Asymmetric and uncertain interactions within mutualisms

Abstract

Although understanding mutualism stability has advanced over the last few decades, two fundamental problems still remain in explaining how mutualisms maintain stable. (i) How does resolve conflict between mutualists over resources? (ii) In the presence of less cooperative and/or uncooperative symbionts, how does prevent symbiont populations from becoming dominated by uncooperative individuals? Many past explanations of mutualism stability have assumed that interactions between mutualists are symmetrical. However, in most mutualisms, interactions between hosts and symbionts show varying degrees of asymmetry at different levels. Here, we review three major types of asymmetric interactions within obligate mutualisms: (i) asymmetric payoffs, which is also defined as individual power differences, (ii) asymmetric potential rates of evolutionary change, and (iii) asymmetric information states between hosts and symbionts. We suggest that these asymmetries between mutualists help explain why cooperation and conflict are inherent in the evolution of mutualisms, and why both hosts and symbionts present diversified phenotypes while cooperation predominates.

摘要

互惠合作系统的非对称性和不确定性

过去的几十年里，尽管对互惠合作系统稳定性的理解取得了较大的进展，但在解释互惠关系如何保持稳定方面仍然存在两个基本问题。(i)如何解决互惠合作双方之间的资源冲突？(ii)当存在不怎么合作和/或完全不合作的共生体时，为什么合作的共生体没有被不合作的个体竞争排斥掉？过去许多关于互惠稳定性的解释都假设互惠合作双方的相互作用是对称的。然而，在大多数共生关系中，宿主和共生体之间的相互作用在不同水平上表现出不同程度的非对称性。本文我们回顾了互惠合作系统中的三种主要的非对称相互作用：(i)收益非对称，也就是合作双方的实力非对称性；(ii)潜在进化速率非对称性；(iii)宿主和共生体之间的信息非对称性。我们认为，互惠合作系统非对称性有助于解释合作和冲突行为的不确定性(合作与竞争的内生不确定性)，以及为什么宿主和共生体都呈现出多样化的表型特征，而合作仍可占主导地位。

INTRODUCTION

Mutualisms are interspecific cooperative interactions, in which each party receives net benefits from the other. Mutualisms are ubiquitous to all ecosystems with almost every species being involved in at least one mutualism (Bronstein 2001b). Although the importance of mutualisms has long been recognized (Kiers et al. 2010), a complete understanding of how mutualisms remain stable over evolutionary time remains elusive. More specifically, it is often unclear why and how non-cooperative individuals (cheaters) (Ghoul et al. 2014), i.e. those less cooperative individuals or parasitic species, coexist with the mutualists (but see Frederickson 2017). Moreover, it remains a challenge to explain how cooperation is maintained when the use of a single resource has to be partitioned between hosts and symbionts in a single system. This is analogous to resolving the “tragedy of the commons” of humans (sensuRankin et al. 2007).

Several mechanisms are likely to contribute to the stability of any one mutualism (Herre et al. 1999), of which niche partitioning (spatial heterogeneity) and reduced virulence (Shi et al. 2006; Sachs et al. 2010, 2011) (e.g. selection for strategies in one mutualist that result in sustainable exploitation of the resources provided by the other mutualist) (Doebeli and Knowlton 1998) have been suggested as being important. However, these processes may not stabilize most mutualisms in the absence of additional mechanisms for two reasons: (i) Why do cheaters, those individuals that provide their partner no or insufficient benefits, fail to invade other populations of cooperative conspecifics, especially in populations in which individuals cannot disperse to another population or are involved in obligate systems? (ii) What prevents individuals that exploit partners unsustainably from outcompeting cooperative conspecifics and thus over-time invading the population? (Doebeli and Knowlton 1998; Hauert and Doebeli 2004). The realization of each of these factors would result in mutualism destabilization, and a shift to a system characterized by parasitism (Herre et al. 1999; Wang et al. 2011; Frederickson 2013). This has clearly not generally happened: although there is evidence of some mutualisms having evolved into completely or partially parasitic interactions (Compton et al. 1991; Pellmyr and Leebens-Mack 2000; Peng et al. 2008), mutualism instability is not a widespread occurrence among and within most mutualism types (Bronstein 2001a; Simms et al. 2006; Smith et al. 2008; Frederickson 2013, 2017).

Until relatively recently limitations to mechanisms posited to explain mutualism stability probably resulted, at least in part, from a general assumption that within individual mutualisms processes and outcomes between mutualists are symmetrical. In other words, each mutualist contributes equally to the stability of their particular mutualism, and if the ecological roles were reversed, the fundamental characteristics of the mutualism will remain unchanged (Rankin et al. 2007; Wang and Shi 2010; Archetti et al. 2011). However, in most mutualisms this is not so, because interactions between hosts and symbionts are inherently asymmetric, as the terms “host” and “symbiont” themselves implies (Pellmyr and Huth 1994; West and Herre 1994; Smith et al. 2008; Wang et al. 2011). The major asymmetry in most mutualisms is that the resources or services exchanged by each of the two partners differ markedly. For example, in Yucca tree–Yucca moth and fig tree–fig wasp obligate mutualisms, host Yuccas and fig trees provide moths or wasps oviposition sites, and food and shelter for the larval offspring of the insects, respectively. In return the moths or wasps pollinate their host’s flowers, and their adult offspring act as pollen vectors for the host plant (Janzen 1979; Pellmyr and Huth 1994; Weiblen 2002; Cook and Rasplus 2003; Herre et al. 2008; Dunn 2020). Rhizobial bacteria fix nitrogen for leguminous plants, while in return the plants provide carbon to the specialized root nodules housing the rhizobia (Long 1989; Zahran 1999; Denison and Kiers 2004; Kiers and Denison 2008; Udvardi and Poole 2013). The payoffs among the partners are thus unequal, and is especially likely in mutualisms with hosts that are able to control access to the resource exploited by their symbionts (Kiers et al. 2011; Wang et al. 2011). Individual symbionts are thus tied to a single host for all benefits but individual hosts receive benefits from multiple symbionts.

In addition to payoff asymmetries between hosts and symbionts, there are other two types of asymmetries that will potentially affect long-term mutualism stability. (i) An asymmetry in the evolutionary rates between hosts and mutualists. This may relate to differences in life histories, with the smaller symbiont species generally have shorter generation times than the larger host species (Table 1). (ii) An asymmetry in information states between hosts and symbionts. For example, hosts may be unable to discriminate between cheater and cooperative symbionts when individuals of each phenotype simultaneously visit the host (Jandér et al. 2012). Conversely, cheater or cooperative symbionts “know” of their own inherent behavioral intentions prior to any interaction with individual hosts (Leigh 2010). Here, we review these asymmetries, focusing on how they may affect variation in the potential mechanisms known to promote stability in some systems.

ASYMMETRIES IN PAYOFFS BETWEEN MUTUALISTS

Payoff inequalities and diversified strategies

Over the last few decades, mutualisms have not been thought of as benign relationships between hosts and symbionts but rather as controlled reciprocal exploitations (Herre et al. 1999; Frederickson 2017). There are several examples of individual symbionts failing to honor totally or partially their side of the mutualism “bargain” (Sachs and Simms 2006; Werner et al. 2018). These uncooperative symbionts can be either: (i) a separate cheater species to the cooperative symbiont species (Compton et al. 1991; Pellmyr and Leebens-Mack 2000; Peng et al. 2008, 2010; Kiers et al. 2011), or (ii) individual cheaters within a symbiont species i.e. otherwise dominated by cooperative individuals (Ghoul et al. 2014).

Almost all mutualisms are subject to “parasites,” species that solely inflict costs but give no direct benefits to one or both mutualists (Bronstein 2001a; Yu et al. 2004). Some parasitic species are commonly referred to as “cheater species.” These are usually closely related (often congeneric) to the “true” cooperative symbiont species, with both species coexisting on a single host species (Pellmyr and Leebens-Mack 2000; Kjellberg et al. 2001). In ant–Acacia mutualisms, host trees are defended against herbivory and leaf-pathogens by mutualist ant species but also harbor non-aggressive “cheater” species that still benefit from the food and shelter rewards given by the tree (Clement et al. 2008). Until relatively recently there has been no clear data as to the mechanisms that enable cheater species to coexist with cooperative ant species, although “cheater” species may in fact benefit hosts via synergistic effects with mutualistic species over the entire lifespans of these relatively long-lived host trees (Palmer et al. 2010). Likewise, in fig tree–fig wasp mutualisms there are several known cases of cheater wasp species that exploit fig tree inflorescences but fail to pollinate their host fig in return (Compton et al. 1991; Peng et al. 2008, 2010; Wang et al. 2015; Zhang et al. 2021). The mechanisms that enable these cheaters to have evolved and to persist in these systems remain unclear (Herre et al. 2008; Dunn 2020), although “free-riding” on the pollination behavior of the cooperative symbionts has been reported in some mutualisms. In fig–fig wasp mutualisms, cheater species are the sister species of the “true” pollinator wasp species but they tend to not be subject to host sanctions (Compton et al. 1991; Peng et al. 2008, 2010; Zhang et al. 2021). However, non-pollinating fig wasps, “cheater species” of different families or subfamilies to the pollinator wasps (Borges 2015), can be subject to host sanctions (Wang et al. 2010).

More commonly, cheating in mutualisms refers to those individuals within an overwhelmingly cooperative symbiont population that obtain benefits from the host species but fail to reciprocate, behavior that can be either facultative or obligate (Ghoul et al. 2014). Facultative cheats simply fail to benefit their host for random reasons, for instance if a pollinating insect is unable to collect pollen before visiting another host plant of the same species to obtain a nectar reward. Obligate cheats are genetically predisposed to cheat by avoiding the costs of benefitting their host and are thus predestined to be uncooperative. To be subject to selection, i.e. for individuals to avoid costs of providing hosts services and/or resources, cheating behavior thus has to be largely obligate (Dunn 2020).

It is noteworthy that in some systems, for instance some insect nursery pollination mutualisms, symbionts have the opportunity to subsequently “cheat” even if they initially provide their hosts benefits (e.g. via pollination). This is because symbionts may have the opportunity to subsequently unsustainably exploit resources that are otherwise optimally partitioned by hosts to be used by both symbionts and hosts, i.e. flowers that are either eaten by the larvae of the pollinating insects or become seeds (Jandér and Herre 2010; Wang et al. 2011). Overexploitation can thus clearly result in the potential destabilization of some mutualisms due to selection for ultra-exploitative symbionts that outcompete their more cooperative conspecifics in the short term, potentially resulting in symbiont population fluctuations (Figs 1 and 2).

Power asymmetries between hosts and symbionts

The host species in most mutualisms typically has the ability to control access to the resources exploited by their symbionts. An individual host may thus be able to “choose” to interact with those symbionts that result in hosts maximising their benefit—cost ratios (Kiers et al. 2003, 2011; Leigh 2010; Archetti et al. 2011; Wang et al. 2015). One mechanism by which this can be achieved is via partner choice, whereby hosts exhibit mechanisms prior to any exchange of resources and/or services to preferentially interact with cooperative, beneficial symbionts. For example, host bobtail squid recruit symbiotic bacteria from the surrounding seawater in order to enable the host squid to bioluminesce. These squid only allow cooperative bacteria that have a bioluminescent capability to enter their bodies and they do this by deploying an effective mucus-based filter (Nishiguchi 2002; McFall-Ngai 2014).

Another mechanism by which individual hosts can increase the likelihood of interacting mainly with cooperative symbionts is via host sanctions (Denison 2000). This can be achieved by either “rewarding” those symbionts that are cooperative and/or “punishing” those that cheat, each strategy being essentially “one side of the same coin.” For example, in some fig tree–fig wasp mutualisms, legume–rhizobia mutualisms, and ant–ant plant mutualisms, symbionts derive the highest net benefits from host plants depending on pollination (Jandér and Herre 2010; Wang et al. 2010, 2014; Jandér et al. 2012; Zhang et al. 2019; Jandér 2021), nitrogen income (Kiers et al. 2011), or protection against browsing (Palmer et al. 2008), respectively.

However, if individuals of a parasitic “cheater” species and/or uncooperative individuals of a normally cooperative symbiont population co-occur with cooperative symbionts, hosts may be unable to accurately target sanctions even though they have a theoretical “power” advantage (see also below for information asymmetries). In such a scenario, the nature of sanctions can vary within and among mutualism systems (Fig. 1). For instance, in fig tree–fig wasp mutualisms when both cheater and cooperative symbionts are present in the same fig, all wasp symbionts tend to escape sanctions (Jandér et al. 2012). This is because in these systems sanctions work via host trees disinvesting from unpollinated figs at the level of the individual fig, so the host reward is maintained when a fig receives pollen investment regardless of whether some of the symbiont wasps failed to pollinate (Jandér and Herre 2016; Zhang et al. 2019; Dunn 2020).

In mutualisms in which the host is able to more freely discriminate between cheaters and cooperative symbionts, sanctions can be more accurately targeted. For example, some grouper fish species are hosts to cleaner wrasse, who remove ectoparasites and small particles of food debris trapped between the teeth of the host fish (Bshary 2002; Soares et al. 2010; Gingins and Bshary 2015). However, there are some “cheater” cleaner symbionts that also eat mucus and bite small pieces of healthy tissue from the host groupers. Individual hosts are cleaned at specific sites around which several cleaner wrasses congregate. Hosts therefore have the propensity to choose to associate with the most cooperative cleaner symbionts according to the benefits gained from previous interactions with individual cleaner fish. Furthermore, host groupers are likely to chase away individual cleaner wrasse that have in the past inflicted costs to the individual host fish (Bshary 2002; Bshary and Schaffer 2002; Bshary and Grutter 2005).

Asymmetries in intraspecific competition

In most mutualisms, a single host will interact with multiple individual symbionts (Wang et al. 2011; Ezoe 2012), with symbiont population densities often greatly exceeding those of hosts. Intraspecific competition is often density dependent, and will thus likely be stronger among symbionts than hosts (Doebeli and Knowlton 1998; Holland et al. 2002). Indeed, density-dependent competition among symbionts for host resources within a single mutualism has been found in several systems involving disjunct taxa (Ferdy et al. 2002; Yu et al. 2004; Wang et al. 2009, 2011; Adam 2010). However, there are fewer examples of competition among hosts for goods and/or services provided by symbionts. Host control over symbiont access to resources can affect the degree to which individual symbionts compete for those resources. This can be achieved by limiting or extending symbiont access in space and/or time to:only a subset of the total available resource, which in turn increases or decreases intraspecific competition among symbionts, respectively (e.g. Wang et al. 2009).

A manipulative experiment in a fig–fig wasp mutualism showed that variation in symbiont competition affects the key interaction between hosts and symbionts: whose offspring uses the plant’s flowers (Janzen 1979). When intraspecific competition for fig flowers among pollinator wasps is intense (multiple foundress wasps are simultaneously introduced into a fig (syconium)) cooperation between host and symbiont occurs: there is a positive correlation between seed and wasp offspring numbers. However, when competition for fig flowers is weak among symbiont wasps (foundress wasps are sequentially introduced into a fig), cooperation between host and symbionts breaks down: there is a negative relationship between seed and wasp offspring numbers (Wang et al. 2011; Fig. 3).

Hosts are able to regulate competition among symbionts. For instance, in at least one fig tree–fig wasp mutualism, trees manage how many pollinating wasp symbionts enter each of their figs, which indirectly sets the rate at which wasps convert flowers into wasp gall, and hence wasp offspring (Wang et al. 2009, 2015). The host trees do this by limiting the time that the wasps have to enter a fig. A field experiment using a Chinese population of Ficus racemosa found that fig (enclosed inflorescences containing many tiny flowers) receptivity to wasps is density dependent. If no wasps or only a single wasp enters, a fig will remain receptive to further wasps for up to a week. However, if multiple wasps enter a single fig within only a few hours, a fig will become unreceptive to further wasp entry by closing the bracts to the entrance tunnel used by wasps to enter the fig (Wang et al. 2009, 2015). This process interacts with seasonal effects on wasp lifespan because wasp longevities are prolonged in cooler, wetter periods (Dunn et al. 2008; Wang et al. 2009; Jeevanandam and Corlett 2013). In addition, figs grow more slowly during the cooler “winter,” with a prolonged period of receptivity to pollinating wasps (Wang et al. 2009). During “winter” wasps thus have increased potential to exploit host flowers by hosts allowing the presence of more foundresses in each fig. However, wasp oviposition rates, and hence rates of flower exploitation, are reduced via interference competition for oviposition sites. The trees are thus able to ensure that their wasp symbionts exploit their flowers sustainably (Wang et al. 2009, 2015).

Restrictions giving symbionts temporal access to host resources in some mutualisms can also be partly achieved using chemical rather than physical barriers. For instance, in fig tree–fig wasp mutualisms, host trees emit species-specific chemical bouquets of volatile organic compounds that attract the female pollinating wasps (Okamoto and Su 2021). In some host tree species, individual figs cease producing such attractants after multiple wasps have entered a fig (Muhlemann et al. 2006; Proffit et al. 2008; Rodriguez-Saona et al. 2011). A narrow time-window of host resource availability may thus directly reduce the degree of resource exploitation due to the presence of fewer symbionts and reflects clearly the power asymmetry between partners. However, this may be offset by relaxed competition among wasps enabling individuals to more fully exploit host resources (Proffit et al. 2008; Wang et al. 2011).

The ant plant Cordianodosa restricts the resources it provides its ant symbionts in space rather than over time. This is achieved by adjusting the volume of the specialist structures it provides to house its ant symbionts (domatia) according to variation in ant densities (Frederickson et al. 2013). This enables individual host trees to only incur the costs of supporting the optimal number of ant symbionts needed for defence against herbivores by maintaining competition for domatium space among ant symbionts (Goncalves-Souza 2016; Pacheco and Del-Claro 2018; Kokolo et al. 2019; Amador-Vargas et al. 2020). Similarly, in the pollinating fly-alpine globeflower nursery pollination mutualism, the morphology of host flowers induces strong density-dependent competition among the larvae of the pollinating flies (Ferdy et al. 2002; Dufaÿ and Anstett 2003; Despres et al. 2007; Ibanez et al. 2009, 2010; Hemborg and Despres 2011). This results in high larval mortality and prevents unsustainable exploitation of the host plant’s seeds by symbionts.

In several species of pollinating fig wasp, competition between ovipositing foundresses in receptive figs for oviposition sites (fig flowers) results in aggression (Ramírez-Benavides 1970; Moore and Greeff 2003; Dunn et al. 2015), that in at least one species manifests into physical injuries and even deaths between competitors (Dunn et al. 2015). Such competitive interactions among symbionts are likely to reduce the potential for unsustainable host resource exploitation and hence facilitate mutualism stability (Holland et al. 2002; Yu et al. 2004). This has been demonstrated experimentally in at least one fig tree–fig wasp mutualism. With reduced competition among wasps by time-spaced sequential introduction into figs, wasps were on average able to exploit more fig flowers than when competition was more intense, i.e. when the same number of wasps were introduced concurrently (Wang et al. 2011; Fig. 3).

In cleaner fish mutualisms, competition between cleaners at cleaning stations can promote cooperation, resulting in host fish receiving higher quality cleaning services (Adam 2010). Pairs of cleaners provide hosts a better service than single cleaners, and because two-cleaner cleaning rates are lower than double that of single cleaners, this is suggestive of competition between cleaners (Bshary et al. 2008). Pairs of cleaners also extend the total interaction time with host fish, which is also in accordance with competition between cleaners benefitting hosts (Gingins and Bshary 2015). Experimental relaxation of competition via the provision of additional food sources to pairs of cleaners shows that decreased food consumption rates on hosts are due to competition for host “provided” food items (Bshary and Grutter 2005; Bshary et al. 2008).

In some mutualisms, different symbiont species may simultaneously interact with a single individual host. This may result in interspecific competition among symbionts for a single host resource. For example, in some Mexican ant–Acacia mutualisms, individual Acacia trees can be colonized by several species of Pseudomyrmex ants, some of which are mutualistic and engage in costly defense of their host, and others that provide hosts no benefits (Heil 2009). Competition between co-colonizing ant species for tree space is strong (Heil 2009). After initial colonization by two interspecific queens, one of a cooperative species and the other an uncooperative species, the section of the tree that receives protection from herbivore attack by its ant colony grows more quickly than the unprotected section that harbors the uncooperative ant species. The cooperative ant colony also grows more relative to the competing colony, which eventually results in the eviction of the uncooperative ant colony. This is because the protected parts of the plant have enhanced growth and attract more insects to the areas of the host plant monopolized by the cooperative ants, providing the cooperative colony with increased resources (Heil 2009). Ultimately, these rewards increase the probability of a cooperative symbiont ant species becoming established on, and interacting over the long term with, an individual host tree compared with an uncooperative species that provides the host no benefits (Archetti et al. 2011).

Asymmetric rates of evolution between mutualists

In most mutualisms, another major asymmetry between partners is that hosts and symbionts are taxonomically disjunct, often widely so. Hosts are also usually much larger than symbionts, often by several orders of magnitude. An asymmetry in host and symbiont generation times is thus unsurprising, with symbiont generation times much shorter than that for hosts. The multiple symbiont generations per host generation might thus give symbionts a greater propensity for evolutionary change than hosts.

For example, in plant–insect pollinator mutualisms the pollinating insects often have multiple generations in a single year. However, the host plants will set seed at a much lower rate and may even take several years to successfully reproduce (see Table 1). Mutation rate is likely to be correlated positively with generation time, demonstrating that these symbionts have higher probability of mutation than their hosts (Frankham 1996; Cook and Rasplus 2003). This may also result in higher genotypic and phenotypic variation in symbionts than hosts, which clearly adds to the challenge in explaining why symbiont populations fail to become dominated by any cheater phenotype, given the likelihood that cheaters will have a theoretical fitness advantage over cooperative conspecifics (but see Frederickson 2017). However, it is theoretical possible that the species with the lowest rate of reproduction, the host, may have higher genotypic and phenotypic variation than the symbionts (e.g. Red King effect) (Frean and Abraham 2004).

Some mutualism types are diverse, presenting an opportunity to explore such predictions. For instance, in fig tree–fig wasp systems there are ~750 species of host fig tree (Ficus) divided into 19 taxonomic sections (Rønsted et al. 2008; Cruaud et al. 2012). Each host tree species is only pollinated by one or a few wasp species of the family Agaonidae, which is comprised of 20 genera and is more species-diverse than Ficus (Cook and Segar 2010). There are multiple cases of several wasp species pollinating a single fig tree species (Lopez-Vaamonde et al. 2002; Molbo et al. 2003; Haine et al. 2006; Compton et al. 2009; Lin et al. 2011; Chen et al. 2012; Conchou et al. 2014; Rodriguez et al. 2017; Sutton et al. 2017; Yu et al. 2019; Zhang et al. 2020), with many more probably unknown due to coexistence on the same host of cryptic pollinator species (Molbo et al. 2003; Haine et al. 2006). Likewise, there are several known cases of a single wasp species pollinating more than one host fig tree species (Compton et al. 2009; Wachi et al. 2016; Wang et al. 2020).

Other insect nursery systems, such as the Yucca tree–Yucca moth (Pellmyr 2003; Smith et al. 2008) and Glochidion tree–(Euphorbiaceae) Epicephala moth (Gracillariidae) (Kawakita and Kato 2006) mutualisms, although not as diverse as fig tree–fig wasp systems, also show similar trends with symbionts being more speciose than hosts. Such patterns are consistent with the predictions that symbionts are more evolutionary labile than hosts (Cook and Rasplus 2003). In a mutualism, the species with the potential to respond most quickly to a given selection pressure should be at an advantage in its interactions with the other mutualist. This is because of the strong selection pressure exerted by the other mutualist via the direct benefits available to the other species in the interaction. In other words, there may be some sort of “arms race” between mutualists in traits in symbionts that maximize the benefits that can be obtained from the host, and traits in hosts that prevent cheating and or unsustainable resource exploitation by symbionts (Herre et al. 1999; Thompson 2013). The faster-evolving species, i.e. the symbiont, is thus likely to be at an advantage via a Red Queen effect (Herre et al. 1999). However, it is possible that the slower responding species, the host, may have an advantage. Because slowly evolving species are constrained by the coevolutionary process (Bergstrom and Lachmann 2003), the slower-evolving species could become dominant due to a Red King effect (Frean and Abraham 2004).

Evidence in support for the Red Queen effect in at least one mutualism type has been found in a recent comparative genomics study. Those ant species involved in a mutualism with a host plant species had higher rates of evolution across their genomes than ant species not involved in mutualisms (Rubin and Moreau 2016). Specifically, genes associated with neurogenesis and muscle activity in those ant species involved in mutualisms, which are key traits associated with behaviors that benefited hosts, were found to evolve at higher rates than in closely related generalist ant species. This suggests that the genes associated with behaviors exhibited in mutualisms are under relatively strong selection pressure, consistent with the importance of Red Queen effects in ant–plant mutualisms. However, a high rate of evolutionary change may only occur during the initial stage of a mutualism, even though selection may be consistent over time (Rubin and Moreau 2016). Red King effects may therefore still play a role after the initiation of a mutualistic relationship.

For a specific mutualism, whether the slower- or faster-evolving species allocates more resources or dominates the mutualistic interaction may depend on both the evolutionary pathway of the mutualism and current ecological conditions (Herre et al. 1999; Gao et al. 2015). In nature, the degree to which resources are allocated by one mutualist to the other so that each mutualist receives net benefits will probably vary according to ecological conditions. This may result in a shift between any Red Queen and Red King effects. Models have shown that either an asymmetry in payoffs or an asymmetry in the number of cooperative partners may result in a shift between a Red Queen and a Red King effect (Bergstrom and Lachmann 2003; Gao et al. 2015). Additionally, shifts between these differing outcomes may be influenced by variation in the number of players and differences in the reward mechanisms used between mutualist species under different environments or selection pressures.

ASYMMETRIC INFORMATION STATES BETWEEN MUTUALISTS

Hosts have incomplete information on the cooperative intentions of individual symbionts

Hosts may be unable to efficiently distinguish between individual symbionts that are either cooperative or cheaters (Friesen and Mathias 2010; Archetti et al. 2011; Jandér et al. 2012). In such cases hosts cannot ensure that cooperative symbionts receive greater net benefits than cheaters. If a host is at a particular point in time only visited by cheater symbionts, “sanctions” can be accurately deployed, e.g. via inflorescence abortion in some nursery pollination systems and the withholding of photosynthate in legume–rhizobia systems (Pellmyr et al. 1996; Kiers et al. 2003; Jandér and Herre 2010). However, cheaters, whether they be uncooperative individuals within a wider predominantly cooperative symbiont population or individuals of a cheater species, are likely to interact with individual hosts simultaneously to cooperative symbionts (Compton et al. 1991; Pellmyr et al. 1996; Denison and Kiers 2004; Kiers and Denison 2008; Peng et al. 2008, 2010; Friesen and Mathias 2010). Under such circumstances, hosts may be unable to accurately deploy sanctions accurately because all symbionts are exploiting the same resource at a relatively small scale. Moreover, it may be unnecessary for hosts to deploy possibly costly sanctioning mechanisms if hosts are provided with adequate benefits from those cooperative symbionts even when cheaters are also present (e.g. Jandér et al. 2012). Cheaters may thus obtain higher net benefits than cooperative symbionts by avoiding any costs of cooperation (Pellmyr and Huth 1994; Wang et al. 2010; Kiers et al. 2011; Jandér et al. 2012; Jander and Steidinger 2017). These cheaters are thus “free-riders” and benefit from the costly behavior of the cooperative symbionts (Wang et al. 2010; Ghoul et al. 2014; see Dunn 2020 for a more detailed discussion on free-riders in fig tree–fig wasp mutualisms).

Symbionts have incomplete information as to whether host sanctions will be deployed or not for a given action/inaction

Uncooperative or less cooperative symbionts are likely to be unaware of the timing and strength of host sanctions. Deploying sanctions may be costly to hosts and/or hosts may obtain additional benefits by not always sanctioning all cheater symbionts (Dunn 2020). Although the net costs or benefits to hosts will usually be the sum of all of the actions of multiple symbionts (Palmer et al. 2010), these may vary in space on a single host, e.g. some flowers of the same tree may receive adequate pollen while others may not. If sanctions are only deployed when costs to hosts exceed the total net benefits, an individual symbiont when and where it interacts with its host will not know if this host threshold will be exceeded, for instance if the cooperative actions of other symbionts are needed, and thus if host sanctions will be deployed (see Wang et al. 2014; Dunn 2020 for more detailed discussions on this point in a fig tree–fig wasp mutualism). For example, host Yucca trees may abort fruits that are insufficiently pollinated. This imposes costs to symbiont moths by killing all moth larvae even though some aborted fruits may have resulted from flowers that by chance were pollinated by very few, fully cooperative symbiont moths (Pellmyr and Leebens-Mack 2000; Wang et al. 2010). Host fig trees do not abort all unpollinated figs and do abort some pollinated figs, which may reflect, at least in monoecious and “male” dioecious species, the potential benefits to trees of allowing some wasp offspring to mature as pollen vectors even though no seeds will be produced (Jandér and Herre 2010; Wang et al. 2014; Dunn 2020). The failure of hosts to sanction cheats in plant–pollination mutualisms when host plants are pollinated artificially, also demonstrates the unpredictability of host sanctioning mechanisms to symbionts (Wang et al. 2010; Kiers et al. 2011).

The game-theory strategy “super-rational” (Hofstadter 1983) is relevant to such information asymmetries. In this game, individual hosts and symbionts each seek the best payoff to maximize their own net benefits, and do not necessarily account for the strategy used by the other mutualist. In mutualism research this is often termed “resource allocation” (Jandér and Herre 2016). A super-rational strategy for the host is to reward symbionts if the net benefits accrued increase and to sanction symbionts if net benefits decrease, irrespective of the behavior, whether cooperative or uncooperative, of individual symbionts (He et al. 2015). Mutualism stability in the presence of information asymmetries may thus be maintained by “super-rationalism” (preferential resource allocation by hosts to obtain high net benefits) alone; hosts thus do not need the ability to distinguish cooperative symbionts from cheaters and then to react, i.e. target sanctions accordingly, in order to ensure symbiont cooperation (He et al. 2015). The presence of such mechanisms may also explain why “cheaters” are able to “free-ride” in otherwise cooperative symbiont populations and the presence of totally uncooperative “symbiont” species (Compton et al. 1991; Peng et al. 2008, 2010) in some mutualisms (Bronstein 2001b; Yu et al. 2004).

CONCLUSIONS AND OUTLOOK

Understanding asymmetries in interactions between hosts and symbionts within mutualisms has recently advanced such as the proliferation in studies of host sanctions and host-dominated partner choice. Much previous research has failed to fully address how such asymmetries evolve and contribute to the maintenance of long-term system stability.

We have summarized three major asymmetries between hosts and symbionts, namely in evolutionary rates, power, and information states. These asymmetries might result in uncertain mutualistic interaction through either biological interaction or environmental fluctuation, but in turn might contribute to maintain a higher probability of cooperation. Different asymmetry types are not necessarily mutually independent, and each type of asymmetry may have varying effects on host–symbiont interactions even within the same mutualism. To our knowledge, there are currently no empirical studies that explicitly test how different types of asymmetries interact in producing varying outcomes between hosts and symbionts. These are urgently needed.

Acknowledgements

We thank for Stuart A. West and Shuang-Quan Huang for comments and suggestions, and Zhongyang Quan-Wei, Jun-Zhou He, and Lei Gao for their discussion and help in manuscript preparation.

Funding

This work was supported by NSFC-Yunnan United fund (U2102221), the National Science Fund for Distinguished Young Scholars (31325005), the National Natural Science Foundation of China (31270433, 31370408, 32070453), and Wenner-Gren Foundations.

Conflict of interest statement. The authors declare that they have no conflict of interest.

REFERENCES