Population‐specific effect of Wolbachia on the cost of fungal infection in spider mites

Abstract Many studies have revealed the ability of the endosymbiotic bacterium Wolbachia to protect its arthropod hosts against diverse pathogens. However, as Wolbachia may also increase the susceptibility of its host to infection, predicting the outcome of a particular Wolbachia‐host–pathogen interaction remains elusive. Yet, understanding such interactions and their eco‐evolutionary consequences is crucial for disease and pest control strategies. Moreover, how natural Wolbachia infections affect artificially introduced pathogens for biocontrol has never been studied. Tetranychus urticae spider mites are herbivorous crop pests, causing severe damage on numerous economically important crops. Due to the rapid evolution of pesticide resistance, biological control strategies using entomopathogenic fungi are being developed. However, although spider mites are infected with various Wolbachia strains worldwide, whether this endosymbiont protects them from fungi is as yet unknown. Here, we compared the survival of two populations, treated with antibiotics or naturally harboring different Wolbachia strains, after exposure to the fungal biocontrol agents Metarhizium brunneum and Beauveria bassiana. To control for potential effects of the bacterial community of spider mites, we also compared the susceptibility of two populations naturally uninfected by Wolbachia, treated with antibiotics or not. In one population, Wolbachia‐infected mites had a better survival than uninfected ones in absence of fungi but not in their presence, whereas in the other population Wolbachia increased the mortality induced by B. bassiana. In one naturally Wolbachia‐uninfected population, the antibiotic treatment increased the susceptibility of spider mites to M. brunneum, but it had no effect in the other treatments. These results suggest that natural Wolbachia infections may not hamper and may even improve the success of biological control using entomopathogenic fungi. However, they also draw caution on the generalization of such effects, given the complexity of within‐host–pathogens interaction and the potential eco‐evolutionary consequences of the use of biocontrol agents for Wolbachia‐host associations.


| INTRODUC TI ON
The maternally inherited bacterium Wolbachia is to date the best studied and probably the most common endosymbiont of arthropods. It is estimated to infect up to 52% of arthropod species (Weinert, Araujo-Jnr, Ahmed, & Welch, 2015), a success mainly attributed to its ability to induce various types of reproductive manipulation in hosts to increase the reproductive success of infected females, thereby increasing its own transmission (Werren, Baldo, & Clark, 2008). In particular, the ability of Wolbachia to spread rapidly within and among host populations (Engelstadter & Hurst, 2009) has raised growing interests in using it in biocontrol programs (Bourtzis et al., 2014).
Possible Wolbachia-based biocontrol strategies include the use of Wolbachia as a microbial biocontrol agent, for instance to enhance productivity of natural predators and parasites such as parasitoids (e.g., Grenier et al., 1998;Stouthamer, 1993); as a potential gene-drive vehicle for population replacement strategies through cytoplasmic drive (which provides a mechanism for the autonomous spread of desired genes into targeted populations; e.g., Dobson, 2003;Sinkins & Godfray, 2004;Turelli & Hoffmann, 1999); or for sterile insect techniques (SIT) to suppress target pest populations by repeated sweeps with infected individuals (Calvitti, Marini, Desiderio, Puggioli, & Moretti, 2015;Zhang, Lees, Xi, Gilles, & Bourtzis, 2015;Zhong & Li, 2014). Subsequently, the discovery of the ability of Wolbachia to protect its hosts against a wide array of pathogens, including viruses, protozoan parasites, fungi, or pathogenic bacteria (reviewed by Cook & McGraw, 2010) has provided new avenues for the control of vector-borne diseases (reviewed by Iturbe-Ormaetxe, Walker, & Neill, 2011). For instance, deliberate introductions of Wolbachia into Aedes aegypti mosquito populations are currently being undertaken successfully in several regions worldwide to control dengue virus (e.g. Hoffmann et al., 2014;Nguyen et al., 2015). However, such ability of Wolbachia to interfere with diverse host pathogens may have undesirable effects on biocontrol strategies if, for instance, natural Wolbachia infection interferes with parasitic biocontrol agents, a possibility that has never been addressed. Alternatively, natural Wolbachia infections in several host species may also increase host susceptibility to parasite infection (e.g., Graham, Grzywacz, Mushobozi, & Wilson, 2012;Hughes, Rivero, & Rasgon, 2014), raising the possibility that Wolbachia could also facilitate the action of biocontrol agents. Moreover, potentially variable effects of Wolbachia on host susceptibility to biocontrol agents may have ecological and epidemiological consequences. For instance, artificially introduced pathogens can select Wolbachia variants that increase host resistance and counter-selection variants that increase host susceptibility to infection, thereby potentially driving the spread of defensive Wolbachia variants (e.g., Cattel, Martinez, Jiggins, Mouton, & Gibert, 2016;Jaenike, Unckless, Cockburn, Boelio, & Perlman, 2010;Kriesner & Hoffmann, 2018). Hence, assessing the effect of natural Wolbachia infection on the efficiency of different strains and/or species of parasitic biocontrol agents is a prerequisite for the development of efficient and long-lasting control strategies (Zindel, Gottlieb, & Aebi, 2011).
Spider mites of the genus Tetranychus (Acari: Tetranychidae) are ubiquitous major crop pests of c.a. 1,100 plant species belonging to more than 140 different plant families (Migeon & Dorkeld, 2006. Due to their short generation time and high fecundity, spider mites rapidly develop resistance to pesticides (Van Leeuwen, Vontas, Tsagkarakou, Dermauw, & Tirry, 2010), which has encouraged the development of alternative control strategies such as the use of essential oils or natural enemies (e.g., predators, entomopathogenic bacteria and fungi; Attia et al., 2013).
To examine the effect of the interaction between Wolbachia and fungal infection on spider mite survival, we carried out a fully factorial experiment using two naturally Wolbachia-infected and two naturally Wolbachia-uninfected spider mite populations belonging to two genetically differentiated forms of T. urticae (Auger, Migeon, Ueckermann, Tiedt, & Navajas Navarro, 2013) and treated K E Y W O R D S antibiotic treatment, bacterial community, facilitation, fungi-induced mortality, symbiontmediated protection, Tetranychus urticae or not with antibiotics. We used a strain of two generalist entomopathogenic fungi species, Beauveria bassiana and Metarhizium brunneum, as they are included in genuses that are among the most used fungi in commercial production (Vega et al., 2009), with wide geographical and host ranges (Gurlek, Sevim, Sezgin, & Sevim, 2018;Meyling & Eilenberg, 2007;Roberts & Leger, 2004). The specific aims of this work were to determine: (a) whether infection with a natural Wolbachia strain protects spider mites against fungus-induced mortality, (b) whether this effect varies with different Wolbachia strains and/or the presence of other bacteria in spider mites, and (c) whether this effect depends on the fungus strain. We then discuss possible mechanisms leading to our results, the importance of considering the whole bacterial community of arthropods when assessing the effect of Wolbachia, as well as the potential eco-evolutionary consequences of the presence of Wolbachia for the success of spider mite control strategies using entomopathogenic fungi.

| Spider mite populations and rearing
Four populations were used in this study, two belonging to the "red" form (AlRo and AMP), and two belonging to the "green" form (DEF and TOM) of Tetranychus urticae (Auger et al., 2013). These populations have been collected in the Iberian Peninsula from 2010 to 2017, on different plant species. Upon collection from the field, the populations AMP and TOM were found to be naturally and fully infected by two different strains of Wolbachia. The population AMP is infected by the Wolbachia strain ST481 (isolate "Turt_B_wUrtAmp," id: 1858 in the PubMLST Wolbachia database; http:// www.pubml st.org/wolba chia/), which is very similar to strain ST219 belonging to supergroup B and found in China by Zhang, Ding, Zhang, and Hong (2013); and the population TOM is infected by the Wolbachia strain ST280 (isolate "Turt_B_wUrt-Tom," id: 1857), which has also been previously found in China by Zhang, Ding, et al. (2013). These strains are very closely related, having 1 SNP difference on the sequences of both the fbpA and coxA genes in the multilocus sequence typing (MLST) system developed by Baldo et al. (2006) for Wolbachia. The two other populations, AlRo and DEF, were naturally uninfected by Wolbachia and none of the populations used in this study carried other maternally inherited bacterial endosymbionts (i.e., Cardinium, Rickettsia, Spiroplasma, and Arsenophonus) at the time of the experiment, as confirmed by PCR using the methods described in Zélé, Santos, et al. (2018). All the information concerning these populations is summarized in Table 1. After collection, these populations were reared in the laboratory under standard conditions (24 ± 2°C, 60% RH, 16/8h L/D) at high numbers (c.a. 500-1000 females per population) in insect-proof cages containing bean plants (Phaseolus vulgaris, cv. Contender seedlings obtained from Germisem, Oliveira do Hospital, Portugal).

| Antibiotic treatments
Roughly 2 months (ca. 4 generations) before the onset of the experiment, a rifampicin solution (0.05%, w/v) was used to treat mites (n = 70 adult females initially) from each population for one generation (see Gotoh et al., 2005). This allowed us to obtain Wolbachiauninfected AMP and TOM populations as well as controls for the antibiotic treatment for the naturally uninfected populations AlRo and DEF. During the treatment, mites were maintained in Petri dishes containing bean leaf fragments placed on cotton with the antibiotic solution. After one generation, 100 adult-mated daughters from each treated population were transferred in insect-proof cages containing bean plants, in the same laboratory conditions as the untreated populations, and these new populations were allowed to grow for 3 successive generations in the absence of antibiotics to avoid potential side effects of the treatment (e.g., Ballard & Melvin, 2007). One generation before the onset of the experiment, pools of 100 females were taken from each treated population and checked by PCR to confirm that they were uninfected by Wolbachia (detailed procedure in . This method allows detecting Wolbachia infection even at low frequencies (up to 1/100; ).

| Entomopathogenic fungi strains and preparation of inoculum
We used the strains V275 (= Met52, F52, BIPESCO 5) of Metarhizium brunneum and UPH-1103 of Beauveria bassiana (obtained from Swansea University; UK, and from Siedlce University; Poland, respectively), as they were previously shown to have the potential to suppress T. urticae populations (Dogan et al., 2017). The procedures used for fungal growth, inoculum preparation, and spider mite infection are similar to that described in Dogan et al. (2017). Briefly, the two fungi were grown on Sabouraud Dextrose Agar (SDA) medium at 25°C for 2 weeks. Conidia were harvested from sporulating cultures with the aid of a spatula, washed with sterile distilled water and filtered through 4 layers of gauze (pore size: 20 µm) to remove any hyphae.

| Spider mite infection and survival
The experiment was conducted in a growth chamber under standard conditions (25 ± 2°C, 80% RH, 16/8 hr L/D). Roughly 2 weeks prior to the experiment, 100 females were collected from each mass culture and allowed to lay eggs during 4 days on detached bean leaves placed on water-soaked cotton. One day prior to the onset of the experiment, 20 young adult mated females (hence with similar age) were randomly collected from these cohorts and placed on a 9 cm 2 bean leaf disc on wet cotton with the abaxial surface facing upwards. On the first day of the experiment, the surface of the leaf discs was sprayed using a hand sprayer with 2.5 ml of a spore suspension of M. brunneum or B. bassiana in 0.03% (v/v) aqueous Tween 20 at 1 × 10 7 conidia/ml (which is the most commonly used concentration in laboratory studies; Dogan et al., 2017), or, as a control, with 0.03% aqueous Tween 20 only. Subsequently, female survival was monitored every 24 hr during 10 days by counting both dead and alive individuals. A total of twelve replicates per treatment (fungi infection and antibiotic treatment) and per population were performed within 2 experimental blocks of one day difference (6 replicates of each treatment per block).

| Statistical analysis
Analyses were carried out using the R statistical package (version 3.5.3). The general procedure for building the statistical models was as follows. Spider-mite populations (AlRo, AMP, DEF, and TOM), antibiotic treatment (treated with rifampicin or not), and infection treatment (sprayed with BB: Beauveria bassiana, with MB: Metarhizium brunneum, or with Tween 20 only as control) were fitted in as fixed explanatory variables, whereas discs nested within population and block were fitted as random explanatory variables. When a significant three-way interaction between the three fixed variables was found, each population was analyzed separately with the same model structure, except that the variable population was removed from the model. Survival data were analyzed using Cox proportional hazards mixed-effect models (coxme, kinship package). Hazard ratios (HR) were obtained from these models as an estimate of the difference between the rates of dying (Crawley, 2007) between the untreated controls and the BB or MB treatments for each population. Because the timing of infection is an important parameter for the fitness of parasites, an additional early measurement of survival, the proportion of dead mites at 3 days postinfection (dpi), was obtained from Kaplan-Maier estimates of the survival distribution for each disc.
This timing was chosen as it is close to the median survival upon infection in most of the populations tested, and hence corresponds to a threshold time-point to unravel differences between treatments.
The numbers of dead and alive mites at 3 dpi were computed using the function cbind and analyzed with a mixed model glmmadmb procedure (glmmADMB package) with a negative binomial error distribution to correct for overdispersed errors (family "nbinom1" with a Øµ variance).
Maximal models, including all higher-order interactions, were simplified by sequentially eliminating non-significant terms and interactions to establish a minimal model (Crawley, 2007). The significance of the explanatory variables was established using chi-squared tests (Bolker, 2008). The significant chi-squared values given in the text are for the minimal model, whereas nonsignificant values correspond to those obtained before deletion of the variable from the minimal model.
To explore significant interactions between infection and antibiotic treatment effects on female survival and mortality at 3 dpi, the two factors were concatenated to fit a single fixed factor containing all treatment levels in the models (i.e., 6 levels for infection by antibiotic treatment effects within each population). Multiple comparisons between levels were then performed from these models using General Linear Hypotheses (glht, package multicomp) with Bonferroni corrections, which uses classical Chisq (Wald test) for testing the global hypothesis H 0 .

| RE SULTS
Overall, depending on whether they were naturally infected or uninfected by Wolbachia, the survival of females from different populations was not evenly affected by fungal infection and by rifampicin treatment (fungal infection x rifampicin treatment x population interaction: X 2 6 = 36.16, p < .0001 and X 2 6 = 18.33, p = .005 on the overall survival of spider mites and on their mortality at 3 days postinfection, respectively). Thus, to understand this three-way interaction, we looked at the effect of fungal infection and of the antibiotic treatment in each population separately. TA B L E 1 Populations of spider mites used in the experiment. Mites were collected in Portugal (P) and Spain (S) and were naturally infected, or not, by Wolbachia. The absence of other maternally inherited endosymbionts (Cardinium, Rickettsia, Spiroplasma, Arsenophonus) in these populations was confirmed by PCR before the onset of the experiment (using methods described in Zélé, Santos, et al., 2018;  urticae in China . b Isolate "Turt_B_wUrtAmp"-id: 1858, Wolbachia strain ST481. This is a new strain, very similar to the strain ST219 (they differ by 1 SNP on the fbpA gene: allele 444 instead of allele 4) that was found in China by Zhang, Ding, et al. (2013).

| Effect of fungal infection and of antibiotic treatment in the naturally Wolbachia-uninfected population AlRo
In the population AlRo, the two fungal strains affected differently the survival of spider mites depending on whether they were treated with antibiotics or not (fungal infection x rifampicin treatment interaction: X 2 2 = 9.53, p = .009; Figure 1a). Indeed, M. brunneum (MB) induced a stronger mortality in rifampicin-treated mites than in untreated mites (z = 2.80, p = .05), while B. bassiana (BB) induced the same mortality in both rifampicin-treated and untreated mites (z = −1.20, p = 1.00 Figure 1b; Table 2a). In both cases, however, M. brunneum induced a stronger mortality than B. bassiana (MB vs. BB: z = 6.72, p < .0001 and z = 2.81, p = .05 in rifampicin-treated and untreated mites, respectively). At 3 days postinfection (dpi), however, no significant interaction between fungal infection and rifampicin treatment (X 2 2 = 1.78, p = .41; Figure 1c) and no effect of the antibiotics treatment alone (X 2 1 = 0.90, p = .34) on female mortality were found. Both fungi strains severely increased the mortality of both rifampicin-treated and rifampicin-untreated mites at this early age of infection (X 2 2 = 135.68, p < .00014; see Table 3a for all multiple comparisons).

| Effect of fungal infection and of antibiotic treatment in the naturally Wolbachia-uninfected population DEF
In the population DEF, we did not find a significant interaction between fungal infection and rifampicin treatment (X 2 2 = 0.65, p = .72; Figure 1d), neither a significant effect of rifampicin treatment (X 2 1 = 0.003, p = .96), but only a significant effect of fungal infection (X 2 2 = 879.17, p < .0001). Indeed, both fungi induced the same mortality in rifampicin-treated and in rifampicin-untreated mites, with an overall stronger effect of M. brunneum than of B. bassiana (Figure 1e; Table 2b for all multiple comparisons).

| Effect of fungal infection and of antibiotic treatment in the naturally Wolbachia-infected population AMP
In the population AMP, we found a significant interaction between infection and rifampicin treatment (X 2 2 = 26.61, p < .0001; Figure 1g). This interaction was due to a lower survival of Wolbachia-infected controls compared with rifampicin-treated controls (z = −4.92, p < .0001) only, as Wolbachia-infected and rifampicin-treated mites had the same overall survival upon infection with each fungal strain (for B. bassiana: z = −0.26, p = 1.00; for M. brunneum: z = 1.88, p = .55; Figure 1h and Table 2c). However, accounting for this difference between Wolbachiainfected and rifampicin-treated controls reveals that, relative to their respective control, both fungi induced higher mortality in rifampicintreated mites (HR = 17.87 and HR = 33.39, for BB and MB, respectively) than in Wolbachia-infected ones (HR = 8.60 and HR = 13.21, respectively). A significant interaction between infection and rifampicin treatment was also found at 3 dpi (X 2 2 = 6.5, p = .04; Figure 1i). However, this interaction was relatively weak at this time-point and could not be explained by multiple comparisons between factor levels (i.e., no differences were found between Wolbachia-infected and rifampicin-treated mites when sprayed with Tween 20 only, B. bassiana, or M. brunneum; Table 3c for all multiple comparisons).

| D ISCUSS I ON
In T. truncatus spider mites, Zhu et al. (2018) showed that antibiotic treatment affects the composition of the bacterial community even after more than 20 generations without antibiotics. In particular, bacteria from different families increased in proportion in tetracycline-treated mites in absence of the Anaplasmataceae  interests, and future work should go in that direction.
In conclusion, our results show variable effects of Wolbachia on spider mite susceptibility to fungi-induced mortality using two generalist fungi, B. bassiana and M. brunneum. To our knowledge, this is the first study investigating the interaction between natural Wolbachia infections and widely used biocontrol agents. As Wolbachia was found to have either no effect or to increase spider mite susceptibility to fungal infection, these results suggest that it may improve the success of biological control using entomopathogenic fungi. However, these results also highlight the complexity of within-host-pathogens interaction and caution against the generalization of such effects as (a) the outcome of these interactions may vary depending on the fungal strain, the Wolbachia strain, and the host genetic background, and (b) these interactions may evolve at a rapid pace with potentially important consequences for the ecology and epidemiology of Wolbachia infection in arthropod pests. Finally, our findings also point to the importance of considering the whole bacterial community of arthropods when assessing the effect of Wolbachia in a particular system.

ACK N OWLED G M ENTS
We thank Catarina Pinto and João Alpedrinha for their help in collecting some data, as well as Marta Palma for technical support. We also thank all members of the SM lab for useful discussions and sug-

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R S' CO NTR I B UTI O N S
FZ and SM conceived and designed the experiment; IS was involved in the maintenance of spider mite populations and plants; MA acquired the data; FZ performed statistical analyses; FZ and SM wrote the manuscript, with input from all authors; IC and SM funded the study. All authors have read and approved the final version of the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
Full dataset has been deposited in the Dryad data repository (doi. org/10.5061/dryad.9p8cz8wc4).