ES51CH18_Chomicki ARjats.

cls August 8, 2020 15:50

Annual Review of Ecology, Evolution, and

Systematics
The Evolution of Mutualistic
Dependence
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

Guillaume Chomicki,1 E. Toby Kiers,2

and Susanne S. Renner3
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

1
Department of Bioscience, Durham University, Durham DH1 3LE, United Kingdom;
email: guillaume.chomicki@gmail.com
2
Department of Ecological Science, Vrije Universiteit, 1081 HV Amsterdam, The Netherlands
3
Systematic Botany and Mycology, Department of Biology, University of Munich,
80638 Munich, Germany

Annu. Rev. Ecol. Evol. Syst. 2020. 51:409–32 Keywords

The Annual Review of Ecology, Evolution, and
mutualism, symbiosis, dependence, specialization, mutualism stability,
Systematics is online at ecolsys.annualreviews.org
cheating, exploitation, mutualism breakdown, context dependence
https://doi.org/10.1146/annurev-ecolsys-110218-
024629 Abstract
Copyright © 2020 by Annual Reviews.
While the importance of mutualisms across the tree of life is recognized, it
All rights reserved
is not understood why some organisms evolve high levels of dependence on
mutualistic partnerships, while other species remain autonomous or retain
or regain minimal dependence on partners. We identify four main pathways
leading to the evolution of mutualistic dependence. Then, we evaluate cur-
rent evidence for three predictions: (a) Mutualisms with different levels of
dependence have distinct stabilizing mechanisms against exploitation and
cheating, (b) less dependent mutualists will return to autonomy more often
than those that are highly dependent, and (c) obligate mutualisms should
be less context dependent than facultative ones. Although we find evidence
supporting all three predictions, we stress that mutualistic partners follow
diverse paths toward—and away from—dependence. We also highlight the
need to better examine asymmetry in partner dependence. Recognizing how
variation in dependence influences the stability, breakdown, and context de-
pendence of mutualisms generates new hypotheses regarding how and why
the benefits of mutualistic partnerships differ over time and space.

·.•�-
409

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

1. INTRODUCTION
Mutualistic associations are widespread in nature and linked to major evolutionary transitions
Mutualistic (Bronstein 2015, West et al. 2015). The evolution of eukaryotic cells, the colonization of land
dependence: the by plants associated with mycorrhizal fungi, the development of coral-dinoflagellate symbioses in
success with which a marine environments, and diverse mutualisms between animals and plants form the basis of global
mutualist can perform
ecosystem services. Mutualisms, mutually beneficial exchanges of services and rewards between
in the absence of its
partner(s) different species, can encompass exchanges involving transport, nutrition, and/or defense (for a
historical review, see Bronstein 2015). There is now overwhelming evidence that mutualisms are
Cheater: a species or
key modulators of global biodiversity and play important roles in diversification and coexistence
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

genotype that defects

from a cooperative processes (Bastolla et al. 2009, Weber & Agrawal 2014, Chomicki et al. 2019).
interaction Although the importance of mutualisms across the tree of life is recognized, it is not under-
stood why some organisms evolve high levels of dependency on partnerships with other species,
Context dependence:
a species interaction is while others remain autonomous or show consistently low levels of dependence. As biotic and abi-
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

context dependent otic conditions change, levels of mutualistic dependence can also change, and organisms that had
when the strength, but evolved dependence on a mutualism may revert to a less dependent state. Explaining this variation
also sometimes the is fundamental to understanding the evolutionary trajectories of species interactions and may help
sign, of the interaction
predict which mutualisms are vulnerable to breakdown.
varies according to its
biotic and/or abiotic One problem is that mutualistic dependency is poorly defined. The terms facultative and ob-
conditions ligate are commonly used to describe levels of mutualistic dependence. However, mutualism de-
pendence is frequently confounded with mutualism specialization—the tendency of a partner to
Evolved dependence:
evolutionarily derived interact with a specific mutualist. This means that often, but not always, specialization is corre-
condition where the lated with dependence. Similarly, but perhaps less often, facultative mutualisms are confounded
dependent partner has with generalization.
a reduced ability to The aim of this paper is twofold: first, we identify the main pathways by which mutualistic
perform in the absence
dependence evolves and explore how it can be classified in a way that is applicable to all mutu-
of the other partner
alisms. Second, we use this framework to ask how mutualism dependence affects key aspects of
mutualisms by focusing on predictions dealing with the following topics: (a) the stability of mutu-
alisms in the face of cheaters or exploiters who do not reciprocate, (b) the evolutionary breakdown
of mutualisms, and (c) the context dependence of mutualisms. Our review draws on a wide ar-
ray of mutualisms, encompassing work on both symbiotic and nonsymbiotic systems that vary in
taxonomic breadth and benefits exchanged.

1.1. What is Mutualistic Dependence?

Historically, mutualistic dependence has been defined as a “reduced capacity to thrive in isolation”
(Douglas & Smith 1989, p. 351). But isolation is itself not well defined, as it could mean isolation
from a specific partner or isolation from a mutualistic interaction more generally. Evolved depen-
dence, in contrast, is an evolutionarily derived condition consisting of a reduced ability to perform
in the absence of a partner (de Mazancourt et al. 2005). But again, performance is difficult to define
because it strongly depends on the abiotic and biotic context and on when and how performance
is measured.
Obligate dependency is somewhat easier to define: It applies to organisms for whom the re-
moval of a mutualist entails the loss of their reproductive ability and/or death. In host-bacterial
symbioses, this type of extreme obligate dependency can sometimes lead to physical and genomic
integration between partners over evolutionary time. Examples include Paracatenula flatworms
colonized by intracellular α-proteobacteria Candidatus riegeria symbionts that are essential for di-
gestive functioning (Gruber-Vodicka et al. 2011; Kiers & West 2015), the nested symbiont-within-

·.•�-
symbiont dependence of mealy bugs that is essential for peptidoglycan biosynthesis (Bublitz et al.

410 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

2019), and metabolic codependence in the Hydra-Chlorella symbiosis (Hamada et al. 2018). These
examples display a range of physical dependencies, with integration to organelle-like status being
the most extreme (Keeling et al. 2015). In these cases, dependency evolves to such a degree that
Symbiosis: term
the partnership can be viewed as a physical or genomic hijacking, with the precise benefits to the coined by Anton de
bacterial partner becoming hard to define (Garcia & Gerardo 2014, Kiers & West 2016, Keeling Bary in 1879 to
& McCutcheon 2017). describe the intimate
In theory, variation in dependence, from extreme to weak, should be straightforward to mea- physical living
together of distinct
sure and classify. First, a beneficial effect of the interaction for both partners should have been
species, which can be
observed, as this defines a mutualism. Second, this benefit should be quantified, ideally based on mutualistic, parasitic,
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

measuring a change in fitness when an organism’s partner is removed. This approach has proven or commensalistic
useful in classifying and studying the evolution of dependency in host-bacterial symbioses across
the tree of life (Fisher et al. 2017). In particular, by measuring dependency in terms of the fit-
ness effect of partner removal, it becomes possible to ask what characteristics across symbioses
have influenced the evolution of mutualistic dependency. Using a phylogenetic approach, Fisher
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

et al. (2017) concluded that both mutualistic function (i.e., the type of mutualistic services, such as
defense or nutrition, provided by the symbiont) and population structure (i.e., how partners are
transmitted across generations, vertically versus horizontally) are important in determining the
level of dependency that hosts exhibit for their bacterial symbionts.
Scaling up this approach to mutualisms involving larger partners (i.e., not host-microbe mu-
tualisms) presents significant challenges, the greatest being accurately measuring the fitness con-
sequences of losing a partner in nature when it is embedded in a complex network of interactions,
as is the case for many such mutualisms (Bascompte et al. 2003). While testing this empirically
is difficult, in theory the same factors that drive variation in dependency between hosts and their
microbial symbionts—mutualistic function and population structure—likely drive dependence in
other types of mutualisms as well.
Additionally, mutualism asymmetries may play an important role in the evolution of depen-
dence, but this is less well understood. Mutualistic dependence tends to be highly asymmetric, such
that one partner is more dependent on the partnership than the other. For example, in arbuscular-
mycorrhizal symbioses, the fungus is obligately dependent on the plant for carbon. The plant,
however, can access nutrients directly from the soil, without the aid of the fungus, leading to a
dependency asymmetry in the partnership. Whether both partners are obligately dependent, or
just one partner, may affect mutualism stability and context dependence.
Problems likewise arise in describing the generalist versus specialist nature of mutualistic part-
nerships. A partner may be considered a specialist because it has associated with a specific lineage
of mutualists over its evolutionary history. This may be because of constraints, such as a lack of
opportunities to gain partners. Insects can become trapped into partnerships with poorly func-
tioning but highly specialized symbionts that are difficult to replace (Keeling & McCutcheon
2017). In contrast, aquatic hosts appear to replace their symbionts with high frequencies (Boscaro
et al. 2018) or even to carry multiple symbionts simultaneously as intracellular symbionts (Ansorge
et al. 2019). While cases of symbiont replacement in insects are accumulating (Sudakaran et al.
2017, McCutcheon et al. 2019, Monnin et al. 2020), more research is needed to understand the
boundaries between generalization and specialization.

1.2. An Evolutionary Framework to Classify Mutualistic Dependence

Here we combine the concepts of partner specialization (generalist versus specialist) with de-
pendency (facultative versus obligate) across mutualisms to illustrate key diagnostic features of

·.•�-
these partnerships, and we also tabulate a series of illustrative examples. The facultative/obligate

www.annualreviews.org • The Evolution of Mutualistic Dependence 411

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

dichotomy, which has been in broad use for some time (for a review, see Bronstein 2015), canalizes
thinking about mutualisms in a way that plays down the role of asymmetry in partner dependence.
At the same time, specialization is the basis for the generalist/specialist dichotomy (discussed be-
Facultative generalist
(FG): a species or low in this section). Because specificity and dependence are often asymmetric, we apply our frame-
genotype that can work at the organismal level, classifying partner species rather than the interactions themselves.
survive or reproduce Our scheme focuses on a strict definition of obligate dependence, namely that it implies death or
in the absence of its complete loss of reproduction in the absence of the mutualistic partner. Thus, while facultative
mutualistic partners
mutualisms form a continuum in terms of dependence, obligate ones make up a strict category.
and whose number of
partners is potentially Regarding specialization, we focus on the number of species partners involved in the interac-
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

large (e.g., plants with tion, with generalist referring to a species that interacts with many potential mutualistic partners
extrafloral nectaries) and specialist referring to species whose interactions are restricted to one or a few specific partners.
Facultative specialist Deciding on a cutoff along this continuum, however, can be difficult. Network ecologists came up
(FS): a species or with a solution by generating specialization metrics (e.g., Bascompte et al. 2003). While this ap-
genotype that can proach has proven successful in addressing ecological questions, it is of limited use in evolutionary
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

survive or reproduce biology since in the former, specialization is defined within a group of interacting species in a par-
in the absence of its
ticular time and area, while we are using specialization in an evolutionary sense. We therefore
mutualistic partners
but whose number of propose a framework that focuses on species traits. This trait-based approach can help break the
partners is small (e.g., continuum into discrete categories by focusing on identifying species that can in principle associate
some defense with a large set of partners and those that cannot because of specific traits. While a limitation of this
endosymbioses in approach is that it requires extensive natural history information, it encourages cross-disciplinary
insects)
collaborations with systematists to assemble clade-level data sets of ecological interactions.
Endosymbiosis: an Starting from the least dependent and least specialized, the first category in our framework
interaction in which an comprises facultative generalists (FGs).
organism (typically
Organisms in this category can persist with or without partners, usually because the mutu-
single-celled but can
be multicellular) alistic function can also be provided (a) directly by the environment (e.g., dispersal still takes
resides in the body place in fleshy-fruited species in the absence of animal dispersers, although probably at a very low
(either intracellularly level), (b) by an intrinsic trait (e.g., self-fertilization, hence no obligate reliance on a pollinator), or
or in the guts) of (c) by an interaction of the two [e.g., pollination aided by wind and intrinsic traits in some plant
another (typically)
species (Culley et al. 2002)]. In FG mutualists, benefits are expected to vary according to con-
multicellular organism
text (see Section 5). While global surveys are not available, this category appears to encompass
a very large number of mutualistic partners. The number of evolved traits in FG mutualists can
vary dramatically (see Supplemental Table 1). Some species have obvious traits to attract and re-
ward mutualists, such as fleshy fruits that aid seed dispersal by animals (Estrada & Fleming 2012).
Others have no readily classifiable traits fostering mutualisms; this is the case for generalist ants
forming nutritional mutualisms with plants, where the benefits are a by-product of plant nesting,
a trait common to arboreal ant species (Chomicki & Renner 2019). In some of these generalist
systems, the plants have evolved hollow structures called domatia that are adapted to ant nesting
(Chomicki & Renner 2015). While they may benefit some ant species more than others, most
arboreal ants can use these structures.
The second category is comprised of mutualists with high partner specificity but facultative
mutualistic function. For such facultative specialists (FSs), the mutualistic service is performed by
specific partners, but the mutualism remains facultative and modulates fitness only under certain
biotic or abiotic conditions. For example, plants that depend on ants for defense against herbivory
show strong dependence only when herbivore pressure is high (Palmer et al. 2008, 2010). In these
cases, ecological context determines the benefits of the partnership, but some factors help ensure
partner specificity. Factors promoting specificity may include vertical transmission, as in some
insect defense-related endosymbioses (Oliver et al. 2003). The specificity can also arise from co-

·.•�-
evolutionary history or strong ecological filtering linked to the mutualism function (for example,

412 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

aggressiveness in ants engaging in ant-plant defense mutualisms). Likewise, in the legume-rhizobia

nitrogen-fixing symbiosis, strain-specific legume-rhizobial symbioses exist (Andrews & Andrews
2017), perhaps involving preinteraction signals. However, similar to more promiscuous legume-
Obligate generalist
rhizobial symbioses (e.g., generalist legumes), the symbiosis is facultative and can be shut down in (OG): a species or
high-nitrogen environments (Ferguson et al. 2019). FS mutualists appear to be much rarer than genotype that cannot
FG mutualists. survive or reproduce
The third category is also comprised of generalists, but here dependence is obligate. Obligate in the absence of its
mutualistic partner(s)
generalist (OG) mutualists require a partner to perform a specific function, but the partner species
but whose number of
on which they depend may come from guilds of organisms (i.e., species that are not closely related partners is potentially
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

but that forage for similar resources) rather than a specific partner taxon. Examples include polli- large (e.g., obligate
nation mutualisms in which plants are self-incompatible or have temporally and/or spatially sepa- frugivore bird
rated male and female function and a floral morphology that prevents abiotic pollination such that dispersers)
flowers require pollination by animal mutualists. Bat-pollinated cacti provide a good illustration Obligate specialist
(Nassar et al. 1997). Here, a mutualistic partnership is obligate for the plant’s reproduction, but (OS): a species or
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

several different bat species can all be effective pollen vectors for any one cactus species. Similarly, genotype that cannot
survive or reproduce
seed dispersers that obligately feed on a range of fruits—such as the superb fruit dove, Ptilinopus
in the absence of its
superbus (Schleucher 1999)—are in this category. Generalist symbionts inhabiting the guts of their mutualistic partner(s)
hosts, as well as the diverse communities of intracellular, sulfur-oxidizing symbionts in Bathymodi- and whose number of
olus mussels, are further examples in which a guild of organisms provides a service on which the partners is small (e.g.,
host is obligately dependent (Shapira 2016, Ponnudurai et al. 2017, Ansorge et al. 2019). Part- leafcutter ant)
ners within these guilds can provide starkly different benefits, which can generate hard-to-predict
ecological and evolutionary dynamics ( Johnson & Bronstein 2019).
Lastly, obligate specialists (OSs) are fully dependent upon a specific partner for survival and/or
reproduction. Such interactions tend to be highly specialized. Examples include classic coevolved
mutualisms, including agricultural systems involving leafcutter ants (Schultz & Brady 2008),
fungus-farming termites (Aanen et al. 2009), and plant-farming ants (Chomicki & Renner 2016),
as well as nursery pollination systems in which active pollination is linked to oviposition into floral
tissues, such as those involving figs and fig wasps ( Jandér & Herre 2010), yucca and yucca moths
(Pellmyr & Huth 1994), and Schisandraceae plants and their pollinating midges (Luo et al. 2017,
2018). Obligate dependency can be highly asymmetric, as in the case of the giant marine tube-
worm Riftia that lacks a digestive system as an adult and depends on a nutritional symbiont gained
during the larval stage (Nussbaumer et al. 2006). Here, the bacterial partner retains a free-living
stage, but the host is obligately dependent on the symbiont. An analysis of endosymbiotic mu-
tualisms suggests that OS dependency is more likely to evolve in partners that provide nutrition
compared to defense services (Fisher et al. 2017). OG mutualists appear to be rarer as a category
than OS mutualists.
In Sections 3–5, we use these four categories (FG, FS, OG, OS) to help quantify mutualism
dependence. Before doing so, however, we describe the four pathways to evolving mutualistic
dependence and ask how these mechanisms drive the initiation of mutualistic partnerships.

2. FOUR PATHWAYS TO MUTUALISTIC DEPENDENCE

How does dependence in mutualisms evolve? In this section, we identify four pathways for the
evolution of dependence. These pathways are not necessarily mutually exclusive.
First, dependence can arise from ecological constraints that limit a species’ niche or behavior
(Peay 2016, Batstone et al. 2018). Ecological specialization (different from mutualism specializa-
tion, which relates to partner specificity; see Section 1) is the process by which an organism adapts

·.•�-
to a narrow niche. Ecological specialization emerges as a result of a complex interplay between

www.annualreviews.org • The Evolution of Mutualistic Dependence 413

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

organismal traits, life history, and the covariance between genes and the environment (Poisot et al.
2011). For example, life history constraints can restrict the pool of partners available, potentially
favoring dependence. This mechanism can be seen in the evolution of some ant-plant mutualisms
Compensated trait
loss: the loss of a trait that involve ecological specialization via the evolution of the plants’ epiphytic habit followed by in-
for which an ecological creased dependence on ants as nutrient providers, as demonstrated in the evolution of South-East
interaction Asian ant gardens (Chomicki et al. 2017).
compensates (see Second, increased mutualistic dependence can evolve via specific traits favoring or enforcing
examples in Section 2)
mutualistic interactions with particular species. By favoring particular species, other species are,
by definition, excluded. Such traits can evolve either through unilateral adaptation to mutualistic
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

partners or from coevolution, reciprocal changes in interacting species driven by natural selection
on traits that are directly involved in their interaction. Examples include long-tubed flowers that
make a plant obligately dependent on either long-beaked or long-tongued pollinators, and such
mutualisms are highly asymmetric (Abrahamczyk et al. 2014, Netz & Renner 2017). The evolution
of mutualistic dependence often occurs concurrently with mutualistic specialization, and indirect
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

effects can play a major role in shaping networks and dependence among partners (Guimarães
et al. 2017).
A requirement for the evolution of specialization in mutualisms is heritable variation in partner
quality (Schemske & Horvitz 1984, Heath & Stinchcombe 2014). Variation in partner quality has
been documented across radically different mutualisms, including ant-plant symbioses (Heil et al.
2009, Palmer et al. 2010, Chomicki & Renner 2019), zooxanthellate-coral symbioses (Parkinson
et al. 2015), and legume-rhizobia symbioses (Ehinger et al. 2014, Porter et al. 2019). Mutualis-
tic specialization, whereby individuals interact with a subset of the best partners, could lead to
evolutionarily increased dependence by promoting coevolution between consistently interacting
partners. Similarly, recruiting new mutualists from nonmutualistic but preadapted lineages can
lead to increased dependence in the newly recruited partner (Chomicki et al. 2017).
Third, mutualistic dependence can evolve via trait loss. Trait loss can lock a species into an
obligate relationship with its partner and can relax selection on features and behaviors that have
become outsourced to another species (Lahti et al. 2009). A pivotal concept here is compensated
trait loss, which focuses on traits that are lost because a reliably present partner (or the envi-
ronment) provides the relevant service or trait, and the traits are presumably costly to maintain
(Visser et al. 2010, Ellers et al. 2012). Examples include the Fijian farming mutualism between
the ant Philidris nagasau and the epiphyte Squamellaria that it cultivates. Here, the ant has lost the
ability to make its own (carton) nests, and this loss drives its obligate dependence on plants that
provide suitable nesting cavities (Chomicki & Renner 2016).
Such compensated trait loss is also well known in endosymbioses, in which genome size reduc-
tion in symbionts appears to be driven by increasing dependence on the host (Bennett & Moran
2015, McCutcheon et al. 2019). Some highly dependent endosymbioses have emerged from rela-
tionships in which a microbe has evolved traits to parasitize a host’s resource-rich tissue. Interac-
tions with pathogens may drive the host to gain novel functions, such as new ways to synthesize
nutrients or mount defense responses. This becomes indirectly beneficial if the host then evolves a
way to control the new partner, which is subsequently turned from an antagonist into a mutualist,
for example, via vertical transmission (Sachs & Wilcox 2005). Some of the best evidence comes
from insect-microbe symbioses; for example, in cicadas the beneficial endosymbiotic Ophiocordy-
ceps fungus evolved from pathogenic fungi (Matsuura et al. 2018). This evolutionary trajectory can
be favored if the impact of the infection upon the host is mild. In these cases, minute shifts in the
biotic or abiotic context of the interaction could tip its balance toward a more beneficial state.
In other cases, dependence may emerge from community-level processes. The Black Queen

·.•�-
hypothesis (Morris et al. 2012) is named after the queen of spades in the game of Hearts; having

414 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

the queen drastically raises your points in a game in which the lowest score wins. This hypothesis
posits that gene loss can be favored in microbial communities because many metabolic functions
are leaky, leading to a pool of public goods upon which organisms can evolve a dependency (Morris
Public good:
et al. 2012). An assumption of this hypothesis is that gene loss provides a selective advantage as it a commodity available
can conserve limiting resources, for example, the time and resources needed for DNA replication, to all members of a
RNA transcription, and protein translation. The initial trait loss may then prime an organism to group that is
become dependent upon a mutualistic partner. However, because this trajectory relies on sourcing nonrivalrous, meaning
that the consumption
public goods, there is a higher likelihood that exploitation, via a tragedy of the commons, will
by one agent does not
emerge (West et al. 2007a). diminish the amount
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

Finally, a fourth type of mutualistic dependence is partner manipulation, in which one partner available to another
forces the other to become dependent. The main difference with this class of dependence is that agent
it is induced by a controlling partner; a species can gain a benefit from actively restricting its
partners’ ability to obtain resources (Wyatt et al. 2016, Sørensen et al. 2019). The result is that
the restricted partner cannot survive or reproduce without the help of the controlling mutualist.
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

A textbook example is the obligately specialized Central American ant-acacia mutualism, in which
the plant secretes chitinase in the extrafloral nectar (EFN) on which the Pseudomyrmex ant partner
feeds. The chitinase enzyme inhibits the sucrose-cleaving invertase of the ant, locking the ant into
feeding exclusively on the acacia’s sucrose-poor nectar (Heil et al. 2014). Similarly, some plants
secrete caffeine in floral nectar; this has been shown to enhance the pollinator’s memory of the
plant’s floral scent, leading to increased fidelity (Wright et al. 2013). In a lycaenid-ant mutualism,
wherein ants tend and protect lycaenid caterpillars in return for nutritious secretions from these
caterpillars, Narathura japonica caterpillar secretions alter the Pristomyrmex punctatus ant workers’
dopamine levels, which lowers their locomotory activity and aggressive behavior, increasing their
fidelity to and dependence on the caterpillar (Hojo et al. 2015). Thus, partner manipulation can
induce either facultative (caffeine in nectar) or obligate (chitinase in acacia EFN) dependency.
This type of dependence illustrates how one partner can control the interaction and suggests that
these interactions could shift from mutualism to parasitism, as may be the case in the lycaenid-ant
interaction (Hojo et al. 2015).
The relative importance of these four pathways to mutualistic dependence (ecological con-
straints, (co)adaptation, functional trait loss, and partner manipulation) is poorly known. In en-
dosymbioses involving microbes, functional gene loss is deemed to be especially important, but
this may not apply to nonsymbiotic mutualisms (see Future Issues). Moreover, these mechanisms
are not exclusive and might act in concert to drive the evolution of mutualistic dependence. For
instance, in the plant-farming symbiosis between Squamellaria and Philidris ants (Chomicki &
Renner 2016), the obligate dependence of the ant on the plant is driven by trait loss (nest-building
behavior), but the evolution of this OS system started with ecological specialization.
Using the evolutionary framework for classifying mutualistic dependences described in
Section 1.2, we can examine how mutualistic dependence affects: (a) the stability of mutualisms
in the face of cheaters or exploiters who do not reciprocate, (b) the evolutionary breakdown of
mutualisms, and (c) the context dependence of mutualisms. In the following sections, we set out
and explore three predictions.

3. PREDICTION ONE: MUTUALISM DEPENDENCE INFLUENCES

THE VULNERABILITY OF THE MUTUALISM TO EXPLOITATION
AND/OR CHEATING
Does the degree of mutualistic dependence influence the stability of a mutualism? Because the

·.•�-
costs and benefits of mutualisms change with dependence, patterns of exploitation and cheating

www.annualreviews.org • The Evolution of Mutualistic Dependence 415

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

are expected to differ with mutualistic dependence. In general, both the emergence and stability
of cooperative partnerships among species are difficult to explain because theory predicts that the
benefits of defecting outweigh the costs of cooperation (Trivers 1971, Axelrod & Hamilton 1981).
In this section, we start by reviewing the key concepts of cheating and exploitation, and describe
mechanisms promoting mutualism stability. We then consider the relationship between stabilizing
mechanisms and mutualistic dependence, asking whether mutualism dependency influences the
vulnerability of the partnership to exploitation and/or cheating.

3.1. Cheating and Exploitation in Mutualisms

Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

In line with Ghoul et al. (2014), we define cheating as a trait that is beneficial to one partner (the
cheater) and costly to a cooperator in terms of inclusive fitness when these benefits and costs arise
from the cooperator directing a cooperative behavior toward the cheater, rather than the intended
recipient. The first key part of this definition is that the cheater, like the exploiter, needs to receive
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

a fitness benefit from the act of cheating. This differentiates a cheater from a maladapted part-
ner (Kiers & Denison 2008). In the mutualism literature, there is also a general consensus that
cheaters are species or genotypes that are evolutionarily derived from a cooperative partner. By
contrast, exploiters (sensu Bronstein 2001a) or parasites (sensu Yu 2001) of mutualisms are not
evolutionarily derived from mutualists, but they also exploit the benefits without paying for them.
By this definition, exploiters in Pseudomyrmex ant-plant symbioses originated from nonmutualis-
tic ancestors, and hence are not cheaters; instead, they are parasites or exploiters of mutualisms
(Chomicki et al. 2015). However, nonpollinating yucca moths (Pellmyr et al. 1996), some non-
pollinating fig wasps (Machado et al. 2001, Jandér & Herre 2010), some parasitic lycaenid butter-
flies (Als et al. 2004), and various plant lineages that parasitize fungi (Bidartondo & Bruns 2001,
Merckx & Freudenstein 2010) evolved from mutualistic ancestors and hence are cheaters sensu
stricto.
Whether the emergence of cheating is frequent enough to be an important evolutionary force
in mutualisms is debated (Kiers & Denison 2008; Kiers et al. 2013; Frederickson 2013, 2017; Jones
et al. 2015). More generally, the debate over the existence and importance of cheating in mutu-
alisms revolves around two issues. First, there is a semantic one (as outlined above), with many
ecologists and evolutionary biologists using cheating to refer to exploitation regardless of its evo-
lutionary origins (Frederickson 2013). Second, it is argued that there is an absence of empirical
evidence for cheating in natural environments (Frederickson 2013, 2017). This view is supported
by sparse examples of positive correlations between host fitness and partner fitness in cases in-
volving hosts interacting with a single symbiont genotype (e.g., Friesen 2012, Frederickson 2017).
When there is a single host and a single symbiont, there is little scope for cheating because the
fitness of the partners is aligned (West et al. 2007b). However, many mutualisms are defined by
interactions with multiple partners either in space or time, and these may be more vulnerable to
partner defection (Kiers et al. 2011, 2013). This view is supported by cases of symbiont genotypes
benefiting from adopting cheating strategies (Simonsen & Stinchcombe 2014, Gano-Cohen et al.
2019).
It remains an open question how common cheating is in nature. If it is not common, is this
because it is not an important evolutionary force, as argued by Frederickson (2013, 2017)? Or
is it because stabilizing mechanisms have likewise evolved that prevent cheating from spreading,
for instance, partner choice or sanctions (e.g., Kiers et al. 2003) that diminish cheater success and
occurrence? Local factors can change the frequency of cheaters in a system, as shown in myco-
heterotrophic plant species that cheat the arbuscular mycorrhizal symbiosis, depending on soil

·.•�-
heterogeneity (Gomes et al. 2019). Work on the legume-rhizobia symbiosis suggests the existence

416 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

of an arms-race dynamic between rhizobium genotypes that provide negligible mutualist services
to hosts and hosts with mechanisms that defend against these rhizobia (Gano-Cohen et al. 2019).
Ideally, data on the fitness effects of all interacting partners, coupled with phylogenetic data,
Arms-race dynamic:
would permit tests of cheating in nature (Sachs 2015). One approach is to trace the evolutionary evolutionary
history of stabilizing mechanisms, such as sanctions, to see if they are evolutionarily derived traits mechanism that
or simply preadaptations. But again, cheating can also lead to mutualism breakdown and local typically occurs in
extinction (Sachs & Simms 2006), which is unlikely to be captured in the phylogenetic record. host-parasite
interactions involving
adaptations and
counteradaptations
3.2. Linking Dependence with Mutualism-Stabilizing Mechanisms
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

occurring as positive
To test Prediction One, that mutualistic dependence influences mutualism stability, we first need feedbacks
to understand the three mechanisms that help promote stability (by-product mutualism, partner- By-product
fidelity feedback, and partner choice and/or sanctions) and understand their relationship to mutu- mutualism:
alistic dependence. Our tabulation of 38 examples of mutualisms shows how these three different a mutualism in which
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

stabilizing mechanisms function across our four dependency types (Supplemental Table 1). the benefit to one
partner has no cost for
First, in by-product mutualisms, the benefit to one partner has no cost for the other. Hence,
the other, and thus
there is no benefit to cheating nor can exploiters use such mutualisms. The issue of costs is cen- there is no benefit to
tral to the evolutionary stability of mutualisms, as the cost of cooperation influences the fitness cheating or
of cheaters and exploiters, yet is very difficult to measure (Bronstein 2001b). Two types of by- exploitation
product mutualisms have been distinguished (Sachs et al. 2004): (a) two-way by-product mu-
tualism, wherein both A and B benefit from each other as consequences of their selfish actions
(Hamilton 1971, Queller 1985, Connor 1995), and (b) by-product reciprocity, wherein A evolves
to enhance the benefits of B, which in turn increases its by-product benefits to A (Connor 1986).
In line with Prediction One, by-product mutualisms, which generally show low mutualistic
dependence, suffer little from exploitation and conflict because there is no cost to cooperation that
can be reduced. Partners in by-product mutualisms tend to be FGs. However, FG partnerships
are often the starting point for the evolution of costly rewards and hence greater dependency
(Harcombe 2010). This idea was tested by Harcombe et al. (2018) using a synthetic consortium of
Salmonella enterica and auxotrophic Escherichia coli. The bacterial strains repeatedly evolved traits
for the costly production of resources, which ultimately fueled the formation of a cooperative
partnership. This dependence, however, required high spatial structuring of the community, a
context well known to promote cooperation in microbial communities (Drescher et al. 2014, Yanni
et al. 2019).
Based on our literature survey, five out of eight examples of FG mutualists appear to be stable
owing to some form of by-product mutualism (Supplemental Table 1). This is consistent with
the idea that by-products may be the evolutionary starting point of more complex mutualisms
(Harcombe et al. 2018). This is best illustrated by the taxonomically pervasive examples of indirect
defense of plants by ants. Here, plants are defended by ants feeding on extrafloral nectaries (e.g.,
Weber & Keeler 2013), classic examples of by-product reciprocity (sensu Connor 1986). Nev-
ertheless, exploitation can still occur, because in by-product reciprocity mutualisms, one of the
partners evolves a potentially costly reward in the context of mutualism. Examples include nectar
or pollen robbing from flowers or predation of seeds from generalist animal-dispersed plants (e.g.,
Renner 1983).
The second stabilizing mechanism is partner-fidelity feedback, whereby an increase in the fit-
ness of one partner increases the fitness of the other partner. Key requirements are that partners
interact long enough for feedback to occur (Sachs et al. 2004) and that a coupling of partner fit-
ness occurs. This can happen, for instance, via vertical transmission (Bull & Rice 1991, Bright &

·.•�-
Bulgheresi 2010) or via mechanisms increasing population viscosity and spatial structuring

www.annualreviews.org • The Evolution of Mutualistic Dependence 417

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

(Drescher et al. 2014). Partner-fidelity feedback limits opportunities to cheat since a decrease
in one partner’s fitness is linked directly to the other partner’s fitness.
This stabilizing mechanism is important for Prediction One because vertically transmitted
microbial mutualisms have been associated with the evolution of higher dependencies. This is
because vertical transmission increases relatedness between symbionts sharing a host (Leeks et al.
2019), which is important for the stability of mutualisms as it promotes cooperation between
symbionts (Foster & Wenseleers 2006).
The prediction that vertical transmission is associated with the evolution of dependence has
been tested across 106 symbioses formed in 89 host species (including insects, plants, fungi, mol-
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

lusks, arachnids, and worms); the results revealed that vertical transmission of mutualistic bacteria
was correlated with higher host dependence (Fisher et al. 2017). While this relationship may seem
intuitive, there are also many cases in which hosts evolve obligate dependencies in horizontally
transmitted bacterial symbioses, especially in marine systems (Bright & Bulgheresi 2010), where
horizontal transmission is frequent regardless of dependence (Russell 2019). Vertical transmission
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

in nonsymbiotic mutualisms is rare, with one example in an ant-mealybug mutualism wherein

founder queens carry mealybugs in their mandibles when they found new colonies (Klein et al.
1992). Vertical-like dispersal also occurs in ant-plant farming symbioses in which ant workers dis-
perse the plants (Chomicki & Renner 2016). More frequently, spatial structuring can play a similar
role by creating partner feedback loops in mutualisms not involving microbes (macromutualisms)
(Rodriguez et al. 2017).
The third mechanism that stabilizes mutualistic partnerships is partner choice (Bull & Rice
1991), which can involve sanctioning behaviors, i.e., behavior whereby resources are allocated
based on actual partner performance (Kiers & Denison 2008, Oono et al. 2009, Regus et al. 2017).
In some systems, there is no difference between partner choice and sanctions because preinter-
action cues indicate the quality of a partner. For example, volatile organic compounds (VOCs)
in Acacia trees are used by Pseudomyrmex ant queens to select high-quality hosts (Razo-Belman
et al. 2018). More generally, the mechanisms for choosing cooperative partners vary. Plants may
use molecular, chemical, or physical cues to choose partners, while animals can employ a range of
behaviors (Federle et al. 1997, Bshary 2002, Orona-Tamayo et al. 2013, Chomicki et al. 2016, Noë
& Kiers 2018). The underlying rules governing partner choice may be relative—the behavior of
potential partners is compared to the behavior of others—or absolute, in which case the decision
is based on choosing any individual above a threshold value (West et al. 2002).
The difference between partner choice and sanctions is significant for Prediction One as it
clarifies the relationship between mutualistic dependence and exploitation. If we use the defini-
tion of partner choice as a choice occurring before the interaction takes place, this mechanism
can control the specificity of the mutualism and so is important for determining the number of
mutualistic partners. Mutualisms involving partner choice, however, are likely to be more vul-
nerable to exploitation compared to those involving sanctions, as partners may have evolved the
ability to mimic the signals of cooperative partners (e.g., Letourneau 1990, Meehan et al. 2009).
More generally, partner choice and sanctions remain prominent mechanisms in both FS and OS
mutualisms, occurring in 7 of the 10 types of FS mutualists and over half of the OS mutualists
(Supplemental Table 1).
Asymmetry in partner dependence is likely to play an important role in partner choice and
sanctions, especially in host-symbiont relationships. An organism that is less dependent on its part-
ner may be able to afford to sanction less effective mutualists. For example, the obligate fungus-
farming ant Cyphomyrmex muelleri can afford to discriminate against non-native fungus strains be-
cause the native fungus is transmitted vertically from parents to offspring, limiting losses (Mueller

·.•�-
et al. 2004). Similarly, there can be asymmetry in partner populations: Hosts exposed to large

418 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

populations of symbionts, for example, legume plants in a legume-rhizobia interaction, may sanc-
tion a subset of their symbionts without risking severe fitness consequences. Rhizobia also may be
able to survive in the absence of a legume host, but occupying a root nodule can produce many mil-
Parasite of
lions more descendants. This sets up a regime that selects for rhizobia engaging in the mutualism mutualism:
regardless of host quality (Denison & Kiers 2004). a species that exploits a
In line with Prediction One, partner choice is expected to drive the evolution of specialization mutualism but is not
and dependence, as well as to limit the effect of exploitation. This is because the efficiency of evolutionarily directly
derived from a
partner choice and sanctions should decrease with a decrease in the number of alternative partners
mutualist
since this reduces outside options and choice (Noë & Hammerstein 1994, Noë 2001, Wyatt et al.
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

2014, Akçay 2015).

Supplemental Table 1 illustrates how these three stabilizing mechanisms function across
our dependency framework (defined in Section 1.2) in 38 mutualisms. FG and OG mutualists
are primarily by-product mutualisms, FS mutualisms are mostly stabilized via partner choice,
and OS mutualisms are frequently stabilized by partner choice and/or one or more of the other
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

mechanisms (Supplemental Table 1). These examples support Prediction One, that mutualism
dependency influences the vulnerability of the partnership. The data also suggest that highly
effective—and potentially costly—stabilizing mechanisms against cheating evolve more readily
in highly dependent mutualists. We next explore whether exploiters should target mutualists that
are highly dependent on specific partners.

3.3. Mutualism Exploitation Increases with Mutualistic Dependence

From an exploiter’s perspective, it makes sense to target individuals that offer the highest rewards.
Here, we use the term exploiter as above, meaning parasites of mutualisms that have nonmutualist
ancestors.
We posit that exploiters should target mutualists that are highly dependent on specific part-
ners, especially when they offer abundant rewards, as seen in a number of ant-plant mutualisms
(Heil et al. 2004, Chomicki et al. 2016). As a result, highly dependent mutualists should have more
proficient filtering mechanisms to limit increased exploitation. There is evidence that this is in-
deed the case (Federle et al. 1997, Federle & Rheindt 2005, Orona-Tamayo et al. 2013, Chomicki
et al. 2016) (Figure 1). This is probably because mutualisms characterized by costly benefits com-
bined with high dependence require stabilizing mechanisms that ensure the exclusivity of rewards
(Chomicki et al. 2016) (Figure 1). In epiphytic ant-plant systems, for example, the evolution of
higher partner dependence correlates with the evolution of more rewards for the preferred part-
ner. In this case, one mechanism preventing exploitation consists of concealed sugar rewards that
are accessible only to the desired partner (Chomicki et al. 2016). The evolution of physical struc-
tures that permit (or restrict) access to rewards is another partner-choice mechanism (Chomicki
et al. 2016).
Such mechanisms to deter exploiters but reward cooperators are common in specialized ant-
plant mutualisms where the ant partner is obligately dependent (OS) on its plant host. In cen-
tral American Pseudomyrmex ant-acacia mutualisms, protein-rich food rewards contain a protease
inhibitor that makes rewards digestible only to a specific partner (Orona-Tamayo et al. 2013)
(Figure 1). In the Crematogaster-Macaranga ant-plant mutualism, hidden rewards and slippery
stems exclude exploiters and make the rewards accessible only to the OS mutualist (Federle et al.
1997, Federle & Rheindt 2005) (Figure 1). The new food rewards in these ant-plant mutualisms
and their integrated screening and partner choice mechanism (Figure 1) are tightly correlated at
the species level. This correlation suggests that the selective pressure to evolve antiexploitation

·.•�-
mechanisms is strong when dependence and reward levels increase.

www.annualreviews.org • The Evolution of Mutualistic Dependence 419

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

a Acacia Beltian bodies containing b Hydnophytinae Concealed flower rewards

protease inhibitors Myrmecodia

Extrafloral nectary with

Clades in the sucrose-poor nectar
Myrmephytum
phylogenies
Acacia branch
Extrafloral nectary with
sucrose-rich nectar Hydnophytum
No rewards
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

Specialized mutualists No exclusive rewards Anthorrhiza

Generalist mutualists Exclusive rewards
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

Squamellaria

c Macaranga Macaranga lowii var. lowii

Macaranga brevipetiolata
Macaranga praestans
Macaranga lowii
Macaranga siamensis
Macaranga indica
Macaranga trichocarpa
Macaranga motleyana
Macaranga constricta
Macaranga griffithiana
Macaranga petanostyla
Macaranga trachyphylla
Macaranga hullettii
Macaranga bancana
Macaranga velutiniflora
Macaranga havilandii
Macaranga aetheadenia
Macaranga glandibracteolata
Macaranga indistincta
Macaranga rostrata
Macaranga triloba
Macaranga depressa
Macaranga angulata
Macaranga calcicola
Macaranga diepenhorstii
Evolution of hidden rewards Macaranga winkleri Evolution of waxy stem
Macaranga puncticulata
Exposed extrafloral nectary Macaranga sarcocarpa Non-waxy stem
Macaranga conifera
Macaranga heynei
Macaranga spathicalyx
Macaranga costulata
Extrafloral nectaries Macaranga recurvata
exposed on the Macaranga bicolor
leaves Macaranga beccariana
Macaranga hypoleuca
Macaranga winkleriella
Concealed food bodies Macaranga caladiifolia Waxy stem
Macaranga lamellata
Macaranga kingii
Macaranga gigantea
Macaranga pruinosa
Macaranga pearsonii
Food bodies
Macaranga puberula
concealed under
Macaranga hosei
scale leaves
0 PP (state = 1) 1 0 PP (state = 1) 1
Purple species: non mutualist
Length = 3.866 Gray species: mutualist Length = 3.866
(Caption appears on following page)

·.•�-
420 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

Figure 1 (Figure appears on preceding page)

Correlated evolution of higher reward levels and filtering mechanisms in three highly dependent ant-plant protection mutualisms.
(a) In Central American Acacia (Vachellia), shifts from facultative generalist to facultative specialist mutualists (Supplemental Table 1)
are correlated with an increase in extrafloral nectar production but reduced nectar attractiveness. The protein-rich Beltian bodies at the
tips of leaflets (shown in yellow) are protected by specific protease inhibitors that make them undigestible to exploiters (Orona-Tamayo
et al. 2013). Moreover, the sucrose-poor extrafloral nectar (which is unattractive to exploiters) contains chitinase, which forces the
symbiotic ant partner to feed exclusively on this nectar because they can no longer digest other nectar types (Heil et al. 2014). (b) In the
epiphytic ant plant clade Hydnophytinae, an increase in mutualistic dependence is mirrored by the evolution of exclusive sugar rewards
(Chomicki et al. 2016). In panels a and b, triangles represent clades in the phylogenies; specialized mutualists are depicted in orange,
while generalist mutualists are depicted in black. (c) In Macaranga, the evolution of high dependence on specialized symbiotic partners is
mirrored by the evolution of hidden rewards and waxy stems that are too slippery to be climbed except by the biomechanically adapted
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

specialized mutualist, or both. The colored lines of the evolutionary trees (red and blue on the left and green and yellow on the right)
correspond to the insets with the same colors, i.e., exposed versus concealed rewards and non-waxy versus waxy stems.

Yet, not all mutualists can evolve mechanisms to screen partners. A prime example of mutualis-
tic exploitation is nectar robbing, in which organisms feed upon nectar via holes bitten in flowers,
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

often bypassing pollen and stigmas (Inouye 1980). Experiments have shown that nectar robbing
is correlated with the amount of nectar produced across species (Rojas-Nossa et al. 2016). In con-
trast, predominantly selfing plants, those less dependent on pollinator mutualists, have smaller
flowers that offer less nectar (Sicard & Lenhard 2011). These tend to be robbed less frequently
compared to obligately outcrossing plants, which depend on animal pollinators. Here, the animal
pollinators are in control, and for plants, evolving filtering mechanisms against robbers that pierce
the corolla to obtain nectar can be difficult. Moreover, the net fitness effect of nectar robbers on
plants is highly variable, and in some cases they act as commensals or even mutualists (Maloof
& Inouye 2000), suggesting that the selective pressure to evolve antirobbing mechanisms may be
weak.
Prediction One suggests that mutualism dependency influences the vulnerability of the part-
nership. This prediction is supported by frequent evidence of complex stabilizing mechanisms in
highly dependent mutualisms compared to cost-free mechanisms in FG mutualists (Supplemen-
tal Table 1; Section 3.2). Concerning exploitation, we find evidence of complex filtering mutu-
alisms in OG and OS mutualists, and ant-plant mutualisms clearly show a correlated evolution of
such filtering mechanisms with specialization and dependence (Figure 1; Section 3.3).

4. PREDICTION TWO: LESS DEPENDENT MUTUALISTS WILL

RETURN TO AUTONOMY MORE OFTEN THAN THOSE
THAT ARE HIGHLY DEPENDENT
Mutualisms can break down when the costs of the interaction outweigh its benefits for one or
both partners enough to drive evolutionary change. Such breakdowns can occur via (a) reversion
to autonomy (also called mutualism abandonment), (b) extinction of one or both partners, (c) shift-
ing to parasitism, or (d) partner switching (Sachs & Simms 2006, Kiers et al. 2010, Chomicki &
Renner 2017). Mutualisms can break down when the resources can be acquired more cheaply from
the environment, as in nutritional mutualisms (Allen 1991, Sprent 2001). In this section, we first
examine Prediction Two, that dependence influences reversion to autonomy, and we then discuss
whether mutualistic dependence is expected to influence other pathways of mutualism breakdown,
notably extinction and shifting to parasitism.

4.1. Mutualistic Dependence Influences the Rate of Return to Autonomy

Prediction Two proposes that facultative and generalist mutualists will abandon the interaction

·.•�-
(i.e., revert to autonomy) more easily than will highly dependent mutualists. This prediction is

www.annualreviews.org • The Evolution of Mutualistic Dependence 421

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

based on the idea that the costs of mutualism breakdown should be highest when one or both of
the partners have lost traits essential for survival in autonomy (Ellers et al. 2012), when traits are
silenced by a manipulating partner (see Section 2), or when symbiont replacement is not possi-
ble in vertically transmitted endosymbiosis (McCutcheon et al. 2019). These three scenarios are
characteristic of high levels of mutualistic dependence (Sections 1–3). Moreover, high mutualism
dependence could increase extinction rates if one partner is locked into the interaction and has no
way to opt out, even as the costs of the mutualism increase. Highly dependent mutualists also may
have smaller range sizes, which might increase both partners’ extinction risk (Purvis et al. 2000).
There is some evidence from ant-plant symbioses (Chomicki et al. 2015) and legume-rhizobia
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

symbioses (Harrison et al. 2018) that mutualisms with lower dependencies have larger geographic
ranges than do obligate specialists, but this pattern has not been tested across the full spectrum of
mutualisms. Similarly, there is strong evidence that some obligate endosymbionts constrain their
host’s physiological limits (Zhang et al. 2019).
To test the prediction that facultative mutualists will return to autonomy more often than
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

obligate ones, we tabulated phylogenetic studies of 24 mutualisms cited in reviews by Sachs &
Simms (2006) and Frederickson (2017). We classified these mutualisms into our dependence cat-
egories (FG, OG, FS, and OS; Section 1.1) and recorded the number of mutualism reversions
to autonomy (mutualism abandonment) normalized by taxon species richness and geological age
(Figure 2; Supplemental Data Set). The analysis suggests that FG mutualisms have higher rates
of return to autonomy compared to OG mutualisms. For FS mutualisms, breakdown of mutualism
is much higher than for OS mutualisms (Figure 2). This result is consistent with Prediction Two.
Examples include mutualisms involving (a) epiphytic ant plants, where in a clade of 105 species, 12
ant mutualism abandonment events occurred in FG ant plants but none in specialized ant plants
(Chomicki & Renner 2017), and (b) arbuscular mycorrhiza, where a generally facultative depen-
dence on fungal symbionts is associated with >25 independent mutualism losses (Maherali et al.
2016, Werner et al. 2018). In many of these cases, the abandonment of plant mutualism with ants
(Chomicki & Renner 2017) or fungi (Werner et al. 2018) is mirrored by a partner switch to a
different group of organisms. Interestingly, return to autonomy can also happen in specialized
mutualisms if they are facultative (FS), as in Cecropia ant-plants expanding their geographic range
to high elevations or islands where their specific Azteca ant partners are absent (Gutiérrez-Valencia
et al. 2017). This is consistent with the idea that mutualism abandonment can happen if the initial
partners become scarce only in the absence of obligate dependence and the presence of outside
options such as a different type of partner or an environmental compensation for the mutualistic
traits.

4.2. Mutualistic Dependence Likely Influences Extinction

Obligate and specialized mutualists have been predicted to show higher susceptibility to local or
global extinction. This pattern can be explained by a number of factors, including small range
size or the fact that obligate dependence potentially increases the number of factors negatively
affecting a species (e.g., Renner 1998, Dunn et al. 2009, Colwell et al. 2012). Demonstrating higher
extinction rates in obligate mutualists empirically, however, is difficult because causes of extinction
cannot be estimated from molecular phylogenies. An interesting case is that of scleractinian corals,
for which the fossil record during the Cretaceous-Tertiary mass extinction reveals an extinction
rate of symbiotic corals relying on photosynthetic zooxanthellae nearly four times higher than that
of asymbiotic corals, which derived their nutrients by capturing plankton and would have been
more affected by abrupt environmental change at the Cretaceous-Tertiary boundary (Kiessling &

·.•�-
Baron-Szabo 2004).

422 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

a Facultative generalists b Facultative specialists c Obligate generalists d Obligate specialists

Example
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

Hydnophytum myrtifolium Cecropia angustifolia Chaitophorus truncatus Buchnera aphidicola within

Acyrthosiphon pisum

3.25 × 10–2 2.05 × 10–2 5.14 × 10–4 1.15 × 10–4

Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

Mutualism
breakdown

M ut ualis m ab an d o n men t
Figure 2
Evolutionary returns from mutualisms to nonassociated states (abandonment) in four study systems. The ancestral states and loss rates
were inferred from the molecular phylogenies listed in Supplemental Data Set. Arrow widths are proportional to the number of
breakdowns, normalized by clade richness and clade age, and the numbers on the arrows show the average standardized rate of
mutualism breakdown (number of breakdowns per clade species per million years; see Supplemental Data Set for the raw data). The
photos depict an example species for each category: (a) Hydnophytum myrtifolium, a plant species in New Guinea with a large globose
tuber filled with rainwater that harbors the tree frog Cophixalus riparius, a state that evolved from the breakdown of an ancestral
ant-plant symbiosis (Chomicki & Renner 2017). (b) Cecropia angustifolia, which has likewise lost a mutualism with ants. (c) Chaitophorus
truncatus aphids that have lost their ability to be tended by ants. (d) Buchnera aphidicola bacteria within bacteriocytes of Acyrthosiphon
pisum in an obligate and specialized symbiosis that appears to have persisted for over 200 million years. Photos are reproduced (a) with
permission from Matthew Jebb, (b) from https://commons.wikimedia.org/wiki/File:Cecropia_angustifolia(Tr%C3%A9cul).jpg
(CC BY-SA 4.0), (c) from https://commons.wikimedia.org/wiki/File:Chaitophorus_sp._probably_truncatus_(on_osier_leaf)_-
_Flickr_-_S._Rae_(1).jpg (CC BY 2.0), and (d) from https://en.wikipedia.org/wiki/Buchnera_(bacterium)#/media/File:Journal.
pbio.0050126.g001.png (CC BY 2.5).

Similarly, climate change research provides examples of obligate mutualists that may be faced
with higher extinction rates in the future, including corals and organisms in obligately dependent
pollination mutualisms (Kiers et al. 2010). Corals, in particular, face extinction due to elevated
temperatures that cause the death of their obligate zooxanthellae symbionts (Pernice & Hughes
2019). An increased extinction risk in obligate mutualists could be caused by two kinds of factors
(Chomicki et al. 2019): (a) indirect factors, if the mutualistic partner restricts the other partner’s
niche or range size, as in the case of the coral symbiosis, or (b) direct factors, if the dependence
leads to coextinction cascades, for example, if one partner disperses outside of its partner range.
The latter may happen more frequently in horizontally transmitted mutualisms.

4.3. Mutualistic Dependence and Breakdown into Parasitism

Mutualism breakdown can also occur when cheater populations or species emerge within highly
dependent mutualistic partnerships. The rare incidences of mutualism breakdown into parasitism

·.•�-
tend to occur in clades with OS mutualists, including yucca moths (Pellmyr et al. 1996) and

www.annualreviews.org • The Evolution of Mutualistic Dependence 423

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

fig wasps (Machado et al. 2001) that parasitize seeds without pollinating (nonpollinating cheater
species evolved from pollinating species, but note that there is no breakdown on the plant side),
lycaenid butterflies whose caterpillars do not reward tending ants (Als et al. 2004), and plants that
parasitize their root symbionts (Bidartondo & Bruns 2001). Evolutionary breakdown of mutual-
ism into parasitism has not been observed in mutualisms with low dependence, possibly reflecting
the lower cost of those mutualisms (Bronstein 2001a,b).
While shifts from mutualism to parasitism appear to be rare, the opposite has been thought
to be prevalent in endosymbioses (Sørensen et al. 2019 and references therein). Some phyloge-
netic analyses have shown that highly dependent host-microbial symbioses have parasitic origins
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

(Sachs et al. 2011). Other analyses suggest that obligately dependent endosymbioses started with
exploitation by the host, whereby initially the host benefits at the expense of its symbiotic partner
(Sørensen et al. 2019). Global data sets are not yet available to test the generality of these patterns,
but we argue that endosymbiosis evolution does not follow a consistent, deterministic pattern (i.e.,
it does not always proceed from parasitic, to commensal, to beneficial). Cases in which mutualists
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

arise from nonpathogenic environmental bacteria are also common (McCutcheon et al. 2019). In
the stinkbug Plautia stali, obligate bacterial mutualists have repeatedly evolved from free-living en-
vironmental bacteria, driving the evolutionary transition of bipartite obligate dependency in hosts
and some bacterial lineages (Hosokawa et al. 2016). The reverse can also occur, with autonomy
arising from initially highly dependent mutualistic partnerships; for example, some eukaryotes
have lost mitochondria (Karnkowska et al. 2016).

5. PREDICTION THREE: OBLIGATE MUTUALISMS SHOULD BE LESS

CONTEXT DEPENDENT THAN FACULTATIVE MUTUALISMS
Mutualisms, like most species interactions, tend to be highly context dependent. This means that
the strength, but also sometimes the sign, of the interaction in ecological time may vary accord-
ing to the biotic and abiotic environment (Bronstein 1994, Chamberlain et al. 2014, Hoeksema
& Bruna 2015). While the importance of context dependence is widely acknowledged, it is un-
known to what extent mutualistic dependency influences context dependence. This is an important
question because the extent of context dependence on an ecological time scale will influence the
evolution of cooperative traits on an evolutionary time scale.
Prediction Three, that outcomes of mutualisms may be less context dependent for obligate
mutualists than facultative mutualists, goes back to Bronstein’s (1994) conceptualization. The rea-
soning is that if a mutualist is obligately dependent on a partner for survival or reproduction, a
conditional outcome, i.e., high context dependency, will limit the mutualism in environments that
vary in time or space. Thus, for an obligate mutualism to persist, the partners are expected to
evolve mechanisms to reduce context dependence.
The meta-analysis by Chamberlain et al. (2014) identified widespread context dependence in
mutualisms. The majority of mutualisms included in their analysis were FG mutualists. The eight
obligate mutualists (both OG and OS) included in this survey all showed less context dependence
than the facultative mutualists (both FG and FS). While this is consistent with Bronstein’s (1994)
prediction, more data are needed to formally test her hypothesis. This is especially true given that
there is context dependence even in some obligate mutualisms (Couret et al. 2019).
Pairing with an OG mutualist could help reduce context dependence (on abiotic and biotic
factors) for both partners, for instance by excluding exploiters (Section 3.3). Work on fungus-
farming Ambrosia beetles suggests that farmer beetles rely on ethanol within trees to control the
growth of exploiting fungal competitors and to facilitate the colonization of tree-host cavities

·.•�-
(Ranger et al. 2018). An evolutionary study of bacterial crypts in fungus-farming ants similarly

424 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

revealed more antibiotic-producing bacteria, and hence more efficient suppression of exploiter
fungi, in the most dependent ant species (Li et al. 2018). These mechanisms reduce biotic context
dependence by diminishing the potential impact of third parties.
For abiotic factors, such as light intensity, soil nutrient levels, or air temperature, all of which
vary spatially, an obligate mutualist partner may control context dependence by actively choosing
favorable environments. Two examples are found in ant agriculture mutualisms. Leafcutter ants
at the northern range of their fungal symbiont (in North Texas) consistently move their fungal
symbionts deeper into the ground in the winter and then to shallower depths in summer to fa-
cilitate growth (Mueller et al. 2011). This behavior reduces fluctuations in the abiotic context.
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

Likewise, in the obligate Fijian symbiosis between the plant-farming ant Philidris nagasau and
Squamellaria epiphytes, the ant farmers control their crop’s distribution by selectively planting its
seeds in sun-exposed sites, thereby optimizing crop yield and buffering environmental variation
(Chomicki et al. 2020a). An untested assumption in endosymbiosis research is that microbes that
grow within host cells, for example, in bacteriocytes, benefit from the highly controlled environ-
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

ments produced by the host because they are better protected from the context dependency of
fluctuating abiotic and biotic environments (Chomicki et al. 2020b).

6. CONCLUSION: THE EVOLUTION OF DEPENDENCE CHANGES

THE RULES FOR MUTUALISMS
Mutualisms can evolve, and move toward or away from dependence, depending on a variety of
factors. Our article highlights four pathways to mutualism dependence and their consequences for
mutualism ecology, evolution, and stability. We found support for Prediction One, which proposed
that FG, OG, FS, and OS mutualisms will differ in stability in the face of cheating and exploitation.
Specifically, both (a) the danger of being cheated or exploited and (b) the incentive for cheating
or exploitation vary with partner dependence, with the incentive for exploitation being highest in
highly dependent OS mutualists with high-value rewards.
Prediction Two, that less dependent mutualists will return to autonomy more often than highly
dependent ones, is supported by several well-studied cases, notably in ant-plant mutualisms and
plant-mycorrhizal symbioses. For Prediction Three, that more dependent mutualisms are less
context dependent, we found that obligate mutualists tend to be better protected from biotic and
abiotic variation than facultative mutualists, but more data on exactly how mutualists mitigate
context dependence are needed.
More generally, there is a need to better examine asymmetry in partner dependence. Research
is needed into whether partnerships function differently when both partners are dependent com-
pared to when only one partner is dependent. As mutualism data sets become more populated
with diverse examples and the resolution of phylogenies of mutualistic clades improves, we can
make advances in understanding the basis and consequences of evolving mutualism dependence
(see Future Issues).

FUTURE ISSUES
1. How can we accurately measure and compare variation in dependencies across mutu-
alisms? Measuring and standardizing fitness parameters across species with different mu-
tualistic strategies is difficult. Partner-removal experiments in nature can be problematic
because of complex webs of interactions and because it can lead to the immediate death

·.•�-
www.annualreviews.org • The Evolution of Mutualistic Dependence 425

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

of the partner in the most dependent mutualisms. Is it possible to develop other metrics
for dependency that can be standardized across all mutualisms?
2. Is gene loss a major driver of mutualistic dependence in macromutualisms? There is
accumulating evidence that gene loss is important in driving dependence in microbial
mutualism, especially endosymbioses. But is there a similar trend in macromutualisms?
How widespread is gene loss compared to trait gain in driving mutualistic dependence?
Comparative genomic approaches offer powerful tools to investigate these questions.
3. Can mutualistic dependence inform conservation? Mutualistic dependence can influence
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

extinction risk. Work integrating mutualistic networks with biogeography showed that if
one wanted to preserve species interactions, rather than merely species, up to five times as
much area would be needed (Gilarranz et al. 2015). Integrating mutualistic dependence
into such a framework could improve predictions about the minimal area needed to
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

conserve species and their obligate partners.

DISCLOSURE STATEMENT
The authors are not aware of any affiliations, memberships, funding, or financial holdings that
might be perceived as affecting the objectivity of this review.

ACKNOWLEDGMENTS
G.C. is supported by a National Environment Research Council Independent Research Fellow-
ship (NE/S014470/1) and formerly by a Glasstone Research Fellowship and a Junior Research
Fellowship at the Queen’s College, University of Oxford, United Kingdom. E.T.K. is supported
by the European Research Council (grant ERC 335542) and an Ammodo award. We thank
Judith Bronstein and an anonymous reviewer for their critical reviews that helped us improve
our manuscript.

LITERATURE CITED
Aanen DK, Henrik H, Debets AJ, Kerstes NA, Hoekstra RF, Boomsma JJ. 2009. High symbiont relatedness
stabilizes mutualistic cooperation in fungus-growing termites. Science 326:1103–6
Abrahamczyk S, Souto-Vilarós D, Renner SS. 2014. Escape from extreme specialization: passionflowers, bats
and the sword-billed hummingbird. Proc. R. Soc. B 281:20140888
Akçay E. 2015. Evolutionary models of mutualism. See Bronstein 2015, pp. 57–76
Allen MF. 1991. The ecology of mycorrhizae. Cambridge, UK: Cambridge Univ. Press
Als TD, Vila R, Kandul NP, Nash DR, Yen SH, et al. 2004. The evolution of alternative parasitic life histories
in large blue butterflies. Nature 432:386–90
Andrews M, Andrews ME. 2017. Specificity in legume-rhizobia symbioses. Int. J. Mol. Sci. 18:705
Ansorge R, Romano S, Sayavedra L, Porras MÁG, Kupczok A, et al. 2019. Functional diversity enables multiple
symbiont strains to coexist in deep-sea mussels. Nat. Microbiol. 4:2487–97
Axelrod R, Hamilton WD. 1981. The evolution of cooperation. Science 211:1390–96
Bascompte J, Jordano P, Melián CJ, Olesen JM. 2003. The nested assembly of plant–animal mutualistic net-
works. PNAS 100:9383–87
Bastolla U, Fortuna MA, Pascual-García A, Ferrera A, Luque B, Bascompte J. 2009. The architecture of mu-
tualistic networks minimizes competition and increases biodiversity. Nature 458:1018–20
Batstone RT, Carscadden KA, Afkhami ME, Frederickson ME. 2018. Using niche breadth theory to explain

·.•�-
generalization in mutualisms. Ecology 99:1039–50

426 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

Bennett GM, Moran NA. 2015. Heritable symbiosis: the advantages and perils of an evolutionary rabbit hole.
PNAS 112:10169–76
Bidartondo MI, Bruns TD. 2001. Extreme specificity in epiparasitic Monotropoideae (Ericaceae): widespread
phylogenetic and geographical structure. Mol. Ecol. 10:2285–95
Boscaro V, Fokin SI, Petroni G, Verni F, Keeling PJ, Vannini C. 2018. Symbiont replacement between bacteria
of different classes reveals additional layers of complexity in the evolution of symbiosis in the ciliate
Euplotes. Protist 169:43–52
Bright M, Bulgheresi S. 2010. A complex journey: transmission of microbial symbionts. Nat. Rev. Microbiol.
8:218–30
Bronstein JL. 1994. Conditional outcomes in mutualistic interactions. Trends Ecol. Evol. 9:214–17
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

Bronstein JL. 2001a. The exploitation of mutualisms. Ecol. Lett. 4:277–87

Bronstein JL. 2001b. The costs of mutualism. Am. Zool. 41:825–39
Bronstein JL, ed. 2015. Mutualism. Oxford, UK: Oxford Univ. Press
Bshary R. 2002. Biting cleaner fish use altruism to deceive image-scoring client reef fish. Proc. R. Soc. B
269:2087–93
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

Bublitz DC, Chadwick GL, Magyar JS, Sandoz KM, Brooks DM, et al. 2019. Peptidoglycan production by an
insect-bacterial mosaic. Cell 179:703–12
Bull JJ, Rice WR. 1991. Distinguishing mechanisms for the evolution of co-operation. J. Theor. Biol. 149:63–74
Chamberlain SA, Bronstein JL, Rudgers JA. 2014. How context dependent are species interactions? Ecol. Lett.
17:881–90
Chomicki G, Janda M, Renner SS. 2017. The assembly of ant-farmed gardens: mutualism specialization fol-
lowing host broadening. Proc. R. Soc. B 284:20161759
Chomicki G, Kadereit G, Renner SS, Kiers ET. 2020a. Tradeoffs in the evolution of plant farming by ants.
PNAS 117:2535–43
Chomicki G, Renner SS. 2015. Phylogenetics and molecular clocks reveal the repeated evolution of ant-plants
after the late Miocene in Africa and the early Miocene in Australasia and the Neotropics. New Phytol.
207:411–24
Chomicki G, Renner SS. 2016. Obligate plant farming by a specialized ant. Nat. Plants 2:16181
Chomicki G, Renner SS. 2017. Partner abundance controls mutualism stability and the pace of morphological
change over geologic time. PNAS 114:3951–56
Chomicki G, Renner SS. 2019. Farming by ants remodels nutrient uptake in epiphytes. New Phytol. 223:2011–
23
Chomicki G, Staedler YM, Schönenberger J, Renner SS. 2016. Partner choice through concealed floral sugar
rewards evolved with the specialization of ant–plant mutualisms. New Phytol. 211:1358–70
Chomicki G, Ward PS, Renner SS. 2015. Macroevolutionary assembly of ant/plant symbioses: Pseudomyrmex
ants and their ant-housing plants in the Neotropics. Proc. R. Soc. B 282:20152200
Chomicki G, Weber M, Antonelli A, Bascompte J, Kiers ET. 2019. The impact of mutualisms on species
richness. Trends Ecol. Evol. 34:698–711
Chomicki G, Werner G, West SA, Kiers ET. 2020b. Compartmentalization drives the evolution of symbiotic
cooperation. Philos. Trans. R. Soc. B. In press
Colwell RK, Dunn RR, Harris NC. 2012. Coextinction and persistence of dependent species in a changing
world. Annu. Rev. Ecol. Evol. Syst. 43:183–203
Connor RC. 1986. Pseudo-reciprocity: investing in mutualism. Anim. Behav. 34:1562–66
Connor RC. 1995. The benefits of mutualism: a conceptual framework. Biol. Rev. 70:427–57
Couret J, Huynh-Griffin L, Antolic-Soban I, Acevedo-Gonzalez TS, Gerardo NM. 2019. Even obligate sym-
bioses show signs of ecological contingency: Impacts of symbiosis for an invasive stinkbug are mediated
by host plant context. Ecol. Evol. 9:9087–99
Culley TM, Weller SG, Sakai AK. 2002. The evolution of wind pollination in angiosperms. Trends Ecol. Evol.
17:361–69
De Mazancourt C, Loreau M, Dieckmann ULF. 2005. Understanding mutualism when there is adaptation to
the partner. J. Ecol. 93:305–14
Denison RF, Kiers ET. 2004. Lifestyle alternatives for rhizobia: mutualism, parasitism, and forgoing symbiosis.

·.•�-
FEMS Microbiol. Lett. 237:187–93

www.annualreviews.org • The Evolution of Mutualistic Dependence 427

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

Douglas AE, Smith DC. 1989. Are endosymbioses mutualistic? Trends Ecol. Evol. 4:350–52
Drescher K, Nadell CD, Stone HA, Wingreen NS, Bassler BL. 2014. Solutions to the public goods dilemma
in bacterial biofilms. Curr. Biol. 24:50–55
Dunn RR, Harris NC, Colwell RK, Koh LP, Sodhi NS. 2009. The sixth mass coextinction: Are most endan-
gered species parasites and mutualists? Proc. R. Soc. B 276:3037–45
Ehinger M, Mohr TJ, Starcevich JB, Sachs JL, Porter SS, Simms EL. 2014. Specialization-generalization
trade-off in a Bradyrhizobium symbiosis with wild legume hosts. BMC Ecol. 14:8
Ellers J, Kiers ET, Currie CR, McDonald BR, Visser B. 2012. Ecological interactions drive evolutionary loss
of traits. Ecol. Lett. 15:1071–82
Estrada A, Fleming TH, eds. 2012. Frugivores and seed dispersal. Dordrecht, Neth.: Springer Neth.
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

Federle W, Maschwitz U, Fiala B, Riederer M, Hölldobler B. 1997. Slippery ant-plants and skillful climbers: se-
lection and protection of specific ant partners by epicuticular wax blooms in Macaranga (Euphorbiaceae).
Oecologia 112:217–24
Federle W, Rheindt FE. 2005. Macaranga ant-plants hide food from intruders: correlation of food presentation
and presence of wax barriers analysed using phylogenetically independent contrasts. Biol. J. Linn. Soc.
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

84:177–93
Ferguson BJ, Mens C, Hastwell AH, Zhang M, Su H, et al. 2019. Legume nodulation: The host controls the
party. Plant Cell Environ. 42:41–51
Fisher RM, Henry LM, Cornwallis CK, Kiers ET, West SA. 2017. The evolution of host-symbiont depen-
dence. Nat. Commun. 8:15973
Foster KR, Wenseleers T. 2006. A general model for the evolution of mutualisms. J. Evol. Biol. 19:1283–93
Frederickson ME. 2013. Rethinking mutualism stability: cheaters and the evolution of sanctions. Q. Rev. Biol.
88:269–95
Frederickson ME. 2017. Mutualisms are not on the verge of breakdown. Trends Ecol. Evol. 32:727–34
Friesen ML. 2012. Widespread fitness alignment in the legume–rhizobium symbiosis. New Phytol. 194:1096–
111
Gano-Cohen KA, Wendlandt CE, Stokes PJ, Blanton MA, Quides KW, et al. 2019. Interspecific conflict and
the evolution of ineffective rhizobia. Ecol. Lett. 22:914–24
Garcia JR, Gerardo NM. 2014. The symbiont side of symbiosis: Do microbes really benefit? Front. Microbiol.
5:510
Ghoul M, Griffin AS, West SA. 2014. Toward an evolutionary definition of cheating. Evolution 68:318–31
Gilarranz LJ, Sabatino M, Aizen MA, Bascompte J. 2015. Hot spots of mutualistic networks. J. Anim. Ecol.
84:407–13
Gomes SI, van Bodegom PM, Merckx VS, Soudzilovskaia N. 2019. Environmental drivers for cheaters of
arbuscular mycorrhizal symbiosis in tropical rainforests. New Phytol. 223:1575–83
Gruber-Vodicka HR, Dirks U, Leisch N, Baranyi C, Stoecker K, et al. 2011. Paracatenula, an ancient symbiosis
between thiotrophic Alphaproteobacteria and catenulid flatworms. PNAS 108:12078–83
Guimarães PR Jr., Pires MM, Jordano P, Bascompte J, Thompson JN. 2017. Indirect effects drive coevolution
in mutualistic networks. Nature 550:511–14
Gutiérrez-Valencia J, Chomicki G, Renner SS. 2017. Recurrent breakdowns of mutualisms with ants in the
neotropical ant-plant genus Cecropia (Urticaceae). Mol. Phylogenet. Evol. 11:196–205
Hamada M, Schröder K, Bathia J, Kürn U, Fraune S, et al. 2018. Metabolic co-dependence drives the evolu-
tionarily ancient Hydra–Chlorella symbiosis. eLife 7:e35122
Hamilton WD. 1971. Geometry of the selfish herd. J. Theor. Biol. 31:295–311
Harcombe W. 2010. Novel cooperation experimentally evolved between species. Evolution 64:2166–72
Harcombe WR, Chacón JM, Adamowicz EM, Chubiz LM, Marx CJ. 2018. Evolution of bidirectional costly
mutualism from byproduct consumption. PNAS 115:12000–4
Harrison TL, Simonsen AK, Stinchcombe JR, Frederickson ME. 2018. More partners, more ranges: generalist
legumes spread more easily around the globe. Biol. Lett. 14:20180616
Heath KD, Stinchcombe JR. 2014. Explaining mutualism variation: a new evolutionary paradox? Evolution
68:309–17
Heil M, Barajas-Barron A, Orona-Tamayo D, Wielsch N, Svatos A. 2014. Partner manipulation stabilises a

·.•�-
horizontally transmitted mutualism. Ecol. Lett. 17:185–92

428 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

Heil M, González-Teuber M, Clement LW, Kautz S, Verhaagh M, Bueno JCS. 2009. Divergent investment
strategies of Acacia myrmecophytes and the coexistence of mutualists and exploiters. PNAS 106:18091–
96
Heil M, Greiner S, Meimberg H, Krüger R, Noyer JL, et al. 2004. Evolutionary change from induced to
constitutive expression of an indirect plant resistance. Nature 430:205–8
Hoeksema JD, Bruna EM. 2015. Context-dependent outcomes of mutualistic interactions. See Bronstein 2015,
pp. 181–202
Hojo MK, Pierce NE, Tsuji K. 2015. Lycaenid caterpillar secretions manipulate attendant ant behavior. Curr.
Biol. 25:2260–64
Hosokawa T, Ishii Y, Nikoh N, Fujie M, Satoh N, Fukatsu T. 2016. Obligate bacterial mutualists evolving
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

from environmental bacteria in natural insect populations. Nat. Microbiol. 1:15011

Inouye DW. 1980. The terminology of floral larceny. Ecology 61:1251–53
Jandér KC, Herre EA. 2010. Host sanctions and pollinator cheating in the fig tree–fig wasp mutualism. Proc.
R. Soc. B 277:1481–88
Johnson CA, Bronstein JL. 2019. Coexistence and competitive exclusion in mutualism. Ecology 100:e02708
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

Jones EI, Afkhami ME, Akçay E, Bronstein JL, Bshary R, et al. 2015. Cheaters must prosper: reconciling
theoretical and empirical perspectives on cheating in mutualism. Ecol. Lett. 18:1270–84
Karnkowska A, Vacek V, Zubáčová Z, Treitli SC, Petrželková R, et al. 2016. A eukaryote without a mitochon-
drial organelle. Curr. Biol. 26:1274–84
Keeling PJ, McCutcheon JP. 2017. Endosymbiosis: The feeling is not mutual. J. Theor. Biol. 434:75–79
Keeling PJ, McCutcheon JP, Doolittle WF. 2015. Symbiosis becoming permanent: survival of the luckiest.
PNAS 112:10101–3
Kiers ET, Denison RF. 2008. Sanctions, cooperation, and the stability of plant-rhizosphere mutualisms. Annu.
Rev. Ecol. Evol. Syst. 39:215–36
Kiers ET, Denison RF, Kawakita A, Herre EA. 2011. The biological reality of host sanctions and partner
fidelity. PNAS 108:E7
Kiers ET, Palmer TM, Ives AR, Bruno JF, Bronstein JL. 2010. Mutualisms in a changing world: an evolu-
tionary perspective. Ecol. Lett. 13:1459–74
Kiers ET, Ratcliff WC, Denison RF. 2013. Single-strain inoculation may create spurious correlations between
legume fitness and rhizobial fitness. New Phytol. 198:4–6
Kiers ET, Rousseau RA, West SA, Denison RF. 2003. Host sanctions and the legume–rhizobium mutualism.
Nature 425:78–81
Kiers ET, West SA. 2015. Evolving new organisms via symbiosis. Science 348:392–94
Kiers ET, West SA. 2016. Evolution: Welcome to symbiont prison. Curr. Biol. 26:R66–68
Kiessling W, Baron-Szabo RC. 2004. Extinction and recovery patterns of scleractinian corals at the
Cretaceous-Tertiary boundary. Palaeogeogr. Palaeoclimatol. Palaeoecol. 214:195–223
Klein RW, Kovac D, Schellerich A, Maschwitz U. 1992. Mealybug-carrying by swarming queens of a Southeast
Asian bamboo-inhabiting ant. Naturwissenschaften 79:422–23
Lahti DC, Johnson NA, Ajie BC, Otto SP, Hendry AP, et al. 2009. Relaxed selection in the wild. Trends Ecol.
Evol. 24:487–96
Leeks A, dos Santos M, West SA. 2019. Transmission, relatedness, and the evolution of cooperative symbionts.
J. Evol. Biol. 32:1036–45
Letourneau DK. 1990. Code of ant-plant mutualism broken by parasite. Science 248:215–17
Li H, Sosa-Calvo J, Horn HA, Pupo MT, Clardy J, et al. 2018. Convergent evolution of complex structures
for ant–bacterial defensive symbiosis in fungus-farming ants. PNAS 115:10720–25
Luo S-X, Liu T-T, Cui F, Yang Z-Y, Hu X-Y, Renner SS. 2017. Coevolution with pollinating resin midges
led to resin-filled nurseries in the androecia, gynoecia, and tepals of Kadsura (Schisandraceae). Ann. Bot.
120:653–64
Luo S-X, Zhang L-J, Yuan S, Ma Z-H, Zhang D-X, Renner SS. 2018. The largest early-diverging angiosperm
family is mostly pollinated by ovipositing insects and so are most surviving lineages of early angiosperms.
Proc. R. Soc. B 285:20172365
Machado CA, Jousselin E, Kjellberg F, Compton SG, Herre EA. 2001. Phylogenetic relationships, historical

·.•�-
biogeography and character evolution of fig-pollinating wasps. Proc. R. Soc. B 268(1468):685–94

www.annualreviews.org • The Evolution of Mutualistic Dependence 429

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

Maherali H, Oberle B, Stevens PF, Cornwell WK, McGlinn DJ. 2016. Mutualism persistence and abandon-
ment during the evolution of the mycorrhizal symbiosis. Am. Nat. 188:E113–25
Maloof JE, Inouye DW. 2000. Are nectar robbers cheaters or mutualists? Ecology 81(10):2651–61
Matsuura Y, Moriyama M, Łukasik P, Vanderpool D, Tanahashi M, et al. 2018. Recurrent symbiont recruit-
ment from fungal parasites in cicadas. PNAS 115:E5970–79
McCutcheon JP, Boyd BM, Dale C. 2019. The life of an insect endosymbiont from the cradle to the grave.
Curr. Biol. 29:R485–95
Meehan CJ, Olson EJ, Reudink MW, Kyser TK, Curry RL. 2009. Herbivory in a spider through exploitation
of an ant–plant mutualism. Curr. Biol. 19:R892–93
Merckx V, Freudenstein JV. 2010. Evolution of mycoheterotrophy in plants: a phylogenetic perspective. New
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

Phytol. 185:605–9
Monnin D, Jackson R, Kiers ET, Bunker M, Ellers J, Henry LM. 2020. Parallel evolution in the integration
of a co-obligate aphid symbiosis. Curr. Biol. 30:1949–57
Morris JJ, Lenski RE, Zinser ER. 2012. The Black Queen Hypothesis: evolution of dependencies through
adaptive gene loss. mBio 3:e00036-12
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

Mueller UG, Mikheyev AS, Hong E, Sen R, Warren DL, et al. 2011. Evolution of cold-tolerant fungal sym-
bionts permits winter fungiculture by leafcutter ants at the northern frontier of a tropical ant–fungus
symbiosis. PNAS 108:4053–56
Mueller UG, Poulin J, Adams RM. 2004. Symbiont choice in a fungus-growing ant (Attini, Formicidae). Behav.
Ecol. 15:357–64
Nassar JM, Ramírez N, Linares O. 1997. Comparative pollination biology of Venezuelan columnar cacti and
the role of nectar-feeding bats in their sexual reproduction. Am. J. Bot. 84:918–27
Netz C, Renner SS. 2017. Long-spurred Angraecum orchids and long-tongued sphingid moths on Madagascar:
a time frame for Darwin’s predicted Xanthopan/Angraecum coevolution. Biol. J. Linn. Soc. 122:469–78
Noë R. 2001. Biological markets: partner choice as the driving force behind the evolution of mutualisms. In
Economics in Nature: Social Dilemmas, Mate Choice and Biological Markets, ed. R Noë, JARAM Van Hooff,
P Hammerstein, pp. 93–118. Cambridge, UK: Cambridge Univ. Press
Noë R, Hammerstein P. 1994. Biological markets: Supply and demand determine the effect of partner choice
in cooperation, mutualism and mating. Behav. Ecol. Sociobiol. 35:1–11
Noë R, Kiers ET. 2018. Mycorrhizal markets, firms, and co-ops. Trends Ecol. Evol. 33:777–89
Nussbaumer AD, Fisher CR, Bright M. 2006. Horizontal endosymbiont transmission in hydrothermal vent
tubeworms. Nature 441:345–48
Oliver KM, Russell JA, Moran NA, Hunter MS. 2003. Facultative bacterial symbionts in aphids confer resis-
tance to parasitic wasps. PNAS 100:1803–7
Oono R, Denison RF, Kiers ET. 2009. Controlling the reproductive fate of rhizobia: How universal are legume
sanctions? New Phytol. 183:967–79
Orona-Tamayo D, Wielsch N, Blanco-Labra A, Svatos A, Farías-Rodríguez R, Heil M. 2013. Exclusive rewards
in mutualisms: Ant proteases and plant protease inhibitors create a lock–key system to protect Acacia food
bodies from exploitation. Mol. Ecol. 22:4087–100
Palmer TM, Doak DF, Stanton ML, Bronstein JL, Kiers ET, et al. 2010. Synergy of multiple partners, in-
cluding freeloaders, increases host fitness in a multispecies mutualism. PNAS 107:17234–39
Palmer TM, Stanton ML, Young TP, Goheen JR, Pringle RM, Karban R. 2008. Breakdown of an ant-plant
mutualism follows the loss of large herbivores from an African savanna. Science 319:192–95
Parkinson JE, Banaszak AT, Altman NS, LaJeunesse TC, Baums IB. 2015. Intraspecific diversity among part-
ners drives functional variation in coral symbioses. Sci. Rep. 5:15667
Peay KG. 2016. The mutualistic niche: mycorrhizal symbiosis and community dynamics. Annu. Rev. Ecol. Evol.
Syst. 47:143–64
Pellmyr O, Huth CJ. 1994. Evolutionary stability of mutualism between yuccas and yucca moths. Nature
372:257–60
Pellmyr O, Leebens-Mack J, Huth CJ. 1996. Non-mutualistic yucca moths and their evolutionary conse-
quences. Nature 380:155–56

·.•�-
Pernice M, Hughes DJ. 2019. Forecasting global coral bleaching. Nat. Clim. Change 9:803–4

430 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

Poisot T, Bever JD, Nemri A, Thrall PH, Hochberg ME. 2011. A conceptual framework for the evolution of
ecological specialisation. Ecol. Lett. 14:841–51
Ponnudurai R, Kleiner M, Sayavedra L, Petersen JM, Moche M, et al. 2017. Metabolic and physiological
interdependencies in the Bathymodiolus azoricus symbiosis. ISME J. 11:463–77
Porter SS, Faber-Hammond J, Montoya AP, Friesen ML, Sackos C. 2019. Dynamic genomic architecture of
mutualistic cooperation in a wild population of Mesorhizobium. ISME J. 13:301–15
Purvis A, Gittleman JL, Cowlishaw G, Mace GM. 2000. Predicting extinction risk in declining species. Proc.
R. Soc. B 267:1947–52
Queller DC. 1985. Kinship, reciprocity and synergism in the evolution of social behavior. Nature 318:366–67
Ranger CM, Biedermann PH, Phuntumart V, Beligala GU, Ghosh S, et al. 2018. Symbiont selection via
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

alcohol benefits fungus farming by ambrosia beetles. PNAS 115:4447–52

Razo-Belman R, Molina-Torres J, Martínez O, Heil M. 2018. Plant-ants use resistance-related plant odours
to assess host quality before colony founding. J. Ecol. 106:379–90
Regus JU, Quides KW, O’Neill MR, Suzuki R, Savory EA, et al. 2017. Cell autonomous sanctions in legumes
target ineffective rhizobia in nodules with mixed infections. Am. J. Bot. 104:1299–312
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

Renner SS. 1983. The widespread occurrence of anther destruction by Trigona bees in Melastomataceae.
Biotropica 15:257–67
Renner SS. 1998. Effects of habitat fragmentation on plant pollinator interactions in the tropics. In Dynamics
of Tropical Communities, ed. DM Newbery, HHT Prins, ND Brown, pp. 339–60. Oxford, UK: Blackwell
Sci. Publ.
Rodriguez LJ, Bain A, Chou LS, Conchou L, Cruaud A, et al. 2017. Diversification and spatial structuring in
the mutualism between Ficus septica and its pollinating wasps in insular South East Asia. BMC Evol. Biol.
17:207
Rojas-Nossa SV, Sánchez JM, Navarro L. 2016. Nectar robbing: a common phenomenon mainly determined
by accessibility constraints, nectar volume and density of energy rewards. Oikos 125:1044–55
Russell SL. 2019. Transmission mode is associated with environment type and taxa across bacteria-eukaryote
symbioses: a systematic review and meta-analysis. FEMS Microbiol. Lett. 366:fnz013
Sachs JL. 2015. The exploitation of mutualisms. See Bronstein 2015, pp. 93–104
Sachs JL, Mueller UG, Wilcox TP, Bull JJ. 2004. The evolution of cooperation. Q. Rev. Biol. 79:135–60
Sachs JL, Simms EL. 2006. Pathways to mutualism breakdown. Trends Ecol. Evol. 21:585–92
Sachs JL, Skophammer RG, Regus JU. 2011. Evolutionary transitions in bacterial symbiosis. PNAS
108:10800–7
Sachs JL, Wilcox TP. 2005. A shift to parasitism in the jellyfish symbiont Symbiodinium microadriaticum. Proc.
R. Soc. B 273:425–29
Schemske DW, Horvitz CC. 1984. Variation among floral visitors in pollination ability: a precondition for
mutualism specialization. Science 225:519–21
Schleucher E. 1999. Energy metabolism in an obligate frugivore, the superb fruit-dove (Ptilinopus superbus).
Aust. J. Zool. 47:169–79
Schultz TR, Brady SG. 2008. Major evolutionary transitions in ant agriculture. PNAS 105:5435–40
Shapira M. 2016. Gut microbiotas and host evolution: scaling up symbiosis. Trends Ecol. Evol. 31:539–49
Sicard A, Lenhard M. 2011. The selfing syndrome: a model for studying the genetic and evolutionary basis of
morphological adaptation in plants. Ann. Bot. 107:1433–43
Simonsen AK, Stinchcombe JR. 2014. Standing genetic variation in host preference for mutualist microbial
symbionts. Proc. R. Soc. B 281:20142036
Sørensen ME, Lowe CD, Minter EJ, Wood AJ, Cameron DD, Brockhurst MA. 2019. The role of exploitation
in the establishment of mutualistic microbial symbioses. FEMS Microbiol. Lett. 366:fnz148
Sprent JI. 2001. Nodulation in Legumes. London: R. Bot. Gard., Kew
Sudakaran S, Kost C, Kaltenpoth M. 2017. Symbiont acquisition and replacement as a source of ecological
innovation. Trends Microbiol. 25:375–90
Trivers RL. 1971. The evolution of reciprocal altruism. Q. Rev. Biol. 46:35–57
Visser B, Le Lann C, den Blanken FJ, Harvey JA, van Alphen JJ, Ellers J. 2010. Loss of lipid synthesis as an

·.•�-
evolutionary consequence of a parasitic lifestyle. PNAS 107:8677–82

www.annualreviews.org • The Evolution of Mutualistic Dependence 431

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)
 ES51CH18_Chomicki ARjats.cls August 8, 2020 15:50

Weber MG, Agrawal AA. 2014. Defense mutualisms enhance plant diversification. PNAS 111:16442–47
Weber MG, Keeler KH. 2013. The phylogenetic distribution of extrafloral nectaries in plants. Ann. Bot.
111:1251–61
Werner GD, Cornelissen JH, Cornwell WK, Soudzilovskaia NA, Kattge J, et al. 2018. Symbiont switching
and alternative resource acquisition strategies drive mutualism breakdown. PNAS 115:5229–34
West SA, Diggle SP, Buckling A, Gardner A, Griffin AS. 2007a. The social lives of microbes. Annu. Rev. Ecol.
Evol. Syst. 38:53–77
West SA, Fisher RM, Gardner A, Kiers ET. 2015. Major evolutionary transitions in individuality. PNAS
112:10112–19
West SA, Griffin AS, Gardner A. 2007b. Evolutionary explanations for cooperation. Curr. Biol. 17:R661–72
Access provided by University of Vienna - Main Library and Archive Services on 09/18/20. For personal use only.

West SA, Kiers ET, Pen I, Denison RF. 2002. Sanctions and mutualism stability: When should less beneficial
mutualists be tolerated? J. Evol. Biol. 15:830–37
Wright GA, Baker DD, Palmer MJ, Stabler D, Mustard JA, et al. 2013. Caffeine in floral nectar enhances a
pollinator’s memory of reward. Science 339:1202–4
Wyatt GA, Kiers ET, Gardner A, West SA. 2014. A biological market analysis of the plant-mycorrhizal sym-
Annu. Rev. Ecol. Evol. Syst. 2020.51. Downloaded from www.annualreviews.org

biosis. Evolution 68:2603–18

Wyatt GA, Kiers ET, Gardner A, West SA. 2016. Restricting mutualistic partners to enforce trade reliance.
Nat. Commun. 7:10322
Yanni D, Márquez-Zacarías P, Yunker PJ, Ratcliff WC. 2019. Drivers of spatial structure in social microbial
communities. Curr. Biol. 29:R545–50
Yu DW. 2001. Parasites of mutualisms. Biol. J. Linn. Soc. 72:529–46
Zhang B, Leonard SP, Li Y, Moran NA. 2019. Obligate bacterial endosymbionts limit thermal tolerance of
insect host species. PNAS 116:24712–18

·.•�-
432 Chomicki • Kiers • Renner

, Review in Advance first posted on

August 17, 2020. (Changes may
still occur before final publication.)