Symbiotic microbes have become increasingly recognized to mediate interactions between natural enemies and their hosts. The ecologies of these symbioses, however, are poorly understood in many systems, and a predictive framework is needed to guide future studies. To achieve this, we focus on heritable, defensive microbes of insects. Our review of laboratory-based studies identifies diverse bacterial species that have independently evolved to protect a range of insects against parasitoids, parasites, predators and pathogens. Although defensive mechanisms are typically unknown, some involve toxins or the upregulation of host immunity.
Despite substantial benefits of infection in the presence of natural enemies, the protective symbionts of insects are often found at intermediate levels in natural populations. Using a host-centred population genetics approach made possible by the host restriction and cytoplasmic inheritance of these microbes, we propose that balancing selection plays a major role in symbiont maintenance, with protective benefits in the presence of enemies and infection costs in their absence. Other mediating factors are likely to be important, including temperature, superinfections and transmission dynamics.
While few studies have provided evidence for defence in the field, several studies have shown symbiont infection frequencies to be dynamic, varying across temporal and spatial gradients and food–plant associations. Newly presented data from our pea aphid research reveal that temporal shifts in defensive symbiont prevalence can be quite rapid, with Hamiltonella defensa showing 10–20% shifts around a seasonal average of c. 50%. Such findings contrast with more unidirectional changes seen in laboratory population cages, suggesting temporal changes in the costs and benefits of symbionts in the field.
To frame future research on defensive symbiont ecology, we briefly consider a range of studies needed to test laboratory- and field-derived predictions on defensive symbiosis. Included are investigations of defensive mechanisms, symbiont-driven co-evolution and community-level effects. We also consider the need for more thorough and highly resolved molecular diagnostics of natural infections, laboratory studies on functional differences between symbiont strains and species and studies on the relative costs and benefits of defenders in nature.
The emerging theme of symbiont-mediated defence across eukaryotes suggests that knowledge of the functional mechanisms behind protection and natural symbiont dynamics could be key to understanding many of the world's antagonistic species interactions. Thus, the development of insects as a model for such studies holds promise for these organisms and beyond.
Antagonistic interactions between eukaryotes and their natural enemies are among the most rapidly evolving and diverse types of species interactions on the planet. These associations have favoured impressive levels of polymorphism and numerous instances of adaptive evolution, while providing a compelling explanation for the maintenance of sex (Lively 1989; Hamilton, Axelrod & Tanese 1990; Agrawal & Lively 2002; Thrall & Burdon 2002). Traditionally, research on defence has focused on eukaryote-controlled mechanisms of host protection, but there is an emerging recognition that symbiotic micro-organisms also mediate interactions with natural enemies (e.g. White & Torres 2009).
Since many of the bacteria and fungi engaged in defensive symbioses are heritable (i.e. reliably transmitted from parent to offspring), hosts harbouring protective symbionts can be favoured by natural selection during times when natural enemies are prevalent. Also, just as plant and animal species are often polymorphic for defence-related genes, hosts may also vary in the presence and variety of defensive symbionts. Theoretical and empirical discoveries on the maintenance of host-encoded defensive polymorphisms (Hedrick 1986, 2002) suggest that polymorphic symbiont infection may derive from enemy-driven selective pressures that fluctuate over time, space or with changing frequency. While an understanding of the correlates of symbiont frequency could help identify important agents of selection, outside of plants (e.g. Clay & Holah 1999), few studies have examined infection dynamics, let alone defence itself, in natural populations. Instead, most of our knowledge on symbiont-based defence is obtained from laboratory studies.
In this review, we summarize discoveries on heritable, defensive symbionts infecting insects, focusing largely on aphids, which have received the most study in both laboratory and field. Since the defensive symbionts of insects likely share key characteristics with those found in other groups, their study can broaden our understanding of symbiont-based defence, while highlighting the importance of mutualistic microbes as frequent mediators of antagonistic interactions.
A crash course on heritable symbionts in insects
Insects and other terrestrial arthropods exhibit widespread and diverse interactions with micro-organisms (Buchner 1965). Most species, for example, are infected with heritable bacteria, for which the primary route of new infection occurs via high rates of maternal transmission (Moran, McCutcheon & Nakabachi 2008). Relative to gut associates, which are also prevalent in insects, within-host ‘communities’ of heritable bacteria are less diverse, with none to a few types typically present in a single host (e.g. Mateos et al. 2006; Russell et al. 2013). The relative simplicity of heritable, defensive communities compared with those in other animals (e.g. gut microbes in mammals), combined with tools to transfer or cure them, often results in tractable systems to explore symbiont-based protection.
Theory predicts that associations with heritable microbes have important consequences for the ecology and evolution of infected hosts, as they must either manipulate host reproduction or provide net fitness benefits to spread and persist within host populations (Bull 1983; Werren & O'Neill 1997). Insects feeding solely on nutrient-poor substrates (e.g. plant sap or vertebrate blood), for example, require associations with microbes to obtain nutrients found in insufficient quantities in their restricted diets (Baumann 2005). Associations with such obligate nutritional symbionts have played major roles in the specialization and subsequent diversification of numerous insect groups (c. 10% of insects) by providing biosynthetic pathways allowing the exploitation of previously unusable resources (Douglas 1989; Moran & Telang 1998; Shigenobu et al. 2000).
In most cases, however, associations with heritable bacteria are facultative from the hosts' perspective. Findings to date reveal that prevalent lineages of facultative symbionts, including Wolbachia, Cardinium, Rickettsia and Spiroplasma, employ a range of strategies to manipulate host reproduction and development (Moran, McCutcheon & Nakabachi 2008). Other symbionts, including members of these same lineages (Wolbachia and Spiroplasma), have been found to mediate important ecological interactions, including defence against natural enemies (Oliver et al. 2003; Hedges et al. 2008; Teixeira, Ferreira & Ashburner 2008; Jaenike et al. 2010; Xie, Vilchez & Mateos 2010).
Defensive symbionts and protective mechanisms
Virtually all insects face attack by a wide range of natural enemies, which exert strong selective forces on their insect hosts (Feeny, Blau & Kareiva 1985; Bernays 1997). Since bacteria and other microbes often possess pathogenicity genes and can produce diverse bioactive compounds, which potentially function as toxins or deterrents (e.g. Piel et al. 2004; Degnan & Moran 2008a; Degnan et al. 2009; Kroiss et al. 2010; Hansen, Vorburger & Moran 2012), insects may benefit from novel defensive properties provided by microbial associates. Symbionts may also confer protection through indirect mechanisms, by ‘priming’ the host immune system against subsequent invasions or by competing with natural enemies for limited host resources.
However, since most heritable symbionts are uncultivable and hence not amenable to direct assessments of gene function (e.g. reverse genetics), determining mechanisms of symbiont-based protection has proven difficult for many interactions. Even when specific mechanisms that underlie protection remain unknown, infection itself, which typically varies among members of a given population and is easily diagnosed, provides a useful marker for monitoring resistance in natural populations. In the following section, we review specific examples of heritable, defensive symbionts in insects (Table S1), briefly describing protective mechanisms when known.
Most insect species are attacked by parasitoids, which live on (ectoparasitoids), or within (endoparasitoids), a single host individual, eventually killing the host to complete development (Godfray 1994). Internally developing parasitoids are hyperdiverse and are generally host specific. Given the typical life or death outcome of host–parasitoid interactions, there is strong selection for the evolution and maintenance of offensive (i.e. virulence) traits in parasitoids and defensive (i.e. resistance) traits in hosts. Further, many endoparasitoids employ complex developmental and physiological manipulations of the host to create an environment suitable for their development (Beckage & Gelman 2004). Protection against endoparasitioids may be a widespread phenotype conferred by heritable bacteria given the number, specificity and intimacy of these interactions. Several examples have been reported to date (Table S1).
In one well-studied example, the pea aphid, Acyrthosiphon pisum, receives protection against an important natural enemy, the parasitic wasp Aphidius ervi, through infection with the bacterial symbiont Hamiltonella defensa (Oliver et al. 2003; Oliver, Moran & Hunter 2005). Symbiont-mediated defence in A. pisum is a physiological trait: female wasps readily lay eggs in both H. defensa-infected and uninfected aphids, but wasps fail to complete development in resistant hosts (Oliver et al. 2003). The bacterial symbiont, H. defensa, is itself often infected with bacteriophages (viruses that infect bacteria) called APSEs, which are temperate phages related to the lambdoid phage P22 (Podoviridae) (van der Wilk et al. 1999; Sandstrom et al. 2001). Three APSE variants (APSE1-3) have been characterized in A. pisum. All share a backbone of conserved genes, while also harbouring a variable ‘cassette’ region that contains one of three toxin genes: shiga toxin (stx, APSE-1), cytolethal distending toxin (CdtB, APSE-2) and YD-repeat toxin (YDp, APSE-3) (van der Wilk et al. 1999; Moran et al. 2005; Degnan & Moran 2008a). Not all H. defensa strains carry APSE; field-collected H. defensa-infected aphids are occasionally found to be phage free, and one variant (APSE-3) can be spontaneously lost in laboratory-held populations (Oliver et al. 2009). The strength of protection conferred by H. defensa varies from moderate to complete and correlates with phage variant (APSE-2 moderate; APSE-3 high to complete) (Oliver, Moran & Hunter 2005; Degnan & Moran 2008b; Oliver et al. 2009). Although direct evidence is lacking, this correlation, combined with the finding that the protective phenotype is completely lost when phages are absent (Oliver et al. 2009), has led to the hypothesis that phage-encoded toxins are responsible for harming wasp tissue. When we consider that A. pisum lacks a strong encapsulation response (i.e. the innate cellular response of most insects to large invading entities) (Carver & Sullivan 1988; Laughton et al. 2011) and that individuals not infected with APSE-bearing H. defensa are highly susceptible to parasitism, it appears that this insect relies largely on a heritable symbiont and its bacteriophage for protection against an important natural enemy (Oliver, Moran & Hunter 2005).
While H. defensa occurs in roughly 14% of aphid species (Oliver et al. 2010), only two other studies have unequivocally shown protection against parasitoids. In one case, H. defensa-infected Aphis fabae (black bean aphid) received protection against the wasp Lysiphlebus fabarum (Schmid et al. 2012). Interestingly, protection levels increased with aphid age, and wasps completing development in symbiont-protected hosts suffered sublethal fitness costs (Schmid et al. 2012). In the third case, a strain of H. defensa experimentally transferred between distantly related aphid species (Aphis craccivora to A. pisum), which share host plants, conferred protection against the parasitoid A. ervi in the novel host (Oliver, Moran & Hunter 2005). This work demonstrates the potential for the protective phenotype to be laterally transferred among host species sharing a resource (e.g. food plant) and also suggests a protective role for this symbiont in its natural host, A. craccivora. Further findings suggest that H. defensa may protect against a wider range of parasitoids in A. pisum (Ferrari et al. 2004), although it may not protect in all of its host aphid species (Łukasik et al. 2013a).
Additional symbionts have been implicated in the defence of aphids against parasitoids. For instance, while some strains of the closely related Regiella insecticola do not confer resistance in Ac. pisum (Oliver et al. 2003; Hansen, Vorburger & Moran 2012), a single strain from Myzus persicae conferred protection against the parasitoid Aphidius colemani within both native and novel (Ap. fabae) aphid species (Vorburger, Gehrer & Rodriguez 2010). Interestingly, bacteriophages were absent from this defensive Regiella strain, although its genome did encode pathogenicity pathways that were absent from a non-protective strain (Hansen, Vorburger & Moran 2012).
Serratia symbiotica also reduced rates of successful parasitism by the wasp A. ervi in pea aphids, albeit to a lesser extent than H. defensa (Oliver et al. 2003). However, when competed against genetically identical pea aphids with differing symbiont complements in the presence of parasitoids, hosts with S. symbiotica did not increase in prevalence, losing out to those with H. defensa (Oliver et al. 2008). This is likely because aphids with the latter symbiont produced more offspring after parasitism than their uninfected counterparts (revealing a direct fitness benefit), which was not the case for those harbouring S. symbiotica (Oliver, Moran & Hunter 2006). The lack of a strong protective phenotype and direct benefits in the presence of natural enemies, combined with limited evidence for known defensive factors in the S. symbiotica genome (Burke & Moran 2011), suggest other benefits, such as thermal tolerance (e.g. Montllor, Maxmen & Purcell 2002; Russell & Moran 2006), may be more important in governing this symbiont's fate.
While we expect the number of instances of symbiont-driven defence against parasitoids to grow, outside of aphids, there is just one documented case, involving Drosophila hydei and a heritable Spiroplasma symbiont (Xie, Vilchez & Mateos 2010). This microbe was shown to increase larval-to-adult survival by 40% in the presence of parasitism by Leptopolina heterotoma. Although the mechanism is currently undefined, wasps do not avoid oviposition into Spiroplasma-infected hosts, suggesting a physiological basis for defence, as seen within the aphid system.
Entomopathogenic nematodes are another diverse and abundant group of important natural enemies attacking a wide range of insect species. But to date, just one natural example of symbiont-based defence against nematodes has been reported (see also Kambris et al. 2009). In this instance, heritable Spiroplasma symbionts were found to protect a mushroom-feeding fly, Drosophila neotestacea, from the sterilizing effects of the nematode Howardula aoronymphium (Allantonematidae) (Jaenike et al. 2010). Although defensive mechanisms are not understood for this interaction, it is possible that Spiroplasma act as pathogens of the nematodes, as the size of reproductive nematodes is reduced in flies with Spiroplasma.
A single case of symbiont-mediated protection against predators has also been reported. In this instance, rove beetles (Staphylinidae) from the genus Paederus carry a symbiont related to Pseudomonas aeruginosa, which protects against spider predation (Kellner & Dettner 1996; Kellner 2002). Painful lesions called dermatitis linearis occur when Paederus haemolymph contacts human skin, and because of this, the mechanism of toxicity was characterized before its role in host protection had been identified. The causative agent is a polyketide amide called pederin produced by the Pseudomonas symbiont (Piel 2002). In laboratory assays, several species of wolf spiders (Lycosidae) and one jumping spider (Salticidae) would not eat eggs or larvae containing pederin, although they readily consumed pederin-free prey items (Kellner & Dettner 1996).
Unlike defence against other enemies, symbiont-based antipredator defences may be less common due to the specific conditions required for a selective advantage. For instance, predator sampling and rejection without fatality might be required for the benefits of symbionts to translate to increased reproductive output. Selective benefits could also arise should infection status be detectable through olfactory or visual (e.g. aposematism) cues; and indeed, symbionts have been implicated in modifying quantities of alarm pheromone (Oliver et al. 2012) and the external coloration of hosts. In this latter instance, Rickettsiella symbionts infecting A. pisum were shown to change aphid body colour from pink to green (Tsuchida et al. 2010). This could conceivably reduce predation by ladybird beetles, which have favoured pink aphid morphs in some studies (e.g. Losey et al. 1997).
There are several instances of symbiont-mediated defence against fungi, and pea aphids are again among the beneficiaries. R. insecticola was first found to protect against the prevalent and specific fungal pathogen Pandora neoaphidis (Ferrari et al. 2004; Scarborough, Ferrari & Godfray 2005) through unknown mechanisms. Excitingly, Spiroplasma, Rickettsia and Rickettsiella symbionts (but not H. defensa) have now also been shown to defend pea aphids against this same pathogen (Łukasik et al. 2013b). Even more recent work found that R. insecticola also protects against another aphid-specific fungal pathogen Zoophthora occidentalis, but not a generalist fungal pathogen of insects, Beauveria bassiana (Parker et al. 2013).
In another example, the European beewolf (Philanthius triangulum) is protected from fungi attacking brood cells by an antibiotic-producing symbiont, Streptomyces philanthi (Actinobacteria), which is cultivated within unique antennal glands and maternally transmitted to brood cells (Kaltenpoth et al. 2005, 2006; Kroiss et al. 2010) (See also Kaltenpoth & Engl: Defensive microbial symbionts in Hymenoptera, this volume). Antibiotic-producing bacteria are also harboured by other insects, including leafcutter ants and pine beetles, which employ these microbes to protect their fungal food sources (Currie et al. 1999; Scott et al. 2008; Haeder et al. 2009).
Other microbial pathogens
Although generally thought to invade host populations through reproductive manipulation (Werren, Baldo & Clark 2008), recent discoveries on the role of Wolbachia in defence against pathogens suggest invasion may also result from protective benefits. For instance, Wolbachia infection promotes resistance against RNA viruses within several Drosophila species (Hedges et al. 2008; Teixeira, Ferreira & Ashburner 2008; Osborne et al. 2009; Unckless & Jaenike 2012). Similarly, a naturally occurring Wolbachia strain was associated with reduced infection by West Nile virus in the mosquito Culex quinquefasciatus (Glaser & Meola 2010). Experimental introduction of Wolbachia into novel mosquito hosts also limits infections with Dengue and Chikungunya, along with apicomplexan Plasmodium parasites (Malaria) (Moreira et al. 2009). While Wolbachia genomes may encode antimicrobial genes responsible for resistance phenotypes (Rances et al. 2012), it has also been proposed that protection may reflect general ‘immune priming’, whereby Wolbachia pre-activates the host immune system by inducing the expression of immunity genes (i.e. defensins, cecropins and several Toll pathway genes) (Moreira et al. 2009; Bian et al. 2010; Kambris et al. 2010; Hughes et al. 2011; Pan et al. 2012). Interestingly, Wolbachia can also increase the susceptibility of the African armyworm, Spodoptera exempta, to a baculovirus (Graham et al. 2012), suggesting that symbiont-driven susceptibility should be more broadly investigated.
Gut symbionts, which may show some heritability, can also mediate host interactions with pathogens in a wide range of insects, just as they do in mammals [see (Dillon & Dillon 2004) for review]. However, most studies of gut microbes have focused on pathogen infection in insects with vs. without gut symbionts, a dichotomy that does not typically exist in nature. Hence, a recent finding on socially transmitted gut microbes of bumblebees is especially important, as it implicates natural variability in gut communities as the driver of defensive specificity against a protozoan parasite (Koch & Schmid-Hempel 2012).
Forces shaping the maintenance of defensive symbionts in natural populations
Heritable, defensive symbionts are typically found at intermediate frequencies within insect populations. While this may not be true for all hosts (e.g. Lim & Haygood 2004), infection polymorphisms have been seen in other groups, including plants, where endophyte presence/absence shapes feeding preference of herbivore natural enemies (Crawford, Land & Rudgers 2010). Such patterns raise questions about the range of factors that impact the spread and maintenance of protective, heritable symbionts (Fig. 1). Much of our discussion is centred on aphids, as knowledge of their defenders exceeds that for other groups. Balancing selection is proposed to play a large role in symbiont maintenance, with several environmental variables shaping symbionts' costs and benefits. Non-selective factors, including transmission rates, transmission modes, migration and drift, should also have smaller, yet important, impacts on the fate of defensive microbes.
Selective effects and mediating factors
Given the many instances of defensive symbionts across the insects and the importance of particular natural enemies in many of these systems, we expect that defensive benefits are the major force driving symbiont invasion within host populations (Fig. 1). Protective benefits, however, may vary with environmental conditions. In A. pisum, for example, laboratory studies show that the protection against parasitoids afforded by H. defensa is reduced at higher temperatures (Bensadia et al. 2006). While there are few reports of such ‘incomplete penetrance’ for other symbiotic defenders, it is likely that most function optimally under a limited range of conditions (see (Yule, Woolley & Rudgers 2011) for an example in plants), potentially limiting symbiont invasion into host populations.
Laboratory-based studies also indicate that defensive symbionts can impose costs in the absence of natural enemies, which could serve to limit or slow their spread in natural populations. For example, a population cage study on pea aphids found that despite increases in H. defensa frequencies in cages with wasps, symbiont levels dropped in enemy-free cages, suggesting a trade-off of infection (Oliver et al. 2008). Such costs coupled with seasonally variable enemy abundance (e.g. Hufbauer 2002) could promote the maintenance of infection polymorphisms seen for many heritable defenders.
Other context-dependent costs may exist for defensive heritable symbionts. For instance, costs may also be induced upon attack if symbiont abundances increase or defensive toxins concurrently harm host tissue (Kraaijeveld, Ferrari & Godfray 2002; Kwiatkowski & Vorburger 2012). Furthermore, the severity of harm caused by symbionts may be driven by factors unrelated to the presence or abundance of enemies, shifting across temperatures, host plants, under competition or due to the presence of symbiont superinfections (Chen, Montllor & Purcell 2000; Oliver et al. 2008; Simon et al. 2011).
While there is clear support for context dependence of symbiont costs, infection with protective microbes may also result in constitutive costs due to the services that hosts exchange for protection. The aphid symbiont H. defensa, for instance, appears auxotrophic for most essential amino acids and likely relies on the aphid and its nutritional symbiont Buchnera for growth (Degnan et al. 2009). Accordingly, subtle fitness costs have sometimes (Chen, Montllor & Purcell 2000; Vorburger & Gouskov 2011), but not always (e.g. Russell & Moran 2006), been identified under permissive conditions for H. defensa and other aphid symbionts.
There is a growing recognition that aphids and other insects can be infected with multiple facultative, heritable symbionts (superinfection). Pea aphids, for example, are estimated to harbour up to two facultative symbiont lineages per individual, on average (typical range = 0–4 per individual) (Ferrari et al. 2012; Russell et al. 2013). Superinfections have also been documented for other hosts and symbionts (e.g. Chiel et al. 2007; Toju & Fukatsu 2011), including Wolbachia, which frequently co-occur with other heritable bacteria (e.g. Zchori-Fein & Brown 2002; Russell et al. 2012). Experimental findings suggest that symbiont–symbiont and symbiont–host dynamics stemming from superinfection have the potential to be highly influential in nature.
The first insect-based study to address this for defensive symbionts found that A. pisum infected with both H. defensa and S. symbiotica received increased protection from parasitoids relative to infection with either symbiont alone (Oliver, Moran & Hunter 2006). However, superinfected aphids had drastically lower fitness in the absence of A. ervi wasps, potentially due to increased S. symbiotica densities. Field surveys accordingly found few instances of co-infection with these two particular lineages (Oliver, Moran & Hunter 2006). Another study in A. pisum showed that the presence of the little-studied gammaproteobacterial ‘X-type’ symbiont was associated with strong defence at high temperatures when co-infecting with H. defensa (Guay et al. 2009). While this could stem from the ‘rescue’ of H. defensa-driven protection, it is also possible that these co-infecting symbionts have additive, or even synergistic, effects on host protection. Given that defensive microbes of many plants and animals are members of multispecies communities (e.g. Guarner & Malagelada 2003; Kang et al. 2007), and the growing evidence for superinfections involving heritable, defensive symbionts, there is a need for studies on the antagonistic, additive or synergistic effects resulting from such co-habitations.
Multiple species of heritable, defensive symbionts can be found within some populations, contributing to the diversity of cytoplasmically inherited genetic elements and the potential of hosts to adapt to enemy pressure (Russell et al. 2013). Impressively, the pea aphid harbours six symbiont species with known or suspected defensive roles (Oliver et al. 2010; Tsuchida et al. 2010; Ferrari et al. 2012). At least three of these contain multiple strains that can coexist within single populations (e.g. Degnan & Moran 2008b; Ferrari et al. 2012; Russell et al. 2013). Just as different symbiont species can vary in their defensive roles, strains can also differ in their phenotypes, even varying in the types of natural enemies they protect against (Vorburger, Gehrer & Rodriguez 2010; Hansen, Vorburger & Moran 2012). Variable effects among strains also involve differences in the strength of protection and associated costs (Oliver, Moran & Hunter 2005; Oliver et al. 2008; Łukasik et al. 2013b), the effectiveness against particular enemy genotypes (Schmid et al. 2012) and the degrees of non-defensive benefits (Russell & Moran 2006).
The maintenance of numerous symbiont strains with similar roles, the presence of genetic variation for virulence in A. ervi parasitoids (Henter 1995), the capacity of these wasps to evolve partial resistance to H. defensa-mediated protection (Dion et al. 2011b), and the existence of strain–enemy specificity (Schmid et al. 2012), all hint at the potential for symbiont-driven co-evolution in the pea aphid system. We hypothesize, therefore, that enemy-driven, negative frequency-dependent selection favours the maintenance of diverse symbiont species or genotypes in aphid populations. A prediction of this hypothesis is that multiple genotypes with differential specificity and effectiveness will coexist and vary temporally within enemy populations (see (Kwiatkowski, Engelstadter & Vorburger 2012) for a theoretical treatment).
While strain variation has gone largely unexplored for other defensive symbionts, multiple strains of the same bacterial symbiont species have been found within several host insects (e.g. Kyei-Poku et al. 2005; Ros et al. 2012; Russell et al. 2012). However, the capacity for a single bacterial species to display a range of diverse phenotypes, including protective and non-protective effects (Scarborough, Ferrari & Godfray 2005; Hedges et al. 2008; Werren, Baldo & Clark 2008; Hansen, Vorburger & Moran 2012; Łukasik et al. 2013b), reveals the need for more work to know whether strain diversity commonly results in variable defence and specificity against distinct enemy genotypes. This will help to identify whether symbiont-driven co-evolution, fitting a ‘trench warfare’ model (Stahl et al. 1999), can explain the maintenance of diverse symbionts across multiple systems.
Since protective symbionts may impact more than defence, additional phenotypes and selective pressures are expected to shape natural symbiont frequencies. A single strain of H. defensa, for example, has been shown to confer both thermal tolerance and host protection (Oliver et al. 2003; Russell & Moran 2006), and warm temperatures in the absence of parasitism could select for aphids infected with this strain. Similarly, Wolbachia and Spiroplasma are lineages known to confer defence and to manipulate host reproduction. Some strains may even employ both tactics to invade host populations (Saridaki & Bourtzis 2010; Anbutsu & Fukatsu 2011; Unckless & Jaenike 2012).
Insects may also modify their behaviour to account for infection with protective symbionts. H. defensa-infected aphids exposed to parasitoids, for example, exhibited reduced aggressiveness and fewer escape behaviours than uninfected aphids (Dion et al. 2011a). Escape behaviours, such as dropping from the host plant, are inherently risky, and symbiont-protected aphids may gain additional benefits through reduced investment in warning signals and costly behaviours.
Yet another non-defensive effect of protective symbionts has been seen for two defensive symbionts, which alter aphid development. Specifically, S. symbiotica can inhibit the development of winged pea aphids (Chen, Montllor & Purcell 2000), while infection with R. insecticola may do the opposite (Leonardo & Mondor 2006). Should these phenotypes prove typical, then the resulting effects on host mobility may impact migration- and drift-mediated symbiont dynamics.
Non-selective forces are also expected to contribute to natural symbiont dynamics (Fig. 1). For instance, while the rates of vertical transmission for some heritable symbionts of aphids approach 100% in the laboratory (Chen & Purcell 1997; Darby & Douglas 2003; Weldon, Strand & Oliver 2013), imperfect transmission has been found for defensive Spiroplasma of Drosophila in the field, revealing a limit to symbiont spread (Jaenike et al. 2010). In pea aphids, the APSE phages required for H. defensa-conferred protection against parasitoids exhibit occasional vertical transmission failure (Oliver et al. 2009). Phage loss should reduce H. defensa spread as aphids infected with APSE-free H. defensa receive no protective benefits yet incur substantial fitness costs associated with higher bacterial abundances (Weldon, Strand & Oliver 2013).
Symbiont transmission failure also appears to occur more frequently in individual, superinfected aphids, with losses resulting in stable, single infections (Sandstrom et al. 2001; Moran & Dunbar 2006). Multiple infections are prevalent in nature, and inefficient transmission in multiply infected aphids could slow symbiont spread. Given that high and low temperatures can reduce symbiont transmission in other hosts (Hurst, Jiggins & Robinson 2001; Osaka et al. 2008; Burke et al. 2010; and see Welty, Azevedo & Cooper 1987 for an example in plants), imperfect levels of maternal transfer are plausible under field conditions for many symbionts.
While maternal transmission serves as their primary route of spread, phylogenetic evidence indicates that heritable symbionts occasionally move laterally within and among host species (e.g. Vavre et al. 1999; Sandstrom et al. 2001; Russell et al. 2003; Degnan & Moran 2008b). Effects on novel hosts can be profound as transfection experiments show that the acquisition of protective symbionts via horizontal transmission can instantly transform host defensive phenotypes (e.g. Oliver, Moran & Hunter 2005; Vorburger, Gehrer & Rodriguez 2010). Furthermore, horizontal transmission facilitates the formation of superinfections, providing avenues for recombination and possible exchange of defence genes and mobile elements between symbiont strains.
While laboratory-based studies suggest low rates of plant-mediated transfer for defensive symbionts of aphids (Chen & Purcell 1997; Darby & Douglas 2003; Oliver et al. 2008), in general, the natural rates and mechanisms of horizontal transfer are poorly understood. Recent findings suggest that parasitoids, phloem and sexual transfer are plausible routes for symbionts of hemipteran insects (Moran & Dunbar 2006; Caspi-Fluger et al. 2012; Gehrer & Vorburger 2012).
Drift and migration
Non-selective forces, such as genetic drift and host migration, are also expected to contribute to natural symbiont dynamics. These may exert stronger effects on ‘r-selected’ insects, given the crashes known from their natural populations (e.g. Hufbauer 2002). Drift may be especially important in the loss of rare symbionts, and the reported effects of symbionts on host dispersal (i.e. on the production of winged aphids) further suggest that tendencies for drift-mediated loss could vary among symbionts.
Effects beyond the host level
Many animals have evolved counterstrategies for coping with defences from the trophic level below. Herbivorous insects, for example, employ diverse behavioural and physiological strategies for overcoming chemical plant defences, including avoidance, detoxification, sequestration and rapid excretion (Gatehouse 2002). We should also expect diverse and complex natural enemy responses to symbiont-based protection. As mentioned before, A. ervi wasps were found to evolve increased virulence towards H. defensa-protected aphids over time (Dion et al. 2011b), and should natural enemies prove capable of discriminating between infected and uninfected hosts, responses to symbionts may involve behavioural tactics that unfold over shorter time-scales. Such recognition could involve the interception of volatiles produced or modified by symbionts, or the detection of symbiont-induced changes in internal or external contact cues used to assess host suitability (Boone et al. 2008; Hatano et al. 2008). Accordingly, female A. ervi wasps were shown to discriminate between H. defensa-infected and uninfected aphids, possibly due to lower levels of aphid alarm pheromone. Wasps selectively deposited more than one egg in infected hosts, which led to a significant increase in their ability to overcome symbiont-based protection (Oliver et al. 2012). Another study found a different response – in this case, Ephedrus plagiator wasps exhibited a slight reduction in the attack rates of H. defensa-infected grain aphids when given a choice between these and uninfected hosts (Łukasik et al. 2013b). While these documented effects trend in opposite directions, enemy behaviour should be considered as a factor potentially influencing the benefits and fates of protective symbionts.
Defensive symbionts in the field – correlates and dynamics
While the heritable, defensive symbionts of insects have been studied for only a short time, a rapidly growing body of laboratory research has identified mechanisms and mediating factors likely important in nature. Field studies are currently the key missing ingredient in the functional ecology of these microbes, as known defensive mechanisms (or at least the infections that drive defence) have rarely been shown to be relevant in the field (but see Jaenike et al. 2010). Below, we summarize what is known about defenders in natural populations, focusing on trends of infection prevalence and their implications for defence (Figs 2 and 3).
Real-world studies from insects
When looking across the insects, defensive symbionts have typically been found at intermediate levels within host populations (Kellner & Dettner 1996; Watts et al. 2009; Unckless & Jaenike 2012), often varying over space and time. For instance, polymorphic infections with defensive Spiroplasma were seen in populations of Drosophila neotestacea (Jaenike et al. 2010) (Fig. 2). Eastern populations exhibited infection frequencies of 60–80%, a potential equilibrium frequency that may be established by inefficient maternal transfer. Frequencies were notably lower within Western populations, with evidence suggesting a recent arrival of beneficial Spiroplasma and an ongoing selective sweep driven by nematode parasites. While strong correlations between nematode parasitism and symbiont prevalence have not been reported in this system, field-caught flies with Spiroplasma can avoid castration by nematodes, implicating these symbionts as important natural defenders.
Potential evidence for natural defence can also be derived from correlations between symbiont prevalence and parasitism in the field – an approach transferrable to other systems. In one case, frequencies of an Arsenophonus symbiont were positively correlated with rates of parasitism by a specialist encyrtid parasitoid across populations of the red gum lerp psyllid (Fig. 2) (Hansen et al. 2007). In this case, experimental studies have not assessed whether this symbiont defends its psyllid hosts, and this correlation could result from either (i) parasitism pressure selecting for protected Arsenophonus-infected hosts or (ii) heightened wasp success in Arsenophonus-bearing hosts. As with other systems, field-based correlations will be most comprehensible when coupled with controlled laboratory experiments.
Real-world studies of pea aphid symbioses (prior findings)
The heritable symbionts of A. pisum are among the best studied in the field, and world-wide surveys employing molecular diagnostics have reported substantial variation both within and among host populations. In general, the six known or suspected defensive symbionts of A. pisum are found at intermediate levels, ranging from uncommon (i.e. 10% or lower) to prevalent (>80%) under different contexts (e.g. Ferrari et al. 2012; Russell et al. 2013). APSE bacteriophage are found at high to nearly ubiquitous levels in H. defensa-bearing aphids (Russell et al. 2013), a trend consistent with parasitoid susceptibility and host detriments for APSE-free, H. defensa-bearing pea aphids in the laboratory (Oliver & Moran 2009; Weldon, Strand & Oliver 2013).
Several correlates of defensive symbiont frequencies have been found for the pea aphid, with the best documented among these being host plant. Within the past 16 000 years, pea aphids have diversified into a large number of genetically distinct and specialized host races, which thrive on a variety of human-cultivated legumes (Peccoud et al. 2009). These host races vary in their symbiotic complements (Fig. 2). A survey of symbionts among several food plant species at numerous localities in Japan, for instance, discovered that R. insecticola was enriched in pea aphids from white clover (Trifolium repens) compared to those from vetch (Vicia sativa) (Tsuchida et al. 2002). Several other studies have also reported an association between clover use and R. insecticola (Leonardo & Muiru 2003; Simon et al. 2003; Ferrari et al. 2004, 2012; Bilodeau et al. 2013; Russell et al. 2013), while Hamiltonella and X-type show enrichment in other host races (Ferrari et al. 2012). Interestingly, some, but not all, pea aphid clones exhibit improved performance on clover when infected with R. insecticola (Leonardo 2004; Tsuchida, Koga & Fukatsu 2004; Ferrari, Scarborough & Godfray 2007), indicating complex interactions between this symbiont, aphid genotype and host plant utilization. Given the role of this symbiont in defence, it remains unclear whether pathogen-imposed selective pressures are greater on clover or whether other factors, such as symbiont-mediated interactions with the food plant or historical associations, play the largest role in the enrichment of R. insecticola in clover-using aphids.
Symbiont surveys across pea aphids have also provided evidence for variation across geographical scales. R. insecticola, for example, was found to be more common in northern regions of Japan, which have cooler climates with greater annual precipitation (Tsuchida et al. 2002) (Fig. 2). Pandora pathogens thrive under cool and humid conditions (Yu et al. 1995), suggesting that R. insecticola prevalence could result from spatially varying selection imposed by this natural enemy. A more recent study also found variation in R. insecticola, S. symbiotica and Rickettsiella frequencies across alfalfa-feeding populations from Pennsylvania, Wisconsin and Utah (Russell et al. 2013). While variable selective pressures by pathogens or temperature could conceivably play a role, the driving forces here remain unclear.
Support for temporal shifts in defensive symbiont frequencies – and the factors that drive them – was first obtained by Montllor, Maxmen & Purcell (2002), who recognized that elevated temperatures may select for pea aphids harbouring thermotolerance-conferring S. symbiotica symbionts (Chen, Montllor & Purcell 2000; Russell & Moran 2006). The authors sampled pea aphids from alfalfa fields at different times and locations within California (USA), finding substantial increases (15–49%) in S. symbiotica frequencies from cooler to warmer months. In another study, field-based population cage experiments revealed not only selection for clones with S. symbiotica under higher temperatures, but also a trade-off in which heat-sensitive lines showed higher growth rates at lower, permissive temperatures (Harmon, Moran & Ives 2009).
Real-world studies of pea aphid symbioses (new findings)
While rates of parasitism and pathogen attack may vary across locales, prior findings of their seasonal variation (e.g. Hufbauer 2002) suggest a need to measure the relationship between natural enemy prevalence and defensive symbiont frequency over time. We are currently conducting such a study and to date have generated a data set on the frequencies of H. defensa over a single season across two New York State pea aphid populations – one from alfalfa and one from red clover (Data S1). PCR screening of aphids collected at 3-week intervals, from May through October, showed significant fluctuations in H. defensa frequency on alfalfa (Fig. 3b; Supplemental Results Tables S2 and S3). Notably, the percentage of pea aphids with this symbiont changed from 57%, to 40%, to 61%, to 48% across a 9-week span from June to August. While the magnitude of these shifts (from one time point to the next) may not differ much from those seen in the laboratory-performed population cage experiments (Oliver et al. 2008; see Fig. 3a), the lack of unidirectional change (observed in laboratory cages) over the entire field season suggests costs and benefits may vary temporally. It will be of interest to identify the changing environmental factors that influence these dynamics and whether transmission-driven factors play a role as well.
In contrast to the observations on alfalfa, H. defensa levels were always <25% in pea aphids from red clover (Fig. 3b), with no significant evidence for temporal shifts (Supplemental Results Tables S2 and S3). The lack of parallels between sympatric alfalfa and clover populations, in spite of prior findings suggesting similar wasp densities and seasonal trends across these populations (Hufbauer 2002), suggests unidentified constraints limiting symbiont spread. These could be driven either by differing ecologies across alfalfa and clover fields or by genotypic differences between pea aphid host races or symbionts.
While knowledge of heritable, defensive symbionts continues to expand, future studies in the laboratory, field and in silico are needed to enhance and balance our understanding of these novel immune systems. Of notable importance are the capacity for symbionts to drive rapid, contemporary evolution and antagonistic co-evolution. Also important will be elucidation of the symbiont–symbiont and symbiont–host effects resulting from superinfections, which provide simple models for more complex microbial communities that often also contain defenders. Below, we suggest future research that will advance our understanding of the functional ecology of defensive symbiosis in nature.
Effects on extended communities
The effects of defensive symbionts likely extend beyond their insect hosts and enemies to natural communities. Defensive symbionts may, for example, limit resources available to higher-order natural enemies (e.g. hyperparasitoids) or lower trophic levels (e.g. host plants, mushrooms) through influences on their hosts' abundance. Such community-level consequences have been reported for symbionts protecting grasses from herbivorous insects, with effects cascading across trophic levels (Rudgers & Clay 2007). Among insects, the Drosophila neotestacea–Spiroplasma symbiosis exhibits several features that suggest the capacity for community-wide effects (Jaenike & Brekke 2011), including multiple fly species competing within mushrooms, the limitation of symbiont-mediated defence to just one of these species and the ability of Howardula nematodes to attack and influence competition among these fly hosts (Jaenike & Perlman 2002).
Characterizing the mechanisms of defence
Mechanisms of defence can be difficult to decode for uncultivable symbionts. In cases where defensive symbiont products are externalized, direct determination of mechanism has been possible (e.g. Frank & Kanamitsu 1987; Kroiss et al. 2010). In other systems, insight into mechanism has been obtained through inferences from genomic studies, which have revealed pathways for diverse putative pathogenicity factors and effector molecules (e.g. Degnan et al. 2009; Hansen, Vorburger & Moran 2012). Of future importance will be transcriptome studies, especially those focusing on expression following enemy challenge.
The increasing affordability of high-throughput sequencing will make these goals more feasible, providing insights into mechanism across many more systems. This will allow us to determine how often the repeated origins of defence across microbial lineages have resulted from lateral gene transfer of protective pathways versus repeated de novo invention. The answer has implications for medicine, given the range of bioactive compounds produced by defensive microbes (e.g. Poulsen et al. 2011). Understanding of the mechanism (along with symbiont–enemy specificity) will also be important in explaining why some host species can harbour multiple strains or species of symbionts defending against the same natural enemy (Oliver, Moran & Hunter 2005; Kwiatkowski, Engelstadter & Vorburger 2012; Łukasik et al. 2013b). Such redundancy would not be expected to persist if the mechanisms and specificities of symbionts were equivalent.
Characterizing symbiont genotypes
To date, nearly all symbiont surveys across insects have used diagnostic PCR to assess the presence of particular bacterial species. However, the growing number of symbionts discovered even in the best-studied systems (e.g. Guay et al. 2009; Tsuchida et al. 2010) indicates that broader screening techniques are needed to permit the detection of novel bacteria. T-RFLP, DGGE and next generation sequencing (e.g. amplicon pyrosequencing or Illumina HiSeq/MiSeq) on PCR products amplified with universal 16S rRNA primers comprise promising possibilities (e.g. Haynes et al. 2003; Jones et al. 2011; Caporaso et al. 2012; Russell et al. 2013). Since even these 16S-based techniques may not distinguish among closely related strains, we would also advocate the analysis of more rapidly evolving genes (e.g. Riegler et al. 2005; Degnan & Moran 2008b). In short, the diversity of heritable symbionts inhabiting insect hosts and the known differences in defensive phenotypes among species and strains argue for the more routine use of both broader and higher-resolution diagnostics.
Additional experimental assays
In concert with improved genotyping and understanding of mechanism, we also need experimental studies that characterize the functional differences between symbiont types and strains. Even among the well-studied pea aphid symbionts, the full range of defensive properties has not been characterized for most symbionts. Further, while strains of H. defensa and Spiroplasma vary in their levels of conferred defence (Oliver, Moran & Hunter 2005; Degnan & Moran 2008b), little is known about the diversity of symbiont strains infecting even a single population of pea aphids, much less the entire species.
Population cage experiments should enhance our understanding of the relative costs and benefits of different symbiont types and strains, helping to mimic natural conditions where multiple aphid genotypes with differing symbiont complements compete side by side. Also, cage experiments examining multiple symbiont varieties in the same genetic background across varying environmental conditions (e.g. enemy abundance, temperature) will increase our understanding of costs and benefits under realistic conditions.
Modelling insect, symbiont and natural enemy dynamics
General theoretical models predict that heritable symbionts can spread by conferring protection against horizontally transmitted pathogens and parasites (e.g. Lively et al. 2005; reviewed in Haine 2008). Four recent studies have provided a theoretical treatment on the spread and maintenance of defensive heritable symbionts within insect populations. In the first, inefficient maternal transmission and moderate levels of nematode parasitism were sufficient to account for intermediate Spiroplasma frequencies in D. neotestacea, with the model recovering an equilibrium frequency resembling natural infection levels (Jaenike et al. 2010). Modelling of heritable Wolbachia recently showed that defence can greatly facilitate spread throughout host populations (Fenton et al. 2011). Specifically, cytoplasmic incompatibility inducing strains conferring high levels of protection could invade host populations even when exhibiting low transmission rates or inducing high physiological costs. The abilities of these microbes to spread were increased with heightened enemy virulence and transmissibility. A third study modelled aphid–parasitoid interactions and found that intermediate levels of parasitism, induced costs after parasitism and nearly perfect vertical transmission were important in favouring the maintenance of intermediate defensive symbiont frequencies (Kwiatkowski & Vorburger 2012). Finally, a recent study showed that cycling of defensive symbiont and enemy genotypes could be commonly recovered with greater specificity between these antagonistic partners (Kwiatkowski, Engelstadter & Vorburger 2012). Future models should look to prior literature from other defensive symbiont systems (e.g. Gundel et al. 2008) while considering additional complexities identified in laboratory studies (e.g. Harmon, Moran & Ives 2009), including the mediating effects of temperature on defence and the effects of superinfections.
Additional field-based approaches, including studies that measure relationships between symbiont prevalence and natural enemy densities or attack rates (Hansen et al. 2007), could elucidate whether selection pressures from enemies are driving symbiont persistence, while uncovering the range of environmental conditions that alter these effects. Manipulative field cage experiments exploring responses of symbiont frequencies to introduced natural enemies could produce similar insights. Studies based in the same locations across generations and seasons may add to the growing body of work showing that evolution can happen over ecological time-scales (Carroll et al. 2007). Indeed, beneficial heritable symbionts are predicted to spread faster than beneficial nuclear alleles (Jaenike 2012), suggesting that contemporary evolution of symbiont-mediated defence could be a common theme within insect populations.
Population cage studies also hold promise in the exploration of symbiont-mediated co-evolution between insects and their natural enemies. The pea aphid–A. ervi system meets several conditions required for co-evolution, yet missing are findings of trait matching in virulence and defence (e.g. Brodie, Ridenhour & Brodie 2002).
When we consider the diversity of microbial lineages conferring protection, the prevalence of heritable symbionts across the insects and the importance of natural enemies in shaping their life histories, it is likely that defensive mutualisms are not only widespread, but also important across the insect world. Given the recent interest in these phenomena, we expect many more examples to come to light, with defensive services acting as a prominent route for the invasion and maintenance of heritable bacteria within insect populations. To fully comprehend the significance of symbiont-based defence, we may need to expand our views on insect immunity (Parker et al. 2011), while further considering the roles of symbionts in community structure, host–parasite dynamics, co-evolution and the rapid, contemporary evolution of their insect hosts (White 2011). Findings from other animal groups and plants suggest micro-organisms play a major role in host immunity across the eukaryotes (Clay 1988; White & Torres 2009). Far from being a curiosity, the protective effects of microbes are instead likely a key feature of life on earth.
We thank Piotr Łukasik and the anonymous reviewers for their many helpful comments on the manuscript. We also thank Rachael DiSciullo, Steven Doll, Mickey Drott, Amanda Lee, Tyler Maruca, Garrett Mayo and Nick Tuttle for assistance with aphid collections and symbiont screening. This work was funded by NSF grant #1050128 to KMO and #1050098 to JAR. The authors declare no conflict of interest.