Mutualism Breakdown by Amplification of Wolbachia Genes

1Vertically transmitted bacterial endosymbionts are widespread in arthropods, particularly in insects. Many endosymbionts are mutualists and confer a fitness advantage to the host. The benefits may range from metabolic provisioning to protection against pathogens. Other symbionts act as parasites and manipulate host reproductive biology in order to increase the relative fitness of their carriers. In both cases, the density of endosymbionts within hosts is a crucial factor determining their prevalence in host populations.

2Symbiont densities are determined by host and symbiont genetic diversity and environment. These densities are under selection at the level of the host and the symbiont. Interestingly, there are conflicting selective forces at the level of the symbiont. Higher symbiont densities are associated with higher transmission fidelity and stronger phenotypes induced in the host. Theoretically, this should lead to a selection for higher densities. On the other hand, high symbiont levels may have a negative impact on host fitness. Since vertical transmission leads to dependence of the symbiont on the fitness of the host, it is advantageous for endosymbionts to limit their densities and consequently minimize the cost to their hosts. Thus, a key question in the field of host–microbe interactions is how symbionts regulate their replication and resulting densities to achieve an equilibrium between these opposing selective forces.

3Wolbachia is conceivably the most prevalent bacterial endosymbiont of insects, and its interactions with hosts have been studied extensively. Wolbachia is maternally transmitted and exhibits a range of phenotypes. These include cytoplasmic incompatibility and other reproductive manipulations that potentiate Wolbachia spread in host populations. Some Wolbachia strains have also been shown to be metabolic mutualists or to protect insects from viral infections. The Wolbachia strain infecting Drosophila melanogaster, wMel, exerts only a weak cytoplasmic incompatibility in laboratory conditions, and this reproductive manipulation seems not to be expressed in field conditions. Since cytoplasmic incompatibility cannot explain Wolbachia prevalence in D. melanogaster populations, it was suggested that wMel exerts positive fitness effects on its hosts. More recently, it was shown that wMel provides Drosophila with strong resistance to systemic and oral infection with the natural pathogen Drosophila C virus (DCV). This protection extends to RNA viruses of different families, indicating that wMel protects against a wide range of RNA viruses. Some of the fastest evolving genes in D. melanogaster are involved in antiviral RNA interference and are under strong positive selection. Therefore, viruses seem to be a strong selective force in D. melanogaster. Moreover, several viruses, including DCV, have been isolated from natural populations of D. melanogaster. Although there are no data regarding Wolbachia antiviral protection in natural populations, the D. melanogaster–wMel–DCV interaction fulfills many of the criteria for defensive mutualism. Therefore, antiviral protection may be the cause of wMel maintenance in D. melanogaster natural populations.

4Natural wMel variants have a small effect on host longevity, yet they provide a strong antiviral protection. This protection is positively correlated with Wolbachia density: the higher the titers of Wolbachia, the higher the antiviral protection. On the other hand, high endosymbiont densities can have a cost in the absence of viral infection, and Wolbachia variants conferring strong protection often shorten the lifespan of the flies. There is thus a fine balance between density, benefit, and cost to the host.

5The wMel variant wMelPop breaks this balance and is clearly pathogenic: it over-proliferates and dramatically shortens the lifespan of infected flies. wMelPop is, hence, an exceptional vertically transmitted symbiont. Its uniqueness was immediately recognized as providing a tool to better understand regulation of vertically transmitted symbionts and the biology of Wolbachia.

6Understanding the cause of the wMelPop phenotype and regulation of Wolbachia densities is also important from an applied perspective. Several Wolbachia strains, including wMelPop, have been transinfected into mosquito vectors of human diseases, where they can interfere with arboviruses or other pathogens. The purpose of this research is to release Wolbachia-carrying mosquitoes refractory to human pathogens into natural populations and prevent infections in humans. Aedes aegypti mosquitoes carrying Wolbachia are more resistant to dengue virus and are already being tested in the field. Different variants of wMel transinfected into A. aegypti show a trade-off between host fitness and resistance to dengue virus. A wMelPop-derived strain gives higher resistance to dengue but has a high fitness cost, which may prevent it from stably infecting natural mosquito populations. On the other hand, a wMel-derived strain confers lower protection to dengue virus but is able to stably invade A. aegypti populations. Ideally, a further understanding of the system would allow researchers to use Wolbachia strains with a better ratio of antiviral protection to cost. Moreover, since wMelPop has been transinfected into mosquitoes, understanding the pathogenicity of this Wolbachia variant is crucial for predicting wMelPop dynamics in the released mosquito populations.

7Finding the genetic basis of wMelPop pathogenicity is essential to understanding its phenotype. Difficulty in the functional analysis of Wolbachia lies in its refractoriness to genetic manipulation. Nonetheless, genomic analyses have provided insight into the cause of wMelPop pathogenicity. The first genomic map of wMelPop was published in 2003, while the full genome of the similar wMel was published in 2004. Analyses of polymorphic genomic markers and whole genome assemblies have shown that wMelPop is closely related to wMelCS variants. We have recently identified genetic differences between wMelPop and the closely related non-pathogenic wMelCS_b. The wMelPop genome contains an amplification of a ~21-kb region, named Octomom, that includes eight Wolbachia genes (WD0507 to WD0514) flanked by direct repeats. This amplification in wMelPop was also described by Woolfit and colleagues. Apart from this amplification, we found only one synonymous SNP unique to wMelPop (position 943,443, G>A). Therefore, we hypothesized that Octomom region amplification underlies wMelPop virulence. Gene amplification has previously been reported to change the pathogenicity of other bacteria and viruses.

8Here we show strong evidence that, in support of our original hypothesis, Octomom region amplification is the cause of the wMelPop phenotypes of over-replication and pathogenicity.

9Currently, Wolbachia cannot be genetically manipulated, which hinders functional studies on Wolbachia genes. However, bacterial amplified DNA sequences have been described before as unstable, leading us to test the hypothesis that natural variation in Octomom copy number exists and causes distinct phenotypes. To detect Octomom copy number variation, we tested several single Drosophila females for the copy number of the Octomom gene WD0513 in their Wolbachia bacteria (Fig. 1A). The copy number of WD0513 was determined by quantitative PCR on genomic DNA samples from single flies carrying Wolbachia, using Wolbachia wsp (Wolbachia surface protein) as a reference gene. wMelCS_b samples were used as reference samples for one WD0513 copy, based on the coverage analysis of our previous Wolbachia sequencing data. We analyzed two fly stocks infected with wMelPop: w 1118, derived from the original stock in the Benzer lab, and a DrosDel isogenic w 1118 (iso) stock into which we introgressed wMelPop from the w 1118 stock. All wMelPop samples analyzed had at least a duplication of the Octomom region, with high variation in WD0513 copy number between individual females, ranging from two to ten copies (Fig. 1A). This copy number corresponds to the average WD0513 copy number in the Wolbachia of each individual female (thus, differences in Octomom copy number between Wolbachia cells within each female may exist).

10To check whether the Wolbachia Octomom region is amplified as a unit, we tested WD0507 and WD0513 copy number simultaneously in individual flies. The copy numbers of the two genes are the same in each fly (Figs. 1B and S1), suggesting integrity of the Octomom region. A common mechanism of gene amplification in bacteria leads to tandem duplications and the formation of new junctions between units. We detected the presence of this new predicted WD0514–WD0507 junction by PCR and Sanger sequencing (Fig. 1C; S1 Text). These data show that Octomom copy number is highly variable and that the amplification is consistent with a tandem duplication.

11To test Octomom amplification’s effect on wMelPop virulence, we established Drosophila lines carrying Wolbachia with different Octomom copy numbers. Individual females with the highest and the lowest Octomom copy number were selected throughout several generations in both w 1118 and iso backgrounds (Figs. 2 and S2). Octomom copy number is heritable: Drosophila mothers carrying high-copy Wolbachia produce mostly offspring with high-copy Wolbachia, while the inverse is observed for mothers with low-copy Wolbachia. In the course of selection for low Octomom copy number in the w 1118 background, we recovered a Drosophila line carrying wMelPop with only a single copy of Octomom (Fig. 2C). This single-copy Octomom line had also lost the WD0514–WD0507 junction detected in wMelPop with multiple Octomom copies (S3 Fig.). Therefore, from generation six onwards, we maintained three selection regimes: high, two, and one Octomom copy number. The wMelPop unique synonymous SNP is present in all three selection lines, including the line carrying Wolbachia with a single Octomom copy (S4 Fig.).

12Taking advantage of the different selection lines, we compared the phenotypes of flies with wMelPop with different Octomom copy numbers. We predicted that the higher the copy number, the more severe the pathogenic phenotype, and that the one-copy Octomom line would be phenotypically identical to wMelCS_b. To perform these assays, we used the progeny of females individually tested for Octomom copy number. As Wolbachia wMelCS_b was associated with the iso fly genetic background and the one-copy Octomom line appeared only in the w 1118 background, we used hybrids between iso and w 1118 to directly compare the two (S1 and S2 Tables). All female hybrids resulting from these crosses have the same host genetic background (heterozygous between iso and w 1118) and differ in the Wolbachia inherited from the mother. Two high-copy Octomom lines, one in each genetic background, were used to control for potential host-genotype-specific maternal effects. Survival data demonstrate that differences in Octomom copy number lead to differences in host longevity: the more Octomom copies, the earlier the flies die (Figs. 3A and S5A–G). The line with one Octomom copy derived from wMelPop is indistinguishable from wMelCS_b and Wolbachia-free control (Figs. 3A and S5E–G). Even a single duplication of this region is enough to significantly shorten the host lifespan (median time to death is reduced by 39%) (Figs. 3A and S5E–G). The lifespan of flies from the two high-copy Octomom lines is further reduced, and there is no difference between these two lines (Figs. 3A and S5E–G). To further test the dependence of the phenotype on Octomom copy number, we reversed the direction of the selection in selected iso lines (choosing females with wMelPop with the highest Octomom copy number from the low-copy lines and with the lowest copy number from the high-copy lines, from generation 17 onwards) (S6A Fig.), simultaneously maintaining the forward selection regime as controls (S2 Fig.). Comparison of the lifespans of females from the forward and reverse selections confirmed that Wolbachia Octomom copy number determines wMelPop pathogenicity (Figs. 3B and S6B–D). Overall, Octomom copy number negatively correlates with longevity (S7 Fig.), and by manipulating copy number we can control Wolbachia virulence.

13We next asked whether Wolbachia growth is associated with Octomom copy number. We tested Wolbachia levels in flies carrying Wolbachia with different Octomom copy numbers over time by real-time quantitative PCR (qPCR) (Fig. 3C). The higher the Octomom copy number, the higher the density of Wolbachia. The levels are different at eclosion, and the growth of Wolbachia is faster in flies with higher Octomom copy number. Both high-copy lines have the same Wolbachia growth rate, which is higher than the Wolbachia growth rate of the two-copy line. This growth rate, in turn, is higher than that of one-copy wMelPop and wMelCS_b, which have the same Wolbachia growth rate (Fig. 3C). We confirmed this effect of Octomom copy number on Wolbachia densities by comparing Wolbachia WSP protein abundance between flies harboring wMelPop with three versus ten Octomom copies (Fig. 3D). Flies carrying Wolbachia with ten Octomom copies had more WSP protein than flies harboring Wolbachia with three copies.

14The density of Wolbachia is known to be related with Wolbachia-conferred antiviral protection, and wMelPop provides very strong protection. This protective effect is best analyzed when flies are kept at 18°C, the temperature at which wMelPop is not pathogenic. In flies that are raised from egg to adult at 25°C, Wolbachia levels at the time of infection are still related to Octomom copy number (see Fig. 3C). The survival of virus-infected flies confirmed that the higher the Octomom copy number, the stronger the antiviral protection (Figs. 3E and S5H). As with pathogenicity and growth rate, wMelPop with one Octomom copy is phenotypically identical to wMelCS_b in terms of antiviral protection.

15We showed that Octomom copy number can change rapidly under direct selection (Figs. 2 and S2). Next we questioned whether Octomom copy number would be stable if this selection were relaxed. We observed that releasing our lines from copy number selection and maintaining them at 25°C in crowded vials for five generations caused a decrease in copy number in three out of four lines tested (S8 Fig.). The only line where the copy number did not change over the five generations started with two Octomom copies. Also, examination of another wMelPop stock did not show the expected life-shortening phenotype and, accordingly, Octomom amplification (S9 Fig.). Presumably, Octomom copy number reverted to one copy, and the phenotype was lost in this stock. All these results demonstrate that wMelPop Wolbachia is genetically and, consequently, phenotypically unstable.

16Octomom amplification could promote wMelPop virulence in several ways, including via local or overall gene expression deregulation. The most parsimonious explanation, however, is that Octomom genes are overexpressed and that this causes the phenotype. Thus, we checked the expression of Octomom genes, immediately adjacent genes, and genes distant from the region by reverse transcription real-time qPCR. All Octomom genes, except WD0514, had a statistically significant higher expression in wMelPop than in wMelCS_b, but immediately adjacent genes did not (S10 Fig.). Moreover, analysis of one Octomom gene (WD0511) showed that expression level was dependent on wMelPop Octomom copy number (Fig. 4).

17Here we identify the genetic basis of Wolbachia wMelPop virulence. By selecting for Wolbachia with different Octomom copy numbers, we show a functional link between copy number and wMelPop phenotypes. The more copies of Octomom, the higher the densities of Wolbachia, and the faster the hosts die, but the stronger the antiviral protection. The evidence we provide is stronger than a simple correlation because we are controlling Octomom copy number and determining its effect. Furthermore, all wMelPop phenotypes are reverted in the line selected for one Octomom copy, establishing that Octomom copy number drives these phenotypes. There is evidence that Wolbachia levels determine the strength of the Wolbachia-associated phenotypes. Therefore, the different replication capacities of wMelPop variants with distinct Octomom copy numbers are the likely cause of the differences in the other phenotypes.

18Woolfit and colleagues also identified Octomom amplification in the D. melanogaster wMelPop genome and a deletion of the Octomom region in a mosquito-adapted wMelPop variant, wMelPop-PGYP. As wMelPop-PGYP retained a strong life-shortening effect in A. aegypti, while an A. aegypti–adapted wMel variant was benign, the authors dismissed Octomom as responsible for the high virulence of wMelPop also in D. melanogaster. We argue that the difference between wMelPop-PGYP and wMel phenotypes in mosquitoes may be due to other genetic changes accumulated during their adaptation to a new host, some already described for wMelPop-PGYP. This phenotypic difference may also exist because wMel and wMelPop belong to the two different monophyletic groups of Wolbachia from D. melanogaster: wMel group and wMelCS group. wMelCS-like variants replicate faster than wMel-like variants and sometimes shorten the lifespan of their natural host, and this difference may be exacerbated in mosquitoes. Relatedly, some Wolbachia bacteria transinfected into a new host species induce new pathogenic phenotypes.

19Amplification of Octomom is in agreement with common gene amplification by nonequal recombination in bacteria: (i) Octomom is flanked by direct repeats (see), (ii) it seems to amplify as a unit, since different Octomom genes are equally amplified in the same fly (Figs. 1B and S1A), (iii) we confirmed the predicted novel joint point (Figs. 1C and S3; S1 Text), and (iv) the amplification is unstable.

20The degree of Octomom amplification, and the associated strength of the phenotypes, can rapidly change and is fully reversible. This shows that Wolbachia can evolve rapidly, and adds to the understanding of genome evolution of endosymbionts. Many endosymbionts have evolutionarily dynamic genomes. Genomes of Wolbachia and other endosymbionts (including Hamiltonella defensa, Serratia symbiotica, Sarocladium oryzae principal endosymbiont [SOPE], and Portiera) are rich in mobile elements, prophages or phage-derived regions, and other repetitive DNA sequences. These DNA elements may mobilize, amplify, or reduce in numbers, leading to genomic changes, but they can also mediate recombination and other genomic rearrangements. Comparative genomics of some closely related endosymbionts show extensive genomic rearrangements. The same repetitive DNA elements may serve as a basis for gene amplification, as observed for Octomom. Consequently, gene copy number variation may be a common feature in these endosymbionts and may promote fast but reversible evolutionary changes. Accordingly, gene amplifications in other wMel variants, other Wolbachia strains, and the whitefly endosymbiont Portiera have previously been reported, although without any associated phenotypes.

21Genotype–phenotype links are very rarely established in endosymbionts, as many of them cannot be cultured in vitro. Previous examples include a point mutation in Buchnera aphidicola that affects thermal tolerance provided to the aphid Acyrthosiphon pisum and the loss of a prophage in Hamiltonella defensa, abrogating induced protection to parasitoids in the same aphid. The involvement of Octomom genes in Wolbachia virulence provides a unique point of entry into understanding Wolbachia–host interactions at the molecular level. As Octomom genes are overexpressed and may cause the phenotype, functional analysis of Octomom-encoded proteins is required to better understand the Wolbachia–host interaction. The Octomom region is part of the Wolbachia accessory genome since it is not present in all Wolbachia strains and shows signs of horizontal gene transfer. There are genes putatively encoding mobile elements in the flanking region (WD0506 and WD0515, in the direct repeats) and inside Octomom (WD0511). Because of its structure and associated phenotype, the Octomom region resembles bacterial pathogenicity islands. However, the pathogenicity seems to be expressed only when the region is amplified. The functions of Octomom genes are unknown and can only be speculated about based on the sequence of predicted proteins. Proteins encoded by three genes (WD0512, WD0513, and WD0514) have eukaryotic protein domains or homologs in arthropods (mosquitoes and Daphnia) and therefore may be effector proteins that interact with the host. When highly expressed, these proteins could suppress host control over the symbiont. Other genes (WD0506–WD0511 and WD0515) encode proteins that may be involved in transposition, DNA replication and repair, or transcriptional regulation. Overexpression of these proteins may increase Wolbachia’s replication rate. It is crucial to determine which of these genes are involved in the regulation of Wolbachia density and which structural characteristics of the Octomom region are important.

22Octomom copy number instability may confound past and future analyses of wMelPop phenotypes. For instance, Octomom copy number variation may have contributed to changes in wMelPop pathogenicity over time or associated with different host species or host genetic backgrounds. This instability has to be taken into consideration in future applications of wMelPop-transinfected mosquitoes to prevent transmission of human pathogens. In the dengue vector A. aegypti transinfected with wMelPop-PGYP, currently being tested in the field, Octomom copy number instability is not a factor since this region is deleted. However, wMelPop is also being transfected to other vectors of human diseases, such as the malaria-transmitting Anopheles gambiae and the dengue and chikungunya vector Aedes albopictus.

23We have shown variation in Wolbachia Octomom copy number between individual hosts within a population and across time. Genetic heterogeneity within individual hosts has been previously shown at the nucleotide level in wCer1 and wCer2. The instability of Octomom copy number suggests that there is also a high level of heterogeneity between Wolbachia bacteria within individual insects. Analysis of the dynamics and consequences of heterogeneity in gene copy numbers in somatic or germline tissues may be important to understand host–endosymbiont interactions.

24Vertically transmitted endosymbionts are subjected to different levels of selection. An increase in replication may confer a fitness advantage to the bacteria in intra-host competition but a disadvantage at the inter-host level, as it can have a high cost to the host and reduce symbiont transmission. A Drosophila line harboring wMelPop was most probably isolated in the laboratory because husbandry conditions buffered the cost to flies of pathogenic bacteria and because low host population numbers increased drift. Our results demonstrate that a single mutation (a duplication) can profoundly alter endosymbiont replication. This conversion of a mutualist into a pathogen by a single genomic event suggests that virulent mutations in microbial symbionts may be frequent and constantly counter-selected. Therefore, symbiont titers may be at a labile equilibrium achieved in the course of co-evolution and to a large extent selected at the level of the symbiont.

25D. melanogaster w 1118 stock with Wolbachia wMelPop was kindly provided by Markus Riegler and Scott O’Neill. wMelPop OPL stock was kindly provided by William Sullivan and Laura Serbus. Both wMelPop stocks are derived from Min and Benzer original stock. DrosDel isogenic background (iso) flies with no Wolbachia and with wMelCS_b or wMelPop were described before. The wMelPop and mitochondria of this DrosDel isogenic background line derive from the w 1118 stock.

26DNA was extracted from individual flies (wMelPop) or pools of ten flies (wMelCS_b controls in the selection experiments). Each fly or pool of flies was squashed in 250 μl of 0.1 M Tris HCl, 0.1 M EDTA, and 1% SDS (pH 9.0) and incubated 30 min at 70°C. Next, 35 μl of 8 M CH3CO2K was added, and samples were mixed by shaking and incubated for 30 min on ice. Samples were then centrifuged for 15 min at 13,000 rpm at 4°C, and the supernatant was diluted 100× for qPCR.

27For each sample, ten 3- to 6-d-old flies were pooled and homogenized with a plastic pestle in 1 ml of Trizol Reagent (Invitrogen). RNA was extracted according to manufacturer’s protocol and resuspended in 50 μl of DEPC-treated water (Ambion). RNA concentrations were determined using a NanoDrop ND-1000 Spectrophotometer. cDNA was prepared from 1 μg of total DNAse-treated RNA using Random Primers and M-MLV Reverse Transcriptase (all Promega). Primers were pre-incubated with template RNA for 5 min at 70°C. Next, the enzyme was added, and reactions were placed at 25°C for 10 min, 37°C for 60 min, and 80°C for 10 min. cDNA was diluted 100× for qPCR.

28The real-time qPCR reactions were carried out in the CFX384 Real-Time PCR Detection System (Bio-Rad) as described before. Briefly, each of the reactions was performed with 6 μl of iQ SYBR Green Supermix (Bio-Rad), 0.5 μl of each primer (3.6 mM), and 5 μl of diluted DNA. We performed at least two technical replicates per biological sample for each set of primers. Primer sequences were described before. The following thermal cycling protocol was applied: initial 2 min at 50°C, denaturation for 10 min at 95°C, followed by 40 cycles of 30 s at 95°C, 1 min at 59°C, and 30 s at 72°C. Melting curves were examined to confirm the specificity of amplified products. Ct values were obtained using Bio-Rad CFX Manager software with default threshold settings. Ct values were subjected to a quality check—samples with standard deviation between technical replicates exceeding one were discarded. Relative amounts of transcripts and genes were calculated by the Pfaffl method. To apply the method, the efficiency of each of the primer pairs was predetermined in a separate experiment. For the Octomom expression data, values were normalized to gmk expression. For the determination of the number of genomic Octomom copies, values were normalized to the single-copy wsp gene. For Wolbachia quantification, wsp levels were normalized to Drosophila Rpl32.

29The WD0514–WD0507 junction was amplified using specific primers (Link_seq_1 and Link_seq_2), and Sanger sequencing was performed with these primers and the primers annealing inside the junction (Link_seq_3–7) by Source Bioscience. Primer sequences are listed in S3 Table.

30Selection for high- and low-copy Octomom wMelPop lines in w 1118 and iso backgrounds was initiated with females from a single vial of each background. For each background, ten single females were separated into individual vials and allowed to lay eggs for 5 d before being sacrificed for determination of Wolbachia WD0513 copy number. The offspring of the female with the highest and the lowest Octomom copy number was used to start the next generation. This general procedure was repeated at every generation of selection. Three replicates of high- and low-copy Octomom selection lines for each background were established at generation two. From that point on, we selected one female/line/generation with the desired Octomom copy number (based on real-time qPCR). Female age for egg laying (0–2 d) and qPCR (5–7 d) was controlled from generation four and two for the iso and w 1118 lines, respectively. At generation seven of the w 1118 lines, we started to also select for one-copy Octomom wMelPop. At this point we selected the female with Wolbachia with WD0513 copy number closest to one for this selection regime, and the female with Wolbachia with WD0513 copy number closest to two for the low-copy Octomom lines.

31From generation two to generation 13 of the w 1118 selection lines and from generation two to generation 22 of the iso selection lines, we were selecting from among six to ten females. From these generations on, we selected from three females per line.

32At generation 14 of the w 1118 lines and generation 18 of the iso lines, the selection was not performed.

33For phenotypic analyses of flies carrying wMelPop with different Octomom copy numbers, single females were placed in vials, allowed to lay eggs for 5 d, and sacrificed to determine WD0513 copy number. The progeny of females carrying Wolbachia with the specified Octomom copy numbers were selected for the phenotypic analyses. All lifespan assays were performed at 25°C and 29°C, the temperature regimes applied in the first report on wMelPop phenotypes.

34In order to directly compare flies with wMelPop with the full range of Octomom copy numbers, flies with wMelCS_b, and flies without Wolbachia, we used hybrids between w 1118 and iso genetic backgrounds (S1 and S2 Tables). Females with the desired Wolbachia status, which is transmitted to the next generation, were crossed with males from the other genetic background. Since females were used in the phenotypic analyses, their genetic backgrounds were all equal and heterozygous between w 1118 and iso, irrespective of the direction of the crosses. The mitochondria from these two lines should be identical since they share a very recent common ancestor. We used females with high Octomom copy number from both genetic backgrounds to control for the possible influence of the direction of the cross and maternal effects potentially associated with different backgrounds.

35Females whose mothers’ Octomom copy number was assessed by qPCR were collected at eclosion (ten per tube), allowed to mate for 24 h (five males per tube), separated from males, and either checked for survival at 25°C or 29°C every day or kept at 25°C and sacrificed at the indicated time points for Wolbachia density quantification. Females were maintained on a standard cornmeal diet without live yeast and were passed to fresh vials every 3 d. The mothers of females used for phenotypic analyses were derived from selection lines at the generations indicated in S2 Table.

36DCV was produced and titrated as described before. Infections were performed by pricking 1- to 2-d-old female flies with virus at 109 TCID50 (median tissue culture infectious dose)/ml. After infection, flies were kept in vials without live yeast, ten flies per vial, at 18°C. It has been shown previously that wMelPop is not pathogenic to the flies at this temperature. Flies were checked for survival daily and passed to fresh vials every 5 d.

37Survival data were analyzed by Cox proportional hazard mixed effects models. Octomom copy number was considered a fixed effect, and replicate tube (containing ten flies) within the same experiment was considered random. Model fitting was done using the coxme package in R. Tukey´s test was applied for pairwise comparisons of Cox hazard ratios between flies with all wMelPop lines, flies with wMelCS_b, and flies without Wolbachia.

38Analysis of growth curves of wMelPop lines with different Octomom copy number was performed with log-linear model fits (lm in R). The slopes of different fitted regression lines were compared and corrected for multiple comparisons (Bonferroni correction).

39Spearman correlation between Octomom copy number and median time to death was performed in R (cor.test).

40Comparison of the expression of several Wolbachia genes between wMelCS_b and wMelPop (S10 Fig.) was done with the t-test in R (t.test) and was corrected for multiple comparisons with the Bonferroni correction.

41Comparison of wsp and WD0511 gene expression between fly lines carrying different Wolbachia (Fig. 4) was performed with a log-linear model fit (lm in R), and the different lines were compared pairwise with a Tukey’s test.

42Ten mated females from high- and low-copy iso selection lines, whose mothers were individually tested for Octomom copy number, were aged for 10 d before protein extraction. Flies without Wolbachia were used as a negative control. Anti-WSP rabbit polyclonal antibody was kindly provided by Kostas Bourtzis and pre-absorbed in fixed Wolbachia-free D. melanogaster embryos. Anti-beta-tubulin mouse monoclonal E7 antibody was acquired from the Developmental Studies Hybridoma Bank.