Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents ›...

23
19. B. Wilm, A. Ipenberg, N. D. Hastie, J. B. E. Burch, D. M. Bader, Development 132, 5317 (2005). 20. B. Zhou et al., Nature 454, 109 (2008). 21. B. Lustig et al., Mol. Cell. Biol. 22, 1184 (2002). 22. S. Hayashi, A. P. McMahon, Dev. Biol. 244, 305 (2002). 23. Y. Ruzankina et al., Cell Stem Cell 1, 113 (2007). 24. N. Barker et al., Nature 449, 1003 (2007). 25. L. Topol et al., J. Cell Biol. 162, 899 (2003). 26. A. J. Mikels, R. Nusse, PLoS Biol. 4, e115 (2006). 27. A. Sato, H. Yamamoto, H. Sakane, H. Koyama, A. Kikuchi, EMBO J. 29, 41 (2010). 28. T. Ishitani et al., Mol. Cell. Biol. 23, 131 (2003). 29. G. J. Hannon, D. Beach, Nature 371, 257 (1994). 30. H. L. Moses, E. Y. Yang, J. A. Pietenpol, Cell 63, 245 (1990). 31. S. Dennler et al., EMBO J. 17, 3091 (1998). 32. L. Grumolato et al., Genes Dev. 24, 2517 (2010). 33. M. Yamada et al., Dev. Dyn. 239, 941 (2010). 34. A. Iavarone, J. Massagué, Nature 387, 417 (1997). 35. N. R. Gough, Sci. Signal. 1, eg8 (2008). 36. T. P. Yamaguchi, A. Bradley, A. P. McMahon, S. Jones, Development 126, 1211 (1999). 37. M. Ito et al., Nature 447, 316 (2007). Acknowledgments: We thank R. Kopan for comments on the manuscript, W. Pu for reagents, and R. Head and C. Storer for the microarray analysis. This work was funded by the NIH (grant DK90251); the Pew Scholars Program in the Biomedical Sciences; grant 5T35DK074375 (Trans- National Institute of Diabetes and Digestive and Kidney Diseases Short-Term Training for Medical Students); the Washington Univ. Digestive Disease Research Core (NIH grant P30-DK52574); and the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research. Microarray data are available on Array Express (accession no. E-MTAB-1175). Materials transfer agreements will be required for the acquisition of the Wnt5a f/f mice from the National Cancer Institute and the L-WRN cell line from the Washington Univ. Medical School. The data presented in this manuscript are tabulated in the main paper and in the supplementary materials. Supplementary Materials www.sciencemag.org/cgi/content/full/science.1223821/DC1 Materials and Methods Figs. S1 to S14 Tables S1 to S3 References (3853) 25 April 2012; accepted 21 June 2012 Published online 6 September 2012; 10.1126/science.1223821 Insect Herbivores Drive Real-Time Ecological and Evolutionary Change in Plant Populations Anurag A. Agrawal, 1,2 * Amy P. Hastings, 1 Marc T. J. Johnson, 3 John L. Maron, 4 Juha-Pekka Salminen 5 Insect herbivores are hypothesized to be major factors affecting the ecology and evolution of plants. We tested this prediction by suppressing insects in replicated field populations of a native plant, Oenothera biennis, which reduced seed predation, altered interspecific competitive dynamics, and resulted in rapid evolutionary divergence. Comparative genotyping and phenotyping of nearly 12,000 O. biennis individuals revealed that in plots protected from insects, resistance to herbivores declined through time owing to changes in flowering time and lower defensive ellagitannins in fruits, whereas plant competitive ability increased. This independent real-time evolution of plant resistance and competitive ability in the field resulted from the relaxation of direct selective effects of insects on plant defense and through indirect effects due to reduced herbivory on plant competitors. T he ubiquitous consumption of plants by insect herbivores represents one of the dominant species interactions on Earth and has been hypothesized to play a strong role in the diversification of plant species and their traits (13). The evolution of plant defense has been studied primarily with either a prospective or retrospective approach. Single-generation pro- spective approaches measure contemporary natural selection on resistance traits and make predictions about how those traits should evolve given esti- mates of their heritability (46). By contrast, ret- rospective studies compare populations or species that have diverged over time (79) and make in- ferences about the processes driving evolutionary change. More generally, temporal studies of natu- ral species interactions that influence evolutionary dynamics remain rare, and few have been exper- imental (912). Thus, we lack experimental field studies that quantify how species interactions in- fluence selection on traits and the evolutionary response to this selection. As such, it is not well established how rapidly plants adapt to selection by herbivores, which traits are most important in the evolution of defense, and whether herbivory drives predictable parallel changes due to selec- tion across populations. The ecology and evolution of plant-herbivore relationships is, in part, governed by the recipro- cal nature of their interaction and the complex- ity of communities. For example, plant traits and plant community structure are critical for determining insect occurrence and attack rates ( 13). Conversely, insects can directly reduce focal plant abundance (13) or indirectly affect plant abundance through changes in the density of co- occurring competitors (14). Thus, the evolution- ary dynamics of a focal plant species might be influenced through direct selection on resist- ance traits and/or by indirect selection on traits influencing competitive ability (15, 16). Despite this likely scenario, our understanding of how species interactions influence the evolution of constituent community members is surprisingly limited (17). We conducted a field study of the selective impact of insects and the evolutionary response 1 Department of Ecology and Evolutionary Biology, Cornell Uni- versity, Ithaca, NY 14853, USA. 2 Department of Entomol- ogy, Cornell University, Ithaca, NY 14853, USA. 3 Department of Biology, University of Toronto at Mississauga, Mississauga, ON L5L 1C6, Canada. 4 Division of Biological Sciences, University of Montana, Missoula, MT 59812, USA. 5 Laboratory of Organic Chemistry and Chemical Biology, Department of Chemistry, University of Turku, FI-20014 Turku, Finland. *To whom correspondence should be addressed. E-mail: [email protected] Fig. 1. Fruit loss to seed predator moths (dominated by M. brevivittella) in 18 genotypes of O. biennis in ambient and insect-suppressed plots (data are from the first generation, 20072008, mixed-model analysis of variance, genotype-by-insecticide interaction: c 2 = 57.55, P < 0.001). (Left) an opened fruit with a M. brevivittella larvae consuming immature seeds; (right) a flowering O. biennis. www.sciencemag.org SCIENCE VOL 338 5 OCTOBER 2012 113 REPORTS on October 4, 2012 www.sciencemag.org Downloaded from

Transcript of Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents ›...

Page 1: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

19. B. Wilm, A. Ipenberg, N. D. Hastie, J. B. E. Burch,D. M. Bader, Development 132, 5317 (2005).

20. B. Zhou et al., Nature 454, 109 (2008).21. B. Lustig et al., Mol. Cell. Biol. 22, 1184 (2002).22. S. Hayashi, A. P. McMahon, Dev. Biol. 244, 305 (2002).23. Y. Ruzankina et al., Cell Stem Cell 1, 113 (2007).24. N. Barker et al., Nature 449, 1003 (2007).25. L. Topol et al., J. Cell Biol. 162, 899 (2003).26. A. J. Mikels, R. Nusse, PLoS Biol. 4, e115 (2006).27. A. Sato, H. Yamamoto, H. Sakane, H. Koyama, A. Kikuchi,

EMBO J. 29, 41 (2010).28. T. Ishitani et al., Mol. Cell. Biol. 23, 131 (2003).29. G. J. Hannon, D. Beach, Nature 371, 257 (1994).30. H. L. Moses, E. Y. Yang, J. A. Pietenpol, Cell 63, 245 (1990).31. S. Dennler et al., EMBO J. 17, 3091 (1998).32. L. Grumolato et al., Genes Dev. 24, 2517 (2010).33. M. Yamada et al., Dev. Dyn. 239, 941 (2010).

34. A. Iavarone, J. Massagué, Nature 387, 417 (1997).35. N. R. Gough, Sci. Signal. 1, eg8 (2008).36. T. P. Yamaguchi, A. Bradley, A. P. McMahon, S. Jones,

Development 126, 1211 (1999).37. M. Ito et al., Nature 447, 316 (2007).

Acknowledgments: We thank R. Kopan for comments onthe manuscript, W. Pu for reagents, and R. Head andC. Storer for the microarray analysis. This work was fundedby the NIH (grant DK90251); the Pew Scholars Program in theBiomedical Sciences; grant 5T35DK074375 (Trans- NationalInstitute of Diabetes and Digestive and Kidney DiseasesShort-Term Training for Medical Students); the WashingtonUniv. Digestive Disease Research Core (NIH grant P30-DK52574);and the Intramural Research Program of the NIH, NationalCancer Institute, Center for Cancer Research. Microarray dataare available on Array Express (accession no. E-MTAB-1175).

Materials transfer agreements will be required for theacquisition of the Wnt5af/f mice from the National CancerInstitute and the L-WRN cell line from the Washington Univ.Medical School. The data presented in this manuscript aretabulated in the main paper and in the supplementarymaterials.

Supplementary Materialswww.sciencemag.org/cgi/content/full/science.1223821/DC1Materials and MethodsFigs. S1 to S14Tables S1 to S3References (38–53)

25 April 2012; accepted 21 June 2012Published online 6 September 2012;10.1126/science.1223821

Insect Herbivores Drive Real-TimeEcological and Evolutionary Changein Plant PopulationsAnurag A. Agrawal,1,2* Amy P. Hastings,1 Marc T. J. Johnson,3 John L. Maron,4 Juha-Pekka Salminen5

Insect herbivores are hypothesized to be major factors affecting the ecology and evolution ofplants. We tested this prediction by suppressing insects in replicated field populations of anative plant, Oenothera biennis, which reduced seed predation, altered interspecific competitivedynamics, and resulted in rapid evolutionary divergence. Comparative genotyping and phenotyping ofnearly 12,000 O. biennis individuals revealed that in plots protected from insects, resistanceto herbivores declined through time owing to changes in flowering time and lower defensiveellagitannins in fruits, whereas plant competitive ability increased. This independent real-timeevolution of plant resistance and competitive ability in the field resulted from the relaxation ofdirect selective effects of insects on plant defense and through indirect effects due to reducedherbivory on plant competitors.

The ubiquitous consumption of plants byinsect herbivores represents one of thedominant species interactions on Earth

and has been hypothesized to play a strong rolein the diversification of plant species and theirtraits (1–3). The evolution of plant defense hasbeen studied primarily with either a prospectiveor retrospective approach. Single-generation pro-spective approaches measure contemporary naturalselection on resistance traits and make predictionsabout how those traits should evolve given esti-mates of their heritability (4–6). By contrast, ret-rospective studies compare populations or speciesthat have diverged over time (7–9) and make in-ferences about the processes driving evolutionarychange. More generally, temporal studies of natu-ral species interactions that influence evolutionarydynamics remain rare, and few have been exper-

imental (9–12). Thus, we lack experimental fieldstudies that quantify how species interactions in-fluence selection on traits and the evolutionaryresponse to this selection. As such, it is not well

established how rapidly plants adapt to selectionby herbivores, which traits are most important inthe evolution of defense, and whether herbivorydrives predictable parallel changes due to selec-tion across populations.

The ecology and evolution of plant-herbivorerelationships is, in part, governed by the recipro-cal nature of their interaction and the complex-ity of communities. For example, plant traitsand plant community structure are critical fordetermining insect occurrence and attack rates(1–3). Conversely, insects can directly reduce focalplant abundance (13) or indirectly affect plantabundance through changes in the density of co-occurring competitors (14). Thus, the evolution-ary dynamics of a focal plant species might beinfluenced through direct selection on resist-ance traits and/or by indirect selection on traitsinfluencing competitive ability (15, 16). Despitethis likely scenario, our understanding of howspecies interactions influence the evolution ofconstituent community members is surprisinglylimited (17).

We conducted a field study of the selectiveimpact of insects and the evolutionary response

1Department of Ecology and Evolutionary Biology, Cornell Uni-versity, Ithaca, NY 14853, USA. 2Department of Entomol-ogy, Cornell University, Ithaca, NY 14853, USA. 3Department ofBiology, University of Toronto at Mississauga, Mississauga, ONL5L 1C6, Canada. 4Division of Biological Sciences, University ofMontana, Missoula, MT 59812, USA. 5Laboratory of OrganicChemistry and Chemical Biology, Department of Chemistry,University of Turku, FI-20014 Turku, Finland.

*To whom correspondence should be addressed. E-mail:[email protected]

Fig. 1. Fruit loss to seed predator moths (dominated by M. brevivittella) in 18 genotypes of O. biennisin ambient and insect-suppressed plots (data are from the first generation, 2007–2008, mixed-modelanalysis of variance, genotype-by-insecticide interaction: c2 = 57.55, P < 0.001). (Left) an opened fruitwith a M. brevivittella larvae consuming immature seeds; (right) a flowering O. biennis.

www.sciencemag.org SCIENCE VOL 338 5 OCTOBER 2012 113

REPORTS

on

Oct

ober

4, 2

012

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Page 2: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

to this selection in the native forb common eveningprimrose, Oenothera biennis (Onagraceae) (18).We established 16 replicate plots planted with60 individual seedlings of O. biennis, containingequivalent frequencies of 18 uniquely identi-fied genotypes (19). Half of the plots were treatedbiweekly during each growing season withesfenvalerate, a nonsystemic insecticide (insectsuppression treatment, supplementary text 2a);the remaining plots received an equivalentamount of water as a control (ambient insect treat-ment). Because O. biennis is annual or biennial,primarily selfing, and has a genetic system thatsuppresses recombination and segregation of al-leles, we could track changes in frequencies of eachplanted genotype within each replicate populationover multiple generations (19). Plots were not fur-ther weeded or otherwise manipulated, allowingthe natural assembly and early succession of plantspecies with and without insects (supplementarytext 2b). We previously summarized our sam-pling methods, the effects of flowering phenologyon insect attack, and plant life-history evolutionin the ambient insect plots (19). Here, we addressthe effects of experimental insect suppression onecological and evolutionary change in O. biennis,including changes in genotypic composition, de-fensive chemistry, competitive ability, and flow-ering time.

Although there were three native species ofspecialist seed predator moths at our site, fruit

loss in O. biennis was dominated by Momphabrevivittella, which was >5 times as abundant asM. stellella and Schinia florida combined, theother two common specialists. The 18 O. biennisgenotypes varied substantially in their resistanceto these seed predators (Fig. 1), and insect sup-pression created an environment with potential-ly large differences in natural selection. Despiteincreases in O. biennis abundance in the first2 years of the experiment (reaching up to 4000individuals per plot in 2009), recruitment wassubstantially lower in the insect suppressionplots compared to ambient controls (Fig. 2). Thisreduction in O. biennis density coincided witha rise in the abundance of common dandelion,Taraxacum officinale (Asteraceae), an early suc-cessional weed that dominated our plots andwas twice as abundant in plots where insectswere suppressed compared to controls (Fig. 2and fig. S1).

A specialist seed-feeding beetle of T. officinale,Glocianus punctiger, was significantly reduced byinsect suppression, and a highly abundant gener-alist moth caterpillar at the site, Noctua pronuba,showed a preference for T. officinale overO. biennis(supplementary text 2c). Thus, we hypothesizedthat the release of T. officinale from its herbivoresled to competitive suppression of O. biennis. In-deed, populations of O. biennis and T. officinaleshowed a negative correlation at the populationlevel (simple correlation, N = 16, r = –0.523, P =

0.038), and an independent field experiment atthe same site confirmed that T. officinale inhibitsearly establishment of O. biennis (supplementarytext 2d). Because O. biennis requires light for ger-mination, we hypothesize that T. officinale likelyaffects germination as well as early seedling sur-vival. Therefore, insect suppression substantiallyaffected plant competitors in addition to our focalspecies.

The net evolutionary consequence of in-sect suppression was a change in the geneticstructure of O. biennis populations (Fig. 3, re-peated measures canonical correspondence anal-ysis, permutation test, trace = 0.275, F = 8.511,P = 0.020), with the strongest effect observed inthe final year of the experiment (fig. S2). Thisparallel genetic differentiation among our repli-cated insect suppression plots in comparison tocontrol plots (Fig. 3) implies that selection byinsects, and not genetic drift, drove the observed

Fig. 2. Effects of insect suppression on the densities of focal O. biennis (univariate analysis of variance,F1,14 = 8.97, P = 0.010) and a major colonizing competitor (T. officinale, univariate analysis of var-iance, F1,14 = 12.96, P = 0.003) in 2009 (means T SEM); also shown are representative ambient andinsect-suppressed plots when T. officinale was in peak bloom.

Fig. 3. (A) Results from an unconstrained correspon-dence analysis on frequencies of all genotypes in2011. Each dot represents an experimental plot; am-bient and suppressed treatments were coded afterthe analysis. A constrained canonical correspondenceanalysis revealed substantial genetic differentiationbetween treatments (data from 2011, permutationtest, trace = 0.127, F = 2.211, P = 0.003). (B) Therelative frequency (ambient/suppressed) of eachgenotype over time. The dashed red line indicatesthe null hypothesis of no divergence in genotype fre-quencies between treatments; darker lines highlightgenotypes that represented >2% of the plants in2011. Skull and crossbones indicate two genotypesthat went extinct in one treatment. The y axis is on alogarithmic scale to facilitate equal visualization ofunder- and overrepresented genotype frequencies.

5 OCTOBER 2012 VOL 338 SCIENCE www.sciencemag.org114

REPORTS

on

Oct

ober

4, 2

012

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Page 3: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

genotypic evolution. At the end of the experi-ment, 1 of the original 18 genotypes was entire-ly extirpated, and two additional genotypes wererepresented only in insect suppression plots (Fig.3). Only six genotypes consistently maintained afrequency of 2% or higher within the popula-tions. Three novel genotypes, derived from out-crossing events between known parents, eachalso reached >2% of the populations in 2011 (sup-plementary text 2e). Thus, across both treatments,nine genotypes dominated all plots (>2% each)at the end of the experiment. Genotypic even-ness was >50% higher in insect suppression plotscompared to controls (Smith and Wilson’s even-ness index, Evar, univariate analysis of variance,F1,14 =15.037, P = 0.002), suggesting an erosionof genotypic variance in the presence of insectherbivores. Nonetheless, insect suppression af-fected neither genotypic richness (univariate anal-ysis of variance, F1,14 = 0.287, P = 0.600) nordiversity (Simpson’s index, univariate analysis ofvariance, F1,14 = 0.377, P = 0.549).

On the basis of annual estimates of genotypefrequencies in each plot and genotype-specificmeans of M. brevivittella attack measured incontrol plots during 2007, we found that experi-mental suppression of insects resulted in evolu-tion of relaxed plant defense (Fig. 4). We nextdetermined the evolutionary response of traits re-sponsible for this divergence in resistance. Wehad previously shown that the extent of early-season flowering (number of open or senesced

flowers after 50% of the plants had at least asingle open flower) positively correlated withM. brevivittella attack in 2 years (19) (i.e., later-flowering plants suffered less damage). Therewas no difference between treatments in the firstyear of our study (2007), as early season flower-ing was not different among the two treatments(univariate analysis of variance, F1,14 = 0.320,P = 0.581); by 2010, there was a significant shifttoward earlier flowering in plots with suppressedinsects (univariate analysis of variance, F1,14 =6.105, P = 0.027). This effect was even strongerin 2011 (Fig. 4, univariate analysis of variance,F1,14 = 7.025, P = 0.019) and was predictable onthe basis of genotype frequencies and genotype-specific means for early flowering (Fig. 4). Thus,a predicted response to selection was evident inonly three to four generations. We found no dif-ference in annual versus biennial phenotypes ofO. biennis (repeated measures multivariate anal-ysis of variance, F1,14 = 0.935, P = 0.350).

O. biennis produces diverse hydrolyzable tan-nins, including the largest ellagitannins knownfrom any plant species (20), and these com-pounds have very high oxidative capacity, nega-tively affecting insect herbivores (21). One trimer,oenothein A (fig. S3), was the dominant ellagitan-nin in O. biennis fruits (up to 7.8% dry mass). Wepreviously showed that oenothein A is favoredby natural selection in O. biennis leaves (6), andits production in fruits was negatively geneti-cally correlated with M. brevivittella attack in

the current study (simple correlation, N = 17, r =–0.491, P = 0.045) (supplementary text 2f), im-plicating a role of this compound in defense.Despite a genetic correlation between foliar andfruit ellagitannin chemistry (simple correlation,N = 18, r = 0.848, P < 0.001), only fruit chem-istry significantly diverged in our experimentalplots (Fig. 4), suggesting that selection was stron-gest for defense against flower- and fruit-feedingherbivores likeM. brevivittella. A negative geneticcorrelation between early-season flowering andthe production of oenothein A in fruits (simplecorrelation, N = 23, r = –0.67, P < 0.001) is con-sistent with the rapid and joint evolution of thesetraits (i.e., early flowering with low oenothein Ain insect-suppressed plots) (Fig. 4).

Given that insect suppression not only affectedherbivory on O. biennis, but also resulted in stron-ger plant competition through larger T. officinalepopulations, we hypothesized that O. bienniswould evolve greater competitive ability in theabsence of insects. We measured the relative com-petition index (RCI) in total above- and below-ground biomass of different genotypes in anindependent experiment (fig. S4). We then usedthese genotype-specific values of competitiveability to determine if the competitive phenotypesdiverged between control and insect-suppressedplots. Insect suppression resulted in the evolutionof enhanced competitive ability relative to am-bient insect plots (Fig. 4), and because this effectwas not genetically correlated with early-season

Fig. 4. Phenotypic evolution in O. biennis plots. (A) In the presence of seedpredator insects, higher plant resistance evolved [estimated by multiplyinggenotype-specific means measured in control (ambient insect) plots by genotypefrequencies in each plot] (repeated measures multivariate analysis of variance,F1,14 = 6.322, P = 0.025), and this effect did not vary over time (time-by-treatment interaction, repeated measures multivariate analysis of variance,F3,12 = 1.459, P = 0.275). (B) Earlier flowering was observed in the insectsuppression plots (univariate analysis of variance, F1,14 = 7.025, P = 0.019).We confirmed that this was an evolutionary response by multiplying genotype-specific means by genotype frequencies (univariate analysis of variance, F1,14 =4.530, P = 0.050); the former field assessment includes the possible effects of

phenotypic plasticity, whereas the latter is based on genotypic values. (C) Leafdefensive chemistry did not diverge (univariate analysis of variance, F1,14 =1.716, P = 0.211) but fruit chemistry changed (univariate analysis of variance,F1,14 = 8.917, P = 0.010) as predicted by genotypic means and genotypefrequencies. (D) Competitive ability of O. biennis when grown with T. officinale;plants in the insect-suppressed plots evolved a greater ability to maintain above-and belowground biomass (relative competition index) when grown with a com-petitor than when grown alone (repeated measures multivariate analysis ofvariance, F1,14 = 5.189, P = 0.038); this effect did not vary over time (repeatedmeasures multivariate analysis of variance, time-by-treatment interaction (F3,12 =0.072, P = 0.974). All panels show means T SEM; ns, not significant.

www.sciencemag.org SCIENCE VOL 338 5 OCTOBER 2012 115

REPORTS

on

Oct

ober

4, 2

012

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Page 4: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

flowering (simple correlation, N = 23, r = –0.157,P = 0.474), the production of oenothein A (sim-ple correlation, N = 23, r = 0.035, P = 0.871),or resistance to M. brevivittella (simple correla-tion, N = 17, r = –0.255, P = 0.241), responsesto natural selection were unconstrained by genet-ic correlations. The evolution of increased suscep-tibility to herbivores and enhanced competitiveability was jointly favored by the suppression ofinsects.

Concurrent evolution of herbivore resistanceand competitive ability was only partially predict-able from annual measures of natural selection.We correlated trait values against lifetime seedproduction in each year, hypothesizing that wewould observe divergent slopes concordant withthe divergent evolutionary change that we ob-served between our two treatments. Although wefound significantly divergent natural selectionbetween ambient and insect suppression plotsfor the two resistance traits, this effect was ob-served in only 1 of 4 years (fig. S5). Nonetheless,when differences in natural selection were ob-served, they were in the predicted direction onthe basis of the observed evolutionary change(table S1). Thus, annual measures of natural selec-tion predicted from fecundity were not strongpredictors of evolutionary change, which suggeststhat selection on other components of fitness con-tribute to observed evolutionary responses (19).

Our study provides definitive evidence for therapid evolution of insect-mediated plant resist-ance traits in real time, supporting current interestin reciprocal feedbacks between evolutionarychange and ecological dynamics (17, 19, 22, 23).We highlight how biotic environmental factorssimultaneously alter ecological and evolutionarydynamics. Specifically, our finding that multiple

herbivores attacking co-occurring host plants canalter competitive interactions is likely quite gen-eral in natural ecosystems (14, 24, 25), and therapid evolutionary responses observed confirmpredicted interactive effects of herbivory and com-petition on plant evolution (15, 16). Given thatO. biennis is a poor competitor, we predict thatevolutionary shifts in resistance and competitiveability allow this species to persist for longer pe-riods of time during succession.

References and Notes1. P. R. Ehrlich, P. H. Raven, Evolution 18, 586 (1964).2. G. S. Fraenkel, Science 129, 1466 (1959).3. D. J. Futuyma, A. A. Agrawal, Proc. Natl. Acad. Sci. U.S.A.

106, 18054 (2009).4. J. Fornoni, P. L. Valverde, J. Núñez-Farfán, Evolution 58,

1696 (2004).5. R. Mauricio, M. D. Rausher, Evolution 51, 1435 (1997).6. M. T. J. Johnson, A. A. Agrawal, J. L. Maron, J. P. Salminen,

J. Evol. Biol. 22, 1295 (2009).7. R. F. Bode, A. Kessler, J. Ecol. 100, 795 (2012).8. G. A. Desurmont, M. J. Donoghue, W. L. Clement,

A. A. Agrawal, Proc. Natl. Acad. Sci. U.S.A. 108, 7070(2011).

9. A. R. Zangerl, M. R. Berenbaum, Proc. Natl. Acad. Sci. U.S.A.102, 15529 (2005).

10. P. H. Thrall, J. J. Burdon, Science 299, 1735 (2003).11. P. R. Grant, B. R. Grant, Science 296, 707 (2002).12. D. A. Reznick, H. Bryga, J. A. Endler, Nature 346, 357

(1990).13. J. L. Maron, E. Crone, Proc. Biol. Sci. 273, 2575 (2006).14. W. P. Carson, R. B. Root, Ecol. Monogr. 70, 73

(2000).15. A. A. Agrawal, J. A. Lau, P. A. Hambäck, Q. Rev. Biol. 81,

349 (2006).16. M. Uriarte, C. D. Canham, R. B. Root, Ecology 83, 2649

(2002).17. T. W. Schoener, Science 331, 426 (2011).18. Materials and methods are available as supplementary

materials on Science Online.19. A. A. Agrawal, M. T. J. Johnson, A. P. Hastings, J. L. Maron,

Am. Nat. 10.1086/666727 (2012); www.jstor.org/stable/10.1086/666727.

20. M. Karonen, J. Parker, A. Agrawal, J. P. Salminen, RapidCommun. Mass Spectrom. 24, 3151 (2010).

21. R. V. Barbehenn, C. P. Jones, A. E. Hagerman, M. Karonen,J. P. Salminen, J. Chem. Ecol. 32, 2253 (2006).

22. N. G. Hairston Jr., S. P. Ellner, M. A. Geber, T. Yoshida,J. A. Fox, Ecol. Lett. 8, 1114 (2005).

23. M. T. J. Johnson, M. Vellend, J. R. Stinchcombe, Philos. Trans.R. Soc. Lond. B Biol. Sci. 364, 1593 (2009).

24. J. Lubchenco, S. D. Gaines, Annu. Rev. Ecol. Syst. 12, 405(1981).

25. A. A. Agrawal, Ecology 85, 2118 (2004).

Acknowledgments: We thank A. Alfano, F. Chen, S. Cook-Patton,P. Cooper, T. Dodge, N. Durfee, A. Erwin, M. Falise, J. Goldstein,E. Kearney, A. Knight, S. McArt, J. Peters, S. Rasmann, A. Smith,T. Ramsey, E. Reyes, M. Weber, and E. Woods for help withfield work and discussion; T. Zhao, J. Yi, and A. Koivuniemi forassistance with chemical analyses; and J. Good, N. Hairston,S. McArt, T. Züst, and the Plant Interactions Group at CornellUniversity for comments. Molecular work for this study wasconducted in the Evolutionary Genetics Core Facility at CornellUniversity with support from V. Askinazi, S. Bogdanowicz,R. Harrison, E. Larson, and E. Reyes. Chemical analyses on anultraperformance liquid chromatography–mass spectrometrysystem were made possible by a Strategic Research Grant ofUniversity of Turku. This study was primarily supported by aU.S. NSF grant (EAGER DEB-0950231), and for this we thankL. Gough. M.T.J.J. was supported by the Natural Sciences andEngineering Research Council of Canada, the Governmentof Ontario, and a Connaught Early Researcher Award;J.L.M. was supported by NSF grants DEB 0614406 andDEB-0915409; J.-P.S was supported by the Academy of Finland(grant 258992). Data sets for genotype frequencies describedin this paper are available in the DRYAD data repositoryunder http://dx.doi.org/10.5061/dryad.6331s. This paper isdedicated to Paul Feeny and Dick Root.

Supplementary Materialswww.sciencemag.org/cgi/content/full/338/6103/113/DC1Materials and MethodsSupplementary TextFigs. S1 to S5Table S1References (26–32)

12 June 2012; accepted 13 August 201210.1126/science.1225977

Natural Enemies Drive GeographicVariation in Plant DefensesTobias Züst,1*† Christian Heichinger,2 Ueli Grossniklaus,2 Richard Harrington,3Daniel J. Kliebenstein,4,5 Lindsay A. Turnbull1

Plants defend themselves against attack by natural enemies, and these defenses varywidely across populations. However, whether communities of natural enemies area sufficiently potent force to maintain polymorphisms in defensive traits is largelyunknown. Here, we exploit the genetic resources of Arabidopsis thaliana,coupled with 39 years of field data on aphid abundance, to (i) demonstrate thatgeographic patterns in a polymorphic defense locus (GS-ELONG) are stronglycorrelated with changes in the relative abundance of two specialist aphids; and (ii)demonstrate differential selection by the two aphids on GS-ELONG, using amultigeneration selection experiment. We thereby show a causal link between variationin abundance of the two specialist aphids and the geographic pattern at GS-ELONG, whichhighlights the potency of natural enemies as selective forces.

Intraspecific genetic variation is essential inenabling species to respond rapidly to evolu-tionary challenges such as changing environ-

mental conditions (1) or the emergence of novelpests and pathogens (2). This diversity often re-flects the balance between the strength of local

selection and the current and historical levels ofpopulation substructure and gene flow (3, 4).Geographic analyses of genetic variation inseveral plant species have revealed clear geneticsignals of local adaptation (5), caused by differ-ences in the selective regime among locations.These analyses are further supported by reciprocaltransplant experiments, in which home genotypesgenerally outperform those transplanted fromother populations (6, 7). Although the drivers oflocal adaptation often remain unidentified, there

1Institute of Evolutionary Biology and Environmental Studiesand Zürich-Basel Plant Science Center, University of Zürich,Zürich CH-8057, Switzerland. 2Institute of Plant Biology andZürich-Basel Plant Science Center, University of Zürich, ZürichCH-8008, Switzerland. 3Rothamsted Insect Survey, AgroEcol-ogy Department, Rothamsted Research, Harpenden AL5 2JQ,UK. 4Department of Plant Sciences, University of California atDavis, Davis, CA 95616, USA. 5DynaMo Center of Excellence,Department of Plant Biology and Biotechnology, University ofCopenhagen, Copenhagen DK-1871, Denmark.

*To whom correspondence should be addressed. E-mail:[email protected]†Present address: Department of Ecology and EvolutionaryBiology, Cornell University, Ithaca, NY 14853, USA.

5 OCTOBER 2012 VOL 338 SCIENCE www.sciencemag.org116

REPORTS

on

Oct

ober

4, 2

012

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Page 5: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

5 OCTOBER 2012 VOL 338 SCIENCE www.sciencemag.org 50

PERSPECTIVES

How Insect Herbivores Drive

the Evolution of Plants

ECOLOGY

J. Daniel Hare

The presence or absence of particular herbivore

species infl uences which plant genotypes are

favored by natural selection.

ing fi nancial responsibility for their national

malaria control programs. The international

financial crisis and redirection of politi-

cal and fi nancial attention to other pressing

global health problems such as malnutrition

and family planning are potential threats to

the investment needed to sustain and expand

malaria control and research. Past experience

shows the disastrous consequences of letting

up on effective malaria control. It’s time to

reap the benefi ts of the mosquito and parasite

genomes to aggressively tackle this disease.

Otherwise, this highly adaptable parasite and

its vector will continue to outwit us and con-

tinue to kill millions.

References 1. R. A. Holt et al., Science 298, 129 (2002). 2. M. J. Gardner et al., Nature 419, 498 (2002). 3. World Health Organization, World Malaria Report 2011

(World Health Organization, Geneva, 2011). 4. C. J. Murray et al., Lancet 379, 413 (2012). 5. B. J. D’Souza, R. D. Newman, Malar. J. 11, 28 (2012). 6. A. P. Phyo et al., Lancet 379, 1960 (2012).

7. A. Asidi, R. N’Guessan, M. Akogbeto, C. Curtis, M. Row-land, Emerg. Infect. Dis. 18, 1101 (2012).

8. I. H. Cheeseman et al., Science 336, 79 (2012). 9. S. N. Mitchell et al., Proc. Natl. Acad. Sci. U.S.A. 109,

6147 (2012). 10. Medicine for Malaria Venture (www.mmv.org). 11. Innovative Vector Control Consortium (www.ivcc.com). 12. RTS, S Clinical Trials Partnership, N. Engl. J. Med. 365,

1863 (2011). 13. L. Schwartz et al., Malar. J. 11, 11 (2012). 14. P. L. Alonso et al., PLoS Med. 8, e1000406 (2011). 15. R. G. Feachem et al., Lancet 376, 1566 (2010).

10.1126/science.1229177

The most common

biological interac-

tion among species

on Earth is that between

plants and the insects that

feed on them ( 1). Insect

herbivores are thought to

impose natural selection,

which favors resistant plant

genotypes and drives the

evolutionary diversifi cation

of plant species. Two reports

in this issue—by Züst et

al. on page 116 ( 2) and

Agrawal et al. on page 113

( 3)—independently provide

strong empirical evidence

for the rapid evolution of

plant traits that confer resis-

tance to herbivores when

herbivores are present but

for the evolution of traits

that confer increased com-

petitive ability when herbi-

vores are absent.

If resistance to insects

benefits plants, then why

are not all plants now resis-

tant? Several answers to

this long-standing question

have been proposed. One is based on the

assumptions that plant defenses are costly,

that resistant genotypes are favored when the

probability of insect damage is high, but that

these genotypes pay a cost for resistance and

are disfavored when the probability of herbi-

vore attack is low. In its simplest form, this

hypothesis states that chemical resources

obtained by plants can be allocated maxi-

mally either to growth and reproduction or

to defense, but allocation to both processes

cannot be maximized simultaneously ( 4). A

second hypothesis is that defense polymor-

phisms are the result of variation in selection

regimes due to variation in the size and mem-

bership of herbivore communities at differ-

ent locations ( 5). The two studies in this issue

fi nd strong evidence for variation in plant

defense traits in response to particular herbi-

vore species, but only partial support for the

allocation hypothesis.

Züst et al. studied natural populations of

Arabidopsis thaliana in Europe. They com-

pared the geographic variation in the profi les

of glucosinolates [a class of defensive chemi-

cal compounds in the plant family Brassica-

ceae ( 6)] with the distribution of two aphid

species that feed only on brassicaceous

plants. The frequency of the GS-ELONG

gene, which determines the length of the car-

bon side chain on glucosinolate molecules,

varied both with latitude and longitude. The

aphid Brevicoryne brassicae predominated in

areas where the plants produced glucosino-

lates with four-carbon side chains, whereas

the aphid Lipaphis erysimi predominated in

areas where plants produced glucosinolates

with three-carbon side chains.

The authors next used a synthetic plant

population consisting of genotypes that pro-

duce several different combinations of gluco-

sinolates. After only fi ve generations of selec-

tion, feeding by each aphid species selected

for plant genotypes with glucosinolate pro-

fi les identical to those in the fi eld locations

where each aphid species predominated. By

contrast, the plant genotype that predomi-

nated after fi ve generations in a “no aphid”

treatment produced relatively low levels of

glucosinolates. This genotype was, however,

eliminated from populations exposed to all

aphid treatments (see the fi gure).

In a related but independent study,

Agrawal et al. conducted a four-generation Entomology Department, University of California Riverside, Riverside, CA 92521, USA. E-mail: [email protected]

Selection by

species A

Selection by

species B

Synthetic populations of

equally abundant genotypes

Genotypic distributions

after four to five generations

Selection in

the absence

of herbivores

Divergent selection. Both Züst et al. and Agrawal et al. established syn-thetic plant populations consisting of equal proportions of plant geno-types and then observed changes in genotype frequencies after four or fi ve generations of selection. The results show that natural selection favored different genotypes in the absence of herbivores rather than in their pres-ence, and different genotypes in response to different herbivore species.

Published by AAAS

on

Oct

ober

4, 2

012

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Page 6: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

www.sciencemag.org SCIENCE VOL 338 5 OCTOBER 2012 51

PERSPECTIVES

selection experiment in the fi eld using Oeno-

thera biennis, a plant species that consists of several asexual genotypes. Natural selection by ambient populations of insects favored plant genotypes with delayed fl owering rela-tive to genotypes on which insects were sup-pressed. Insects also selected for plant geno-types having different mixtures of ellagitan-nins, a group of hydrolyzable tannins pro-duced by O. biennis implicated in chemical defense ( 7).

In the presence of insects, plant geno-types that produced relatively more ellagi-tannin trimers were favored over plants that produced relatively more ellagitannin dimers, even though all genotypes produced simi-lar quantities of total hydrolyzable tannins. However, suppression of herbivorous insects on Oenothera also suppressed herbivores on competing plant species. As a result, suppres-sion of insect herbivores indirectly selected for Oenothera genotypes that grew larger and competed better with other plant spe-cies (see the fi gure). In contrast to the allo-

cation hypothesis, the traits favoring growth were genetically independent of those favor-ing defense.

Both studies are consistent with ear-lier findings that herbivorous insects can impose natural selection on plant genotypes for defensive chemicals with specifi c struc-tural attributes ( 8). For Arabidopsis, plant resistance to each aphid species was related to the length of the side chain of glucosino-lates and not the total quantities of glucosino-lates. For Oenothera, resistance was associ-ated with selective synthesis of tannin trimers at the expense of tannin dimers. Both studies reinforce concerns that plant defense theories based on differential allocation of resources to broad classes of chemical compounds are of limited value ( 9).

Finally, the studies by Züst et al. and Agrawal et al. illustrate the importance of ecological context in predicting evolution. In the case of Arabidopsis, different special-ist aphid species produced different evolu-tionary outcomes. In Oenothera, the evo-

lutionary outcome would surely have been different if the investigators had eliminated the effects of competition by maintaining their plots free of competing weeds, as other researchers often do. Because levels of her-bivore pressure, the composition of herbi-vore communities, and levels of competition within plant populations all differ widely between habitats, evolution is expected to match plant genotype to habitat ( 8).

References 1. D. J. Futuyma, A. A. Agrawal, Proc. Natl. Acad. Sci. U.S.A.

106, 18054 (2009). 2. T. Züst et al., Science 338, 116 (2012). 3. A. A. Agrawal, A. P. Hastings, M. T. J. Johnson, J. L. Maron,

J.-P. Salminen, Science 338, 113 (2012). 4. D. A. Herms, W. J. Mattson, Q. Rev. Biol. 67, 283 (1992). 5. J. N. Thompson, The Geographic Mosaic of Coevolution

(Univ. of Chicago Press, Chicago, IL, 2005). 6. R. J. Hopkins et al., Annu. Rev. Entomol. 54, 57 (2009). 7. R. V. Barbehenn et al., J. Chem. Ecol. 32, 2253 (2006). 8. A. R. Zangerl, M. R. Berenbaum, Evolution 57, 806

(2003). 9. J. G. Hamilton et al., Ecol. Lett. 4, 86 (2001).

10.1126/science.1228893

Intestinal Wound Healing Requires a Wnt Balancing Act

DEVELOPMENTAL BIOLOGY

Terrence A. Barrett

Control of wound healing via balancing

signaling regulators may have implications

for treating infl ammatory bowel disease.

Infl ammatory bowel disease (IBD) encom-passes a group of disorders of the colon and small intestine including Crohn’s dis-

ease and ulcerative colitis. It affects roughly 396 per 100,000 persons worldwide ( 1) and in the United States is responsible for more than $1.7 billion in overall health care costs. The chronic or recurrent infl ammation associ-ated with this disease causes severe damage to the epithelial lining of the intestine. On page 108 of this issue, Miyoshi et al. ( 2) present a model for wound healing in the intestinal tract that may have clinical relevance to mucosal repair in disorders of intestinal ulceration.

Within the epithelial lining of the small intestine and colon is a gland composed of subunits called intestinal crypts or crypts of Lieberkühn. After mucosal wounding, chan-nels of epithelial cells move under exposed surfaces to begin to recreate normal crypt architecture. Epithelial proliferation at wound

edges is driven by the Wnt signaling pathway, particularly the canonical Wnts, which signal through β-catenin.

Miyoshi et al. have found that creation of new crypts requires release of noncanonical (β-catenin–independent) Wnt5a. Mesenchy-mal Wnt5a-secreting cells are derived from serosal mesothelial (WT1+) stem cells and appear to migrate into areas to participate in tissue repair. Wnt5a lowers epithelial prolif-

eration rates and induces epithelial channel invaginations or clefting under the wounded surface. Miyoshi et al. suggest that Wnt5a potentiates transforming growth factor–β (TGF-β) signaling (via Serpine1 and Smad3) to reduce epithelial proliferation and cause clefting of epithelial channels. Clefting alters the polarization of highly proliferative crypt structures at wound margins, allowing them to branch into new crypt units. The authors did

not detect Wnt5a-mediated inhibi-tion of canonical Wnt signaling as reported by others ( 3). This inhi-bition, when observed, is context dependent ( 3), suggesting that non-canonical Wnt5a may affect canon-ical Wnt signaling in other mucosal healing processes in the intestine.

CR

ED

IT: G

UA

NG

-YU

YA

NG

/NO

RT

HW

EST

ER

N U

NIV

ER

SIT

Y M

ED

ICA

L S

CH

OO

L

Department of Medicine and Microbiology/Immunology, Division of Gastroenterology, Northwestern University Medical School, Chicago, IL 60611, USA. E-mail: [email protected]

1.0 mm

Ulcer

EMEMEMEM

Wnt5a is needed for wound healing. Image of transmural ulcer from a Crohn’s disease resection specimen showing an area of hyperproliferative branching crypts producing an epithelial mono-layer (EM). An area of epithelial channel clefting is highlighted (dark arrow).

Published by AAAS

on

Oct

ober

4, 2

012

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

Page 7: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

www.sciencemag.org/cgi/content/full/338/6103/113/DC1

Supplementary Materials for

Insect Herbivores Drive Real-Time Ecological and Evolutionary Change in Plant Populations

Anurag A. Agrawal,* Amy P. Hastings, Marc T. J. Johnson, John L. Maron, Juha-Pekka Salminen

*To whom correspondence should be addressed. E-mail: [email protected]

Published 5 October 2012, Science 338, 113 (2012)

DOI: 10.1126/science.1225977

This PDF file includes:

Materials and Methods Supplementary Text Figs. S1 to S5 Table S1 References (26–32)

Page 8: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

1. Materials and methods

a. Experimental design b. Population trends for Oenothera biennis and Taraxacum officinale c. Competition between Oenothera biennis and Taraxacum officinale and

impacts of herbivores on T. officinale d. Genotyping e. Phenolic fruit chemistry of Oenothera biennis f. Genotypic variation in competitive ability in Oenothera biennis g. Evidence for divergent natural selection among treatments

2. Supporting online text

a. Effects of insecticide on Oenothera biennis and Taraxacum officinale b. Impacts of insect suppression on other plants c. Herbivores of Taraxacum officinale d. Impacts of removing Taraxacum officinale on Oenothera biennis e. Outcrossing in experimental plots of Oenothera biennis f. Phenolic fruit chemistry of Oenothera biennis

3. Supporting figures

Figure S1. Population trends for O. biennis and T. officinale Figure S2. Correspondence analysis and visualization of genetic differentiation among experimental field plots analyzed each year Figure S3. The structure of ellagitannin trimer oenothein A Figure S4. Heritable variation in plant competition among the 18 genotypes of Oenothera biennis Figure S5. Divergent natural selection among treatments

4. Supporting tables

Table S1. Evidence for divergent natural selection among treatments

2

Page 9: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

1) Materials and Methods a. Experimental design

In May 2007, we established 16 13.5 m2 plots in an old field in Tompkins County, Ithaca, NY. Plots were spaced at least 10 m apart. The field encompassing the plots had homogeneous vegetation and shared an identical land use history (abandoned agriculture). Plots were plowed and sprayed twice with herbicide prior to commencing the experiment, and were protected from deer herbivory by mesh fencing. In the center of each plot, we established a 1x1 m focal area that was planted with experimental O. biennis populations composed of 60 equally spaced individual O. biennis seedlings at the 1st true leaf stage. There was no further weeding or manipulation of the plots, and seedling recruitment in subsequent years occurred throughout the 13.5 m2 plot area. The nearest natural populations of O. biennis were over 1 km from our plots and therefore the potential for outside contamination was low.

We initially planted three individuals from each of 16 genotypes and 6 individuals from each of two additional genotypes (3*16 + 6*2 = 60 total individuals per plot) into each plot, with the location of each genotype randomly assigned. The latter two genotypes were originally thought to be four distinct genotypes, but extensive additional analyses indicate that these were in fact indistinguishable (see details in ref. 19). The 18 genotypes were selected to span the range of phenotypic characteristics from a total of 40 genotypes, which were all grown in a common garden in 2006 (23). The 40 original collections were all from early successional habitats (one per site, separated by at least 0.5 km; average and maximum distance between sites was 12 km and 30 km, respectively) within Tompkins County, New York, USA. O. biennis grows in recently disturbed successional habitats, where populations are typically small (<100 individuals) and discretely dotted across the landscape; our experimental populations attempted to mimic such natural populations. Although our field experiment was also located in Tompkins County, the closest collection site was ≈2 km away. To reduce the impact of maternal environment effects, we grew all genotypes for a generation in a common garden at the same site and used their offspring in the current experiment.

Half of the 16 plots were randomly assigned to the insect suppression treatment. We applied 0.425 % esfenvalerate (Bug-B-Gon, Ortho, EPA Reg. No. 1021-1645-239; 2007-2009, or Asana XL, Dupont, EPA Reg. No. 352-515; 2010-2011) at a rate of 1 U.S. fluid oz. per gallon of water to the vegetation in each of the insect suppression plots every two weeks (April through October) during each of the five growing seasons. Control plots were sprayed with an equal amount of water.

In each year we assessed seed production and seedling recruitment. We counted fruits for all original 60 plants in each focal area (2007 and 2008). However, due to the large episode of recruitment beginning in 2008, we sub-sampled populations in each year between 2008 and 2011 by counting fruits on at least 50 individuals from each plot (and identifying their genotypes as well, see below). We estimated seed production based on genotypically variable mean fruit size (analysis of variance, F17,44=3.61, P<0.001) and the regression slope of fruit size (volume) vs. seed number (Seed number= 2.33(fruit volume) -102.47, r2=0.60, F1,26=38.47, P<0.001) (further details are provided in ref. 19).  

Our experimental populations were reflective of natural populations. O. biennis typically has discrete patchy populations, representing meta-populations ranging in size

3

Page 10: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

from a few individuals to thousands of plants. The genetic diversity of O. biennis populations has been estimated at up to 14 genotypes (26), although this estimate is likely conservative given the nature of the genetic tools employed in 1977. b. Population trends for Oenothera biennis and Taraxacum officinale

T. officinale was the dominant competitor of O. biennis in the spring and early summer throughout the five year experiment (see images in Fig. 2). We estimated the size of each of the 16 O. biennis populations in July of each year (2008-2011) by counting the number of rosettes and flowering plants within each of nine quadrats (625cm2), and then multiplying the mean abundance per m2 by plot area. Seedlings with fewer than four true leaves were not included in our demography estimates. T. officinale was censused annually in May by counting all reproductive stems in all plots. c. Competition between Oenothera biennis and Taraxacum officinale and impacts of herbivores on T. officinale

During October and November 2009, we established 28 small plots (930 cm2) at the same field site as our larger experimental plots in order to estimate competitive impacts of T. officinale on germination and establishment of O. biennis. All plots had naturally occurring T. officinale at a minimum density of three large rosettes per plot, and plots were weeded of all other existing vegetation. In half the plots, we carefully weeded T. officinale (dandelion exclusion plots), cutting stems at soil level to avoid soil disturbance. In late January 2010, we added 160 O. biennis seeds to each plot. Seeds were added to 2 mL of dried sand, mixed, and sprinkled with a salt shaker to allow for even application of seeds within each plot. Beginning in early April 2010, we maintained the dandelion exclusion plots by weeding, as above, once every 1-2 weeks and conducted regular seedling counts through mid-June.

A specialist seed predator weevil Glocianus punctiger was abundant in the developing seed heads of T. officinale in spring (May-June). In 2012, we surveyed 10 randomly selected nearly mature seed heads of T. officinale from each of the 16 experimental plots and opened each to assess the number of G. punctiger larvae.

A choice test was conducted with nearly mature larvae of the abundant generalist Noctua pronuba in May 2010. Larvae were collected from the old field habitat surrounding field plots. Pairs of leaf discs (one each of O. biennis and T. officinale, 2.5 cm2) were offered to individual larvae on a bed of soil in small ventilated containers (600 mL) and damage was assessed after 12 hours (N=24 pairs). d. Genotyping

Approximately 190 plants were genotyped per plot per year by randomly sampling rosettes and flowering plants according to their relative proportion of the total population of recruits within the plot in that year. Plot level genotype frequencies were used in the canonical correspondence analysis. Genotyping methods are given in ref. 19. Briefly, genomic DNA was obtained from freeze-dried leaf tissue using a standard cetyltrimethylammonium bromide (CTAB) extraction protocol (total N=11,731 samples). For each sample, four microsatellite loci (Oenbi2tri2, Oenbi2tri3, Oenbi2tri6 (27), Oenbi 102 (19)) were amplified in a single multiplex PCR reaction (Type-It Microsatellite PCR kit; Qiagen, Valencia, CA). Samples were analyzed on a 3730xl DNA Analyzer (Applied

4

Page 11: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

Biosystems, Foster City, CA) at the Cornell University Life Sciences Core Laboratories Center. Allele sizes were determined using Genemapper v3.5 (Applied Biosystems), with all calls checked by eye. Plants were assigned to a particular genotype only if their alleles at all four loci were representative of that genotype. To evaluate the drivers of trait evolution we conducted several genetic correlations (based on genotype means) between phenotypic traits measured (flowering phenology, leaf and fruit chemistry, competitive ability, resistance to attack, relative competition index). These analyses were conducted with simple Pearson product moment pairwise correlations, with sample sizes that ranged from 16-23 genotypes. Sample size was variable based on available data; late in the experiment, no phenotypic data was available for rare or extinct genotypes. In some cases, we evaluated phenotypes not only on the 18 original genotypes, but the additional 8 outcrossed genotypes that became abundant in the plots. In these analyses, as in others involving analysis of variance, the residuals were checked for normality and homoscedasticity. e. Phenolic fruit chemistry of Oenothera biennis

For 900 samples, taken from the leaves and fruits of the 16 field plots in 2010, we collected samples on dry ice, lyophilized tissues, and ground them to a fine powder. Samples were analyzed by ultra-performance liquid chromatography/diode-array detection coupled to a triple quadrupole mass spectrometer with an electrospray ionization interface (Waters Acquity Xevo TQ) (21, 23). Individual polyphenols were separated into hydrolyzable tannins (HTs) and flavonoids. f. Genotypic variation in competitive ability in Oenothera biennis

In 2011, we conducted a competition experiment between 26 genotypes of Oenothera biennis (18 original genotypes used in the experimental evolution plots plus the eight most common outcrossed genotypes) and T.officinale. All plants were germinated from seed in plug trays containing potting soil, then transplanted at the cotyledon stage into field soil (500 mL plastic pots) either alone or with a single T. officinale) competitor (competing plants were planted 0.8 cm apart). Plants were grown in the absence of herbivores on an open air roof-top patio in full sun (watered and weeded as needed). After 60 days of growth plants were harvested, roots were washed and separated from shoots by species (26 genotypes x 2 treatments x 5 replicates = total n=260), dried at 40-50 °C until reaching a constant weight and then weighed. For each genotype, we calculated competitive ability (RCI) (28) as [( Bo - Bi)/ Bo] X 100, where Bo was the total biomass of the plant at the end of the experiment growing alone and Bi was the total biomass of the same genotype growing with T. officinale. g. Evidence for divergent natural selection among treatments

For each year (2007 and 2008 are combined for the first generation of plants, some of which did not reproduce until the second year), we calculated total genotypic selection separately for early flowering, oenothein A, and competitive ability using analysis of covariance. Genotype means of traits were regressed against mean genotypic fitness (lifetime seed production) (N=18 genotypes for each analysis). In particular, an interaction between traits and our experimental manipulation of insects would indicate divergent selection caused by insect suppression.

5

Page 12: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

2. Supporting online text a. Effects of insecticide on Oenothera biennis and Taraxacum officinale To test for any direct effects of the insecticide (esfenvalerate) on plant germination or growth, we conducted four trials, outlined below, using the same methods for insecticide application as in the main experiment.

Trial #1: Effects on O. biennis shoot and root growth. Forty-two plants were grown in pots on an open air roof-top patio free of insect herbivores. A randomly chosen set of 21 plants were sprayed with insecticide (esfenvalerate, trade name Bug-B-Gon) every two weeks for a total of three applications. All plants were harvested one week after the final application. There was no effect of the insecticide on shoot (mean ± SEM grams dry mass, control 1.20 ± 0.03, insecticide 1.15 ± 0.03, univariate analysis of variance, F1,40=1.18, P=0.28) or root biomass (mean ± SEM grams dry mass, control 0.73 ± 0.04, insecticide 0.71 ± 0.04, univariate analysis of variance, F1,40=0.26, P=0.61).

Trial #2: Effects on O. biennis shoot and root growth controlling for plant genotype. 60 plants from the 18 genotypes from our experiment were randomly assigned to be sprayed with insecticide (esfenvalerate, trade name Asana XL) or water controls every two weeks exactly as in the experimental evolution plots. Plants were 2-3 months old, grown in 500 mL pots with field soil, and were maintained on an open air roof-top patio. After three applications, above and below-ground tissue were harvested and there were no effects of insecticide on shoot (mean ± SEM grams dry mass, control 1.13 ± 0.04, insecticide 1.15 ± 0.04, mixed model analysis of variance F1,40=1.18, P=0.67) or root biomass (mean ± SEM grams dry mass, control 0.88 ± 0.06, insecticide 0.92 ± 0.06, mixed model analysis of variance, F1,40=0.33, P=0.57). Plant genotypes varied substantially in shoot (mixed model analysis of variance, P=0.001) but not root (mixed model analysis of variance, P=0.13) biomass. There was no genotype-by-treatment interaction.

Trial #3: Effects on O. biennis germination. During the spring of 2012, six standard 1020 greenhouse trays were filled with field soil and moistened thoroughly. Three hundred primrose seeds (from a homogenous mixture of five genotypes) were sprinkled on top of the soil of each flat. Flats were watered as needed and germination was permitted to occur for 17 days prior to the first insecticide application (very few seedlings emerged during this period). Esfenvalerate (Asana XL) was applied at one-quarter the rate per area applied to our evolution plots, as a conservative estimate for the amount of chemical that actually reaches the soil surface during each application. Insecticide was applied every two weeks (3 applications in total) and did not affect percent germination (William’s corrected G-test = 1.85, P=0.174).

Trial #4: Effects on T. officinale germination. During the spring of 2012, flats were prepared for dandelion germination, as above (Trial #3) with the exception that 100 seeds were sown in each flat. Again, germination was not affected by treatment (William’s corrected G-test < 0.001, P=1). b. Impacts of insect suppression on other plants

Beginning in 2008, >20 other plant species also recruited into the plots, but most were relatively rare. In 2011 we surveyed the vegetation in each of the plots (species identity and abundance) of those plants that naturally recruited into the plots. We

6

Page 13: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

assessed the impacts of insect suppression treatment on species richness (number of plant species per plot), evenness (Smith and Wilson’s Evar) (29), and diversity (Simpson’s index). Species richness was not impacted by our treatments (mean ± SEM plants per plot in 2011, insect suppression 13.3 ± 1.2, ambient insect controls 13.3 ± 1.2, univariate analysis of variance, F1,14<0.001, P=1.0). In addition to O. biennis and T. officinale, one other species, Solanum carolinense, was significantly impacted by our insect suppression treatment in 2011. Solanum carolinense populations were larger in insect suppression plots (mean ± SEM plants per plot, insect suppression 22.8 ± 2.3, ambient insect controls 5.8 ± 2.3, univariate analysis of variance, F1,14=26.401, P<0.001). c. Herbivores of T. officinale

Weevils (Glocianus punctiger) were half as abundant in the insect suppression plots (mean ± SEM number of larvae per 10 seed heads, control plots 9.50 ± 1.30, insect suppression plots 4.63 ± 1.30, F1,14=7.01, P=0.019).

In 18 out of the 24 choice tests (see Methods above under 1.c), larvae of an abundant generalist moth, Noctua pronuba, consumed more of T. officinale than O. biennis tissue (Wilcoxon signed rank test, Z=-2.779, P=0.005). d. Impacts of removing T. officinale on O. biennis By mid-June, establishment of O. biennis was >3.5 times higher where T. officinale was removed compared to control plots (mean ± SEM number of established O. biennis, control 2.64 ± 0.97, removal plots 9.64 ± 0.97, univariate analysis of variance, F1,26=5.09, P<0.0001). See 1.d above for methods. e. Outcrossing in experimental plots of Oenothera biennis

Despite the fact that O. biennis typically produces seeds that are clonal replicates of the parents (via selfing and a near complete suppression of recombination and free segregation caused by a permanent translocation heterozygote genetic system) (30, 31), a low level of outcrossing can occur. In 2008, 7.3% of the naturally recruited plants in our plots were outcrossed genotypes, and our insecticide treatment did not affect the proportion of outcrossed recruits (analysis of variance, F1,14=0.593, P=0.454). By 2011, 36% of the plants across plots were outcrossed genotypes, either produced through the repeated outcrossing events or the clonal propagation of outcrossed genotypes in 2008. Four parental genotypes of abundant outcrossed genotypes were rare by 2011 (>1% of the final population), suggesting that the success of outcrossed genotypes was not simply dependent on the abundance of parentals.

In a canonical correspondence analysis (CCA) where we only analyzed genotype frequencies of the original genotypes, ambient insect and insect suppression plots still show a high level of genotypic divergence (CCA on 2011 genotype frequencies, Trace = 0.077, F = 2.590, P=0.007) and differentiation in insect resistance (F1,14=22.117, P<0.001). In other words, the outcrossed genotypes were not critical for the evolutionary change observed. f. Phenolic fruit chemistry of Oenothera biennis

Because the main herbivores of O. biennis are seed predators and there is a genetic correlation between leaf and fruit chemistry (simple correlation, N=18, r=0.848, p<0.001), here we focus on fruit chemistry. Among the total phenolics in the fruits, 97%

7

Page 14: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

were hydrolysable tannins (HTs), and 75% of HTs were ellagitannins. The ellagitannins were dominated by a trimer (oenothein A, which comprised 56% of all ellagitannins) and a dimer (oenothein B, 33%).

O. biennis produces ellagitannins that are dimers, trimers, tetramers, pentamers, hexamers, and heptamers. In our dataset, there were 15 possible pairwise genetic correlations among these classes, and all but one were highly statistically significant. Remarkably, dimers were negatively correlated with each of the larger classes (4 of the 5 genetic correlations had r-values between -0.71 and -0.38, all P-values < 0.02), while all larger classes (trimers to heptamers) were significantly positively genetically correlated (r-values ranged from 0.62 to 0.95, all P-values <0.001). We conclude that in environments with ambient insects, natural selection has favored larger ellagitannin oligomers, while the relaxation of selection by herbivores favors the basic dimer structure.

There was no difference among our treatment plots in total phenolics (univariate analysis of variance, F1,14=0.023, P=0.882), total hydrolyzable tannins (univariate analysis of variance, F 1,14<0.001, P=0.977), or total ellagitannins (univariate analysis of variance, F 1,14=0.383, P=0.546). Rather, the suppression of insects appears to have driven evolution of specific fruit ellagitannin oligomers. For example, in the main text we report the evolution of lower oenothein A in fruits in response to insect suppression. This effect appears to come at the expense of oenothein B (which is increased in insect suppressed plots, univariate analysis of variance, F 1,14=16.384, P=0.001). Hence, the ratio of dimers to trimers substantially increased (85%) in our insect suppression plots (univariate analysis of variance, F 1,14=6.636, P=0.022). In this and several previous experiments, we see a negative genetic correlation between plant allocation to dimers and trimers (simple correlation, N=17 , r=-0.693, P<0.001), with the trimers being more strongly genetically correlated with insect resistance (simple correlation of trimers and Mompha attack: N=17, r=-0.491, P=0.045, simple correlation of dimers and Mompha attack: N=17, r=0.275, P=0.286). Furthermore, total concentrations of hydrolysable tannins in fruits were not genetically correlated with Mompha attack (simple correlation, N=17 , r=-0.259, P=0.316). It is expected that specific phenolic compounds will show strong negative effects on insect performance even when total phenolic concentrations show little effect. Most previous studies of plant phenolics measured total phenolic concentrations, or the protein binding capacity of phenolics. Yet, it is now recognized that these measures of phenolic chemistry often show little relationship with resistance to insects (32), even though specific phenolic metabolites can have large negative effects on resistance to insects. In particular, ellagitannins produced by Oenothera have a demonstrated negative impact on insect herbivores (21).

8

Page 15: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

3. Supporting figures

Figure S1. Population trends for Oenothera biennis (A) and Taraxacum officinale (B) in experimental plots. Shown are means ± SEM. N=8 plots per treatment.

9

Page 16: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

Figure S2. Correspondence analysis and visualization of genetic differentiation among experimental field plots analyzed each year. Numbers indicate plot identifications, and polygons were drawn to represent treatments (purple=ambient insects, black=insect suppression) after running an unconstrained model. Statistical analysis was conducted by assigning treatments using canonical correspondence analysis and a permutation test reported for each year in the panels. Repeated measures canonical correspondence analysis, trace=0.275, F=8.511, P=0.020.

10

Page 17: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

Figure S3. The structure of ellagitannin trimer oenothein A.

11

Page 18: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

Figure S4. Heritable variation in response to plant competition among the 18 genotypes of Oenothera biennis from the field experiment. Shown is the combined above and belowground biomass of O. biennis when grown alone or with a major competitor (Taraxacum officinale). Genotypes showed heritable differences in their competitive ability (genotype-by-competition interaction: likelihood ratio test for random factors in mixed model analysis of variance, χ2=3.99, P=0.023).

12

Page 19: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

Figure S5. Divergent natural selection among treatments. Divergent selection was detected for each trait in at least one year and is depicted above: A) selection on early flowering (2009) and B) selection on fruit oenothein A concentrations (2011). In both cases there was a significant interaction term between insect suppression treatment and the trait for predicting lifetime seed production (see table S1).

13

Page 20: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

14

Table S1. Evidence for divergent natural selection among our treatments for traits observed to evolve in the experimental evolution plots. Shown are analysis of covariance; an interaction between treatment (insect suppression) and the trait for predicting fruit production indicates divergent selection. Shown below are F-values, * indicates P<0.05, while † indicates P<0.1. 2007/8 2009 2010 2011 Early flowering (EF) 0.396 1.044 0.027 1.952 Insect suppression (IS) 3.813† 0.945 1.719 0.002 EF x IS 0.058 5.724* 0.457 4.230* Oenothein A (OA) 0.585 1.807 0.9 2.536 Insect suppression (IS) 0.467 1.28 1.884 3.759† OA x IS 0.14 0.1337 0.009 4.000* Competitive ability (CA)

0.090 0.154 0.218 3.201†

Insect suppression (IS) 0.556 0.343 0.026 1.900 CA x IS 0.879 0.307 0.340 0.071

Page 21: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

15

References and Notes

1. P. R. Ehrlich, P. H. Raven, Butterflies and plants: A study in coevolution. Evolution

18, 586 (1964). doi:10.2307/2406212

2. G. S. Fraenkel, The raison d’ĕtre of secondary plant substances. Science 129, 1466

(1959). doi:10.1126/science.129.3361.1466 Medline

3. D. J. Futuyma, A. A. Agrawal, Macroevolution and the biological diversity of plants

and herbivores. Proc. Natl. Acad. Sci. U.S.A. 106, 18054 (2009).

doi:10.1073/pnas.0904106106 Medline

4. J. Fornoni, P. L. Valverde, J. Núñez-Farfán, Population variation in the cost and

benefit of tolerance and resistance against herbivory in Datura stramonium.

Evolution 58, 1696 (2004). Medline

5. R. Mauricio, M. D. Rausher, Experimental manipulation of putative selective agents

provides evidence for the role of natural enemies in the evolution of plant defense.

Evolution 51, 1435 (1997). doi:10.2307/2411196

6. M. T. J. Johnson, A. A. Agrawal, J. L. Maron, J. P. Salminen, Heritability, covariation

and natural selection on 24 traits of common evening primrose (Oenothera

biennis) from a field experiment. J. Evol. Biol. 22, 1295 (2009).

doi:10.1111/j.1420-9101.2009.01747.x Medline

7. R. F. Bode, A. Kessler, Herbivore pressure on goldenrod (Solidago altissima L.,

Asteraceae): Its effects on herbivore resistance and vegetative reproduction. J.

Ecol. 100, 795 (2012). doi:10.1111/j.1365-2745.2012.01958.x

8. G. A. Desurmont, M. J. Donoghue, W. L. Clement, A. A. Agrawal, Evolutionary

history predicts plant defense against an invasive pest. Proc. Natl. Acad. Sci.

U.S.A. 108, 7070 (2011). doi:10.1073/pnas.1102891108 Medline

9. A. R. Zangerl, M. R. Berenbaum, Increase in toxicity of an invasive weed after

reassociation with its coevolved herbivore. Proc. Natl. Acad. Sci. U.S.A. 102,

15529 (2005). doi:10.1073/pnas.0507805102 Medline

10. P. H. Thrall, J. J. Burdon, Evolution of virulence in a plant host-pathogen

metapopulation. Science 299, 1735 (2003). doi:10.1126/science.1080070 Medline

11. P. R. Grant, B. R. Grant, Unpredictable evolution in a 30-year study of Darwin’s

finches. Science 296, 707 (2002). doi:10.1126/science.1070315 Medline

12. D. A. Reznick, H. Bryga, J. A. Endler, Experimentally induced life-history evolution

in a natural-population. Nature 346, 357 (1990). doi:10.1038/346357a0

13. J. L. Maron, E. Crone, Herbivory: Effects on plant abundance, distribution and

population growth. Proc. Biol. Sci. 273, 2575 (2006). doi:10.1098/rspb.2006.3587

Medline

14. W. P. Carson, R. B. Root, Herbivory and plant species coexistence: Community

regulation by an outbreaking phytophagous insect. Ecol. Monogr. 70, 73 (2000).

doi:10.1890/0012-9615(2000)070[0073:HAPSCC]2.0.CO;2

Page 22: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

16

15. A. A. Agrawal, J. A. Lau, P. A. Hambäck, Community heterogeneity and the

evolution of interactions between plants and insect herbivores. Q. Rev. Biol. 81,

349 (2006). doi:10.1086/511529 Medline

16. M. Uriarte, C. D. Canham, R. B. Root, A model of simultaneous evolution of

competitive ability and herbivore resistance in a perennial plant. Ecology 83, 2649

(2002). doi:10.1890/0012-9658(2002)083[2649:AMOSEO]2.0.CO;2

17. T. W. Schoener, The newest synthesis: Understanding the interplay of evolutionary

and ecological dynamics. Science 331, 426 (2011). doi:10.1126/science.1193954

Medline

18. Materials and methods are available as supplementary materials on Science Online.

19. A. A. Agrawal, M. T. J. Johnson, A. P. Hastings, J. L. Maron, A field experiment

demonstrating plant life-history evolution and its eco-evolutionary feedback to

seed predator populations. Am. Nat. 10.1086/666727 (2012);

www.jstor.org/stable/10.1086/666727.

20. M. Karonen, J. Parker, A. Agrawal, J. P. Salminen, First evidence of hexameric and

heptameric ellagitannins in plants detected by liquid chromatography/electrospray

ionisation mass spectrometry. Rapid Commun. Mass Spectrom. 24, 3151 (2010).

doi:10.1002/rcm.4756 Medline

21. R. V. Barbehenn, C. P. Jones, A. E. Hagerman, M. Karonen, J. P. Salminen,

Ellagitannins have greater oxidative activities than condensed tannins and galloyl

glucoses at high pH: Potential impact on caterpillars. J. Chem. Ecol. 32, 2253

(2006). doi:10.1007/s10886-006-9143-7 Medline

22. N. G. Hairston Jr., S. P. Ellner, M. A. Geber, T. Yoshida, J. A. Fox, Rapid evolution

and the convergence of ecological and evolutionary time. Ecol. Lett. 8, 1114

(2005). doi:10.1111/j.1461-0248.2005.00812.x

23. M. T. J. Johnson, M. Vellend, J. R. Stinchcombe, Evolution in plant populations as a

driver of ecological changes in arthropod communities. Philos. Trans. R. Soc.

Lond. B Biol. Sci. 364, 1593 (2009). doi:10.1098/rstb.2008.0334 Medline

24. J. Lubchenco, S. D. Gaines, A unified approach to marine plant-herbivore

interactions. I. Populations and communities. Annu. Rev. Ecol. Syst. 12, 405

(1981). doi:10.1146/annurev.es.12.110181.002201

25. A. A. Agrawal, Resistance and susceptibility of milkweed: Competition, root

herbivory, and plant genetic variation. Ecology 85, 2118 (2004). doi:10.1890/03-

4084

26. E. Steiner, D. A. Levin, Allozyme, SI gene, cytological, and morphological

polymorphisms in a population of Oenothera biennis. Evolution 31, 127 (1977).

doi:10.2307/2407551

27. E. L. Larson, S. M. Bogdanowicz, A. A. Agrawal, M. T. J. Johnson, R. G. Harrison,

Permanent genetic resources: Isolation and characterization of polymorphic

microsatellite loci in common evening primrose (Oenothera biennis). Molecular

Page 23: Ecology & Evolutionary Biology at Cornell - Insect … › agrawal › documents › agrawaletalScience...impact of insects and the evolutionary response 1Department of Ecology and

17

Ecology Resources 8, 434 (2008). doi:10.1111/j.1471-8286.2007.01983.x

Medline

28. D. E. Goldberg, T. Rajaniemi, J. Gurevitch, A. Stewart-Oaten, Empirical approaches

to quantifying interaction intensity: Competition and facilitation along

productivity gradients. Ecology 80, 1118 (1999). doi:10.1890/0012-

9658(1999)080[1118:EATQII]2.0.CO;2

29. B. Smith, J. B. Wilson, A consumer's guide to evenness indices. Oikos 76, 70 (1996).

doi:10.2307/3545749

30. R. E. Cleland, Oenothera: Cytogenetics and Evolution (Academic Press, New York,

1972).

31. U. Rauwolf, H. Golczyk, J. Meurer, R. G. Herrmann, S. Greiner, Molecular marker

systems for Oenothera genetics. Genetics 180, 1289 (2008).

doi:10.1534/genetics.108.091249 Medline

32. J. P. Salminen, M. Karonen, Chemical ecology of tannins and other phenolics: We

need a change in approach. Funct. Ecol. 25, 325 (2011). doi:10.1111/j.1365-

2435.2010.01826.x