER 2001 321 S C I E N C E ’ S C O M P A S S to test the specific hypothesis of reciprocal quently produce tubules. After additional re- strength of predaceous crabs: crabs eating phenotypic change in ecological time, several ciprocal signaling, rhizobia differentiate into mussels without shells grew smaller and examples suggest that it is common. Interac- bacteroids that fix atmospheric nitrogen. Sub- weaker claws than crabs eating intact mussels tions between probable mutualists such as sequent reciprocity dictates the level to which with shells. Conversely, mussels respond to leguminous plants (Fabaceae) and nitrogen- legumes and rhizobia cooperate and ex- cues from predators, including crabs, by in- fixing (rhizobia) demonstrate recip- change resources (10). ducing increased shell thickness, abductor rocal phenotypic change (9). Bacteria near Reciprocal phenotypic change has also muscle strength, and byssal threads (12, 13). start by producing lipo-oligosaccharides been indicated in largely antagonistic interac- In the relationship between plants and herbi- (termed Nod factors). Plants then distort tions. Smith and Palmer (11) elegantly dem- vores, plants may induce defenses that are that curl around the bacteria and subse- onstrated plasticity in morphology and claw dependent on the density of attackers, and herbivores may induce counterdefenses that Fig. 1. Two individuals are dependent on the concentrations of of a single clone of the defenses consumed (14–17) (Fig. 4). The Asian and African wa- continuous range of phenotypes induced by ter flea, Daphnia lum- each partner exemplifies a hallmark require- holtzi. The individual ment of the ecological arms race hypothesis on the left was ex- posed to chemical because it allows for escalating phenotypic cues from predaceous change. If these responses are not continuous, fish (induced); the in- then “arms races” in ecological time are less dividual on the right likely. If reciprocal phenotypic change is the was not (control). The result of adaptive plasticity in both partners, sharp helmet and ex- then it is predicted that coevolution may re- tended tail spine of the induced morph sult in phenotypic plasticity, as opposed to, or protect D. lumholtzi in addition to fixed . from fish predators. Phenotypic changes in species interac- The uninduced form tions typically revert back to the original was formerly de- phenotypic state after each extended bout of scribed as a different interactions (e.g., each year). However, four species (D. monacha Brehm 1912). Green possibilities may delay this resetting of the (83), in an accurate interaction: (i) Maternal environment effects and prophetic study, may persist across generations and years (18, related the occurrence 19); (ii) some phenotypic modifications, such of both morphs to dif- as tree responses to defoliation, occur across ferences in fish preda- years (20); (iii) some phenotypic responses, tion. The induction of this morphological de- especially morphological responses, are can- fense has now been alized and cannot revert to the original state implicated as a key factor in the success of D. lumholtzi invading North America (84). even if the interaction is over (21); and (iv) some environments have reduced seasonality, which may prolong the duration of the inter- Fig. 2. In the mutualism action. Alternatively, reciprocal phenotypic between many herbivo- responses, especially in behavioral interac- rous insects and ants, her- bivores produce a sugar- tions, may begin and end in a matter of and amino acid–rich nec- seconds. Thus, the time scale of reciprocal tar as food for ants, and phenotypic responses between species varies ants protect the herbi- from seconds to years, and it may depend on vores from predators and whether the responses are chemical, physio- parasitoids. Here an ant is logical, morphological, or behavioral. drinking from the nectar gland of a scale insect on Phenotypic plasticity as a determinant an aspen tree. Because of food chain structure. The distribution the herbivores can alter and abundance of organisms in a multitrophic the level of food rewards community context can also be influenced by offered depending on the phenotypic plasticity. For example, there are biotic and abiotic envi- manifold plastic responses of prey to preda- ronment and the ants can similarly alter their pro- tors, some of which may affect other species tective services, reciprocal in an ecological community (Table 1). The phenotypic change deter- consumption of prey by predators can obvi- mines the ecological ously have strong ramifications for the com- outcome of the species munity; however, nonconsumptive effects of interaction. [Photograph predators on the phenotype of prey and lower by A. A. Agrawal] trophic levels may be important, but they have only recently been examined. One com- mon response of prey to the scent or visual presence of predators is to hide and/or to reduce feeding (22, 23). These behavioral responses have the potential to affect not only

322 12 OCTOBER 2001 VOL 294 SCIENCE www.sciencemag.org S C I E N C E ’ S C O M P A S S the predator and prey but also the prey’s food. regulatory role is still unknown (15). fend caterpillars against predators and para- Indeed, in both aquatic and terrestrial sys- Plants may also affect herbivores through sitoids. Various environmental factors in- tems, it has now been shown that noncon- phenotypic plasticity in indirect defense. Her- cluding caterpillar group size and threat of sumptive effects of predators on prey behav- bivore-damaged plants emit volatile com- predation influence the cost-benefit ratio of ior can cascade through the trophic web to pounds that attract natural enemies of herbi- providing rewards to ants and, thus, dictate a have effects on standing plant biomass (22, vores (26, 27). The production of plant vola- caterpillar’s signaling and subsequent re- 23). In other words, the mere threat of tiles elicits responses from natural enemies of wards (31, 32). The level of the caterpillars’ predation can result in decreased feeding herbivores in two ways. Predators and para- signaling and rewards reciprocally influences to the point where plant biomass increases. sitoids are innately attracted to some plant the number and attentiveness of mutualistic For example, predator-induced morphologi- volatiles and learn to associate other volatiles ants. Here, the multilevel responses of both cal structures in an intertidal barnacle (Chtha- with the presence of prey (28, 29). Hence, the partners are essential for maintaining stability malus anisopoma) resulted in complex inter- reciprocal plastic responses (i.e., volatile pro- in the interaction. The presence of rewards actions whereby the abundance of mussels duction and associative ) can be im- affects the abundance and distribution of was reduced and algal cover was increased portant for the ultimate benefit to the plants. ants; however, ineffective defense by ants (24). Because mussels and algae compete for Inappropriate volatile signals or repeated ex- may result in reduction in the rewards offered settlement space, and the plastic response of posure to volatiles without prey can backfire by caterpillars. Plants display similar flexibil- the barnacles reduced settlement space for for the plant resulting in natural enemies of ity in interactions with pollinating mutualists, mussels, algae indirectly benefited from herbivores that learn to avoid plants (29). with the ability to decrease (30) or increase higher trophic interactions. Thus, reciprocal phenotypic responses may (33) rewards based on previous pollination Phenotypic plasticity in plants can affect mediate the abundance and distribution of success. Thus, reciprocal phenotypic change the abundance and distribution of herbivores trophic levels in food chains. may be an answer to the often-discussed evo- and the predators and parasitoids of herbi- In species interactions, especially mutual- lutionary instability of mutualisms. Recipro- vores. In 1975, Haukioja (25) proposed that isms, there is often extensive signaling be- cal phenotypic change may promote fidelity the well-documented dramatic multiyear cy- tween partners that may cause reciprocal phe- among mutualists, because it allows for part- cles of herbivores could, in part, be mediated notypic change (Fig. 2). This reciprocal ner choice, retaliation, or a general adjust- by plant responses that depended on the den- change may ameliorate the conflict of interest ment of rewards or services (34, 35). Geno- sity of herbivores. In his model, herbivore that often ensues between the partners in types that cheat may be cut off by their populations rise until inducing a plant de- mutualism: The increase in fitness of one partner and thus disfavored by natural selec- fense strong enough to precipitate a crash in species that comes at the expense of the other tion (10, 30). the herbivore population. Although density- (30). For example, the interaction between dependent induction of defenses has been lycaenid caterpillars and ants has been well documented in some systems (Fig. 4) and studied: Caterpillars use vibrations and chem- theoretical arguments suggest that plasticity ical signals to attract ants and then feed them in plant defense could cause cycles, its true sweet nectar; consequently, tending ants de-

Fig. 3. (A) One possible course of reciprocal phenotypic change between individuals in an antagonistic species interaction. Key: Circles, levels of defense in a host organism; triangles, levels of counterdefense in the parasite. The dynamic nature of such increase in defense following attack and decrease following remov- al of parasites has been recently demonstrated for spines on Acacia drepanolobium that are Fig. 4. Potential for an ecological arms race induced by vertebrate herbivores (75). Ecolog- between plants and herbivores. (A) Phenotypic ical reciprocity may take place in all interac- escalation in a plant defense (pigmented tions, irrespective of the sign of effect on an glands containing hemigossypolone in cotton) individual species. However, phenotypic change that is dependent on the number of attacking in response to a species interaction need not be mite herbivores [redrawn from (16) with per- directional (5). (B) Colorado potato beetle (Lep- mission from Kluwer Academic Publishers]; and tinotarsa decemlineata) with eggs. Although (B) phenotypic escalation in a herbivore’s damage to potato by this herbivore (Trichoplusia ni) counterdefense (production of causes an induction of plant defenses (- inhibitor insensitive proteases) dependent of ase inhibitors), following induction beetles cleverly adjust their arsenal of digestive proteases to be the concentration of plant defense (proteinase insensitive to plant defense (14). [Photograph by Jack Kelly Clark, ANR, University of California at inhibitors) consumed. [Redrawn from (17) with Davis, ©The Regents, University of California] permission from Elsevier Science]

www.sciencemag.org SCIENCE VOL 294 12 OCTOBER 2001 323 S C I E N C E ’ S C O M P A S S Ecological and evolutionary conse- ful invaders? Two studies highlight the fact this shift (41). By definition, organisms in quences of phenotypic plasticity in novel that organisms rely on plasticity in novel novel habitats will experience new challenges habitats. How does phenotypic plasticity habitats: (i) Lizards introduced to islands that may cause the expression of potentially affect ecological success and evolutionary with different vegetation types and (ii) snails beneficial traits through plasticity. divergence in new or novel habitats, where in habitats invaded by exotic predators were Two empirical examples demonstrate the “habitat” refers to a biotic or abiotic neigh- thought to have genetically adapted to these possible role of phenotypic plasticity in evo- borhood? As an example consider host-use in new environments (37, 38). However, in both lutionary divergence. Phenotypic plasticity parasitoids, insects with a well-developed of these cases, the respective phenotypes in may have facilitated the host-shift of Rhago- ability for associative learning. Learning can the new habitats (long hind limbs and in- letis flies from hawthorn to apple fruits (42, be a general form of phenotypic plasticity creased shell thickness) were ultimately ex- 43). Naı¨ve adults emerging from apple hosts that organisms can apply to different environ- plained by phenotypic plasticity, not genetic showed a strong preference for hawthorn in mental stimuli. Can the ability to learn, and change (39, 40). One reason why plasticity both choice and no-choice tests (42). How- therefore associate particular environmental was not strongly considered in the original ever, those adults that experience apple in cues with particular hosts, influence the po- studies is that the respective native habitats their ecological neighborhood after emerging tential to use new hosts and potentially the were apparently not particularly variable in from pupae overwhelmingly chose apple over ability for a race of parasitoids to specialize vegetation structure and predation risk (37, hawthorn in subsequent oviposition. The con- on that new host? Such questions about the 38). In another study, Fox has discovered tribution of this induction of host-fidelity to ecological and evolutionary consequences of that the host-range of a seed beetle has restricted flow and genetic substructur- plasticity have been raised in the past, as far expanded to a new tree species and that ing has been demonstrated in nature for this back as Baldwin in 1896 (1, 36). Yet, empir- maternal environmental effects (i.e., phe- system (43). Thus, phenotypic plasticity may ically these questions remain completely un- notypic plasticity cued in the maternal have played a role in this sympatric specia- answered. For example, are generation and expressed in progeny) on tion. Early abiotic experience in the form of more phenotypically plastic than unsuccess- progeny size and composition are facilitating light conditions of adult melano- gaster resulted in assortative mating in labo- ratory trials (44). Although the phenotype Table 1. Apparently adaptive phenotypic responses of organisms in species interactions. In few of these itself was unknown, the environmentally interactions have reciprocal responses been investigated. Because of the variable nature of phenotypic plasticity, the sign of the interaction may change depending on the respective phenotypes of the cued change resulted in organisms from sim- interacting organisms. ilar environments preferentially mating with each other, a potential mechanism for facili- Response Reference tating evolutionary divergence. In both of these studies a plastic response led to assor- Competition tative mating among individuals that experi- Increased production of defensive caste in social insects (55) ence the same environment. Production of defensive structures in aquatic (56, 57) The idea that plasticity may lead to eco- Production of dispersal morphs (58) logical success in a novel habitat is perhaps Trophic specialization in tadpoles (59) intuitive, but the idea that plasticity may also Stem elongation in plants (54) lead to evolutionary divergence in novel hab- Adjustment of progeny vigor in plants (60) itats is less so. The hypothesis posits that Mutualism plasticity may itself lead to genetic differen- Rewards dependent on the presence of mutualists in plants and animals (31, 61, 62) tiation, a seeming contradiction. Here, the Behavioral avoidance/attraction to partner depending of rewards in (63, 64) key is that plasticity allows for the coloniza- animals tion and success in a novel habitat, and other Plants punishing cheaters (10, 30) forces (e.g., allopatry or induced preference Adjustment of rewards according to need in plants and animals (31–33, 65) for the novel habitat) cause restricted gene Predation risk (animals) flow to organisms in the original habitat. If plastic organisms are restricted to or favor the Hiding, reduced activity and feeding (22, 23) novel habitat, a new host race may be formed. Diet or habitat induced camouflage (66, 67) Induction of morphological defenses (18, 21, 39, 68) The eventual loss of plasticity may not be Production of dispersal morphs (69) required in all cases, as it seems that Rhago- Behavioral escape (70, 71) letis has retained its diet-induced phenotypic Transformation into a parasite of the predator (72) plasticity even in the new host race (42). Adjustment of progeny defenses (18, 73) The ecological and evolutionary conse- /herbivory quences of phenotypic plasticity depend, in part, on whether plasticity evolves in re- Increased immune function in animals (68) sponse to particular environmental variation Induced chemical and/or morphological defense in plants (15, 74, 75) Increased heterophylly (in this case, producing leaves under water) (76) (i.e., specialized sun versus shade leaves in Behavioral shifts in the host (adaptive for host) (77) plants) or evolves as an overall strategy of the Behavioral shifts in the host (adaptive for parasite) (78) organism (i.e., associative learning). For most Gall formation on plants (adaptive for galler) (15) plastic traits, the response is probably inter- Food quality (predators and herbivores) mediate, in that the trait may be responsive to at least a few different environmental stimuli. Adjustment of gut (14, 17, 79) Whether this range of response to stimuli is Adjustment of feeding structures (11, 80) adaptive or not, I predict that the larger the Adjustment of progeny size and quality (41, 81) Adjustment of behavioral preferences (82) range of stimuli to which a trait will respond, the more likely this plastic trait will affect the

324 12 OCTOBER 2001 VOL 294 SCIENCE www.sciencemag.org S C I E N C E ’ S C O M P A S S ecology and evolution of interactions in novel genotypes (48). These types of costs may nipulations where the mechanisms of plastic- habitats. Learning and other general forms of result in selection disfavoring plasticity. For ity are unknown. The ecological costs of phenotypic plasticity with responses to di- example, when transferred from antibiotic- phenotypic plasticity will be evident as long verse stimuli have enormous potential as free to antibiotic-containing media, inducible as the ecological arena for the experiment is mechanisms to increase the success of - resistance to tetracycline in representative of the typical (variable) habi- isms in novel habitats (45). If phenotypic is associated with a much longer lag phase in tats that organisms experience. plasticity is involved in the ecological and growth compared with that of bacteria con- Conclusion. Understanding the interac- evolutionary success of organisms in novel stitutively expressing resistance (50). tions between species is an important goal of habitats, then it might be expected that sub- The genetic costs of plasticity involve ecology and evolution. In addition to the disciplines, from invasion biology to specia- trade offs between the degree or pattern of perhaps obvious consumptive interactions be- tion, may benefit from considering the con- plasticity and other traits that increase fitness. tween a consumer and a resource, or two sequences of plasticity. These costs include the maintenance of phys- exploitative competitors, we now have a firm A unique form of phenotypic plasticity iological machinery to sense or regulate phe- grip on the sometimes more subtle interac- that may result in evolutionary change is notypic plasticity (i.e., an allocation cost), tions that organisms mediate through altered in response to stress. Transposable and are based in , linkage disequi- phenotypes when exposed to interaction part- element activity increases in response to librium, or influencing other traits. ners. Because phenotypically plastic adapta- stress in both plants (6) and animals (7). The signature of genetic costs is lower fitness tions are more likely to evolve in variable Similarly, under stressful conditions organ- of plastic organisms compared with less plas- environments than fixed adaptations, and isms can produce DNA polymerases (muta- tic conspecific genotypes in a single environ- species interactions are intrinsically variable ses) that replicate faulty or mutated DNA, ment. In the example of resistance to tetracy- in space and time, the (co)evolution of spe- potentially introducing solutions in an envi- in E. coli, possession of an inducible cies interactions has certainly resulted in phe- ronment where novelty is required (46). Such itself (in the absence of its expression notypic plasticity. These phenotypic chang- environmentally induced have par- or expressed and compared with a constitu- es can result in the structuring of food ticularly potent evolutionary potential in tively expressed genotype) has a very low chains, reciprocal ecological changes in an- clonal organisms such as microbes and fitness cost (51). tagonistic and mutualistic interactions, and plants. Even in nonclonal plants, stress-in- Both ecological and genetic costs of plas- the evolution of species. Although costs are duced somatic mutations may be transferred ticity may impose an evolutionary constraint likely to constrain the evolution of adaptive to progeny because reproductive tissues may on responses to favoring phenotypic plasticity, the ubiquity of plas- be differentiated from mutated somatic tis- plasticity, although little empirical progress ticity in species interactions suggests that sues. The adaptive benefits of high mutation has been made in this area. The most power- the benefits outweigh costs under a wide rate in a novel interaction between species ful approach to study the costs of phenotypic variety of conditions. have now been demonstrated (47). Thus, if plasticity is to start with heritable variation induced transposons and mutases have for a plastic trait within a species. Tradition- References and Notes evolved by natural selection, then this form ally, quantitative within a 1. M. J. West-Eberhard, Annu. Rev. Ecol. Syst. 20, 249 of phenotypic plasticity may be a new mech- population, selection experiments, or mutants (1989). anism for generating variation subject to nat- that under- or overexpress a trait of interest 2. S. A. Dudley, J. Schmitt, Am. Nat. 147, 445 (1996). 3. J.-B. Lamarck, Zoological Philosophy: An Exposition ural selection. have been employed. Hybridization of close- with Regard to the Natural History of Animals (Univ. Evolution of phenotypic plasticity: ly related species (52) and genetic engineer- of Chicago Press, Chicago, 1984). Costs? Although phenotypic plasticity ap- ing (53) has also been employed to manipu- 4. J. N. Thompson, Annu. Rev. Ecol. Syst. 19, 65 (1988). 5. F. R. Adler, R. Karban, Am. Nat. 144, 813 (1994). pears to be ubiquitous in species interactions, late the level of phenotypic responses in spe- 6. M. A. Grandbastien, Trends Plant Sci. 3, 181 (1998). costs of plasticity must constrain its evolu- cies interactions. 7. M. G. Kidwell, D. R. Lisch, Evolution 55, 1 (2001). tion. The following questions arise from this Knowing the mechanism of phenotypic 8. J. N. Thompson, in Herbivores: Between Plants and Predators, H. Olff, V. Brown, R. Drent, Eds. (Blackwell, fundamental assumption: (i) Do costs of plas- plasticity will be of utmost importance in Oxford, 1999), pp. 7–30. ticity prevent the evolution of plasticity for detecting genetic costs. Take for example, 9. D. J. Gage, W. Margolin, Curr. Opin. Microbiol. 3, 613 particular traits; and (ii) given that plasticity two individuals of a plant species, one that is (2000). has evolved in a trait, why does it not con- plastic in its response to competitors and one 10. R. F. Denison, Am. Nat. 156, 567 (2000). 11. L. D. Smith, A. R. Palmer, Science 264, 710 (1994). tinue to evolve to the point where the species that is not (54). The unresponsive individual 12. O. Reimer, M. Tedengren, Oikos 75, 383 (1996). could be successful in all environments (48, may lack the receptors necessary to respond 13. G. H. Leonard, M. D. Bertness, P. O. Yund, Ecology 80, 49)? to competitors and also may lack all of the 1 (1999). 14. C. J. Bolter, M. A. Jongsma, J. Insect Physiol. 41, 1071 Ecological costs of plasticity are defined necessary physiological machinery to re- (1995). by a reduction in fitness of plastic organisms spond to the signal from the receptors; con- 15. R. Karban, I. T. Baldwin, Induced Responses to Her- compared with less plastic organisms, where versely, the unresponsive individual may bivory (Univ. of Chicago Press, Chicago, 1997). the fitness difference is only realized in a have all of the receptors and machinery nec- 16. A. Agrawal, R. Karban, Entomol. Exp. Appl. 96, 39 (2000). particular set of environments (i.e., with par- essary to respond, but may simply be defec- 17. R. M. Broadway, J. Insect Physiol. 41, 107 (1995). ticular interaction partners present or during tive in some final step of the response path- 18. A. A. Agrawal, C. Laforsch, R. Tollrian, Nature 401, 60 rapid environmental shifts). An imperfect way. These two extremes of the continuum (1999). 19. T. A. Mousseau, C. W. Fox, Eds., Maternal Effects as match between a phenotype and the environ- represent potential pitfalls in detecting the Adaptations (Oxford Univ. Press, New York, 1998). ment that results in relatively low fitness genetic costs because the mechanistic basis 20. E. Haukioja, Annu. Rev. Entomol. 36, 25 (1991). exemplifies an ecological cost. Because the for why some organisms are more plastic 21. C. M. Lively, Evolution 40, 232 (1986). success of plasticity is dependent on the pre- than others is not known. In the former case, 22. A. P. Beckerman, M. Uriarte, O. J. Schmitz, Proc. Natl. Acad. Sci. U.S.A. 94, 10735 (1997). dictability of a changing environment, lags in genetic costs are likely detectable; in the 23. J. Van Buskirk, B. R. Schmidt, Ecology 81, 3009 the response to environmental stimuli or latter case, genetic costs may be more diffi- (2000). changes in the environment that are not pre- cult to detect. The ecological, rather than 24. P. Raimondi, S. Forde, L. Delph, C. Lively, Oikos 91, 353 (2000). dictable can impose a significant ecological genetic, costs of plasticity may be readily 25. E. Haukioja, T. Hakala, Rep. Kevo Subarctic Res. Stat. cost compared with that experienced by fixed determined using genetic (or phenotypic) ma- 12, 1 (1975).

www.sciencemag.org SCIENCE VOL 294 12 OCTOBER 2001 325 S C I E N C E ’ S C O M P A S S

26. C. M. De Moraes, W. J. Lewis, P. W. Pare, H. T. Alborn, 49. M. vam Kleunen, M. Fischer, B. Schmid, Evolution 54, 72. J. O. Washburn, M. E. Gross, D. R. Mercer, J. R. J. H. Tumlinson, Nature 393, 570 (1998). 1947 (2000). Anderson, Science 240, 1193 (1988). 27. A. Kessler, I. T. Baldwin, Science 291, 2141 (2001). 50. R. E. Lenski et al., Mol. Ecol. 3, 127 (1994). 73. R. Shine, S. J. Downes, Oecologia 119, 1 (1999). 28. L. Vet, W. Lewis, R. Carde´, in Chemical Ecology of 51. T. N. M. Nguyen, Q. G. Phan, L. P. Duong, K. P. 74. A. A. Agrawal, Science 279, 1201 (1998). Insects 2, R. Carde´, W. Bell, Eds. (Chapman & Hall, Bertrand, R. E. Lenski, Mol. Biol. Evol. 6, 213 (1989). 75. T. P. Young, B. D. Okello, Oecologia 115, 508 (1998). New York, 1995), pp. 65–101. 52. J. M. Scriber, R. L. Lindroth, J. Nitao, Oecologia 81, 76. J. Kouki, Funct. Ecol. 7, 21 (1993). 29. B. Drukker, J. Bruin, M. W. Sabelis, Physiol. Entomol. 186 (1989). 25, 260 (2000). 53. G. W. Felton et al., Curr. Biol. 9, 317 (1999). 77. R. Karban, G. English-Loeb, Ecology 78, 603 (1997). 30. O. Pellmyr, C. J. Huth, Nature 372, 257 (1994). 54. J. Schmitt, S. A. Dudley, M. Pigliucci, Am. Nat. 154, 78. J. Moore, BioScience 45, 89 (1995). 31. O. Leimar, A. H. Axen, Anim. Behav. 46, 1177 (1993). S43 (1999). 79. M. J. Snyder, J. I. Glendinning, J. Comp. Physiol. A. 32. A. H. Axen, N. E. Pierce, Behav. Ecol. 9, 109 (1998). 55. J. A. Harvey, L. S. Corley, M. R. Strand, Nature 406, 179, 255 (1996). 33. A. H. Ladio, M. A. Aizen, Oecologia 120, 235 (1999). 183 (2000). 80. G. G. Mittelbach, C. W. Osenberg, P. C. Wainwright, 34. O. Leimar, Proc. R. Soc. London Ser. B. 264, 1209 56. C. D. Harvell, D. K. Padilla, Proc. Natl. Acad. Sci. Evol. Ecol. Res. 1, 111 (1999). (1997). U.S.A. 87, 508 (1990). 81. C. W. Fox, M. S. Thakar, T. A. Mousseau, Am. Nat. 35. M. Doebeli, N. Knowlton, Proc. Natl. Acad. Sci. U.S.A. 57. E. Chorensky, Biol. Bull. 165, 569 (1983). 149, 149 (1997). 95, 8676 (1998). 58. R. F. Denno, G. K. Roderick, Ecology 73, 1323 (1992). 82. A. A. Agrawal, C. Kobayashi, J. S. Thaler, Ecology 80, 36. B. W. Robinson, R. Dukas, Oikos 85, 582 (1999). 59. D. W. Pfennig, P. J. Murphy, Evolution 54, 1738 37. R. H. Seeley, Proc. Natl. Acad. Sci. U.S.A. 83, 6897 (2000). 518 (1999). (1986). 60. K. Donohue, J. Schmitt, in Maternal Effects as Adap- 83. J. Green, E. African J. Zool. 151, 181 (1967). 38. J. B. Losos, K. I. Warheit, T. W. Schoener, Nature 387, tations, T. A. Mousseau, C. W. Fox, Eds. (Oxford Univ. 84. R. Tollrian, personal communication. Scanning elec- 70 (1997). Press, New York, 1998), pp. 137–158. tron micrograph courtesy of C. Laforsch and R. 39. G. C. Trussell, L. D. Smith, Proc. Natl. Acad. Sci. U.S.A. 61. S. J. Risch, F. R. Rickson, Nature 291, 149 (1981). Tollrian. 97, 2123 (2000). 62. J. Stimson, Marine Biol. 106, 211 (1990). 85. This article was inspired by J. Gurevitch and N. 40. J. B. Losos et al., Evolution 54, 301 (2000). 63. R. Irwin, A. K. Brody, Oecologia 116, 519 (1998). Huntley’s challenge to address three questions for 41. C. W. Fox, U. M. Savalli, Ecology 81, 3 (2000). 64. D. Wagner, L. Kurina, Ecol. Entomol. 22, 352 (1997). the next century of ecology and evolution ( pre- 42. R. J. Prokopy, A. L. Averill, S. S. Cooley, C. A. Roitberg, 65. A. A. Agrawal, M. T. Rutter, Oikos 83, 227 (1998). sented at the Ecological Society of America meet- Science 218, 76 (1982). 66. W. N. Hazel, D. A. West, Biol. J. Linn. Soc. 57, 81 ing in 2000). I thank L. Adler, anonymous review- 43. J. L. Feder et al., Proc. Natl. Acad. Sci. U.S.A. 91, 7990 (1996). ers, T. Bao, S. Barrett, A. Bell, J. Bronstein, M. (1994). 67. E. Greene, Science 243, 643 (1989). Dorken, R. Dukas, J. Fordyce, M. Johnson, R. Karban, 44. M. Barth, H. V. B. Hirsch, M. Heisenberg, Anim. Behav. 68. R. Tollrian, C. D. Harvell, The Ecology and Evolution of N. Kurashige, A. Mooers, K. Rotem, J. J. Thaler, J. S. 53, 25 (1997). Inducible Defenses (Princeton Univ. Press, Princeton, Thaler, P. Van Zandt, L. Vet, D. Viswanathan, M. 45. R. Dukas, E. A. Bernays, Proc. Natl. Acad. Sci. U.S.A. 1999). West-Eberhard, and A. Winn for comments and 97, 2637 (2000). 69. W. W. Weisser, C. Braendle, N. Minoretti, Proc. R. discussion. My laboratory and current work (www. 46. M. Radman, Nature 401, 866 (1999). Soc. London Ser. B. 266, 1175 (1999). botany.utoronto.ca/researchlabs/agrawallab/index. 47. A. Giraud et al., Science 291, 2606 (2001). 70. K. M. Warkentin, Proc. Natl. Acad. Sci. U.S.A. 92, stm) is being supported by the Botany Department 48. T. J. DeWitt, A. Sih, D. S. Wilson, Trends Ecol. Evol. 13, 3507 (1995). at the University of Toronto and the Natural Sci- 77 (1998). 71. J. E. Losey, D. R. F, Ecology 79, 2143 (1999). ences and Engineering Research Council of Canada.

326 12 OCTOBER 2001 VOL 294 SCIENCE www.sciencemag.org