Variation in Pseudonocardia antibiotic defence helps govern parasite-induced morbidity in Acromyrmex leaf-cutting ants

Summary

Host–parasite associations are potentially shaped by evolutionary reciprocal selection dynamics, in which parasites evolve to overcome host defences and hosts are selected to counteract these through the evolution of new defences. This is expected to result in variation in parasite-defence interactions, and the evolution of resistant parasites causing increased virulence. Fungus-growing ants maintain antibiotic-producing Pseudonocardia (Actinobacteria) that aid in protection against specialized parasites of the ants’ fungal gardens, and current evidence indicates that both symbionts have been associated with the ants for millions of years. Here we examine the extent of variation in the defensive capabilities of the ant–actinobacterial association against Escovopsis (parasite-defence interactions), and evaluate how variation impacts colonies of fungus-growing ants. We focus on five species of Acromyrmex leaf-cutting ants, crossing 12 strains of Pseudonocardia with 12 strains of Escovopsis in a Petri plate bioassay experiment, and subsequently conduct subcolony infection experiments using resistant and non-resistant parasite strains. Diversity in parasite-defence interactions, including pairings where the parasites are resistant, suggests that chemical variation in the antibiotics produced by different actinobacterial strains are responsible for the observed variation in parasite susceptibility. By evaluating the role this variation plays during infection, we show that infection of ant subcolonies with resistant parasite strains results in significantly higher parasite-induced morbidity with respect to garden biomass loss. Our findings thus further establish the role of Pseudonocardia-derived antibiotics in helping defend the ants’ fungus garden from the parasite Escovopsis, and provide evidence that small molecules can play important roles as antibiotics in a natural system.

Introduction

The biology of all living organisms, including humans and human-domesticated plants and animals, is influenced by parasites (Jaenike, 1978; Price et al., 1986; Ewald, 1994; Cohen, 2000; Palumbi, 2001). The constant selective pressures imposed by parasites on hosts have made them a major force influencing evolution and species diversification (Jaenike, 1978; Ewald, 1994; Cohen, 2000; Palumbi, 2001). On an evolutionary time scale, natural host–parasite associations are shaped, at least in part, by arms race-like dynamics: parasites evolve to overcome host defences and hosts are selected to counteract these through the evolution of new defences (Van Valen, 1973; Hamilton, 1980). In humans, the dynamics of host– parasite interactions are modified through the application of bacteria-derived (Clardy et al., 2005) antimicrobial chemicals. However, the efficacy of these compounds in human medicine and agriculture is continuously threatened by the emergence of resistant parasite phenotypes, which arise as a consequence of selective pressures imposed by the applied antimicrobials (Cohen, 2000; Palumbi, 2001; Davies, 2007). Recent work has revealed that some insects also employ bacteria-derived antibiotics for protection against parasites (fungus-growing ants: Currie et al., 1999a; Oh et al., 2009; European beewolves: Kaltenpoth et al., 2005; and Southern Pine beetles: Scott et al., 2008). Theory predicts that the target parasites of the bacteria-derived antibiotic compounds should evolve resistance, but the presence and impact of such resistant parasites have not been evaluated in any of these natural systems. Here, we determine the potential for the evolution of parasite resistance by determining whether: (i) there is variation in parasite resistance, and (ii) this variation leads to differential parasite success. Fungus-growing ants in the tribe Attini (Hymenoptera: Formicidae) engage in an obligate mutualism with fungi they cultivate for food (Agaricales: Lepiotaceae and Pterulaceae) (Chapela et al., 1994; Munkacsi et al., 2004; Schultz and Brady, 2008). This beneficial ant–fungus symbiosis is parasitized by microfungi in the genus Escovopsis that attack and consume the ants’ fungal garden (Hypocreales: Ascomycota) (Currie et al., 1999a; 2003b; Reynolds and Currie, 2004). The cultivated fungi can suppress the growth of Escovopsis strains not known to infect them in nature, but cannot inhibit isolates of naturally occurring parasites (Gerardo et al., 2006a,b; Gerardo and Caldera, 2007; Supporting information; Fig. S1). Instead, to defend their fungus gardens from Escovopsis, the ants rely on behavioural removal of Escovopsis spores and mycelium (Currie and Stuart, 2001), secretions from worker’s metapleural glands (Bot et al., 2002; Fernández-Marín et al., 2009), and mutualistic antibiotic-producing filamentous bacteria in the genus Pseudonocardia (Actinomycetales: Pseudonocardiaceae; Currie et al., 1999b; Cafaro and Currie, 2005) (Fig. 1A). The relative contribution of these different defence mechanisms likely depends on the genus of fungus-growing ant. In the ant genus Acromyrmex, artificially removing Pseudonocardia from the ant cuticle results in significantly larger negative effects of Escovopsis, indicating that Pseudonocardia is an important defence against these parasites (Currie et al., 2003a). Actinobacteria in other genera have also been isolated from ant fungus gardens (Mueller et al., 2008; Haeder et al., 2009), and in culture some strains have been shown to produce small molecules that inhibit the growth of Escovopsis (Haeder et al., 2009). However, it remains to be established if these bacteria are more than transient community members (i.e. being isolated because they are present as non-metabolizing inocula introduced to the garden with the leaf material added by workers), or instead are capable of growth and production of these small molecules within the garden matrix (see Results and discussion). Individual Acromyrmex ant colonies appear to associate with a single Pseudonocardia strain (Poulsen et al., 2005), which is persistently present within nests (at least in laboratory colonies) for extended periods of time (years) (••. Marsh, M. Poulsen and C.R. Currie, unpubl. data), but there is genetic diversity in this bacterial symbiont both within and between ant species (Cafaro and Currie, 2005; Poulsen et al., 2005; Mikheyev et al., 2008). The between-colony diversity in ant-associated Pseudonocardia leads to the prediction that different Pseudonocardia strains have different inhibitory capabilities against different Escovopsis strains (Currie et al., 2006).

Fig. 1.

A. Leaf-cutting ant worker with Pseudonocardia-covered cuticle. Micrograph of an Acromyrmex octospinosus major worker illustrating the abundance of mutualistic bacteria on the ant cuticle. B. Acromyrmex-associated Pseudonocardia–Escovopsis bioassay. Results of a symbiont pairing bioassays between 12 strains of Pseudonocardia (vertical) and Escovopsis (horizontal) associated with five species of Acromyrmex leaf-cutting ants. These included three Argentinean species: A. niger (colony code CC030327-02), A. hispidus fallax (SP030327-01 and UGM030327-02) and A. laticeps (UGM030330-04), and two Panamanian species: A. echinatior (Ae291, Ae292, Ae295 and CC031212-01) and A. octospinosus (CC011010-04, CC031210-22, ST040116-01, UGM020518-05). Pure culture Pseudonocardia were point inoculated in the centre of plates containing PDA without antibiotics and were grown for 3 weeks until reaching a diameter of c. 1–1.5 cm, after which Escovopsis was inoculated at the edge of the plate. Plates were examined bi-weekly and when a clear zone of inhibition (ZOI) had formed in a given pairing (typically within 2–3 weeks after Escovopsis inoculation), the ZOI between symbionts (Cafaro and Currie, 2005) was recorded. Each box represents the average ZOI (n = 3) of a given pairing and different colours indicate the degree of inhibition: white: ZOI = 0 cm, light yellow: ZOI = 0.01–0.50 cm, yellow: ZOI = 0.51–1.0 cm, orange: ZOI = 1.01–1.5 cm, and red: ZOI > 1.51 cm. Bioassay pairings enclosed in boxes represent the pairings from which subcolonies were created and inoculated with Escovopsis (Fig. 1C). C. Subcolony evaluation of Escovopsis virulence. The in vivo virulence of Escovopsis measured as fungus garden biomass loss when miniature colonies, created using a standard set-up (Bot et al., 2001; Poulsen and Boomsma, 2005) with 80 mg of fungus garden and two major Acromyrmex workers covered with the Pseudonocardia bacterium (abundance scores 9–12 in Poulsen et al., 2003), were exposed to Escovopsis infection. Subcolonies were housed in 60 ml plastic cups with small holes in the lids to provide air, moist tissue in the bottom to assure high humidity, and a folded oak leaf to provide forage for the ants. To assure that Escovopsis spore inoculations did not differ between subcolonies, we quantified spore concentrations in aqueous suspensions containing filter-sterilized water and 0.05 μl Tween 20 per ml water prior to application, and subsequently inoculated individual subcolonies with an amount of suspension corresponding to 11 500 ± 252 spores (mean ± SE; n = 17). Negative control subcolonies were applied the same volume of filter-sterilized water and Tween 20 without Escovopsis spores. Fungus garden weight was measured 1 week after Escovopsis application. Fungus garden loss is averaged across three strength levels of in vitro Escovopsis inhibition by Pseudonocardia (strong: ZOI > 1.0 cm, n = 48; weak: ZOI = 0.01–1.0 cm, n = 42; or no: ZOI = 0 cm, n = 36; Fig. 1B).

In order to build on our knowledge of the role of Pseudonocardia in the association with fungus-growing ants, we examine how differences in inhibitory capabilities between different Pseudonocardia strains affect Escovopsis-induced morbidity. We do this by linking observed Pseudonocardia–Escovopsis interactions in vitro to in vivo defensive capabilities during subcolony infection. Capitalizing on the ability to culture the microbial symbionts associated with Acromyrmex leaf-cutting ants, we examine interactions between the antibiotic-producing bacteria and the target parasites in a bioassay experiment, crossing strains associated with 12 colonies from five species of Acromyrmex leaf-cutting ants (Fig. 1A and B). Based on the variation in Escovopsis–Pseudonocardia (parasite-defence) interactions in vitro, we test the prediction that the inhibitory capabilities of Pseudonocardia-derived antibiotics correlate with parasite-induced morbidity in fungus-growing ant nests in vivo. We do this by performing an infection experiment pairing colonies with Escovopsis strains susceptible or resistant to the antibiotics produced by the resident bacterium.

Results and discussion

Our symbiont pairings, evaluating the variation in parasite-defence interactions in a Petri plate bioassay experiment of Pseudonocardia and Escovopsis from 12 colonies distributed across five Acromyrmex species, revealed abundant variation in suppression of Escovopsis (Fig. 1B). The degree of Escovopsis inhibition observed depended on the specific combination of strains of Pseudonocardia and Escovopsis paired. The reason for this variation is likely a combination of (i) the susceptibility of Escovopsis, (ii) compound differences between Pseudonocardia strains, (iii) compound production rate differences between Pseudonocardia strains, and (iv) potential differences between bacterial strains in their composition of compounds when more than a single compound is produced. This is supported by our findings of a statistically strong effect of the interaction between Pseudonocardia inhibitory capabilities and the susceptibility of Escovopsis (likelihood ratio statistics for a Type III two-way ANOVA: χ² = 202.5, d.f. = 104, P < 0.0001) (Table 1). The analysis further identified significant strain effects of both Escovopsis (likelihood ratio statistics for a Type III nested ANOVA: χ² = 14.50, d.f. = 7, P = 0.0429) and Pseudonocardia (χ² = 114.9, d.f. = 7, P < 0.0001) (Table 1). The variation observed included pairings in which the parasite Escovopsis exhibited significant resistance to the Pseudonocardia-derived antibiotics. Escovopsis growth was completely suppressed by the antibiotic-producing Actinobacteria in 54.2% of bioassay pairings, but Escovopsis was not suppressed, indicating antibiotic resistance, in 11.8% of the bioassay pairings (Fig. 1B). Despite abundant variation in interactions within and between ant species, there were notable effects of the ant species origin of both symbionts (Pseudonocardia χ² = 41.41, d.f. = 4, P < 0.0001; Escovopsis: χ² = 21.21, d.f. = 4, P = 0.0003; Table 1). Our analysis also indicated a possible dose effect (χ² = 9.59, d.f. = 1, P = 0.002), but this correlation explained only 1.5% of the variation (r = 0.12; Fig. S2). The presence of resistance in the parasite and variation in bacteria-derived chemistry suggests that interactions on this finer phylogenetic scale shape host–parasite dynamics within this system. In fact, 45.3% (29 out of 64) of pairings between symbionts isolated from the two sympatric Acromyrmex species (A. echinatior and A. octospinosus) showed weak or no inhibition [zone of inhibition (ZOI) < 1.0 cm] in the bioassay, indicating that resistant Escovopsis phenotypes do infect leaf-cutting ant nests in nature.

Table 1.

Results of the statistical evaluation of the within-Acromyrmex Pseudonocardia–Escovopsis bioassay.

	χ 2	d.f.	P
Nested model
Pseudonocardia strain (ant species)	114.9	7	< 0.0001
Escovopsis strain (ant species)	14.50	7	0.0429
Ant species of Pseudonocardia	41.41	4	< 0.0001
Ant species of Escovopsis	21.21	4	0.0003
Ant species of Pseudonocardia * ant   species of Pseudonocardia	28.14	15	0.0207
Size of Pseudonocardia	9.590	1	0.002
Two-way ANOVA
Pseudonocardia strain	233.8	11	< 0.0001
Escovopsis strain	77.83	11	< 0.0001
Pseudonocardia strain * Escovopsis   strain	202.5	104	< 0.0001
Size of Pseudonocardia	1.930	1	0.1616

Due to the non-normality of data and the high sensitivity of imputing, zero-values in log-linear models, we employed a generalized linear model in which zero-values were replaced with the value, 0.00001 and significance of terms in the models was determined through likelihood ratio statistics for Type III ANOVAs. The analysis was performed in SAS v. 9.1.3 (SAS Institute, 1994). Two models were built: the first model contained the respective ant species from which Pseudonocardia and Escovopsis strains originated from, Pseudonocardia and Escovopsis strains nested within their ant species origins, the interaction term between the ant species symbionts originated from, and the size of Pseudonocardia at the time inhibition of Escovopsis was scored (top section). Since this nested design did not allow for an evaluation of the interaction between Pseudonocardia and Escovopsis strains, we performed a separate analysis including only Pseudonocardia and Escovopsis strains and their interaction (bottom section).

In order to examine whether the patterns of inhibition of Escovopsis observed reflect general antifungal properties of the Pseudonocardia strains tested, we performed a supporting experiment testing the antifungal effects of the 12 strains against a strain of Candida albicans (strain 1002, see Supporting information for more details). We used the same experimental Petri plate bioassay set-up as in the Pseudonocardia–Escovopsis pairings, except that C. albicans was inoculated in a liquid suspension, and found that none of the 12 Pseudonocardia strains displayed visible inhibition of C. albicans. Although we cannot rule out that inhibition of C. albicans could occur at higher concentrations of Pseudonocardia-derived secondary metabolites, this set-up allows for a direct comparison with the Pseudonocardia–Escovopsis assay. This suggests that the variation in Escovopsis inhibition between Pseudonocardia strains is not a result of differences in the levels of general antifungal activity between different bacterial strains (for details, see Supporting information).

Based on the Petri plate bioassay results, we predicted that the extent of parasite-induced garden biomass loss would depend upon the combination of Pseudonocardia and Escovopsis parasite strain. To test this prediction, we performed a subcolony experiment pairing six different ant colonies with three different Escovopsis strains, choosing pairings that included the range of antagonistic displays of Pseudonocardia symbionts towards Escovopsis in vitro (Fig. 1B). Escovopsis impact was evaluated as fungus garden mass loss 1 week after Escovopsis application. Escovopsis-induced garden biomass loss indeed depended on the combination of the resident Pseudonocardia associated with the host ant colony and the strain of Escovopsis (Fig. 1C). The in vivo evaluation revealed that Escovopsis rapidly overgrew all cultivar fragments in the absence of ants (one-way ANOVA: F_7,752 = 11 240, P < 0.0001; Table 2). In the presence of ants, the impact of Escovopsis was strongly correlated with the inhibition, or lack thereof, displayed in the in vitro bioassay by the resident Pseudonocardia bacterium, indicating that reactions in vitro represent the effectiveness of bacterial defence in vivo (F_2,123 = 9.66, P < 0.0001) (Fig. 1C; Table 2). Of particular interest were situations where Escovopsis strains were only weakly inhibited by the bacterium in vitro (ZOI < 1.0 cm; 25.7% of the pairings), as these combinations, on average, showed virulence levels in vivo that were comparable to pairings with strongly resistant Escovopsis strains (Fig. 1C) (F_1,83 = 0.2819, P = 0.5969). A two-way ANOVA confirmed the isolated effects of ant colony (F_5.135 = 5.74, P < 0.0001) and Escovopsis treatment (F_6,135 = 23.53, P < 0.0001), and further showed a marginally non-significant interaction between the two (F_15.135 = 1.64, P = 0.071) (Table 2). The results of this experiment indicate that the inhibitory capabilities of the antimicrobials produced by the resident Pseudonocardia bacterium directly influence the impact of Escovopsis during colony infections. Thus, strains of the garden parasite exhibiting in vitro resistance to Pseudonocardia secondary metabolites cause greater morbidity to the ants’ fungus garden in vivo, confirming that the symbiont bioassay pairings reflect interactions within colonies.

Table 2.

Statistical analyses of subcolony evaluation of Escovopsis virulence in Acromyrmex.

	d.f.	F	P
Two-way ANOVA
Colony origin	5,135	5.740	< 0.0001
Escovopsis strain	6,135	23.53	< 0.0001
Colony origin * Escovopsis strain	15,135	1.640	0.071
One-way ANOVAs
Absence of ants	11,240	7.752	< 0.0001
No, weak or strong actinobacterial   inhibition in vitro	2,123	9.660	< 0.0001
No or weak actinobacterial inhibition   in vitro	1,83	0.2819	0.5969
Escovopsis strain	5,120	2.532	0.0323
Ant colony	5,120	6.331	< 0.0001

The top section of the table shows the results of a two-way ANOVA evaluating the overall effects of Escovopsis on subcolonies of Acromyrmex ants in vivo. The bottom section represents four one-way ANOVAs testing the isolated effects of (i) ant presence in subcolonies, (ii) whether the Pseudonocardia symbiont associated with the ant colony exhibited strong (ZOI > 1.0 cm), weak (ZOI = 0–0.9 cm) or no (ZOI = 0 cm) in vitro inhibition of the Escovopsis strain, (iii) whether there was a difference in Escovopsis virulence between subcolonies harbouring Pseudonocardia strains exhibiting weak (0 < ZOI < 0.9 cm) or no (ZOI = 0) in vitro inhibition of the Escovopsis strain, (iv) individual Escovopsis strains, and (v) ant colony.

Further, our infection experiment indicates that even strains of Escovopsis exhibiting some, but not strong, resistance in vitro (i.e. growth suppression < 1.0 cm ZOI) cause higher levels of fungus garden morbidity in vivo. This indicates that more than a third of the parasite-defence pairings screened in the bioassay exhibit a degree of resistance to the antibiotics produced by the bacteria that would result in increased garden morbidity. Importantly, resistance was present in Escovopsis– Pseudonocardia combinations found in nature, and even occurred, although only rarely, in pairings between symbionts from the same individual ant colony (Fig. 1B). While it may appear counterintuitive to find surviving colonies inhabited by resistant parasite strains, other defences in the ants (e.g. grooming and weeding; Currie and Stuart, 2001) likely compensate for this lack of efficient Actinobacteria-defence and thus help prevent complete colony collapse in such rare cases. Despite the presence of resistant parasite strains, a large portion of pairings in the bioassay experiments exhibited strong suppression of the parasite (62.5%) (Fig. 1B), and infection with parasite strains susceptible to the Actinobacteria exhibited less garden biomass loss (Fig. 1C). This implies that the costs incurred by the ants to maintain Pseudonocardia, through specialized exocrine glands that secrete the substrate for Pseudonocardia growth (Poulsen et al., 2003; Currie et al., 2006), are outweighed by the benefits obtained by the ability of Pseudonocardia to suppress Escovopsis during colony infections.

Naturally occurring resistant parasite strains, like those identified in this study, should impose a selective pressure on the ant host for the evolution of novel defences. The ant–Pseudonocardia association appears to maintain efficient defence against Escovopsis despite the evolution of antibiotic resistance. The mechanisms behind the successful application of Actinobacteria-derived antibiotics by fungus-growing ants are unclear, but several scenarios can be envisioned. The first conceivable scenario is that parasite-defence dynamics in the fungus-growing ant symbiosis are shaped by antagonistic co-evolution, involving the evolution of resistance in the parasite, counteracted by the evolution of novel compounds by Pseudonocardia. Specifically, this posits that the selective pressure imposed by Escovopsis on Pseudonocardia results in the Pseudonocardia-derived antibiotics evolving in parallel with the target parasites. In turn, the antibiotic secretions by Pseudonocardia targeted at Escovopsis strains are expected to impose a selective pressure on Escovopsis to evolve resistance (Cohen, 2000; Palumbi, 2001; Davies, 2007). The presence of such antagonistic co-evolution would result in variability in the chemistry of Pseudonocardia secretions, as our bioassay pairings suggest (Fig. 1B). The relatively short generation time of bacteria may allow for a fast rate of change, but our understanding of changes in the metabolic capacities of the bacterium residing in individual nests is, as of yet, limited. Such change either may be restricted to include only minor modifications of existing secondary metabolites, and not major and elaborate alterations, or may involve the possibility of acquiring new antibiotic-coding gene clusters through horizontal gene transfers (Ginolhac et al., 2005; Jenke-Kodama et al., 2005; Schmitt and Lumbsch, 2009). An alternative mechanism for rapid modifications is the acquisition of a novel Pseudonocardia strains, with novel secondary metabolites, by an ant colony. Co-phylogenetic patterns between Pseudonocardia and fungus-growing ants suggest relatively frequent switching of the Pseudonocardia symbiont between species, and even genera (Poulsen et al., 2005; Mikheyev et al., 2008), as well as several acquisitions of Pseudonocardia from free-living (non-fungus-growing ant-associated) counterparts over the evolutionary history of the association (Cafaro and Currie, 2005). Both evolution of novel Pseudonocardia antibiotics and acquisition of novel antibiotic-producing bacteria from the environment likely play a role in mediating host–pathogen dynamics in this system.

It is generally recognized that just because one microbe inhibits another in Petri plate challenges, this may imply, but does not prove, that the compound produced causes inhibition in nature (Waksman, 1961; Davies, 2006; Clardy et al., 2009). In fact, it has been argued that what we consider antibiotics do not function as such in nature, but instead are signalling compounds (Davies, 2006; Yim et al., 2007). Recently, it has been shown that Actinobacteria genera other than Pseudonocardia (mostly Streptomyces) have the potential to inhibit Escovopsis (Mueller et al., 2008; Haeder et al., 2009). However, these studies are only supported by bioassays, not infection experiments, and thus implies, but does not show, that these Actinobacteria could play a similar role in the fungus-growing ant symbiosis. In contrast, the role of Pseudonocardia-derived antibiotics for dealing with Escovopsis infections is now supported through both removal and infection experiments (Currie et al., 2003a) and through the evaluation of the role inhibitory capabilities play in governing parasite-induced morbidity (this study). For the two main species of Acromyrmex focused on in this study, A. octospinosus and A. echinatior, it is well established that Pseudonocardia is vertically transmitted between generation by incipient queens (Currie et al., 1999a), and that specific Pseudonocardia strains are associated with these ant species (Cafaro and Currie, 2005; Poulsen et al., 2005). In addition, the same Pseudonocardia strain can be isolated from the exoskeleton of workers in individual Acromyrmex colonies maintained in the lab for several years, indicating consistent and persistent associations with Pseudonocardia (••. Marsh, M. Poulsen and C.R. Currie, unpubl. data). Haeder and colleagues (2009) also isolated Pseudonocardia as the main bacterium from workers of A. octospinosus and A. echinatior; they only found Streptomyces in the fungus garden matrix. Although these Streptomyces strains potentially inhibit Escovopsis in Petri plates, it remains to be determined if they represent more than transient microbes that do not actively metabolize within the ant symbiosis (see Caldera et al., 2009).

Given the metabolic capacity of Actinobacteria to produce antibiotic (e.g. Demain, 1999), in addition to the relatively short generation time of bacteria allowing for faster rates of change, it is perhaps not surprising that the defensive mutualist in the association is an Actinobacterium. This suggests that the ant–Pseudonocardia mutualism is an example of symbiosis as evolutionary innovation: the ants benefit from the mutualism by gaining access to a source of rapidly changing metabolites with potent antimicrobial properties. The development of novel antibiotics by Pseudonocardia strains may evolve in response to Escovopsis parasites, thereby counteracting the evolution of parasite resistance. The availability of potent antibiotic producers to modify their secondary metabolism and/or the ants’ ability to acquire new sources of antibiotics from other ant-associated or free-living bacteria may have allowed for a more efficient defence against Escovopsis than would have been achieved by the ants alone. Long-term antibiotic use in ants may only have remained stable and efficient over time due to their ability to access microbes from the environment that harbour diverse antimicrobial properties. Whether or not novel chemistries are obtained from the resident ant-associated bacterium, or via acquisition forming novel associations, our findings illustrate the benefits of this type of defensive mutualism in the ability to counteract resistance evolution in the target parasite.