Virus evolution in Wolbachia-infected Drosophila

Abstract

Wolbachia, a common vertically transmitted symbiont, can protect insects against viral infection and prevent mosquitoes from transmitting viral pathogens. For this reason, Wolbachia-infected mosquitoes are being released to prevent the transmission of dengue and other arboviruses. An important question for the long-term success of these programmes is whether viruses can evolve to escape the antiviral effects of Wolbachia. We have found that Wolbachia altered the outcome of competition between strains of the DCV virus in Drosophila. However, Wolbachia still effectively blocked the virus genotypes that were favoured in the presence of the symbiont. We conclude that Wolbachia did cause an evolutionary response in viruses, but this has little or no impact on the effectiveness of virus blocking.

1. Introduction

Wolbachia is a maternally transmitted intracellular bacterium found in many insects [1]. Its ability to spread rapidly through insect populations by inducing a sperm–egg incompatibility called cytoplasmic incompatibility [2–4], coupled with its inhibitory effect on the replication of RNA viruses [5–8], make it a promising control agent to prevent the transmission of mosquito-borne diseases [9]. In several parts of the world, the bacterial symbiont is being introduced into natural populations of the mosquito Aedes aegypti, the main vector of dengue and Zika viruses [10–13]. Preliminary field releases of Wolbachia-infected Ae. aegypti females have demonstrated that the bacterial infection is able to spread and be stably maintained [4,10,11], turning susceptible populations of mosquitoes into virus-resistant ones [14].

Like other control methods, there is a risk that the release of Wolbachia-infected mosquitoes may not be evolution-proof. It may promote adaptive changes in the mosquito vector, Wolbachia or virus that could hamper the long-term success of field interventions. Therefore, there is an urgent need to understand and predict what genetic changes might follow the introduction of Wolbachia, especially because such introductions are likely to be irreversible [15,16]. For instance, high antiviral resistance is associated with high densities of the symbiont within the insect tissues [17,18], and this leads to reductions in the fecundity, lifespan and other fitness-related traits of the insect host [11,19–22]. These costs may lead to the evolution of lower Wolbachia densities and thus a reduction or loss of the antiviral phenotype. A second concern is the evolution of the virus itself. Since Wolbachia blocks the transmission of the virus by inhibiting its replication, virus populations should be selected to overcome such inhibition. Potentially, virus strains that are able to replicate at a higher rate in the presence of Wolbachia could be advantaged and spread. For example, the intensity of Wolbachia's effect on dengue transmission varies between virus serotypes [14], and the magnitude of these differences is sufficiently large that it is predicted to alter the outcome of control programmes [23]. Therefore, if viruses can escape the resistance conferred by Wolbachia, this would threaten the sustainability of symbiont-based interventions.

Here, we passaged Drosophila C virus (DCV) through Wolbachia-infected Drosophila melanogaster and examined how the symbiont affected the evolution of the virus. DCV is a positive-strand RNA virus of the family Discistroviridae that naturally infects D. melanogaster [24]. It is highly pathogenic in laboratory experiments, leading to fly death within a few days [25]. Wolbachia leads to large reductions in DCV titres and increases survival after DCV infection, which has led to DCV becoming a common model to study Wolbachia-mediated antiviral protection [5,17,18]. We found that populations of the virus became genetically differentiated from controls in Wolbachia-free flies, with the same viral genotype being favoured across replicate populations. However, despite these parallel genetic changes providing evidence of adaptive evolution in the viral populations, we could not detect any reduction in Wolbachia's antiviral effect or any increase in DCV virulence.

2. Methods

(a). Fly husbandry and virus isolates

Drosophila melanogaster lines previously described in [21] were kindly provided by Luis Teixeira. The DrosDel w¹¹¹⁸ isogenic background was used as the Wolbachia-free control. The Wolbachia-infected line was created in [21] by introgressing the DrosDel w¹¹¹⁸ nuclear background into a cytoplasm infected with the Wolbachia strain wMelCS_b through chromosome replacement using balancers for the first, second and third chromosomes (the fourth chromosome was not replaced). Flies were maintained on a cornmeal diet (1200 ml water, 13 g agar, 105 g dextrose, 105 g maize, 23 g yeast, 35 ml Nipagin) at 25°C under a 12 h light–dark cycle and 70% humidity.

DCV isolates were previously described in [26] and kindly provided by Karyn N. Johnson. Isolates DCV-C and -G originate from France, DCV-EB and -CYG from Australia, and DCV-M, -O, -T and -Z from Morocco.

(b). Virus production

All DCV isolates were passaged once in Schneider Drosophila Line cells (DL2) before the series of experiments. Cells were cultured at 26.5°C in Schneider's Drosophila medium with 10% fetal bovine serum, 100 U ml⁻¹ penicillin and 100 mg ml⁻¹ streptomycin (all Invitrogen, UK). Cells were then freeze–thawed twice to lyse cells and centrifuged at 4000g for 10 min at 4°C to remove cellular debris. Finally, the supernatant containing DCV was aliquoted and frozen at −80°C.

For infection assays, aliquots of virus solutions were defrosted on the day of infection. Virus aliquots were diluted in Ringer's solution [27] to standardize the concentration of DCV RNA measured by quantitative PCR (qPCR, see DCV titre for primers and amplification cycles). The concentration of DCV RNA was used instead of the TCID50 method [28] as the cytopathic effects of the eight DCV isolates differed considerably. Total RNA from the eight virus solutions was extracted and a standard Wolbachia spike-in added during the RNA extraction. The DCV critical threshold values were then normalized relative to the Wolbachia gene atpD using primers atpDQALL_F (5'-CCTTATCTTAAAGGAGGAAA-3') and atpDQALL_R (5'-AATCCTTTATGAGCTTTTGC-3').

(c). Sequencing of Drosophila C virus genome and phylogenetic analysis

Viral RNA was extracted using TRIzol reagent (Invitrogen) and reverse-transcribed with the Maxima H Minus Reverse Transcriptase (ThermoFisher Scientific) and oligo dT primers. The DCV genome was then amplified by PCR using a set of 22 primers distributed along the 9264 bp genome (electronic supplementary material, table S1). For each genome, the 11 PCR products were Sanger-sequenced and the reads assembled into a consensus genomic sequence using Sequencher v. 4.5 software (GenBank accession numbers: MK645238–MK645245). DCV genomes were aligned with ClustalW in BioEdit v. 7.0.9 [29]. A maximum-likelihood phylogenetic tree was built using the aligned genomes and node supports were assessed with 1000 bootstrap replications.

(d). Selection experiment 1 on genetically diverse virus populations

To investigate virus adaptation from standing genetic variation, we created a virus population composed of equal proportions of each of the eight DCV isolates based on the relative amount of DCV RNA in the original virus solutions. The DCV mixture was then passaged in flies with or without Wolbachia by infecting 3–6-day-old female flies (passage 1). For this, flies were anaesthetized on CO₂ and stabbed on one side of the thorax with a steel needle (Austerlitz Insect Pins) dipped into the virus solution as explained in [30]. Ten flies were then transferred to a vial containing cornmeal food. Twelve vials per Wolbachia treatment were prepared, representing 12 biological replicates. Virus populations were harvested 3 days post-infection by homogenizing the 10 virus-infected flies from each vial in 25 µl of Ringer's solution. The homogenate was then centrifuged at 12 000g and 10 µl of the supernatant was frozen at −80°C and later used as the inoculum for further passage. The remaining 20 µl containing the fly tissues were diluted in 250 µl of TRIzol reagent and frozen at −80°C for later RNA extraction. The virus populations were serially passaged two more times (passages 2 and 3) by repeating the steps above and infecting new flies from the respective Wolbachia treatment. Note the amount of virus in the inoculum was only controlled at the start of selection (passage 1). In order to avoid cross-contamination between virus populations, different tools (needles, handling brushes) were used for each replicate population at each passage.

(e). Pool-sequencing, read processing, mapping and variant identification

Total RNA was extracted from flies in passage 3 of the selection experiment 1. Since the DCV genome contains a poly(A) tail, we isolated the virus genomic RNA along with fly transcripts by capturing the polyadenylated RNAs from the extracted total RNA using the KAPA Stranded mRNA-Seq kit (Kapa Biosystems) and a different indexed adapter for each RNA library. Twenty-four libraries were prepared (one per virus population), quantified by qPCR and pooled in equal proportions into a multiplexed library. The pool was sequenced in one lane of Illumina HiSeq4000 to generate single-end 50 bp reads (SRA study accession number PRJEB21984).

We used Trimmomatic v. 0.32 [31] to trim reads. We first removed three bases from the 3′ end of the read. Reads were quality trimmed from the 3′ end, cutting when the average quality scores in sliding windows of four bases dropped below 15. We required reads have a minimum length of 36. Using BWA MEM [32], we mapped reads to the genome of a genetically homogeneous DCV population (DCV-ref) previously produced from the isolate DCV-C by endpoint dilution [33]. We removed optical duplicate reads using Picard tools (https://broadinstitute.github.io/picard/). We realigned reads close to indels using GATK [34]. Following that, we used Samtools [35] to remove reads with mapping quality scores lower than 40. We also used Samtools to generate a multi-pileup file to assist with variant identification. Finally, we used PoPoolation2 to identify the allele frequency of the variants in the pooled genomic library samples [36]. To make the statistical power and influence of different variants and libraries similar, we down-sampled sites to a maximum coverage of 50×/library.

(f). Recombination analysis

In order to investigate the extent of recombination of the DCV genome in selection experiment 1, we resequenced the 24 RNA libraries using MiSeq to generate paired-end 250 bp reads (SRA study accession number PRJEB21984). The software bcl2fastq on BaseSpace (Illumina Inc., San Diego, CA, USA) was used to demultiplex the base call files to Fastq files, trim adaptor sequences using a sliding window with an adapter stringency of 0.9. We then used Trimmomatic [31] to cut reads at the first base where the quality score (Q) dropped below 30, retaining only reads with a minimum length of 35 bp. These were mapped to the genome of isolate DCV-C (GenBank accession number MK645242) using BWA MEM [32]. We used Picard tools (https://broadinstitute.github.io/picard/) to reorder reads, add read groups and sort them by coordinates. Mapped reads were converted into a table of variants where each row is a separate sequence read using sam2tsv in JVarkit [37]. The reads from all libraries were combined. We kept properly paired reads, as identified by their SAM flags, with at least one SNP of interest in each of the forward and reverse reads using custom scripts that are deposited in the Dryad Data Repository (http://dx.doi.org/10.5061/dryad.18j31ch). When a site had the DCV-C allele, we then counted the proportion of read-pairs where the other SNP either had the DCV-C allele (non-recombinant) or the alternate allele (recombinant). As polymorphisms segregating in the founding viral populations can give a spurious signal of recombination in this analysis [38], we removed any pairs of SNPs where the ratio of reads carrying the two possible products of recombination deviated from the expected 50 : 50 ratio (binomial test, p < 0.01). We tested whether the fraction of recombinant reads between pairs of SNPs increased with distance between the SNPs using logistic regression, accounting for overdispersion using a quasi-binomial model. Retaining only pairs of SNPs that were represented by at least 25 read-pairs, we regressed the proportion of recombinant reads against the distance between the SNPs to estimate the fraction of recombinants between adjacent nucleotides.

(g). Drosophila C virus titre

Total RNA was extracted 3 days post-infection from pools of 10 DCV-infected flies. Five biological replicates (10 flies each) were performed per Wolbachia treatment, DCV isolate (single infection experiment) or DCV population (selection experiments). The extracted RNA was reverse-transcribed with Promega GoScript reverse transcriptase (Promega) and random hexamers. The amount of virus RNA was quantified with qPCR by amplifying a 135 bp region of the DCV genome with primers DCV_S (5′-GACACTGCCTTTGATTAG-3′) and DCV_AS (5′-CCCTCTGGGAACTAAATG-3′) targeting regions that are conserved among the DCV isolates used in this study. Additionally, we quantified the fly gene actin 5C in a separate reaction (forward: 5′-GACGAAGAAGTTGCTGCTCTGGTTG-3′; reverse: 5′-TGAGGATACCACGCTTGCTCTGC-3′; 193 bp product). 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. Two reactions per sample and per target gene were performed and the mean Ct value of the two technical replicates was used to calculate the relative amount of DCV RNA per fly as 2^ΔCt, with ΔCt = Ct_{fly gene} – Ct_DCV.

(h). Virus-induced mortality

Flies were infected with each DCV isolate or with Ringer's solution (mock-infected controls) as above except that biological replicates consisted of vials with 20 females. Following infection, flies were transferred onto fresh food every 3 days and survival was recorded daily for 15 days post-infection.

(i). Selection experiment 2 on genetically homogeneous virus populations

In order to study virus evolution from de novo mutation, we used a similar protocol as in the selection experiment on genetically diverse virus populations. We passaged the homogeneous DCV population DCV-ref. [33] for ten passages in either Wolbachia-free or Wolbachia-infected flies (25 replicate populations in each treatment). The only difference was that male flies were used instead of females and viruses were harvested two days post-infection instead of three. Flies were initially infected in passage 1 with a virus concentration of 6.32 × 10⁸ TCID50 ml⁻¹.

(j). Statistical analysis

All statistical analyses were done in R software v. 3.2.3 [39] and R scripts deposited in the Dryad Data Repository (http://dx.doi.org/10.5061/dryad.18j31ch). Pairwise genetic distances between virus populations of the selection experiment 1 were calculated as the probability of drawing different alleles from two different populations P_diff = 1 − (p_pop1 × p_pop2 + q_pop1 × q_pop2), where p_pop1 and p_pop2 are the mean frequencies of reference alleles in population 1 and 2 and q_pop1 and q_pop2 the mean frequencies of the alternative allele. We conducted a Mantel test with 1000 permutations by randomly attributing Wolbachia treatments to populations. The p-value was obtained by comparing the observed mean genetic distance between treatments with the null distribution of mean genetic distances obtained by permutation.

The principal component analysis was performed with the function dudi.pca (R package ade4). For each Wolbachia treatment, 95% confidence ellipses were computed with the assumption of multivariate normal distribution of the data using the function stat_ellipse (R package ggplot2). The discriminant analysis of principal components was performed with the function dapc (R package adegenet [40]).

Viral titres were analysed with a linear model after log₁₀ transformation of the data to meet the assumptions of normality and homoscedasticity. For the experiment measuring viral titres of evolved populations, the virus population was treated as a random effect in a linear mixed-effect model (package LmerTest). For the time-course analysis of DCV titre, three-parameter asymptotic exponential growth curves were fitted to the log₂ viral titre using the function nls (R package stats). To test whether there was an effect of Wolbachia on the growth curve, we used a likelihood ratio test to compare the fit of a single curve to all the data with the fit of separate curves to the data from Wolbachia-infected and Wolbachia-free flies. To test whether there were either differences in the growth of the three viruses or there were virus-specific effects of Wolbachia, we compared our two-curve model (Wolbachia + and −) to a six-curve model (a separate curve for each of the three viruses, with and without Wolbachia). When plotting the curves, 95% confidence intervals were estimated by Monte Carlo simulation.

Fly survival was analysed with a Cox's proportional hazard mixed-effect model (R package coxme). Flies that were alive at the end of the experiment were treated as censored data. Multiple pairwise comparisons were performed with the function glht (R package multcomp [41]).

3. Results

(a). Virus populations evolve in response to Wolbachia

In order to test whether DCV adapts to the presence of Wolbachia, we passaged a genetically diverse population of viruses through Wolbachia-infected or Wolbachia-free flies (selection experiment 1). The viral population was founded by mixing eight DCV isolates collected from both laboratory stocks and wild D. melanogaster from around the world [26]. The mixture initially contained equal concentrations of viral RNA from each DCV isolate and after three passages, we sequenced the polyadenylated RNA from 12 independent replicates per Wolbachia treatment. The mean depth of coverage of each replicate ranged from 65 to 89× (electronic supplementary material, table S2). After filtering out variants with a mean minor allele frequency below 5%, there were 167 Single Nucleotide Polymorphisms (SNPs). Among these, 161 were found among the genomes of the eight DCV isolates used to found the viral population. There were 703 SNPs among the founding DCV genomes, suggesting that approximately 77% had minor allele frequencies below 5% by the end of the selection experiment.

Parallel evolution, where the same genetic changes evolve independently in response to the same selection pressure, provides evidence of adaptation. To test for parallel evolution in our experiment, we calculated the genetic distance between all possible pairs of populations. We found that the mean genetic distance between DCV populations from the same Wolbachia treatment (Wolbachia-infected or Wolbachia-free flies) was less than between populations from different Wolbachia treatments (Mantel test, p = 0.028; electronic supplementary material, figure S1A). Parallel evolution of the DCV populations was also apparent in a principal component analysis (PCA) on DCV allele frequencies, where the second principle component separated DCV populations depending on whether or not they had evolved in flies infected with Wolbachia (figure 1a).

Figure 1.

Effect of Wolbachia on virus allele frequencies when selecting on a genetically diverse viral population. A genetically diverse population of DCV was passaged through Wolbachia-infected or Wolbachia-free flies and then sequenced. (a) Principal components analysis on allele frequencies of SNPs, where each point is an independent virus population (biological replicate). (b) Differences in the frequency of SNPs along the viral genome. The sequence reads were mapped to the DCV-C reference genome, and the heatmap shows the difference in the frequency of the allele carried by a given DCV strain between Wolbachia treatments (frequency in Wolbachia-infected flies minus frequency in Wolbachia-free flies; red is a higher allele frequency in Wolbachia-infected flies). The tree was computed from the Euclidian distance computed from these differences in allele frequencies. (c) DCV phylogeny with bootstrap support for the nodes. (d) Differences in the frequency of isolate- or clade-specific variants between viral populations that had evolved in Wolbachia-infected and Wolbachia-free flies (positive numbers are more common in the presence of Wolbachia). The number of SNPs is shown in parentheses, and the number in bold is the mean difference in frequency across all those SNPs. NA stands for cases where no isolate- or clade-specific SNPs could be found. (Online version in colour.)

(b). Wolbachia alters the frequency of viral genotypes in the population

The evolutionary response to Wolbachia could either involve changes in the frequency of the eight founding viral genotypes or selective sweeps of specific SNPs through a recombining population of viruses. We calculated the difference in the frequency of SNPs between the populations that had evolved in flies with and without Wolbachia (figure 1b). Across the viral genome, alleles from the DCV-C isolate consistently showed higher frequencies in the presence of Wolbachia (figure 1b), while the DCV-EB, -CYG and -G alleles had lower frequencies. All of the variants that were at a substantially higher frequency in the Wolbachia-infected flies were present in DCV-C, and these were scattered across the genome (figure 1b). Alleles specific to the DCV-C isolate have increased frequency in both the Wolbachia-infected and Wolbachia-free flies, reaching mean frequencies of 77% and 67%, respectively (based on DCV-C specific alleles; electronic supplementary material, figure S1). Therefore, DCV-C was successful in all the populations, but its competitive advantage has been increased with Wolbachia.

These analyses suggest that there may have been limited recombination in the population, and the frequency of the founding viruses has changed in response to selection by Wolbachia. To confirm this, we generated longer sequence reads from the same samples and used read-pairs containing multiple SNPs to examine the rate at which DCV-C had recombined with other viral isolates. We found that there were more putative recombination events between SNPs that were further apart in the genome, as expected if there is recombination (logistic regression: t = 4.45, p = 0.0001). By estimating a per base pair recombination rate and extrapolating this to the whole genome, we estimate that 86% of DCV-C genomes will have survived intact without recombination by the end of the experiment. This supports the conclusion that we are primarily looking at changes in the frequency of the founding viruses.

As there is limited recombination between DCV isolates, we can use SNPs as markers to track changes in the frequency of different viral isolates. We first reconstructed the phylogeny of the DCV isolates using their Sanger-sequenced genomes (figure 1c), finding similar relationships to published analyses of restriction fragment length polymorphisms [26]. In the evolved populations, we identified isolate-specific alleles for four isolates (figure 1d; DCV-C, -E, -M and -T), while DCV-G and DCV-O are polymorphic since their genomes contain high numbers of ambiguous bases (electronic supplementary material, table S3). Larger numbers of SNPs were found that defined clades of viruses on the phylogeny (figure 1c,d). Comparing the frequency of these SNPs among our evolved populations confirmed that DCV-C was favoured in Wolbachia-infected flies, while alleles specific to the clade containing DCV-EB, -CYG and -G decreased in frequency by around 10% in the presence of Wolbachia. There was little difference between Wolbachia treatments in the frequency of isolates in the other clade on the phylogeny (DCV-M, -Z, -T and -O; figure 1c,d).

To further examine the genetic basis of differentiation between the DCV populations in Wolbachia-infected and Wolbachia-free flies, we used discriminant analysis on the principal components (DAPC). The virus populations showed a bimodal distribution which separates the Wolbachia treatments (electronic supplementary material, figure S2A). The genetic differentiation is driven by SNPs across the viral genome, which is consistent with there being limited recombination (electronic supplementary material, figure S2B). Alleles specific to the isolate DCV-C and the EB-CYG-G clade consistently contributed the most to the genetic differentiation between Wolbachia treatments (electronic supplementary material, figure S2C). This confirms that DCV-C was favoured in the presence of Wolbachia and that this was at the expense of viruses in the EB-CYG-G clade.

(c). The Drosophila C virus isolate favoured in Wolbachia-infected flies does not evade the symbiont's antiviral effects

Wolbachia may be selecting for viruses that evade its antiviral effects. In order to investigate this, we inoculated Wolbachia-free and Wolbachia-infected flies with the DCV isolate that increased the frequency in the presence of Wolbachia (DCV-C), one isolate that decreased in frequency (DCV-EB) and one isolate which frequency was little affected by Wolbachia (DCV-M). Viral titres were measured over 6 days to allow an asymptotic exponential growth curve to be fitted to the data. We found that Wolbachia altered the growth curve of DCV (figure 2a; χ² = 36.8, d.f. = 4, p < 0.0001), reducing both the viral growth rate and the final viral titre. However, there was no difference in the growth curves of the three viral isolates, regardless of whether the flies were infected with Wolbachia (figure 2a; χ² = 20.6, d.f. = 16, p = 0.20). Therefore, we found no evidence that Wolbachia has favoured viral isolates that overcome the symbiont's antiviral properties.

Figure 2.

Wolbachia's effect on viral titres and virus-induced mortality in single virus infections. (a) Growth curves of three DCV isolates in Wolbachia-free and Wolbachia-infected flies. The lines are asymptotic exponential curves and the shaded areas are 95% confidence intervals. (b) Survival curves following infection with DCV. (c) Wolbachia effect on virus-induced mortality expressed as –ln(hazard ratio) where the hazard ratio is the probability of flies dying in Wolbachia-infected flies relative to their Wolbachia-free counterpart. Error bars are standard errors and ***p < 0.001. (Online version in colour.)

We also measured the ability of the DCV isolates to kill Wolbachia-free and Wolbachia-infected flies. Levels of mortality varied among virus isolates, with flies infected with DCV-CYG showing the lowest mortality, regardless of whether the flies carried Wolbachia (figure 2b; main effect virus: ${\chi }_i^2=125.29,$ d.f. = 10, p < 0.0001). Wolbachia had no effect on the survival of mock-infected flies (control in figure 2b), but in all cases increased survival of DCV-infected flies (figure 2b,c; main effect Wolbachia: ${\chi }_i^2=132.7,$ d.f. = 6, p < 0.0001). The magnitude of the protective effects of Wolbachia depended on the DCV isolate (figure 2b,c; Wolbachia-by-virus interaction: ${\chi }_i^2=16.7,$ d.f. = 2, p = 0.005). However, there was no association between the extent to which Wolbachia protected flies against the virus (figure 2b) and whether that virus increased the frequency in the presence of Wolbachia (figure 1).

(d). Virus populations evolved with Wolbachia did not adapt to counteract its antiviral effect

To test whether viruses had adapted to overcome the antiviral effects of Wolbachia, we randomly chose five virus populations from each of the selection treatment and infected flies with equal concentrations of viral RNA (see Methods). The viral titre 3 days post-infection was lower in Wolbachia-infected flies, regardless of the selection regime (Wolbachia effect: ${\chi }_i^2=158.68;$ p < 0.0001; electronic supplementary material, figure S3 for pairwise comparisons). Surprisingly, the effect of Wolbachia on viral titres was slightly greater for the viral populations that had been passaged through Wolbachia-infected flies (Wolbachia-by-selection effect: ${\chi }_i^2=4.34;$ p = 0.04; electronic supplementary material, figure S3). Therefore, viruses evolved with the symbiont were still susceptible to the inhibitory effect of Wolbachia.

(e). A genetically homogeneous virus population did not evolve to escape the antiviral effects of Wolbachia

The high mutation rates, replication rates and population sizes of many viruses mean that they can frequently evolve to overcome host resistance during selection experiments through de novo mutations. To test whether this was the case for Wolbachia and DCV, we serially passaged a genetically homogeneous population of the DCV-C isolate through Wolbachia-free and Wolbachia-infected flies (selection experiment 2). While DCV was maintained at high titres in Wolbachia-free flies, titres tended to progressively decrease in the presence of the symbiont and frequently became undetectable (figure 3a). After 10 passages, only 2 of 25 replicates in Wolbachia-infected flies had DCV titres close to those observed in the absence of the symbiont (figure 3a). We tested whether these two virus populations had maintained high titres by evolving to counter the antiviral effects of Wolbachia by infecting new flies with equal concentration of viral RNA. There was no significant difference in the effects of Wolbachia on these viruses compared to control populations passaged through Wolbachia-free flies (figure 3b; Wolbachia effect: F_1,44 = 113.95; p < 0.0001; selection effect: F_1,44 = 0.72; p = 0.4; Wolbachia-by-selection interaction: F_1,44 = 3.25; p = 0.08).

Figure 3.

Presence of DCV and Wolbachia effect on viral titres in selection experiment 2. (a) Ct values obtained from qPCR reactions targeting DCV RNA in virus populations at different passages during selection. Populations passaged in Wolbachia-free flies were only checked for DCV infection at passage 10. Fractions at the top of the plot indicate the number of biological replicates for which DCV was detected relative to the total number of replicates (n = 25 in each selection treatment). (b) Virus titre of the DCV populations 3 days post-infection in Wolbachia-free (black) and -infected (red) flies. Horizontal bold lines and dots indicate mean titres and values per biological replicate, respectively. Error bars are standard errors. (Online version in colour.)

4. Discussion

Wolbachia is able to block the replication of RNA viruses, and this is being harnessed by public health programmes to control mosquito-borne diseases [42]. The impact of such interventions on evolution of the viruses is unknown. If viruses could evolve to escape Wolbachia's antiviral effects, or Wolbachia selects for increased viral virulence in the mosquito or human host, this would have important implications for control programmes. For example, interventions that reduce the growth rate of a pathogen without clearing the infection can select for compensatory increases in replication rates that in turn increase virulence [43]. We found that Wolbachia can alter the evolution of a virus—when DCV evolved in the presence of Wolbachia, there were replicable genetic changes in viral population. This was mediated by Wolbachia modifying competition between virus strains within the insect. However, these changes neither allow the virus to escape Wolbachia's blocking effect nor alter the virus's virulence.

In our first experiment, we passaged a genetically diverse viral population through Wolbachia-infected flies in an attempt to select for viruses that escaped the symbiont's antiviral effects. At the genetic level, there was a clear response to selection, with the DCV-C genotype reaching higher frequencies in the presence of Wolbachia. However, we were unable to find any evidence that DCV-C either had higher virulence or escaped Wolbachia's antiviral effects. It is possible DCV-C was favoured because of Wolbachia altering the strength of competition between viruses. While the mechanisms of the Wolbachia-mediated antiviral effect remain elusive, previous studies have shown that competition between the symbiont and viruses for resources such as cholesterol might be involved [44,45]. By reducing the availability of these resources, Wolbachia might exacerbate fitness differences that already exist between virus isolates. Consistent with this, DCV-C increased the frequency in Wolbachia-free flies, but to a lesser extent than in the presence of Wolbachia.

In the second selection experiment, we tested virus adaptation from de novo mutation. Selection was strong, leading to the loss of most viral populations across serial passages. The few virus populations that managed to persist in the presence of the symbiont still suffered large reductions in titre in Wolbachia-infected flies. Therefore, we were unable to select for major-effect Wolbachia-escape mutations in the virus, and it is possible these viruses simply persisted due to stochastic processes. This is similar to a recent study where dengue virus populations evolved in Wolbachia-infected mosquito cell lines showed rapid decline and frequently went extinct [46]. Strikingly, the few dengue infections that persisted in the presence of Wolbachia showed an almost complete inability to replicate in both Wolbachia-free and Wolbachia-infected cells following selection. We did not observe this with DCV, but our mixed DCV populations evolved with Wolbachia showed slightly lower replication in Wolbachia-infected flies (electronic supplementary material, figure S3), i.e. they appeared to be more affected by Wolbachia than populations evolved without symbiont. It is possible that the presence of Wolbachia leads to the production of more defective viral genomes [47,48]. However, the size of this effect was modest, and whether this is a general pattern remains to be investigated.

The reason why viruses do not readily evolve to escape the antiviral effects of Wolbachia is unclear. However, clues come from two patterns that are recurrently observed. First, strong antiviral protection is associated with high symbiont density within cells [17,18,21]. Second, Wolbachia provides protection against a diverse array of distantly related RNA viruses, including viruses whose natural hosts are Wolbachia-infected [17,49,50]. This is reminiscent of ‘quantitative’ plant defences against insect herbivores [51]. These are secondary metabolites that are produced in large quantities, and protect against a broad array of herbivores by reducing the digestibility of the plant. These contrast with ‘qualitative’ toxin defences that are produced in low quantities [51]. While specialist herbivores frequently evolve to escape qualitative defences, quantitative defences are thought to be more ‘evolution proof’ [51,52]. As discussed above, one of the leading hypotheses to explain the antiviral effects of Wolbachia is that the symbiont competes with the virus for resources such as cholesterol [44,45]. If these resources are essential to viral replication, such a mechanism may be a form of quantitative defence that viruses cannot readily evolve to escape from. This contrasts with ‘qualitative’ forms of antiviral protection, such as restriction factors, changes to surface receptors or drugs, where virus escape may evolve by altering the molecular target of the antiviral effector.

The final verdict on whether Wolbachia-based control of vector-borne disease is more ‘evolution-proof’ than drugs or insecticides awaits its long-term deployment in the field. However, there are grounds for optimism. Wolbachia is maintained at high frequency and retains its antiviral properties years after it is released into populations [11,14]. RNA viruses, including DCV, frequently show considerable responses to selection in short-term laboratory experiments [33,53]. Therefore, our results, together with similar work in cell culture, demonstrate that viruses do not readily evolve to escape Wolbachia's effects [14]. Nonetheless, in the wild, viral population sizes are larger and selection will act for longer periods of time, so our results may not reflect the outcome of this interaction in nature. However, Wolbachia effectively protects against viruses that will probably have experienced many decades of selection because their natural hosts are Wolbachia-infected, suggesting that our experiments may reflect the outcome of evolution in nature [17,49,50] (although it is unclear whether the symbiont is an important selection pressure in natural populations of D. melanogaster [54,55]). Long-term monitoring of field populations will be essential to test whether this reflects fundamental biological constraints that prevent viruses evolving to escape from the effects of Wolbachia.

Data accessibility

Data and R scripts are available at the Dryad Digital Repository: https://doi.org/10.5061/dryad.18j31ch [56] and Sequence Read Archive (accession number PRJEB21984) [57].

Authors' contributions

J.M., G.B.-M., S.C.L.S., J.P.D. and B.L. carried out the experimental work. J.M., R.A., G.B.-M. and F.M.J. analysed the data. J.M. and F.M.J. designed the study and drafted the manuscript. All authors gave final approval for publication.

Competing interests

The authors have no competing interests.

Funding

This study was funded by the Wellcome Trust grant no. WT094664MA (http://www.wellcome.ac.uk/) and the European Research Council (ERC) grant no. 281668 DrosophilaInfection. G.B.-M. was supported by SENACYT-IFARHU.