Insects and other arthropods are frequently infected with heritable bacterial symbionts [1]. Theory predicts that these strictly inherited microbes must confer net benefits or manipulate host reproduction to invade and persist in host populations [2–4]. While documented fitness benefits of symbionts have been primarily nutritional, a number of recent studies have reported protective effects of symbiont infection, suggesting that defense against natural enemies may be another major route facilitating invasion of heritable symbionts into host populations [5–7]. Insect symbionts of diverse bacterial lineages have been shown to provide substantial protection against fungal pathogens [8, 9], viruses [10, 11], predators [12, 13], parasitoids [14–17] and parasitic nematodes [18]. Despite increasing awareness of the roles of these bacteria in protecting hosts, little is known about natural enemy responses to this line of defense. Just as herbivorous insects have evolved strategies to overcome plant chemical defenses, natural enemies may employ strategies to counter symbiont-based defense.

The pea aphid, Acyrthosiphon pisum, has become a model for characterizing the effects of infection with symbionts [19]. In addition to carrying the obligate nutritional symbiont, Buchnera aphidicola, this aphid is often infected with additional facultative bacteria, called 'secondary' symbionts, which have been found to mediate important ecological interactions. These effects include thermal tolerance [20, 21], resistance to fungal pathogens [9] and parasitoid wasps [14]. In the latter case, A. pisum acquires partial to complete immunity to parasitism by the common parasitoid wasp Aphidius ervi via infection with the γ-proteobacterial symbiont Hamiltonella defensa [14, 22]. In no-choice assays, wasps readily parasitized H. defensa-infected aphids, but failed to complete development in resistant aphids [14]. While the details of the protective mechanism of H. defensa remain unclear, a bacteriophage called APSE (Acyrthosiphon pisum secondary endosymbiont) is required to produce the protective phenotype, suggesting a key role for eukaryotic toxins carried by the phage [23, 24]. Among examined lines, A. pisum lacking H. defensa (or those carrying H. defensa that lack APSE) are highly susceptible to parasitism by A. ervi, while lines carrying APSE-infected H. defensa receive partial to complete protection depending on H. defensa strain and associated phage haplotype [22–24].

Given the low probability of survival for wasp eggs laid in aphids infected with H. defensa and APSE, wasps should benefit from strategies that allow them to avoid or overcome symbiont-mediated defense. One potential strategy involves the use of dose-dependent factors associated with the number of oviposition events in a particular host [25]. Endoparasitoid wasps such as A. ervi typically employ both maternal and embryonic factors to counter host defenses and create an environment suitable for wasp development [26]. During oviposition, for example, female A. ervi wasps inject a γ-glutamyl transpeptidase-containing venom, which targets and degenerates pea aphid ovarioles, leading to a reduction in aphid fecundity and presumably enhancing resource availability for developing wasps [27, 28]. Also, fast-growing polyploid cells called teratocytes dissociate from the serosal membrane surrounding the wasp embryo and circulate within the aphid hemocoel during larval development [29]. In A. ervi, teratocytes are known to synthesize and release two proteins that are likely involved in redirecting host resources from aphids to developing wasps: a fatty acid binding protein (Ae-FABP), and an extracellular enolase (Ae-ENO), which may degrade aphid tissue [30, 31]. If an A. ervi female oviposited twice in a single host, she would then double the quantity of both venom and teratocytes present in the host, potentially increasing the likelihood of successful parasitism in symbiont-defended hosts. Aphidius ervi is a solitary endoparasitoid; regardless of the numbers of eggs deposited, only one parasitoid will emerge from an aphid host. Nonetheless, in a host with more than one egg, the surviving wasp may gain an advantage in the host-parasitoid conflict from additional complements of venom and teratocytes.

In this study we examined potential wasp responses to symbiont-mediated defense in A. pisum. We conducted experiments to determine if superparasitism, defined here as the deposition of more than one egg, increases the likelihood of successful parasitism in symbiont-defended aphids. We also conducted a choice test to determine whether female A. ervi wasps' oviposition behavior differs when presented with both H. defensa infected and uninfected A. pisum. Lastly, we analyzed the volatile signature of H. defensa-infected and uninfected aphids as wasps could potentially exploit volatile differences occurring between symbiont infected and uninfected hosts to modify oviposition behavior when faced with symbiont-protected hosts. In bark beetles, for example, natural enemies use symbiont-derived volatiles to locate hosts [32, 33].

Recent studies indicate that many insects are protected by heritable symbionts [7], yet little is known about how natural enemies respond to symbiont-based defenses. We report here that the parasitoid wasp, A. ervi, can partially overcome H. defensa-mediated protection in A. pisum by superparasitizing aphids (Figure 1, Table 2). Further, we find in choice assays that A. ervi can discriminate between H. defensa-infected and uninfected A. pisum and selectively superparasitizes infected aphids, while laying single eggs in uninfected hosts (Figure 2). Taken together, these results indicate that A. ervi has likely evolved successful strategies to partially counter symbiont-based defenses employed by A. pisum. Counter-strategies to overcome defense on the trophic level below are well known. Herbivorous insects have devised strategies for countering plant-based defenses (for example, [39–41]), and predators and parasitoids may overcome herbivore behavioral [42, 43] and chemical defenses as well. This report provides evidence that natural enemies are sometimes able to counteract symbiont-based defense. Recent reports have documented: 1) the evolution of increased A. ervi virulence when faced with symbiont-carrying resistant A. pisum lines [44] and 2) genotypic variation in the parasitoid Lysiphlebus fabarum's ability to successfully parasitize H. defensa-infected Aphis fabae (black bean aphid) [16]. The experimental design of these studies, however, does not allow conclusive partitioning of increased virulence towards symbiont- versus host-based defensive factors, but these reports do indicate that parasitic wasps likely have additional mechanisms for overcoming symbiont-based resistance.

Consistent with previous reports, we found that aphids infected with H. defensa, when singly parasitized, received moderate to high levels of protection from attack by A. ervi, depending upon strain [22]. We also found that the same strain of H. defensa (82B) in two backgrounds (A2E and 5A) conferred similar levels of protection, indicating little symbiont X aphid interaction [22]. When doubly-parasitized, however, we found that the likelihood of successful parasitism of H. defensa-infected aphids increased in every instance, in both pink (5A) and green (A2E) aphid backgrounds and among all three H. defensa strains (82B, A1A and A2F). Strains A1A and A2F carry the YD-repeat toxin-encoding phage variant ASPE-3 and strain 82B carries the cdtB-encoding APSE-2 [23]. Not only does superparasitism increase successful parasitism in symbiont defended hosts, but wasps gain a relatively greater benefit of superparasitism when attacking H. defensa infected aphids compared to uninfected aphids. Thus, superparasitism appears to be an effective counter strategy to both phage variants hitherto detected in N. American populations in A. pisum. One caveat regards the possibility that wasps emerging from successfully superparasitized aphids are not fit enough to reproduce themselves. While we did not explicitly examine wasp fitness, we did verify that adult wasps emerge from superparasitized hosts and bear no obvious abnormalities. A previous study in this system found slight increases in dry weight, and no differences in developmental time, in wasps emerging from superparasitized aphids relative to singly parasitized A. pisum [45], indicating that superparasitism does not generally lead to reductions in wasp fitness. In future experiments we will examine costs associated with superparasitism as a strategy to overcome defensive symbionts.

The mechanisms underlying the increase in successful parasitism of superparasitized aphids are not known, but the additional maternal and embryonic factors found in superparasitized hosts may contribute. In aphids infected with 82B→5A, dead or moribund larvae appear four to five days post parasitism - long after teratocytes have dissociated from the extra-embryonic membrane [46]. We confirmed that an additional complement of teratocytes was present in doubly-parasitized aphids. We also found reductions in the number and irregularities in the morphology of teratocytes deposited in H. defensa infected aphids suggesting a possible interaction. Polyploid teratocytes are believed to be involved in redirecting resources from aphid to wasp development [28, 29]. Fewer, or defective, teratocytes may prevent the creation of an environment suitable for wasp development. The extra complement of teratocytes in superparasitized aphids may shift the balance in favor of successful parasitoid development.

Some authors have suggested that the likelihood of one solitary parasitoid surviving host encapsulation responses may increase when multiple eggs are deposited in a single host -the so-called 'multiple target hypothesis' (for example, [47, 48], rev. in [25]). While there are some accounts that host encapsulation responses are not as effective when faced with multiple parasitism events (for example, [49–51]), Vinson [52] reviewed the evidence for the multiple target hypothesis and concluded that target proliferation is probably not generally a viable strategy for endoparasitoids in overcoming host encapsulation defenses. While the aphid encapsulation response is very weak in our system, the notion of multiple parasitoid targets, especially teratocytes, diluting the effectiveness of symbiont-based defenses, may resurrect this hypothesis for cases when protective symbionts are present.

We found that not only does superparasitism increase the odds of successful parasitism, but that A. ervi females appear to discriminate between H. defensa-infected aphids and uninfected aphids (Figure 2) and preferentially superparasitize symbiont-defended aphids. This indicates that wasps likely employ superparasitism as a strategy to overcome symbiont-based defense. In choice tests, females were more likely to lay single eggs in uninfected aphids, but more than one egg in H. defensa infected aphids. Since only one wasp can complete development in a single aphid, self-superparasitism (more than one egg laid by the same mother) by solitary endoparasitoids like A. ervi has generally been considered maladaptive, yet wide-ranging conditions have been proposed under which this strategy may be adaptive [51]. If the net fitness of offspring is higher under conditions of superparasitism compared to single-parasitism then selection may favor the sacrifice of an additional egg for a parasitoid such as A. ervi, which is limited by their time to find hosts rather than the number of eggs they have [25, 53]. Aphidius ervi has a large number (96 to 567) of small, yolk-free eggs during adulthood (for example, [54, 55]). While we did not compare the net fitness of wasps when singly parasitizing versus superparasitizing hosts, we did find large increases in the numbers of mummies (32 to 200%) produced by superparasitism in each of the H. defensa-infected lines. This sharp increase in successful parasitism combined with the relatively low expected cost of A. ervi eggs suggests that superparasitism may produce net benefits. Thus, superparasitism may be an adaptive strategy when faced with symbiont-defended hosts.

We found that levels of trans-β-farnesene (EBF) were lower in H. defensa-infected aphids (Figure 3), and differences in the abundance of this volatile compound are a possible basis for wasp discrimination between infected and uninfected hosts. EBF is the major component of aphid alarm pheromone and is secreted from the cornicles (siphunculi). EBF in aphids has been found to serve as a short-range (due to high reactivity with ozone) attractant to a wide range of natural enemies, including both predators and parasitoids (rev. [38]). Cornicle secretions, although not specifically EBF, from A. pisum have been reported to be a contact kairomone for A. ervi [56]. Aphids release alarm pheromone when confronted with natural enemies, which often leads to escape behaviors, such as dropping from the host plant [36, 37], and inducing the production of winged morphs [57]. In groups of symbiont-defended aphids, the release of lower amounts of EBF could be adaptive, as dropping from the plant, or dispersing is risky [58] while staying put may be a better strategy if one is likely to survive a parasitoid attack. A recent report found that A. pisum infected with H. defensa exhibited fewer escape behaviors and reduced aggressiveness relative to uninfected counterparts [59], consistent with reduced levels of EBF. More generally, heritable variation in escape behavior among A. pisum clones has been reported [60]; and heritable symbiont influences on EBF titers could explain some of this variation.

In addition to documenting parasitoid behavioral responses to symbiont-based defense in herbivores, it is also remarkable that aphid secondary symbionts appear to be detectable by foraging parasitoids. Oviposition by Aphidius ervi, and aphidiine parasitoids in general, is extraordinary in its speed; the parasitoid faces the aphid, coils its abdomen underneath its head and thorax so that it too faces forward, and appears to strike the aphid with the ovipositor at the tip of its abdomen, generally laying an egg in that instant. Aphidiine wasps are not generally known to lay two eggs with a single strike [45, 61], so superparasitism involves a second strike and oviposition. Detection of contact or even internal compounds likely involve sensilla on the ovipositor [62] presumably after contact chemicals have been detected and evaluated. That detection of the host's condition can occur during this very short interval has been shown by several authors [61, 63, 64]; female A. ervi were shown to avoid parasitism of aphids already containing a parasitoid egg. We show that the levels of EBF are generally lower in symbiont-defended relative to uninfected A. pisum. Whether EBF is the compound used to discriminate between symbiont-defended and uninfected aphids, however, is not certain. Infection with H. defensa and other symbionts has been shown to alter the pool of metabolites in A. pisum [65], and these changes may also serve as host selection cues. It would be interesting, however, if symbiont-defended aphids use reduced EBF to increase the threshold for dropping behavior, but parasitoids have co-opted the signal for their own benefit.

Recent reports indicate that a wide range of insects are protected from natural enemies via infection with heritable symbionts, yet little is known about natural enemy responses to symbiont-based protection [7]. Here we report that the parasitic wasp A. ervi discriminates between symbiont-defended and undefended (that is, uninfected) aphids, and modifies its oviposition behavior when faced with protective symbionts in ways that increase the likelihood of successful parasitism. Wasps selectively deposit two or more eggs in symbiont-defended hosts, even though only one wasp can complete development within a single aphid. The discovery of wasp behavioral responses to symbiont-based defense in aphids sets the stage for further investigations of the coevolutionary dynamics of the host A. pisum and its parasitoid A. ervi, as mediated by a bacterial symbiosis. In general, we expect that the study of natural enemies that are confronted with defensive symbionts will yield insights into various counter-strategies to circumvent or overcome the protection these symbionts provide.