Symbiont strain is the main determinant of variation in Wolbachia‐mediated protection against viruses across Drosophila species

Abstract Wolbachia is a common heritable bacterial symbiont in insects. Its evolutionary success lies in the diverse phenotypic effects it has on its hosts coupled to its propensity to move between host species over evolutionary timescales. In a survey of natural host–symbiont associations in a range of Drosophila species, we found that 10 of 16 Wolbachia strains protected their hosts against viral infection. By moving Wolbachia strains between host species, we found that the symbiont genome had a much greater influence on the level of antiviral protection than the host genome. The reason for this was that the level of protection depended on the density of the symbiont in host tissues, and Wolbachia rather than the host‐controlled density. The finding that virus resistance and symbiont density are largely under the control of symbiont genes in this system has important implications both for the evolution of these traits and for public health programmes using Wolbachia to prevent mosquitoes from transmitting disease.

used as a biocontrol agent. Strains of Wolbachia have been transferred from Drosophila to the mosquito Aedes aegypti with the aim of preventing the transmission of dengue virus (Frentiu et al., 2014;Joubert et al., 2016;Moreira et al., 2009;Walker et al., 2011;Yeap et al., 2011). Understanding what governs changes in phenotype following a host shift can thus help predict the success of such symbiont-based applications, and will determine whether model species like Drosophila melanogaster can be used to identify the best symbiont strains to transfer to mosquitoes.
The role of the host genome in determining the phenotype of Wolbachia infections has been investigated by experimentally moving Wolbachia between host species. Many of these studies have investigated reproductive manipulations such as cytoplasmic incompatibility and sex ratio distortion (Fujii, Kageyama, Hoshizaki, Ishikawa, & Sasaki, 2001;Jaenike, 2007;Poinsot, Bourtzis, Markakis, & Savakis, 1998;Sakamoto et al., 2005;Veneti et al., 2012). Here, host shifts have been shown to be associated with changes in the intensity of the phenotype (Poinsot et al., 1998), a complete loss of the phenotype (Veneti et al., 2012) or even a switch in the type of reproductive alteration (Jaenike, 2007).
Here, we compared Wolbachia strains in their native host and a new host to test whether the host and/or symbiont genome determines the level of antiviral protection. We first assessed the frequency of antiviral protection in 16 natural host-symbiont associations. We then compared the level of protection induced by eight of these Wolbachia strains in both their original host and a line of D. simulans to which they have been artificially transferred. We find that the level of antiviral protection is largely determined by the Wolbachia strain rather than the host species.
2 | METHODS 2.1 | Drosophila stocks and Wolbachia strains All Drosophila species were maintained on a cornmeal diet (see recipe in Longdon et al., 2015) at 25°C, under a 12-hr light/dark cycle and 70% relative humidity. Ten Drosophila species infected with their native Wolbachia strains were used in this study (Tables 1   and S1). Of these, more than one line of D. melanogaster and D. simulans was used, each infected with a different Wolbachia strain. For each Wolbachia-infected fly line, we had a matching Wolbachia-free control. Wolbachia-infected D. melanogaster and the uninfected control were created using balancer chromosomes to homogenize their nuclear background as described in Chrostek et al. (2013). For all the other fly lines, a Wolbachia-free line was created from Wolbachiainfected flies by raising them on Ready Mix Dried Food (Philip Harris) supplemented with 0.03% w/v tetracycline for two generations.
In order to homogenize the gut microbiota between Wolbachiainfected lines and their tetracycline-treated counterparts, the tetracycline-treated lines were then raised for one generation on standard cornmeal food on which ten males of the respective Wolbachia-infected line had been kept for 1 day and removed (as in Wolbachia strains that were also used in the D. simulans line STCP (see Methods). Martinez et al., 2016). Experiments were all performed more than twenty generations after tetracycline treatment. The Wolbachia infection status of all fly lines was checked by PCR and Sanger sequencing as described below.
In order to compare the Wolbachia strains in their original host and in a new host, we also used the D. simulans line STCP into which some of the Wolbachia strains were previously transferred through backcrossing or microinjection (Martinez et al., 2014;Poinsot et al., 1998;Zabalou et al., 2008;

| Virus production
To test antiviral protection, we used Flock House virus, which has a positive-sense single-stranded RNA genome. FHV belongs to the family Nodaviridae and was initially isolated from a beetle (Scotti, Dearing, & Mossop, 1983). We chose to use FHV instead of a native virus such as Drosophila C virus (Comendador et al., 1986;Plus, Croizier, Jousset, & David, 1975) as we have found that there is less genetic variation among hosts in susceptibility to FHV (Magwire et al., 2012). FHV was produced in Schneider Drosophila line 2 (DL2) cells. Cells were cultured at 26.5°C in Schneider's Drosophila medium with 10% foetal bovine serum, 100 U/ml penicillin and 100 mg/ml streptomycin (all Invitrogen, UK). Cells were then freeze-thawed twice to lyse cells and centrifuged at 4,000 g for 10 min at 4°C to remove cellular debris. Virus was then aliquoted and frozen at À80°C. Virus infectivity was calculated using serial dilutions of virus in Schneider's medium added to wells of a plate of DL2 cells as described in Longdon, Cao, Martinez, and Jiggins (2013). After 7 days, the wells were visually examined under the microscope and classed as "infected" when cell death (presence of cell debris) and cytopathic effects were visible (lysing, shrinking or losing of compartmentation of cells). The Tissue Culture Infective Dose 50 (TCID50) was calculated by the Reed-Muench endpoint method (Reed & Muench, 1938).

| Wolbachia screening
The Wolbachia infection status of fly lines was checked by PCR using the diagnostic primers wsp81F and wsp691R (Zhou, Rousset, & O'Neill, 1998). DNA from ten female flies per fly line was first extracted by crushing the flies in 150 ll of a 5% w/v suspension of Chelex 100 resin (Sigma-Aldrich) and 1 ll of proteinase K (20 mg/ ml, Fermentas). Extracts were incubated for 5 hr at 56°C. After 10 min at 95°C, samples were centrifuged and stored at À20°C. PCR conditions were as described in Ref. (Zhou et al., 1998). For the Wolbachia-infected lines, the PCR products of the genes wsp and 16S (16Swol F: 5 0 -TTGTAGCCTGCTATGGTATAACT-3 0 ; 16SWol Karr, & Robertson, 1992) were Sanger-sequenced to identify the Wolbachia infections at the strain level.

| Survival assay
To infect flies with FHV, 3-to 6-day-old females were anaesthetized with CO 2 and then stabbed into the left pleural suture of the thorax

| Wolbachia density
To measure the Wolbachia density within fly tissues, DNA was extracted using the Gentra Puregene kit (Qiagen) from a pool of ten 2-to 5-day-old females reared at 25°C. Flies from each Wolbachiainfected fly line were collected every day from the same cohorts used in the second survival experiment. Five biological replicates (independent pools of females) were extracted for each Wolbachiainfected line and the DNA was then diluted 1:10 with nuclease-free water. For each Drosophila species, we sequenced the fly gene RpL32 as in Longdon, Hadfield, Webster, Obbard, and Jiggins (2011) and designed species-specific primers in two conserved regions for quantitative PCR (qPCR) ( Table S2). The copy number of the Wolbachia gene atpD (atpDQALL_F: 5 0 -CCTTATCTTAAAGGAGGAAA-3 0 ; atpDQALL_R: 5 0 -AATCCTTTATGAGCTTTTGC-3 0 ) relative to the endogenous Drosophila control gene RpL32 (species-specific primers; Table S2) was quantified with the SensiFAST SYBR and Fluorescein kit (Bioline). For each gene, all samples were run on the same qPCR plate and a second technical replicate was performed on a different plate. The efficiency with which each set of primers amplified the product was checked using a dilution series. In all cases, the efficiency was >95% (with 100% efficiency equating to a doubling of the PCR product concentration every cycle). The Wolbachia density was estimated as 2 DCt , where Ct is the mean cycle threshold of the two technical replicates and DCt = Ct RpL32 À Ct atpD . The qPCR cycle was 95°C for 2 min, followed by 40 cycles of 95°C for 5 s and 60°C for 30 s.

| Viral titre
In order to estimate FHV titre, flies were raised and females were infected with virus under the same conditions as the survival experiments. As the first experiment was performed at 22°C and the second one at 25°C, flies were collected 5 and 3 days post-infection, respectively, in order to allow sufficient time for the viral replication before any significant mortality occurs. Flies were snap-frozen in liquid nitrogen in pools of ten females (five to ten biological replicates from separate pools of flies per fly line). Flies were then homogenized in TRIzol.
Total RNA was extracted using TRIzol (Invitrogen) and reverse-transcribed with Promega GoScript reverse transcriptase (Promega) and random hexamer primers, and then diluted 1:10 with nuclease-free water. The FHV RNA copy number (forward: 5 0 -ACCTCGATGGCAG GGTTT-3 0 ; reverse: 5 0 -CTTGAACCATGGCCTTTTG-3 0 ) relative to the endogenous control gene RpL32 (species-specific primers; Table S2) was quantified as for the Wolbachia density with two technical replicates per sample. As for the Wolbachia density, the speciesspecific primers for RpL32 were designed in two conserved regions, except that the forward primer was designed on an exonexon junction in order to amplify only mRNA. This exon-exon junction was previously confirmed in several Drosophila species (Longdon et al., 2011). For a given sample, the Ct values were averaged between the two technical replicates and the relative FHV titre was calculated as DCt = Ct RpL32 À Ct FHV .

| Statistical analysis of survival data
Statistical analyses were performed in the R software (R Core Team 2013). Survival rates were analysed using Cox's proportional hazard mixed-effect models (package COXME). This allowed estimating the antiviral protection conferred by each Wolbachia strain as a hazard ratio. The hazard ratio for a given Wolbachia-infected line is the probability of death occurring at a given time point divided by the probability of death in the respective Wolbachia-free line. Flies that were alive at the end of the experiment were treated as censored data.
To estimate the level of antiviral protection provided by each Wolbachia strain in the original hosts, we fitted the model: where k 0 is a baseline hazard, H i is a fixed effect of host species i, W j is a fixed effect of Wolbachia infection status j (infected or Wolbachia-free), and H i :W j is an interaction between host species and infection status. The vial in which each fly was found was treated as a random effect (v k ) and e ijkl was the residuals.
The survival data in D. simulans STCP line were analysed with the simpler model: Viral titres (T) in the original host species were analysed as: where the parameters are defined in Model (1). Viral titres (T) in D. simulans STCP were analysed as: where the parameters are defined in Model (2). Wolbachia density (D) in D. simulans STCP and the original hosts was analysed as: 2.9 | The relative importance of host and symbiont genomes To quantify the relative importance of the host and symbiont genomes, we used an ANOVA to analyse trait data from both the original hosts and D. simulans STCP. The response variable was either relative survival (see below), relative viral titre (see below) or Wolbachia density of Wolbachia-infected flies (R). This allowed us to fit the linear model: In this data set, there is no cross-factoring of the different hosts and symbiont strains. Therefore, we cannot distinguish a main effect of the host (an effect of the host on all Wolbachia strains) from a host-by-Wolbachia interaction (an effect of the host on specific Wolbachia strains). For this reason, the effect of the host j (H j ) was nested within the effect of Wolbachia strain i (S i ). As these were both treated as fixed effects, this is equivalent to fitting one main effect (S i ) and one interaction (S i : H j ).
For the survival data, the response variable R ijk was the hazard of a vial of Wolbachia-infected flies relative to the mean hazard of the  F 15,95 = 2.51, p = .004). Four of the symbiont strains significantly reduced titres (Figure 1b). Furthermore, the reduction in viral titre caused by Wolbachia was positively correlated with increases in survival after infection (Figure 1c). Overall, the change in titre explained 70% (r 2 ) of the variance in protection (Figure 1c).

| Most variation in antiviral protection is
explained by the symbiont strain rather than the host species  (Table 2a). However, while the Wolbachia strain explained more than 90% of the variance in protection, less than 5% was explained by symbionts behaving differently in the different hosts (Model 6,

| Symbiont density is conserved when strains are transferred between host species and explains most of the variation in antiviral protection
Within a single host species, Wolbachia-mediated protection is known to be tightly linked to the density of the symbiont in host tissues (Chrostek et al., 2013;Martinez et al., 2014;Osborne, Iturbe-Ormaetxe, Brownlie, O'Neill, & Johnson, 2012;Osborne et al., 2009). To explain why the host genetic background has little effect on the level of antiviral protection that a given Wolbachia strain provides, we tested whether symbiont densities were conserved when a Wolbachia strain was moved between different hosts. We found significant variation in density between Wolbachia strains in both the original hosts and the common genotype of D. simulans (Figure 4a).
As for protection, the Wolbachia strain explained far more of the variance in symbiont density than the host genetic background

| DISCUSSION
The extent to which Wolbachia protects insects against viruses varies greatly among host-symbiont associations. While symbiont strain is known to be a key determinant of protection (Chrostek et al., 2013;Martinez et al., 2014;Osborne et al., 2009), the role of the host genome has been poorly investigated. By comparing several Wolbachia strains in different host species, we found that the symbiont genome was far more important than the host genome in determining the level of protection. This was due to the density that a given Wolbachia strain reaches being conserved when it is moved to a new host.
In natural host-symbiont associations, we found that Wolbachia commonly protects Drosophila against viral infection. Wolbachia significantly decreased virus-induced mortality in more than half (10/ 16) of the Drosophila-Wolbachia associations tested, although in most cases the increase in survival was only modest. This is similar to the patterns we have reported from a panel of Wolbachia strains that we transferred into D. simulans, where about half of the strains F I G U R E 2 Wolbachia-mediated protection in original hosts and a common genotype of D. simulans. (a) Survival curves following infection with FHV. p-values for the comparisons of Wolbachia-infected and Wolbachia-free flies after infection with FHV are shown (Model 1 and 2, see Methods). When this analysis was repeated on the mock-infected flies, none of the Wolbachia strains significantly affected survival (models 1 and 2; p > .05 in all cases). (b) Correlation in Wolbachia-mediated increases in survival after FHV infection in the original hosts and the common genotype of D. simulans. The dashed lines show predicted values from linear regressions with (black) and without (red) wAu. r is Pearson's correlation coefficient between traits [Colour figure can be viewed at wileyonlinelibrary.com] T A B L E 2 Statistical analysis of Wolbachia-mediated protection in original hosts and the common genotype of D. simulans   provided protection (Martinez et al., 2014). Other studies of single species found that Wolbachia protects against FHV in Drosophila innubila (Unckless & Jaenike, 2011), D. suzukii (Cattel et al., 2016) and D. melanogaster (Hedges et al., 2008;Teixeira et al., 2008), but not D. bifasciata (Longdon, Fabian, Hurst, & Jiggins, 2012). Unlike our results, interpreting these different studies can be difficult due to publication biases towards positive results and differences in experimental conditions and statistical power. For example, antiviral protection and symbiont density are affected by temperature and diet (Caragata et al., 2013;Mouton, Henri, Charif, Boul etreau, & Vavre, 2007;Serbus et al., 2015;Ulrich, Beier, Devine, & Hugo, 2016). This may explain why the strain wHa was previously found to be nonprotective (Osborne et al., 2009) but conferred low levels of protection in our study using the same fly stock. Similarly, Cattel et al. (2016) found weak protection against FHV in D. suzukii but we did not.
By comparing the same symbionts in different hosts, we found that the symbiont strain was far more important than the host species in determining whether Wolbachia protects Drosophila against FHV. This was true in terms of both survival and viral titre. The large differences between Wolbachia strains have been reported before (Chrostek et al., 2013;Martinez et al., 2014;Osborne et al., 2009), but the small effect of the host was unexpected given that the host genetic background is critical to the expression of other Wolbachia phenotypes (Jaenike, 2007;Poinsot et al., 1998;Veneti et al., 2012). As found in previous studies (Chrostek et al., 2013;Martinez et al., 2014;Osborne et al., 2009), the symbiont strains varied greatly in their densities and this correlated with antiviral protection.

Reduction in relative FHV
Critically, when the symbionts were transferred between host species, the strain-specific densities were mostly conserved. Therefore, symbionts appear to regulate their density independently of the host, and this in turn determines the level of antiviral protection.
Previous work found that the host genotype affects Wolbachia density (Kondo, Shimada, & Fukatsu, 2005;Mouton et al., 2007;Veneti et al., 2012). Our study does not contradict these results as we also found host effects on symbiont density, but these were small compared with the Wolbachia strain effect. Our results differ from studies of another Drosophila symbiont. As is the case for Wolbachia, the density of Spiroplasma symbionts that protect some Drosophila species against parasitic nematodes is similar between the native and the novel hosts (Haselkorn et al., 2013). However, the protective effect of different symbiont strains is decoupled from their density and strongly depends on the host species (Haselkorn et al., 2013). Therefore, the host genetic background is a strong determinant of the protective phenotype of Spiroplasma but not Wolbachia.
During its evolution, Wolbachia has frequently jumped between host species (Vavre et al., 1999;Werren et al., 1995;Zhang et al., 2013). Our results suggest the protective phenotype will often be transferred to the newly infected host. This could drive up the frequency of Wolbachia in the new host, potentially making protective strains more likely to move between species. This may be especially important for CI-inducing Wolbachia strains, as these need to reach a minimum frequency in the population to be able to spread (Turelli, 1994). The benefit conferred by antiviral protection to the new host may promote the spread of the newly acquired Wolbachia infection allowing it to reach this threshold. The importance of this effect will depend on RNA viruses being a strong selective pressure, as highly protective Wolbachia strains are costly for the insect owing to their high density within host's tissues (Chrostek et al., 2013;Martinez et al., 2015). RNA viruses are extremely prevalent in Drosophila populations (Webster et al., 2015), but their effects on fitness in nature are unknown.
The observation that the host genome has comparatively little effect on antiviral protection or Wolbachia density is interesting. It seems likely that there is selection on hosts to control Wolbachia density to some optimal level, as RNA viruses are common Drosophila pathogens (Webster et al., 2015) and high Wolbachia densities substantially reduce host fitness (Martinez et al., 2015). Our finding that hosts have not evolved to modulate symbiont densities suggests there may be constraints that prevent flies from altering Wolbachia density. For example, Wolbachia may occupy an intracellular niche that protects it from insect immune defences (Bourtzis, Pettigrew, & O'Neill, 2000;Siozios, Sapountzis, Ioannidis, & Bourtzis, 2008). This could mean that hosts might be more likely to evolve tolerance to Wolbachia infections rather than mechanisms controlling the symbiont density.
Being able to predict the antiviral effects of a Wolbachia strain in a new host is useful for public health programmes that are releasing Wolbachia-infected mosquitoes to prevent disease transmission. The Zika and dengue vector Ae. aegypti does not harbour Wolbachia in nature and therefore needs to be artificially infected with Wolbachia strains found in other host species (Hoffmann et al., 2015). These transfers are laborious and time-consuming. Finding the optimal Wolbachia strains for disease control would be greatly facilitated by screening symbiont strains in Drosophila where the artificial transfer of Wolbachia between species has become routine (Chrostek et al., 2014;Martinez et al., 2014;Poinsot et al., 1998;Veneti et al., 2012). Our results suggest that such studies are likely to be a powerful way to select symbiont strains to be used as biocontrol agents.
We conclude that the extent to which Wolbachia protects different Drosophila species against viral infection depends primarily on the symbiont strain and not the host genome. This is due to Wolbachia regulating its density to similar levels in different host species.
Wolbachia density in turn determines whether Wolbachia protects the host against viruses.