Developing Wolbachia-based disease interventions for an extreme environment

Aedes aegypti mosquitoes carrying self-spreading, virus-blocking Wolbachia bacteria are being deployed to suppress dengue transmission. However, there are challenges in applying this technology in extreme environments. We introduced two Wolbachia strains into Ae. aegypti from Saudi Arabia for a release program in the hot coastal city of Jeddah. Wolbachia reduced infection and dissemination of dengue virus (DENV2) in Saudi Arabian mosquitoes and showed complete maternal transmission and cytoplasmic incompatibility. Wolbachia reduced egg hatch under a range of environmental conditions, with the Wolbachia strains showing differential thermal stability. Wolbachia effects were similar across mosquito genetic backgrounds but we found evidence of local adaptation, with Saudi Arabian mosquitoes having lower egg viability but higher adult desiccation tolerance than Australian mosquitoes. Genetic background effects will influence Wolbachia invasion dynamics, reinforcing the need to use local genotypes for mosquito release programs, particularly in extreme environments like Jeddah. Our comprehensive characterization of Wolbachia strains provides a foundation for Wolbachia-based disease interventions in harsh climates.

mosquitoes, with 100/100 progeny from ten Wolbachia-infected mothers for each strain testing positive for Wolbachia by PCR (lower 95% confidence interval: 0.963). This was corroborated in a second experiment with wAlbB S showing complete maternal transmission to 75 progeny from 17 Wolbachia-infected mothers (lower 95% confidence interval: 0.952). These results are consistent with previous work on these strains in an Australian background [63,64].
We then tested the ability of both strains to induce cytoplasmic incompatibility in Saudi Arabian mosquitoes. Both wAlbB-infected and wMelM-infected males induced complete cytoplasmic incompatibility with uninfected females, with no eggs hatching ( Table 1). All other crosses were fertile, but egg hatch proportions were lower in crosses with wAlbB-infected females, consistent with later experiments (see below).
To test for potential changes in the Wolbachia genome in Saudi Arabian mosquitoes, the Wolbachia genomes of the wAlbB Au and wAlbB S populations were Illumina-sequenced, resulting in coverage of 91X for the genome in the Saudi Arabian genetic background and 143X in the Australian genetic background. The two genomes are identical, both differing from the reference genome by four single nucleotide variants and one insertion (S1 Table). These results point to the stability of the wAlbB genome across time and in different mosquito genetic backgrounds.

Wolbachia infections block DENV2 replication and dissemination in Ae. aegypti with a Saudi Arabian background
We previously demonstrated that wAlbB and wMelM reduce DENV2 virus titers in Australian Ae. aegypti [64]. Here, we estimated the effects of these Wolbachia strains on DENV2 blocking in Ae. aegypti from a Saudi Arabian background. Both wAlbB S and wMelM S showed reduced midgut (Two-tailed Fisher's exact test, wAlbB: P = 0.025, wMelM: P < 0.001, Fig 1A) and carcass (wAlbB: P < 0.001, wMelM: P < 0.001, Fig 1B) DENV2 infection frequencies compared to Uninfected S. For samples that tested positive, wAlbB S had fewer DENV2 copies compared to Uninfected S in midgut samples (Mann-Whitney U: Z = 4.431, P < 0.001). DENV2 blocking appeared to be particularly strong for wMelM, with reduced midgut infection frequencies compared to wAlbB (Two-tailed Fisher's exact test, P < 0.001). Near-complete DENV2 blocking by wMelM was further demonstrated in an additional comparison of the wMelM S and Uninfected S populations (S1 Fig). We were unable to perform other comparisons of DENV2 copy numbers owing to the very low infection frequencies in Wolbachia-infected mosquitoes as noted above. We also had low power to estimate effects of Wolbachia infection on DENV2 transmission due to low saliva infection frequencies in the uninfected population (13.3% positive, Fig 1C), however all wAlbB S and wMelM S saliva samples were negative for DENV2. In a  separate experiment, wAlbB also showed blocking in an Australian background, with wAlbB Au having fewer DENV2 copies in the midgut (Z = 2.748, P = 0.006) and carcass (Z = 4.935, P < 0.001) compared to Uninfected Au (S2 Fig). Together, these results demonstrate DENV2 blocking of both wAlbB and wMelM in Saudi Arabian Ae. aegypti. Saliva infection assays likely underrepresent transmission potential [65], but our analyses of midgut and carcass infections suggest that Wolbachia infections will likely reduce DENV2 transmission.

3 Proliferation of Wolbachia in adult mosquitoes
Wolbachia genotypes that reach a high density in their insect hosts tend to block viral infection more strongly but impose greater fitness costs on their hosts. We found a

Mosquito background influences body size but not larval development
We measured larval development time, survival to pupa, sex ratio and adult wing length for each Wolbachia infection type across the Australian and Saudi Arabian mosquito backgrounds. We found no significant effect of Wolbachia infection type for any trait tested when significance thresholds were adjusted for Bonferroni correction (GLM: all P > 0.012, adjusted α: 0.008 given that six traits were scored). However, wing length was strongly influenced by background in both females (F 1,130 = 41.378, P < 0.001) and males (F 1,122 = 76.234, P < 0.001), with Ae. aegypti with a Saudi Arabian background being consistently larger than Ae. aegypti with an Australian background ( Table 2).

Severe cumulative fitness costs of wAlbB infection following egg storage
We measured the viability of wAlbB-infected and uninfected eggs from Australian and Saudi Arabian backgrounds following storage. Egg viability decreased across time, particularly in Saudi Arabian populations in the early time points, where egg hatch proportions were below 0.5 at week 5, but costs of wAlbB infection also became apparent in later time points (Fig 2A). Both Wolbachia infection (GLM: all P < 0.014) and background (all P < 0.001) had significant effects on (logit transformed) egg hatch proportions at each time point. At week 11, we also

Wolbachia in extreme environments
We then measured the fertility of females hatching from stored eggs. wAlbB-infected females stored as eggs became increasing infertile with storage time (Fig 2B). After 9 weeks of storage, 77% (binomial confidence interval: 63-87%) of females from the wAlbB Au and wAlbB S populations were infertile (with no eggs laid). In contrast, females from uninfected populations remained fertile, with no infertile females at week 9 (binomial confidence interval: 0-7%). The fecundity and (logit transformed) egg hatch proportion of fertile females stored as eggs for 1 week was influenced by wAlbB infection (GLM: fecundity: F 1,190 = 7.369, P = 0.007, egg hatch: F 1,190 = 39.279, P < 0.001) but not background (fecundity: F 1,190 = 0.238, P = 0.626, egg hatch: F 1,190 = 0.319, P = 0.573). We found similar effects after 9 weeks of storage, but with a stronger effect of wAlbB infection on egg hatch (F 1,117 = 124.292, P < 0.001). The egg hatch of uninfected populations remained above 95% across all time points, but this declined to 50-60% in both wAlbB-infected populations (Fig 2D). In a scenario where eggs are stored for 9 weeks, wAlbB-infected populations had a relative fitness of 8.15% (Australian background) and 4.99% (Saudi Arabian background) compared to their uninfected counterparts when combining the effects on fecundity, female fertility and egg hatch across two generations.
In a second experiment on the populations with a Saudi Arabian background including wMelM S, egg viability decreased across time in all populations ( Fig 2E). We found a significant effect of Wolbachia infection type on (logit transformed) egg hatch proportions (GLM: F 3,125 = 12.995, P < 0.001), where the wAlbB S population had a lower egg hatch than the other populations. Costs of the wAlbB infection on egg hatch were similar regardless of storage time, as there was only a marginally significant interaction between Week and Wolbachia infection type (F 10,125 = 1.911, P = 0.049). We then measured the fertility of females hatching from stored eggs by dissecting ovaries (Fig 2F). After 11 weeks of storage, 34% (binomial confidence interval: 22-48%) of wAlbB S females were infertile (with no visible ovaries) compared to only 12% (6-23%) and 2% (0-10%) for wMelM S and Uninfected S respectively. These measures may be an underestimate of infertility compared to the previous experiment where female egg laying was scored directly since some females with ovaries may still be infertile. Overall, these results show substantial fitness costs for the wAlbB infection and much weaker costs for the wMelM infection.
In a separate experiment, we estimated water loss by weighing mosquitoes before and after an 8 hr period at low RH. We found significant effects of background on water loss (expressed as a percentage of water mass lost per hr) in males (GLM: F 1,27 = 25.357, P < 0.001) but not females (F 1,27 = 3.035, P = 0.093). Males with an Australian background had a~50% higher rate of water loss compared to males from the Saudi Arabian background ( Table 3). The same pattern was observed for females, but differences were marginal, at~6% higher.

Loss of Wolbachia infection and costs to egg survival at high temperatures
We measured the effect of wAlbB infection and background on heat tolerance in eggs by exposing them to a range of cyclical temperatures for one week. Egg hatch proportions declined in all populations as temperatures increased, with near-zero hatching when eggs were exposed to a 32-42˚C for one week (Fig 4A and 4B). Egg hatch proportions were lower in Wolbachia-infected populations compared to uninfected populations at most high temperatures,   suggesting an effect of Wolbachia on egg thermal tolerance (Fig 4A and 4B). However, we also found significant effects of Wolbachia infection (GLM: F 2,99 = 10.305, P < 0.001) and background (F 1,99 = 9.411, P = 0.003) on (logit) egg hatch proportion in the 26˚C control temperature, indicating that Wolbachia infection had negative effects independent of temperature. We also found an effect of background in the 31-41˚C treatment (F 1,97 = 11.074, P = 0.001), suggesting that eggs from the Saudi Arabian background had increased tolerance to temperatures near their upper thermal limit. We then measured Wolbachia density in adults emerging from eggs exposed to heat stress (Fig 4C and 4D). We found strong overall effects of Wolbachia strain (GLM: females: F 1,137 = 417.422, P < 0.001, males: F 1,138 = 627.602, P < 0.001) and temperature (females: F 4,137 = 111.474, P < 0.001 males: F 4,138 = 76.812, P < 0.001) on Wolbachia density, with an interaction between Wolbachia strain and temperature (females: F 3,137 = 92.940, P < 0.001, males: F 3,138 = 71.569, P < 0.001). Wolbachia densities of both strains decreased with increasing temperature, but the decline was much sharper for wMelM (Fig 4C and 4D). We observed complete loss of wMelM in some individuals at high temperatures, with 13/36 individuals exposed to 29-39˚C testing negative and 34/36 negative at 30-40˚C across males and females from both backgrounds. In contrast, all individuals from wAlbB-infected populations were positive for

Wolbachia in extreme environments
Wolbachia across all treatments. We found no overall effect of background (females: F 1,137 = 0.460, P = 0.499, males: F 1,138 = 0.014, P = 0.907), with wMelM and wAlbB densities decreasing to a similar extent with temperature both the Australian and Saudi Arabian backgrounds ( Fig  4C and 4D). We then tested whether the effects of temperature and wAlbB infection on egg hatch were modulated by RH. Eggs from the wAlbB S and Uninfected S populations were stored for 3 weeks under two temperature (20-30˚C and 29-39˚C) and two RH (80% and 20%) conditions. We found significant effects of both temperature (GLM: F 1,139 = 88.702, P < 0.001) and RH (F 1,139 = 39.703, P < 0.001) on egg hatch. There was also an interaction between temperature and RH (

Discussion
With Wolbachia releases taking place in different climate regions, it is important to understand variation in mosquito tolerance to environmental stress and any interactions with Wolbachia infections. In this study, we developed and characterized two Wolbachia strains for use in population replacement programs under the extreme conditions of Jeddah, Saudi Arabia. Both strains are likely to be suitable for field release due to their strong virus blocking, maternal transmission, cytoplasmic incompatibility and other characteristics. However, each strain has advantages and disadvantages. While wMelM exhibited very strong blocking in the Saudi Arabian background and had lower costs during egg quiescence, the wAlbB infection showed greater stability at higher temperatures. These differences may make them suited to different contexts.
Our results highlight phenotypic differences between Ae. aegypti with genetic backgrounds from Cairns, Australia and Jeddah, Saudi Arabia. The Saudi Arabian mosquitoes showed a higher fecundity and increased adult desiccation tolerance compared to Australian mosquitoes, but the ability of their eggs to withstand long periods of quiescence was surprisingly relatively poor. This geographic variation in stress tolerance may reflect local adaptation to climate [67,68] and the type of larval habitats that are common in these locations. Perhaps the unexpected lower quiescent egg viability in Saudi Arabian mosquitoes suggests that eggs are not exposed to long dry periods, despite the very low annual rainfall in this city (S5 Fig). The Ae. aegypti in this location may use permanent sources of water for oviposition such as underground car parks where we have found evidence of mosquito breeding even in the middle of the dry and hot summer period. In these locations, temperatures can be relatively cool (<30˚C) even when ambient temperatures at the surface exceed 40˚C. There is very little shade in Jeddah residential areas outside of buildings due to low vegetation cover and this would make it unlikely that quiescent eggs would survive in summer [69]. In contrast, in Cairns there is shade around buildings provided by lush vegetation and across the dry winter period survival of quiescent eggs can be high [69].
The increased desiccation tolerance in Saudi Arabian mosquitoes suggests that adult mosquitoes are adapted to low RH conditions. In Cairns, relative humidity averages 63% and the

PLOS PATHOGENS
Wolbachia in extreme environments humidity peaks in the summer wet season (www.bom.gov.au) when mosquito numbers are highest but average temperatures are still below 30˚C. In Jeddah, the average humidity is similar but is around 60% when Ae. aegypti numbers are high in February-March which coincides with higher average temperatures of 30˚C or more (S5 Fig). Adult insect desiccation stress will likely depend on a combination of temperature and humidity at times when the adults are active [70] and combined temperature/humidity metrics such as the drying power of air are often used in assessing insect species differences in desiccation resistance [71]. We would therefore expect the adults from Saudi Arabia to be exposed to additional desiccation stress as they move about beyond built structures; based on genomic data we have previously found that Ae. aegypti do move tens or even hundreds of meters across a generation although this can be limited by the presence of wide roads [58].
The phenotypic variation that exists in Ae. aegypti across its current distribution is still poorly understood given that large scale comparisons across geographic regions have not been completed, but there are indications that population variation exists for several traits [11,12,15,16]. In other Diptera like various Drosophila species, geographic variation in traits has been linked to climatic factors [72,73]. There is also a rich literature on climate adaptation in insect populations associated with quiescence and diapause at various developmental stages [74,75] which have developed in many mosquito species where these phenotypes aid in population persistence and spread [2]. Our data suggest that it may be worthwhile assessing the relative desiccation resistance of different developmental stages of Ae. aegypti and linking these not only to climate but also to the nature of breeding sites present in different locations that affect the dynamics of the local populations.
Endosymbionts can modulate the desiccation [76,77] and thermal [78,79] tolerance of their hosts. Hence, we investigated the phenotypic effects of Wolbachia under a range of stressful environmental conditions. We found clear costs of Wolbachia infection to several traits, including egg hatch, particularly in quiescent eggs and greatly reduced female fertility. These costs were largely independent of genetic background, but their importance will depend on background and environment. As demonstrated previously [43], effects of Wolbachia on viability and fertility after egg storage are substantial, but may be less relevant in areas with continuously available larval habitats or in locally-adapted mosquitoes where egg viability is already poor. Given that Wolbachia establishment can vary substantially between locations [34,40], even in similar climates, the local environment and mosquito adaptation to this environment are likely to be key drivers of Wolbachia establishment success.
We investigated novel traits, adult desiccation tolerance and water loss, in this study and found no clear effect of Wolbachia on either trait. Consistent with our previous work [41], we also found costs of Wolbachia infection when eggs were heat-stressed. However, it is unclear if Wolbachia reduces thermal tolerance when there are also general fitness costs, especially given that the slope of the decrease in hatch was similar between Wolbachia-infected and uninfected lines. Our experiments showing substantial costs of wAlbB infection after storage under different temperature and RH conditions also support the idea that the costs of Wolbachia infection are not strongly temperature-dependent.
Dengue hemorrhagic fever has been reported in Saudi Arabia since 1994 [80] with cases fluctuating between 1500 up to 7000 cases spanning a decade from 2006 to 2016, resulting in an economic burden of US$110 million annually. The most common dengue virus serotype has been DENV2. Two more serotypes were reported (DENV1 and DENV3); however, there is no record of DENV4 among published serological studies [81]. The strong blockage of dengue virus (DENV2) by Wolbachia and particularly wMelM in the Jeddah genetic background provides confidence for mosquito releases aimed at population replacement to proceed in the Jeddah area. At present, dengue suppression in Jeddah is reliant on regular and widespread

PLOS PATHOGENS
Wolbachia in extreme environments applications of pesticides, which is not sustainable in the long run due to both environmental costs and the evolution of resistance [57,62]. It is still unclear how easily Wolbachia invasion will proceed within the Jeddah environment, and whether the replacement process can be facilitated through the selection of a particular Wolbachia strain and releases targeting a particular time of the year. Nevertheless, even partial replacement is likely to have beneficial outcomes in terms of dengue suppression [19,34], and there is clearly the added benefit of fitness costs incurred by the Wolbachia strains leading to a reduction in mosquito abundance which can benefit both dengue transmission [82] and nuisance mosquito biting, potentially reducing the pesticide load in urban environments.

Ethics statement
Blood feeding of female mosquitoes on human volunteers for this research was approved by the University of Melbourne Human Ethics Committee (approval 0723847). All adult subjects provided informed written consent (no children were involved).

Mosquito strains and colony maintenance
We performed experiments with laboratory populations of Ae. aegypti mosquitoes from two geographic origins (Cairns, Australia and Jeddah, Saudi Arabia) and three Wolbachia infection types (wAlbB, wMelM and uninfected). The wAlbB infection was originally transferred from Ae. albopictus via microinjection [83] and is a wAlbB-Hou variant of the strain [84]. The wMelM infection is a clade I variant of wMel that was transferred from temperate D. melanogaster via microinjection [64]. Wolbachia infections were introduced to Ae. aegypti with an Australian mitochondrial haplotype and nuclear background through microinjection as described previously [63,64]. wAlbB-infected mosquitoes were then treated with tetracycline [63]

PLOS PATHOGENS
Wolbachia in extreme environments feeders (Hemotek Ltd., Blackburn Lancashire, Great Britain). For experiments in the United Kingdom, larvae were reared in pots (Azlon Straight-sided Beakers; 500mL) or trays (Addis Seal Tight rectangular; 3000mL) at controlled densities 100 or 200 larvae respectively and fed liver powder ad libitum. Adults were kept in 27 L BugDorm-1 cages (Megaview Science C Ltd., Taichung, Taiwan) and fed with a solution of 10% fructose and 0.1% para-aminobenzoic acid (PABA). Females were fed donated human blood from Blood Transfusion Services at Addenbrooke's Hospital, Cambridge, UK via Hemotek membrane feeders. Eggs were hatched in reverse osmosis water with a small pinch of desiccated liver and a few grains of desiccated yeast. Each experiment (i.e. each subheading in the methods) was performed with separate cohorts of mosquitoes.

Wolbachia genome sequencing
We sequenced the Wolbachia genome of the wAlbB S population and compared this to published sequences of wAlbB Au [63] to investigate potential genetic changes following the introgression of the wAlbB infection into Saudi Arabian Ae. aegypti. Two DNA sequencing libraries were prepared from two wAlbB S mosquitoes using the NEBNext Ultra II FS DNA Library Prep Kit for Illumina (New England Biolabs, E7805L) and NEBNext Multiplex Oligos for Illumina (New England Biolabs, E7335G) according to the manufacturers' recommendations with variations as follows: all samples were fragmented for 10 minutes at 37˚C. A post-ligation size selection step was performed on both libraries, 25ul of KAPA Hyperpure beads (Roche 8963835001) were used for the first step and 10ul of beads for the second step. 7 cycles of postadaptor ligation PCR enrichment were used. The libraries were quantified using a Qubit HS DNA assay kit (Thermofisher Q32851) and then combined to give a final quantity of 950ng in a volume of 45ul. Samples were enriched for Wolbachia target DNA using a KAPA Hypercapture kit (Roche 09075810001) and KAPA Hypercapture Bead kit (Roche 09075780001) with custom designed KAPA Hyperexplore Max probes (Roche 09063633001) according to the manufacturer's recommendations. Alternative options in the protocol were available, where this was the case, 20ul of COT human DNA (1mg/ml) was used to block the hybridization reaction and 10 cycles of post-capture PCR amplification was used. Captures were eluted from the final bead wash in 20ul of EB, quantified using Qubit HS DNA assay kit and analyzed using a Tapestation D1000 kit (Agilent 5067-5582). Sequencing was performed using Illumina MiSeq with 2 x 250bp reads (University of Cambridge, Department of Biochemistry sequencing service).

Processing of sequencing data
As preliminary analysis revealed that the two replicate wAlbB Au libraries were identical, they were combined before analysis. We also reanalysed raw sequence reads of wAlbB Au [63]. AdapterRemoval was used to trim adapters sequences from paired-end FASTQ reads, which also merged the overlapping reads using collapse option. The merged and unmerged sequence reads were aligned separately to the Wolbachia pipientis wAlbB chromosome (GenBank accession no. CP031221.1) using BWA-MEM with default options. BAM files belonging to the same sample were then merged using SAMtools. MarkDuplicates module in Picard was used to remove PCR and duplicates from the BAM files. Reads with a mapping quality of <30 were removed from the alignment, and coverage metrics were subsequently calculated using module DepthOfCoverage in GATK. Variant calling was performed for each individual using the HaplotypeCaller tool from GATK, and genotyping was then conducted using tool Genoty-peGVCFs. The raw variant calling files were filtered by the set of hard filters to remove potential false variants, using the standard parameters recommended by GATK.

PLOS PATHOGENS
Wolbachia in extreme environments

Cytoplasmic incompatibility
We performed reciprocal crosses to test the ability of wAlbB and wMelM to induce cytoplasmic incompatibility in Ae. aegypti with a Saudi Arabian background. wAlbB S and Uninfected S males were crossed to wAlbB S or Uninfected S females (both 2 d old) and females were blood fed at 5 d old after being sugar-starved for 24 hr. Crosses were also repeated with wMelM S and Uninfected S individuals in each combination. Twenty females from each cross were isolated for oviposition in 70 mL specimen cups with mesh lids lined with sandpaper strips and filled with 15 mL of larval rearing water. Eggs were collected 4 d after blood feeding, partially dried, then hatched 3 d post-collection. Egg hatch proportions were determined 3 d after hatching by counting the number of hatched (egg cap detached) and unhatched (intact) eggs.

Wolbachia maternal transmission
We assessed the fidelity of wAlbB and wMelM maternal transmission in Ae. aegypti with a Saudi Arabian background. Females from the wAlbB S and wMelM S populations were crossed to Uninfected S males and fifteen females per population were blood fed and isolated for oviposition. We then screened ten offspring from ten females (110 individuals per population in total) for the presence of Wolbachia (see below). In a second experiment, we measured the maternal transmission of wAlbB S infection at a later date by testing 75 female offspring from 17 female parents.

Wolbachia proliferation in adult mosquitoes
We investigated the proliferation of the wAlbB and wMelM infections in adult female mosquitoes with a Saudi Arabian background. Larvae were reared to the pupal stage and separated by sex. Virgin wAlbB S females were fed on fructose and PABA, then collected at days 0, 7, 14, 21, 28 and 35 days post-eclosion. We then measured wAlbB densities in 40 females from each time point and wMelM densities in 20 females from each time point across two independent experimental replicates (see Wolbachia quantification).

DENV2 virus challenge
We tested the effect of wAlbB and wMelM strains on dengue virus serotype 2 (DENV2) infection in Saudi Arabian Ae. aegypti through an artificial membrane feeding assay [85]. DENV2 isolate ET300 (MT921572.1), a commonly used isolate in studies of Wolbachia virus blocking [50,86], was passaged in Vero cells and viral supernatant was used in experiments. The final virus stock titer was 10 6 TCID 50 /ml in Vero cells. In the first experiment, we compared DENV2 infection frequencies and titers between uninfected, wAlbB-infected and wMelMinfected populations with a Saudi Arabian background (Uninfected S, wAlbB S and wMelM S). In a second experiment, we compared uninfected and wAlbB-infected populations with an Australian background (Uninfected Au and wAlbB Au). In a third experiment, we performed a repeat of the Uninfected S and wMelM S comparison.
Female mosquitoes (6-12 d old) were artificially fed a meal of human blood (agreement number 20-04VIC-09) for 30 mins using a collagen membrane and a Hemotek membrane feeding system. For each infection, human blood was centrifuged at 2500 rpm for three minutes on a tabletop centrifuge, plasma discarded, then DENV2 viral supernatant was mixed with an equal volume of red blood cells for a final DENV2 titer of 10 6 TCID 50 /ml. Mosquitoes were then chilled and anesthetized with carbon dioxide. Engorged females were transferred to plexiglass cages and housed at 27˚C, 65% relative humidity with a 12:12 hr light/dark cycle and provided with 10% sugar ad libitum and an oviposition substrate.

PLOS PATHOGENS
Wolbachia in extreme environments

DENV2 quantification
Samples from blood-fed female mosquitoes were collected as previously described [87]. Briefly dissected mosquito midguts or mosquito carcasses (mosquito body after removing head, wings and legs) were homogenized, RNA extracted using RNeasy RNA isolation kit (Qiagen, Australia) and cDNA prepared using Superscript-III reverse transcriptase (Thermo Fisher Scientific Inc., Australia). Mosquitoes were anaesthetised with CO 2 and saliva was collected as previously described with slight modifications [85]. Briefly, after removing mosquito's legs and wings, the proboscis was inserted into a 20ul pipette tip containing 10ul foetal bovine serum (FBS) and left in place for 15 mins. The mix of FBS and saliva were transferred to individual tubes and processed for RNA extraction. Quantitative PCR (qPCR) reactions were prepared in triplicate using TB Green Premix Ex Taq II (Tli RNase H Plus) (Takara, USA) with a modified protocol. Each reaction consisted of 2μL cDNA, 10μL TB Green Premix Ex Taq II, 0.4μL ROX, 0.4μL 10μM forward primer, 0.4μL 10μM reverse primer and 6.8μL nuclease free water in a total volume of 20μL. Quantitative PCR was performed on QuantStudio6 Flex PCR System (Applied Biosystems, USA) using DENV2 non-structural protein 1 primers (S2 Table) with the following cycle parameters-Initial denaturation at 95˚C for 30 seconds, 40 cycles at 95˚C for 5 seconds, 60˚C for 30 seconds plus melt curve. Ae. aegypti Ribosomal protein S17 (RpS17) were used to normalize total RNA. After each run, threshold was set to 0.04 with an automatic baseline and data exported for detailed analysis.
To determine the absolute number of DENV2 genomic copies in each mosquito sample, a recombinant plasmids containing DENV2 ET300 fragment (2489bp-2598bp) or Aedes aegypti RpS17 fragment (228bp-295bp) (Integrated DNA Technologies, Singapore) were used to generate a standard curve. Briefly, serial 10-fold dilutions of each plasmid were prepared in 10mM TE buffer with 50ng/μL yeast RNA (Sigma, USA) and each dilution tested via qPCR as previously described. Limits of detection (LOD) was 2 plasmid copies per reaction. Linear equations were derived from the standard curve, enabling normalized DENV2 genomic copy numbers to be calculated from C T values obtained in qPCR of a DENV2 infected mosquito sample. For calculation, samples with a negative log value were interpreted as below LOD and were assigned a value of 0.

Larval development and body size
We measured larval development time, survival to pupa and adult wing length for each population. Eggs (< 1 week old) were hatched in plastic trays filled with RO water and a few grains of yeast. Larvae (< 8 hr old) were then counted into 750 mL plastic trays filled with 500 mL RO water and provided with fish food ad libitum throughout their development. We set up six replicate trays with 100 larvae per tray for each population. Pupae were counted, sexed and removed from trays twice per day (at 9 am and 5 pm). Adults emerged in cages and twenty adults of each sex were then measured for their wing length according to previous methods 51 .

Female fertility following egg storage
We measured the viability of quiescent eggs stored for up to 11 weeks and the fertility of females hatching from these eggs in two experiments. Eggs were collected on sandpaper strips from colony cages of wAlbB Au, wAlbB S, Uninfected Au and Uninfected S populations. Sandpaper strips were partially dried for 2 d then placed in a sealed chamber with a saturated solution of potassium chloride to maintain relative humidity (RH) at~80%. Eggs were then hatched after 1, 3, 5, 7, 9 and 11 weeks of storage by submerging sandpaper strips in trays of RO water with a few grains of yeast. Twelve replicate batches of eggs (range: 39-129 eggs per replicate) were hatched per time point per population. Egg hatch proportions were determined 3 d after hatching as described above. Larvae were controlled to a density of 50 larvae in 500

PLOS PATHOGENS
Wolbachia in extreme environments mL RO water and reared to adulthood. We then assessed fertility in females hatching from stored eggs by blood feeding and isolating 50 females per population at each time point. We scored the proportion of infertile females (that did not lay eggs after one week) and the number of eggs laid by fertile females and their hatch proportions. Females that died before egg laying were excluded from the analysis. Female fertility was not measured after 11 weeks of storage due to low egg viability in some populations.
In a second experiment, we measured quiescent egg viability in the wAlbB S, wMelM S and Uninfected S populations with eight replicate batches of eggs according to the procedure above. We then scored the proportion of infertile females hatching from quiescent eggs by rearing females to adulthood and dissecting ovaries [88]. Infertile females were defined as those lacking visible ovaries. Ovaries which were incomplete or could not accurately be scored as present or absent were excluded from the analysis.

Adult desiccation tolerance
We measured the survival of adult Ae. aegypti from the wAlbB Au, Uninfected Au, wAlbB S and Uninfected S populations under low (~15%) and high (~80%) RH as an estimate of desiccation tolerance. In each experiment, six (low RH) or eight (high RH) adults (5 d old, sugarstarved for 1 d) were aspirated into numbered glass vials with mesh lids, with eight replicate vials per sex per population. Vials were placed in sealed chambers in a random order around the perimeter so that mosquitoes could be observed without opening the chamber. We determined mortality by the tapping chamber and observing movement; mosquitoes that did not move after 10 s were considered dead. In the low RH experiment, mortality was scored every two hr, while in the remaining trials mortality was scored two to three times per day (four hr apart). In the low RH experiment, vials were placed in a sealed chamber containing trays with silica beads which maintained RH at~15% within 3 hr of sealing the chamber. In the high RH experiment, containers filled with a saturated solution of potassium chloride were placed in the chamber to maintain RH at~80%. Temperature and RH in each experiment was monitored with data loggers (Thermochron).

Adult water loss
We estimated water loss in adults from the wAlbB Au, Uninfected Au, wAlbB S and Uninfected S populations using methods adapted from previous studies [89,90]. Adults (5 d old, sugar-starved for 1 d) were aspirated into numbered glass vials covered with a mesh lid, with eight adults per vial and eight replicate vials per sex per population. Vials were weighed on an analytical balance (Sartorius BP 210 D) then placed in a sealed chamber under conditions identical to the adult desiccation experiments at 15% RH. After 8 hr, vials were removed from the chamber and immediately weighed again. No mortality was observed during this period. Mosquitoes were killed by placing vials at -20˚C for 48 hr then dried by placing vials in an oven at 70˚C for 3 d. Vials were weighed before and after removing the dried mosquitoes. Fresh mass per mosquito was calculated by subtracting the empty vial mass from the mass of vials containing mosquitoes prior to desiccation, divided by eight. Dry mass per mosquito was subtracted was calculated by subtracting the empty vial mass from the mass of vials containing dried mosquitoes, divided by eight. Water mass was calculated by subtracting dry mass from fresh mass, while percent water content was calculated by dividing water mass by fresh mass. Percent water loss per hr was calculated by subtracting the mass of vials after 8 hr at 15% RH from the mass of vials containing mosquitoes prior to desiccation, then dividing by eight. Then, this value was divided by water mass and multiplied by 100. Two vials were excluded from the analysis due to escapees during the experiment.

PLOS PATHOGENS
Wolbachia in extreme environments

Egg viability and Wolbachia densities following exposure to cyclical heat stress
We measured egg hatch and Wolbachia titres at high temperatures by hatching eggs from the wAlbB Au, wMelM Au, Uninfected Au, wAlbB S, wMelM S and Uninfected S populations following exposure to a range of cyclical temperature regimes. Eggs from each population were collected on filter paper (Whatman 125 mm qualitative circles, GE Healthcare Australia Pty. Ltd., Parramatta, New South Wales, Australia), partially dried, then stored for 3 d at room temperature. Batches of eggs (range: 15-60) in 0.2 mL tubes were placed in blocks of TProfessional TRIO 48 thermocyclers (Biometra, Gottingen, Germany). Eggs were exposed to a constant 26˚C or daily cycles of 27-37˚C, 28-38˚C, 29-39˚C, 30-40˚C, 31-41˚C or 32-42˚C according to methods described previously [41,91]. Between 12 and 20 replicate tubes per population were exposed to each temperature cycle; low numbers of eggs laid by some populations resulted in variable numbers of replicates. Eggs were hatched after one week of exposure by brushing eggs from each replicate tube into wells of Costar 12-well cell culture plates (Corning, NY) filled with 4 mL of RO water and a few grains of yeast. Egg hatch proportions were determined one week after hatching according to methods described in the section above. Larvae hatching from the Wolbachia-infected populations at each temperature were pooled across replicates and reared at a density of 50 larvae in trays of 500 mL water. Adults were stored in absolute ethanol within 24 hr of emergence to measure wAlbB and wMelM and densities (see Wolbachia quantification). Nine adults per sex, per population, per temperature cycle were screened for Wolbachia density. Adults from the 31-41˚C or 32-42˚C treatments were not screened due to near-zero hatch proportions under these conditions.

Egg viability under different temperature and humidity conditions
We tested the effects of wAlbB infection on the hatch proportion of stored eggs under different temperature and RH conditions in two experiments. Blood-fed females from the Uninfected S and wAlbB S populations were transferred to cups in groups of five and eggs were collected every two days over a six day period. Each collection of eggs was divided evenly between the experimental treatments by cutting the paper. We hatched eggs at 48-72 hours after collection ('control'; 80% RH and 26˚C) or 21 days after the beginning of the experiment for each treatment under different temperature and humidity conditions. In the first experiment, eggs were stored at cycling temperatures of 20-30, 23-33, 26-36 or 29-39˚C and 80% RH, with 12 hr at each temperature. In the second experiment, eggs were stored at 20-30˚C or 29-39˚C and 80% or 30% RH. To control RH, the eggs were placed in a sealed plastic box that contained an open beaker of a saturated hygroscopic salt solution. Saturated magnesium chloride hexahydrate (MgCl 2 �6H 2 O) solution maintained RH at~30% while saturated potassium chloride (KCl) solution maintained RH at~80%. We monitored the temperature and humidity in all the eggs storage boxes using a data logger to ensure these treatments were successful. Egg hatch proportions were determined for 15-25 replicate batches of eggs per treatment by counting the number of eggs and larvae 4-5 days after hatching.

Wolbachia detection and quantification
Aedes aegypti from colonies and experiments were screened to confirm their Wolbachia infection status and estimate relative Wolbachia density. We confirmed the presence or absence of Wolbachia in mothers and offspring from the maternal transmission experiment and screened colonies used in experiments routinely (approximately every 3 months). We also quantified Wolbachia densities in adults of different ages and in adults following exposure of eggs to cyclical heat stress.

Wolbachia in extreme environments
For experiments in Australia, genomic DNA was extracted using 250 mL of 5% Chelex 100 resin (Bio-Rad Laboratories, Hercules, CA) and 3 mL of proteinase K (20 mg/mL) (Roche Diagnostics Australia Pty. Ltd., Castle Hill, New South Wales, Australia). Tubes were incubated for 60 min at 65˚C and then for 10 min at 90˚C. We ran real-time PCR assays [92,93] using a Roche LightCycler 480 using primer sets to amplify an Ae. aegypti-specific marker (aRpS6_F [5'-ATCAAGAAGCGCCGTGTCG-3'] and aRpS6_R [5'-CAGGTGCAGGATCTT CATGTATTCG-3']) and a Wolbachia-specific marker (wMwA_F [5'-GAAGTTGAAGCAC AGTGTACCTT-3'] and wMwA_R ) that can distinguish between wAlbB and wMelM infections based on differences in melting temperature (80.4 ± 0.02˚C and 82.6 ± 0.03˚C and respectively). Relative Wolbachia densities were determined by subtracting the wMwA crossing point (Cp) value from the aRpS6 Cp value. Differences in Cp values were averaged across 2 or more consistent replicate runs and then transformed by 2n.
For experiments in the United Kingdom, mosquitoes were homogenized in 1.7 mL tubes with 50 μL of water and approximately ten 1.0 mm diameter zirconia beads (Thistle Scientific 11079110zx) by shaking at 30 Hz for 2 minutes in a Tissue Lyzer II homogenizer (Qiagen). DNA was extracted using a PureLink Genomic DNA Mini kit (Thermofisher Scientific, K1820-02) by adding 180 μL of PureLink Genomic Digestion Buffer and 20 μL of Proteinase K solution, followed by mixing and incubation at 55˚C for 2 hours. Samples were centrifuged at 13,000 g for 5 minutes to pellet debris and the supernatant was removed to a fresh tube. 400 μL of a 1:1 mixture of PureLink Genomic Lysis/Binding Buffer: Ethanol was added, and samples were mixed well by vortexing. Genomic DNA was purified using the standard spin column protocol according to the manufacturers recommendations and eluted from the column with 50ul PureLink Genomic Elution Buffer.
Quantitative PCR was carried out using Hi-Rox SensiFAST Sybr reagent (Meridian BIO-92020). For each qPCR plate a master mix was made containing 5 μL of SensiFAST master mix, 2 μL of nuclease free water and 1 μL of 2.5uM of each primer per reaction. The primers to amplify wAlbB were qwAlbB_Fw (5'-CCTTACCTCCTGCACAACAA) and qwAlbB_Rv (5'-GGATTGTCCAGTGGCCTTA) [42]. The primers to amplify mosquito DNA were mRpS6_Fw (5'-AGTTGAACGTATCGTTTCCCGCTAC) and mRpS6_Rv (5'-GAAGTG ACGCAGCTTGTGGTCGTCC) [42]. After adding 8 μL of the master mix to wells of a 384-well MicroAmp Optical Reaction Plate (Applied Biosystems 4309849), 2 μL of genomic DNA sample which had been diluted 2:1 with nuclease free water was added. Plates were sealed using MicroAmp Optical Adhesive film (Applied Biosystems 4311971), vortexed to mix, and centrifuged briefly. The plates were run on a 384-well QuantStudio 5 Real-Time PCR System (Applied Biosystems A28575) using the pre-set comparative-CT-Sybr with melt curve program with the reaction volume adjusted to 10 μL. Data were analyzed using QuantStudio Design and Analysis Software v1.5.2. The CT threshold value was set to 0.1 for each plate.

Statistical analysis
All statistical analyses were performed in SPSS Statistics 26. Proportional data such as egg hatch were logit transformed before analysis [94], treating each replicate batch of eggs (or emerged adults) as an independent data point rather than each egg given that different eggs in a batch may have come from the same female and that each batch may have experienced slightly different exposure conditions. Wolbachia density data were analyzed prior to 2 n transformation. We ran general linear models for each trait, with genetic background and Wolbachia infection included as factors. Where multiple traits were extracted from the same data set, P values were corrected by the Bonferroni method based on the number of traits scored. When

PLOS PATHOGENS
Wolbachia in extreme environments both sexes were tested, data for each sex were analyzed separately. For the adult heat tolerance experiment, we performed an analysis with a trial by population interaction. Because there were apparent differences in the relative performance of populations across trials, we reanalyzed data separately for each trial. Adult survival at low and high relative humidity was analyzed with Cox regression, with genetic background and Wolbachia infection included as factors. In experiments with multiple time points or temperature treatments, we analyzed data separately for each time point/treatment. For analyses of Wolbachia density, we included temperature, Wolbachia strain, background and their interactions as factors. For comparisons of female infertility, we computed 95% binomial confidence intervals on the proportion of females that did not lay eggs using the Wilson method [95]. We compared DENV virus copy numbers between populations using non-parametric Mann-Whitney U tests, with negative samples excluded from the analysis. We also compared DENV2 infection frequencies (i.e. the number of samples testing positive for DENV2 out of the total number tested) between populations with two-tailed Fisher's tests.