Download EVOLUTIONARY DYNAMICS OF MUTUALISM: THE ROLE OF

EVOLUTIONARY DYNAMICS OF MUTUALISM:
THE ROLE OF EXPLOITATION AND COMPETITION
by
Emily Isobel Jones
_____________________
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF ECOLOGY AND EVOLUTIONARY BIOLOGY
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
2009
2
THE UNIVERSITY OF ARIZONA
GRADUATE COLLEGE
As members of the Dissertation Committee, we certify that we have read the dissertation
prepared by Emily Isobel Jones
entitled Evolutionary Dynamics of Mutualism: The Role of Exploitation and Competition
and recommend that it be accepted as fulfilling the dissertation requirement for the
Degree of Doctor of Philosophy
_______________________________________________________________________
Date: 07/17/2009
Judith Bronstein
_______________________________________________________________________
Date: 07/17/2009
Régis Ferrière
_______________________________________________________________________
Date: 07/17/2009
Anna Dornhaus
_______________________________________________________________________
Date: 07/17/2009
John Pepper
Final approval and acceptance of this dissertation is contingent upon the candidate’s
submission of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and
recommend that it be accepted as fulfilling the dissertation requirement.
________________________________________________ Date: 07/17/2009
Dissertation Director: Judith Bronstein
________________________________________________ Date: 07/17/2009
Dissertation Director: Régis Ferrière
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an
advanced degree at the University of Arizona and is deposited in the University Library
to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided
that accurate acknowledgment of source is made. Requests for permission for extended
quotation from or reproduction of this manuscript in whole or in part may be granted by
the author.
SIGNED: Emily Isobel Jones
4
ACKNOWLEDGEMENTS
My last fortune cookie before my dissertation defense read “You are not a failure
because you did not make it; you are a success because you tried.” While I managed to
evade the ill omen of the cookie, it was due to the good luck of having the guidance and
friendship of many people over many years. My advisors, Judie Bronstien and Régis
Ferrière, balanced freedom and high expectations in a way that encouraged me to become
independent and to set the same high standards for myself. I could not have hoped for
better advisors to set the example of what a scientist should be and I deeply appreciate all
the faith they have had in me. I am also very grateful to Anna Dornhaus, who enabled my
experiments on bees and was always available to provide support and motivation, and to
John Pepper, for his thoughtful insights into my work.
I was very fortunate that my undergraduate advisors, Joan Strassmann and David
Queller, invited me to work in their lab at Rice University. I thank Joan and Dave, as well
as Chandra Jack, Natasha Mehdiabadi, Kevin Foster and Tom Platt for making the
experience intellectually stimulating, satisfying and fun, thus confirming that graduate
school was the right choice for me.
During graduate school, I took many opportunities to explore the world beyond
Tucson. I’m thankful to those who were generous with their guidance during those times:
Erika Deinert and Chris Ivey (OTS, Costa Rica), Richard Law (University of York), and
Bill Morris (Duke University), as well as to my co-advisor, Régis Ferrière, who made it
possible for me to spend several months visiting the École Normale Supérieure and the
Université Pierre et Marie Curie in Paris. I also thank the friends who made sure I wasn’t
lonely, even so far from home: Hannah Lewis, Dom Carval, Violaine Cotté, and Sandra
Gillespie.
When back in Tucson, I benefited greatly from both the scientific discussions and
distractions provided by the EEB graduate students, postdocs and visitors, particularly in
the Bronstein and Dornhaus labs. I thank Nick Waser and Mary Price, who provided
invaluable feedback on my research ideas and writing, and Joel and Betsy Wertheim,
Tammy Haselkorn, Jenny Jandt, Latifa Jackson, Michele Lanan, and Maggie Couvillon,
among many others, who provided invaluable friendship. In addition, I am especially
grateful to Ali Ongun for his constant encouragement that never failed to lift my spirits.
Many thanks have to go to Josefa Bleu, Julia Olszewski, Patty Stamper, and Mike
Willey, who assisted in my experiments on honey bees and bumble bees, and to Maren
Friesen, Tammy Haselkorn, Wendy Isner, Jenny Jandt and Michele Lanan, who
volunteered their time to participate in field collections and bee research. I applaud their
courage in the face of stinging bees and stabbing yuccas. My work was also greatly
enhanced by funding from the Department of Ecology and Evolutionary Biology and the
National Science Foundation.
5
DEDICATION
To my parents, Keith and Renata Jones, for their enthusiastic encouragement of my
“scientist-ing” from a very early age.
6
TABLE OF CONTENTS
LIST OF FIGURES……………………………………………………………………….7
ABSTRACT……………………………………………………………………………….8
INTRODUCTION…………………………………………………………………...…..10
PRESENT STUDY……………………………………………………………………....16
REFERENCES…………………………………………………………………………..19
APPENDIX A. ECO-EVOLUTIONARY DYNAMICS OF MUTUALISTS AND
EXPLOITERS……………………………………………………………………23
APPENDIX B. COEVOLUTION AND THE TRANSIENCE OF SPECIES IN AN
EXPLOITED MUTUALISM……………………………………………………82
APPENDIX C. OPTIMAL FORAGING WHEN PREDATION RISK INCREASES
WITH PATCH RESOURCES: AN ANALYSIS OF POLLINATORS AND
AMBUSH PREDATORS………………………………………………………119
APPENDIX D. PREDATION RISK MAKES BEES REJECT REWARDING
FLOWERS AND REDUCE FORAGING ACTIVITY………………………...141
7
LIST OF FIGURES
FIGURE 1………………………………………………………………………..………11
FIGURE 2………………………………………………………………………..………13
8
ABSTRACT
Species exist in complex biotic environments, engaging in a variety of
antagonistic and cooperative interactions. While these interactions are generally
recognized to be context-dependent, varying in outcome in the presence of other
interactions, studies tend to focus on each interaction in isolation. One of the main classes
of species interaction is mutualism, in which partner species gain a net benefit from their
interaction. However, mutualisms are beset by a variety of species that can reduce or
even eliminate the benefits of mutualism through exploitation of and competition for the
resources and services offered by mutualists. These exploiter species potentially threaten
the ecological stability of mutualisms and may alter selection on mutualistic traits. Thus,
understanding the ecology and evolution of mutualisms requires consideration of
interactions with exploiter species. In this dissertation, I investigated the effects of
exploiter species on mutualisms between plants and pollinators using a combination of
eco-evolutionary modeling, optimization theory, and behavioral studies. Using two
adaptive dynamics models of coevolution in exploited pollinating seed parasite
mutualisms, I found that exploiters reduce mutualist densities and select for more
parasitic mutualists. Nevertheless, the models demonstrate that intraspecific competition
for host resources and host defense of those resources restrict the ecological conditions
that lead to extinction of the mutualism, as well as the chances of evolution to extinction.
Thus, exploiters are unlikely to be the threat to mutualisms that has been assumed
previously. On the other hand, in another type of exploitation, exploitative predators may
9
pose a greater threat to investment in mutualism than has been presumed. Through both
optimal foraging theory and behavioral experiments on bumble bees, I found that the risk
from ambush predators can change pollinator floral preferences when predators
preferentially use high-quality flowers to locate their prey. This research suggests that
predators of mutualists may have important top-down effects and that further research is
needed to investigate the effects of exploitative predators on selection on mutualist traits.
10
INTRODUCTION
Explanation of the problem and a review of the literature
Mutualisms are defined by the net benefit partner species derive from the
interaction. Mutualisms play a key role in ecosystem function and the delivery of major
ecological services, including nitrogen fixation, pollination, and protection against
herbivores, predators and parasites (Agrawal et al. 2007; Holland and Bronstein 2008).
Mutualists generally must invest resources into the interaction (Bronstein 2001a);
consequently, individuals of mutualist species that cheat their partners by investing little
or nothing are expected to have an advantage over their more cooperative conspecifics.
However, recently, considerable progress has been made in explaining how mutualists
are able to resist this ‘temptation to defect’ (Frank 1994; Ferrière et al. 2002; West et al.
2002; Foster 2004; Holland et al. 2004; Foster and Kokko 2006; Foster and Wenseleers
2006; Kiers and van der Heijden 2006).
In addition to cheating by individuals of the mutualist species, mutualisms are
vulnerable to the loss of commodities to exploiter species which do not provide any
benefit in return (Bronstein 2001b; Yu 2001). Exploiters appear to be nearly ubiquitous
across mutualisms, including rhizobia that infect plants but fail to fix nitrogen, insects
and hummingbirds that take floral resources without pollinating, and ants that take
rewards from their partners without protecting against herbivores or predators (e.g., Cook
and Rasplus 2003; Tillberg 2004; Dedej and Delaplane 2005; Sachs and Simms 2008).
These exploiter species have the potential to affect the population dynamics of
11
mutualists, as well as selection on mutualist traits. Therefore, instead of assuming
pairwise interactions between two mutualist species or guilds, studies of mutualism need
to incorporate antagonism from exploiters, which can be visualized as in fig. 1.
+
M1
+
M2
_
+
E
Figure 1: Diagram of interactions between mutualist and exploiter species. The net
effect of the interaction between the mutualist species, M1 and M2, is positive for both
species. Exploiters, E, take resources from at least one mutualist species, inflicting a cost
while extracting a benefit. Exploiters may have a variety of direct interactions with the
second mutualist species (e.g., fig. 2), or may interact only indirectly through the first
mutualist species.
If the costs of exploitation outweigh the benefits of mutualism, the mutualism
could collapse, or in the case of obligate mutualism, the mutualist species could go
extinct. Consequently, it has been argued that mutualist-exploiter associations should be
rare in nature for both ecological reasons (e.g., exploiters would invade and drive
mutualists to extinction) and evolutionary reasons (e.g., under the pressure of
exploitation, natural selection would disfavor investments into mutualistic traits)
12
(Axelrod and Hamilton 1981; Bull and Rice 1991; Maynard Smith and Szathmáry 1995;
Doebeli and Knowlton 1998; Herre et al. 1999; Denison 2000; Edwards and Yu 2007;
Johnstone and Bshary 2008). However, as exploiters are in fact common both within and
across mutualisms, as well as persistent across time (Machado et al. 1996; Després and
Jaeger 1999; Pellmyr and Leebens-Mack 1999; Currie et al. 2003), additional study is
required to explain the surprising frequency of mutualist-exploiter associations,
especially incorporating coevolution of exploiters with mutualists.
Furthermore, while exploiters are defined in terms of their parasitic interaction
with one of the mutualist species, the ecological and evolutionary effects of exploiters are
likely to depend on the nature of the interaction between the exploiter and the second
mutualist species. Previous models have assumed that the exploiter competes with the
second mutualist for resources, and that this competition may be asymmetric in favor of
the mutualist (Bronstein et al. 2003; Morris et al. 2003; Wilson et al. 2003; Ferrière et al.
2007). In this case, the interaction effects are as shown in fig. 2a. However, the second
mutualist may not be an antagonist of the exploiter in all mutualisms. Predators and
parasites may act as exploiters of mutualisms when they use one mutualist species to gain
access to their victims, as in the dispersal of anther smut disease by pollinators (Shykoff
and Bucheli 1995) and the ambush of pollinators on flowers by crab spiders (Morse
2007) and predaceous bugs (Greco and Kevan 1995). In such cases, the exploiter gains a
benefit from its interaction with the second mutualist, as shown in fig. 2b.
13
(a)
(b)
+
M1
M2
+
_
_
+
+
M1
+
_
0
E
M2
+
+
0
E
Figure 2: Interaction diagram for mutualisms with (a) traditional exploiters and (b)
exploitative predators. Mutualists, M1 and M2, interact for a mutual net benefit.
Exploiters gain a benefit by (a) taking resources from or (b) through predation of one
mutualist species, M1. However, the effect of the interaction between exploiters and the
second mutualist species depends on the type of exploiter. While M2 can regulate the
density of (a) traditional exploiters, M2 may augment the densities of (b) exploitative
predators.
As can be seen in fig. 2, exploiter population densities can be either regulated or
augmented by the second mutualist species. Meanwhile, the effects of exploiters on the
ecological and evolutionary dynamics of the mutualism are likely to depend on the type
of interaction between the exploiter and the second mutualist. Thus, I propose that in
addition to a closer integration of exploiters into the study of the ecological and
evolutionary dynamics of mutualisms, a distinction is necessary between traditional
exploiters, which compete for mutualist rewards, and exploitative predators, which
benefit from interactions with both mutualists.
14
Explanation of Dissertation Format
In this dissertation, I investigated how exploiters change the behavior, ecological
dynamics and coevolution of mutualists, using two classes of pollination mutualisms as
model systems for a combination of empirical and theoretical studies. These types of
pollinating mutualisms are the group of pollinating seed parasite mutualisms, typified by
yuccas and yucca moths, and generalized pollination, such as by bumble bees. Pollinating
seed parasite mutualisms are highly specialized, obligate interactions in which the service
of pollination by insects is traded for seeds as a resource for the insect larvae (Dufaÿ and
Anstett 2003). Exploiter species are also seed parasites, but do not pollinate. In contrast,
generalized pollination involves many plant and pollinator species, making the
interaction facultative (at least between any pair of plant and pollinator species). Ambush
predators can exploit the mutualism by using flowers to locate their prey, and may even
prefer the more pollinator-attractive and rewarding flowers (Heiling et al. 2004; Heiling
and Herberstein 2004).
The document is presented in four appendices, each formatted as an independent
manuscript. Appendix A describes the role of intraspecific competition in mediating the
coevolutionary dynamics between mutualist and exploiter birth rates in a pollinating seed
parasite mutualism. Appendix B extends the model presented in Appendix A to allow for
coevolution of plant defense against seed predation. It demonstrates that coevolution of
the plant with the mutualist and exploiter decreases the probability of extinction of the
mutualism. Appendix C introduces the idea that ambush predators can not only make
pollination mutualisms more costly for pollinators, but may also have top-down effects
15
on selection on floral traits by changing pollinator floral preferences in favor of less
rewarding flowers. Finally, Appendix D tests the prediction that bumble bees should
switch to less rewarding flowers when predation risk is associated with higher rewards.
16
PRESENT STUDY
The methods, results, and conclusions of this study are presented in the
manuscripts appended to this dissertation. The following is a summary of the most
important findings in this document.
Appendix A models coevolution between pollinating mutualists and nonpollinating exploiters in a pollinating seed parasite mutualism in order to determine the
effects of exploiters on mutualism dynamics. Mutualist and exploiter birth rates are
allowed to coevolve in an adaptive dynamics model with a fixed host plant. Evolution of
mutualist birth rate is found to determine which exploiters can invade successfully.
Subsequent coevolution with an exploiter has a strong, predictable influence on long-term
mutualist-exploiter coexistence, mutualist and exploiter phenotypes, and host, mutualist,
and exploiter abundances. Weak mutualist competition promotes ‘evolutionary purging’
of the exploiter, while weak exploiter competition leads to ‘evolutionary suicide’ of the
system. Thus, the degree of intraspecific competition in each species is critical for the
outcome of these interspecific interactions.
Appendix B moves beyond a pairwise perspective in coevolutionary dynamics to
determine how the community dynamics of an exploited pollinating seed parasite
mutualism are changed by coevolution between these interacting species, as compared to
independent species evolution when the traits of interacting species have fixed values. An
eco-evolutionary dynamic model incorporating evolving plant defense and mutualist and
exploiter oviposition rates is used to predict the ecological outcome after trait evolution
17
and coevolution. The analysis demonstrates that extinction of the exploiter or other all
three species is most frequently a consequence of host over-exploitation by the mutualist,
rather than the exploiter species. However, the chances of extinction can be decreased by
the effectiveness of plant defense and a fecundity-competition trade-off in mutualists and
exploiters. Overall, coevolution in communities is more likely to enable species
persistence than is independent evolution of species.
Appendix C uses optimization of a model of pollinator lifetime foraging gains to
generate predictions of how pollinators should respond to different distributions of
ambush predators on flowers. Lifetime foraging gains are modeled as the sum of the
gains from a series of visits to flowers of the preferred reward level, with the number of
visits a pollinator can make depending on the level of predation risk. Predation risk
depends on pollinator vulnerability to capture, the density of ambush predators, and the
relationship between flower reward level and the probability of harboring an ambush
predator. For pollinator species that are difficult for predators to capture, the optimal
strategy is to visit the most rewarding flowers as long as predator density is low. At
higher predator densities and for pollinators that are more vulnerable to predator capture,
the optimal strategy depends on the predator distribution. In some cases, a wide range of
floral rewards provides near-maximum lifetime resource gains for pollinators, which is
predicted to favor generalization if searching for flowers is costly. In other cases, a low
flower reward level provides the maximum lifetime resource gain and so pollinators are
predicted to specialize on less-rewarding flowers. Therefore, the model suggests that
predators can have qualitatively different top-down effects on plant reproductive success
18
depending on the pollinator species, the density of predators, and the distribution of
predators across flower reward levels.
Appendix D tests the response of bumble bee foraging preferences to predation
risk in relation to flower reward level. In the laboratory, foraging bumble bees were
subjected to simulated predator attacks on artificial feeders that differed in sugar
concentration and color. In response to a simulated predator attack, bees foraging on a
low-reward artificial flower were more likely to cease foraging for at least an hour or to
extend the time between flower visits. Bees attacked on a high-reward artificial flower
were more likely to increase visitation of low-reward artificial flowers. Forager body
size, which is thought to affect vulnerability to capture by predators, did not have an
effect on response to an attack. This study confirms that predation risk can alter pollinator
foraging behavior in ways that influence the number and reward level of flowers that are
visited.
19
REFERENCES
Agrawal, A. A., D. D. Ackerly, F. Adler, A. E. Arnold, C. Caceres, D. F. Doak, E. Post,
P. J. Hudson, J. Maron, K. A. Mooney, M. Power, D. Schemske, J. Stachowicz, S.
Strauss, M. G. Turner, and E. Werner. 2007. Filling key gaps in population and
community ecology. Frontiers in Ecology and the Environment 5:145-152.
Axelrod, R., and W. D. Hamilton. 1981. The Evolution of Cooperation. Science
211:1390-1396.
Bronstein, J. L. 2001a. The costs of mutualism. American Zoologist 41:825-839.
—. 2001b. The exploitation of mutualisms. Ecology Letters 4:277-287.
Bronstein, J. L., W. G. Wilson, and W. E. Morris. 2003. Ecological dynamics of
mutualist/antagonist communities. American Naturalist 162:S24-S39.
Bull, J. J., and W. R. Rice. 1991. Distinguishing Mechanisms for the Evolution of
Cooperation. Journal of Theoretical Biology 149:63-74.
Cook, J. M., and J. Y. Rasplus. 2003. Mutualists with attitude: coevolving fig wasps and
figs. Trends in Ecology & Evolution 18:241-248.
Currie, C. R., B. Wong, A. E. Stuart, T. R. Schultz, S. A. Rehner, U. G. Mueller, G. H.
Sung, J. W. Spatafora, and N. A. Straus. 2003. Ancient tripartite coevolution in
the attine ant-microbe symbiosis. Science 299:386-388.
Dedej, S., and K. Delaplane. 2005. Net energetic advantage drives honey bees (Apis
mellifera L) to nectar larceny in Vaccinium ashei Reade. Behavioral Ecology and
Sociobiology 57:398-403.
Denison, R. F. 2000. Legume sanctions and the evolution of symbiotic cooperation by
rhizobia. American Naturalist 156:567-576.
20
Després, L., and N. Jaeger. 1999. Evolution of oviposition strategies and speciation in the
globeflower flies Chiastocheta spp. (Anthomyiidae). Journal of Evolutionary
Biology 12:822-831.
Doebeli, M., and N. Knowlton. 1998. The evolution of interspecific mutualisms.
Proceedings of the National Academy of Sciences of the United States of America
95:8676-8680.
Dufaÿ, M., and M. C. Anstett. 2003. Conflicts between plants and pollinators that
reproduce within inflorescences: evolutionary variations on a theme. Oikos 100:314.
Edwards, D. P., and D. W. Yu. 2007. The roles of sensory traps in the origin,
maintenance, and breakdown of mutualism. Behavioral Ecology and Sociobiology
61:1321-1327.
Ferrière, R., J. L. Bronstein, S. Rinaldi, R. Law, and M. Gauduchon. 2002. Cheating and
the evolutionary stability of mutualisms. Proceedings of the Royal Society of
London Series B-Biological Sciences 269:773-780.
Ferrière, R., M. Gauduchon, and J. L. Bronstein. 2007. Evolution and persistence of
obligate mutualists and exploiters: competition for partners and evolutionary
immunization. Ecology Letters 10:115-126.
Foster, K. R. 2004. Diminishing returns in social evolution: the not-so-tragic commons.
Journal of Evolutionary Biology 17:1058-1072.
Foster, K. R., and H. Kokko. 2006. Cheating can stabilize cooperation in mutualisms.
Proceedings of the Royal Society B-Biological Sciences 273:2233-2239.
Foster, K. R., and T. Wenseleers. 2006. A general model for the evolution of mutualisms.
Journal of Evolutionary Biology 19:1283-1293.
Frank, S. A. 1994. Genetics of Mutualism - the Evolution of Altruism between Species.
Journal of Theoretical Biology 170:393-400.
21
Greco, C. F., and P. G. Kevan. 1995. Patch choice in the anthophilous ambush predator
Phymata americana: Improvement by switching hunting sites as part of the initial
choice. Canadian Journal of Zoology-Revue Canadienne De Zoologie 73:19121917.
Heiling, A. M., K. Cheng, and M. E. Herberstein. 2004. Exploitation of floral signals by
crab spiders (Thomisus spectabilis, Thomisidae). Behavioral Ecology 15:321-326.
Heiling, A. M., and M. E. Herberstein. 2004. Floral quality signals lure pollinators and
their predators. Annales Zoologici Fennici 41:421-428.
Herre, E. A., N. Knowlton, U. G. Mueller, and S. A. Rehner. 1999. The evolution of
mutualisms: exploring the paths between conflict and cooperation. Trends in
Ecology & Evolution 14:49-53.
Holland, J. N., and J. L. Bronstein. 2008. Mutualism, Pages 2485-2491 in S. E.
Jørgensen, and B. D. Fath, eds. Encyclopedia of Ecology. Oxford, Elsevier.
Holland, J. N., D. L. DeAngelis, and S. T. Schultz. 2004. Evolutionary stability of
mutualism: interspecific population regulation as an evolutionarily stable strategy.
Proceedings of the Royal Society of London Series B-Biological Sciences
271:1807-1814.
Johnstone, R. A., and R. Bshary. 2008. Mutualism, market effects and partner control.
Journal of Evolutionary Biology 21:879-888.
Kiers, E. T., and M. G. A. van der Heijden. 2006. Mutualistic stability in the arbuscular
mycorrhizal symbiosis: Exploring hypotheses of evolutionary cooperation.
Ecology 87:1627-1636.
Machado, C. A., E. A. Herre, S. McCafferty, and E. Bermingham. 1996. Molecular
phylogenies of fig pollinating and non-pollinating wasps and the implications for
the origin and evolution of the fig-fig wasp mutualism. Journal of Biogeography
23:531-542.
Maynard Smith, J., and E. Szathmáry. 1995, The Major Transitions in Evolution. Oxford,
W.H. Freeman Spektrum.
22
Morris, W. F., J. L. Bronstein, and W. G. Wilson. 2003. Three-way coexistence in
obligate mutualist-exploiter interactions: The potential role of competition.
American Naturalist 161:860-875.
Morse, D. H. 2007, Predator upon a Flower: Life history and fitness in a crab spider.
Cambridge, MA, Harvard University Press.
Pellmyr, O., and J. Leebens-Mack. 1999. Forty million years of mutualism: Evidence for
Eocene origin of the yucca-yucca moth association. Proceedings of the National
Academy of Sciences of the United States of America 96:9178-9183.
Sachs, J. L., and E. L. Simms. 2008. The origins of uncooperative rhizobia. Oikos
117:961-966.
Shykoff, J. A., and E. Bucheli. 1995. Pollinator Visitation Patterns, Floral Rewards and
the Probability of Transmission of Microbotryum violaceum, a Venereal-Disease
of Plants. Journal of Ecology 83:189-198.
Tillberg, C. V. 2004. Friend or foe? A behavioral and stable isotopic investigation of an
ant-plant symbiosis. Oecologia 140:506-515.
West, S. A., E. T. Kiers, E. L. Simms, and R. F. Denison. 2002. Sanctions and mutualism
stability: why do rhizobia fix nitrogen? Proceedings of the Royal Society of
London Series B-Biological Sciences 269:685-694.
Wilson, W. G., W. F. Morris, and J. L. Bronstein. 2003. Coexistence of mutualists and
exploiters on spatial landscapes. Ecological Monographs 73:397-413.
Yu, D. W. 2001. Parasites of mutualisms. Biological Journal of the Linnean Society
72:529-546.
23
APPENDIX A
EVO-EVOLUTIONARY DYNAMICS OF MUTUALISTS AND EXPLOITERS
In press at The American Naturalist
24
Eco-Evolutionary Dynamics of Mutualists and Exploiters
Emily I. Jones1,*, Régis Ferrière1,2,3, Judith L. Bronstein1
1. Department of Ecology and Evolutionary Biology, University of Arizona, Tucson,
Arizona 85721;
2. Laboratoire Ecologie & Evolution, UMR 7625 CNRS/UPMC/ENS, École Normale
Supérieure, 46 rue d'Ulm, 75005 Paris, France;
3. CEREEP – ECOTRON Ile De France, UMS 3194 CNRS/ENS, Rue du Chateau, 77140
Saint-Pierre-les-Nemours, France
* Corresponding author; e-mail: [email protected].
Phone: (520) 621-3534
Fax: (520) 621-9190
Running head: Mutualist-Exploiter Coevolution
Keywords: mutualism, exploitation, competition, coexistence, adaptive dynamics,
coevolution.
25
ABSTRACT: Following upon growing recognition of exploiters as a prominent and
enduring feature of many mutualisms, there is a need to understand the ecological and
evolutionary dynamics of mutualisms in the context of exploitation. Here, we model
coevolution between mutualist and exploiter birth rates, using an obligate pollinating seed
parasite mutualism associated with a non-pollinating exploiter as a reference system. In
this system, mutualist and exploiter larvae parasitize the host plant, competing for and
consuming seeds. Evolution of the mutualist determines which exploiters can invade
successfully. Subsequent coevolution with an exploiter has a strong, predictable influence
on mutualist-exploiter coexistence, mutualist and exploiter phenotypes, and species
abundances. Weak mutualist competition promotes ‘evolutionary purging’ of the
exploiter, while weak exploiter competition leads to ‘evolutionary suicide’ of the system.
When stable, long-term coexistence occurs, we identify two main ‘trait-abundance
syndromes’ that have three novel implications. (1) Persistent, highly parasitic exploiters
can be favored by coevolution. (2) Even then, the density of coevolved mutualists can be
high. (3) Low plant density results primarily from the evolution of mutualist, not
exploiter, birth rate and density. To evaluate these predictions, studies are needed that
identify and compare populations with and without exploiters and that compare lifehistory traits of mutualists and exploiters.
26
Introduction
Mutualisms are increasingly recognized as playing a key role in ecosystem function and
the delivery of major ecological services (Agrawal et al. 2007; Holland and Bronstein
2008). However, many if not most mutualisms are beset by other, non-mutualistic
species. These ‘exploiters’ take advantage of the commodities provided by one or more
of the mutualists without providing any commodity in return (Bronstein 2001; Yu 2001).
Well-studied examples of exploiters include bees, ants, and hummingbirds that feed on
floral nectar but that do not pick up or deposit pollen; non-pollinating seed parasites of
obligately insect-pollinated yuccas and figs; ants that consume food rewards produced by
plants and other insects without defending their associates; and Rhizobium bacteria that
fail to fix nitrogen for their legume hosts (e.g., Cook and Rasplus 2003; Dedej and
Delaplane 2005; Sachs and Simms 2008; Tillberg 2004). In addition to being numerically
abundant and taxonomically diverse, there is evidence that exploiters are evolutionarily
persistent in both facultative and obligate mutualisms (Currie et al. 2003; Després and
Jaeger 1999; Machado et al. 1996; Pellmyr and Leebens-Mack 1999).
This empirical evidence for the ubiquity of exploitation, however, clashes with
conceptual expectation. The net effect of exploiters can be distinctly detrimental to one or
both mutualists. Consequently, it has been argued that mutualist-exploiter associations
should be rare in nature for both ecological reasons (e.g., exploiters would invade and
drive mutualists to extinction) and evolutionary reasons (e.g., under the pressure of
exploitation, natural selection would disfavor investments into mutualistic traits)
27
(Axelrod and Hamilton 1981; Bull and Rice 1991; Denison 2000; Doebeli and Knowlton
1998; Edwards and Yu 2007; Herre et al. 1999; Johnstone and Bshary 2008; Maynard
Smith and Szathmáry 1995).
While progress has been made in explaining how mutualists are able to resist the
‘temptation to defect’ (Ferrière et al. 2002; Foster 2004; Foster and Kokko 2006; Foster
and Wenseleers 2006; Frank 1994; Holland et al. 2004; Kiers and van der Heijden 2006;
West et al. 2002), there is a comparative paucity of theory explaining the evolutionary
persistence of mutualisms in the presence of separate exploiter species. Nonetheless,
several theoretical steps have been taken towards solving this ecological and evolutionary
conundrum. The mathematical analyses of Law et al. (2001), Ferrière et al. (2002),
Bronstein et al. (2003), Morris et al. (2003), and others have identified conditions based
on trait values and on ecological details under which exploited mutualisms are
ecologically viable. An emerging principle from these studies is that intra- and
interspecific competition for commodities provided by the partner species is key for
mutualists and exploiters to coexist.
The ecological interactions of mutualists and exploiters are expected to generate
strong selective pressures on the mutualists (Ferrière et al. 2007), and to cause reciprocal
selection on the exploiters. In turn, the coevolutionary process is likely to play a
significant role in shaping the ecological dynamics of the system (Fussmann et al. 2007;
Thompson 1998). As competition, both intra- and interspecific, is a cornerstone of natural
selection, it is likely to propagate feedbacks between the ecological and evolutionary
dynamics of mutualist-exploiter systems. Thus, our understanding of the ecology and
28
evolution of exploited mutualisms requires an integrated theory of eco-evolutionary
feedbacks mediated by intra- and interspecific competition.
Pollinating seed parasite mutualisms provide an ideal model system for an ecoevolutionary theory of exploited mutualisms. These are obligate interactions between
flowering plants (the host), pollinating seed parasites (the mutualist), and non-pollinating
seed parasites (the exploiter) (Dufaÿ and Anstett 2003). The best-known examples
involve figs and yuccas, in which seed production by the plant is dependent on the
pollination services of (with few exceptions) a single species of insect (fig wasp, yucca
moth) that both pollinates and deposits offspring on its host plant. In addition, both figs
and yuccas are exploited by one or more species of non-pollinators, each of which is
associated with one or at most a few species of plant hosts. Exploiters depend on the
pollinators either because inflorescences that have not been pollinated are aborted by the
plant along with developing exploiter larvae (in figs), or because pollinators fertilize the
seeds on which exploiter offspring feed (in yuccas). Pollinating seed parasite mutualisms
are among the best-studied examples of exploitation. Most of the currently available
phylogenetic evidence for coevolution in mutualist-exploiter systems has been drawn
from these systems (Després and Jaeger 1999; Machado et al. 1996; Pellmyr et al. 1996),
and enough is known about their natural history to permit realistic features of their
biology to be built into mathematical population models. The species-specific, obligate
nature of the interaction for all three species means that, unlike for more diffuse
interactions, models that include only three species can account for significant features of
the ecological and evolutionary dynamics.
29
In pollinating seed parasite mutualisms, pollinator and exploiter larvae develop at
the expense of seeds within the same fruit or inflorescence; as a result, larvae of both
species may experience exploitation competition and interference competition from both
conspecific and heterospecific larvae (Jaeger et al. 2001; Marr et al. 2001; Peng et al.
2005). The role of such intra- and interspecific competition in promoting three-species
ecological viability has been demonstrated by ecological models (Bronstein et al. 2003;
Morris et al. 2003; Wilson et al. 2003). In these models, the critical parameters for
species coexistence are the birth rates of mutualists and exploiters, parameters that also
determine the number of larvae potentially competing for seeds within a fruit or
inflorescence. We build on this body of ecological theory by treating the mutualist and
exploiter birth rates as adaptive phenotypic traits subject to heritable variation and
selection generated by ecological interactions. Given the intensity of intra- and
interspecific competition, mutualist and exploiter birth rates determine the ecological
state of the system, and thus the selection pressures acting on heritable variation in the
birth rates. As the birth rates respond to selection and evolve, the ecology of the system
changes, closing the eco-evolutionary feedback loop. The analysis of this feedback loop
enables us to answer three general questions: How does the evolutionary history of the
mutualism influence its ability to sustain invasion by exploiters and form an ecologically
viable interaction with the exploiters? How and by how much does the coevolutionary
process change the phenotypes of the mutualist and the exploiter? What are the
consequences of mutualists’ and exploiters’ coevolution for their ecological stability and
population dynamics?
30
Methods
Morris et al. (2003) presented an ecological model of an obligate pollinating seed
parasite mutualism exploited by a non-pollinating seed parasite species and found the
conditions, with respect to pollinator and exploiter birth rates, under which coexistence of
all three species is possible. We extended this ecological model by expanding the analysis
of the role of competition in the population dynamics. We then embedded the ecological
model into an evolutionary framework in order to analyze the patterns and consequences
of coevolution between the pollinator and exploiter birth rates.
Our analysis rests on three important assumptions. First, mutations affecting
pollinator and exploiter birth rates are rare and have small effects, nevertheless, all trait
values are accessible through mutation. This assumption allows us to link population
dynamics and trait evolution by using the adaptive dynamics framework (Champagnat et
al. 2006; Dieckmann and Law 1996; Metz et al. 1992). Second, the evolution of birth
rates is constrained by the fundamental tradeoff between egg number and egg
provisioning (e.g., Smith and Fretwell 1974); as a consequence, larvae produced at a
higher (birth) rate are assumed to suffer a cost during competition with their conspecifics,
due, e.g., to their smaller size (e.g., Fox et al. 2001). Third, because of the difference in
pollinator and exploiter phenology (Kerdelhue and Rasplus 1996; Marr et al. 2001;
Pellmyr et al. 1996), pollinator larvae enjoy a competitive advantage over exploiter larvae
31
that is independent of the species’ birth rates (Morris et al. 2003). Thus, the pattern of
interspecific competition is fixed, and our analysis focuses on eco-evolutionary dynamics
with respect to variation in pollinator and exploiter intraspecific competition.
Based on these assumptions, we construct a two-species (plant and pollinator)
model to study the evolution of pollinator birth rate in the absence of an exploiter; by
doing so we identify the likely state of the mutualism prior to exploiter invasion. Next,
we extend this model to include ecological interaction and coevolution with the exploiter.
This three-species models allows us to predict (i) the phenotypes of exploiters that could
invade and subsequently coevolve with the pollinator, (ii) the consequences of the
pollinator-exploiter coevolutionary dynamics for long-term coexistence, and (iii) the
phenotypic and population outcomes of the coevolutionary process; and to investigate
how these predictions are affected by the intensities of pollinator and exploiter
intraspecific competition.
Eco-Evolutionary Model in the Absence of an Exploiter
In the absence of an exploiter, the plant/pollinator interaction involves pollination and
oviposition by one pollinator species visiting the flowers of one plant species. Larvae
develop at the expense of seeds. The ecological dynamics of plant and pollinator are
given by the per capita growth equations (1):
(1a)
1 dP
= bP M(1− M)(1− P) −1,
P dt
32
(1b)
1 dM
= bM P(1− µM) − dM .
M dt
P and M are the population densities of the plant and pollinator (mutualist), respectively;
bP, bM, dP, and dM are the intrinsic birth and death rates; and µ is the intensity of mutualist
intraspecific competition. The densities and birth rates incorporate sub-parameters such
as numbers of sites for plants, potential oviposition sites, sites searched by pollinators,
and the proportion of pollination events that are combined with oviposition. Furthermore,
time and the pollinator death rate have been rescaled according to the plant death rate
(see Morris et al. 2003). In the dimensionless version given above, the per capita
population growth of plants is determined by the intrinsic birth rate (bP), availability of
pollinators (M term), seed destruction by pollinator larvae ((1 – M) term), competition
between plants for space ((1 – P) term), and death (at normalized rate 1). Since
competition among plants is directly related to the proportion of free space in the
environment, no coefficient is used to modify the intensity of competition. The per capita
population growth of pollinators is determined by the intrinsic birth rate (bM), availability
of plants (P term), competition between larvae for seeds ((1 – µM) term), and death (at
rate dM).
Equations (1) were solved to find the equilibrium densities of the two species. All
cases had a non-trivial two-species equilibrium. At this equilibrium, the invasion fitness
(i.e., population growth rate when rare, Metz et al. 1992) of a pollinator with a mutant
intrinsic birth rate, bMmut , competing with the ‘resident’ population of pollinators with the
resident type, bM, is:
33
(2)
ˆ ]− d ,
W M (bMmut , bM ) = bMmut Pˆ [1− (µ + µ'(bMmut − bM )) M
M
ˆ are the densities of plants and pollinators at the ecological equilibrium
where Pˆ and M
given by equations (1), which are also functions of the trait value bM. Equation (2)
incorporates the assumed tradeoff between fecundity and competitive ability. The
competitive effect of any resident-type individual on a focal individual is measured by µ
if the focal individual is of the resident type, and by (µ + µ'(bMmut − bM )) if the focal
individual is a mutant. Thus, if the mutant’s birth rate is larger than the resident’s, the
competitive effect experienced by the mutant is increased by an amount taken to be
proportional (for simplicity) to the difference between birth rates; if the mutant birth rate
is smaller than the resident’s, the competitive effect experienced by the mutant is reduced
by that amount. The proportionality coefficient, µ’, measures the slope of the fecunditycompetition tradeoff that constrains the evolution of the mutualist birth rate; it is assumed
to be species-specific and constant.
Values of bM where the selection gradient
∂WM
(bM ,bM ) is equal to zero are
∂bMmut
called evolutionary singular states (Geritz et al. 1997). An evolutionary singular state at
which WM is a maximum is an evolutionary stable state (ESS). At the ESS, the resident
population cannot be invaded by any (nearby) mutant. Evolutionary stability (as opposed
to evolutionary branching) and evolutionary convergence were systematically tested by
constructing and inspecting pairwise invasibility plots (Geritz et al. 1997).
Eco-Evolutionary Model in the Presence of an Exploiter
34
Once the system has been invaded by an exploiter species, both pollinator and exploiter
larvae destroy seeds, but the exploiter does not contribute any additional pollination. The
ecological dynamics of the three-species community are given by equations (3):
(3a)
1 dP
= bP M (1 − M )(1 − E )(1 − P ) − 1 ,
P dt
(3b)
1 dM
= bM P (1 − µM ) − d M ,
M dt
(3c)
1 dE
= bE PM (1 − M )(1 − αE ) − d E ,
E dt
where the symbols for plants and mutualists are the same as in equations (1); E denotes
the exploiter population density and α is the intensity of exploiter intraspecific
competition. The per capita population growth of exploiters is determined by the intrinsic
birth rate (bE), availability of pollinated plants (PM term), interspecific competition with
pollinator larvae for seeds ((1 – M) term), intraspecific competition between exploiter
larvae for seeds ((1 – αE) term), and death (at rate dE). Plants suffer additional seed loss
from exploiter larvae ((1 – E) term), compared to pollinator seed destruction alone.
Importantly, the exploiter species requires both plant and pollinator for its own survival.
The model incorporates our assumption about interspecific competitive superiority of
mutualists over exploiters: the pollinator growth rate is not directly affected by the
presence of an exploiter (equation (3b)), whereas the exploiter growth rate is limited by
interspecific competition with mutualists (equation (3c)). The interspecific competition
coefficient is normalized to 1 (hence the term (1 – M) in equation (3c)), which sets the
35
scale for the coefficients of intraspecific competition in mutualists (µ) and exploiters (α).
The ecological model specified by equations (3) has three potential types of
solution: coexistence of all three species, exclusion of the exploiter, and extinction of all
three species. Each outcome corresponds to a region in the space of the evolving traits
bM and bE. In all combinations of positive intraspecific competition intensities µ and α
explored, coexistence was possible over some region of pollinator and exploiter birth
rates and the coevolution of pollinator and exploiter species could be investigated. Again,
the intrinsic birth rates were allowed to evolve according to the invasion fitnesses:
(4a)
ˆ ]− d ,
W M (bMmut , bM ) = bMmut Pˆ [1− (µ + µ'(bMmut − bM )) M
M
(4b)
ˆ (1− M
ˆ )[1− (α + α '(b mut − b )) Eˆ ] − d .
W E (bEmut , bE ) = bEmut Pˆ M
E
E
E
ˆ are
Equation (4a) is identical to equation (2), except here the densities Pˆ and M
evaluated at the three-species ecological equilibrium given by equations (3), and thus
depend on both bM and bE. Invasion fitness for a mutant exploiter with trait value bEmut is
affected by the same fecundity-competition tradeoff, this time with the tradeoff slope α’.
Eö is the density of the exploiter at the ecological equilibrium given by equations (3). In
the two-dimensional trait space bM, bE, the curves where the selection gradients
∂WM
∂W
(bM ,bM ) and mutE (bE , bE ) become zero define the selective isoclines of the
mut
∂bE
∂bM
coevolutionary system. Points of intersection of the isoclines define the system’s
‘coevolutionary singularities’; they are potential coevolutionary stable states (coESSs),
that is, combinations of bM and bE that are evolutionarily stable for both the mutualist and
the exploiter. The attractivity and stability of the coevolutionary singularities can be
36
studied by constructing the ‘canonical equations’ of the trait dynamics based on the
selection gradients (Dieckmann and Law 1996, see Online Appendix).
Results
As described by Morris et al. (2003), the population dynamics of the three-species
community given by equations (3) depend on the birth rates of the pollinator and
exploiter, bM and bE. Three-species extinction occurs at low bM values, three-species
coexistence at intermediate bM and bE values, and exclusion of the exploiter at low bE
values (fig. 1). Morris et al. (2003) demonstrated that the equilibria are stable except in a
narrow band of the three-species coexistence region bordering the extinction region, in
which there is oscillation around an unstable equilibrium.
Varying the intensities of pollinator and exploiter intraspecific competition, µ
and α, changes the range of pollinator and exploiter birth rates that lead to each type of
ecological outcome (fig. 1). In agreement with Morris et al.’s (2003) conclusions, we find
that the three-species coexistence region widens into the exclusion region with increases
in pollinator intraspecific competition, µ (e.g., compare fig. 1A, B and C) and into the
extinction region with increases in exploiter intraspecific competition, α (e.g., compare
fig. 1I, F and C). Thus, the major ecological effect of increasing pollinator competition is
to make the mutualism more amenable to ecological coexistence with exploiters that
would otherwise be excluded; the major ecological effect of increasing exploiter
37
competition is to ‘tame’ exploiters that would otherwise cause three-way extinction.
Altogether, the likelihood that an invading exploiter can coexist with a given pollinator
increases with increasing competition in both species.
What we learn from this purely ecological analysis is that both the adaptive traits
bM and bE and the intraspecific competition intensities µ and α influence the ecological
outcome of the mutualism-exploiter interaction. Therefore, evolution of the pollinator
birth rate in the absence of exploiters sets the stage for an ecological invasion by
exploiters, the outcome of which will depend on the exploiter birth rate and both
intraspecific competition coefficients. While exploiter speciation is not explicitly
modeled, the exploiter species may either arise from within the mutualism by a shift in
oviposition time (described below), or arise elsewhere and then invade. In the case where
invasion and three-way coexistence is possible initially, subsequent coevolution of the
pollinator and the exploiter will change their birth rates as well as the ecological state of
the interaction, possibly moving the system into the exclusion or even extinction region
of adaptive traits bM and bE.
Pollinator Evolution and Exploiter Invasion
How does pollinator evolution affect the success of exploiter invasion? To answer this
question, we consider the evolution of pollinator birth rate in the absence of exploiters,
and examine the outcome of exploiter invasion in a population at the pollinator’s
38
evolutionary equilibrium. We call the stable evolutionary equilibrium reached by the
pollinator’s birth rate in the absence of exploiters the ‘mutualism-only’ evolutionarily
stable state (ESS).
Equation (2) yields the mutualism-only ESS as a function of pollinator
competition. Evolution to an ESS, and generally to a low birth rate, is a consequence of
the fecundity-competition tradeoff. Except when the intensity of pollinator competition is
low, pollinator evolution drives the system near the tapered end of the region of
ecological coexistence with potential exploiters (fig. 1). More precisely, the mutualismonly ESS is always at the point where the pollinator coevolutionary isocline intersects the
exploiter exclusion boundary (along this boundary, exploiter population density goes to
zero making the three-species system and the two-species system formally equivalent).
At the mutualism-only ESS, exploiters with a lower birth rate than the pollinator are
always excluded. Which exploiters can invade at the mutualism-only ESS depends on the
intensity of intraspecific competition in both pollinators and exploiters. At lower
intensities of pollinator intraspecific competition, exploiters must have a higher birth rate
in order to successfully invade, while at higher intensities of exploiter intraspecific
competition, exploiters with a wider range of birth rates can invade (fig. 2).
The case of an exploiter that has the same birth rate as the pollinator is interesting
because it highlights the scenario in which an exploiter species arises via a change in the
timing of oviposition, with no difference in birth rate. Delaying oviposition changes the
competitive environment experienced by larvae, which is the only feature that
distinguishes pollinator and exploiter growth rates in our model. Later oviposition also
39
eliminates the opportunity for pollination; the lack of selection on active pollination could
then lead to secondary loss of pollination behavior in nascent exploiters. The model
shows that invasion by such a ‘derived exploiter’ is essentially determined by the
intensity of pollinator competition, and is possible only above a certain threshold
(estimated to be µ = 1.1, fig. 3; see also fig. 1C, F, I).
Mutualism-Exploiter Evolutionary Persistence
Given the successful invasion and ecological persistence of an exploiter, we now
examine the subsequent coevolution of pollinator and exploiter birth rates and its
consequences for the long-term viability of the three-species interaction. Evaluating
equations (4) for different intensities of pollinator and exploiter intraspecific competition
(µ and α between 0 and 2), we find that selection on birth rate can lead to a
coevolutionary stable state (coESS) lying within the three-species coexistence region. In
this case, there is evolutionary persistence of the three-species community (fig. 1A-F, H).
However, there are two alternatives to evolutionary persistence. In one, coevolution leads
to the boundary of the exploiter exclusion region (fig. 1G): the exploiter goes extinct
while the mutualism remains intact. We refer to the extinction of the exploiter after
coevolution as ‘evolutionary purging’ of the exploiter, since the exploiter is driven to
competitive exclusion by the coevolving pollinator. Once evolutionary purging occurs,
the mutualism will begin to return to the mutualism-only ESS. In the second alternative
40
to evolutionary persistence, coevolution leads to the boundary of the three-species
extinction region and consequently to the extinction of all three species (fig. 2I). This is a
case of ‘evolutionary suicide’ or evolution to extinction, in which the selection gradient
drives a species to acquire nonviable trait values (Dieckmann and Ferrière 2004; Ferrière
2000; Gyllenberg and Parvinen 2001).
Evolutionary coexistence occurs over a wide range of pollinator and exploiter
competition intensities (fig. 1 and fig. 3). Nevertheless, figure 3 demonstrates that
evolutionary purging of the exploiter is the expected outcome for weak intraspecific
competition in both pollinator and exploiter, and evolutionary suicide is expected for
strong pollinator competition and weak exploiter competition. These findings show that
intensities of intraspecific competition that allow ecological coexistence between
mutualists and exploiters do not guarantee evolutionary persistence (fig. 1G, I). On the
other hand, intensities of intraspecific competition that restrict ecological coexistence to a
small range of phenotypes may yet promote the evolutionary persistence of the threespecies community (fig. 1A, D).
Phenotypic and Population Outcomes of Pollinator-Exploiter Coevolution
The intensity of intraspecific competition in pollinators is the main determinant of the
phenotypic and population outcomes of coevolution with the exploiter (fig. 4). The
coevolutionary process always drives the birth rate of the pollinator to a coESS value
41
larger than the mutualism-only ESS value (fig. 4B). Pollinator and exploiter coESS birth
rates vary predictably, yet non-monotonically with pollinator intraspecific competition
(fig. 4A): they decrease as pollinator competition intensity µ increases to ca. 1, and then
increase gradually. The difference between pollinator and exploiter coESS birth rates is
about two-fold when pollinator competition is strong (µ > 1), and is greatest, up to tenfold, when pollinator competition is weak (µ < 1) (fig. 4A). In some cases, the coevolved
exploiter birth rate is so high that the coevolved exploiters would have caused threespecies extinction had they invaded the ancestral mutualism. Thus, coevolution has the
potential to generate ‘kamikaze exploiters’, i.e., those that can cause local extinctions of
the mutualism upon invading naïve, non-coevolved populations (fig. 3).
The equilibrium densities of the coevolved pollinator, exploiter, and plant
populations also show strong non-linear patterns of covariation with the intensity of
pollinator competition, µ (fig. 4C, E). Below µ = 1, relaxing pollinator competition
results in a strong increase of pollinator coESS population density (fig. 4C), while both
exploiter and plant coESS population densities experience a marked decline (fig. 4C, E).
Above µ = 1, intensifying pollinator competition results in a decrease of the pollinator
coESS population density (fig. 4C), while both exploiter and plant coESS population
densities decrease more slowly (fig. 4C, E). Coevolution between pollinators and
exploiters reduces plant population density from the ESS to the coESS (fig. 4F), but leads
to remarkably little change in the pollinator population density from the ESS to the
coESS (fig. 4D) in spite of having a significant impact on pollinator birth rate (fig. 4B).
42
The phenotypic and population outcomes of coevolution with respect to pollinator
competition intensity are summarized in figure 5A. The observed patterns (fig. 5A) result
from a combination of ecological and evolutionary effects (fig. 5B), the relative
importance of which can be evaluated from our model. Ecological effects of varying
pollinator competition intensity, µ, can be predicted from the ecological model, equations
(3). Although no analytical results are available, a numerical approach is straightforward
and the results are displayed in figure A2. Evolutionary effects can be predicted from the
selective gradients, equations (4), which yield expressions of the coESS birth rates bM*
and bE* with respect to intraspecific competition intensities, µ and α, and pollinator and
ˆ and Eˆ :
exploiter equilibrium densities, M
(5a)
bM* =

1 1
 ˆ − µ ,

µ′  M
(5b)
bE* =

11
 ˆ − α.

α′ E
By varying µ and holding the birth rates bM and bE constant in equations (3), we
find that µ has primary (i.e., direct) ecological effects on the equilibrium population
ˆ and Eˆ (fig. A2); meanwhile, by varying µ and forcing the ecological
densities M
ˆ and Eˆ to be constant in equations (5), we find that µ
equilibrium population densities M
has primary evolutionary effects on the coESS birth rates bM* and bE* . Additionally, the
primary ecological effects of µ cause secondary (i.e., indirect) evolutionary effects on the
coESS birth rates bM* and bE* that can be predicted from equations (5) by varying the
ˆ and Eˆ while keeping µ constant. Furthermore, the
equilibrium population densities M
43
primary evolutionary effects of µ cause secondary ecological effects on the equilibrium
ˆ and Eˆ that can be predicted from equations (3) by varying the
population densities M
pollinator and exploiter birth rates bM and bE while holding µ constant (fig. A2).
How much variation in equilibrium traits and population densities is explained by
ecology versus evolution as pollinator competition intensity varies? This question can be
answered by comparing the observed outcomes of the coevolutionary process (fig. 5A)
with the predicted primary and secondary ecological and evolutionary effects (fig. 5B).
Primary and secondary ecological effects appear to be entirely congruent in their
direction, both at low and high µ. However, they do not explain the strong decrease in
plant population density Pˆ when µ decreases at low µ, or the quasi-absence of change in
exploiter population density Eˆ at high µ. Thus, more complex effects of ecoevolutionary feedbacks must play a significant role. The observed response of the
pollinator birth rate at low µ is consistent with a strong primary evolutionary effect,
which opposes and overcomes the secondary evolutionary response. In contrast, at high
µ, it is the secondary evolutionary effect of changing µ that explains the observed change
in pollinator birth rate; the primary evolutionary effect only has a moderating effect.
Likewise, only the secondary evolutionary effect can explain the observed dramatic
change in exploiter birth rate at low µ. At high µ, the decrease of exploiter birth rate as µ
decreases is explained neither by primary nor by secondary evolutionary effects. Again,
more complex effects of eco-evolutionary feedbacks must be involved.
44
Discussion
The mutualism-exploitation interaction is an important model system because it provides
a test case in which to move coevolutionary theory beyond a pairwise perspective.
Coevolution is increasingly recognized to involve groups rather than pairs of species;
furthermore, the presence of other species is being shown empirically to alter the
outcome of pairwise, coevolving interactions (e.g., Currie et al. 2003; Strauss et al. 2005;
Thompson and Fernandez 2006; Thrall et al. 2007). Thus, multi-species coevolution has
the potential to be a powerful organizing force within ecological communities. A logical
way to begin to incorporate multiple species within coevolutionary models is to explore
the eco-evolutionary effects of third species that closely interact with well-understood,
coevolving, pairwise interactions, a research direction that has been opened only recently
by theorists (Ferrière et al. 2007; Gandon 2004; Nuismer and Doebeli 2004).
In order to investigate the eco-evolutionary dynamics of an exploited mutualism,
we constructed an adaptive dynamics model of the coevolution of mutualist and exploiter
birth rates given a fecundity-competition tradeoff and using the ecological dynamics
described by Morris et al. (2003). Our model assumes complete competitive dominance
of the mutualist species over the exploiter due to a gap between oviposition times.
Therefore, we fixed the intensity of interspecific competition from mutualists
experienced by exploiters, and focused on examining the role of mutualist and exploiter
intraspecific competition on the state of the mutualism at the time of exploiter invasion,
on the coevolutionary process, and on the phenotypic and population consequences of
45
eco-evolutionary feedbacks. We found that intraspecific competition does make longterm coexistence possible. Once an exploiter has successfully invaded the mutualism,
high intensities of mutualist and exploiter intraspecific competition make it more likely
that a stable coevolutionary equilibrium will be reached at which the host, mutualist and
exploiter coexist. When the intensity of exploiter intraspecific competition is low,
evolutionary suicide will occur with all three species going extinct. When the intensity of
mutualist intraspecific competition is low, the exploiter is excluded from the community
through evolutionary purging, a type of coevolutionary dynamic not previously identified
in mutualism-exploiter interactions (but see Kooi and Troost 2006 for evolutionary
purging of a competitor).
When long-term coexistence occurs, ecological and evolutionary processes
interact to determine phenotypic and population outcomes, with feedbacks between
ecological and evolutionary responses playing a significant role. Under weak mutualist
competition, the exploiter’s high birth rate is produced by secondary evolutionary effects
while the host’s low density results from more complex effects of eco-evolutionary
feedbacks. Under strong mutualist competition, coevolution leads to a minimal difference
in birth rate between mutualist and exploiter, and the host density is largest. We have
shown that neither primary nor secondary ecological or evolutionary effects suffice to
explain this pattern.
Below we discuss the implications of our results for understanding the dynamics
of mutualist-exploiter communities. We emphasize insights that can be applied to
coevolution in general.
46
Origin of Exploiters: Assembled or Derived?
The model predicts that in the absence of exploiters, the mutualist species evolves to a
relatively low birth rate as a consequence of a fecundity-competition tradeoff. Evolution
of the mutualist is likely to produce an interaction that is ecologically more stable than
the ancestral mutualism, as the low cost imposed on hosts by evolved mutualists results
in a larger host population. On the other hand, evolving to a low birth rate also means
evolving to a trait region in which three-species extinction is more likely following an
exploiter invasion. However, early invasion of the mutualism by exploiters can
‘immunize’ the mutualism against more harmful exploiters that would cause threespecies extinction, a process that is emerging as an important mechanism maintaining
cooperation and mutualism (Ferrière et al. 2007; Foster and Kokko 2006; Nowak and
Sigmund 1998). These results suggest that exploiters should often be anciently associated
with mutualisms, such as in the attine ant-fungus mutualism (Currie et al. 2003) and the
fig-fig wasp mutualism (Machado et al. 1996), in which exploiters are known to have
persisted for millions of years. Unrelated exploiters that have invaded a mutualism
recently may have already been ‘tamed’, as they have coevolved with a closely related
mutualist pair (Currie et al. 2003; Lopez-Vaamonde et al. 2001; Segraves and Pellmyr
2004).
Our model also suggests that exploiters may be derived from the mutualist
species, for example by a shift in oviposition time as has been proposed in the yucca
(Pellmyr and Leebens-Mack 2000; Pellmyr et al. 1996) and globeflower (Després and
47
Cherif 2004) mutualisms (see also Law et al. 2001). Invasion by a derived exploiter is not
predicted to cause the immediate extinction of the mutualism at any point during
evolution of the mutualist. The model predicts that strong mutualist intraspecific
competition is the key condition for derived exploiters to be able to persist in the
mutualism. When mutualist competition is weak, derived exploiters can never arise, and
exploiters are predicted to be assembled. When mutualist intraspecific competition is
strong, derived exploiters can arise at any time in the mutualist's evolutionary history.
This prediction agrees with the suggestion that escape from strong intraspecific
competition was the mechanism behind the radiation of non-pollinators from pollinators
in the globeflower-globeflower fly mutualism (Després and Cherif 2004; Després and
Jaeger 1999).
Dynamics of Mutualist-Exploiter Coevolution
In our three-species model, the long-term persistence of the mutualism is achieved
through evolutionary purging of the exploiter, or else through mutualist-exploiter
coevolution towards an ecologically viable three-species equilibrium. Surprisingly, wide
ranges of trait values that permit ecological coexistence between mutualists and
exploiters do not guarantee evolutionary persistence; meanwhile, a narrow range of trait
values allowing for ecological coexistence is no indication that evolutionary persistence
is impossible.
48
Our results suggest that systems in which both mutualist and exploiter
intraspecific competition are weak should be rare, and at best evolutionarily transient, due
to evolutionary purging of the exploiter. Systems in which exploiter competition is weak
and mutualist competition is strong should also be rare, in this case due to evolutionary
suicide. Thus, the intensity of intraspecific competition can be critical for long term
mutualist-exploiter coexistence. However, while there is empirical evidence that
intraspecific competition in mutualists and exploiters does exist (Anstett et al. 1996;
Bronstein et al. 1998; Jaeger et al. 2001; Jousselin et al. 2001), and theory suggesting that
there should be selection on hosts to increase competition in mutualists and exploiters
(Ferdy et al. 2002), there is as yet little information available about the relative intensities
of intraspecific competition in mutualists and exploiters.
The model makes the key assumption of a spatially homogenous community.
However, spatial variation in environment, community composition and evolutionary
history can influence coevolutionary dynamics (Thompson 1999; Thompson 2005).
Inspection of the model results suggests that the three-species system may be maintained
as a meta-community even when coevolution should lead to extinction of the exploiter
(evolutionary purging) or of all three species (evolutionary suicide). When evolutionary
purging occurs, the exploiter goes extinct locally and might be expected to go extinct
globally as well. However, if populations linked by migration are at different
evolutionary stages, exploiters may be able to persist by invading populations that are
still susceptible. In this case, a species that should be eliminated by evolutionary
dynamics (purging) could be maintained by an ecological mechanism (dispersal).
49
Similarly, an entire system that should be eliminated by evolutionary suicide might be
rescued by frequent re-colonization of patches by hosts and mutualists, allowing
temporary escape from exploiters. In either case, changes in migration rate or patch
fragmentation could cause large changes in the eco-evolutionary outcome of the threespecies interaction, as suggested by Urban et al. (2008).
Migration within meta-communities could also lead to local extinctions of the
mutualism even when there is a viable three-species coESS. When exploiter intraspecific
competition is low but exploiters are not purged, exploiters are expected to coevolve into
'kamikazes' that will cause extinction of communities composed of mutualists that
evolved in the absence of exploiters. Whether the entire system is driven to local
extinction may depend on whether mutualists are obligate or facultative. For example, in
the seed dispersal mutualism between conifers and crossbills, introduction of red squirrels
(seed predators) from neighboring communities has driven crossbills (seed dispersers),
but not black spruce, to local extinction (Parchman and Benkman 2002).
Another model assumption is a constant environment. Examination of figure 1
yields insight into how gradual environmental change or evolution of the host species can
affect the eco-evolutionary dynamics of the system. If the intensity of mutualist
intraspecific competition is altered by a changing environment (e.g., Memmott et al.
2007) or by evolution of host morphology (e.g., Ferdy et al. 2002), the system can move
between evolutionary purging of the exploiter and evolutionary persistence. Intensifying
mutualist intraspecific competition could also open the opportunity for invasion by
derived exploiters. Meanwhile, changes in the intensity of exploiter intraspecific
50
competition can shift the system between evolutionary suicide and evolutionary
persistence.
Phenotypic Consequences of Mutualist-Exploiter Coevolution
We predict that mutualists coevolving with exploiters should always have a higher birth
rate than mutualists that have evolved without exploiters. Similarly, Ferrière et al. (2007)
predicted that mutualists evolve lower rates of investment into the mutualism when in the
presence of exploiters. While there are mutualisms with only partial geographic overlap
with exploiters (Anderson 2006; Pellmyr and Leebens-Mack 2000), comparisons between
populations are still needed in order to determine the effect the presence of exploiters has
had on mutualist evolution (e.g., Benkman 1999).
Exploiters with a lower birth rate than the mutualist can exist transiently;
however, our model shows that coevolution always results in coevolved exploiters with a
higher birth rate than that of the coevolved mutualist. The difference in coevolved birth
rates is predicted to be minimal and fairly constant in systems where mutualist
competition is strong, but to diverge dramatically as mutualist competition decreases. In
the globeflower and yucca mutualisms, more exploiter than pollinator eggs are found per
fruit (Addicott 1996; Pellmyr 1989). However, as both mutualists and exploiters can
spread their eggs among multiple fruits, the differences in eggs per fruit may reflect
differences in pollinator and exploiter population densities and in numbers of
51
inflorescences/fruits available at the time of oviposition, rather than differences in
fecundity.
In our model, intermediate birth rates evolve as a consequence of the fecunditycompetition tradeoff experienced by mutualists and exploiters. While we have used an
intraspecific tradeoff to understand the evolution of individual species, fecunditycompetition tradeoffs, and competition-colonization tradeoffs more generally, have
previously only been applied to mutualistic communities in order to explain interspecific
coexistence (Stanton et al. 2002; Yu et al. 2001). Consequently, the shape and slope of
this tradeoff within mutualist and exploiter species requires empirical investigation. In the
absence of more empirical data, we assumed the same tradeoff for both the mutualist and
the exploiter (i.e. equal slopes, µ′ = α ′ ) and we varied this slope (between 0 and 0.5) to
test the robustness of our results; the qualitative ecological and evolutionary patterns
reported here were unaffected except when the slope is zero (results not shown).
However, different eco-evolutionary dynamics might develop if the tradeoff were to
depart significantly from the assumed linearity (cf. equations (2) and (4)). With a
sufficiently large degree of concavity, evolutionary branching might occur, as indicated
by general theory (de Mazancourt and Dieckmann 2004; Rueffler et al. 2004) and a
previous model of mutualism evolution (Ferrière et al. 2002). This is an interesting
possibility that warrants future theoretical investigation, as it would predict persistent trait
variation within mutualist and exploiter species, and reveal new ways in which ecoevolutionary feedbacks mediated by phenotypically different populations within species
could affect eco-evolutionary interactions between species.
52
Population Consequences of Mutualist-Exploiter Coevolution
The model shows that the response of coevolutionary equilibrium population densities of
hosts, mutualists and exploiters to variation in the intensity of competition is mediated by
ecological and evolutionary forces, and their interactions. Such a partitioning of
ecological and evolutionary effects due to environmental change is germane to the traitpopulation time series analysis proposed by Hairston et al. (2005).
A significant response of relaxing mutualist intraspecific competition is the sharp
decline of host density. This response is best explained by a synergistic interaction
between primary ecological and primary evolutionary effects (i.e., the increase in
mutualist density and in mutualist birth rate as mutualist competition becomes weaker).
The decrease in host population density at coevolutionary equilibrium in response to
relaxed mutualist competition is part of a two-fold 'trait-abundance syndrome' that could
not have been predicted without accounting for eco-evolutionary feedbacks. Under weak
mutualist competition, the model predicts coevolution of associations of hosts with low
density, mutualists with high density, and high birth rate exploiters with low density.
Under strong mutualist competition, coevolution is expected to result in associations of
hosts with high density, and mutualists and exploiters with similar densities and a
relatively small difference in birth rates. These trait-abundance syndromes have three
counterintuitive implications. (1) Persistent, highly parasitic exploiters can be favored by
coevolution. (2) Exploiters can evolve strong parasitism (i.e., high birth rate) and yet the
density of coevolved mutualists can be high. (3) Low plant density is caused primarily by
53
the mutualist’s coevolved birth rate and high density, not by highly parasitic coevolved
exploiters. Indeed, under weak mutualist competition, the coevolutionary process leaves
host population density almost unchanged from its mutualism-only equilibrium
population density.
Concluding Remarks
Our eco-evolutionary model of a mutualism-exploiter system highlights that exploiters
are not just opportunistic visitors of little ecological or evolutionary relevance. The
success or failure of their invasion reflects specific features of mutualist evolution.
Persistent exploiters have significant evolutionary effects on mutualists, and themselves
change as mutualists evolve. The phenotype of mutualists and the abundance of hosts
cannot be understood without taking coevolution with exploiters into account, and the
ecological characteristics of each species (mutualist and exploiter) have eco-evolutionary
repercussions on the other species. By accounting for interactions between ecological and
evolutionary processes, our model has delivered new predictions about the evolutionary
history of mutualisms and the characteristics of coevolved mutualist-exploiter
communities, and new insights into the coevolutionary process in a community context.
In order to evaluate these predictions, we are in need of empirical studies that identify
and compare populations with and without exploiters and that directly compare lifehistory traits of mutualists and exploiters.
54
Extensions of the current theory are needed to determine whether exploited
mutualisms can be maintained spatially when coevolution should otherwise lead to the
extinction of one or more species, and whether coevolution can still lead to stable threespecies coexistence when the intensity of intraspecific competition changes as a result of
environmental change or host evolution. An even broader theoretical perspective is
opened by connecting our model to host-parasite coevolutionary theory. The model
studied here has the structure of a one host/two parasite system, assuming a
superinfection process between parasite populations, and a single-infection process
among genetic variants within parasite populations (Levin and Pimentel 1981; May and
Nowak 1995; Mosquera and Adler 1998; Nowak and May 1994; van Baalen and Sabelis
1995). Our conclusions about long-term coexistence are consistent with host-parasite
evolutionary theory, although the latter would predict coevolution of less, rather than
more, parasitic mutualists and exploiters (Boldin and Diekmann 2008; Mosquera and
Adler 1998; Pugliese 2002). We hypothesize that this fundamental difference is rooted in
the mutualistic relationship that one of the parasites has with the host.
Current evolutionary theory views parasitism and mutualism as two ends of a
phenotypic continuum (Genkai-Kato and Yamamura 1999; Yamamura 1993). Our
analysis suggests that mutualism and parasitism should be investigated as separate
characters that can evolve in a multi-dimensional trait space. This will pave the way for a
deeper understanding of the ecology and evolution of complex, partly beneficial, partly
antagonistic ‘liaisons dangereuses’ (Selosse et al. 2006; van Baalen and Jansen 2001) in
natural communities.
55
Acknowledgments
EIJ thanks Bill Morris for valuable discussion of the original ecological model and
Richard Law for guidance and for hosting EIJ at the University of York during the initial
stages of this project. RF is grateful to Michael Donoghue, Tom Near and Steve Stearns
for providing an excellent working environment in their laboratories at Yale University
where part of this research was conducted. We thank Benjamin Bolker, Ruth Geyer Shaw
and two anonymous reviewers for helpful comments on the manuscript. EIJ’s research
was supported by the Department of Ecology and Evolutionary Biology at the University
of Arizona and the National Science Foundation grant DEB 0806836. RF’s research was
supported by the French Agence Nationale de la Recherche (Stochastic Models for
Evolution project) and the National Science Foundation Frontiers in Integrative
Biological Research grant EF0623632.
56
Online Appendix A
Coevolutionary dynamics of mutualist and exploiter birth rates
The ecological model given by equations (3) in the Article,
1 dP
= bP M(1− M)(1− E)(1− P) −1,
P dt
1 dM
= bM P(1− µM) − dM ,
M dt
1 dE
= bE PM(1− M)(1− αE) − dE ,
E dt
ˆ , Eˆ ) . At this
was found numerically to possess at most one positive equilibrium ( Pˆ , M
equilibrium, the invasion fitnesses of mutant phenotypes bMmut and bEmut are defined by
equations (4) in the Article:
ˆ ]− d ,
W M (bMmut , bM ) = bMmut Pˆ [1− (µ + µ'(bMmut − bM )) M
M
ˆ (1− M
ˆ )[1− (α + α '(b mut − b )) Eˆ ] − d ,
W E (bEmut , bE ) = bEmut Pˆ M
E
E
E
hence the selection gradients of traits bM and bE,
∂W M
∂W
(bM , bM ) and mutE (bE , bE ). Under
mut
∂bM
∂bE
the assumption of small and rare mutations, the eco-evolutionary process that links the
ˆ , Eˆ ) and in trait values (b , b ) can
changes through time in ecological equilibria ( Pˆ , M
M
E
be approximated by the ‘canonical equation of adaptive dynamics’ (Champagnat et al.
2006; Dieckmann and Law 1996):
(A1a)
dbM
ˆ ∂W M (b , b ) ,
= kM M
dt
∂bMmut M M
57
(A1b)
dbE
∂W
= k E Eˆ mutE (bE , bE ) ,
∂bE
dt
where kM and kE are rates of evolutionary variation in bM and bE, which combines the peroffspring probability of trait mutation and the variance of the distribution of mutational
effects. We assumed that mutualists and exploiters had similar evolutionary variation
rates (kM = kE, normalized to 1), due to phylogenetic relatedness or physiological
similarity (e.g., similar body size, Allen et al. 2006).
Coevolutionary singularities are obtained as the non-negative equilibria of
equations (A1), that is, non-negative bM and bE values such that
dbM
db
= 0 and E = 0.
dt
dt
The attractivity and the evolutionary stability of coevolutionary singularities are
evaluated numerically by computing the eigenvalues of the Jacobian matrix and the
Hessian matrix, respectively, which are associated with equations (A1) (Dieckmann and
Law 1996). Across the ranges of intraspecific competition intensities and for the
parameter values considered in this study, coevolutionary singularities were always found
to be (locally) attractive and evolutionarily stable.
The trait selective field and trait trajectories towards coevolutionary singularities
were obtained by numerical integration of the system of equations (A1), and an example
is displayed in figure A1. This is figure 1E from the Article, on which we superimposed
the numerically estimated trait vector field. As in all cases analyzed in this study, the
coevolutionary singularity appears to be globally attractive. This example also reveals
that under our assumption of equal rates of evolutionary variation in both traits, the
evolutionary dynamics of traits bM and bE seem decoupled. With ancestral conditions
58
sufficiently distant from the coESS, first the pollinator trait evolves near its final value at
coESS while the exploiter trait remains almost constant; the exploiter trait then evolves
toward its final value at coESS while the pollinator trait value changes little. This
decoupling of phenotypic evolution in the two species may add to the general difficulty
of detecting reciprocal selection and coevolution in natural systems.
Population densities at ecological equilibrium
The effect of mutualist intraspecific competition and mutualist birth rate on
population densities at ecological equilibrium was evaluated numerically by performing
simulations over the relevant parameter range (fig. A2).
59
Literature Cited
Addicott, J. F. 1996. Cheaters in yucca/moth mutualism. Nature 380:114-115.
Agrawal, A. A., D. D. Ackerly, F. Adler, A. E. Arnold, C. Caceres, D. F. Doak, E. Post,
P. J. Hudson, J. Maron, K. A. Mooney, M. Power, D. Schemske, J. Stachowicz, S.
Strauss, M. G. Turner, and E. Werner. 2007. Filling key gaps in population and
community ecology. Frontiers in Ecology and the Environment 5:145-152.
Althoff, D. M., K. A. Segraves, J. Leebens-Mack, and O. Pellmyr. 2006. Patterns of
speciation in the yucca moths: Parallel species radiations within the Tegeticula
yuccasella species complex. Systematic Biology 55:398-410.
Anderson, B. 2006. Inferring evolutionary patterns from the biogeographical distributions
of mutualists and exploiters. Biological Journal of the Linnean Society 89:541549.
Anstett, M. C., J. L. Bronstein, and M. HossaertMcKey. 1996. Resource allocation: A
conflict in the fig/fig wasp mutualism? Journal of Evolutionary Biology 9:417428.
Axelrod, R., and W. D. Hamilton. 1981. The Evolution of Cooperation. Science
211:1390-1396.
Benkman, C. W. 1999. The selection mosaic and diversifying coevolution between
crossbills and lodgepole pine. American Naturalist 153:S75-S91.
Boldin, B., and O. Diekmann. 2008. Superinfections can induce evolutionarily stable
coexistence of pathogens. Journal of Mathematical Biology 56:635-672.
Bronstein, J. L. 2001. The exploitation of mutualisms. Ecology Letters 4:277-287.
60
Bronstein, J. L., D. Vernet, and M. Hossaert-McKey. 1998. Do fig wasps interfere with
each other during oviposition? Entomologia Experimentalis Et Applicata 87:321324.
Bronstein, J. L., W. G. Wilson, and W. E. Morris. 2003. Ecological dynamics of
mutualist/antagonist communities. American Naturalist 162:S24-S39.
Bull, J. J., and W. R. Rice. 1991. Distinguishing Mechanisms for the Evolution of
Cooperation. Journal of Theoretical Biology 149:63-74.
Champagnat, N., R. Ferrière, and S. Méléard. 2006. Unifying evolutionary dynamics:
From individual stochastic processes to macroscopic models. Theoretical
Population Biology 69:297-321.
Cook, J. M., and J. Y. Rasplus. 2003. Mutualists with attitude: coevolving fig wasps and
figs. Trends in Ecology & Evolution 18:241-248.
Currie, C. R., B. Wong, A. E. Stuart, T. R. Schultz, S. A. Rehner, U. G. Mueller, G. H.
Sung, J. W. Spatafora, and N. A. Straus. 2003. Ancient tripartite coevolution in
the attine ant-microbe symbiosis. Science 299:386-388.
de Mazancourt, C., and U. Dieckmann. 2004. Trade-off geometries and frequencydependent selection. American Naturalist 164:765-778.
Dedej, S., and K. Delaplane. 2005. Net energetic advantage drives honey bees (Apis
mellifera L) to nectar larceny in Vaccinium ashei Reade. Behavioral Ecology and
Sociobiology 57:398-403.
Denison, R. F. 2000. Legume sanctions and the evolution of symbiotic cooperation by
rhizobia. American Naturalist 156:567-576.
61
Després, L., and M. Cherif. 2004. The role of competition in adaptive radiation: a field
study on sequentially ovipositing host-specific seed predators. Journal of Animal
Ecology 73:109-116.
Després, L., and N. Jaeger. 1999. Evolution of oviposition strategies and speciation in the
globeflower flies Chiastocheta spp. (Anthomyiidae). Journal of Evolutionary
Biology 12:822-831.
Dieckmann, U., and R. Ferrière. 2004. Adaptive dynamics and evolving biodiversity,
Pages 188-224 in R. Ferrière, U. Dieckmann, and D. Couvet, eds. Evolutionary
Conservation Biology. Cambridge, Cambridge University Press.
Dieckmann, U., and R. Law. 1996. The dynamical theory of coevolution: A derivation
from stochastic ecological processes. Journal of Mathematical Biology 34:579612.
Doebeli, M., and N. Knowlton. 1998. The evolution of interspecific mutualisms.
Proceedings of the National Academy of Sciences of the United States of America
95:8676-8680.
Dufaÿ, M., and M. C. Anstett. 2003. Conflicts between plants and pollinators that
reproduce within inflorescences: evolutionary variations on a theme. Oikos 100:314.
Edwards, D. P., and D. W. Yu. 2007. The roles of sensory traps in the origin,
maintenance, and breakdown of mutualism. Behavioral Ecology and Sociobiology
61:1321-1327.
62
Ferdy, J. B., L. Després, and B. Godelle. 2002. Evolution of mutualism between
globeflowers and their pollinating flies. Journal of Theoretical Biology 217:219234.
Ferrière, R. 2000. Adaptive responses to environmental threats: evolutionary suicide,
insurance, and rescue. Options Spring 2000:12-16.
Ferrière, R., J. L. Bronstein, S. Rinaldi, R. Law, and M. Gauduchon. 2002. Cheating and
the evolutionary stability of mutualisms. Proceedings of the Royal Society of
London Series B-Biological Sciences 269:773-780.
Ferrière, R., M. Gauduchon, and J. L. Bronstein. 2007. Evolution and persistence of
obligate mutualists and exploiters: competition for partners and evolutionary
immunization. Ecology Letters 10:115-126.
Foster, K. R. 2004. Diminishing returns in social evolution: the not-so-tragic commons.
Journal of Evolutionary Biology 17:1058-1072.
Foster, K. R., and H. Kokko. 2006. Cheating can stabilize cooperation in mutualisms.
Proceedings of the Royal Society B-Biological Sciences 273:2233-2239.
Foster, K. R., and T. Wenseleers. 2006. A general model for the evolution of mutualisms.
Journal of Evolutionary Biology 19:1283-1293.
Fox, C. W., M. E. Czesak, and R. W. Fox. 2001. Consequences of plant resistance for
herbivore survivorship, growth, and selection on egg size. Ecology 82:2790-2804.
Frank, S. A. 1994. Genetics of Mutualism - the Evolution of Altruism between Species.
Journal of Theoretical Biology 170:393-400.
63
Fussmann, G. F., M. Loreau, and P. A. Abrams. 2007. Eco-evolutionary dynamics of
communities and ecosystems. Functional Ecology 21:465-477.
Gandon, S. 2004. Evolution of multihost parasites. Evolution 58:455-469.
Genkai-Kato, M., and N. Yamamura. 1999. Evolution of mutualistic symbiosis without
vertical transmission. Theoretical Population Biology 55:309-323.
Geritz, S. A. H., J. A. J. Metz, E. Kisdi, and G. Meszéna. 1997. Dynamics of adaptation
and evolutionary branching. Physical Review Letters 78:2024-2027.
Gyllenberg, M., and K. Parvinen. 2001. Necessary and sufficient conditions for
evolutionary suicide. Bulletin of Mathematical Biology 63:981-993.
Hairston, N. G., S. P. Ellner, M. A. Geber, T. Yoshida, and J. A. Fox. 2005. Rapid
evolution and the convergence of ecological and evolutionary time. Ecology
Letters 8:1114-1127.
Herre, E. A., N. Knowlton, U. G. Mueller, and S. A. Rehner. 1999. The evolution of
mutualisms: exploring the paths between conflict and cooperation. Trends in
Ecology & Evolution 14:49-53.
Holland, J. N., and J. L. Bronstein. 2008. Mutualism, Pages 2485-2491 in S. E.
Jørgensen, and B. D. Fath, eds. Encyclopedia of Ecology. Oxford, Elsevier.
Holland, J. N., D. L. DeAngelis, and S. T. Schultz. 2004. Evolutionary stability of
mutualism: interspecific population regulation as an evolutionarily stable strategy.
Proceedings of the Royal Society of London Series B-Biological Sciences
271:1807-1814.
64
Jaeger, N., F. Pompanon, and L. Després. 2001. Variation in predation costs with
Chiastocheta egg number on Trollius europaeus: how many seeds to pay for
pollination? Ecological Entomology 26:56-62.
Johnstone, R. A., and R. Bshary. 2008. Mutualism, market effects and partner control.
Journal of Evolutionary Biology 21:879-888.
Jousselin, E., M. Hossaert-McKey, D. Vernet, and F. Kjellberg. 2001. Egg deposition
patterns of fig pollinating wasps: implications for studies on the stability of the
mutualism. Ecological Entomology 26:602-608.
Kerdelhue, C., and J. Y. Rasplus. 1996. Non-pollinating Afrotropical fig wasps affect the
fig-pollinator mutualism in Ficus within the subgenus Sycomorus. Oikos 75:3-14.
Kiers, E. T., and M. G. A. van der Heijden. 2006. Mutualistic stability in the arbuscular
mycorrhizal symbiosis: Exploring hypotheses of evolutionary cooperation.
Ecology 87:1627-1636.
Kooi, B. W., and T. A. Troost. 2006. Advantage of storage in a fluctuating environment.
Theoretical Population Biology 70:527-541.
Law, R., J. L. Bronstein, and R. G. Ferrière. 2001. On mutualists and exploiters: Plantinsect coevolution in pollinating seed-parasite systems. Journal of Theoretical
Biology 212:373-389.
Levin, S., and D. Pimentel. 1981. Selection of Intermediate Rates of Increase in ParasiteHost Systems. American Naturalist 117:308-315.
65
Lopez-Vaamonde, C., J. Y. Rasplus, G. D. Weiblen, and J. M. Cook. 2001. Molecular
phylogenies of fig wasps: Partial cocladogenesis of pollinators and parasites.
Molecular Phylogenetics and Evolution 21:55-71.
Machado, C. A., E. A. Herre, S. McCafferty, and E. Bermingham. 1996. Molecular
phylogenies of fig pollinating and non-pollinating wasps and the implications for
the origin and evolution of the fig-fig wasp mutualism. Journal of Biogeography
23:531-542.
Marr, D. L., M. T. Brock, and O. Pellmyr. 2001. Coexistence of mutualists and
antagonists: exploring the impact of cheaters on the yucca - yucca moth
mutualism. Oecologia 128:454-463.
May, R. M., and M. A. Nowak. 1995. Coinfection and the Evolution of Parasite
Virulence. Proceedings of the Royal Society of London Series B-Biological
Sciences 261:209-215.
Maynard Smith, J., and E. Szathmáry. 1995, The Major Transitions in Evolution. Oxford,
W.H. Freeman Spektrum.
Memmott, J., P. G. Craze, N. M. Waser, and M. V. Price. 2007. Global warming and the
disruption of plant-pollinator interactions. Ecology Letters 10:710-717.
Metz, J. A. J., R. M. Nisbet, and S. A. H. Geritz. 1992. How Should We Define Fitness
for General Ecological Scenarios. Trends in Ecology & Evolution 7:198-202.
Morris, W. F., J. L. Bronstein, and W. G. Wilson. 2003. Three-way coexistence in
obligate mutualist-exploiter interactions: The potential role of competition.
American Naturalist 161:860-875.
66
Mosquera, J., and F. R. Adler. 1998. Evolution of virulence: a unified framework for
coinfection and superinfection. Journal of Theoretical Biology 195:293-313.
Nowak, M. A., and R. M. May. 1994. Superinfection and the Evolution of Parasite
Virulence. Proceedings of the Royal Society of London Series B-Biological
Sciences 255:81-89.
Nowak, M. A., and K. Sigmund. 1998. Evolution of indirect reciprocity by image
scoring. Nature 393:573-577.
Nuismer, S. L., and M. Doebeli. 2004. Genetic correlations and the coevolutionary
dynamics of three-species systems. Evolution 58:1165-1177.
Parchman, T. L., and C. W. Benkman. 2002. Diversifying coevolution between crossbills
and black spruce on Newfoundland. Evolution 56:1663-1672.
Pellmyr, O. 1989. The Cost of Mutualism - Interactions between Trollius-Europaeus and
Its Pollinating Parasites. Oecologia 78:53-59.
Pellmyr, O., and J. Leebens-Mack. 1999. Forty million years of mutualism: Evidence for
Eocene origin of the yucca-yucca moth association. Proceedings of the National
Academy of Sciences of the United States of America 96:9178-9183.
—. 2000. Reversal of mutualism as a mechanism for adaptive radiation in yucca moths.
American Naturalist 156:S62-S76.
Pellmyr, O., J. Leebens-Mack, and C. J. Huth. 1996. Non-mutualistic yucca moths and
their evolutionary consequences. Nature 380:155-156.
67
Peng, Y. Q., D. R. Yang, and Q. Y. Wang. 2005. Quantitative tests of interaction between
pollinating and non-pollinating fig wasps on dioecious Ficus hispida. Ecological
Entomology 30:70-77.
Pugliese, A. 2002. On the evolutionary coexistence of parasite strains. Mathematical
Biosciences 177:355-375.
Rueffler, C., T. J. M. Van Dooren, and J. A. J. Metz. 2004. Adaptive walks on changing
landscapes: Levins' approach extended. Theoretical Population Biology 65:165178.
Sachs, J. L., and E. L. Simms. 2008. The origins of uncooperative rhizobia. Oikos
117:961-966.
Segraves, K. A., and O. Pellmyr. 2004. Testing the out-of-Florida hypothesis on the
origin of cheating in the yucca-yucca moth mutualism. Evolution 58:2266-2279.
Selosse, M. A., F. Richard, X. H. He, and S. W. Simard. 2006. Mycorrhizal networks: des
liaisons dangereuses? Trends in Ecology & Evolution 21:621-628.
Smith, C. C., and S. D. Fretwell. 1974. Optimal Balance between Size and Number of
Offspring. American Naturalist 108:499-506.
Stanton, M. L., T. M. Palmer, and T. P. Young. 2002. Competition-colonization tradeoffs in a guild of African Acacia-ants. Ecological Monographs 72:347-363.
Strauss, S. Y., H. Sahli, and J. K. Conner. 2005. Toward a more trait-centered approach
to diffuse (co)evolution. New Phytologist 165:81-89.
Thompson, J. N. 1998. The population biology of coevolution. Researches on Population
Ecology 40:159-166.
68
—. 1999. Specific hypotheses on the geographic mosaic of coevolution. American
Naturalist 153:S1-S14.
—. 2005, The Geographic Mosaic of Coevolution. Chicago, IL, USA, The University of
Chicago Press.
Thompson, J. N., and C. C. Fernandez. 2006. Temporal dynamics of antagonism and
mutualism in a geographically variable plant-insect interaction. Ecology 87:103112.
Thrall, P. H., M. E. Hochberg, J. J. Burdon, and J. D. Bever. 2007. Coevolution of
symbiotic mutualists and parasites in a community context. Trends in Ecology &
Evolution 22:120-126.
Tillberg, C. V. 2004. Friend or foe? A behavioral and stable isotopic investigation of an
ant-plant symbiosis. Oecologia 140:506-515.
Urban, M. C., M. A. Leibold, P. Amarasekare, L. De Meester, R. Gomulkiewicz, M. E.
Hochberg, C. A. Klausmeier, N. Loeuille, C. de Mazancourt, J. Norberg, J. H.
Pantel, S. Y. Strauss, M. Vellend, and M. J. Wade. 2008. The evolutionary
ecology of metacommunities. Trends in Ecology & Evolution 23:311-317.
van Baalen, M., and V. A. A. Jansen. 2001. Dangerous liaisons: the ecology of private
interest and common good. Oikos 95:211-224.
van Baalen, M., and M. W. Sabelis. 1995. The dynamics of multiple infection and the
evolution of virulence. American Naturalist 146:881-910.
69
West, S. A., E. T. Kiers, E. L. Simms, and R. F. Denison. 2002. Sanctions and mutualism
stability: why do rhizobia fix nitrogen? Proceedings of the Royal Society of
London Series B-Biological Sciences 269:685-694.
Wilson, W. G., W. F. Morris, and J. L. Bronstein. 2003. Coexistence of mutualists and
exploiters on spatial landscapes. Ecological Monographs 73:397-413.
Yamamura, N. 1993. Vertical Transmission and Evolution of Mutualism from Parasitism.
Theoretical Population Biology 44:95-109.
Yu, D. W. 2001. Parasites of mutualisms. Biological Journal of the Linnean Society
72:529-546.
Yu, D. W., H. B. Wilson, and N. E. Pierce. 2001. An empirical model of species
coexistence in a spatially structured environment. Ecology 82:1761-1771.
Online Appendix References
Allen, A. P., J. F. Gillooly, V. M. Savage, and J. H. Brown. 2006. Kinetic effects of
temperature on rates of genetic divergence and speciation. Proceedings of the
National Academy of Sciences of the United States of America 103:9130-9135.
Champagnat, N., R. Ferrière, and S. Méléard. 2006. Unifying evolutionary dynamics:
From individual stochastic processes to macroscopic models. Theoretical
Population Biology 69:297-321.
70
Dieckmann, U., and R. Law. 1996. The dynamical theory of coevolution: A derivation
from stochastic ecological processes. Journal of Mathematical Biology 34:579612.
Morris, W. F., J. L. Bronstein, and W. G. Wilson. 2003. Three-way coexistence in
obligate mutualist-exploiter interactions: The potential role of competition.
American Naturalist 161:860-875.
71
Figure 1: Mutualist-exploiter eco-evolutionary dynamics for different intensities of
intraspecific competition in mutualists (µ) and exploiters (α). The white regions
signify that a stable three-species ecological equilibrium exists, with three-species
extinction in the gray region to the upper left and exclusion of the exploiter in the gray
region to the lower right. Within the three-species coexistence regions, the evolutionary
isoclines for the mutualist (quasi-vertical line) and exploiter (curved line) are shown.
Where the isoclines intersect, there is a coevolutionary stable state, coESS (solid circle)
(A-F, H). The mutualism-only ESS is also shown on the x-axis (triangle). In cases
without an intersection of the isoclines, there is either evolutionary purging of the
exploiter, when the mutualist isocline is to the right of the exploiter isocline (G), or
evolutionary suicide, when the mutualist isocline is to the left of the exploiter isocline (I).
Parameter values: bP = 10, dM = dE = 3, µ’ = α’ = 0.03; bP , dM and dE held constant from
Morris et al. (2003).
Figure 2: Range of exploiter birth rates capable of invading the mutualism at the
mutualism-only ESS. The lower surface (opaque) and upper surface (transparent)
represent the lowest and highest, respectively, exploiter birth rates than can invade and
persist in the mutualism. Below the lower limit, exploiters will be excluded from the
mutualism. Above the upper limit, invading exploiters will cause three-species extinction.
The range of exploiter birth rates that can invade and coexist with the mutualist species is
affected by intraspecific competition, such that the range is widened as the intensity of
72
exploiter intraspecific competition increases and the entire range is shifted to lower birth
rates as mutualist intraspecific competition increases. Parameter values as in figure 1.
Figure 3: Summary of eco-evolutionary outcomes. Lettering A-I refers to examples
shown in figure 1. Parameter combinations are evaluated at grid intersection points.
White: evolutionary persistence (e.g., A-F); gray: evolutionary purging of the exploiter
(e.g., G); dark gray: evolutionary suicide (e.g., I); light gray: evolution of kamikaze
exploiters (e.g., H). Exploiters can be derived from a pollinator lineage above the dotted
black line (µ > 1.1). Parameter values as in figure 1.
Figure 4: Phenotypic and population outcomes of mutualist-exploiter coevolution,
with respect to the intensity of mutualist intraspecific competition (µ). Exploiter
competition is held constant at α = 0.6 (gray lines) and α = 1.4 (black lines). (A)
Mutualist (solid lines) and exploiter (dashed lines) birth rates at the coevolutionary stable
state (coESS). (B) Difference between mutualist birth rates at the coESS and at the
mutualism-only evolutionary stable state (ESS). (C) Mutualist (plain lines) and exploiter
(dashed lines) population densities at the coESS. (D) Difference between mutualist
population densities at the coESS and at the ESS. (E) Host population density at the
coESS. (F) Difference between host population densities at the coESS and at the ESS.
Parameter values as in figure 1.
73
Figure 5: Ecological and evolutionary effects of decreasing mutualist intraspecific
competition (µ) on adaptive traits (mutualist birth rate bM, exploiter birth rate bE)
and population densities (mutualist density M, exploiter density E, host density P) at
coESS. (A) Observed ecological and evolutionary effects, summarized from figure 4. (B)
Predicted primary and secondary ecological and evolutionary effects. Primary ecological
effects are responses of equilibrium population densities to varying µ while holding traits
constant; they can be predicted from equations 3 (see also figure A2). Primary
evolutionary effects show the selective effect on the adaptive traits of varying µ while
holding population densities constant; they can be predicted from equations (5).
Secondary ecological effects are changes in equilibrium population densities resulting
from changes in trait values, independent of the direct changes caused by varying µ; they
can be predicted from equations 3 (and figure A2). Secondary evolutionary effects are
changes in adaptive traits caused by changes in population densities, estimated
independently of the direct selective effect of varying µ; they can be predicted from
equations (5).
Online Figure A1: Coevolutionary dynamics of mutualist-exploiter traits. The white
region signifies that a stable three-species ecological equilibrium exists, with threespecies extinction in the gray region to the upper left and exclusion of the exploiter in the
gray region to the lower right. In the absence of an exploiter, the mutualist evolves to its
ESS (open circle, dashed line). After successful exploiter invasion, the mutualist and
exploiter evolve according to the selective field (arrows). The quasi-vertical and the
74
curved solid lines are evolutionary isoclines for the mutualist trait and the exploiter trait,
respectively. Where the evolutionary isoclines intersect, there is a coevolutionary stable
state (solid circle). Parameter values: bP = 10, dM = dE = 3, µ = α = 1, µ’ = α’ = 0.03; bP ,
dM and dE held constant from Morris et al. (2003).
Online Figure A2: Population densities at ecological equilibrium with respect to the
intensity (µ) of mutualist intraspecific competition and the mutualist birth rate (bM).
Population densities of the (A) host, (B) mutualist and (C) exploiter are zero in the white
regions and are highest in the darkest bands. Parameter values: bP = 10, bE = 50, dM = dE =
3, α = 1.
75
Figure 1:
76
Figure 2:
77
Figure 3:
78
Figure 4:
79
Figure 5:
80
Online Figure A1:
81
Online Figure A2:
82
APPENDIX B
COEVOLUTION AND THE TRANSIENCE OF SPECIES IN AN EXPLOITED
MUTUALISM
83
Coevolution and the Transience of Species in an Exploited Mutualism
Emily I. Jones1,*, William F. Morris2, Régis Ferrière1,3,4
1. Department of Ecology and Evolutionary Biology, University of Arizona, Tucson,
Arizona 85721;
2. Biology Department, Duke University, Durham, North Carolina 27708;
3. Laboratoire Ecologie & Evolution, UMR 7625 CNRS/UPMC/ENS, École Normale
Supérieure, 46 rue d'Ulm, 75005 Paris, France;
4. CEREEP – ECOTRON Ile De France, UMS 3194 CNRS/ENS, Rue du Chateau, 77140
Saint-Pierre-les-Nemours, France
* Corresponding author; e-mail: [email protected].
Phone: (520) 621-3534
Fax: (520) 621-9190
Running head: Host-mutualist-exploiter coevolution
Keywords: mutualism, exploitation, competition, coexistence, adaptive dynamics,
coevolution.
84
ABSTRACT:
The dynamics of species interactions depend on both the community context and on
coevolution with species in the community. However, nearly all coevolutionary models
have taken a pairwise perspective. Here, we analyze an adaptive dynamics model of an
exploited pollinating seed parasitism mutualism, such as the yucca-yucca moth
mutualism, in order to distinguish between the population and trait consequences
interactions with non-evolving species versus coevolution with the interacting species.
Specifically, we investigate the evolution of mutualist and exploiter oviposition rates and
host plant defense against seed destruction by larvae of the mutualists and exploiters;
each of these traits determines the degree of parasitism of the species. By exploring how
these traits evolve across different ecological and evolutionary parameters, we find that
increasing the asymmetry in evolutionary rates (and thus the degree of coevolution)
greatly increases the chances of the evolutionary suicide of the mutualism. Surprisingly,
the major threat to the evolutionary stability of the mutualism comes from
overexploitation of the host by the mutualist, not the exploiter. This threat is largest when
evolutionary rates are asymmetric in favor of the mutualist (at the extreme, independent
evolution of the mutualist). Nevertheless, effective plant defenses can limit the conditions
that lead to extinction and increase population densities at the coevolutionary
equilibrium.
85
Introduction
Evolution often alters interactions between pairs of species by driving population
dynamics, changing the interaction type (Gomulkiewicz et al. 2003), leading to speciation
(Law et al. 2001; Ferdy et al. 2002), or even leading to species extinction (Dieckmann et
al. 1995; Ferrière et al. 2002). Due to eco-evolutionary feedbacks, coevolutionary models
of pairwise interactions may make very different predictions about the evolutionary
outcomes of an interaction than when one species is considered to evolve at a time (e.g.,
Best et al. 2009). Coevolution within communities should also lead to different dynamics
than pairwise coevolution (Thompson 1998; Strauss and Irwin 2004), but so far this
problem has only received limited theoretical attention (e.g., Gomulkiewicz et al. 2003;
Nuismer and Doebeli 2004). Mutualisms provide the opportunity to model community
coevolutionary dynamics involving multiple interaction types: cooperation, exploitation
and competition.
Mutualisms are a key feature of many ecosystems, resulting in services including
nitrogen fixation (legumes and rhizobia), pollination and seed dispersal (plants and
insects, birds, and mammals), and protection against herbivores, predators and parasites
(ants and plants, ants and other insects) (Agrawal et al. 2007; Holland and Bronstein
2008). Exploitation, in which individuals of non-mutualistic species take rewards from
the mutualist species, are increasing recognized as common both within and across
mutualisms (Bronstein 2001b; Yu 2001). Meanwhile, exploiters compete with mutualists
for access to these rewards, and asymmetry in this competition favoring mutualists is
86
considered to be an important factor promoting the persistence of mutualisms (Ferrière et
al. 2002; Morris et al. 2003; Ferrière et al. 2007).
The balance of benefits received versus costs expended determines selection on
investment in mutualism (Ferrière et al. 2002) and defenses against exploitation (Kiers
and van der Heijden 2006). Consequently, the state of the interaction between each pair
of species is likely to have indirect effects on coevolution with the third species.
However, previous models of exploited mutualisms have only considered host-mutualist
coevolution in the presence of non-evolving traits (Ferrière et al. 2007) or mutualistexploiter coevolution with non-evolving hosts (Jones et al. in press).
In order to investigate how the ecological and evolutionary dynamics in an
exploited mutualism are affected differently by independent species evolution, pairwise
coevolution, and three-species coevolution, we build and analyze an eco-evolutionary
model of the interactions between a host, mutualist and exploiter. Our model is based on
one of the best-studied types of exploited mutualism: pollinating seed parasite
mutualisms, which include the fig-fig wasp and yucca-yucca moth mutualisms (Dufaÿ
and Anstett 2003). In these interactions, plants (hosts) rely on pollinators (mutualists) for
fertilization, but lose a proportion of their potential seeds to pollinator and non-pollinator
(exploiter) larvae. Both pollinators and non-pollinators are dependent on the plant for
larval survival. The non-pollinators are also dependent on pollinators to ensure that seeds
are produced (yuccas) or that fruits are not aborted (figs). Larvae of both species face
exploitation and interference competition from both conspecific and heterospecific larvae
(Jaeger et al. 2001; Marr et al. 2001; Peng et al. 2005). These interactions are generally
87
obligate for all species involved, and have long phylogenetic histories (Machado et al.
1996; Pellmyr et al. 1996a; Després and Jaeger 1999), making coevolution between the
species likely.
Here, we model the evolution of mutualist and exploiter oviposition rates and
constitutive plant defense against seed destruction. In all three species, higher trait values
can be seen as 'less cooperative' or 'more parasitic', yet they are beneficial for the species
possessing the trait. Not all combinations of these traits are viable; therefore, evolution of
these traits determines whether the species persist. Oviposition rates can vary between
populations (Thompson 1997), and high oviposition rates increase mutualist and exploiter
fecundity, while reducing the number of seeds that survive in each fruit (e.g., Jaeger et al.
2001). However, oviposition rates are expected to trade off with larval competitive
ability, as egg provisioning is constrained (Smith and Fretwell 1974). Meanwhile, at high
values of plant defense, fewer of the larvae oviposited in each fruit are able to survive.
Plant traits that may act in this way include: the closed globeflower corolla, which has
been proposed to increase larval competition of the pollinating Chiastocheta flies (Ferdy
et al. 2002); the oviposition constriction zones (Shapiro and Addicott 2003), as well as
the number and distribution of infertile seeds (Addicott 1986; Ziv and Bronstein 1996;
Bao and Addicott 1998) in yuccas, which could restrict access of yucca moth larvae to
viable seeds; and ovary position in Lithophragma which has recently diversified (Kuzoff
et al. 2001), possibly in response to selection from the oviposition behavior of the
pollinating Greya moths (J.N. Thompson, pers. comm.).
88
We investigate the coevolutionary dynamics of these three traits, plant defense
and mutualist and exploiter oviposition rates, in order to answer the following three
questions. (1) How do ecological parameters affect the equilibrium population densities
and the potential for species coexistence? (2) What conditions are necessary for the
species to reach a stable coevolutionary equilibrium? (3) Do the relative rates of
evolution in the three species change the likelihood of species persistence?
Methods
Below, we first derive a model of the population dynamics of hosts, mutualists and
exploiters in order to solve for the equilibrium population densities of each species.
Analysis of these equations allows us to determine the effects of the traits of interest on
population densities, and more generally, what trait combinations result in coexistence
versus extinction. We then use the ecological equations to construct invasion fitness
equations and selection gradients for the evolving traits in each of the three species in
order to identify and classify any trait combinations that are coevolutionary singular
states. Lastly, we incorporate evolutionary rates in order to track the coevolutionary
trajectories of the evolving traits and to determine whether they lead to a convergencestable coevolutionary singular state or to extinction of one or more species.
89
Ecological Model Derivation
We first describe the changes in numbers of plants (p), mutualists (m), and exploiters (x)
over discrete time steps (∆t) in a landscape with K sites.
1
1



λ m βτ m 
λ x ετ x 
− βτ m 


c ν
1 − c xν
1 − p  − δ ∆t p
∆p = p (1 − ν )θ∆tσ 1 − e K 1 − m
p




K
K
K







(1a)
(1b)
− c νλ βτ m

 λ m βτ m  m Km
 − δ m ∆t m
∆ m = pθ∆t 
 e

 K 


∆x = pθ∆t 1 − e

− βτ m
K
1


λ m βτ m 

− c xνλ xε x

cν
 λ x ετx  e K  − δ ∆t x
 1− m
x


 K 
K






(1c)
In a unit of time, ∆t, each plant produces θ locules, each containing σ ovules,
reduced by the cost of imposing competition on larvae (1-ν). Mutualists and exploiters
search for plants at a number of sites β and ε, respectively, a proportion of the time, τ,
such that plant visits are Poisson distributed and occur at a mean rate of βτ m
ετ x
K
K
and
, respectively. During a pollinator visit to a plant, the pollinator fertilizes all the
plant’s ovules and deposits λm eggs per locule; meanwhile, each exploiter deposits λx eggs
into each fertilized locule encountered. Thus, the probability that a locule receives at least
one pollination visit is 1 − e
for mutualists and
λ x ετ x
K
− βτm
K
and the mean number of larvae per locule is
for exploiters.
λm βτ m
K
90
A number of seeds per locule is destroyed by larvae at a rate proportional to the
number of ovipositions, but reduced due to larval mortality. Larval mortality is partially
determined by constitutive plant defense, ν, which has species-specific effects on
mutualists, cm, and exploiters, cx. The probability that a mutualist larva survives the plant
defense and intraspecific competition is e
− cmνλm βτ m
K
. The probability that an exploiter
larva survives the plant defense and intraspecific competition is e
− c xνλ x βτ x
K
. Exploiter
larvae also face interspecific competition, as they feed on the proportion of seeds that

 1
 c ν λ m βτ m
have not been consumed by mutualist larvae, 1 −  m 
mutualists destroy a proportion
further proportion
the 1 − p
K
(1 c xν )λ x ετ x
(1 cmν )λm βτ m
K
K
K
. Consequently,
of seeds and exploiters destroy a
of seeds. Finally, surviving seeds must land in one of
unoccupied sites in order to germinate, and individuals have a species-
specific probability of dying δp, δm, and δx, for plants, mutualists and exploiters,
respectively.
Assuming that ∆t is small, equations (1) can be approximated with continuous
ordinary differential equations. With the additional assumption e z ≈ 1 + z , for small z,
equations (1) become:
1
1



λm βτ m 
λ x ετ x 

cν
dp
 βτm 
1 − p  − δ p
1 − c xν
= p(1 − ν )θσ 
1 − m
p
 K 

dt
K
K
 K 






(2a)
91
dm
 λ βτ m  cmνλm βτ m 
= pθ  m
 −δm m
1 −
dt
K

 K 
(2b)
1


λm βτ m 

dx
 βτ m  cmν
 λxετx 1 − c xνλ xε x  − δ x
= pθ 
 1−
x
 K 

dt
K
K 
 K 




We then group parameters, with the definitions T ≡ δ p t , P ≡ p
X ≡ ετ x
K
, bP ≡ θσ
(2c)
K
, M ≡ βτm
K
,
θβτ , b ≡ θετ , d ≡ δ m , and d ≡ δ x ,
M
X
δ p , bM ≡
δp X
δp
δp
δp
leaving the following, dimensionless per-capita growth rates:
 λ M
1 dP
= b P (1 − ν )M 1 − M
cMν
P dT

 λ X X
1 −
c Xν


(1 − P ) − 1

1 dM
= λ M bM P (1 − c M νλ M M ) − d M
M dT
 λ M
1 dX
= λ X b X PM 1 − M
c Mν
X dT


(1 − c Xνλ X X ) − d X

(3a)
(3b)
(3c)
Solving equations (3) when each is set to zero yields the equilibrium population
densities. Equations (3) can be evaluated with the density of exploiters, X, set to zero to
model the plant-mutualist interaction in the absence of exploiters. In this case, the
 λX X
1 −
c Xν


 term is removed from equation (3a) and equation (3b) is unchanged.

92
Evolutionary Dynamics
The evolving traits that we model are traits that increase the benefit of the interaction for
the possessing species while increasing the cost of the interaction for one or more of the
interacting species. In the plant, the constitutive defense, ν, decreases the proportion of
seeds lost to larvae while increasing larval mortality. However, constitutive defense also
imposes a cost for the plant in terms of the total number of ovules that can be
produced. In the mutualist and exploiter species, the traits λM and λX are the number of
eggs deposited per locule by mutualists and exploiters, respectively, and increased
numbers of eggs result in a greater number of seeds destroyed. Due to resource limitation,
the number of eggs trades off with size of eggs, and as larvae compete for survival, larger
larvae have an advantage over smaller larvae.
At the ecological equilibrium found by solving equations (3), the population
densities are given by P̂ for the plant, M̂ for the mutualist, and X̂ for the exploiter. We
can investigate invasion by mutant phenotypes at the ecological equilibrium using
invasion fitnesses (Metz et al. 1992), which give the growth rate of a rare mutant (mut)
trait in the resident (res) population.

λ Mˆ
W P ν mut = b P 1 − ν mut Mˆ 1 − M mut
 cMν
(
(
)
(
)
)
( (
(

λ Xˆ
1 − X mut
 c ν
X

))

 1 − Pˆ − 1


(
)
)
res
mut
res
'
ˆ
ˆ
νλ res
W M λ mut
= λ mut
M , λM
M P 1 − cM + cM λM − λM
M M − dM
(4a)
(4b)
93
 λM Mˆ
res
mut ˆ ˆ

WX (λmut
X , λ X ) = λ X PM 1 −
cMν


res
res ˆ
 1 − (c X + c X' (λmut
X − λ X ))νλ X X − d X


(
)
(4c)
The tradeoff between egg number and size is assumed to be linear, with slope c M' in
mutualists and c 'X in exploiters. Equations (4) can be used to investigate plant-mutualist
evolution in the absence of exploiters by setting the equilibrium density of exploiters, X̂ ,

λ Xˆ
to zero; the 1 − X mut
 c Xν

 term is lost from equation (4a) and equation (4b) is unchanged.


From equations (4), the selection gradients are
∂WP
(ν ) for the plant,
∂ν mut
∂WM
∂W
(λ M ,λ M ) for the mutualist, and mutX (λ X ,λ X ) for the exploiter. To approximate
mut
∂λM
∂λ X
the change in trait values over time, with feedback from the ecological dynamics, we use
the ‘canonical equations of adaptive dynamics’ (Dieckmann and Law 1996; Champagnat
et al. 2006):
dν
∂WP
= k P Pˆ mut
(ν )
dt
∂ν
∂WM
dλ M
= k M Mˆ mut
(λ M , λ M )
dt
∂λM
∂W
dλ X
= k X Xˆ mutX (λ X , λ X )
dt
∂λ X
(5a)
(5b)
(5c)
94
in which kP, kM and kX are evolutionary rate constants for the plant, mutualist and
exploiter, respectively. These rates account for differences among the three species in the
probability of mutation, the variance of mutation effects, and migration rates.
We find a coevolutionary singularity when
dλM
dλ X
dν
= 0,
= 0 , and
= 0.
dt
dt
dt
These singularities are coevolutionary stable states (coESSs) when both evolutionarily
and convergence stable. All potential coESSs were tested for evolutionary stability and
evolutionary convergence by inspection of pairwise invasibility plots (Geritz et al. 1997).
Results
Ecological Outcomes and Equilibrium Population Densities
In the absence of exploiters, the plant and mutualist can coexist across a range of
combinations of constitutive plant defense, ν, and the mutualist oviposition rate, λM (fig.
1). The highest population densities for both species occur at very low values of λM. At
higher values of λM, fewer seeds survive per fruit and larval mortality due to intraspecific
competition is higher, thus both plant and mutualist densities are decreased. Meanwhile,
the maximum density of plants occurs at slightly higher values of ν than that of
mutualists. At higher values of ν, fewer of the larvae developing in each fruit are able to
survive, and mutualist density is decreased. The size of the ecological coexistence region
95
increases with both species' intrinsic growth rates, such that it becomes larger as birth
rates of each species are increased, and as the death rates are decreased (not shown).
Meanwhile, the shape of the ecological coexistence region depends on the how strongly
plant defense affects larval survival and seed destruction. A stronger effect results in
increased larval mortality at high values of ν, but also in increased plant tolerance of high
oviposition rates. Thus, as the effectiveness of plant defense on mutualists, cM, is
increased, the coexistence region is shifted to a lower, smaller range of ν values, but
expands to higher λM values (fig. 1).
Exploiters can coexist with plants and mutualists across a range of values of
constitutive plant defense, ν, and mutualist oviposition rate, λM, that depends on the
exploiter oviposition rate, λX (fig. 2). This three-species coexistence region exists within a
sub-region of the plant-mutualist coexistence region (fig. 3). The region is widest for
exploiters with intermediate oviposition rates, λX, and narrows at high and low values of
λX (fig. 4). At higher values of constitutive plant defense, ν, the boundary of the
coexistence region is shared with three-species extinction region, while at lower values of
ν the boundary is shared with the exploiter-only extinction region (fig. 3). Plant and
mutualist equilibrium densities follow the same relationship with ν and λM as in the
absence of exploiters (fig. 3A,B), and decrease as exploiter oviposition rate, λX, increases
(fig. 4A,B). Exploiter density is highest at low values of ν and λM (fig. 3C), and decreases
as λX increases (fig. 4C). As with the plant-mutualist system, the coexistence region
widens as birth rates are increased, and shrinks as death rates are increased for any of the
three species. Similarly, the shape of the coexistence region is determined by the
96
effectiveness of plant defense on mutualists and exploiters, cM and cX, respectively. The
coexistence region shrinks over ν while widening over λM and λX when either cM or cX is
increased.
Evolutionary Singularities
Whether a convergence-stable coevolutionary stable state (coESS) exists within the
species coexistence region depends on the slope of the fecundity-competition tradeoff in
the mutualist, c M' . In the plant-mutualist interaction, each species evolves towards its
evolutionary isocline, the curve along which its selection gradient equals zero. At high
values of c M' , there is only one combination of constitutive plant defense, ν, and
mutualist oviposition rate, λM, values at which the plant and mutualist evolutionary
isoclines intersect, and it is always convergence stable. As the slope, c M' , becomes
shallower, a second intersection exists that is a "Garden of Eden" ESS (a point that
cannot be invaded by mutants, but that is not convergence stable) for the mutualist.
Selection at nearby points will lead away in either direction, resulting in either evolution
to the convergence-stable coESS or to extinction (fig. 5A). As cM' is decreased further,
the mutualist’s two evolutionary singularities collide (fig. 5B) and are both lost (fig. 5C).
Consequently, when the fecundity-competition tradeoff is weak, evolution will always
lead to extinction.
In the three-species system, each species evolves towards its evolutionary
isosurface, the surface in ν, λM, λX space over which its selection gradient equals zero.
97
These three surfaces tend to have one convergence-stable intersection when the slope of
the fecundity-competition trade-off is steep in both mutualists and exploiters (i.e., high
values of both cM' and c 'X ). A second intersection exists, which is a “Garden of Eden”
ESS for the exploiter (see fig. 4). At lower values of cM' and c 'X , the evolutionary
isosurfaces of the mutualist and exploiter do not intersect. When the effectiveness of
plant defense on mutualist and exploiter larvae is weak (i.e., low cM and cX), the slope of
the fecundity-competition trade-off must be steeper in order for a coESS to exist.
Equilibrium and non-Equilibrium Evolutionary Outcomes
Whenever there is no coESS, coevolution leads to the extinction of all the interacting
species. However, the existence of a convergence-stable coESS does not guarantee that
species will persist as they coevolve. When there is a "Garden of Eden" ESS for the
mutualist (e.g., fig. 5A), evolution from initial values of λM higher than this ESS point
will always result in extinction of the plant and mutualist. Similarly, when there is a
"Garden of Eden" ESS for the exploiter (e.g., fig. 4), evolution from initial values of λX
higher than this ESS point lead to extinction of all three species.
Asymmetry in the evolutionary rates of the plant and the mutualist results in an
even greater proportion of initial combinations of ν and λM leading to extinction of the
plant-mutualist interaction. When the mutualist evolves faster than the plant, the
mutualist oviposition rate, λM, can reach non-viable values before the plant evolves to
compensate. Trait combinations further from the coESS, and especially those close to the
98
boundaries of the coexistence region, evolve to extinction at lower levels of asymmetry
(fig. 6). The proportion of initial trait combinations that evolves to extinction depends on
the shape of the coexistence region, which is affected by the effectiveness of plant
defense on mutualist larvae, cM. When the coexistence region is contracted at lower cM
values, more of the initial trait combinations result in evolution to extinction (fig. 6A).
In the presence of exploiters, asymmetry between plant and mutualist
evolutionary rates can result in either three-species extinction or extinction of the
exploiter alone. Extinction of all three species occurs at high values of constitutive plant
defense, ν, at which the boundary between coexistence and extinction is shared between
the two- and three-species systems (fig. 3). On the other hand, purging of the exploiter
occurs at low values of ν, at which the boundary of the three-species coexistence region
lies within the plant-mutualist coexistence region (fig. 3). Purging of the exploiter is not
an equilibrium outcome; after extinction of the exploiter, the mutualism always returns to
a point at which the exploiter can reinvade.
Asymmetry between mutualist and exploiter evolutionary rates can lead to either
purging of the exploiter or three-species extinction, depending on whether mutualist or
exploiter evolution is faster, respectively. When the mutualist evolves significantly faster
than the exploiter and the initial combinations of mutualists and exploiters have low λM
and λX values, mutualist evolution of increased λM values drives exploiters extinct (fig. 4).
Exploiters, which still have low oviposition rates, go extinct as a result of outcompetition by mutualists for seeds. However, as with the exploiter purging outcome
described above, exploiters can reinvade after extinction. On the other hand, when the
99
exploiter evolves faster than the exploiter and the initial mutualists have high values of
λM, selection on exploiters can lead to very high values of λX, and thus high proportions of
seeds destroyed, leading to three-species extinction.
Discussion
Empirical (Thompson and Pellmyr 1992; Thompson 1998) and theoretical (Thrall et al.
2007) studies have shown the importance of the community context of pairs of
coevolving species for understanding patterns of selection on traits and the evolution of
interaction specificity. Here, we have moved beyond the typical pairwise coevolutionary
perspective in order to investigate how the population dynamics within the community
are affected by coevolution with interacting species as opposed to interactions with nonevolving species. In our adaptive dynamics model, we analyzed the eco-evolutionary
dynamics of a host plant, its pollinating seed predator mutualist and a seed predator
exploiter. We focused on traits in each species that alter the costs and benefits of the
interaction: constitutive plant defense that decreases larval seed destruction and larval
survival, at a cost to the total seeds produced; and mutualist and exploiter oviposition
rates that increase both insect fecundity and seed destruction. As not all combinations of
these traits are ecologically viable, we find that evolution of these traits can lead to the
persistence of the community, purging of the exploiter, or extinction of the entire
community.
100
The coevolutionary outcome depends on both interspecific and intraspecific
factors. The species-specific effectiveness of plant defense on the mutualist and on the
exploiter determine the range of trait values in each species that allow species
coexistence, with stronger constitutive defense enabling plants to tolerate higher numbers
of mutualist and exploiter larvae. Meanwhile, the slope of the trade-off between fecundity
and competition in mutualists and exploiters determines the endpoint of coevolutionary
trajectories. Steeper slopes of this trade-off result in selection for lower oviposition rates,
decreasing the destruction of host plant seeds.
Species extinction can occur through two routes. First, there may be no
coevolutionary stable state (coESS), with selection always favoring nonviable trait
combinations. Second, the coESS may be unreachable from some initial trait
combinations due either to asymmetric evolutionary rates or to a second, repelling
evolutionary singular state. Surprisingly, the major threat to the persistence of the
mutualism in both cases is from evolution of the mutualist, not the exploiter. Despite the
fact that in this model the mutualism is constrained to have a net effect for the plant that
is either positive or neutral, extinction can occur when the net benefits of the interaction
are sufficiently reduced since the interaction is obligate.
Limiting the costs of mutualists and exploiters
Mutualisms are defined by their net beneficial effects, but mutualisms often involve
significant costs (Bronstein 2001a). Moreover, it is to the advantage of individuals to get
101
as much as they can from their mutualist partners, such that mutualisms are better
described as mutual exploitation than as cooperation. In our model, the host plant trait
can evolve to reduce exploitation by mutualists and exploiters. Meanwhile, mutualists
and exploiters can evolve to increase the level of their exploitation. Our analysis
demonstrates that higher species-specific effectiveness of plant defense against mutualist
and exploiter larvae enables coexistence over a wider range of plant, mutualist and
exploiter traits, especially at trait combinations likely at the origin of the interaction (low
constitutive defense, high oviposition rates). Furthermore, evolution from these initial
trait value combinations is more likely to lead to the coESS when plant defense is
effective.
While not modeled as such, the strength of constitutive plant defense could be
related to the slope of the fecundity-competition trade-off in the mutualist and exploiter.
Thus, in addition to increasing the size of the ecological coexistence region, higher levels
of defense would increase the chance of the selection gradients of each species
intersecting at a convergence-stable coESS. Partner choice mechanisms may act in just
such a way, as they have been suggested to decrease the costs of mutualism by excluding
(Kiers and van der Heijden 2006) or decreasing cooperation (Simms et al. 2006) with
non-rewarding partners, and thus selecting for more beneficial partners (Denison 2000;
Foster and Wenseleers 2006). However, we propose that explicit choice between partners
is not necessary as long as defense mechanisms give a competitive advantage to the more
beneficial partners.
102
Fitness trade-offs and evolutionary persistence
Fitness trade-offs result in trait values that are evolutionary equilibria. The stability and
convergence properties of the equilibrium depends on the shape of the trade-off, with a
convergence-stable ESS, a "Garden of Eden" ESS, an evolutionary branching point, or an
invasible repeller as possible outcomes (de Mazancourt and Dieckmann 2004). We
assumed a linear trade-off between fecundity and competition in mutualists and
exploiters as the simplest case, particularly for a scenario in which competition is for a
limiting resource (seeds) that cannot be substituted. This trade-off generally leads to a
convergence-stable coESS, which is joined by a "Garden of Eden" ESS at higher
oviposition rates as the trade-off slope is decreased. However, when the trade-off is too
weak and the disadvantages of decreased competitive ability do not overcome the
advantages of higher larval number, these two singularities collide, leading to
evolutionary suicide as a result of runaway selection on oviposition rate.
Under certain conditions, competition could be non-linear, with individuals
possessing distant trait values having reduced competitive effects on each other. This
configuration is likely to lead to evolutionary branching. Indeed, ancestral seed predators
may have diversified to reduce competition for seeds. In the globeflower mutualism,
larval competition has been proposed to have driven the adaptive radiation of
Chiastocheta flies through changes in oviposition timing that result in changed egg
mortality rates (Ferdy et al. 2002; Després and Cherif 2004). While we assumed larvae to
be obligately dependent on seeds, it is possible that competition could cause
103
diversification onto other resources. In the Prodoxidae, which include yucca moths and
Greya, larval feeding on vegetative tissue in some exploiter species appears to be derived
from the ancestral state of feeding on seeds (Brown et al. 1994; Pellmyr et al. 1996b); this
diversification to a nutritionally inferior resource may also have followed intense
competition for seeds.
Causes and consequences of asymmetric evolutionary rates
Asymmetries in the rates of evolution between species can result from differences in
mutation rates, migration rates and genetic system. Additionally, evolutionary rates may
vary within a species across time and space. Asymmetric evolutionary rates have mostly
been considered in the context of host-parasite (e.g., Kaltz and Shykoff 1998) and
predator-prey (e.g., Dieckmann et al. 1995) interactions, which demonstrate Red Queen
dynamics (Van Valen 1973). However, a previous investigation of the effects of
asymmetric evolutionary in a mutualism by Bergstrom and Lachmann (2003) predicted
that the species with the slower evolutionary rate should gain the larger benefit from the
interaction. Importantly, their model considered evolution of investment into mutualism.
In our model of evolution of exploitation traits, we see a very different consequence of
asymmetric evolutionary rates: asymmetry increases the chance of extinction. Thus,
significance of asymmetric evolutionary rates depends not only on the type of interaction
under consideration, but on the types of evolving traits.
104
In our host-mutualist model, we have focused on the case in which mutualists
evolve faster than plants. This case may be the most likely, due to the shorter generation
time of the mutualists. Additionally, oviposition rate is likely to evolve through changing
the resource distribution between egg size and number, which in some insect species is a
plastic behavior (reviewed in Fox and Czesak 2000), and thus should be fast in
comparison to evolution of plant defense traits requiring changes in physical structure.
For the three-species interaction, mutualists are still likely to be the fastestevolving species. While exploiters and mutualists are generally closely related and have
the same evolving trait in our model, exploiters are predicted to have high migration rates
(Wilson et al. 2003), which could limit their potential for local adaptation.
Acknowledgements
E.I.J.’s research was supported by the National Science Foundation DEB-0806836.
References
Addicott, J. F. 1986. Variation in the costs and benefits of mutualism - the interaction
between yuccas and yucca moths. Oecologia 70:486-494.
Agrawal, A. A., D. D. Ackerly, F. Adler, A. E. Arnold, C. Caceres, D. F. Doak, E. Post,
P. J. Hudson, J. Maron, K. A. Mooney, M. Power, D. Schemske, J. Stachowicz, S.
Strauss, M. G. Turner, and E. Werner. 2007. Filling key gaps in population and
community ecology. Frontiers in Ecology and the Environment 5:145-152.
Bao, T., and J. F. Addicott. 1998. Cheating in mutualism: defection of Yucca baccata
against its yucca moths. Ecology Letters 1:155-159.
105
Bergstrom, C. T., and M. Lachmann. 2003. The Red King effect: When the slowest
runner wins the coevolutionary race. Proceedings of the National Academy of
Sciences of the United States of America 100:593-598.
Best, A., A. White, and M. Boots. 2009. The implications of coevolutionary dynamics to
host-parasite interactions. American Naturalist 173:779-791.
Bronstein, J. L. 2001a. The costs of mutualism. American Zoologist 41:825-839.
—. 2001b. The exploitation of mutualisms. Ecology Letters 4:277-287.
Brown, J. M., O. Pellmyr, J. N. Thompson, and R. G. Harrison. 1994. Phylogeny of
Greya (Lepidoptera, Prodoxidae), based on nucleotide sequence variation in
mitochondrial cytochrome oxidase I and cytochrome oxidase II: Congruence with
morphological data. Molecular Biology and Evolution 11:128-141.
Champagnat, N., R. Ferrière, and S. Méléard. 2006. Unifying evolutionary dynamics:
From individual stochastic processes to macroscopic models. Theoretical
Population Biology 69:297-321.
de Mazancourt, C., and U. Dieckmann. 2004. Trade-off geometries and frequencydependent selection. American Naturalist 164:765-778.
Denison, R. F. 2000. Legume sanctions and the evolution of symbiotic cooperation by
rhizobia. American Naturalist 156:567-576.
Després, L., and M. Cherif. 2004. The role of competition in adaptive radiation: A field
study on sequentially ovipositing host-specific seed predators. Journal of Animal
Ecology 73:109-116.
Després, L., and N. Jaeger. 1999. Evolution of oviposition strategies and speciation in the
globeflower flies Chiastocheta spp. (Anthomyiidae). Journal of Evolutionary
Biology 12:822-831.
Dieckmann, U., and R. Law. 1996. The dynamical theory of coevolution: A derivation
from stochastic ecological processes. Journal of Mathematical Biology 34:579612.
Dieckmann, U., P. Marrow, and R. Law. 1995. Evolutionary cycling in predator-prey
interactions: Population dynamics and the Red Queen. Journal of Theoretical
Biology 176:91-102.
Dufaÿ, M., and M. C. Anstett. 2003. Conflicts between plants and pollinators that
reproduce within inflorescences: Evolutionary variations on a theme. Oikos
100:3-14.
106
Ferdy, J. B., L. Després, and B. Godelle. 2002. Evolution of mutualism between
globeflowers and their pollinating flies. Journal of Theoretical Biology 217:219234.
Ferrière, R., J. L. Bronstein, S. Rinaldi, R. Law, and M. Gauduchon. 2002. Cheating and
the evolutionary stability of mutualisms. Proceedings of the Royal Society of
London Series B-Biological Sciences 269:773-780.
Ferrière, R., M. Gauduchon, and J. L. Bronstein. 2007. Evolution and persistence of
obligate mutualists and exploiters: Competition for partners and evolutionary
immunization. Ecology Letters 10:115-126.
Foster, K. R., and T. Wenseleers. 2006. A general model for the evolution of mutualisms.
Journal of Evolutionary Biology 19:1283-1293.
Fox, C. W., and M. E. Czesak. 2000. Evolutionary ecology of progeny size in arthropods.
Annual Review of Entomology 45:341-369.
Geritz, S. A. H., J. A. J. Metz, E. Kisdi, and G. Meszéna. 1997. Dynamics of adaptation
and evolutionary branching. Physical Review Letters 78:2024-2027.
Gomulkiewicz, R., S. L. Nuismer, and J. N. Thompson. 2003. Coevolution in variable
mutualisms. American Naturalist 162:S80-S93.
Holland, J. N., and J. L. Bronstein. 2008. Mutualism, Pages 2485-2491 in S. E.
Jørgensen, and B. D. Fath, eds. Encyclopedia of Ecology. Oxford, Elsevier.
Jaeger, N., F. Pompanon, and L. Després. 2001. Variation in predation costs with
Chiastocheta egg number on Trollius europaeus: How many seeds to pay for
pollination? Ecological Entomology 26:56-62.
Jones, E. I., R. Ferriere, and J. L. Bronstein. in press. Eco-evolutionary dynamics of
mutualists and exploiters. American Naturalist.
Kaltz, O., and J. A. Shykoff. 1998. Local adaptation in host-parasite systems. Heredity
81:361-370.
Kiers, E. T., and M. G. A. van der Heijden. 2006. Mutualistic stability in the arbuscular
mycorrhizal symbiosis: Exploring hypotheses of evolutionary cooperation.
Ecology 87:1627-1636.
Kuzoff, R. K., L. Hufford, and D. E. Soltis. 2001. Structural homology and
developmental transformations associated with ovary diversification in
Lithophragma (Saxifragaceae). American Journal of Botany 88:196-205.
107
Law, R., J. L. Bronstein, and R. G. Ferrière. 2001. On mutualists and exploiters: Plantinsect coevolution in pollinating seed-parasite systems. Journal of Theoretical
Biology 212:373-389.
Machado, C. A., E. A. Herre, S. McCafferty, and E. Bermingham. 1996. Molecular
phylogenies of fig pollinating and non-pollinating wasps and the implications for
the origin and evolution of the fig-fig wasp mutualism. Journal of Biogeography
23:531-542.
Marr, D. L., M. T. Brock, and O. Pellmyr. 2001. Coexistence of mutualists and
antagonists: Exploring the impact of cheaters on the yucca - yucca moth
mutualism. Oecologia 128:454-463.
Metz, J. A. J., R. M. Nisbet, and S. A. H. Geritz. 1992. How should we define fitness for
general ecological scenarios. Trends in Ecology & Evolution 7:198-202.
Morris, W. F., J. L. Bronstein, and W. G. Wilson. 2003. Three-way coexistence in
obligate mutualist-exploiter interactions: The potential role of competition.
American Naturalist 161:860-875.
Nuismer, S. L., and M. Doebeli. 2004. Genetic correlations and the coevolutionary
dynamics of three-species systems. Evolution 58:1165-1177.
Pellmyr, O., J. Leebens-Mack, and C. J. Huth. 1996a. Non-mutualistic yucca moths and
their evolutionary consequences. Nature 380:155-156.
Pellmyr, O., J. N. Thompson, J. M. Brown, and R. G. Harrison. 1996b. Evolution of
pollination and mutualism in the yucca moth lineage. American Naturalist
148:827-847.
Peng, Y. Q., D. R. Yang, and Q. Y. Wang. 2005. Quantitative tests of interaction between
pollinating and non-pollinating fig wasps on dioecious Ficus hispida. Ecological
Entomology 30:70-77.
Shapiro, J. M., and J. F. Addicott. 2003. Regulation of moth-yucca mutualisms: Mortality
of eggs in oviposition-induced 'damage zones'. Ecology Letters 6:440-447.
Simms, E. L., D. L. Taylor, J. Povich, R. P. Shefferson, J. L. Sachs, M. Urbina, and Y.
Tausczik. 2006. An empirical test of partner choice mechanisms in a wild legumerhizobium interaction. Proceedings of the Royal Society B-Biological Sciences
273:77-81.
Smith, C. C., and S. D. Fretwell. 1974. Optimal Balance between Size and Number of
Offspring. American Naturalist 108:499-506.
108
Strauss, S. Y., and R. E. Irwin. 2004. Ecological and evolutionary consequences of
multispecies plant-animal interactions. Annual Review of Ecology Evolution and
Systematics 35:435-466.
Thompson, J. N. 1997. Evaluating the dynamics of coevolution among geographically
structured populations. Ecology 78:1619-1623.
—. 1998. The population biology of coevolution. Researches on Population Ecology
40:159-166.
Thompson, J. N., and O. Pellmyr. 1992. Mutualism with pollinating seed parasites amid
co-pollinators: Constraints on specialization. Ecology 73:1780-1791.
Thrall, P. H., M. E. Hochberg, J. J. Burdon, and J. D. Bever. 2007. Coevolution of
symbiotic mutualists and parasites in a community context. Trends in Ecology &
Evolution 22:120-126.
Van Valen, L. 1973. A new evolutionary law. Evolutionary Theory 1:1-30.
Wilson, W. G., W. F. Morris, and J. L. Bronstein. 2003. Coexistence of mutualists and
exploiters on spatial landscapes. Ecological Monographs 73:397-413.
Yu, D. W. 2001. Parasites of mutualisms. Biological Journal of the Linnean Society
72:529-546.
Ziv, Y., and J. L. Bronstein. 1996. Infertile seeds of Yucca schottii: A beneficial role for
the plant in the yucca-yucca moth mutualism? Evolutionary Ecology 10:63-76.
109
Figure 1: Plant-mutualist coexistence regions and equilibrium population densities.
The ecological outcome of the plant-mutualist interaction depends on the combinations of
constitutive plant defense, ν, and the mutualist oviposition rate, λM. The species coexist in
the colored region and are both extinct in the grey region. The shape of the coexistence
region is determined by the effectiveness of plant defense on mutualists, cM; (A,B) cM =2,
(C,D) cM =5. The population densities of the plant (A,C) and mutualist (B,D) are shown
within the coexistence region, with areas of higher densities having warmer colors (red)
and areas of lower densities having cooler colors (blue). The species evolve to their
evolutionary isoclines (dashed = plant, solid = mutualist), which do not necessarily lead
to higher population densities. Parameter values: bP=50, bM=10, dM=5, c M' =0.5.
Figure 2: Plant-mutualist-exploiter coexistence region. All three species coexist at
combinations of constitutive plant defense, ν, mutualist oviposition rate, λM, and exploiter
oviposition rate, λX, within the enclosed region. Each species evolves to a surface
determined by its selection gradient; these surfaces intersect at a coESS (red point). The
coESS is convergence stable for all three species, as shown in the pairwise invasion plots
for the plant (B), mutualist (C), and exploiter (D). Mutant traits can invade when in the
black region, but not in the grey region; consequently, for each species, selection leads to
the coESS from both higher and lower trait values. Parameter values: bP=50, bM= bX =10,
dM= dX =5, cM = cX =5, cM' = c 'X =0.3.
110
Figure 3: Plant-mutualist-exploiter equilibrium population densities and
evolutionary outcomes at coESS exploiter oviposition rate, λX. Slices of the three-
species coexistence region (fig. 2) are superimposed on the plant-mutualist ecological
outcomes. The population densities of the plant (A), mutualist (B) and exploiter (C) are
shown within the coexistence region, with areas of higher densities having warmer colors
(red) and areas of lower densities having cooler colors (blue). In the light gray region, the
exploiter is out-competed by the mutualist and goes extinct, leaving the mutualism intact.
In the dark gray region, all three species go extinct. The two- and three-species systems
share a boundary between coexistence and extinction at high values of constitutive plant
defense, ν; otherwise, evolution out of the three-species coexistence regions leads to the
extinction of the exploiter alone. The species evolve to their evolutionary isoclines
(dashed = plant, solid = mutualist) and reach a convergence stable coESS at the
intersection. Parameters as in fig. 2.
Figure 4: Plant-mutualist-exploiter equilibrium population densities and
evolutionary outcomes at coESS constitutive plant defense, ν. Slices of the three-
species coexistence region (fig. 2). The population densities of the plant (A), mutualist
(B) and exploiter (C) are shown within the coexistence region, with areas of higher
densities having warmer colors (red) and areas of lower densities having cooler colors
(blue). The species evolve to their evolutionary isoclines (solid = mutualist, dot dashed =
exploiter) and reach a convergence stable coESS at the lower left intersection. At the
upper right intersection, the exploiter evolves away from the intersection, towards either
111
extinction or the coESS. Parameter values: bP=50, bM= bX =10, dM= dX =5, cM = cX =2,
cM' = c 'X =0.5.
Figure 5: Plant-mutualist evolutionary outcomes depending on the slope of the
mutualist's fecundity-competition tradeoff. The top row shows the ecological and
evolutionary outcomes of the plant-mutualist interaction across combinations of
constitutive plant defense, ν, and the mutualist oviposition rate, λM, when the slope of the
trade-off between mutualist oviposition rate and larval competitive ability is (A) c M' =0.2,
(B) c M' =0.15, and (C) c M' =0.1. The species coexist in the white region and are extinct in
the grey region. The species evolve to their evolutionary isoclines (dashed = plant, solid
= mutualist), and reach a coESS when the intersection of the isoclines is convergence
stable for both species (e.g., Ai). The second and third rows show the pairwise
invasibility plots for the plant trait and the mutualist trait, respectively. Mutant traits can
invade when in the black region, but not in the grey region. When the singular point
cannot be invaded, it is an evolutionary stable state (e.g., Aii, Aiii, Bii, Biii, Cii).
However, the upper ESS in Aiii and the ESS in Biii are not convergence stable;
evolution from nearby the former, "Garden of Eden" singularity will lead away from it in
either direction, and evolution from above the latter singularity will lead to higher trait
values. Parameter values: bP=50, bM=10, dM=5, cM=5.
Figure 6: Evolutionary outcomes depending on the relative evolutionary rates. As in
fig. 1, the plots show combinations of constitutive plant defense, ν, and the mutualist
112
oviposition rate, λM, that lead to ecological coexistence (inner region) or extinction (solid
grey region), and the evolutionary isoclines (dashed = plant, solid = mutualist), when the
effectiveness of plant defense on larval survival is (A) cM=2, and (B) cM=5. Within the
coexistence region, the shading shows whether evolutionary trajectories lead to the
coESS or to extinction. Initial trait combinations in the white regions always evolve to
the coESS when the evolutionary rate of the plant, kP, is varied from 10-1 (equal to the
mutualist's rate) to 10-5 (highly asymmetric evolution). Darker shading signifies that less
asymmetry in evolutionary rate is required for evolution to extinction. Parameter values:
bP=50, bM=10, dM=5, c M' =0.3, kM=10-1.
113
Figure 1
A
B
C
D
114
Figure 2
A
B
C
D
115
Figure 3
A
B
C
116
Figure 4
A
B
C
117
Figure 5
Ai
Bi
Ci
Aii
Bii
Cii
Aiii
Biii
Ciii
118
Figure 6
A
B
119
APPENDIX C
OPTIMAL FORAGING WHEN PREDATION RISK INCREASES WITH PATCH
RESOURCES: AN ANALYSIS OF POLLINATORS AND AMBUSH PREDATORS
In revision for Oikos
120
Optimal foraging when predation risk increases with patch resources: an analysis of
pollinators and ambush predators
Emily I. Jones
Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona 85721
E-mail: [email protected]
Fax: (520) 621-9190
121
ABSTRACT:
Pollinators and their predators share innate and learned preferences for high quality
flowers. Consequently, pollinators are more likely to encounter predators when visiting
the most rewarding flowers. I present a model of how different pollinator species can
maximize lifetime resource gains depending on the density and distribution of predators,
as well as their vulnerability to capture by predators. For pollinator species that are
difficult for predators to capture, the optimal strategy is to visit the most rewarding
flowers as long as predator density is low. At higher predator densities and for pollinators
that are more vulnerable to predator capture, the lifetime resource gain from the most
rewarding flowers declines and the optimal strategy depends on the predator distribution.
In some cases, a wide range of floral rewards provides near-maximum lifetime resource
gains, which may favor generalization if searching for flowers is costly. In other cases, a
low flower reward level provides the maximum lifetime resource gain and so pollinators
should specialize on less rewarding flowers. Thus, the model suggests that predators can
have qualitatively different top-down effects on plant reproductive success depending on
the pollinator species, the density of predators, and the distribution of predators across
flower reward levels.
122
Introduction
Pollinator preferences play an important role in both selection on floral traits (e.g., Medel
et al. 2003, Gomez et al. 2006) and plant community assembly (Sargent and Ackerly
2008). Thus, understanding the basis of pollinator preferences will advance our
understanding of the ecological and evolutionary dynamics acting in plant communities.
As a rule, pollinators attempt to visit the most rewarding flowers (i.e., those with the
highest nectar volume, sugar concentration and/or pollen availability) both within and
between plant species, through innate and learned preferences for floral traits that
correlate with reward level, such as color (Menzel 1985), display size (Ashman and
Stanton 1991, Cohen and Shmida 1993), symmetry (Møller 1995), odor (Andersson
2003), age (Higginson et al. 2006), and sex (Shykoff and Bucheli 1995).
However, recently it has been found that ambush predators such as crab spiders
(Thomisidae) can use some of the same floral characters to recognize flower quality,
including flower symmetry (Wignall et al. 2006), odor (Heiling et al. 2004), and age
(Chien and Morse 1998). Using this information, they will move to the more rewarding
plant individuals (Heiling and Herberstein 2004) and species (Schmalhofer 2001).
Consequently, the most rewarding flowers may become the most dangerous flowers.
How then should pollinators forage to maximize the rewards they collect over their
lifetime?
Crab spiders are often camouflaged, making detection by pollinators difficult
(Chittka 2001, Théry and Casas 2002) and increasing vigilance is likely to decrease
foraging efficiency (Ings and Chittka 2008). Indeed, field and laboratory experiments
123
have shown that pollinators sometimes, but do not always avoid plants harboring an
ambush predator (Heiling and Herberstein 2004, Robertson and Maguire 2005, Reader et
al. 2006). Since identifying individual flowers that harbor crab spiders appears to be
either unreliable or costly, another solution would be for pollinators’ flower preferences
to shift after encounters with crab spiders. Honey bees will avoid locations where they
have encountered a spider, a simulated attack, or a dead honey bee (Dukas 2001) and
bumble bees will avoid locations with dead bumble bees (Abbott 2006), although the
extent to which bees generalize estimates of predation risk to other locations is not
known. However, experiments by Ings and Chittka (2008) have demonstrated that after
experiencing attacks by camouflaged predator models, bumble bees are more likely to
falsely reject similar-appearing feeders lacking a predator model. Moreover, bumble bees
will switch to low-reward feeders if high-reward feeders are perceived as dangerous, and
will slow or quit foraging if even low-reward feeders are perceived as dangerous (Jones
and Dornhaus, in prep).
If pollinators avoid flowers that have a higher probability of harboring a predator,
predators of pollinators should have a much more widespread effect on plant
reproduction within the community than would be expected if pollinators avoid only
those flowers where a predator has been detected. Indeed, game theoretic models of patch
choice by predators and prey have suggested that predators should prefer patches that are
rich in the resources consumed by their prey, while prey themselves should be uniformly
distributed among patches (Hugie and Dill 1994, Sih 1998). However, empirical studies
124
of the top-down effects of predators on plant reproduction have not taken into account the
consequences of a relationship between flower reward level and predation risk.
Here, I demonstrate that in order to maximize pollinator lifetime resource gain,
pollinator foraging preferences should depend on the type of relationship between
predation risk and flower reward level, as well as on the density of predators and the
vulnerability of pollinators. To do this, I present a model of the lifetime foraging gains of
pollinators visiting flowers of differing reward levels. I then evaluate the model for a
range of potential predator distributions, using field data on the risk of predation by crab
spiders, a major predator of pollinators on flowers, to show that the differences between
distributions are relevant for natural systems. Finally, I discuss the results in the context
of the methodology of measuring top-down effects of crab spiders and of the
development of more advanced predator and prey patch-choice games.
Methods
I modeled the lifetime resource gain of pollinators foraging at a specific reward level
when the probability of encountering a crab spider at a flower depends on the reward
level. The probability of a pollinator encountering a crab spider (E(r)) depends on the
proportion of crab spiders per flower (s) and the influence of flower reward level on crab
spider choice:
(1)
E (r ) = csr c −1 ,
in which the constant c determines the crab spiders’ preference function. If crab spiders
are indifferent to flower reward level, c=1 and the distribution of crab spiders across
125
flowers in uniform. When c>1, crab spiders have a preference for more rewarding
flowers; this preference decelerates with reward level for 1<c<2, increases linearly with
reward level for c=2, and accelerates with reward level for c>2. As E(r) must not exceed
1 for any value r, equation (1) can only be evaluated over 0≤r≤1 and 0≤s≤1/c.
The expected lifetime of a pollinator, and thus the lifetime resource gain,
decreases as the probability of encountering a crab spider at the given reward level (csrc1
), the prey capture probability (p), and the external mortality probability (m) increase.
Lifetime resource gain, as in Clark and Dukas (1994), is given as the sum of the resource
gain from a series of flower visits that is limited only by a constant chance of mortality:
(2)
L(r ) = r (1 − cpsr c −1 − m) + r (1 − cpsr c −1 − m) 2 + Κ =
r (1 − cpsr c −1 − m)
.
cpsr c −1 + m
When the overall chance of mortality per flower visit is low, equation (2) can be
approximated as:
(3)
L(r ) ≈
r
cpsr
c −1
+m
.
To investigate the effect of potential predator distributions on pollinator lifetime
resource gain, equation (2) was evaluated at four values of the constant c, so that the
chance of encountering a crab spider on a flower, from equation (1), was either uniform
(c=1), decelerated with reward level (c=1.5), increased linearly with reward level (c=2),
or decelerated with reward level (c=2.5) (fig. 1). In addition, equation (2) was evaluated
for patches with two levels of predator density and for two types of pollinators. Estimates
of the percentage of pollinator foraging sites occupied by crab spiders vary for crab
126
spider species, population, and year, with reports as low as <1% (Morse 2007), between
6.8% and 12.5% (Suttle 2003), and between 2% and 30% (Robertson and Maguire 2005).
Thus, ‘low predator’ patches with s=0.01 and ‘high predator’ patches with s=0.10 reflect
natural conditions. The vulnerability to crab spider attack varies greatly between
pollinator species and may depend on a combination of morphological and behavioral
variables. In particular, body size of the pollinator appears be a significant component of
vulnerability; in a study by Morse (1979), bumble bees were only captured in 1.1% of
opportunities for attack, while the much smaller syrphid flies were captured in 21.4% of
opportunities for attack. Therefore, I contrasted relatively ‘invulnerable’ pollinators,
characterized by p=0.01, and ‘vulnerable’ pollinators, characterized by p=0.20. While
here parameterized for individual flowers of different reward levels, the model could also
be applied to patches with different reward and spider occupation levels.
Results
As an increasing proportion of foraging sites are occupied by predators, the lifetime
resource gain achieved by pollinators declines. For relatively invulnerable pollinators
foraging in a patch in which very few predators are present, all four predator distributions
give the same optimal foraging strategy for pollinators: specialize on the most rewarding
flowers (fig. 2A, Table 1). More interestingly, increasing the density of predators also
changes which flower reward level yields the maximum lifetime resource gain (fig. 2B,
Table 1). A uniform (c=1), decelerating (c=1.5), or linear (c=2) relationship between
predators and flower reward level continues to favor the pollinator strategy of
127
specializing on the most rewarding flowers. Meanwhile, if there is an accelerating (c=2.5)
relationship between predators and reward level, the maximum of the lifetime resource
gain curve occurs at a lower reward level and a wider range of reward levels provide
nearly the maximum lifetime resource gain (Table 1). Thus, the optimal pollinator
strategy may become generalization over a range of lower-reward flowers, especially if
searching costs (Heinrich 1972) are high enough to negate the slight advantage of more
rewarding sites.
More vulnerable pollinators show qualitatively similar responses to predator
density and distribution (fig. 2C, D). However, the effect of predator distribution remains
significant at the lower predator density. Additionally, the difference between optimal
pollinator strategies is even more pronounced than for the less vulnerable pollinators. At
both predator densities considered, the range of pollinator reward preferences, r, that
approach the maximum lifetime gain increases as the relationship between predators and
flower reward level shifts from uniform to linear (Table 1). Thus, even when predators
are relatively rare, vulnerable pollinators may benefit from generalizing across reward
levels. Moreover, if the chance of encountering a predator accelerates with flower reward
level, low-reward flowers always provide the maximum lifetime gain for vulnerable
pollinators (Table 1).
Discussion
The analysis presented here demonstrates that when predation risk is positively correlated
with flower reward level, optimal pollinator foraging preferences should shift to less
128
rewarding flowers. These shifts should be greater for more vulnerable pollinator species,
at higher predator densities, and when the distribution of predators is more weighted
towards high-reward flowers. Under the same circumstances, pollinators should generally
also increase the range of visited reward levels. However, for vulnerable pollinators when
predators are at a high density and have an accelerating relationship with flower reward
level, specialization on low-reward flowers is optimal.
While Dukas et al. (2005) suggested that pollinators often do not have enough
experience to accurately assess predator abundance within a patch, bumble bees may be
able to learn optimal foraging preferences indirectly by observing more experienced
foragers (Baude et al. 2008). Moreover, empirical results from studies with similar
predator densities and pollinator vulnerabilities as used to parameterize the model are in
agreement with the model predictions on when foraging sites with predators should be
avoided. In areas with low crab spider density, patches with crab spiders were visited by
large bumble bees, but were avoided by the smaller, and thus more vulnerable, honey
bees (Dukas and Morse 2005), or by both honey bees and small bumble bees (Dukas and
Morse 2003). Additionally, Schmalhofer (2001) found that small pollinators were more
likely to visit smaller and less rewarding flower patches than those preferred by both
large pollinators and crab spiders; however, Schmalhofer suggested that this pattern could
be explained by competitive exclusion. Experiments in areas with higher crab spider
densities have found avoidance of flowers with crab spiders across pollinator species
(Suttle 2003, Robertson and Maguire 2005). However, it is not clear from these studies
129
whether pollinator avoidance of flowers with predators resulted from increased vigilance
or from a change in preferences for flowers.
Previous experiments on the effects of ambush predators on pollinator behavior
and plant reproductive success have used pairs of plants (e.g. Suttle 2003) or patches (e.g.
Dukas and Morse 2003, Dukas and Morse 2005) with and without crab spiders. However,
the model presented here suggests that within a patch, the presence of crab spiders on
some high-reward flowers may decrease pollinator visitation to all the high-reward
flowers, a foraging response that might or might not change the amount of pollinator
visitation at the patch level. Therefore, studies comparing paired plants within a patch
may have underestimated the effect of crab spiders, while studies comparing paired
patches may have ignored changes of pollinator preferences within the patches with
predators. Additionally, while predators may reduce pollination of both paired plants,
they sometimes also capture plant enemies and thus can have positive effects on the
occupied plant by reducing insect herbivory and seed predation (Louda 1982, Romero
and Vasconcellos-Neto 2004). Consequently, in order to determine the effects of crab
spiders specifically on pollination, it may be necessary to compare the distributions of
pollinator visits across flowers of different reward levels between patches with and
without crab spiders.
The game between pollinators and ambush predators differs from traditional
models of predator and prey patch choice (reviewed in Brown and Kotler 2007), as the
latter assume that predators are equally or more mobile than their prey. Ambush predators
such as crab spiders, while mobile, move more slowly and over shorter distances than
130
their prey (Morse 2007), and thus cannot react to short-term changes in pollinator
preference. While a number of studies have suggested that the probability that a flower is
occupied by a crab spider should increase with the reward level of the flower (Morse and
Fritz 1982, Morse and Stephens 1996, Chien and Morse 1998, Schmalhofer 2001,
Heiling, Cheng and Herberstein 2004, Heiling and Herberstein 2004, Wignall, Heiling,
Cheng and Herberstein 2006), there has been no quantitative investigation of the
distribution of predators across flower reward levels. As pollinators shift their foraging
preferences in response to predation risk, it should become advantageous for some
predators to move to less rewarding flowers. Indeed, a minority of crab spiders
consistently choose lower quality foraging sites (Morse and Fritz 1982, Morse and
Stephens 1996) and prey abundance generally plays a role in choice of ambush site
(Morse 1988).
However, the optimal distribution for predators is likely to depend on a
combination of many factors that affect the distribution of their principal prey, requiring a
more formal game-theoretic analysis. These factors are likely to include the vulnerability
of prey, the experience levels and learning abilities of prey, the degree to which naïve
foragers are easier to capture (as in Armitage 1965), and whether foragers are attempting
to maximize lifetime resource gain or rate of resource gain (Cartar 1991). The
pollinator/ambush predator system also provides an interesting extension to predator-prey
patch choice games through the feedback between predator-prey interactions on the
patches themselves. As pollinator preferences change, the distribution of flower reward
levels will change over the short-term through depletion of pollen and nectar. Unlike
131
other systems, however, pollinators will increase the long-term abundance of the
preferred flower types through increasing their reproduction.
The model results described here extend the growing body of literature on
nonconsumptive effects of predators (e.g., Orrock et al. 2008, Peckarsky et al. 2008,
Schmitz et al. 2008), which highlights the important consequences of antipredator
behavior for community and evolutionary dynamics. In addition to decreasing pollinator
density and causing pollinators to avoid individual plants, ambush predators may lead to
shifts in pollinator preferences, such as towards generalization across reward levels or
specialization on low-reward flowers. These preference changes could affect competition
among plant species and trait selection within plant species. In order to better understand
these potential consequences, studies are needed on ambush predator distribution and the
level at which pollinators respond to predation risk, whether at the level of individual
plants, plant phenotypes, plant species, or patches.
Acknowledgements
I thank A. Houston, N. Waser, A. Dornhaus, D. Papaj, J. Bronstein and members of the
Bronstein lab for discussion of the model and helpful comments on the manuscript.
132
References
Abbott, K. R. 2006. Bumblebees avoid flowers containing evidence of past predation
events. - Canadian Journal of Zoology-Revue Canadienne De Zoologie 84: 1240-1247.
Andersson, S. 2003. Foraging responses in the butterflies Inachis io, Aglais urticae
(Nymphalidae), and Gonepteryx rhamni (Pieridae) to floral scents. - Chemoecology 13:
1-11.
Armitage, K. B. 1965. Notes on the biology of Philanthus bicinctus (Hymenoptera:
Sphecidae). - Journal of the Kansas Entomological Society 38: 463-470.
Ashman, T. L. and Stanton, M. 1991. Seasonal-Variation in Pollination Dynamics of
Sexually Dimorphic Sidalcea-Oregana Ssp Spicata (Malvaceae). - Ecology 72: 9931003.
Baude, M. et al. 2008. Inadvertent social information in foraging bumblebees: effects of
flower distribution and implications for pollination. - Animal Behaviour 76: 1863-1873.
Brown, J. S. and Kotler, B. P. 2007. Foraging and the Ecology of Fear. - In: Stephens, D.
W., Brown, J. S. and Ydenberg, R. C. (eds.), Foraging: Behavior and Ecology. The
University of Chicago Press, pp. 437-480.
Cartar, R. V. 1991. Colony energy-requirements affect response to predation risk in
foraging bumble bees. - Ethology 87: 90-96.
Chien, S. A. and Morse, D. H. 1998. The roles of prey and flower quality in the choice of
hunting sites by adult male crab spiders Misumena vatia (Araneae, Thomisidae). - Journal
of Arachnology 26: 238-243.
133
Chittka, L. 2001. Camouflage of predatory crab spiders on flowers and the colour
perception of bees (Aranida : Thomisidae/Hymenoptera : Apidae). - Entomologia
Generalis 25: 181-187.
Clark, C. W. and Dukas, R. 1994. Balancing foraging and antipredator demands - an
advantage of sociality. - American Naturalist 144: 542-548.
Cohen, D. and Shmida, A. 1993. The evolution of flower display and reward. Evolutionary Biology 27: 197-243.
Dukas, R. 2001. Effects of perceived danger on flower choice by bees. - Ecology Letters
4: 327-333.
Dukas, R. and Morse, D. H. 2003. Crab spiders affect flower visitation by bees. - Oikos
101: 157-163.
Dukas, R. and Morse, D. H. 2005. Crab spiders show mixed effects on flower-visiting
bees and no effect on plant fitness components. - Ecoscience 12: 244-247.
Dukas, R. et al. 2005. Experience levels of individuals in natural bee populations and
their ecological implications. - Canadian Journal of Zoology-Revue Canadienne De
Zoologie 83: 492-497.
Gomez, J. M. et al. 2006. Natural selection on Erysimum mediohispanicum flower shape:
Insights into the evolution of zygomorphy. - American Naturalist 168: 531-545.
Heiling, A. M. et al. 2004. Exploitation of floral signals by crab spiders (Thomisus
spectabilis, Thomisidae). - Behavioral Ecology 15: 321-326.
Heiling, A. M. and Herberstein, M. E. 2004. Floral quality signals lure pollinators and
their predators. - Annales Zoologici Fennici 41: 421-428.
134
Heiling, A. M. and Herberstein, M. E. 2004. Predator-prey coevolution: Australian native
bees avoid their spider predators. - Proceedings of the Royal Society of London Series BBiological Sciences 271: S196-S198.
Heinrich, B. 1972. Energetics of temperature regulation and foraging in a bumblebee,
Bombus terricola Kirby. - Journal of Comparative Physiology 77: 49-&.
Higginson, A. D. et al. 2006. Morphological correlates of nectar production used by
honeybees. - Ecological Entomology 31: 269-276.
Hugie, D. M. and Dill, L. M. 1994. Fish and Game - a game-theoretic approach to habitat
selection by predators and prey. - Academic Press (London) Ltd, pp. 151-169.
Ings, T. C. and Chittka, L. 2008. Speed-accuracy tradeoffs and false alarms in bee
responses to cryptic predators. - Current Biology 18: 1-5.
Louda, S. M. 1982. Inflorescence spiders - a cost/benefit analysis for the host plant,
Haplopappus venetus Blake (Asteraceae). - Oecologia 55: 185-191.
Medel, R. et al. 2003. Pollinator-mediated selection on the nectar guide phenotype in the
Andean monkey flower, Mimulus luteus. - Ecology 84: 1721-1732.
Menzel, R. 1985. Learning in honeybees in an ecological and behavioral context. - In:
Holldobler, B. and Lindauer, M. (eds.), Experimental behavioral ecology. Fischer, pp. 5574.
Møller, A. P. 1995. Bumblebee preference for symmetrical flowers. - Proceedings of the
National Academy of Sciences of the United States of America 92: 2288-2292.
Morse, D. H. 1979. Prey capture by the crab spider Misumena calycina (Araneae,
Thomisidae). - Oecologia 39: 309-319.
135
Morse, D. H. 1988. Cues associated with patch-choice decisions by foraging crab spiders
Misumena vatia. - Behaviour 107: 297-313.
Morse, D. H. 2007. Predator upon a Flower: Life history and fitness in a crab spider. Harvard University Press.
Morse, D. H. and Fritz, R. S. 1982. Experimental and observational studies of patch
choice at different scales by the crab spider Misumena vatia. - Ecology 63: 172-182.
Morse, D. H. and Stephens, E. G. 1996. The consequences of adult foraging success on
the components of lifetime fitness in a semelparous, sit and wait predator. - Evolutionary
Ecology 10: 361-373.
Orrock, J. L. et al. 2008. Consumptive and nonconsumptive effects of predators on
metacommunities of competing prey. - Ecology 89: 2426-2435.
Peckarsky, B. L. et al. 2008. Revisiting the classics: Considering nonconsumptive effects
in textbook examples of predator-prey interactions. - Ecology 89: 2416-2425.
Reader, T. et al. 2006. The effects of predation risk from crab spiders on bee foraging
behavior. - Behavioral Ecology 17: 933-939.
Robertson, I. C. and Maguire, D. K. 2005. Crab spiders deter insect visitations to
slickspot peppergrass flowers. - Oikos 109: 577-582.
Romero, G. Q. and Vasconcellos-Neto, J. 2004. Beneficial effects of flower-dwelling
predators on their host plant. - Ecology 85: 446-457.
Sargent, R. D. and Ackerly, D. D. 2008. Plant-pollinator interactions and the assembly of
plant communities. - Trends in Ecology & Evolution 23: 123-130.
136
Schmalhofer, V. R. 2001. Tritrophic interactions in a pollination system: impacts of
species composition and size of flower patches on the hunting success of a flowerdwelling spider. - Oecologia 129: 292-303.
Schmitz, O. J. et al. 2008. From individuals to ecosystem function: Toward an integration
of evolutionary and ecosystem ecology. - Ecology 89: 2436-2445.
Shykoff, J. A. and Bucheli, E. 1995. Pollinator Visitation Patterns, Floral Rewards and
the Probability of Transmission of Microbotryum violaceum, a Venereal-Disease of
Plants. - Journal of Ecology 83: 189-198.
Sih, A. 1998. Game theory and predator-prey response races. - In: Dugatkin, L. A. and
Reeve, H. K. (eds.), Game Theory and Animal Behavior. Oxford University Press, pp.
221-238.
Suttle, K. B. 2003. Pollinators as mediators of top-down effects on plants. - Ecology
Letters 6: 688-694.
Théry, M. and Casas, J. 2002. Predator and prey views of spider camouflage - Both
hunter and hunted fail to notice crab-spiders blending with coloured petals. - Nature 415:
133-133.
Wignall, A. E. et al. 2006. Flower symmetry preferences in honeybees and their crab
spider predators. - Ethology 112: 510-518.
Proportion of flowers with a predator
137
Reward level of flower
Figure 1: The relationship between the reward level of a flower and the proportion
of such flowers occupied by a predator. The distribution of predators across sites
depends on whether the probability that a predator will choose a site with a given reward
level is random, c=1 (dotted line), increases less than linearly with reward level, c=1.5
(thick dashed line), increases linearly with reward level, c=2 (solid line), or increases
more than linearly with reward level, c=2.5 (thin dashed line). Overall proportion of
flowers occupied by a predator: s = 0.1.
Lifetime resource gain by forager
138
A
B
C
D
Preferred reward level
Figure 2: The total amount of resources collected by pollinators depending on their
reward preference. The optimal foraging strategy is represented by the maximum of the
lifetime resource gain curve and depends on whether the probability that a predator will
be on a flower with a given reward level is random, c=1 (dotted line), decelerates with
reward level, c=1.5 (thick dashed line), increases linearly with reward level, c=2 (solid
line), or accelerates with reward level, c=2.5 (thin dashed line). (A) ‘Invulnerable
pollinator/Low predator density’ patch: p=0.01, s=0.01. (B) ‘Invulnerable pollinator/High
predator density’ patch: p=0.01, s=0.10. (C) ‘Vulnerable pollinator/Low predator density’
patch: p=0.20, s=0.01. (D) ‘Vulnerable pollinator/High predator density’ patch, p=0.20,
s=0.10. For all graphs, m=0.001. Maxima for lifetime resource gain and optimal reward
levels are listed in Table 1.
139
Population
Distribution
L(r)max
ropt
Range (%)
A.
Uniform
908
1
5.0
Low vulnerability
decelerating
869
1
5.3
Low density
Linear
832
1
5.9
accelerating
799
1
6.9
B.
Uniform
499
1
5.0
Low vulnerability
decelerating
399
1
7.1
High density
Linear
332
1
13.7
accelerating
286
0.857
44.7
Uniform
332
1
5.0
249
1
7.9
Linear
199
1
21.1
accelerating
180
0.540
51.6
Uniform
47
1
5.0
31
1
9.7
Linear
23
0.766
77.5
accelerating
39
0.116
11.1
C.
High vulnerability decelerating
Low density
D.
High vulnerability decelerating
High density
140
Table 1: Characteristics of the lifetime resource gain curves. For each combination of
prey vulnerability, predator density and predator distribution, the maximum lifetime
resource gain (L(r)max), the reward level that yields the maximum lifetime resource gain
(ropt), and the percent range of reward levels that provide at least 95% of the maximum
lifetime resource gain are given.
141
APPENDIX D
PREDATION RISK MAKES BEES REJECT REWARDING FLOWERS AND
REDUCE FORAGING ACTIVITY
142
Predation risk makes bees reject rewarding flowers and reduce foraging activity
Emily I. Jones* and Anna Dornhaus
Department of Ecology and Evolutionary Biology, University of Arizona, 1041 East
Lowell Street, Tucson, AZ, 85721, USA
* Author for correspondence ([email protected])
143
SUMMARY: In the absence of predators, pollinators can often maximize their foraging
success by visiting the most rewarding flowers. However, if predators use those highly
rewarding flowers to locate their prey, pollinators may benefit from changing their
foraging preferences to accept less rewarding flowers. Previous studies have shown that
some predators, such as crab spiders, indeed hunt preferentially on the most pollinatorattractive flowers. In order to determine whether predation risk can alter pollinator
foraging behaviour, we conducted laboratory experiments on the response of bumble bees
(Bombus impatiens) to simulated predator attacks when faced with a tradeoff between
predation risk and flower reward (measured here as sugar concentration). In response to a
simulated attack, bees foraging on a low-reward artificial flower were more likely to
cease foraging for at least an hour or to extend the time between flower visits. Bees
attacked on a high-reward artificial flower were more likely to increase visitation of lowreward artificial flowers. Forager body size, which is thought to affect vulnerability to
capture by predators, did not have an effect on response to an attack. Predation risk can
thus alter pollinator foraging behaviour in ways that influence the number and reward
level of flowers that are visited.
Keywords: Bombus impatiens, bumble bees, foraging, predation risk
144
1. INTRODUCTION
Foraging pollinators face danger on the flowers they visit from ambush predators such as
crab spiders (Thomisidae), predaceous bugs (Reduviidae), and praying mantids
(Mantidae). Crab spiders, the best studied of these predators, are more likely to be
encountered on the more pollinator-attractive flowers. Crab spiders will move to flowers
receiving more pollinator visits (e.g., Morse & Fritz 1982; Morse 1988), as well as to
flowers with cues that they provide more or better resources, both within (Heiling &
Herberstein 2004a) and between (Schmalhofer 2001) plant species. These cues, which
are preferred by both bees and crab spiders, include floral symmetry (Møller 1995;
Wignall et al. 2006), odour (Andersson 2003; Heiling et al. 2004), and age (Chien &
Morse 1998; Higginson et al. 2006).
As a result of the higher chance of encountering a predator on the most preferred
flowers, pollinators face a tradeoff between the amount of resources they can collect in a
single flower visit and predation risk, which ultimately affects how many flower visits
they can make in their lifetime. If colony fitness depends on long-term resource gain, it
can be optimal for foragers to accept lower flower rewards or even to specialize on lowreward flowers (Jones, submitted). When predation risk is associated with a high reward
level, will pollinators demonstrate the predicted behaviour of switching to a lower reward
level?
Bees tend to avoid crab spiders (Dukas & Morse 2003; Robertson & Maguire
2005; Reader et al. 2006; Gonçalves-Souza et al. 2008). However, crab spiders are often
well camouflaged (Chittka 2001; Théry & Casas 2002) and, indeed, bees often fail to
145
react to a crab spider on a flower (Heiling & Herberstein 2004b; Robertson & Maguire
2005; Reader et al. 2006). If vigilance is not very successful, bees may benefit from
avoiding locations that are more likely to harbour crab spiders. For example, after
experiencing a simulated attack on a flower with a camouflaged predator model, bees are
more likely to falsely reject similar-appearing flowers without a predator model (Ings &
Chittka 2008; Ings & Chittka 2009).
Previous studies of bee response to predation risk have found that foragers avoid
(artificial) flowers where either they have experienced simulated predation attempts or
encountered dead conspecifics (Dukas 2001a; Abbott 2006); however, in these studies,
safe and dangerous flowers were of equal reward level. Therefore, it is not known
whether bees will avoid locations that are more likely to have predators even at a cost to
their short-term foraging gains. Additionally, forager response to predation risk may
depend on individual traits, since individuals that are more vulnerable to capture by
predators are predicted to benefit more from preferentially visiting lower reward flowers
(Jones, submitted).
One of the determinants of vulnerability to capture by crab spiders appears to be
forager body size relative to crab spider size. However, it has been argued both that
smaller species are more vulnerable, as observed by Morse (1979), and that smaller
individuals should be less vulnerable due to higher manoeuvrability (Dukas 2001b). In
bumble bees, forager body sizes vary widely, with consequences for flower choice
(Cumber 1949; Morse 1978) and foraging success (Goulson et al. 2002; Spaethe &
146
Weidenmüller 2002). Thus, response to predation risk may also vary between foragers
within a colony as a result of differences in vulnerability.
Here we examine the response of bumble bee (Bombus impatiens) foragers to
simulated predator attacks when artificial flowers differ in reward level. We measured
whether predation risk affects foraging activity and preference for flower type, and
investigated whether there is any interaction between response to predation risk and
forager body size. Foragers were given access to two foraging arenas, representing
separate flower patches, each provisioned with one high-reward and one low-reward
artificial flower, representing flowers of different species. Given this setup, an attacked
forager could react by (1) continuing to visit the same flower type, but switching to a new
resource patch, or (2) switching to the other flower type. Switching patches, but not
flower type, suggests that bees associate the experience of danger with an individual
flower, not the flower type. Alternatively, switching flower types suggests that danger is
associated with flower type, not just an individual flower.
2. MATERIAL AND METHODS
(a) Setup and training
Two Bombus impatiens colonies (obtained from Koppert Biological Systems, MI, USA)
containing approximately 150 to 200 workers each were housed in wooden nest boxes
(39 x 23 x 8 cm). Each colony was connected to three wooden foraging arenas (each 24 x
20 x 8 cm) by transparent PVC tubes 152 cm long (figure 1). Nest boxes and foraging
arenas had transparent acrylic covers to allow viewing, and foraging arenas had trap
147
doors on the rear wall to allow for simulated predator attacks. All bees were marked with
coloured, numbered plastic tags for individual identification. Bees were given ground
pollen daily and, except for training and trials, were provided with a low concentration
sugar solution in artificial flowers placed in the centre foraging arena. The sugar solution
was 25% v/v ‘Beehappy’ in water solution (Koppert Biological Systems) in Colony 1 and
20% w/w sucrose solution in Colony 2. The artificial flowers were constructed from Petri
dishes (4cm diameter x 1cm) sealed with hot glue, with a slit drilled through the top, and
placed on top of a gray foam (Crafters Square) circle.
Training of all active foragers took place for the 24 hours before experimental
trials were carried out. During training, a low-sugar artificial flower on a yellow circle
and a high-sugar artificial flower on a blue circle were put in the centre foraging arena.
Blue is preferred over yellow by bees (Menzel 1985), so colour and sugar preferences
were matched. In Colony 1, ‘low’ and ‘high’ sugar solutions were 25% and 50% v/v
‘Beehappy’ in water, respectively; in Colony 2, ‘low’ and ‘high’ sugar solutions were
20% and 40% w/w sucrose, respectively. These sugar concentrations span a typical range
found in bumble bee visited flowers (Cruden et al. 1983), and honey bees have been
shown to discriminate against sucrose solutions that are 50% lower than the alternative
(Bachman & Waller 1977). In the ‘Beehappy’ trials of colony 1, bees had additional
olfactory cues for reward level compared to sugar solution trials of colony 2; however,
data were combined after no significant difference was found between colonies.
(b) Testing
148
During experimental trials, artificial flowers were removed from the centre (training)
arena and pairs of new, clean artificial flowers were placed in the experimental arenas
(which had not previously contained food). Each pair included one low-sugar flower on a
yellow circle and one high-sugar flower on a blue circle. Trials lasted between two and
four hours, during which every forager visit was recorded by forager identity, flower
visited, and time. Bees only received simulated attacks in their very first visit to a flower
in the experimental arena. Since we had a complete record of flower visits by all bees for
the experimental arenas, we could identify which bees were visiting a flower for the first
time in that trial. Every other forager on its first visit to an artificial flower received a
simulated attack; however, foragers were not attacked if there was another forager
present on or near the artificial flower. In a crab spider attack, the spider grabs the bee
with its raptorial forelimbs (Morse 2007). We simulated this type of attack by lightly
squeezing the bee’s abdomen with forceps for approximately 2 seconds before releasing
the bee. Forceps were immediately removed from the foraging arena after an attack. This
may have led to an underestimate of predation risk, since some bees searched for the
source of the attack and resumed feeding after failing to find the source (pers. obs.). After
all trials were completed, the thorax widths of foragers were measured with digital
callipers as a measure of body size (Goulson et al. 2002; Jandt & Dornhaus 2009).
Body size data were normally distributed and thus analyzed with parametric tests.
All other data were non-normal and analyzed with non-parametric tests. Analyses of
proportions of return visits to the original feeder type or to the original foraging arena
149
included only those bees that had made at least four or two return visits, respectively,
within the analysis time period.
3. RESULTS
For bees that first visited a high-reward artificial flower, there was no significant
difference between bees that were attacked and those that were not attacked in the
likelihood to return to forage in the hour after the first foraging trip (χ2 test: n=89,
χ2=0.67, p=0.41; figure 2a), the time to return for a second flower visit (Mann-Whitney
U-test: n=76, U=525.0, p=0.34; figure 3a), and the average time between flower visits
(Mann-Whitney U-test: n=74, U=569.0, p=0.97; figure 3b). However, for bees that first
visited a low-reward artificial flower, bees that were attacked were less likely to return to
forage in the hour after the attack (χ2 test: n=72, χ2=5.53, p=0.019; figure 2a), took
significantly longer to return for a second foraging trip (Mann-Whitney U-test: n=55,
U=210.0, p=0.0060; figure 3a), and showed a trend toward longer gaps between flower
visits (Mann-Whitney U-test: n=52, U=234.0, p=0.075; figure 3b) than those bees that
were not attacked. Forager body size had a weak negative correlation with both the time
to second flower visit (linear regression: attacked, n=39, p=0.30, R2=0.029; not attacked,
n=70, p=0.033, R2=0.065; figure 4a) and average time between flower visits (linear
regression: attacked, n=36, p=0.41., R2=0.020; not attacked, n=69, p=0.045, R2=0.059;
figure 4b), but neither of these relationships was affected by whether or not bees had
experienced an attack (comparison of regression coefficients: time to second flower visit,
t=0.29, p=0.77; average time between flower visits: t=0.27, p=0.79). Furthermore, for the
150
bees that had experienced an attack, body size did not affect whether the bees returned to
forage within an hour of the attack (unpaired t-test: n=57, t=-0.25, p=0.80).
While those foragers that were attacked on a high-reward flower did not reduce
their foraging activity, their foraging preferences did change. These bees were
significantly less likely to return to the same type of flower on the second foraging trip
than bees that were not attacked (χ2 test: n=76, χ2=14.09, p<0.0001; figure 2b). There was
a trend for foragers attacked on a high-reward flower to keep avoiding high-reward
flowers over their first hour of foraging (Mann-Whitney U-test: n=30, U=58.0, p=0.058;
figure 3c). On the other hand, foragers that were attacked on a low-reward flower did not
show a difference in preference on the second trip (χ2 test: n=56, χ2=2.70, p=0.10; figure
2b) or during the first hour (Mann-Whitney U-test: n=24, U=55.0, p=0.57; figure 3c)
compared to foragers that were not attacked. After the first hour of foraging, there was no
longer a significant difference in preference for flower type for bees that were attacked
either on a high-reward flower (Mann-Whitney U-test: n=25, U=59.5, p=0.53; figure 3d)
or on a low-reward flower (Mann-Whitney U-test: n=20, U=41.5, p=0.59; figure 3d).
Attacks did not affect foragers’ choice of foraging arena on the visit after an
attack. Overall, 27.6% of bees switched foraging arenas between the first and second
foraging trips, and bees that had experienced an attack were not more likely to switch
between arenas (χ2 test: first visit to high-reward feeder, n=78, χ2=0.24, p=0.62; first visit
to low-reward feeder, n=56, χ2=1.95, p=0.16). There was also no significant effect of an
attack on the proportion of trips that foragers made to the originally chosen foraging
151
arena during the first hour (Mann-Whitney U-test: first visit to a high-reward feeder,
n=45, U=216.0, p=0.98; first visit to a low-reward feeder, n=32, U=94.0, p=0.39).
4. DISCUSSION
In our experiment with Bombus impatiens, we found that simulated predator attacks have
an effect on the overall foraging activity and foraging preferences of bees. Bees attacked
on a low-reward artificial flower spent more time between flower visits and were less
likely to return to forage within an hour of the attack. Bees attacked on a high-reward
artificial flower were more likely to switch flower types on subsequent foraging trips,
even though that meant switching to an inferior resource. The change in preference
decayed within about one hour. Forager body size did not affect the response to predation
risk.
By avoiding dangerous flowers, foragers may be able to increase their lifespan
and therefore lifetime resource gain. Dukas (2001b) estimated that, even in areas with
relatively low crab spider densities, between 17.5% and 46.7% of forager losses each day
are due to predation by crab spiders. Therefore, avoiding crab spiders may be adaptive
despite a cost to short-term resource gain. Since a single crab spider attack is unlikely to
be successful in capturing a bumble bee (Morse 1979), bumble bees may often be able to
use unsuccessful attacks to inform their future foraging behaviour. Given the positive
relationship between flower reward level and crab spider presence that may occur in
many situations (Chien & Morse 1998; Schmalhofer 2001; Heiling et al. 2004; Heiling &
Herberstein 2004a; Wignall et al. 2006), foragers may be able to infer crab spider
152
densities, and thus optimal foraging strategies, through limited encounters with crab
spiders. Theory predicts that a forager encountering a crab spider on a high-reward flower
is likely to be able to increase its lifetime foraging gains by switching to lower-reward
flowers (Jones, submitted), the forager response we observed here. However, if a forager
encounters a crab spider on a low-reward flower, it is likely that there is a high crab
spider density and that the forager may be better off searching for another patch. While
we provided two patches in our experiment, it is possible that they were not considered
sufficiently independent by the foragers, which reduced their foraging activity rather than
switching between patches.
We found that smaller foragers generally took longer to forage, as has been
reported previously (Spaethe & Weidenmüller 2002). However, contrary to our
expectations, smaller foragers were not more timid with respect to their response to
predation risk. Previous studies compared the vulnerability of pollinator species of
different sizes (Morse 1979; Morse 1981), rather than conspecific individuals of different
sizes. It is possible that although bumble bee foragers vary greatly in size, the size
differences are not great enough to affect the vulnerability of the foragers. Alternatively,
there might be behavioural constraints for small foragers to react in the same way as large
foragers, especially since it is usually the large bees that do most of the foraging
(Goulson et al. 2002; Jandt & Dornhaus 2009). Finally, smaller individuals may
compensate for their size with higher manoeuvrability, as has been suggested by Dukas
(2001b).
153
Predation risk is recognized to decrease the fitness advantages of high net rates of
energetic gain (reviewed in Lima & Dill 1990; Lima 1998). Furthermore, maximizing
short-term foraging rate or efficiency may not be optimal for colony fitness when there is
a tradeoff between predation risk and flower reward level. Our research suggests that bee
foragers may sacrifice rate of collecting floral rewards in order to lengthen lifespan, a
behaviour which has been predicted to be increase lifetime resource gain under some
conditions (Jones, submitted). Avoidance of mortality risk at a cost to foraging rate has
also been observed in ants (Nonacs & Dill 1991; Nonacs & Calabi 1992) and this
behaviour has been found to have a net benefit for colony growth (Nonacs & Dill 1990).
On the other hand, sensitivity to predation risk may depend on colony energy
requirement, as which foraging behaviour maximizes colony fitness is predicted to
depend on colony state (Beauchamp 1992) and starving bumble bee colonies have been
found to have riskier foraging behaviour (Cartar 1991).
By changing pollinator foraging behaviour, crab spiders may impose costs on
colony foraging efficiency in addition to the more obvious mortality costs from predation
of foragers, as has been observed in ants (Nonacs & Calabi 1992). Furthermore, these
behavioural changes potentially affect plant reproduction. Changes in pollinator
behaviour such as we have shown could both decrease levels of pollination and shift
pollination toward less rewarding plant individuals and species. However, there is
conflicting evidence on whether predation risk has a large enough effect on plant fitness
to lead to selection on floral traits. A number of field studies have found negative effects
of predator presence on plant fitness (Suttle 2003; Dukas 2005), whereas others have
154
failed to find any effect (Dukas & Morse 2005), or have even found a positive effect due
to predation of herbivores (Romero & Vasconcellos-Neto 2004) or seed predators (Louda
1982). The results from our laboratory experiments, that predation risk does significantly
alter pollinator behaviour, both in terms of foraging activity and foraging preferences,
suggest that predators could affect plant reproduction, provided that predator densities are
sufficiently high. Therefore, we suggest that studies of top-down effects of crab spiders
need to consider the possibility that selection acts not just on plants occupied by crab
spiders, as has been investigated previously, but also on plants phenotypically similar to
occupied plants.
Acknowledgments
We thank Josefa Bleu and Julia Olszewski for their enthusiastic assistance with
preliminary experiments, the members of the Bronstein and Dornhaus labs for helpful
comments on the manuscript, and the Department of Ecology and Evolutionary Biology
at the University of Arizona for funding.
References
Abbott, K. R. 2006 Bumblebees avoid flowers containing evidence of past predation
events. Can. J. Zool.-Rev. Can. Zool. 84, 1240-1247. (DOI 10.1139/Z06-117.)
155
Andersson, S. 2003 Foraging responses in the butterflies Inachis io, Aglais urticae
(Nymphalidae), and Gonepteryx rhamni (Pieridae) to floral scents. Chemoecology
13, 1-11. (DOI 10.1007/s000490300000.)
Bachman, W. W. & Waller, G. D. 1977 Honeybee responses to sugar solutions of
different compositions. J. Apic. Res. 16, 165-169.
Beauchamp, G. 1992 Effects of energy-requirements and worker mortality on colony
growth and foraging in the honey-bee. Behav. Ecol. Sociobiol. 31, 123-132. (DOI
10.1007/BF00166345.)
Cartar, R. V. 1991 Colony energy-requirements affect response to predation risk in
foraging bumble bees. Ethology 87, 90-96.
Chien, S. A. & Morse, D. H. 1998 The roles of prey and flower quality in the choice of
hunting sites by adult male crab spiders Misumena vatia (Araneae, Thomisidae).
J. Arachnol. 26, 238-243.
Chittka, L. 2001 Camouflage of predatory crab spiders on flowers and the colour
perception of bees (Aranida : Thomisidae/Hymenoptera : Apidae). Entomol. Gen.
25, 181-187.
Cruden, R. W., Hermann, S. M. & Peterson, S. 1983 Patterns of nectar production and
plant-pollinator coevolution. In The Biology of Nectaries (ed. B. Bentley & T.
Elias). New York: Columbia University Press.
Cumber, R. A. 1949 The biology of humble-bees, with special reference to the
production of the worker caste. Trans. R. Entomol. Soc. Lond. 100, 1-45.
156
Dukas, R. 2001a Effects of perceived danger on flower choice by bees. Ecol. Lett. 4, 327333. (DOI 10.1046/j.1461-0248.2001.00228.x.)
Dukas, R. 2001b Effects of predation risk on pollinators and plants. In Cognitive Ecology
of Pollination (ed. L. Chittka & J. D. Thompson), pp. 214-236. Cambridge:
Cambridge University Press.
Dukas, R. 2005 Bumble bee predators reduce pollinator density and plant fitness.
Ecology 86, 1401-1406. (DOI 10.1890/04-1663.)
Dukas, R. & Morse, D. H. 2003 Crab spiders affect flower visitation by bees. Oikos 101,
157-163. (DOI 10.1034/j.1600-0706.2003.12143.x.)
Dukas, R. & Morse, D. H. 2005 Crab spiders show mixed effects on flower-visiting bees
and no effect on plant fitness components. Ecoscience 12, 244-247. (DOI
10.2980/i1195-6860-12-2-244.1.)
Gonçalves-Souza, T., Omena, P. M., Souza, J. C. & Romero, G. Q. 2008 Trait-mediated
effects on flowers: Artificial spiders deceive pollinators and decrease plant
fitness. Ecology 89, 2407-2413. (DOI 10.1890/07-1881.1.)
Goulson, D., Peat, J., Stout, J. C., Tucker, J., Darvill, B., Derwent, L. C. & Hughes, W.
O. H. 2002 Can alloethism in workers of the bumblebee, Bombus terrestris, be
explained in terms of foraging efficiency? Anim. Behav. 64, 123-130. (DOI
10.1006/anbe.2002.3041.)
Heiling, A. M., Cheng, K. & Herberstein, M. E. 2004 Exploitation of floral signals by
crab spiders (Thomisus spectabilis, Thomisidae). Behav. Ecol. 15, 321-326. (DOI
10.1093/beheco/arh012.)
157
Heiling, A. M. & Herberstein, M. E. 2004a Floral quality signals lure pollinators and
their predators. Ann. Zool. Fenn. 41, 421-428.
Heiling, A. M. & Herberstein, M. E. 2004b Predator-prey coevolution: Australian native
bees avoid their spider predators. Proc. R. Soc. Lond. Ser. B-Biol. Sci. 271, S196S198. (DOI 10.1098/rsbl.2003.0138.)
Higginson, A. D., Gilbert, F. S. & Barnard, C. J. 2006 Morphological correlates of nectar
production used by honeybees. Ecol. Entomol. 31, 269-276.
Ings, T. C. & Chittka, L. 2008 Speed-accuracy tradeoffs and false alarms in bee
responses to cryptic predators. Curr. Biol. 18, 1-5. (DOI
10.1016/j.cub.2008.07.074.)
Ings, T. C. & Chittka, L. 2009 Predator crypsis enhances behaviourally mediated indirect
effects on plants by altering bumblebee foraging preferences. Proc. R. Soc. BBiol. Sci. 276, 2031-2036. (DOI 10.1098/rspb.2008.1748.)
Jandt, J. M. & Dornhaus, A. 2009 Spatial organization and division of labour in the
bumblebee Bombus impatiens. Anim. Behav. 77, 641-651. (DOI
10.1016/j.anbehav.2008.11.019.)
Lima, S. L. 1998 Stress and decision making under the risk of predation: Recent
developments from behavioral, reproductive, and ecological perspectives. In
Stress and Behavior, vol. 27, pp. 215-290. San Diego: Academic Press Inc.
Lima, S. L. & Dill, L. M. 1990 Behavioral Decisions Made under the Risk of Predation a Review and Prospectus. Can. J. Zool.-Rev. Can. Zool. 68, 619-640. (DOI
10.1139/z90-092.)
158
Louda, S. M. 1982 Inflorescence spiders - a cost/benefit analysis for the host plant,
Haplopappus venetus Blake (Asteraceae). Oecologia 55, 185-191. (DOI
10.1007/BF00384486.)
Menzel, R. 1985 Learning in honeybees in an ecological and behavioral context. In
Experimental behavioral ecology (ed. B. Holldobler & M. Lindauer), pp. 55-74.
Stuttgart: Fischer.
Møller, A. P. 1995 Bumblebee preference for symmetrical flowers. Proc. Natl. Acad. Sci.
U. S. A. 92, 2288-2292.
Morse, D. H. 1978 Size-related foraging differences of bumble bee workers. Ecol.
Entomol. 3, 189-192. (DOI 10.1111/j.1365-2311.1978.tb00918.x.)
Morse, D. H. 1979 Prey capture by the crab spider Misumena calycina (Araneae,
Thomisidae). Oecologia 39, 309-319. (DOI 10.1007/BF00345442.)
Morse, D. H. 1981 Prey capture by the crab spider Misumena vatia (Clerck)
(Thomisidae) on three common native flowers. Am. Midl. Nat. 105, 358-367.
Morse, D. H. 1988 Cues associated with patch-choice decisions by foraging crab spiders
Misumena vatia. Behaviour 107, 297-313. (DOI 10.1163/156853988X00395.)
Morse, D. H. 2007 Predator upon a Flower: Life history and fitness in a crab spider.
Cambridge, MA: Harvard University Press.
Morse, D. H. & Fritz, R. S. 1982 Experimental and observational studies of patch choice
at different scales by the crab spider Misumena vatia. Ecology 63, 172-182.
159
Nonacs, P. & Calabi, P. 1992 Competition and predation risk - their perception alone
affects ant colony growth. Proc. R. Soc. Lond. Ser. B-Biol. Sci. 249, 95-99. (DOI
10.1098/rspb.1992.0089.)
Nonacs, P. & Dill, L. M. 1990 Mortality risk vs. food quality trade-offs in a common
currency - ant patch preferences. Ecology 71, 1886-1892.
Nonacs, P. & Dill, L. M. 1991 Mortality risk versus food quality trade-offs in ants - patch
use over time. Ecol. Entomol. 16, 73-80. (DOI 10.1111/j.13652311.1991.tb00194.x.)
Reader, T., Higginson, A. D., Barnard, C. J. & Gilbert, F. S. 2006 The effects of
predation risk from crab spiders on bee foraging behavior. Behav. Ecol. 17, 933939. (DOI 10.1093/beheco/arl027.)
Robertson, I. C. & Maguire, D. K. 2005 Crab spiders deter insect visitations to slickspot
peppergrass flowers. Oikos 109, 577-582. (DOI 10.1111/j.00301299.2005.13903.x.)
Romero, G. Q. & Vasconcellos-Neto, J. 2004 Beneficial effects of flower-dwelling
predators on their host plant. Ecology 85, 446-457. (DOI 10.1890/02-0327.)
Schmalhofer, V. R. 2001 Tritrophic interactions in a pollination system: impacts of
species composition and size of flower patches on the hunting success of a
flower-dwelling spider. Oecologia 129, 292-303. (DOI 10.1007/s004420100726.)
Spaethe, J. & Weidenmüller, A. 2002 Size variation and foraging rate in bumblebees
(Bombus terrestris). Insect. Soc. 49, 142-146. (DOI 10.1007/s00040-002-8293-z.)
160
Suttle, K. B. 2003 Pollinators as mediators of top-down effects on plants. Ecol. Lett. 6,
688-694. (DOI 10.1046/j.1461-0248.2003.00490.x.)
Théry, M. & Casas, J. 2002 Predator and prey views of spider camouflage - Both hunter
and hunted fail to notice crab-spiders blending with coloured petals. Nature 415,
133-133. (DOI 10.1038/415133a.)
Wignall, A. E., Heiling, A. M., Cheng, K. & Herberstein, M. E. 2006 Flower symmetry
preferences in honeybees and their crab spider predators. Ethology 112, 510-518.
(DOI 10.1111/j.1439-0310.2006.01199.x.)
161
Figure 1: Experimental setup. The position (left or right) of the low-reward and highreward artificial flowers was randomized in each box for each training day and during
each trial. The experimental arenas were accessible to bees at all times, but were empty
except during test periods.
Figure 2: The proportion of bees that (a) returned to forage within an hour of their first
foraging trip and (b) returned to their first flower choice on their second foraging trip. On
the first foraging trip, bees either went to a low or high-reward artificial flower and either
foraged safely (white bars) or were attacked (gray bars). (a) An attack decreased the
probability that a bee would return to forage in the hour after the first foraging trip for
bees that first visited low-reward flowers (χ2-test: n=72, χ2=5.53, p=0.019), but not for
bees that first visited high-reward flowers (χ2-test: n=89, χ2=0.67, p=0.41). (b) Both the
bees that first visited low-reward flowers and those that first visited high-reward flowers
showed a tendency to be less likely to return to the same flower type after an attack; this
effect was significant for bees that first visited a high-reward flower (χ2-test: n=76,
χ2=14.09, p=0.0001), but not for bees that first visited a low-reward flower (χ2-test: n=56,
χ2=2.70, p=0.10).
Figure 3: The effect of artificial flower reward level and simulated predator attacks on (a)
the time between first and second flower visits, (b) the average time between flower visits
over the duration of the trial, (c) the proportion of return visits to a forager’s first flower
choice during the first hour of foraging, and (d) the proportion of return visits to a
162
forager’s first flower choice after the first hour of foraging. Bars show the median and
lower and upper quartiles, with whiskers demarking the 10th and 90th percentiles and
outliers shown (solid circles). On the first foraging trip, bees either went to a low or highreward artificial flower and either foraged safely (white bars) or were attacked (gray
bars). (a) An attack increased the time taken to return to forage for bees that first visited a
low-reward flower (Mann-Whitney U-test: n=55, U=210.0, p=0.0060), but not for those
that first visited a high-reward feeder (Mann-Whitney U-test: n=76, U=525.0, p=0.34).
(b) Over the duration of the trial, an attack tended to increase the average time taken by
bees between foraging trips for bees that first visited a low-reward feeder (Mann-Whitney
U-test: n=52, U=234.0, p=0.075), but not for those that first visited a high reward feeder
(Mann-Whitney U-test: n=74, U=569.0, p=0.97). (c) Over the first hour, an attack tended
to decrease the proportion of returns to high-reward flowers (Mann-Whitney U-test:
n=30, U=58.0, p=0.058), but did not affect returns to low-reward flowers (Mann-Whitney
U-test: n=24, U=55.0, p=0.57). (d) After the first hour of foraging, no effect of an attack
remained on returns to either low (Mann-Whitney U-test: n=20, U=41.5, p=0.59) or high
(Mann-Whitney U-test: n=25, U=59.5, p=0.53) reward flowers.
Figure 4: The relationship between forager body size and (a) time until first return visit
and (b) average time between artificial flower visits. Bees that were attacked (filled
circles, solid regression line) and not-attacked (open circles, dashed regression line) had a
weak negative relationship between thorax width and both time until the first return
(linear regression: attacked, n=39, p=0.30, R2=0.029; not attacked, n=70, p=0.033,
163
R2=0.065) and overall foraging rate (linear regression: attacked, n=36, p=0.41., R2=0.020;
not attacked, n=69, p=0.045, R2=0.059). The slopes of the linear regressions were not
significantly different between attacked and not-attacked bees in either the time until first
return (t=0.29, p=0.77) or overall foraging rate (t=0.27, p=0.79), showing that the
response to an attack is not sensitive to body size.
Figure 1:
expt.
arena
nest box
training
arena
divider
artificial
flower
expt.
arena
trap door
164
Figure 2:
proportion of bees returning to forage
within the first hour
(a)
1.00
*
n. s.
0.75
0.50
0.25
0.00
low reward
high reward
reward level of first flower visited
proportion of bees returning to the
original flower type on the second visit
(b)
1.00
n. s.
*
0.75
0.50
0.25
0.00
low reward
high reward
reward level of first flower visited
165
Figure 3:
time until the first return (minutes)
(a)
100
75
50
25
0
low reward
high reward
reward level of first flower visited
average time between returns (minutes)
(b)
100
75
50
25
0
low reward
high reward
reward level of first flower visited
166
(c)
proportion of returns to
the original flower type
1.00
0.75
0.50
0.25
0.00
low reward
high reward
reward level of first flower visited
(d)
proportion of returns to
the original flower type
1.00
0.75
0.50
0.25
0.00
low reward
high reward
reward level of first flower visited
167
Figure 4:
time until first return visit (minutes)
(a)
100
75
50
25
0
3.5
4.5
5.5
6.5
thorax width (mm)
average time between visits (minutes)
(b)
100
75
50
25
0
3.5
4.5
5.5
thorax width (mm)
6.5