Loss of cytoplasmic incompatibility in Wolbachia-infected Aedes aegypti under field conditions

Wolbachia bacteria are now being introduced into Aedes aegypti mosquito populations for dengue control. When Wolbachia infections are at a high frequency, they influence the local transmission of dengue by direct virus blocking as well as deleterious effects on vector mosquito populations. However, the effectiveness of this strategy could be influenced by environmental temperatures that decrease Wolbachia density, thereby reducing the ability of Wolbachia to invade and persist in the population and block viruses. We reared wMel-infected Ae. aegypti larvae in the field during the wet season in Cairns, North Queensland. Containers placed in the shade produced mosquitoes with a high Wolbachia density and little impact on cytoplasmic incompatibility. However, in 50% shade where temperatures reached 39°C during the day, wMel-infected males partially lost their ability to induce cytoplasmic incompatibility and females had greatly reduced egg hatch when crossed to infected males. In a second experiment under somewhat hotter conditions (>40°C in 50% shade), field-reared wMel-infected females had their egg hatch reduced to 25% when crossed to field-reared wMel-infected males. Wolbachia density was reduced in 50% shade for both sexes in both experiments, with some mosquitoes cleared of their Wolbachia infections entirely. To investigate the critical temperature range for the loss of Wolbachia infections, we held Ae. aegypti eggs in thermocyclers for one week at a range of cyclical temperatures. Adult wMel density declined when eggs were held at 26–36°C or above with complete loss at 30–40°C, while the density of wAlbB remained high until temperatures were lethal. These findings suggest that high temperature effects on Wolbachia are potentially substantial when breeding containers are exposed to partial sunlight but not shade. Heat stress could reduce the ability of Wolbachia infections to invade mosquito populations in some locations and may compromise the ability of Wolbachia to block virus transmission in the field. Temperature effects may also have an ecological impact on mosquito populations given that a proportion of the population becomes self-incompatible.


Introduction
Aedes aegypti mosquitoes are the principal vectors of dengue and are widespread in the tropics where they live near humans [1,2]. Chemical insecticides have historically been used to control Ae. aegypti populations during disease outbreaks, but this approach is unlikely to be sustainable as insecticide resistance is now widespread in many parts of the world [3,4]. There is increasing interest in 'rear and release' programs where mosquitoes modified with desirable traits are released into natural populations as an alternative approach to disease control [5]. At the forefront of these programs is the deployment of mosquitoes infected with Wolbachia bacteria. Wolbachia occur naturally in many insects but have been introduced experimentally into Ae. aegypti where they can interfere with the transmission of dengue and other pathogens [6][7][8]. Wolbachia are transmitted maternally and typically affect host reproduction [9] or provide other advantages [10] to facilitate their spread into populations. These phenotypes have been utilized in disease control programs where Wolbachia-infected mosquitoes have been deployed to replace natural populations [11][12][13] or suppress populations through the release of only males [14,15]. Both approaches rely on cytoplasmic incompatibility induced by Wolbachia, where uninfected females that mate with infected males do not produce viable offspring, but viability is restored if the female is also Wolbachia-infected [16,17].
Over ten Wolbachia strain associations have now been generated in Ae. aegypti and they exhibit a diverse range of phenotypes. Some Wolbachia strains are relatively benign and have little impact on host fitness or virus blockage such as the wRi strain [18]. Others impose large fitness costs but also strongly reduce virus transmission including wMelPop [17] and wAu [19] while others like wAlbB fall somewhere in between [20,21]. There are also superinfections where two or more Wolbachia strains infect the same host, which can have combined or unexpected effects [21,22]. Aedes aegypti infected with the wMel strain of Wolbachia have been or are now being released in over ten countries (https://www.worldmosquitoprogram.org/) and have successfully established in suburban areas in Cairns and Townsville in Queensland, Australia [11][12][13] and in Brazil [23]. The wAlbB strain has also been deployed successfully in Malaysia for population replacement (http://www.imr.gov.my/wolbachia/) and in several countries for population suppression, where the release of only infected males has reduced population sizes by more than 80% due to cytoplasmic incompatibility (https://debug.com/; https://www.nea.gov.sg/corporate-functions/resources/research/wolbachia-aedes-mosquitosuppression-strategy).
Despite these successes, there are limitations of Wolbachia infections that may affect their utility as disease control agents (reviewed in Ritchie et al. [24]). The majority of Wolbachia infections in Ae. aegypti reduce mosquito fitness and these costs tend to be exacerbated in stressful environments such as when larvae are starved [25] or in quiescent eggs [18,19,26,27]. Fitness costs can have enormous effects on invasion success. For example, the wMelPop infection failed to persist in release zones in Australia and Vietnam despite reaching frequencies above 90%, likely due to the massive fitness costs of this strain [28]. Wolbachia infections that occur naturally in mosquitoes can interfere with patterns of cytoplasmic incompatibility and limit the potential for population replacement and suppression [29]. Density-dependent interactions [30] and spatially heterogeneous environments [31] can also slow the rate of invasion, as can pesticide susceptibility in released mosquitoes [23].
For population replacement programs to be successful, Wolbachia infections should persist at high frequencies in the environment and block virus transmission under field conditions for many years following deployment [24]. There is a risk that Wolbachia infections, viruses or mosquitoes will evolve following the establishment of Wolbachia in populations, leading to less effective virus protection in Wolbachia-infected mosquitoes in the long-term [32]. However, the wMel infection has remained stable so far in terms of virus blockage [33] and its effects on fitness [34]. After seven years in the field, wMel has retained a high titer and continues to induce complete cytoplasmic incompatibility in the laboratory [35], indicating that attenuation is unlikely for at least several years following deployment.
While the wMel infection in northern Queensland Ae. aegypti populations does not appear to have changed phenotypically since release, environmental conditions in the field such as temperature can have transient effects on Wolbachia infections, influencing their ability to suppress virus transmission or establish in populations. This issue is particularly important as climate change is leading to warmer average conditions and higher temperature extremes, including in the tropics [36]. Wolbachia infections in Ae. aegypti are vulnerable to high temperatures; heat stress during larval development reduces Wolbachia density in adults [37] and decreases the fidelity of cytoplasmic incompatibility and maternal transmission [35,38]. The fidelity of cytoplasmic incompatibility and maternal transmission are two key parameters for Wolbachia spread [39] while Wolbachia density is positively associated with the strength of virus blockage in both Drosophila [40,41] and mosquitoes [42].
Wolbachia strains in Ae. aegypti differ in their response to heat stress; the wMel and wMel-Pop strains are relatively susceptible while wAlbA, wAlbB and wAu are more robust, retaining high densities when larvae are reared at cyclical temperatures of 26-37˚C [19,38]. These laboratory studies demonstrate the potential for heat stress to affect the success of Wolbachia interventions, but conditions experienced by mosquitoes in field situations are more complex than in an incubator. To understand the effects of high temperatures under more natural conditions, we reared Ae. aegypti larvae infected with wMel in the field with varying levels of exposure to sunlight. We then performed crosses to test for effects on cytoplasmic incompatibility and measured Wolbachia density. Finally, we performed experiments with Ae. aegypti eggs in the laboratory to determine the range of temperatures that adversely affect Wolbachia infections.

Ethics statement
Blood feeding on human subjects was approved by the Human Research Ethics Committee, James Cook University (approval H4907). All adult subjects provided informed oral consent (no children were involved).

Mosquito strains and colony maintenance
Aedes aegypti mosquitoes infected with the wMel strain of Wolbachia were collected in 2013 from locations near Cairns, Australia where wMel had successfully established [11]. Aedes aegypti infected with wMelPop were collected from Cairns, Australia in 2012 following field releases and local field breeding [28]. Aedes aegypti infected with wAlbB were derived from laboratory colonies and are described in Xi et al. [16] and Axford et al. [20]. Uninfected Ae. aegypti were collected in 2016 from locations where Wolbachia-infected mosquitoes had not been released. Mosquitoes were maintained in an insectary at the University of Melbourne according to methods described by Ross et al. [43]. Females with each Wolbachia strain were crossed to males from the uninfected population for at least three generations prior to the start of experiments to ensure that genetic backgrounds between populations were similar [27].

Larval rearing under field temperature conditions
We conducted two experiments during the wet season in Cairns, Australia in 2018 to test the stability of the wMel Wolbachia infection under field temperature conditions. Experiments took place at James Cook University under a protective awning with different levels of shade as described in Ritchie et al. [44]. The relative reduction in light intensity compared to direct sunlight was determined using an EasyView EA30 light meter (Extech Instruments Corporation, Waltham, MA 02451 U.S.A.); two shade levels were chosen for the experiments which reduced light intensity by 50% and 99% respectively. We then established water-filled containers of various sizes to simulate a range of field larval habitats. The wMel-infected eggs were hatched in a single tray in 99% shade, then approximately 100 1 st instar larvae were placed into each container. Larvae were provided with TetraMin tropical fish food tablets (Tetra, Melle, Germany) ad libitum throughout their development. Water temperatures were recorded every 30 minutes for the duration of larval development with data loggers (Thermochron; 1-Wire, iButton.com, Dallas Semiconductors, Sunnyvale, CA, USA) placed in zip-lock bags and submerged in the centre of each container. Pupae were collected one week after hatching and were returned to the laboratory for adult emergence.
In the first experiment, Ae. aegypti eggs infected with the wMel strain of Wolbachia were hatched on the 10 th of January 2018 and pupae were collected on the 17 th of January. We used containers of three types with a wide range of water volumes which were expected to experience a variety of temperature conditions. Black buckets (13 cm radius, 25 cm height) were filled with 8 L of tap water, plant pots (5 cm radius, 10 cm height) with 500 mL and cups (2 cm radius, 6 cm height) with 60 mL. Containers were covered with mesh or stockings to prevent wild mosquitoes from ovipositing, and experimental mosquitoes from escaping. This setup was repeated for both the 99% and 50% shade levels. wMel-infected larvae were also reared in a single bucket filled with 8 L of water and placed in direct sunlight.
In the second experiment, we repeated this procedure but used two container types: black buckets filled with 8 L of water and small round clear plastic containers (10 cm radius, 7 cm height) filled with 400 mL of water, with each container replicated three times at 99% and 50% shade levels. Eggs were hatched on the 26 th of January 2018 and pupae were returned to the laboratory on the 1 st of February. Populations of wMel-infected and uninfected Ae. aegypti were also reared in the laboratory concurrently at 26˚C ± 1˚C according to Ross et al. [43] for experimental crosses.

Cytoplasmic incompatibility after field rearing
We tested the ability of wMel-infected males to induce cytoplasmic incompatibility and wMelinfected females to restore compatibility after being reared under field temperature conditions. Adults emerging from each container type and shade level were added to 15 cm 3 cages (Bug-Dorm-4M1515, Megaview Science Co., Taichung, Taiwan) where sexes were maintained separately. Crosses were performed two to three days after adults emerged by aspirating approximately 50 females into cages with an equal number of males. Females were blood fed three days later and then isolated in plastic cups containing 15 mL of larval rearing water and a strip of sandpaper for oviposition. Eggs were collected four days after blood feeding, partially dried, and then hatched four days after collection. Eggs were then counted under a dissecting microscope and hatch rates were determined by counting the proportion of eggs that had a clearly detached cap.
In the first experiment, we performed crosses with adults reared in buckets at either 99% or 50% shade. Field-reared wMel-infected males were crossed to uninfected females to determine their ability to induce cytoplasmic incompatibility. We also crossed field-reared wMel-infected females to either uninfected males or wMel-infected and laboratory-reared males to determine the ability of females to restore compatibility. Infected males and uninfected females both reared under laboratory conditions were crossed to each other to confirm that wMel induces complete cytoplasmic incompatibility at 26˚C. Twenty females were isolated for oviposition in each of these crosses, but individuals that died or did not lay eggs were excluded from the analysis.
In the second experiment we performed a similar set of crosses but only for adults emerging from buckets held in 50% shade. In crosses where both sexes were infected we used males and females from the same container rather than using males reared at 26˚C. This was done to see if populations became self-incompatible when both sexes were reared at warmer temperatures. Thirty females were isolated for oviposition in each cross. Crosses between males and females reared under the same conditions were also performed with adults that were reared in buckets held in 99% shade and small containers held at 50% and 99% shade. We determined egg hatch proportions and Wolbachia densities of females in each of these crosses to see if there was a relationship between Wolbachia density and hatch rate.

Egg thermal tolerance and Wolbachia density
We performed two experiments using thermocyclers (Biometra, Göttingen, Germany) to test the thermal tolerance of Wolbachia-infected Ae. aegypti eggs and the density of Wolbachia under a range of temperature conditions. We followed methods described in Kong et al. [45] with some modifications. Eggs from uninfected, wMel, wAlbB and wMelPop colonies were collected on sandpaper strips which were then partially dried, wrapped in paper towel and held in sealed zip-lock bags. Four days after collection, eggs were brushed onto filter paper with a small paint brush and then tipped into 0.2 mL PCR tubes using a funnel. Batches of 15 to 39 eggs (mean 25.7) were added to each tube. Tubes were closed and then tapped on the bench to ensure that eggs sank to the bottom of the tube where temperature control in the thermocycler is the most accurate [45]. Tubes were then placed in heat blocks of Biometra TProfessional TRIO 48 thermocyclers with tubes from each population arranged randomly in each block.
In both experiments we used three thermocyclers, each with three heated blocks that can run independently for a total of 9 temperature regimes. In the first experiment we chose a broad range of temperature cycles to cover the entire range of temperatures that Ae. aegypti may experience in the field (http://www.bom.gov.au/climate/averages/tables/cw_031011_All. shtml). Each regime had a fluctuation of 10˚C between the minimum and maximum temperature; the lowest being 8-18˚C and the highest being 32-42˚C, with a difference of 3˚C between each regime (S3A Fig). In the second experiment we chose a narrower temperature range based on when egg hatch and Wolbachia density started to decline in the previous experiment.
The lowest regime was set to 24-34˚C and the highest was 32-42˚C, with difference of 1˚C between each regime (S3B Fig). In the first experiment there were six replicate tubes of eggs for each temperature cycle and Wolbachia infection type and the second experiment had 12 replicates. After all tubes were added to the thermocyclers we closed the lids and started programs simultaneously. Eggs in tubes were also maintained at 26˚C in a controlled temperature room in both experiments.
After one week, tubes were removed from the thermocyclers and eggs were hatched by holding PCR tubes sideways above 70 mL specimen cups and then pipetting water into the tubes so that eggs fell into the cup. Each cup was filled with 40 mL of water and provided with a small amount of TetraMin and a few grains of yeast. Two days after hatching we determined egg hatch proportions by dividing the number of larvae by the number of eggs. We counted larvae again every 2 days as some eggs were slow to hatch, allowing one week in total for larvae to appear before we ceased counting. All larvae that hatched were added to plastic containers filled with 500mL of RO water and reared to adulthood. Multiple replicate cups of larvae were combined into trays for rearing, but the larval density was controlled to 100 larvae per tray or fewer to account for effects of larval competition and development time on Wolbachia density [46]. All adults were stored in ethanol for Wolbachia density measurements.

Wolbachia detection and density
In each experiment, random subsets of adults were stored in ethanol within 24 hours of emergence for Wolbachia screening. For both field experiments we extracted DNA from 16 males and 16 females from each container type and shade level. For experiments with eggs held in thermocyclers we extracted DNA from up to 10 (first experiment) or 12 (second experiment) males and 10 or 12 females from each Wolbachia infection type and treatment. Some treatments had lower sample sizes due to low egg hatch proportions. DNA was extracted from whole adults with 150 μL of 5% Chelex 100 resin (Bio-Rad Laboratories, Hercules, CA) and 3 μL of Proteinase K. We then conducted qPCR to detect and estimate the density of Wolbachia in each whole adult using methods described previously [47]. Individuals were considered uninfected if the Ae. aegypti-specific marker amplified successfully (Cp value < 35) but the Wolbachia-specific marker did not (Cp value of 35 or no Cp value) in two independent runs. For individuals that were positive for Wolbachia, (Cp value < 35 for both markers), differences in Cp between the two markers were transformed by 2 n to provide an estimate of Wolbachia density, averaged from at least two independent runs. For the second field experiment, we also estimated the Wolbachia density of females after they had laid eggs to see if there was a relationship between Wolbachia density and egg hatch rate when crossed to infected males reared under the same conditions.

Statistical analysis
All data were analyzed using SPSS statistics version 24.0 for Windows (SPSS Inc, Chicago, IL). Hatch proportion and Wolbachia density data were often not normally distributed, and we therefore compared treatments for these variables with Kruskal-Wallis and Mann-Whitney U tests. We also used Spearman's rank-order correlation to test the relationship between egg hatch and female Wolbachia density.

First field experiment
We monitored water temperatures experienced by larvae in each container type at different levels of shade. Maximum temperatures differed between container types at 99% shade, with cups having average maximum daily temperatures that were 2.5˚C higher than buckets, though average temperatures were similar because smaller containers reached cooler temperatures at night (S1 Fig). Temperature cycles were similar between containers in 50% shade, which was unexpected given the large differences in water volume. The level of shade affected temperature substantially, with buckets in direct sunlight experiencing average maximum temperatures of 38.3˚C, while containers in 50% shade (average minimum: 23.7˚C, average maximum: 35.3˚C) were much warmer than containers in 99% shade (23.2-29.6˚C) (Fig 1A).
We tested the ability of wMel-infected males reared under field temperature conditions to induce cytoplasmic incompatibility with uninfected females. In a control cross, wMel-infected males reared in the laboratory at 26˚C caused complete cytoplasmic incompatibility (no eggs hatched) with uninfected females (Fig 1). wMel-infected males reared in buckets at 99% shade induced almost complete cytoplasmic incompatibility, though 2/18 females produced a single viable progeny each (Fig 1B). wMel-infected males reared in buckets at 50% shade induced weaker cytoplasmic incompatibility, with 9/16 females producing some viable progeny ( Fig  1C).
We also tested the ability of wMel-infected females to retain their compatibility with wMelinfected males reared in the laboratory. When females were reared in buckets at 99% shade, there was no difference in hatch rate between crosses with uninfected males and crosses with wMel-infected males (Mann-Whitney U: Z = 0.348, P = 0.726, Fig 1B). In contrast, wMelinfected females reared in buckets at 50% shade had a 47.6% reduction in egg hatch rate when crossed to wMel-infected males (Z = 3.612, P < 0.001). This indicates partial incompatibility with Wolbachia-infected males, suggesting a substantial loss of Wolbachia infection.
We estimated the Wolbachia density of a subset of adults from each container type and level of shade (Fig 2). Wolbachia density was not consistently affected by container type for both females (Kruskal-Wallis χ 2 = 2.598, df = 2, P = 0.273) and males (χ 2 = 4.419, df = 2, P = 0.110), likely because the container types experienced similar temperature fluctuations at 50% shade. Conversely, Wolbachia density was affected substantially by shade level for both females (χ 2 = 71.261, df = 1, P < 0.001) and males (χ 2 = 68.563, df = 1, p < 0.001). Females reared under 50% shade had a median Wolbachia density that was just 0.32% of the laboratory control, while males had a density of 8.09% of the control. This reduction likely reflects the substantially higher maximum daily temperatures experienced in containers at 50% shade. In contrast, the Wolbachia density of adults reared at 99% shade was not significantly different to laboratory-reared adults (females: χ 2 = 0.650, df = 1, P = 0.420, males: χ 2 = 0.085, df = 1, P = 0.771).
All adults screened from containers in 99% shade, 50% shade and the laboratory were positive for Wolbachia. However, we were unable to detect any Wolbachia infection in a sample of 11 adults taken from a bucket placed in direct sunlight. This indicates a complete loss of infection which is likely due to the extreme temperatures experienced in that container (up to 43˚C, Fig 1A). Though we did not score survival to adulthood in containers directly, the bucket placed in direct sunlight experienced high mortality since only 11 adults emerged out of the approximately 100 larvae added initially.

Second field experiment
We conducted a second experiment later in the month where we also tested cytoplasmic incompatibility and measured Wolbachia density in adults. Temperatures were affected substantially by the location of containers where average maximum temperatures were nearly 7˚C warmer in 50% shade compared to 99% shade (Fig 3A). Maximum temperatures also differed between container types; at 50% shade small containers reached 39.26˚C on average while buckets reached 36.54˚C, but average temperatures did not differ much between containers at the same level of shade because of warmer minimum temperatures in buckets.
We set up crosses with adults emerging from buckets held in 50% shade to test for any effects on cytoplasmic incompatibility. wMel-infected males induced strong but incomplete cytoplasmic incompatibility with uninfected females; 7/18 females produced some viable progeny, compared to 0/20 in the control (Fig 3B). wMel-infected adults reared in buckets at 50% shade that were crossed to each other experienced a 79.9% reduction in egg hatch rate when crossed to each other relative to crosses with uninfected males (Mann-Whitney U: Z = 5.615, P < 0.001), suggesting a greatly reduced ability of females to restore compatibility under these rearing conditions.
We estimated the Wolbachia density of adults that emerged from each treatment in the second experiment. We found that many individuals had lost their Wolbachia infection (no detectable infection) when reared in containers held in 50% shade (Fig 4), particularly in small containers where larvae experienced higher maximum daily temperatures (Fig 3A). Of the adults reared at 50% shade that were still infected with Wolbachia, their density had been reduced to 0.19% and 0.23% of the 26˚C control in females and males respectively. In this experiment, Wolbachia density was also reduced at 99% shade relative to the 26˚C control (females: χ 2 = 14.828, df = 1, P < 0.001, males: χ 2 = 16.519, df = 1, P < 0.001), with densities being approximately 50% the level of the control.
wMel-infected adults emerging from the two container types and shade levels were returned to the laboratory and allowed to mate with individuals from the same container. We then scored egg hatch proportions of individual females and measured their Wolbachia density after oviposition to determine the relationship between Wolbachia density and egg hatch proportion. Females with high Wolbachia densities exhibited high hatch proportions while females with lower densities tended to have very low hatch proportions or produced no viable offspring, with the correlation between density and egg hatch being highly significant (Spearman's rank-order correlation: ρ = 0.899, P < 0.001, n = 65, Fig 5). This indicates that females with low densities had partially or completely lost their ability to restore compatibility, but males reared under the same conditions had largely retained their ability to induce cytoplasmic incompatibility. The strong relationship between Wolbachia density and egg hatch indicates that a high density in females is important for restoring compatibility with infected males.

Egg thermal tolerance and Wolbachia density
We tested the tolerance of Wolbachia-infected and uninfected eggs to a broad range of temperature conditions. When eggs were held at 26˚C for one week, Wolbachia-infected eggs did not  Fig 6A). wAlbB-infected eggs also had reduced hatch proportions relative to uninfected eggs at 29-39˚C (Z = 2.162, P = 0.031) but wMel-infected eggs did not differ from uninfected eggs under any temperature cycle (all P > 0.05).
In a second experiment, we used a narrower temperature range to investigate egg thermal tolerance and the loss of Wolbachia infections on a finer scale and with greater replication. Egg hatch proportions declined for all Wolbachia infection types as maximum temperatures increased, with the effect being most severe for the wMelPop infection (Fig 7A). Egg hatch proportions of the wMel and wAlbB strains did not differ significantly from uninfected eggs at maximum temperatures of 37˚C and below (Mann-Whitney U: all P > 0.05). However, at maximum temperatures of 38-41˚C both the wMel and wAlbB strains had lower hatch proportions than uninfected eggs held at the same temperature (all P � 0.026). This indicates that Wolbachia infections in Ae. aegypti lower the tolerance of eggs to high temperatures, particularly in the case of wMelPop.

Loss of cytoplasmic incompatibility under field conditions
Consistent with the previous experiment, Wolbachia density declined as eggs were exposed to increasing maximum temperatures, beginning at 35˚C for wMelPop and 36˚C for wMel (Fig 7B and 7C). The wMel and wMelPop infections were lost from some individuals when eggs were exposed to 29-39˚C and absent from all adults at 30-40˚C (Fig 7B and 7C). In contrast, all wAlbB adults were infected across all temperature cycles, though density was reduced in both females (Mann-Whitney U: Z = 3.204, P = 0.001) and males (Z = 3.897, P < 0.001) at 30-40˚C relative to 26˚C.

Discussion
We tested the stability of the wMel Wolbachia infection in Ae. aegypti under field temperature conditions and performed laboratory experiments to determine the range of temperatures that affect different Wolbachia strains. Our experiments demonstrate three main outcomes of heat stress on Wolbachia-infected mosquitoes. Firstly, there are direct costs of Wolbachia infections on Ae. aegypti thermal tolerance, at least during the egg stage. Secondly, heat stress under partial shade conditions in the field reduces cytoplasmic incompatibility fidelity in wMel-infected males, while infected females become partially incompatible with infected males. Thirdly, heat stress reduces Wolbachia density and may impair the ability of Wolbachia to block virus transmission for a subsection of the mosquito population reared under specific field conditions. Heat stress could therefore adversely affect the success of disease control programs depending on the location and nature of the field breeding sites.
There are relatively few examples of symbionts affecting the thermal tolerance of their hosts [48]. In Drosophila melanogaster, the wMelCS strain of Wolbachia increases the survival of adults under heat stress while the wMelPop infection decreases survival [49], though wMel appears to have no effect on high temperature tolerance [50]. Here we show that Wolbachia infection reduces the tolerance of Ae. aegypti eggs to high temperatures, with the severity of the effect depending on Wolbachia strain. In addition, we have determined the temperature range where deleterious effects on Wolbachia infections start to occur and where the infections  (8 L) or small containers (400 mL) held at 50% or 99% shade were crossed to individuals from the same rearing container before females were isolated for egg hatch rate and Wolbachia density measurements. https://doi.org/10.1371/journal.pntd.0007357.g005 Loss of cytoplasmic incompatibility under field conditions are lost, at least during the egg stage. For the wMelPop and wMel infections, Wolbachia density declined beginning at temperatures of 25-35˚C (30˚C mean) and 26-36˚C (31˚C mean) respectively, while for wAlbB this occurred at a much higher temperature range (30-40˚C, 35˚C mean). The higher tolerance of wAlbB to heat stress is consistent with prior studies in Ae. aegypti larvae [19,35,38], but the increased resolution in this experiment provides a better estimate of the maximum daily temperatures that could affect Wolbachia interventions. In field situations the temperature ranges where Wolbachia infections are adversely affected will depend on the duration and timing of heat stress (see below).
In our field experiments we found substantial effects of heat stress on cytoplasmic incompatibility that could limit the potential of wMel to invade natural populations during disease control programs and persist following releases. When reared in partial shade, wMel-infected males partially lost their ability to induce cytoplasmic incompatibility while females partially or completely lost their ability to restore compatibility when crossed to infected males reared in the lab. Infected females reared in partial shade had greatly reduced fertility in crosses with Loss of cytoplasmic incompatibility under field conditions infected males from the same container. Female Wolbachia density was positively associated with egg hatch, consistent with a study in Drosophila [51]. High densities therefore appear needed for females to restore compatibility with infected males, but the density required for males to induce cytoplasmic incompatibility appears to be lower. Heat stress conditions in the field could greatly diminish or even reverse the reproductive advantage provided by Wolbachia, making invasion challenging, particularly when Wolbachia is at a low frequency, when its fitness relative to uninfected individuals is relatively lower and where it is susceptible to stochastic [52,53] and density related [30,54] effects. Where wMel has already established in a population, reduced egg hatch in wMel-infected mosquitoes that mate with each other could provide an opportunity for an increase in the frequency of uninfected mosquitoes although once wMel invaded areas of North Queensland it appears to have been stable [34]. Fitness costs and self-incompatibility between infected mosquitoes could also have unexpected ecological effects; a decline in the Ae. aegypti population could lead to shifts in species composition [24] which could be beneficial for disease control efforts. Though we attempted to rear mosquitoes under realistic temperature conditions, our field experiments will only be relevant to a subset of natural breeding sites. We provided abundant food to speed up and synchronise larval development to facilitate experimental crosses between strains. Larval development times in nature are variable and can exceed two months under competitive conditions [30]. Increasing the rate of larval development in this experiment likely underestimated the effect of heat stress; longer development times increase the chance that larvae will experience a heat wave and increased durations of heat stress may further reduce Wolbachia density, though density may also recover over time in the absence of heat stress [37]. wMel-infected larvae provided with a low level of food have a greatly reduced Wolbachia density when reared at 26-32˚C compared to 26˚C, indicating that even moderate temperatures can reduce Wolbachia density when combined with nutritional stress (see Figure S4 of Ross and Hoffmann [35]). The effects of heat stress on Wolbachia density can carry over into the next generation [55] which may lead to reduced virus blockage or cytoplasmic incompatibility across a generation after a heat wave. In our laboratory experiments eggs were maintained for one week before hatching, but in the field the egg stage can be shorter or much longer. During the dry season eggs can remain quiescent for months before hatching [56], increasing their potential exposure to high temperatures. Wolbachia infections reduce the viability of quiescent Ae. aegypti eggs [20,26,27] and under high temperatures these fitness costs will likely be exacerbated.
A further limitation of our field experiment is that the containers used for larval rearing were not colonized naturally. Aedes aegypti seem to prefer laying eggs in shaded areas but will also utilize containers in sunlight [57][58][59]. Wolbachia infections may also affect thermal preference; adult Drosophila melanogaster infected with Wolbachia prefer cooler temperatures than uninfected flies [60, 61]. Nevertheless, data from sentinel containers indicates that wMelinfected mosquitoes will lay eggs in containers placed in direct sunlight. Sentinel buckets and small containers placed within the wMel release zone in Cairns were all colonized by Ae. aegypti despite some of these experiencing similar temperatures to the experimental containers held at 50% shade (S2 Fig). Ae. aegypti tend to lay eggs during cooler parts of the day [62, 63] and therefore may be unable to discriminate against habitats that reach high maximum temperatures later. Unlike adult mosquitoes, immature stages cannot easily escape heat stress as they are unable to move beyond the container. Since Wolbachia density and egg hatch in wMel-infected mosquitoes appears to depend strongly on the level of shade, temperature and productivity surveys of larval habitats could be conducted in release areas if there are concerns around heat stress impacts in a release area.
Despite the substantial effects on Wolbachia density and fertility in our experiments, Ae. aegypti mosquitoes infected with wMel have successfully established in Cairns [11,12] and Townsville [13], Australia and in Brazil [23], with the infection persisting at a high frequency in most locations. In areas where the releases succeeded, the costs of heat stress observed here were clearly not prevalent or severe enough to prevent the establishment of wMel. Once a Wolbachia infection has attained a high frequency in a population it may stay high unless the fitness costs are extreme, as is the case for wMelPop [28]. Nevertheless, heat stress will likely slow the rate of Wolbachia invasion and spread, increasing the number of mosquitoes required for releases, and potentially creating an unstable situation around critical invasion points that must be exceeded for Wolbachia to invade [52]. Heat stress could partially explain why infection frequencies have persisted at intermediate levels in some suburbs [13] and may also contribute to the incomplete maternal transmission fidelity of wMel observed in Cairns [64] given that some individuals were cleared of their Wolbachia infections in our experiments. wMelinfected mosquito releases outside of Australia in locations where maximum daily temperatures are warmer may be more challenging. Reduced Wolbachia densities may also reduce virus protection provided by Wolbachia even if infection frequencies remain high in a population, though we do not demonstrate this effect directly. The wMel strain has retained its susceptibility to heat stress for seven years after field deployment in Australia [35], indicating that alternative strains may be needed in areas where wMel has difficulty establishing or where viral blockage is insufficient.