The Wolbachia strain wAu provides highly efficient virus transmission blocking in Aedes aegypti

Introduced transinfections of the inherited bacteria Wolbachia can inhibit transmission of viruses by Aedes mosquitoes, and in Ae. aegypti are now being deployed for dengue control in a number of countries. Only three Wolbachia strains from the large number that exist in nature have to date been introduced and characterized in this species. Here novel Ae. aegypti transinfections were generated using the wAlbA and wAu strains. In its native Ae. albopictus, wAlbA is maintained at lower density than the co-infecting wAlbB, but following transfer to Ae. aegypti the relative strain density was reversed, illustrating the strain-specific nature of Wolbachia-host co-adaptation in determining density. The wAu strain also reached high densities in Ae. aegypti, and provided highly efficient transmission blocking of dengue and Zika viruses. Both wAu and wAlbA were less susceptible than wMel to density reduction/incomplete maternal transmission resulting from elevated larval rearing temperatures. Although wAu does not induce cytoplasmic incompatibility (CI), it was stably combined with a CI-inducing strain as a superinfection, and this would facilitate its spread into wild populations. Wolbachia wAu provides a very promising new option for arbovirus control, particularly for deployment in hot tropical climates.


Introduction
The mosquito Aedes aegypti (Linneaus) is the most important vector of human arboviruses. Although native to Africa it now has a broad distribution throughout the tropics and subtropics and is peridomestic, often laying its eggs in man-made water containers, and displaying a strong preference for feeding on humans. Attempts to reduce the global incidence of dengue fever and stem the spread of recent chikungunya, Zika and yellow fever virus outbreaks have focused on Ae. aegypti control [1,2], which has proven challenging. An emerging vector control strategy utilizes mosquitoes artificially transinfected with virus-blocking strains of the alpha-proteobacterium Wolbachia pipientis [3]. Wolbachia are obligate intracellular endosymbionts naturally found infecting a wide range of terrestrial arthropods. The natural abundance of Wolbachia can be partly attributed to its capacity to spread through naïve populations by manipulating host reproduction. Although several forms of reproductive manipulation are found across different arthropod species, the only form observed in mosquitoes is a type of crossing sterility known as cytoplasmic incompatibility (CI). Wolbachia modifies the sperm of infected males [4], which results in the generation of non-viable progeny when mated to uninfected females. Infected females, in contrast, 'rescue' this sperm modification, producing viable progeny and resulting in a relative fitness advantage that can drive and maintain Wolbachia at high population infection frequencies [5].
While Ae. aegypti is not a natural Wolbachia host, stable transinfections with the wAlbB strain from Aedes albopictus and wMelPop/wMel strains from Drosophila melanogaster have been generated in the laboratory using embryonic microinjection, with the resulting lines showing reductions in vectorial capacity for a number of arboviruses and other pathogens [6][7][8][9][10][11]. Ae. aegypti transinfected with wMel have significantly reduced vector competence for dengue virus [7,12], yellow fever virus [10], chikungunya [10] and Zika [13] viruses in laboratory challenges. However, mosquito challenges with patient-derived dengue infected blood have indicated that wMel-mediated blocking is incomplete, and modelling predicts that wMel would be insufficient to achieve complete control in some settings [14]. Field trials aimed at spreading Wolbachia in Ae. aegypti for dengue control have to date focused primarily on wMel [15,16].
Different strains of Wolbachia reach varying intracellular densities and display divergent tropism within host tissues; the magnitude of the pathogen inhibition effect shows a positive correlation with Wolbachia intracellular density in several species [17][18][19]. The wMelPop strain reaches very high densities in Ae. aegypti, which probably contributes to an almost complete blocking of dengue virus transmission [6,12]. However, wMelPop imposes significant costs on a variety of traits including reduced longevity, fecundity and egg survival in quiescence [20][21][22][23]. These negative fitness effects have made the introduction of wMelPop into wild host populations problematic, despite the presence of strong uni-directional CI-recent field trials in Vietnam and Australia failed to achieve population replacement using this strain [24].
Recently several studies have reported the influence of a variety of factors on Wolbachia intracellular density. Larval rearing temperature in particular has a significant impact on the densities of wMel and the over-replicating wMelPop strain in Ae. aegypti [25,26]: exposure of larvae to diurnal rearing temperatures cycling between 27-37˚C resulted in dramatic reductions in total Wolbachia density, and rates of maternal transmission-ultimately leading to the loss of the wMel and wMelPop infections when the high temperature regimes were maintained for more than one generation [26]. In addition to environmental factors, a genetic basis to density determination has been postulated based on duplications of a set of eight genes in the wMelPop genome, with copy number reported to correlate with wMelPop density in Drosophila melanogaster [27]. However, further studies failed to find a straightforward causal role for copy number in Wolbachia density regulation [28], but see [29] and [30].
So far, only a few of the vast repertoire of naturally-occurring Wolbachia strains have been introduced into Ae. aegypti. It is important to create and characterize further transinfections in this species since they might provide improved characteristics such as viral blocking under particular environmental conditions, especially in hot climates [3], and offer insights into the regulation of intracellular density and its role in inducing pathogen inhibition and effects on host fitness. Limitations are imposed by the technical demands of embryo cytoplasmic transfer by microinjection and the need for robust lab colonies of the insects to be used as the source of Wolbachia. While wAlbA and wAlbB are naturally found superinfecting Ae. albopictus, wAlbA is maintained at around 10% of the density of wAlbB [31] and only wAlbB established itself following previous embryo cytoplasm transfers from Ae. albopictus into Ae. aegypti [32]. Strain wAu does not induce CI in its native host Drosophila simulans [33], but confers a notably high degree of protection from pathogenic viruses of Drosophila [34,35]. We therefore aimed to generate and characterize Ae. aegypti lines containing wAlbA and wAu for a variety of traits relevant to transmission-blocking and population-replacement potential, in comparison with the previously reported wAlbB and wMel transinfections.

Strain generation and transmission
Using embryonic cytoplasmic transfer and taking advantage of incomplete maternal inheritance we generated Wolbachia transinfected lines carrying strains wAlbA, wAlbB, wMel, and wAu in the same host background of Ae. aegypti. Each of the Wolbachia strains apart from wAu was capable of inducing full unidirectional cytoplasmic incompatibility with wild-type mosquitoes, and therefore showed population replacement potential. wAu produced no detectable CI (Table 1), consistent with observations in its native host (Drosophila simulans)  [33] and providing further evidence that it is genetically incapable of generating CI, as opposed to a strain-specific suppression of the phenotype in its native host. Wolbachia-infected lines were crossed to determine the crossing types between strains. For lines that induced unidirectional CI with wild-type mosquitoes, no hatching of the resulting eggs was observed, in other words between-strain crosses resulted in complete bidirectional CI. Rates of Wolbachia maternal inheritance were determined by PCR of progeny from compatible crosses between wild-type males and infected females. All lines showed complete (100%) maternal transmission of all strains in 200 progeny assessed. Since wAu does not induce CI, its maintenance in the wAu line is facilitated by high rates of maternal inheritance, and it is hypothesized to produce positive host fitness effects under some conditions based on increases in its frequency in native D. simulans host populations [36]. To assess its stability in Ae. aegypti populations, 200 individuals from the wAu colony were randomly selected and tested for the presence of Wolbachia at the fourth, seventh, and tenth generations post initial establishment. Colonies of this line had been maintained at relatively high numbers (>2,000 individuals per generation from G 4 ) with no direct selection for wAu infection from G 1 onwards. All individuals tested positive at each generation, indicating that wAu is maternally transmitted at very high fidelity under these laboratory conditions.
Wolbachia intracellular density and tropism. Total Wolbachia density in each line was monitored over the initial post-transinfection generations by qPCR. Once densities were stable (after five generations for each line) a time-course study was performed to monitor total densities in females over the first 15 days post adult eclosion ( Fig 1A). Although all lines showed increasing Wolbachia density with adult age, there was significant variation in the total densities of the individual Wolbachia strains.
Although wAlbA reaches only modest densities in its native host Ae. albopictus, it showed the highest overall density of all the strains assessed at the majority of time-points in Ae. aegypti. wAu also reached comparatively high densities (no significant difference with wAlbA at all time points, p>0.01, t-test), several fold higher than that of the other native Drosophila strain, wMel. wAlbB showed the lowest total Wolbachia density of the strains at all time points.
It is highly likely that Wolbachia tissue distribution plays a key role in determining levels of pathogen inhibition since this phenotype has been reported to be cell-autonomous rather than systemic [6]. Total Wolbachia densities were therefore assessed in dissected ovary, midgut and salivary gland tissues (Fig 1B). In ovarian tissue wAu and wAlbA reached similarly high densities (p = 0.192, t-test) with wAlbB showing the lowest density. In midguts all strains showed relatively low densities compared to the other tissues assessed; in salivary glands wAlbA showed the highest density while wMel and wAlbB reached relatively low density.
Virus inhibition. To provide an initial indication of the virus blocking potential of the different Wolbachia single-infection lines in Ae. aegypti, the titres of Semliki Forest Virus (SFV), an arbovirus model system, were assessed in whole adult females following intrathoracic microinjection and a ten-day incubation period (Fig 2A). wAu was more effective in reducing viral load than wMel and wAlbB, although all three strains resulted in significantly reduced viral loads compared to wild-type (p<0.01, 1-way ANOVA with Dunnett's). It is notable that the highest density strain (wAlbA) was not the most efficient virus blocker, producing no detectable differences in levels of SFV compared with wild-type mosquitoes (p>0.3, 1-way ANOVA with Dunnett's) To further assess levels of virus transmission blocking, females of the wild-type, wAu, wMel and wAlbB lines were orally challenged with either a DENV or ZIKV-infected blood meal. Engorged 5-day old females were allowed to incubate virus for 12 days, at which point salivary glands and abdomens were dissected and the presence of viral RNA was quantified by reversetranscriptase quantitative PCR. Additionally, saliva was collected prior to dissection for the ZIKV infected females, and levels of infectious virus were quantified by fluorescent focus assay (FFA).
For DENV, significant differences in rates of replication and dissemination were observed across the different Wolbachia lines. Females of the wAu, wAlbB, wMel and wild-type lines contained salivary glands positive for DENV RNA, at rates of 0%, 26.3%, 36.8% and 41.2%, respectively ( Fig 2B). This represents a significant reduction in infection rate of the wAu infection compared to wild-type, and for wAu compared to the wAlbB and wMel lines. In abdomen tissue, infection rates were 22%, 47.4%, 31.6% and 58.8% in the wAu, wAlbB, wMel and wildtype lines, respectively, and were significantly different between the wAu and wild-type lines ( Fig 2C).
All of the Wolbachia infected lines showed significant decreases in ZIKV transmission in saliva compared to wild-type: the wAlbB and wAu lines completely blocked infectious virus transmission, while 6.3% of the wMel saliva samples were positive, compared to 39% of the wild-type ( Fig 2D). Similarly, the wAlbB and wAu females contained no detectable ZIKV in salivary gland tissue, while 12.5% and 50% of the wMel and wild-type were positive, respectively, representing a significant decrease for the wAu and wAlbB infected lines compared to wild-type (S1 Fig). In abdomen tissue 27.7%, 37.5% and 72.2% were ZIKV positive in the wAlbB, wMel and wild-type lines, respectively, while none of the wAu abdomens were positive for ZIKV RNA (Fig 2E).
Effects of high temperature on Wolbachia density. Recent reports have shown that exposure of larvae to higher rearing temperatures can significantly affect the density of Wolbachia in adults [25,26], leading to reduced rates of maternal transmission [26]. As Ae. aegypti larvae are often found in bodies of water experiencing day-time heating, temperature susceptibility could potentially limit the invasive capacity of a Wolbachia strain, as well as its ability to inhibit virus transmission. Interestingly, comparisons of the densities of wAlbB and wMel in Ae. aegypti found that the latter was particularly susceptible to cyclical heat treatment, while wAlbB was more resilient [26].
The densities and maternal transmission rates of Wolbachia in the wAlbA, wAlbB, wMel and wAu lines were examined in adults following exposure of larvae to a rearing temperature regime fluctuating between 27˚C and 37˚C (12hr:12hr). Consistent with previously published results [25,26], fluctuating heat regimes during larval rearing resulted in reductions in Wolbachia density in emerging adult males and females. Although significant decreases in density were observed for all strains under heat treatment (p<0.05 for each comparison, t-test) ( Fig  3A), the effect was most dramatic for wMel, with a drop in average Wolbachia levels in females to 0.017 ± 0.015 (mean ± SD) Wolbachia/host cell, an infection density 0.49% that of newly emerged wMel controls raised at a constant 27˚C. wAu, another native Drosophila strain,

Fig 2. Virus inhibition in Wolbachia-infected Aedes aegypti lines. (A) Semliki
Forest virus (SFV) genome copies per host cell following thoracic injection into Wolbachia-infected lines and wild-type Ae. aegypti. Females were left for 10 days prior to total RNA extraction and virus quantification by qPCR. Levels of target RNA sequences were normalized against the RPS17 house-keeping gene. 17, 16, 18, 17 and 17 females were PCR'd for the wAlbA, wMel, wAu, wAlbB and wt, respectively. Statistical analysis was performed using a one-way ANOVA with a Dunnett's post-hoc test. Dengue-2 (DENV) (B and C) and Zika (ZIKV) (D and E) viruses were orally administered to 5-day old females. After an incubation period of 12 days, females were salivated (Zika only) and salivary glands and abdomens dissected. Viral RNA in salivary glands (SG) and abdomens were quantified by reverse-transcriptase qPCR, with viral RNA levels normalized to host RNA using the RpS17 house-keeping gene. A value of zero for normalized virus levels, indicates no amplification for virus cDNA in that sample. Zika viral titers in saliva were quantified by fluorescent focus assay with results show focus forming units (FFU). Proportions underneath each graph indicate the infection rate for a given strain. Statistical analyses for panels B, C, D and E were performed using a one-tailed Fisher's exact test comparing rates of virus-positive to virus-negative samples. Black lines indicate median of non-zero values.
https://doi.org/10.1371/journal.ppat.1006815.g002 proved to be more resilient to high temperature treatment than wMel, retaining an infection density of 4.6 ± 1.31 (mean ± SD) Wolbachia/host cell in emerging females, 4.47% that of wAu controls. The two native mosquito strains tested, wAlbA and wAlbB, proved to be the most resilient to high temperature fluctuation, retaining infection densities of 4.9 ± 4.4 (mean ± SD) and 0.2 ± 0.06 (mean ± SD) Wolbachia/host cell in females, representing 10.72% and 32.12% that of controls, respectively. Strain-specific reductions in Wolbachia density were found to be similar in males and females (Fig 3A).
Despite the reductions in density, the females in all lines displayed complete transmission to progeny (Fig 3B). The effects of two consecutive generations of larval temperature treatment on infection rates was also examined. Females having undergone larval heat treatment were crossed with wild-type males to avoid the effects of CI, and the resulting progeny were also reared under fluctuating high temperature to pupation, at which point infection status was assessed by PCR. We found that each strain contained uninfected individuals, although rates varied widely between Wolbachia strains (Fig 3B). The wMel line showed an almost complete loss of detectable infection, with only 2% of individuals PCR positive for Wolbachia; infection rates were 58%, 66% and 82% for the wAlbB, wAlbA and wAu strains, respectively. These results are consistent with previous findings showing complete loss of wMel and wMelPop infections in Ae. aegypti following two generations of heat cycling [26].
Wolbachia strains and host fitness. Another important factor when assessing the comparative utility of Wolbachia strains for disease control is their effect on host fitness [37]. The dynamics of cytoplasmic incompatibility, where reproductive advantage afforded to Wolbachia-infected females is frequency-dependent, dictate that invasion cannot occur until Wolbachia frequency has exceeded a threshold level, and this threshold is in part determined by the effects of the bacteria on host fitness. We assessed important fitness parameters previously shown to be influenced by Wolbachia infection: longevity (Fig 4A), fecundity ( Fig 4B) and egg hatch following a period of desiccated quiescence (Fig 4C).  Out of all the lines tested, wMel was the only Wolbachia strain that did not cause a significant reduction in adult female longevity compared to uninfected wild-type in these laboratorycage experiments (p>0.1). wAlbB caused a slight but significant reduction in the longevity of females. The wAu and wAlbA infected lines resulted in the most significant reductions in female longevity, consistent with the hypothesis that the higher density Wolbachia strains cause the highest fitness costs (p<0.001). Similar trends in longevity were observed in males (S2 Fig). No influence of any of the Wolbachia strains on female fecundity was detected when compared to wild-type hatch rates (p = 0.91, 1-way ANOVA with Dunnett's) (Fig 4B), with all strains producing approximately 90 eggs per female on average. A strain-dependent effect on the ability of eggs to survive in desiccated quiescence was observed. The wild-type, wMel and wAlbB lines showed no significant reductions in egg hatch rates after 50-days of quiescence compared with their respective 5-day hatch rate (p>0.4 for each comparison, t-test). However, the wAu and wAlbA containing lines showed reductions in egg survival with time over this period. The effect was strongest in the wAlbA line, with hatch rates dropping to 50% after approximately 13 days of desiccated quiescence. The hatch rate of the wAu infected line dropped to 50% after approximately 18 days.
A wAu-wAlbB superinfection causes uni-directional CI. The wAu strain was able to successfully establish itself in wild populations of Drosophila simulans without inducing CI [36]. Given that densities and associated fitness costs in Ae. aegypti are likely to be lower in wild mosquitoes than in the lab, and that its maternal transmission fidelity is very high, it is possible that wAu could maintain itself in field Ae. aegypti following introduction. Alternatively, it could be driven into an uninfected population by combining wAu with a Wolbachia strain capable of causing unidirectional CI. As a proof-of-concept, we created a wAu superinfection using the wAlbB strain as the 'driver', since wAlbB combines unidirectional CI with temperature stability and relatively strong viral inhibition. A superinfected line was generated by transferring cytoplasm from wAu-carrying Ae. aegypti embryos to embryos carrying wAlbB. As expected, the wAuwAlbB line produced full unidirectional CI when crossed with wild-type mosquitoes (Fig 5A). Analysis of adult females showed that the wAuwAlbB line possesses very similar over-all densities to the wAu single-infection ( Fig 5B). Moreover, a comparison of the wAu, wAlbB and wAuwAlbB lines revealed that the presence of wAlbB did not significantly reduce the density of wAu in wAuwAlbB ovaries (Fig 5C) (p>0.1, t-test), suggesting that wAlbB will not affect the maternal transmission rate of wAu in the superinfected line, and no reductions in wAu density were found in the midgut or salivary gland tissues of the wAuwAlbB line compared to the wAu-only line (S3 Fig), strongly suggesting that wAuwAlbB will display a similar virus blocking and fitness profile to the wAu-only line. However, wAlbB ovary density was significantly reduced in the presence of wAu (Fig 5C) (p<0.03, t-test), although this does not appear to impact the capacity of wAlbB to rescue CI (Fig 5A). The infections in the wAu and wAlbB single and superinfected lines were visualized by whole-mount fluorescence in situ hybridization, using separate wAu (green) and wAlbB-specific (red) probes ( Fig 5D). The images obtained show that wAu is present in a greater number of ovarian cells and occupies a greater volume within the cells compared with wAlbB, which is notably more restricted in its distribution.

Discussion
In light of the failure to establish wMelPop in wild populations [24], and with the finding that wMel densities are unstable under high temperature treatments [25,26], it is important to investigate the properties of additional Wolbachia strains in Ae. aegypti. The novel lines generated here highlight the variability in phenotypic effects caused by different Wolbachia strains in a common host background, and emphasizes the difficulties in making reliable predictions of phenotype based solely on observations of the strain in single host species.
The high level of virus inhibition by wAu observed here is consistent with results obtained in Drosophila species. A comprehensive assessment of Wolbachia-mediated antiviral protection, comparing Drosophila C virus (DCV) and Flock House virus (FHV) inhibition by 19 Wolbachia strains in same background of D. simulans, found that wAu caused the strongest blocking of both viruses, greater than both wMel and a higher density wMel variant, wMelCS [38]. Similar observations of stronger blocking of DCV and FHV by wAu relative to wMelCS have also been made in D. melanogaster [34]. Although wAu produces some costs to host fitness, these were modest compared to the wMelPop strain. In previous experiments carried out by McMeniman and colleagues [23], the median longevity of wMelPop-infected females was found to be approximately 26 days, compared to 62 days for wild-type controls. The wAu line produced median female longevity of 40 days, compared to 60 days for wild-type controls.
It was surprising to find significantly higher densities of wAlbA than wAlbB in Ae. aegypti, given that wAlbA is maintained at much lower densities than wAlbB in its native host Ae. albopictus [31], and is strongly suggestive of the presence of host factors/interactions determining Wolbachia density in a strain-specific fashion, rather than simple differences in replication rates between Wolbachia strains. The influence of host factors has been previously suggested, when densities of wMelPop were found to vary significantly between the native host D. melanogaster and the closely related Drosophila simulans [39]. Studies comparing wMel with the high-density variant wMelPop have correlated duplications of a region of eight genes with increases in Wolbachia density in the native host Drosophila melanogaster [27,29,30]. However, this region is completely deleted in other wMelPop sub-strains [40]. Moreover, wAu lacks this locus [34], but reaches higher densities than wMel and provides greater pathogen protection in D. simulans [35]. The apparent strain-specific nature of Wolbachia density control is encouraging in terms of maximizing the long-term effectiveness of Wolbachia-based strategies for virus control. Even if mean Wolbachia density reduction occurs over time due to selection on the host, and thus amelioration of virus transmission-blocking, other strains could subsequently be introduced to restore high density and thus the effectiveness of disease control.
We show that the effects of high temperature on density can vary dramatically between Wolbachia strains, and confirm previous studies showing that wMel is particularly susceptible to maternal leakage over consecutive generations of heating. The upper temperature used here is high but realistic for larvae in tropical regions [41]. Choosing Wolbachia strains that show the greatest density stability of natural environments where releases take place should therefore be a key concern when considering the suitability of strains in a given location. Higher temperature conditions may result in lower Wolbachia densities in the field, which could cause reduced pathogen inhibition. However lower densities also correlate with lower fitness costs; high temperatures may therefore also result in improved fitness characteristics and population spread capacity. Likewise, in hot tropical regions without a marked dry season, reduced embryo hatch after quiescence may have little impact on spread dynamics. The direct comparison of Wolbachia strains presented here also highlights the utility of wAlbB, which combines similar levels of virus blocking to wMel, with greater temperature stability-suggesting it may be more effective at spreading and blocking virus transmission in very hot climates.
The demonstration of a stable superinfected line carrying wAu and wAlbB demonstrates one of several possible methods by which wAu could be spread through populations. When used in combination with a 'driver' strain, there is always the risk that a decoupling of the strains may occur over time in the field, although the rate at which this would occur is difficult to predict, and may vary between environmental conditions and thus locations. Further experiments can explore different strain combinations with wAu to maximize co-transmission stability under field-approximating conditions, but the driver strain should also reduce or block virus transmission in case wAu is lost, as is the case for wAlbB. The combination of two Wolbachia strains was previously reported in Ae. aegypti, where wAlbB was stably combined with wMel, resulting in a superinfected strain that showed unidirectional CI with wt, wAlbB-only and wMel-only lines [12]. Interestingly the superinfected line showed increased levels of pathogen inhibition compared to the constituent strains.
The recent discovery in wMel that at least two of the genes required for CI induction form an operon located in an integrated WO prophage region [42,43], which is notably absent in the wAu genome [44], opens the intriguing possibility that wAu could be converted into a CI-inducing stain following integration of a suitable WO phage element. Crossing-type conversion has been previously reported in Nasonia wasps, whereby an incompatibility phenotype was transferred between different strains following innoculation with a 0.23μm porefiltred pupal homogenate [45]. Overall fitness benefits are also possible under some field conditions, perhaps including protection from harmful viruses, as hypothesized for wAu in its native host Drosophila simulans-where it is capable of spreading and maintaining high population infection frequencies [36]. Little is known about the frequency of natural entomopathogens to which wAu could provide protection in Ae. aegypti. wAu could also potentially be spread through a mosquito population by applying suitable selection pressures such as bacterial, fungal or viral entomopathogens; Wolbachia wMelPop has for example been shown to provides resistance to several such agents [46]. In addion to the applied potential of wAu, the differences in virus inhbiiton between wAu and wAlbA, despite reaching similar densities in Ae. aegypti, provide excellent in vivo systems for comparative studies to better understand the mechanistic basis of the phenotype.

Mosquito strains and rearing
The Ae. aegypti wild-type line used was colonized from Selangor State, Malaysia in the 1960s. All mosquito colonies were maintained at 27˚C and 70% relative humidity with a 12-hour light/dark cycle. Larvae were fed tropical fish pellets (Tetramin, Tetra, Melle, Germany) and adults were given access to a sucrose meal ad libitum. Blood meals were provided using a Hemotek artificial blood-feeding system (Hemotek, UK) using defribrinated sheep blood (TCS Biosciences, UK). Eggs were collected by providing damp filter-paper (Grade 1 filter paper, Whatman plc, GE healthcare, UK) for oviposition. Eggs were desiccated for 5-10 days prior to hatching in water containing 1g/L bovine liver powder (MP Biomedicals, Santa Ana, California, USA).

Generation of Wolbachia-infected lines
wMel, wAlbA and wAlbB Ae. aegypti lines were generated by transferring cytoplasm from superinfected Ae. albopictus (origin Indonesia) embryos carrying wMel, wAlbA and wAlbB to wild-type Ae. aegypti embryos. Microinjections were performed using methods described previously [17]. Female G 0 survivors were back-crossed to wild-type males, blood-fed and separated individually for oviposition. G 0 females were analysed for Wolbachia infection by strain specific PCR (see primer table in Supporting Information for sequences) and eggs from Wolbachia negative G 0 females were discarded. Eggs of positive females were hatched and G 1 's were assessed for Wolbachia G 0 -G 1 germ-line transmission. Injections from the superinfected Ae. albopictus line initially resulted in the generation of a triple-infected Ae. aegypti line (wMel-wAlbAwAlbB), which showed unstable maternal inheritance of Wolbachia strains. Individualizing the progeny of triple infected females resulted in the isolation and establishment of the wAlbA-only, wAlbB-only and wMel-only lines. The wAu line was generated as above, but involved transfer of cytoplasm from wAu infected Drosophila simulans embryos (origin Australia). The wAuwAlbB line was generated as above but involved the transfer of cytoplasm from the wAu-infected Ae. aegypti line into embryos of the wAlbB-infected line.

Maternal inheritance and CI
To assess rates of maternal inheritance, females from each Wolbachia transinfected line were crossed to wild-type males in pools of 20 males and 20 females. A blood-meal was provided and females were individualised for oviposition. The resulting eggs were hatched and DNA from a selection of 10 of these (200 assessed for each line in total) was extracted at the pupal stage and a PCR for Wolbachia was performed.
Rates of CI induction and rescue both with wild-type mosquitoes and between infected lines were assessed by crossing 20 males and 20 females of each line. A blood-meal was provided and females were individualised for oviposition. Eggs were collected on damp filter paper, which was subsequently desiccated for 5 days at 27˚C and 70% relative humidity. Eggs were counted and hatched in water containing 1g/L bovine liver powder. Larvae were counted at the L2-L3 stage to provide hatch rates. Females with no egg hatch were dissected to check spermathecae for successful mating. Unmated females were excluded from hatch rate evaluations.

Wolbachia strain-specific PCR and density qPCR
For PCR analysis, genomic DNA was extracted from mosquitoes using the Livak method [47]. For primer sequences see primer table in supporting information. For measurements of Wolbachia density by qPCR, genomic DNA was extracted from mosquitoes using phenol/chloroform. Unless stated otherwise, mosquitoes used in density experiments were adults 5-days post pupal eclosion. gDNA was diluted to 100ng/μl using a NanoDrop spectrophotometer (Thermo Scientific, Waltham, Massachusetts, USA). A BioRad CFX-96 real-time PCR detection system was used (Bio Rad, Hercules, California, USA) with 2 x SYBR-Green mastermix (Biotool, Houston, Texas, USA). Total Wolbachia density was analysed by absolute quantification against a dilution curve of a vector containing single copies of the homothorax (HTH) gene and Wolbachia surface protein (wsp).

Fluorescent in situ hybridization
Ovaries were dissected from 5-day old adult females in a drop of PBS buffer, and were immediately transferred to a tube containing Carnoy's fixative (chloroform:ethanol:acetic acid, 6:3:1) and fixed at 4˚C overnight. Samples were then rinsed in PBS and transferred to a 6% hydrogen peroxide in ethanol solution for 72 hours at 4˚C. Samples were then incubated in a hybridization solution containing: 50% formamide, 25% 20xSSC, 0.2% (w/v) Dextran Sulphate, 2.5% Herring Sperm DNA, 1% (w/v) tRNA, 0.015% (w/v) DTT, 1% Denhardt's solution, and 100ng/ml of each probe. Probe sequences were as follows: wAu (green) 5'-ACCTGTGTGAAA CCCGGACGAAC-(Alexa flour 488)-3'; wAlbB (Red) 5'-TAGGCTTGCGCACCTTGCAGC-(Cyanine3)-3'. Samples were left to hybridize overnight in a dark-damp box at 37˚C. Samples were washed twice in a solution containing: 5% 20xSSC, 0.015% (w/v) DTT, and twice in a solution of 2.5% SSC, 0.015% (w/v) DTT in dH2O, with each wash performed at 55˚C for 20 minutes. Samples were then placed on a slide containing a drop of VECTASHIELD Antifade Mounting Medium with DAPI (Vector Laboratories, California, USA) and were visualized immediately using a Zeiss LSM 880 confocal microscope (Zeiss, Oberkochen, Germany). Both the red and green probes were added to the hybridization solution to produce the images of wAlbB, wAu, wAuwAlbB and wild-type ovaries.

Thoracic injection of SFV
Twenty 5-day old female mosquitoes of each Wolbachia-infected line and wild-type were injected with the respective virus in the thorax using a pulled glass capillary and a Nanoject II (Drummond Scientific, Pennsylvania, USA) hand-held microinjector. Injected mosquitoes were immediately transferred to an incubator set at 27˚C and a 12-hour light/dark cycle for recovery. SFV injected females were left for ten days prior to RNA extraction and virus quantification by qRT-PCR. RNA was extracted using TRI Reagent (Sigma-Aldrich, Missouri, USA). cDNA was synthesized using 1μg of total RNA and the All-In-One cDNA Synthesis SuperMix (Biotool, Houston, Texas, USA). qRT-PCRs were performed on a 1 to 20 dilution of the cDNAs. Virus levels were normalized to the RPS17 house-keeping gene.
Semliki Forest virus was sub-type C (catalogue number 1112041v) obtained from Public Health England culture collections. SFV was propagated on C6/36 cells to a final injection concentration of 1.78x10 13 FFU/ml. Primers used for viral detection were: SFV4-F and SFV4-R.

Oral infections with ZIKV and DENV
Five day-old females were fed an infectious blood-meal containing a mixture of 800μl defibrinated sheep blood and 400μl viral suspension supplemented with a phagostimulant (ATP) at a final concentration of 5mM. Dengue virus was serotype 2, New Guinea C strain, obtained from Public Health England culture collections. Zika virus was strain MP1751, obtained from Public Health England culture collections. The final concentration of dengue virus in the blood meal was 8.3x10 7 FFU/ml. The final concentration of Zika virus in the blood meal was 1.6x10 8 FFU/ml. Engorged females were separated and maintained in a climactic chamber at 27˚C and 75% humidity. After 12 days females were salivated by inserting the proboscis into a capillary containing mineral oil and placing a drop of 1% pilocarpine nitrate onto the thorax. Collected saliva was ejected into tubes containing Dulbecco's Modified Eagle Medium (DMEM) medium supplemented with 2% fetal bovine serum (FBS), 10-fold serially diluted, and added to pre-seeded Vero cells for fluorescent focus assay (FFA). Primary antibody for dengue was the MAB8705 Anti-Dengue Virus Complex Antibody clone D3-2H2-9-21 (Millipore, Massachusetts, USA). Primary antibody for Zika was the MAB10216 Anti-Flavivirus Virus Complex Antibody clone D1-4G2-4-15 (Millipore, Massachusetts, USA). Secondary antibody for both viruses was the Goat anti-mouse Alexa Fluor 488, A-11001 (Thermo Scientific, Waltham, Massachusetts, USA). Plates were imaged using a Typhoon 9400 plate reader (GE Healthcare, Little Chalfont, UK) and images were analysed using Ima-geJ (NIH, USA).
Once saliva was collected, mosquito salivary glands were dissected and RNA was extracted using the QIAamp Viral RNA Mini kit (Qiagen, Hilden, Germany) according to manufacturers guidelines. Abdomens were removed and placed into tubes containing RNAzol reagent (Sigma-Aldrich, Missouri, USA). RNA was extracted according to manufacturers guidelines. cDNA synthesis was performed using the All-In-One cDNA Synthesis SuperMix (Biotool, Houston, Texas, USA), and qPCRs were run using NS5-F and NS5-R primer set for dengue and the ZIKV 835 and ZIKV 911c primers for Zika virus. For Zika infected mosquitoes, the numbers of samples analysed were 22, 21, 16 and 18 for wAlbB, wAu, wMel and wild-type, respectively. For dengue infected mosquitoes the numbers of samples analysed were 19, 18, 19 and 17 for wAlbB, wAu, wMel and wild-type, respectively. Levels of target cDNA sequences were normalized against the RpS17 house-keeping gene using the Pfaffl method. Primer sequences can be found in S1 Table. Wolbachia response to temperature fluctuations Eggs of the wAlbA, wAlbB, wAu and wMel strains were hatched under either constant 27˚C (control) or 27-37˚C at a 12:12hr cycle (heat-stressed) in a Panasonic MLR-352-H Plant Growth Chamber incubator (Panasonic, Osaka, Japan) and corresponding light and dark photoperiods (light during 37˚C). Immediately upon hatching, larvae were picked and placed into trays containing 1L of water and larval food at a density of 50 larvae per tray (three replicate trays per strain). Larvae were reared until pupation with water temperatures monitored daily using a glass thermometer placed inside a water-filled beaker. Water in the larval trays was replaced every 2 days to reduce bacterial growth. A selection of adults was removed upon emergence and split into two groups with a batch of approximately 15 (pooled into 5 repeats each containing 3 pooled adults) analysed by qPCR for Wolbachia density using the WSP and HTH primer sets. The remaining adults were set up in cages maintained at a constant 27˚C and allowed to recover for 7 days. These were also split into two groups with a sub-set analysed by qPCR for Wolbachia density, and the remainder (five females from each line) mated to wild-type males and blood-fed at day 5-post emergence. Eggs were collected and hatched at constant 27˚C and reared to pupation. A selection of 10 pupae from each female were chosen at random and assessed for Wolbachia infection by PCR.
A subset of adult females emerging from heat-stressed larvae were maintained under temperature cycling, mated to wild-type males, blood fed at day 5 post emergence and allowed to oviposit. Eggs were hatched and reared to pupation under heat stress at which point a selection of 10 pupae from each female were chosen at random and assessed for Wolbachia infection by PCR.

Adult longevity
Adult survival was assessed using groups of 50 individuals at a sex ratio of 1:1, with four replicates for each line. Experiments were performed in 24.5x24.5x24.5cm insect rearing cages inside an incubator set to 27˚C and 70% relative humidity with a 12-hour light/dark cycle. Cages were blood-fed once a week from day 5 onwards and damp filter paper was provided for oviposition. A sucrose meal was accessible ad libitum. Cages were checked daily for mortality. Experiments ran for 70 days at which time approximately 10% of the wMel and wild-type females remained alive.

Fecundity
Female fecundity was assessed by feeding 5-day old males and females of Wolbachia-infected and wild-type mosquitoes on a hemotek feeder containing defribrinated sheep blood. 20 fully engorged females (considered fully engorged when a female had a full abdomen and voluntarily dropped off the blood source) were isolated using an aspirator. Females were placed individually inside up-turned cups on top of a circle of filter paper. Cotton-wool soaked in a 10% sucrose solution was made available through a hole in the cup. 3 days post-feeding the filter paper was wetted and left overnight. The filter paper was replaced the next day and the process was repeated for a second night. Eggs from each filter paper were counted using a clickercounter and a dissecting microscope.

Egg survival
Egg survival in desiccated quiescence was assessed by feeding one week old Wolbachiainfected or wild-type females in cages and collecting eggs 3 and 4 days after feeding by placing three separate damp filter-paper cones in each cage-each cone collected >1,000 eggs. Egg papers were stored at 27˚C and 70% relative humidity. At 5, 10, 20, 35 and 50-days post oviposition a section of each of the egg papers containing approximately 200-300 eggs was cut from the original paper, the eggs counted using a clicker-counter and dissection microscope, and hatched by placing in water containing 1g/L bovine liver powder. Hatch rates were assessed 10 days later by counting larvae using a Pasteur pipette and a clicker-counter.

Statistical analysis
All statistical analyses were performed in the RStudio interface (version 0.99.489) (RStudio Inc., Boston, Massachusetts, USA) of the R software (version 3.4.0). Graphics were generated using the 'ggplot2' package. Normality of data distributions were assessed using a Kolmogorow-Smirnov Test prior to hypothesis test selection. Multiple comparisons were performed using the 'multcomp' package and used the Bonferroni multiple comparisons p value correction method. Survival analyses were performed using the 'Survival' and 'SurvMiner' packages. Survival analyses were performed using a Cox proportional hazard regression model with cage repeats clustered as a random effect.
Primer sequences. A list of primers and primer sequences used in this manuscript can be found in supplementary Table 1.